Epistemology: Contexts, Values, Disagreement: Proceedings of the 34th International Ludwig Wittgenstein Symposium in Kirchberg, 2011 9783110329018, 9783110328646

Table of contents :
Contents
Preface
I. Contextualism and Invariantism
Knowledge, Contextualism, and Moorean Paradox
Context-Sensitivity without Indexicality?
Non-Relativist Contextualism about Knowledge
Knowing about Other Contexts
Epistemic Variantism and the Factivity of Knowledge
Explaining Variation in Knowledge by Full Belief
The Contextualist Promise
II. Epistemic Virtues
Intellectual Virtues and Their Place in Philosophy
Understanding’s Tethers
Knowledge as Achievement and the Value Problem
Apriority and Virtue: How Successful a Relationship?
III. The Nature and Value of Knowledge
Epistemology without the Concept of Knowledge
Reliabilism as Explicating Knowledge: A Sketch of an Account
How to Take Truth as a Goal?
Knowledge as Achievement – Greco’s Double Mistake
IV. Testimony
Norms of Trust, De Re Trust, and the Epistemology of Testimony
Jennifer Lackey on Non-Reductionism: A Critique
V. Disagreement
“A Troubled Area”. Understanding the Controversy over Screening Mammography for Women Aged 40-49
Is Philosophical Knowledge Possible?
The Significance of Disagreement in Epistemology
The Role of the Uniqueness Thesis in the Epistemology of Disagreement
Counterfactual-Peer Disagreement
Meta-Induction and the Problem of Fundamental Disagreement
Rival Logics, Disagreement and Reflective Equilibrium
VI. Wittgenstein
The Builders
First Person Authority without Glamorous Self-Knowledge
Subjective and Objective Certainty as Regards Knowledge and Action
Zu einigen Bemerkungen Wittgensteins über die Seele
Conjecture, Proof, and Sense in Wittgenstein’s Philosophy of Mathematics
Through Pictures to Problems: Cognitive Epistemology and Therapeutic Philosophy
Werte und Vernunft in Wittgensteins Spätwerk
Wittgenstein über Gedankenexperimente
Wittgenstein, Ethics and Therapy

Citation preview

Christoph Jäger | Winfried Löffler (Eds.) Epistemology: Contexts, Values, Disagreement

Publications of the Austrian Ludwig Wittgenstein Society. New Series Volume 19

Christoph Jäger | Winfried Löffler (Eds.)

Epistemology: Contexts, Values, Disagreement Proceedings of the 34th International Ludwig Wittgenstein-Symposium in Kirchberg am Wechsel, Austria 2011

Bibliographic information published by Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliographie; detailed bibliographic data is available in the Internet at http://dnb.ddb.de

North and South America by Transaction Books Rutgers University Piscataway, NJ 08854-8042 [email protected] United Kingdom, Ire-, Iceland, Turkey, Malta, Portugal, Spain by Gazelle Books Services Limited White Cross Mills Hightown LANCASTER, LA1 4XS [email protected]

Livraison pour la France et la Belgique: Librairie Philosophique J.Vrin 6, place de la Sorbonne ; F-75005 PARIS Tel. +33 (0)1 43 54 03 47 ; Fax +33 (0)1 43 54 48 18 www.vrin.fr

2012 ontos verlag P.O. Box 15 41, D-63133 Heusenstamm www.ontosverlag.com ISBN 978-3-86838-160-3 2012 No part of this book may be reproduced, stored in retrieval systems or transmitted in any form or by any means, electronic, mechanical, photocopying, microfilming, recording or otherwise without written permission from the Publisher, with the exception of any material supplied specifically for the purpose of being entered and executed on a computer system, for exclusive use of the purchaser of the work Printed on acid-free paper ISO-Norm 970-6 FSC-certified (Forest Stewardship Council) This hardcover binding meets the International Library standard Printed in Germany By CPI buch bücher.de

Contents

Preface

9 

I. Contextualism and Invariantism Knowledge, Contextualism, and Moorean Paradox (ELKE BRENDEL)

15 

Context-Sensitivity without Indexicality? (WAYNE A. DAVIS)

41 

Non-Relativist Contextualism about Knowledge (MARCUS WILLASCHEK)

53 

Knowing about Other Contexts (PETER BAUMANN)

63 

Epistemic Variantism and the Factivity of Knowledge (WOLFGANG FREITAG)

81 

Explaining Variation in Knowledge by Full Belief (NICHOLAS SHACKEL)

95 

The Contextualist Promise (SEBASTIAN SCHMORANZER)

103 

II. Epistemic Virtues Intellectual Virtues and Their Place in Philosophy (JOHN GRECO)

117 

Understanding’s Tethers (CATHERINE Z. ELGIN)

131 

Knowledge as Achievement and the Value Problem (BRUNO NIEDERBACHER)

147 

Apriority and Virtue: How Successful a Relationship? (NENAD MIŠČEVIČ)

155 

III. The Nature and Value of Knowledge Epistemology without the Concept of Knowledge (ANSGAR BECKERMANN)

171 

Reliabilism as Explicating Knowledge: A Sketch of an Account (ERIK J. OLSSON)

189 

How to Take Truth as a Goal? (MARIAN DAVID)

203 

Knowledge as Achievement – Greco’s Double Mistake (CHRISTIAN PILLER)

215 

IV. Testimony Norms of Trust, De Re Trust, and the Epistemology of Testimony (SANFORD GOLDBERG)

229 

Jennifer Lackey on Non-Reductionism: A Critique (MARTIN KUSCH)

257 

V. Disagreement “A Troubled Area”. Understanding the Controversy over Screening Mammography for Women Aged 40-49 (MIRIAM SOLOMON)

271 

Is Philosophical Knowledge Possible? (HILARY KORNBLITH)

285 

The Significance of Disagreement in Epistemology (DIRK KOPPELBERG)

305 

The Role of the Uniqueness Thesis in the Epistemology of Disagreement (MATTHEW LEE)

319 

Counterfactual-Peer Disagreement (KATHERINE MUNN)

329 

Meta-Induction and the Problem of Fundamental Disagreement (GERHARD SCHURZ)

343 

Rival Logics, Disagreement and Reflective Equilibrium (GEORG BRUN)

355 

VI. Wittgenstein The Builders (MEREDITH WILLIAMS)

371 

First Person Authority without Glamorous Self-Knowledge (ANDREAS KEMMERLING)

399 

Subjective and Objective Certainty as Regards Knowledge and Action (MICHAEL KOBER)

429 

Zu einigen Bemerkungen Wittgensteins über die Seele (RICHARD RAATZSCH)

439 

Conjecture, Proof, and Sense in Wittgenstein’s Philosophy of Mathematics (SEVERIN SCHROEDER)

459 

Through Pictures to Problems: Cognitive Epistemology and Therapeutic Philosophy (EUGEN FISCHER)

475 

Werte und Vernunft in Wittgensteins Spätwerk (ANDREAS KORITENSKY)

493 

Wittgenstein über Gedankenexperimente (JOACHIM BROMAND)

507 

Wittgenstein, Ethics and Therapy (EDWARD HARCOURT)

523 

Preface The present volume collects papers that were presented at the 34th International Wittgenstein Symposium “Epistemology: Contexts, Values, Disagreement” 2011 in Kirchberg. Only a few years ago, influential philosophers proclaimed the death of epistemology. Their verdict proved to be false. Today, the theory of knowledge flourishes as perhaps never before. The Kirchberg Symposium 2011 focused on some of the most recent debates and developments in current epistemology: (epistemic) contextualism and invariantism; epistemic virtues; the value of knowledge; testimony; and the structure and importance of rational disagreement. Finally, a large section whose topics were not required to focus on general epistemology was devoted to the work of the great intellectual patron of the annual symposia, Ludwig Wittgenstein. Epistemology not only enjoys great prominence among professional philosophers. The discipline has become so popular in recent years that even influential politicians have been seduced by the attractions of epistemological reasoning. Recall a by now famous statement from a public figure who – wittingly or not – introduced the intricacies of epistemological reasoning to a world wide audience. “There are known knowns”, the person in question reasoned, “there are things we know we know. We also know there are known unknowns. That is to say, we know there are some things we do not know. But there are also unknown unknowns – the ones we don’t know we don’t know. – And ... it is the latter ... that tend to be the difficult ones.” (Donald Rumsfeld, quoted from “The Economist”, December 4, 2003; for live footage see www.youtube.com.) Throughout the world, this statement quickly became a most useful means in introductory classes for piquing the epistemology student’s interest. What does it mean to know something? What are the conditions for knowing that one knows, and for knowing that one fails to know? What makes an ordinary knowledge attribution true? And, we may ask, is it really true that only “unknown unknowns” are “the difficult ones”? Epistemology aims at answering such questions, and it helps us disentangle arguments and statements presented outside the philosophy classroom as well. Though not all statements about knowledge in non-philosophical contexts are as accessible as might be desirable, knowledge itself may be eas-

10

ier to achieve in everyday contexts than in the seminar. This, at least, is what epistemic contextualists maintain. Ordinary knowledge attributions, they claim, are typically true in quotidian contexts, yet the same knowledge attributions are false in demanding, and especially in sceptical, contexts. The pros and cons of this view are the topic of the section “Contextualism and Invariantism”. – It was a comforting experience that many speakers and guests seemed to have no trouble in switching from demanding philosophy classroom contexts to quotidian contexts, and that knowledge attributions to the effect that people knew that the local pubs exist proved to be true in the evenings. Discussions such as these take us into the normative realm of questions about good epistemic (and linguistic) behaviour. In recent years, epistemologists have broadened the perspective on such questions by investigating what they call “epistemic” or, more broadly, “intellectual virtues”. What is the nature of such virtues, and how are they connected to our cardinal epistemic goal to “reach the truth” (or, more precisely, to acquire and maintain true beliefs while avoiding false ones in matters of significance)? When should we credit people for what they believe, and how are epistemic virtues conceptually connected with other, especially moral or other practical virtues? Such are the questions of our section on virtue epistemology. Many epistemologists will not agree with the claim from the above quotation that only “unknown unknowns” are “the difficult ones”. Many would agree however that it is good to have knowledge, at least concerning nontrivial matters of practical significance. Often it seems better to have knowledge than mere true belief. (In other words: it would often seem better to possess knowns than to possess unknowns, even if the latter are constituted by true beliefs.) But why exactly would that be so? In particular, why does knowledge seem to be more valuable than mere true belief? This question, too, has recently gained centre stage in analytical epistemology. It is extensively discussed in the contributions represented in the section “The Nature and Value of Knowledge”. Proceeding from within a broadly Cartesian perspective, epistemology traditionally focused on issues such as the nature and content of subjective epistemic obligations, the structure of internal justification, and the methods of individual knowledge acquisition. Only fairly recently philosophers have begun to take seriously the fact that in many, if not most epistemic settings knowledge acquisition is essentially a social enterprise. Most of our (true) beliefs rely upon the testimony of others. Social epistemology

11

has begun to systematize such observations, and in recent years epistemologists have developed detailed theories of testimonial knowledge. This is the topic of our section on testimony. Philosophers, too, must for the greater part of their belief-forming lives rely on others. Yet they very much like to quarrel, and often only agree that they disagree. What epistemologists don’t agree upon is when it is rational to disagree. Even more interesting and intricate is the question whether there can be rational disagreement among (mutually acknowledged) epistemic peers – people whom one takes to be more or less as well informed, competent, etc., regarding a given topic as one takes oneself to be. How should we react if, as it often happens, we find ourselves in persisting disagreement with acknowledged epistemic peers? Should we let this weaken or maybe even defeat our justification (or epistemic entitlement, warrant, etc.) for what we believe, or are we entitled in such situations to stick to our guns? This is one of the hottest topics in current epistemology, thus our section on “disagreement”. Finally, there is Wittgenstein. For decades, Wittgenstein scholars have been coming to Kirchberg in order to discuss the work and life of this outstanding figure of analytical philosophy. In this tradition, the biggest section of this year’s symposium was once more the Wittgenstein section, with many contributions devoted to epistemological issues in Wittgenstein’s Œuvre. May the present volume contribute to creating demanding epistemological contexts in which we value knowledge and understanding, explore intellectual virtues, learn from the testimony of our colleagues and engage in constructive peer disagreement. We thank all the speakers and contributors to this volume and congratulate Matthew Lee and Wolfgang Freitag for having won the Leinfellner award, sponsored by the Austrian Ludwig Wittgenstein Society and Dr. Rafael Hüntelmann from Ontos publishers. Thanks to all our co-organizers and supporters in Kirchberg for their great engagement. Special thanks go to Mag. Monika Datterl’s super-efficient and competent lay-outing and typesetting; to Dr. Rafael Hüntelmann for his patience; to Mag. Margret Kronaus and her crew for their excellent organisation of all the small and big practical issues in Kirchberg; and to the board of the Austrian Ludwig Wittgenstein Society for committing to us the society’s 2011 symposium. Innsbruck, June 2012

Christoph Jäger & Winfried Löffler

I. Contextualism and Invariantism

Knowledge, Contextualism, and Moorean Paradox ELKE BRENDEL, University of Bonn

Abstract The paper gives a comparative overview of contextualist, relativist and subject-sensitive invariantist accounts of knowledge. It is argued that none of these theories can provide a satisfying semantics for knowledge ascriptions. It is shown that there are crucial dissimilarities between “know” and indexical or context-sensitive terms. In particular, contextualists, relativists, and subject-sensitive invariantists are committed to some kind of Moorean paradoxes. 1. Introduction In contemporary epistemology, there is an ongoing debate about the proper semantic analysis of the term “know”. Many recent epistemologists defend the view that the truth conditions of knowledge ascriptions of the form “S knows that p (at t)” (where “S” refers to an epistemic subject, “p” to a proposition, and “t” to a point in time) depend inter alia on certain epistemic standards for “know”, i.e., besides the truth of p and S’s belief that p, it is also required that S meets certain epistemic standards in order to know that p. Epistemologists, however, strongly disagree about the exact features that determine epistemic standards and about whose epistemic standards are decisive for the truth conditions of knowledge ascriptions. There are three different epistemic contexts with respect to knowledge ascriptions: First of all, there is the context of the epistemic subject S herself which is determined by, for example, S’s interests, expectations, S’s practical costs of being wrong about p, etc. Second, there is the context of the knowledge ascriber, i.e., the context of the speaker or the context of the person who utters the knowledge-ascription. In the following, I will call this context the context of utterance, abbreviated with cu. Apart from cases of knowledge self-ascription where the epistemic subject is identical with

16

Elke Brendel

the knowledge ascriber, the speaker’s epistemic context can be quite different from S’s context. For example, the practical interests of the speaker with regard to p might be different to the interests of the would-be-knower S. Furthermore, the speaker might have different background knowledge with respect to p and might consider other error-possibilities to p than the epistemic subject. Third, in cases where a knowledge ascription uttered by a speaker is evaluated from the perspective of an assessor, we also have to take the context of assessment – “cA” for short – into account. According to contextualism, one of the most prominent and influential semantic accounts of knowledge, the truth values of knowledge ascriptions depend on the epistemic standards operative in the context of utterance cu. Another recent account is relativism, according to which knowledge ascriptions are assessment-sensitive. Relativists thus claim that the truth values of knowledge ascriptions vary with the context of assessment cA. In contrast to contextualism and relativism, defenders of subject-sensitive invariantism argue that the truth values of knowledge ascriptions are neither sensitive to the varying epistemic standards operative in cu nor to the varying epistemic standards operative in cA. They contend that the epistemic standards required for knowledge are exclusively determined by the practical interests of the epistemic subject. Contextualists, relativists, and subjectsensitive invariantists argue for the superiority of their respective views. They are convinced that the positions they embrace are best suited for accounting for linguistic intuitions with regard to “know” and for addressing certain notorious epistemic problems, such as skepticism. In the following, I will provide arguments against contextualist, relativist and subject-sensitive invariantist semantics for knowledge ascriptions. Counterintuitive results of these accounts are discussed, and it is shown that there are crucial dissimilarities between “know” and indexical or context-sensitive terms. In particular, it will be argued that contextualism, relativism, and subject-sensitive invariantism are confronted with problems of Moorean paradoxes. 2. Contexualist Accounts of Knowledge According to contextualism, the truth value of a knowledge ascription varies according to the context of utterance cu. In Kaplanian semantics (Kaplan 1989), the context of utterance is modeled as a sequence of parameters including a parameter for the world in which the utterance is made, as well

Knowledge, Contextualism, and Moorean Paradox

17

as a parameter for the speaker, the time and a location. Sentences like “Peter is sleeping”, “I am tired”, “Tomorrow is Monday”, “It is raining here”, etc., can thus have different truth values due to varying states of the world, speakers, times and locations at different contexts of utterance. Contextualists contend that a context of utterance also includes a parameter for the speaker’s epistemic standard for knowledge. Due to these varying epistemic standards operative in the context of utterance, statements of the form “S knows that p (at t)1” can also have varying truth values at different contexts of utterances. Contextualism comes in two different variants: indexical contextualism and non-indexical contextualism. In indexical contextualism, championed, in particular, by Stewart Cohen, Keith DeRose and David Lewis, “know” is conceived as an indexical. “Know” is therefore semantically analogous to other indexicals, such as “I”, “here, “tomorrow”. Due to the indexicality of “I”, for example, the sentence “I am female” uttered by Angela Merkel expresses a different proposition than uttered by Barack Obama. Uttered in these different contexts, one and the same sentence can therefore have different truth values: In the first context of utterance it expresses a true proposition, and in the second context of utterance it expresses a false proposition. Consequently, Angela Merkel can truthfully claim “I am female” and Barack Obama can truthfully claim “I am not female”, without contradicting each other. The proposition expressed by Obama is simply not the negation of the proposition expressed by Merkel. Indexical contextualists, maintain that the term “know” semantically also behaves like an indexical: Because of the assumed indexicality of “know”, a knowledge ascription can express different propositions at different contexts of utterance cu. Knowledge ascriptions in contextualist accounts are therefore sometimes called “use-indexical” (MacFarlane 2005b, 326). According to indexical contextualism, the proposition expressed by “S knows that p” at cu depends on the epistemic standards of the speaker operative in cu and is true only if S’s epistemic situation allows S to meet the epistemic standards for knowledge operative in cu. As a consequence, a 1

For reasons of simplicity, I will omit the time index in the following discussion of contextualism and relativism. Since the time of knowing plays an important role in subject-sensitive invariantism, I will later use time indices again when discussing subject-sensitive invariantism.

18

Elke Brendel

knowledge ascription “S knows that p” can be true at cu1, but false at a different context cu2 in which the epistemic standards are higher or more stringent than in cu1 – for the same subject S and the same true proposition p. Furthermore, “S knows that p” can be true at cu1 and “S does not know that p” can also be true at cu2, without any contradiction. Cohen explicitly holds such an indexical contexualist view by claiming: “[…] the theory I wish to defend construes “knowledge” as an indexical. As such, one speaker may attribute knowledge to a subject while another speaker denies knowledge to that same subject, without contradiction.” (Cohen 1988, 97).

In addition, he semantically compares “know” with context-sensitive gradable adjectives: “The standards that determine how good one’s reasons have to be in order to know are determined by the context of ascription. Of course, many predicates in natural languages are such that the truth value of sentences containing them depend on contextually determined standards, e.g., flat, bald, rich, happy, sad …” (Cohen 2000, 97).

Just as “The Taunus is flat” is true uttered by an ambitious professional mountain climber and false uttered by a hobby hiker due to different standards for “flat” at play in the respective contexts of utterance, a knowledge ascription can, according to Cohen, also have varying truth values if there are different epistemic standards operative in different contexts of utterance. Non-Indexical Contextualism (see, for example, Kompa 2002) is the slightly different view, according to which knowledge ascriptions are usesensitive without being use-indexical, i.e., the proposition expressed by “S knows that p” is invariant across contexts of use cu, but its truth value nevertheless varies with the epistemic standards operative in cu. The features that determine the epistemic standards for knowledge operative in cu are, in particular, the “purposes, intentions, expectations, presuppositions etc., of the speaker” (Cohen 2000, 94), “the importance of being right” (DeRose 1992, 914), and the consideration of relevant alternatives or error-possibilities (DeRose 1992, 915, and Lewis 1996, 559). According to an indexical contextualist, like DeRose, the knowledge standards can rise if the practical costs for the speaker of being wrong rise or if the speaker considers error-possibilities hitherto ignored. Consequently, the

Knowledge, Contextualism, and Moorean Paradox

19

proposition expressed by a knowledge ascription varies with such changing standards: “[…] there seems to be a fairly straightforward and important sense in which one does mean something different if the range of relevant alternatives has been changed by attributor factors […].” (DeRose 1992, 920)

What are the advantages of (indexical and non-indexical) contextualism? First of all, it seems that these positions can account for certain contextualist intuitions with regard to knowledge ascriptions. Many examples and test cases put forward by contextualists seem to support the view that assuming a context-dependent variability of knowledge ascriptions due to varying epistemic standards of the knowledge attributor is intuitively highly plausible. In DeRose’s famous “bank cases” (DeRose 1992, 913), Keith and his wife have to deposit their paychecks. They drive past a bank. It is Friday afternoon and the lines inside the bank are very long. Keith’s wife, realizing that she has been to that bank on a Saturday morning just a couple of weeks ago, says: “I know the bank will be open tomorrow”. Let us assume that the bank is indeed open on Saturdays and that it is not very important that they deposit their paychecks before the beginning of the next week. In this “low stakes” context it seems that the wife’s claim that she knows that the bank will be open on Saturday is true. Now consider a different context in which it is very important that the couple deposits their paychecks by Saturday because the check will otherwise bounce, leaving them in a “very bad situation”. In this “high stakes” context, the standards for knowledge have been raised such that the wife’s claim “I know that the bank will be open tomorrow” turns out to be false – although the bank is indeed open on Saturdays and the evidence the wife has for her belief that the bank will be open the next day (namely that she has been to the bank on a Saturday a couple of weeks ago, where the bank was open) is still the same. A similar situation arises, according to contextualists, if an important errorpossibility is considered that the epistemic subject cannot rule out. If, for example, Keith considers the possibility that the bank could have changed its hours, his wife’s claim “I know that the bank is open tomorrow” is again false in this high standards context. These and other cases seem to support the contextualist thesis that we deny knowledge to ourselves and to others in situations where the costs of error are high or relevant alternatives become salient.

20

Elke Brendel

Second, one of the main advantages of contextualism is its alleged ability to offer an answer to the skeptical challenge. The skeptical challenge can be described in terms of the following inconsistent triad: (1) (2)

(3)

S knows that p (where p is an ordinary empirical proposition, e.g., that S has hands). S knows that: if p, then not-h (where h is a skeptical hypothesis, e.g., that S is a handless brain in a vat whose neurons are stimulated by electrical impulses of a supercomputer, such that a perfect illusion of an external world is induced). S does not know that not-h.

Each of the claims (1)-(3) seem to be correct, but they are mutually inconsistent: (1) and (2), together with the plausible principle that knowledge is closed under known logical implication,2 lead to the negation of (3), whereas (2) and (3) imply the negation of (1). In light of this inconsistent triad, the skeptic, of course, argues that since we cannot know that skeptical hypotheses do not hold (i.e., (3) is true), we have to conclude that knowledge about empirical propositions is impossible (i.e., (1) is false). The contextualist, however, has an easy way out of the skeptical challenge. According to contextualism, there is an unnoticed context change in the reasoning from the three above claims to inconsistency. (1) can be true in a low or ordinary standards context in which skeptical hypothesis are not salient, and (3) is only true in a high standards context where the speaker considers skeptical hypotheses. Keeping the standards fixed, the three claims are not mutually inconsistent: In a low or ordinary standards context, we can know that not-h, i.e., in such a context, we can have the following truth value assignments: (1) and (2) are true, but (3) is false. Only in a skeptical standards context, we cannot have knowledge about ordinary propositions, since we cannot rule out the skeptical error-possibility that is salient in this context. That is why in such a high standards context, the following holds: (1) is false, whereas (2) and (3) are true. So, the inconsistency disappears if we keep the context of utterance fixed. Contextualists contend that by their analysis of the “inconsistent triad” a satisfying solution to the skeptical challenge is achieved without rejecting 2

This principle of epistemic closure can be expressed as follows: If S knows that p, and if S knows that p implies q, then S knows that q.

Knowledge, Contextualism, and Moorean Paradox

21

skepticism wholesale. Contextualists argue against skepticism by claiming that there are contexts in which knowledge of the empirical world is possible. But they still make a concession to the skeptic insofar as in skeptical contexts positive knowledge claims about empirical propositions are always false. 3. Problems of Contextualism Contextualism has been criticized for various reasons. In particular, it could be argued that contextualists are still too concessive to the skeptic and that it is implausible that the mere and unmotivated consideration of a far-fetched error-possibility always raise the standards for “know” such that we land in a high standards context in which positive knowledge ascriptions turn out false (see, for example, Williams 2005). Furthermore, it is highly contestable whether, as DeRose claims, the “lack of contradiction is the key to the sense in which the knowledge attributor and the knowledge denier mean something different by ‘know’” (DeRose 1992, 920). According to indexical contextualism, an apparent dispute between, for example, a common-sense philosopher and a skeptic can simply be settled by pointing out that the common-sense philosopher’s knowledge ascription does not express the same proposition that the skeptic denies. If, for example, Cypher, the anti-skeptic, claims in low standards context of utterance: “Anna knows that she has hands”, and Neo, the skeptic, claims in high standards context: “Anna does not know that she has hands”, Cypher and Neo mean something different by ‘know’. Therefore, both claims can be truthfully uttered. But in contrast to what contextualism predicts, we have, as John Hawthorne contends, “very few devices in ordinary life for implementing the clarification technique when it comes to ‘knows’” (Hawthorne 2004, 105). There is a strong intuition that we have in many cases a real and genuine disagreement between a knowledge ascriber and a knowledge denier that cannot be put to an end by simply clarifying possible different uses of ‘know’. Thus, contextualists are committed to a kind of error theory, because they have to admit that many of our intuitions of real disagreement between a knowledge ascriber and a knowledge denier are actually erroneous. In the following, I would like to point out some further problems for contextualism. In particular, I will argue that contextualism makes counterintuitive predictions in certain cases of context shift. The first problem

22

Elke Brendel

arises when a speaker refers to a knowledge claim uttered in a context different from her own context. For example, Neo, in his high standards context, should, according to contextualism, accept both that Anna does not know that she has hands and that Cypher spoke truthfully when claiming that Anna knows that she has hands. This result seems to be highly implausible. It would be much more natural for Neo to regard Cypher’s knowledge ascription as false. Similar counterintuitive results for contextualism can occur when we reflect on our own knowledge claims that we made when we were in different contexts. In these cases contextualism does not seem to be able to account for the phenomenon of retracting knowledge claims. Consider the case where Anna just parked her bicycle in front of the logic department. Some minutes later her colleague Bart asks her whether she knows where her bicycle is. Anna replies: “Yes, I know that my bicycle is parked in front of the logic department.” Bart now tells her, that there have just been bicycle thefts in front of the logic department and that many bicycles have already been stolen. This information about the bicycle thefts provides Anna with an error-possibility that puts her in a higher standards context. Consequently, she now claims: “I do not know that my bicycle is parked in front of the logic department.” Let us assume that Anna’s bicycle has not been stolen and is still standing in front of the logic department. According to contextualism, Anna should now, in her higher standards context, accept both that she does not know that her bicycle is parked in front of the logic department, but that she spoke truthfully when she claimed (in her lower standards context) that she knew that her bicycle is parked in front of the logic department. Again, this seems to be inadequate because Anna would most probably rather retract her knowledge claim she made in the lower standards context. So, she would claim that she does not know that her bicycle is parked in front of the logic department and that she did not knew it before. If we consider an important error-possibility to what we claimed to know, we normally retract our former knowledge claim – even when we learn that what we claimed to know was true. The reason for this is that we tend to regard the high standards context as the better or the most adequate one. We therefore cannot simply lower the standards for knowledge again if we become aware of an error possibility. We have, as Hawthorne writes, a strong “inclination to reckon ourselves more enlightened with regard to our former self (on the topic of knowledge) when possibilities of error be-

Knowledge, Contextualism, and Moorean Paradox

23

come salient” (Hawthorne 2004, 106). Contextualism cannot satisfactorily account for this inclination. A similar problem for contextualism arises in cases in which the knowledge attributor is in a low standards context because he does not attend to an error-possibility that is relevant given the situation in which the knowledge ascription is made. Assume that Carl, another of Anna’s colleagues, is completely ignorant of the fact that there have been bicycle thefts in front of the logic department. So, according to contextualism, Carl’s utterance “Anna knows that her bicycle is parked in front of the logic department” is true in his low standards context. It would be true even if Anna were aware of bicycle thefts, and would therefore refrain from claiming to know that her bicycle is parked in front of the logic department. But again, this is not what intuition tells us. If there have been bicycle thefts in front of the logic department such that it could have easily been the case that Anna’s bicycle was stolen, the error-possibility of bicycle thefts is relevant and thus makes the knowledge ascription “Anna knows that her bicycle is parked in front of the logic department” false – independent of the epistemic standards that happen to be operative in the context of the speaker.3

3

DeRose tries to address some of the problems for contextualism that I have pointed out above in the framework of what he calls “single scoreboard semantics” (see, for example, DeRose 2009, ch. 4). This account, which is in some ways a departure of pure indexical contextualism, has, to my mind, its own problems and shortcomings. First of all, as John MacFarlane has pointed out, the single scoreboard semantics does not succeed in addressing the contextualist’s problem of “lost disagreement” in all cases. In particular, DeRose’s single scoreboard approach cannot adequately account for cases of inter-conversational disagreement, i.e., cases of disagreement between two parties who are not involved in a direct conversation with each other (see MacFarlane 2007, 20f.). Furthermore, I find it quite counterintuitive to claim that in disputes where the subject fails to meet the standards of at least one disputant or in cases of “great divergence” (as in the debate between a skeptic and her opponent), knowledge ascriptions simply lack truth-value (see, for example, DeRose 2009, 145). In addition, the objection to contextualism that a relevant alternative given in the situation of the knowledge ascription has an effect on the truth value of the knowledge ascription, independent on whether the knowledge ascriber (or even a whole party to a conversation) is attending to this alternative or not, is not answered by DeRose’s single scoreboard semantics.

24

Elke Brendel

4. Moorean Paradoxes for Contextualists In the following, I would like to elaborate on problems for contextualists with regard to specific knowledge self-ascriptions. In particular, I will argue that a contextualist is committed to accept certain self-ascriptions of knowledge that cannot be sincerely uttered without self-contradiction. Thus, contextualists are committed to paradoxes similar to the famous selfcontradictory paradoxes of George Edward Moore. One of the selfcontradictory paradoxes Moore discusses consists in assertions of the form “p, but I do not know that p”, such as “Dogs bark, but I don’t know that they do” (Moore 1962, 277). Although these sentences can be true (since a fact p might not be known by a person), they cannot sincerely be asserted without self-contradiction by any speaker S. In asserting a sentence of the form “p, but I do not know that p” S violates the knowledge rule of assertion (KRA) which can be explicated by the following claims (KRA1), (KRA2), and (KRA3): (KRA1) (KRA2) (KRA3)

S should assert that ψ only if S knows that ψ. If it is impossible for S to know a proposition ψ, then ψ cannot be sincerely asserted by S. S cannot properly assert that ψ if S does not reasonably take herself to know that ψ.

Moore argues that “[…] by asserting p positively you imply, though you don’t assert that you know that p (Moore 1962, 277). In asserting “Dogs bark”, the speaker S therefore implies that she knows that dogs bark – or that she reasonably takes herself to know that dogs bark. In asserting that she does not know that dogs bark in the same breath, S thereby retracts the knowledge implication of the first conjunct of the assertion, which is, of course, self-contradictory. A similar knowledge rule of assertion holds for contextualist accounts of knowledge. The following contextualized knowledge rule of assertion (CKRA1) is due to DeRose: (C-KRA1) “One must assert only what one knows according to the standards for knowledge that are in place as one makes one’s assertion.” (DeRose 2009, 99) The contextualized versions of (KRA2) and (KRA3) can now be formulated by the following principles (C-KRA2) and (C-KRA3):

Knowledge, Contextualism, and Moorean Paradox

25

(C-KRA2) If it is impossible for S to know that ψ according to the standards for knowledge that are in place as S asserts that ψ, S cannot sincerely assert that ψ. (C-KRA3) S cannot properly assert that ψ if S does not reasonably take herself to know that ψ according to the standards for knowledge that are in place as S asserts that ψ. As in the non-contextualized version, a speaker S would commit a Moorean paradox if S claimed in the context of utterance cui: “p, but I don’t know in cui that p”. In contextualism, the distribution principle of knowledge over conjunction as well as the factivity principle of knowledge holds. Let “KiSφ” stands for “‘S knows that φ’ is true in the context of utterance cui”. The parameter i indicates the epistemic standards of the speaker operative in the context of utterance cu. If i = l, standards are low; if i = h, standards are high.4 The distribution principle of knowledge over conjunction and the factivity principle of knowledge can thus be formalized by the following principles (DC) and (F) Distribution of contextualized knowledge over conjunction: (DC) KiS(φ  ψ) → (KiSφ  KiSψ) for all subjects S, propositions φ and ψ, and standards i  {l,h} Factivity of contextualized knowledge: (F)

KiSφ → φ for all subjects S, propositions φ, and standards i  {l,h}

Of course, even in contextualist accounts of knowledge, a knowledge ascription “S knows that p” is true in any context of utterance only if p is true.

4

For reasons of simplicity, I will restrict the standards to two distinctive epistemic standards that are operative in the context of utterances: a low standard and a high standard. I do not want to make any decisive claims about how many standards there are or how these standards are ordered. It is sufficient for the following arguments to presuppose that there at least two different standards that have an influence on the varying truth values of knowledge ascriptions.

26

Elke Brendel

It can now easily be seen that it is impossible for a speaker S to sincerely assert in context cui: “p, but I don’t know in cui that p”. Such an assertion violates the (contextualist) knowledge rule of assertion since “p, but I don’t know in cui that p” cannot be known by the speaker in cui – as formulated in the following general Moorean principle for contextualism (MP-C): (MP-C)

¬KiS(φ  ¬KiSφ) for all subjects S, propositions φ, and standards i  {l,h}.

Proof of (MP-C): Assume that KiS(φ  ¬KiSφ). With (DC) we get KiSφ  KiS¬KiSφ. Applying (F) to the second conjunct, we get the contradiction: KiSφ  ¬KiSφ. So, for example, if Anna claimed in cuh (where she is aware of recent bicycle thefts in front of the logic department): “My bicycle is parked in front of the logic department, but I don’t know (in cuh) that it is parked there”, she would commit a self-contradictory Moorean paradox since it is impossible for her to know this in her context of utterance. But according to contextualism, it is true and it should also be possible for Anna to sincerely assert in cuh that she does not know that her bicycle is parked in front of the logic department, and that she would know it in the low standards context cul in which the possibility of bicycle thefts is not salient: (4)

Anna in cuh: “I do not know that my bicycle is parked in front of the logic department, but I know it in cul” – or: “I do not know that my bicycle is parked in front of the logic department, but if I were not aware of recent bicycle thefts in front of the logic department, I would know it.”

In a similar vein, it should be possible for Neo, according to contextualism, to properly assert the true proposition in his high standards context cuh (in which he is considering skeptical hypotheses) that he does not know that he has hands, but that he knows it (or would know it) in low standards contexts in which the hypothesis of being a brain in a vat is not salient: (5)

Neo in cuh: “I do not know that I have hands, but I do know it in cul” – or: “I do not know that I have hands, but if I were less sceptical, I would know that I have hands.”

Knowledge, Contextualism, and Moorean Paradox

27

After all, contextualism has it that “know” is an indexical or contextsensitive term. If the concept of knowledge is thus subject to some kind of degree, such as “flat”, “tall” etc., Anna’s and Neo’s claims in (4) and (5) should not be defective because such claims are, according to contextualism, semantically analogous with, for example, the following claim about tallness made by a person who uses “tall” in the sense of “tall for an NBA basket player”: (6)

“I am not tall, but I am tall for a first-grader”. (Or with semantic ascent: “The sentence: ‘I am tall’ is false, but if ‘tall’ meant ‘tall for a first grader’, this sentence would be true.”)

However, the knowledge claims made in (4) and (5) seem to be defective, whereas (6) is perfectly in order. Moreover, (4) and (5) are versions of Moorean paradoxes since they violate the contextualist rule of assertion. The knowledge claims made in (4) and (5) cannot be known in cuh by their respective speakers – which is shown by the following general theorem (MP-C*) that holds in contextualist accounts of knowledge: (MP-C*)

¬KhS(KlS φ  ¬KhSφ) for all subjects S, and propositions φ.

In order to prove (MP-C*), we need two further principles, (DI) and (F*), that contextualists surely embrace: Distribution of contextualized knowledge over implication: (DI) KiS(φ → ψ) → (KiSφ → KiSψ) for all subjects S, propositions φ and ψ, and standards i  {l,h}. Knowledge of the factivity of contextualized knowledge: (F*) KiS(KjSφ → φ) for all subjects S, propositions φ, and standards i, j  {l,h}. A subject knows in any context that knowledge is (in any context) factive. Proof of (MP-C*): Assume that KhS(KlS φ  ¬KhSφ). With (DC) we get (*) KhSKlSφ  KhS¬KhSφ. Because of (F*) it holds that KhS(KlS φ → φ), from which we get with (DI):

28

Elke Brendel

KhSKlSφ → KhSφ. From this and the first conjunct of (*), we get: KhSφ. From (F) and the second conjunct of (*), we get: ¬KhSφ – and thus the contradiction: KhSφ  ¬KhSφ.5 So, on the one hand, speaker S cannot (without violating the contextualist rule of assertion) sincerely assert, in a high standards context, that “S knows that p” is true in low standards context but false in high standards context. Even if this claim is true and is exactly what contextualism predicts, it cannot in principle be known by S in high standards context. But on the other hand, if “know” were a context-sensitive expression this is exactly what S should be able to properly assert. Analogous constructions with other context-sensitive gradable expressions do not result in self-contradictory assertions. Therefore, in contrast to what contextualists claim, there is an important semantic difference between “know” and such expressions. Philosophers have already provided some linguistic evidence against the contextualist assumption that “know” can be treated as an indexical or a context-sensitive term. For example, it has been pointed out that, unlike a pure indexical (such as “I”), “know” does not block inter-contextual disquotational reports (see Cappelen/Lepore 2005, 89). Furthermore, in contrast to gradable expressions, “know” does not seem to admit of degree modifiers (such as “really” or “very”) and comparative constructions (see Stanley 2004, 124f.). But this linguistic evidence against the gradability of “know” does not necessarily rule out the possibility of knowledge claims being sensitive to the varying epistemic standards operative in the context of the speaker. “Know” might still be subject to varying epistemic degrees, even though “know” is not gradable by a scale (see Stainton 2010, 152).6 However, my above argument against the semantic analogy between “know” and context-sensitive terms does not necessarily presuppose that knowledge comes into linearly ordered degrees. It just presupposes that the meaning of “know” is subject to varying epistemic standards – just as the 5

See also the so-called factivity problem in Brendel 2005, 2009. Robert Stainton, for example, argues that there are many other expressions that are subject to degrees but not gradable. Consider, for example, the term “vegetarian”: “John is vegetarian” can mean that John does not eat red meat, or that he does not eat meat or fish, or that he does not eat any animal products, etc. (see Stainton 2010, 152).

6

Knowledge, Contextualism, and Moorean Paradox

29

meanings of “tall” or “flat” are subject to varying standards of tallness or flatness. The argument shows that, in contrast to claims with such other context-sensitive terms, a speaker cannot, on pain of self-contradiction, utter in high standards context a true proposition about her knowledge in low standards context. 5. Relativist Accounts of Knowledge Relativist semantics of knowledge ascriptions, most prominently championed recently by John MacFarlane, share with contextualist accounts the idea that the truth values of knowledge ascriptions are sensitive to contextually dependent epistemic standards. Unlike contextualism, relativism holds that the truth values of knowledge ascriptions do not vary with the context of utterance but with the context of assessment. According to MacFarlane, knowledge ascriptions are thus assessment-sensitive. For MacFarlane “know” is not an indexical, i.e., a knowledge ascription expresses the same proposition in all contexts of utterance. However, its truth value depends on the epistemic standards that are operative in the context from which the knowledge claim is being assessed. MacFarlane uses the framework of Kaplan’s double index semantics, but adds to the context of utterance another context: the context of assessment cA. Whether a knowledge ascription expresses a true proposition depends, according to MacFarlane, not only on the world and time of the context of utterance, but also on the epistemic standards in play at the context of assessment (see, for example, MacFarlane 2005a, 223). Thus, MacFarlane treats “know” semantically on a par with, for example, aesthetic or taste predicates. Just as in aesthetic relativism or relativism with regard to taste predicates, where the truth values of the claims “Jonny Depp is attractive” or “Vegemite taste great” depend on the aesthetic leanings or the taste preferences of the person who assesses those claims, the truth value of a knowledge ascription depends, according to knowledge relativism, on the epistemic standards of the person who assesses the knowledge ascription. Like contextualism, relativism also seems to be able to account for the apparent context-dependent variability of knowledge-ascriptions due to varying epistemic standards. In cases where the epistemic standards of the assessor are identical to those of the knowledge ascriber or where the context of assessment is identical to the context of utterance, there is no difference in the semantic evaluation of knowledge ascriptions between contex-

30

Elke Brendel

tualism and relativism anyway. But there is a crucial difference between contextualism and relativism with regard to problems of context shift and the retraction of knowledge claims. Recall the example of Neo, the skeptic, who should, according to contextualism, counter-intuitively accept both that Anna does not know that she has hands and that Cypher, the anti-skeptic, spoke truthfully (in his lower standards context in which no skeptical hypotheses are considered) when claiming that Anna knows that she has hands. In relativist semantics for knowledge ascriptions this counterintuitive result does not occur since it is the assessor’s context, i.e., Neo’s context, which decides on the truth values of the knowledge claims. Thus, assessed from Neo’s context, Anna does not know that she has hands, and Cypher spoke falsely when claiming that Anna knows that she has hands. Consider again the example where Anna has just parked her bicycle in front of the logic department and thus claims to know that her bicycle is parked in front of the logic department. After she learned from her colleague Bart about recent bicycle thefts in front of the logic department, she no longer claims to know that her bicycle is parked there. According to contextualism, instead of retracting her former knowledge claim, Anna should accept both that she does not know that her bicycle is parked in front of the logic department and that she spoke truthfully when she claimed (in her former lower standards context) that she knew that her bicycle is parked in front of the logic department. Unlike contextualism, relativism is able to account for the phenomenon of knowledge retraction. Assessed from her high standards context (in which the possibility of bicycle thefts is salient), her former claim made in low standards context that she knows that her bicycle is parked in front of the logic department turns out to be false – in accordance to intuition. 6. Problems of Relativism How does relativism handle intuitions of disagreement? We have seen that contexualists run the risk of “explaining away” many forms of seemingly genuine disagreement if the disputants are in different attributor contexts with different standards of “know”. In such cases, the disputants use the term “know” with different meanings such that the proposition one disputant expresses is not the negation of the proposition her opponent expresses. So there is no disagreement at all. According to MacFarlane’s truth

Knowledge, Contextualism, and Moorean Paradox

31

relativism with respect to “know”, a knowledge ascription expresses the same proposition in every context of utterance. Therefore, if an anti-skeptic claims “S knows that p” and a skeptic responds by claiming “S does not know that p”, both disputants disagree about the same proposition, i.e., they disagree about whether S knows that p or not. However, their disagreement seems to be, according to relativism, faultless. There is no fact of the matter that decides whose claim is the correct one. Just as the sentence “Vegemite taste great” can vary in truth value when assessed from persons with different taste preferences, a knowledge ascription in relativist accounts can also vary in truth value when assessed from different persons with different epistemic standards. And just as opposing judgments about questions of taste seem to be without fault on anyone’s part, opposing assessments about knowledge are also faultless, according to relativism. Thus, while contextualist accounts dismiss many forms of seemingly real disagreement by pointing out that the disputants are using “know” with different meanings, in relativism these forms of disagreement turn out to be faultless if the disputed knowledge ascription is assessed from different contexts. In both accounts, intuitions of genuine real disagreement are explained away. Relativizing the truth of knowledge ascriptions to contexts of assessment is a more radical departure from objective conceptions of knowledge than relativizing its truth to contexts of utterance. In contextualist accounts, a knowledge claim made in a certain context of utterance always has a fixed truth value, depending, inter alia, on the epistemic standards operative in that given context of utterance. In relativist accounts, however, a knowledge claim “S knows that p” uttered at a context cu does not have a fixed truth value. It only gets its truth value when it is assessed from a context cA – and depending on this context cA, it can always change its truth value. I find this relativistic consequence hard to swallow. If “know” is an assessment-sensitive term it does not seem to make any sense to make judgments about knowledge ascriptions and engage in disputes about their truth. Aaron Zimmerman even claims that “[…] entities that are only relatively true are like rocks and chairs and bottles of beer – they are not the kinds of things that can be judged” (Zimmerman 2007, 342). We have seen so far that in treating “know” as an assessment-sensitive term, relativism has problems in accounting for intuitions of genuine real disagreement. Furthermore, it seems no longer to be rational to engage in a dispute about the truth of a knowledge ascription, since a knowledge as-

32

Elke Brendel

cription’s truth value can always shift due to a different context of assessment. But compared to contextualism, relativism seems to fare better with regard to cases in which an assessor in a high standards context assesses a knowledge ascription in a low standards context, and, in particular, with regard to retracting a former knowledge claim in light of new countervailing evidence. However, relativism faces problems in dealing with certain cases where the assessor is more ignorant than the speaker. Assume that Bart, after he had heard about the bicycle thefts in front of the logic department, claimed in his (high standards) context of utterance: “Anna does not know that her bicycle is parked in front of the logic department”. Now consider again Anna’s colleague Carl, who is completely ignorant about the bicycle thefts in front of the logic department. Assessed from his (low standards) context, the following sentence is true: “Anna knows that her bicycle is parked in front of the logic department and Bart spoke falsely when claiming that Anna does not know that her bicycle is parked in front of the logic department.” Given Carl’s epistemic situation, it is, of course, understandable from his perspective to claim that Anna knows that her bicycle is parked in front of the logic department – and that a person who denies this utters something false. But since the fact of bicycle thefts in front of the logic department is a relevant error possibility to Anna’s belief that her bicycle is parked in front of the logic department, Anna’s belief could have easily been wrong – and as a result, would not count as knowledge. So, despite the assessor’s ignorance about this relevant error-possibility, his claim that Anna knows that her bicycle is parked in front of the logic department, and that Bart falsely denied this, would simply be wrong. 7. Moorean Paradoxes for Relativists In addition to the above mentioned problems, it seems that relativists are also committed to a kind of Moorean paradox. Since “know” is an assessment-sensitive term in relativist accounts, it should be possible for an assessor to claim in high standards context of assessment that she does not know that p, but that she knows (or would know) that p, if she assessed p from a low standards context of assessment. So, for example, the following knowledge claims are true and should be properly assertable, according to relativism:

Knowledge, Contextualism, and Moorean Paradox

(7)

(8)

33

Anna in cAh: “I do not know that my bicycle is parked in front of the logic department, but I know it in cAl” – or “I do not know that my bicycle is parked in front of the logic department, but assessed from a perspective in which I were not aware of recent bicycle thefts in front of the logic department, I would know it.” Neo in cAh: “I do not know that I have hands, but I know it in cAl” – or “I do not know that I have hands, but assessed from a perspective in which skeptical scenarios are not salient, I would know that I have hands.”

Recall that in relativist accounts, “know” is an assessment-sensitive term, like an aesthetic term or a taste predicate. So, Anna’s and Neo’s claims in (7) and (8) are, according to relativists, analogous to the following claim made by a person who detests Vegemite: (9)

“Vegemite does not taste great, but assessed from the perspective of someone who loved Vegemite, it would taste great.” (Or with semantic ascent: “The sentence: ‘Vegemite taste great’ is false according to my taste preferences, but the sentence is true, assessed from the perspective of someone who loves Vegemite.”)

But whereas (9) can be properly claimed by a Vegemite hater, the knowledge claims in sentences (7) and (8) are Moorean paradoxes. It is easily seen that an analogous version of (MP-C*) – (MP-R*) – also holds for relativism: (MP-R*)

¬KhS(KlS φ  ¬KhSφ), for all subjects S, propositions φ, where “KiSφ” stands for “‘S knows that φ’ is true in the context of assessment cAi”, and i  {l,h}.

Thus, although the knowledge claims in (7) and (8) are true, they cannot be sincerely uttered by Anna and Neo in their respective contexts of assessment, since they cannot be known in these contexts of assessment – as stated in (MP-R*). So, whereas the contextualist versions of Moorean paradoxes show an asymmetry between “know” and indexical terms that are subject to certain degrees, the relativist versions of Moorean paradoxes show a substantial disanalogy between “know” and assessment-sensitive terms.

34

Elke Brendel

8. Subject-Sensitive Invariantism Subject-sensitive invariantism (SSI), defended by, for example, John Hawthorne (Hawthorne 2004) and Jason Stanley (Stanley 2005), is another important semantic account of knowledge ascriptions in recent epistemology. According to SSI, “know” is neither use-indexical nor assessmentsensitive. “Know” is rather sensitive to the subject’s practical situation at the time of knowing. The subject’s practical situation at the time of knowing is determined by the subject’s practical interests with respect to p, i.e., in particular, the subject’s practical costs of being wrong about the proposition in question. The more that hinges on the truth of p for the subject S at the time of knowing t, the higher the epistemic standards are that S has to satisfy in order for the knowledge ascription “S knows that p at t” to turn out true: “[T]he greater the practical investment one has in a belief, the stronger one’s evidence must be in order to know it […]” (Stanley 2005, 88). So, at any given time t, a knowledge ascription “S knows that p” has a fixed truth value due to the fixed practical interests of S at t. Therefore, “know” is a univocal knowledge relation in SSI. “Know” can only change its extension if S’s practical interests with respect to p change with the time of knowing. Stanley thus contends that “there is no specifically epistemological sense in which knowledge attributions are context-sensitive” (Stanley 2005, 90). According to SSI, “know” is just like other verbs that can have different semantic contents at different times, such as “sit”, “read”, “sleep” etc. Since in SSI knowledge ascriptions can change their truth values depending on the varying practical situation of the subject at different times of knowing, SSI can also account for the seemingly context-dependent variability of knowledge ascriptions. In DeRose’s “low stakes” bank case, for example, where the practical costs for the couple of being wrong about whether the bank will be open on Saturday are low, the wife’s claim that she knows that the bank will be open on Saturday is true, according to SSI. Whereas in the “high stakes” context in which it is very important for the couple to deposit their paychecks by Saturday the knowledge claim turns out to be false. Compared to contextualism and relativism, SSI does not have to explain away intuitions of real disagreement. According to SSI, if a person A claims: “S knows that p at t”, and if another person B denies this by claiming: “S does not know that p at t” only one of them makes a true assertion.

Knowledge, Contextualism, and Moorean Paradox

35

9. Problems of Subject-Sensitive Invariantism SSI is confronted with various problems. Jonathan Schaffer, for example, has pointed out that SSI cannot adequately account for the social role of knowledge as a means of transmitting information (Schaffer 2006). It seems that, as Schaffer claims, knowledge transmission of p is only successful in SSI if the practical interest in p of the recipient is not higher than the practical interest of the informant: “If testimony could only transmit knowledge to subjects with comparable stakes (or at least to hearers with no more at stake than testifiers), then our practice of relying on testimony would be troubled” (Schaffer 2006, 97). In knowledge self-ascriptions where the speaker and the subject are identical, a stakes-related contextualism and SSI converge. As a consequence, SSI runs into similar problems as contextualism with respect to the retraction of (first-person) knowledge claims. Assume, for example, that Bart at time t1 (let us say, December 2011) lives in Bonn, Germany, and believes that a particular nuclear power plant in India won’t ever explode. His evidence for this belief is based on the very good safety track record of this nuclear power plant and on the fact that some experts consider the earthquake risk in the area of the nuclear power plant as relatively low. Bart lives far away from this Indian power plant, he doesn’t have any friends or relatives in India and he does not care much about people from other countries he doesn’t know. Thus, given his practical situation at t1, Bart’s claim at t1: “I know that this Indian nuclear power plant won’t ever explode” is true, according to SSI (let us also assume that this Indian power plant won’t indeed explode). Now, assume that in January 2012 Bart gets a surprising job offer from an Indian company, which happens to be seated right next to this nuclear power plant, and happens to move to India by the end of the month. After his move, his evidential situation with respect to the safety of this power plant has not changed. He still believes that this Indian power plant will be safe. Nevertheless, since his practical interest in the safety of the power plant has changed dramatically, he now claims at t2 (end of January 2012 when he lives right next to this power plant): “I do not know that the power plant won’t ever explode.” According to SSI, Bart should now at time t2 accept both, that he does not know that the nuclear power plant won’t ever explode, and that he spoke truthfully when he claimed at t1 that he knew that the power plant won’t ever explode. But this seems to be counterintuitive. It would be much more plau-

36

Elke Brendel

sible if Bart at t2 rejected his former knowledge claim at t1 by claiming: “I do not know that this nuclear power plant won’t ever explode, and when I earlier claimed that I know it, I was wrong.” Furthermore, SSI makes counterintuitive predictions in certain cases where the knowledge ascriber refers to the knowledge of a subject whose stakes with respect to the truth of a given proposition are lower than that of the ascriber. Consider, for example, the case where Anna, who lives next to the above mentioned nuclear power plant, refers to the following knowledge claim made my Bart, who lives in Bonn, Germany: “I know that this Indian nuclear power plant won’t ever explode.” Recall that, according to SSI, it is always the subject’s practical situation on which the truth conditions of a knowledge ascription depend. Thus, if Anna claimed: “Bart knows that this Indian nuclear power plant won’t ever explode”, her claim should be true – disregarding the fact that Anna’s practical interests with respect to the safety of this power plant are quite different from Bart’s. 10. Moorean Paradoxes for Subject-Sensitive Invariantists Last, but not least, we can also construct certain forms of Moorean paradoxes for subject-sensitive invariantists. Consider the following claim made by Bart at a time t2, where he lives next to the nuclear power plant: (10)

Bart at t2: “I do not know that this Indian nuclear power plant will never explode, but if I lived in Bonn, Germany, I would know it.”

If “know” were like a verb that can receive different semantic values at different times, (10) should be as properly assertable as the following sentence (11): (11)

Anna (at 5 a.m. on a Tuesday morning during spring term): “Now, I am not standing in front of my logic class, but in 4 hours I will be standing in front of my logic class” – or: “I am not standing in front of my logic class, but if it were 9 a.m. I would be standing in front of my logic class.”

But, unlike (11), Bart’s claim in (10) clearly sounds very odd. The reason for this is, that (10) is an instance of a Moorean paradox in SSI, i.e., a

Knowledge, Contextualism, and Moorean Paradox

37

claim that is true, but nevertheless cannot be sincerely uttered by the subject. Although, “know” is neither use-indexical nor assessment-sensitive, the following general principle (MP-SSI*) holds in SSI (and can be proved analogous to (MP-C*)): (MP-SSI*) ¬KhS(KlSφ  ¬KhSφ), for all subjects S, propositions φ, where “KhSφ“ stands for “‘S knows that φ at a time where S’s stakes with respect to φ are high’ is true”, and “KlSφ“ stands for “‘S knows that φ at a time where S’s stakes with respect to φ are low’ is true”. Since a subject S cannot know in a high stakes context that she does not know that p, but knows it in low stakes context, she cannot, without violating the knowledge rule of assertion, sincerely assert that she does not know in high stakes context, but knows in low stakes context that p. So, SSI runs into similar problems of Moorean paradoxes as contextualist and relativist accounts of knowledge. 11. Conclusion The above arguments suggest that neither contextualism nor relativism nor SSI provide completely satisfying semantic accounts of knowledge ascriptions. We have seen, in particular, that there are substantial differences between “know” and use-indexical as well as between “know” and assessment-sensitive terms. Contextualism cannot adequately account for intuitions of disagreement and for the phenomenon of knowledge retraction. Contextualism makes counterintuitive predictions in cases where the knowledge ascriber is in context with lower standards than the standards ofthe subject. Furthermore, contextualism is confronted with the problem of Moorean paradoxes. While relativism seems to fare better with respect to knowledge retraction, it also has problems in accounting for intuitions of real disagreement. In addition, the ability to judge knowledge ascriptions seems to get lost, since knowledge ascriptions are only relatively true and can always change their truth values when assessed from different perspectives. Relativism has counterintuitive consequences in cases where the assessor is in a lower

38

Elke Brendel

standards context than the knowledge ascriber. Like contextualism, relativism is also committed to some kind of Moorean paradoxes. While SSI seems to be able to account for the intuition of real disagreement, SSI runs into similar problems of knowledge retraction as contextualist accounts. Furthermore, SSI cannot adequately account for cases in which a person A makes a judgment about a knowledge claim “S knows that p” where S’s stakes are lower with respect to p than A’s own stakes. Lastly, SSI is also confronted with problems of Moorean paradoxes. It seems to me that the apparent context-dependent variability of knowledge ascriptions might be better addressed in terms of the pragmatics of knowledge ascriptions. To be sure, stakes or salient error-possibilities can have an effect on our willingness to attribute or deny knowledge to others or to ourselves. Unless in cases of knowledge self-ascriptions where stakes or error-possibilities can have an effect on S’s belief that p, the epistemic standards that the speaker or assessor happens to have in their contexts of utterance or assessment, do not have a bearing on the truth value of knowledge claims. In a similar vein, Kent Bach argues that “[…] what varies is not the truth conditions of knowledge attributions but the knowledge attributions people are prepared to make. It is not the standards for the truth of knowledge attributions that go up but the attributor’s threshold of confidence regarding the relevant proposition” (Bach 2005, 85). So, for example, it seems that it simply does not matter whether the subject, the knowledge ascriber or the assessor is aware of the fact of bicycle thefts in order for the knowledge ascription “Anna knows that her bicycle is parked in front of the logic department” to turn out false. But, of course, if the speaker (or the assessor or the subject herself), after learning about the bicycle thefts in front of the logic department, is considering the possibility that Anna’s bicycle might have been stolen, their willingness to attribute the knowledge to Anna that her bicycle is still parked in front of the logic department decreases. So, a moderate invariantist semantic account of knowledge (as, for example, an externalist safety account of knowledge) combined with a pragmatic account that explains the context-dependent variability of knowledge ascriptions might be a more promising linguistic theory about knowledge ascriptions than contextualism, relativism, or subject-sensitive invariantism. However, how such a theory is spelled out in detail is a separate and highly challenging topic in its own right.

Knowledge, Contextualism, and Moorean Paradox

39

References Bach, K., 2005: The Emporer’s New ‘Knows’, in: Contextualism in Philosophy, ed. G. Preyer and G. Peter, Oxford: Oxford University Press, 51-89 Brendel, E., 2005: Why Contextualists Cannot Know They Are Right: Self-Refuting Implications of Contextualism, Acta Analytica 20, 38-55 Brendel, E., 2009: Contextualism, Relativism, and Factivity, in: Reduction – Abstraction – Analysis, ed. A. Hieke and H. Leitgeb, Frankfurt am Main: Ontos, 403-416 Cappelen, H. and E. Lepore, 2005: Insensitive Semantics: A Defense of Semantic Minimalism and Speech Act Pluralism, Oxford: Blackwell Cohen, S., 1988: How to be a Fallibilist, Philosophical Perspectives 2, 91-123 Cohen, S., 2000: Contextualism and Skepticism, Philosophical Issues 10, 94-107 DeRose, K., 1992: Contextualism and Knowledge Attribution, Philosophy and Phenomenological Research 52, 913-929 DeRose, K., 2009: The Case for Contextualism, Oxford: Oxford University Press Hawthorne, J., 2004: Knowledge and Lotteries, Oxford: Clarendon Press Kaplan, D., 1989: Demonstratives: An Essay on the Semantics, Logic, Metaphysics, and Epistemology of Demonstratives and Other Indexicals, in: Themes from Kaplan, ed. J. Almog, J. Perry, and H. Wettstein, Oxford: Oxford University Press, 481-564 Kompa, N., 2002: The Context Sensitivity of Knowledge Ascriptions, Grazer Philosophische Studien 64, 1-18 Lewis, D., 1996: Elusive Knowledge, Australasian Journal of Philosophy 74, 549-567 MacFarlane, J., 2005a: The Assessment Sensitivity of Knowledge Attributions, in: The Oxford Studies in Epistemology (Vol. 1), ed. T. Gendler and J. Hawthorne, Oxford: Oxford University Press, 197-233 MacFarlane, J., 2005b: Making Sense of Relative Truth, Proceedings of the Aristotelian Society 105, 321-340 MacFarlane, J., 2007: Relativism and Disagreement, Philosophical Studies 132, 17-31 Moore, G.E., 1962: Commonplace Book: 1919-1953, London: Allen & Unwin Schaffer, J., 2006: Against Subject-Sensitive Invariantism, Philosophical Studies 127, 87-107 Stainton, R., 2010: Contextualism in Epistemology and the Context-Sensitivity of ‘Knows’, in: Knowledge and Skepticism (Topics in Contemporary Philosophy, Vol. 5), ed. J.C. Campbell, M. O’Rourke, and H. Silverstein, Cambridge, Mass.: MIT Press, 137-163

40

Elke Brendel

Stanley, J., 2004: On the Linguistic Basis for Contextualism, Philosophical Studies 119, 119-146 Stanley, J., 2005: Knowledge and Practical Interests, Oxford: Oxford University Press Williams, M., 2005: Knowledge, Reflection, and Sceptical Hypotheses, in: Contextualisms in Epistemology, ed. E. Brendel and Chr. Jäger, Dordrecht: Springer, 173-201 Zimmerman, A.Z., 2007: Against Relativism, Philosophical Studies 133, 313-348

Context-Sensitivity without Indexicality? WAYNE A. DAVIS, Georgetown University

John MacFarlane (2007, 2009) argues that “nonindexical contextualism” has significant advantages over the standard indexical form. MacFarlane maintains that the extension of an expression may depend on an epistemic standard variable even though its content does not. Focusing on ‘knows,’ I will argue against the possibility of extension dependence without content dependence when factors such as meaning, time, and world are held constant, and show that MacFarlane’s nonindexical contextualism provides no advantages over indexical contextualism. The discussion will shed light on the definition of indexicals as well as the meaning of ‘knows.’1 I. Context Sensitivity vs. Indexicality As MacFarlane (2007, 247-248; 2009, 232) defines his terms, an expression is context-sensitive iff its extension at a context of use depends on features of the context, and indexical iff its content at a context depends on contextual features. MacFarlane takes contexts of use to be characterized by a speaker, a place, a time, a world, an epistemic standard, and possibly other factors. He allows that the world characterizing a context may be non-actual. ‘Indexical’ is commonly used as a synonym of ‘context sensitive.’ Indeed, indexicals are usually defined as “linguistic expressions whose reference shifts from context to context” (Braun 2007, 1), with ‘reference’ and ‘extension’ used interchangeably. But it is hard to find any term in natural language whose extension does not depend on and vary with the context in some way. First, all contingent expressions have different extensions when used in different possible worlds. Thus the extension of ‘the 43rd president of the United States’ is George W. Bush at any context in the actual world, 1

I thank the audience at the 34th International Wittgenstein Symposium, especially John Greco and Stewart Cohen, for comments that helped improve my argument significantly. For an expanded version of this article, see Davis 2012.

42

Wayne A. Davis

but Al Gore at any context in a world in which Gore won the 2000 presidential election. Second, non-eternal terms have different extensions at different times. Thus when used in 1920, the extension of ‘airplane’ did not include any jets, whereas its extension in 2010 included many jets. Third, every natural language expression I’ve ever examined is ambiguous, in the broadest sense of having multiple meanings or interpretations. The extension of an ambiguous term at a context depends on the sense with which it is used in that context. Thus the extension of ‘plane’ in an actual context depends on whether it means “airplane, ” “wood plane,” or “geometric plane” on that occasion. Since ‘plane’ could have had, or could come to have, different meanings, its extension in a non-actual context depends on what meaning it has in that world at the time. The terms ‘context sensitive’ and ‘indexical’ are customarily applied to terms like ‘I,’ ‘here,’ ‘now,’ ‘he,’ ‘local,’ and ‘foreign’ to denote the way in which they differ markedly from expressions like ‘plane’ and ‘the 43rd president of the United States.’ What is special about the terms commonly classified as indexical or context sensitive is that their extension varies with the context of use even when their linguistic meanings and the world and time of evaluation are held constant. The extension of ‘the 43rd president of the United States’ does not vary with the context when those three factors are held constant. As long as we hold fixed its meaning and the world it is evaluated in, ‘the 43rd president of the United States’ will have the same referent no matter which world it is used in. Can the extension of a term or sentence vary with the context of use without its content varying, when meaning, world, and time are held constant? MacFarlane answers “Yes.”2 MacFarlane further argues that epistemic contextualists would be on firmer ground if they claimed that what varies with the context is the extension of ‘know’ rather than its content―if they adopted a nonindexical contextualism. II. Content According to MacFarlane’s (2009, 232, fn. 2) definition, “the content of a sentence at a context is a proposition, and ... the content of a subsentential expression is the contribution it makes to the content of the sentences con2

MacFarlane cites Kompa 2002 and Ludlow 2005, 27 as possible antecedents.

Context-Sensitivity without Indexicality?

43

taining it.” He indicates that propositions are “what is said or asserted or believed” (2009, 248), and that the word ‘know’ expresses a relation, which has an intension. Relations and intensions are thus candidates for the contents words contribute to the contents expressed by sentences. MacFarlane follows Kaplan (1977, 494) in taking intensions to be functions from circumstances of evaluation to extensions. When we ask whether a sentence is true in a given context, the circumstance of evaluation includes the world within which that context occurs. We can also ask whether the sentence, as used in that context, is true in other circumstances. As used by me now, ‘I am here’ is true because it is true in the actual world; but it is not true in a world in which Wayne Davis is at the beach. Kaplan (1977, 512-3) was clear that when ‘I am here’ as I used it is evaluated in another world, we hold fixed the linguistic meaning and content it had in my context, and see whether that content is true at the other world. If factors other than worlds influence the extension of a sentence, then they need to be included in circumstances of evaluation. Kaplan included times as well as worlds, while allowing that additional parameters may be needed. MacFarlane’s view is that circumstances of evaluation include an epistemic standard parameter for epistemic terms. On contextualist views, the value of the epistemic standard parameter is set by the context of use. The thesis that extension may vary independent of content even when meaning, world, and time are held constant may now seem to follow directly from three premises: (i) circumstances of evaluation include parameters other than worlds and times; (ii) intensions are functions from circumstances of evaluation to extensions; and (iii) intensions are or represent contents. But an additional premise is necessary: (iv) the value of an intension function is sometimes affected by the values of parameters other than the world and time. That is, intension functions are not constant over the other parameters. Thus inclusion of an epistemic standard parameter in circumstances of evaluation will entail that the extension of an instance of ‘S knows p’ may vary independently of its content even when world and time are held constant only if the values of that parameter affect the truthvalue of the sentence. In order to isolate the effect of the standards parameter, we will focus on the actual world and the present time. On contextualism, the epistemic standard parameter is determined by factors such as the interests of the speaker, what is at stake for the speaker, and what is salient to the speaker. The problems confronting contextualism are due to the fact that it makes the truth-value of knowledge claims de-

44

Wayne A. Davis

pendent on features of the speaker’s context that are independent of the truth of ‘p.’ So the problems facing contextualism are evidence that the extension of ‘S knows p at t’ does not vary with such truth-independent features of the speaker’s context. III. Knowledge Claims The indexical contextualist believes that the extension of ‘S knows p at t’ varies from context to context because its content does. The thesis can be formulated well-enough for present purposes as follows: (1)

When used in a context C, ‘S knows p at t’ expresses the proposition that ‘p’ is true and S’s epistemic position with respect to ‘p’ at t meets e(C).3

MacFarlane characterizes e(C) as the epistemic standard “in play” at C. On indexical contextualism, despite having only three explicit argument places, the sentence ‘S knows p at t’ expresses a four-place relation in any context, the fourth relata e being supplied by the context. In the same way, sentences of the form ‘S knows p’ are well-formed with only two argument places because context supplies the time of utterance as the third relata via the present tense. Indexical contextualism faces a number of problems. Suppose Sam is in one context, and Barry another, and imagine them saying the following. (2)

Sam: Alice knows her car is in the driveway. Barry: Sam said that Alice knows her car is in the driveway.

Given what Sam said, Barry’s report seems perfectly correct. On indexical contextualism, Barry’s report is or at least might be false if Barry is in a high standards context while Sam is in a low standards context. In that case, Sam said Alice meets a low epistemic standard, but Barry may have described Sam as saying that she met a high standard. MacFarlane (2009, 240) believes that nonindexical contextualism can avoid this problem by saying the following: 3

This formulation is adapted from MacFarlane 2009, 236. For it to be plausible, we also need to assume that S’s epistemic position with respect to p meets e(C) only if S believes p, and is not in a Gettier situation.

Context-Sensitivity without Indexicality?

45

[T]he word “know” expresses the same relation at every context of use: the relation a subject S stands in to a proposition p and a time t just in case S knows that p at t. This relation, like others, has an intension: a function from circumstances of evaluation (world/epistemic standard pairs) to extensions. We can specify this intension roughly as follows: a subject S stands in the knowing relation to p at t at (w, e) iff p is true at (w, e) and S is in a strong enough epistemic position at w and t with respect to p to satisfy the standard e. (MacFarlane 2009, 236)

But if the truth-value of the proposition expressed by ‘S knows p at t’ in every context depends on a variable e that is independent of S, p, and t, then there is no unique relation a subject S stands in to a proposition p and a time t just in case S knows p at t. For S’s epistemic position at t with respect to p may meet the standard e(C1) without meeting the standard e(C2). The contextualist cannot maintain that ‘knows’ expresses a four-place relation between S, p, t, and e without being an indexicalist, holding that the value of e, and therefore the proposition expressed, is determined by features of the context of use. To make this concrete, suppose the bank will be open on Saturday, and suppose Alice has moderately good evidence that it will be. The time is the present, and the world is the actual world. Suppose Alan is talking about Alice in a context A with low standards, while Bob is talking about Alice in a context B with high standards. Then Alice meets the epistemic standards of A but not of B. Can we say whether Alice stands in the knowing relation to the proposition that her car is in the driveway? Not given the truthconditions MacFarlane provided. According to those truth-conditions, all we can say is that Alice stands in the knowing relation to that proposition and e(A), but does not stand in the knowing relation to that proposition and e(B). The intension MacFarlane has provided is that of a four-place relation between S, p, t, and e. So he has not described a form of contextualism according to which ‘knows’ expresses the same three-place relation in every context. To see the problem in another way, consider the following definition of a relation, in which q(x) is the measure of quantity q possessed by x. To make sure we rely on the definition provided, and not our antecedent understanding, I use the nonsense verb ‘grear.’ (3)

x grears y at t in q iff q(x) > 1.25q(y) at t.

We can thus say that Alan grears Bill at the present time in height, while Bill grears Alan in net worth. These truth-conditions define a four-place

46

Wayne A. Davis

relation, not a three-place relation. Consequently, Alan grears Bill at the present time makes no sense unless a quantity is understood from the context. We cannot say that ‘x grears y at t’ expresses the relation of x grearing y at a time t. MacFarlane says that ‘know’ expresses “the relation a subject S stands in to a proposition p and a time t just in case S knows that p at t.” His truthconditions may seem to tell us when S knows p at t given that they tell us when S knows p at t at any circumstance of evaluation (w, e). But (4)(b) cannot be validly inferred from (4)(a). (4)

(a) S knows p at t at (w, e). (b) S knows p at t.

This is particularly clear given that w may be a non-actual world. ‘Alice knows she has three hands at a world in which she has three hands’ does not entail that Alice knows she has three hands. But even when we let w be the actual world, there is no way to get from (4)(a) to (4)(b) on contextualism. There is no uniquely “correct” epistemic standard according to contextualism. If S knows p at t at e1 but not at e2, we have no basis for inferring that S either does or does not know p at t. IV. Conceptual Contents If MacFarlane maintains that propositions are composed of concepts rather than relations, he could hold that the verb ‘know’ expresses a concept that combines with a subject concept, a propositional concept, a temporal concept, and an epistemic standard concept to form the proposition that S knows p at t by e. What works best for MacFarlane’s uniformity claim is the thesis that ‘S knows p at t’ expresses the same complex of concepts on each occasion of use. MacFarlane would have to maintain that this complex is incomplete in the sense that it (i) does not have a truth-value and is not an object of belief in itself, but (ii) can be a part of propositions that are completed by the addition of concepts representing standards.4 There are two problems with this conceptual approach. First, as MacFarlane (2009, 240) notes following Stanley (2004: 138), if ‘S knows p 4

MacFarlane could fill this out by using the theory of thoughts and propositions I developed in Davis 2003.

Context-Sensitivity without Indexicality?

47

at t’ did express such an incomplete proposition, we would expect to be able to expand sentences of the form ‘S knows p at t’ by adding clauses that express the different standards concepts. But we never say things like (5). (5)

Alice knows her car is in the driveway by low standards, but not by high standards.

This sounds like an oblique way of saying that Alice does not really know. On the suggested version of nonindexical contextualism, though, we should be able to hear sentence (5) as reconciling the apparently contradictory claims in (6), just as we can hear ‘Alice is strong by little league standards but not by big league standards’ as reconciling the apparent contradiction between utterances of ‘Alice is strong’ and ‘Alice is not strong.’ (6)

Sam: Alice knows her car is in the driveway Barry: Alice does not know her car is in the driveway.

But we cannot. We hear the two claims in (6) as intractably contradictory. The second problem, not noted by MacFarlane, is that if what ‘S knows p at t’ expresses on all occasions is a propositional part that is not an object of belief and lacks a truth-value, then it is not what Sam said in (2), which is what Barry said he said. As (2) is understood, what Sam said is either true or false, and can be believed or disbelieved. On nonindexical contextualism, ‘Alice knows her car is in the driveway’ should be much like (7): (7)

Alice finished.

‘Alice finished’ is a complete sentence, unlike ‘Alice completed.’ There is a complex of concepts that ‘Alice finished’ expresses in every context. But that conceptual complex is not a proposition and has no truth value. ‘Alice finished’ is also used to express a proposition, but different ones on different occasions. ‘Alice knows her car is in the driveway’ does not need to be “completed” by reference to an epistemic standard the way ‘Alice finished’ needs to be completed by reference to an object. V. Sentence Occurrences MacFarlane says that the key to understanding nonindexical contextualism lies in a rule provided by Kaplan.

48

Wayne A. Davis

(8)

If c is a context, then an occurrence of φ in c is true iff the content expressed by φ in this context is true when evaluated with respect to the circumstance of the context. (Kaplan 1977, 522) (MacFarlane 2009, 236; see also MacFarlane 2007, 246)

Let φ be ‘Alice knows her car is in the driveway (at the present time).’ According to nonindexical contextualism, this sentence expresses the same content in every context. Let that content be K. We imagined that φ occurs in both A and B. MacFarlane wants us to suppose that K is true relative to e(A) and false relative to e(B). Why then should we say that the occurrence of φ in A is true just because the content K it expresses is true relative to e(A) when K is false relative to e(B)? On contextualism, no standard can be singled out as the correct standard. And on nonindexical contextualism, the meaning of φ has no indexical element referring to features of the context the way ‘I am here’ does. Moreover, what φ expresses entails nothing about A, B, or standards of evidence. How can the occurrence of a sentence―a mere string of letters or speech sounds―get a truth-value independent of its meaning and content? If the speakers in A and B meant the very same thing by φ, and used it to express exactly the same content, then how could one occurrence be true while the other is false? We noted in §III that ‘φ is true’ cannot be inferred from ‘φ is true at E’ when E contains a non-actual world or an epistemic standard parameter. Kaplan’s rule works fine for Kaplan provided we apply it to actual occurrences. Thus the following inference is valid when the circumstances of evaluation contain nothing but a time and a world: (9)

(a) Wayne Davis uttered ‘I am in Kirchberg’ on August 9, 2011. (b) The occurrence of ‘I am in Kirchberg’ in that context was true at the circumstances of the context. (c) ‫ ׵‬The occurrence of ‘I am in Kirchberg’ in that context was true.

The premises of (9) imply the conclusion because premise (a) entails that the world of the context is the actual world. The premises in (10) imply the conclusion for the same reason if the circumstances of evaluation contain nothing but a world and a time. (10)

(a) Sam uttered ‘Alice knows her car is in the driveway’ on August 9, 2011.

Context-Sensitivity without Indexicality?

49

(b) The occurrence of ‘Alice knows her car is in the driveway’ in that context was true at the circumstances of the context. (c) ‫ ׵‬The occurrence of ‘Alice knows her car was in the driveway’ in that context was true. But on nonindexical contextualism, the circumstances of evaluation contain an epistemic standard parameter. From the fact that a knowledge claim is true at a particular standard we cannot infer that it is true. So (10) is invalid on nonindexical contextualism. Let us assume for the sake of argument that the nonindexical contextualist could adopt Kaplan’s rule. Identify the utterance of an expression with the production of an occurrence, so that the utterance of φ in C is true iff the occurrence of φ in C is true, as MacFarlane (2009: 247, fn. 20; 248) maintains.5 The truth value of the occurrence of φ in C is presumably the truth value of the proposition expressed by φ in C. But then nonindexical contextualism with Kaplan’s rule would allow occurrences of (11) to be true when Sam is in a low standards context and the speaker is in a high standards context (cf. Kompa 2002). (11)

Sam’s utterance of ‘Alice knows her car is in the driveway,’ and the occurrence of that sentence in his context, are true; but Alice does not know her car is in the driveway.

In fact, such occurrences seem contradictory.6 Kaplan’s rule (8) would make nonindexical contextualism acquire all of the well-known problems for indexical contextualism that arise from its making the truth of an utterance depend on standards prevailing in the context of utterance. It would entail that Sam’s utterance of ‘Alice knows her car is in the driveway’ does not contradict Tom’s utterance of its negation when their contexts differ only because the standards in Tom’s context are higher than those in Sam’s. And it would allow that occurrences of (12) may be true when the speaker is in a low standards context while Alice is in a high standards context. 5

Things are less obvious if a sentence occurs more than once in a context, which is why I changed MacFarlane’s ‘an’ to ‘the.’ 6 See Cappelen & Lepore 2005, 105ff; MacFarlane 2005, 202ff; Hawthorne 2004, 107, fn. 125; Davis 2005, 39; 2007, 400-01; Bach 2005, 60; Stanley 2005, 54ff, 115, 119ff; Fantl & McGrath 2009, 180ff. Contrast DeRose 1992, 925-6; 2009, 207-12; Cohen 2005, 205-6; Blome-Tillmann 2008, 34, 37.

50

Wayne A. Davis

(12)

Alice knows that her car is in the driveway, but she does not have enough evidence to assert that her car is in the driveway.

Sentences like (12) sound as inconsistent as the others. To recap, the nonindexical contextualist cannot use Kaplan’s rule for sentence occurrences, because truth at a circumstance does not entail truth when the circumstance includes an epistemic standard. If Kaplan’s rule could be used, nonindexical contextualism would inherit many of the objectionable consequences of indexical contextualism. Without Kaplan’s rule, however, nonindexical contextualism provides no account of when knowledge claims are true or false. It is not possible for the truth-value of a sentence to vary with the context of use when it expresses the same proposition and is evaluated at the same world. References Bach, K., 2005: The emperor’s new ‘knows’, in: Contextualism in Philosophy: Knowledge, Meaning, and Truth, ed. G. Preyer & G. Peter, Oxford: Clarendon Press, 51-90 Blome-Tillmann, M., 2008: The indexicality of ‘knowledge’, Philosophical Studies 138, 29-53 Braun, D., 2007: Indexicals, in: Stanford Encyclopedia of Philosophy, ed. E. N. Zalta, Stanford, CA: 1-33 Cappelen, H. and J. Hawthorne, 2009: Relativism and Monadic Truth, Oxford: Oxford University Press Cappelen, H. and E. Lepore, 2005: Insensitive Semantics: A Defense of Semantic Minimalism and Speech Act Pluralism, Oxford: Blackwell Publishing Cohen, S., 1986: Knowledge and context, Journal of Philosophy 83, 574-583 Cohen, S., 1988: How to be a fallibilist, Philosophical Perspectives 2, 91-123 Cohen, S., 2005: Knowledge, speaker and subject, Philosophical Quarterly 55, 199212 Davis, W. A., 2003: Meaning, Expression, and Thought, New York: Cambridge University Press Davis, W. A., 2005: Contextualist Theories of Knowledge, Acta Analytica 20, 29-42 Davis, W. A., 2007: Knowledge claims and context: Loose use, Philosophical Studies 132(3), 395-438 Davis, W. A., 2012: On Nonindexical Contextualism, Philosophical Studies, forthcoming

Context-Sensitivity without Indexicality?

51

DeRose, K., 1992: Contextualism and knowledge attributions, Philosophy and Phenomenological Research 52, 913-929 DeRose, K., 2009: The Case for Contextualism: Knowledge, Skepticism, and Context, Vol. 1, Oxford: Oxford University Press Fantl, J. and M. McGrath, 2009: Critical Study of John Hawthorne’s Knowledge and Lotteries and Jason Stanley’s Knowledge and Practical Interests, Noûs 43, 178-192 Hawthorne, J., 2004: Knowledge and Lotteries, Oxford: Clarendon Press Kompa, N., 2002: The context sensitivity of knowledge ascriptions, Grazer Philosophische Studien 64, 1-18 Kaplan, D., 1977: Demonstratives, in: Themes from Kaplan, ed. J. Almog, J. Perry, and H. Wettstein, Oxford: Oxford University Press (1989), 481-563 Lewis, D., 1970: Anselm and Actuality, Nous 4, 175-188 Ludlow, P., 2005: Contextualism and the new linguistic turn in epistemology, in: Contextualism in Philosophy, ed. G. Preyer and G. Peter, Oxford: Oxford University Press, 11-50 MacFarlane, J., 2005: The assessment sensitivity of knowledge attributions, in: Oxford Studies in Epistemology, Vol. 1, ed. T. Gendler & J. Hawthorne, Oxford: Oxford University Press, 197-233 MacFarlane, J., 2007: Semantic minimalism and nonindexical contextualism, in: Context-Sensitivity and Semantic Minimalism, ed. G. Preyer and G. Peter, Oxford: Oxford University Press, 240-250 MacFarlane, J., 2009: Nonindexical contextualism, Synthese 166, 231-250 Stanley, J., 2004: On the linguistic basis for contextualism, Philosophical Studies 119, 119-146 Stanley, J., 2005: Knowledge and Practical Interests, Oxford: Oxford University Press

Non-Relativist Contextualism about Knowledge MARCUS WILLASCHEK, University of Frankfurt/Main

1. Introduction In epistemology contextualism is the view that the epistemic standards for determining the truth of knowledge-attributions vary with the context. Epistemic standards specify what S’s epistemic position with respect to p must to be like for S to know that p. (The term “epistemic position” here is meant to be neutral with respect to epistemic internalism and externalism. As will become apparent, my own favored account combines both internalist and externalist elements.) Contextualism thus defined can be developed in a variety of different ways. Currently, the most discussed version of contextualism is what I will call Semantic Contextualism. Different versions of this position have been defended, e.g., by Steward Cohen, Keith DeRose and David Lewis. (For classical statements of this view, see Cohen 1988, DeRose 1995, Lewis 1996). Glossing over many differences and details, we can characterize Semantic Contextualism very roughly as follows: Semantic Contextualism is primarily an account of the semantics of the verb “to know” and only secondarily, and indirectly, an account of knowledge. The context for determining the truth of knowledge-attributions is not the subject’s context, but the attributor’s. Depending on which errorpossibilities are salient in the context of attribution, different sets of errorpossibilities have to be eliminated in order for the subject to know; thus, the strength of epistemic standards varies from high to low. The salience of error-possibilities can be changed in a number of ways e.g., by raising certain error-possibilities or by changing what is ‘at stake’. Hence, on this account, the truth-value of knowledge-ascriptions as well as the epistemic standards varies with the conversational context. Semantically, these truthvalue-variations are accounted for by claiming that the meaning of “knows” contains a hidden indexical element. A standard criticism of Semantic Contextualism is that it implies a kind of relativism about knowledge attributions according to which for same S, t

54

Marcus Willaschek

and p, “S knows (at t) that p” can be true in one context and false in another. For instance, if I claim in an everyday context to know that I have two hands, and in a ‘skeptical’ context (e.g. in an epistemology class) someone else claims that I don’t know this, both claims may be true. This result is counterintuitive, since it denies what seems correct, namely that there is a contradiction between the claims “S knows that p” and “S does not know that p” even across contexts. However, if the truth of knowledgeascriptions depends on the context of attribution, this kind of relativism about knowledge is unavoidable. The only way to avoid it while retaining some version of contextualism would seem to be to focus on the subject’s context instead. This is what I suggest we should do. The kind of Non-Relativist Contextualism I now would like to introduce is a version of a relevant-alternatives-account where relevance is determined, in part, by the subject-context. The view I propose here is influenced by the work of Peirce, Wittgenstein and Austin, and more directly, by that of Michael Williams (cf. Williams 2001), but it differs from Williams’ view in various respects, including the fact that Williams’ position also seems to have relativist consequences, which my account seeks to avoid. I will proceed first by giving a brief sketch of my account as whole (section 2 – for a more detailed presentation, cf. Willaschek 2007.) I will then consider three widely discussed epistemological puzzle cases (the zebra case, the airport case and the skeptic), all three of which have been claimed by Semantic Contextualists to speak in favor of their view, and argue that Non-Relativist Contextualism does better in accounting for them than Semantic Contextualism (section 3). 2. Non-Relativist Contextualism The view I want to defend has four main elements. First, there is contextualism: the view that epistemic standards vary depending on different contextual factors. These contextual elements can be subsumed under two headings: (i) the subject’s physical, biological, geographical, social etc. environment and (ii) an epistemic practice. Epistemic standards vary along various dimensions (including strength, methods, relevance, kinds of evidence etc.) for both of these kinds of contextual factors. I will say more about this below. Second, there is an anti-skeptical condition on the relevance of error-possibilities, namely that one requires a positive reason, not just a logical possibility, for an error-possibility to count as relevant. What

Non-Relativist Contextualism about Knowledge

55

is required, in other words, is a reason to believe that the error-possibility might in fact obtain. Third, epistemic justification follows a default and challenge-pattern: if there are no relevant error-possibilities in a given situation, S is not required to possess evidence in favor of her belief in order to have knowledge. Rather, S has knowledge by default. Only if there are challenges to p based on relevant error-possibilities, is S required to be able to rule out these error possibilities in order to know that p. And finally, anti-relativism: Even though the standards for knowledge vary with the context (insofar as knowledge depends on the subject’s environment and the error-possibilities that must be ruled out, and what counts as ruling them out which is dependent on the relevant epistemic practice), each individual knowledge-ascription is appropriately evaluated relative to exactly one context, namely the subject’s factual environment and the strictest epistemic practice applicable in that context. Let me now say a bit more about these two kinds of contextual factors. An epistemic practice is a communal practice of assigning epistemic status, in particular the practice of ascribing knowledge and epistemic justification. This can be either an expert practice specific to a particular domain or subject matter (e.g. the epistemic practice of biology, law, or stampcollecting) or it can be our general non-specific lay-practice. (For an epistemic practice to be an expert practice it must generally be considered successful in its knowledge attributions; hence, for instance, astrology is not an expert practice.) Since epistemic practices will not play a major role in the three cases discussed below, I will not expand on this claim here. Let me just point out that this plurality of epistemic practices does not imply relativism about knowledge-ascriptions because of what might be called the demandingness of knowledge: If an error-possibility is relevant according to at least one epistemic practice, it is relevant for knowledge; if an error-possibility has not been ruled out according to at least one epistemic practice, then it has not been ruled out. This means that if a knowledge claim concerns a subject matter for which there are experts, it is to be evaluated according to the standards of that expert practice. If the knowledge claim is whether there is (a bottle of) milk in the fridge, the appropriate standards will be those of our common sense everyday practice of epistemic appraisal; knowing that there is a bottle of milk there normally requires no more than opening the door of the fridge and looking inside. By contrast, if the knowledge claim is about whether there is a specific kind of germ in the fridge, this requires methods of investigation available only to

56

Marcus Willaschek

experts (the microbiologist, say, or specialists in household hygiene). So if I claim to know that a specific germ is present in our fridge, I must either be an expert myself or base my claim on the testimony of experts. Assuming that expert practices are typically stricter – more demanding – than our all-purpose everyday practices, we can say that each knowledge claim is to be evaluated according to the strictest standards that can be appropriately applied to it. The reason that the demandingness of knowledge does not lead to skepticism has to do with the other contextual factor involved, viz. the subject’s environment. Whether a given error-possibility is relevant also depends on facts about the subject’s specific circumstances. In particular, for an errorpossibility to be relevant, there must be a reason to believe that the errorpossibility might in fact obtain. This condition permits both an externalist and an internalist reading (both of which capture something correct): An error possibility can be relevant either if there is a reason, based on facts, to believe that the error possibility might in fact obtain, or if the subject rationally believes that there is such a reason (or both). What counts as a reason for believing that an error possibility might obtain, will often depend on the epistemic practice involved. Thus, error-possibilities that may not seem relevant according to our ordinary lay-practice may well be relevant according to the standards of some science or craft. Consider an example: Assume that I hold the true belief that there is an oak tree before me and that I believe that it is an oak tree because of the shape of the tree’s leaves. Assuming that I can distinguish an oak leaf from an elm, beach etc. leaf, I think in general I could be said to know that there is an oak tree before me. But now imagine that there is some other kind of tree, a pseudo-oak, the leaves of which look exactly like those of oak trees. Further assume that only botanists can distinguish between this pseudo-oak tree and real oak trees. Do I still know that the tree in front of me is an oak? This depends, among other things, upon whether the possibility that it is a pseudo-oak is relevant according to the standards of botany, which in turn will depend on whether pseudo-oaks grow around here at all, whether they are common or nearly extinct etc. For instance, if pseudo-oaks are indigenous to Scandinavia and do not flourish where I am (at the Mediterranean, say), the botanist may well judge that the chances that the tree I’m standing in front of is a pseudo-oak are too remote to be relevant. In this case, I can tell that the tree is an oak tree by the shape of its leaves, whether or not I know about pseudo-oaks. If, however, the possibility that the tree

Non-Relativist Contextualism about Knowledge

57

before me is a pseudo-oak is relevant according to botanist’s standards, then I don’t know that it is an oak-tree unless I can rule out that it’s a pseudo-oak. So in this case, my having knowledge depends on the epistemic standards of the relevant experts plus certain facts about my environment. To be sure, even if the pseudo-oak possibility was relevant, most people would mistakenly ascribe knowledge to me, given that they, too, do not know about pseudo-oaks. But there is a simple, and quite plausible, errortheory available for this kind of mistake: People mistakenly ascribe knowledge when there are relevant error-possibilities they don’t know about; this is in fact a very common phenomenon. This must suffice as a very rough characterization of the Non-Relativist Contextualism I wish to defend. I will now compare it to Semantic Contextualism by very briefly discussing three classical puzzle cases from the epistemological literature. Semantic contextualists have argued for their view in part by claiming that it can handle these cases in a satisfactory way. As I will argue now, this is not obviously true. In any case, the NonRelativist Contextualism just outlined can give a better account of them than Semantic Contectualism. (Changes in context resulting from different epistemic practices will not play a role in the following discussion; what does matter in these cases though are changes in the subject’s environment.) 3. The three cases The zebra case Alfred sees a zebra in a pen at the zoo. Bertrand asks whether Alfred knows what kind of animal this is. Alfred reads the sign at the pen, which reads “Zebra” and says: “Yes, I know. It’s a zebra.” Bertrand asks: “But do you know that it’s not a painted mule?” – Alfred answers: “No, I don’t know that it’s not a painted mule” (adapted from Dretske 1970). Dretske’s original analysis was that Alfred knows that it’s a zebra, even though he does not know that it is not a painted mule. Famously, Dretske argued that this example shows that knowledge is not closed under known implication. Semantic Contextualism is partly motivated by the desire to avoid this consequence. According to the semantic contextualist, “Alfred knows that it’s a zebra” is true relative to a context of attribution in which the possibility that the animal is a painted mule is not salient; and since

58

Marcus Willaschek

one’s epistemic position with respect to the possibility that it is not a painted mule is at least as strong as one’s epistemic position with respect to the possibility that it is a zebra, in such a context, one also knows that it’s not a painted mule (cf. DeRose 1995). But as soon as that possibility is salient, “Alfred knows that it’s a zebra” is false, since Alfred would have to be able to rule out that it’s a painted mule, which he can’t (at least not without taking further steps inappropriate for a visitor at the zoo). In this more demanding context, Alfred neither knows that the animal is a zebra nor that it is not a painted mule. Since in both contexts, Alfred knows that it’s a zebra if and only if he knows that it’s not a painted mule, closure is saved, but, as we have seen before, at the price of relativism, since now it is both true that Alfred knows that it’s a zebra and that he does not know that it’s a zebra. By contrast, on the Non-Relativist Contextualism I have sketched, the case turns out to be severely underdescribed. Whether Alfred knows that the animal is a zebra depends, among other things, on whether there is a reason to believe that the animal might be a painted mule. The story we are told is silent about this possibility. Assuming that there is no such reason, the possibility that the animal is a painted mule is not relevant and thus does not have to be ruled out. Put differently, if there is no reason to distrust the sign at the pen, then Alfred knows that the animal is a zebra. By contrast, if there are facts that constitute a positive reason to assume that the animal might in fact be a painted mule (e.g. the zoo being in a financial crisis, the zoo director having being sued for fraud, the animal behaving strangely for a zebra, wet paint dripping from its side etc.), then this may well be a relevant error-possibility and Alfred does not know that the animal is a zebra as long as this possibility hasn’t been ruled out. Remember that the mere logical possibility that the animal is a painted mule, on my account, does not constitute a reason for believing that it might in fact be a painted mule. To be sure, the concept of a reason to believe that some error-possibility might obtain is vague. There are clear cases in which there are such reasons, and equally clear cases in which there are not, and there are many cases in between. But I don’t think it is implausible to say that our concept of knowledge is vague in just this way. So this kind of vagueness does not tell against an account of knowledge in terms of relevant error-possibilities and against an account of relevant error-possibilities in terms of reasons for believing that an error-possibility might obtain.

Non-Relativist Contextualism about Knowledge

59

Assuming that Alfred knows that the animal is a zebra, does he also know that it is not a painted mule? Well, according to Non-Relativist Contextualism, Alfred does know this: The only error-possibility with respect to the claim “The animal is not a painted mule” that is relevant is the possibility that the animal is a painted mule. But if Alfred knows that the animal is a zebra, this means either that the painted-mule-possibility is not relevant in the given context or that Alfred can rule out that possibility. So in either case, if Alfred knows that the animal is a zebra, Alfred knows that the animal is not a painted mule. Closure is saved here in much the same way as it is by the semantic contextualist: In both kinds of context, Alfred knows that it’s a zebra if and only if he knows that it’s not a painted mule. – Note that in our example Alfred does not infer, in quasi-Moorean fashion, “The animal is not a painted mule”, from “The animal is a zebra” (an inference most philosophers do not consider to be knowledgeconferring). Rather, the same aspects of Alfred’s epistemic position that allow him to know that the animal is a zebra (namely, either that the possibility that it’s a painted mule is not relevant or that this possibility can be ruled out), also allow him to know that it is not a painted mule. The airport case Anne and Beth are at the airport waiting to board a plane to New York. They wonder whether their flight will have a stopover in Chicago, where they have an urgent business appointment at the airport. They overhear Carl saying to David: “I know that the flight stops in Chicago, I’ve checked the itinerary”. Now Anne says to Beth: “The itinerary might contain a misprint; or they might have changed the schedule at the last minute. So Carl doesn’t know that the flight stops in Chicago” (adapted from Cohen 1999). According to Semantic Contextualism, Carl is correct in claiming to know that the plane stops in Chicago relative to a low-stakes context such as Carl’s, but Anne is correct in claiming that Carl does not know this since hers is a high-stakes context in which the standards for knowledge are more demanding than in a low-stakes one. Again, on the Semantic Contextualist account, we have the relativist consequence that it is both true that Carl knows that the flight stops in Chicago and that he does not know this. On the non-relativist version of contextualism I’ve proposed, the case is severely underdescribed. First of all, nothing in Cohen’s original story indicates that Carl in fact is in a low-stakes context. For all we know, it may

60

Marcus Willaschek

be more important to him to stop in Chicago than it is for Anne and Beth. The fact that, for the example to work, nothing needs to be said about the stakes for Carl already suggests that the question of what is at stake might not be as relevant as the semantic contextualist thinks. In fact, I think that in this case, nothing at all depends on how much is at stake. Here is why: Imagine, first, that nothing important is at stake for Anne and Beth, but both correctly believe that it may well be the case that their flight schedule has been changed in the very last minute. It seems that they still would be correct in denying knowledge to Carl. Now imagine that it is very important for them to stop in Chicago, but assume that a misprint in Carl’s itinerary or a last-minute change in their schedule is no more likely, say, than that their plane crashes, and that Anne and Beth know this. It seems to me that in this situation, it would be correct for them to accept Carl’s knowledge-claim based on the itinerary even though much is at stake for them. And similarly for Carl himself: If there is no particular reason to doubt the itinerary, then Carl knows that the flight stops in Chicago; if there is such a reason, he must be able to rule it out in order to know this; and both of these claims hold quite independently of what is at stake for him. Hence, contrary to what both semantic contextualists and subject-sensitive invariantists (cf. Hawthorne 2004; Stanley 2005) claim, whether Carl has knowledge or not depends only on which error-possibilities are relevant and not on what is at stake. According to my non-relativist version of contextualism, which error-possibilities are relevant in turn depends on facts about the subject’s environment, for instance on the reliability of the itinerary, the likelihood of a last-minute change in the schedules etc. Note that according to my account, people who disagree about these facts – facts about the subject’s environment – will also disagree about whether Carl has knowledge or not, which seems to be the correct result. Semantic Contextualism, as such, cannot account for this phenomenon, since these are facts about the subject’s context and not about the context of attribution. The skeptic’s case Adam looks at his hands and then claims to know to he has two hands. Bill accepts Adam’s knowledge-claim. Then Chuck objects that they all might be brains in a vat and therefore Adam does not know that he has two hands. According to Semantic Contextualism, Bill is correct in ascribing knowledge to Adam. But once Chuck mentions the possibility that Adam is a

Non-Relativist Contextualism about Knowledge

61

brain in a vat, a possibility Adam cannot rule out, Chuck is equally correct in denying that Adam knows he has two hands. Semantic contextualists pride themselves on their being able to account for the prima facie plausibility of skeptical arguments. Their account is simple: Skeptical arguments seem cogent because, in a skeptical context, they are valid and sound; hence, their conclusions are true. In other words, Semantic Contextualism simply gives the game away to the skeptic. According to the semantic contextualist, the skeptic is right to deny knowledge by invoking skeptical scenarios. I think this concedes too much to the skeptic. First note again that on my account, the case, again, is underdescribed. Everything depends on whether there is a reason to believe that the error-possibility Chuck mentions might be a real one. There would be such a reason for doubt, for instance, if putting people’s brains in vats and having them believe that their lives continued on as before was a common medical treatment or a generally recognized government scheme, and there was reason to believe that Adam might have been subjected to this kind of treatment. But assuming that this is not the case, the possibility Chuck mentions is not relevant and hence Adam need not be able to rule it out in order to know he has two hands. Again, allowing that Adam knows that he is not a brain in a vat saves closure. Assuming that Adam is not a brain in a vat, he knows this simply by virtue of the fact that in the given situation, his being a brain in a vat is not a relevant error-possibility. The semantic contextualist can hardly object to this, since he makes a similar move to account for closure in the low-standards scenario of the example. How can the non-relativist contextualist account for the apparent cogency of skeptical arguments? The short version of an answer to this question is this: By pointing to the fact that in doing philosophy, we abstract from the rich and complicated contexts on which the relevance of errorpossibilities depends. I think that it is no coincidence that all three cases we have considered turn out to be underdescribed. On the account advanced here, in our real-life ascriptions of knowledge we rely on a broad range of contextual factors in order to determine which error-possibilities are relevant and which are not. Since we abstract from these factors in philosophy (or rather, in traditional epistemology), all error-possibilities seem equally relevant; and that’s why skeptical arguments can seem cogent. Admittedly, this answer provides only the barest outline of a diagnosis of skeptical arguments, but I think that it is a more promising route than an account that

62

Marcus Willaschek

simply lets the skeptic win whenever he cares to trump his opponent’s claims by invoking hypothetical skeptical scenarios. 4. Conclusion I conclude that Non-Relativist Contextualism not only avoids relativism, but is also more successful than Semantic Contextualism in dealing with all three of the puzzle cases discussed in the literature. A version of contextualism that focuses on the subject’s context instead of the context of attribution can avoid relativism because there is only a single context for the subject whereas there are potentially many contexts of attribution. A version of the relevant-alternatives-account that requires positive reasons for alternatives to be relevant can avoid skepticism in a way that is much closer to our actual epistemic practices than Semantic Contextualism. The views that have inspired his type of contextualism have long been neglected in favor of Semantic Contextualism. I think it is time to reconsider them.1 References Cohen, S., 1988: How to be a Fallibilist, Philosophical Perspectives 2, 91-123 Cohen, S., 1999: Contextualism, Skepticism, and the Structure of Reasons, Philosophical Perspectives 13, 57-89 DeRose, K., 1995: Solving the Sceptical Problem, Philosophical Review 104, 1-52 Dretske, F., 1970: Epistemic Operators, Journal of Philosophy 67, 1007-1023 Hawthorne, J., 2004: Knowledge and the Lotteries, Oxford/New York: Oxford University Press Lewis, D., 1996: Elusive Knowledge, Australasian Journal of Philosophy 74, 549-567 Stanley, J., 2005: Knowledge and Practical Interests, Oxford/New York: Oxford University Press Willaschek, M., 2007: Contextualism about Knowledge and Justification by Default, Grazer Philosophische Studien 74, 251-272 Williams, M., 2001: Problems of Knowledge, Oxford/New York: Oxford University Press 1

I would like to thank Sebastian Boll, Michel de Araujo Kurth, Andreas Maier, Hannes Ole Matthiessen, Andreas Müller and Michael Parker for very helpful comments on earlier versions of this paper.

Knowing about Other Contexts PETER BAUMANN, Swarthmore College

1. Introduction One of the most discussed recent views in epistemology is contextualism. Contextualists about “knowledge” (see Cohen 1987, 1988, DeRose 1995, 2009, Lewis 1979, 1996, Unger 1984, 1986) hold that the truth conditions or meanings of knowledge attributions can vary with the attributor’s context. The sentence “Paul knows that Amsterdam is the capital of the Netherlands” could be true in the mouth of the first year geography teacher, Dr. Basic (given less demanding standards for “knowledge”), but false in the mouth of the fourth year geography teacher, Dr. Soph (given more demanding standards for “knowledge”): Paul is confused about the status of The Hague. I want to leave aside the attractions of contextualism here - be it a good fit with the data about how we use the word “knowledge”, be it the explanatory potential and ability to solve certain problems like the problem of Cartesian scepticism or be it something else altogether. Rather, I want to discuss a problem for contextualists which is, I think, more serious than the ones usually discussed (lacking fit with linguistic data, inability to deliver on the promised solutions of problems, etc.). Here is a rough and intuitive illustration of the problem. In the weather station I, the meteorological layperson, would say something false were I to claim that I “know” that we’re in for some rain later today. However, given my contextualist leanings, I would also concede that outside the weather station it would be true to say of me that I “know” that we’re in for some rain later today. However, given plausible epistemic principles, I am thereby also committed to the claim that we’re in for some rain later today. Even worse, I’m even committed to the claim that I “know” that. In other words, my contextualist tendencies commit me to an outright inconsistency. This problem has been “in the air” for some time, with very few discussion in print (but see Brendel 2003, 2005 and Wright 2005). Our problem is a very serious one because inconsistency is a vice for any view and the

64

Peter Baumann

contradiction arises at the centre of the contextualist view: I also take the ability to account for cross-context attributions of truth values to other knowledge attributions to be part of the adequacy conditions for any account of knowledge (or “knowledge”). Below I will first give a more precise formulation of the problem and second propose a contextualist solution to the problem. It turns out that a relationalist view of knowledge as a 3-place relation plus a particular contextualist version of the closure principle for knowledge will do the trick. But first back to the problem. 2. The Problem Suppose two subjects, Joseph and Ann, both believe that Ann has hands. While Joseph finds himself in a not so demanding context O, Ann finds herself in a more demanding context D. More precisely, the standards in O which determine whether it is true to say (in O) that Joseph “knows” that Ann has hands are less demanding than the standards in D which determine whether it is true to say (in D) that Ann “knows” that she has hands. It is easy to imagine circumstances under which, according to contextualism, the following holds: (1)

“Joseph knows that Ann has hands” is true in O

and (2)

“Ann knows that she has hands” is not true in D.

(1) and (2) simply result from the application of contextualism to our little toy situation. Let us further assume that Ann herself is a contextualist and that D is not a sceptical but just a very demanding context. So, Ann can be truly said (in D) to “know”, e.g., that there are other people in other contexts, making utterances, etc. Furthermore, Ann can apply her contextualist views to particular cases. She can attribute a truth value to knowledge attributions made in other contexts. For instance, she can attribute a truth value to an utterance in O of “Joseph knows that Ann has hands”. Nothing keeps us from constructing our toy scenario by further assuming that (3)

“Ann knows that (1)” is true in D.

Knowing about Other Contexts

65

More will have to be said about (3) below. – For the next step in the argument, we need two very plausible principles: a principle of disquotation and a principle of the factivity of knowledge. Here are their noncontextualist versions combined (with the arrow for material implication): (Disquotation & Factivity) Necessarily for all subjects S and propositions p: “S knows that p” is true  p. Contextualists need a different principle. Especially disquotation poses a challenge to the contextualist because it will not work to use something like the following principle: Necessarily for all subjects S and propositions p: “S knows that p” (as uttered in some context) is true  S knows that p? The challenge can, however, be met (see Sosa 2004, 43-4; Bach 2005, 589, Cohen 2005 201-4 and in particular Schaffer 2004, 2005) by making the context-sensitivity explicit and accepting a “relationalist” view according to which the knowledge-relation is a ternary, three-place relation between a person, a proposition and a context (or standard or some other kind of parameter). We will get back to this below. The contextualist can accept the following combined contextualist principle of disquotation and factivity: (DF) Necessarily for all subjects S, contexts C, and propositions p: “S knows that p” (as uttered in C) is true  p. Now, one of the things Ann can certainly be assumed to “know” in our toy scenario is that (DF) as applied to (1) is true. How should she not be able to disquote and use the factivity principle in concrete cases? In other words, not only is it the case that (4)

“Joseph knows that Ann has hands” is true in O  Ann has hands.

but furthermore Ann can “know” this: (5)

“Ann knows that (4)” is true in D.

The last principle we need for the derivation of a contradiction is a principle of closure of knowledge under known entailment. There is very little

66

Peter Baumann

disagreement that some such principle is both plausible and important. The most straightforward version of such a principle says that (Closure)

Necessarily for all subjects S and propositions p, q: S knows that p and S knows that (p  q)  S knows that q.

There are problems with this simple version of a closure principle even for non-contextualists (see Baumann 2011) but we can leave them aside here and now. More pressing is the fact that the contextualist needs a contextualist closure principle. Here is a plausible proposal (but for more on this see below): (Clos)

Necessarily for all subjects S, contexts C, and propositions p, q: [“S knows that p” is true in C and “S knows that (p  q)” is true in C]  “S knows that q” is true in C.

But now we can derive our contraction. (3), (5) and (Clos) entail (6)

“Ann knows that she has hands” is true in D.

(6) and (2) gives us the contradiction (7)

“Ann knows that she has hands” is true and not true in D.

This is the core version of our problem (for related versions of the problem see Baumann 2008 and Wright 2005; see also Brendel forthcoming; Jäger forthcoming; Freitag forthcoming). What can or should we do about this? Which premise has to go? If we don’t want to touch the very plausible principles of closure, factivity and disquotation or their knowability by Ann or their applicability to our case, then only (1)-(3) remain. If we don’t want to doubt that the contextualist Ann can apply her contextualist principles to particular cases (3), then only the contextualist assumptions behind (1) and (2) seem to remain. Should we then give up contextualism? Is this a refutation of contextualism?

Knowing about Other Contexts

67

3. The Nature and Scope of the Problem Simply giving up contextualism won’t help much because the same basic problem appears in slightly different forms for most (if not all) main alternative views of knowledge (or “knowledge”) currently discussed: for the orthodox view of classical invariantism according to which “knowledge” is context-invariant and knowledge only depends on truth plus epistemic parameters (like belief, warrant, etc.) as well as for subject-sensitive invariantism (see Fantl and McGrath 2002; Hawthorne 2004; Stanley 2005) according to which “knowledge” is context-invariant but also dependent on and relative to non-epistemic parameters (like practical stakes); whether the problem also arises for recent forms of relativism (see MacFarlane 2005) depends a lot on the details of the view and shall be left open here (but see Brendel 2009, 411-4). I also won’t go into subject-sensitive invariantism here (but see Baumann Ms. and Wright 2005 and Brendel 2007, 313). For the purpose of the illustration of the wide scope of the problem, I just want to briefly go into classical invariantism (see also Freitag 2011, fn.19 who to my knowledge is the first to point out that if there is a problem for contextualism here, then there is also one for invariantism; Freitag does not believe there is a problem in the first place). According to the classical invariantist, it should certainly be possible that (I)

Mary knows that the sun is out

while (II)

Jack doesn’t know that the sun is out.

In addition, Jack could know (I) (if you have doubts about this, wait and see the discussion below): (III)

Jack knows (I).

Factivity tells us in this case that (IV) If (I) then the sun is out. And there is nothing standing in the way of Jack knowing this: (V)

Jack knows (IV).

68

Peter Baumann

Given (Closure), (III) and (V) entail (VI) Jack knows that the sun is out which contradicts (II). A solution to this problem is not hard to find. Even if one decides to stick with (III) (again, see below), one can simply replace (Closure) by a more plausible, dynamic version of a closure principle: (Closure*) Necessarily for all subjects S, times t-1 and t-2 (t-1 earlier than t-2) and propositions p, q: S knows at t-1 that p, S knows at t-1 that (p  q), S competently infers between t-1 and t-2 that q and thus comes to believe at t-2 that q  S knows at t-2 that q (see Williamson 2000, 117; Hawthorne 2004, 29, 33). More details would have to be added (for instance, that S keeps knowing both premises at t-2) but they are not relevant here and can thus be left aside. What is important is that as soon as we make the temporal indices explicit the inconsistency disappears. All we get is something like this: (I*) (II*) (III*) (IV*) (V*)

Mary knows at t-1 that the sun is out Jack doesn’t know at t-1 that the sun is out Jack knows (I) at t-1 If (I*) then the sun is out Jack knows (IV*) at t-1.

Given that Jack makes the relevant competent inference between t-1 and t2 and thus comes to have the relevant belief at t-2, and given (Closure*), (III*) and (V*) entail (VI*) Jack knows at t-2 that the sun is out which certainly does not contradict (II*). Jack now (at t-2) knows what he did not know before (at t-1). No problem. A similar strategy has been proposed as a way out for contextualism (as well as subject-sensitive invariantism) (see Brueckner and Buford 2009, 2010); it will become apparent below why such an easy solution is not available to the contextualist.

Knowing about Other Contexts

69

Many people are reluctant to accept that there really is a consistency problem for contextualism. I therefore think it is helpful to compare our problem here with another problem which is nowadays often referred to as “the sceptical puzzle”. Many people accept the latter as a real problem (whatever the solution, if any, might turn out to be). Given a remarkable structural similarity with our problem, it is, I think, pretty hard for these people to refuse to accept that there is at least prima facie a consistency problem for contextualism. What exactly do I have in mind? The sceptical puzzle can be found in Descartes’ 1. Meditation and consists in an inconsistent triad of equally and extremely plausible propositions: (a) (b) (c)

Peter knows that he has hands; Peter knows that if he has hands then he is not a handless brain in a vat; Peter doesn’t know that he is not a handless brain in a vat.

Given (Closure), (a) and (b) entail (d)

Peter knows that he is not a handless brain in a vat

which contradicts (c) (nothing much changes if we replace (Closure) by something like (Closure*)). So (given closure), (a), (b) and (c) form an inconsistent triad of extremely plausible propositions. It is not surprising that this is considered by many to be a tough and serious philosophical problem. But now compare the sceptical puzzle with our problem. Here, too, we have three plausible assumptions: (2) (3) (5’)

“Ann knows that she has hands” is not true in D. “Ann knows that (1)” is true in D; “Ann knows that [(1)  Ann has hands”] is true in D;

Given (Clos), (3) and (5) entail (6)

“Ann knows that she has hands” is true in D

which contradicts (2). So (given closure), (2), (3) and (5’) form an inconsistent triad of very plausible propositions. There is a strong similarity of structure between the sceptical puzzle and our consistency problem for contextualism: (3) corresponds to (a); (5’)

70

Peter Baumann

corresponds to (b) – both have conditional content; finally (2) corresponds to (c). Both cases instantiate the following schema (with “K” for the knowledge operator): Kp (see (a) and (3)); K (p  q) (see (b) and (5’)); Not Kq (see (c) and (2)). So, why is it so plausible to accept that there is a sceptical puzzle involving (a), (b) and (c) but no real puzzle involving (3), (5’) and (2)? How can one accept the one but not the other puzzle? If it is not surprising that the sceptical puzzle is accepted by many as a serious problem, then it is surprising why our problem is not accepted by many as a serious problem either. Isn’t our consistency problem just another sceptical puzzle? All the above gives us, I think, good reason to look for a solution of our problem which does not consist in simply giving up contextualism. What then could be a contextualist way out? 4. A Proposal that Doesn’t Work: Knowability Restrictions Many people have, either in conversation and discussion or in published articles (see Brueckner and Buford 2009, 2010) proposed to skip (3): Roughly speaking, Ann cannot, according to them, “know” in her demanding context that Joseph “knows” in his lax context that she has hands. Generalized: Attributors in demanding contexts cannot know that knowledge attributions made in lax contexts are true. The question is: Why should one think so and deny (3)? One reason could be the following one: (3) entails (6) (together with some plausible assumptions); however, (6) is false (given (2)); hence, (3) must be false, too. This is not a good reason to doubt (3) because it begs the question against the whole argument that there is a contradiction for contextualism. Let us therefore put this argument aside and look for good independent reasons to deny (3). One could, for instance, argue that knowing the truth value of a knowledge ascription made in another context requires that one has prior and independent knowledge of the proposition said to be known in that other context. More precisely (see Baumann 2010, forthcoming):

Knowing about Other Contexts

(Prior)

71

Necessarily for all ascribers A and subjects S, contexts O and D, and propositions p: If ‘A knows that “S knows that p” is true in O’ is true in D, then “A knows that p” is true in D independently from and prior to the truth in D of ‘A knows that “S knows that p” is true in O’.

It is easier to see that (Prior) is false if one considers a parallel noncontextualist principle like (Priority)

Necessarily for all subjects B and A, and propositions p: If B knows that A knows that p, then B has antecedent knowledge that p independently from and prior to the knowledge that A knows that p.

Take a case of testimony. I can come to know by reading the papers that Wiles has found out and come to know that Fermat was right. Certainly, I can know this without having independent and prior knowledge that Fermat was right. Otherwise, the whole point of testimony would be lost. The case is similar, mutatis mutandis, with the contextualist (Prior). Another attributor, for instance, who shares my demanding context with me but is in a much better epistemic position with respect to “p” than I (actually “knows” it) could reliably testify to me that it is true in O to say that S “knows” that p. There is no reason to deny that I could thus come to “know” what my fellow attributor tells me – even if I don’t “know”, given our high standards, that “p” is true. Sure, I can deduce by competent deduction that “p” is true but that does not give me prior and independent knowledge. So, (Prior) does not constitute a good reason to deny (3) either. Denying (3) or, more generally, denying that attributors in demanding contexts can be truly said to “know” that knowledge-attributions made in less demanding contexts are true, constitutes an awkward and implausible restriction of what can be truly said to be “knowable” in demanding contexts. There would be a systematic and big blindspot for attributors in demanding contexts. Contextualism would only be statable within certain weird limits. Sure, contextualists could still state their contextualist views in conditional or modal form:

72

Peter Baumann

Necessarily for all subjects S, contexts O and propositions p: If “p” is/ were true and if S meets/ met the epistemic standards relevant in O, then it is/were true to say in O that S knows that p. However, this does not amount to a direct application of contextualism to knowledge attributions made in other contexts (like in (3)). Knowledge attributors in demanding contexts would turn out to suffer from systematic and widespread blindspots and lack of knowledge (or better: what “knowledge” refers to in their context) and suffer from statability problems. However, that cross-context attributions of truth values to knowledge attributions are possible is a very plausible adequacy condition for any theory of knowledge (or “knowledge”). Violating this condition requires good reasons – which I cannot see in our case. Contextualists should therefore go for a view of “open” contexts instead of “closed” contexts, that is, accept and explain that and how one can make statements (of the relevant kind) across contexts. Finally, one might wonder what someone like Ann in our example should say or think, according to the critic of (3). Should she affirm or deny that (1) “Joseph knows that Ann has hands” is true in O? Or suspend judgment? Or just not think about it? If Ann affirms (asserts or believes) (1), then she affirms something which is, according to the critic, true but “unknowable” for her. This raises questions about what one ought to affirm, assert or belief (see for the knowledge rule of assertion, e.g., Williamson 2000). May one affirm what one does not know? Apart from that, Ann would be entitled to infer from (1) and thus come to believe that she has hands. Again, this is something, that she cannot truly be said to “know” in her context, both the critic and any contextualist will agree. And how does this fit with Ann’s selfknowledge that she does not “know” by the standards in her context that she has hands? She would, it seems, be committed to affirming a Mooreparadox: “I have hands but I don’t know it”. Perhaps, given enough power of reflection, she would even be entitled to affirming a contradiction: “I know I have hands and I don’t know I have hands”. There is something for the contextualist to say here, as we will see below, but anyone who simply reacts to our problem by denying (3) will be in a very difficult position here if they recommend that Ann should affirm (1). Should the critic of (3) then recommend that Ann deny (1)? But this would mean that we recommend to Ann that she affirm something false, namely that (1) is not true. But how could that be good advice? Finally, to recommend suspending judgment about whether (1) is true or simply not thinking about it, seems odd, too.

Knowing about Other Contexts

73

Ann would affirm that she, in her demanding context, does not “know” that she has hands while at the same time leaving it open whether Joseph can, in his less demanding context, be said to “know” that she has hands. What reason could Ann have to come to a conclusion about her own epistemic situation with respect to that proposition while leaving everything open concerning Joseph’s epistemic relation to it? It seems that those who deny (3) are in a difficult predicament here. However, this does not mean that those who accept (3) won’t have a lot to explain. 5. A Solution: Contextualist Closure So, what can one say if one accepts (1) and (2) as well as (3)? Is there something wrong with (5) or with the principles of disquotation and factivity involved? I cannot see what could be wrong with them. Hence, the only assumption left to question is our closure principle. People often take something like (Closure) or (Clos) to be so plausible that the principle should be beyond doubt. However, there are good reasons, trivial and not so trivial, for having doubts about (Closure) or (Clos). Trivial ones involve the fact that we often do not “put 2 and 2” together; this insight motivates principles like (Closure*) Necessarily for all subjects S, times t-1 and t-2 (t-1 earlier than t-2) and propositions p, q: S knows at t-1 that p, S knows at t-1 that (p  q), S competently infers between t-1 and t-2 that q and thus comes to believe at t-2 that q  S knows at t-2 that q or its contextualist cousin. But there are also more substantial reasons to modify the simple principles of closure (see, e.g., Baumann 2011). One major reason which is of direct relevance here has to do with what has been called “failure of the transmission of warrant” (see Davies 1998; Wright 2000). Some authors have argued that the warrant for the premises in the following argument do not transmit to the conclusion: (A) (B) (C)

I have hands If I have hands, then I am not a handless brain in a vat Hence, I am not a handless brain in a vat.

74

Peter Baumann

While I can have warrant for (A) as well as (B) and even know both, the inference to (C) will not transfer the warrant: I have no warrant for and thus do not know (C). One way to make sense of this and explain it is to say that knowledge of (A) and (B) is based on presuppositions which include the proposition that I am not in weird sceptical circumstances (like being an ignorant brain in a vat). However, so the idea goes, one cannot by making an inference to a presupposition gain warrant or knowledge concerning that presupposition (see Baumann 2011 and Davies 1998 and Wright 2000). I don’t want to say much more about this in general but rather get to our problem about contextualism concerning “knowledge” and the question which closure principle applies to it. Consider a case and let us, for the sake of simplicity, perform semantic descent and use the explicit ternary, relational form of contextualist knowledge ascriptions. Let us thus take instances of the schema “S knows-d that p” as equivalent and short for “”S knows that p’ is true in (demanding context) D” and instances of the schema “S knows-o that p” as equivalent and short for “”S knows that p’ is true in (not so demanding context) O”. Now, suppose that ()

Mary knows-d that there is no life on Mars.

Suppose further that ()

Fred knows-d that ()

and ()

Fred knows-d that if () then there is no life on Mars.

Does it follow that ()

Fred knows-d that there is no life on Mars?

The following closure principle would sanction the step from () and () to (): (Clos-1)

Necessarily for all subjects S, knowledge-relations know-x, know-y and know-z, and propositions p, q: [S knows-x that p and S knows-y that (p  q)”]  S knows-z”.

Knowing about Other Contexts

75

But should we accept something like (Clos-1)? Is it really plausible in our case that Fred knows-d that there is no life on Mars? I think this need not be the case. Fred might only have knowledge-o that there is no life on Mars. He might lack the relevant expertise concerning Mars and life on other planets. However, this is compatible with being an expert on Mary’s epistemic situation concerning, say, topics from work (she is a specialist on Mars). It is also compatible with having knowledge-d of the relevant entailment (whatever it means to be an expert on entailment relations). So, it seems perfectly possible that () () ()

Mary knows-d that there is no life on Mars; Fred knows-d that (); Fred knows-d that if () then there is no life on Mars.

and ()

Fred knows-o that there is no life on Mars.

Put differently and in more general terms: Take a conditional p  q. That one knows-d the conditional as well as its antecedent does not mean that one knows-d the consequent of the conditional. It does not help to object that the premises and the conclusion of an argument must be known by standards of the same degree of demandingness. How should one compare the levels of demandingness for different propositions? This suggest the following contextualist closure principle for “knowledge-sentences” (we could call this a principle of “downwards closure”): (Clos-K)

Necessarily for all ascribers A and subjects S, knowledgerelations know-o and know-d (where know-o is less demanding than know-d), and propositions p: (a) [A knows-d that S knows-d that p and A knows-d that (S knows-d that p  p)]  A knows-o that p. and (b) [A knows-d that S knows-o that p and A knows-d that (S knows-o that p  p)]  A knows-o that p.

I am restricting myself to a special principle for certain cases of “knowledge” here; a more general contextualist closure principle is not at issue

76

Peter Baumann

here. (Clos-K) is all we can expect from closure here if we’re contextualists. And it is enough. It does, of course, not exclude that possibly A comes to know-d that p; it just doesn’t guarantee it. (Clos-K-a) explains our judgments about the Fred-Mary-Mars-case while (Clos-k-b) takes care of our initial Ann-Joseph-hands-case. In the former case, Fred’s knowledge-d of the two premises does not “translate” into knowledge-d but only knowledge-o of the conclusion. Similarly, Ann’s knowledge-d that Joseph knows-o that she has hands plus her knowledge-d that the latter entails that she has hands does not “translate” into knowledge-d but only knowledge-o of the conclusion that she has hands. It is a subtle but farreaching mistake to think that warrant or knowledge transfer without restrictions as to the kind of warrant or knowledge-relation. The crucial point is that if we use the appropriate (Clos-K) rather than the subtly misleading (Clos) or (Clos-1), then our contradiction disappears. In our Ann-Joseph-case, only a meta-linguistic version of Clos-K-b is relevant. Given it, (3) and (5) do not entail (6)

“Ann knows that she has hands” is true in D

but only (6’)

“Ann knows that she has hands” is true in O

or, in other words (6’’) A knows-o that she has hands. Hence, we don’t get a contradiction but only (7’)

“Ann knows that she has hands” is true in O but not in D

or (7’’) A knows-o that she has hands but she does not know-d that she has hands. There is nothing problematic about that. Our problem is solved; the contradiction is gone. Furthermore, the above allows us to give a contextualist account of cross-context attributions of truth values to knowledgeattributions made in other contexts. This is an adequacy condition for any account of knowledge (or “knowledge”). The ideas used here are a rela-

Knowing about Other Contexts

77

tionalist, ternary view of the knowledge-relation and, more importantly, a modified closure principle. The latter is also interesting for more general debates about epistemic closure (see Baumann 2011). The former is also remarkable given that many contextualists tend to treat “knows” like an indexical. If one does that, then it is not easy to see how our problem can be solved. The fact that a relationalist version of contextualism can contribute to a solution of our problem speaks in its favor (over “indexicalist” versions). Finally, one has, I think, good reasons to be optimistic about the ability of a relationalist contextualism cum modified closure to deal with other issues, like belief reports or several “abominable conjunctions” the contextualist is supposedly committed to (see DeRose 1995). References Bach, K., 2005: The Emperor’s New ‘Knows’, in: Contextualism in Philosophy: Knowledge, Meaning, and Truth, ed. G. Preyer and G. Peter, Oxford: Clarendon, 51-89 Baumann, P., 2008: Contextualism and the Factivity Problem, Philosophy and Phenomenological Research 76, 580-602 Baumann, P., 2010: Factivity and Contextualism, Analysis 70, 82-89 Baumann, P., 2011: Epistemic Closure, in: The Routledge Companion to Epistemology, ed. S. Bernecker and D. Pritchard, London/New York: Routledge, 597608 Baumann, P., forthcoming: A Contradiction for Contextualism, in: Epistemology, Context, Formalism, ed. F. Lihoreau and M. Rebuschi Baumann, P., Subject-Sensitive Invariantism, Ms. Brendel, E., 2003: Was Kontextualisten nicht wissen, Deutsche Zeitschrift für Philosophie 51, 1015-1032 Brendel, E., 2005: Why Contextualists Cannot Know They Are Right: Self-Refuting Implications of Contextualism, Acta Analytica 20-2 (35), 38-55 Brendel, E., 2007: Kontextualismus oder Invariantismus? Zur Semantik epistemischer Aussagen, in: Referenz und Realität, ed. A. Rami and H. Wansing, Paderborn: Mentis, 11-37 Brendel, E., 2009: Contextualism, Relativism, and Factivity, in: Reduction and Elimination in Philosophy and the Sciences, ed. H. Leitgeb and A. Hieke, Frankfurt am Main: Ontos, 403-416 Brendel, E., 2012: Knowledge, Contextualism, and Moorean Paradox, this volume

78

Peter Baumann

Brueckner, A. and C.T. Buford, 2009: Contextualism, SSI and the Factivity Problem, Analysis 69, 431-438 Brueckner, A. and C.T. Buford, 2010: Reply to Baumann on Factivity and Contextualism, Analysis 70, 486-489 Cohen, S., 1987: Knowledge, Context, and Social Standards, Synthese 73, 3-26 Cohen, S., 1988: How to Be a Fallibilist, Philosophical Perspectives 2, 91-123 Cohen, S., 2005: Knowledge, Speaker and Subject, Philosophical Quarterly 55, 199212 Davies, M., 1998: Externalism, Architecturalism and Epistemic Warrant, in: Knowing Our Own Minds, ed. C. Wright, B. Smith, and C. Macdonald, Oxford: Clarendon, 321-361 DeRose, K., 1995: Solving the Skeptical Problem, The Philosophical Review 104, 1-52 DeRose, K., 2009: The Case for Contextualism. Knowledge, Skepticism, and Context, Vol.1, Oxford: Clarendon Press Fantl, J. and M. McGrath, 2002: Evidence, Pragmatics, and Justification, The Philosophical Review 111, 67-94 Freitag, W., 2011: Epistemic Contextualism and the Knowability Problem, Acta Analytica 26, 273-284 Freitag, W., 2012: Epistemic Variantism and the Factivity of Knowledge, this volume Hawthorne, J., 2004: Knowledge and Lotteries, Oxford: Clarendon Jäger, C., 2012: Contextualism, and the Knowledge Norm of Assertion, Analysis 72 (in press) Lewis, D., 1979: Scorekeeping in a Language Game, Journal of Philosophical Logic 8, 339-359 Lewis, D., 1996: Elusive Knowledge, Australasian Journal of Philosophy 74, 549-567 MacFarlane, J., 2005: The Assessment Sensitivity of Knowledge Attributions, Oxford Studies in Epistemology, Vol. 1, ed. T. Szabó Gendler and J. Hawthorne, Oxford: Oxford University Press, 197-233 Schaffer, J., 2004: From Contextualism to Contrastivism, Philosophical Studies 119, 73-103 Schaffer, J., 2005: Contrastive Knowledge, Oxford Studies in Epistemology, Vol.1, ed. T. Szabó Gendler and J. Hawthorne, Oxford: Oxford University Press, 235271 Sosa, E., 2004: Relevant Alternatives, Contextualism Included, Philosophical Studies 119, 35-65 Stanley, J., 2005: Knowledge and Practical Interests, Oxford: Clarendon Unger, P., 1984: Philosophical Relativity, Oxford: Blackwell

Knowing about Other Contexts

79

Unger, P., 1986: The Cone Model of Knowledge, Philosophical Topics 14, 125-178 Williamson, T., 2000: Knowledge and Its Limits, Oxford: Oxford University Press Wright, C., 2000: Cogency and Question-Begging: Some Reflections on McKinsey’s Paradox and Putnam’s Proof, Philosophical Perspectives 10 (Skepticism), 140-163 Wright, C., 2005: Contextualism and Scepticism: Even-Handedness, Factivity and Surreptitiously Raising Standards, Philosophical Quarterly 55, 236-262

Epistemic Variantism and the Factivity of Knowledge WOLFGANG FREITAG, University of Heidelberg

Epistemic variantism is the position that there are, or may be, different standards for the evaluation of knowledge ascriptions. While a statement of the form ‘S knows that p’ may be true according to one such standard, it may be false if measured against another. The theory comes essentially in two versions, differing with respect to the kinds of context assumed to be relevant for standard determination. According to Epistemic Contextualism (EC) the standards may vary with the context of the knowledge attributor, while Subject-Sensitive Invariantism (SSI) binds standard alteration to change of the context of the subject of knowledge ascriptions. Epistemic variantism has met with many criticisms concerning its motivation and application, but recently there have been doubts about its very tenability. A number of authors, amongst them Timothy Williamson (2001), Elke Brendel (2003, 2005, 2007, 2009), Crispin Wright (2005), and Peter Baumann (2008), have proposed that epistemic contextualists face a kind of performative contradiction. According to their arguments, contextualism cannot be known to be true and therefore cannot be knowledgeably stated. The argument for the factivity problem, as the difficulty has come to be called, has subsequently been extended to cover SSI by Anthony Brueckner & Christopher Buford (2009, 2010) and Peter Baumann (2010). If the argument is correct, any form of epistemic variantism seems to be threatened by the factivity problem. Two types of reaction have been proposed. Brueckner & Buford, and perhaps Williamson, hold that the (alleged) fact that variantism cannot be knowledgeably stated has no further problematic consequences. Baumann, Brendel, and Wright, on the other hand, take this to be an unsatisfactory response. They claim that either one of the principles needed to derive the problem has to be given up or epistemic variantism itself must be abandoned. In this paper I argue that, given a clear view on the commitments of epistemic variantism, the difficulty turns out to be apparent only. There is no factivity problem for epistemic variantism. The question of how to react

82

Wolfgang Freitag

to the alleged performative contradiction does therefore not even arise in this case. Frequent claims to the contrary are based on a misunderstanding of the commitments of epistemic variantism. Since my discussion involves a review of these commitments, this paper is meant to contribute not only to the understanding of the dialectical situation with respect to the factivity problem, but also to a clarification of the position of epistemic variantism itself. I will, however, remain neutral on the further issues of whether variantism is correct as a linguistic theory or whether it is significant for epistemology at all. My argument develops as follows. In the first section, I give a unified description of epistemic variantism covering both EC and SSI. Section 2 describes the argument for the factivity problem as it is usually presented. I will comment on the argument in section 3 and reject it in section 4. Section 5 addresses another worry, stemming from the fact that epistemic variantism suffers from certain statability limitations. I claim that, to the extent that there are such limitations for epistemic variantism at all, epistemic invariantism has analogous statability limitations. Even if such limitations should be considered problematic, they do not militate against epistemic variantism. I conclude with discussing a general result on the factivity problem. The necessary and sufficient conditions for the factivity problem will be identified. I then claim that it is highly implausible that epistemic variantism meets these conditions. 1. A unified description of epistemic variantism Let “K(S, T, P)” stand for “S knows at T that P”. Since we are concerned with EC and SSI, we must make room for context-dependent concepts of knowledge. In the following, let  be the set of possible contexts and i   be the set of possible contexts with knowledge standard i (for varying i). In particular, let h   be the set of high-standard contexts and l   the set of low-standard contexts. I will make the simplifying assumption that  = h  l and thus that there are only high- and low-standard contexts.1 Let “Kh(S, T, P)” stand for “ ‘S knows at T that P’ is true according 1

Epistemic variantism is neither limited to only two possible standards, nor is it tied to the claim that these standards can be ordered in strength. Different standards might make demands which differ in incommensurable ways. The mentioned assumptions

Epistemic Variantism and the Factivity of Knowledge

83

to high standards” (or “S knowsh at T that P”, for short), and “Kl(S, T, P)” stand for “ ‘S knows at T that P’ is true according to low standards” (“S knowsl at T that P”).2 As opposed to the meaning of “know”, the meanings of “knowl” and “knowh” are context-invariant. “S knowsi that p” says that S’s epistemic position with respect to p is strong enough to meet standard i for the correct application of “knows”.3 John knowsi that p if and only if (i) knowledgei demands a certain strength of epistemic position, and (ii) John’s epistemic position with respect to the proposition p is of at least that strength. Change of context affects neither (i) nor (ii). Knowledgei is not “lost” if the attributor moves into a context with higher standards. Knowledgei-claims are invariantly true or false.4 Both EC and SSI claim it to be possible that Kl(S, T, P) and nonKh(S, T, P), for certain S, T, and P. They also agree on the view that the standards for correct attribution of “knows” may vary from context to context; they differ only with respect to whose context is relevant for the determination of knowledge standards. EC takes this to be a function of the context of the knowledge attributor; “Ki(S, T, P)” is then to be read as a short version of “the utterance ‘S knows at T that P’ is true if made in an attributor context with standard i”.5 SSI, by contrast, takes the strength of the epistemic position required to be a function of the interests, stakes, and salient error possibilities of the epistemic subject, S. “Ki(S, T, P)” is then to be read as a short version of “the utterance ‘S knows at T that P’ is true in a context with standard i determined by S’s interests, stakes, and salient error reflect the most prominent versions of epistemic variantism discussed in the literature. Anyway, my arguments below do not rest on these assumptions. 2 For a similar way of representing contextualism, see also Bach 2005. By indexing “know” in the way described, we are not liable to the fallacy of semantic descent, which we would be if we were to use disquotation without indexing (see Brendel 2005, 46, and Baumann 2008, 588 ff.). Note that we can retranslate our way of description into proper metalinguistic form without loss of content. 3 I borrow the term “strength of epistemic position” and its cognates from DeRose (1995, e.g., 34). Note that the strength of the epistemic position is independent of the context of knowledge ascription. 4 Here and in the following, I simplify by assuming that knowledge claims are about complete propositions and thereby ignore the possible involvement of other indexical terms. 5 Prominent contextualists along these lines are Stewart Cohen (e.g., 1988 and 1998), Keith DeRose (e.g., 1992 and 1995) and David Lewis (1996).

84

Wolfgang Freitag

possibilities”.6 For present purposes, we may ignore the differences between EC and SSI on the question of standard-relevant contexts and speak of them both as forms of (epistemic) variantism. 2. The factivity problem The factivity problem derives from the following considerations. Let  be the set of subjects,  the set of times, and  the set of empirical propositions. The following principles are hard to give up; in particular they are usually accepted by proponents of epistemic variantism: (Fact) x s t p: Kx(s, t, p)  p;7 and (Clos-KE) x s t p, q: ([Kx(s, t, p)  Kx(s, t, (p  q))]  Kx(s, t, q)).8 (Fact) is a variantist version of the factivity condition for knowledge. (Clos-KE) is the claim that knowledge is closed under known entailment, given that the context of knowledge ascription remains constant. My policy in this paper will be to grant all such ‘transitional’ principles to those who argue that variantism has the factivity problem. For ease of presentation, I will even grant a stronger principle than (Clos-KE), namely, (Clos)

x s t p, q: ([Kx(s, t, p)  (p  q)]  Kx(s, t, q)).

(Clos) states that knowledge is closed under entailment simpliciter. Given (Fact), (Clos) implies (Clos-KE). If a theory turns out to be unproblematic given (Clos), it will also be unproblematic given the weaker principle (Clos-KE). Now to the argument for the factivity problem, as it is usually presented. Let S and S* be subjects,9 T some point in time, and let ‘HANDS’ stand for 6

Proponents of SSI are, for example, John Hawthorne (2004) and Jason Stanley (2005). 7 The symbol ‘’ stands for entailment. 8 Obviously (Clos-KE) is very problematic (see, for some critical points, Williamson 2000 and Hawthorne 2004). My point is that, usually, some form of closure is accepted. Which form this might be is rather irrelevant to my present concerns. 9 It is not necessary that S and S* are distinct subjects. Some proponents of the factivity problem use examples in which S and S* are identical.

Epistemic Variantism and the Factivity of Knowledge

85

the proposition ‘S has hands’. Given (Clos) and (Fact), the allegedly variantist propositions (1) and (2) entail, with the additional assumption of (3), a contradiction: (1) (2) (3)

Kl(S, T, HANDS), Not: Kh(S*, T, HANDS), Kh(S*, T, (1)).

Proposition (4) follows from (3) with (Fact) and (Clos), but it contradicts (2): (4)

Kh(S*, T, HANDS).10

The argument is meant to show that epistemic variantism cannot be knownh, and hence not knowledgehably stated, by S*. As Brueckner & Buford understand the argument: “This seems to be a reductio of [epistemic variantism] given [(Clos) and (Fact)]. [(3)] appears to be the culprit, and since [epistemic variantism] seems to be committed to cross-context claims like [(3)], so much the worse for [epistemic variantism]” (2009, 432). Brueckner & Buford defend variantism against the threat of contradiction by claiming that, after all, “the theories are not committed to the possibility of such asymmetrical knowledge attribution” (2009, 434) and that therefore (3) can be rejected. The fact that a variantist “cannot ‘knowledgeably’ state the contextualist thesis [(1)]” (Brueckner & Buford 2009, 436) is, according to their view, not a problem for variantism: a true theory need not be knowledgeably statable. Baumann objects, however, that statability limitations as those accepted by Brueckner & Buford are “bad enough for [variantism] to not deserve acceptance” (2010, 87). This then is the dispute between the two parties characterising the different takes on the problem. Is the (putative) fact that variantism is not knowledgeably statable detrimental for the theory? Bruckner & Buford (and Williamson) give a negative answer to this question. According to them, failure of knowing a theory does not touch upon its tenability. Philosophers of the opposing camp give a positive answer and ask for drastic measures. Either 10

See Brueckner & Buford 2009, 431–434; Baumann 2010, 83–84. Similar arguments can be found in Williamson 2001, Brendel 2003, 2005, 2007, 2009, and Wright 2005.

86

Wolfgang Freitag

they reject one of the principles used in inferring the contradiction (Baumann rejects (Clos) and (Clos-KE)) or they take epistemic variantism itself to be disproved (Brendel and Wright reject EC on grounds of the factivity problem). I think that it is a very interesting question how to deal with a theory that has the factivity problem,11 but not one with which we must be concerned in the present context. I will argue that, contrary to the received view, variantism is knowledgeably statable. There is no factivity problem for variantism in the first place. 3. Comments on the argument Some comments on the argument are in order. Firstly, premise (1) is not really needed, since it follows with (Fact) from premise (3). Secondly, the problem arises only with (3), where knowability of epistemic variantism enters stage. The problem would therefore more aptly be called knowability problem. Thirdly, proposition (3) concerns S*’s high-standard knowledge. As a consequence, the argument shows at most that knowledgeh of epistemic variantism entails a contradiction. The possibility of knowingl epistemic variantism is left intact. So we would have to speak more properly of the knowhability problem. In the following, I will, however, stick to the established label ‘factivity problem’. Fourthly, and most importantly, the argument presented above is a problem for epistemic variantism only if it has both (1) and (2) as consequences. So the following entailment thesis is essential: (ET) Epistemic variantism  (1)  (2). Without (ET) the aforementioned argument does not demonstrate the existence of the factivity problem. Unsurprisingly, all proponents of the factivity problem endorse (ET). For example, Brueckner & Buford say, “According to contextualism, [(1)] is true” and “[(1)] is a consequence of SSI” (2009, 431 and 433), and implicitly ascribe (2) to epistemic variantism. Baumann concurs: “According to [the epistemic variantist] views, it would

11

A theory which holds, for example, that nothing whatsoever is known or knowable would not be knowledgeably statable. Does that render the theory untenable?

Epistemic Variantism and the Factivity of Knowledge

87

then be true that [(1)] and that [(2)]” (2010, 83).12 The entailment thesis is the common ground for the participants of the factivity debate, which might explain why there is hardly any explicit argument in its favour. But I disagree with this consensus: the entailment thesis is not only unwarranted; it is even false. No form of epistemic variantism known to me entails propositions (1) and (2). 4. Rejecting the entailment thesis Whether (ET) is true or false depends on the claims of epistemic variantism. Nothing said so far about epistemic variantism makes (ET) true; the possibility of different standards for ‘knows’ does not entail that (1) and (2) are true. It might be suggested that (ET) is a consequence of the position’s ulterior epistemological motives, but this is going to be shown as mistaken. The major motivation for variantism is its promise to undermine the position of ‘global’ scepticism represented by the following proposition: (GS) x s t p: Kx(s, t, p).13 Global scepticism says that knowledge of any standard is impossible. Given the variantist reconstruction, scepticism relies on the following assumptions: (Invariantism)

x, y s t p: (Kx(s, t, p)  Ky(s, t, p)); (High Standard) h ≠ ; (Scepticismh) xh s t p: Kx(s, t, p). Proposition (Invariantism) says that, necessarily, a knowledge ascription true according to some standard is true according to any standard. In other 12

We find the entailment thesis already in Baumann 2008. “The sceptic typically argues that we do not know because we cannot know” (Williamson 2001, 27). Sometimes the sceptic position is described as stating the much weaker claim that all knowledge ascriptions are false as a matter of (contingent) fact. Such a characterisation, however, is not really distinctive of scepticism. A moderate (non-sceptical) invariantist may hold that all knowledge ascriptions are false, if he thinks that, as a matter of contingent fact, people do not fulfil his moderate demands. Surprising as this empirical fact would be, it would not turn moderate invariantism into a form of scepticism.

13

88

Wolfgang Freitag

words, it says that there are no substantially different knowledge standards. Proposition (High Standard) says that there are high knowledge standards. Proposition (Scepticismh), finally, states that it is impossible to fulfil high knowledge standards. These three assumptions jointly entail (GS). The traditional reaction to scepticism is moderate invariantism, which accepts (Invariantism), but denies (High Standard). The distinctive variantist reaction is to allow for (High Standard), but to deny (Invariantism). As DeRose says with respect to EC: [T]he fact that the sceptic can […] install very high standards which we don’t live up to has no tendency to show that we don’t satisfy the more relaxed standards that are in place in ordinary conversations. Thus, it is hoped, our ordinary claims to know will be safeguarded from the apparently powerful attacks of the sceptic, while, at the same time, the pervasiveness of the sceptical argument is explained. (DeRose 1992, 914)

According to DeRose, the high standards are indeed sometimes operative, but only in very special contexts of knowledge attribution. In ordinary, non-sceptical contexts, more moderate standards are at work. Analogously for SSI; the high standards are at work only given certain interests, stakes and salient error possibilities of the subject. In other situations, low standards may be operative. The variantist answer to the problem of scepticism is hence the negation of (Invariantism), resulting in the following form of minimal epistemic variantism: (MEV)

x, y s t p: (Kx(s, t, p)  Ky(s, t, p)).

If (MEV) is held true, (GS) cannot be derived from (Scepticismh) and (High Standard), and therefore the argument for (GS) is blocked. (MEV) is all that epistemic variantists need to claim in response to the sceptical argument.14 If reduced to just two distinct sorts of standards, l and h, MEV turns into (MEVh/l) s t p: (Kl(s, t, p)  Kh(s, t, p)). (MEVh/l) states that low-standard knowledge claims need not have the same truth-values as the corresponding high-standard knowledge claims. It 14

I have sympathies with those who claim that this answer to scepticism is not adequate (see, e.g., Sosa 2000, Kornblith 2000 and Williams 2001). But this debate is only tangential to the purposes of this essay.

Epistemic Variantism and the Factivity of Knowledge

89

is obvious, however, that (MEVh/l) entails neither (1) nor (2). If epistemic variantism does not claim more than (MEVh/l), (ET) is false. Do epistemic variantists have to endorse a position stronger than (MEV), such that (ET) becomes true? While epistemic variantism of the usual sort allows for actual variantist scenarios like the conjunction of (1) and (2), it does not entail the existence of any such scenario, let alone any such scenario with respect to HANDS. Consider (1): No form of epistemic variantism known to me entails that subject S knowsl HANDS. It is compatible with variantism that S is not in the right epistemic position with respect to HANDS, or that HANDS is false. It may even be that S does not exist. All this is possible without variantism being falsified. Likewise, no variantist known to me claims that (2) is part of the variantist theory. Epistemic variantism as such does not claim that high standards cannot be fulfilled; nor does it claim that they are not fulfilled in certain cases. To be sure, epistemic variantists often do accept (Scepticismh) and therefore (2). But this acceptance is not demanded by their position; at most it is required by the dialectical setting in which they find themselves: if (Scepticismh) were abandoned, the cited argument for (GS) could be rejected without endorsing (MEV) and hence without endorsing epistemic variantism. Even versions of epistemic variantism stronger than (MEV), e.g., (MEV*) below, do not make true the entailment thesis: (MEV*) x, y s t p: (Kx(s, t, p)  Ky(s, t, p)). (MEV*) states that it is possible for (1) and (2) to be true, but this entails neither (1) nor (2). If the entailment thesis is false, however, the contradiction derived from (1), (2) and (3) is pointless as an argument against (MEV*) – and a fortiori against the (MEV). Even though the contradiction shows that (1) and (2) are not jointly knowledgehably statable, this has no bearing on (MEV*). If, therefore, epistemic variantism claims no more than (MEV*), then it has not been shown to be vulnerable to the factivity problem. There is ample textual evidence for the view that epistemic variantists do not, as a matter of fact, claim significantly more than (MEV*). Stewart Cohen, for example, characterizes contextualism as follows: [A]scriptions of knowledge are context sensitive. […] The truth-value of sentences containing the words “know” and its cognates will depend on contextually determined standards. Because of this, such a sentence can have different truth-values in different contexts. (Cohen 2000, 94, my italics; similarly DeRose 1992)

90

Wolfgang Freitag

On a very plausible reading, this quote provides an informal statement of (MEV*). Even the proponents of the factivity problem render epistemic variantism in this modal form. Let me give just two quotations relating to epistemic contextualism. According to Brendel (2005, 40), the “main thesis” of contextualism is as follows: “Let S be an epistemic subject, p a proposition and ci, cj contexts with i, j  and i  j, then it is possible that: ‘K(S, p)’ is true in ci and ‘K(S, p)’ is true in cj” (my italics). Baumann (2008, 581) concurs: “The contextualist says that the following is possible: O’s utterance of ‘S knows that he has hands’ is true [= Kl(S, HANDS)] and S’s utterance of the same sentence is false [= Kh(S, HANDS)]” (my italics). So both take ((1)  (2)) to be part of contextualism, which suggests that they consider (MEV*) to be the proper description of epistemic variantism. But then the entailment thesis (ET) is false and the argument for the facitivity problem irrelevant for the case of epistemic variantism. 5. Statability limitations The involvement of the modal operator in the definition of epistemic variantism is a sign of sanity. If we were to drop it, epistemic variantism would evolve into an empirical theory making claims about the contingent epistemic positions of contingently existing individuals and the – equally contingent – existence of their hands. Given this situation, why do the proponents of the factivity problem ignore the modality operator when they endorse the entailment thesis? The trace of a reason can be found in the following remark from Baumann 2009, which is a reaction to the argument for the factivity problem: The contextualist would have to retreat to something like “It is possible that [(1)].” This strikes me as a reluctance to take one’s contextualism seriously and apply it to concrete cases. What is the attraction of contextualism if one cannot (at least as a contextualist) coherently say or think that knowledge attributions made in a lower context are in fact true? (Baumann 2008, 583)

To be attractive, Baumann seems to think, contextualism must allow that one can “coherently say or think” that attributions of knowledgel are in fact true. Even granted this, however, there is no reason to suppose that the factivity problem arises. Firstly, Baumann shifts the topic from the statability of contextualism to the statability of (1). Secondly, in opposition to Baumann’s wording, (1) can be coherently stated. At most, it cannot be

Epistemic Variantism and the Factivity of Knowledge

91

knowledgehably stated. Thirdly and most importantly, it can be demonstrated that (1) is knowledgehably statable. Even (1) and (2) jointly are knowledgehably statable, as I will now show. Call a scenario a variantist scenario if (1) and (2) obtain. Is it possible to knowh that a variantist scenario obtains? Surely it is – if (Scepticismh) is denied; epistemic variantism does not preclude a subject S** (with S**  S*) from knowingh (1) and (2). Even S* herself can knowh that the variantist scenario obtains, if this type of knowledge occurs at T* (with T*  T). To be sure, there are statability limitations. Given (Fact) and (Clos), S* cannot at T knowh that the variantist scenario obtains. A subject S* cannot knowh, and therefore cannot knowledgehably state, that S knowsl that HANDS without herself being in possession of high-standard knowledge of HANDS. More generally, a subject who does not knowh that P at a certain time T, cannot at T knowledgehably state of anybody that she knowsl that P. It is worth noting, however, that such limitations equally apply to epistemic invariantism. As the propositions (1), (2), and (3) are (together with (Clos) and (Fact)) jointly inconsistent, so are (5) (6) (7)

K(S, T, HANDS), Not: K(S*, T, HANDS), and K(S*, T, (5)).

A subject cannot, at T, knowledgeably state that somebody (else) knows a certain proposition P, if she does not, at T, herself know that P. Failure of knowing/knowingh some proposition entails limitations on what is knowledge(h)ably statable about (other) persons’ knowledge. However, this has nothing to do with the issue of variantism and invariantism. Statability limitations are a consequence of the factivity of knowledge alone, and hence a consequence for any epistemology that respects the factivity of knowledge. In particular, it is a consequence for both variantism and invariantism. Even if know(h)ability limitations of the sort identified above should be considered problematic, they cannot be held against epistemic variantism. But of course one may fail to see why these knowhability limitations should be held problematic in the first place.

92

Wolfgang Freitag

6. Concluding remarks I have argued that (MEV*) – and therefore the weaker (MEV) – has not been shown by the contradiction resulting from (1), (2), and (3) to possess the factivity problem. Moreover, it has been demonstrated that epistemic invariantism is subject to statability limitations analogous to those of epistemic variantism. Before we close the book on the factivity problem, however, it remains to address one further worry: To undermine an argument for the factivity problem is one thing, to reject the factivity problem as such is another. It would be desirable to have a general proof to the effect that epistemic variantism is free of the factivity problem. The following general thesis on the factivity problem can be proven (see Freitag 2011):15 For arbitrary subject S and arbitrary point in time T: Theory Θ has the factivity problem – i.e., S cannot, at T, know(h) Θ to be true – iff (FP) p: (Θ  [p  K(h)(S, T, p)]). Given this result, we can easily check for a given theory Θ whether it does or does not have the factivity problem. The following can be concluded from this result: a. A theory Θ committed to (1) and (2) has the factivity problem; that is, if Θ  (1)  (2), (FP) is true. The general result on the factivity problem provides the right verdict on theories committed to (1) and (2). b. (MEV*) does not have the factivity problem. For Θ = (MEV*), (FP) is not true. Of course, the weaker (MEV) does then not have the factivity problem either. c. Since (FP) refers to a single knowledge standard only, the factivity problem turns out to be independent of the question of variantism vs. invariantism. This explains why variantism and invariantism have analogous know(h)ability limitations.

15

The proof assumes (Fact) and (Clos).

Epistemic Variantism and the Factivity of Knowledge

93

d. There are even non-modalised versions of epistemic variantism which don’t incur the factivity problem. For example, (FP) is not true for Θ = p: [Kl(S*, T, p)  Kh(S, T, p)]. To conclude: Proponents of the factivity problem must demonstrate that epistemic variantists endorse a theory Θ such that (FP) is true. I have indicated in this paper that a successful demonstration is very unlikely. At any rate, there are as yet no convincing arguments to the contrary, and thus there is no reason to suppose that epistemic variantism has the factivity problem.16 References Bach, K., 2005: The Emperor’s New ‘Knows’, in: Contextualism in Philosophy: Knowledge, Meaning, and Truth, ed. G. Preyer and G. Peter, Oxford: Clarendon Press, 51-89 Baumann, P., 2008: Contextualism and the Factivity Problem, Philosophy and Phenomenological Research 76, 580-602 Baumann, P., 2010: Factivity and Contextualism, Analysis 70, 82-89 Brendel, E., 2003: Was Kontextualisten nicht wissen, Deutsche Zeitschrift für Philosophie 51, 1015-1032 Brendel, E., 2005: Why Contextualists Cannot Know They Are Right: Self-Refuting Implications of Contextualism, Acta Analytica 20-2, 38-55 Brendel, E., 2007: Kontextualismus oder Invariantismus? Zur Semantik epistemischer Aussagen, in: Referenz und Realität, ed. A. Rami and H. Wansing, Paderborn: Mentis, 11-37 Brendel, E., 2009: Contextualism, Relativism, and Factivity: Analyzing ‘Knowledge’ after the New Linguistic Turn in Epistemology, in: Reduction and Elimination in Philosophy and the Sciences, ed. H. Leitgeb and A. Hieke, Frankurt a. M.: Ontos Press, 403-416 Brueckner, A., and C.T. Buford, 2009: Contextualism, SSI and the Factivity Problem, Analysis 69, 431-438 Brueckner, A., and C.T. Buford, 2010: Reply to Baumann on Factivity and Contextualism, Analysis 70, 486-489 16

I am grateful to Elke Brendel, Jochen Briesen, Christopher von Bülow, Wolfgang Spohn, Holger Sturm, and especially to Alexandra Zinke for helpful discussions on the present topic.

94

Wolfgang Freitag

Cohen, S., 1988: How to Be a Fallibilist, Philosophical Perspectives 2, 91-123 Cohen, S., 1998: Contextualist Solutions to Epistemological Problems: Scepticism, Gettier, and the Lottery, Australasian Journal of Philosophy 76, 289-306 Cohen, S., 2000: Contextualism and Skepticism, Philosophical Issues 10, 94-107 DeRose, K., 1992: Contextualism and Knowledge Attributions, Philosophy and Phenomenological Research 52, 913-929 DeRose, K., 1995: Solving the Skeptical Problem, Philosophical Review 104, 1-52 Freitag, W., 2011: Epistemic Contextualism and the Knowability Problem, Acta Analytica 26, 273-284 Hawthorne, J., 2004: Knowledge and Lotteries, Oxford: Clarendon Press Kornblith, H., 2000: The Contextualist Evasion of Epistemology, Philosophical Issues 10, 24-32 Lewis, D., 1996: Elusive Knowledge, Australasian Journal of Philosophy 74, 549-567 Sosa, E., 2000: Scepticism and Contextualism, Philosophical Issues 10, 1-18 Stanley, J., 2005: Knowledge and Practical Interests, Oxford: Clarendon Press Williams, M., 2001: Contextualism, Externalism and Epistemic Standards, Philosophical Studies 103, 1-23 Williamson, T., 2000: Knowledge and Its Limits, Oxford: Oxford University Press Williamson, T., 2001: Comments on Michael Williams’ ‘Contextualism, Externalism and Epistemic Standards’, Philosophical Studies 103, 25-33 Wright, C., 2005: Contextualism and Scepticism: Even-Handedness, Factivity and Surrepticiously Raising Standards, Philosophical Quarterly 55, 236-262

Explaining Variation in Knowledge by Full Belief NICHOLAS SHACKEL, Cardiff University

Suppose we consider the noon train from Vienna to Payerbach-Reichenau. Does it stop at Gloggnitz? I looked at the timetable a while ago and seem to remember that it does. My memory is reasonably reliable. When it is just the question of taking it myself we are happy to say that I know it stops. When, however, Professors Goldman and Hintikka ask me, I am under the strong moral obligation not to mislead eminent philosophers on practical matters and we are not happy to say that I know. In both cases I believe truly that it stops at Gloggnitz and have the same degree of justification by memory. What explains the difference in knowing, or in knowledge attribution?1 Contextualists says that the semantic value of ‘know’ varies with context. In the low stakes context where it is just a matter of myself taking the train, ‘know’ predicates easier knowledge: only moderate justification is required and I have that. In the high stakes context of a query from eminent philosophers, ‘know’ predicates harder knowledge: stringent justification is required but this I lack. Invariantists need to explain away such cases. Sceptics say that the high stakes cases reveal that we don’t even know in low stakes cases. Pragmatists say that knowledge is a norm of practical reasoning and rational action (Hawthorne 2004; Stanley 2005). Because I know something only if it is practically rational to act as if that were the case, the shift to high stakes removes the rationality of acting as if the belief were true and so I don’t know the train stops at Gloggnitz.

1

This example takes the relevant act to be assertion, which is of course a very special kind of act. The link of assertion to belief can introduce complications which I am going to ignore on the whole. If we must avoid them we can do so by taking the context to be one in which I am responsible for taking Professors Goldman and Hintikka to the conference rather than informing them.

96

Nicholas Shackel

I want to explore a different explanation in aid of invariantism: that the change in knowledge is due to a loss of full belief in the high stakes context. Since knowledge implies full belief, loss of full belief leads to loss of knowledge. I concede that this may lack initial appeal but I think it has more going for it than might have been expected. Before I turn to it’s defence, however, let me mention a significant virtue of the explanation. What distinguishes the sceptic from the pragmatist is the former thinks stakes are not epistemically normative whereas the pragmatist accepts that they are. In so doing the pragmatist gives up on intellectualism, that ‘what matters for knowledge are truth-relevant factors, not practical stakes’ (McGrath 2010, 384). De Rose regards a virtue of contextualism to be that ‘it allows one to uphold intellectualism’ (2009, 189) and I think he is right. The full belief explanation, if successful, permits us to retain intellectualism without becoming sceptics or contextualists. There is controversy over full belief and some would like to do away with it altogether.2 I shall not discuss this controversy but characterise what of full belief is sufficient for the full belief explanation. It is sufficient that full belief be distinct from partial belief insofar as it is treated differently in these two ways  downstream rationality of full belief in A: poised to reason on the basis of A without reconsidering A  downstream practicality of full belief in A: poised to act as if A without reconsidering A These conditions are individually necessary for full belief. I think they may also be jointly sufficient but we don’t need that here. Corresponding to rationality and practicality are the probability of A and the stakes of A, which I define to be how bad it is to act as if A when ¬A. Our degree of confidence in A is responsive to these two dimensions and we fully believe A only if we are confident enough. This sounds like Foley’s Lockean thesis: ‘one believes … just in case one is sufficiently 2

For eliminativism see Jeffrey 1970; Stalnaker 1984; Maher 1993; Vogel 2008. For reductionism see Kaplan 1996; Fantl and McGrath 2002; Christensen 2004; Weatherson 2005; Ganson 2008; Foley 2009. For non-reductionism see Weisberg 2010; Ross and Schroeder online; Wedgwood online.

Explaining Variation in Knowledge by Full Belief

97

confident’ (Foley 2009, 37). However, he identifies confidence with credence whereas I am saying that confidence depends on both credence and stakes, increasing with increases in credence but decreasing with increases in stakes. The upshot of this is that full belief is a partly theoretical and partly practical state with the cognitive direction of fit and it is flighty in the face of what is practically salient. It is the flightiness of full belief that suffices for the full belief explanation. In our case my credence in the train stopping at Gloggnitz remains the same but when Professors Goldman and Hintikka ask me the increase in stakes reduces my confidence sufficiently to make me hesitate in saying it does. So I have lost the downstream practicality of my earlier full belief and no longer fully believe that it stops, hence no longer know. I shall now turn to a number of challenges to the full belief explanation. Feldman (2001) drew to our attention the phenomenon of constancy of evaluation. MacFarlane’s example (2005, 202) is particularly clear. A judge quizzes a witness who says he knows the car was on the driveway. When asked if he could rule out it being stolen he concedes that he can’t. The constancy here is that he doesn’t now say that a moment ago he knew but now he doesn’t. He says he never knew. Taking my Gloggnitz example again, on my account it seems that I should say ‘I knew it stopped at Gloggnitz but now I don’t’, contravening constancy. It would sound odd to say that but it would be true all the same. It might merely be unassertable by me. For example, under a norm that requires knowledge for assertion, to say ‘I knew it’ requires that I know that I knew it, which entails that I know that it was true, which entails that I know it. Whereas saying ‘I don’t know it’ requires that I know that I don’t know it which entails that I don’t know it. So the assumption that ‘I knew it but now I don’t’ could be properly asserted entails a contradiction. This response is one a contextualist might use in the face of this problem. It works as well here and I suspect that most of the defences contextualists can use to constancy can be applied to the full belief account. Furthermore, constancy is a different kind of problem for the contextualist than it is for me. In the full belief explanation I have stopped believing that the train stops at Gloggnitz and that in itself explains why I would be reluctant to say that I knew it. The contextualist does not have this resource. So I doubt that constancy is serious problem. The next challenge for the full belief explanation is that even if it succeeds, it is simply another variety of pragmatic invariantism, whereas I

98

Nicholas Shackel

advertised it as a defence of intellectualist invariantism. This objection is correct if all that is meant by pragmatism is that stakes affect whether you know. Ross and Schroeder (online) seem to take it this way. Similarly, if we take pure intellectualism to be the denial of any practical influence over knowledge then we don’t have pure intellectualism. However, pragmatism can also be taken to be a matter of the ‘epistemic standard relevant for determining extension of ‘know’ [varying] with context’ (MacFarlane 2005, 218). It is this kind of pragmatism, pragmatic encroachment on the normative aspect that distinguishes true belief from knowledge, that it is my object to avoid. So the full belief explanation permits intellectualism where it matters to me. The pragmatic encroachment on knowledge can be confined to the practicality of full belief and thereby excluded from normative involvement. One of McGrath’s arguments poses a clear problem for the compatibility of the full belief explanation with intellectualism: 1. If assertibility varies across Low and High [stakes contexts], then so does believability. 2. If believability varies across Low and High, then generalized intellectualism is false. 3. If assertibility varies across Low and High, then generalized intellectualism is false (McGrath 2010, 389-90). That assertibility varies is widely accepted and hence by modus ponens this argument takes us to the falsity of intellectualism. Furthermore, since I seem to be committed to believability varying the second premiss alone gets me into trouble. However, we need to consider the question of modality: is it metaphysical or normative? If it is metaphysical then I deny the second premiss. The mere fact that it is metaphysically possible for belief to vary says nothing about whether ‘what matters for knowledge are truth-relevant factors, not practical stakes’ McGrath 2010, 384. The variance could be merely accidental, or an error, whereas I take intellectualism to deny pragmatic encroachment on the normative aspect that distinguishes true belief from knowledge. So we need the modality to be normative, and in fact, that is what McGrath intends. If the modality is normative then my reply is more complicated. McGrath characterises believability thus:

Explaining Variation in Knowledge by Full Belief

99

P is believable for you iff believing P is proper at least as far as your strength of epistemic position with respect to P is concerned, i.e., iff you don’t need a stronger epistemic position with respect to P for believing P to pass epistemic muster. (McGrath 2010, 394)

Now here we need to distinguish two positions, one is evidentialism for epistemic justification and the other is evidentialism in the ethics of belief. It is often assumed that these go together but this assumption is unwarranted. Pragmatism in the ethics of belief is compatible with evidentialism for epistemic justification simply because it denies that evidence is sufficient for determining what ought to be believed.3 If we take believability to be a matter of pragmatism for epistemic justification the argument is question begging. If we take believability to be a matter of evidentialism for epistemic justification then it is incompatible with the first premiss. In the cases of interest it is stipulated that epistemic justification as the evidentialist understands it remains constant, so believability remains constant. Yet assertibility can vary for the pragmatic reasons. If we take believability to be a matter of evidentialism in the ethics of belief then the first premiss fails in a similar way. What you ought to believe is determined solely by the evidence whereas what you ought to say can be subject to practical constraints. Hence assertibility can vary whilst believability does not. If we take believability to be a matter of pragmatism in the ethics of belief then for at least some such pragmatisms the first premiss will be true. However, the second will not be. Pragmatism, in denying that epistemic justification is sufficient for determining what ought to be believed, is not denying that what matters normatively for knowledge is evidence. In short, and only apparently paradoxically, pragmatists in the ethics of belief can be evidentialists about epistemic justification. So what you ought to believe can vary for practical reasons even though the evidence hasn’t changed. Hence pragmatism in the ethics of belief allows that believability can vary whilst intellectualism for knowledge is true.

3

Unless you define epistemic justification in terms of what ought to be believed, but that is merely a verbal evasion of the problem and muddles what we mean by evidence.

100

Nicholas Shackel

We have now examined the metaphysical and normative for believability and McGrath’s argument fails under each one. So whether downstream rationality and practicality are taken to be about the metaphysics of belief, or to be normative constraints on belief, the full belief explanation can be defended. What strikes me as a bigger problem for the full explanation account is a case I call the heedless believer. This is someone who, because they are heedless of stakes, doesn’t lose their full belief. Under my explanation they still know. My answer here is that yes, they know, but they ought not to believe. In the broader scheme of things such a possibility is something which I wish to allow since I do not accept that epistemology settles in any straightforward way the ethics of belief. The heedless believer is defective, certainly. They believe what they should not, but their failure is not a failure of response to epistemic standards but to practical standards. There is nothing wrong with their credence but with their continuing to be poised to act without reconsidering. If they were as they ought to be they would be practically sensitive, not poised to act, thereby not believe and so not know. Defects have multiple consequences and since what is epistemically in place is sufficient for knowledge (given a true belief) the defect in belief results in ethically defective knowledge. I regard this as a feature not a bug but I concede that on this issue intuitions may diverge. I want to finish by illustrating broader explanatory powers of the full belief explanation by considering a couple of cases in which competitor explanations have some degree of success. The first is the problem of ordinary knowledge in conflict with lottery based propositions. So I know that I can’t afford an expensive holiday but I don’t know that I won’t win the lottery (cf. Hawthorne 2004, 2). If I don’t know I won’t win then I don’t know that I won’t be rich so I don’t know that I can’t afford an expensive holiday. Towards the end of his book Hawthorne briefly considers what he calls the Belief Removal Model for explaining the ‘knowledge-undermining role of salience’: salience might destroy knowledge that p by destroying belief that p. It seems to one that it might be that non-p (where the ‘might’ is here one of epistemic modality). This in turn induces one to stop believing that p’ (Hawthorne 2004, 169)

Explaining Variation in Knowledge by Full Belief

101

He gives this short shrift because ‘there seems to be a perfectly reasonable sense of “belief” in which one believes one will lose a lottery even when the possibility of error is salient’ (Hawthorne 2004, 169). It seems clear that my belief that I can’t afford an expensive holiday satisfies both downstream rationality and practicality. However, what about my attitude to whether I won’t win the lottery? It looks as if it satisfies rationality but I think it fails practicality. If I fully believe it then practicality requires that I am poised to act without reconsidering. But I am not so poised. If I were then, for example, in clearing out my wallet of receipts and scrap paper the lottery ticket would join them in the bin, but it doesn’t. So I suspect that Hawthorne’s ‘perfectly reasonable sense of “belief”’ is just high credence. We can therefore explain the difference in knowledge by the difference in full belief. The second is the challenge of scepticism. Contextualism explains the conflict between, for example, knowing I have hands but not knowing I am not a brain in a vat, by a shift in epistemic standards. In the ordinary context ‘know’ means easy knowledge but once sceptical possibilities have been introduced ‘know’ means hard knowledge. In this case my belief I have hands satisfies both rationality and practicality. My attitude to not being a brain in a vat satisfies practicality but fails rationality (so this case is the other way round from the lottery case). I remain poised to act as if I am not a brain in a vat. The very fact that, once raised, we engage in the dispute over scepticism, shows that we are no longer poised to reason as if we are not a brain in a vat without reconsidering it (this may also apply to my attitude to having hands). So once again, we can explain the difference in knowledge by a difference in full belief. In conclusion, it has been my purpose to explore the full belief explanation of the cases used to motivate contextualism. The explanation itself is straightforward. I think I have shown that it can deal effectively with some obvious objections (granted some assumptions in the ethics of belief with which I am sympathetic) and illustrated briefly its wider explanatory potential. So despite the fact that the full belief explanation is perhaps not initially appealing I think it deserves further attention.

102

Nicholas Shackel

References Christensen, D.P., 2004: Putting Logic in Its Place: Formal Constraints on Rational Belief, Oxford: Oxford University Press DeRose, K., 2009: The Case for Contextualism: Knowledge, Skepticism, and Context, Oxford: Oxford University Press Fantl, J. and M. McGrath, 2002: Evidence, Pragmatics, and Justification, Philosophical Review 111, 67-94 Feldman, R., 2001: Skeptical Problems, Contextualist Solutions, Philosophical Studies 103, 61-85 Foley, R., 2009: Beliefs, Degrees of Belief and the Lockean Thesis, in: Degrees of Belief. ed. F. Huber and C. Schmidt-Petri, Springer Ganson, D., 2008: Evidentialism and Pragmatic Constraints on Outright Belief, Philosophical Studies 139, 441-458 Hawthorne, J., 2004: Knowledge and Lotteries, Oxford: Oxford University Press Jeffrey, R.C., 1970: Dracula Meets Wolfman: Acceptance Vs. Partial Belief, in Induction, Acceptance, and Rational Belief, ed. M. Swain, Dordrecht: Reidel Kaplan, M., 1996: Decision Theory as Philosophy, Cambridge: Cambridge University Press MacFarlane, J., 2005: Assessment Sensitivity of Knowledge Attributions, in: Oxford Studies in Epistemology, ed. T. Gendler and J. Hawthorne, Oxford: Clarendon Press Maher, P., 1993: Betting on Theories, Cambridge: Cambridge University Press. McGrath, M., 2010: Contextualism and Intellectualism, Philosophical Perspectives 24, 383-405 Ross, J., and M. Schroeder, online: Belief, Credence, and Pragmatic Encroachment. Online at http://www-bcf.usc.edu/~jacobmro/ppr/Belief_Credence_and_ Pragmatic_Encroachment.pdf Stalnaker, R., 1984: Inquiry, Cambridge, Mass.; London: MIT Press Stanley, J., 2005: Knowledge and Practical Interests, Oxford: Oxford University Press Vogel, J. 2008: Epistemic Bootstrapping, Journal of Philosophy 105, 518-539 Weatherson, B. 2005: Can We Do without Pragmatic Encroachment? Philosophical Perspectives 19, 417-443 Wedgwood, R. online: Outright Belief. Online at http://www-bcf.usc.edu/~wedgwood/ Outright_Belief.pdf Weisberg, J. 2010: Bootstrapping in General, Philosophy and Phenomenological Research 81, 525-548

The Contextualist Promise SEBASTIAN SCHMORANZER, University of Cologne

Abstract One basic reason for subscribing to contextualism is that it seems to promise well in solving the skeptical puzzle. One contextualist theory arguing this way is a certain version of a relevant alternatives approach whose proponents maintain that skeptical hypotheses are irrelevant in ordinary contexts but relevant in skeptical contexts. I argue that this approach cannot keep its promise because skeptical alternatives can be shown to be always relevant. I I take the basic idea of epistemological contextualism to be expressed in the following thesis: It is possible that a person S knows that p relative to one context whereas the person does not know that p relative to another context although the epistemic subject truly believes that p in both cases and her evidential situation is the same. And I take the invarientist about knowledge to deny that contention. An example might illustrate and motivate the contextualist position: Carla believes truly that John was at home on September 1st because Peter called her the following day and told her so. Surely, in this ordinary situation Carla has knowledge. A month later Carla is asked to testify as a witness in court. John is accused of having killed his wife in their commonly shared apartment on September 1st. A lot hinges on the question whether John was at home that day or not. Carla tells her story. John’s attorney brings up the possibility that Peter might not have told the truth. In that situation Carla’s evidence is not good enough to justify a verdict. Carla no longer knows that John was at home on September 1st. Why should one adopt a contextualist position? There are two main reasons to do so. On the one hand contextualism fits well with cases like the

104

Sebastian Schmoranzer

Carla-example. On the other hand contextualism seems to promise well in solving the skeptical puzzle.1 The latter is what I am interested in here. Can the specific version of contextualism I want to focus on keep that promise? In order to answer that question we have to ask what the skeptical puzzle consists in and what kind of solution to that puzzle is at issue. Almost everyone agrees that we know a lot about the external world. Yet, once we are entering a debate with the skeptic the following skeptical argument seems compelling – at least if one is not convinced from the outset that knowledge does not presuppose evidence: (1) (2) (3) (4)

In order to know that one has a hand, one has to be evidentially justified in believing this to be the case. But one is justified in believing to have a hand only if one is evidentially justified in believing that one is not a cleverly deceived handless brain in a vat. However, one is not evidentially justified in denying such a skeptical hypothesis. Therefore, one does not know that one has a hand.

We are puzzled. How can it be, that on the one hand we know a lot while on the other hand the skeptical argument looks so convincing? Contextualism seems to offer a neat solution to this puzzlement by arguing as follows: Within ordinary contexts our standards for knowledge are such that one of the premises is false. Yet, within a debate with the skeptic he somehow manages to raise the standards such that the premises are all true and the argument is beyond reproach. One brand of contextualism arguing this way is a certain version of the contextualist relevant alternatives approach. According to this theory knowledge presupposes evidential justification in every context and the first premise is true no matter what context we are in. Furthermore, it is also the case that we are never evidentially justified in denying skeptical hypotheses since there is absolutely no evidence that we are not brains in a vat. The third premise of the skeptical argument is always true too. Yet, evidential justification and thereby knowledge is a context-relative matter. In order for someone to be justified in believing p to be the case it is necessary that she be justified in denying all

1

Cf. Brady/Pritchard 2005, 162.

The Contextualist Promise

105

relevant alternatives2 to p. But it is not necessary that one be able to rule out all possible alternatives. The relevance of an alternative in turn is context-sensitive. And it seems fairly obvious that within ordinary contexts skeptical hypotheses are irrelevant. When two people at the market are discussing whether there are cucumbers on the shelves it is inappropriate for the customer to put the seller’s positive knowledge claim into doubt by asking whether he is able to rule out the possibility that they are all just brains in a vat. With respect to situations such as these the second premise of the skeptical argument has to be rejected. However, the defendant of the relevant alternatives approach (for short: the relevantist) continues, in skeptical contexts in which we are philosophizing about our epistemic access to an external world in general we tend to be quite cooperative and we are willing to consider skeptical hypotheses to be relevant after all.3 As convincing as this solution to the skeptical puzzle might look at first sight, I do not think that it can stand closer scrutiny. Instead, I want to argue that skeptical hypotheses are always relevant (II). Since my argument essentially hinges on the general validity of a certain closure principle for justification under justified entailment, I will thereafter reject a well-known alleged counter-example to that principle (III). The upshot of the entire discussion amounts to this: If your reason for being a contextualist about knowledge is that you think that the relevant alternatives approach sketched above offers a convincing solution to the skeptical puzzle, you should reconsider that motivation (IV). II One part of the relevantist’s solution consists in denying that skeptical hypotheses are relevant in ordinary contexts. However, there is a straightforward argument to the contrary:

2

A proposition q is an alternative to a proposition p if it is impossible that p and q be both true. An alternative q to a proposition p is a relevant alternative if and only if a justification for p presupposes a justification for non-q. 3 Who is defending this kind of relevant alternatives approach? One can interpret Austin (1962) as holding this position as far as the irrelevance of skeptical hypotheses in ordinary contexts is concerned. One has to interpret Rheinwald (2004) as defending exactly this view. See especially Rheinwald 2004, 350, 358-361, 365.

106

(i) (ii)

(iii)

Sebastian Schmoranzer

For every context: In order to be justified in believing that I have a hand I have to be justified in believing: if I have a hand, then I am not a cleverly deceived handless being. For every context: If I am justified in believing that I have a hand and if I am justified in believing that this entails that I am not a cleverly deceived handless being then I am justified in believing that I am not a cleverly deceived handless being. For every context: In order to be justified in believing that I have a hand I have to be justified in believing that I am not a cleverly deceived handless being.

Premise (i) is quite plausible. Justification presupposes understanding. And understanding what it means to have a real hand implies understanding that this entails that one is not erroneously believing that one has a hand. And this understanding in turn provides conceptual evidence for the view that having a hand implies not being a handless brain in a vat.4 Now, premise (ii) might be more controversial since it presupposes the general validity of the following closure principle for justification: (CJ) If S is justified (in context C) in believing that p and if S is justified (in context C) in believing that p entails q, then S is justified (in context C) in believing that q. Although this might be obvious, it is important to notice with respect to what follows that the entailment relation in CJ does not hold if p is true and q is false. Why should we accept CJ? One reason lending prima facie plausibility is that there is a huge number of cases in which the principle obviously holds. Another rationale can be spelled out as follows: Given the way the entailment relation has been specified, the following principle seems plausible: I am justified in believing that p entails q only if I am justified in believing that it is not the case that p is true and q is false. Otherwise I am not even capable of ruling out one of the clear cases in which the entailment

4

If you deny this, the argument still holds for all of us who are justified in believing that having a hand entails not being a brain in a vat.

The Contextualist Promise

107

relation does not hold. Differently put we get the following principle of justified entailment: (PJE) If I am justified in believing that p entails q, then I am justified in believing that p is false or q is true. Furthermore, the following principle of justified disjunction is intuitively plausible as well: (PJD) If I am justified in believing that p or q is the case then it is not the case that I am justified in believing that p is false although I am not justified in believing that q is true. Suppose that I have no reason to believe that Clara has eaten the cake. Furthermore, I have every reason to believe that Peter has not eaten the cake. How could I then possibly be justified in believing that one of the two has eaten it? Maybe I know that no other person could have been the culprit. But then, since I already know that Peter was not the perpetrator, I am certainly justified in believing that Clara is to blame. Now, both principles taken together, but only taken together,5 imply CJ as can be seen by the following argument (with “J” for “I am justified in believing ...” and “” for the sentential operator “... entails__”): {(PJE)} {(PJD)} {(PJE), (PJD)} {(A1)}

5

(1) J(p  q)  J(¬p v q) (2) J(¬p v q)  ¬(Jp & ¬Jq) (3) J(p  q)  ¬(Jp & ¬Jq) (4) J(p  q)

(PJE) (PJD) (1), (2) by transitivity assumption (A1)

Does the principle of justified entailment (PJE) by itself imply CJ? That is not the case. J(p  q)  J(¬p v q) can be true if J(¬p v q) is true. At the same time it is possible that Jq is false while Jp and J(p  q) are true. Now, you may ask how it is possible that ¬Jq and Jp while J(¬p v q). This cannot be if J(¬p v q) implies ¬(Jp & ¬Jq). But this is an instance of the principle of justified disjunction. Does the principle of justified disjunction (PJD) by itself imply CJ? That is not the case either. J(¬p v q)  ¬(Jp & ¬Jq) can be true if J(¬p v q) is false. At the same time it is possible that Jq is false while Jp and J(p  q) are true. Of course, that J(¬p v q) is false but J(p  q) is true cannot be maintained if one also accepts the principle of justified entailment. But this just shows that the principle of justified disjunction does not by itself imply CJ.

108

{(PJE), (PJD), (A1)} {(A2)} {(PJE), (PJD), (A1), (A2)} {(PJE), (PJD)} {(PJE), (PJD)}

Sebastian Schmoranzer

(5) ¬(Jp & ¬Jq) (6) Jp

(3), (4) by modus ponens assumption (A2)

(7) Jq (5), (6) by logic (8) Jp  (J(p  q)  Jq) (4), (6), (7) by cond. (CJ) (Jp & J(p  q))  Jq (8) by law of importation

Of course, the two reasons in favor of CJ are no knock-down-arguments. But they shift the burden of proof onto those, like the relevantist, who deny the universal validity of the closure principle. They have to provide us with a convincing case in which closure of justification under justified entailment does not hold. Are there such counter-examples? III This cannot be the place to discuss all counter-examples that are to be found in the appropriate literature.6 I therefore want to focus on one case that seems to accommodate well with the general idea of the relevant alternatives approach: Dretske’s famous zebra-example.7 Dretske presents his example as a case against closure of knowledge under known entailment. And he does not understand justification in terms of evidential justification.8 In what follows, however, I take the liberty to discuss his example as a case against closure concerning evidential justification. I am interested in it only in so far as it could be used to sustain the relevantist’s position. Dretske’s story goes as follows: A father visits the local zoo with his son. They are standing in front of a pen with zebras. Seeing a black-andwhite striped, horse-like animal the father says to his son that there is a zebra. It is plausible to suppose that the visual evidence justifies the father’s belief. The father is also justified in believing that zebras are not cleverly painted mules. But, since a cleverly painted mule would look exactly like the animal in the pen the father is not justified in believing that 6 7 8

For an extended discussion see Schmoranzer 2010, chapter II. Dretske (1970). Dretske (1971).

The Contextualist Promise

109

the animal is not a cleverly painted mule. It therefore looks as if the following three propositions hold: (a) (b) (c)

The father is justified in believing that there is a zebra. The father its justified in believing: if it is a zebra, then it is not a cleverly painted mule. The father is not justified in believing that the animal is not a cleverly painted mule.

We seem to be confronted with a plausible case against CJ. Yet, I think that this evaluation of the example is problematic. Following the only interpretation according to which it has any chance of being plausible it is α) still unconvincing and β) useless for the relevantist.9 Let us ask first what evidence is supposed to justify the zebrahypothesis. It is that zebras look exactly like the animal in the pen. Yet, and this is crucial, this visual evidence speaks in favor of the zebra-hypothesis only if the father has some general background information about how likely it is that – in zoos – something looking like a zebra usually is a zebra. He has to know something like “Zoo authorities normally do not exhibit painted animals”. Why is that? Imagine a case in which a Martian lands in the zoo. Before he arrives he has been shown a picture of a zebra subscribed “zebra or cleverly painted mule”. And the Martian has been told that zebras and mules are of different kinds. Furthermore, the Martian has absolutely no idea about the frequency of cleverly painted mules and zebras on earth or in earthly zoos. Now, seeing the animal in the pen he would neither be justified in believing that it is a zebra. Nor would he be justified in believing that it is a cleverly painted mule. Appearance in itself gives him no clue to favor any one of the two hypotheses. That is why he should abstain from 9

In what follows I try to be as charitable as possible. I have to admit that I do not find the example plausible at all. Consider the following dialogue between the father and his son: Father: “Look, there is a zebra. You can tell by its stripes.” Son: “Oh, but it could be a cleverly painted mule.” Father: “You are right, we are not justified in believing that it is not a painted mule since it would look exactly the same. Still, since it has black and white stripes I am justified in believing that it is a zebra.” The father’s reply is extremely awkward. To explain this awkwardness by means of an unwarranted assertibility maneuver à la Dretske (2005, 17f.) is not convincing. I here agree with Hawthorne (2005): 30.

110

Sebastian Schmoranzer

judging.10 If on the other hand, the Martian knew that it is unlikely that zoos exhibit painted animals, the situation would be different. Then he would be allowed to take the appearance as indicator of the fact that the animal is a zebra. What goes for the Martian goes for the father as well. He is justified in believing that the animal is a zebra only if he has some background information speaking against the hypothesis that it is a cleverly painted mule. This information is a precondition in order for the visual evidence to substantiate the zebra-hypothesis.11 Dretske seems to admit this when he writes: You have some general uniformities on which you rely, regularities to which you give expression by such remarks as „That isn’t very likely“ or „Why should the zoo authorities do that?“ Granted, the hypothesis [i.e. that it is a painted mule] (if we may call it that) is not very plausible, given what we know about people and zoos. But the question here is not whether this alternative is plausible, not whether it is more or less plausible than that there are real zebras in the pen, but whether you know that this alternative hypothesis is false. I don’t think you do. (Italics Dretske)12

Given that Dretske already grants the father knowledge concerning the zebra-hypothesis, the passage should be interpreted as follows: The father is justified in believing that the animal is a zebra to an extend that is sufficient for knowledge. Say that he is justified+. But he is only justified in believing that the animal is not a painted mule to a lesser extent which is not sufficient for knowledge. Say that he is only justified-. The example should therefore be evaluated as follows: (a*) (b*) (c*)

10

The father is justified+ in believing that there is a zebra. The father is justified+ in believing: if it is a zebra, then it is not a cleverly painted mule. The father is not justified+ in believing that the animal is not a cleverly painted mule – even though he is justified- to do so.

Can we not say that both hypotheses are justified? Then we would have a situation in which Jp, J(p  q) and J¬q and we would have to deny the following even more plausible principle: Jp & J(p  q)  ¬J¬q. 11 The idea that visual justification presupposes certain independent general background information can be found in Wright 2002, 335f. 12 Dretske 1970, 1016.

The Contextualist Promise

111

But this interpretation of the example is problematic for two reasons. First: As things stand right now, the relevantist has to admit two things. On the one hand he has to say that the zebra-hypothesis is justified to a greater extend than the negation of the mule-hypothesis. On the other hand he has to accept that it is a precondition for the visual justification of the zebrahypothesis that there is some justification speaking against the mulehypothesis.13 Yet, if a justification for the proposition that p presupposes a justification for the proposition that q, then the justification for p cannot be any better than the justification for q. Suppose I was using a tape measure to identify the length of my desk. In order for me to be justified in believing that my desk is 2 meters long I have to be justified in believing that the measuring device is reliable. But I cannot be more certain of the length of my desk than I may be confident that the measuring device is correct.14 The proposed interpretation of Dretske’s zebra-example is therefore unconvincing. Second: Even if it was convincing, the evaluation of the example would be of no use for the relevantist. It would not speak against the following weaker closure principle: (CJ-) If one is justified+ that p and if one is justified+ that p entails q, then one is at least justified- that q. Relying on this principle the skeptic can still stick to the claim that, no matter what context we are in, it is always necessary that we be to some extent justified in believing that skeptical hypotheses are false. But the relevantist already admitted that we are not in the least justified in dismissing skeptical hypotheses because there is absolutely no evidence against them. Therefore, he also has to admit that our external world beliefs are never justified. His promise to save our everyday certainties from skepticism could not be kept. IV To sum up: One reason to hold a contextualist position of the relevantist sort could consist in the belief that the relevantist view presented here 13 14

That was the lesson learned from the Martian example. Cf. Wright 2005, 255.

112

Sebastian Schmoranzer

seems to promise well in solving the skeptical puzzle. According to that approach, skeptical alternatives are relevant and have to be ruled out only in skeptical but not in ordinary contexts. I argued that this solution is unconvincing. Given the plausibility of a certain version of the closure principle for justification under justified entailment one can show that skeptical hypotheses are always relevant. The alleged zebra-counter-example to that principle turned out to be implausible and of no use for the relevantist. Where does this leave us in the contextualism-versus-invarientismdebate? If your motivation to be a contextualist consists in believing the relevant alternatives approach to offer a neat solution to the skeptical puzzle, you should reconsider that motivation. This does not mean that the approach has conclusively been discredited. There may be other, more convincing examples against the closure principle.15 This does not mean either that other contextualist positions fare no better in keeping the contextualist promise. Finally, even if those approaches turn out not to solve the skeptical puzzle either, this does not force one to adopt an invarientist stance. For one thing, the question remains which position best explains 15

If one assumes degrees of evidential justification, there is a vexing “degree”problem for CJ. Suppose that the kind of evidential justification necessary for knowledge has to have a degree of more than ½. Furthermore, suppose that p is justified to a degree of 6/10 and p  q is justified to a degree of 6/10. Then q may be justified only to a degree of less than ½. The skeptic might reply that in the situation at hand p  q is certain and justified to a degree of 1. A different reply would go as follows: Even if CJ cannot be held hostage any more in order to defend the general relevance of skeptical hypotheses, the skeptic could argue as follows by means of a slightly different principle: (i) For every context: I am justified in believing that I have a hand only if I am justified in believing that I have a hand and that this entails that I am not a handless brain in a vat. (ii) For every context: If I am justified in believing that I have a hand and that having a hand entails that I am not a handless brain in a vat, then I am justified in believing that I am not a handless brain in a vat. (iii) For every context: If I am justified in believing that I have a hand, then I am justified in believing that I am not a handless brain in a vat. Now, the first premise might be more controversial than its counterpart in the original argument (see section II). But I maintain that it can still be defended. Justification for p presupposes a justification for p and the obvious semantic implications. The second premise presupposes the universal validity of the following principle: J(p & (p  q))  Jq. With respect to this principle the “degree”-problem does not arise. And Dretske’s example is unconvincing as far as this principle is concerned too.

The Contextualist Promise

113

cases like the Carla example. Furthermore, there is reason to doubt whether invarientists have a better solution to the skeptical puzzle than their contextualist opponents. I conjecture that they have not and that skepticism is not the pitch on which the struggle between contextualism and invarientism will be decided.16 References Austin, J.L., 1962: Other Minds, in: Philosophical Papers by the late J. L. Austin ed. J.O. Urmson and G.J. Warnock, Oxford: Oxford University Press, 44-84 Brady, M. and D. Pritchard, 2005: Epistemological Contextualism: Problems and Prospects, The Philosophical Quarterly 55, 161-171 Dretske, F.I., 1970: Epistemic Operators, The Journal of Philosophy 67, 1007-1023 Dretske, F.I., 1971: Conclusive Reasons, Australasian Journal of Philosophy 49, 1-22 Dretske, F.I., 2005: The Case against Closure, in: Contemporary Debates in Epistemology, ed. M. Steup and E. Sosa, Oxford: Prentice Hall, 13-26 Hawthorne, J., 2005: The Case for Closure, in: Contemporary Debates in Epistemology, ed. M. Steup and E. Sosa, Oxford: Prentice Hall, 26-43 Rheinwald, R., 2004: Skeptische Herausforderung – Eine Diagnose, Zeitschrift für Philosophische Forschung 58, 347-372 Schmoranzer, S., 2010: Realismus und Skeptizismus, Paderborn: Mentis Wright, C., 2002: (Anti-)Scepticism Simple and Subtle: G. E. Moore and John McDowell, Philosophy and Phenomenological Research 65, 330-348 Wright, C., 2005: Contextualism and Scepticism: Even-Handedness, Factivity and surrepetitiously raising standards, The Philosophical Quarterly 55, 236-262

16

In memory of my teacher Rosemarie Rheinwald. I thank Nikola Kompa, Flavia Mormann and Ansgar Seide for comments on earlier drafts of this paper.

II. Epistemic Virtues

Intellectual Virtues and Their Place in Philosophy JOHN GRECO, Saint Louis University

Part One of the paper reviews some early thinking about how to characterize virtue epistemology and about how virtue epistemology is related to virtue ethics. In that context I review a dispute about how to understand the intellectual virtues, and I reconsider a consensus that has emerged on that question. Part Two of the paper defends a neo-Aristotelian account of understanding as knowledge of “causes,” and an analogous account of wisdom as understanding regarding the most important things. Part Three compares the neo-Aristotelian account of understanding to Ernest Sosa’s notion of Reflective Knowledge. Part One. Virtue Theories in Ethics and Epistemology: The Place of Intellectual Virtues 1. Virtue Ethics and Virtue Epistemology One popular way of characterizing virtue ethics is fairly weak: A virtue ethics is one that makes moral virtues such as courage and justice a focus of moral inquiry. In sum, a virtue ethics gives the moral virtues pride of place. Virtue epistemology may be given an analogous weak formulation, giving the intellectual virtues pride of place in epistemic inquiry. A stronger way of characterizing virtue ethics is more substantive: A virtue ethics makes the moral virtues theoretically fundamental; that is, the normative properties of persons (virtues) are understood to be the most fundamental moral properties, in the sense that other moral properties are to be understood in terms of these. Most famously, this sort of virtue theory defines right action in terms of virtuous character: right action is equivalent to virtuous action, or action that manifests virtuous character. A “pure” virtue theory also defines moral goodness in terms of virtuous character. A “eudaimonistic” virtue theory makes the good and the virtuous equally fundamental, or inter-definable. Virtue epistemology may be given an

118

John Greco

analogous strong formulation. Whereas a strong virtue ethics takes the moral virtues as fundamental, and then uses these to understand other moral categories, such as right action, moral good, and moral responsibility, a strong virtue epistemology takes the intellectual virtues as fundamental, and then uses these to understand other epistemic categories, such as justified belief, good evidence, and intellectual responsibility. Accordingly, the stronger kind of virtue theory, in ethics or epistemology, adopts a substantive thesis about the “direction of explanation” or “direction of analysis”: The less fundamental is to be understood in terms of the more fundamental, and the normative properties of persons (the moral or intellectual virtues) are the most fundamental normative properties (Sosa 1980; Zagzebski 1996). Here is a second thesis about the direction of explanation, often adopted by strong virtue theorists: Normative properties (fundamental and non-fundamental) can be understood in terms of non-normative properties. This generates the following ontological picture: Non-fundamental normative  Fundamental normative  Non-normative Counting against this second thesis about direction of analysis, we have philosophy’s track record; that is, no reduction of the normative to the nonnormative has been convincingly displayed. In favor of the thesis is R. M. Hare’s worry about unanchored normative properties (Hare 1952, 118ff). Suppose we have two cars, Hare suggests, that are exactly alike in terms of their physical properties: for example, they are exactly alike in terms of power, acceleration, gas mileage, color, shape, etc. Would it then be coherent to hold that the two cars are not alike in their normative properties – that one is a “good” car whereas the other is a “bad” or “poor” one? That seems absurd, for it would seem that the first car must be good in virtue of its physical properties. And of course analogous reasoning would seem to hold for morally normative properties (Hare 1952) and epistemically normative properties (Sosa 1980). In any case, it is common among strong virtue theorists to embrace this second thesis about direction of analysis. Specifically: The normative properties of persons (the virtues) are fundamental for the normative realm.

Intellectual Virtues and Their Place in Philosophy

119

These, in turn, can be understood in terms of non-normative properties, such as psychological properties, causal properties, motivational structure, dispositions to act, etc. 2. Two Kinds of Intellectual Virtue Virtue epistemologists have talked about two major kinds of intellectual virtue. First, some authors have focused on “character virtues,” such as intellectual courage and intellectual fairness, which they suppose to be closely analogous to the moral virtues (eg. Zagsebski 1996). Second, some authors have focused on “faculty virtues,” such as accurate perception and sound reasoning, which they suppose to be intellectual abilities or powers rather than character traits (eg. Sosa 2007; Greco 2010). Early on in the literature, a dispute arose concerning what now looks to be a bad question: What are the intellectual virtues really like–are they intellectual character traits or are they intellectual powers? Now with hindsight, the clear answer seems to be this: There is more than one kind of intellectual virtue, insofar as there is more than one kind of person-level intellectual excellence. That is, both intellectual character traits such as intellectual courage and intellectual carefulness, and powers of the mind such as accurate perception and keen memory, are person-level excellences and therefore deserve the label of “intellectual virtue.” Putting aside what looks to be a terminological question, the more substantive question is this: What can these different kinds of virtue do for us in epistemology? In other words, what kind of theoretical work can they do (Greco 2002)? In this context, somewhat of a consensus has emerged. Namely, it is now widely agreed among virtue epistemologists that knowledge can be understood in terms of the faculty virtues. Roughly, knowledge is true belief produced by reliable intellectual faculties, such as good vision, or sound reasoning, or keen memory. Somewhat less roughly: In cases of knowledge, S believes the truth because S believes from some such ability (Sosa 2007; Greco 2003; 2010). A second part of the consensus is that the character virtues do not plausibly figure into an account of knowledge. Not essentially, in any case. For example, one might know by the power of sight that one is in the path of an oncoming truck, and even if one’s character is not intellectually virtuous. In fact, one might strain to be closed-minded about the prospect, and yet fail to avoid knowing about the truck (Greco 2000).

120

John Greco

But then what place in epistemology is there for the character virtues? Here a number of proposals have been made. For example, Jason Baehr has argued that, at least in many cases, the character virtues are necessary to turn mere faculties into faculty virtues, or powers or abilities proper (Baehr 2011). Alternatively, Christopher Hookway has emphasized the role of the character virtues in the “ethics of inquiry” and “ethics of belief” (e.g. Hookway 1994). Finally, a number of theorists have argued that the character virtues are essential for understanding “higher” epistemic goods such as understanding and wisdom. Whereas mere knowledge requires only sound cognitive powers, such as powers of perception and reasoning, these more valuable epistemic goods require virtuous character as well. (e.g. Zagzebski 1996) In Part Two I will argue against this last suggestion. There I will defend a neo-Aristotelian account of understanding and wisdom, on which both of these are species of knowledge. The argument is roughly this: Just as the character virtues do not figure essentially into an account of knowledge in general, neither do they figure essentially into an account of understanding or wisdom in particular. True enough, the character virtues often are necessary to bring about or enable understanding and wisdom, but that is more a fact about us than about the nature of these higher epistemic goods. In that respect, the character virtues have the same relation to these goods as they do to knowledge in general. Part Two. A Neo-Aristotelian Account of Understanding and Wisdom 1. Aristotle’s Account of Episteme According to Aristotle, episteme (variously translated as “knowledge,” “scientific knowledge,” “understanding”) consists in knowledge of causes. To have episteme is to know the cause of a thing. An important aspect of Aristotle’s account concerns the relationships among a) episteme, b) having the answer to “Why” questions, and c) being able to cite causes. Specifically, to have episteme regarding some fact that p is to have an account or explanation regarding why p is the case. To have such an account, in turn, is to be able to (knowledgeably) cite causes. So again, episteme consists in knowledge of causes. Explicating Aristotle’s account, R. J. Hankinson writes,

Intellectual Virtues and Their Place in Philosophy

121

To have scientific knowledge, then, is to have explanatory understanding: not merely to ‘know’ a fact incidentally, to be able to assent to something which is true, but to know why it is a fact. The proper function of science is to provide explanations ... (Hankinson 1995, 110).

Famously, Aristotle thought that there are four kinds of cause: efficient, material, formal, and final. Aristotle’s notion of efficient cause is closest to our own notion of cause. Roughly, an efficient cause is a source or agent of change. For example, fire is the efficient cause of the wood’s burning. The explosion was the efficient cause of the house’s catching fire. But Aristotle recognizes other kinds of cause as well. A material cause is, roughly, the material out of which a thing is made. For example, the material cause of the house’s burning is that it is made out of wood. Here again we note the close connection between Aristotle’s four causes and the various answers we can give to “Why” questions. For example, someone might ask, “Why is that house over there in ruins?” In some contexts, we will be inclined to cite the efficient cause – it was a fire, an explosion. But in other contexts we might cite the material cause – the house was made out of wood, or of straw. (For example, that is why the first two little pigs’ houses are in ruins, whereas the third little pig’s house is still standing.) A formal cause is a thing’s “nature” or “essence” or “what-it-is.” For example, we might say that “the cause” of the dog’s barking is that it is a dog. In other words, that’s what dogs do – they bark! Lastly, a final cause is an end or goal. The easiest place to see what Aristotle has in mind is in the case of human action. Thus, we commonly answer “Why” questions by citing what a person is trying to do or trying to achieve. For example, “Why is she running down the road?” “Because she is trying to lose weight.” Or: “Because she wants to get home in time for dinner.” Notice, finally, that we can answer the same “Why” question by citing any one of Aristotle’s four causes. Why did the house burn down? There was an explosion (efficient cause). It was made of wood (material cause). The owner wanted to collect on the insurance (final cause). We might even cite a formal cause here: “Sometimes houses burn down,” said in answer to the insurance agent, trying to understand why this happened, just now, in this economy. What do Aristotle’s four causes have in common? One way to think of it is that each constitutes some kind of “dependence relation.” In other words, each amounts to some way in which one thing can depend on another. Thus the house’s burning down depended on there being an explo-

122

John Greco

sion. But it also depended, in various ways, on the house’s being made of wood, the owner wanting to collect insurance, and the fact that houses are the sort of thing that can burn down. Put differently, each of these things is relevant to the fact that the house burned down. Consider: not everything can burn down, and not everything that can burn down does burn down. To understand why this house burned down – to understand it fully – requires knowing how the house’s burning down depended on these various factors. Think of a complex net, constituted by these various kinds of dependence relations. According to Aristotle, to have episteme regarding some thing is to know its location in such a net. 2. A Neo-Aristotelian Account of Understanding Notice the tight relations between a) Aristotle’s episteme, b) knowing the cause, c) being able to cite the cause, c) having an account or explanation, and d) having the answer to a “Why” question. Our contemporary concept of understanding displays similar tight relations. This is a clue to a close relationship between Aristotle’s episteme and our understanding. In this section we update Aristotle’s account so as to make the relationship even tighter. That is, we give a neo-Aristotelian account of understanding. First, we may replace Aristotle’s “four causes” with dependence relations in general. As we saw above, Aristotle recognizes a) efficient causal relations, b) constitutive (or “material”) relations, c) essential (or “formal”) relations, and d) teleological (or “final”) relations. These, we noted, are various sorts of dependence relations – they are ways in which one thing (or process, or event) can depend on another. But there are other dependence relations as well. For example, there are e) part-whole or “mereological” relations, f) logical and mathematical relations, g) conceptual relations, and h) supervenience relations of varying strength. The list is not meant to be exhaustive. Rather, the substantive point is that there are many and various dependence relations, and understanding centrally involves knowledge of these. Again, think of a complex net of many and various modally strong dependence relations. According to our neo-Aristotelian account, to have understanding regarding some thing is to know its location in such a net. The present account preserves close relations between a) understanding, b) knowing the cause, c) being able to cite the cause, c) having an account or explanation, and d) having answers to “Why” questions. Another feature

Intellectual Virtues and Their Place in Philosophy

123

of the account is that it makes causal explanation (in our more restricted sense of “cause”) a species of explanation in general. To have an explanation is to be able to cite appropriate dependence relations. To have a causal explanation is to be able to cite causal relations. In similar fashion, the account makes scientific understanding (in our more restricted sense of “science”) a species of understanding in general, including mathematical understanding, philosophical understanding, and practical understanding. A second way to update Aristotle’s account is to stress that understanding consists in systematic knowledge of dependence relations. Put differently, understanding consists in knowledge of a system of dependence relations. This accommodates the idea that understanding, unlike mere knowledge, cannot be isolated. (Riggs 2003; Kvanvig 2003, 205) It also accommodates the idea that understanding comes in degrees, in terms of both breadth and depth. We can think of “depth of understanding” in terms of “depth of knowledge,” where the latter corresponds to knowledge of more fundamental dependence relations. Likewise, we can think of “breadth of understanding” in terms of “breath of knowledge,” where the latter corresponds to knowledge of more diverse dependence relations. Finally, the present account accommodates the idea that understanding can have diverse objects of understanding (Grimm 2010). In particular, it accommodates the idea that understanding can have “non-propositional” objects, such as maps, graphs, pictures, and models, as well as “propositional” objects such as theories, narratives and mathematical equations. This is because all of these involve complex representations of relations, or representations of complexes of relations. So far, we have argued that understanding consists in a systematic knowledge of dependence relations, where both relations and relata (the objects of the relations) may be of various sorts. We may now draw a further distinction among the various objects of understanding, one that will become important later. Namely, understanding can take as its object: (a) (b) (c)

A system of “real” relations, or relations “in the world.” For example: an ecosystem, an economy, a machine, a historical event. A representation of a real system. For example: a theory, a narrative, a model, a set of equations. The relations between a real system and a representation. For example: relations between a model and the economy that it represents, relations between a theory and a causal process that it

124

John Greco

represents, relations between a diagram and a machine that it represents, relations between a narrative and a historical event that it represents. In each case, we can make a distinction between the object of understanding and the vehicle of understanding, i.e. between the thing understood and its representation. In case (a), understanding will involve a representation of some part of “the world”. In case (b), understanding will involve a representation of a representation. In case (c), understanding will involve a representation of a relation between representation and world. To summarize, understanding consists in a systematic knowledge of dependence relations, where dependence relations can be of various sorts, including “real” relations between parts of the world, conceptual and logical relations between parts of a theory, and semantic relations between theory and world. Our neo-Aristotelian account also explains why scientific explanation is only one kind of explanation, and it locates scientific understanding within a unified account of understanding in general. Finally, this “understanding of understanding” accommodates various pretheoretical data nicely. In particular: it explains why one can have knowledge without having understanding; why understanding cannot be isolated or episodic; why understanding is closely tied to explanation and answering “Why” questions; and why understanding involves coherence, especially explanatory coherence. (Greco, forthcoming) 3. A Neo-Aristotelian Account of Wisdom With this account of understanding in hand, we are now in a position to give a neo-Aristotelian account of wisdom. Roughly, Aristotle thought that wisdom (sophia) is understanding of the “highest” or “most important” things. Plugging this into the present account of understanding: Wisdom is systematic knowledge of the most important dependence relations. How should we understand “highest” or “most important” here? I don’t want to say too much. I lack the wisdom to do so. But I am happy to follow Aristotle here in rough outline. Specifically, we can make a distinction between practical wisdom, or understanding concerning how to live well, and theoretical wisdom, or understanding concerning ultimate causes. In this respect, wisdom is understanding, and therefore knowledge, concern-

Intellectual Virtues and Their Place in Philosophy

125

ing the fundamental structure of the world, both moral/practical and ontological. 4. An objection to the neo-Aristotelian account of understanding The neo-Aristotelian account has much to recommend it, but it does face objections. The one I will focus on here is that understanding cannot be a species of knowledge because understanding does not require truth. Put differently, knowing that p entails that p is true, but understanding that p does not. Accordingly, understanding cannot be knowledge of dependence relations, since it is not a species of knowledge at all. Catherine Elgin has developed a powerful version of this kind of objection. According to Elgin, understanding need only be “true enough,” as opposed to strictly true. In fact, Elgin argues, this feature is essential to an adequate account of understanding. First, if we restrict understanding to what is true, then much of what we count as understanding falls by the wayside. In fact, some of what seem to be paradigm instances of understanding will no longer count. Despite the fact that Copernicus’s central claim was strictly false, the theory it belongs to constitutes a major advance in understanding over the Ptolemaic theory it replaced. Kepler’s theory is a further advance in understanding, and the current theory is yet a further advance. The advances are clearly cognitive advances. With each step in the sequence, we understand the motion of the planets better than we did before. But no one claims that science has as yet arrived at the truth about the motion of the planets (Elgin 2007, 37-8).

Second, if we think that understanding must be true, then we will miss much about the nature and role of scientific theories, and the relationship of such theories to scientific understanding. ... science routinely transgresses the boundary between truth and falsehood. It smooths curves and ignores outliers. It develops and deploys simplified models that diverge, sometimes considerably, from the phenomena they purport to represent. Even the best scientific theories are not true (Elgin 2004, 113). The problem comes with the laws, models, idealizations, and approximations which are acknowledged not to be true, but which are nonetheless critical to, indeed constitutive of, the understanding that science delivers. Far from being defects, they figure ineliminably in the success of science (Elgin 2004, 113-114).

The point can be generalized.

126

John Greco

Nor is science the only casualty. In other disciplines such as philosophy, and in everyday discourse, we often convey information and advance understanding by means of sentences that are not literally true. An adequate epistemology should account for this as well (Elgin 2004, 114).

Elgin uses the ideal gas law to make her point. Some laws never obtain. They characterize ideal cases that do not, perhaps cannot, occur in nature. The ideal gas law represents gas molecules as perfectly elastic spheres that occupy negligible space and exhibit no mutual attraction. There are no such molecules. Explanations that adduce the ideal gas law would be epistemically unacceptable if abject fidelity to truth were required (Elgin 2004, 118).

I want to reply to Elgin’s objection by exploiting a distinction made above, regarding the various objects of understanding. Specifically, recall the distinction between (a) a system of real relations in the world, (b) a representation of a real system, and (c) the relations between a real system and a representation. Any of these, we said, can be a proper object of understanding. Keeping this in mind allows us to see that understanding indeed tracks knowledge and is therefore factive: understanding that p always entails knowing that p, and hence that p is true. To illustrate, consider two cases in which a student of chemistry is studying the ideal gas law. Case 1. Jill knows what the ideal gas laws says (i.e. she knows relevant facts about the representation), Jill knows that the ideal gas law is an idealization of how actual gases behave in the world (i.e. she knows relevant facts about the representation-world relation), and Jill knows that actual gases behave so as to approximate the ideal gas law (i.e. she knows relevant facts about the world). In all of these instances, Jill has understanding as well. In each instance, she understands the relevant “object,” and understands the relevant facts about the object of understanding. Case 2. Jack knows what the ideal gas law says, but does not know that it is supposed to be an idealization. Accordingly, Jack knows relevant facts about the representation, but he does not know relevant facts about the representation-world relation, and he does not know relevant facts about the world (for example, that actual gases behave only so as to approximate the ideal gas law).

Intellectual Virtues and Their Place in Philosophy

127

With these distinctions in place, we can maintain that knowledge and understanding indeed come and go together. Specifically, Jack understands the representation (the law-statement) insofar as he has systematic knowledge of what it means. But Jack does not understand the relation between the law-statement and the world insofar as he lacks relevant knowledge (that the law is an idealization of what goes on in the world). Likewise, Jack does not understand the behavior of gases in the world insofar as he lacks relevant knowledge (that gases in the world do not instantiate the law). In fact, it seems right to say that Jack misunderstands the relation between the law-statement and the world, and misunderstands the behavior of gases in the world. Again, once our distinction between different objects of understanding is in place, knowledge and understanding seem to come and go together in just the way that the neo-Aristotelian account predicts. We should consider another of Elgin’s points, however. Namely, she notes that we often talk of understanding that is not strictly true, but true enough. For example, we would be happy to say that Jack understands something of the behavior of gases, even if what he believes about the gases is strictly false. But notice that we talk about knowledge in the same way. Thus, let p be that it is 3 o’clock. We are happy to say, “S knows that p,” even when p is strictly false, because it is in fact 3:01. Why? Because p is “true enough.” Similarly for “You know he never declines an invitation” (when in fact he almost never does) and “I know the table is level” (when in fact it is not perfectly level). In other words, we sometimes speak loosely, but in a way that is acceptable (even useful) for present purposes. But then it is hard to see why we cannot handle non-factive uses of understanding claims in the same way. In other words, we can maintain that understanding entails truth, strictly speaking, while acknowledging Elgin’s point about usage. (Greco, forthcoming) Part Three. The Place of Reflective Knowledge Virtue epistemology was introduced into contemporary epistemology by Ernest Sosa, and he remains that approach’s most prominent defender. In this final part of the paper I consider the relationship between Sosa’s notion of “reflective knowledge” and our neo-Aristotelian account of understanding and wisdom.

128

John Greco

Sosa famously defends a two-tier epistemology – one that makes central a distinction between mere “animal knowledge” and more impressive “reflective knowledge” (Sosa 2007; 2009; 2011). His account of animal knowledge follows the general approach to knowledge outlined above: to know is to believe the truth by means of one’s reliable cognitive faculties. In Sosa’s most recent terminology, knowledge is true belief that manifests competence. For example, one has animal knowledge if one believes the truth through reliable perception or memory, or through reliable reasoning from these. Reflective knowledge requires, in addition to this, that one have a perspective on one’s belief and its reliable source. Reflective knowledge, then, involves a kind of second-order knowledge, or knowledge that one knows. Reflective knowledge, then, is characterized by a coherent perspective on one’s reliable cognition. In that sense, the reflective knower understands why she has a true belief, and hence can defend her belief both to herself and to interlocutors. Reflective knowledge is a matter of degree, however. Just as one’s faculties can be more or less reliable, one’s perspective can be more or less detailed and more or less explicit. Finally, reflective knowledge requires that one’s perspective is not by accident – it must itself be produced by a reliable cognitive competence, in this case coherenceseeking reason. On Sosa’s view, human knowledge is typically reflective knowledge, but in many cases the degree of detail and explicitness involved in the relevant perspective will be minimal. Put differently, human knowledge is always at least animal knowledge, but it is also typically reflective to at least some degree. That degree can increase in both detail and articulation, however, so that in some cases reflective knowledge is both extensive and explicit, regarding both the source of one’s belief in reliable cognition and one’s location in favorable circumstances. Sosa’s two-tiered epistemology is designed to address various Pyrrhonian concerns. For one, Sosa proposes a structure of knowledge that stops the regress of reasons in animal knowledge. More specifically, the regress stops in non-reasoning competences such as perception and memory that are in fact reliable in the conditions in which they are exercised. Reflective knowledge, however, requires a coherent perspective on one’s belief and its reliable source. Moreover, this perspective amounts to second-order knowledge of one’s first-order knowledge, and in many cases will be available to the reflective knower as such.

Intellectual Virtues and Their Place in Philosophy

129

In this regard, Sosa endorses the Pyrrhonian point that grasping gold in the dark is inferior to grasping it in the light. That is, grasping gold (or truth) as such is inferior to grasping gold (or truth) that is recognized as such. Reflective knowledge involves the desired light, however, since it involves a second-level perspective on truth and knowledge on the first level. This kind of knowledge is never complete, and for some it is quite minimal. On the other hand, it can be increased indefinitely, both in breath and in depth, as we build up our knowledge of the world and of our place within it (Sosa 1997; 2007, 113-133). On the present account, then, Sosa’s reflective knowledge is a species of understanding. Specifically, it is understanding regarding the relations among one’s beliefs, the cognitive dispositions that give rise to them, and the world. Moreover, Sosa argues, this is a species of understanding that is of considerable importance to the philosophical tradition. Importantly, it is just the sort of understanding that the Pyrrhonian values, and reasons incorrectly that we cannot have (Sosa 2004, 291-292). More recently, Sosa has argued that reflective knowledge can involve another species of understanding as well; namely, understanding regarding how one ought to manage one’s cognitive resources, how to “choose well” regarding what and when to believe (Sosa 2011, especially ch. 1). If this constitutes understanding of “the most important things,” then reflective knowledge is sometimes a species of wisdom as well.1 References Baehr, J., 2011: The Inquiring Mind: On Intellectual Virtues and Virtue Epistemology, Oxford: Oxford University Press Elgin, C., 2004: True Enough, Philosophical Issues 14, 113-131 Elgin, C., 2007: Understanding and the facts, Philosophical Studies 132, 33-42 Greco, J., 2000: Two Kinds of Intellectual Virtue, Philosophy and Phenomenological Research LX, 179-184 Greco, J., 2002: Virtues in Epistemology, in: The Oxford Handbook of Epistemology, ed. P. Moser, Oxford: Oxford University Press, 287-315 1

Thanks to Jason Baehr, Catherine Elgin, Stephen Grimm, Anna Kim, Ernest Sosa and Linda Zagzebski for conversations on relevant material, and to Jonathan Reibsamen for comments on an earlier draft.

130

John Greco

Greco, J., 2003: Knowledge as Credit for True Belief, in: Intellectual Virtue: Perspectives from Ethics and Epistemology, ed. M. DePaul and L. Zagzebski, Oxford: Oxford University Press, 111-134 Greco, J., 2010: Achieving Knowledge: A Virtue-theoretic Account of Epistemic Normativity, Cambridge: Cambridge University Press Greco, J., forthcoming: Episteme: Knowledge and Understanding, in: Virtues and Their Vices, ed. K. Timpe and C. Boyd, Oxford: Oxford University Press Grimm, S., 2010: Understanding, in: The Routledge Companion to Epistemology, ed. D. Pritchard and S. Bernecker, New York: Routledge, 84-94 Hankinson, R.J., 1995: Philosophy of science, in: The Cambridge Companion to Aristotle, ed. J. Barnes, New York: Cambridge University Press, 109-139 Hare, R.M., 1952: The Language of Morals, Oxford: The Clarendon Press Hookway, C., 1994: Cognitive Virtues and Epistemic Evaluation, International Journal of Philosophical Studies 2, 211-227 Kvanvig, J., 2003: The Value of Knowledge and the Pursuit of Understanding, Cambridge: Cambridge University Press Riggs, W., 2003: Understanding ‘Virtue’ and the Virtue of Understanding, in: Intellectual Virtue: Perspectives from Ethics and Epistemology, ed. M. DePaul and L. Zagzebski, Oxford: Oxford University Press, 203-226 Sosa, E., 1980: The Raft and the Pyramid: Coherence versus Foundations in the Theory of Knowledge, Midwest Studies in Philosophy 5, 3-25 Sosa, E., 1997: How to Resolve the Pyrrhonian Problematic: A Lesson from Descartes, Philosophical Studies 85: 229-249 Sosa, E., 2004: Replies, in: Ernest Sosa and his Critics, ed. J. Greco, Oxford: Blackwell Publishing Sosa, E., 2007: A Virtue Epistemology, Apt Belief and Reflective Knowledge, Volume I, Oxford: Oxford University Press Sosa, E., 2009: Reflective Knowledge, Apt Belief and Reflective Knowledge, Volume II, Oxford: Oxford University Press Sosa, E., 2011: Knowing Full Well, Princeton, New Jersey: Princeton University Press Zagzebski, L., 1996: Virtues of the Mind: An Inquiry into the Nature of Virtue and the Ethical Foundations of Knowledge, Cambridge: Cambridge University Press

Understanding’s Tethers CATHERINE Z. ELGIN, Harvard University

Introduction Understanding is holistic. Although we can know isolated facts, understanding a topic – be it the rise of the working class, the defensive strategy of the Boston Celtics, or the function of chlorophyll in photosynthesis – involves an interconnected network of commitments that suitably bear on the relevant facts. This characterization is hideously murky. Much needs to be said about the identity of and interconnections among the networked commitments, the criteria for suitably bearing on the facts, and the cognitive stance that an epistemic agent must take to both the network and the facts. Indeed, my initial claim might seem unnecessarily murky. Many take it as obvious that the strands in the network must be beliefs (perhaps justified or reliably formed beliefs) and that the tie to the facts is truth. Plenty of work would remain to be done, but at least we would have familiar semantic and epistemological resources to draw on. However obvious it might seem, this answer is wrong for at least two reasons. One derives from Gettier cases; the other from modern science. Gettier cases disclose that justified true belief is not sufficient for knowledge because the factors that afford the justification for a belief may have nothing to do with what makes that belief true. In such cases, it is just by luck (and too much luck) that justification and truth align. Understanding is vulnerable to the same difficulty. If the Copenhagen interpretation of quantum mechanics (C) is justified but false, and the theory that the universe is endlessly expanding (E) is unjustified but true, someone who believes C or E has a constellation of justified true beliefs but lacks understanding. It is just by luck (and too much luck) that her constellation of beliefs is true. Mature science generates a different problem. It involves idealizations, models and thought experiments which are not and are not supposed to be true, but which figure centrally and ineliminably in the understanding that science provides. The falsity of such devices does not discredit them. Plato maintained that the difference between doxa and episteme is that, unlike doxa, episteme has a tether. Gettier cases reveal that truth – even justified

132

Catherine Z. Elgin

truth – is an inadequate tether; modern science reveals that falsehoods are not always untethered. I will argue that exemplification supplies a tether that connects epistemologically acceptable theories or accounts to the phenomena they pertain to. Because exemplification involves instantiation, accounts that afford understanding are tied to the facts. Because exemplars need not be true, idealizations, models and thought experiments may figure ineliminably in understanding. To make my case, I need to begin by discussing understanding and exemplification. Understanding The term ‘understanding’ has a variety of uses. I am concerned with understanding as a type of epistemic success. The condition at issue in a sentence like ‘The parties to the lawsuit came to an understanding’ is of no interest to epistemology, for the so-called understanding is a mere matter of agreement. When Ben ventures, ‘As I understand things, the committee has the authority to decide’, the word ‘understand’ is used as a hedge. He thinks so, but he is not sure. Such a use, although of some interest to epistemology, is far from central. Things are more complicated with a sentence like ‘Lesley understands Policur cosmology’, if we take it to mean that she is cognizant of, but not that she endorses, the commitments of Policur cosmology. Her cognizance is a proper topic for epistemological investigation. But it is not, under the reading I gave the sentence, a cognizance of cosmology. When we say ‘Mark understands photosynthesis’, we mean that he understands not, or not just, the theory, but the phenomenon. This is the sort of understanding I am primarily interested in. Typically such understanding is embodied in a theory, so the agent is apt to understand a theory as well. But my topic is the understanding of the phenomena, not merely the understanding of the theory. It might be nice to start (or even to end) with a real definition of ‘understanding’, or a demarcation criterion that sharply distinguishes among understanding, misunderstanding and failing to understand. I have none to offer. But even without them, we have verdicts about what we take to be clear cases. Astronomy affords an understanding of the motions of the planets; astrology does not. Chemistry affords an understanding of the constitution of matter; alchemy does not. An adequate epistemology should largely respect these verdicts. It should count most of what we consider

Understanding’s Tethers

133

clear cases of understanding to be understanding, most of what we consider clear cases of failing to understand to be failing to understand, most of what we consider clear cases of misunderstanding to be misunderstanding; and it should yield principled verdicts on the cases we consider currently undecided. I am not going to venture such an explication here. For my purposes, the main point is that we can identify instances where we are pretty sure that people understand things and instances where we are pretty sure that they do not. These provide touchstones for our epistemological theorizing. One such touchstone is mature science. I take it that our best science embodies an understanding of its subject matter, and that any adequate epistemology should reflect the fact that it does. This is more problematic than it might initially seem. Understanding is an epistemic achievement. Ceteris paribus, our cognitive situation vis à vis a topic is better if we understand the topic than if we do not. The issue is what sort of epistemic achievement it is. Unlike knowledge, understanding is holistic. It seems possible to know isolated facts. One can know (as solvers of American crossword puzzles do) that Edo was the capital of Japan and that Ott played for the Giants, but know virtually nothing else about the history of Japan or the Giants. But understanding involves a body of mutually supportive cognitive commitments. The elements of an understanding must hang together. Moreover, the understander should grasp or appreciate how they hang together. This might suggest that coherence is the hallmark of understanding. But delusions and conspiracy theories are often admirably coherent. The madman who thinks he is Louis XIV interprets the evidence, whatever it might be, to support his delusion; the conspiracy theorist who believes that the UN is plotting to promote vegetarianism does the same. In such cases, each belief nestles comfortably in a network of other beliefs; but many of the beliefs that form the network are unfounded. Coherence then is achieved at the price of plausibility. Seemingly obvious considerations are denied, discredited or warped to shield the belief system from objections. Coherence theories of knowledge block such cases by imposing a truth requirement which discredits delusions and (false) conspiracy theories (Lehrer 1974). An epistemology of understanding might do the same. Then the difference between knowledge and understanding would be one of perspective. Where the epistemological status of an individual proposition p is at issue, the topic is knowledge. Where what is at issue is the cluster of propositions that cohere and collectively vindicate p and one another, or at

134

Catherine Z. Elgin

least those that remain when the false beliefs are excised, the topic is understanding. On such a view, coherence alone is not sufficient for knowledge or understanding.1 Truth is required as well. Still the worry remains. Even when coherence and truth align, a coherent body of truths might float free of the facts, not be tied to them. Truth, as the Gettier cases show, is an inadequate tether. I have argued that understanding consists of a system of cognitive commitments in reflective equilibrium (Elgin 1996). These commitments consist of beliefs, methods, standards, perspectives, and so forth. Together they constitute what van Fraassen calls a stance (2002). Some of the commitments are antecedently acceptable. We have good reason to endorse them independently of the system in question. Others are acceptable only because of their role in the system. For example, the scientific community was originally committed to the existence of positrons, not because it had any direct evidence of positrons, but because it was strongly committed to symmetry principles and to the existence of electrons. If electrons exist and symmetry holds, then there exist positively charged counterparts to electrons – that is, positrons. Because the elements of the system are reasonable in light of one another, they are in equilibrium; because the system as a whole is as reasonable as any available alternative in light of the relevant antecedent commitments, its equilibrium is reflective. It is a system that the community of inquiry can, on reflection, endorse. The commitments a community starts with need not be true (or even truth evaluable) and are not assured a place in the system it eventually endorses. Members of the community realize that their antecedent commitments are to some degree inadequate. But being the current best guesses about the matter at hand, the proper ways to investigate that matter, and the appropriate standards for evaluating results, those commitments have some claim on the community’s epistemic allegiance; investigators need a reason to revise or reject them. Such reasons are often readily available, so a system under construction need not, and typically does not, incorporate all 1

Neither coherence theorists of knowledge nor coherence theorists of understanding insist that every belief in the relevant coherence cluster must be true. According to Lehrer (1974), a true belief that p is known just in case it coheres with the believer’s doxastic system and coheres with the system that would result if all false beliefs in her doxastic system were excised. According to Kvanvig (2003), an agent who understands a topic may harbor a few, relatively peripheral false beliefs about it.

Understanding’s Tethers

135

initial, seemingly relevant commitments. But if the revised system corrects or rejects them, it should, for the most part at least, show why they seemed as reasonable as they did when they did. The resulting system is fallible and is accepted provisionally. Reflective equilibrium can always be upset by further findings. Nevertheless, a system in reflective equilibrium – one does as good a job as any available alternative at accommodating a community of inquiry’s cognitive commitments about a topic, the proper methods for investigating that topic, and the standards that findings should satisfy – constitutes, I maintain, an understanding of that topic. Although there may subsequently be reason to abandon it, such a system is reasonable in the epistemic circumstances. My account is not a pure coherence theory, since a tenable system is answerable to something outside itself – antecedent commitments, many of which are commitments about the topic. But it might seem too close to a coherence theory to be palatable. The worry is this: The tether I focused on in Considered Judgment seems to consist largely of beliefs or other intentional states. The community’s evolving understanding must be, in a suitable sense, supported by what its members were already committed to. This may insure that they are not making up their accounts out of whole cloth, but it seems to provide no strong reason to think their conclusions are correct. Perhaps then we should follow philosophers like John Greco (2012), Jonathan Kvanvig (2003) and Stephen Grimm (2009) and construe understanding, like knowledge, as factive. Just as one cannot know that Ott played for the Giants unless it is true that Ott played for the Giants, one cannot understand photosynthesis unless at least most of the beliefs that comprise one’s network of beliefs about photosynthesis are true. Understanding, like knowledge, then reflects the facts. If this is correct, it is easy to see why understanding is an epistemic accomplishment. An understanding of a topic consists of a constellation of mutually supportive truths about that topic. Since the mutual support would confer justification, this is very close to what a coherence theory of knowledge maintains (with the possible exception that in the case knowledge, the support should involve no false beliefs essentially, while understanding might tolerate a few). On a factivist account, the reason we should take a theory to afford an understanding of the rise of the working class, or the role of chlorophyll in photosynthesis, or whatever, is that it consists of

136

Catherine Z. Elgin

a sufficiently large cluster of mutually supportive truths about the topic. Understanding then would be knowledge plus coherence. The Challenge Posed by Science This is an attractive picture. But it has a fatal flaw: it does not accommodate natural science (Elgin 2004). Natural science standardly and unabashedly deploys models and idealizations that are not, and are not supposed to be, true. The ideal gas law describes a gas that consists of dimensionless spherical molecules that are not subject to friction and exhibit no mutual attraction. The Hardy-Weinberg formula describes the distribution of genes in an infinite population whose members mate randomly. Not only are there in fact no such things, according to our best theories, there could be no such things. But the ideal gas law figures in the understanding provided by statistical mechanics and the Hardy-Weinberg formula figures in the understanding provided by population genetics. Such modeling is not treated as a temporary expedient; science considers it a good way to encapsulate, convey and advance its understanding. Although scientists expect particular models to be superseded as science progresses, they neither expect to, nor think they ought to, abandon the practice of modeling. To deny that modern science affords an understanding of nature would be mad. So epistemology should explain how scientific models and idealizations figure in and contribute to understanding. Doing so, I maintain, requires acknowledging that understanding is non-factive. By this I do not mean that understanding, in science or elsewhere, is wholly indifferent to facts, but that its relation to facts is more complex and circuitous than factivist epistemological theories maintain. The position I sketched earlier (and developed in Considered Judgment) does not privilege the literal, the descriptive, or the true. Rather, it maintains that different sentential elements of a theory or system of thought perform different functions. Some are ostensible truth-tellers. They are defective if they are not true. And their not being true undermines the contention that the system they belong to affords an understanding of the phenomena. Others are what I call felicitous falsehoods (Elgin 2004). Although they are not true, they are not defective on that account; nor is a theory or system of thought defective for containing such elements. So long as (1) the theory or system as a whole is suitably connected to the phenomena, (2) the requisite epistemic standards

Understanding’s Tethers

137

are met, and (3) the felicitous falsehoods perform their epistemic functions, the system affords an understanding of the domain. Not all models are propositional. Besides equations and verbal models, science is rife with diagrams, such as the harmonic oscillator depictions in physics texts, and three dimensional models, such as tinker toy models of proteins. Non-propositional models are not, strictly speaking, false. But if interpreted as realistic representations of their targets, they are inaccurate in much the way that false descriptions of an object are inaccurate. All represent their targets to be as they are not. For ease of exposition, I label all such models falsehoods; if despite (or even because of) their inaccuracy, they afford epistemic access to their objects, they are felicitous falsehoods. Actual populations are not infinite; actual gas molecules are not spherical. If a model is a felicitous falsehood, its inaccuracy stems from its target’s failure to completely mirror the model. Its felicity consists in its affording genuine insight into the target. So felicitous falsehoods are false in respects that do not interfere with, and may even facilitate, their affording such insight. How can they do this? Suppose Jill wants to know whether Fred is in town. Fiona says, ‘He is still in China’. As it happens, Fred is in Tibet. Although false, Fiona’s statement is true enough in the conversational context, since if Fred is anywhere in Asia, he is not in Cambridge this afternoon. Given the interests of the interlocutors, it is a mildly felicitous falsehood. This is a trivial case, for the falsehood in question is no better than the truth. Fiona could have said, ‘He is out of town’ or ‘He is still in Asia’, or just ‘No’, thereby serving the immediate purpose of the exchange without running any danger of fostering false belief. In other cases, however, felicitous falsehoods are more useful. They serve legitimate epistemic purposes more effectively than readily available truths. Thought experiments are cases of this kind. According to the gas laws, pressure, temperature, and volume are interdependent. So, we are told, as the temperature in an enclosed container of gas increases, pressure increases to infinity. In reality, that would not happen; eventually, the container would burst. A thought experiment designed to manifest the interdependence prescinds from that material inconvenience; although this is not stated, it pretends that the walls of the container are infinitely strong. In effect, the thought experiment is a fiction. In entertaining it, we suspend disbelief and play out the consequences that figure in the fictional scenario. Going along with the fiction fosters understanding. The thought experi-

138

Catherine Z. Elgin

ment discloses something about the behavior of gases – namely, the interdependence of temperature, pressure and volume – which remains constant regardless of the strength of the container. This is clearly a good thing. But the worry is that this can go too far. If we start admitting fictions into the realm of science, where do we draw the line? To make the worry vivid, consider “The Elephant’s Child”, Kipling’s story about how the elephant got its trunk (Kipling 1978). Once, the story says, elephants had ordinary noses. But an insatiably curious elephant’s child wanted to know what crocodiles eat for breakfast. He leaned down to ask a crocodile. The crocodile, smacking his lips and thinking that young elephant would be a delicious answer to that very question, bit the elephant’s nose and tugged on it. As the elephant (with the help of a python) pulled back, its nose elongated until it took the shape of a trunk. That basically is the story. What is wrong with it? As a children’s story, perhaps nothing. But there is surely something wrong with it as a contribution to the understanding (even to a child’s understanding) of biology. Since I allow science to incorporate fictions and falsehoods, I am in no position to say, as others would, that what is wrong with the story is that it is false. So is the ideal gas law. The difference is that while the ideal gas law affords some understanding of the behavior of actual gases, Kipling’s story affords no understanding of the phylogenesis of elephant trunks. If the story, with its tacit commitment to the heritability of acquired traits, were true, it might yield some understanding of how the elephant got its trunk. But once we acknowledge that both it and its Larmarckian presuppositions are false, the story seems to have nothing to do with the way the elephant got its trunk. We need to appeal to natural selection for an explanation, and that explanation will be entirely unsympathetic to any tugging-on-noses hypothesis. The challenge then is to vindicate thought experiments and models, while excluding just-so stories. Exemplification To meet the challenge I appeal to Goodman’s notion of exemplification. I will argue that effective models and thought experiments exemplify features they share with their targets, and make reference to their targets via that exemplification; just-so stories and other infelicitous falsehoods do not.

Understanding’s Tethers

139

Exemplification is the relation of a sample or example to whatever it is a sample or example of (Goodman 1968). Soap manufacturers distribute free samples of laundry detergent to entice customers to buy; logic texts include examples to illustrate rules of implication; politicians appeal to focus groups to learn voter preferences; safety inspectors take air samples to assess air quality in mines. All such cases involve attending to something specific in order to find out something more general. To fix ideas, let us begin with a familiar case: a sample of laundry detergent. The first thing to notice is that it is in fact laundry detergent. Unlike other advertisements, which simply inform potential customers about a product, the sample is an instance of the product. It is, however, not a mere instance of the product, as a cup of detergent borrowed from a neighbor might be. The sample is a telling instance. In Wittgenstein’s terms, it shows rather than says something about the product (Wittgenstein 1947). It serves to point up and make manifest some of the detergent’s properties, such as its capacity to remove stains. In so doing, it refers to those properties.2 An exemplar then is a symbol that refers to some properties it instantiates. But it does not refer to all of its properties. Exemplification is selective. An exemplar highlights some of its features by overshadowing or downplaying others. Even though the detergent sample was free, it would be a mistake to conclude that that brand of detergent is free. Price is not one of the properties that, under its standard interpretation, such a sample exemplifies. Nor does it exemplify properties like being manufactured in Cleveland or being inedible, although it also instantiates these properties. Because commercial samples typically belong to regimented symbol systems, their standard interpretation is straightforward. Consumers normally know what properties such symbols exemplify. But not all exemplars are so regimented. The results exemplified by a public opinion poll may be difficult to interpret. It might, for example, be unclear whether a poll taken outside a farmers’ market reveals dissatisfaction with food prices generally or dissatisfaction with the price of vegetables in particular. Moreover, many exemplars are sui generis; they belong to no even roughly regimented system. Any item comes to exemplify simply by being used as a sample or example. An unassuming rock in the woods comes to exemplify 2

As I use the term, a property is whatever the members of an extension have in common.

140

Catherine Z. Elgin

‘feldspar’ simply by being pointed out as an example of the mineral. A teacher can take a student’s paper to provide an example of the sort of paper he wants his students to write, or the sort wants them not to write. Depending on how he uses it, it might exemplify its content, its form, or even its format. Whichever of these it exemplifies, it does so with some measure of generality. Even if the paper exemplifies its content, the teacher is not recommending that other students submit papers with the very same content. Rather, he is recommending that they do ‘this sort of thing’, where context, it is hoped, supplies values for the missing parameters. These are characteristics that exemplars share with indexicals; words like ‘here’ and ‘there’, ‘now’ and ‘then’ are similarly dependent on context. To complicate matters further, the same item can exemplify different properties in different contexts. The detergent sample might exemplify ‘free’ in a symposium on effective advertising, even though it does not exemplify ‘free’ in the context of the customer’s deliberations about the desirability of switching brands. In yet other contexts, the packet of soap powder does not exemplify at all. If it remains in the warehouse because the advertising campaign is scuttled, although its contents still instantiate the capacity to remove stains, the packet of detergent does not exemplify that capacity or any other feature. Interpreting exemplars then requires ascertaining the dimensions along which an item exemplifies, and the grain with which it does so. In this regard, exemplars are no different from other context sensitive symbols. Exemplification involves a dual representational relationship. An exemplar refers to a property or cluster of properties it instantiates and thereby to the extension of that property or cluster. The properties in question might be properties that the exemplar alone instantiates. If there are such things as haecceities, an individual who exemplified her haecceity would exemplify a feature that nothing else could share. If, as a matter of contingent fact, Trixie is the only dog who walks on two legs, she is the only individual who, as a matter of contingent fact, can exemplify the property of being a canine biped. Such exemplars represent unit classes. Others function as representatives of wider classes. By exemplifying greenness, a patch on a color chart refers greenness and thereby to green things generally. Putting the matter this way may make the twofold reference look trivial. One green thing serves as a representative sample of green things generally, and only a green thing can do so. Sometimes exemplification is that

Understanding’s Tethers

141

trivial. But often it is not. For an exemplar can exemplify a cluster of properties. If the detergent sample simultaneously and without change of context exemplifies its brand, its fabric softening capacity and its capacity to remove grass stains, then it refers to the extension consisting of instances of that brand of detergent which can soften fabric and remove grass stains. If the sample is a representative sample, in relevant respects there is nothing special about it. In terms of cleaning power nothing differentiates the sample from any other equal size portions of that brand of detergent. With respect to exemplified properties then, what holds of the sample holds of that brand of detergent generally. In determining that there is no relevant difference between the detergent sample and the soap powder it represents, we rely on background assumptions about what differences there are, and which of them would make a sample unrepresentative. If the soap in the sample has the same chemical constitution as the rest of the detergent of that brand, we are apt to conclude that it is representative. Then if the sample removes grass stains, it is reasonable to expect that other similarly sized portions of the detergent do too. In making this inference we rely on the background belief that whatever has the same chemical constitution has the same cleaning power. The case of a focus group is slightly more complicated. A focus group is chosen on the basis of demographic features, such as age, race, gender, and income, which are supposed to be proxies for the factors investigators are interested in. With some trepidation, they take the focus group’s opinions to be representative, because collectively the group constitutes a demographically representative sample. If the demographic factors are well chosen, investigators have reason to believe that what holds of the focus group holds of the class they are interested in. There is then an epistemically well grounded basis for projecting from the exemplar to that class. They can of course be wrong. The conviction that an exemplar is representative of a particular extension is undermined if the relevant assumptions are unfounded. Some public opinion polls before the 2008 US presidential election were unrepresentative because the pollsters relied on telephone directories in deciding whom to survey. The driving idea was that to find out how people in a given region feel about the issues, you get the local phone book, call people and ask them. Home address was taken to be a reliable indicator of other demographically relevant factors. Phone numbers were thought to be keyed to home addresses. Such a polling procedure used to work. But now, unlike their elders, many young people have no

142

Catherine Z. Elgin

land lines and no directory listings; their mobile phone numbers bear no correlation to where they live. So the polls that relied on telephone directories were skewed toward older voters. The falsity of the belief that most potential voters had land lines and were listed in local telephone directories undermined the reliability of the polling practice. The opinions gleaned were not representative of the population of interest. To summarize: a well chosen exemplar can afford epistemic access, not only to some of its own properties, but also to a wider class of cases. Because exemplars are symbols, they require interpretation. Because their reference depends on context, interpretation is keyed to circumstances. Although an exemplar refers to the extension consisting of items that share the exemplified properties, that extension can be described in multiple ways. So an exemplar can be informative. But because a non-trivial identification of that extension relies on fallible background assumptions, although exemplars can afford fairly direct epistemic access to the properties they exemplify, the epistemic access they provide to wider classes of cases is sensitive to the adequacy of the background assumptions. If we focus only on commercial or pedagogical exemplars, we might conclude that the function of exemplification is primarily heuristic. Commercial samples and textbook examples are designed to display what is already known. But not all exemplars are like that. In the course of product development, the manufacturer might test a detergent sample to find out whether it is effective against grass stain. Until the test is run, no one knows whether the sample exemplifies ‘eliminates grass stain’ or ‘fails to eliminate grass stain’. Until the public opinion poll is carried out, no one knows whether it will exemplify concern or complacency about corruption. So exemplars have the capacity to advance as well as exhibit knowledge or understanding. Exemplification in Science In principle any exemplar can exemplify any property it instantiates and any property that is instantiated can be exemplified. But what is feasible in principle may not be straightforward in practice. Exemplification of a particular property is not always easy to achieve, for not every instance of a property affords an effective example of it. Considerable stage setting is often needed to bring a rare or recondite property to the fore.

Understanding’s Tethers

143

Exemplification is sometimes achieved simply by directing, or redirecting attention. A psychologist can bring test subjects to exemplify automaticity by redirecting attention from the contents of their responses to their reaction times. In effect, she just reframes the phenomena, reversing figure and ground. In other cases exemplification requires isolating aspects of phenomena. To exemplify the electrical conductivity of water, a scientist distills out impurities before running her experiment. Elaborate experiments may be required to exemplify subtle, difficult to discern properties. In yet other cases things are even more complicated. Some properties fuse in such a way that they cannot in fact be prized apart. Preferential mating, migration, natural selection and genetic drift affect the distribution of genes in a population. They are always present in nature and two of them – natural selection and genetic drift – are ineliminable, even in the lab. So biologists can neither discover in nature nor create in a lab a population that is not subject to these contingencies of fortune. They are, as we say, facts of life. But to measure genetic change, biologists need a baseline – a measure of how alleles would redistribute if no change other than redistribution was taking place. Here is where felicitous falsehoods reenter the picture. To obtain the baseline, biologists create a model – the HardyWeinberg model – which describes the distribution of genes in an infinite, isolated population of organisms that mate randomly and are not subject to natural selection. That the population is infinite counteracts the effects of genetic drift; that mating is random insures that neither physical proximity nor genes that give rise to attractive phenotypes have a preferential advantage; that neither selection nor migration takes place insures that no novelties are introduced into the gene pool. The distribution of genes in this imaginative scenario exemplifies the factor in genetic redistribution is not a manifestation of genetic change. The conditions described in the model do not – indeed could not – occur in nature. Nevertheless, the model highlights a factor in what does occur in nature, and enables scientists to discern it and play out its consequences in a way that they could not do if they were restricted to accurate factual representations. Models are felicitous falsehoods that exemplify features they share with their targets. They diverge from their targets in unexemplified features. That divergence enables them to make manifest features that are normally obscured. But because the features with respect to which they diverge are not the features with respect to which they symbolize, models refer to, hence are models of their targets.

144

Catherine Z. Elgin

As is the case with more mundane exemplars, the interpretation of models depends heavily on background assumptions that might be false. So we go out on a limb when we take the target to instantiate the properties its model exemplifies. But that limb is no shakier than the limbs we go out on when we generalize inductively from a limited body of evidence. Just as we are vulnerable to ill chosen or misleading evidence in inductive reasoning, we are vulnerable to poorly designed or misleading models. This is unfortunate. But we’ve long realized that Cartesian certainty is not in the offing. This discussion reveals something about the cognitive function of models but how does it relate to understanding? Because exemplification plays a central role in systems of thought in reflective equilibrium, the role of exemplification in my account blocks the charge that my epistemology is, or is a near relative of, a coherence theory. There is no worry that a tenable theory just floats above the phenomena, that its observational and predictive success could, for all we can tell, be due to brute luck or to preestablished harmony. Exemplification requires instantiation. So any exemplar (whether or not it is a felicitous falsehood) that exemplifies a property instantiates that property. This secures a referential connection to the world.3 Because the exemplar affords epistemic access to the property it exemplifies and a basis for believing that the target has that property, this referential connection is also an epistemological connection. Observational and predictive success are evidence that we have correctly identified the extension that the exemplar is representative of. Moreover, felicitous falsehoods can show not only that something obtains but also why it is significant. So they contribute to understanding by highlighting features that would otherwise be obscured, thereby enabling us to see subtle, but potentially significant patterns in the phenomena. By interpreting the behavior of actual gases as variants on the supposed behavior of an ideal gas, we may be able to understand aspects of actual gas be3

The issue is slightly more complicated than this suggests, for models can be layered. A theoretical model can take as its immediate referent a phenomenological model which in turn refers to a data model which in turn refers to the facts. Each of the models exemplifies a feature it shares with its target, but only at the end of the sequence is the the target comprised of real-world phenomena. See Morrison and Morgan 1999.

Understanding’s Tethers

145

havior that would be overshadowed by the complex interactions among differently sized, differently shaped, actual gas molecules. The epistemological difference between felicitous falsehoods and just-so stories consists in this: because felicitous falsehoods exemplify features they share with their targets, they afford epistemic access to features of those targets. These features are typically not simple monadic properties, but intricate structural and, in many cases, dynamic properties. So felicitous falsehoods enable us to apprehend how things are beneath the welter of complexities that we typically encounter. Just-so stories do not exemplify features they share with the phenomena they purportedly concern. If they were true, such stories would be informative. This, such a story would say, is in fact how the elephant’s trunk developed. But given that they are not true, they contribute nothing to the understanding of the phenomena. Felicitous falsehoods have strong tethers to the phenomena they concern; just-so stories lack such tethers. The conviction that understanding is factive rests on the (typically tacit) assumption that truth is the only secure link between theories and the world. The driving idea is that if the terms in a theory denote real things and the claims made by the theory accurately characterize those things and (enough of) the relations among them, the theory embodies an understanding of those things. I have argued that exemplification supplies another strong and secure tether. Because exemplification requires instantiation, an exemplar is guaranteed to instantiate the properties it exemplifies. The tie to real, existent properties is thereby assured. Because exemplification requires reference to the extension of those properties, the exemplar provides an avenue of access to other members of that extension. An exemplar, as we have seen, need not be truth evaluable. Neither the detergent sample nor the diagram of a harmonic oscillator has a truth value. And even if an exemplar, such as the ideal gas law or the Hardy-Weinberg formula, has a truth value, neither its accuracy nor its adequacy need be undermined by its falsity. So long as the properties it exemplifies are properties of the members of the extension it represents, an exemplar is accurate. So long as its serving as a representative of that extension promotes our epistemic ends, the exemplar is adequate. Questions of interpretation may arise about just which properties a given exemplar exemplifies and what (non-trivially characterized) extension it represents. But these are not obviously more troublesome than the questions of interpretation that arise for any statement of scientific fact. Still, exemplification is no epistemological

146

Catherine Z. Elgin

panacea. The usual problems surrounding induction, misleading evidence, and so forth still arise. Nevertheless, exemplification enables us to vindicate felicitous falsehoods, and thus to explain how scientific accounts that are not and do not purport to be true figure in a genuine understanding of the way the world is.4 References Elgin, C., 2004: True Enough, Philosophical Issues 14, 113-131 Elgin, C., 1996: Considered Judgment, Princeton: Princeton University Press Goodman, N., 1968: Languages of Art, Indianapolis: Hackett Greco, J., 2012: Intellectual Virtues and Their Place in Philosophy, this volume, 115128 Grimm, S., 2009: Reliability and the Sense of Understanding, in: Scientific Understanding, ed. H. de Regt, S. Leonelli, and K. Eigner, Pittsburgh: University of Pittsburgh Press, 83-99 Kipling, R., 1978: The Elephant’s Child, in: Just So Stories, New York: Weathervane Books, 49-63 Kvanvig, J., 2003: The Value of Knowledge and the Pursuit of Understanding, New York: Cambridge University Press Lehrer, K., 1974: Knowledge, Oxford: Clarendon Press Morrison, M. and M. Morgan, 1999: Models as Mediating Instruments, in: Models as Mediators, ed. M. Morrison and M. Morgan, Cambridge: Cambridge University Press, 10-37 Van Fraassen, B., 2002: The Empirical Stance, New Haven: Yale University Press Wittgenstein, L., 1947: Tractatus Logico-Philosophicus, London: Kegan Paul, Trench, Trubner & Co

4

I am grateful to Jonathan Adler and Brad Armendt for discussions of an earlier draft of this paper.

Knowledge as Achievement and the Value Problem BRUNO NIEDERBACHER, University of Innsbruck

In his book, Achieving Knowledge. A Virtue-Theoretic-Account of Epistemic Normativity, John Greco offers a version of virtue epistemology that claims to answer several questions concerning the value of knowledge: why is knowledge valuable, why is knowledge more valuable than true belief, why is knowledge more valuable than any subset of its constituents, and finally, why is knowledge more valuable than the value of all its parts taken separately? (Greco 2010, 98) The key to his answer to all these questions is the thesis that knowledge is a kind of success that arises from cognitive abilities or cognitive virtues of the agent. In this paper I am going to raise one problem for Greco’s thesis. 1. Greco’s argument in a nutshell Greco writes: “[…] knowledge is a kind of success from ability, and in general success from ability is both intrinsically valuable and constitutive of human flourishing, which is also intrinsically valuable. Moreover, both success from ability and human flourishing have ‘final’ value, or value as ends in themselves, independently of any instrumental value that they might also have. Therefore knowledge has value over and above the practical value of true belief.” (Greco 2010, 99)

Greco’s answer can be put in the following argument: (1) (2)

Knowledge is a kind of success from ability. Success from ability is a complex state of affairs that has both intrinsic and final value and is furthermore constitutive of human flourishing which has intrinsic and final value.

Therefore

148

(3) (4)

Bruno Niederbacher

Knowledge has both intrinsic and final value and is furthermore constitutive of human flourishing which has intrinsic and final value. Other kinds of true belief that fall short of knowledge, although practically valuable, lack these kinds of value.

Therefore (5)

Knowledge has more value than other kinds of true belief that fall short of knowledge.

My objection concerns premise (1). The question is: what does “success from ability” amount to in the case of knowledge? 2. What “success from ability” amounts to in the case of knowledge According to Greco, “success from ability” means in the cognitive domain that one gets a true belief (success) through the exercise of one’s cognitive abilities: one’s intellectual powers or intellectual virtues. According to Greco (2010, 10), abilities are reliable dispositions of the cognitive agent. In current epistemology there are two ways to understand cognitive virtues. Some epistemologists, sometimes called virtue-responsibilists, understand these intellectual abilities as character traits, akin to Aristotles’s moral virtues. For them, intellectual virtues are, for example, openmindedness, intellectual carefulness, and intellectual courage (see: Zagzebski 1996). Greco, however, belongs rather to those epistemologists who can be named virtuereliabilists. For them, cognitive virtues are such cognitive powers as perception, memory, introspection, and inferential reasoning (Greco 2003, 116; Greco 2010, 72). Although Greco thinks that the virtues of the virtueresponsibilists should be contained in a complete epistemology, he does not think that they are needed to define knowledge (Greco 2010, 10). Nevertheless, and not entirely consistent with this claim, Greco uses the Aristotelian account of moral virtue when he defines the knowledge-relevant status of a belief (Greco 2010, 43-44). Similar to Linda Zagzebski’s interpretation of Aristotle, Greco says that moral virtues involve both a motivational component and a reliable success component (Greco 2010, 43). And he continues: “Now it seems to me that the Aristotelian model is the better one for theories of epistemic normativity. This is because, it seems to me, knowledge requires both responsibility in one’s cognitive conduct and re-

Knowledge as Achievement and the Value Problem

149

liability in achieving epistemic ends” (Greco 2010, 43). He defines the responsibility in question in terms of proper motivation for believing a proposition, whereby proper motivation is understood as being motivated to believe the truth.1 And believing reliably a proposition is understood in terms of the belief’s resulting from intellectual dispositions that reliably produce true belief. Putting the two together, Greco defines epistemically virtuous belief as follows: “EV. S’s belief that p is epistemically virtuous if and only if both (a) S’s belief that p is epistemically responsible; and (b) S is objectively reliable in believing that p.” (Greco 2010, 43) Central to Greco’s account of knowledge is that if a belief is knowledge then this belief is not only true and virtuously formed, but the belief is true because it is virtuously formed (Greco 2010, 44; 99-101). For it is possible that one exercises one’s intellectual virtues and that one acquires a true belief, but that the acquired true belief lacks the proper connection to the virtuous exercise. To be successful because of the exercise of one’s ability is an achievement. And achievements in general have intrinsic value, final value, and are constitutive of human flourishing. Thus, if knowledge is an achievement, then knowledge has intrinsic value, final value, and is constitutive of human flourishing. This is Greco’s proposal to solve the value problem. One can ask whether it is true that knowledge is an achievement, and whether it is true that achievements generally do have these kinds of value. Duncan Pritchard (2010) asks these questions, and there is no need to repeat his critical remarks. Instead, I would like to raise another question that concerns the meaningfulness of the distinction between “a belief’s being true and virtuously formed” and “a belief’s being true because virtuously formed” (Greco 2010, 100). This distinction is crucial for Greco: he employs it not only in his solution to several kinds of Gettier-cases, but also in his solution to the value-problem. For what constitutes the surplus value of knowledge over and above other forms of true belief that fall short of knowledge is, according to Greco, the proper connection between the actu1

Greco 2010, 43. Later, on p. 167, Greco calls the condition of responsibility “a kind of subjective justification: S is subjectively justified in believing that p just in case S is epistemically responsible in believing that p.”

150

Bruno Niederbacher

alization of a responsible and reliable disposition and the resulting true belief. A true belief that falls short of knowledge is either a true belief that is formed without the actualization of the cognitive dispositions in question or a true belief that is not properly connected with the actualization of the cognitive dispositions in question. Greco (2010, 44) writes: “It is a good thing that one’s belief is epistemically responsible, and it is a good thing that one is objectively reliable in forming one’s belief. But it is even better if responsibility and reliability bring success – if one’s belief is true because it is responsibly and reliably formed.” Thus, whether Greco succeeds in solving the value-problem depends on whether the phrase “one’s belief is true because it is responsibly and reliably formed” makes sense. 3. Whether the phrase “S’s belief is true because it is virtuously formed” is meaningful According to Greco, his account is able to explain the paradigm Gettier cases, for example Lehrer’s Nogot Case and Chisholm’s Sheep Case. “In Gettier cases, S believes the truth, and S believes from an ability, but S does not believe the truth because S believes from an ability.” (Greco 2010, 74) In Chisholm’s Sheep Case, a “man takes there to be a sheep in the field and does so under conditions which are such that, when a man does thus take there to be a sheep in the field, then it is evident to him that there is a sheep in the field. The man, however, has mistaken a dog for a sheep and so what he sees is not a sheep at all. Nevertheless, unsuspected by the man, there is a sheep in another part of the field.” (Chisholm 1977, 105) The man believes the truth, and he believes from an ability, since he forms his belief that there is a sheep in the field with his usual cognitive abilities of perception, etc. But, as Greco would say, he does not believe the truth because S believes from an ability. The “because” marks a causal explanation. It is not just the belief, but the belief’s being true that is supposed to be explained by the exercise of cognitive virtues.2 Greco says that some Gettier 2

The same solution proposes Zagzebski 1996, 296-297 in Zagzebski 2009, 123: “Knowledge is belief in which the believer gets to the truth because of her good epistemic behavior”. Similarly, Sosa 2003, 175: “What we prefer is the deed of true believing, where not only the believing but also its truth is attributable to the agent as his or her own doing.”

Knowledge as Achievement and the Value Problem

151

cases “involve something akin to a deviant causal chain” (Greco 2010, 75; 76). We know the problem of deviant causal chains from action theory. The standard example is the following: Tim wants to kill his uncle, and he believes that he can find him at home. His decision to kill his uncle so agitates him that he drives recklessly. He hits and kills a pedestrian, who by chance is his uncle. Another example, often used in epistemological contexts, is the case of the archer (Sosa 2003, 163-164) who shoots skillfully and hits the bull’s eye. However, the hitting of the bull’s eye is not due to his skillful shot but rather to some gusts of wind. Thus, the agent has the intention to achieve something, he has an ability to do so, he exercises this ability, and he achieves what he intends to achieve. However, the success does not occur because of these features of his act. My first question is this: is it really something akin to wayward causation that happens in the paradigm Gettier cases? Let us look at Chisholm’s Sheep Case again. What goes wrong in this case is that one believes the true proposition (p) because of its seeming to one (or one’s believing) the false proposition that (q): (p) (q)

There is a sheep in the field. This (thing in front of me) is a sheep.

I do not see any wayward causation involved in the belief-production. There is just a false proposition involved. Greco should slice the beliefforming process into smaller pieces. Then he could say, in his terminology, that the belief that q was from “ability but without success.” My second question is this: Do formulations like “S’s belief is true because it is virtuously formed” or “S believes the truth because S’s belief is epistemically virtuous” (Greco 2010, 44) make sense? The idea is that the causal chain of the particular belief-production leads in a non-interrupted or non-bypassed way from the motivation to believe the truth via actualizations of intellectual virtues straight into “the truth” like into the bull’s eye. But is this idea consistent when spelled out properly? I do not think so. In epistemology “getting the truth” has to be spelled out in the following way: “forming a belief with a content p, and p is true”. I think that the particular causal chain leads only to the particular mental state of believing that p. The belief’s being true cannot be connected with the particular causal chain. What we explain by the actualization of intellectual virtues is only

152

Bruno Niederbacher

why somebody came to believe that p (which happens to be true) rather than to believe that not-p (which happens to be false). Why the belief that p is true cannot be explained by the actualization of cognitive ability. Thus, I doubt that the formulation “the belief is true because it is from ability” makes sense. Greco (2010, 71) makes a start to define knowledge by writing: “S knows that p if and only if S believes the truth (with respect to p) because S’s belief that p is produced by intellectual ability.” And he assumes that the complex state of affairs of something’s being a success from ability can be specified in the cognitive domain as something’s being a true belief that results from an exercise of intellectual virtue. Moreover, Greco thinks that this state of affairs has intrinsic value. Intrinsic value is the value a state of affairs has in virtue of its intrinsic properties. However, I think that a belief’s being true is not an intrinsic property of a belief. That p is true depends also on the relation to an external existing state of affairs. If so, however, one would need to change the definition of knowledge into: “S knows that p if and only if S’s belief that p is produced by intellectual ability, and p is true.” This definition would fall prey to Gettier cases. But could it nevertheless be of some help for dealing with the value problem? 4. Conclusion The requirement that S’s belief that p is produced by the exercise of an intellectual virtue is similar to the requirement that a belief needs to be based on adequate grounds. And similarly to Greco one could say: It is a good thing that one possesses an adequate ground for a belief, but it is even better, if one’s belief is based on this adequate ground. The basing relation could be understood in different ways. First, it could be understood as a causal relation, where the adequate ground causes the belief in question. Second, it could be understood as a sort of evidential relation, where the cognizer “sees” the connection between the ground and the proposition believed. Greco, being an agent reliabilist and externalist (Greco 2010, 7), would endorse the first way. Linda Zagzebski has recently endorsed the second way, a way that would be open only for internalists.3 3

See: Zagzebski (2009, 114): “The fact that she bases her belief on the evidence is good, not only because doing so leads to the truth in general, nor because it has led to

Knowledge as Achievement and the Value Problem

153

One question is how one understands the “seeing” of the connection between grounds and proposition. Probably it would amount to S’s believing that these grounds make it probable that the proposition is true. But our main question is: Why would it be better that the belief be based on the adequate grounds, whether one understands the basing-relation causally or evidentially? The most natural answer would be: Because that leads more reliable to true belief. However, if this were the answer, we would still be entangled in the value problem as Zagzebski posed it. References Chisholm, R., 1977: Theory of Knowledge. Englewood Cliffs, NJ: Prentice Hall Greco, J., 2003: Knowledge as Credit for True Belief, in: Intellectual Virtue. Perspectives from Ethics and Epistemology, ed. M. DePaul and L. Zagzebski, Oxford: Oxford University Press, 111-134 Greco, J., 2010: Achieving Knowledge. A Virtue-Theoretic Account of Epistemic Normativity. Cambridge: Cambridge University Press Pritchard, D., A. Millar, and A. Haddock, 2010: The Nature and Value of Knowledge. Three Investigations. Oxford: Oxford University Press Sosa, E., 2003: The Place of Truth in Epistemology, in: Intellectual Virtue. Perspectives from Ethics and Epistemology, ed. M. DePaul and L. Zagzebski, Oxford: Oxford University Press, 155-179 Zagzebski, L., 1996: Virtues of the Mind. An Inquiry into the Nature of Virtue and the Ethical Foundations of Knowledge, Cambridge: Cambridge University Press Zagzebski, L., 2000: From Reliabilism to Virtue Epistemology, in: Knowledge, Belief, and Character. Readings in Virture Epistomology, ed. G. Axtell, Lanham: Rowman and Littlefield, 113-122 Zagzebski, L., 2009: On Epistemology. Belmont: Wadsworth

the truth on this occasion, but because on this occasion she has seen the connection between the evidence and the true proposition she believes, and has thereby acquired a level of epistemic status she would not have had otherwise.”

Apriority and Virtue: How Successful a Relationship? NENAD MIŠČEVIČ, University of Maribor

Abstract The ability-based virtue epistemology championed by authors like Greco and Sosa is increasingly popular in epistemology. The paper argues that such virtue epistemology goes with a degree of aposteriority of “armchair beliefs”. Cognitive integration, oof first or of second order is crucial for turning abilities into virtues. But in the case of armchair intuition, this demand for integration stresses the connections between it and the nonintuitional, presumably empirical rest. The integration of any candidate a priori source or piece of knowledge with some heterogeneous source or piece brings in aposteriority. So, ability based epistemic virtue goes with aposteriority and virtue epistemology should admit aposteriority into the armchair in order to obtain a well-integrated armchair. 1. Introduction The ability-based virtue epistemology championed by authors like Greco (2010), and Sosa (2007) is increasingly popular in epistemology.1 How does such a view epistemology, treat intuitional or armchair knowledge, “armchair” in the descriptive sense of being relatively independent from particular experiences?2 Does it confirm its a priori status, or does it of necessity bring in a posteriori elements? Being myself very much in favor of virtue epistemology (to be shortened to VE in the sequel), I shall argue for the later, and in particular for the following claim: the full justification of the deliverances of intuition is a highly structured one with some a priori elements, but with an important a posteriori component. 1

Big thanks go to John Greco, Kati Farkas, Istvan Bodnar and Gabor Betegh. I shall use the terms “armchair” and “intuitional” as synonyms, in a neutral fashion, without prejudging its nature and epistemic status. 2

156

Nenad Miščevič

Here is the preview: Our question concerns intuition as a cognitive capacity, grounded in competence(s). whose proper functioning makes it an epistemic virtue (where epistemic virtue is taken in the sense of ability rather than as an epistemico-moral character-trait.). What is the normative status of its deliverances, on virtue-theoretic premises? In the next section I discuss Greco’s view; he stresses the reliability of cognitive competence or ability, but also the fact that reliability is not sufficient, and proposes cognitive integration as the specific difference of the virtue-based approach. I then argue that integration of a priori items with the relevant heterogeneous ones will result in the change of their normative status: they become partly a posteriori. In Section Three I very briefly address the alternative VE proposal of integration on reflective level due to E. Sosa (with a promise to address it in more detail elsewhere). The conclusion recapitulates the discussion and offers the idea of structured justification with both a priori and a posteriori components. In order to start with a wider map of VE, let me sketch in the remainder of the section my own preferred view of virtues (for more, see Miščevič 2007). I would divide epistemic virtues into the narrowly cognitive ones on the one side, and the moral-like ones on the other. The narrowly cognitive ones can be divided into motivational and executive ones. The motivational one is good curiosity (I leave the characterization of goodness open), that sets the goal of inquiry: getting reliably at truth and avoiding error, and possibly achieving understanding as a particularly valuable kind of knowledge. The executive ones are abilities and capacities, perception, intuition, reason, memory and the like. I agree with Greco and Sosa that the executive ones can and perhaps should figure in the definition of knowledge. The moral-like cognitive virtues, like, for instance, openness, intellectual modesty or humility, critical mind and intellectual courage, get part of their value from their cognitive achievements, and part from indicating a general moral quality: intellectual courage is a sign of courage in general, and so on for modesty and others. They can play both motivating and executive roles (and are being put forward by authors like Zagzebski 1996).3 The motivating cognitive virtue(s) as well as the moral-like ones have little to do with a priori/a posteriori contrast. The relevant ones are capaci3

Thanks to I. Bodnar for reminding me of this possibility.

Apriority and Virtue: How Successful a Relationship?

157

ties and abilities, championed by Greco and Sosa. However, we value cognitive capacities for their reliability. So, capacity-VE is very close to (mere?) reliabilism, and threatens to collapse into it. It turns out that the only plausible difference would be linked to cognitive integration: we require that the capacity in question, in order to be virtuous, be integrated with others. This can happen already at the low level of spontaneous cognition – as suggested by Greco (2010) or on the reflective level – as Sosa would have it (2009, 2011). But now, suppose there is a capacity that is relatively experience-independent, say intuition. In order to be virtuous and produce knowledge (knowledge simpliciter, or sophisticated, reflective knowledge) it has to be integrated with its very different, experiencedependent cousins. But, such an integration brings in aposteriority. More cautiously, virtue-theoretic justification of armhcair (intuitional) beliefs is a structured one, containing important a posteriori elements. The pure intuition-derived justification might be (to some extent) a priori, but the full justification will involve heterogeneous elements, and bring in aposteriority. This is the main argument I want to propose. In order to develop it in more detail, I will concentrate on John Greco’s (2010) proposal to understand knowledge as success from ability. 2. Integrating our cognitive capacities: from Norman to Nash Why do we need cognitive integration? Well, a promising line for a virtue theorist is to build the notion of intellectual ability into the very definition of knowledge: a person’s believing truth is explained by her exercise of her intellectual ability.4 This gives one a fine diagnosis of Gettier cases, Smith’s believing out of ability is not connected with the truth of his beliefs in a productive and explanatory way. However, reliability is not enough for virtue. Greco first discusses “fleeting” cognitive processes (in Greco 2010, ch. 9), like for instance, Serendipitous Brain Lesion, “one efect of which is to reliably cause the true belief that one has a brain lesion.”(2010, 149). Such a capacity is not a virtue since it is not integrated with the rest. He next focuses upon the clairvoyance counterexample to reliabilism, that bring into considerations strange (and fleeting) processes 4

Greco: “S knows that p if and only if s believes the truth (with respect to p) because s’s belief that p is produced by intellectual ability.” (Greco 2010, 74)

158

Nenad Miščevič

of cognizing. Famously, Norman the Clairvoyant, introduced three decades ago by L. Bonjour (Bonjour 1980), has reliably true beliefs, but they do not always amount to knowledge Facing the Clairvoyant squarely, Greco offers a solution: He suggests to “divide clairvoyance cases among those that involve good cognitive integration and those that do not. The present account of knowledge can then handle the latter in the same way that it handles the brain lesion case. it remains to be seen, however, how the account can handle the former. (Greco 2010,154). The key to knowledge is cognitive integration: “… to evaluate how the present account rules on clairvoyance cases, we have to make sure that the cases are not underdescribed regarding the issue of cognitive integration. If the power of clairvoyance is not well integrated with the remainder of S’s cognition, then it will not count as cognitive character and will therefore not count as one of S’s abilities.” (…) one aspect of cognitive integration concerns the range of outputs – if the clairvoyant beliefs are few and far between, and if they have little relation to other beliefs in the system, then the power of clairvoyance is less well integrated on that account. Another aspect of integration that we noted is sensitivity to counterevidence. if clairvoyant beliefs are insensitive to reasons that count against them, then this too speaks against cognitive integration.” (Greco 2010, 153)

How does this apply to armchair beliefs, say elementary mathematical ones, to inferential moves we spontaneously make, and to elementary logical intuitions about the correctness of such moves? Let us agree that logical and elementary mathematical intuitions are reliable (and other intuitions hopefully as well). If reliability were enough, this would warrant apriority. However, not all intuitions. Some people are eccentric. The famous mathematician John Nash once explained that he is receiving messages from extraterrestrials which come to him with the same degree of persuasiveness (and he probably meant obviousness and immediate compellingness) as his mathematical theorems, and this sentence on math and extraterrestials is printed on the cover of the “A Beautiful mind” (Nasar 1998), his bestselling biography. let me call the problem The Beautiful Mind Problem. The problem introduces the need for going beyond mere reliability and success in justification: why do I trust my use of Conjunction elimination, if geniuses like Nash found their logical and mathematical reasoning as persuasive as messages from extraterrestrials? If a Nash could fall in for extraterrestrials, then obviousness and compellingness are not enough. (It is a more moderate problem than the related extreme Cartesian problem of

Apriority and Virtue: How Successful a Relationship?

159

madness.) The Beautiful mind problem is, as far as extraterrestrials are in question, parallel to the Norman problem, though the parallel is not perfect. His beliefs about extraterrestrials are “few and far between”, and have little relation to other beliefs in the system, Moreover, when it comes to extraterrestrials Nash shows little sensitivity to counterevidence. So, even if the beliefs had been, God forbid, true, they would not constitute knowledge. But now, if he experiences his mathematical beliefs as being produced in manner similar to the extraterrestrial ones, we should ask about their integratedness. Not all obvious and compelling beliefs & procedures are justified. And indeed, they are much better integrated: his mathematical discoveries have been vastly successful, his notion of equilibrium in games has contributed to explanations of actual behavior of people, and of course, mathematics as a whole is much better integrated with our other practices. Kati Farkas has in the discussion raised an interesting objection: mathematical capacities are sufficiently diverse to warrant confidence once they are well integrated between themselves. Some people are better at numbers, others at equations involving variables, still others in geometry, and some do math as applied logic. Ingenious as it is, the objection is not true to the spirit of Greco’s VE; it demands integration across the board, so that differences within the mathematical family, interesting as they might be, simply do not offer a sufficiently wide area to be integrated. The integration proposal reminds one of the classics. Kant famously claimed that “objective reality” of mathematical judgments comes only from reference to the objects of senses: “Pure mathematics, and especially pure geometry, can have objective reality only under the single condition that it refers merely to objects of the senses.” (Kant 2004, 38).

This might have beenn his reaction to the integration problem. Of course, his Copernican solution is somewhat ironic, namely that the objects of senses are phenomena, dependent on our forms of intuition. “…with regard to which objects, however, the principle remains fixed, that our sensory representation is by no means a representation of things in themselves , but only of the way in which they appear to us.” (Ibid.)

But this need not detain us. We need a different solution. Well, as Kant himself has realized, and many contemporary authors were stressing, mathematical beliefs have wide range of application, and logical beliefs are compelling (Horwich 2005, Boghossian 2008) and necessary for any cog-

160

Nenad Miščevič

nitive project, and they don’t mislead us in our actions (Wright, 2001).. Let me develop this point for mathematical and logical beliefs (and procedures). Take logic: we need a well-integrated armchair So, how is simple and naïve logical reasoning justified? (e.g. where does the entitlement of a naïve thinker who spontaneously and unthinkingly uses Modus Ponens or Conjunction Elimination come from?) Maybe from obviousness and compellingness. However, here is a worrying analogy with sensory modalities. When my spectacles are not around, I might be unsure about a shape of a thing in front of me, and I can help myself to feeling the object, in order to make sure that the visual information I am receiving is reliable. This is a very practical method for a practical context, but of little use in theoretical debate about reliability of sense-perception. One component of perceptual ability is tested by its conformity with another, within the same perceptual domain, and the circle is too close for more ambitious epistemic justification. The same holds for justifying one area of logic (say, proof-theory) with element from another (say, model theory). The literature on coherentism has familiarized us with a recipe that could solve the problem: if a circle of justification is too narrow, try to widen it. Why is wide circularity better, wonder some of my colleagues. Well, for one, a wider data-base answers more objections, as my former student Zsolt Novak put it in a discussion. Of course, if you are a foundationalist, no amount of coherence will by itself justify a belief for you; but there is a widespread agreement in the literature that a wide circle of coherent beliefs is in a good shape to be justified. Now, how could the circle under discussion be widened? An obvious proposal is to widen the areas involved. First, one should look at relation to other beliefs in the system; ie. the non-intuitional, empirical ones. Let me paraphrase the quotation from Greco, replacing his target with our present one: If the power of logical or mathematical insight is not well integrated with the remainder of S’s cognition, then it will not count as cognitive character and will therefore not count as one of S’s abilities. One aspect of cognitive integration concerns the range of outputs – if one’s the logical and mathematical beliefs are few and far between, and if they have little relation to other beliefs in the system, then the power of intuition is less well integrated on that account. Another aspect of integration that we noted is sensitivity to counterevidence. If the relevant beliefs are insensitive to reasons that count against them, then this too speaks against cognitive integration.

Apriority and Virtue: How Successful a Relationship?

161

Integration of course, involves capacities distinct from the original capacity that has produced the belief-candidate for being justified. It has to go beyond justifiers that are of the same-kind (“homogenous”), in order to enlarge the basis of justification. Consider now how the two aspects of integration interact with apriority. Take the argument from success: the reasoning of this kind has proved to be massively successful, so why could not a reflective thinker appeal to this past success as one reason for her trust? This widens the basis for legitimating the belief, bringing in some a posteriori elements. Our philosopher is not thereby becoming a fanatical Quinean. She can, in defiance to reliability, retain the obviousness and compellingness as her immediate reason. But if questioned further, about credential of such immediate obviousness, she might reply that it has not deceived her ever, and has been empirically successful. So, these more holistic, success-involving considerations might be taken in the usual way to indicate objective reliability of one’s methods. This is the way realism functions in science and in commonsense. A good, probably the best, explanation of success is truth(-likeness) and reliability. Logic is, of course used in the process of integration, and is at the same time justified by its role in it; it integrates itself by helping the thinker to integrate the whole of one’s beliefs. But the circle is virtuous, not vicious, the alleged circularity being the usual circularity of holistic, coherentist justification. Notice the symmetrical or opposite case: Perceptual beliefs get integrated with the rest with the help of massive use of deductive logic and (often, but not always) inductive schemas which are to a large extent a priori. So, the empirical gets mixed with the a priori, the same way in which the armchair gets integrated with the empirical. What about sensitivity to counterevidence from a different domain? Remember that, according to Greco, clairvoyant beliefs have to be sensitive to deliverances of other sources (presumably perception and testimony). In parallel, armchair beliefs should be sensitive to counterevidence from a different domain. But this involves refutability by non-armchair evidence, and works against the assumption that these beliefs are immune to empirical refutation, thus depriving them of one of the hallmarks of apriority. At the conference in Kirchberg Greco’s immediate reply was that he just demands de facto coherence of beliefs and integration of abilities, in particular their causal connectedness. This is in line with his demanding “a high degree of cooperative interaction with other aspects of the cognitive

162

Nenad Miščevič

system” (Greco 2010, 154), not reflection about their integratedness. (See also his chapter Eight on Pyrrhonian problematic). So, But why would such de facto coherence and causal connectedness induce aposteriority? Well, here is a dilemma for this line of reply. On the one hand, the philosopher offering it might argue that the integration account is merely causal. (He might add that it bypasses the a priori/a posteriori contrast, since it does not propose inferential justification as the central type of justification. He might appeal to quotations like the following one: “: … the reliabilist can deny(…) that all knowledge must be grounded in good reasons. Some reliable processes involve grounding in good reasons. But not all reliable processes do. Perception, for example, might be highly reliable, but involve nothing by way of inference from good reasons. introspection is plausibly like that as well, as is logical intuition and memory.” (Greco 2010, 125)

On the other hand, he might accept that the integration account is justificational. Good causal connections contribute to the normative status of the belief, in particular its status as an item of knowledge. Consider the first horn first. If the relation is merely causal, than it does not add to the normative account. The link to other capacities is inert in regard to justification (or warrant, or whatever normative property that is specific to knowledge). Then, it is irrelevant for the normative status of the belief, and the proposal collapses back into reliabilism: the only normatively property of the belief its reliable origin. The addition of the remark against inferential justification is irrelevant. Suppose that the integrative connection of intuitional belief to the empirical capacities, like perception, is non-inferential, but still justificatory. Then, the justification of intuition is partly empirical, and the addition does no useful work. Similarly, if we take the worst option, and claim that logic is groundless, that it does not operate on reliable grounds at all” things do not change. This would only make integration with non logical abilities even more needed, and only non-logical, presumably partly empirical abilities will do the job. The second horn plays in our hands: if a capacity (and the belief) is justified because it is connected with more empirical capacities (or belief coheres with empirical beliefs), than a part of its justification is not independent from empirical sources, but positively dependent on them for their normative status. then a priori intuition capacity(and related ability etc.) Gets a part of its full justification from other, empirically oriented abilities (prominently perceptual ones).

Apriority and Virtue: How Successful a Relationship?

163

This seems to be part of the notion of integration. No matter how we descriptively characterize the integration-ties (possibly as non-inferential), they contribute to justification, and some of their crucial components are experiential. But then, the crucial role of integration supports the presence of an a posteriori component. 3. Holism of reflective justification My next worry, which will take us to the competing VE account, is that I am not sure how unreflective a thorough integration can be. Remember that Greco takes sensi-tivity to counterevidence as an important aspect of integration. Doesn’t this sensitivity to reasons that count against one’s beliefs induce a certain amount of reflectivity? In principle one could imagine a totally extroverted thinker who follows G. Evans’s advice and formulates his questions and answers only in the world-oriented fashion (“Here are some drops of water! So, it’s raining, after all.”), never in a reflective fashion (“Gee, I was wrong, after all!”). But this is going too far; normal cognizers admit of mistakes, and often of the need to revise their opinion. So, integration might turn out to be more reflective in the normal case than Greco is willing to admit. This brings us to the VE alternative to Greco’s immediate integration proposal is the view that integration should be reflective, most famously defended by E. Sosa (2007, 2009, and 2011). According to the general picture, immediate justification might derive from various sources, either external, like reliability, or internal, like obviousness or just from being perceptually appeared to in a certain way. Such an immediate justification is ubiquitous, and might accrue to very simple cognizers. In contrast, a sophisticated thinker is capable to raise the issue of how reasonable it is for her to trust her own abilities and their reliability and is able to reflect about it. This level of reflection is the second level in relation to the first, spontaneous one. Here, the reasons are accessible, or "internal" to the thinker herself, and bestow an "internal" justification, the same way obviousness does at the first level. If a cognitive capacity is reliable (i.e. has external justification) and the cognizer has good reason available to herself to trust it (i.e. has internal justification) then the cognizer is justified (virtuous) in an absolute sense. Unfortunately, I have to be very brief here. Reflective justification of a given belief can proceed by appealing to beliefs of the same kind: my be-

164

Nenad Miščevič

lief that this table is brown can be justified by appeal to my other perceptual beliefs. It is the same-kind justification.. This strategy has been made famous under the name of (search for) narrow reflective equilibrium, in the tradition of moral and political philosophy founded by Rawls (see DePaul 1993). Now, thoughtful cognizers want to know whether their intuitional beliefs cohere with the rest of their belief system, both positively and negatively. Both types of coherence are important: positive, being supported by the system, and negative, not being undermined by it. Such a concern is commanded by the requirement of total evidence. And the check is feasible. But such a reflective justification typically mobilizes capacities distinct from the original capacity that has produced the belief-candidate for being justified, in order to assess the reliability of the original capacity. It has to go beyond justifiers that are of the same-kind (“homogenous”) as first-level immediate ones, in order to enlarge the circle of justification (and thus avoid viciousness), and is, therefore, holistic and coherentist. A serious reason for appeal to empirical evidence is that there is an important asymmetry between deliverance of senses (perception) and intuition. In the case of perception even the most naive cognizers have a viable quasi-theory about its reliability: the things outside interact with us and produce in us the relevant perceptual states. The quasi-theory is then further developed in a detailed scientific account, capable of supporting our trust. There is nothing of comparable weight available in the case of intuitions, as testified by the persistence of Benacerraff’s problem. When you don’t know where the information comes from, you have a right, and perhaps a duty to doubt and double-check. So, waiting for an explanation, we better support our a priori data with some a posteriori material. Next, a full reflective justification will have to involve some explanation of our armchair intuitions, that will be much more difficult to find than the one for perceptual data. That explanation will be a posteriori, as noted by John Skorupsky more than two decades ago: “If minds are only a part of nature, there can be no a priori inference from a state of mind, a belief – be it innate or otherwise, be its negation inconceivable or not – to the truth of the state of mind.” (Skorupsky 1989, 159)

In short, reflective justification of armchair beliefs, presumably produced by intuition and some reasoning, should revert to empirical considerations testifying to their reliability.

Apriority and Virtue: How Successful a Relationship?

165

An objection immediately comes to one’s mind, and it has been reiterated at least since Kant. Suppose that some a posteriori account assures the thinker that her mathematical competence is in order. Does the thinker really have to believe the a posteriori account in order to justifiably trust her elementary mathematics? After all, simple mathematical truths are much more certain than any complicated a posteriori account we might ever hope to get. The answer is that no such nonsense is entailed or even suggested by the two premises. One should be careful to establish the proper order of justification: First, the obviousness and indubitability gives the thinker a prima facie reason for accepting one’s intuitions. The explanation-based doubts make it vivid that this reason might not be sufficient. But then, secondly, the success comes in. On can later add indispensability. These features are the ones capable of completely justifying the reliance on intuitional knowledge. They come in very handy since they point to the massive empirical success of everyday knowledge and of science in which such beliefs are essentially used. The success does a posteriori vindicate the certainty of elementary logical and mathematical intuitions, for which there is a massive overlap with factual domain. Therefore, reflective justification of armchair beliefs typically combines a posteriori elements in an articulated way, on order to strengthen thinker’s reflective trust in her armchair capacities. I am, of course, painfully aware, that this is a merest sketch, but I have to leave it as it is for this occasion (for more on epistemology of intuitions see Miščevič 2004, 2005, 2006). 4. Conclusion The paper has argued for a quite surprising conclusion that ability-based VE goes with aposteriority. So, why does virtue theory pull this way? Probably because of a particular combination: the externalist-reliabilist motivation guiding our judgment of relevant armchair ability or abilities, and the integration and availability of facts about one’s own reliability as specifically virtue-involving addition. In short, we value capacities for their reliability. So, ability-VE comes very close to (mere?) reliabilism. It minimizes the importance of subjective traits: obviousness with its kin, and requires cognitive integration, either direct (Greco) or reflective (Sosa). The only naturalistically plausible difference going beyond reliability would be cognitive integration. Demand for integration stresses the con-

166

Nenad Miščevič

nections between a given competence-virtue, e.g. intuition, and the rest. This is common to Greco-style demand and to Sosa’s reflectively integrated knowledge. But, the integration of any candidate a priori source or piece of knowledge with some heterogeneous source or piece brings in aposteriority. Some might think it is bad, whereas I think it is good. Be it as it may, virtue goes with aposteriority and virtue epistemology should admit aposteriority into the armchair in order to obtain a well-integrated armchair. My own preference is for a two-level structure, of the kind proposed by Sosa, with more externalist virtue-based justification at the basis and a more internalist-coherentist on the second, reflective level. Of course, reflective justification of a given belief can proceed by appealing to beliefs of the same kind, and this “homogenous” style of reflective justification has been proposed for all sorts of sources of beliefs. Unfortunately, it results in a rather narrow circle of justification, the narrow reflective equilibrium. The “heterogeneous” style, on the contrary, consists in deploying beliefs of different origin in order to justify the target one(s), and thus relies on a wider range of capacities-virtues. It is argued that this style is preferable. A consequence of the proposal is that the second-level virtue-based justification of intuitional beliefs will have an important reflective a posteriori component. So, what is the final verdict? The demand for cognitive integration of the deliverances of intuition yields a structured justification with some a priori elements, but with an a posteriori component, perhaps sufficient to make it predominantly a posteriori. A traditional principle says that if justification (or entitlement) contains a posteriori elements, then it is ultimately a posteriori (e.g. if it is mixed and contains one a posteriori element, it is ultimately a posteriori).So, if you accept the principle, you might talk about structured aposteriority of armchair intuitions. If not, just about their structured justification and entitlement, comprising an important a posteriori component.

Apriority and Virtue: How Successful a Relationship?

167

References Alston, W.P., 1989: Epistemic Justification, Ithaca: Cornell University Press Boghossian, P. and C. Peacocke (ed.), 2000: New Essays on the A priori, Oxford: Clarendon Press Boghossian, P. 2008: Content and Justification, Oxford: Oxford University Press Bonjour, L., 1980: Externalist Theories of empirical Knowledge, Midwest Studies in Philosophy 5: 53-73 DePaul, M., 1993: Balance and Refinement, Beyond Coherence, Method of Moral Inquiry, London: Routledge Greco, J., 2010: Achieving Knowledge, A Virtue-Theoretic Account of Epistemic Normativity, Cambridge: Cambridge University Press Horwich, P., 2000: Stipulation, Meaning, and Apriority, in: New Essays on the A priori, ed. P. Boghossian and C. Peacocke, Oxford: Clarendon Press, 150-169 Horwich, P., 2005: Reflections on Meaning, Oxford: Clarendon Press Kant, I., 2004: Prolegomena to Any Future Metaphysics, Cambridge: Cambridge University Press Miščevič, N., 2004: Is apriority context sensitive, Acta Analytica 20, 55-80 Miščevič, N., 2006: Intuitions: The discrete voice of competence, Croatian Journal of Philosophy 16, 69-96 Miščevič, N., 2007: Virtue -Based Epistemology and the Centrality of Truth Towards a Strong Virtue-Epistemology, Acta Analytica 22, 239-266 Nasar, S., 1998: A Beautiful Mind: The Life of Mathematical Genius and Nobel Laureate John Nash, Austin: Touchstone Skorupsky, J., 1989: John Stuart Mill, London: Routledge Sosa, E., 2007: A Virtue Epistemology, Apt Belief and Reflective Knowledge, Vol. 1, Oxford: Oxford University Press Sosa E., 2009: Reflective Knowledge, Apt Belief and Reflective Knowledge, Vol. 2, Oxford: Oxford University Press Sosa, E., 2011: Knowing full well, Princeton, Princeton University Press Wright, C., 2001: On basic logical knowledge, Philosophical Studies 106, 41-85 Wright, C., 2004a: On Epistemic Entitlement: I Warrant for Nothing and Foundations for Free? Proceedings of the Aristotelian Society 104, 167-212 Wright, C., 2004b: Intuition, Entitlement and the Epistemology of Logical Laws, Dialectica 58, 155-175 Zagzebski, L., 1996: Virtues of the Mind: An Inquiry into the Nature of Virtue and the Ethical Foundations of Knowledge, Cambridge: Cambridge University Press

III. The Nature and Value of Knowledge

Epistemology without the Concept of Knowledge ANSGAR BECKERMANN, University of Bielefeld

1. In epistemological discussions about the concept of knowledge, people have time and again considered why knowledge should not simply be defined as true belief. Crispin Sartwell has defended this position especially vigorously. Sartwell even claims that every definition of the form “knowledge = true belief + X” is in danger of becoming incoherent or even inconsistent. He thinks that this is especially true if we assume that the additional condition X is the requirement that the belief should be justified. What reasons have led Sartwell to this conviction? In his paper ‘Why Knowledge is Merely True Belief’ Sartwell notes, first of all, that it seems more than reasonable to assume that the word ‘knowledge’ refers to the goal for which we aim when we are trying to obtain information: “knowledge is our epistemic goal in the generation of particular propositional beliefs” (Sartwell 1992, 167). If this is right, all attempts to find the correct definition of ‘knowledge’ only aim at finding out what this goal ultimately consists in. If we apply this to the present discussion, we are led to ask: Is it more plausible to say that the goal of our epistemic efforts lies in the acquisition of true beliefs or in the acquisition of beliefs that are true and justified? For Sartwell, the case is clear: All our epistemic efforts are ultimately aimed at nothing but the truth. In order to justify this, he first points out that virtually all epistemologists agree on the fact that justification has to be truth-conducive. Laurence BonJour puts especially strong emphasis on this point: If epistemic justification were not conducive to truth in this way, if finding epistemically justified beliefs did not substantially increase the likelihood of finding true ones, then epistemic justification would be irrelevant to our main cognitive goal and of dubious worth. It is only if we have some reason for thinking that epistemic justification constitutes a path to truth that we as cognitive beings have any motive for preferring epistemically justified beliefs to epistemically unjustified ones. Epis-

172

Ansgar Beckermann

temic justification is therefore in the final analysis only an instrumental value, not an intrinsic one. (BonJour 1985, 8)

According to BonJour, and many others, justification has thus only an instrumental value, consisting in its tendency to bring us closer to our main goal of having true beliefs. All authors who subscribe to this view have, according to Sartwell, already accepted (at least implicitly) that knowledge consists merely in true belief, and that justification serves as a criterion rather than as an additional constituent of the concept of knowledge. For it is in his view simply incoherent to include in the definition of a goal something that only serves as a means of reaching it. Another way of putting the matter is this. If we describe justification as of merely instrumental value with regard to arriving at truth, as BonJour does explicitly, we can no longer maintain both that knowledge is the telos of inquiry and that justification is a necessary condition of knowledge. It is incoherent to build a specification of something regarded merely as a means of achieving some goal into the description of the goal itself; in such circumstances, the goal can be described independently of the means. So, if justification is demanded because it is instrumental to true belief, it cannot also be maintained that knowledge is justified true belief. (Sartwell 1992, 174)

Sartwell thinks that this conclusion can be further supported by another consideration. Like others, he does not deny that we have some interest in justification; that we try to have reasons for our beliefs; and that we are not content in simply having those beliefs. Evidently, however, one could perfectly well ask why that is the case. If justification were part of the goal at which our epistemic efforts are aimed, this question would be completely out of place. “For there is no good answer to the question of why we desire our ultimate ends.” (Sartwell 1992, 175) But in fact, that question is not out of place. And this, too, yields a clear indication that justification is not the goal of our epistemic efforts, but merely a means of reaching it. Sartwell in effect proposes the following equation: knowledge = the goal of our epistemic efforts = true belief. His reasons for the second part of this equation can be easily summarized: If anyone were to dispute this part and maintain that the goal of our epistemic efforts lies rather in justified true beliefs, he or she would be immediately faced by a dilemma: either justification is (a) a means of acquiring

Epistemology without the Concept of Knowledge

173

true beliefs, or (b) it is a means of achieving other goals, if not an end in itself. (a) If justification is only a means of acquiring true beliefs, then the goal of our epistemic efforts cannot be justified true belief. For it would be incoherent to include a means of achieving a certain goal in the definition of that goal itself. (b) If, on the other hand, justification is a means of achieving other goals, or if it is an end in itself, then the tripartite concept of knowledge also becomes incoherent. For in that case, we will not be aiming for a single goal when we strive for knowledge, but rather for two goals, which need not always be realizable by the same course of action. On the other hand, if justification is valued not for its truth conduciveness, but for its conduciveness to some other goal, for example, successful adaptation, or for that matter, if justification is itself proposed as an intrinsic goal (a demand of reason, for example), then knowledge is an incoherent notion. It gives us two goals for inquiry, which cannot always be realized simultaneously. (Sartwell 1992, 180)

For these reasons, it is in Sartwell’s view impossible to avoid the following conclusion: […] either justification is instrumental to truth or it is not. If it is, then knowledge is merely true belief. If it is not, there is no longer a coherent concept of knowledge. Thus knowledge is mere true belief. Q.E.D. (ibid.)

However: though Sartwell’s arguments are quite interesting, his two conclusions are nevertheless unwarranted. Consider first case (a). Sartwell does unfortunately not distinguish sufficiently clearly between, on the one hand, the view that justification is a means of acquiring true beliefs and, on the other hand, the view that justification serves as a test or a criterion that we use to check whether a given belief is true. In both cases, however, it would not be fundamentally incoherent or inconsistent to combine knowledge and justification in a single definition. For, in the first place, it may well be that someone would like to achieve some goal, but also cares about the means by which that goal is achieved. E.g., it is possible that someone might want to stand on the summit of Mount Kilimanjaro, but, rather than to have been flown there by helicopter, wants to have reached the summit on his or her own feet. And it is possible that someone might want to own a certain Edward Hopper painting, but, rather than to steal it, wants to pur-

174

Ansgar Beckermann

chase it from his or her own money. The same holds for criteria. It admittedly sounds odd to define the concept of acid as follows: An acid is a chemical compound which, due to electrolytic dissociation, acts as a donor of protons and acid anions, and which causes litmus paper to be colored red.

But this definition, too, would be neither incoherent nor inconsistent. If we want, we could certainly agree to call ‘acid’ only those substances that meet both criteria. After all, not all chemical substances do so. Hence: even if we accept that justification is a means of acquiring true beliefs or a criterion that we use in order to check whether a given belief is true, it would be neither incoherent nor inconsistent if we combine knowledge and justification in a single definition. Analogous remarks can be made for (b): Why should it be incoherent or inconsistent to include X and Y in the definition of a goal if Y is either a means of achieving something else or an end in itself? Of course I might be interested in antique coins, even though not all coins are antique and not all antique objects coins. And of course, I might have the goal to acquire Louis XVI furniture that once belonged to Marie Antoinette, even though not all Louis XVI furniture was at some point in Marie Antoinette’s possession, and even though not every piece of furniture that was ever owned by her belongs into that period. In a word, there is no reason to believe that it would be incoherent or inconsistent to assume that the goal of our epistemic efforts lies not just in true beliefs but rather in justified true beliefs – or more generally: beliefs that are not merely true, but which also have a further property X. 2. From this perspective, Sartwell’s arguments must probably be seen as a failure. But his considerations can also be interpreted from a different perspective. As we have already seen, Sartwell emphasizes time and again that very many authors agree in taking the real goal of our epistemic efforts to lie in truth, while regarding justification as having only instrumental value. Perhaps it is therefore better to understand him as follows: Irrespective of the question of whether it is incoherent to assume that the goal of our epistemic efforts does consist in the having of justified true beliefs, in fact the real goal of our epistemic efforts consists in nothing but truth. Whether our beliefs are justified is of interest to us only as long as we are unsure as to

Epistemology without the Concept of Knowledge

175

whether they are actually true. Is at least this claim supported by convincing reasons? Consider the following case: Someone, let’s call her Anna, claims that p. Should I believe what she says? Should I adopt this belief myself? Should I make it the basis of my future deliberations and decisions? As soon as I become firmly convinced that p is true (or false), all these questions are put to rest. However: How can I find out whether p is true? Truth is epistemically inconspicuous; beliefs don’t tend to wear their truth values on their sleeves. Precisely for this reason, we are interested in criteria for truth – conditions to which we have readier epistemic access than to the respective belief’s truth value, and which indicate that the belief in question is (at least probably) true. Justification is one such criterion. If Anna claims that p, I will ask, “And why do you believe that?”. If she then mentions some plausible circumstances that lend support to her claim, i.e., if it emerges that her belief is justified, this will also support p itself; and so an answer of this sort will usually lead me to accept p as well. If, however, I am (for whatever reasons) already beforehand firmly convinced that p is the case and, hence, that Anna’s claim is true, then I will no longer ask the question, “And why do you believe that?”. This, in turn, shows that what I am really interested in is the question of whether Anna’s belief is true, and not whether it is true and justified. For the question of whether her belief is justified will interest me only as long as I am not yet sure whether it is true. This example first of all shows that my epistemic efforts primarily concern my own beliefs; I aim at a system of beliefs that only comprises true beliefs, or at least as many true beliefs and as few false beliefs as possible. Second, the example shows a way how to decide whether the goal of our epistemic efforts lies in true beliefs or in justified true beliefs. Obviously, this question has the form: Does our goal lie in F? Or does it rather lie in F and G? And questions of this sort can be answered by a simple test: Will we still be interested in the question of whether something is G if we are already convinced that it is F? Suppose that we are interested in particularly hard diamonds, that this kind of hardness, however, is a property that is very difficult to be determined. Suppose further that, by empirical investigation, we have found out that all (or almost all) pink diamonds possess the specific degree of hardness we are interested in. As a result, the color pink has some greater significance for us. Whenever we will be presented with diamonds, we will henceforth pay special attention to the question of whether those diamonds

176

Ansgar Beckermann

are pink. But still, it would be wrong to say that our interests have shifted from especially hard diamonds to especially hard pink diamonds. And this can be seen from the fact that the question of whether a given diamond is pink will no longer interest us as soon as we can be sure – independently of the diamond’s color – that it possesses the desired degree of hardness. It seems to me that the relationship between truth and justification is quite similar. We ask whether a belief is justified, because we (a) are interested in whether it is true, because (b) truth is epistemically inconspicuous and because (c) justification is a reliable truth criterion. The just mentioned example shows this very clearly. I ask Anna whether she can adduce any reasons for her belief that p because and as long as I am not already sure whether p is the case. If (for whatever reasons) I am already convinced that Anna’s belief is true (or false) I am no longer interested in whether Anna can adduce any reasons. What I am really interested in therefore is whether Anna’s belief is true and not whether it is true and justified. Even if we replace the justification requirement by some other condition, things will not look radically different. Thomas Grundmann, e.g., holds that the real goal of our epistemic efforts lies not in true opinions but in non-accidentally true opinions. What we strive to attain in our epistemic efforts (Erkenntnisbemühen) is not truth simpliciter but non-accidental truth. And according to the analyses of the postGettier generation, this is precisely the correct definition of knowledge! So it is knowledge that appears to be the real goal of our epistemic efforts. (Grundmann 2002, 121)

But is this true? Grundmann does not make use of the simple test just described, but instead tries to show that non-accidentally true opinions are better, more valuable than merely true opinions. But why should that be the case? Practical considerations are here evidently irrelevant. “It is no doubt correct that the practical utility of non-accidental truths [nicht-zufällige Wahrheiten] is no greater than that of truths simpliciter” (ibid., 123). But what, then, should convince us of the higher value of non-accidentally true opinions? At first, Grundmann is somewhat at a loss: “Admittedly I can give no truly compelling argument for this goal.” But then he begins the following line of thought: If we assume that the real goal of our epistemic efforts lies in non-accidental truth, and if we thus define ‘knowledge’ as non-accidentally true belief, then we will in that way have accommodated two intuitions at the same time: first, the externalist intuition that knowl-

Epistemology without the Concept of Knowledge

177

edge requires a reliable belief-forming process, and second, the intuition that knowledge must on the whole be accorded a higher value than mere true belief. For it is precisely this latter intuition that becomes intelligible if we take knowledge to be non-accidentally true opinion and assume that the latter has indeed a higher epistemic value than mere true belief. Grundmann concludes: “In this way we will have saved knowledge as a coherent fundamental concept of epistemology”. But he is honest enough to add: “However, I don’t want to hide the fact that I have not explained why nonaccidental truth is our epistemic goal” (ibid.). In other words, Grundmann does not answer the central question: Does the goal of our epistemic efforts really consist in non-accidentally true opinions? Or only in true opinions? Clearly, this question can only be answered if we first clarify what exactly a non-accidentally true opinion might be. I will base my considerations in the following on a proposal that should be acceptable to most readers, if not to all: (*)

A belief b is non-accidentally true just in case b is true and some conditions obtain that make it probable that b is true.

Thus, the non-accidental truth of a belief b depends, according to this proposal, on the obtainment of conditions that make it probable that b is true. These conditions might consist in the fact that b has been formed by a reliable process, but they might also consist in the fact that the subject is justified in having b – at least if one assumes, in accordance with most advocates of justification-based accounts, that justification has to be truthconducive or truth-indicative. Already at first glance, however, the question arises why nonaccidentally true beliefs in the sense of (*) should be more interesting than beliefs that are merely true. If we are already convinced that b is true, why should we be further interested in the question of whether any conditions obtain that make it probable that b is true? The situation is no different from the case of the justification requirement: the question of whether the additional condition X obtains will interest us only as long as we have not yet ascertained that the truth requirement is satisfied. In other words: the additional condition that Grundmann proposes – non-accidentality – also fails the test mentioned above. For here, too, we are no longer interested as to whether this condition is satisfied if we are already sure that b is true. This becomes especially clear if we ask the following question: What if

178

Ansgar Beckermann

truth were not epistemically inconspicuous? Suppose we had direct, noninferential access to the truth values of our beliefs; that we could as it were ‘see’ whether a given belief is true or not. How would this change our epistemic situation? Well, I would much like to find out if I’m going to win the lottery this year. So I visit a fortune-teller, who readily bends over her crystal sphere. After a long while (and some unintelligible muttering), she looks up and tells me, “I’m very sorry – no such luck”. Since I have a direct, noninferential access to the truth values of beliefs, I straightaway ‘see’ that she is right. So what would I do next? I think I would say: “Too bad, maybe next year”. And that would be it. I would not be interested in whether she really knows her craft, whether she is a reliable informant. I would not be interested in whether she has perhaps only ventured a guess, or whether she really ‘knows’ that what she has told me is true. After all, I can ‘see’ that it is true. And thus all those questions can be put to rest. To emphasize once more the crucial point. The intuition that beliefs also have other intrinsically interesting features besides their truth values stems solely from the fact that truth is epistemically inconspicuous. We are interested in these other features because (and as long as) they provide us with information as to whether the respective beliefs are (probably) true. If we had the ability to be non-inferentially sure as to what is true and what isn’t, all these features would have no significance. In other words: They really have only instrumental value. But that this is so is an empirical fact – it could have been otherwise. Our considerations could also have yielded the result that besides their truth values, beliefs have still other features that are of intrinsic interest to us. But as a matter of fact, this is not the case. As a matter of fact, all those other features are of interest to us only as means towards our real goal. A further point must be added. Suppose that someone who wants to do me a favor gives me a small bag with diamonds – without any further comment. I have no idea whether these diamonds possess the hardness I am looking for. In fact, however, they do. Have I now reached my goal? In a certain sense, I have. I now possess a couple of diamonds with the desired hardness. But, on the other hand: I have no idea that I have reached my goal. This is not satisfactory. If I pursue a certain goal G, I will not be content if G simply comes true, but only if I also have a means to ascertain that G has come true. With regard to our beliefs the situation is exactly similar. Even if all my beliefs were in fact true I would still be unsatisfied

Epistemology without the Concept of Knowledge

179

if I had no idea that that is the case. For, as far as I am informed, all my beliefs could be false. I will therefore be satisfied only if my beliefs are not only true but if I am also in a position to tell that they are true. Does this not indicate, after all, that the goal of our epistemic efforts is not just to have true beliefs, but true beliefs for which we have good reasons to think that they are true? I don’t think so. For even in the diamond-case it would be wrong to say: What I am really interested in are not particularly hard diamonds, but particularly hard diamonds for which I have good reasons to believe that they are particularly hard. That we, in pursuing a certain goal G, are going to be satisfied only if we not only have achieved G but are also in a position to be sufficiently sure that G has come true, does not change the fact that our goal simply is G. 3. What consequences – terminological and otherwise – should we draw from this result? Sartwell’s position is as follows: Since ‘knowledge’ is the word that we use in order to refer to the goal of our epistemic efforts, we should simply equate knowledge with whatever that goal consists in. If it consists only in true beliefs, rather than in true beliefs + X, we should consequently equate knowledge with true belief. I think, however, that this would be dialectically unwise, since it would inevitably lead to various misunderstandings (cf. Beckermann 2001, 577f.). My own proposal is that at least in epistemology, we should completely abstain from using the concept of knowledge. There is, in my view, no interesting epistemological question that could not be discussed equally well – and perhaps better – without this concept. This proposal might raise the immediate worry that in practice it is apparently quite difficult to get along without any use of the verb ‘to know’. In making this point, one could even refer to my own writings, in which I have not always seemed able to do so myself. For instance, take the passage from my essay Zur Inkohärenz und Irrelevanz des Wissensbegriffs in which I deal with the relationship between intrinsic and instrumental goals. The verb ‘to know’ occurs here no fewer than five times: Basically the situation is as follows. We are interested in everything that has the property F – in short: we are looking for Fs. But F is a property of which we cannot immediately tell whether it is instantiated by any given object. However, we know that objects that have the property G usually also have the property F; and we have

180

Ansgar Beckermann

easier epistemic access to G than to F. So we will be looking for objects that have the property G, or more precisely: for objects of which we know that they have this property. Once we have found such an object, say, a – once we know, in other words, that a is G –, we can be pretty sure that a is also F. For this reason, we are also interested in the property G. But this does not mean that in reality we are not looking for Fs at all, but rather for Fs that are also Gs. This becomes clear if, in particular, we consider that we would not in the least be interested in the question of whether a is G if we had already known that a is F. And exactly the same holds for our beliefs: whether they are justified is of interest to us only because, and as long as, we do not know whether they are true. (2001, 579; emphases in boldface added)

I confess that these formulations are the result of barely excusable negligence. However: they do not show that the verb ‘to know’ is, even if only in practice, indispensable. For there is an alternative. In the rejoinder that was, one year later, published under the title, Lässt sich der Wissensbegriff retten? (‘Can the concept of knowledge be saved?’), I again employ – almost verbatim – the considerations just quoted. This time, however, without any uses of the verb ‘to know’: Suppose we are interested in everything that has the property F – in short: we are looking for Fs. But F is a property of which we cannot immediately tell whether it is instantiated by any given object. However, we have sufficient evidence for assuming that objects that have the property G usually also have the property F; and we have easier epistemic access to G than to F. So we will be looking for objects that have the property G. Once we have found such an object, say, a – once we can be sure, in other words, that a is G –, we can also be pretty sure that a is F. For this reason, we are also interested in the property G. But this does of course not mean that in reality we are not looking for Fs at all, but rather for Fs that are also Gs. This becomes clear if, in particular, we consider that we would not in the least be interested in the question of whether a is G if we were already sure that a is F. (2002, 589; emphases in boldface added)

As this example makes clear, we can dispense with the verb ‘to know’; and this is not because we can replace each occurrence of this word by its definiens (cf. Hofmann 2002, 124), but rather because we can use other terms instead, which may even be better suited to the purpose at hand. I do in fact think that we will be better able to conduct epistemological research if we refrain from using the concept of knowledge. Once we stop using this concept, many things will become clearer and simpler. First of all, it is of course clear that all discussions about how to define or explicate the concept of knowledge will become superfluous. This will save us a lot of time and effort though perhaps some will fear to have labored in vain or to be put out of work. For example, if the concept of

Epistemology without the Concept of Knowledge

181

knowledge is renounced, all work on the topic of contextualism will be rendered obsolete. But let us here not focus on that. Instead, let us consider, as an example, the problem of skepticism. Skeptics, we are told, are philosophers who defend the thesis that we cannot (really) know anything. But one should not suppose that skeptics are engaged in the defense of positive claims (their position would then be immediately open to attack); they are better conceived of as questioners: Can we (really) know anything at all? In both formulations, however, the verb ‘to know’ plays a crucial role. Would the skeptic’s position even be intelligible if this term were omitted? I think that it would; and this can be seen precisely by considering two of the main arguments from the skeptics’ repertoire. In Pyrrhonian skepticism we can find two main lines of argument: one to the effect that every claim can be just as well justified as some other claim incompatible with it;1 and the other being based on Agrippa’s trilemma. A tower looks round from afar, but square from nearby. We have no reason to favor the one appearance over the other; hence, the proposition that the tower is round is exactly as well justified as the proposition that it is square. From our sense experience, we learn that everything around us is in motion and subject to change; but the arguments of the Eleatic philosophers purport to show that there can be no such thing as movement or change. We have no reason to invest greater trust in our sense experience than in the Eleatic arguments, or vice versa; and so the proposition that the things around us are in motion and subject to change is exactly as well justified as the proposition that it is not so. The orderly motion of the heavenly bodies is a reason to believe in providence; but the fact that virtuous people often have to suffer while the wicked enjoy their lives is a reason not to believe in any such thing. We 1

According to Sextus, however, it will not be correct to say that the skeptic claims that every claim can be just as well justified as some other claim that is incompatible with the former (nor even that it appears to him this way). From Sextus’ point of view, skepticism is rather a form of art − the art (or ability) to find, whenever a justification of some claim is proposed, an equally strong justification for a claim that is incompatible with the former. What does one gain by mastering this art? Well, Sextus holds that we will not attain ataraxia (tranquility of the soul) if we do our best to answer our own questions; rather, ataraxia is (as may be empirically shown) the result of abstaining from all judgment. In order to attain it, we therefore have to achieve just this: to abstain from all judgment, in all matters. And this will be best achieved if we are able, for any justification of any given proposition, to think of an equally strong justification that supports the contradictory of that proposition (Sextus 2000, I, 4 & 12).

182

Ansgar Beckermann

have no reason to favor the one reason over the other; and so the proposition that there is such a thing as providence is exactly as well justified as the proposition that there isn’t. Agrippa’s trilemma is based on the assumption that any proposition A can be considered justified only if it is possible to produce reasons for A that are themselves justified. But for any such chain of reasons there are only the following three possibilities. (i) The chain terminates after some finite number of steps, i.e., it ends in a reason that does not itself have a justification; in this case, A will not be justified, either. (ii) The chain forms a circle, i.e., it contains reasons that have already been used earlier (we thus have an attempt to justify some proposition by itself); here again, A is not justified. (iii) The chain continues ad infinitum, i.e., it never reaches a point at which everything that requires a justification has in fact been justified; in this case, too, we have no justification of A. Evidently, the Pyrrhonian arguments mentioned here do not concern knowledge but justification. Their main claim is apparently this: that we are not able satisfactorily to justify even a single proposition.2 And behind this idea, we can discern the following picture of the human quest for truth. Many things seem of interest to us: Is the tower actually round, or is it square? Are things constantly changing or do they remain the same? Is there such a thing as providence, or is there not? Which one of the corresponding propositions is true, and which one is false? Truth is epistemically inconspicuous; we cannot ‘see’ whether a given claim is true. Hence, we require criteria – properties to which we have easier epistemic access, and which are such that any claim that instantiates them is (probably) true. The only reliable criterion of this sort is justification: a claim is (probably) true if it can be justified and if no claim that is incompatible with it can be equally well justified. With this picture in place, the rest of the argument can be easily completed. The criterion is never fulfilled: on the one hand there is, for every proposition A that can be justified, some other proposition that is incompatible with A but can be equally well justified; and on the other hand, for the reasons adduced by Agrippa it is in any event im2

Pyrrhonian skeptics do not claim anything to be one way or other, they only claim that things seem to be such and such. So it would be better to say: Pyrrhonians skeptics claim that it seems that we are not able satisfactorily to justify even a single proposition.

Epistemology without the Concept of Knowledge

183

possible to come up with satisfactory justifications. We thus have a reliable criterion for truth, but one that is never met. So we have no way of deciding for any given proposition whether it is true or not. (Hence we have to abstain from judgment; and when we do so, ataraxia will take hold all by itself.) At the beginning of the Meditations, Descartes, too, seems to be concerned with knowledge. Can we really know anything at all? Is there at least one proposition that we can be sure of with absolute certainty? Descartes tries to shake the naïve confidence of ordinary people by asking: Could it not be that much (or almost all) of what we take to be true is in fact false, even though from our perspective, we have every reason to believe that it is true? Could it not be that the world around us is (almost) never the way that, based on our sensory experience, it seems to be? Yes, he answers, that could be, e.g., if we were dreaming all the time, i.e., if our sensory experiences were not caused by the things in our environment in the way we normally assume they are. Descartes further asks: Could it not be that a square does not have four sides, after all, or that the square of the hypotenuse of a right triangle is not equal to the sum of the squares of the two other sides – even though this seems so clearly to be the case, and can even be proved? Yes, he answers, even this might happen, e.g., if even our capacity to reason were defective due to a malicious demon who “has employed all his energies in order to deceive” us (Meditations AT 22, CSM II, 15). So, what is it that Descartes is really concerned with? In my view, it is evident that Descartes is first of all interested in the reliability of certain epistemic methods. Is perception a reliable method of acquiring true beliefs? Is reasoning a reliable method? And, as his further arguments show, he is also interested in reliable criteria. Having found out that the thought that he himself exists cannot be false – at least as long as he entertains this thought – he uses this first example of a belief that he can be absolutely sure of to find out which feature of this belief is responsible for its absolute certainty. His answer is more than well known: In this first item of knowledge there is simply a clear and distinct perception of what I am asserting; this would not be enough to make me certain of the truth of the matter if it could ever turn out that something which I perceived with such clarity and distinctness was false. So I now seem to be able to lay it down as a general rule that whatever I perceive very clearly and distinctly is true. (Descartes 1984, 24; AT VII, 35).

184

Ansgar Beckermann

Using this absolute fool proof criterion, Descartes goes on to prove the existence of god. And this proof enables him to answer the question of whether sense perception and reasoning are reliable epistemic methods: Yes, both methods are in principle reliable, because a perfect god (and not some malicious demon) has equipped us with epistemic powers that in principle (and provided we are careful enough) allow us to partake of the truth. It seems to me that the skeptical arguments just reviewed concern two questions which in my view are the most important questions of epistemology – two questions that have little to do with the concept of knowledge: Are there reliable criteria on the basis of which we can tell true from false beliefs? Are there reliable methods that would help us to acquire true beliefs? And if so, what are they? I think we will be much better able to understand the skeptic if we realize that he is not primarily concerned with the possibility of knowledge, but rather with criteria for truth and reliable methods for the gathering of true beliefs.3 In encyclopedia articles and introductory texts one often reads that the three basic questions of epistemology concern the nature, the sources, and the scope of knowledge. This is certainly not completely wrong, but on the other hand it does not offer a very clear picture of what epistemology is really all about. Instead, it seems to me that the two central questions of epistemology can be put as follows: 1. 2.

Are there reliable criteria for truth? And if so, what are they? Are there reliable methods for the acquisition of true beliefs? And if so, what are they? Do they include perception, memory, reasoning, rational intuition?4

This is not to say that the two questions about the nature and scope of knowledge (or rather: about the nature and scope of reliably attainable true belief) play no role at all. But one should not misunderstand the question 3

Perhaps it is a little unfair to the Pyrrhonian to say that he is concerned with these two questions. For he rather seems to claim that there clearly is a truth criterion – justification –, but that this criterion is never met. 4 These two questions are of course not independent from each other; for if a given belief has been formed by a reliable process, this will clearly be a reason to think that it is true.

Epistemology without the Concept of Knowledge

185

about the ‘nature of knowledge’ as a question about the concept of knowledge. In my view, this question is rather to be understood as asking what the real goal of our epistemic efforts consists in. (And if I am right, the answer is precisely not ‘Knowledge’, but instead, ‘Truth’.) The question concerning the ‘scope of knowledge’, on the other hand, may be best understood as a more specific version of the second central question. We can wonder about reliable methods not only in a general way, but also more specifically, e.g.: Are there reliable methods for the acquisition of true beliefs about the past, the future, about other minds, in the field of mathematics, about morality? And of course, it need not always be the case that a method that is reliable in one field will also prove reliable in other fields. 4. Finally, I would like to return once more to the topic of justification. Although I have here defended the thesis that justification should not be counted among the goals of our epistemic efforts, this does of course not entail that justification is, on the whole, not an interesting epistemological topic – on the contrary. I merely urge that this topic be placed in the right context. According to the thesis that I have been defending here, justification is properly regarded as a criterion for true beliefs. Reflecting on what diverse beliefs I have, I notice that, among other things, I am convinced that there are craters on the far side of the moon. Or suppose that I see someone on TV who claims something to this effect. I wonder if this belief or claim is really true: Are there really craters on the far side of the moon? Once again: truth is epistemically inconspicuous; we cannot simply ‘see’ whether a given belief or claim is true. So how can I find out whether the proposition in question is true? Criteria for truth must be more readily accessible than truth itself. If someone is searching for criteria that would enable him to ascertain whether his own beliefs are true, he will therefore have to take recourse to criteria that are more accessible to himself. But this does not provide unambiguous support for internalism. If I want to find out whether I am right in believing that there are craters on the far side of the moon, I may ask myself what reasons I have for holding this belief, and whether these reasons are really compelling. But I may also ask how I have come to hold this belief. Have I read it somewhere? Did the source maybe even contain a photograph of the far side of the moon? Is the source reli-

186

Ansgar Beckermann

able? Are there reasons to believe that the photograph has been manipulated? And so on. In the same way, I may use two ways to find out whether someone else’s claim is true. I may ask her, “Why do you believe that?” and see if she can offer a compelling justification. Or I may try to find out whether the belief behind her claim has been formed by a reliable process (independently of whether she herself is aware of it). The crucial point is that internalism and externalism are not at all incompatible once they are seen in the right context. The debate about the concept of knowledge has focused on the question which condition X has to be added to true belief so as to yield real knowledge. For this X, Plantinga has introduced the useful term ‘warrant’ (Plantinga 1993). Internalists take the view that an epistemic subject has to be able to justify a true belief in order to qualify as having knowledge. Externalists, by contrast, hold that a belief will count as knowledge only if it is true and has been formed by a reliable process. I think, however, that we would be mistaken to regard justification and reliability as rivals for the role of ‘warrant’. Much rather, justification and reliability should be seen as two criteria that we can use – be it singly or in combination – to ascertain whether a given belief is true. And consequently, they do not by any means rule each other out, but may instead even complement each other. That this view of the matter compares favorably with the received view has already been argued by Sartwell in his first essay, ‘Knowledge is Merely True Belief’: A more fundamental advantage of the view that knowledge is merely true belief is that, on it, we are under no apparent pressure to choose between a broadly externalist and a broadly internalist account of justification. The pressure to choose between these views arises largely because proponents of each argue that their account is an account of the sense of justification that is logically required for knowledge. But justification is in no sense logically required for knowledge. (Sartwell 1991, 163)

But while Sartwell claims that we should equate knowledge with true belief, I argue that when doing epistemology we should refrain from using of the concept of knowledge, on the ground that this will help us to have a clearer view on the central epistemological questions. Hume’s Enquiry Concerning Human Understanding ends on the following drastic remark: When we run over libraries, persuaded of these principles, what havoc must we make? If we take in our hand any volume; of divinity or school metaphysics, for instance; let us ask, Does it contain any abstract reasoning concerning quantity or number? No. Does it contain any experimental reasoning concerning matter of fact

Epistemology without the Concept of Knowledge

187

and existence? No. Commit it then to the flames: For it can contain nothing but sophistry and illusion. (Hume 1999, 211)

Analogously, maybe I could say: When we run over libraries, persuaded of these principles, what havoc must we make? If we take in our hand any volume of epistemology, let us ask, Does it contain any reasoning concerning criteria for the truth of beliefs? No. Does it contain any reasoning concerning reliable methods for gathering information? No. Commit it then to the flames: for it can contain nothing but sophistry and illusion.

This would probably be too radical. But why don’t we try it out one time? Why don’t we try for a while to concern ourselves only with possible criteria for truth and with the search for reliable methods of belief acquisition? In the end, we will perhaps notice that we are not missing anything at all – from an epistemological point of view – if all other questions (and in particular those that concern the concept of knowledge) are left aside. References Beckermann, A., 1997: Wissen und wahre Meinung, in: Das weite Spektrum der Analytischen Philosophie, ed. W. Lenzen, Berlin / New York: Walter de Gruyter, 24-43 Beckermann, A., 2001: Zur Inkohärenz und Irrelevanz des Wissensbegriffs. Plädoyer für eine neue Agenda in der Erkenntnistheorie, Zeitschrift für Philosophische Forschung 55, 571-593 Beckermann, A., 2002: Lässt sich der Wissensbegriff retten? Zeitschrift für Philosophische Forschung 56, 586-594 BonJour, L., 1985: The Structure of Empirical Knowledge, Cambridge, Mass.: Harvard University Press Descartes, R., 1984: Meditations on First Philosophy, in: Descartes: Selected Philosophical Writings, ed. J. Cottingham and R. Stoothoff, Cambridge: Cambridge University Press, 3-62 (‘AT VII’ refers to the 7th volume of the standard edition of Descartes’ works by Ch. Adam and P. Tannery.) Grundmann, T., 2002: Warum wir Wissen als einen wichtigen Begriff der Erkenntnistheorie betrachten sollten – Eine Antwort auf Ansgar Beckermann, Zeitschrift für Philosophische Forschung 56, 118-124 Hofmann, F., 2002: Die Rolle des Wissens und des Wissensbegriffs in der Erkenntnistheorie, Zeitschrift für Philosophische Forschung 56, 125-131 Hume, D., 1999: An Enquiry concerning Human Understanding, ed. T.L. Beauchamp, Oxford University Press

188

Ansgar Beckermann

Plantinga, A., 1993: Warrant: The Current Debate, Oxford: Oxford University Press Sartwell, C., 1991: Knowledge is Merely True Belief. American Philosophical Quarterly 28, 157-165 Sartwell, C., 1992: Why Knowledge is Merely True Belief. The Journal of Philosophy 89, 167-180 Sextus Empiricus, 2000: Outlines of Scepticism, ed. J. Annas and J. Barnes, Cambridge: Cambridge University Press

Reliabilism as Explicating Knowledge: A Sketch of an Account ERIK J. OLSSON, University of Lund

1. Introduction According to simple reliabilism, S knows that p if and only if (1) p is true, (2), S believes that p, and (3) S’s belief that p was acquired through (sustained by) a reliable process. Henceforth “reliabilism” means “simple reliabilism”. The first clear statement of reliabilism apparently stems from F. P. Ramsey (1931). Since Ramsey, many distinguished researchers have advocated reliabilism in one form or another, including David Armstrong (1973), Fred Dretske (1969, 1971) and Alvin I. Goldman (1979, 1986). For a recent overview of reliabilism and the issues surrounding it, see Goldman (2011). In this paper I will sketch a defense of reliabilism from a certain methodological perspective, namely that of the theory of explication due to the great Austrian philosopher Rudolf Carnap. I will defend reliabilism (a) as a highly promising explication of knowledge, and also (b) as an explication of knowledge that is likely to be better than any other standard proposal on the market, although the latter claim will continue to have the status of a mere plausible conjecture. 2. Reliabilism as an explication of knowledge An explication is a kind of definition satisfying the following conditions (Carnap 1950, 7): 1. The explicatum [the thing that explicates] is to be similar to the explicandum [the thing that is explicated] in such a way that, in most cases in which the explicandum has so far been used, the explicatum can be used; however, close similarity is not required, and considerable differences are permitted. 2. The characterization of the explicatum, that is, the rules of its use (for instance, in the form of a definition), is to be given in an exact form, so as to introduce the explicatum into a well-connected system of scientific concepts.

190

Erik J. Olsson

3. The explicatum is to be a fruitful concept, that is, useful for the formulation of many [true] universal statements (empirical laws in the case of a nonlogical concept, logical theorems in the case of a logical concept). 4. The explicatum should be as simple as possible; this means as simple as the more important requirements (1), (2) and (3) permit.

Thus, to take Carnap’s example, in zoology the artificial concept piscis has come to replace the common sense concept of fish, although piscis is a narrower concept that excludes several kinds of animal which were subsumed under the concept fish, e.g. whales and seals. The main reasons for the replacement was that zoologists found that piscis is a much more fruitful concept than fish in the sense that it allows for the formulation of a greater number of interesting general truths: that all piscis are coldblooded, extract oxygen from water using gills, lay eggs, and so on. At the same time, the concept piscis, apart from being a precise and simple one, is sufficiently close to the common sense concept of fish to replace it at last in a scientific context. This seems to be a typical case of a successful scientific definition. I will proceed on the assumption that scientific definitions of ordinary concepts generally should be seen as explications. Now the most coherent/unified view on philosophical definitions is arguably that such definitions are, in principle, no different from scientific definitions. This follows for instance if one thinks that philosophy is a branch of science or is at least closely related to science or some variation on that theme. Since scientific definitions are explications, it would follow that philosophical definitions are, or should be seen as, explications as well. In fact, not only Carnap but also Quine (1960) held this view. In particular, definitions of knowledge are, or should be seen as, explications. There is an alternative route to this conclusion over a weak form naturalized epistemology stating that epistemology, but perhaps not philosophy in general, is a branch of science. Quine went on, of course, to advocate the naturalization of epistemology (Quine, 1969) which is significant in this context because reliabilism, as we all know, is usually subsumed under that epistemological program (although researchers tend to have different ideas about what exactly the program involves). For a recent defense of the method of explication, see Maher (2007). For reliabilist knowledge to be a good explication of the ordinary concept of knowledge the following must be satisfied to a high degree:

Reliabilism as Explicating Knowledge: A Sketch of an Account

191

1. Reliabilist knowledge must be similar the ordinary concept of knowledge in such a way that, in most cases in which the ordinary concept has so far been used, reliabilist knowledge can be used; however, close similarity is not required, and considerable differences are permitted. 2. The characterization of reliabilist knowledge must be given in an exact form, so as to introduce the concept into a well-connected system of scientific concepts. 3. Reliabilist knowledge must be a fruitful concept, that is, useful for the formulation of many universal statements (empirical laws, since it is a nonlogical concept). 4. Reliabilist knowledge should be a simple concept, e.g. not requiring a complicated definition. That these conditions should be satisfied to a higher degree for reliabilism to be a good explication follows directly from Carnap’s account of explication. We understand the claim that reliabilist knowledge is a good explication, and probably better than other standard accounts, to mean that it is competitive in that regard among a certain class of definitions, all of which assume that knowledge involves true belief plus something more, e.g. JTB. Thus we leave the non-standard TB account out of the picture in this paper. It follows that we need not discuss the exactness, simplicity and so on of the concepts of truth and belief; they are after all common to all competing definitions. Rather, we can focus entirely on the characteristic third clause of the reliabilist explication: that the belief has been acquired through a reliable process. We will deal with the claims 1-4 in the order 4, 2, 3, 1, moving from what I take to be the least to the most contentious. 3. Simplicity Is the third reliabilist clause simple? It states that S’s belief that p was acquired through a reliable process. Yes, that does sound quite simple. Are any of the competing accounts of knowledge simpler, e.g. JTB? No, probably not. It would seem that the reliabilist account of knowledge is at least as simple as any competing account of knowledge, and much simpler than most.

192

Erik J. Olsson

But hold on a second: isn’t the notorious generality problem precisely an argument against the simplicity of the reliabilist account? Reliabilism does not include a method for deciding how a belief formation process should be classified and for deciding whether the process was reliable or not; it is therefore “seriously incomplete” (Conee and Feldman 1998). While reliabilism appears to be a simple theory on first sight, once it is supplemented with such methods it is bound to be extremely complex! Plausible as it may sound this line of reasoning is mistaken. An explication of knowledge states necessary and sufficient conditions for knowledge. Providing additional criteria for deciding, in concrete cases, whether those conditions are satisfied is not part of what an explication of knowledge needs to deliver. Thus, in order to qualify as an explication a JTB style theory does not have to include additional criteria for deciding, in concrete cases, whether a given belief is justified. Nor does a reliabilist style explication have to provide a method for deciding whether the process was reliable (which would involve providing a further method for deciding how the process should be classified). Hence, from the point of view of the current methodology Conee and Feldman’s reasoning fails: the fact that a reliabilist account of knowledge fails to include a method for identifying the process type and for deciding on the reliability of the process does not make that account incomplete as an explication of knowledge. An explication of knowledge is complete once it provides necessary and sufficient conditions for knowledge, which does not mean of course that the proposal cannot have other flaws but that’s irrelevant in the present context. There is a second point to be made in this connection. Let us by a supplemented explication of knowledge mean a package consisting of an explication of knowledge stating necessary and sufficient conditions in the usual way plus additional criteria for deciding when those first conditions obtain. Reliabilism conceived as a supplemented explication will be more complex than reliabilism conceived as an (unsupplemented) explication. But the same is true mutuatis mutandis of JTB, evidentialism or any other competing accounts of knowledge. For instance, a JTB theory which supplies criteria for deciding when a given belief is justified will be more complex than a JTB theory in the usual sense. Be that as it may, the focus in this paper is on reliabilism as providing an (unsupplemented) explication, as opposed to a supplemented explication, of knowledge.

Reliabilism as Explicating Knowledge: A Sketch of an Account

193

The bottom line is that our conclusion still stands: reliabilism provides a strikingly simple explication of knowledge. We will consider a more interesting version of the generality problem in section 6, in connection with Carnap’s first condition, that of similarity with the ordinary concept. 4. Exactness We recall that “the characterization of reliabilist knowledge must be given in an exact form, so as to introduce the concept into a well-connected system of scientific concepts”. Consider again the characteristic reliabilist clause stating that S’s belief that p was acquired through a reliable process. The crucial concept here, obviously, is that of a reliable process. This concept is a concept of statistics: a random process is reliable with regard to an outcome type (in this case “truth”), if outcomes of that type occur more often than not, or often enough for the purposes at hand. Hence, the reliabilist explication of knowledge introduces the concept of knowledge into the well-connected system of scientific concepts represented by statistical theory. I am not aware of any competing account of knowledge that to a similar high extent satisfies the condition of introducing the concept defined into a well-connected system of scientific concepts. JTB for sure does not because ‘justified belief’ is not a concept that occurs in exact science. There is, to be sure, some vagueness here regarding how reliable a process must be to be reliable simpliciter. However, the same kind of vagueness pertains to many applications of statistics to real life, and it does not make statistics less exact. The context determines how reliable something must be in order to be reliable simpliciter. Thus, a reliable method for disposing of nuclear waste is hopefully more reliable than a reliable method for disposing of garden waste. And of course, competing definitions of knowledge show the same kind of vagueness: how justified must a belief be in order to be justified simpliciter? We must conclude that the reliabilist explication of knowledge satisfies Carnap’s desideratum of exactness to considerable degree due to its close connection to statistical theory, and it seems unlikely that its competitors could do better in this regard. This seems particularly unlikely for JTB, evidentialism and virtue epistemology, which all rely on concepts (‘justification’, ‘evidence’ and ‘virtue’) that arguably do not occur in exact science.

194

Erik J. Olsson

5. Fruitfulness “Reliabilist knowledge must be a fruitful concept, that is, useful for the formulation of many universal statements (empirical laws, since it is a nonlogical concept)”. On my reconstruction, challenges to the fruitfulness of the reliabilist concept of knowledge purport to show that such knowledge is no more valuable than mere true belief: the universal inference from knowledge to surplus value is rejected. The famous swamping argument by Jonathan Kvanvig (2003), Linda Zagzebski (1996) and others attempts to accomplish just this. I have rejected this argument in several previous papers, arguing for the following: (A) (B) (C)

If you have reliabilist knowledge of p, then you are more likely to have more true belief (and knowledge) in the future than if you have a mere true belief that p (Goldman and Olsson 2009). If you have reliabilist knowledge of p, then your true belief that p is more likely to be stable than if you have a mere true belief that p (Olsson 2007, 2008) As Williamson (2000) has observed, stability of true belief is positively correlated with practical success when acting over time (Olsson 2007, 2008).

I take these three claims to express true statistical laws relating reliabilist knowledge to cognitive and practical success. The laws are true in our world because our world satisfies a number of empirical regularities such as: (Non-uniqueness) Once you encounter a problem of a certain type, you are likely to face other problems of the same type in the future. (Cross-temporal access) A method that was used once is often available when similar problems arise in the future. (Learning) A method that was unproblematically employed once will tend to be employed again on similar problems in the future. (Generality) A method that is (un)reliable in one situation is likely to be (un)reliable in other similar situations in the future. The regularities hold normally (but not always) and this is sufficient for (A) to be true (i.e. for the relevant conditional probabilities to be in place). To see this, consider (A) again: if you have reliabilist knowledge of p, then you are more likely to have more true belief (and knowledge) in the future

Reliabilism as Explicating Knowledge: A Sketch of an Account

195

than if you have a mere true belief that p. Consider the following train of events: 1. 2. 3. 4. 5.

S acquires the true belief that p through method M as a solution to problem P (assumption) S is confronted with a problem of the same type as P in the future (non-uniqueness) S still has access to method M (cross-temporal access) S makes use of M again (learning) S acquires a new true belief q through method M

Given the assumption of generality, this fortunate sequence of events is more likely, if M is reliable, than it is if M is unreliable due to the fact that the step from 4 to 5 is more likely in that case. The other steps are likely, and equally so, for reliable and unreliable methods. In other words, under the condition stated, the probability that S will acquire new true beliefs is greater conditional on reliabilist knowledge than it is conditional on mere true belief. For a recent exchange regarding the learning condition, see Jäger (2011a, 2011b), Olsson and Jönsson (2011). For replies to other criticisms, see Olsson (2009) and Olsson (2011a). Carnap’s account of explication requires of a fruitful empirical concept that it be useful for the formulation of many universal empirical laws. There are at least three empirical laws that involve reliabilist knowledge, namely those just mentioned. I am not aware that the competing accounts of knowledge, JTB etc., allow for the formulation of (true) empirical laws. I conjecture, therefore, that the reliabilist conception of knowledge satisfies Carnap’s desideratum of fruitfulness to a uniquely high degree. 6. Similarity to the ordinary concept “Reliabilist knowledge must be similar the ordinary concept of knowledge in such a way that, in most cases in which the ordinary concept has so far been used, reliabilist knowledge can be used; however, close similarity is not required, and considerable differences are permitted.” In the literature, there are two main challenges to the claim that reliabilist knowledge is sufficiently similar to the ordinary concept of knowledge: the generality problem and the Gettier problem. To set the stage for the generality problem, as I take it to be most fruitfully and compellingly understood, consider the following train of thought:

196

1. 2. 3. 4. 5.

Erik J. Olsson

Each concrete (token) process can be classified as belonging to a great many different types. It is not obvious how to single out one of these types rather than another as the unique associated type of the process in question. Furthermore, depending on what type is singled out as special, we may get different verdicts as regards the reliability of the process. Hence, there is no fact of the matter whether a given process is reliable or not. Hence, there is no fact of the matter whether a given person has reliabilist knowledge and not.

As for the third premise, a concrete process may be categorized both as “reading The Economist” and as “reading” (regardless of what is being read). In the first case is it probably reliable, in the second case probably not. Under what conditions would the generality problem be a threat to the claim that reliabilist knowledge is reasonably similar to the ordinary concept of knowledge? It would be if it would show that it is not true that, in most cases in which the ordinary concept of knowledge has so far been used, reliabilist knowledge can also be used. This is indeed what Earl Conee and Richard Feldman have argued as I reconstruct them (Conee and Feldman 1998, Feldman and Conee 2002). Here is their argument in a nutshell: 1. 2. 3. 4. 5.

People will agree on process typing and reliability only if a particular process type is salient in the conversational context A particular process type is salient in the conversational context only if it has been explicitly mentioned In many cases in which no type has been explicitly mentioned, we still agree on the corresponding knowledge claims Hence, in many cases people will agree on the corresponding knowledge claims without agreeing on process typing and reliability Hence: it is (arguably) not true that, in most cases in which the ordinary concept of knowledge has so far been used, reliabilist knowledge can be used.

It would follow that reliabilist knowledge is not sufficiently similar to the ordinary concept of knowledge and that it therefore fails to satisfy Carnap’s first criterion on an explication of knowledge.

Reliabilism as Explicating Knowledge: A Sketch of an Account

197

Unfortunately, Conee and Feldman’s argument is empirically unsound. The second premise is false: it is not true that a particular process type is salient in the conversational context only if it has been explicitly mentioned. I have questioned that premise on theoretical grounds drawing on the influential basic level tradition in cognitive psychology (Olsson 2011b, Olsson forthcoming). According to that tradition, our conceptual taxomomies are largely culturally inherited and for many taxomonies there is a salient “basic level”, i.e. a level which is more naturally, and more often, used than other levels. For instance, it is more natural to classify the thing on which I am now sitting as a “chair” than as, say, a “wooden structure”, “piece of furniture”, or a “designer chair”, even though these descriptions can all be truthfully applied to the thing in question. Basic level categories stand out as being highly informative, yet economical. From membership in the chair category, many things can be (defeasibly) inferred: that you can sit on it, that it has four legs, and so on. You cannot infer those things if you classify the thing as a “piece of furniture”. You can do it if you classify it as a designer chair. However, that would be a more complex, and therefore cognitively less economical, classification. Note that salience here has nothing to do with what has been mentioned or not in a conversational context. Similarly, I have proposed, categories like “seeing”, “hearing”, and so on are naturally viewed as salient basic level categories for belief forming processes. The claim that there are basic level categories for belief formation processes was recently empirically confirmed in an experimental study (Jönsson forthcoming a, forthcoming b). The study found, among other things, that people tend to agree on how to classify belief forming processes even when no classificatory type has been explicitly mentioned. Let us move on to the Gettier problem, which can also be seen, or reconstructed, as an objection to the claim that the reliabilist concept of knowledge satisfies Carnap’s criterion of similarity to the ordinary concept. Under what conditions would the Gettier problem be a threat to the claim that reliabilist knowledge satisfies this desideratum? It would be if it would show that it is not true that, in most cases in which the ordinary concept of knowledge has so far been used, the reliabilist concept of knowledge can be used. But clearly the Gettier problem does not show this: Gettier cases are too rare to threaten the claim that reliabilist knowledge can be substituted for ordinary knowledge in most cases. They are too rare for that purpose because they involve the consecutive occurrence of two improbable events: a proposition (Brown owns a Ford) that is strongly supported by

198

Erik J. Olsson

evidence turns out nonetheless to be false and yet by sheer luck (Brown is in Barcelona) the target proposition (Brown owns a Ford or is in Barcelona) comes out true anyway. Hence, the Gettier problem can never constitute a fatal challenge for a reliabilist explication of knowledge. A critic might still want to entertain the thought that reliabilist knowledge cannot replace the ordinary concept in Gettier cases because (i) according to reliabilism the person in a Gettier scenario knows whereas (ii) according to the folk concept of knowledge, that person doesn’t know. But both (i) and (ii) can be contested. As for (i), reliabilism by itself doesn’t predict anything about Gettier scenarios; it gives a definite prediction only against the background of a specific categorization of the belief forming process involved. And who knows how the “folk” will categorize the belief forming processes in Gettier cases? For instance, will the folk think of Henry’s belief that there is a barn over there, referring to Goldman’s barn façade case, as having been acquired through a process of “seeing” or through, say, a process of “seeing in deceptive circumstances”? We don’t know. There is room for empirical work here. As for (ii), experimental work has undermined the supposed stability of intuitions about knowledge in Gettier cases (Weinberg et al 2001, Swain et al 2008). Swain et al (2008) report that the barn case elicited among a majority of subjects the “wrong” response: that the person before the barn (“Susan” in their example) knows that she is looking at a barn. Gettier cases are the whales of epistemology: just as whales are like fish in some respects and like mammals in others, so too Gettier cases are like knowledge in some respects and like ignorance in others. They are like knowledge e.g. because they involve a (usually) reliable method of belief formation. They are like ignorance e.g. because they involve elements of randomness and error. So, are Gettier cases more like knowledge or more like ignorance? Consider the corresponding zoological question: are whales more like fish or more like mammals? Zoologists didn’t attempt to answer this question by consulting their zoological intuitions. Rather they considered the theoretical costs and benefits of various classificatory options. I suggest that it is time that we do the same in epistemology. 7. Conclusion My starting point was the methodology of explication which has strong independent standing in analytic philosophy deriving from its association

Reliabilism as Explicating Knowledge: A Sketch of an Account

199

with two of its founding fathers: Rudolf Carnap and W. V. O. Quine. It turned out that (simple) reliabilism looks very attractive indeed from that methodological perspective. The reliabilist explication of knowledge is strikingly simple, exact and fruitful. Furthermore, here is no compelling reason to believe that the reliabilist concept of knowledge falls short of being sufficiently similar to the ordinary concept. Even longstanding issues like the generality problem, in its most compelling form, and the Gettier problem were unable to undermine that conclusion. Future work includes the reconsideration, from an explicationist perspective, of some further objections to reliabilism, including the problem of easy knowledge (Cohen 2002) and possibly Bonjour’s clairvoyance cases (Bonjour 1980), although the latter target the reliabilist theory of justification rather than the reliabilist theory of knowledge. The present article has focused on JTB as the main competitor to reliabilism. We need to look more closely at other competing accounts, such as evidentialism and virtue epistemology, and how their fare as explications of knowledge before any more definite conclusions can be drawn. References Armstrong, D.M., 1973: Belief, Truth and Knowledge, Cambridge: Cambridge University Press Bonjour, L., 1980: Externalist Theories of Empirical Knowledge, Midwest Studies in Philosophy 5, 53-74 Carnap, R., 1950: Logical Foundations of Probability, Chicago: Chicago University Press Cohen, S., 2002: Basic Knowledge and the Problem of Easy Knowledge, Philosophy and Phenomenological Research 65, 309-329 Conee, E., and R. Feldman, 1998: The Generality Problem for Reliabilism, Philosophical Studies 89, 1-29 Dretske, F., 1969: Seeing and Knowing, Chicago: Chicago University Press Dretske, F., 1971: Conclusive Reasons, Australasian Journal of Philosophy 49, 1-22 Feldman, R., 1985: Reliability and Justification, The Monist 68, 159-174 Feldman, R., and E. Conee, 2002: Typing Problems, Philosophy and Phenomenological Research 65, 98-105 Goldman, A.I., 1979: What is Justified Belief? In: Justification and Knowledge, ed. G. Pappas, Dordrecht: D. Reidel, 1-23

200

Erik J. Olsson

Goldman, A.I., 1986: Epistemology and Cognition, Cambridge Mass.: Harvard University Press Goldman, A.I., 2011: Reliabilism, The Stanford Encyclopedia of Philosophy (Summer 2011 Edition), ed. E.N. Zalta, URL = Goldman, A.I., and E.J. Olsson, 2009: Reliabilism and the Value of Knowledge, in: Epistemic Value, ed. A. Haddock, A. Millar and D.H. Pritchard, Oxford: Oxford University Press, 19-41 Jäger, C., 2011a: Process Reliabilism and the Value Problem, Theoria 77, 201-213 Jäger, C., 2011b: Reliability and Future True Belief: Reply to Olsson and Jönsson, Theoria 77, 223-237 Jönsson, M.L., forthcoming a: A Reliabilism Built on Basic Levels Jönsson, M.L., forthcoming b: Linguistic Convergence on Abstract Verbs: A Challenge to the Standard Explanation of Basic Levels Kvanvig, J.L., 2003: The Value of Knowledge and the Pursuit of Understanding, Cambridge: Cambridge University Press Maher, P., 2007: Explication Defended, Studia Logica 86, 331-341 Olsson, E.J., 2007: Reliabilism, Stability, and the Value of Knowledge, American Philosophical Quarterly 44, 343-355 Olsson, E.J., 2008: Knowledge, Truth, and Bullshit: Reflections on Frankfurt, Midwest Studies in Philosophy 32, 94-110 Olsson, E.J., 2009: In Defense of the Conditional Probability Solution to the Swamping Problem, Grazer Philosophische Studien 79, 93-114 Olsson, E.J., 2011a: Reply to Kvanvig on the Swamping Problem, Social Epistemology 25, 173-182 Olsson, E.J., 2011b: The Generality Problem Naturalized, in: Festschrift for Erik Carlson. Uppsala Philosophical Studies, in press Olsson, E.J., forthcoming: A Naturalistic Approach to the Generality Problem, in: Goldman and His Critics, ed. H. Kornblith and B. Maclaughlin, Blackwell Olsson, E. J., and M.L. Jönsson, 2011: Kinds of Learning and the Likelihood of Future True Beliefs: Reply to Jäger on Reliabilism and the Value Problem, Theoria 77, 214-222 Swain, S., J. Alexander, and J. Weinberg, 2008: The Instability of Philosophical Intuitions: Running Hot and Cold on Truetemp, Philosophy and Phenomenological Research 126, 138-155 Ramsey, F.P., 1931, The Foundations of Mathematics and other Logical Essays, ed. R.B. Braithwaite, London: Routledge and Kegan Paul

Reliabilism as Explicating Knowledge: A Sketch of an Account

201

Rosch, E., C.B. Mervis, W.D. Gray, D.M. Johnson, and P. Boyes-Braem, 1976: Basic Objects in Natural Categories, Cognitive Psychology 8, 382-439 Weinberg, J.M., S. Nichols, and S. Stich, 2001: Normativity and Epistemic Intuitions, Philosophical Topics 29, 429-460 Zagzebski, L., 1996: Virtues of the Mind: An Inquiry into the Nature of Virtue and the Ethical Foundations of Knowledge, Cambridge: Cambridge University Press

How to Take Truth as a Goal? MARIAN DAVID, University of Notre Dame

Of the intellect which is contemplative, not practical nor productive, the good and the bad state are truth and falsity… Aristotle, Nicomachean Ethics, vi.2 As to the preference which most people – as long as they are not annoyed by instances – feel in favor of true propositions, this must be based, apparently, upon an ultimate ethical proposition: ‘It is good to believe true propositions, and bad to believe false ones’. This proposition, it is to be hoped, is true; but if it is not, there is no reason to think that we do ill in believing it. Bertrand Russell (1904, 76).

I The idea that truth is a positive value, a good, is quite prominent in philosophy, also in science (or at least in popular science), and in religious thinking. In these fields, truth tends to be referred to as a great good, a worthy goal, as something desirable to strive after, and talk of “the pursuit of truth” or “the search for truth” as eminently worthwhile is all but commonplace, especially in philosophy. To my knowledge, Nietzsche was the only major philosopher who dared to deprecate, even ridicule, “the will to truth”, if only at times. With this exception, paying homage to truth is pretty much the norm in philosophy – though maybe not always easily distinguished from paying lip service. When one talks about truth as a value, what is meant is not actually truth per se; what is meant, rather, is possessing truth. There are at least two important ways of possessing truth that are often not clearly separated: (a) having true beliefs, and (b) having knowledge. They are distinct, but not entirely: the second entails the first, but not the other way round. A perceived difference in value between the two prompted the question: “Why is knowledge better than mere true belief?” First raised in Plato’s Meno, the question is under discussion again in recent and very recent epistemology; it is part of the so-called “value problem”. However, here I will focus just

204

Marian David

on (a), which is typically referred to simply as the truth-goal, even though it would be more accurate to call it the true-belief goal. If true belief is a positive value, then it is good, or a good, at least prima facie, to have true beliefs. This leads naturally to the view that it ought to be a goal, ought to be one of our goals, to have true beliefs, that we ought to want to have true beliefs. This normative thought, in turn, leads one naturally to look more closely at the truth-goal (the true-belief goal) itself, bracketing normative concerns, if only temporarily: What is it, anyway, to have having true beliefs as a goal? What is it to want to have true beliefs? Though I will take occasional glances at the idea that having true beliefs is good, in this paper, I will be mostly concerned with the non-normative (preliminary) aspect of our topic, with having true beliefs as a goal, with desiring or wanting to have true beliefs.1 II I assume that belief (believing) is an attitude towards a proposition which encapsulates the content of the belief. To say that a belief is true is best understood as saying that what is believed is true, and the what is believed is a proposition. In this framework, talk of true (false) belief comes out as talk of believing true (false) propositions; and formulations pertaining to the truth-goal in sense (a) come out as: It is a goal, one of my goals, to believe true propositions; I want to believe true propositions. But the truthgoal is supposed to have two parts, a positive part and a negative part. For starters, the contents of the two parts might be put like this – using ‘p’ as an objectual variable ranging over propositions: positive to believe p, if p is true Tp  Bp

1

negative and

not to believe p, if p is not true ~Tp  ~Bp

Probably, ‘goal’, ‘desire’, and ‘want’-talk should be distinguished more than they usually are in this context; I will nevertheless follow widespread custom and be a bit cavalier about the distinctions.

How to Take Truth as a Goal?

205

Below the (almost) ordinary language formulations, we find logical formulas, widely used as abbreviations. The formulas have limitations. The one intended to represent the content of the positive goal does not make clear that the wanting is – in some sense not easy to make precise – directed at the consequent, rather than the antecedent, of the conditional; the wanting points at the believing rather than the truth of the proposition: I want to believe p, if p is true; I want to be such that (I believe p, if p is true). These are alternative acceptable formulations of what is intended. If, on the other hand, we read the formula with the wanting pointing at the formula’s antecedent instead, we get unintended results: *I want p to be true, only if I believe p; *I want to be such that (p is true, only if I believe p). This is not what is intended. In fact, these formulations seem a bit crazy. They suggests that, in case the world goes beyond my beliefs, I expect it to change, to back down as it were, rather than being prepared to change my beliefs. This is the wrong direction of fit. From the examples one might surmise that the ‘want’ always points at that part of the ordinary language conditional that is closer to it; however, that can’t be quite right, because: I want to be such that (if p is true, then I believe p) is an acceptable rendering of the content of the positive goal, with the ‘want’ properly pointed at the believing, even though the ‘believe’ is more distant than the ‘true’. Similar considerations apply to the formula intended to represent the content of the negative goal. Here, too, the wanting should be taken to point at the consequent of the conditional formula: ‘I want to not believe p, if p isn’t true’; not: ‘I want p to be not true, only if I don’t believe it’. The two parts of the goal are often combined into one with the formula: (Bp  Tp). This looks acceptable, provided it is read with the wanting pointing firmly at the believing: ‘I want to believe p, if p is true, and only if p is true’. Note that this requires reading the right-to-left direction of the formula, i.e. (Tp  Bp), as before, with the wanting pointed at the consequent (wanting to believe p, if p is true), while reading the left-to-right

206

Marian David

direction, i.e. (Bp  Tp), with the wanting pointed at the antecedent (wanting to believe p, only if p is true). Concerning the latter, note also that ‘goal’ and ‘want’-talk give rise to opaque (intensional) contexts: even though (~Tp  ~Bp) is logically equivalent to (Bp  Tp), it does not follow that the goal (~Tp  ~Bp) is the same as the goal (Bp  Tp). But this issue is a bit confusing because the logical formulas can be read in different ways. As I have in effect suggested above, if the goal (Bp  Tp) is read as wanting to believe p, only if p is true, there doesn’t seem to be any important difference between this and the explicitly negative formulation: wanting to not believe p, if p is not true, i.e. (~Tp  ~Bp); both will serve as abbreviations of the negative part of the goal. But if the goal (Bp  Tp) is read along the lines of wanting to be such that, if one believes p, then p is true, there is an important difference between this and the explicitly negative formulation, even though the formulas are logical equivalents. As Christian Piller (2009, 198) points out, wanting to be such that, if one believes p, then p is true, wanting one’s beliefs to be true, exhibits more of a concern for oneself – wanting to be right – than a genuine concern with the truth. A weather forecaster, having predicted bad weather, may want her prediction to be true, she may want to be right; but this is not a concern for the truth, it is more like wanting to win. Someone who wants her beliefs to be true, who wants to be right, would, in case of mismatch, prefer that the world conform to her beliefs, rather than make her beliefs conform to the world. Again, this is the wrong direction of fit (see also David 2005, 296). So, on an apparently natural reading of (Bp  Tp), it specifies a goal importantly different from the goal of not believing p, if p is not true, in spite of the logical equivalence with the formula (~Tp  ~Bp). III Let me focus now on the positive part of the goal. The formulations used above are deficient; they don’t pay sufficient attention to quantifiers. Once we think about quantifiers, we notice that there are three different forms associated with the positive truth-goal, distinguished by the scope of the goal-operator relative to the quantifier ‘p’, i.e. ‘for all propositions p’: 1. 2.

Goal: (p) (Tp  Bp) (p) Goal: (Tp  Bp)

How to Take Truth as a Goal?

3.

207

(p) (Tp  Goal: Bp)

Note that only item 1 can be taken as specifying a possible goal, the goal of believing all true propositions, or the goal of being such that one believes all true propositions. 2 and 3 can’t be taken as specifications of a goal at all, they make sense only when taken as statements – absurdly false statements, I should say. But let us look at 1 first. I said that 1 can be taken as specifying a possible goal. This is not to say that it specifies a plausible goal. It is frequently pointed out that there are simply way too many, too trivial and uninteresting truths for 1 to be a plausible goal: no one in their right mind would want to believe all the truths available from the local phone book, and all the truths available by counting the flowers on the wallpaper of one’s hotel room, and similar trivialities. Consider especially the mindboggling redundancy of trivialities involved: every pair of boring truths of the ‘At least n flowers’-variety gives rise to an equally boring conjunctive truth and an equally boring disjunctive truth; and there is much more “garbage” of this sort. Still, 1 specifies a (in principle) possible goal. Use the goal thus specified to form a statement ascribing it to some subject S, viz.: ‘S has the goal of believing all true propositions’. The complaint is not that statements of this form cannot possibly be true. The complaint is, rather, that they will in fact be false for almost all S; that most subjects won’t have this goal and shouldn’t be expected to have it either; that most of the few who may profess to have this goal will, on reflection, come to realize fairly quickly that they don’t really have it; and that the remaining few, if any, are not quite of sound mind. Maybe this is a bit too quick. Yes, we don’t have desires for believing trivial truths and don’t have desires for believing large sets of trivial truths (the phone-book truths, the wallpaper truths). It doesn’t follow, though, that we don’t have some desire for believing the whole truth. There is some allure to the idea of believing all the truths – the air of insanity flows, not so much from the goal itself, but from the all too obvious unrealizability of the goal: even if we had an infallible method for distinguishing truths from falsehoods, we couldn’t possibly achieve this goal. Still, it does seem alluring somehow. In any case, remember that we are talking about the goal of having true beliefs in part because we are interested in the idea of true belief being a value, a good. And remember that many people ascribe to God attributes entailing that God believes all truths, apparently including the most stultifyingly trivial and redundant ones. This is evidence that many

208

Marian David

regard believing all truths as a good, as an in principle valuable state to be in, even though they may not be much inclined to adopt it as a goal for themselves. This being said, it does seem that the goal of believing all true propositions is a bit extravagant. Looking for a more plausible goal, one will propose restricting the quantifier in 1 to propositions that are interesting. Most likely, this means that a subjective element is being imported into the goal. At least, it is difficult to see how being interesting could be an entirely objective property of propositions: what is interesting seems to be relative to, well, our interests. It also means giving up on any idea of a “pure” truth goal. This is not an objection; just a reminder that there is now a second value in play: what is (we find) interesting. Alternatively, one might propose restricting the quantifier in 1 to propositions one has considered; more precisely, to propositions p with respect to which one has asked oneself whether p. By my lights, this is too restrictive. Insofar as I have the goal to have true beliefs, I also want to find truths beyond my horizon, truths concerning questions I have not considered. Aristotle at one point mentions “indefinite” statements: statements that lack an explicit quantifier (cf. De Int., 17a38). I surmise that an indefinite formulation of the truth-goal would be the one people would be most likely to assent to when queried; something like: “I want to believe true propositions and not believe false ones”. Putting in an explicit quantifier tends to end up sounding too bold (universal quantifier) or too meek (existential quantifier). Recently, there has been some interest in generics, i.e. claims like: Tigers have stripes (cf. Leslie 2012). Maybe a truth-goal plausibly attributable to people, a psychologically realistic truth-goal, is one with a generic content: believing true propositions. Let us consider more briefly item 2 from above. I pointed out that it does not specify a goal; instead, it makes a statement: it says that, with respect to each and every proposition, a subject S has the conditional goal to believe it, if it is true. This cannot be right; there are way too many propositions for this to be humanly possible, non-denumerably many; and even if we restrict the quantifier to interesting propositions, there are still way too many. We cannot have that many goals, that many wants. Of course, there may be many correct instances of 2. For many propositions p, a subject S may have the following conditional goal/want: to believe p, if it is true. Consider 3. It too makes a statement, and again a false one. It says that, with respect to each and every true proposition, S has the non-conditional

How to Take Truth as a Goal?

209

goal to believe it. This assumes, implausibly, that S has somehow managed to single out the true propositions; in any case, there are still way too many true propositions (and way too many interesting true propositions) for this to be humanly possible.2 Item 3 has a special feature that makes even its instances problematic, or at least unusual. Each instance ascribes the non-conditional goal, with respect to a particular proposition, to believe it; each instance talks about wanting to believe a certain proposition. Now, wanting to believe a particular proposition is rather extraordinary. Not impossible – we can think of cases. Having witnessed, say, great suffering, a believer may have lost his faith in God’s goodness; realizing this, he may very much want to believe this particular proposition, that God is good, but may find it hard going (at least initially). This sort of thing is possible, but it is a rare and rather special condition to be in. Believing is not under our voluntary control, at least not in any direct manner. We are to some extent aware of this, if only darkly, which goes some way towards explaining why we don’t typically find ourselves wanting to believe particular propositions – except for extraordinary circumstances. Cynics may add the suspicion that our subconscious often enough manages to circumvent this problem; that below the threshold of conscious control, we often want to believe particular propositions (ones that make us feel good about ourselves) and proceed to believe them because “we” want to.3 With respect to both, 2 and 3, one may observe that, while they are plainly false, attributing absurdly impossible numbers of goals, the results of replacing ‘goal’ with ‘good’ are not equally impossible. It is not impossible that every proposition is such that it is good to believe it, if it is true; nor is it impossible that every true proposition is such that it is good to

2

Concerning the relations between 1, 2, and 3. Use 1 to form a statement ascribing the specified goal to S. We then have: 1. S has the goal to believe every true proposition. 2. For every proposition, S has the goal to believe it, if it is true. 3. For every proposition, if it is true, S has the goal to believe it. There are no entailments here: 1 doesn’t entail 2 or 3; nor does 2 entail 3; and 3 doesn’t entail 2 or 1, nor does 2 entail 1. 3 ““I did that,” says my memory. “I could not have done that,” says my pride, and remains inexorable. Eventually – the memory yields” (Nietzsche 1886, §68).

210

Marian David

believe it. Indeed, Michael Lynch (2004) holds that all these things are good, at least prima facie good. IV I now want to consider a puzzle that arises with respect to the positive part of the true-belief goal. To be honest, I am not entirely sure that there really is a puzzle – some are trying to tell me that there isn’t one. Still, it seems to me that there is something puzzling, and I want to bring it up. Consider first what appears to be a fairly plausible transition principle for conditional desires (wants), proposed by Piller (2009, 196). Here is how Piller puts it: “Wanting if A then B and noticing that A commits me to wanting B or, at least, gives me a reason to want B”. Note that the conditional desire is presented in the somewhat artificial form ‘wanting if A then B’; the more natural form is “wanting B, if A’. Note also that ‘noticing that A’ involves believing that A; indeed, nothing more demanding, epistemically, is required here than belief. Now consider the following scenario, for comparison. Imagine I have recently developed a nostalgic longing for the music of the 70s glamorrock band T. Rex. I want to buy their CDs (and only their CDs, but I’ll suppress this part for simplicity). With this longing, I enter a music store: In the music store: I want to buy all CDs by T. Rex

I want: (x)(Tx  Bx), quantifier restricted to CDs

I go through the CDs in the bargain bin, for each one: I want to buy it, if it is by T. Rex

I want: (Tx  Bx), for each of the x’s

I find one by T. Rex: I want to buy it.

I want: Bx.

This progression involves analogues of items 1, 2, and 3 discussed earlier. The transition from 2 to 3 goes by Piller’s principle. The earlier transition from 1 to 2 goes by a closely related principle. Considering a CD from the bin, the global goal turns into a conditional goal with respect to the CD under consideration; and when I come to believe of such a CD that it is by T. Rex, the conditional goal turns into a non-conditional goal with respect

How to Take Truth as a Goal?

211

to that particular CD: I want to buy it (unless it is too expensive, or broken, etc.). So far so good. Now, imagine I have the goal of believing true propositions, and imagine that I am presented with a list of propositions (or if you want, a list of sentences expressing propositions): A list of propositions: I want to believe all the ones that are true

I want: (p)(Tp  Bp), quantifier restricted to the list

I go through the propositions on the list, for each one: I want to believe it, if it is true

I want: (Tp  Bp), for each of the p’s

I notice one that is true: I want to believe it

?

I want: Bp

?

Unlike the first scenario, where there is nothing at all remarkable, the last row of this second scenario strikes me as odd. Considering an item on the list, my global goal (form 1 from above) is converted into the conditional goal, with respect to the proposition under consideration, to believe it, if it is true (an instance of 2). Now I notice of the proposition under consideration that it is true. What happens now? Remember, noticing that p is true amounts to believing that p is true: Given that I already believe that p is true, does it make sense to say that I want to believe p, or that I am committed to wanting to believe p, or that my believing p gives me a reason to want to believe p? I am not among those who subscribe to the strong deflationary view about truth, that believing that p is true = believing p. Still, the following does hold: if one believes that p is true, one believes p. How does believing p go together with wanting to believe p, or being committed to wanting to believe p, or having a reason to want to believe p? Admittedly, my setup is somewhat artificial. However, we certainly can do what I described. We can go through a list with the global goal in mind to believe the items that are true (the common activity of reading nonfiction literature is a bit like that). But when we do this, what happens is not at all like what happens when we go through some CDs with the global goal in mind to buy the ones by T. Rex. Why is that?

212

Marian David

Coming to believe that a CD has the “search property” being by T. Rex is quite different from buying it, so there is something additional left for me to do which I can want to do. But coming to believe that a proposition has the search property being true entails believing it, so there is nothing further for me to do and, it seems, nothing for me to want to do. The desire for music by T. Rex is action guiding; the desire for true belief is not. Ernest Sosa (2001, 50) remarks briefly that the desire for safety cannot guide how we form beliefs. Sosa’s safety is of the form (Bp  Tp), which is the positive-formula representation of the negative part of the truth-goal (discussed above in Section II). If this is read properly, as wanting to believe p, only if p is true, i.e. wanting not to believe p, if p is not true, his remark concerns a closely related point, namely the absence of guidance for non-belief: believing that p is not true, already entails not believing p, so once one believes that p is not true, there is nothing further left to do or to want, as far as one’s desire for safety is concerned.4 One feature that makes the second scenario odd is that it is a case of wanting to believe a particular proposition. I noted earlier (with respect to instances of 3) that this is strange or unusual at best, probably because believing is not under our direct voluntary control. However, I think this is not really the source of what is strange about the second scenario. One can remove this feature using a science-fiction example. Imagine I acquire a program that gives my computer access to my brain, specifically to the belief-forming module of my brain, say, via a thick cable plugged into my forehead. The program displays a list of propositions on my computer screen. By clicking the little “button” next to a proposition, I can cause the program to cause me to believe that proposition. In this scenario, there is no in-principle problem with going through the list wanting, for example, to believe particular propositions that will make me feel good about myself. Say I encounter the proposition that I am a better than average driver/lover. Will believing this make me feel good about myself? Yes, at least for a while. I want to believe it. Click. I believe it. In this scenario, that source of oddness has been removed: here it makes sense for me to want to believe a given proposition under consideration, 4

If safety is read (improperly, by my lights) along the lines of wanting to be such that, if one believes p, then p is true, talk of belief or action guidance is of course entirely out of the question: p’s being true is not a belief or an action.

How to Take Truth as a Goal?

213

and my beliefs are (very close to being) under my direct voluntary control: I can turn on the light by flipping a switch; I can believe that I am a better than average driver/lover by clicking a button. In this science-fiction scenario, coming to believe, which normally isn’t much like an action, is much like an action, at least like an action of the ‘turning on the light by flipping a switch’-sort. In this scenario – and this is the feature I want to emphasize – my desires can guide (almost) directly how I form beliefs: I can believe because I want to believe (without subconscious subterfuge). Nevertheless, the case I am wondering about remains odd. I go through the list of propositions displayed on the screen wanting to believe the propositions that are true. I encounter one that I believe is true. What would be the point clicking the button? And even if I do (as I probably would), nothing happens. G. E. Moore once observed “that I cannot at any given moment distinguish what is true from what I believe” (1903, §80). Here is a final thought experiment by way of illustration. I present you with a list of propositions and ask you: Mark the ones that are true! You comply. Imagine now that, concerning the very same list of propositions, I had asked you: Mark the ones that you believe! You would have marked the very same propositions (which, of course, does not even begin to indicate that being true = being believed). I find Moore’s observation highly suggestive; it strikes me as clearly related to the feature of the true-belief goal that I find puzzling; they are two sides of the same coin. For the moment, though, I must leave it at that. I have not yet found anything deeper to say on this topic. References David, M., 2005: On ‘Truth is Good’, Philosophical Books 46, 292-301 Leslie, S.J., 2012: Generics, The Routledge Companion to the Philosophy of Language, ed. D.G. Fara and G. Russell, New York: Routledge Lynch, M., 2004: True to Life: Why Truth Matters, Cambridge, Mass: The MIT Press Moore, G.E., 1903: Principia Ethica, Cambridge: Cambridge University Press Moore, G.E., 1944: Moore’s Paradox, in: G. E. Moore: Selected Writings, ed. T. Baldwin, London and New York: Routledge, 1993, 207-212 Nietzsche, F., 1886: Beyond Good and Evil; tr. H. Zimmern, Tribeca Books 2012 Piller, C., 2009: Desiring the Truth and Nothing But the Truth, Noûs 43, 193-213

214

Marian David

Russell, B., 1904: Meinong’s Theory of Complexes and Assumptions, reprinted in: B. Russell, Essays in Analysis, ed. D. Lackey, London: Allen and Unwin 1973, 21-76 Sosa, E., 2001: For the Love of Truth? In: Virtue Epistemology, ed. A. Fairweather and L. Zagzebski, Oxford: Oxford University Press, 49-62

Knowledge as Achievement – Greco’s Double Mistake CHRISTIAN PILLER, University of York

Abstract John Greco claims that knowledge is a kind of achievement. The value achievements have (as such) shows, according to Greco, why knowledge is better than mere true belief. I argue that, for a variety of reasons, it is not always good to know. Furthermore, it is wrong to think that achievements are always good – think of achieving what is bad. Greco is mistaken twice; this leaves the idea that knowledge is a kind of achievement intact. 1. Introduction In virtue epistemology, the traditional concern about the nature of knowledge goes hand in hand with another concern, the quest to show why knowledge is a good thing. These concerns are obviously connected. The value of anything will depend on its nature: whether what you carry around in your pocket is valuable depends on what it is that you carry around in your pocket. Contemporary virtue epistemology draws inspiration from parallels between the theoretical domain of belief and knowledge and the practical domain of intention and action. Successful action has become the model on which to understand success in our epistemic endeavours. John Greco makes this parallel central to his recent book Achieving Knowledge. ‘The central thesis of this book’, Greco says, ‘is that knowledge is a kind of success from ability’ (Greco 2010, 3). The lucky guess which happens to be true does not amount to knowledge. Although you succeed in what is, let us suppose, your aim, namely to get things right, that you get it right is not due to your cognitive abilities. The archer’s lucky shot, which hits the target because a sudden gust of wind diverts it from its given path, is, similarly, not a success due to ability. Successful actions count as achievements only if the agent’s success is due to the agent’s abilities. According to

216

Christian Piller

Greco, the notion of knowledge captures our achievements within the theoretical domain. Thereby epistemic achievements participate in a general category we are familiar with from the practical domain. The notions of value and normativity as they apply to beliefs and knowledge are illuminated by their participation in the general category of achievements. The idea that knowledge is a kind of achievement not only helps us with the nature question – Greco argues that it gives the right verdict on Gettier cases – it also issues a direct answer to the value question. I want my children to be happy but even more so do I value a state in which their happiness is to some extent due to my own contributions. I want my paper to be published but even more so do I value a state in which its publication is due (in the right way) to my own doing. In general, we value being active participants in the fortunes of our lives over being passive recipients of goods. If we achieve something through effort and due to our abilities we satisfy this need for active involvement. Greco expresses this idea as follows. ‘Knowledge is a kind of success from ability, and in general success from ability is both intrinsically valuable and constitutive of human flourishing, which is also intrinsically valuable. Moreover, both success from ability and human flourishing have ‘final’ value, or value as ends in themselves, independently of any instrumental value they might also have. Therefore, knowledge has value over and above the practical value of true belief’ (Greco 2010, 174f). Let me separate the claims Greco is making. First, we encounter the idea that epistemological notion are instances or analogues of notions from the practical domain. This claim is what is distinctive of virtue epistemology. It is also central to Ernest Sosa’s view. Sosa understands knowledge as apt belief, i.e. as a belief the truth of which is due to an agent’s competence. Sosa’s AAA terminology renders knowledge an instance of a general kind, namely of performances which are accurate because adroit and, thus, apt. His favourite example of an apt performance is the archer’s shot. Shots that hit the target are accurate; those that manifest an agent’s skill are adroit; shots are apt and are, thus, good shots, if they are accurate because they are adroit. Similarly, an agent knows when the fact that his belief is true is due to an agent’s intellectual skill. Animal knowledge, which contrasts with reflective knowledge, is apt belief. The Gettier Problem is a test for this view. Greco’s second claim is that understanding knowledge as success from ability issues the right verdict in Gettier case. Virtue epistemology, according to Greco, solves the Gettier Problem. The third of Greco’s claim

Knowledge as Achievement – Greco’s Double Mistake

217

is that it is a good thing to know. It is better to know something than to guess, even if guessing correctly. Finally, Greco’s fourth claim is that the nature of knowledge explains its value. Achievements are good. Knowledge is an achievement. Thus, it is a good thing to know. In this discussion I will start with Greco’s third claim: it is better to know than to have a merely true belief. Contrary to Greco I will argue that it’s good to know some things; other things, however, are better not known. And then there are things were it really does not matter whether we know them or not. Thus, I disagree with Greco about the target, i.e. to explain that knowledge is good or that it is better than mere true belief. He tried to explain the value of knowledge via the value of achievements. Achievements, I claim, are not always good either. Achieving the bad, for example, is not good. Greco, thus, makes a double mistake which seems to leaves the idea that the nature of knowledge can be illuminated via the idea of achievements on the table, or so I will argue. 2. Greco’s First Mistake Greco’s first mistake is to assume that the value of knowing is independent of what it is that is known. Contrary to Greco, I claim that sometimes it is good to know and sometimes it is not – it depends, amongst other things, on what it is that is known. Knowledge is important in the pursuit of important projects. However, not always is it good to know. (A) You can know too much. You might become an obstacle or a danger to others if you know what they do not want you to know. Some knowledge, like knowing when and how you are going to die, would often interfere with and hinder a pursuit of a normal life’s project, especially if you’d know that you are going to die prematurely. In such cases it doesn’t matter much whether you know or truly believe. These are cases in which a lack of any epistemic attitude is advantageous. Similarly, even if true, it might be better not to know some bad things. It would be very bad to know that no one loves you, bad that none of your colleagues respects you. In such cases a belief that p (namely that the bad thing is not true) is better than its opposite as well as better than withholding, even if it would amount to having a false belief. (B) Some things are simply not your business. For example, there are limits on what you are allowed to know about your neighbours. Keeping a diary about when they go to bed and when they get up does not serve any

218

Christian Piller

legitimate epistemic interest. Legitimate curiosity is limited and nosiness is a vice. (C) Think about the category of experiences described by knowing what it is like. Knowledge of what it is like to kill someone, knowledge of what it is like to have your fingernails pulled out – it has, for most of us, nothing to be said for it. I have a weird friend. When his cat died, he didn’t bury it in the garden. He keeps it in his shed in a sealed transparent plastic bag. Do you want to know what a dead cat looks like after 2 weeks or after 2 months? He gave me some pictures. Would you like to see them? Free knowledge! No? I thought so. (D) Think about all the matters others take interest in. My auntie is interested in what her neighbour reads and in where it was she has last seen her cousin. I understand her concern but I am myself not interested in any of these things. I might want to know when I last visited my dentist. But do you want to know? Should I tell you? It’s no effort for me and if it does you some good – I can easily tell you. I need not bother? It would not do you any good? Alright then. I need to know my own date of birth and that of my children. No one needs to know everyone’s date of birth. Dates of birth are important pieces of biographical information. Still you don’t want to know. I don’t know what happened to my old tennis shoes I threw away years ago and I don’t know what Kevin Costner is doing at the moment. I DON’T CARE. Now think about all the things no one has ever taken any interest in. The history of a leave on a tree provides a lifetime worth of study. It grows, it moves as the wind blows – first this way, then that way. A bee flies past at a certain distance. And even if nothing happens that’d be worth recording to, if truth in itself mattered. If God would whisper only the tiniest amount of all there is to know in our ears we would die of that roar. I conclude that we do not need to explain why knowledge is always good because, for most things, it is not good to know them. What would it mean if we said that knowledge is good? Do we mean it’s just plain and simply good or do we mean that it is good for the person who knows (and for others who might need the information he has)? The second option is less obscure and, for that reason, more plausible. Let’s leave the matters devoid of any interest aside – these matters don’t make it into Wikipedia – and concentrate on what is of interest to some people. Would you like to know everything that one can find in Wikipedia? What good would that do for you, the Wikipedia man? If you are around and someone wants to know the four districts of Tomsk and you say ‘Kirovsky,

Knowledge as Achievement – Greco’s Double Mistake

219

Leninsky, Oktyabrski, Sovetski’, you are weird. ‘Famous resident, the philosopher Gustav Shpet – translator of the Phenomenology of Spirit, executed 16/11 1937.’ One is not supposed to know these things (unless one is from Tomsk or related to Shpet). In which way would this be good for you? There is no good answer, I can think of. In my paper ‘Valuing Knowledge’ (Piller 2009), I have tried to make fun of philosophers who assert that knowledge is always a good thing. They are the majority and they say outrageous things. In between we find philosophers who do not put their common sense on the hanger when they arrive at the office. Jane Heal (1987/88) who compares the project of matching every fact with its corresponding belief with the project of providing every garden with a suitable garden gnome – a useless and absurd project – is one of them. If someone says something simply because it’s funny, we understand what the person saw in saying it (even if, overall, it was wildly inappropriate). If, in contrast, someone says something simply because it is true (or because he knows it), we do not yet understand why he said it. Assert what you know would, if implemented, be a sufficient reason not to leave one’s room anymore. What about believing? Is it not enough to believe something because it is true? Once you think it’s true, you believe it already, so its being true won’t explain (in this sense) why you believe it. How do we express a concern for truth then? We are careful in our enquiry; we choose methods which are likely to lead to the truth in respect to the issue which interests us. The care we take in such cases is, however, often explained by what explains our interest in truth. I need to be at the meeting in time. So I better double check its starting time. I need to know who did what in order to reward or punish those who deserve it. And so on. What about curiosity? Isn’t it to be concerned about truth for truth’s sake? Heal talks about a person who is writing the numbers of the cars parked on the street in his notebook. A police investigation? A traffic survey? Someone from the Office for Statistical Research? No. None of those. He is simply and plainly a truth-seeker! We are happy to acknowledge that what we find fascinating (philosophy, let’s say), others might just yawn at. They might subscribe to the Heraldry Gazette or study for the spelling championship. It’s a wide church but there’s a dogma: one must be capable of explaining why some subject matter is interesting – at least interesting to oneself, even if not to others. Sharing a human nature, we want to understand what gets you going and what

220

Christian Piller

you are up to. The truth-seeker without anything further to say is not admitted. The Wikipedia man doesn’t make sense to us; what is the project in which knowing everything could play a role? Even a much more limited project like following the being-a-partner-city relation from some given subset doesn’t make any sense unless there is, for example, a competition in which such knowledge might win a prize. Any interest in this relation would have to be motivated by concerns we could understand, e.g. a limited project of this kind might serve some sociological interest. If you just want to know which city partners which other without such explanation, you don’t make sense. 3. Greco’s Second Mistake I said that Greco makes a double mistake. His second mistake is to think that achievements are good. Despite Greco’s claim to the contrary, I cannot believe that there is anyone who would really think that every success from ability has something good in it. Think of the achievement of what is distasteful, brutal or, in some other way, horrible. Greco’s value claim is intended to contrast the lucky success with the success due to one’s abilities. However, the fact that the terrible result is the result of competent agency does not diminish its negative value; to the contrary, the intended brutal assault is worse than the accident which, otherwise, has the same disastrous results. Greco borrows the idea that knowledge is success from ability (or that knowledge is the manifestation of a competence) from Ernest Sosa. For Sosa knowledge is apt belief; apt belief is belief which is accurate because it is adroit. Though agreeing on the nature of knowledge, Sosa, like me, finds Greco’s answer to the value question implausible. Let me make my own use of Sosa’s archery example to illustrate this point. Think of an archer who practices his skills on the children’s playground. Not that he himself is on the children’s playground, no, the playground and its children are his target. He himself is 100 yards away. Despite this distance, his arrows, let us assume, nail five-year old Charlie to the climbing frame in exactly the way he intended. The detectives, who investigate this horrible act, will be able to reduce the list of suspects in virtue of the fact that hitting a child from such a distance, if intended, would be a good, actually a truly excellent shot: a good shot and a wicked and horrible act with nothing good in it.

Knowledge as Achievement – Greco’s Double Mistake

221

Greco, however, seems unimpressed. He insists, ‘... knowledge is a kind of success from ability, and in general success from ability is both intrinsically valuable and constitutive of human flourishing, which is also intrinsically valuable.’ (Greco 2010, 137) ‘I follow Aristotle’, he says ‘in holding that success from ability is constitutive of human flourishing, which has final value’ (Greco 2010, 180). Aristotle, however, was talking about the exercise of the virtues. Greco, surprisingly, seems to fail to realize that virtues are not the whole but only a subset of abilities. If one is able to use the faults and weaknesses of one’s children to deeply embarrass and humiliate them in front of their peers, such skill is, needless to say, not a virtue. Greco does not engage with this point. Duncan Pritchard does. Considering the achievement of something pointless or wicked, Pritchard writes, ‘Are even achievements of this sort of final value? Note, however, that the value of achievement thesis, properly construed, is only that achievements have final value qua achievements’ (2010, 45). Achievements, according to Greco and Pritchard, always contribute something positive to overall states of affairs. It is compatible with this idea that these overall states of affairs containing achievement are overall bad. Furthermore, the laws of how values combine can throw up all sorts of things. Putting some perfectly nice bright red ketchup on the yellow dress will spoil it. Similarly, if I hit the target, which happens to be the child on the playground, then the overall value is lower if I hit it ‘virtuously’, i.e. because of my skill, than had I hit it accidentally. Pritchard says, ‘It is important to recognize that the value of achievement thesis when properly understood is entirely consistent with this possibility’ (ibid). As long as we understand that achievements are only good as achievements, Pritchard suggests, we are free from trouble. The badness of horrible acts is compatible, he says, with the idea that achievements are good qua achievements. What does it mean to say of something that it is good as such and such? (i) The attributive use of ‘being good as such’. Suppose I tell a student, ‘There’s good and bad news. As a philosophy essay your paper is rather poor. However, as something the reading of which is a complete waste of time it is, actually, very good.’ ‘X is good as such and such’ often means that X is a good such and such. ‘You might think of Idi Amin whatever you

222

Christian Piller

want but as a poker player, he was very good.’ He was good as a poker player means he was a good poker player. We have met this idea before. Shooting little Charlie required an excellent shot and was a horrible act with noting good in it. Peter Geach has famously argued, ‘There is no such thing as being just good or bad, there is only being a good or bad so-and-so’ (Geach 1956, 34). In Geach’s preferred terms ‘good’ is a logically attributive adjective, whereas adjectives like ‘red’ are logically predicative. There are good dentists, good strawberries, good lawnmowers, but there is nothing – no property of goodness – in virtue of which they would all be good. Attributive goodness is not the same as contributory goodness, which is what Greco was after. A good shot is good as a shot but might have nothing that speaks in its favour and everything against. Sosa’s views are best seen as taking epistemology to deal in its own notion of attributive goodness. A justified and, in this sense, good belief need not have anything good about it – the subject matter might lack any legitimate interest. ‘Silly beliefs about trivial matters can attain the very highest level of justification and knowledge even if these are not beliefs that one should be bothering with, not even if one’s concerns are purely epistemic’ (Sosa, 2007, 66.) (ii) The contrastive and, thus, predicative use of ‘being good as such’. Achievements are good as achievements – does not, at first sight, say very much except that achievements are good. Compare ‘Friendship is good qua friendship’. This means, I guess, no more than that friendship is good. In both cases we use ‘good’ predicatively. Compare ‘Pleasure is good as pleasure’. As in the two previous cases, this says that pleasure is good. Why do we say these apparently trivial things? In identifying something as something we identify its evaluative dimension and we contrast this evaluative aspect with some of its other aspects. Being pleased, let us assume, is being in a certain physiological state. Under this assumption, the statement that being pleased is good qua being pleased tells us that being pleased rather than being in its corresponding physiological state is what is of evaluative importance. Let us apply this idea to achievements. Saying that achievements are good as achievements, is identifying the way in which they are good (predicatively). The contrast in this case is with what has been achieved. On this view, hitting the child skilfully has an aspect to it which is good. I find it hard to believe that anyone could, on reflection, really believe such a thing.

Knowledge as Achievement – Greco’s Double Mistake

223

4. Conclusion and Looking Ahead I conclude that Greco’s view is indefensible. Greco was right that we can think about the value of knowledge in terms of the value of achievements. He was wrong, however, to think that achievements are always good. It turns out that knowledge, like achievements, are not generally good either and these two points are related. It is not good to know when knowing would be imprudent, when it would be morally dubious, unpleasant or when, and this is the vastest category, it would simply be pointless. Similarly, succeeding, because of one’s abilities, in a stupid project, exercising one’s abilities in pursuit of the bad and the wicked, having creditable success in the absolutely pointless has nothing to be said for it. The fact that these criticisms of Greco’s view run parallel confirms the basic idea of the virtue epistemological approach. The practical might still be able to illuminate the theoretical. There are two ways in which this idea can be pursued further. I think it is not always good to know. Nevertheless, I accept a restricted value-of-knowledge problem. What we need to explain is why, when we are interested in a subject matter, and legitimately so, we want to know how things stand and not merely have a true belief about them? Why do we prefer knowing to believing truly in matters of interest? I have defended the idea (Piller 2009) that when we know we believe as we ought to believe and that, generally, when we do something to achieve an end we want to do what we do well because, thereby, we maximize our active involvement in what is going on. I see the conditions of theoretical as well as of practical rationality as determining what it is to believe and to act well. To satisfy these conditions, which I see as independent deontological requirements, is something we want. Thus, I agree with the virtue epistemological method of trying to illuminate the value and normativity of believing and knowing by regarding them as instances of general phenomena which are at home in the practical domain. Knowing that p is better than believing truly that p, when the question whether p or not-p is a legitimate concern, because in knowing we believe with justification and, thus, believe as we ought to believe. Thereby we satisfy our concern for active involvement because in doing things well we minimize our dependency on luck in the achievement of our aims. To know is a conditional good; it is conditional on the legitimacy of our interest in the question we are trying to answer.

224

Christian Piller

A second route to answering the value-of-knowledge question is taken by Sosa. Knowledge, for Sosa, is apt belief, i.e. a belief the truth of which is saliently explained by an agent’s competence. Like Greco, Sosa thinks that via a parallel to action this account tells us something about the nature of epistemic value and epistemic normativity. Unlike Greco, however, he accepts the point I emphasized in section 2, namely that knowledge need not always be good (or good for the person who has it). There is knowledge not worth having. There is no point in counting the blades of grass, even if doing it carefully would give us knowledge. How then should we understand epistemic normativity and value? Sosa’s answer seems to be: epistemic value has to be understood attributively. ‘ ... the supposed normativity of epistemology seems rather like the normativity of a good gun or a good shot. This normativity is restricted to the sphere of guns and shots in some way that isolates it from other important concerns, even from whether there should be guns at all, or shots. ... If ours is the right way to understand such normativity, then in speaking of a justified belief we are saying something rather like “Good shot!” which someone might sincerely and correctly say despite being opposed to gun possession and to shooting’ (Sosa 2007, 66). It need not have been a good thing to shoot even when the shot was excellent. Here is not the space to engage critically with Sosa’s theory. Nevertheless I want to end by raising a general concern about the idea of founding epistemic value and normativity on attributive goodness. The fact that we talk about the category of epistemic value suggests by itself that we are after a predicative notion of goodness. When we divide the good in values of various kinds, we usually think about the good in predicative terms. Liberty and equality of opportunity are political values; it is how the good comes to us in the field of politics. Honesty and loyalty are moral values which determine the good of social interaction. Creativity and sensitivity are personal values; harmony and beauty are how the good comes to us in experience and so on. Compare this to attributive goodness. There are good cars and there is good gardening; however, we do not talk about automotive or gardening values. If we call understanding and knowing ‘epistemic values’, this seems to suggest that in understanding something we realize the good in the domain of doxastic states. Thus, the way we talk about epistemic value raises the doubt whether a purely attributive notion can deliver an account that would fit our pre-theoretic demands and commitments that are in play when we rely on theories of epistemic justification.

Knowledge as Achievement – Greco’s Double Mistake

225

Calling someone unreasonable is not the same as finding out that someone fails to share one’s sensibility as a ‘belief-connoisseur’. Epistemologists mean to talk to a wider audience than those who signed up for this rather peculiar connoisseurship – they mean to talk to all of us. References Geach, P., 1956: Good and Evil, Analysis 17, 33-42 Greco, J., 2010: Achieving Knowledge, Cambridge: Cambridge University Press Heal, J., 1987/88: The Disinterested Search for Truth, Proceedings of the Aristotelian Society 88, 79-108 Piller, C., 2009: Valuing Knowledge: A Deontological Approach, Ethical Theory and Moral Practice 12, 413-428 Pritchard, D., 2010: Knowledge and Understanding, in: The Nature and Value of Knowledge: Three Investigations, ed. A. Haddock, A. Miller, and D. Pritchard, Oxford: Oxford University Press, 10-118 Sosa, E., 2007: A Virtue Epistemology. Oxford: Oxford University Press

IV. Testimony

Norms of Trust, De Re Trust, and the Epistemology of Testimony1 SANFORD GOLDBERG, Northwestern University

Abstract The thesis of this paper is to argue, first, that trust itself, as opposed to the reasons one has for trusting on a given occasion, has no distinctly epistemological payoff, and second, that despite this, epistemologists ought to be interested in the phenomenon anyway. In particular, (interpersonal) trust is of interest because it manifests the distinctly interpersonal nature of testimonial transactions – something that must be accommodated by any adequate epistemology of testimony. 1. As I will be discussing it in this paper, the sort of trust I have in mind is the sort that is exemplified in communicative (especially testimony-involving) exchanges. To get at this notion it will be helpful to contrast trust with another (non-trust-involving) way of reacting to testimony. Suppose that you hear a dozen different people tell you that p. You don’t have a particularly high opinion of the epistemic character of any one of the speakers, and you would not be inclined to accept any of these testimonies taken singly (the topic is one on which you are especially careful in taking the word of another). But at the same time you have good reasons to think that none of them conferred with the others, and that they each got their information 1

With thanks for helpful comments on this paper to members of audiences at the various places I have given this paper as a talk: the 34th Wittgenstein Symposium in Kirchberg, Austria (August 2011), the “Testimony and Trust” conference at Reading University, Reading, UK (July 2011), and the Northwestern Epistemology Brownbag, Northwestern University, Evanston, US (2009). Special thanks to Alvin Goldman, Peter Graham, Thomas Grundmann, Andrea Kruse, Jennifer Lackey, and Baron Reed for extensive comments on earlier versions. (Neither they nor audience members are responsible for any problems with the views expressed here.)

230

Sanford Goldberg

from different sources (some first-hand, some second-hand, but all from different sources). You reason that the likelihood that all twelve of them would be telling you the same thing if it weren’t the case is extraordinarily low, and so, on this basis, you come to believe that p. Here your attitude is not one of trust; on the contrary, you are treating the collection of testimonies as evidence, where on the basis of this evidence you are drawing the conclusion that p.2 This sort of case contrasts with one in which someone tells you that p and, “taking it on her authority,” you come to believe that p. The latter involves the sort of trust I am interested in. I suppose that it will be relatively uncontroversial to say that trust in this yet-to-be-specified sense is part of the characteristic way we respond to testimony. Even at this level of schematic abstraction, however, this idea must be handled with care. For one thing, the claim that testimony should not be regarded as a piece of evidence is a controversial one.3 And even those who accept it disagree over precisely how to characterize the contrast between testimony and evidence.4 Secondly, not all epistemologists who think about testimony agree that trust in this schematic sense is involved in the paradigmatic reaction to testimony;5 and even among those who think it might be involved, it is unclear what the distinctly epistemic significance of trust is.6 Nevertheless, I want to persist in using ‘trust’ to designate what I regard as a typical way to react to testimony in those cases in which one accepts the testimony: a hearer H trusts a speaker S on an occasion O in this sense when H comes to believe that p on the strength of S’s having asserted that p (on O).7 The key question is how believing that p “on the 2

If I read him right, Lipton (2007) would regard this case as no different in kind from a normal case of testimonial belief; his view was that testimony-based belief is always belief based on an inference to the best explanation. Obviously, I will be disagreeing with him on this matter. 3 See again Lipton (2007), who regards the fact of the testimony as the evidential basis of the inference underwriting testimonial belief. 4 See e.g. Grice (1957), Ross (1987), Goldberg (2006), and Moran (2007). 5 In addition to Lipton (2007), see also Fumerton (2007). 6 Among those who think it epistemically significant are Baker (1987), Pettit (1995), and Faulkner (2007b, 2007c, 2010, 2011). Among those who think it is epistemically insignificant is Lackey (2008). 7 That trust is involved in cases in which the hearer comes to believe the testimony is important. As my colleague Jennifer Lackey reminded me, one can rely on a piece of testimony, and in particular rely on its truth, even if one doesn’t believe it – as when,

Norms of Trust, De Re Trust, and the Epistomology ofTestimony

231

strength of” another’s say-so differentiates the case from the case of the dozen speakers above. On this question I offer the following conditions as individually necessary conditions on trust in my sense. (I would venture that they are jointly sufficient, but nothing hangs on this claim.) First, trust involves Epistemic Dependence: if H trusts S on O, then H epistemically relies on S on O, where, in virtue of this reliance, at least some of the epistemic properties of S’s trust-involving belief on O counterfactually depend othe quality of S’s say-so on O.8 If, on the basis of trusting you on the occasion of your stating that p, I come to believe that p, then e.g. whether or not I know that p depends on the quality of your say-so.9 Second, trust involves Exposure to Risk: in virtue of this dependence, there is an element of risk involved in trust, whereby the trustee (the one who is trusted) can fail to deliver the expected goods to the truster (the one doing the trusting). This can happen as a matter of S’s evil intention, as when S exploits H’s trust for her own gain, or as a matter of incompetence, as when S unwittingly offers insufficiently warranted testimony to H. Third, trust involves Semantic Guidance: if H trusts S on O, then H is guided in the matter of what to believe by what S linguistically represents to be the case. (Trust in the sense in which I am interested in it involves trust in someone regarding what is the case.10) Fourth, trust involves Hearer’s Awareness: if H trusts S on O, then in some

lost in some remote place and encountering a stranger who tells one that the town lies to the east, one might walk east without trusting the stranger at all (one has nothing else to go on). 8 In saying that at least some of the epistemic properties of the belief depend on the properties of the say-so, I am side-stepping the controversial matter regarding whether the justification or rationality of the belief depends on the properties of the say-so. Many people who deny that the justification or rationality of a testimonial belief depends on the de facto quality of the testimony agree that whether the hearer knows through testimony so depends. See Goldberg (2012) for a discussion. 9 This dependence is more complicated than one might suppose. Various people have discussed this. See e.g. Burge (1993), Faulkner (2000), Graham (2000a and 2000b), Lackey (2008), Goldberg (2010). 10 Practical trust offers a clear contrast. I might trust you to take out the trash; here, my trust in you is directed, not at what is the case, but what I would like to be made to be the case.

232

Sanford Goldberg

sense H is aware of this reliance.11 To be sure, H’s regarding her reliance on S as epistemically appropriate is itself something that can be assessed epistemically: perhaps H shouldn’t have trusted S, there were various reasons for doubt etc. There is one final condition that I want to discuss, concerning the speaker’s awareness of being relied upon in this way. I regard this Speaker Awareness condition – S is aware that H is depending on her for the truth – as something that, though satisfied in a special class of cases of trust, is not required by trust as such. As we will see, however, several authors writing on trust appeal to the Speaker Awareness condition to draw conclusions about the epistemic benefits of trusting another speaker.12 Since I will be denying that trust carries these benefits, a word is in order regarding my denial that the Speaker Awareness condition states a necessary condition on trust itself. It seems to me to be pretty clear that a hearer can take something on a speaker’s authority even when the speaker is unaware that the hearer is doing so. Consider cases in which one overhears another speaker, or reads someone’s diary without her knowledge. Pretheoretically these are cases in which the hearer/reader acquires beliefs on the authority of the speaker/writer, even though the speaker/writer is unaware of being relied upon in this way. To be sure, that such beliefs amount to testimonial beliefs has been denied by some.13 But I think that such a denial is entirely theorydriven: pre-theoretically, these are cases of testimonial belief, belief formed on the speaker’s authority. Consequently it seems to me that if ‘trust’ is to designate the (implicit or explicit) attitude involved in taking something on someone’s authority, the Speaker Awareness condition is not a necessary 11

It is perhaps controversial whether trust involves awareness of this sort. For example, one might think that very young children, who do not yet have awareness of their dependence on others, nevertheless trust their loved ones. Perhaps. But if this is so, then I would simply say that the sort of trust I have in mind is a more explicit, selfconscious sort – the sort that is in play in the mature subject’s response to certain testimonies. 12 This isn’t quite right. What they appeal to is the condition that the hearer make manifest to the speaker that he is relying on her in this way – something that entitles the hearer to assume that the speaker is aware of this reliance. (See e.g. Pettit 1995, 216.) However, nothing in particular hangs on this distinction in what I have to say below, so I will continue to speak of the Speaker Awareness condition. 13 Those who defend the “assurance view” of testimony typically deny that these are cases of testimonial belief. See e.g. Moran (2007).

Norms of Trust, De Re Trust, and the Epistomology ofTestimony

233

condition on trust as such. Still, its satisfaction (together with the other four conditions above) does pick out an interesting class of cases – what I will be calling cases of mutually recognized trust. I will have more to say about such cases in the next section; those unhappy with my denial that Speaker Awareness is necessary for trust will be addressed there, where I argue that, even in such cases, trust has no significant epistemic upshot. 2. I want to defend a very simple, and rather traditional, view regarding the epistemic payoff of trust. My simple view is that trust itself – the fact that H trusts S on O, as opposed to H’s reasons for placing his trust in S on O – has no distinctive epistemic payoff. This is a view that is shared by very many epistemologists who agree on little else. Thus one finds this view among motivated by an old-school epistemic internalism,14 as well as by externalists who endorse a reliability condition on epistemic justification.15 I think this is the right way to think about the epistemic payoff associated with trust. Recent epistemology, however, has seen a variety of authors advocating the view that trusting another speaker does have a distinctive epistemically payoff. Although early versions of this claim were perhaps overly focused on the non-epistemic reasons trust itself generates,16 more recent versions appear to be more epistemically focused. Thus we find some theorists arguing for the following: the fact that a hearer H places his trust in a speaker S can give S a motive to be reliable, and this is something that can be (and typically is) anticipated by H – with the result that the trust itself generates a reason in support of the belief H forms through that trust.17 A familiar example is the store owner who, hiring an ex-convict recently released from prison to run the cash register, and aiming to rehabilitate the ex-con’s moral character, trusts the ex-con even in the face of the strong evidence that there is for high rates of recidivism among ex-cons in general. The storekeeper’s trust in the ex-con, when recognized, can generate a motive for the ex-con to 14

See Fumerton (2007). See Lackey (2008, 240-248), Schmitt (2010). 16 See Baker (1987). 17 Versions of this idea can be found in Gambetta (1988), Pettit (1995), and Faulkner (2007 and 2011). 15

234

Sanford Goldberg

be reliable – namely, the desire for the continued good graces of the shopkeeper who trusts him (Pettit: 1995: 216). What is more, the shopkeeper, who recognizes that the ex-con is aware that he is being relied upon in this way, can appreciate the motive that he himself, ex-con has for speaking truly. And so it would appear that the shopkeeper has a reason – an epistemic reason – to regard ex-con as trustworthy, where this reason is generated by the fact that the shopkeeper trusts the ex-con.18 If that is the case, of course, then trust itself can have an epistemic payoff, since when its presence is mutually known it can generate reasons to regard a speaker as trustworthy. I believe that this sort of argument fails. Since my reasons for thinking this are based on a criticism Lackey levels against this view (Lackey 2008: 240-248), I begin with Lackey’s criticism. Lackey’s objection takes the form of a dilemma directed against anyone who endorses what she calls the ‘generative nature of trust.’ This is the feature (or alleged feature) of trust whereby a hearer, recognizing that trusting another speaker gives the speaker a motive to act in the expected way, thereby generates an epistemic reason to trust the speaker, even in the absence of having any pre-existing reasons to regard the speaker as trustworthy. Lackey suggests that anyone who endorses such a view faces the following dilemma: [E]ither the relevant hearer has independent reasons for believing the speaker in question to be trust-responsive, in which case beliefs acquired on the basis of her testimony may be epistemically justified, but not via an awareness of trust’s generative nature, or the hearer does not possess reasons for believing the speaker to be trust-responsive, in which case beliefs acquired on the basis of her testimony may be grounded in an awareness of trust’s generative nature, but they are not epistemically justified. Either way, epistemically justified testimonial beliefs cannot be grounded purely in an awareness of the generative nature of trust. (Lackey 2008: 242-43)

Now I think Lackey’s central point here is correct: trust itself, as opposed to the reasons one has for trusting a speaker on a given occasion, has no distinctive epistemically payoff. But I need to tweak her argument before it 18

See Pettit (1995, 216) for an explicit defense of this sort of argument. Faulkner (2007) contains a variant. Faulkner’s position might be stronger, in that he allows for the possibility that the reasons generated by what he calls “affective trust” can trump the reasons the hearer has for thinking that a given speaker is not trustworthy in a case like that of the ex-con; see Faulkner (2007, 549).

Norms of Trust, De Re Trust, and the Epistomology ofTestimony

235

will be suitable for my purposes. This is because the point I wish to make (asserting that trust itself has no epistemic payoff) gets somewhat obscured in the second horn of Lackey’s dilemma, in which the hearer lacks reasons to believe that the speaker is trust-responsive. (Call such a case, in which the hearer has no positive reasons for regarding the speaker as trustworthy or trust-responsive, “no positive reasons” cases.) Lackey’s claim is that the hearer’s belief in a “no positive reasons” case is epistemically unjustified. But the charge of unjustifiedness here is not required to show that trust itself has no distinctive epistemic payoff; and it would be best to recast Lackey’s dilemma so that it does not require this charge. To see why, consider that anti-reductionism about testimonial justification – a widely popular if not uncontroversial view in the literature – holds that, in the absence of reasons to think a piece of testimony untrustworthy (insincere or incompetent), one is entitled to accept it. Such a view can allow that a testimonial belief can be epistemically justified even in a “no positive reasons” case.19 Now suppose that anti-reductionism requires for its motivation a commitment to the generative nature of trust, in the sense that if the doctrine asserting the generative nature of trust is false, then anti-reductionism is without motivation. In that case, Lackey’s formulation of the dilemma would be correct as it stands, since in that case, the falsity of the doctrine asserting the generative nature of trust ensures that antireductionism is unmotivated – with the further result that a hearer who forms a testimonial belief with “no positive reasons” has formed an unjustified belief. But suppose that anti-reductionism does not require a commitment to the generative nature of trust. In that case, even if trust itself has no distinctive epistemic payoff, this need have no bearing on the justification enjoyed by a testimonial belief in a “no positive reasons” case. And since there is ample reason to think that anti-reductionism does not depend on the generative nature of trust (see the arguments in the references above), I conclude that we should focus on the issue of the epistemic payoff of trust itself, not on the doxastic justification of testimonial beliefs formed in “no positive reasons” cases.20 19

Versions of such a view are found e.g. in Coady 1992, Burge 1993, and Goldberg (2007, Chapter 5). 20 I should perhaps make clear that this is not intended as a criticism of Lackey, as she has what she takes to be reasons to think that anti-reductionism is false. And if anti-

236

Sanford Goldberg

Still, I think a small tweak in the formulation of the second dilemma will do the trick. Currently, the accusation Lackey levels in this case is that, while the “beliefs acquired on the basis of her testimony may be grounded in an awareness of trust’s generative nature,” even so “they are not epistemically justified.” The key point to my mind is not that they are not epistemically justified in such cases, but rather that if they are epistemically justified it is not in virtue of their being grounded in an awareness of trust. This is a minor tweak to Lackey’s argument, as is seen in the fact that the conclusion she herself draws acknowledges this point: “Either way, epistemically justified testimonial beliefs cannot be grounded purely in an awareness of the generative nature of trust.” The key task, then, is to show that in “no positive reasons” cases the hearer’s testimonial belief is not justified on the basis of its being grounded in her awareness of trust’s generative nature. We can put the point conditionally: if a testimonial belief in a “no positive reasons” case is justified at all, its justification derives from something else (more on which below). To our task, then: why think that a testimonial belief in a “no positive reasons” case cannot be justified merely in virtue of its being grounded in the hearer’s awareness of trust’s generative nature? Here I think that the sorts of considerations to which Lackey points make the case.21 For one thing, even in cases in which the trust is mutually recognized it is only sometimes true that a speaker (recognizing the hearer’s dependence on her) will want to preserve the good graces of her audience; in one-off cases, where the speaker will not again encounter the hearer, there would seem to be little reason to suppose that a speaker has such a desire, and yet these constitute a good many of the testimony cases that there are. Second, even given a hearer who both recognizes that an audience depends on her and does have such a desire relative to this audience, it is only sometimes true that this will motivate her to behave as her audience is relying on her to behave. After all, presumably she regards the decision whether to tell the truth as an all-things-considered matter, and her desire to preserve the good graces of her audience might be trumped by other considerations – considreductionism is false, then her dilemma, as stated, is cogent. It’s just that I disagree with her regarding the status of anti-reductionism. 21 Before giving these, Lackey pointed out that it is unclear how often it is the case that a speaker recognizes the hearer’s dependence on her.

Norms of Trust, De Re Trust, and the Epistomology ofTestimony

237

erations of self-interest perhaps, or of political motives, or motives of politeness or the desire not to offend, or … . The result of these points is straightforward: if the only reason that the hearer has to go on is the trustgenerated reason, then, even assuming that this does provide some minimal reason in support of the testimonial belief, it will not be enough to justify that belief.22 Lackey puts the key point in terms of luck. She writes that “it will be simply a matter of luck if the hearer happens upon someone who is appropriately trust-responsive,” concluding from this that the trust-based reason is not sufficiently strong to establish that the “subject’s belief is likely to be true.” And since the hallmark of justification-level reasons is that they establish that the subject’s belief is likely to be true, Lackey concludes that the subject’s testimonial belief in such a case is not justified (Lackey 2008, 246). To this I would like to add two other types of consideration to those Lackey presents, in defense of the claim that a grounding on the generative nature of trust is not sufficient to justify a testimonial belief. First, where Lackey defends this point by appeal to an anti-luck condition on justification, we might also defend it by appeal to an epistemic propriety condition on justification. A belief is justified only if it was formed in a way that was epistemically proper (in some intuitive sense). But given the points Lackey made (noted above), it would be epistemically improper to form a testimonial belief merely on the basis of the trust-generated reason. To see this, consider the sorts of worries that figure prominently in the literature on the epistemology of testimony: these have to do not only with speakers’ potential incompetence, but also speakers’ potential mendacity, and the variety of motives they might have other than speaking truth. Epistemologists rightly worry that in light of these possibilities, and given the distribution of the incompetent and the mendacious in our world, it is epistemically improper to accept what people say on blind trust.23 (Even anti-reductionists agree: see Goldberg and Henderson 2006.) To be sure, one who forms a testimonial belief on the basis of the trust-generated reason is not forming it ‘blindly.’ 22

Lackey herself does not address whether she regards this as providing any reason in support of the belief; but I don’t see any reason to doubt this. The key point is that this reason is not sufficient for justification. 23 Call trust ‘blind’ if it is neither based on any reasons to regard the speaker as trustworthy, nor sensitive to factors that indicate untrustworthiness.

238

Sanford Goldberg

But if accepting through blind trust is epistemically improper, it is hard to see how it could be proper to accept what another says merely on the basis of the generative nature of trust. After all, such a reason is perfectly generic, in the sense that it holds in any case of mutually recognized trust, in a way that is indifferent to variations between speakers. For this very reason it seems that the trust-generated reason is too insensitive and coarse-grained to render acceptance justified all by itself. Of course, if it is not proper to accept on this basis, then one who does so thereby forms a belief that is epistemically improper, hence unjustified. The conclusion is precisely that of Lackey’s dilemma: “epistemically justified testimonial beliefs cannot be grounded purely in an awareness of the generative nature of trust.” To this I would like to add one further, related point. (This might be a special case of the point I just made.) Those who in fact are trustresponsive are those who, recognizing that they are being trusted, do their best to live up to that trust, and so to behave in the fashion in which they are being trusted to behave. This involves doing their best to speak truly. Now a speaker who is trying to speak truly will avoid testifying when she recognizes herself to be insufficiently competent. Still, she might fail to recognize when she is insufficiently competent. For this very reason it would seem that a belief’s being grounded on the generative nature of trust is a belief whose basis is compatible with massive incompetence (albeit unrecognized by the speaker herself) in the speaker on whom the hearer is relying. To be sure, most hearers themselves are sensitive to massive incompetence on the part of the speaker; but acknowledging this point takes us beyond the reason deriving merely from the generative nature of trust. It continues to seem that such a reason is too insubstantial a basis on which to argue that the testimonial belief is justified. Again, Lackey’s conclusion holds: “epistemically justified testimonial beliefs cannot be grounded purely in an awareness of the generative nature of trust.” Still, we might ask what would provide the justification for a testimonial belief in a “no positive reasons” case? This question acquires added force given that I, unlike Lackey, allow that there can be justification in such cases. My answer, presented in detail in Goldberg (2007 and 2010), is this: the reliability of the testimonial belief-forming process itself justifies the belief. (As I think of this process, it includes not only the process by which the hearer comprehends the testimony and monitors it for trustworthiness, but also the testimonial production process itself.) I do not here argue for this view, so much as point out that it is view that has been defended. The

Norms of Trust, De Re Trust, and the Epistomology ofTestimony

239

take-home lesson is simply this: even after we dismiss the existence, or perhaps merely the substance, of the reasons generated by trust, we are not restricted to having to endorse the (reductionist) view that requires positive reasons to regard a speaker as trustworthy, as a condition on justified testimonial belief – as Lackey’s formulation of the dilemma might suggest. We are free to acknowledge the force of the case against the generative nature of trust, without having to resort to reductionism about testimonial justification. 3. It is tempting to conclude from the foregoing that, at least from the perspective of epistemology, the phenomenon of trust (as implicated in testimonybased belief) is of little or no interest. One who draws such a conclusion can allow that the phenomenon of trust is of interest for those doing ethics, and/or for those with an interest in social norms; the conclusion is that trust is of little or no interest to the epistemology of testimony. But such a reaction, I submit, would be wrong. Even though the phenomenon of trust has no significant epistemic payoff, it should still be of interest to epistemologists, albeit for another reason: it points us in the direction of the distinctive epistemic features of beliefs formed through our epistemic reliance on others. The point I would like to make here, concerning the distinctive epistemic features of beliefs formed through our epistemic reliance on others, is one I pursued in great detail in Goldberg (2010). What I would like to add here concerns the role that the phenomenon of trust itself can play in helping us to appreciate the nature of our epistemic reliance on others. My claim is this: the phenomenon of trust can be used to argue for a distinctly interpersonal treatment of the epistemology of testimony. Further, the resulting position is a very different sort of interpersonal treatment than is currently on offer by those who stress the interpersonal dimension of testimony-based belief (such as the proponents of the “assurance view” of testimony). To defend this I will introduce, characterize, and then appeal to what I will be calling the de re dimension of trust.

240

Sanford Goldberg

I submit that trust is a relation to a particular information source. The point here can be put in terms of the nature of trust itself:24 trust is an attitude directed at a particular information source on a particular occasion on which the source presents some piece of information as true. The point can also be put in terms of the correct analysis of ‘trust’-involving sentences: ‘trust’ denotes a relation, one of whose relata is the truster and another is the trustee. I do not expect either of these points to be very controversial; I think it is obvious once one gives these matters any thought. In having this sort of trust in mind, I will be speaking of de re trust in a source, or de re trust for short. The controversial point I am making concerns not the existence of de re trust, but its significance for epistemology. Before moving on to consider that significance, I want to clear up a possible confusion one might have even at this early point of my discussion of de re trust. We sometimes speak of individuals who exhibit a high degree of trust, who are (as we say) “highly trusting” individuals. When we use ‘trust’ in this way we are describing a person’s character trait, not something that relates her to any particular person. We need to distinguish, then, what I am calling de re trust from what we might call characterological or dispositional trust. Unlike de re trust, dispositional trust is not a relation to a particular information source. Of course not: it is a disposition or character trait. And this might lead us to a question: if we are interested in the epistemology of testimony, which if either of these notions of trust – de re vs. dispositional – is fundamental? And how do they relate to one another? My view is this. One who is highly trusting – one who has a high degree of dispositional trust, as we might say – has a disposition to accept what others tell her, where the disposition in question is more easily elicited than the disposition we would see in someone of a more ‘skeptical’ bent. Such a disposition, by itself, does not tell us not very much about the epistemological features of the beliefs formed through testimony. This said, high dispositional trust does have an indirect bearing on the epistemic features of beliefs formed through testimony; only, as I will presently argue, this indirect bearing only serves to highlight the fundamental importance of de re trust.

24

It is to be borne in mind that the sort of trust I am discussing is that implicated in certain communicative exchanges. I do not deny that there are other kinds of trust, as when e.g. I trust that this bridge will not collapse beneath the weight of our car.

Norms of Trust, De Re Trust, and the Epistomology ofTestimony

241

Suppose that we are in a world in which, as a matter of counterfactuallysupporting regularity, most testimonies are true. In that world, a high degree of dispositional trust would seem to be an epistemically good thing, as it would conduce one to accept most of the testimonies one encountered, and so (since most testimonies are true) it would conduce one to having a set of beliefs most of which were true. Now suppose that we are in a world in which, as a matter of counterfactually-supporting regularity, most testimonies are false.25 In that world, a high degree of dispositional trust would seem to be an epistemically bad thing, as it would conduce one to having a set of beliefs most of which were false. Clearly, the epistemic goodness or badness of dispositional trust, as measured by the truth-related outcome of having such a disposition, depends on how the world is, and in particular on the regularities pertaining to the ratio of true to total testimonies. This point might tempt us to think that the epistemology of testimony, at least for externalist theories, ought to give pride of place to dispositional trust. The reasoning that leads to this conclusion would be as follows. What we care about is how modally robust our beliefs are: how reliable, sensitive, safe, etc. But how modally robust our testimonial beliefs are will reflect (among other things) how trusting we are, how high a degree of dispositional trust we have. The more trusting we are, the better our testimonial beliefs are in worlds where most testimonies are true, and the worse our testimonial beliefs are in worlds where most testimonies are false, with intermediate goodness values in worlds where the truth-ratio of testimonies is mixed. Generalizing from this, one might think that a subject’s degree of dispositional trust is part of her on-board process for managing testimony, and that it is the goodness or badness of this on-board process, relative to the prevailing testimonial environment, that determines the epistemic goodness or badness of resulting testimonial beliefs.26 But whatever attractiveness such a proposal has dissipates on reflection. In particular, such a proposal does not seem equipped to accommodate the phenomenon of de re trust at all; it cannot capture the reliance a hearer has 25

Some have argued that such a world is an impossibility; see e.g. Coady 1992. Such a view is controversial; see Graham 2000c. In any case, let the world in question be a world in which the ratio of false to total testimonies is as high as possible. 26 Something like this view appears to be the way Alvin Goldman approaches testimony. (Goldman has expressed his views to me in conversation.)

242

Sanford Goldberg

on the particular speaker whose particular speech contribution she observes on a given occasion. Suppose I trust you when you tell me that p. Here I am depending on the goodness of your testimony; this much seems obvious. Can the proposal under consideration acknowledge this? Suppose, with this proposal, that the epistemic goodness of my testimonial belief is determined by the epistemic goodness of the process through which I manage testimonies generally, and its suitability to the normal or standard testimonial environment. On such a view, the epistemic goodness of the testimonial belief I acquire through your testimony is determined by the epistemic goodness of the processes that I employ as I react to the various testimonies I get throughout some suitably extended portion of my adult life. It should be obvious that such a view simply ignores the contribution you in particular make on this particular occasion of testimony.27 In short, the proposal on offer does not acknowledge my dependence on you, in particular, on this occasion. Perhaps it will be thought that the sort of dependence the hearer has on the speaker on a given occasion of testimony is akin to the sort of dependence a would-be knower has on the compliance of the external world in those cases in which there are no Gettier conditions present. To develop this proposed, consider the following two claims: (1) (2)

I depend on the world’s cooperation in a given scenario if my true, justified belief is to be knowledge: the world must be such that there are no Gettier factors. I depend on your cooperation on a given occasion in which I de re trust you if my true, justified belief formed through your testimony is to be knowledge: your testimony must itself be suitably reliable.

Call the sort of dependence described in (1) ‘anti-Gettier dependence.’ The proposal on the table is that the dependence in (2) is a special case of antiGettier dependence. One who endorses this proposal will claim that it is this that captures the relevance of de re trust. 27

This is because the epistemic goodness of the processes I employ are determined by the totality of testimonies that I have encountered (might encounter?) throughout the relevant interval. Your testimony on this occasion is among that totality, of course; but it is only one testimony among many, many others.

Norms of Trust, De Re Trust, and the Epistomology ofTestimony

243

In response, I want to question whether (2), which I accept as true, captures the totality of our epistemic dependence on others in testimony cases. In particular, I question whether (2) captures the sort of epistemic dependence manifested in de re trust.28 My reasons are as follows. First, when I depend on your word, you are not a mere piece of earthly furniture whose regularities I rely on. Rather, you are an agent whose intentional (speech) act I am relying on. A dimension of responsibility and blameworthiness is present as a matter of course in cases involving de re trust,29 while the same is not true as a matter of course in cases involving anti-Gettier conditions.30 In addition, our anti-Gettier dependence on the world is a kind of dependence on normal background conditions; but when I depend on your testimony, your testimony is front-and-center, in the foreground. This use of the metaphor of background and foreground points to an important difference between the two cases. In particular, it highlights the de re nature of the dependence in testimony, which is absent in our anti-Gettier dependence on the world. In testimony, there is a particular entity (the speaker) whose particular act (speech contribution) is being epistemically relied upon by the hearer; in our anti-Gettier dependence on the world, there need be no particular entity or particular act or event being relied upon, rather the subject is relying on the obtaining of normal conditions, and on the non-intrusion of any luck-introducing factors. For these reasons, it is unreasonable to assimilate our epistemic dependence in testimony cases to our anti-Gettier dependence on the world. And this just points to my previous claim: the de re nature of the trust implicated in testimony cases makes

28

If I am correct in this regard, we should not assimilate that dependence to antiGettier dependence. 29 Indeed, proponents of so-called “assurance view” have made a big deal of the element of agency and susceptibility to blame; see Ross (1987), Hinchman (2006), Moran (2007), and McMyler (2012). But one need not be a proponent of the assurance view to see the importance of the blame dimension, or to account for it; see Goldberg (2010b). 30 That is, while there may be blame in cases of Gettierization – you may be a devious epistemologists who brought me to a hillside where you knew I would take that sheepdog to be a sheep, knowing as well that the sheepdog obscured a sheep – this is a special case, not something characteristic of all cases in which I depend on the world to protect me against Gettierization. “(With thanks to Baron Reed for raising this sort of point, in conversation.)

244

Sanford Goldberg

the sort of dependence involved epistemically distinctive, unlike the antiGettier dependence we have on the world in empirical belief generally.31 I have been arguing for the centrality of de re trust, as against dispositional trust, to the epistemology of testimony.32 My case so far has assumed a decidedly externalist orientation. But in fact the same point can be used to generate difficulties for an internalist epistemology. Let us assume that we want to think of epistemic justification and knowledge in terms of reasons or evidence that is/are accessible to the subject under assessment. It is clear that any view that regards this as the totality of the basis for epistemic assessment will utterly fail to be able to acknowledge the phenomenon of de re trust. This is for a simple reason: when I de re trust you on an occasion of your testimony, the epistemic goodness of your testimony – its de facto reliability – is not something that is accessible to me. What is accessible to me are all sorts of things I take to be indications or evidence of your competence and sincerity: your apparent confidence, your looking me in the eyes, your past track-record, and so forth. Of course, there could be two cases that are exactly alike with respect to all of these features, yet where in one case the testimony is reliable and in the other not. So far as an internalist epistemology of testimony cares, these cases should not differ at all (at least not at the level of justification). And yet, the phenomenon of de re trust suggests otherwise: one who de re trusts a speaker on an occasion of reliable testimony is better-off, epistemically, than one who de re trusts a speaker on an occasion of unreliable testimony – even if both cases the hearers share the same “internally accessible” reasons and evidence. Of course, the internalist can try to accommodate this point by suggesting that it has to do with knowledge, not justification. But we have already seen that this sort of point cannot be supported by any analogy between our dependence in testimony cases and our anti-Gettier dependence on the world; so at the very least the internalist who makes this move owes some sort of motivation for it.

31

See Goldberg (2010, Chapter 4) for a distinct argument to the same effect. What, then, is the proper way to incorporate the phenomenon of dispositional trust into an adequate epistemology of testimony? This question turns out to be more vexed than many have appreciated. See Goldberg (2010, Chapter 5).

32

Norms of Trust, De Re Trust, and the Epistomology ofTestimony

245

I conclude, then, that there are reasons to acknowledge the central importance of the phenomenon of de re trust to the epistemology of testimony. 4. In section 2 I argued that in itself, trust has no significant epistemic payoff. In section 3 I argued that de re trust should nevertheless be of interest to epistemologists. How to square these two claims? Here I suggest that the importance of the phenomenon of de re trust, at least for those interested in epistemology, is in what it tells us about trust-based belief. In particular, it tells us that a full accounting of the epistemic credentials of the belief formed through de re trust must involve an assessment of the epistemic goodness of the (de re-trusted) source. Although this point can seem obvious, in fact it is quite controversial. I begin with a few words on why it is controversial before proceeding to argue for it. When it is present in testimony cases, de re trust is a form of interpersonal trust. So if the point I wish to defend is correct, the result would be the following: (T)

A full accounting of the epistemic credentials of a testimonial belief formed through de re trust must involve an assessment of the epistemic goodness of the source speaker on whom the hearer relied, and in particular on the process by which that speaker produced the testimony.

Now one might be inclined to accept (T) as a more-or-less obvious point about the nature of testimonial knowledge. Those who think of (T) in this fashion will regard it as reflecting the following point: whether a hearer acquires knowledge through testimony involves assessing the process by which the speaker produced the testimony. And this point would appear to be obvious: if the speaker wasn’t actually competent on the matter at hand, then, even if (a) she got it right and (b) the hearer had excellent reasons for relying on her say-so, the hearer doesn’t acquire knowledge. This is for the simple and well-appreciated reason that there would be a knowledgeundermining element of luck in the process by which the hearer came to believe truly.

246

Sanford Goldberg

In response I concede the foregoing but deny that this point about testimonial knowledge exhausts the full significance of (T). Indeed, my reasons on this score will have been anticipated. For the foregoing claim makes a point regarding Gettierization. But above, in section 3, I argued that our epistemic dependence in testimony cases goes beyond anti-Gettierization dependence. If this is correct, then the full significance of (T) goes beyond a point about the conditions on testimonial knowledge. And herein lies the controversial nature of (T). I submit that (T) should be read in such a way that it implies the following: (T1) How epistemically well-off H’s belief is can depend on facts that determine how epistemically well-off another subject S is, in ways that go beyond whether H’s belief amounts to knowledge. Though it is not entirely unheard of in recent epistemology,33 (T1) is nevertheless still quite controversial. In what follows, I want to suggest how the phenomenon of de re trust might be used as the basis of a novel argument for this idea. I want to argue for (T1) in two steps. First, I argue that (T2) How epistemically well-off H’s epistemic position is regarding whether p, in those cases in which H relies on the speech contributions of another speaker (S), depends on facts that determine how epistemically well-off S’s speech contribution is, in ways that go beyond whether H’s epistemic perspective on whether p amounts to knowledge. Then I will argue that (T2) is a special case of a still more general claim regarding reliance on information sources, to the effect that (T3) How epistemically well-off H’s epistemic position is regarding whether p, in those cases in which H relies on an information source φ, depends on facts that determine how epistemically welloff φ’s contribution is, in ways that go beyond whether H’s epistemic perspective on whether p amounts to knowledge.

33

Versions of it are defended in Owens (2004), Schmitt (2007), Lackey (2008), and Goldberg (2010).

Norms of Trust, De Re Trust, and the Epistomology ofTestimony

247

I regard (T3) as a fundamental thesis regarding the epistemic significance of reliance on an information source; (T2) as the special case of (T3) in which that reliance takes the form of reliance on the speech of another speaker on a given occasion; and (T1) as that special case of (T2) in which that reliance on another speaker takes the form of de re trust of another speaker’s contribution on a given occasion. My ultimate claim, then, is that (T1) holds because (T3) does (the former being a special case of the latter). I want to begin, however, not with (T3), but with (T2). Here I want to start with a series of cases that may or may not be cases of de re trust, and may or may not eventuate in something we would want to acknowledge as an actual belief.34 What makes this series interesting to me is that the cases themselves manifest the phenomenon that I am calling epistemic reliance on another speaker on a given occasion. (In this way they manifest what I will argue are one of the crucial features of de re trust, namely, the candidate risks and payoffs involved in the typical case of de re trust.) Let us begin with one case in the series. S and H are lost deep in the woods of a very large forest that is not familiar to either one of them, in the middle of the hot summer, with their water supply running dangerously low. T himself is not an experienced hiker; S is a Park Ranger, with years of experience hiking in forests in this part of the country. On the basis of what S takes to be some signs in the local environment that point to the possibility of a water source off to their west, she says that the best thing for them to do is to walk west. On hearing this H decides that, since they have to do something given the little water they have left, and since S is in a better epistemic position than he, H, himself is to determine what it would be best to do, he will rely on S’s proposal. And so he does, and the two of them start the long, hot westward journey. Now consider variations on this case where the only dimension of variation concerns the degree to which S’s judgment (regarding the possible presence of a water supply to the west) is competent: in some cases she is more competent, and in cases slightly less. We can now arrange the cases according to the degree of competence manifested in S’s judgment. Now it seems to me that the following aptly describes this series of cases: insofar as H relies on S’s judgment, H himself is better-off, epis34

The point I wish to make with this series does not depend on any assumptions on either score.

248

Sanford Goldberg

temically speaking, in proportion to the degree of competence that S manifests in that judgment. To see this, let’s be clear about what such reliance involves. H relies on S’s judgment at least in the minimal sense that H accepts this judgment as determining how they should proceed. Such reliance need not involve de re trust: perhaps H is a skeptical sort, and so while he is willing to rely on this proposal (having nothing better to go on), he would not claim to trust S on this occasion. Stronger, such reliance need not even involve (full) belief: it might be that, while H is prepared to act as if the best thing to do is to proceed westward, he does not actually believe this (although he might assign this proposition some suitably appropriate credence that falls short of outright belief). But whether or not this case involves de re trust or (full) belief on H’s part, the point remains: how epistemically well-off H’s epistemic position is regarding whether it is best for them to walk west depends on how epistemically competent S’s judgment to that effect was. This is perhaps best seen in the extreme cases, where S is extremely competent in water-detection, on the one side, and extremely incompetent in water-detection, on the other. In the former case, we can imagine that what S takes to be evidence of water to the west is just that (and there is water to the west, nearby); in the latter case, we can imagine that what S takes to be evidence of water to the west is no such thing (and there is no water to the west, even extending out dozens of miles in that direction). But it is worth emphasizing that the same point holds even in the more intermediate cases: insofar as H relies on S’s (manifestation of) competence on this occasion, the more competent S is, the better off, epistemically, H is. And (this is key) this last point remains true even if we consider two cases that are otherwise alike with respect to whether or not S knows of the presence of water several miles westward. So imagine two cases in which S does know, but where in one of the cases S’s competence approximates epistemic certainty whereas in the other it is just barely over the threshold for knowledge; or, alternatively, imagine two cases in which S does not know, but where in one of the cases S’s competence is just barely under the threshold for knowledge, whereas in the other she is incompetent (though takes herself to be competent, and acts accordingly). I submit that in these cases, H’s epistemic perspective is better or worse, according to the degree of competence that is manifested in S’s judgment. If so, then there can be differences in H’s epistemic perspective that are not traceable to any difference in whether S knows, and so, presumably, are not to be represented as differences in whether H himself knows. Since these

Norms of Trust, De Re Trust, and the Epistomology ofTestimony

249

differences reflect differences in S’s epistemic competence, I conclude that (T2) is true. So far, the point that I am making concerns a minimal sort of epistemic reliance on another speaker – one that is in place whether or not the hearer forms a (full) belief in what the relied-upon speaker says, and whether or not the hearer de re trusts the speaker on whom she is relying. But it seems that the point that I am using this example to illustrate – namely, how the epistemic goodness of the hearer’s perspective reflects facts about the speaker’s perspective, in ways that go beyond the question of whether the hearer knows – can be illustrated in other cases of epistemic reliance. In fact, they can be illustrated in cases in which this reliance is a reliance, not on another person, but on oneself (one’s own faculties). For example, consider a child’s reliance on her perceptual faculties. I submit that this sort of reliance can be used to illustrate much the same point. Once again, it will be helpful to have a series of cases before us, where the only difference between them is in the perceptual acuity of the child’s perceptual faculties. At the extremes, imagine one case in which the child has perfect vision, where her visual faculties are nearly perfectly reliable, and another case in which the child’s vision is awful. Suppose that in both cases the child relies on her visual system for getting about in the world: in the case of perfect vision, she negotiates her immediate environment well, whereas in the case of awful vision, she is often bumping in to things. Clearly, her epistemic position regarding the presence or nonpresence of an object before her depends on the epistemic goodness of her visual faculties. But this same point can be made even in the intermediate cases: the epistemically better her perceptual faculties, the epistemically better is her epistemic position regarding the information she acquires through those faculties. And again, we can set up pairs of cases in which the difference between them is not a difference at the level of knowledge: either she knows in both, or does not know in either. I think it is important to highlight that the same point that we saw above in an interpersonal case can be illustrated in an intrapersonal case (of selfreliance) as well. What this suggests is that the point has to do, not with the interpersonal nature of the reliance, but with the reliance itself. In particular, it has to do with the nature of reliance on an information processing system – whether that system is an extractor of information (as in the case

250

Sanford Goldberg

of one’s perceptual system) or a transmitter of information (as in the case of another speaker).35 Either way, one’s reliance on an information system affects the epistemic goodness of one’s epistemic position towards the information one acquired from that system. (T3) captures this point in its full generality. It is at this point that we can return to cases involving de re trust, to discern their relevance for epistemology. I want to claim that its relevance to epistemology reflects the fact that de re trust of another speaker is a special case of epistemic reliance, and that as such beliefs formed through this sort of trust require an interpersonal kind of epistemic assessment. In what follows I develop this point. I submit that de re trust of another speaker is a special case of the phenomenon I have been calling ‘epistemic reliance.’ It is that special case of epistemic reliance in which (i) one’s epistemic reliance is on (the speech contribution of) another speaker, where (ii) having recognized (if only implicitly) that one is so relying, (iii) one explicitly endorses one’s reliance on the speaker’s contribution, and on the basis of this endorsement (iv) one comes to believe what was presented-as-true in the speech contribution. Suppose that I am correct in this, so that de re trust is a special case of epistemic reliance. And assume as well that I am correct that epistemic reliance in general points to the need to assess the process by which the relied-upon information-processing system made its contribution. These two points underscore the relevance to epistemology of the phenomenon of de re trust: it establishes a constraint on any adequate epistemology of testimonial belief. In particular, no epistemic assessment of a testimonial belief is adequate unless it assesses (not only the process by which the hearer observes and comprehends the testimony, and assesses it for credibility, but also) the process by which the source speaker came to produce the testimony. That is, the phenomenon of de re trust points to the fundamentally interpersonal nature of the epistemology of testimonial belief. 5. I am not the first one to point to, or to argue for, the fundamentally interpersonal nature of the epistemology of testimonial belief. On the contrary, 35

I owe this distinction to Peter Graham, who suggested it to me in conversation.

Norms of Trust, De Re Trust, and the Epistomology ofTestimony

251

many others before me have done so. Most saliently, those advocating the so-called “assurance view” of testimony have done so.36 However, unlike the argument that is typically offered by proponents of that view, my argument has tried to distinguish very sharply between the ethical dimension of testimonial exchanges, and the epistemological dimension of such exchanges. In this last section I want to draw this contrast between my argument, and that presented by proponents of the “assurance view” of testimony. (As several people have presented concerns regarding the assurance view elsewhere,37 here I will be brief.) According to Richard Moran’s version of the “assurance view,” to testify that p is to give one’s assurance that p. (It is, in the words of Ted Hinchman – another proponent of the assurance view – to “invite” another to trust one on the matter.) In assuring another person that p, one generates for oneself a responsibility to have the epistemic goodies that warrant one’s so assuring her. Correspondingly, a hearer who accepts one’s testimony does so in virtue of accepting one’s assurance; the hearer who does so is thereby entitled to hold the speaker responsible for having the warranting epistemic goodies. Relatedly, the hearer is entitled to complain if it should turn out that the testimony she accepted was no good (either because the speaker was incompetent or insincere). And these features of testimony – the feature whereby the testifier offers her assurance (and so generates for herself the responsibility to have the relevant warranting authority), and the hearer accepts that assurance (and in so doing inherits an entitlement to hold the speaker responsible for having that authority) – form the core part of what we might call the assurance view’s account of the responsibilities bound up with testimony. Interestingly, the assurance view does not stop at an account of the responsibilities of testimony, but aims to build an account of the epistemology of testimonial belief from its account of these responsibilities. The basic idea is this. A speaker S who assures another person that p is responsible for having the epistemic authority needed to be warranted in so assur36

Proponents include Hinchman 2005, Moran 2006, and McMyler (2012). Arguably, an early version of what would become the assurance view is found in Ross 1987. All of these versions take inspiration from Grice 1958. I will focus on Moran’s version, as it is the version that has attracted the most critical attention. 37 See Lackey (2008: 219-40), Schmitt (2010), and Goldberg (2010b).

252

Sanford Goldberg

ing her. A hearer H who observes S testify that p acknowledges S’s assumption of responsibility, and (assuming H has no reasons for concern regarding this assurance) forms a belief on the basis of regarding S as satisfying this responsibility. That S has assumed responsibility for the truth of p is itself a reason that S makes available to another subject – it is a reason to believe that p. When H acknowledges that S has assumed this responsibility, H thus has this reason as his own, and so has a reason to believe that p. When all goes well, the testimonial belief H acquires through accepting S’s assurance that p – H’s testimonial belief that p – is justified. In this way the assurance view holds out the tantalizing prospect that the responsibilities which are part of the institution of testifying are themselves implicated in the very epistemic support for the testimonial belief. Many philosophers have welcomed this prospect, thinking it a way to unite the epistemology of testimony with the ethics of testimony.38 But trouble looms when we consider precisely what does the justifying. Consider for example the claim that another’s assurance, together with the hearer’s having no reason to think that an act of assurance is somehow deficient, suffices to justify the hearer’s testimonial belief. Such a view would be a clean account of the epistemology of testimony alright. But it leaves open a crucial question: how can the fact that another has assured one that p, even when one acknowledges this fact, serve to justify one in believing that p? Isn’t such a view guilty of holding a “magical view” of assurance, according to which the mere fact of assurance itself is epistemically significant?39 Recognizing something like this worry, Moran himself notes that if the hearer’s assurance-based belief is to be justified by another’s assurance, the assurance itself must actually be warranted. And to my mind, this just raises the question: why think that the assurance itself,

38

The distinctly moral dimension itself was highlighted as early as Ross (1986, 7778). But it is also a feature in the thinking of several who aim to draw parallels between practical reason and theoretical reason; see e.g. Darwall 2006 (The Second Person Perspective). 39 Moran invites this objection when he characterizes the speaker as “invest[ing] his utterance with a particular epistemic import,” where this investment reflects the speaker’s “explicit assumption of responsibility for his utterance’s being a reason for belief.” (2006, 291; italics added)

Norms of Trust, De Re Trust, and the Epistomology ofTestimony

253

as opposed to whatever it is that warrants the assurance, is what does the justificatory work?40 The present account, appealing to the epistemic reliance implicated in cases involving the de re trust in another speaker, does a better job of accounting for the phenomenon here. In extending de re trust to a speaker on an occasion, one is exhibiting epistemic reliance on her speech contribution. It is a particularly explicit and self-conscious form of reliance, to be sure, but it is a form of epistemic reliance. And it is for this reason that one has opened oneself up to a certain kind of risk, whereby the epistemic status of one’s trust-based belief reflects the de facto epistemic goodness of the (process eventuating in the offering of the) testimony. This risk is on a par with the sort of risk that we run when we rely on our perceptual faculties (as we must, if we are to have any hope of warranted empirical belief). In particular, if our perceptual faculties are not reliable, then our perception-based beliefs are (correspondingly) epistemically deficient. The point here has nothing whatsoever to do with the assumption of responsibility on the part of the speaker. This, I think, is the key lesson of a focus on de re trust: while it points to the distinctly interpersonal nature of the epistemology of testimony, it does not motivate the interpersonal account through considerations of ethics or responsibility, and in this way allows us to continue to make a sharp contrast between the epistemology of testimony and the ethics of testifying. One might still wonder about the ethics of testifying: how can a speech act have an ethical dimension? I think that this is a very good question; and I think that it is a tribute to proponents of the assurance view that they have brought this to our attention. However, I think that the answer to their question is very different from what they imagine. I think that the best way to understand the ethics of testifying is in terms of the hypotheses (one) that testifying involves assertion, (two) that assertion itself has an epistemic norm, and (three) that this feature of assertion is mutually familiar to all people who are competent in the practice. However, it is a matter for another occasion to say how these three hypotheses enable us to account for the ethics of testimony.41 40

I develop this criticism at length in Goldberg (2010b, 187-192). I offer the outline of such an account in Goldberg (2010b). I offer the full details in Goldberg (manuscript).

41

254

Sanford Goldberg

References Baker, J., 1987: Trust and Rationality, Pacific Philosophical Quarterly 68, 1-13 Burge, T., 1993: Content Preservation, Philosophical Review 102, 457-488 Coady, C., 1992: Testimony: A Philosophical Study, Oxford: Oxford University Press Darwall, S., 2006: The Second-Person Standpoint: Morality, Respect and Accountability, Cambridge: Harvard University Press Faulkner, P., 2000: The Social Character of Testimonial Knowledge, Journal of Philosophy 97, 581-601 Faulkner, P., 2007a: What is Wrong with Lying? Philosophy and Phenomenological Research 75, 535-557 Faulkner, P., 2007b: A Genealogy of Trust, Episteme 4, 305-321 Faulkner, P., 2007c: On Telling and Trusting, Mind 116, 875-902 Faulkner, P., 2010: Norms of Trust, in: Social Epistemology, ed. A. Haddock, A. Millar, and D. Pritchard, Oxford: Oxford University Press, 129-147 Faulkner, P., 2011: Knowledge on Trust, Oxford: Oxford University Press Fumerton, R., 2007: The Epistemic Role of Testimony: Internalist and Externalist Perspectives, in: The Epistemology of Testimony, ed. J. Lackey and E. Sosa, Oxford: Oxford University Press, 77-92 Gambetta, D., 1988: Can We Trust Trust? In: Trust: Making and Breaking Cooperative Relations, ed. D. Gambetta, Oxford: Blackwell, 213-37 Goldberg, S., 2006: Testimony as Evidence, Philosophica 78: 29-52 Goldberg, S., Anti-Individualism: Mind and Language, Knowledge and Justification, Cambridge: Cambridge University Press Goldberg, S., 2010: Relying on Others: An Essay in Epistemology, Oxford: Oxford University Press Goldberg, S., 2010b: Putting the Norm of Assertion to Work: The Case of Testimony, in: Assertion: New Philosophical Essays, ed. J. Brown and H. Cappelen, Oxford: Oxford University Press, 175-196 Goldberg, S., 2012: Preserving Sensitivity, in: Sensitivity: New Essays, ed. K. Becker and T. Black, Cambridge: Cambridge University Press Goldberg, S., (manuscript): The Ethics of Assertion as an angle on the Ethics of Belief. Graham, P., 2000a: Conveying Information, Synthese 123, 365-392 Graham, P., 2000b: Transferring Knowledge, Nous 34, 131-152 Graham, P., 2000c: The Reliability of Testimony, Philosophy and Phenomenological Research 61, 695-709 Grice, P., 1957: Meaning, Philosophical Review 66, 377-388

Norms of Trust, De Re Trust, and the Epistomology ofTestimony

255

Hinchman, T., 2005: Telling as Inviting to Trust, Philosophy and Phenomenological Research 70, 562-87 Lackey, J., 2008: Learning From Words, Oxford: Oxford University Press McMyler, B., 2012: Testimony, Trust, and Authority, Oxford: Oxford University Press Moran, R., 2006: Getting Told and Being Believed, in: The Epistemology of Testimony, ed. J. Lackey and E. Sosa, Oxford: Oxford University Press, 272-306 Owens, D., 2000: Reason without Freedom: The Problem of Epistemic Normativity, London: Routledge Pettit, P., 1995: The Cunning of Trust, Philosophy and Public Affairs 24, 202-225 Ross, A., 1986: Why do we believe what we are told? Ratio 28, 69-88 Schmitt, F., 2006: Testimonial Justification and Transindividual Reasons, in: The Epistemology of Testimony, ed. J. Lackey and E. Sosa, Oxford: Oxford University Press, 193-224 Schmitt, F., 2010: The Assurance View of Testimony, in: Social Epistemology, ed. A. Haddock, A. Millar, and D. Pritchard, Oxford: Oxford University Press, 216242

Jennifer Lackey on Non-Reductionism: A Critique MARTIN KUSCH, University of Vienna

1. Introduction In this paper I shall try to explain why I am unconvinced by Jennifer Lackey’s criticism, in her recent book Learning from Words (2008), of nonreductionism in the epistemology of testimony. I shall use Tyler Burge’s papers “Content Preservation” (1993) and “Interlocution, Perception, and Memory” (1997) as my paradigm of the non-reductionist spectrum of positions. 2. Lackey on Non-Reductionism: Its Definition and Main Argument Lackey defines non-reductionism (= NR) as follows: NR: (NR 1) (NR 2) (NR 3) (NR 4) (NR 5) (NR 6)

For every speaker, A, and hearer, B, B knows that p on the basis of A’s testimony if and only if: B believes that p on the basis of the content of A’s testimony, B has no undefeated (psychological or normative) defeaters for A’s testimony, it is true that p, A’s testimony is reliable or otherwise truth-conducive, B is a reliable or properly functioning recipient of testimony, and the environment in which B receives A's testimony is suitable for the reception of reliable testimony (2008, 167).

Non-reductionism differs from its two opponents – reductionism, and Lackey’s own position, called “dualism” – in that it rejects the principle “PR-N”. PR-N says that “positive reasons” are necessary for the hearer to have a justified or warranted testimonial belief. Reductionism in turn differs from dualism is that the former, but not the latter, regards such positive

258

Martin Kusch

reasons as not only necessary but also sufficient for justification or warrant in the case of a testimonial belief (2008, 148). As Lackey reconstructs non-reductionism, its main argument against PR-N is the "infant/child objection": 1. 2. 3. 4. 5.

According to PR-N … B’s belief that p is justified or warranted on the basis of A’s testimony only if B has appropriate positive reasons for accepting A’s testimony. However, infants and young children lack the cognitive capacity for acquiring and possessing appropriate positive reasons. Therefore, infants and young children are incapable of satisfying PR-N. Infants and young children do have at least some testimonial knowledge. Therefore reductionism and dualism are false. (2008, 196)

Or, if you prefer a shorter version: We all agree that young children do have some testimonial knowledge. But young children are unable to acquire positive reasons. Therefore, and contrary to PR-N, positive reasons are not necessary for testimonial knowledge. 3. Lackey’s Criticism of Non-Reductionism Lackey first line of criticism of non-reductionism seeks to establish a mismatch between non-reductionism and our epistemic intuitions. Central here are two examples that I shall quote in full: ALIEN: Sam, an average human being, is taking a walk through the forest one sunny morning and, in the distance, he sees someone drop a book. Although the individual's physical appearance enables Sam to identify her as an alien from another planet, he does not know anything about either this kind of alien or the planet from which she comes. Now, Sam eventually loses sight of the alien, but he is able to recover the book that she dropped. Upon opening it, he immediately notices that it appears to be written in English and looks like what we on Earth would call a diary. Moreover, after reading the first sentence of the book, Sam forms the corresponding belief that tigers have eaten some of the inhabitants of the author’s planet during their exploration of Earth. It turns out that the book is a diary, the alien does communicate in English, and it is both true and reliably written in the diary that tigers have eaten some of the inhabitants of the planet in question. Moreover, Sam is not only a properly functioning recipient of testimony, he is also situated in an environment that is suitable for the reception of reliable reports. (2008, 168-169)

Jennifer Lackey on Non-Reductionism: A Critique

259

The second example is meant to show that the case against non-reductionism does not hinge on aliens specifically. Lackey does not give the second example a name; I shall call it “SALLY” in honour of its main character: SALLY: Sally has been in a coma for the past two months and, upon waking, discovers that she has lost all of her previous knowledge except for her competence with the English language. Upon leaving the hospital, she stumbles upon a diary of an unknown author and begins reading it. Now, ex hypothesi, Sally no longer has common-sense beliefs about human psychology, she no longer has beliefs about the general reliability of humans as testifiers, she no longer has beliefs about how diaries function in our society, and so on. Is Sally justified in accepting the contents of the diary? (2008, 174)

Lackey claims that these two examples refute non-reductionism. As she sees it, the non-reductionist is committed to saying that Sam’s or Sally’s testimonial beliefs (on the basis of the respective diaries) are in the extension of the predicate “justified or warranted testimonial belief”. According to Lackey, our intuitions about the two cases contradict the nonreductionist on this point: Sam’s and Sally’s testimonial beliefs are part of the mentioned extension only if Sam and Sally have positive reasons for their testimonial belief. Sam’s testimonial belief for example is justified only if he has positive reasons ruling out the possibilities that the aliens do not after all (despite appearance) speak English, that their folk psychology is totally different from ours, or that their law statues look like our diaries (2008, 169-173). Lackey also offers a critique of the Infant/Child Objection. She begins by rejecting the idea that while infants and young children are capable of grasping defeaters, they are unable to grasp positive reasons. Defeaters are, Lackey insists, simply negative reasons. And since reasons are reasons, never mind whether they are positive or negative, children either grasp both of them or neither of them. And now the non-reductionist faces a dilemma. Obviously, he cannot opt for “both”, for then the argument against PR-N collapses. But "neither" is not an option either. If infants and young children are unable to grasp reasons of any form, then infants and young children are “a-rational”, that is “completely outside the domain of reasons”, that is, they are epistemically like paper-clips, and unable to have any knowledge or justified beliefs. Thus if the non-reductionist opts for “neither” then he can no longer maintain that infants and young children can have knowledge. And thus again the argument against PR-N collapses (2008, 199-200).

260

Martin Kusch

Finally, Lackey considers the following reply by the non-reductionist. According to this reply “defeaters” can be spelled out “externalistically” as in the following condition: The presence of negative evidence, E, for a subject, S, defeats the positive epistemic status of S’s belief that p just in case E sufficiently supports a proposition indicating that p is either false or unreliably formed or sustained, regardless of whether S has the capacity to appreciate E. (2008, 201)

Lackey presents two objections. The first is that the condition fails to capture the idea that someone who ignores an undefeated normative defeater acts epistemically irrationally. And the second objection is that the typical way of making sense of the expression “the presence of negative evidence, E, for a subject S” refers to S’s cognitive capacities involved in identifying this evidence (2008, 201-3). 4. The Problem of Consistency Having introduced Lackey’s understanding and criticism of nonreductionism at some length, I can now turn to explaining my objections to her views. I have three problems with her position: the problem of consistency, the problem with the classical view of concepts, and the problem of low argumentative resolution. Consistency is a problem insofar as Lackey’s treatment of ALIEN and SALLY jars with the discussion in her later section entitled “The Generalization Problem” (2008, 188-193). The issue is this: if positive reasons are needed for justifying testimonial beliefs, then why are they not also needed for justifying, say, perceptual beliefs? But if they are needed for justifying perceptual beliefs then we obviously face “all the problems facing traditional internalist theories of epistemic justification or warrant, such as infinite regresses, circularity, foundations, and so on” (208, 189). Lackey’s way out goes as follows: ... testimonial beliefs are acquired from persons. Persons, unlike other sources of belief, have all sorts of different intentions, desires, goals, motives, and so on. Some of these desires and goals make it very advantageous to lie, to exaggerate, to mislead, and to otherwise deceive. (2008, 189)

And hence even though we need positive reasons in the case of testimony, we do not need them in the case of perception. Put differently, Lackey’s claim is that for me to identify a linguistic or gestural item as a case of

Jennifer Lackey on Non-Reductionism: A Critique

261

(purported) testimony I need to be able to identify the producer of the item as a person. I think this observation is exactly right, and I do not know of any non-reductionist who would disagree. However, Lackey’s entirely correct insight causes difficulties for her discussion of Sam and Sally. For the example to work as an objection to non-reductionism, Sam must not know anything about the kind of alien he is encountering in the forest. In particular, he must not know anything about the alien’s psychology. Not knowing anything about the alien’s psychology means not knowing whether the alien has beliefs, intentions or desires. But then, by Lackey’s own criteria, Sam does not know whether the seemingly linguistic items produced by the alien are testimony at all – never mind whether they are good or bad testimony. And something similar holds for Sally. Since she has forgotten everything about folk psychology she is not in a position to distinguish between creatures having beliefs and desires, and creatures without these intentional attitudes. In other words, Lackey has given us two cases in which the hearer is unable to determine – without further empirical evidence, or without positive reasons – whether he or she is confronted with a person and thus with testimony. Two different presumptions are involved in Lackey’s examples. According to the first – the presumption of personhood – encountering a seemingly meaningful linguistic or gestural item entitles us to presume that its producer is a person (rather than, say, a machine). According to the second

262

Martin Kusch

– the non-reductionist’s presumption of truth – when we encounter an assertion produced by a person we are entitled to presume that the assertion is true. Lackey’s examples run together the two presumptions and thus fail to properly isolate and test the non-reductionist view. Put differently, the nonreductionist might agree that the hearer needs positive reasons for the belief that he is confronted with a person, and thus with testimony, while still denying that the hearer need positive reasons for being prima facie entitled to his or her testimonial belief. Picture 1 might clarify the distinction. 5. The Problem with the Classical View of Concepts My second objection is that Lackey’s use of her two examples reveals her commitment to an outdated view of concepts. Lackey assumes that we have clear and unequivocal linguistic intuitions about whether Sam’s testimonial belief falls within, or outside of, the extension of “justified or warranted testimonial belief”. And we have these clear intuitions, Lackey thinks, regardless of the highly unusual scenario of meeting and identifying aliens in the forest, or of being similar to an autistic person in not understanding human folk psychology at all. Picture 2 sums up the central features of Lackey’s implicit understanding of words and concepts. For Lackey non-reductionism is an attempt to offer an explicit intension of our intuitive extension for “justified or warranted testimonial belief”. And

Jennifer Lackey on Non-Reductionism: A Critique

263

this extension knows of no distinction between central and marginal cases, no distinction between cases that are prototypical and cases that are not. Lackey’s view fits nicely with what is sometimes referred to the “the classical view of concepts”, for example in Gregory L. Murphy’s The Big Book of Concepts (2002). According to the classical view, every (conceivable) object is either inside or outside of given categories or extensions; there are no borderline cases; all category members are equally good instances of the category; there is no divide between typical and atypical cases; and there are necessary and sufficient conditions for membership in the category, that is, there are no family resemblance concepts (Murphy 2002, 15-16). Unfortunately, the classical view of concepts has been proven untenable by forty years of psychological and linguistic research “... so that it is not now the theory of any actual researcher in this area ...” (2002, 16). For instance, actual researchers agree that almost all concepts have borderline cases. This is indicated by the facts that for many items people are unsure whether they belong within, or outside of, the category, or that people disagree both with one another and with themselves over time. There is general agreement only over typical – frequently occurring – cases. To bring out the contrast as clearly as possible, Picture 3 represents how the contemporary researchers would modify Lackey’s implicit semantic understanding.

264

Martin Kusch

Consider Lackey’s examples ALIEN and SALLY in light of the distinction between the classical and the non-classical view of concepts. It seems to me a very questionable methodology to try to refute a widely held view such a non-reductionism on the basis of such a-typical borderline cases. ALIEN and SALLY are precisely the kinds of cases con-cerning which intuitions are instable and highly variable. I can only report that I fail to have any clear intuitions about these examples. Of Sam we are told that he knows nothing about the aliens even though he is able to identify them as aliens. My mind goes blank already at this point: I cannot make sense of this possibility. Or how are we supposed to imagine the alien? Like Mr. Spock or Mr. Data from Spaceship Enterprise? Or in the shape of a photocopier? Or as a pile of slime? I find that my intuitive judgements about the case are highly sensitive to such differences. I also do not to have firm intuitions on what to say about the abilities of, and demands upon, autistic people when they receive testimony. Sally obviously falls in that category. I do not think that my (lack of) intuitions about ALIEN and SALLY are in any way unique; the students in my courses on Lackey’s book share my misgivings. Another way to bring out the atypical character of Lackey’s scenarios is to think about our response if the alien were replaced by an ordinary human. Under those circumstances the error possibilities on which Lackey focuses – does the alien speak English or Twenglish, is he psychotic, does he have intentions and beliefs – could safely be ignored: they would be too far from the actual world. Only the sceptic would take them seriously. Lackey’s alien is a device to turn the sceptical error possibilities into nonsceptical error possibilities. But this move does not alter the fact that the scenario is very far from the actual world. 6. The Problem of Low Argumentative Resolution I am struck by the fact that Lackey does not discuss in detail any specific non-reductionist proposal; for instance, John McDowell and Tyler Burge are given no more than half a dozen lines each, and the considerable differences between their proposals, and theirs and others’ positions, are all ignored. Instead Lackey chooses to give us an account of what she thinks the non-reductionist ought to be saying. I fear that this methodology leads to a low level of argumentative resolution: the specific argumentative resources of specific non-reductionist theories are set aside and ignored. In order to

Jennifer Lackey on Non-Reductionism: A Critique

265

make this accusation stick, I shall briefly explain how Burge might respond to Lackey’s criticism. What would Burge say about ALIEN and SALLY? This question is difficult to answer since Burge is sensitive not only to the distinction between typical and atypical scenarios, but also to the distinction between the presumptions of personhood and of truth. In order to overcome this distance between Burge and Lackey, we need to adjust ALIEN: assume that Sam is entitled to the thought that the aliens are persons, and that they thus are able to produce instances of (true or false) testimony. In that case, Burge maintains, we have the strong intuition – linked to a norm of our practices of interlocution – that Sam is prima facie entitled to the testimonial belief that he forms on the basis of the text dropped by the alien. And this even though Sam has no positive empirical reasons that allow him to rule out the possibility of lies or incompetence on part of the source. The norm of our practices of interlocution Burge is referring to is the "acceptance principle”. The following version is from his 1997 paper (the brackets are his): The Acceptance Principle is: A person is [a priori] entitled to accept a proposition that is [taken to be] presented as true and that is [seemingly] intelligible to him, unless there are stronger reasons not to do so. (Burge 1997, 45)

Burge tries to reconstruct a rationale for the Acceptance Principle along roughly the following line; he tries to make plausible that we accept the following licensing of presumptions: If a linguistic item seems like an indicative sentence, then (in the absence of counterevidence) one is prima facie entitled to presume that the item is an indicative sentence. If one is prima facie entitled to presume than a linguistic item is an indicative sentence, then one is prima facie entitled to presume that it is an assertion. If one is prima facie entitled to presume that a linguistic item is an assertion, then one is prima facie entitled to presume that it comes from a rational source. If one is prima facie entitled to presume that a linguistic item comes from a rational source, then one is prima facie entitled to presume that it is true. Thus Burge would say that we have the intuition that – provided the aliens are persons – Sam has a defeasible prima facie entitlement to his testimonial belief (i.e. that the aliens were eaten by tigers). Lackey thinks her example “clearly” shows that Sam does not have any such entitlement. But here we have no more than a standoff of intuitions: at best Lackey has pitted intuition against intuition. I share Burge’s intuition, and I am not

266

Martin Kusch

alone. It is interesting to note that most philosophical writers on the issue are with Burge and not with Lackey – and even when they do not share his reconstruction the non-reductionist intuitions. Lackey herself lists Austin, Coady, Dummett, Evans, Foley, Goldman, Goldberg, Hardwig, McDowell, Owens, Plantinga, Reid, Ross, Schmitt, Sosa, Strawson, Weiner, Welbourne, Williamson (Lackey 2008, 155). Of course I am not saying that the issue should be decided by number of votes. I am using the list of philosophers simply to make plausible that Burge’s intuition is pretty widespread and not easily dismissed as an idiosyncrasy. No doubt, it is possible that the aliens do not speak English but Twenglish, that they are psychotic, or that their law statues look like our diaries. But the question is whether one has to rule out all of these possibilities to be even prima facie entitled to take the report as true. Burge’s position can also be used to present the Infant/Child Objection in a more favourable light than Lackey allow for. Here too Burge would start from the observation that he has the intuition according to which infants and young children can gain testimonial knowledge and entitlements to testimonial beliefs even in the absence of abilities to consider counterevidence. Burge takes it to be the case that his intuition is linked to a norm implicit in our practice of interlocution: Children may not conceive of counterconsiderations as possible. It is normally enough that there be no available counter reasons. In such cases, they may be a priori entitled to accept what they are told. (1997, 23)

As we have seen, Lackey insists that we do not have this norm or intuition. I can only report that my intuitions and those of my philosophically-trained and untrained informants are with Burge. But could Lackey not argue that Burge’s intuition, though widespread, is better given up since it is in tension with other, more important, intuitions such as the intuition that paperclips cannot be knowers? Indeed, we might read Lackey in that way. But if we do, we should note that on Burge’s specific proposal infants and young children are not outside the domain or space or reason, that is, children are not epistemically like paper clips. The crucial difference is that children have minds: … the role of the mind … is relevant in this way: it makes possible an a priori, conceptual, connection between seeming understanding of a putative assertion, and being rationally entitled to form beliefs. (1997, 33)

Jennifer Lackey on Non-Reductionism: A Critique

267

For Burge the infant or young child is in the space or domain of epistemic reasons in virtue of being someone who is able to understand linguistic items as instances of testimony. This capacity of understanding is conceptually linked to the capacity for reason. Or put differently, the young child is in the space of reasons insofar as she can understand – and is dependent upon – linguistic and gestural items coming from a rational source. This may or may not be the correct view – though I think it comes close. But the important point is that Lackey offers no consideration that speaks against this view.1 References Burge, T., 1993: Content Preservation, Philosophical Review 102, 457-488 Burge, T., 1997: Interlocution, Perception, and Memory, Philosophical Studies 86, 2147 Lackey, J., 2008: Learning from Words: Testimony as a Source of Knowledge, Oxford: Oxford University Press Murphy, G. L., 2002: The Big Book of Concepts, Cambridge, Mass.: MIT Press

1

For critical comments and discussions I am grateful to Jennifer Lackey, Veli Mitova, Sandy Goldberg, Cheng Kai-Yuan, Adela Roszkowski and Sebastian Kletzl.

V. Disagreement

“A Troubled Area”. Understanding the Controversy over Screening Mammography for Women Aged 40-49 MIRIAM SOLOMON, Temple University

Recommendations for screening mammography have been controversial ever since the technology was developed in the late 1960s. Despite the fact that mammography has been more extensively evaluated by randomized trials than any other screening method (Wells 1998, 1224-1229), it continues to be controversial, especially for routine use in women aged 40-49. The length and persistence of controversy – over forty years, with the same issues frequently rehashed – is extraordinary. The Canadian Preventative Services Task Force, re-established in 2010, selected screening mammography as its first topic for discussion, indicating its continuing importance. Currently, in the USA1, the American Cancer Society, the National Cancer Institute, the Society for Breast Imaging and the American College of Radiologists all recommend annual screening starting at age 40, while the United States Preventative Services Task Force (USPSTF) and the American College of Physicians recommend against routine mammography for women aged 40-49. These recommendations are made by groups of experts – physicians, researchers, statisticians – who typically construct a consensus statement after examination and discussion of the evidence. Judgments on the topic differ sharply. Michael Baum, a breast surgeon from the UK, writes that screening mammography is “one of the greatest deceptions perpetrated on the women of the Western World” (quoted in (Ehrenreich 2001, 43-53)), while the American College of Radiology has an statement on its website about the “ill advised and dangerous USPSTF mammography recommendations” which claims that “The [USPSTF] recommendations make unconscionable decisions about the value of human life” and predicts that the recommendations will lead to more women dy1

There is similar controversy in other developed countries, but I am most knowledgeable about the USA.

272

Miriam Solomon

ing of breast cancer.2 In response to such comments, Jane Wells, a British public health physician, astutely states “The debate over the necessity of screening for breast cancer among women in their 40s has assumed an importance out of proportion to its potential impact on public health” (Wells 1998, 1224-1229). I also like Vivian Coates’s evenhanded assessment that mammography recommendations are “a troubled area,”3 and use it for the title of this essay. Expert consensus statements are intended to speak with the authority of a group, and thereby have persuasive force. The institution of consensus conferences began at the National Institutes of Health (NIH) in 1978, and has been widely adopted and adapted. Consensus conferences play an important role in creating authoritative knowledge in medicine (Solomon 2007; Solomon 2011). However, when there is more than one consensus statement on a topic, and the statements disagree with each other, the authority of one or more of them – usually all of them – is undermined. While expert agreement is often taken as a marker of truth or objectivity, expert disagreement is typically explained as the consequence of some failure(s) to be objective. For example, last year’s New England Journal of Medicine article by Quanstrum and Hayward (2010, 1076-1079) suggests that self-interest of the community of radiologists lies behind their recommendations for more frequent and earlier mammograms. The article uses this suggestion to discredit the recommendations of the Society for Breast Imaging and the American Society of Radiologists. This is a typical scientific reaction to peer disagreement: to attribute conflicting positions to nonnormative “social factors” while maintaining that one’s own position is unbiased.4 Quanstrum and Hayward’s analysis has some truth to it, as a critique of the position of radiologists. But it does not go very far to explain and assess a complex epistemic situation in which primary care practitioners, oncologists, patients, public health experts, epidemiologists, politicians, 2

http://www.acr.org/MainMenuCategories/media_room/FeaturedCategories/ other/PositionStatements/UPSTFDetails.aspx accessed May 31, 2011. 3 Vivian Coates is ECRI Institute’s Vice President for Information Services and Health Technology Assessment. ECRI Institute is located in Plymouth Meeting, Pennsylvania. The comment “a troubled area” was made in an e-mail communication with me, September 21, 2010. 4 I call this the “Me Hempel, you Kuhn” reaction.

Understanding the Controversy over Screening Mammography

273

and health insurers have different perspectives both as groups and as individuals. We understand very little of the debate when we analyze only the position of radiologists. In this paper I am going to explore several sources of the disagreements over screening mammography for women aged 4049. It turns out that multiple methods, as much as self-interested motives and other “biasing factors,” are responsible for the professional dissent. This means that even if everyone was “perfectly objective,” (if that was possible), we would still have some expert disagreement resulting from methodological variability. I have often argued that dissent is valuable in science, largely because of the division of labor that results from scientific disagreement (Solomon 2001; Solomon 2006, 23-36; Solomon 2006, 28-42).5 However, in the case of medicine and other publicly relevant science and technology, the epistemic benefits of dissent are offset by the political and rhetorical costs of expert disagreement. Nonscientific claims to knowledge (such as those produced by self-interested corporations, or political or religious agendas) can gain power – rhetorically speaking – when scientists are in disagreement. Logically speaking, scientific disagreement should not make the claims of cigarette companies, climate change deniers or intelligent design theorists any stronger than they actually are. The proper response to scientific disagreement is to do more science. But practically (and rhetorically) speaking, the response to scientific disagreement has been to give more credence to other kinds of claims. Hence, the consensus statements of the Intergovernmental Panel on Climate Change (IPCC) are important not for science but for the epistemic authority of science. Unless science has epistemic authority, its recommendations for co-operative human actions – from regular mammograms to environmental treaties – are unlikely to be taken seriously. When medical experts disagree about practice recommendations, other voices – such as political voices, financial interests, and so forth – play more of a deciding role. In the next section I will give an overview of the screening mammography controversy. There is more scientific uncertainty in this case than in the climate change case, in which the uncertainty is intentionally “manu5

Other philosophers of science, such as Helen Longino (1990), argue that dissent is valuable largely because of the mutual criticism and refinement of views that it makes possible.

274

Miriam Solomon

factured” by special interests outside of science (Oreskes and Conway 2010). That is why I selected the screening mammography case for discussion: it is a case of scientific (= “real”), rather than manufactured uncertainty.6 My approach will be to strive for what I call “normative equipoise,” avoiding as far as possible any a priori favoring of one method, or position, over another. I shall be looking for the reasons and causes of the existing controversy while also thinking about why the controversy is not greater – or lesser – than it is. It turns out, I will argue, that some controversy is avoided by implicit negotiation. At least four different methods are used in making judgments about the value of screening mammography: evidence-based medicine, pathophysiological reasoning, consensus of experts and “clinical judgment.” I go through these in turn to see what they deliver: 1. Evidence-based medicine results Nine randomized controlled trials of screening mammograms have been performed in the United States, Canada and Europe, beginning in the early 1960s. Most of the trials, especially in the early years, have methodological flaws, ranging from post-randomization exclusions to inadequate randomization to unblinded cause of death attributions (Goodman 2002, 363-365). The most recent trial, called the Age trial (conducted in the UK, begun in 1991 and reported in 2006), was designed to look specifically at the question of the efficacy of mammography in women under 50. In all the studies, any benefit of mammographic screening is small, even for women over age 50. The typical effect found is a reduction in breast cancer mortality of about 20% in women aged 50-75. The most recent Cochrane review (Gotzsche and Nielsen 2011, CD001877) estimates that approximately 2000 women need to be screened annually for ten years to prevent one “premature death” from breast cancer; this amounts to a life extension of two days, on average, for every woman who undergoes ten years of mammography). For women under 50, the effect size is smaller still, with the Canadian National Breast Screening Study (Miller et al. 2002, 305-312)

6

Cases of “manufactured uncertainty” have much to teach us about the ways in which scientific authority works; they employ these in order to achieve their goals.

Understanding the Controversy over Screening Mammography

275

and the recent Age trial showing no statistically significant benefit in either breast cancer or overall mortality (Moss et al. 2006, 2053-2060). When effect size is small or uncertain, meta-analysis is sometimes used to combine the results of several trials in order to discern any overall effect. A meta-analysis needs to make principled decisions about which trials to include, and which to exclude, based on trial quality. But there is no universally accepted set of such principles.7 There is a “gray zone” in which judgments of reasonableness differ, and the benefits of routine mammography for women aged 40-49, if any, fall in this gray zone. This is why some meta-analyses show a small benefit, and others show no benefit. Steven Goodman (2002, 363-365) concludes that the lack of a univocal answer constitutes a “crisis for evidence-based medicine”; he thinks that EBM should be able to deliver an exact and unambiguous result. Another way of looking at the situation is to say that EBM is not as fine grained a tool as some people would like it to be. The data are what I call “plastic;” I mean by this that they are dependent on the standards used for including a study in a meta-analysis. Recommendations for health care interventions need to take into account harms as well as benefits. Screening mammography produces harms that are uncertain and difficult to quantify. The greatest concern is that the small cancers often detected by screening mammography are less harmful in terms of both quantity and quality of life than the treatments used to eradicate them. It is well known that the incidence of mastectomies and lumpectomies is 30% to 40% higher among women screened by mammography without a corresponding increase in survival, suggesting that some DCIS and even invasive cancers would never become clinically relevant. The most recent Cochrane review estimates that 10 women will be treated unnecessarily for every life saved (Gotzsche and Nielsen 2011, CD001877). Surgery, radiation treatments and chemotherapy produce both short term and long term harms. Even false positive mammography results – there are more than 200 false positives for every life saved – cause emo7

There is no universally accepted evidence hierarchy and there is no universal standard for the minimum quality of a trial necessary for use in a meta-analysis. There are attempts to come to a practical consensus on this (e.g. the GRADE working group). Even if we get this, there is still individual variability in judgments about how to apply the standards in particular cases.

276

Miriam Solomon

tional distress and sometimes unnecessary biopsies. On average, half of all women who get annual screening mammography will receive a false positive result in ten years. Harms and benefits can be framed differently, evoking different preferences. For example, the Cochrane report frames the benefits of mammography in terms of average life extension (2 days for every 10 years of mammography), but others might frame the benefit of mammography in terms of avoidance of an unlikely but great harm, premature death (one person will lose 30 or more years). Sometimes it helps to find an analogous case in which our assessments of risk are more settled and we choose to either accept a small risk (e.g. we choose not to put seat belts in American school buses, even though that increases the small risk of premature death) or to pay the price of preventing an unlikely but catastrophic outcome (e.g. we choose to pay the cost of tamper-resistant packaging for over-thecounter drugs). 2. Pathophysiological Reasoning When mammography was invented, cancer was understood as a progressive disease in which malignant cells multiply at first locally and then distantly (as metastases). The prognosis was thought to be better the earlier that the malignant cells are discovered and then eradicated by surgery, radiation or chemotherapy. The mantra “early detection saves lives” has been widely embraced and has evidence in its favor, especially for some types of cancer such as melanoma, colon cancer, cervical cancer and testicular cancer. According to this reasoning, there is an expectation that screening mammography will save lives, since it detects cancers too small to be felt in a clinical breast examination, and thus presumably at an earlier stage of development. This expectation was so strong that when in the early years of the Canadian National Breast Screening Study the data appeared to show that 40-49 year old women had more deaths from breast cancer than controls who were not screened, the result was dismissed as a statistical fluke (Baum 2004, 66-67; discussion 69-73). Since the 1960s and 1970s, we have learned that cancer is a more complex class of diseases, and that early detection does not always save lives because even in early stages cancer can be a systemic disease. Prognosis is affected by the type of cancer cell and by the host immune response to the cancer. One well-respected (although somewhat maverick) breast cancer

Understanding the Controversy over Screening Mammography

277

surgeon, Michael Baum, has extended the angiogenesis work of Judah Folkman and suggested that mammography causes breast cancers, not only through exposure to radiation but also through the unforeseen effects of biopsy of suspicious lesions. Baum suggests that the injury caused by biopsy produces increased blood flow and stimulates angiogenesis in the area, and that this nourishes the cancer and enables it to spread. In the current state of knowledge about breast cancer and its mechanisms, this is a reasonable suggestion. We do not have any evidence for it at this point. But recall that we do not have evidence that screening mammography saves lives in women aged 40-49, either. 3. Consensus of experts Expert consensus is frequently taken as evidence for a position, the reasoning being that if those most knowledgeable about a topic agree, then any lay disagreement lacks credibility. As I argue elsewhere (Solomon 2011, 451-466), I think that this is a contemporary version of the traditional argument from authority. We have switched from using past intellectual giants as sources of authority to using groups of up-to-date expert professionals. In the 1970s the NIH developed a social epistemic institution for producing expert consensus knowledge, naming it the NIH Consensus Development Conference Program. This program became the model for consensus conferences worldwide. The NIH Program is distinguished by its requirement for neutral panelists; this supports its claims to having an objective process. Panelists are required to be free of both financial and intellectual bias. Panelists who have prior publications on the topics of discussion are judged to have intellectual bias. (Not all consensus conference programs have this requirement of freedom from intellectual bias.) In 1997 an NIH CDC was convened on precisely the question of recommendations for screening mammography in women aged 40-49. It reached the surprising (to the planners) conclusion that such screening offers no benefits. Interestingly, two of the panelists changed their minds after the conference, and the written record shows a minority opinion dissenting. It is very rare for NIH CDCs to fail to reach unanimous agreement (only three times in well over a hundred such meetings). However, it is also well known from the early days of NIH CDCs that they are not good at resolving controversy; they are much better at reflecting general agreement (Wortman, Vinokur, and Sechrest 1988, 469-498).

278

Miriam Solomon

More recently, as mentioned at the beginning of this paper, different expert groups have produced conflicting recommendations for screening mammograms in women aged 40-49. Thus, expert consensus does not help resolve the uncertainty. If anything, it intensifies it (“if groups of experts disagree with each other then we really don’t know”). 4. Clinical judgment After the 1997 NIH CDC in which consensus was not sustained, John Ferguson, the Director of the Program, wrote an interesting piece about the “great divide” between what he calls “curative medicine” and what he calls “population medicine” (1999, 111-119). “Curative medicine” is the perspective of the clinician who treats patients individually. “Population medicine” is the perspective of the epidemiologist. Ferguson claims that those who have these different perspectives have difficulty understanding each other. This is a polite way of framing his actual opinion that nonepidemiologists don’t understand epidemiology. In particular, he thinks that clinicians and the public have difficulty with understanding the mathematics of statistics and probability theory. He also thinks that clinicians and the public focus on sick people while epidemiologists are often concerned with healthy populations. Ferguson observes that the two panelists who declined to endorse the consensus statement were practicing clinicians. (According to the published conference statement, they are a radiologist and an obstetrician/gynecologist.) Ferguson makes a point that has been made in a number of different ways, and usually from the perspective of the clinician. My own primary care physician believes that epidemiologists are young Washington bureaucrats without clinical experience who are not in a position to make judgments about how to take care of individual patients.8 Her view (perhaps minus the claims about youth and government!) is widely shared by clinicians. “Clinical judgment” is a common although somewhat vague term that generally refers to particular judgments about actual clinical cases, based on training and extensive clinical experience. Clinical judgment has a number of well known biases, such as salience bias (the frequency of salient cases, such as cancers apparently caught by mammography, is often 8

Conversation with Dr Margaret Lytton, December 14, 2011.

Understanding the Controversy over Screening Mammography

279

overestimated), the aggregate fallacy (belief that one’s patients are atypical, which leads to overruling clinical practice guidelines), commission bias (better to do something than to do nothing), and the tendency to overlook iatrogenic harms. Defensive medicine may also play a role in clinician’s recommendations for screening mammography. Also, patient preferences often influence screening decisions. Clinical judgment is thus often a source of arguments for recommending screening mammography for women aged 40-49. I have not come across cases in which clinical judgment has been used to argue the other side, although I think it is worth considering that it would not take much to produce such a case. Here’s a plausible scenario in which a clinician might use “clinical judgment” to recommend against screening mammography for women aged 40-49. The clinician has several patients in the 40-49 age group who receive false positive screening tests. They undergo expensive work-ups and take up the clinician’s time with requests for reassurance and explanations of risk. They show considerable distress during the period of uncertainty. The clinician has no experience of true positive screening tests in women aged 40-49. Both Ferguson and I have been critical of the perspective of the clinician, Ferguson pointing out that clinicians sometimes do not understand statistics and probability theory, and me pointing out some of the typical biases in clinical decision making. For balance, we need acknowledge the biases and errors that are frequent in epidemiological approaches.9 I have discerned four different ways of thinking about screening mammography: the perspectives of evidence-based medicine, pathophysiological reasoning, consensus conferences and clinical decision making. In practice, arguments for or against screening can appeal to more than one of these ways, for example, clinicians might use pathophysiological reasoning, clinicial judgment, and the recommendations of the American Cancer Society to support their recommendations to patients. Or, for example, epidemiologists might use pathophysiological reasoning (with different theories of cancer development), evidence based medicine, and the recommendations of the Cochrane Collaboration to support their recommendations. None of these four ways of thinking, alone or in combination, gives an unequivocal result. 9

I have summarized some of these in (Solomon 2011, 451-466).

280

Miriam Solomon

I have omitted at least two important factors in this paper, which focuses on expert disagreement. The first is the perspective of the patient, and the frequent construction of patient narratives in which screening mammography is given a life-saving role.10 The second is the political atmosphere around breast cancer management, which is seen as the touchstone of women’s health care. Perceived reductions in effort to detect breast cancer easily produce distrust. Both of these factors can also affect experts (i.e. physicians and scientists), although they are not “methods,” strictly speaking. (In (Solomon 2001) I refer to them as “decision vectors.”) The case of screening mammography is unusual in that no consensus has been reached on clinical recommendations. Consensus on clinical recommendations is more typical for health care interventions. Why? Evidence-based medicine, pathophysiological reasoning, consensus conferences and clinical judgment are methods used throughout medicine, and surely it is to be expected that different methods, or even different applications of the same method, may lead to different conclusions. So why is the dissent that we find over screening mammography not more common? I will approach this question by asking another question: why isn’t there even more dissent over screening mammography? I am struck less by the dissent than by the reflection that there could be much more dissent than there is. I will suggest that conflict and uncertainty are “managed,” in general, by social processes that reduce or minimize them. In many contexts, this results in consensus, but in the case of mammography there is still some dissent. Here are some examples of the ways in which dissent is limited in medicine, with screening mammography as a prime example: A. Reappropriating areas of uncertainty as areas of personal preference and clinical judgment It is common to exploit the plasticity of the data and the uncertainty about harms as a locus for the “individualized” judgments that support both physician autonomy (“clinical judgment”) and patient self-determination. 10

A recent article in the Archives of Internal Medicine (Welch and Frankel 2011, 2043-2046) estimates that chance that mammographic detection of a particular cancer has saved a life is around 13%.

Understanding the Controversy over Screening Mammography

281

Hence, we find a repeated recommendation that each woman discuss with her physician whether or not to have screening mammography in her 40s. This turns the area of epistemic uncertainty into an area of personal judgment and preference, in my view, making a virtue out of the uncertainty by using it as an area in which to bow to “patient centered medicine” and physician autonomy. It also, subtly, encourages confining the area where personal judgments and preferences make a difference to women in their 40s. But personal judgments and preferences could be used for women in their 50s and beyond, to weigh benefits and risks for each woman. For example, for some women the anxiety associated with false positive results may be severe, and for them the potential harms of mammography will be greater. The magnitude of benefits and risks is not the same as the magnitude of uncertainty in the data. However, this expansion of patient and physician discretion is not encouraged by any party to the debate. In my view, there is a compromise here between the desire to be guided by the evidence and the desire to acknowledge both clinical judgment and patient selfdetermination. A “gray area” is defined where guidelines are not governing. Quanstrum and Hayward are a particularly clear example of endorsement of this strategy when they write of “…a gray area of indeterminate net benefit, in which clinicians should defer to an individual patient’s preferences...” (Quanstrum and Hayward 20101077). The strategy is not limited to the case of mammography; similar recommendations have been made recently for prostate cancer screening with the PSA test. B. Maintaining trust by drawing no fiscal consequences It is typical for guidelines that do not recommend routine screening mammography for women aged 40-49 to insist, at the same time, that such mammography be covered by health insurance or government reimbursement programs. For example, the 1997 Consensus Development Conference statement says, “For women in their forties who choose to have mammography performed, the costs of the mammograms should be reimbursed by third-party payers or covered by health maintenance organizations so that financial impediments will not influence a woman’s decision.” Without such a statement (and perhaps even with it) dissent might be amplified by distrust.

282

Miriam Solomon

C. Producing guideline syntheses that minimize or omit dissent The National Guidelines Clearinghouse (NGC), managed by the Agency for Health Care Research and Quality (AHRQ), lists up-to-date clinical guidelines that are produced by a recognized organization and include an examination of relevant scientific evidence. Sometimes, this results in listing guidelines that conflict with one another; in these cases NGC often publishes a “Guideline Synthesis” that addresses the areas of dispute. For screening mammography, NGC currently includes the American College of Physicians 2007 guideline, the US Preventative Services Task Force Guideline from 2009, the American Cancer Society 2007 guidelines and the American College of Obstetricians and Gynecologists 2006 guideline. The first two guidelines are in agreement that routine screening mammography should not be offered to women aged 40 to 49 and the second two guidelines are in agreement that it should not be offered. However, the “guideline synthesis” includes only the first two guidelines, and thus presents more consensus than exists. I sent an inquiry to the NGC, asking why other guidelines were not included in the synthesis. Their response was that the other guidelines are not up-to-date with the evidence. I sent another query, asking why criteria can be available on the website yet not used for guideline syntheses. This query did not receive a response.11 Obviously, if guidelines are dropped from syntheses, it will look like there is less dissent. When uncertainty is “managed” in these ways, there is more unanimity and, along with that, more epistemic authority. In other cases that I have looked at, there is even more unanimity. One way in which this is regularly achieved is by holding consensus conferences after systematic review and meta-analysis; in this way the consensus conferences tend to be constrained by the technical results. All NIH CDCs, as well as several other consensus programs, are now preceded by a formal evidence report. This process did not work to produce consensus in the mammography screening case at least in part because of the plasticity of the evidence in this case.12

11

My contact was Vivian Coates at the ECRI Institute, which manages the NGC for AHRQ. Our communications were by e-mail in Fall 2010. 12 Other causes may include belief perseverance phenomena; positions were staked out a long time ago.

Understanding the Controversy over Screening Mammography

283

In conclusion, the case of screening mammography is a case of genuine (techno)scientific controversy (rather than manufactured controversy). Methodological pluralism (different applications of the same method as well as different methods) contributes to the controversy. The disagreement is “managed” by various strategies; this distinguishes clinical medicine from more “pure” research contexts in which disagreement is typical and tolerable. Peer disagreement is epistemically good for science, but rhetorically risky for the authority of science. In the case of medicine, minimizing rhetorical risk has outweighed the benefits of disagreement. Even then, disagreement cannot always be managed away.13 References Baum, M., 2004: Commentary: False premises, false promises and false positives – the case against mammographic screening for breast cancer, International Journal of Epidemiology 33, 66-67; discussion 69-73 Ehrenreich, B., 2001: Welcome to cancerland. Harper‘s Magazine Ferguson, J.H., 1999: Curative and population medicine: Bridging the great divide, Neuroepidemiology 18, 111-119 Goodman, S.N., 2002: The mammography dilemma: A crisis for evidence-based medicine? Annals of Internal Medicine 137, 363-365 Gotzsche, P.C. and M. Nielsen, 2011: Screening for breast cancer with mammography. Cochrane Database of Systematic Reviews (Online), CD001877 Longino, H.E., 1990: Science as social knowledge: Values and objectivity in scientific inquiry. Princeton, N.J.: Princeton University Press Miller, A.B., T. To, C.J. Baines, and C. Wall, 2002: The Canadian national breast screening study-1: Breast cancer mortality after 11 to 16 years of follow-up. A randomized screening trial of mammography in women age 40 to 49 years, Annals of Internal Medicine 137, 305-312 Moss, S.M., H. Cuckle, A. Evans, L. Johns, M. Waller, L. Bobrow, and Trial Management Group, 2006: Effect of mammographic screening from age 40 years on breast cancer mortality at 10 years’ follow-up: A randomised controlled trial, Lancet 368, 2053-2060

13

I am grateful to audiences at the BiCoDa conference at Bielefeld in June 2011, the Wittgenstein conference in Kirchberg in August 2011 as well as to Robyn Bluhm, Vivian Coates, Gretchen Condran, Julia Ericksen, Marya Schechtman and Susan Wells for helpful discussions on the material in this paper.

284

Miriam Solomon

Oreskes, N. and E.M. Conway, 2010: Merchants of doubt: How a handful of scientists obscured the truth on issues from tobacco smoke to global warming, New York: Bloomsbury Press Quanstrum, K.H. and R.A. Hayward, 2010: Lessons from the mammography wars. The New England Journal of Medicine 363, 1076-1079 Solomon, M., 2011: Group judgment and the medical consensus conference, in: Elsevier handbook on philosophy of medicine, ed. F. Gifford, London: Elsevier Solomon, M., 2011: Just a paradigm: Evidence based medicine meets philosophy of science, European Journal of Philosophy of Science 1, 451-466 Solomon, M., 2007: The social epistemology of NIH consensus conferences, in: Establishing medical reality: Methodological and metaphysical issues in philosophy of medicine, ed. H. Kincaid and J. McKitrick. Dordrecht: Springer Solomon, M., 2006: Groupthink versus the wisdom of crowds: The social epistemology of deliberation and dissent, The Southern Journal of Philosophy 42 (Supplement), 28-42 Solomon, M., 2006: Norms of epistemic diversity. Episteme 3, 23-36 Solomon, M., 2001: Social empiricism. Cambridge, Mass.: MIT Press Welch, H.G. and B.A. Frankel, 2011: Likelihood that a woman with screen-detected breast cancer has had her “life saved” by that screening, Archives of Internal Medicine 171, 2043-2046 Wells, J., 1998: Mammography and the politics of randomised controlled trials, BMJ (Clinical Research Ed.) 317, 1224-1229 Wortman, P.M., A. Vinokur, and L. Sechrest, 1988: Do consensus conferences work? A process evaluation of the NIH consensus development program, Journal of Health Politics, Policy and Law 13, 469-498

Is Philosophical Knowledge Possible?1 HILARY KORNBLITH, University of Massachusetts

No reasonable person can doubt that we have a great deal of knowledge. In particular, the achievements of science over the past few hundred years show progress of three very striking sorts: we are able to make predictions which are both more accurate and more wide ranging over time; we have better and better explanations for a wider and wider range of phenomena; and we have ever more elaborate technological innovations. There is a single underlying explanation for these progressive features of science: our theories tend to be at least approximately true, and the enterprise of science, the greatest intellectual achievement of the human species, thus provides us with an ever expanding body of knowledge of the world around us. The achievements of philosophy, by comparison, may seem rather paltry. What I have in mind here is not that philosophy has failed to yield explanations or predictions of physical phenomena, let alone that it has failed to produce exciting technological innovations. Philosophy does not aspire to such things, and its success or failure cannot be measured by such standards. Philosophy does, however, aspire to achieve knowledge and understanding – or so it seems – and there is reason to worry that, despite its long and illustrious history, and despite the many great minds which have been devoted to the philosophical enterprise, we have made little if any progress in answering philosophical questions, and little if any philosophical knowledge is thus to be had. It is not just that philosophy looks bad by comparison with science. It is rather that philosophy looks bad by any measure at all. It is just not clear that the philosophical enterprise has served as a source of knowledge. And if that is so, then there is reason to

1

An earlier version of these ideas was presented at the University of St. Andrews. Discussion on that occasion was very helpful and had a large influence on this paper. I particularly want to thank Jessica Brown, Jesse Prinz, and David Sosa. I have been influenced as well by the excellent discussion in Kirchberg.

286

Hilary Kornblith

worry that philosophy, at least as it has thus far been practiced, has been an intellectual failure. There is no shortage, to be sure, of confidently held opinion among philosophers. I myself have strongly held views on quite a number of philosophical topics, and, in this respect, I am not at all unusual among philosophers. But confidently held opinion, of course, need not amount to knowledge, and it is a noteworthy fact about the state of current opinion in philosophy that there is, beyond doubt, a great deal of diversity of opinion among philosophers. The dynamics of opinion in philosophy looks utterly different from the dynamics of opinion in scientific fields. There is disagreement within the sciences, to be sure. But in the sciences, disagreements tend to be resolved over time. Opinions in the sciences tend to converge, and, as I remarked at the outset, there is a great deal of reason to believe that these opinions tend to converge to the truth, or at least to closer and closer approximations to the truth. In philosophy, however, we see no such convergence of opinion over time. Issues debated at the time of Plato and Aristotle, Descartes, Hume, and Kant, are still debated today. We teach these philosophers not only in history of philosophy courses, but in philosophy courses generally. We do not find a similar pattern in the sciences, with two thousand year old physics books serving to instruct beginning students. The comparison, it seems, is revealing. There has been progress in the sciences. The sciences have been a source of a substantial amount of knowledge. But nothing like this seems to be true of philosophy. I believe that something very much like this is right about the philosophical enterprise, and I want to elaborate on this argument here.2 In addition, I want to consider a number of objections to this line of argument. Much as I wish I could find some satisfactory way of responding to these worries, I do not believe that any of the available responses to them are successful.

2

I first presented a version of this argument in “Belief in the Face of Controversy,” in Richard Feldman and Ted Warfield, eds., Disagreement, Oxford University Press, 2010, 29-52. Richard Fumerton takes a similar line in his contribution to that volume, “You Can’t Trust a Philosopher,” 91-110, as does Sanford Goldberg in “Reliabilism in Philosophy,” Philosophical Studies 142 (2009), 105-117.

Is Philosophical Knowledge Possible?

287

I. Let me begin by considering a common kind of epistemological problem, a problem involving deference to experts. Suppose that Frieda wonders what she ought to believe about global climate change. Frieda is not a climate scientist, and has no expertise in that field. She is, however, an intelligent layperson, and she reads the newspapers; she has an intelligent layperson’s understanding of the issues about climate change. She can explain some of the issues involved here, and she has some knowledge of the relevant evidence, so she can explain the state of play, in rough and ready terms, to someone who is entirely new to the issue. Now consider some claim – call it p – about global climate change. Suppose Frieda knows that the vast majority of climate scientists believe that p is true. It seems to me that this is extremely good reason for Frieda to believe that p. It is not the case, of course, that the majority of scientists in any given field are incapable of error. Nevertheless, in a mature scientific discipline which has a substantial history of progressive results – ever better predictions, explanations, and, where relevant, technological applications – there is extremely good reason to side with the opinion of the majority of reputable experts in the field. Anyone who regularly made bets about scientific claims in this way, and here I’ll assume that the resolution of the bets is decided by an omniscient deity, would occasionally lose some money. But these would be extremely good bets to make, and they would pay off handsomely. Barring special circumstances, Freida would be well advised to believe as the majority of experts believe. There can be situations in which this would not be a good kind of bet to make. If one knows, for example, that the vast majority of scientists have been taking pay-offs from some megalomaniacal billionaire industrialist who is trying to influence global climate policy in ways which will serve to enrich him, then one will have good reason to ignore the opinion of the majority. There are less extreme situations as well in which a simple strategy of following the majority would be unwise. If the field in question is extremely small, and the majority of the qualified members of the field has all been trained by a single charismatic individual, one might have good reason for caution. At a minimum, in cases of this sort, the fact that a majority of a field converges on a certain opinion does not carry the weight it does when the field is large and there is a greater diversity in the training of the members of the field. So the rule to believe as the majority of experts

288

Hilary Kornblith

do is not exceptionless, but it is, ceteris paribus, an excellent bit of epistemic advice. A very highly idealized version of this case is modeled by the Condorcet Jury Theorem.3 If one has a group of individuals each answering a yes-orno question, then if (a) each individual is better than chance at the relevant class of questions; and (b) the opinions of the individuals are each independent of the others, the likelihood of a unanimous opinion on the question being correct approaches 1 as the size of the group increases or the competence of the individuals increases. It is worth playing around with some numbers in order to see just how rapidly the likelihood approaches 1. More than this, one can use the same formalism to examine cases in which there are disagreements within the group. Here, too, one can see how a majority of competent judges, under the conditions described, are very likely to be right; the likelihood increases sharply as the size of the majority increases. The Condorcet Theorem involves, as I’ve said, highly idealized circumstances. In real life cases of the sort Frieda confronts, the individual scientists do not have opinions which are entirely independent of one another. This condition may be approximated in actual cases, but the implications of the Theorem when the idealization is not satisfied are not entirely clear. I have thus avoided giving any precise account of the degree of confidence Frieda should have in the opinion of the majority. Nevertheless, I believe that what I have said thus far about what Frieda should believe should be uncontroversial. The advice to believe as the majority of experts believe, ceteris paribus, is excellent advice. I have said that Frieda should be guided, ceteris paribus, by the opinion of the majority of experts. I have not said anything at all about any attempt Frieda might make to evaluate what one might call “the direct evidence” about global climate change, the evidence which the experts themselves evaluate, leaving aside the facts about expert opinion. So let us examine this issue. I find myself in the very situation I described Frieda as being in. I’ve read the reports in reputable newspapers. I can describe some of the state 3

For a useful discussion of the Theorem and its implications, see Christian List and Robert Goodin, “Epistemic Democracy: Generalizing the Condorcet Jury Theorem,” Journal of Political Philosophy 9 (2001), 277-306.

Is Philosophical Knowledge Possible?

289

of play in this particular field. But I am not, of course, a climate scientist myself, nor have I ever had any training in this field. How should I evaluate the direct evidence on questions about climate change? Suppose, in particular, that some new bit of evidence is discovered that bears on some particular question about global climate change. I read about it in The New York Times, and I think about how it fits in with the various other bits of data I’m familiar with, as well as relevant background theory. When I think about these kinds of issues, it seems to me that this new bit of evidence strongly supports some interesting claim that p. I also discover, on reading the Times, that the vast majority of climate scientists do not think that this new evidence supports the claim that p. Unfortunately, the Times does not explain why they think this. Let me suppose that I should have complete confidence in the claim which the Times makes about the opinions of climate scientists. How should I weigh my own evaluation of the direct evidence with respect to p? It seems to me that I should assign it almost no weight at all. As I said, I am not myself a climate scientist; I have no training in the field. I should recognize that rank amateurs such as I have only the most casual acquaintance with the relevant evidence and the relevant background theories. If I think that the new evidence counts strongly in favor of the truth of p, but the vast majority of climate scientists – that is, the very people who have expert knowledge of the relevant evidence and background theory – think otherwise, then it is overwhelmingly likely that there is something here that I have missed or something that I don’t understand. It would be extraordinarily presumptuous of me to think that I, someone without any training in this field and with only the most casual layperson’s understanding of it, have understood something about this new evidence which the experts have somehow missed. So although I should be interested in understanding why it is that this new evidence doesn’t show what it seems to me to show, I should be very confident that what the experts believe here is correct and that my own assessment of the evidence is in error. Were I to rely on my own evaluation of the direct evidence rather than the evaluation of the experts, or even assign any substantial weight to it, I would be making a very large error. I should show some epistemic humility, and recognize my own limitations. While it is admirable for laypeople to try to understand the issues involved in fields such as climate science, we need to remember that the body of relevant evidence here is huge, as is the body of relevant background theory, and that even those of us who keep up with newspaper re-

290

Hilary Kornblith

ports on these things are in a vastly inferior epistemic situation to the experts in the field. So when I say that we should, ceteris paribus, believe as the vast majority of experts do, I mean that we should do this even in those cases where our own evaluation of the direct evidence would lead us to believe otherwise. Our own evaluation of the direct evidence, in cases of this sort, should count for almost nothing. II. Suppose now that we consider a different case. Helga, unlike Frieda, is a climate scientist. Indeed, she is one of the most distinguished climate scientists in the world. She is recognized throughout her field as someone who is extraordinarily accomplished in the field, and someone whose works has always met the very highest standards. Helga has examined the new evidence that is reported in the Times, and she too has considered how it fits in with other evidence and with relevant background theory, and she too has formed an opinion about how it bears on the claim that p. She is also well aware of what other climate scientists think about this very issue. I want to suppose that Helga’s source of information here is not, of course, the New York Times, but professional journals, papers in progress, conferences, and discussion with other experts. How should Helga address this sort of question? In particular, how should she weigh her own assessment of the importance of the new evidence, and to what extent should she take into account the opinions of other experts about this very question? One might think that she should pay no attention at all to what other experts think. Helga is no amateur here. She is perfectly capable of weighing the evidence for herself. Her judgment on such questions is as good as anyone’s, and it is, indeed, very, very good. If after scrutinizing the direct evidence for and against p, Helga things that the evidence gives her good reason to form some particular belief, then that is what she should believe, and the opinion of others in the field is just not relevant here. She should follow the evidence where it seems to her to lead, and ignore the opinions of others about this question, even if the others are experts as well.4 4

Something very much like this view is endorsed by Thomas Kelly in “The Epistemic Significance of Disagreement,” Oxford Studies in Epistemology 1 (2005), 167-

Is Philosophical Knowledge Possible?

291

Now I wouldn’t be surprised if this correctly describes, or approximately describes, the actual behavior of experts.5 But for purposes of this paper, I am not interested in the actual epistemic behavior of experts. I am instead interested in what that epistemic behavior ought to be. To what extent should Helga take into account the opinions of other experts? It is also important here to make clear that I am not addressing the question of whether it would be a good thing professionally for Helga to ignore the opinions of others, or whether it would be a good thing for the progress of science for Helga to ignore the opinions of other experts. It might be that the best thing for Helga to do, if she is interested in professional advancement, is to pursue those theories that seem to her to be best supported by the evidence, whatever others may believe. And it might be that it would be a good thing for the scientific enterprise if Helga, and every other expert, were to ignore the opinions of their peers. Science progresses most quickly if many different theories are being examined and developed, and individual investigators may pursue various possibilities with the greatest vigor if they actually believe in them, rather than if they merely take them seriously. It might be that we would all converge on the truth more quickly if this were how we worked than if we tried, at each point in an investigation, to factor in the opinions of our peers. But these are pragmatic questions about how one’s career is best advanced, or about how inquiry is best pursued. For purposes of this paper, I am not interested in these questions. What I am interested in here is what Helga is epistemically justified in believing given her total evidence, and, in particular, whether her opinion would be epistemically justified if she were to ignore the opinions of other experts. And on this question, I believe, as opposed to the pragmatic questions, it is clear that Helga ought to take the opinions of other experts into account. If Helga ignores the opinions of others, she may be taking into account all of her direct evidence, but indirect evidence is evidence too. The fact 196. Kelly modifies this view in important ways in “Peer Disagreement and HigherOrder Evidence,” in Feldman and Warfield, eds., op. cit., 111-174. 5 I also wouldn’t be entirely surprised if it did not. Experts may well show some real sensitivity to the opinions of other experts in ways that affect their own judgments. Note that I am addressing the question here of the factors which influence expert judgment, not the question of what experts believe about the factors which influence their judgment.

292

Hilary Kornblith

that an expert on a subject holds a certain belief is indirect evidence that the belief in question is true. And if the majority of experts in a certain field all hold a particular belief, that is very strong strong evidence that the belief is true. Helga’s opinion counts too, of course, since she is an expert as well. So the situation here is not the same as the situation involving Frieda. But although Helga’s opinion counts no less than the opinions of other experts, it also counts for no more. And if Helga finds herself in a situation where she disagrees with the vast majority of experts in her own field, then she ought to think that, in light of all the available evidence, it is most likely that the other experts are right and that her own prior assessment of the evidence is mistaken. After all, consider the sort of reasoning which we said that Frieda ought to engage in. Frieda knows that the vast majority of experts believe that p. There is some disagreement among the experts, but the dissenters here are small in number. Frieda should recognize that, given the progressive nature of the discipline, very substantial agreement among the experts on any issue within their area of expertise provides one with extremely good reason for agreeing with that opinion. Minority dissenters are, of course, sometimes right, but anyone who regularly bet on the minority in these sorts of situations would lose a great deal of money. Barring extremely unusual circumstances, one would be well advised to believe as the majority of experts do. Now Frieda herself is not an expert on this issue, and so her own assessment of the bearing of the direct evidence carries extremely little weight. She should simply believe as the majority of experts do. But how is Helga’s situation different? Helga too should recognize that, given the progressive nature of her discipline, the majority opinion of the experts counts for a great deal. Just as Frieda needed to recognize that betting against this kind of majority is a losing epistemic strategy, Helga should recognize this as well. Now if Helga has found new evidence of which the other climate scientists are unaware, then that, of course, changes the situation dramatically. Similarly, if she has discovered some new argument for her view which has not yet been aired in the scientific community, this too would change the situation. But these are not the kinds of situations I’m interested in here. Rather, there are many cases in which there is disagreement within an expert community, and where all of the participants in the disagreement are well aware of the relevant evidence, and where they are also well aware of the

Is Philosophical Knowledge Possible?

293

arguments on either side of the disputed issue. These scientists read one another’s papers; they regularly attend conferences together; they talk about the issues face-to-face. Each of them could give a perfectly accurate summary of how the state of play in the discipline seems to other investigators. So Helga has no special evidence, or special arguments, that are not familiar to others. In a situation such as this, I believe that Helga should reason precisely as Frieda does. If the question at issue is what she ought to believe given her total evidence, then she ought to believe that the majority here is very probably right, even though her own assessment of the direct evidence would lead her to believe otherwise. Were she to follow the direct evidence where it seems to her to lead, she would be guilty of epistemic hubris. We all need to show a certain amount of humility in the face of such strong disagreement from others who are, as best we can tell, at least as wellplaced as we ourselves are for determining the truth. Anything less than this amounts to assigning extra weight to our own opinion merely in virtue of the fact that it is ours. Such a policy is never legitimate. III. Let us now turn to the situation in philosophy. The first thing to note about philosophy is that there is an extraordinary amount of disagreement among experts in the field. Not only is there disagreement among the experts on virtually every philosophical question; this sad situation is one that has characterized the field of philosophy from its beginning to the present day. In this respect, the dynamics of opinion within philosophy looks very different from that in the sciences. The characteristic convergence of opinion which we see in the sciences is largely absent from philosophy. There are some exceptions. Opinion in logic and set theory, despite areas of real disagreement, seems to have the same general convergent character as the empirical sciences. There are other formal areas of philosophy, such as decision theory, which, arguably, have a similar dynamic of opinion. But there can be little doubt that most of philosophy does not look like this at all. The central questions of epistemology, metaphysics, ethics, and so on, are very much areas of dispute. And the striking thing here is that, although there are serious disputes at the cutting edge of all of the sciences, we do not seem to have, in philosophy, what is characteristic of the sciences: an ever increasing body of accepted opinion among the experts. In the sci-

294

Hilary Kornblith

ences, there are large bodies of well-established results, and smaller areas of dispute. Over time, by and large, disputes tend to be resolved and many matters of dispute from fifty years ago, or even five years ago, have now been resolved and added to the body of established results. We simply don’t see this in philosophy. Philosophical opinion among the experts does not display anything like the convergent pattern that we see in the sciences. This has an important bearing on how we should view disputes within philosophy and what we should think about the possibility of philosophical knowledge. I said that within the sciences, even the experts should have opinions which are, ceteris paribus, dictated by the dominant opinion within the field. Where there is no dominant opinion, there is good reason to withhold belief. If we were to apply this same approach to philosophical issues, on which there is so little agreement, the result would be that we should all, to a first approximation, withhold belief on philosophical questions across the board. The extent of disagreement thus serves to defeat any claims to knowledge within those broad areas of philosophy where disagreement is widespread. Now one might think that this conclusion can be avoided by pointing out that our reasons for following the opinion of the majority in the sciences, and for withholding opinion when there is no clear majority, depended precisely on the convergent dynamic of opinion within the sciences. Since we have just finished pointing out that there is no such convergence in philosophy, the argument seems to be sidestepped. But far from saving philosophical knowledge, this point merely serves to underline just how bad the epistemic situation is in philosophy. The term ‘expert’ may be used in either of two different ways. On the one hand, we may use this to denote a certain social status, indicating an individual who is widely treated with a certain sort of epistemic deference. Experts in this sense may be picked out by their appointments at prestigious universities; their numerous publications and high citation count; the number and size of the grants and awards which they receive; and so on. To say that someone is an expert in this sense makes no commitment to any claim about the extent to which such an individual is likely to form true beliefs as a result of that person’s investigations. But the term ‘expert’ is more frequently used in a more substantive sense in which it denotes someone who is particularly well-placed for getting at the truth about certain matters. In a well-organized scientific community, being an expert in

Is Philosophical Knowledge Possible?

295

the first sense is a very good proxy for being an expert in the second, more substantive, sense. Nevertheless, the two may obviously come apart. Now there is no question at all that there are experts in philosophy in the first, social status, sense. There are philosophers who have appointments at distinguished universities, who command a great deal of respect within the profession, and so on. But the point about the lack of convergence of opinion within our discipline over time raises the very real question of whether there are any experts within philosophy in the second, and more substantive, sense. I have no doubt that professional philosophers have a great many intellectual skills to a very high degree. And I don’t doubt, of course, that they are frequently knowledgeable about the content of various texts. But the question at issue here is whether there is reason to believe that philosophers are well-placed to form true beliefs about philosophical questions, and the problem with the attempt to evade the worrisome argument about the status of philosophy by pointing out that philosophy does not have a convergent pattern of opinion over time is that it evades that particular argument only by calling into question whether anyone in philosophy is genuinely deserving of the title of expert in the substantive, and more important, sense. Experts within the sciences are highly reliable. Individual experts tend, to a very high degree, to form true, or at least approximately, true beliefs about the matters they investigate. It is for primarily this reason, of course, that opinions within the sciences tend to converge over time. By the same token, the lack of convergence over time within philosophy provides us with good reason to believe that the social status of being an expert is not well correlated with the property of having beliefs which are reliably formed. What this means, of course, is that we can safely ignore the opinions of those who disagree with us about philosophical issues, since there is no reason to believe that their beliefs are reliably formed, and so the fact that some particular expert has a certain belief gives us no reason to believe as the expert does. This is not to say, of course, that we may therefore hold fast to the philosophical views which we ourselves hold, for, unfortunately, the same considerations apply to our own beliefs. There is no reason to believe that we ourselves have philosophical beliefs which are reliably formed, and, as a result, our own beliefs are cast into doubt as well. The dynamics of opinion in philosophy gives us reason to believe that there are no experts in philosophy in the substantive sense, and therefore, at least for the foreseeable future, no philosophical knowledge seems to be possible.

296

Hilary Kornblith

This is, to be sure, a depressing conclusion, and it is one I am not eager to endorse. But I see no way out of it. The current state of our discipline does not seem to leave room for philosophical knowledge. IV. I want to consider a number of different replies to this line of argument, and I will begin with a reply which I take to be fairly radical. One might think that the comparison between philosophy and science is unfair, not because one could fail to measure up to science and still have reliable methods6, but rather because the comparison with science wholly misunderstands the nature of the philosophical enterprise. Such a view is held, I believe, by Barry Stroud. It is worth quoting fairly extensively from the preface to his book, The Significance of Philosophical Skepticism: This book is written in the belief that the study of philosophical problems can itself be philosophically illuminating. Of course no one would deny the need for a clear understanding of the problem at hand if there is to be real intellectual progress. But I do not just mean that solving or answering philosophical questions can be illuminating. Of course it could be, if you happened to get the right answer and knew that you had, or even if you failed and knew that you had failed, and perhaps even had some idea why. I mean that the study of the very nature of a philosophical problem can be an illuminating activity quite independently of whether it ever leads to a better answer. The attempt to understand what I am calling the nature of a philosophical problem can be expected to illuminate not only the problem itself, but also the very ‘phenomenon’ – morality, religion, knowledge, action, or whatever it might be – out of which the philosophical problem arises. It is surprising to me how few people writing in philosophy in this day and age actually concentrate on the problems themselves and where they come from. There seems to be widespread confidence about what the problems are, what sort of thing a successful philosophical doctrine or theory would be, and what it would take to give us the kind of understanding philosophy can give us of the phenomena it has traditionally been concerned with. I do not share this confidence. I think that whatever we seek in philosophy, or whatever leads us to ask philosophical questions at all, must be something pretty deep in hu6

This would not, I believe, be a very good objection. It is certainly true that one might conduct investigations in ways which are reliable and yet fall short of the degree of reliability found in the sciences. As I hope I made clear above, however, the problem with philosophy is not that it is less reliable than scientific investigation. It is, instead, that philosophical investigations seem not to be reliable at all.

Is Philosophical Knowledge Possible?

297

man nature, and that what leads us to ask just the questions we do in the particular ways we now ask them must be something pretty deep in our tradition. Studying the sources of philosophical problems as they now present themselves to us can therefore perhaps be expected to yield some degree of understanding, illumination, satisfaction, or whatever it is we seek in philosophy, even if we never arrive at something we can regard as a solution to a philosophical problem.7

Stroud is no more sanguine than I about our ability to solve philosophical problems or answer philosophical questions. Quite clearly, he believes that we are in no position to do that. Nevertheless, Stroud does think that we are in a position to make intellectual progress in philosophy, and that is where he and I differ. If one gives up, however, on the possibility of answering philosophical questions, then one must say something about the kind of intellectual progress which remains possible in its absence. Instead of focusing on the answers to philosophical questions, Stroud focuses on the questions themselves. Now the idea that we can make progress in a field by asking better questions is not at all mysterious. Great progress was made in understanding the nature of combustion when we stopped asking questions about phlogiston, and replaced them with questions about oxidation. But this familiar sort of progress can only be explained in terms of our understanding the answers to certain other questions. Chemists came to recognize that there was no such thing as phlogiston, and the discovery of oxygen, and a recognition of its role when various substances are burned, proved to open up whole new areas of profitable investigation. The new questions were better ones to ask than the old ones because they did not have false presuppositions, and also because they pointed the way toward tractable areas of investigation, areas in which chemists had reliable means of discovering important truths. The respects in which questions about phlogiston were bad questions to ask, and questions about oxidation were good ones, can only be explained in terms of our reliable access to truths about the chemical processes involved in combustion, and our ability to answer various questions by way of reliable methods. In this case, talk of the good and bad questions to ask in a field does not point us away from talk of having reliable methods for answering questions, but just the opposite. Here we see how the measure of the quality of various questions is wholly dependent on how it is that asking vari7

Oxford: Oxford University Press, 1984, x-xi.

298

Hilary Kornblith

ous questions will reliably lead to correct answers. So when Stroud talks about the importance of good questions, and of understanding the questions we ask, he cannot have in mind this very straightforward way in which ordinary inquiry certainly depends on asking good questions and understanding the questions we ask. So if this is not what Stroud has in mind, then what sort of progress does he see as genuinely possible in philosophy? How is the possibility of progress tied to an understanding of philosophical questions, in a way that is wholly independent of reliably answering them? Stroud does provide a few hints here. He does remark that we should be interested in what it is that prompts us to ask philosophical questions, and he conjectures that it is either something very deep in human nature or deep in our intellectual tradition. Once again, I have some trouble understanding just what this might mean. Thus, consider the kind of work done by certain cognitive anthropologists which sounds very much like the sort of thing that Stroud says. One might be an atheist, and yet be struck by the fact that religious belief is exceptionally widespread, and this might lead one to wonder what it is deep within human nature, or deep in our intellectual tradition, which prompts us to ask certain sorts of religious questions – questions about God’s nature, or interests, or activity in the world – and what it is that prompts so many to have religious beliefs. This is a perfectly coherent research program, and it could, in principle, illuminate all sorts of things. But this sort of research program is, of course, a certain sort of debunking program. In at least one obvious way of approaching it, it clearly depends on the prior claim that there are no gods, that religious beliefs are mistaken, and that theological questions at least typically have false presuppositions. So this approach depends on the view that we have a good deal of background knowledge here. It presupposes that we have been able to provide accurate answers to important questions. When this sort of research program is thus pursued, it not only seeks to provide illumination by seeing what the motivation for certain sorts of questions might have been; it is predicated on the assumption that it has already identified various false presuppositions of those questions. But this cannot be what Stroud means either. Stroud does not take himself to have discovered that philosophical questions have false presuppositions and that we should therefore stop trying to answer them, in the way that our cognitive anthropologist would have us stop asking questions

Is Philosophical Knowledge Possible?

299

about God’s intentions. Stroud cannot think this because it would itself involve the suggestion that we have already achieved some important philosophical knowledge, and this is precisely the view that Stroud rejects. We must stop thinking about philosophical progress in terms of answers to any philosophical questions, according to Stroud, because, on his view, the supposition that we can actually answer such questions reliably is simply mistaken. But this makes the analogy with the very similar sounding inquiries of certain cognitive anthropologists wholly unilluminating. Stroud does say that his approach, which involves a better understanding of philosophical questions, might lead serve to illuminate “the very ‘phenomenon’ – morality, religion, knowledge, action, or whatever it might be – out of which the philosophical problem arises.” What is puzzling about this, however, is that Stroud seems to think that we might understand such phenomena better without, at the same time, coming to solve any philosophical problems. Of course, this is exactly what would happen if there were some sort of debunking story to tell, as in the case of the cognitive anthropologists. But we have seen that this does not seem to be what Stroud has in mind. It is thus quite difficult to say how it is that we might achieve a better understanding of various philosophical phenomena without, thereby, gaining the very sort of philosophical knowledge which Stroud clearly thinks is impossible. I am thus at a loss to say what kind of intellectual progress Stroud believes we can make in philosophy, once he accepts, as I do here, the view that we are in no position to answer philosophical questions reliably. V. A very different sort of response to the challenge I have posed would involve telling some sort of progressive story about the philosophical enterprise, thereby challenging my claim that the dynamics of opinion in philosophy is importantly different from the dynamics of opinion in the sciences. Hegel, of course, had a grand story about the history of philosophy which presented it as a progressive discipline leading, ultimately, to his own philosophical view. Many of us, at times, may have ways of telling the history of philosophy, or at least the history of recent philosophy, in some such progressive manner. There are recent notable accounts of this sort to

300

Hilary Kornblith

be found in work by Scott Soames8, Timothy Williamson9, and Derek Parfit10, for example, but, of course, there are a great many such accounts to be found in the philosophical literature. The problem with these accounts, however, is that there is no more agreement on them than there is on other philosophical issues.11 What this means, of course, is that the very same argument I gave about philosophical issues generally applies equally well to the account of philosophical progress. When the experts in a field disagree quite radically on what a progressive account of the discipline looks like, we should all remain agnostic about such accounts. In a similar vein, one might suggest that a certain approach to philosophical problems is available which allows for the possibility of progress of the very sort needed to leave room for philosophical knowledge. One such suggestion, a suggestion which, I have to acknowledge, I find quite appealing, is that a naturalistic methodology would achieve precisely this result.12 Why might one think that naturalism could provide a solution to the problem under discussion here? The idea, I think, is quite straightforward. Traditional philosophical methods have failed to result in the kind of consensus over time which is characteristic of progressive disciplines. If, however, we allow philosophical questions to be pursued in a manner which is 8

Philosophical Analysis in the Twentieth Century, volume 1, Princeton: Princeton University Press, 2003; and volume 2, Princeton: Princeton University Press, 2005. 9 The Philosophy of Philosophy, Malden / Oxford / Victoria: Blackwell, 2007, 27892. 10 On What Matters, vols. 1 and 2, Oxford: Oxford University Press, 2011. 11 Thus, for examples, see these remarks by Mark Schroeder about Parfit’s attempt to reconcile Kantian and rule consequentialist approaches to ethics: If the background assumptions that Parfit has used to evaluate which Kantian principle is most defensible and his derivation of Rule Consequentialism are rejected by Kantians, then it looks like a large overstatement to say that Parfit’s convergence argument shows that there are no “deep disagreements” between Kantians and Rule Consequentialists. For what the argument actually shows is that there is a principle that Kantians don’t necessarily actually believe, but which is more defensible than what they actually believe by standards they don’t actually accept, such that, given background assumptions Kantians reject, it is equivalent to Rule Consequentialism. To say that this shows that the apparent disagreement between Kantians and Rule Consequentialists is not deep appears to overstate the case considerably. [Review of On What Matters, vols. 1 and 2, Notre Dame Philosophical Reviews, 2011.] 12 Sally Haslanger suggested this when I discussed these ideas at MIT.

Is Philosophical Knowledge Possible?

301

continuous with the sciences, then the progressive features of science will carry over to philosophy, thereby making philosophical knowledge a genuine possibility. The kinds of consensus one finds in the sciences, both about methods and about background theory and relevant data, can lead philosophy on the path to progress. No one will be surprised to hear that this is a view I am tempted by, and, on many occasions, a view I actually believe. But that is mere autobiography. The problem with the view about progressive histories of philosophy, such as those of Hegel, or Soames, or Williamson, clearly applies equally to the would-be naturalistic solution. Sad to say, the naturalistic approach to philosophical problems is itself a matter of substantial controversy among philosophers. Many, famously, see it as misguided, or a way of avoiding, rather than addressing, philosophical issues.13 Whatever the virtues of naturalism, and I believe there are many, it cannot provide a solution to the problem under discussion here. One might attempt, at this point, to lower one’s sights somewhat. Perhaps the big issues in philosophy are ones on which there simply is no adequate consensus. But even if we cannot achieve consensus on the larger issues, there are still smaller points on which consensus can be, and actually has been, achieved. Thus, for example, when an article appears in some prominent philosophical journal, there may be little consensus on the largest issues it takes on, but responses to the paper sometimes succeed in showing that certain arguments in the paper do not achieve what was

13

To cite just one such example, we need look no further than the work of Stroud already cited for a passage which clearly targets a certain sort of naturalistic response to skepticism: The attempt to meet, or even understand, the skeptical challenge to our knowledge of the external world is regarded in some circles as an idle academic exercise, a willful refusal to abandon outmoded forms of thinking in this post-Cartesian age. When this attitude is not based on ignorance or a philistine impatience with abstract thought it often rests on the belief that we already understand quite well just how and why traditional philosophical skepticism goes wrong. One aim of this book is to suggest that that comfortable belief is not true. [Op. cit., viii.] It’s safe to say that Stroud would not accept the proposed naturalistic approach to philosophical questions. More than this, Stroud has a great deal of company here.

302

Hilary Kornblith

claimed for them, and points of this sort may be ones on which there is sufficiently wide agreement to leave room for philosophical knowledge.14 I don’t believe that this suggestion really helps. It seems to me to lower the ambitions of philosophy so much as to make it difficult to see why it is worth engaging with at all. If it is granted that the large issues in philosophy are really ones that we have no reliable means of addressing, then the suggestion that we can at least know that some recent journal article contains an error, even if this knowledge will still be of no use in addressing the larger issues in the field, utterly demeans the entire discipline. Imagine what science would be like if it were to find itself in a similar position. If we had no idea whether the atomic theory of matter was even roughly correct, or, instead, a variety of alternative views, but we were able to establish that some journal article advancing the atomic theory contained a logical error, and if this were characteristic of the sort of knowledge that was possible within the sciences, then I think it is safe to say that we would regard the sciences as being in an extraordinarily primitive state, where scientific knowledge was no more than a fond wish, rather than a real achievement. It would be cold comfort if philosophy could legitimately make claim to achievements of this sort. There is thus, I believe, little prospect of painting philosophy as a progressive discipline in the way that the sciences clearly are, and this leaves us open to the problem initially presented. VI. I want to consider one final challenge to the argument that I have presented, and it is the suggestion that my argument is self-undermining. One might think that the argument is self-undermining in two quite different ways. First, one might think that one or another of the premises in the argument is itself a subject of philosophical controversy, and if, as the argument suggests, we have no right to rely on assumptions which are controversial within the discipline, then anyone who endorses the argument should therefore reject one of its premises, and the argument is thereby 14

Jessica Brown suggested something like this when I presented these ideas at St. Andrews, although I do not know whether, in the end, she would endorse this suggestion.

Is Philosophical Knowledge Possible?

303

defeated. For example, one might suggest that the very approach to disagreement I presuppose, which requires a great deal of conciliation in the face of disagreement from peers, is far too controversial to do the work it does in the argument under discussion. If claims which are a subject of controversy in the field cannot be relied upon, then this claim must be dispensed with, and the argument never even gets started. Alternatively, one might think that, even if all of the premises were immune from controversy, the conclusion itself is surely controversial. But then anyone who accepts the conclusion must, for that very reason, reject it, and once again, the argument is defanged. Either way, the argument undermines itself. These are, to be sure, interesting suggestions, and I want to respond to each. On the first point, let me say something about my conciliatory approach to disagreement among peers. The literature on disagreement has evolved very rapidly in recent years.15 There is, to be sure, a great deal of disagreement within this literature about theoretical approaches to the issue. Nevertheless, while early on in the literature there was also some substantial disagreement about how particular cases should be handled, there is an emerging consensus about many of them. Early on, some suggested that the opinion of one’s peers might simply carry no weight at all in determining what belief one ought to hold.16 This suggestion, however, is now nearly universally rejected. Peer disagreement carries some real weight. More than this, there is a great deal of agreement about a number of cases brought forward by David Christensen17 – and here what I have in mind is that there is agreement about the verdicts in these cases even when there is disagreement about the theory which should deliver the verdict – and what is agreed upon here is that in the face of substantial disagreement from one’s peers, one must be quite conciliatory. This is all I need to get my argument going. So at least on this premise, there is, I believe, no threat that the argument might be self-undermining. 15

For just a small sampling, see the volume cited in note 2 above, and also Jennifer Lackey and David Christensen, eds., The Epistemology of Disagreement: New Essays, New York: Oxford University Press, forthcoming. 16 See, for example, the first of the two Kelly papers cited in note 4 above. 17 See, for example, “Epistemology of Disagreement: The Good News,” Philosophical Review 116 (2007), 187-217.

304

Hilary Kornblith

Matters are quite different, however, when it comes to the conclusion of the argument, for here I think it is quite clear that the conclusion is itself controversial, and since the argument suggests that we should not accept claims which are matters of controversy among the experts in a field, the worry that the argument is self-undermining is particularly powerful. And I have to admit that, in one very straightforward sense, I have nothing to say in reply to this. The argument does have this feature, and there is no denying it. I offer this argument, however, in the spirit in which I think we must offer any philosophical argument. Here is an argument. The premises seem to be true. They seem to lead to a certain conclusion. If one wishes to reject the conclusion, it seems that one ought to have something to say about the argument: one ought either to offer a reason for thinking that one of the premises is false, or, alternatively, that the premises do not, in fact, support the conclusion. At the moment, I cannot see a way to defend either of these claims. And so I find this argument deeply disturbing. It is not a conclusion I wish to accept, but there is an argument for it which seems to force us to accept the conclusion. This argument leaves us all, I believe, in an uncomfortable position. It would be nice to know how to get out of it.

The Significance of Disagreement in Epistemology DIRK KOPPELBERG, Freie Universität Berlin

Disagreement shows up almost everywhere in our life. Whenever we meet colleagues, friends or strangers and we talk to each other, it usually won’t be long until we disagree on a certain topic. We disagree on many things indeed. We disagree, for instance, on the content of an ordinary visual perception, on the division of a collective bill in a restaurant, on the location of a certain bookstore in downtown, but also on more complicated matters such as art, politics, religion and philosophy. Some of these disagreements can be dissolved rather quickly by correcting easily detectable errors or mistakes, but some of them seem to be rather stable. Among the most stable disagreements are disagreements in philosophy and especially in one of its central theoretical disciplines, in epistemology. The epistemology of disagreement deals with all kinds of different cases in ordinary life, in theoretical domains and even in epistemology itself. Out of this vast field of disagreement I pick up the following three questions which I will try to answer in this paper. First, why does disagreement matter to epistemology? Second, what characterizes disagreement in epistemology? And third, how ought we to deal with disagreement in epistemology? I will close with my replies to some important objections to my view. 1. Why does disagreement matter to epistemology? – Some basic distinctions To answer this basic question, we need to answer three further questions. First, with whom can and do we disagree? Second, what kind of disagreement matters mostly to epistemology? Note that this question is not to be confused with the quite different question of what kind of disagreement matters mostly in epistemology. This second question I’m going to address in the second section of the paper. And third, which levels of disagreement are to be distinguished? Only by answering these three questions we will

306

Dirk Koppelberg

arrive at a better understanding of why disagreement matters to epistemology and what kinds of challenges have to be met. So, first, with whom can and do we disagree? This question concerns the kinds of different intellectual relations which are in principle possible between two disagreeing persons. There are three of them: (i) it might be a disagreement between me and an intellectual inferior, (ii) it might be a disagreement between me and an intellectual superior, and (iii) it might be a disagreement between epistemic peers. To describe these three different kinds of intellectual relations, we need at least the following two criteria: first, relative to the question whether p, familiarity with the evidence and arguments which bear on p; second, relative to the question whether p, intelligence, competence and carefulness in assessing the evidence and arguments. Respective different kinds of inequality in both areas can then be used to characterize intellectual inferiors and intellectual superiors. Here I will not say more about inferiors and superiors, because it is the third kind of relation, namely peer disagreement which matters mostly to epistemology – at least according to the dominant view of those philosophers who work on this topic. Why does peer disagreement play such a central role? It poses a serious challenge to epistemology. How can this challenge be circumscribed? A preliminary and rather global answer would be the following: it puts our conceptions of epistemic rationality and responsibility to a severe test and it demands of us to spell out how we should conceive of the relation between epistemic and social factors in our cognitive life. Now before going into the details of the challenge triggered by peer disagreement, let’s take a closer look at the two notions involved here: the notion of an epistemic peer and the notion of the relevant disagreement. Let’s begin with the concept of an epistemic peer. What has to be taken into account for characterizing an epistemic peer? Building upon and modifying some proposals by David Christensen and Jennifer Lackey I put tentatively forward the following two necessary conditions which jointly might turn out to be sufficient (cf. Christensen 2007 and Lackey 2010, 302 f. for rather similar conditions, but note that I have supplemented them by adding “presuppositions” to the list which will turn out to be especially important for my view; in former versions I added a third condition but due to convincing criticism especially by Thomas Grundmann, Andrea Kruse and David Löwenstein I have meanwhile dropped it.).

The Significance of Disagreement in Epistemology

307

(C1) Evidential equality: A and B are evidentially on a par to the question whether p when A and B are equally familiar with the presuppositions, the evidence and the arguments that bear on the question whether p. (C2) Intellectual equality: A and B are intellectually on a par relative to the question whether p when A and B are equally intelligent, competent and careful in their assessment of the presuppositions, the evidence and the arguments that bear on the question whether p. When there is evidential and intellectual equality between A and B with regard to the question whether p, A and B are epistemic peers in my sense. (In the following I will often drop the adjective “epistemic” when no misunderstanding is to be feared.) After having defined what an epistemic peer is, let us now distinguish two kinds of disagreement which might matter to epistemology: idealized disagreement on the one hand and presumptive disagreement on the other hand (Here I elaborate on a proposal of Lackey 2010, 303 f.). (ID) Idealized disagreement: A and B disagree in an idealized sense if and only if, relative to the question whether p, (1) A and B acknowledge that they hold differing doxastic attitudes, (2) prior to recognizing that this is so, A and B have taken themselves to be epistemic peers to this question, and (3) A and B are in fact epistemic peers. (PD) Presumptive disagreement (I use the adjective “presumptive” in its technical sense, according to which a presumptive disagreement exists so long until good reasons are found in virtue of which a crucial presupposition for maintaining the disagreement has to be rejected. Cf. the presumption of innocence at a court: The accused has to be regarded as innocent until good, if not even conclusive reasons for his debt are presented.) A and B disagree in a presumptive sense if and only if, relative to the question whether p, (1) A and B acknowledge that they hold differing doxastic attitudes, (2*) prior to recognizing that this is so, A and B have taken themselves to be epistemic peers to this question which means that they have not – perhaps not yet? - discovered any relevant epistemic asymmetry between their situations, but (3*) such a relevant epistemic asymmetry cannot be excluded before having re-examined all the evidence, the arguments and the presuppositions for holding differ-

308

Dirk Koppelberg

ent doxastic attitudes. (I have inserted condition (3*) due to some helpful criticism of Andrea Kruse. – Note that condition (2*) does not mean that the respective opponents are justified in their respective beliefs about their epistemic peerhood – as Grundmann seems to read my proposal –, they have presumed it because they had no reason to doubt it. One ought not to identify presumption with justification.) Despite its importance, the crucial distinction between idealized and presumptive disagreement has not been sufficiently taken into account by many epistemologists. In the following I’ll focus exclusively on presumptive disagreement because the fulfillment of condition (3) in the idealized case cannot be taken for granted at least under empirically realistic constraints from the perspective of A and B. I think, this difficulty is not restricted to A and B. For any observer from a third-person-perspective – leaving an omniscient God aside – it is hard to imagine how it could be ascertained that condition (3) is actually satisfied. After having introduced the distinction between idealized and presumptive disagreement, I now turn to answer the second question of this section, namely which kind of disagreement matters mostly to epistemology. Is it disagreement being based on mutually presumed peerhood, or is it disagreement which involves factual peerhood? I think, that it is presumptive disagreement which is the adequate basis for the epistemological challenge. Why? Only on the basis of a mutually acknowledged peerhood of the disagreeing subjects the specific epistemological problem arises. Why does it arise just on that basis? Only if it’s not supposed to solve the problem of disagreement with an opponent by making explanatory use of his alleged inferiority or his alleged superiority, instead of which I assume that she is a peer, I am confronted with the specific epistemological question of how to behave epistemically rational and responsible in the face of such a disagreement. Thus presumptive peer disagreement is at least a good starting point for the analysis of disagreement in epistemology – the topic in the next section. Before turning to the next section, we have to address the third question, namely which levels of disagreement are to be distinguished. (cf. Lynch 2010, 264) Usually disagreements begin with disagreement over some facts. Therefore, let’s call this level of disagreement factual disagreement. Disagreement over facts mostly turn into disagreements over whose view of those facts is best supported or justified. Let’s call this level of disa-

The Significance of Disagreement in Epistemology

309

greement doxastic disagreement. “And sometimes we move still further up what might be called the epistemic ladder: we begin to disagree over how we ought to support our view of the facts, about the sort of evidence that should be admitted, and whose methods more accurately track the truth. When we do that, we are engaged in a truly epistemic disagreement: a disagreement over epistemic principles.” (Lynch 2010, 264) After having thus distinguished factual, doxastic and epistemic disagreement, we see more clearly that truly epistemic disagreement consists in disagreement over epistemic principles or strategies which should be used to deal with doxastic disagreement among us. We are now in a good position to address the crucial question of the first part of my paper, namely: what is the relation between epistemic peerhood and epistemic disagreement (that is disagreement over epistemic principles or strategies)? Epistemic disagreement is, of course, a genuine topic in epistemology. Epistemologists seem to be the relevant experts for dealing with questions concerning disagreement over epistemic strategies. (Sebastian Watzl has stressed that non-philosophers also disagree about epistemic strategies. That’s true. Nevertheless I hold that epistemologists are supposed to be experts for these questions.) Now the crucial question is whether prima facie genuinely competing epistemologists are epistemic peers in the strict sense that I have just introduced. Here I do not deny that different epistemologists are on a par in several respects but nevertheless it might turn out that apparent genuinely competing epistemologists are only peers in a restricted sense or even that they are no peers at all. I have already argued for the thesis that it is presumptive peer disagreement which plays the crucial role in the epistemology of disagreement. To make further progress I now want to examine more closely how disagreement in epistemology has to be characterized. 2. Interlude – What characterizes disagreement in epistemology? In the first section of my paper I answered the question what kinds of disagreement matter mostly to epistemology. We found out that it is both presumptive peer disagreement and epistemic disagreement which can be taken together as presumptive peer disagreement on the epistemic level. Now it seems plausible that epistemologists are the relevant experts to examine this kind of disagreement that might easily amount to a kind of disagreement in epistemology.

310

Dirk Koppelberg

What are the most important or - to be more cautious – at least the most often acknowledged features of disagreement in epistemology? I think there are the following two of them: first, disagreement in epistemology seems to be rather stable; second, disagreement in epistemology seems often to be merely verbal. (I believe that these two features may be generalized to almost all areas of philosophy, but here I cannot go into that general thesis.) What is the relation between the supposed two main features of disagreement in epistemology? Concerning this question, there are two rather influential views. First, there is what I will call the merely-verbal-view according to which stable disagreements in epistemology depend on the fact that disagreeing philosophers talk past each other and that their alleged opposition is finally to be reduced to merely verbal misunderstandings among them. Second, there is what I will call the irreconcilable-view according to which stable disagreements in epistemology go beyond merely verbal misunderstandings and cannot satisfactorily be dissolved even in the long run. According to this view, stable disagreements seem to be essential features both of epistemology in particular and of philosophy in general. My own proposal tries to maintain what seems to be right in both views and, of course, it tries to avoid what seems to be wrong in them. It is – in a certain sense – an irenic view which consists in a theoretical diagnosis about the errors of its rivals. According to my proposal, disagreements in epistemology are - at least more often than commonly recognized - only apparently stable, because they involve hidden and not easily detectable theoretical confusions among only seemingly competing epistemologists. This is not to say that the relevant rivalry depends upon no more than merely verbal misunderstandings. On the contrary – it is itself an important philosophical task to give a convincing theoretical diagnosis accounting for the enduring and apparently indissoluble stability of the respective philosophical disagreements. And my claim is that my proposal is able to do that. 3. How ought we to deal with disagreement in epistemology? – A case study To defend my irenic point of view, I now turn to a little case study about the epistemological disagreement over epistemic justification (thereby tak-

The Significance of Disagreement in Epistemology

311

ing up a suggestion by Sosa 2010, 281). Let’s listen to what a committed externalist, Hilary Kornblith, has to say about the ongoing debate between internalists and externalists: “I have argued for this position [he means externalism; D.K.] at length and on numerous occasions. My arguments for the position are not merely a pose. I have not presented arguments merely to serve as a gadfly, provoking discussions. I sincerely believe that the arguments I have presented are good arguments, and I sincerely believe their conclusions. At the same time, I recognize, of course, that there are many philosophers who are equally committed internalists about justification, and that the arguments they offer are ones not only to which they are committed, but which they believe are good arguments, and whose conclusion they sincerely believe. It would be reassuring to believe that I have better evidence on this question than those who disagree with me, that I have thought about the issue longer than internalists, or that I am simply smarter than they are, my judgment superior to theirs. It would be reassuring to believe these things, but I do not believe them; they are all manifestly untrue. So, on the question of whether externalism or internalism is correct, I find that I have an opinion, but there are others who disagree with me, who are, to adopt a useful tem, my epistemic peers…” (Kornblith 2010, 31) Contrary to what Kornblith claims here and in defense of my irenic view, I now want to show that internalists and externalists are not epistemic peers at all, that they miss their mutual genuine insights, that they sometimes might not even understand themselves properly, and that nevertheless there exists not just a merely verbal misunderstanding between them. There is room for an interesting theoretical diagnosis with regard to the alleged genuine debate between externalists and internalists which might finally dissolve it. In a former paper (Koppelberg 1999) I have discussed several different kinds of epistemic justification. For my present purposes I concentrate on the relevant distinction between doxastic justification and personal justification. Doxastic justification means that S’s belief that p is actually justified. Personal justification means that S is justified in believing that p. Many philosophers do not distinguish the concepts of personal justification and doxastic justification. For instance, in exposing the concept of epistemic justification William Alston writes: “It applies to beliefs, or alternatively to a cognitive subject’s having a belief. I shall speak indifferently of S’s

312

Dirk Koppelberg

belief that p being justified and of S’s being justified in believing that p.” (Alston 1985, 83) Alston and a lot of other epistemologists either subscribe explicitly to what Mylan Engel (1992) has called the equivalency thesis of personal justification and doxastic justification or at least they implicitly presuppose its validity. The thesis asserts: S is justified in believing p if and only if S’s belief that p is justified. According to the equivalency thesis, there exists no theoretical difference between S’s belief that p is justified and S being justified in believing that p. This thesis is supposed to guarantee that the debate between internalists and externalists turns out to be a discussion between genuine opponents with regard to the same topic. Unfortunately, the equivalency thesis is false. Many externalists and internalists write indifferently about the justification of a belief and the justification of a believer. Nevertheless it is noteworthy that externalists normally speak of the justifiedness of a belief whereas internalists seem to be occupied with the justifiability of a believer. But often the alleged opponents seem to presuppose that if the believer is justified, then her belief is, and if her belief is justified, then the believer is. Thus, subscribers to the equivalency thesis exclude the possibility that the believer may be justified when her belief is not and that her belief be justified when the believer is not. In contrast to the adherents of the equivalency thesis I hold that these just mentioned and excluded possibilities should be admitted because personal justification and doxastic justification deal with two different objects of epistemic appraisal and their respective adherents are committed to different theoretical goals. In the first case we evaluate whether a person is justified in adopting or maintaining a certain belief, whereas in the second case we evaluate whether that belief itself is justified. Externalists are interested in doxastic justification, the kind of justification which most epistemologists take to be necessary for propositional knowledge. (Of course, some philosophers deny that any kind of justification, doxastic justification included, is necessary for knowledge; e.g. the early Goldman, Sartwell, Kornblith. Here I cannot deal with this position.) Internalists are mainly interested in personal justification, the kind of justification accounting for the subject’s theoretical rationality. Personal justification might not be necessary for knowledge, even if some epistemologists like Laurence BonJour, Robert Fogelin and Michael Williams hold that it is.

The Significance of Disagreement in Epistemology

313

What makes a person justified in a certain belief can be fleshed out both in deontological and in virtue-theoretical terms. The justifiability of a person might even depend – at least prima facie – both on the exercise of his intellectual virtues and on his blameless cognitive behavior which resist easy summarizing. Which intellectual virtues and which epistemic duties are relevant here, is itself a complex epistemological problem which I cannot discuss here. Nevertheless, I think that asking the right questions, considering the relevant evidence and maybe trying to exclude obvious defeaters are at least important candidates for epistemically virtuous and laudable behavior. Even if it is not easy to give a generally convincing characterization of good epistemic behavior, we should agree on the point that a person is justified in believing something if he has done what is feasible within his ken to acquire or to sustain the belief rationally and responsibly. However, a belief can be epistemically justified without any specific epistemically virtuous or laudable behavior on the part of the believer. Many of our justified perceptual, introspective and memory beliefs are formed automatically or routinely without any deliberate action on the believer’s part. So I think, it is of great epistemological importance to distinguish between the internalist’s concern with a person being justifiable in a belief and the externalist’s concern with the belief itself being justified. Even if a certain belief is not justified in the sense that its truth-conduciveness is thereby increased, this might not be a good reason to criticize the respective epistemic behavior of its believer. And even if the believer’s epistemic behavior is vicious or irresponsible, its end product in form of his belief may be epistemically valuable because it has been produced by independent reliable means. Thus the enduring stability of the alleged opponents’ disagreements over internalism and externalism can be dissolved by my theoretical diagnosis concerning the question of why they talk past each other. As already indicated, my diagnosis shows on the one hand that the allegedly genuine competition between externalists and internalists is based on a kind of deep theoretical confusion on both sides, but this confusion is by no means merely verbal. On the other hand it shows that the relevant kind of theoretical confusion and the only apparently stable disagreement is based on the fact that internalists and externalists deal with different objects and pursue divergent goals without seeming to be sufficiently clear about that. I hope that my diagnosis succeeds in illuminating three points: first, it elaborates that the alleged stable disagreement between externalists

314

Dirk Koppelberg

and internalists is in fact based on theoretical confusion. But second, it shows why more is at issue than a merely verbal misunderstanding; it is a kind of mutual error which consists in ascribing the same project to the alleged opponent and not having recognized that the alleged opponent is engaged in pursuing a quite different aim. And third, my diagnosis is able to make sense of the alleged stability of the disagreement and seems to be successful in dissolving it by showing that externalists and internalists are not genuine opponents at all because they are engaged in different tasks. According to my diagnosis, even a friendly reconciliation of these only apparently competing positions is possible; they need not mutually exclude each other if it is conceded that they address different questions. Thus, my irenic view has avoided both the shortcomings of the merely-verbal-view and the irreconcilable-view and it has preserved and explained what seems to be right in both of them. 4. Some objections and replies Of course, here I could present and defend my view on disagreement in epistemology only with reference to a single short case study. For its further development and defense some more detailed case studies have to be delivered. This is indeed what I’m going to do within a larger project (cf. Koppelberg forthcoming). The candidates for these case studies are disagreements over different conceptions of knowledge and the relation among those conceptions, the recent controversy on the value of knowledge and current discussions about the proper place of intellectual virtues in epistemology. Hence much work still needs to be done to elaborate my view on disagreement in epistemology. Until now it is far from clear that all epistemological disagreements can be treated along the same line. But let’s stick to the case that I have presented here and let me recapitulate my two main theses before I am going to face some important objections to my account. There are two main features of disagreement in epistemology that have been emphasized: disagreement in epistemology seems to be rather stable and often it seems to be merely verbal. I have introduced two received views that try to account for these features: the merely-verbal-view and the irreconcilable-view. My first thesis consists in a further view, according to which at least some important disagreements in epistemology are only apparently stable because they involve not easily detectable theoretical confusions among only seemingly competing epistemologists. My second

The Significance of Disagreement in Epistemology

315

thesis, a corollary of the first one, is the claim that in the case study I presented here, the disagreement between externalists and internalists is not to be understood as an instance of genuine peer disagreement. In earlier discussions of this paper three important objections have been put forward against these two theses. First, in discussion David Löwenstein has raised the objection that - contrary to what I have claimed - externalists and internalists are genuine peers because both of them subscribe to the equivalency thesis and hence they fulfill both conditions for epistemic peerhood: they are equally familiar with a crucial presupposition of their dispute and they are equally intelligent, competent and careful in their assessment of this presupposition. But this is by no means settled. From the observation that many epistemologists and even dedicated followers of externalism respectively internalism take the equivalency thesis for granted it doesn’t follow that all of them do or that they are committed to do so. Quite the contrary. I have argued that both internalists and externalists have good reasons to distinguish their different endeavors; they pursue different projects which should not be mixed up. If this is true, externalists and internalists are not genuine peers who disagree on the same topic but adherents of different epistemological projects which might even turn out to be compatible. Second, it has been objected in several discussions that internalists would not agree with my diagnosis concerning their proper object of analysis within their project of developing an adequate account of epistemic justification. My opponents have claimed that Laurence BonJour, for instance, is not interested in personal justification, but in doxastic justification. This is just false. Recall BonJour’s famous example of Norman, the completely reliable clairvoyant with regard to certain subject matters without having any evidence for his extraordinary faculty (cf. BonJour 1985, 41). One day he comes to believe that the President stays in New York, though he has no reasons for or against this belief. But the belief is true and it is the result of Norman’s clairvoyant power under circumstances in which it is absolutely reliable. Now it is crucial to hear BonJour’s appraisal of the case in his own words: “Is Norman epistemically justified in believing that the President is in New York City, so that his belief is an instance of knowledge? […] Aren’t there still sufficient grounds for a charge of subjective irrationality to prevent Norman from being epistemically justified?” (BonJour 1985, 41; my emphasis; D.K.) Here we have a really nice example for the tacit identification of personal justification, the charge of

316

Dirk Koppelberg

the missing subjective rationality, with doxastic justification which consists in the truth-conduciveness necessary for knowledge. And it is exactly this tacit identification that a strict externalist won’t or at least shouldn’t accept. (For detailed arguments see Koppelberg 1999.) Third and finally it has been claimed that my proposed irenic view is not a genuine alternative to the other two options. Especially it has been objected that my position amounts to nothing more than a version of the merely-verbal-view. Here I concede that this criticism may look superficially appealing but nevertheless it overlooks the deep differences between the two accounts. Of course, both of them agree on the general point that disagreements on controversial issues in epistemology are to some extent verbal. But my irenic view is able to explain why they are to some extent verbal without being uninterestingly superficial. My account gives an explanatory diagnosis; it shows that the allegedly irreconcilable and stable disagreement between externalists and internalists can be traced back to a deep theoretical confusion about their proper targets. Thus, with regard to its explanatory power my view supersedes its rivals and maintains their partial insights. The title of my contribution is “The Significance of Disagreement in Epistemology”. Of course, this is an elliptical way to put it. One might ask: significant for what? And then the answer will be that I am wondering about the significance of disagreement in epistemology for the epistemology of disagreement. At the beginning of my paper I remarked that current epistemology of disagreement has mainly focused on peer disagreement. In order to look at this phenomenon more closely, I have chosen a case of disagreement in epistemology. If my analysis of disagreement among epistemologists is correct, we arrive at the surprising result that disagreement in epistemology turns out to be different from peer disagreement. Of course, this does not show that peer disagreement is not worth further research. It just shows how much work is still to be done to get a clear picture of the relation between the epistemology of disagreement and disagreement in epistemology. Note I have presented earlier and quite different versions of this paper at the Ruhr-Universität Bochum on June, 7th 2011, at the Freie Universität Berlin on June, 30th 2011, at the Universität zu Köln on July, 13th 2011 and at the

The Significance of Disagreement in Epistemology

317

Wittgenstein Symposium in Kirchberg on August, 11th 2011. I’d like to thank Helen Bohse, Daniel Friedrich, Joachim Horvath, Katharina Keil, Jens Kipper, Peter Kirschenmann, Albert Newen, Bruno Niederbacher, Erik Olsson, Anna Rosenbaum, Pedro Schmechtig, Holger-Jens Schnell, Gerhard Schurz, Tobias Starzak, Holm Tetens, Lars Weisbrod and Markus Werning for their critical questions and constructive proposals. I’ve especially benefited from the detailed written comments by Thomas Grundmann, Andrea Kruse, David Löwenstein and Sebastian Watzl; many thanks to them! References Alston, W., 1985: Concepts of Epistemic Justification, in: Epistemic Justification, Ithaca and London: Cornell University Press 1989, 81-114 BonJour, L., 1985: The Structure of Empirical Knowledge, Cambridge, Mass.: Harvard University Press Christensen, D., 2007: Epistemology of Disagreement: the Good News, The Philosophical Review 116, 187-217 Engel, M., 1992: Personal and Doxastic Justification in Epistemology, Philosophical Studies 67, 133-150 Grundmann, T., forthcoming: Das Problem stabiler Dissense in der Philosophie, in: Erkenntnistheorie – wie und wozu? Ed. D. Koppelberg and St. Tolksdorf, Münster: Mentis Koppelberg, D., 1999: Justification and Causation, Erkenntnis 50, 447-462 Koppelberg, D., 2012: Epistemology of Disagreement and Disagreement in Epistemology, Ms. Koppelberg, D., forthcoming: Dissense – Grundprobleme sozialer Erkenntnistheorie Kornblith, H., 2010: Belief in the Face of Controversy, in: Disagreement, ed. R. Feldman and T.A. Warfield, Oxford: Oxford University Press, 29-52 Lackey, J., 2010: A Justificationist View of Disagreement’s Epistemic Significance, in: Social Epistemology, ed. A.Haddock, A. Millar, and D. Pritchard, Oxford: Oxford University Press, 298-325 Lynch, M., 2010: Epistemic Circularity and Epistemic Incommensurability, in: Social Epistemology, ed. A. Haddock, A. Millar, and D. Pritchard, Oxford: Oxford University Press, 262-277 Sosa, E., 2010: The Epistemology of Disagreement, in: Social Epistemology, ed. A. Haddock, A. Millar, and D. Pritchard, Oxford: Oxford University Press, 278297

The Role of the Uniqueness Thesis in the Epistemology of Disagreement MATTHEW LEE, University of Notre Dame

Abstract The Uniqueness Thesis is the thesis that there is always a unique doxastic attitude that is rational to have toward a given proposition when in possession of a given body of evidence. This thesis has featured prominently in recent epistemology of disagreement literature, largely because of its supposed connections with Conciliationism: the view that a person should suspend judgment (or at least decrease her confidence) when she finds herself in disagreement with an acknowledged epistemic peer. I present some problems for the alleged connections between the Uniqueness Thesis and Conciliationism and argue that the Uniqueness Thesis should be expected to play only a lesser (though still significant) role in the epistemology of disagreement. 1. Introduction The Uniqueness Thesis is the thesis that there is always a unique doxastic attitude that is rational to have toward a given proposition when in possession of a given body of evidence.1 This thesis has come to play a central role in the epistemology of disagreement since its formulation by Richard Feldman half a decade ago.2 The Uniqueness Thesis has the obvious con1

Some writers on Uniqueness have been concerned with the “coarse-grained” doxastic attitudes: belief, disbelief, and suspension of judgment. Others have focused on degrees of confidence (or “credences”). 2 Feldman first formulated the thesis in a manuscript that would become Feldman 2007. The first published appearance of the Uniqueness Thesis was in White 2005, but Feldman’s manuscript was temporally and causally prior to White’s paper. For further discussion of the Uniqueness Thesis as it pertains to the epistemology of disagreement, see Christensen 2007, 2009, Conee 2010, Feldman and Warfield 2010, Goldman 2010, Kelly 2010, and Ballantyne and Coffman 2011, forthcoming.

320

Matthew Lee

sequence that there can be no rational disagreements between epistemic peers (understood as intellectual equals who have the same evidence).3 But the Uniqueness Thesis is also thought to be intimately connected with Conciliationism – the view that a person should suspend judgment (or at least decrease her confidence) when she finds herself in disagreement with an acknowledged epistemic peer.4,5This paper, however, will cast doubt on the usefulness of invoking the Uniqueness Thesis in discussions of Conciliationism. In Section 2 I will explain one of two ways in which the Uniqueness Thesis is supposed to be connected with Conciliationism. I will then give reason to doubt both the existence of the alleged connection and its usefulness to the epistemologist of disagreement if it does exist. In Section 3 I will explain a second way in which the Uniqueness Thesis is thought to be connected with Conciliationism, and I will show how the main argument for this alleged connection can be undermined. 2. From Uniqueness to Conciliationism Epistemologists of disagreement distinguish “coarse-grained” doxastic attitudes (belief, disbelief, suspension of judgment) from “fine-grained” doxastic attitudes (degrees of belief or “credences”) and, accordingly, recognize the following two kinds of disagreement: s is in coarse-grained disagreement with s* about p =df s and s* have different coarse-grained doxastic attitudes toward p (where s and s* are persons and p is a content). 3

“Epistemic peer” has a semi-technical meaning in the recent literature on the epistemology of disagreement. On one of the most prominent construals, a person s and a person s* are “epistemic peers” relative to a proposition p if and only if s and s* are equally familiar with the evidence that is relevant to whether p is true and are equals with respect to the intellectual virtues (intelligence, reflectiveness, etc.) relevant to weighing the evidence for and against p. See Kelly 2005 for discussion. 4 A further condition is often included in statements of Conciliationism – namely, that the person has no reasons independent of the disagreement itself to suspect that it is her peer who is making a mistake (e.g. reasons for thinking that her peer inattentive or intoxicated). See Feldman and Warfield 2010 for discussion. 5 Alleged connections between the Uniqueness Thesis and Conciliationism are discussed in Feldman 2007, Christensen 2007, 2009, Kelly 2010, Feldman and Warfield 2010, and Ballantyne and Coffman forthcoming.

The Role of the Uniqueness Thesis in the Epistemology of Disagreement

321

s is in fine-grained disagreement with s* about p =df s and s* have different credences toward p (where s and s* are persons and p is a content). The distinction between coarse-grained and fine-grained doxastic attitudes also gives rise to two versions of the Uniqueness Thesis:6 Coarse-Grained Uniqueness (CGU): Necessarily, for any body of evidence e and proposition p, there is a coarse-grained doxastic attitude d such that, for any person s whose total evidence is e, it is rational for s to have d toward p and irrational for s to have any coarse-grained doxastic attitude other than d toward p. Fine-Grained Uniqueness (FGU): Necessarily, for any body of evidence e and proposition p, there is a credence (i.e. degree of belief) c such that, for any person s whose total evidence is e, it is rational for s to have c toward p and irrational for s to have any credence other than c toward p. Suppose Alvin and Bob are in coarse-grained disagreement about whether capital punishment deters crime. Suppose also that they have the same total evidence. If CGU is true, then there is exactly one coarse-grained attitude that is rational for someone with Alvin and Bob’s evidence to have toward the proposition that capital punishment deters crime. Whether that attitude is belief, disbelief, or suspension of judgment, Alvin and Bob do not both have it, since they are in coarse-grained disagreement. So they are not in rational disagreement. More generally, if CGU is true and if any two people have the same total evidence but are in coarse-grained disagreement over any proposition, then at least one of the two holds an irrational coarse-grained attitude toward that proposition. Similarly, if FGU is true and if two people have the same total evidence but different credences toward the same proposition, then at least one of the two has an irrational credence toward that proposition. If it is stipulated that “epistemic peers” have the same total evidence, then CGU has the immediate consequence that there can be no rational coarse-grained 6

Lee MS contends that epistemologists should recognize a large family of uniqueness theses that are interestingly related to one another and each of which has interesting consequences. For the sake of simplicity, only two are formulated here.

322

Matthew Lee

peer disagreement, and FGU has the immediate consequence that there can be no rational fine-grained peer disagreement.7 With these relatively trivial results in mind, epistemologists of disagreement have claimed to find two important connections between the Uniqueness Thesis and Conciliationism. In order to state these alleged connections clearly, we should distinguish two relevant versions of Conciliationism: Coarse-Grained Conciliationism (CGC): Necessarily, for any persons s and s* and proposition p, if s realizes that s* is her peer with respect to p and that she herself believes p while s* disbelieves p, and if s has no independent reason to think that s* has made a performance error8, then it is irrational for s not to suspend judgment on p.9 Fine-Grained Conciliationism (FGC): Necessarily, for any persons s and s* and proposition p, if s realizes that s* is her peer with respect to p and that she is in fine-grained disagreement about p with s*, and if s has no independent reason to think that s* has made a performance error, then it is irrational for s not to change her credence toward p.10 The first alleged connection between Uniqueness and Conciliationism is a support relation running from Uniqueness to Conciliationism. Feldman (2007) argues for CGU as an intermediate step in securing the truth of CGC. Christensen (2007) argues for FGU as an intermediate step in securing the truth of FGC. Neither Feldman nor Christensen is explicit about how the argument from Uniqueness to Conciliationism is supposed to go. But one way it 7

If it is stipulated only that “epistemic peers” have the same evidence relevant to the proposition under dispute, then a further step is needed to secure the same results. It must be assumed that a difference in evidence irrelevant to a proposition can make no difference in the attitudes which it is rational to take toward that proposition. 8 A performance error is an error due not to incompetence, but to disruptions of the person’s competence arising from, e.g., uncharacteristic inattentiveness, distraction, fatigue, or intoxication. 9 See Lackey 2008 for a similar formulation of Conciliationism (what she calls “conformism”). 10 Feldman 2006, 2007 advocates Coarse-Grained Conciliationism. Christensen 2007 and Elga 2007 advocate Fine-Grained Conciliationism.

The Role of the Uniqueness Thesis in the Epistemology of Disagreement

323

might go is this. Given the Uniqueness Thesis, it is guaranteed that when two peers disagree, at least one of them has responded irrationally to their shared evidence. With that in mind, I should not react complacently to the revelation that a peer disagrees with me, provided that I have no independent reason for thinking it is my peer who has made the error. Agreeing to disagree is not a rational option. At least one of us has responded irrationally to our evidence, and I have no good reason for thinking that it isn’t I who has done so. It would be hubristic of me to be unwilling to budge; I should suspend judgment on the matter, or at least readjust my credence toward the disputed proposition. But one reason to doubt that the Uniqueness Thesis will be helpful to the Conciliationist is that advocates of anti-Conciliationist views sometimes grant (at least for the sake of argument) that the Uniqueness Thesis is true. For example, Kelly (2010) presupposes the Uniqueness Thesis when setting out his anti-Conciliationist “Total Evidence View.” On Kelly’s view, “what is reasonable to believe depends on both the original, first-order evidence as well as on the higher-order evidence that is afforded by the fact that one’s peers believe as they do” (Kelly 2010, 142). Kelly allows (in conformity with the Uniqueness Thesis) that one’s total evidence – firstorder plus higher-order evidence – may always recommend a unique doxastic attitude toward any given proposition. But he insists (in opposition to Conciliationism) that there are cases in which the higher-order evidence gained from revelation of peer disagreement “effectively counts for nothing in virtue of being overwhelmed by the first-order considerations” (Kelly 2010, 149). In such cases, the one doxastic attitude licensed by the person’s total evidence may be the same as the attitude that she held before learning of the disagreement; no conciliation is required of her. The opponent of Conciliationism, then, might grant that, in light of the Uniqueness Thesis, revelation of disagreement generally does produce some pressure to conciliate. But it is a further question whether this pressure to conciliate is ever overwhelmed or nullified by other features of the disagreement situation. The Conciliationist must still argue that it is not. The truth of the Uniqueness Thesis, then, is not enough to secure Conciliationism.

324

Matthew Lee

Furthermore, as several writers have pointed out11, the Uniqueness Thesis is an extremely strong thesis. It is incompatible with a wide range of views in traditional epistemology, as well as with virtually every occupied form of Bayesianism. But perhaps the best way to appreciate the extremeness of the Uniqueness Thesis is to compare it with evidentialism. The evidentialist holds that rationality supervenes on evidence (Feldman and Conee 2004, 101). That is to say, if any two persons are evidential duplicates, then they are also rationality duplicates – whatever attitudes are rational for one are rational for the other. Evidentialism is already a controversial thesis, and its detractors are numerous.12 But the Uniqueness Thesis represents a particularly radical form of evidentialism. For the Uniqueness Thesis entails evidentialism (the same attitudes are rational for persons with the same total evidence). But it is open to the evidentialist to allow that a given body of evidence might rationalize a range of different credences toward a given proposition (though always the same credences for anyone who possesses that body of evidence). This possibility is not allowed by the Uniqueness Thesis. The Uniqueness Thesis implies that only a single credence toward a given proposition is rational when in possession of a given body of evidence; all other credences are irrational. It is doubtful, therefore, that the Uniqueness Thesis will be especially helpful for the purpose of motivating Conciliationism. Rather powerful argumentation would be needed to establish the truth of the Uniqueness Thesis. And even if such argumentation could be supplied, we would need further reason to adopt Conciliationism over anti-Conciliationist rivals (such as Kelly’s Total Evidence View) that are compatible with the Uniqueness Thesis. 3. Is the Conciliationist Committed to the Uniqueness Thesis? The second alleged connection between Uniqueness and Conciliationism is a commitment relation: it is alleged that proponents of Conciliationism are implicitly committed to the Uniqueness Thesis. Kelly (2010) presses this claim as follows. If the Uniqueness Thesis were false, then there would be cases in which two peers are in a rational disagreement, wherein neither 11 12

See White 2005, Kelly 2010, and Ballantyne and Coffman 2011, forthcoming. See, for example, Foley 1993, Fantl and McGrath 2002, and Bergmann 2006.

The Role of the Uniqueness Thesis in the Epistemology of Disagreement

325

holds an irrational doxastic attitude toward the disputed proposition. If they realize that they are in such a disagreement, then the two peers have no reason to change their doxastic attitudes (by suspending judgment or decreasing credence). This would give us a counterexample to Conciliationism – a revealed peer disagreement in which the peers are not rationally required to conciliate. Thus, the Conciliationist is committed to denying that there can be such rational disagreements; the Conciliationist must endorse the Uniqueness Thesis. Given the radical character of the Uniqueness Thesis, the charge that Conciliationism incurs a commitment to the Uniqueness Thesis is a quite serious one. And there is a good deal of plausibility in Kelly’s argument that the Conciliationist is committed to Uniqueness. Why, after all, should I be required to conciliate if I know that I am in a rational peer disagreement in which my doxastic attitude is permitted by my evidence? And mustn’t the Conciliationist affirm the Uniqueness Thesis in order to rule out the possibility of such no-fault disagreements? It can be discerned upon reflection that the answer to the latter question is “No.” One need not embrace the Uniqueness Thesis in order to reject the possibility of the disagreement situation that Kelly describes. The situation at issue is one in which there is a body of evidence that (i) licenses two different doxastic attitudes toward the same proposition, (ii) is the total evidence of two different persons, and (iii) is known by both persons to license multiple attitudes toward the same proposition. Rejecting the Uniqueness Thesis only commits one to countenancing the possibility of bodies of evidence with feature (i).13 It is a further claim that there can be 13

Strictly speaking, rejecting the Uniqueness Thesis does not even commit one to this much. For one way to reject the Uniqueness Thesis is to hold that some bodies of evidence license no doxastic attitudes at all toward some propositions. We might say a person is in an “epistemic quandary” with respect to a proposition when there is no doxastic attitude that is rational for her to have toward that proposition. If epistemic quandaries are possible, then the Uniqueness Thesis is false, since the Uniqueness Thesis guarantees that there will always be exactly one attitude toward a given proposition that is licensed by a person’s evidence. But obviously someone who countenances epistemic quandaries is not thereby committed to the possibility of a body of evidence that licenses multiple attitudes toward some proposition. For simplicity’s sake, though, I will assume that epistemic quandaries are impossible. (And even if they are possible, there are theses in the neighborhood of Uniqueness that would serve Kelly’s purposes just as well.)

326

Matthew Lee

bodies of evidence that have features (ii) and (iii) in addition to (i). Kelly says nothing to motivate this further claim. There are a number of ways in which the Conciliationist might try to exploit this lacuna in Kelly’s argument. Perhaps the least plausible would be to maintain that permissive bodies of evidence (ones that license multiple credences toward some propositions) are essentially private – no two people can have the same permissive body of evidence. The situation Kelly describes would then be impossible, even though the Uniqueness Thesis is false, because the situation is one in which two parties have the same permissive body of evidence. A more promising strategy would be to claim that it is impossible to know of a permissive body of evidence that it is permissive. Or, more conservatively: it is impossible to know of a permissive body of evidence just which credences it licenses toward the propositions with respect to which it is permissive. Either way, the situation Kelly describes would be impossible, since it involves subjects knowing just which credences are licensed by their permissive evidence. More promising still would be to maintain that it is impossible to have as one’s total evidence a permissive body of evidence and, at the same time, to know which credences that body of evidence permits. This line of thought might be developed in various ways. But here is one way of developing it: When, given what you know, one credence c1 for a proposition p has just as much going for it epistemically as a different credence c2 (and vice versa), then rationality forbids you to arbitrarily choose c1 over c2 (or vice versa). Instead, you should split the difference by adopting a credence that is the arithmetic mean of c1 and c2. So, for example, if one advisor tells you that your total evidence licenses c1 for p and another equally reliable advisor tells you that your evidence licenses c2 for p, you should not arbitrarily follow the advice of the first and dismiss that of the second. Now suppose you somehow knew that your total evidence licenses having credence c1 for p. Then c1 would have to have something going for it epistemically. And suppose you somehow knew that your total evidence also licenses having c2 for p. Then c2 would have to have just as much going for it as c1 (if it didn’t, having c2 for p would not be permissible for you). But since c2 has just as much going for it as c1 (and vice versa), it would not be rational for you to arbitrarily choose c1 over c2 (or vice versa); you would need to split the difference. So your total evidence would not, after all, rationally license both c1 and c2 for p, contrary to the

The Role of the Uniqueness Thesis in the Epistemology of Disagreement

327

supposition. It is not possible, then, to have as one’s total evidence a permissive body of evidence and, at the same time, to know which credences that body of evidence permits. To reiterate: this is just one way for the Conciliationist to develop the idea that, even if permissive evidence is possible and even if such evidence can be possessed by multiple parties and even if it can be known which credences toward which propositions are permitted by such evidence, it is not possible to have as one’s total evidence a permissive body of evidence and, at the same time, to know which credences that body of evidence permits. There may be many other ways to develop that idea. If so, so much the worse for Kelly’s argument. We have found a significant weakness in Kelly’s argument, and there are potentially many ways for the Conciliationist to exploit that weakness. It is thus doubtful that Conciliationism incurs a commitment to the Uniqueness Thesis. 4. Concluding Remarks It is by no means clear, then, that the Uniqueness Thesis can play the roles that epistemologists of disagreement have hoped it could play. Using the Uniqueness Thesis to argue for Conciliationism does not seem to be a particularly promising strategy, given the radical character of the Uniqueness Thesis and given the compatibility of the Uniqueness Thesis with certain anti-Conciliationist views. And it seems that the Conciliationist is not, after all, committed to the Uniqueness Thesis (anyway, the main argument for that claim fails). Nevertheless, as I observed at the outset, the Uniqueness Thesis does have the important consequence that there can be no mutually rational disagreements between parties with the same total evidence. And that is enough to guarantee the Uniqueness Thesis an important role in the epistemology of disagreement, even if it cannot play the other roles that it has been invoked to play. References Ballantyne, N., and E.J. Coffman, forthcoming: Conciliationism and Uniqueness, Australasian Journal of Philosophy Ballantyne, N., and E.J. Coffman, 2011: Uniqueness, Evidence, and Rationality, Philosophers’ Imprint 11, 1-13

328

Matthew Lee

Bergmann, M.A., 2006: Justification without Awareness: A Defense of Epistemic Externalism, Oxford: Oxford University Press Christensen, D., 2009: Disagreement as Evidence: The Epistemology of Controversy, Philosophy Compass 4, 756-767 Christensen, D., 2007: Epistemology of Disagreement: The Good News, Philosophical Review 116, 187-217 Conee, E., 2010: Rational Disagreement Defended, in: Disagreement, ed. R. Feldman and T.A. Warfield, Oxford: Oxford University Press, 69-90 Conee, E., and R. Feldman, 2004: Evidentialism: Essays in Epistemology, Oxford: Oxford University Press Elga, A., 2007: Reflection and Disagreement, Noûs: 41, 478-502 Fantl, J., and M. McGrath, 2002: Evidence, Pragmatics, and Justification, Philosophical Review 111, 67-94 Feldman, R., 2007: Reasonable Religious Disagreement, in: Philosophers Without Gods: Meditations on Atheism and the Secular Life, ed. L.M. Antony, Oxford: Oxford University Press, 194-214 Feldman, R., and T.A. Warfield, 2010: Introduction, in: Disagreement, ed. R. Feldman and T.A. Warfield, Oxford: Oxford University Press, 1-9 Feldman, R., 2006: Epistemological Puzzles about Disagreement, in: Epistemology Futures, ed. S. Hetherington, Oxford: Clarendon Press, 216-235 Foley, R., 1993: Working Without a Net: A Study of Egocentric Epistemology, New York: Oxford University Press Goldman, A., 2010: Epistemic Relativism and Reasonable Disagreement, in: Disagreement, ed. R. Feldman and T.A. Warfield, Oxford: Oxford University Press, 187-215 Kelly, T., 2010: Peer Disagreement and Higher-Order Evidence, in: Disagreement, ed. R. Feldman and T.A. Warfield, Oxford: Oxford University Press, 111-174 Kelly, T., 2005: The Epistemic Significance of Disagreement, in: Oxford Studies in Epistemology, Vol. 1, ed. T.S. Gendler and J. Hawthorne, Oxford: Clarendon Press, 167-196 Lackey, J., 2010: What Should We Do When We Disagree? In: Oxford Studies in Epistemology, Vol. 3, ed. T.S. Gendler and J. Hawthorne, Oxford: Clarendon Press, 274-293 Lee, M., Ms.: Sharpening Up the Uniqueness Thesis White, R., 2005: Epistemic Permissiveness, Nous-Supplement: Philosophical Perspectives 19, 445-459 Williamson, T., 2000: Knowledge and its Limits, Oxford: Oxford University Press

Counterfactual-Peer Disagreement KATHERINE MUNN, University of Oxford

Abstract The peer-disagreement scenario assumes that agents share evidence: a peer is someone whom you think shares your evidence and is just as likely as you are to evaluate that shared evidence correctly. But often evidence isn’t shared. In this case, the problem of counterfactual-peer disagreement looms: How should you rationally respond to the disagreement of someone whom you think is as likely as you are to evaluate the evidence correctly had you shared it, and whom you think is as likely to evaluate her evidence correctly as you are to evaluate yours correctly? This problem is more intractable and widespread than peer disagreement. I’ll suggest a way to mitigate it, the limitations of which underline the urgency of greater research into the problem of counterfactual-peer disagreement. Introduction: Peer and Counterfactual-Peer Disagreement The peer-disagreement thought experiment isolates an epistemological puzzle. Holmes and Poirot share evidence about whether p. Before comparing notes on their respective evaluations of the evidence, Poirot counts Holmes as his epistemic peer: he believes1 (a) that he and Holmes are equally likely to evaluate evidence correctly,2 and (b) that he and Holmes share evidence. 1

Things become swiftly more complicated if we take belief to admit of degree; for present purposes we won’t. Here, beliefs about what a body of evidence supports are beliefs which ascribe probability values to propositions given only that evidence. 2 Cf. Elga’s (2007) usage of ‘epistemic peer’. This paper assumes that agents think themselves highly likely or even certain to evaluate evidence correctly.

330

Katherine Munn

Concerning (a): An evaluation of a body of evidence is a statement of how probable some proposition p is given only that evidence. As far as evaluating evidence correctly goes, the peer-disagreement setup assumes that some probability assignments for a proposition p given a body of evidence are correct and that others are not – or that some probabilities are more correct than others. But it is committed to no particular way of spelling out the notion of correctness at play. Finally, presumably an agent’s estimation of her own and hence her peer’s likelihood to evaluate evidence (about whether p) correctly is itself, indirectly, evidence about whether p. Concerning (b): precise conditions for evidence sharing will be given below. The thought experiment continues: Poirot enters into it taking Holmes to be his epistemic peer. But then, after he and Holmes independently evaluate their evidence about whether p, he discovers that he and Holmes disagree: Holmes assigns p a high probability on their evidence, whereas Poirot assigns it a low one. The only difference between Holmes and Poirot, as far as Poirot can tell, is the way in which each has evaluated the shared evidence. How should Poirot rationally respond? In other words, what is the correct evaluation of p’s probability, given the shared evidence plus the evidence that Holmes disagrees with Poirot? One suggestion is that Poirot should re-assess his original assumption that Holmes was his peer: Perhaps the discovery that Holmes has evaluated their evidence differently itself counts as evidence, for Poirot, that they are not peers. This would stand to reason, if Poirot’s assessment of the probability of agreement depends in part on statistical inference from actual cases. But suppose that Poirot doesn’t take this isolated disagreement to change his assessment of their probability of evaluating the evidence correctly. In this case, the peer-disagreement puzzle arises: Given that Poirot still judges Holmes to be his peer, what is the correct way for Poirot to evaluate the shared evidence about whether p, plus the evidence supplied by the peer disagreement itself?3 In particular, what should his all-thingsconsidered probability assignment be for p? The peer-disagreement scenario requires that at least one of the agents, the agent from whose point of view the scenario is narrated, takes (a) and 3

Since the puzzle is presented from Poirot’s point of view, and is hence a puzzle for Poirot, it doesn’t matter here how Holmes has reacted to the disagreement, if he has even discovered it, or whether he takes Poirot to be his peer.

Counterfactual-Peer Disagreement

331

(b) to obtain. But often they do not obtain and there is little reason for him to think that they do. In this case, there are logically three other possible views he may take (assuming he has any opinion on the matter). He may think: I. II. III.

that he and his interlocutor are not equally likely to evaluate evidence correctly, but that they do share it (i.e., that (a) fails to obtain but that (b) obtains), that they are not equally likely to evaluate evidence correctly, and that they do not share it (that neither (a) nor (b) obtains), or that they are equally likely to evaluate evidence correctly but that they do not share it (that (a) obtains but (b) fails to obtain).

Let’s focus on situations of type III, which we’ll call counterfactual-peer disagreements. (The notion of counterfactual peerhood will be defined below.) In such situations, another problem looms: how should you rationally respond to the disagreement of someone whom you think would be as likely as you to evaluate the evidence correctly were you to share it, and whom you think is as likely to evaluate his evidence correctly as you are to evaluate yours correctly? In other words: what is the correct evaluation of the total evidence, which now includes the proposition that your interlocutor is as likely as you are to have evaluated evidence correctly, but that you and she have different evidence? This problem is more intractable and widespread than peer disagreement. After clarifying and motivating it, I’ll suggest a way to mitigate it, but it cannot be completely overcome. The upshot will be an appeal for urgent further research into the problem of counterfactual-peer disagreement. The Evidence-Sharing Problem The counterfactual-peer disagreement problem arises when evidence is not shared. This section clarifies the problem by giving some minimal necessary conditions on evidence-sharing, which, it turns out, are very hard to meet. An agent’s evidence is the set of propositions which, minimally, the agent believes. Most accounts take evidence to be subject to additional constraints, requiring for example that evidence be true, justifiedly believed, or known (see e.g. Williamson, 2000; Conee and Feldman, 2004). The particular constraints do not matter for my general point that evidence sharing is very rare,

332

Katherine Munn

but the more constraints there are, the rarer evidence sharing will be. As for propositions, I am thinking of them in the minimalist sense as the referents of ordinary ‘that’-clauses. Some may object that a propositional construal of evidence is too narrow, excluding the possibility that an agent’s experiences provide her with evidence, but this objection is off the mark. Propositions can contain demonstratives which refer to experiential content, e.g., My experience was like that. Since any experience can receive such treatment, any experience can provide evidence (see Williamson, 2000; Millar, 1991). Propositions are a neat way to encapsulate the content of an agent’s evidence, regardless of where that evidence came from.4 Evidence-sharing, in the sense needed for peer disagreement, is rare, since this sense must pinpoint disagreements caused only by differences in how agents evaluate evidence, not in the evidence itself. In order for two agents to share evidence about whether p, both must at least: (i) (ii) (iii) (iv)

use all and only the same propositions to reason about whether p; fulfill any constraints on evidence for each of those propositions; grasp each of these propositions; think that each of these propositions is just as probable5 as the other agent does.

Two agents fulfill (i) just in case they are think that p’s probability conditional on each proposition is either higher or lower than p’s unconditional probability, or, if they lack beliefs about probability, if they are disposed, should the propositions in question obtain, to be more or less confident that p is the case than they would be if they did not obtain. (i) excludes disagreements which arise because agents appeal to different considerations in their reasoning about whether p. (ii) ensures that the propositions which each agent uses to reason about whether p are contained in his evidence. Thus both agents must minimally believe each proposition, and, if there are other constraints (which arguably there are), such as that those beliefs are justified or count as knowledge, or that those propositions are true, then both agents must fulfill these too. 4

Thus I am not thinking of propositions as sets of logical possibilities. I am individuating them more finely, as sets of epistemic possibilities. 5 By ‘probable’, I mean probable unconditional on any of the other evidence which either agent uses to reason about whether p, or indeed unconditional on p itself.

Counterfactual-Peer Disagreement

333

As for (iii), an agent grasps a proposition p just in case p represents, for him, a way things could be. I may not grasp a proposition about your experience of (say) childbirth, especially not if you can’t describe it to me using locutions other than demonstratives (you may be able, breathlessly, to express yourself only by saying, ‘It was like that’). In this case I won’t grasp the proposition you have expressed by this utterance. (iii) excludes disagreements which arise because, although both agents agree that (e.g.) the proposition Holmes’s experience was like that evidentially affects whether p, Poirot does not understand the referential content of ‘that’ and so cannot directly judge the evidential force of Holmes’s experience was like that. He may be able to judge it indirectly, that is, via mediation by Holmes’s testimony: Holmes could utter the proposition My experience has led me to raise (lower) my probability that p to such-and-such a value. But there is little reason to think that such a testimonial proposition has the same evidential force vis-à-vis p as the proposition Holmes’s experience was like that does. (iv) is needed because a difference in the agents’ probability assignments to their evidence will result in differing probability assignments to p. Imagine that both detectives think that rain would be 75% likely given that the weather forecaster has said ‘Rain is 75% likely’. Even if both trust the weather forecaster completely, if Holmes is certain that the forecaster said this, whereas Poirot is only half-sure, Poirot’s probability assignment to rain should be much lower than Holmes’s (barring other evidence). Given these constraints, I take it that, if evidence is shared, it is only in very simple situations. First, the more constraints which condition (ii) places on evidence, the harder it will be for two agents to share evidence. Moreover, when a disagreement pertains to the moral, political, and religious questions motivating much of the disagreement literature (see e.g. Lackey, 2010 and forthcoming; Christensen, 2007; Feldman, 2007; Kelly forthcoming and 2010) (call them tough-nut issues), agents never share all of their evidence. Call this the evidence-sharing problem. The reason is that part of their evidence will take the form of propositions about experiences which one agent has had but which the other agent does not grasp, violating condition (iii). This prevents both agents from being able to evaluate all of their combined evidence for themselves, violating condition (i). Condition (iv) will be hard to meet too, because many reasons may prevent an agent from assigning the same probability to a proposition which he cannot grasp as an agent who can grasp it assigns. There may

334

Katherine Munn

well be other necessary conditions for evidence-sharing; if so, evidencesharing only becomes more difficult.6 The evidence-sharing problem is important: if evidence is not shared, peer disagreement does not arise, and an account of how to handle it is useless (although perhaps theoretically interesting). One might object, claiming that accounts of rational peer disagreement can be applied ‘roughly’. The idea, I take it, is something like this: if an account of rational peer disagreement prescribes that both agents, say, ‘conciliate’ by adjusting their probability assignments to be closer to each other’s by suchand-such a value, then to apply such an account “roughly” to a counterfactual-peer case is to recommend that they move their probability assignments closer to each other’s but perhaps not by precisely that value. Or if an account of peer disagreement lists factors which determine the extent to which agents should conciliate or ‘stick to their guns’, this recommendation should be followed in non-peer cases too. But this strategy is deficient. First, the problem of peer disagreement itself is still a subject to fierce debate. The correct account to apply to counterfactual-peer cases would depend on the correct account to apply to peer cases, and this is far from settled. Second, even if there is a correct account of peer disagreement, it is not at all clear that such an account can or should be applied ‘roughly’ to a counterfactual-peer case. Its results would be either significantly distorted or highly non-specific: it would not be able to zero in on any particular probability assignment to p which either agent should adopt; if anything it will give a rough ballpark, and all the rougher the more evidence is unshared. Counterfactual-peer disagreement, as the next section will argue more closely, is a problem in its own class and hence requires a unique solution. Counterfactual Peerhood The evidence-sharing problem reveals that peer disagreement is only the tip of a giant iceberg, with counterfactual-peer disagreement still submerged. This problem arises as follows: Imagine that Holmes and Poirot 6

An additional necessary condition, which might derive from the assumption that there are correct and incorrect ways to evaluate evidence, might be that the evidence which the agents use to reason about whether p in fact be evidence about whether p.

Counterfactual-Peer Disagreement

335

share some evidence about whether p, but not all. Poirot thinks that Holmes is his counterfactual-peer, i.e.: (a*) (b*) (c*)

that Holmes is just as likely to evaluate Holmes’s evidence correctly as Poirot is to evaluate Poirot’s correctly; that, were Poirot and Holmes to share evidence, they would be equally likely to evaluate the shared evidence correctly, and hence that they are equally likely correctly to evaluate whatever evidence they do share.

Counterfactual-peer disagreement always has at least one cause, namely: (A) differences in the agents’ evidence. But it often has a second cause, too, namely: (B) differences in how agents evaluate whatever evidence they do share. This makes counterfactual-peer disagreement much thornier than peer disagreement, which by definition is only ever caused by (B). It leaves untouched disagreements caused by (A) or (even worse) by both (A) and (B). Although, because evidence-sharing is so rare, counterfactual-peer disagreement is more widespread than peer disagreement and perpetrates tough-nut disagreements, it might take a moment to appreciate. After all, perhaps both agents evaluated their respective evidence correctly, in which case the unshared components of their evidence are to blame for their different probability assignments to p. Thus both parties may be equally reasonable in spite of the disagreement; what’s the problem? The problem is p. Holmes assigns it a high probability and Poirot a low one. If Poirot wants high probabilities in truths, rather than just reasonableness relative to his evidence, he should worry. For the disagreement provides evidence that there is powerful evidence favoring p (namely, Holmes’s) which Poirot does not share. Since Holmes is Poirot’s counterfactual peer, if Poirot shared Holmes’s evidence, his probability assignment to p would likely be much higher than Poirot takes his own evidence currently to warrant. Never mind that Poirot has evidence favoring not-p which Holmes lacks. If Poirot wants a high probability in the truth about whether p, he should want Holmes’s evidence too.7 7

This is so even if Poirot has strong evidence that Holmes’s evidence is misleading. Poirot’s evidence that it is misleading will safeguard him against taking it too seri-

336

Katherine Munn

Thus behind the woolly problem of peer disagreement looms the woollier problem of counterfactual-peer disagreement. It is woollier because more realistic; because it involves two possible causes, not one; and because, unlike peer disagreement, it infects tough-nut issues. I merely call attention to this underappreciated problem; I don’t resolve it. But I’ll also deliver hope, arguing that the evidence-sharing problem underlying counterfactual-peer disagreement can be mitigated: Poirot can come closer to sharing at least some of Holmes’s evidence, in particular his experiential evidence. Moving Toward Evidence-Sharing? Suppose Poirot perfectly shared Holmes’s evidence;8 then he could assess it himself. If so, either he would assess Holmes’s evidence in the same way as Holmes does, in the sense of assigning p the same probability conditional on it as Holmes assigns, or he would assess it differently. If Poirot assessed Holmes’s evidence in the same way as Holmes does, then their disagreement would likely be fully explicable in terms of (A), differences in their evidence. Poirot would have some evidence about whether p which Holmes lacks, even though Holmes has evidence which Poirot lacks. Now that Poirot has both his evidence and Holmes’s, he may adjust his probability assignment for p in light of it and rest easy. But if Poirot perfectly shared Holmes’s evidence (regardless of whether Holmes shared Poirot’s) and disagreed with Holmes’s assessment of Holmes’s evidence, then their disagreement is explicable only in terms of (B), differences in how they evaluate their evidence. In this case, Poirot may determine the correct response by applying the right account of peer disagreement (pending discovery thereof). In summary: if the evidencesharing problem could be overcome, then parties to counterfactual-peer disagreements could either resolve the disagreement by sharing evidence, or, if this fails to bring resolution, focus exclusively on the remaining problem of peer disagreement.

ously, but it will put him in a better position than he will be in if his evidence about its alleged misleadingness is ever undermined. 8 This is regardless of whether Holmes also shares Poirot’s evidence.

Counterfactual-Peer Disagreement

337

Even though (as I’ve argued) sharing evidence is very rare, I’ll now argue that agents can at least sometimes come closer to sharing at least one kind of evidence: the sort which refers to each agent’s experiences. I’ll do so by arguing that agents can come closer to meeting condition (iii), which will at least facilitate meeting some of the other conditions, too. But how can agents share, or even come closer to sharing, evidence which refers to agents’ individual experiences? Sharing experiential evidence, one might object, is absurd; condition (iii) on evidence-sharing prevents it. Poirot cannot possibly grasp the proposition E

Holmes had that experience,

in the sense of entertaining the content of that. For grasping E requires phenomenological duplication of Holmes’s experience (arguably including the phenomenology of being Holmes). I have two responses. First, it isn’t clear that grasping E requires phenomenological duplication of Holmes’s experience; it may require only that Poirot have the same type of experience (minus Holmesphenomenology, inter alia). If so, then, if Poirot can have a similar experience, he might yet fulfill criterion (ii). Second, even if grasping E does require that Poirot duplicate Holmes’s experience, little hangs on this. I claim only that Poirot can come closer to sharing E with Holmes. The idea is this: If Poirot has an experience similar to the one referred to by Holmes’s evidence E, then he is closer to sharing E than (say) Watson, who has had no such experience. Poirot may not grasp E itself, but he grasps the proposition: E*

Holmes had an experience like that,

where ‘that’ refers, this time, to Poirot’s experience.9 Acquiring E* will still put Poirot in a better position vis-à-vis condition (iii) (that both grasp E) than he would enjoy, all else equal, without grasping E*: he would have at 9

How can he and Holmes ascertain that they have had similar experiences? Possibly they cannot for certain, but it is reasonable to suppose that, by describing their experiences to the extent that they can, and comparing notes about what they take their experiences to support evidentially (which will be more effective since both take each other to be peers), they can come reasonably to believe that their experiences were relevantly similar.

338

Katherine Munn

least some sense of what Holmes’s experiential evidence is, even if he did not grasp it. And he would certainly be better off vis-à-vis condition (iii) than Watson (regarding E). Grasping E* might bring Poirot closer to fulfilling condition (iv) (that agents must assign the same probability to their evidence), if doing so moves him to assign the same probability to E as Holmes does, or at least a probability which is closer to Holmes’s than Poirot’s pre-experience probability assignment. For if Poirot has an experience which resembles Holmes’s, he may be rationally entitled to raise the probability which he assigns to E itself (the proposition Holmes had that experience). As for condition (ii), believing E*, and meeting any other constraints on having E* as evidence, will perhaps make it less of a rational leap for him to believe E.10 Finally, if Poirot’s experience resembles Holmes’s enough, Poirot comes closer to fulfilling condition (i), for having his experience enables Poirot to reason about whether p if not from E, then from E*, which has similar content to E. The objector may persist: But why think that one can move closer to evidence-sharing – why think that Poirot can bring himself to have an experience which is even remotely similar to Holmes’s? In answer, assume that much of the description-defying phenomenology of certain experiences, especially experiences providing evidence for tough-nut opinions, is emotional. Assume further that emotional experience can sometimes legitimately provide evidence. It is reasonable to think that emotions can be indicators of the way things really are, if fallible ones. If Poirot can experience emotions relevantly similar to those which Holmes’s experience involved, he may come to grasp either E or E*, and thus be able to incorporate one of these into his reasoning about whether p. For example, take the tough-nut proposition P 10

Unregulated laissez-faire economics is right.11

This assumes of course that it is possible to believe a proposition which one does not grasp, let alone be justified in doing so or have one’s belief count as knowledge, all of which are admittedly dubious. 11 I assume there are truths about morality. Anyone who doesn’t should anyway care little about peer or counterfactual-peer disagreement, at least insofar as these problems pertain to probabilities about moral truths. But such a person might also think that there are no truths about reasonableness relative to one’s evidence, in which case the disagreement problems won’t interest her at all.

Counterfactual-Peer Disagreement

339

Poirot assigns a high probability to P, whereas Holmes assigns a low one. Assume that Holmes’s low probability does not demote him from peerhood in Poirot’s estimation. They share the evidence E1

Unregulated laissez-faire economics leads to exploitation,

but they have different assessments of the evidential import of E1 relative to P. Both think that E1 lowers P’s probability, but Poirot thinks that it lowers that probability by much less than Holmes thinks it does. (Poirot thinks that exploitation results equally from any economic policy and thus disconfirms the rightness of each by only a little; this may be additional evidence which he and Holmes do not share.) Holmes, by contrast, thinks that E1 lowers P’s probability significantly. The reason is that he has additional evidence E, where ‘that’ refers to a phenomenology of witnessing factoryworker exploitation. (Holmes reasons that no policy can be right which permits such suffering, and if all policies do, then none is right.) Poirot, who has had no such experience, is at best compromised in his ability to assign P (or any other proposition) a probability given E. Thus Holmes has evidence which Poirot lacks and which, if Poirot shared it, might dramatically change the probability which Poirot assigns to P.12 One obstacle to Poirot’s sharing E – the one on which we shall focus – is (iii), grasping E. If grasping E requires identical phenomenology to Holmes’s, then Poirot cannot grasp E but might grasp a proposition, E*, about a similar experience. And if grasping E is compatible with having a similar if not identical experience to Holmes’s, then so much the better. Thus Poirot may move toward sharing E, say, by taking measures to witness laissez-faire-induced worker exploitation, or similar suffering which (based on Holmes’s testimony) compares with Holmes’s experience. Failing this, he might imagine such suffering, prompted for example by memoirs or films, to trigger the emotions he would have were he to witness it (for a fascinating discussion of the power of the imagination and its interaction with belief see Gendler, 2006; and Gendler and Kovakovich, forth12

Although Poirot doesn’t share E, he does have the evidence that Holmes witnessed laissez-faire-induced suffering. One might think that this evidence-of-evidence should itself lower the probability which Poirot assigns to P. This view, however, presupposes a solution to the counterfactual-peer disagreement problem, which is precisely the problem of whether and how Poirot should change his probability assignment to P.

340

Katherine Munn

coming). I suggest that triggering such emotions can bring Poirot closer to sharing E by bringing him to grasp either E or E*, and thus closer to meeting condition (i) for evidence-sharing, that agents reason from the same propositions. And if Poirot doubted whether Holmes did have such an experience, thus assigning E a lower probability than Holmes and thereby violating (i), perhaps Poirot’s experience will generate trust in Holmes, which may provide evidence (via the proposition Holmes is trustworthy) that Holmes’s assessment of Holmes’s own experiential evidence is correct. This in turn may be evidence which moves Poirot’s probability assignment for E closer to Holmes’s. Condition (ii), that agents meet the same evidential conditions, remains the stickiest to fulfill, depending on what the conditions on evidence are, but fulfilling (i), that agents reason from the same propositions, at least gives Poirot a shot at doing so. This suggestion for moving toward evidence-sharing doesn’t solve the evidence-sharing problem, except perhaps for very simple cases. By and large, counterfactual-peer disagreement still haunts us. In tough-nut cases especially it is a real problem, unlike the merely theoretical problem of peer disagreement. I’ll thus end by urging further discussion of counterfactual-peer disagreement: How, if at all, should agents’ probability assignments change in light of the disagreement of a counterfactual peer? A solution may (but need not) take its cue from the peer-disagreement literature, where two broad approaches suggest themselves. Conciliationism says that a peer disagreement provides evidence in favor of moving one’s probability assignment closer to one’s interlocutor’s (see e.g. Christensen, 2007; Elga, 2007). Recalcitrance (as represented e.g. by Kelly, forthcoming) by contrast, says that peer disagreement provides no such evidence at all.13 Analogous approaches suggest themselves for counterfactual-peer disagreement. Conciliationism might look like this: Poirot lacks some of Holmes’s evidence, but Holmes’s testimony assures him that that there is evidence which supports a high probability assignment to p. Poirot should thus adjust his low probability for p in the direction of Holmes’s, Holmes being (in Poirot’s estimation) just as likely as Poirot to evaluate evidence correctly (and Poirot thinking himself skilled in evidence13

Compromise views, such as Lackey’s (2010 and forthcoming) and Kelly’s (2010) suggest conditions under which the disagreement provides more or less evidence to this effect.

Counterfactual-Peer Disagreement

341

handling). Another take, however, suggests that counterfactual-peer disagreement is not evidence for Poirot for p; that Poirot should be recalcitrant, maintaining his own probability assignment. The reason is that, although Poirot has evidence that there is evidence (namely, Holmes’s) speaking loudly for p, he is also aware of evidence which Holmes lacks (namely, Poirot’s own) which speaks loudly against p. If Holmes shared Poirot’s evidence, Holmes would or should see its force and move his probability assignment toward Poirot’s. Surely these two pieces of evidence-of-evidence cancel out, and Poirot should maintain his own probability assignment. Conclusion Counterfactual-peer disagreement is a real problem, since disagreements are most often caused not just by differences not just in how agents evaluate evidence but in their evidence itself. Accounts of peer disagreement only address disagreements caused by the former. One reason why evidence-sharing may fail is that at least one agent’s evidence refers to experiences which the other agent has not had, and hence she cannot even grasp that evidence. I have suggested that this obstacle to evidence-sharing can be mitigated, if not overcome, by the gaining of new experiences which resemble the experiences which provide one’s interlocutor with the unshared evidence. But counterfactual-peer disagreement is a can of worms. Manifold questions remain about both counterfactual-peer disagreement and how my suggestion about moving toward evidence-sharing might be developed or play out in practice. I offer such questions for future research. References Christensen, D., 2007: Epistemology of Disagreement, The Philosophical Review 116, 187-217 Conee, E., and R. Feldman, 2004: Evidentialism: Essays in Epistemology, Oxford: Oxford University Press Elga, A., 2007: Reflection and Disagreement, Noûs 41, 478-502 Feldman, R., 2007: Reasonable Religious Disagreements, Philosophers without Gods ed. L. Antony, Oxford: Oxford University Press Gendler, T.S., 2006: Imaginative Contagion, Metaphilosophy 37, 183-203

342

Katherine Munn

Gendler, T.S., and K. Kovakovich, forthcoming: Genuine Rational Fictional Emotions, in: Contemporary Debates in Aesthetics, Oxford: Blackwell Kelly, T., forthcoming: The Epistemic Significance of Disagreement, in: Oxford Studies in Epistemology, ed. J. Hawthorne and T. Gendler Kelly, T., 2010: Peer Disagreement and Higher-Order Evidence, Disagreement, ed. T.A. Warfield and R. Feldman, Oxford: Oxford University Press, 111-174 Lackey, J., 2010: What Should We Do When We Disagree? Oxford Studies in Epistemology 3, 274-293 Lackey, J., forthcoming: A Justificationist View of Disagreement’s Epistemic Significance, Social Epistemology, ed. A. Haddock, A. Millar, and D. Pritchard, Oxford: Oxford University Press Millar, A., 1991: Reasons and Experience, Oxford: Clarendon Press Swinburne, R., 2001: Epistemic Justification, New York: Oxford University Press Williamson, T., 2000: Knowledge and Its Limits, Oxford: Oxford University Press

Meta-Induction and the Problem of Fundamental Disagreement GERHARD SCHURZ, University of Düsseldorf

Abstract Proponents of competing word-views, such as science versus religion or naturalism versus constructivism, are in fundamental disagreement. They disagree in their underlying cognitive system. How can fundamental disagreement be overcome? Expert-opinions don’t provide a solution, because agreement among experts is relative to shared cognitive systems. Is there a way out of the resulting world-view-relativity? In this paper I argue that the method of meta-induction provides a way out. While in earlier papers I developed the method of meta-induction for prediction games, I generalize this method in this paper to arbitrary action games. In the final section I sketch some applications and some limitations of my results. 1. Introduction: Remarks on ‘Reasonable Disagreement’ 1.1 Justification and cognitive (background) systems. Prima facie, the justifiedness of a belief, or an action, of a given person or subject P, depends not only on P’s total evidence. It depends also on P’s accepted cognitive (background) system. This system contains fundamental principles of reasoning that determine the criteria for justification which are accepted by the person. An epistemologist who is oriented towards objective truth (I myself belong to this species of epistemologists) will object that what is decisive for justification are not accepted but correct principles of reasoning. However, there is sustainable disagreement in epistemology about the ingredients of the correct  or ‘right’, ‘best’  cognitive system. Therefore, to be as general as possible, I start with a notion of justification which makes this notion dependent on the accepted cognitive system. Assuming that one has a definition of the ‘right’ cognitive system, one may define justification proper, in distinction from subjectively accepted justification,

344

Gerhard Schurz

as justification based on this ‘right’ cognitive system. But as I said, I want a most general and presupposition-free epistemological starting point. A cognitive system corresponds to what Goldman (2006) has called an “epistemic system”, to what Kuhn (1962) has called a “paradigm”, or also to what is called a “world-view”. Cognitive systems include fundamental principles of theoretical and practical reasoning that are accepted by the given person, or are accepted standards in the culture with which the person identifies herself. This characterization allows that a person’s reasoning may be fallacious, i.e., it may contradict the standards that are accepted by the person or by the culture to which the person’s cognitive system belongs. Cognitive systems are usually metaphysically loaded, i.e., they make implicit assumptions about the world in which one is in (e.g., that the world is uniform, or that it has been created by a benevolent being, etc.). Moreover, a cognitive system determines, at least partially, what counts as an admissible evidence (e.g., only observation and proof, or also authority, or even mere intuition, etc.). Based on this characterization of justification, we discuss in the next subsections the possibility of reasonable disagreement, i.e., of situations in which persons disagree in regard to mutually contradicting beliefs, know that they disagree, and yet are justified (and believe that they are justified) in upholding their mutually contradicting beliefs. 1.2 Why reasonable disagreement among epistemic peers is hardly possible. Epistemic peers are, by definition, persons who share both their cognitive system and their total evidence, in regard to a given subject or quest. Moreover, they assume that their cognitive capacities are mutually comparable (at least qualitatively). There exists a strong argument, going back to Feldman (2006), why disagreement among peers cannot be reasonable: if two persons P1 and P2 are epistemic peers, then P2’s belief that she is justified in believing B constitutes a defeater for P1’s justification of his belief in non-B. For P2 knows that P1 has exactly the same evidence as she has; moreover, he shares with her all cognitive standards (and has comparable cognitive abilities). This implies that there is a non-negligible chance that P2 is mistaken in her belief in B, and, hence, is a reason for P2 to withdraw here belief that she is (strongly) justified in her belief in B; and the same applies to P1. I think this argument resists various objections (cf. Conee 2010) which cannot be discussed here. In conclusion, reasonable disagreement among peers is not possible, at least not in the strong form in

Meta-Induction and the Problem of Fundamental Disagreement

345

which the peers uphold their belief that their (conflicting) beliefs strongly justified. 1.3 Disagreement because of different evidence. Two persons who share their cognitive system may disagree because they have different evidence. What they should do in this case is, of course, to exchange their evidence and include the evidence of the other persons as part of their own total evidence. Under the assumption that they share their cognitive system, this is easily possible, with the effect that the persons will then share their evidence.1 So also in this case, reasonable disagreement is not possible. 1.4 Fundamental disagreement: can it be reasonable? If two persons disagree because they rely on different cognitive systems, I speak of fundamental disagreement. A paradigm example of competing cognitive systems or world-views are the cognitive systems of a religious culture, with holy scriptures and authorities as the fundamental source of knowledge, versus the cognitive system of a scientific culture, with empirical induction (and the like) as the fundamental source of knowledge. Can fundamental disagreement be reasonable? I think, yes: Upholding a fundamental disagreement can be reasonable if and only if one of the following assumptions is true : (a) (b)

either there don’t exist any objective standards for better or correct cognitive systems at all (a position that has been developed, e.g., by Boghossian 2006), or such standards exist, but they are cognitively inaccessible (a position that has been suggested as a possibility by Goldman 2010).

Both cases of reasonable fundamental disagreement imply a kind of epistemological relativism: (a)’s relativism is based on a kind of radical constructivism (“all ‘knowledge’ is subjective construction”), and (b)’s relativ-

1

Conee (2010) has objected that “my evidence” counts for me more than “your evidence”. But given that “I” and “you” share the cognitive system, “my” ignorance of “your” evidence would be a clear sign of irrationality. So I don’t think that Conee’s point constitutes a strong argument against the impossibility of rational disagreement in this case.

346

Gerhard Schurz

ism results from a radical justificational skepticism (“objective justifications are out of the reach of human minds”). In conclusion, fundamental disagreement can be reasonable when the standards of evaluation are clearly subjective. In such a situation, the value of subjective freedom requires from us to accept the given diversity of subjective perspectives, and the only intersubjectively justifiable metarules of reasoning are pluralism, tolerance, and mutual respect. However, in domains where one possesses intersubjectively accepted goals (such as truth in theoretical or well-being in practical reasoning), and intersubjectively shared standards of evaluation, fundamental disagreement seems to be a sign of unreasonableness, and an indicator of error. Disagreement in problems which are broadly accepted to be not a matter of mere subjective preferences leads to ongoing controversies, or even fights, and thus should be overcome by finding a reasonable agreement. 1.5 But how can fundamental disagreement be overcome? The answer that I suggest in this paper is: by cognitive methods that are universal in the sense of being reasonable in every cognitive system. Or in other words, by methods that have a justification that is world-view-neutral. 2. Meta-Induction in Action Games In this paper I suggest meta-induction as a candidate for a universal method, or more precisely, a universal meta-method. Meta-induction recommends to trust an epistemic or practical method  which is part of a given cognitive system  to the extent that it has been successful in the past. Meta-induction is a success-based social learning method: the meta-inductivist compares the success rates of all accessible methods (applied by other agents) and applies an optimal combination of these methods, in terms of their so-far observed success-rates (precise definitions see below). The thesis that I defend is that the application of meta-induction is optimal in the sense that in every possible world or situation, the application of this method can only improve and never diminish one’s success  apart from small short run losses. Two important features are the following: (1.)

Meta-induction is world-view neutral. For example, if an esoteric healing method turns out to be more successful than a standard

Meta-Induction and the Problem of Fundamental Disagreement

(2.)

347

medical method, the meta-inductivist sticks to esoteric healing method. Meta-induction, being a meta-method (or social learning method), is not recommendable in isolation. It is equally important to have good self-sustained object-level methods, which (by definition) do not imitate other methods, and among which the metainductivist can select. We assume that the meta-inductivist has always at least some object-level methods at hand which she can perform in the absence of other agents whose method could be exploited.

There is only one condition that must hold for meta-induction to be applicable: they must exist intersubjectively shared standards for observable success of a given method over a given domain. I argue that typically human cultures do have such intersubjectively and interculturally shared standards of success. Important examples are, e.g., all kinds of forecasting techniques (e.g. weather forecasting), all kinds of healing practices, of hunting or agricultural planting techniques, warfare techniques, etc. I argue that in all these domains, meta-induction is a cross-culturally applicable universally optimal method. This is of particular importance in cultures and domains in which there don’t exist monopolistic scientific standards, but rather a heterogeneous variety of methods stemming from different cognitive systems. Generally speaking, my account of meta-induction belongs to the framework of learning theory (cf. Kelly 1996), but it is enriched by the multi-agent perspective of game theory. In Schurz (2008, 2009a) I introduced meta-induction as the basis of a new approach to Hume’s problem of induction. In these papers I have applied meta-induction only to prediction methods. In this paper, I will apply meta-induction to arbitrary methods of action. For this purpose, I develop the notion of an action game, in analogy to the notion of a prediction game as developed in Schurz (2008). My approach will include predictions as a special kind of actions, but will be more general than my account of prediction games. A action game (,(n),p) consists of (1) (2)

An infinite sequence (n) := 1,2, of rounds. A finite set of players P whose task it is to perform an action an(P) in each round n (e.g., a weather forecasting action (i.e., a

348

(3)

(4)

Gerhard Schurz

prediction), a healing action, planting action, hunting action, etc.). We assume that the player’s actions are taken from a fixed (and arbitrary large) set of possible actions, that need not all be realized by the players. Payoff function: The “reaction of the environment” is reflected in terms of a payoff function p. The payoff pn(P) of the action of a player P in round n is measured within the normalized interval [0,1] (we write pn(P) short for pn(an(P))). We assume that the payoffs of actions may change over time; in other words, we admitarbitrarily non-uniform environments. Success evaluation: The maximal payoff of all possible actions in a given round n (with a given state of environment) is denoted as maxn  Based on this notion, we define the natural lossfunction of a given player P in round n as follows: lossn(P) := |maxnpn(P)|. In other words, our ‘natural’ loss-functions are linear. However, our major theorem below does not only admit linear loss-functions, but all loss-functions loss(maxn,pn(P)) that are convex in the sense that for all max [0,1], weights  [0,1] and two possible payoffs p, q [0,max] the following holds: loss(max, (p + (1)q))  loss(max,p) + (1)loss(max,q). In words, the loss of a weighted average of the payoffs of two actions is smaller-equal than the weighted average of the losses of the payoffs of the two actions.

Based on the loss-function, we define   

the score of P’s action in round n: scoren(P) := 1  loss(Pn), the absolute success of P’s actions until round n: absn(P): = the sum of P’s scores until time n , and finally the (relative) success rate of P’s actions until time n, sucn(P):= absn(P)/n.

Because our loss-function may be (improperly or properly) convex, our scoring rules cover not only linear but all scoring rules with decreasing slope; this class covers in particular the so-called ‘proper’ scoring rules, which are important in Bayesian epistemology and decision theory. The players of our action games include one or several meta-inductivists (MIs), and arbitrary other (so-called non-MI) players, who are abbreviated as P1,,Pm, and include, for example, scientists or other object-level in-

Meta-Induction and the Problem of Fundamental Disagreement

349

ductivists, but also alternative players, such as prophets, religious leaders, etc. We admit possible worlds (so-called ‘paranormal’ worlds) in which alternative players are more successful than scientists. Each type of player corresponds to a method of action (which is part of a cognitive system). Of course, the actions performed by players may change. In particular, players may learn inductively from the observed past success records which action will be the presumably best (though, as noted above, the best action may change due to a changing environment). Moreover, the actions of all players and their payoffs are assumed to be accessible to all players of the game, and in particular, to all meta-inductivist(s). In other words: social learning is assumed to be (1) global (each player has access to all other players) and (2) cost-free. Both assumptions are, of course, idealizations (for a relaxation of assumption (1) cf. Schurz 2012). Prediction games, as introduced in Schurz (2008a, 2009c) are pairs (,(e)) consisting of a finite set  of players (forecasters) and an infinite sequence of events (e) that are coded by real numbers between zero and one; i.e. (e) := e1, e2, with ei [0,1]. The actions of the players of a prediction game consist simply in predictions of real numbers predn  [0,1], and the loss function is defined as a convex function of the linear or ‘natural’ loss |enpredn|. Action games are related to prediction games in two ways. Prediction games as a special kind of action games: Action games subsume prediction games, by assuming that “maxn” takes over the role of the events en to be predicted, and the actions consist of the predictions predn(P)  [0,1]. We don’t need a payoff-function in prediction games: the linear loss-function is directly determined as the absolute difference between en and predn. (We could formally define the payoff pn(P) as en|enpredn|, to obtain the structural isomorphism that pn  en and lossn = |enpn|.) Action games as special kind of prediction games: By letting “maxn” take over the role of the events en we can, however, also consider every action game as a special kind of prediction game, at least from a formal point of view. In this case we have to assume that the “predictions” of a proper action game are not generated by letting each player P assert a real number predn(P)  [0,1], but by letting each player P perform a real action an(P), and by identifying the obtained payoff pn(P) with P’s prediction of the event en = maxn. (As a side effect, the predictions generated in this way

350

Gerhard Schurz

will always be smaller-equal as maxn.) This consideration is not philosophically helpful, but formally very useful, insofar it allows us to transfer all mathematical theorems proved for prediction games to action games. The fact that both prediction games are a special kind of action games, and action games are a special kind of prediction games, is not a contradiction, because the relation of specialization is different in the two cases. We now define the actions of the so-called weighted meta-inductivist, abbreviated as wMI. Without restricting the assumptions we assume that all the non-MI-players players Pi play differently; so each player corresponds to a method of action. The weighted meta-inductivist wMI performs the following “weighted combination” of the so-far best methods (resp. players): nan+1(wMI) = P

atn(P)  an  1(P)



P

atn(P)

,

where atn(P) is player P’s so-called attractivity until round n, and is defined as the surplus success of P over wMI, i.e. atn(P) := max(sucn(P) – sucn(wMI), 0) (provided n > 1, else := 0.5). Thus, wMI’s actions are a weighted combination of the actions of all attractive players, with their so-far attractivities as weights. Since the attractivities refer to past success records, their time index (n) is one unit delayed compared to the time index of the performed action (n+1). There are three different ways in which actions of different players can be combined into a mixed action: (1.)

If the actions are numerically graded  e.g., driving with graded velocity, or predicting a real-valued magnitude  then their weighted combination is simply their weighted average in the literal mathematical sense.

If the actions are discrete, their combination is a probabilistic mixture, and the attractivity of a player, atn(P), corresponds either (2.)

to the probability with which a single meta-inductivist wMI imitates this player in the next round n+1, or

Meta-Induction and the Problem of Fundamental Disagreement

(3.)

351

to the relative frequency of members of a population of metainductive players that imitate this player in round n+1.

We now turn to our main theorem: Optimality theorem for weighted meta-induction wMI over action games: For every action game ({wMI,P1,,Pm},(n),p), sucn(wMI) approximates the maximal success rate of the non-MI-players in the long run, and wMI’s short run loss is smaller than m/n , where m = number of competing methods and n = number of rounds. The proof of this theorem follows from the proof of theorems 4 and 5 in Schurz (2008a; see also 2009c) for prediction games, based on the observation above that every action game can be considered as a special kind of prediction game. The worst-case bound m/n for the short-run loss of the optimality theorem can be improved to ln(m)/2n if one uses exponentiallyweighted-average meta-induction with a known prediction horizon (cf. Bianchi and Lugosi 2006, 16f). Depending on which method of combining the attractive actions (among(1.)-(3.) above) is applied, the interpretation of this theorem is slightly different. For combination method (1.), the theorem holds for every single wMI-player, without any further assumptions. For combination method (2.), the theorem holds for every single wMI-player, but only under the additional assumption that the environment or payoff-function does not react to wMI’s player-choice in the given round n; i.e., the payoff pn(P) is independent from whether wMI chooses P’s action or not. Finally, for combination method (3.) the theorem holds without any further assumption, but now it applies to the entire population of wMI-players, who approximate with their (relative) imitation-frequencies the attractivities of the attractive non-MI-players.2 In this case theoptimality theorem holds or the mean success rate of the entire population of wMI-players. However, under the assumption that the wMI’s cooperate, i.e,. that they share their 2

This approximation is defined as follows. Assume Q1,,Qr are all non-MI-players with positive attractivity in a given round n; the population has k wMI-players (with k > r); and [r] denotes the integer rounding of a real number r. Then for each i with 1i(r1), [katn(Qi)] wMI’s imitate Qi’s action and the remainder wMI’s imitate Qr’s action. The additional short-run loss die to finite approximation of a real number is 1/2k, which goes to zero for k  (cf. Schurz 2008a, theorem 5).

352

Gerhard Schurz

joint success to equal amounts, the optimality result for the mean success rate implies also an optimality result for every single member of the wMIpopulation. 3. Some Applications and Limitations In this final section we briefly mention some applications and some limitations of our results. 3.1 Intelligent design theory versus evolution theory. This is a paradigm example of two world-views or cognitive systems that are ontologically and methodologically incompatible, and in part even incommensurable. Yet the two are comparable by the neutral method of meta-induction. Evolution theory is so far predictively much more successful than intelligent design theory, whence (by the above optimality theorem) one should trust evolution theory much more than intelligent design. This conclusion is not only reasonable for the evolution theorist but also for the intelligent design theorist, provided that (s)he accepts the evaluation criterion of predictive success. Of course, a religious person could object that God requires from his people “thou shalt not try me”, which means in the effect that God forbids to test Him by the method of meta-induction. My optimality result suggests, however, that a (divine) principle that forbids the method of meta-induction, is unreasonable within the viewpoint of every cognitive system  even within a religious system. 3.2 Reliability of testimonial evidence and meta-induction. The view that testimonial evidence makes up an irreducible source of knowledge  not reducible to ordinary observation and induction  has recently been defended by several epistemologists (e.g. Coady 1992 or Insole 2000). This view seems to be hardly tenable in general. It may be justified in the domain of perceptual testimonies3 in which the error rates of humans are normally small (due to genetically determined skills). But the irreduciblesource-of-knowledge view is unjustified in domains that are in the focus of 3

In the recent debate the notion of “testimony” is understood in the broad sense, including not only perception reports but communicated information of any sort (cf. Lackey and Sosa 2006). In my view this notion of “testimony” is too broad to be useful.

Meta-Induction and the Problem of Fundamental Disagreement

353

this paper  e.g., weather forecasting or healing methods  in which error rates of untrained humans are typically large, their judgments typically biased by wishful thinking, and where critical tests based on success records are essential for making progress. In these domains one needs inductive reasons for the reliability of a purported informant (cf. Lackey 2006). In other words, one needs the evaluation of the reliability of testimonial evidence by meta-induction over available success-records. 3.3 Some limitations of meta-induction. As already explained above, if the competing cultures or world-views don’t possess a common measure of success that is intersubjectively observable or recognizable, the method of meta-induction cannot be applied. This is, so to speak, a minimal assumption for meta-induction to take hold. It may be that success records are intersubjectively observable in principle, but they are not globally but only locally accessible within so-called “epistemic neighborhoods”. In Schurz (2012) it is shown that metainduction works even in local epistemic neighborhoods, provided that these neighborhoods are sufficiently overlapping. A more severe restriction is given when success records are not only locally unavailable, but also temporally delayed. i.e., the payoff result arrives long after the respective action has been performed. Under this condition, meta-induction becomes less and less efficient. Last but not least, meta-induction applies only to some but not to all problems of theory-confirmation in philosophy of science. In particular, meta-inductive optimality justifies only the preference for theories that are empirically most adequate. It does not justify the abductive inference from the empirical adequacy of a theory to its to realistic truth. Moreover, metainduction over predictive success rates cannot discriminate betweenempirically equivalent theories. Preference for theories whose theoretical principles are most simply has to be justified by other means (here I favor Kelly’s 2007 method of minimizing the worst case number of retractions). Acknowledgement: This paper was supported by DFG-financed project “The Role of Meta-Induction in Human Reasoning”, as part of the Priority Program SPP 1516: “New Frameworks of Rationality”. For valuable discussions I am indebted to Kevin Kelly, Alvin Goldman, Christoph Jaeger and Oliver Scholz.

354

Gerhard Schurz

References Boghossian, P., 2006: Fear of Knowledge, New York: Oxford University Press Cesa-Bianchi, N., and G. Lugosi, 2006: Prediction, Learning, and Games, Cambridge: Cambridge University Press Coady, C.A., 1992: Testimony: a Philosophical Study, Oxford: Oxford University Press Conee, E., 2010: Rational Disagreement Defended, in: Disagreement, ed. R. Feldman and T. Warfield, Oxford: Oxford University Press 2010, 69-90 Feldman, R., 2006: Epistemological puzzles about disagreement, in: Epistemology Futures, ed. S. Hetherington, Oxford: Oxford University Press Goldman, A., 2010: Epistemic Relativism and Reasonable Disagreement, in: Disagreement, ed. R. Feldman and T. Warfield, Oxford: Oxford University Press, 187-215 Kelly, K.T., 1996: The Logic of Reliable Inquiry, New York: Oxford University Press Kelly, K. T., 2007: A New Solution to the Puzzle of Simplicity, Philosophy of Science 74, 561-573 Kuhn, T., 1962: The Structure of Scientific Revolutions, Chicago: University of Chicago Press (3rd edition 1996) Lackey, J., 2006: It Takes Two to Tango: Beyond Reductionism and Non-Reductionism, in: The Epistemology of Testimony, ed. Lackey and Sosa, Oxford: Clarendon Press, 160-189 Lackey, J., and E. Sosa, (eds.), 2006: The Epistemology of Testimony, Oxford: Clarendon Press Schurz, G., 2008a: The Meta-Inductivist’s Winning Strategy in the Prediction Game: A New Approach to Hume’s Problem, Philosophy of Science 75, 278-305 Schurz, G., 2009a: Meta-Induction and Social Epistemology: Computer Simulations of Prediction Games, Episteme 6, 200-220 Schurz, G., 2009b: Meliorative Reliabilist Epistemology: Where Externalism and Internalism Meet, Grazer Philosophische Studien 79, 41-62 Schurz, G., 2009c: Meta-Induction. A Game-Theoretical Approach to the Problem of Induction, in: Logic, Methodology and Philosophy of Science. Proceedings from the 13th International Congress, ed. C. Glymour, W. Wang, and D. Westerståhl, London: College Publications, 241-266 Schurz, G., 2012: Meta-Induction in Epistemic Networks and Social Spread of Knowledge, Episteme 9, 151-170

Rival Logics, Disagreement and Reflective Equilibrium GEORG BRUN, ETH Zurich and University of Zurich

The method of reflective equilibrium (henceforth “RE”) has been conceived as pluralistic right from the start. In Goodman’s original account, we read: “When we speak of the rules of inference we mean the valid rules – or better, some valid rules, since there may be alternative sets of equally valid rules” (Goodman 1983, 63). Later on, the method of RE was invoked in order to analyse the controversy between classical and intuitionistic logic (Prawitz 1977; Daniels 1996; Haack 1982) and in this context it seems promising to appeal to the pluralistic nature of the method of RE in order to account for the factual disagreement among logicians. This paper investigates the claim that rival logics can simultaneously be justified by the method of RE. Specifically, I analyse a dispute between Shapiro and Resnik. Against Resnik’s (1985; 1996; 1997, ch. 8.3) extensive discussion of RE as the methodology for logic, Shapiro (2000) argues that the method of RE can only account for a pluralist position if we accept that there is a “core” of logical notions outside the scope of the method of RE and knowable a priori. After briefly discussing some aspects of the method of RE (sect. 1), I will analyse that dispute and suggest a defence of the possibility of reasonable disagreement between proponents of rival logics (sect. 2-4). This paper provides neither a detailed analysis and defence of the method of RE nor a comprehensive discussion of reasonable disagreement about logic; rather, it explores in what sense and to what extent the method of RE can underwrite pluralism in logics. 1. Reflective Equilibrium At the core of the method of (so-called “wide”) RE are two ideas (Goodman 1983, 63-67). Epistemic justification is a matter of whether judgements, systematic principles and background theories are in equilibrium, and this state is reached through a process of mutual adjustment of judgements, principles and in some cases also background theories. To avoid

356

Georg Brun

process-result ambiguities, I will distinguish between “(reflective) equilibrium” (referring to a state), “RE-process” and “method of RE” (referring to the whole account of justification). Various accounts of the method of RE spell out the two core ideas differently and include additional elements. So far, the most elaborated general account is Elgin’s (1996); Resnik (1985; 1996; 1997; 2004) has presented the most thorough discussion of the application to logic. For present purposes, we need to briefly address three points which are crucial but often overlooked in the literature. Firstly, in the case of logic, an equilibrium is sought between judgements (or more generally, “commitments”) about logical properties (e.g. validity, logical truth or consistency) of (sets of) sentences in some given language and a logical system (“a logic”) which includes: a logical formalism, which is a formal language with a semantic or proof-theoretic definition of validity and further logical notions; an informal interpretation of the formalism, which relates, e.g., “” with “follows from”; and a theory of formalization, which regulates the relation of ordinary language arguments and sentences to expressions of the formalism (cf. Haack 1982, 223; Resnik 1985, 225; 1997, 159). Without the last two elements, there is no way of deciding whether the judgements follow from the logical system (Brun 2004; 2008). Secondly, all accounts of the method of RE refer to a contrast between principles and judgements. Although it is standardly described in terms of particular vs. general, the relevant contrast is that principles are elements of a logical system (“systematic elements” for short) whereas judgements constitute extra-systematic commitments.1 In fact, however, there are two contrasts involved. At every stage in a process of developing a RE (abbreviated “RE-stage”), there is a contrast between the systematic elements at that stage and our judgements at that stage. A second contrast is between the resulting account and the judgements the RE-process started out with; the latter may in turn be the result of a previous development of a logical system, but this need not be so. To fix the distinction terminologically, judgements (and similarly commitments) will be characterized as “antece-

1

Speaking of “judgements” is not meant to imply that the relevant commitments need to be stated explicitly; they may be expressed in any behaviour of treating inferences as (in)valid.

Rival Logics, Disagreement and Reflective Equilibrium

357

dent” in the context of the second contrast; in the context of the first contrast, I will use “current judgement” or simply “judgement”. Thirdly, as Elgin made clear, justification by RE involves several criteria. Judgements, logical system and background theories must be in equilibrium. This is usually characterized in terms of coherence. For present purposes, it suffices to note that coherence requires at least that judgements, logical system and background theories are consistent, and that the judgements follow from the logical system.2 Moreover, the resulting logical system must do justice to relevant epistemic desiderata in order to ensure that it provides a systematization, not merely a list of our commitments. We expect a logic to be a system of formal and general principles that is well-organized, exact, comprehensive and as simple as possible. In particular, we aspire for a logical system which permits transparent representations of logical forms and allows for rigorous proofs of validity which should be sound complete and maybe even decidable. Finally, the resulting account (i.e. the ordered pair judgements, logical system) must respect antecedent judgements adequately. Otherwise, there is no guarantee that the RE-process does not change the subject by effecting revisions of judgements which are so drastic that the resulting system will not count as a logical system anymore. To simplify, I will often leave implicit that a RE includes background theories and that being in RE is relative to antecedent commitments and to desiderata. 2. The Plausibility of Pluralism The method of RE has been claimed to be pluralistic in the following sense (e.g. Resnik 1997, 160, 162): It is not uniquely determined what logic results if the method of RE is applied to a given set of antecedent judgements. Different RE-processes may lead to different logical systems, each in equilibrium with the respective (current) judgements. Hence rival logics – for example classical, intuitionistic and paraconsistent logics – may be 2

The requirement that the judgements follow from the logical system has to be understood in a sense that permits developing partial theories, which cover only part of a range of commitments. A system L of zero-order logic, for example, is expected to sanction only valid inferences as valid, but not to cover all valid inferences. Consequently, if a RE is reached, only those judgements of validity which follow from L will be justified.

358

Georg Brun

simultaneously justified according to the method of RE. This is plausible because there are several reasons why different states of RE may be reached from the same antecedent judgements. The process of mutual adjustments not only involves creative moves but it also permits that conflicts be dealt with in more than one way because judgements, systematic elements, background theories or desiderata may be weighed differently, or because the various conflicts are addressed in a different order. Typical examples of conflicts which allow multiple resolutions can be found in debates about conditionals. At least three strategies are available for resolving the conflict between the classical account of material conditionals and the judgement that “If I put sugar in this cup of tea it will taste fine” is true while “If I put sugar and diesel oil in this cup of tea it will taste fine” is false (Harper 1981, 6; cf. Resnik 1985, 228). One can either keep to classical logic and revise the judgement, or modify the logical system so that it no longer includes     (  )  , or formalize the “if-then” sentences in question with a different type of conditional. In other cases, there is a conflict which involves background theories. According the Prawitz’s (1977) and Daniels’s (1996) analysis, Dummett’s arguments for intuitionistic logic rely on a theory of meaning to argue against accepting classical logic, which includes such theorems as   ¬ and inference rules such as ¬¬   . A more precise formulation of pluralism is the following: (P1) Given a set of antecedent judgements A, it is possible that two RE-processes lead to rival logics L1 and L2 and to sets of judgements J1 and J2, such that (relative to A): L1 and J1 are in RE, and L2 and J2 are in RE. Two qualifications are needed. First, I limit the discussion to rival logics that are developed from the same set of antecedent judgements. Second, I will not try to give a definition of “rival”; it suffices to say that rival logical systems have the same scope yet yield different sets of judgements (which are not merely the result of an equivocal use of logical vocabulary; cf. Haack 1996, ch. I.1-2; Resnik 1996, 497-498). Examples of rival logics are classical, intuitionistic and paraconsistent logics; or the different modal logics such as S5 and T. They are non-equivalent, in contrast to, for example, mere notational variants, different axiomatizations and systems with truth-tables vs. systems with semantic tableaux. Also, more and less com-

Rival Logics, Disagreement and Reflective Equilibrium

359

prehensive logical systems (such as zero- vs. first-order logic, or logics with and without modal operators) and systems designed for application to different languages do not count as rival logics (this excludes pluralism in the vein of Carnap’s principle of tolerance; Carnap 2002, § 17). 3. The Own-Lights Principle We are now in a position to analyse the arguments about the pluralistic nature of RE which can be found in the debate between Resnik (1996, 493494, 502-504; 1997, 160-162) and Shapiro (2000, 349-351) (which primarily revolves around cognitivism and realism and not pluralism). The claim (P1) can be challenged by pointing out that a proponent of, say, classical logic will find that intuitionistic and paraconsistent logics are not in RE because they are incomplete and inconsistent. As long as an epistemic subject is committed to a particular logic, she will find that at least one of two rival accounts is not in RE and hence that (P1) is false. Against this charge, one can argue that the notion of RE has logical components and therefore the criteria for being in RE are not independent of the account they are applied to. As Resnik points out, RE requires that a logical system be consistent by its own lights and the judgements follow from the logical system in the sense of “follow” defined in the logic under consideration. Hence, we cannot argue against rival logics that they are not in RE according to our standards. The question is rather whether they are in RE according to their own standards. For an analysis of this argument, we need to distinguish four ways of using expressions with a logical meaning, such as “follows from” and “is consistent”: (1) (2)

In current and antecedent judgements, “follows from”, for example, is used extra-systematically as a relation between sentences. In a logical formalism, there is typically a symbol such as “”, which expresses a relation between formulas. It is read “follows from” because it is – according to the informal interpretation of the formalism – intended to be a systematic counter-part to the extra-systematically used “follows from”.

As long as RE has not been reached, we cannot assume that the extrasystematic use of, say, “follows from” and the systematic use of the corresponding symbol agree (cf. the sugar-and-diesel-oil example mentioned

360

Georg Brun

above).3 Furthermore, use (1) and (2) typically change during a RE-process as a result of revisions which affect the judgements or the logical system. (3)

(4)

In applying the method of RE, an epistemic subject may use logical expressions extra-systematically with reference to judgements and logical systems (not as in (1) and (2) as parts of judgements or expressions of a logical formalism). For example, “is consistent” may be used to express a property of the logical system or of a set of judgements; “follows from” may be used to express a relation between the logical system and a judgement. Logical expressions are also used extra-systematically when the method of RE or one of its particular applications is discussed in philosophy of logic or epistemology, as in this paper.

With respect to use (3), Resnik argues that an epistemic subject must use at every RE-stage the logical notions developed up to that point. In determining whether a set of judgements and a logical system are in RE, ”the logic contained in one’s own evolving logical theory” determines whether the required coherence has been achieved (Resnik 1997, 160). Thus, Resnik’s remark that an account needs to be coherent by its own lights amounts to the claim that type-(3) uses of “is consistent” and “follows from” must agree with their corresponding type-(2) uses. Consequently, these type-(3) uses will change together with the respective type-(2) uses in the course of a RE-process. Resnik’s position leads to two difficulties. First, the relation of agreement between the systematic use (2) and the extra-systematic uses (1) and (3) is not at all straightforward. We cannot use “follows from” and “” in exactly the same way because the two expressions do not belong to the same language; “” is meaningless when flanked by expressions that are not formulas in the language of the formalism in question. It would thus be more straightforward if use (3) was tied to use (1) instead of use (2). Secondly, Resnik claims that type-(3) use of, say, “follows from” should agree with use (2) of a corresponding systematic expression. However, this seems odd. Unless a RE has been reached, use (2) can differ from use (1), 3

To be fully explicit, we would need to index occurrences of “follows from” and “” with the RE-stage they belong to, and occurrences of “” additionally with the logical system they are part of.

Rival Logics, Disagreement and Reflective Equilibrium

361

but use (2) is just the systematic use, whereas use (1) is the use the epistemic subject is actually committed to. This commitment gives us reason to insist that the extra-systematic uses (3) and (1) agree. This move also dovetails with the reason why one should accept an “own-lights principle” in the first place. If an epistemic subject sincerely wants to appeal to considerations of consistency or valid inference in arguments about the justification of a logical system she is considering adopting, she must rely not on just any notions of consistency and validity, but on those she is actually committed to. The general picture driving this line of argument is the following: When an epistemic subject develops a logical system, she simultaneously makes her logic explicit and adapts it according to her epistemic desiderata; she reconstructs the logic she already has and during this process she uses at every stage the logical commitments she has at that stage. These considerations motivate the following own-lights principle, which ties use (3) to use (1): (OL) At every RE-stage, the logical notions used in criteria for being in RE must agree with the logical notions used extra-systematically in the current account. This principle gives us reason to rewrite thesis (P1) about pluralism in a way that incorporates the fact that the notion of being in RE can vary with the account it is applied to: (P2) Given a set of antecedent judgements A, it is possible that two RE-processes lead to rival logics L1 and L2 and to sets of judgements J1 and J2, such that (relative to A): L1 and J1 are in RE1 (but not in RE2), and L2 and J2 are in RE2 (but not in RE1). According to (OL), “in REi” in (P2) must be understood in agreement with the logical notions as used in Ji. (P2) now evades the objection to (P1). If an epistemic subject S1 argues against a proponent S2 of a rival logic that S2’s logic is not in RE by S1’s standards, this is beside the point. S2’s logic must be in RE by standards according with S2’s account. As an aside, it may be worth noting that the element of self-application captured in (OL) is peculiar to the application of the method of RE to logical systems. The situation is different if we deal with, for example, moral theories. Whether a theory, moral or otherwise, is in equilibrium with the

362

Georg Brun

relevant judgements is not a moral issue. When an epistemic subject assesses whether her moral theory is in equilibrium with her moral judgements, she must use logical notions – ones she is actually committed to. But this assessment will not use her moral commitments. It will rather apply to them. In what follows, I want to defend the claim that (P2) opens up the possibility of reasonable disagreement: (RD) Given two rival logics as described in (P2), then two epistemic subjects S1 and S2 who adopt the accounts J1, L1 and J2, L2 respectively are in reasonable disagreement because both accounts are in RE. The two subjects S1 and S2 are in disagreement because they adopt rival logics and hence their respective sets of judgements J1 and J2 are different. The disagreement is reasonable because the method of RE gives both subjects a justification for their commitments (cf. Goldman 2010, 189-190 on the notion of reasonable disagreement). 4. Transcendent and Immanent Logical Notions Against (RD), one may argue that the expression “in RE” is problematic since it is not tied to a particular account. Resnik and Shapiro insist that such a use is incompatible with the immanence of logical notions, which can be used only in the context of a particular account, in contrast to transcendent notions, which can be used in the context of various (or even all) accounts. They consequently hold that also the notion reflective equilibrium is immanent to an account because it is partly constituted by logical notions such as is consistent and follows from. On the background of the discussion in the preceding section, this objection may be elaborated as follows: (RD) quite obviously raises the question of whether using the expression “in RE” without an index is legitimate. Do the arguments for replacing (P1) by (P2) not apply equally well to (RD)? Maybe the plausibility of (RD) merely rests on using “in RE” ambiguously. However, (OL) does not straightforwardly apply to the use of “in RE” in (RD) since (RD) as well as (OL), (P2) and the comments made in section 3 are not put forward from the point of view of an epistemic subject who is actually developing a logical system. Rather, we are engaged in investigat-

Rival Logics, Disagreement and Reflective Equilibrium

363

ing this epistemic practice, taking the point of view of epistemology. So the incriminated uses of “in RE” involve type-(4) uses of logical notions, which, in contrast to type-(3) uses, are not covered by (OL) (similar considerations apply to other expressions such as “agree with” in (OL) and “it is possible” in (P2); cf. Shapiro 2000, 349-350). Nonetheless, the arguments that spoke in favour of the own lights principle could still be put forward against using “in RE” in (RD) as applicable to two rival accounts of logic. After all, an epistemologist or philosopher of logic will have logical commitments just like any other epistemic subject (and there is also no reason why the same person could not take both the perspective of the epistemic subject developing a logical system and that of the epistemologist reflecting on the possibility of reasonable disagreement). If we accept these points, (RD) has to be given up and we are back with (P2). But (P2) is not a claim of reasonable disagreement because it merely asserts that an account may have some property and lack another, whereas it is the other way around for its rival. For Resnik, considerations in this vein are a reason to be sceptical about the possibility of reasonable disagreement. (RD) and other claims invoking transcendent logical notions are in danger of making no sense. Shapiro turns the argument against Resnik by pointing out that transcendent logical notions are needed if we want to acknowledge for the intuitively plausible possibility of reasonable disagreement. Moreover, he argues, without transcendent logical notions, speaking of the method of RE is meaningless.4 To thwart the argument from immanence against (RD), one could attack the distinction between immanent and transcendent notions, dispute the need for transcendent notions, or argue in favour of transcendent logical notions. For present purposes, I shall adopt the last strategy. The basic idea is that distinguishing two contrasts between judgements and systematic elements (sect. 1) allows us to defend the possibility of reasonable disagreement while acknowledging what is convincing about Resnik’s and Shapiro’s arguments. To begin with, we distinguish, within the extra4

In response to Shapiro’s critique, Resnik explicitly withdraw an important aspect of his position and switched to an interpretation of the notion of RE according to which RE is a state of “triples consisting of LOGICIANS [i.e. practitioners of the discipline of logic in contrast to practitioners of inference], logical theories, and considered judgements of so-called facts of logic” (Resnik 2004, 192-193). Analysing this move is outside the scope of this paper.

364

Georg Brun

systematic use, between antecedent and current logical notions. Current notions result from a RE-process and are immanent to the current account. But some antecedent notions are transcendent, namely the pre-systematic notions. These are not the product of some previous development of a logical system, but the notions we use “before” we start doing logic. Paradigmatically, they are informally used as part of ordinary language, without any basis in an explicit account in the form of, for example, necessary and sufficient conditions. More often, pre-systematic logical notions remain implicit when specific arguments and (sets of) sentences are treated as valid, consistent and so on. Even if expressed explicitly, antecedent commitments involving pre-systematic logical notions need not include terms such as “is logically valid” expressing some formal, “thin” logical notion. Frequently, they will employ “thick” notions such as agrees with or is coherent which have a logical aspect but are not reducible to a formal logical notion.5 However, there is still room for explicitly explaining presystematic logical notions and for regulating the use of terms expressing them (as, for example, in the discussion of validity in Beall / Restall 2006, ch. 2)6. In contrast to current extra-systematic notions, pre-systematic logical notions are “open”; that is, they are not precisely defined and cannot be classified as, say, classical or intuitionistic because such distinctions are simply not available without recourse to more or less developed logical systems. Hence, pre-systematic logical notions are not tied to a particular logical system. Rather, they are the common starting point for developing logical systems. In the course of developing a logic, more precise extrasystematic notions are developed, which then are tied to the respective system and hence no longer transcendent. Pre-systematic notions are therefore not only transcendent because they can be used with reference to different logical accounts, but they can also be made precise in different ways in different accounts. All in all, the relation between transcendent, presystematic logical notions and immanent, current extra-systematic logical 5

Thanks to Catherine Elgin for drawing my attention to these points. In the approach of Beall and Restall, current logical notions of validity are not only explications of the pre-systematic notion of validity and hence less vague, but the presystematic notion of validity is also a generic notion of which the current notions are more specific instances (see Beall / Restall 2006, 29). According to my explanation of “pre-systematic” this need not be the case, but it is not excluded either. 6

Rival Logics, Disagreement and Reflective Equilibrium

365

notions is a relation of explication in Carnap’s sense (Carnap 1962, §§ 23). Similarly, the informal characterization of the method of RE can be explicated in various ways in the context of different logical systems. Transcendent notions are the basis for giving an account of the method of RE. Framed in pre-systematic terms, such an account is not tied to a particular logic. Precise but immanent notions of being in RE (and being consistent, following from etc.) become available to an epistemic subject if she develops a logic, and they will be subject to the arguments of Resnik and Shapiro. Such immanent uses of logical expressions are covered by (OL), in contrast to the general account of the method of RE which is given in type-(4) uses of logical terms. Furthermore, an account of the method of RE is also a basis for defending the possibility of reasonable disagreement. The arguments about immanence can be avoided if we interpret “both accounts are in RE” in (RD) as follows: each of the two accounts falls under some specific (current extra-systematic, not presystematic) notion of being in RE. The method of RE settles what counts as a specific RE-notion by specifying how such notions are developed from pre-systematic logical notions. Once a RE has been reached for an account of logic, there are two ways of discussing the justification of rival logics. On the background of the commitments an epistemic subject has in the context of a particular account of logic, rival logics are unjustified, because they are not in RE in terms of these commitments. But in terms of pre-systematic notions, there can be reasonable disagreement because rival logics can be justified as the result of applying the method of RE to the same antecedent judgements. Proponents of rival logics can take this point of view as well and reasonably discuss their differences on the basis of shared pre-systematic notions. Let us now look briefly at two objections. Firstly, when Quine introduced the relevant distinction between immanent and transcendent notions, he wanted to cast doubt on whether there are any transcendent logical notions rather than just “a loosely related family of mutually more or less analogous immanent notions” (Quine 1986, 60; cf. 87) which create the mere appearance of a transcendent notion where in fact there is only “some felt family resemblance whereof no capital need be made.” (Quine 1986, 19; cf. 20, 21). It is not clear to me whether Quine’s attack is directed at pre-systematic or current extra-systematic logical notions (or both). If the latter, there simply results no challenge to the view I have proposed here, which takes current extra-systematic logical notions to be immanent any-

366

Georg Brun

way. Furthermore, the claim that there is at best a family resemblance between rival (current extra-systematic) logical notions is clearly compatible with the view that they are different explicata for the same pre-systematic explicandum. If the former, the worry is that the pre-systematic logical notions exhibit a family-resemblance structure. This claim is not only compatible with the view that they are explicanda which admit for various explications, it provides in fact a neat explanation for it. Secondly, there is the worry that pre-systematic notions are more transcendent the more vague they are. However, even if such a correlation should prove unavoidable, it would not necessarily trivialize the method of RE, which can and should be made more precise. This will lead to disagreement about the exact nature of the method and pre-systematic logical notions will play the important role of providing the background against which such disagreement can be understood as reasonable and be discussed. 5. Concluding Remarks In sum, the arguments of Resnik and Shapiro show that in terms of a particular account of logic in RE, we can argue that rival logics are unjustified. Furthermore, differences between rival accounts of logic may also lead to differences with respect to the exact conditions which determine whether an account of logic or some other field of inquiry is in RE. On the other hand, we can defend the claim that the method of RE permits reasonable disagreement about rival logics if we rely on pre-systematic logical notions. These serve as a common starting point for developing various logical systems and they provide the resources for a meaningful characterization of the method of RE. To put things into perspective, I would like to mention three points. Firstly, the arguments discussed are specific to the application of the method of RE to logic and do not bear in the same way on its application in other contexts. In moral philosophy, for example, pluralism and reasonable disagreement do not raise the issues discussed here because the method of RE can be explained without (explicitly or implicitly) appealing to moral notions. On the other hand, any reasonably ambitious justification of a moral theory by RE will need to rely on more or less precise logical notions (for example, when evaluating whether the judgements follow from the principles) and hence draw on some particular account of logic. It may

Rival Logics, Disagreement and Reflective Equilibrium

367

well be that the choice of a logical system makes a difference with respect to the question of whether a given set of moral commitments counts as being in RE with some given moral theory. Secondly, the possibility of reasonable disagreement raises further issues which cannot be dealt with here. For example, should the fact that epistemic peers endorse rival logics lower the epistemic standing of one’s logical system? Should an epistemic subject accept that rival logics are promoted? (Cf. Resnik 1997, 162.) What is the relation between reasonable disagreement and pluralism in the sense of Beall and Restall? And in what sense can or should we accept more than one logic? (Cf. Resnik 1996; Beall / Restall 2006) Thirdly, defending presystematic logical notions as transcendent does not amount to claiming that the pre-systematic background is immune to criticism. On the contrary, the method of RE calls for critically reworking pre-systematically used logic.7 References Beall, J.C., and G. Restall, 2006: Logical Pluralism, Oxford: Clarendon Press Brun, G., 2004: Die richtige Formel. Philosophische Probleme der logischen Formalisierung, 2nd ed., Frankfurt a. M.: Ontos Brun, G., 2008: Formalization and the Objects of Logic, Erkenntnis 69, 1-30 Carnap, R., 1962: Logical Foundations of Probability, 2nd ed., Chicago: University of Chicago Press Carnap, R., 2002: The Logical Syntax of Language, Chicago: Open Court Daniels, N., 1996: On Some Methods of Ethics and Linguistics, in: Justice and Justification. Reflective Equilibrium in Theory and Practice, Cambridge: Cambridge University Press, 66-80 Elgin, C.Z., 1996: Considered Judgment, Princeton: Princeton University Press Goodman, N., 1983: Fact, Fiction, and Forecast, 4th ed., Cambridge, Mass.: Harvard University Press Goldman, A.I., 2010: Epistemic Relativism and Reasonable Disagreement, in: Disagreement, ed. R. Feldman and T.A. Warfield, Oxford: Oxford University Press, 187-215 Haack, S., 1982: Dummett’s Justification of Deduction, Mind 91, 216-239

7

I am grateful to Christoph Baumberger, Dale Jacquette and Timm Lampert for helpful comments.

368

Georg Brun

Haack, S., 1996: Deviant Logic, Fuzzy Logic. Beyond the Formalism, Chicago: University of Chicago Press Harper, W.L., 1981: A Sketch of Some Recent Developments in the Theory of Conditionals, in: Ifs. Conditionals, Belief, Decision, Chance, and Time, ed. W.L. Harper, R. Stalnaker, and G. Pearce, Dordrecht: Kluwer, 3-38 Prawitz, D., 1977: Meaning and Proofs. On the Conflict Between Classical and Intuitionistic Logic, Theoria 43, 2-40 Quine, W.V.O., 1986: Philosophy of Logic, 2nd ed., Cambridge, Mass.: Harvard University Press Resnik, M.D., 1985: Logic: Normative or Descriptive? The Ethics of Belief or a Branch of Psychology? Philosophy of Science 52, 221-238 Resnik, M.D., 1996: Ought There to be but One Logic? In: Logic and Reality. Essays on the Legacy of Arthur Prior, ed. B.J. Copeland, Oxford: Clarendon Press, 489-517 Resnik, M.D., 1997: Mathematics as a Science of Patterns, Oxford: Clarendon Press Resnik, M.D., 2004: Revising Logic, in: The Law of Non-Contradiction. New Philosophical Essays, ed. G. Priest, J.C. Beall and B. Armour-Garb, Oxford: Clarendon Press, 178-194 Shapiro, S., 2000: The Status of Logic, in: New Essays on the A Priori, ed. P. Boghossian and Ch. Peacocke, Oxford: Clarendon Press, 333-366

VI. Wittgenstein

The Builders1 MEREDITH WILLIAMS, Johns Hopkins University

There are many ways to discuss Wittgenstein’s iconic language-game the Builders. Yet all doing so would agree that it is a springboard into his discussion of language if only negatively. Rush Rhees, in his paper on “Wittgenstein’s Builders”, argues that the builders’ game Wittgenstein describes in PI §2 is not a language. He cites a number of reasons for holding this view. The central problem with this game is “to imagine that they spoke the language only to give these special orders on this job and otherwise never spoke at all. I do not think it would be speaking a language”2. In elaborating on this point, Rhees identifies three further shortcomings. These are: the calls of A are only signals “which cannot be used in any other way”3; the builders game does not show how speaking is “related to the lives which people live”4; and lastly for A and B “there would be no distinction between ‘that is not what we generally do’ and ‘that makes no sense.’”5 More recently, Robert Brandom has urged that the Builders’ Sprachspiel “is a vocal, but not yet verbal practice.”6 Language is essentially discursive, that is, language has an assertoric structure that supports and enables its fundamental task, that of asking for and giving reasons, to be realized. Where Rhees emphasizes the complexity of human life as a ground for repudiating the relevance of the Builders, Brandom emphasizes the episte1

I thank David Stern and Danièle Moyal-Sharrock for their helpful comments on an earlier version of this paper. 2 Rush Rhees, “Wittgenstein’s Builders,” Proceedings of the Aristotelian Society 1959-60; reprinted in Discussions of Wittgenstein, New York: Schocken Books 1970, p. 76. 3 Ibid., 77. 4 Ibid. 5 Ibid. 6 Robert Brandom, Articulating Reasons, Cambridge, Mass.: Harvard University Press 2000, 14. Warren Goldfarb, “I Want You to Bring Me a Slab,” Synthese 82, (1983) also rejects the idea that the Builders could be a language on the grounds that it fails to have the requisite syntactic and logical structure.

372

Meredith Williams

mological and logical complexity of discursive practice. Relative to the expansiveness and systematicity of these conceptions of linguistic practice, the builders’ use of language is rigidifying. Indeed as Stephen Mulhall puts it in his excellent discussion of the Builders game, it “stultify[ies] the human imagination and depriv[es] itself of a future” (p. 57).7 Anyone familiar with the builders recognizes the relevance of these objections. Valuable though they are in allowing us to reflect on what it is to live in the medium of language, these objections do not focus on the primary use of the Builders game. It plays a methodological role in Wittgenstein’s diagnostic critique of the Augustinian picture of language, which is a crude expression of the representational picture of language: words name objects and are learned through ostensive definition. The Builders game initiates a diagnostic journey from the crudest form of the word as call to our sophisticated philosophical account of the explanatory work that name and reference are to provide. For this diagnostic journey to succeed, we must see in the Builders game a similarity to language. This similarity seems to me to be genuine. I am going to follow Stephen Mulhall’s suggestion that we treat the first 40 or so passages as the story of the development of natural language, “each episode” of which “adds to the original tale an element of the structure or content of natural languages whose absence had raised doubts about the linguistic nature of the builders’ calls.”8 The epi7

Stephen Mulhall, Inheritance & Originality, Oxford: Clarendon Press 2001, 57. Mulhall treats PI §2 and the passages following (§§2-46) as a kind of abbreviated natural history of the evolution of language users. Anyone interested in Wittgenstein’s builders should consult this interpretation. See 52-87. 8 Ibid., 70. Mulhall’s full view of the Builders is more complex than this suggests. Mulhall does hold that the Builders tale consists of a story of apparent development. But the episodes of that story can be viewed as either continuous, and so a development of natural language, or discontinuous, each episode offering the opportunity for an examining a feature of language. My methodological use of the language-game emphasizes the diagnostic contribution of the “episodes”. There is considerable overlap in the identification of the stages in the evolution of the Builders game. But their relation to the philosophical issues differs. For Mulhall, the complex character of the language-game is revealed through the differing perspectives we can take with respect to the Builders. My primary concern is to reveal the theoretically motivated changes in the relations between word-and-object and features of systematicity. Also see Stanley Cavell, “Declining Decline: Wittgenstein as a Philosopher of Culture,” in S. Mulhall, ed.: The Cavell Reader, Oxford: Blackwell 1996. I would like to thank David Stern for pointing out this complexity in Mulhall’s discussion of the Builders.

The Builders

373

sodes as I identify them are in the service of a critical diagnosis of the philosophical conception of name and its meaning. The stages follow the evolution of language from the simple calls of the Builders to the philosophical idea of names as referring only to simple objects. The journey as I understand it is a philosophical one, not an anthropological one. Thus, in the final critical examination of the idea that the demonstrative is the only “real name”, I will discuss both B. Russell’s defense of this idea and T. Burge’s contemporary defense in his important book Origin of Objectivity.9 1. Introduction: PI §§2 - §39 In the set of passages to be considered here, Wittgenstein moves from saying “[l]et us imagine a language for which the description given by Augustine is right…Conceive this as a complete primitive language” (PI §2) to judging that “[n]aming appears as a queer connection of a word with an object….a name ought really to signify a simple” (PI §§38-39). What is the connection, then, between a primitive language that associates calls with building blocks and a sophisticated philosophical theory in which the demonstrative “this” is deemed the only real name within a language because it alone can name a simple object. This notion of a real name requires increased complexity of the systematic features of language. Thus, issues of systematicity are arguably in play at the outset of the Investigations.10 For any language-game, some form of systematicity is required. In the simplest games, patterns of behavior relating the players to each other and perhaps to some object(s) in the world. Such simple (in the sense of being primitive) games do not contain real names. In sophisticated games, names cannot exist without syntactic and logical structure. These two forms of systematicity are typically associated with a third kind of systematicity, that of semantic compositionality. The device that brings these three together is what Wittgenstein calls “Frege’s Idea”, the idea that assertoric form is fundamental in any language. There are four intervening stages that take us from animal-like calls (PI §2) to real names (PI §39). These are (i) the expanded Builders game con9

Tyler Burge, Origin of Objectivity, Oxford: Clarendon Press 2010. See Meredith Williams, Blind Obedience, London: Routledge 2010, ch. 2, for a fuller discussion of this point.

10

374

Meredith Williams

sisting of articulated phrases (PI §8): “A gives an order like “d–slab– there”. (ii) The tagging game of PI §15 concludes that “naming something is rather like attaching a name tag to a thing”. (iii) The jump to full systematicity of PI §§19-22 opens with the question “is the call ‘Slab!’ in example (2) a sentence or a word?”.” (iv) And lastly Wittgenstein introduces the proper place for ostensive definition and its object the name: “Now one can ostensively define a person’s name, the name of a colour, the name of a material, a number-word…and so on” (PI §28). The first thing to note is that this journey to PI §38 alternates between discussing names and discussing matters of systematicity: from calls (2) to articulated phrases (8), from name-tags (15) to sentences (19). Names and systematic features are seen to evolve together. This raises questions for the role played by ostensive definition (PI §28) in fixing the meaning of names. Wittgenstein sums up the problems by pointing out that “…an ostensive definition can be variously interpreted in any case” (PI §28). It is at this point that Wittgenstein raises his fundamental philosophical question: How is the problem of normative similarity to be satisfactorily addressed? In virtue of what is there a semantic relation between name and object (or property) such that judgments of sameness are secured over time. This problem both explains why it is important to see the language-like features in the Builders game and why the philosophical turn to “real names” became imperative. We begin by accepting that the Builders game counts for Wittgenstein, and so for us, as a language, “a complete primitive language”. The most important feature of the Builders game that justifies calling it a “language”, albeit a primitive one, is the normativity of the game. Normative practices, which include assertoric practice, can be distinguished both from associative and causally explicable behaviors. There are four features that mark any normative practice. These are: (1) The use of linguistic items or signs is an indispensable part of the behavior of the practitioners and the coordination of their behavior. (2) The behavior is normative in that it is appropriate or not, correct or not. In other words, we have a normative grounding in the world, not an epistemic one. (3) In being normatively constrained, the behavior is evaluable and thus subject to sanctions. (4) And lastly such normatively constrained behavior is culturally heritable; it is not

The Builders

375

just acquired, it is learned.11 There are many different kinds of norms and normative practice, but we shall be concerned primarily with the normative bedrock at which we judge for sameness, where it is obvious and certain how to go on in the same way. Wittgenstein’s diagnostic work in these passages is to show how philosophers have tried to identify the identity conditions for names, through use of a demonstrative “this” or “that”. This fundamental philosophical problem concerns how we specify the identity conditions for categorizing or going on in the same way. For Russell and Burge, in their different ways, the demonstrative “this” is the only real name. 2. Calls: PI §§2 - 12 The most notable feature of the movement from animal-like calls to articulated speech is Wittgenstein’s appeal to the learning situation. The linguistic repertoire can be learned. He describes this first for the children of (2) who learn the use of the calls for the four distinct building blocks (§6), and then for the children of the expanded language game, who acquire demonstratives, numerals, and color words (§8). They are able to say things like “4–slabs–there” or “this–red–beam”. Important philosophical work is done in both learning situations. Builders (2) focuses on the relation between word and object. Expanded Builders (8) focuses on the creation of a complexity that engenders an enhanced space of possible uses of language that are internally related to each other. This is, of course, a kind of systematicity within the language, but it is not syntactic. It derives from the techniques learned in using the words of the language. In the context of the Builders (2) children learning their four calls Wittgenstein identifies a key conflation that contributes to the philosophical misunderstanding of the relation of word to object. The children of the Builders tribe are brought up to perform certain actions and use certain 11

There are different ways to contrast learning from acquisition. One is that learning must involve some kind of hypothesis formation and testing process. A broader conception, and one more congenial to Wittgenstein, is the idea that learning extends to being trained or acquiring a technique for the exercise of a role within a game. Such training into a practice is contrasted with a purely causal account like that of conditioning theory. Ostensive teaching is used with training into a practice. That is learning though it is not a kind of hypothesis formation and confirmation model.

376

Meredith Williams

words. What actions and what words involve the use of the ostensive gesture. This ostensive gesture is not ostensive definition, which requires the ability to ask for names. The novice does not have that cognitive capability. It is rather a kind of ostensive teaching of words. In displaying the logic of ostensive teaching, Wittgenstein asks Doesn’t someone who acts on the call “Slab!” in such-and-such a way understand it? – No doubt it was the ostensive teaching that helped to bring this about; but only together with a particular kind of instruction. With different instruction the same ostensive teaching of these words would have effected a quite different understanding (PI §6).

If we think of ostensive teaching as creating an association between the call and an object (type), the point here is that such an association will effect different understandings of the word depending on the instruction or training the child receives.12 The training constitutes the practical context within which the word is associated with its object. “Ostensive teaching has [an] effect” in training of the child, the initiate learner, “to perform these actions, to use these words as they do so, and to react in this way to the words of others” (PI §6). The demonstratives in this sentence are all italicized, to make the point that what is fundamental to this language game is the acting and the saying and the reacting. For the novice, the training does not include any theoretical component or he exercise of any second-order competence, like asking for a name. Causal language is used: the teaching effects a change in acting. Wittgenstein is making a three-fold distinction between ostensive definition (or explanation), ostensive teaching and training or instruction. Ostensive definition presupposes that the inquirer has the concept of names and naming. Ostensive teaching effects an association between word and object. And training or instruction is that which initiates the child or novice into a normative social practice. To make the distinction between ostensive teaching and training clearer, let us consider the following simple lan12

In PI §6, Wittgenstein uses both “training” and “instruction” to describe the pedagogic context in which ostensive teaching is used. “Training” has connotations of animal training. That is quite deliberate on Wittgenstein’s part. He is focusing on the situation of the young child or novice first acquiring certain cognitive abilities. “Instruction” does not have the conditioning connotations that “training” does. It is more readily used in the context of schooling. But there too Wittgenstein seeks to emphasis the important differences between teacher and novice or pupil.

The Builders

377

guage-game. Tribe A has a simple ritual that must be performed each morning. Each tribe member must face the east and call out “lulu” as the sun rises. It is a new beginning, and the ritual expresses the joy and reverence each feels for the rising sun. The sanction for failing to do this is to be beaten with a stick. The children of this tribe must learn the ritual since they don’t naturally face the sun each morning. The children are ostensively taught to associate the rising sun with the word “lulu”. This could be a causally induced association. The training within which it is embedded communicates a reverence for the sun and various other features that inform this practice. Tribe B instills, through the same kind of ostensive teaching, the association of the word “lulu” and the rising sun. But the training within which this association is acquired is quite different from that of Tribe A. The rising of the sun is the mark of another day of toil in the effort to raise the sun by chanting “lulu” while harnassed to a grinding stone. The training, and so the beliefs, actions and emotions that are constitutive of that practice, are quite different in the two cases while the ostensive teaching is the same. Displaying the role of ostensive teaching plays in learning a practice shows how it differs from ostensive definition. Ostensive definitions can be asked for and given only within more sophisticated language games, more particularly the game of asking for a name. Understanding the Builders’ calls is acting in the ways that builders A and B do. One consequence of conflating these two forms of ostension is to treat the ostensive teaching of the initiate learner as requiring the conceptual resources appropriate to ostensive definition. Wittgenstein famously gives voice to this when he says: “Augustine describes the learning of human language as if the child came into a strange country and did not understand the language of the country; that is, as if it already had a language, only not this one” (PI §32). Wittgenstein uses the learning situation as the natural lens through which to watch the progress of human normativity, for it is normativity that is the fundamental feature of our language games. Initiate learning is the child’s actions being “shaped”, not by stimulus and reinforcement, but by adults using ostensive teaching to train the child in norms. Ostensive teaching exploits the child’s causal embeddedness in the world to train him into a normative practice, the one already mastered by the teacher. The call is powerfully associated with the stone; it cannot be meaningfully used independently of the stone itself. That is the training. To hear the call “Slab!” is to pick up the slab. There is no room for discussing slabs or debating the

378

Meredith Williams

relative merits of each or any other more sophisticated game. There is only logical space, as it were, for picking up a slab or for calling “Slab!”, that is, calling for a slab to be picked up. And, of course, there can be failure to pick up the slab or to deliver it to A. This is enough space for normativity – behaving correctly or incorrectly – but not enough to free the call from the presence of slabs. Words are part of the Builders’ actions; and these words are taught the Builders’ children. These are three of the four key features of normativity, identified earlier: the word is integral to the action, the linguistic action is correct or incorrect, and the linguistic practice is culturally heritable. This important methodological work of the learning situation helps illuminate Wittgenstein’s characterization of “language-game” in the passage immediately following the discussion of ostensive teaching. In PI §7, he idenfies three ways in which “language-game” is used: (1) the urlanguage game, the Builders, and by extension any other artificial or appropriately simple natural language-game, constructed to examine a philosophical problem or theory; (2) the games by which children learn their native language; and (3) the language-game is “the whole, consisting of language and the actions into which it is woven” (PI §7). The methodological use of the simplified language-game and its kinship to the young child acquiring a first language makes it natural to associate the third kind of language-game with those who have mastered language. And it is equally natural to ask how training in calls can be transformed into a mastery of language. The next stage in the narrative is how to create a more complex language-game. This brings with it a greater normative space of possible linguistic acts. It is this that bridges the gap between the Builders game and full-fledged language. At PI §8, Wittgenstein introduces an articulated language game that builds on the primitive Builders game. This expanded game is also discussed through the lens of the initiate learner. Articulated speech comes with the addition of demonstratives, numerals, and color terms. By combining these, the speaker expands the range of what can be said. This complication of the language is achieved through the kind of training received by the child. This is a case in which ostensive teaching is combined with different kinds of training to produce different ways of acting: the child learns to use numerals by counting out, he learns the color words by using a color chart as his sample for different colors; and he learns the demonstrative in connection with ostensive gestures. Different techniques govern

The Builders

379

the use of different words. The differing techniques create the complexity suggestive of full-fledged language. It is not a complexity that is achieved by introducing more words of the same kind. That would keep us in the Builders game. The complexity of the Expanded Builders game introduces words that differ in the very techniques of use. It is here that Wittgenstein introduces his analogies with tools in a toolbox and handles of a locomotive (PI §§11-12). On the surface the tools and handles look the same, but they must be used in different ways if the engine is to function properly or carpentry project is to be completed successfully. The different roles of words, and so the different import of ostension, is captured in the different training associated with the words. By looking at how the children of the extended practice learn, we see that the apparent uniformity of words, “4–slabs–there”, is belied by the very different roles they play in the linguistic act. These differences are made visible by considering the training required for each kind of sign. Ostensive teaching would be used as part of the training in mastering each kind of word. The grocers language-game has already revealed this (PI §1). But only now do we see the significance of the very different skills that are involved. This is the bedrock for the modest degree of articulation, and thereby systematicity, achieved within the Extended Builders game. That systematicity just is the range of possible uses to which the signs of the language can be put. The gap between the Builders and the Extended Builders is large. What was required was not more names associated with new objects but signs used in very different ways, also involving features of the world in different ways. What must be given up, of course, is the idea that they are all like the calls of the Builders. Treating these signs like the primitive names of the Builders is obviously a mistake. The basis for differentiating signs is the kind of training it takes to acquire their use. And the range of possible uses of language has noticeably multiplied. The Extended Builders have achieved systematicity without syntax. 3. PI §15-18: Name-Tags At PI §15 Wittgenstein replaces the calls with names: “naming something is rather like attaching a name tag to a thing”. The name becomes detachable from the object. This was not so either for the simple Builders game or the expanded game. For these games, to call out “Slab!” is for B to take

380

Meredith Williams

hold of a slab. On Certainty includes remarks that resonate with just this feature of the primitive use of language: 510. If I say “Of course I know that that’s a towel” I am making an utterance. I have no thought of verification. For me it is an immediate utterance. …. It is just like directly taking hold of something, as I take hold of my towel without having doubts 511. And yet this direct taking-hold corresponds to a sureness, not to a knowing. But don’t I take hold of a thing’s name like that, too?

The Builders take hold of words in this manner as they take hold of the building blocks. Wittgenstein suggests that this primitive use of words takes the name to be a property of the object by being engraved on the object. Thus, the name and object are one. This primitive foundation of language is never eliminated. It is the bedrock of even the most sophisticated uses of language. But when “slab” becomes a name tag for a kind of object, that tag can be used independently of the presence of its object. Instead of “slab” being a call to action, the name-tag “slab” can be used independently of the action of getting a slab. For example, it can be used in merely looking for a slab. Gazing on a slab deliberately while saying “slab” is not possible for the Builders. Their words are not detachable from their objects or actions. Most of the arguments against the idea that the Builders game is a language appeal to the systematicity of mature languages.13 Before engaging this issue, Wittgenstein introduces the metaphor of language as an ancient city: PI §18 …a maze of little streets and squares, of old and new houses, and of houses with additions from various periods; and this surrounded by a multitude of new boroughs with straight regular streets and uniform houses.

Already we are invited to compare and contrast the Builders game (an ancient game) with “the straight regular streets and uniform houses” of ma13

See G.P. Baker and P.M.S Hacker, Wittgenstein, Understanding and Meaning: An Analytical Commentary on the Philosophical Investigations, Oxford: Blackwell 1980, 65; Robert Brandom, Making It Explicit, Cambridge, Mass.: Harvard University Press 1994, ch. 6; and Warren Goldfarb, “I Want You to Bring Me a Slab,” Synthese 82, (1983).

The Builders

381

ture language. The point is not to deny the differences, especially with respect to systematicity, but to critique how these differences are to be understood. We might hold that the meaning of the Builder’s call “Slab!” can only be captured by what we would translate as “Bring me a slab!” (cf. PI §19). But Wittgenstein asks …And why should I translate the call “Slab!” into a different expression in order to say what someone means by it? And if they mean the same thing – why should I not say: “When he says ‘Slab!’ he means ‘Slab!’? Again, if you can mean “Bring me the slab”, why should you not be able to mean “Slab!”? (PI §19)

Issues of meaning are taken to be bound up with issues of systematicity, at least, the productivity of syntactic structure. What lies behind the view that meaning must mirror syntactic form is a picture of meaning as essentially compositional. This is the semantic counterpart of the systematicity of language at both the syntactic level and logical level. The structure of meaning will be appropriately isomorphic to syntactic structure and will reliably track logical movement.14 The name-as-tag is an intermediary between the subject and the object. It is the first step from primitive language to language as a representational system, a system defined by the primacy of assertoric form, the topic of PI §§19-22. It is in this context that Wittgenstein identifies a second great philosophical confusion: Conflating assertoric form with the proposition and mistakenly taking assertion to consist of two actions, entertaining a content and asserting that content. The product of these mistakes is the philosophical concept of the proposition, the fundamental medium of representational content. 4. Sentences: PI §§19-22 What is worrying, or perhaps useful, about “Slab!” is precisely that it can be treated (by us) both as a word and as a sentence, a holophrastic sentence. The objection to the Builders is that if it can only be translated as “Slab!”, it doesn’t really mean anything. Meaning, even meaningfulness, requires the right sort of complexity. Wittgenstein takes this way of theo14

This is a fundamental methodological and theoretical requirement for the computational theory of mind. See J.A. Fodor and E. Lepore, The Compositionality Papers, Oxford: Clarendon Press 2002 and J.A. Fodor, LOT2, Oxford: Clarendon Press 2008.

382

Meredith Williams

rizing about meaning to be a mistake. He pursues his diagnostic inquiry first with respect to syntactic complexity (PI §20), then semantic complexity (PI §§21-22), and then a return to the name, properly understood (PI §§26-39). Wittgenstein is concerned with why philosophers insist that meaning requires the systematicity of logico-syntactic form. Wittgenstein himself seems to disagree. The Builders mean what they say, and so do we when we take a shortcut, as it were, calling out “Help!” or “Fire!” (PI §27). In any case, the representationalist counter is to say “…when I call out “Slab!”, then what I want is that he should bring me a slab!” Wittgenstein responds in a deflationary manner: “Certainly, but does ‘wanting this’ consists in thinking in some form or other a different sentence from the one you utter” (PI §19). The representationalist insists that syntactic complexity structures meaning. Indeed, compositionality of meaning mirrors the syntactic structure of language. Wittgenstein’s objection is that this is far too narrow a conception of how words mean what they do. Whether someone means one word or four when he says “Bring me a slab” depends on the character of the language more generally: “…we mean the sentence as one consisting of four words when we use it in contrast with other sentences such as “Hand me a slab”, “Bring him a slab”, “Bring two slabs”, etc. …We say that we use the command in contrast with other sentences because our language contains the possibility of those other sentences” (PI §20). The capacity for meaning different things depends, at least in part, on the possibilities our language supports. Broadly speaking, this aligns Wittgenstein more closely to an inferentialist semantics.15 Syntactic structure just is the possibilities for substituting different words to create well-formed sentences. Mastery of language, even with respect to its systematic features, is not the explicit mastery of such syntactic possibilities, not even in thought. It certainly is not the case that being able to navigate syntactic possibilities requires the formulation of the relevant sentences in speech or thought. Wittgenstein’s target here is the simple picture of speech being backed up with syntactically well-formed thoughts 15

Robert Brandom’s inferentialist comes immediately to mind. There is an affinity here, but there are also important differences. Wittgenstein does not think that language is restricted to the logical space of reason. Wittgenstein’s grammar has inferentialist structure in the sense that the grammar of a language game consist of an array of internal relations that obtain among the words in their application.

The Builders

383

that determine just what the speaker really means. Syntactic structure doesn’t contribute to meaning in that sense.16 Having raised doubts about the existence of a mirroring relation between syntactic structure and the compositionality of meaning, Wittgenstein turns to the key philosophical idea that underwrites these views about systematicity and language. This picture is what he calls “Frege’s idea”.17 Frege’s idea makes two claims about the structure of language: The primary thesis is that “…every assertion contains an assumption, which is the thing asserted” (PI §22). The secondary thesis is that assertion consists of two actions, entertaining and asserting” (PI §22). The first thesis expresses a theory of sense, and the second a supplementary theory of force. It is not clear whether Wittgenstein completely repudiates these ideas, but he does certainly repudiate a philosophical account of these two dimensions of language. Very briefly, he objects to the first, that every assertion contains an assumption, with the diagnostic remark that this idea “…really rests on the possibility found in our language of writing every statement in the form: “It is asserted that such-and-such is the case”” (PI §22). And he objects to the supplementary thesis that assertion consists of two actions, holding that this is a mistake. Wittgenstein has no objection to treating language in a way that separates the force with which one says something from the sentence to which the force is applied. So, of course, we can “distinguish an assertion from a fiction or a supposition” or a question, for that matter. The mistake is in thinking that for any linguistic act we must entertain some propositional content as well as assert it or question it or com-

16

It is interesting to note that Chomsky has given up the hypothesis that a large set of sentences, under conditions of transformation and generative grammar, are necessary to the production of meaningful well-formed sentences. His minimalist project eliminates completely that picture of deep grammar. See Noam Chomsky, The Minimalist Program, Cambridge, Mass.: MIT Press 1995. 17 Here is one place that I do not agree with P.M.S. Hacker and J. Schulte’s translation. They translate “Freges Ansicht” as “Frege’s opinion”. In the Anscombe translation it is translated as “Frege’s idea”. This is the translation I have used. The standard translation uses “view”. This standard translation seems the best to me, but I will continue to use “idea” so as not to introduce too many variants. “Opinion” is far too weak for the position Frege defends. He may have been of the opinion that parsnips are best eaten in the winter, but he wasn’t of the opinion that the Begriffschrift is applicable to language. It was his view or picture of language.

384

Meredith Williams

mand it. So, our question is, just what are these two errors and what are the consequences of making them? Earlier I characterized the first philosophical error as a conflation between assertoric form and proposition: “every assertion contains an assumption, which is the thing asserted”. The conflation results from the use of an innocuous expression that Wittgenstein takes to be the source of the thesis. This is “the possibility found in our language of writing every statement in the form: “It is asserted that such-and-such is the case”.” The conflation occurs in using two apparently identical sentences as one, thereby using features from each to create a new and potent philosophical idea, that of the proposition as the medium of meaning. The two banal sentences are: “It is asserted that such-and-such is the case” and “It is asserted: such-and-such is the case.” That these are really different sentences is seen when we delete the phrase “it is asserted” from each of them. What we get is a sentence fragment in the first case, “that such-and-such is the case”, and in the second case we get the sentence “such-and-such is the case.” In this second case, the phrase “it is asserted” is fully superfluous. So, “it is asserted” either introduces a linguistic fragment (which is not meaningful) or it is superfluous (and so explanatorily idle). But in conflating them new logical space seems to open: a sentence that isn’t a sentence seems to have the meaning of an ordinary sentence. It is this combination that gives rise to an explanatorily robust notion of the proposition. The proposition is not a sentence yet it has the constituent structure that mirrors an assertoric sentence and the meaning of an assertion. It is the second mistake, however, that secures a place for an explanatory role for proposition. This second mistake is the view that the proper analysis of the assertoric sentence reveals two actions: entertaining the sentence and asserting the sentence. Frege’s assertion sign “|—” introduces a notational device that shows that the two actions are implicit in every assertion and ultimately every use of language. The horizontal dash represents entertaining a propositional content and the vertical dash represents asserting that content. In hypothesizing two linguistic actions, Frege ensures a unified account of content for any speech act. The meaningful content is the same for each linguistic act while allowing for different forces to be applied. The explanatory notion that comes out of these mistakes is a metaphysically robust notion of the proposition. As a result of Frege’s confusion and mistake, the syntactic structure of language enlists a semantic structure that

The Builders

385

mirrors that syntactic structure precisely because the propositional content is identified through the syntactic assertoric sentence. 5. Ostensively Defined Names: PI §§23-28 Wittgenstein who is suspicious of such easy success immediately turns to the variety of sentences that we use and the variety of speech acts in which we engage (PI §23). “It is interesting,” Wittgenstein concludes, “to compare the multiplicity of the tools in language and the ways they are used, the multiplicity of kinds of word and sentence, with what logicians have said about the structure of language. (Including the author of the Tractatus Logico-Philosophicus.)” (PI §23). The reference to the Tractatus is directly pertinent since the hypothesis of the proposition is designed to ensure that the content of a representational language mirrors its assertoric syntactic structure. The conflation guarantees this, and the mistake extends it to all speech acts. What happens to the name as a consequence of this Fregean sophistication of our simple language games? Naming remains “…something like attaching a name tag to a thing” (PI §26). Now we can ask for the name of something: “One can call this a preparation for the use of a word. But what is it a preparation for?” (PI §26). We supplement languages (2) and (8): “(i)n languages (2) and (8) there was no such thing as asking something’s name. This, with its correlate, ostensive explanation [ostensive definition], is, we might say, a language game on its own” (PI §27). We need a distance between the name and its object. The language game of ostensive definition permits this distance, indeed relies upon it. The importance of this is seen in Wittgenstein’s long discussion of ostensive definition (PI §§26-38). In this new language-game, that of explaining what an object is by giving its name-tag, two players are essential even on the representationalist picture. And this is not just because a question is asked; rather two players are required for the speech act of the inquirer to have meaning. When A asks “what is that called?”, he has not said anything that is meaningful on the representationalist picture. On that picture, the structure of the query is the following: Question (Entertain) (this is called “x”) and (Assert)(this is called “x”) Question (Entertain) (this is called “x”) Question (proposition that this is called “x”)

386

Meredith Williams

But a propositional content cannot involve variables without losing its explanatory role. The whole point of the proposition, if it is to be explanatory, is that it does not stop short of the fact that it expresses. As he says later, “…we – and our meaning [propositional content] – do not stop anywhere short of the fact; but we mean: this–is–so” (PI §95). For this languagegame (ostensive definition), the account of meaning as propositional content cannot apply. Further it is no defense to say that a propositional form is retained because we can express the query as involving “this is called x”, where this could be translated in some appropriate way into the form of an existential quantification: x (x is called “x”)”. That is not a propositional content. It is the form of a proposition within which there is no problem about what “this” is called: It is “ x(x)”. So the propositional analysis defeats the point of the language-game of asking for names. Wittgenstein directs our attention to another place altogether if we wish to know how this language-game is played: This, with its correlate, ostensive definition, is, we might say, a language-game on its own. That is really to say: we are brought up, trained to ask: “What is that called?” – upon which the name is given (PI §27). We cannot be in a practice that is the asking for names unless we can engage in a two-party action. Ostensively defining a word is constitutionally social. The propositional content, in other words, is distributed over the two players. The content “this is called “sepia”” consists of the speech act by A “What is this called?” and the speech act by B “sepia”. Both players are needed. If it is thought that A is eliminable from this game of ostensive defining, then B does not provide an ostensive definition. The language-game isn’t being played. The game of ostensively defining or explaining is a specialized language-game. It requires the name as tag, used by one who is fully competent in the language. Thus, the name’s independence is an illusion. Nonetheless it provides the prototype for the rarified philosophical notion of a “real name” that Wittgenstein introduces in PI §38. 6. The Problem of Normative Similarity: PI §§29-37 The most repeated feature of Wittgenstein’s treatment of ostensive definition is that the language-game requires considerable background for the definition to be successful: “One has already to know (or be able to do) something in order to be capable of asking a thing’s name” (PI §30). This is because “…an ostensive definition can be variously interpreted in every

The Builders

387

case (PI §28). Wittgenstein concludes “(t)his explanation again only tells him the use of the piece because, as we might say, the place for it was already prepared” (PI §31). This is Wittgenstein’s first construction of the problem of normative similarity. The demonstrative or ostensive definition is to fix or set the meaning of a word by relating it to an object or property. Yet closer examination of particular instances of ostensively defining some object or property reveals how difficult it is to conclusively identify the object or property named. To explain “two” to someone, one might point to two nuts. But the pupil might take the expression “two” to name, as a proper name, those two nuts (PI §28). In order to clarify the ostensive situation, the master of language might use sortal expressions – the number “2” or the color “sepia” – to disambiguate the situation (PI §29). Yet for this to help the pupil, the pupil must be clear about the roles played by the identifying concepts “number” or “color”: So, one could say: an ostensive definition explains the use – the meaning – of a word if the role the word is supposed to play in the language is already clear…One has already to know (or be able to do) something before one can ask what something is called. But what does one have to know? (PI §30)

What one has to know is the detailed context within which the object plays a role. Only against such a rich background can an ostensive definition be given successfully. One might be competent in participating in a language game without explicitly knowing the rules for the game or knowing the names for all the objects that play a role in the game. The ostensive explanation of an object does nothing to explain the role of the object within the language game. It provides the name-tag for an object that is already individuated, in large part, by its role (PI §31). Ostensive definition provides very little new understanding of the object that is named. This is Wittgenstein’s primary point in these first passages critically examining what ostensive explanation achieves. Wittgenstein’s own position is that ostensive definition is important especially in providing exemplars or paradigms, but an object can only function in that way within an appropriate context. Being name-tagged does not create the status or role of the object as paradigm. Despite these commonsensical observations about ostensive definition, philosophers search for the essential act of meaning that determines the referent of the name. Mentalistic accounts are the obvious alternative, but they are notoriously elusive and problematic. “Mentalistic” is understood

388

Meredith Williams

in a phenomenological way, the feeling of tracing out the shape of the object (PI §34-35) or the different ways one might mean blue (PI §33). What is needed, it would seem, is the identification of the “…“characteristic experiences” of pointing, say, to the shape” (PI §35). Yet, here too, Wittgenstein argues to identify the characteristic experience of pointing to shape or to color cannot be done independently of the circumstances – that is, on what happened before and after the pointing” – whether we would say “He pointed at the shape and not at the colour” (PI §35). We cannot identify “…any one bodily action which we call pointing at the shape (as opposed to the colour, for example)” (PI §36). In sum, successful ostensive definition adds little to our understanding of the role words play in our languagegames. Ostensive definition is explanatorily shallow. It cannot solve the problem of normative similarity, that is, the problem of identifying or constituting the similarity or sameness of the named objects. The naming of them per se cannot identify them. The philosophical challenge is to find a way of fixing the semantic relation between name and object (or property) that avoids these difficulties. In other words, the challenge is to make naming semantically significant. Naming must be the means whereby language is pinned to the world, as it were, in some especially secure manner. This will be the search for “real names”. 7. “Real Names”: PI §§37-39 The question “What is the relation between name and thing named?” (PI §37) is raised in the context of failure, failure to find any stable or enduring physical or mental property that anchors the name in the world. For those who raise the question, none of the ordinary ways in which names are related to objects or properties. These are just the physical and mental properties that have been found wanting: [B]ecause we cannot specify any one bodily action which we call pointing to the shape (as opposed, for example, to the colour), we say that a spiritual [mental, intellectual] activity corresponds to these words (PI §36).

The time is ripe for a philosophical interpretation of the name relation. This will lead, as is only to be expected, to a strange result: The word “this” has been called the real name; so that anything else we call a name was one only in an inexact, approximate sense (PI §38).

The Builders

389

“This” is called the only real name in spite of the fact that our ordinary uses of the demonstratives “this” and “that”, as in the Extended Builders game (PI §8) and in giving ostensive definitions, are clearly not names at all. It is a philosophical demand that makes their transformation possible, indeed inevitable. The philosophical requirement is a simplicity norm: “a name ought really to signify a simple” (PI §39). The demonstrative “this” seems the only linguistic or representative device that can respect this semantic norm, precisely because it is not an ordinary name. The philosophical use of “this” is a name, sublimated from its coarse ordinary use to its idealized role in language qua logical form.18 We turn to two philosophers, each of whom thinks that the demonstrative plays an indispensable and fundamental role in linking language to the world. These are B. Russell’s treatment of this subject in his The Philosophy of Logical Atomism (1918) and T. Burge’s more recent account of perceptual representation in his Origin of Objectivity (2010). First Russell. Russell’s logical atomism requires, as all atomistic theories do, a specification of the simple building blocks which provide the material content to the theory and a specification of the laws or rules in virtue of which the simples combine to constitute more complex higher level structures. Russell’s theory is directed to language. Formal logic provides the structure of language and names are the simplest linguistic units. Complexity and hier18

Compare PI §§38, 89, and 94. Three passages, in the Anscombe translation, speak of logic as sublime: “This queer conception springs from a tendency to sublime (sublimieren) the logic of our language” (PI §38), speaking of treating the demonstrative as the real name. “These considerations bring us up to the problem: In what sense is logic something sublime (Sublimes)?” (PI §89). “These considerations” are Wittgenstein’s initial characterization of rule following issues. “A ‘proposition’ is a queer thing!’ Here we have in germ the subliming (Sublimierung) of our whole account of logic” (PI §94). The words in bold type are all the English translations of the German. The new Hacker-Schulte translation distinguishes between “sublim” and “Sublimierung”, rendering the first “sublime” and the second “sublimate”. The question in interpreting these passage is how significant the difference in translation is. Anscombe took their use to be the same, all pointing to the way in which philosophers idealize logic. It would matter greatly if Wittgenstein intended his readers to construe “Sublimierung” as psychoanalytic sublimation. I think that Wittgenstein used this expression in a metaphorical or analogical manner as he does when he says that engaging with a philosophical problem is like being subject to mental illness. What he takes from the idea of sublimation is that of replacing something of a coarse or vulgar character (ordinary language) with an ideal (formal logic). That ideal is itself sublime.

390

Meredith Williams

archy within language is revealed through linguistic analysis that eventuates in sentences consisting of proper names only. The name, that is, a real name, cannot be further analyzed nor can its object be anything but simple. For Russell, the relation that obtains between proper name and simple object is that of sensing, and the expression ‘this’ is the only proper name: The only words one does use as names in the logical sense are words like ‘this’ or ‘that’…It is only when you use ‘this’ quite strictly, to stand for an actual object of sense, that it is really a proper name.19

Russell’s mentalistic approach to subliming proper names leads to some understandably comic moments, as when a questioner asks “(i)f the proper name of a thing, a ‘this’, varies from instant to instant, how is it possible to make any argument?”. Russell answers, “(y)ou can keep ‘this’ going for about a minute or two….If you argue quickly, you can get some little way before it is finished.”20 It takes the author of the Tractatus to complete the task of subliming the referential relation and its relata, stripping it of any mentalistic nuances. He describes this project in the Investigations in normative terms: “a name ought really to signify a simple” (PI §39). This is what Russell was attempting to do when he was forced into acknowledging the inevitable consequences of applying that norm.21 The subliming of reference satisfies the norm, by providing the only link possible between proper names and simple objects. Experiential events or objects, even of the sort hypothesized by Russell like a white chalk dot, have multiple properties, being white, being roughly circular, being spatially located. This is a complex object, and so subject to description and the application of general features. The proper name assigned to this object, ‘this’, is inevitably ambiguous in its application. Does it name the color, the shape, the spatial location? In the ordinary sense, the sense that Wittgenstein intended 19

B. Russell, “The Philosophy of Logical Atomism” (1918) in R.C. Marsh, ed.: Logic and Knowledge, New York: Macmillan 1956, 201. 20 Ibid., 203. 21 This is not the only occasion in which philosophical commitments lead to untoward consequences. Consider one example from the philosophy of consciousness: The phenomenal properties of consciousness force the philosopher to admit that states of consciousness are metaphysically aloof from our embodied lives: Qualia have no causal or other relation to our bodies and yet are of the first importance in being in pain or thirsty or seeing red. See D.J. Chalmers, The Conscious Mind, Oxford: Oxford University Press 1996. Chalmers has modified his views in later work.

The Builders

391

with the extended builders game when he added ‘this’ and ‘there’ along with numbers and color words, ‘this’ denotes objects or properties. The demonstratives make a contribution only in the context of the languagegame as a whole. They don’t operate in isolation. But to think of ‘this’ as the only real name is to hold that it is a device of designation that does not rely on any general features of its object or the object’s history or its surrounding context. One cannot explain the referential connection; one cannot even point to the connection. This is because the demonstrative and its object are not detachable. Interestingly this is true of the four calls of the Builders game and their objects. The calls are embedded in the actions of A and B. Removing all slabs while the game is played can only result in confusion and a disruption of the game. This is why their calls are not names. Names must be separable from the presence of their objects. Wittgenstein sees this as the beginning of an objection. He says: The demonstrative “this” can never be without a bearer. – ….But that does not make the word into a name. On the contrary: for a name is not used with, but only explained by means of, the gesture of pointing (PI §45).

In this passage, Wittgenstein breaks the connection between the fact that “this” can only be used indexically in pointing out a present object and the idea that it is used as a name to talk about that object. To combine these two roles in the use of “this” is a misunderstanding of the names. “This” must be used with a gesture of pointing (or some substitute) whereas the whole point of a name is that an object can be picked out without pointing, without its even being present. The demonstrative is a device whereby names are explained. For the representationalist, these differences between names and demonstratives must be rethought or restructured under pressure from the semantic norm that proper names must name simples. What Russell saw as available to our consciousness, indeed what finds its locus in our consciousness, is subject to just the implicit criticism of PI §45. He treats “this” as a name because it is necessarily tied to its object and so has its status as name in virtue of being so bound in the act of ostending; in being so used ‘this’ is itself an ostending. Wittgenstein’s criticism is that the explanation of the referential tie by means of ostension precisely rules “this” out as a name, for, to repeat, “a name is not used with, but only explained by means of, the gesture of pointing” (PI §45). For those still in the grip of the semantic norm that names ought to name simples, an accommodation must found in a different way.

392

Meredith Williams

We have a new “Russell” in Tyler Burge’s theory of autonomous perceptual content. In his Origins of Objectivity22, Burge argues against the pervasive view of perception held by most 20th century philosophers, one he calls “Compensatory Individual Representation.” Early proponents of the view, which include Russell and most other philosophers of the first half of the century, think that perception of the physical environment is possible only via some more primitive form of perceptual awareness, awareness of sensa or sense data. This is clearly true of the Russell of The Philosophy of Logical Atomism. Burge calls this “the first family of Individual Representation”. Far from seeing himself as a new “Russell”, Burge takes Russell to represent just the kind of philosophy of perception that he opposes. So it will take some explaining to show why there is a philosophical affinity between the work of these two very dissimilar philosophers. There is a “second family” of philosophers which Burge also describes as proponents of Compensatory Individual Representation. This second family of theories of perception holds that some supporting or supplementary capacities or techniques are required for perception, say, beliefs or concepts. Wittgenstein is included among these.23 This is surprising since Wittgenstein is neither an individualist nor a representationalist. Indeed if the Builders game does nothing else, it displays the social and nonrepresentational character of primitive language, which is the bedrock for our sophisticated language games. Burge acknowledges the questionable fit, but urges that Wittgenstein too holds that perception requires supplementary capacities on the part of the individual. Burge says: [Wittgenstein’s] emphasis on having criteria for the applications of words and his focus on language and on the complex background to any linguistic reference led others to seek in linguistic usage a basis for understanding the representational aspects of experience.24

Whether we think of the use of “Slab” in the Builders game or what is required for asking for a name or the use of words that have different roles in our perceptual games, some supplementary capacities and skills are in play. Whether these compensatory capacities are engagements in patterns of behavior (the Builders) or the mastery of various techniques for using 22

T. Burge, Origins of Objectivity, Oxford: Clarendon Press 2010. Ibid., 127-129. 24 Ibid.,128. 23

The Builders

393

words (as with the extended Builders), or the dyadic skill of ostensively defining names, they supplement our perception of the physical environment in essential ways. It is for this cluster of considerations that Burge includes Wittgenstein in his second family of Individual Representationalists. (Others include Quine, Davidson, and Sellars.) Burge seeks to articulate and defend a theory of perception in which the representational content of the perceptual state does not depend on any supplemental capacities nor on more primitive forms of phenomenal sensing. Nor is it reducible to something causal like the registrations created by patterns of retinal stimulation. Though perception has its origin in visual registration, the emergence of semantically rich perceptual states cannot be explained in terms of that de facto origin. Indeed it creates the key problem of perception: the problem of underdetermination of perceptual content by causally induced registration.25 How can we explain the emergence of representational perception and how can the objectivity of that perception be guaranteed? The answer is to be found in the conjunction of the following two claims: …whether there are any perceptual states or perceptual systems is an empirical matter. But it is a priori that where there are perceptual states or systems, their representational function is to be accurate… [R]epresentational contents…set veridicality conditions.26

“Perception is a product of objectification,” a process that, in some unknown way, transforms registrations into perceptions.27 Further, “[p]erceptual objectification is carried out automatically and unconsciously.” 28 Objectification thus occurs at the subpersonal level. This is an empirical matter. But given that perception emerges at the personal level, it is a priori that its representational content is its veridicality conditions. We must look to the claim that representational contents set veridicality conditions. It is here, I would suggest, that Burge engages in subliming the semantic content of perceptual states in a way that is analogous to Russell’s subliming the proper name by identifying it with ‘this’. Both involve securing an essential role for the demonstrative ‘this’. In either philosophical 25 26 27 28

See ibid., 90-92 and 344-345. Ibid., 310. The last sentence occurs two paragraphs earlier on the same page. Ibid., 400. Ibid., 401.

394

Meredith Williams

context, the point is to tie the perceptual state to a particular environmental situation. For Russell “this” denotes a particular white dot. Russell’s subject, in some sense, uses “this”. In sensing the dot, the subject denotes the dot. But not so for Burge. Burge’s subject does no pointing. Rather the demonstrative component combines with an essential attributive component to constitute any given perceptual constancy. This use of the attributive avoids the inherent ambiguity of Russell’s “this→dot”, i.e., “this→white” or “this→round”. The very fact that in sensing, the use of “this” disambiguates the chalk dot just is the subliming of the demonstrastive. The logical form of Burge’s demonstrative content is: “this ↓F” or “↓as of red”. This demonstrative content underwrites the perceptual competence of the perceiver: Systematic, repeatable, usually diverse principles for objectification fitted to specific attributes in the environment mark the competences of a perceiver. These structures differ from the generalized weighting and averaging techniques characteristic of sophisticated non-perceptual sensory systems.29

Burge does not appear to have sublimed the demonstrative “this” as Russell did when he claimed “this” is the only name. For Burge, the demonstrative is embedded in a nonpropositional structure that includes an attribute, e.g., red, round, concave, etc., as well as the demonstrative. Whereas Russell clearly did look for a simple object to be the bearer of his name, Burge does not. But Burge is still in the grip of a related philosophical picture. Representational content can be set only by appeal to the veridicality conditions, that is, by semantic conditions, not by causal conditions, nor by techniques of public use. The veridicality conditions are to solve the problem of determining sameness over time and space, where there is no simple causal formula for doing so. This is Burge’s problem of the underdetermination of perceptual content.30 The perceptual constancies seem to him just right for doing this work. They are semantically too rich for causation to solve, but too thin presumably to call for any supplementary capacities or compensatory contexts. What Burge gives us is an a priori claim that the identity conditions of perceptual constancies must be given by their veridicality conditions: “The 29

Ibid., 410. This is one case of a general problem that Wittgenstein recognizes in his later work. I have called it the problem of normative similarity. See my Blind Obedience, op. cit.

30

The Builders

395

representational function of a perceptual system is to represent veridically.”31 This is not an empirical hypothesis, but a conceptual requirement. Secondly, Burge gives us an empirical argument that perception scientists assume just such a view of the perception of the perceptual constancies: “Empirical science has found that explanations that make essential explanatory reference to representational states are fruitful.”32. The philosophical picture in play is something like the “Fido”-FIDO picture of meaning, a crude representationalism. In the perceptual system, we have analogous pairs: representational content “red”-RED, “concave”-CONCAVE, and so on. This is the place at which names find their simple correspondent, namely, the property each name means. Ex hypothesi they are simple in that no compensatory capacities or more primitive perceivings are required or permitted to individuate the relevant property. Rather the representational state just is its veridicality condition. That is a conceptual requirement. Which environmental states realize these various veridicality conditions presumably can only be revealed contingently. The conceptual necessity must be tied in some way to the contingencies of the environment. The perceptual state ‘↓as of red’ is necessarily as-ofred. It is not as-of-red due to being caused by light reflectance properties of a certain band-width. That is contingently associated with being as-of-red. The problem is the identity of the representational content of any particular perceptual state. If the content is as-of-red, then necessarily it is as-of-red. This is true of any representational content. It is necessarily true that any representational content is as-of-what it represents. What in the world distinguishes one veridicality condition from another? A ripe tomato contingently satisfies the veridicality condition for the perceptual state as-of-red. But what are the identity conditions for any given veridicality condition as that is realized in the world? Is that supplied by the representational content? But the representational content is to be determined by its veridicality condition. The a priori truth doesn’t solve the contingent problem on the ground. The a priori truth is necessarily true by conceptual fiat. The bridge from the necessity to the contingency is the demonstrative. Whatever attribute is demonstratively identified constitutes the veridicality condition for the perceptual state. The perceptual state red, the state as-of-red, de31 32

Burge, op.cit., 309. Ibid., 310.

396

Meredith Williams

notes and is necessarily related to the environmental property of being red. This is a confabulated space between causal systems of registrations and the normative practices of our language games.33 This kind of demonstrative does not exist. It is a figment of the philosophical imagination. It is a kind of sublimed denotation. It is for this reason that we can say that Burge is the Russell of our times. 8. Conclusion The norm “a name ought to name a simple object” develops from a mistaken understanding of the complementary roles of names and deonstratives. They cannot be identical. A demonstrative can be used to explain the meaning of a word, such as “red” or “round” or “chair”. But these names can be used independently of their respective objects. That is part of the point of names. The simplicity norm is such that the name cannot be independent of its object. It is this idea of a necessary connection between name and object that lies behind the norm. Russell finds the connection in the nondetachable relation between “this” and its simple object. Burge finds that relation in the demonstrative: It links the perceptual state to its contingently realized veridicality condition, guaranteeing the preservation of the a priori conceptual link between the representational content and veridicality condition. In a sense, the demonstrative cannot but link state to veridical condition. It is this necessary, yet substantive, connection, secured by the demonstrative, that is the subliming of the name, that is, the representational content, as-of-red. It is not that Wittgenstein opposes the idea of necessary or conceptual connection. On the contrary, but for him they can only be understood in the context of the compensatory capacities, skill, and techniques that Burge argues are irrelevant to the representational content of our perceptual states. What Wittgenstein’s narrative of the changing status of names in relation to features of systematicity shows is that the simplicity norm “a

33

Burge attempts to address this by introducing what he calls “natural norms”. Natural norms are standards for fulfilling functions. “Goodness” for the perceptual system “just is success” (ibid., 311). This needs to be addressed more careful than room allows for here.

The Builders

397

name ought to name a simple” is a profound mistake. And replacing a name with the indexical or the demonstrative does not rectify that mistake.

First Person Authority without Glamorous Self-Knowledge ANDREAS KEMMERLING, University of Heidelberg

In this talk, I want to discredit the view that we have a very special sort of knowledge about our own current mental states, events or processes. Of course, I do not deny that we have self-knowledge, but I do want to cast serious doubt on the idea that some of our self-knowledge is far out of the ordinary. Today, I shall focus exclusively on the case of our current conscious thoughts. This case would be a paradigmatic example of the special self-knowledge in question, if there were such a thing. I want to attack what I take to be the best reason for believing in extraordinary self-knowledge – the assumption that it is needed for explaining the special authority we have when it comes to saying what our current mental phenomena are – “first person authority”, as it is called. I accept without reservation that we do have it. But although I accept special authority, I find it hard to believe in a very special sort of knowledge – knowledge of contingent matters more profound than any other. I shall not attack here the general thesis that there is such knowledge. What I shall try to refute is the claim that it is only in virtue of such special knowledge that first person authority can be explained. For this purpose I shall sketch a different kind of explanation which does not presuppose self-knowledge and which is meant to show this: All that is needed, in order to say (with authority) what one thinks, is an ability every normal speaker has – namely, the ability to think out loud. Or so I shall argue. A spin-off of my way of explaining first person authority is that it reveals why it is so strange to speak of knowledge at all in this context. Let me anticipate this one point. It is this: first person authority bears on truth guaranteed by honesty. But if honesty is sufficient for truth-thatcannot-be-doubted, then knowledge is uncalled-for.

400

Andreas Kemmerling

1. The target: Glamorous self-knowledge of sayable thoughts We have self-knowledge of various sorts: knowledge of things we have done or suffered, for example, and some knowledge of who we are: of our character-traits, our temper, our inclinations, weaknesses, feelings, addictions, worries, lusts and so on. Most of this knowledge is human knowledge of the regular kind, nothing exciting about it, epistemologically speaking. Even where we actually do know, we could have been mistaken. But many philosophers assume that some of our self-knowledge is of a very special kind: it is such that we cannot be mistaken about it. It is infallible self-knowledge in the following sense: If it seems to a subject that he so-and-sos (or if he believes that he so-and-sos), then it is somehow guaranteed that he actually so-and-sos. Our current conscious thoughts are one striking example. If it seems to a human subject S, at time t, that he is entertaining a particular thought (e.g., the thought he would express by the words “Harvey is stupid”), then, in this exact moment, he incontestably entertains this thought. Another possibly more controversial example would be one’s current beliefs: If it seems to S (or if he believes) that he holds the belief which he would express by assenting to “Harvey is stupid”, then, incontestably, he does hold this belief, in the exact moment in which it seems so to him. As I said, I shall focus today on the case of current thoughts, and leave the issue of our alleged self-knowledge concerning our beliefs for another occasion. Let me warn you about the slightly technical way in which I shall use the phrase “to think that …”. In saying that S is thinking that such-andsuch, I do not mean to suggest that S judges, or is at least inclined to judge, that such-and-such. What I have in mind is some broader (and, possibly, artificial) sense of thinking that, namely: having or entertaining, in whatever mode, the thought in question. (Modes of entertaining a thought are, for example: judging, doubting, possibly merely “perceiving” it in Descartes’s sense, or “grasping” it in Frege’s sense; but there are many more modes, and I shall briefly return to this point later on). One more technicality. When I say that S thinks that such-and-such, the correctness of what is inserted for the place-holder phrase “such-and-such” is determined by the way S himself would express the thought he is entertaining, if he were a normal competent speaker of English who expresses his thought in English. The reason for this is to avoid side-issues which otherwise would have to be considered very carefully (for example issues about the compatibility

First Person Authority without Glamorous Self-Knowledge

401

of anti-individualism and infallible self-knowledge). – In brief, for the purposes of this paper, whenever I say something like “S (at time t) thinks that Harvey is stupid”, please understand this as an abbreviation for: At t, S is entertaining, in whatever mode, a thought which he would express at t, if he were to express it, at t, in English as a normal competent speaker of this language, by using the very words “Harvey is stupid”.1 Many philosophers have assumed that we have knowledge about what we are currently thinking (in the sense of “think” I’ve just explained). This kind of alleged self-knowledge has been showered with highly honorific epithets like direct, a priori, infallible, self-evident, incorrigible, self-verifying, immediate, arrived at via privileged access, resulting from self-intimation, etc. And it is widely assumed that in voicing one’s current thoughts, one has socalled first person authority.2 Moreover, this is substantial (non-trivial) selfknowledge. By this I mean that the truth of what is known is contingent and is not guaranteed solely by one’s sheer existence or one’s essential properties (like the knowledge of one’s being self-identical, or one’s having a mind). If somebody were to know that he is thinking that Harvey is stupid, then he would know a substantial fact; for he might as well have had another thought instead at this particular moment. – Self-knowledge, which deserved these characterizations (or at least sufficiently many of them), would be what I shall call (epistemologically) glamorous, as opposed to the more standard sorts of knowledge which we have about ourselves. I have certain misgivings about the applicability of some of the attributes I’ve just mentioned, and about the appropriateness of speaking, in this context, of knowledge at all. After all, to say such a thing as “Birgit knows 1

Note that the language is English, actual English, not a language spoken by twinearthlings. – As a consequence of this explication, if S is an entity of which it doesn’t make much sense to assume that it could express itself competently in English, then, to the extent in which this doesn’t make sense, it is left open what is said, if anything, in saying that S thinks that Harvey is stupid. 2 Here are two examples, more or less randomly selected. According to Tyler Burge, what he calls basic self-knowledge is direct, authoritative, non-empirical and selfverifying (Burge 1988), and in his paper “Our Entitlement to Self-Knowledge” (Burge 1996) he says that cogito-like thoughts are direct, self-verifying, infallible and selfevident. Ted Warfield says that knowledge of the contents of our own thoughts is a priori and privileged (Warfield 1992, 232).

402

Andreas Kemmerling

that she is thinking that Harvey is stupid” sounds somewhat strange. To exercise a certain amount of conceptual caution here, is, as far as I can tell, not just a silly quirk. You don’t have to have been under the influence of philosophers like Wittgenstein, Austin or Ryle, to feel that something may be conceptually amiss with such ways of talking. Even Descartes, assumed by many to have been a grand champion of self-knowledge about our current thoughts, would not have put it this way; he would have preferred to say that Birgit is conscious of her thought/thinking, or that she ‘experiences’ that she thinks. (What you find in Descartes’ writings is not: scio me cogitare, but instead phrases like: meae cogitationis conscius sum or experior me cogitare.3) According to Kant’s explication of the concept of knowledge in his Kritik der reinen Vernunft (1787, A 822/B 850), we have no knowledge about our own current mental states. For knowledge entails “objective certainty” and that is, as he puts it, certainty for everybody. Wittgenstein’s opposition to speaking of knowledge in this context is public knowledge. And, to return to the current discussion, Fred Dretske argues that there is no knowledge, and certainly no glamorous knowledge, about the fact that one is thinking, even if it were taken for granted that one knows what one thinks.4 The reasons Dretske puts forward for his, as he labels it, “skepticism” about self-knowledge are clearly different from those which might be found in Descartes, Kant or Wittgenstein. So conceptual caution as to the appropriateness of speaking of knowledge at all, in this context, should not be simply shrugged off. Such caution does not betray undue fondness for being linguistically pernickety. First rate thinkers, past and present, with quite different outlooks on philosophy and the human mind, have given us reason to take this issue seriously. – So we should not take it for granted that we have glamorous knowledge, or even only regular knowledge, of our current thinking. But to question this is, of course, not to question that we are aware of thinking the thoughts we are currently thinking. Rather, it is to doubt that the awareness, or consciousness, we have of our current thinking is a sort of knowledge. *** 3

Cf., for example, his Fourth and Fifth Responses (Descartes 1897–1910, vol. VII, 246, 358 and 359). 4 In a series of recent papers and some unpublished material he was kind enough to make available to me. As to the published papers, cf. Dretske 2003a, 2003b and 2004.

First Person Authority without Glamorous Self-Knowledge

403

As a starting point for the following discussion let us assume the following conditional: If there is glamorous self-knowledge, then some of our current thoughts are items about which we can have it. Which of our current thoughts? Well, at least those which I shall label as sayables, i.e., thoughts which one may formulate in words, completely and exactly as one has them. I assume that some of our conscious thoughts are sayables. And I intend this to be a modest thesis. I am not making any grand claim here about the nature of thinking. I am not saying, like Kant, that “thinking is talking with oneself … and therefore it is also hearing oneself inwardly” (1800, Teil I, § 36). Neither do I want to claim that we, at least sometimes, ‘think in language’. Let me add that having a sayable thought is not the same as merely having a fully meaningful linguistic item going through one’s mind – even if one fully understands the item. A while ago, I couldn’t get rid of a certain line from a song by an English pop group; sadly enough, the words “Reason is treason” lingered in my mind for weeks. But it happened only very rarely during this period that I ever thought, even if only in the sense explained, that reason is treason. – I have no theory to offer about sayable thoughts, but I shall assume that they exist and that we could have glamorous selfknowledge about some of them, if we could have such knowledge at all. To sum up these preliminary remarks: If there is glamorous selfknowledge (in the following often abbreviated as GSK), it is knowledge about a contingent fact about oneself. It is substantial, i.e., cannot be derived from knowing that one exists and has such-&-such essential properties. And it is at least infallible in the sense explained (maybe it has further epistemological qualities like self-evidence etc.) The friend of GSK holds that there is such knowledge and that we have it, inter alia, about our current conscious sayable thoughts. I doubt that there is GSK. But I shall take issue only with a certain claim about it, namely that it is needed in order to explain first person authority. 2. Supreme authoritativeness in saying what one thinks Now let’s turn to our ability to say what we think. Let me first mention an ambiguity of the phrase “(being able) to say what one thinks”. In one read-

404

Andreas Kemmerling

ing the subordinate wh-clause is an indirect interrogative and in another it is a relative clause. There are languages, like English or German, which blur this difference by not marking it syntactically. But there are also languages, like Latin, in which the difference is marked syntactically. In Latin, the indirect-interrogative reading of the wh-clause of “Harvey says what he thinks” is given by a “quid”-clause in the subjunctive mood (“Herveus dicit quid se ipse cogitet”), the relative-clause reading by a “quod”-clause in the indicative mood (“Herveus dicit quod se ipse cogitat”). For the sake of brevity, I shall refer to the two readings as the quid-reading and the quod-reading, respectively.5 The difference between these readings is quite manifest in sentences like “Harvey knows what he thinks”. If this sentence, in the quod-reading, is taken in combination with “What Harvey thinks is that it’s raining”, it follows that Harvey knows that it’s raining. This consequence cannot be drawn if the first sentence is taken in its quid-reading. (Interestingly, “to believe what one thinks” seems to have no quid-reading.) – The ambiguity may be less perspicuous in “being able to say what one thinks”. But it is there nevertheless. In ascribing to S the ability to say what he thinks, we may commit ourselves either to the claim: (Quod)

If S thinks that ---, then he is able to say “…” (where his uttering “…” is a way of exactly saying that ---);

or instead, more naturally, to the claim (Quid)

If S thinks that ---, then he is able to say that (or whether) he thinks that ---.

In this paper, I shall be primarily concerned with the quid-sense of our ability to say what we think. But I shall argue that, very roughly speaking, the ability mentioned in (Quod) is explanatory of the ability (Quid) is about. The way in which we can say what we currently think has some epistemological splendour too. At least if the thought in question is a sayable one. Consider a case in which you say what you think by using a sentence 5

For philosophical applications of this distinction in the theory of perception and theory of meaning see J. L. Austin (1946, 96f.) and E. von Savigny (1969, 35f., 268f.).

First Person Authority without Glamorous Self-Knowledge

405

of the type “I’m thinking that such-&-such”. What you thereby say is that you are thinking that such-&-such. Assume that the thought that such-&such is one of your sayable thoughts. Now in such a case (a) you yourself, given that you speak honestly, cannot doubt what you say, (b) nobody else, except those who have reservations about your honesty, can reasonably doubt what you say6, and (c) nobody else but you could say, in such a doubt-excluding way, what you think. Let us say that S, in a certain context c, says that p with supreme authoritativeness, if the following holds: Whoever assumes that S is honest in his saying that p (in c), cannot rationally doubt that p; but if somebody other than S were to honestly say the same,7 it could be coherently doubted that p. This sort of authority is marked by a distinctive exclusiveness: what one person can say with supreme authoritativeness, only he or she can say in this way. Put differently: It is marked by a distinctive asymmetry, between 6

Well, maybe not nobody else. There are people (“skeptics”) who doubt everything. Or at least philosophers are fond of talking as if there were such people. More interestingly, there are people who conceive of saying what one thinks on a translational model. They may object: “Thoughts are mental representations; and in order to say what one thinks, one has to translate one’s mental representation in question into a sentence of the respective public language. But translation can go wrong. Hence there is always leeway for doubt. Even if I’m honest in my attempt to tell you what I think, I may have mistranslated my thought.” And there may be even others who feel that they could doubt (c). – Thing is, we don’t have to care. For with regard to the dialectical setting of this paper, first person authority is common ground. If those who don’t believe in it anyway have misgivings about the proposed analysans, why worry? They are –on principle, as it were– not party in our discussion, which is a dispute between two opponents who accept that there is first person authority. After all, my explication is not meant to convince anyone who rejects first person authority. Rather than this it is an attempt to clarify a concept which both the friend of glamorous knowledge and I accept in good faith. – Whatever, you are welcome to read “nobody” as short for “nobody except those who don’t believe in first person authority anyway” 7 I am assuming here that there is a good sense in which, e.g., Birgit who says (at t) “I am thinking that Harvey is stupid” says the same as Albert when he says “Birgit (at t) thinks that Harvey is stupid”. – If you have misgivings about the appropriateness of speaking of same-saying here, you are welcome to replace the phrase, “to say the same” by a more convoluted phrase you take to be apt (e.g., “to say something the truth condition of which is the same as the one of Birgit’s saying that she thinks that Harvey is stupid”).

406

Andreas Kemmerling

him, who says something with this sort of authority, and everybody else. His honesty is sufficient for the indubitability of what he says; whereas honesty would not be sufficient, if anybody else said it. I think that my characterization of supreme authoritativeness captures what is meant by the more common term “first person authority”. So I shall use these two terms interchangeably. What can be said with this special authority? Well, clearly anything we can say by using a sentence of the type “I am thinking that p” as a report about oneself and one’s current thinking. Such a ‘real-time autobiographical’ use may be rare or even somewhat recherché. But I accept that it exists and is of some philosophical interest. There may be other things which we can say with this special authority (“I’m in pain”, “I believe …”, “I feel …”, etc.). But, as I said, I shall focus on first person thought ascriptions. Here is an example of what I shall consider as an uncontroversial starting point for my ensuing considerations: Given that the thought that someone would express by saying “Harvey is stupid” is one of her sayable thoughts, then what she would say in using the words “I am thinking that Harvey is stupid” is something she would say with supreme authoritativeness. My central concern in the rest of this talk will be the question: How is this authority to be explained? Or, more specifically: What could plausibly be offered as an explanation, if we abstain from the usual one (“We have this ability because we have self-knowledge”)? I shall restrict myself in this talk to cases in which the subject is a normal speaker of a natural language,8 where the thought in question is sayable and where the overall circumstances of the utterance are normal. I think it’s a good idea to concentrate on such cases, because it is primarily within these limits that we should have some confidence in our judgments concerning the matters at issue. Given these restrictions, there is an intimate connection between alleged GSK about our current sayable thoughts on the one hand, and the ability to 8

A normal speaker of L is, roughly, an adult ‘fluent’ in L. The crucial thing is that he has a high degree of mastery of L: speakers of L who use L as their common language would consider him as of equal linguistic competence. – So, being a normal speaker of L turns on what kind of linguistic ability a person has. (It is not required that he is an actual member of an L-speaking community. Nor is it required that he is a normal person; being linguistically competent to the relevant degree is compatible with having foibles, quirks and abnormalities.)

First Person Authority without Glamorous Self-Knowledge

407

say authoritatively what one thinks, on the other. In fact, I assume that the statements (1) and (2) below are conceptual truths. Let S be a normal speaker of a natural language, who (at time t) has a sayable thought that p. (1) (2)

If S at t has GSK about (his having) the thought in question, then, ceteris paribus, S at t has the ability to say authoritatively that he, at this moment, is thinking that p. If S at t has the ability to answer authoritatively the question whether he, at this moment, is thinking that p, then: If there is GSK at all, then, ceteris paribus, S at t has GSK about (his having) this thought.

Notice that (1) and (2) contain a ceteris paribus clause. This is crucial for their being conceptual truths. Without this qualification, they wouldn’t even be true. Statements that are conceptually true if they contain such a qualification, but may well be false if they don’t, I call cp-necessities.9 – Also notice that the existence of GSK is presupposed neither by (1) nor (2). Both can be readily accepted even by someone, like me, who doesn’t believe in GSK. – I consider these statements as common ground between the friends of GSK and me. *** But given such a close conceptual connection between first person authority and GSK, how can one believe in the one but not in the other? Isn’t the existence of GSK suggested by an inference to the best (in fact, the only available) explanation? “You say you don’t deny first person authority. But how on earth could this authority be explained – except by the fact that we have it in virtue of our GSK about our current thoughts?” – This is a fair question which I shall try to answer. For this purpose, I shall introduce an ability we all have, describe some of its characteristic peculiarities, and try to show that it opens a way to account for the authority with which we can say what our current sayable 9

The cp-clause must of course not be understood as an ad hoc device for immunizing the generalization against refutation (“except when it’s not so”). For more on cpnecessities see Kemmerling (2006, 122f. and 133ff.).

408

Andreas Kemmerling

thoughts are. What makes this sort of explanation interesting is that it does not presuppose self-knowledge, glamorous or not. Later on, in section 5 below, I shall argue that in acting on this ability we may indeed acquire self-knowledge properly so-called, but that the knowledge acquired in this way is regular knowledge: neither infallible nor otherwise glamorous. 3. The ability to think out loud No, I must not try to think, simply utter. Samuel Beckett

Now what explains our ability to say, authoritatively, what we think? I suggest that it is our ability to think out loud. By this I mean the ability to ‘put out’, manifest, or display the thought one is consciously entertaining, in whatever mode, in this exact moment – i.e., the thought one is having as one puts it out. Thoughts can be entertained in various modes, many more than those which are usually considered in the philosophical literature. We can entertain a thought curiously (“what does it amount to?”), critically (“what are its weaknesses?”), affirmatively (as accepted premises in a process of consideration), playfully (as a toy, as it were), daydreamingly, fictionally (as something which may go into the movie-script one is writing), hedonistically (as something which to think is delightfully pleasant), etc., etc. In thinking out loud, we put out the very thought we are having.10 The question in which mode the thought is entertained is not settled (and not 10

It should be noted that what is thought out loud must be a thought. This is to say that what is uttered, in a case of thinking out loud, has to be something which is syntactically and semantically appropriate to complete the lacuna of “the thought that …”. Thinking out loud is indeed a kind of monological talking which is not audiencedirected (or, if you wish, a kind of soliloquizing, even if performed in the presence of an audience). But not each and every kind of speech noise which is produced in this manner constitutes a thinking out loud, in the sense in which I am using this term here. The propositional wheat has to be separated from the chaff of the utterance. – This restriction is unpleasantly strong, in at least one sense: many meaningful parts of what is said in this monological manner might have to be excluded. Questions, exclamations, imperatives, etc. should also be considered as linguistic forms in which some of our sayable thoughts can be revealed exactly as we entertain them. A full-fledged

First Person Authority without Glamorous Self-Knowledge

409

even addressed) by what we do when we think out loud. Metaphorically speaking, in thinking out loud, we turn the mute function off and present sayable thoughts of our current thinking in their original version (without subtitles of any kind). It should be noted that if and when we act on this ability in public – which, by the way, is something we do only very rarely, fortunately–, we do not perform constative or assertive speech acts. In fact, there need no commitment at all concerning the truth of what we utter. We say what we think, in the quod-sense of verbally exposing that which we think. But this does not entail that the thought we are thinking is one we hold true, nor does it mean that by our utterance we report, or assert, that we are thinking the thought we put out. Moreover, the honesty (or frankness) which it characteristically takes a sane adult human being to think out loud, at least in the presence of an audience, is one of its conceptual ingredients. Someone who acts as if he were thinking out loud but in fact tries to hide several of the sayable thoughts he is consciously entertaining, is not really engaged in what he pretends to be doing. He would not be thinking out loud; rather he’d be acting dishonestly. Note also that the honesty involved in thinking out loud does not concern one’s believing to be true what one says, but rather one’s voicing what one is thinking – even if the thought voiced is not one which one holds to be true. So this sort of honesty has more to do with unreserved undisguisedness than with ‘honest truth’. Taken to extremes it amounts to mental exhibitionism, not to unyielding veracity.11 In pure acts of thinking out loud12, it might be argued, we do not even perform any communicative speech acts. At least, we clearly do not, then,

theory of thinking out loud should cover such thoughts too. Lacking such a theory, I grudgingly exclude them. – I hasten to add that none of this affects my case against glamorous self-knowledge. 11 In the next section I shall return to the issue of the conceptual connection between thinking out loud and honesty. 12 Notice the word “pure”. – It might be worth mentioning in passing that impure acts of thinking out loud are much more common. Performing regular speech acts of various sorts, in what Ryle calls normal unstudied talk, can be a way of thinking out loud. (See Ryle 1949, 181 ff.) But it is characteristic of many such ‘impure’ acts that we are aware of the fact that we commit ourselves, in making the utterance, to holding true the thought which we put forward, and sometimes even to the thought’s truth. – In this paper, I shall use “thinking out loud” as a short way of referring to a pure act of think-

410

Andreas Kemmerling

mean anything by our utterances, in the Gricean sense of “to mean something by one’s utterance”. The famous Gricean mechanism is not at work when we think out loud. We don’t intend, when we’re thinking out loud, to get our audience to recognize our higher order intention to recognize our basic intention to bring about, in them, particular beliefs. In fact, we do not even intend to produce beliefs in our audience in the first place, neither beliefs about the topics of our thoughts on display (e.g., Harvey’s being stupid or not) nor beliefs about our own beliefs concerning these topics. The speech act of thinking out loud seems not to fit in the usual taxonomies which focus on communicative speech acts. Superficially, it may seem to bear some similarity with what Searle has called declarations: speech acts like declaring war or making a gift by uttering sentences like “It is war” or “This is yours”.13 But clearly the similarity breaks down at a crucial point: the performance of a declarative speech act brings into existence the very state of affairs which corresponds to the propositional content of the speech act. But when someone says, in the course of thinking out loud, “Harvey is stupid”, he thereby neither brings it about that Harvey is stupid, nor does he thereby bring about his own thought that Harvey is stupid. Furthermore, a thinking out loud is not an expressive like thanking, apologizing, congratulating etc. In performing an expressive the speaker presupposes the truth of the expressed proposition; but it is an essential characteristic of a thinking out loud that it has no such presupposition. A speaker actually expresses some of his attitudes in performing a common speech act (e.g., an assertive, commissive, or directive) which is not formally an expressive. That is to say, he represents himself as having certain beliefs, desires, intentions, preferences, etc. Whereas in a pure thinking out loud nothing is re-presented as being the case. Moreover, it would be misleading to say that the speaker expresses that he is thinking the very thought which he thinks out loud. For to express an attitude (in the ing out loud in the presence of an audience (where it is usually common knowledge between speaker and audience that such an act is being performed). 13 “It is the defining characteristic [of a declaration] that the successful performance of one of its members brings about the correspondence between the propositional content and reality, successful performance guarantees that the propositional content corresponds to the world … Declarations bring about some alteration in the status or condition of the referred to object or objects solely in virtue of the fact that the declaration has been successfully performed.” (Searle 1979, 16f., my italics)

First Person Authority without Glamorous Self-Knowledge

411

sense which is at issue here) is to commit oneself to having the attitude one represents oneself to have. No such commitment is involved in a thinking out loud. – Maybe the term Exhibitive would be apt for labelling the category of speech acts to which thinking out loud belongs. It is the characteristic point of such speech acts that the speaker, in making his utterance, exhibits (presents, rather than represents) himself as being so-and-so. The speaker, in thinking out loud, does not express (in the relevant sense) that he is thinking this-&-that. His utterance is a manifestation – like putting the money in cash on the table, which manifests your having it, as opposed to signing a cheque. One’s entertaining the thought that Harvey is stupid, or rather one’s entertaining it, can be expressed (or attested to) in many ways, for example by raising one’s arm in reaction to the request “Who of you is, right now, entertaining the thought that Harvey is stupid? Please raise your arm exactly when you are entertaining this thought, in whatever mode!”. But in doing so, the arm-raising person does not manifest his having this thought. Even if he is honest, there is leeway for doubt: he may have misheard the request, his arm-raising may be due to a sudden cramp, and so on. A manifestation, sensu stricto, of X is something that leaves no doubt at all about the existence of X. If honesty is granted, thinking out loud counts as a manifestation of the entertaining of the very thought which is displayed acoustically. In normal circumstances, further evidence is neither needed nor would it be clear what it could consist in. Possibly, thinking out loud (with honesty granted) is the only way to manifest, in the realm of the sayable, that we are thinking the particular thought which we think out loud. It’s imaginable that people who are very quick scribblers could do the same in writing. (Remember the automatic writing movement among the Dadaists and some of the Surrealists.) Maybe someone like Jackson Pollock was able to do something of this kind with oil on canvas14; someone like Charlie Parker with a saxophone, someone like Sonny Sharrock with a guitar, etc. – I’m not sure about it, but it seems to me that some of our non-sayable thoughts can also be manifested, in some way analogous to thinking out loud. We don’t do a lot of (pure) thinking out loud, at least not in the presence of an audience. But whether we ever do it or not, we are able to do it. Any 14

Concededly, “thinking out loud” may not be appropriate for describing anything which is done in painting.

412

Andreas Kemmerling

normal speaker has this ability. And when we do it, we are often not aware of doing it. A friend once told me he never does it and more than that: he simply couldn’t do it; he couldn’t even try. A little later when we were playing chess, I heard him mumbling things like: “I move it there, then he will capture it. But that’s O.K. Then I can threaten his queen. Ahem. But that’s no good. Ahem. No, this won’t work. Once again, I move it there, …” Sound volume is not important. Thinking out sotto voce is thinking out loud, albeit in a low voice. 4. The ability to think out loud explanatory of the ability to say what one thinks Now here is one crucial fact about the ability to think out loud, in virtue of which it is explanatory of the ability to say, with absolute authority, what one thinks: Given overall normal circumstances, it is conceptually sufficient. This is to say: (3) is a conceptual truth. (3)

Ceteris paribus: If S is able to think out loud (in L), then he is able to say authoritatively (in L), what he thinks.

Why should this be a cp-necessity? And why should its antecedent be explanatory of the consequent? As to the first question, and leaving the authority issue aside for a moment, it needs no argument to see that anyone who is able to think out loud is normally able to say what he thinks. If he has the ability to present, by uttering “So-&-so”, what he’s thinking –and is, by assumption, a normal speaker and has therefore mastered the use of “I am thinking that …”–, then, in overall normal circumstances, he is also able to say “I’m thinking that so-&-so”, thereby telling us what he he is thinking. – If you don’t think this is obvious, see if the following consideration persuades you: If S is able to think out loud, he is able to verbally present (to quod-say) what he thinks – e.g., by uttering “Harvey is stupid”, thereby performing the exhibitive speech act of thinking out loud. But for anyone who has mastered the self-ascriptive use of “I’m thinking that …”, it follows from this (excuse me for being excessively banal) that he is also able to quid-say what he thinks – by uttering “I’m thinking that Harvey is stupid”, therein performing a truth-evaluable speech act of self-ascription. So (3) is plainly true, if the occurrence of “authoritatively” is left aside. Moreover, in order

First Person Authority without Glamorous Self-Knowledge

413

to recognize this plain truth we don’t have to engage in empirical observation or speculation; it is enough that we have mastered the relevant concepts (of a normal speaker, the ability to think out loud, of a situation in which overall normal conditions obtain, etc.). Now to the authority, which I claim to be a cp-conceptual consequence of the ability to think out loud. Where is it supposed to come from? Well, basically from two things: (4) (5)

Ceteris paribus, in saying what he thinks, the speaker could, if he wished, act on his ability to think out loud. Ceteris paribus, if in fact he acts on his ability to think out loud, his saying what he says has all the marks of supreme authoritativeness in saying what he thinks.

As to (4), I shall sketch my reasons for it in section 7 below. Let me at this point merely try to forestall one objection: Yes, there may be people who are psychologically inhibited, people who in overall normal circumstances could not bring themselves to think out loud (not even under their breath), however hard they try. Yet if they are normal competent speakers, they still have the ability. What they’d lack would be the second order ability to implement it. As to (5), let us first consider why the cp-proviso is required. In acting on the ability to think out loud, a speaker is not immune to the usual linguistic mishaps. He may fall victim to a slip of the tongue, momentary linguistic confusion, etc. In such a case, although acting on the ability, he would not say what he thinks. A fortiori he wouldn’t authoritatively say what he thinks. This has to do with a general fact about abilities. Acting on an ability doesn’t guarantee that the desired result is achieved. Even if I fail in my attempt to jump a metre-high fence, I may have nevertheless brought to bear, on that occasion, my ability to jump much higher than that: somebody may have pulled me down when I was high up in the air. Similarly, when in trying to say what I think, I act on my ability to think out loud, I may nevertheless fail to say what I think: I may misspeak. The ability to think out loud isn’t magical. Something may intervene. Maybe not only slips of the tongue, but also other disruptive factors. Since there is no hope for a complete list of possibly relevant abnormalities (there may be ones we haven’t even thought of), a cp-clause is needed to exclude them. The best we can safely say is this:

414

Andreas Kemmerling

Given overall normal conditions, when a normal speaker, in acting on his ability to think out loud, says what he thinks, then he does so with supreme authoritativeness. And this is nothing but a reformulation of (5). Let us focus now on cases in which we presume overall normal conditions. In such cases it holds that given the speaker’s honesty in saying what he thinks, the truth of what he says cannot be doubted. For as long as we, the others, presume that he, in saying “I’m thinking that so-&-so”, acts on his ability to think out loud, we can have no reason to doubt that he is thinking that so-&-so. Moreover, in acting on his ability to think out loud, he simply cannot mistakenly give us something he actually does not think, in this exact moment, as something he’s thinking. (Remember that we are not talking about what S holds true or judges.) There is just no scope for involuntary mistakes on his part which are not covered by the cp-clause. Errors of memory, for example, are out of the question because of the simultaneity. Such errors would not be out of the question, if instead he were to say: “One moment ago, I had exactly the same thought that you just mentioned: Harvey is stupid”. Even if he’s an honest person to the bone, he may err in this. His actual thought then may have been that Harvey is silly. But if in acting on his ability to think out loud, he says “I’m thinking that Harvey is stupid” (and the situation is normal, as we presume), he simply cannot be mistaken in some such way; he puts out the thought as and how he is entertaining it. So if he acts on this ability, then (presuming overall normality) the question “But was it really what you were thinking at this particular moment?” has only one correct answer: Yes. – In a nutshell, deceitfulness is excluded by the speaker’s honesty, involuntary mishaps are ruled out by the normality constraint. In the light of the exhibitive nature of the act, these are the only sources for coherent doubt of the relevant kind. That is to say, there can be no coherent doubt about the truth of what he says (when he utters “I’m thinking that so-&-so”), thereby acting on his ability to think out loud. But if we said the same (i.e., that he is thinking that so-&-so), the truth of what we say can be doubted. Therefore he speaks with supreme authoritativeness. Hence (5) is true. ***

First Person Authority without Glamorous Self-Knowledge

415

It may be asked: But how, then, could we ever know in a given case whether it was a thinking out loud? We would have to know that the speaker wasn’t being dishonest. And even if we knew this somehow, how could we know that he didn’t misfire – that the circumstances of his utterance were in fact normal? These are fair questions. (And I tend to think that the answer is: We cannot know.) But if such questions were meant as an objection, they’d be off the mark. My claim is that we cannot coherently doubt that we are facing a case of thinking out loud, if we presume the absence of dishonesty and abnormalities. The question whether we can ever know that our presumptions are true is of no concern here. But again, one may feel uneasy about the requirement, or stipulation, that honesty be a conceptual ingredient of thinking out loud. Two remarks on this. First, I hold that this is simply the right way of explicating the concept. Just as the appropriate analysis of informing includes truth (otherwise one has, at best, been misinformed),15 the analysis of thinking out loud should require the speaker’s honesty. If the speaker is not honest, if he, e.g., recites sentences with his mind elsewhere, his utterance ought not to be described as a thinking out loud. Just as there are not two types of informing (truly and falsely), there are not two types of thinking out loud (honest and dishonest). Secondly, no vicious circle is involved in such an explication. The relevant honesty is not to be characterized as saying what one thinks, or suchlike. It is simply the absence of any pertinent intention to conceal, mask or withhold the sayable thought in question. Put in old fashioned terminology, it’s not a ‘real and positive’ cognitive attitude but rather a ‘privation’: the absence of a certain family of such perfidious attitudes. *** So much in support of (4) and (5), which in turn support (3). Let’s assume then that someone’s being able to think out loud is conceptually sufficient16 for his being able to say authoritatively what he thinks. The more ambitious 15

Fred Dretske straightened me out on this in a lecture he gave in Heidelberg 2008. To be fully accurate, I should say “cp-conceptually sufficient”, but I trust no-one will begrudge my not always marking this distinction. The reader should know by now that I consider almost any interesting sort of conceptual relationship as one in which overall normality is essentially involved. 16

416

Andreas Kemmerling

claim is this: The fact that a speaker has the first ability explains that he has the second. What sort of explanation is this? Given the conceptual truth of (3), it’s not a causal explanation. What then is it? A conceptual explanation. What is this? Consider the sentence “S is able to say authoritatively what he thinks, in virtue of his being able to think out loud”. For it to be true, the truth of (3)

Ceteris paribus: If S is able to think out loud (in L), then he is able to say authoritatively, (in L), what he thinks.

is not enough. Mere conceptual sufficiency is not explanatory. (Mr Bird is neither unmarried, nor male, nor adult in virtue of his being a bachelor.) It may be even denied that a conceptual relationship can be explanatory at all. Admittedly, in real life conceptual explanations are rare; usually they are not an appropriate reaction to the request for an explanation.17 I shall not try to argue here that such relationships can be genuinely explanatory (in particular in philosophy), but rather assume that they are.18 Yet, even if the admissibility and potential value of conceptual explanations is granted in general, a question might be raised about the particular explanation I am proposing. It can be put as follows: “Might we not just as well do it the other way round, i.e., ‘explain’ the ability to think out loud in terms of the ability to say what one is thinking? But if this works equally well (or equally poorly), then this shows that your alleged conceptual explanation is no genuine explanation. To pick up your own example: Mr Bird’s bachelorship can be explained in terms of his being adult, unmarried, etc. only because we cannot explain the fact that he is adult, unmarried etc. in terms of his being a bachelor. If there is as much reason for the truth of ‘A in virtue of B’ as there is for the truth of ‘B in virtue of A’, neither sentence expresses a genuine explanation”. – I think that this is cor17

But think of a typical learner’s questions. A novice in chess may ask “Why has Black won this game?” Answers like “Because Black checkmated White” or “Because Black challenged the white king in such a way that he could not avoid being captured” may give him exactly the sort of explanation which was hoped for. More ‘informative’ or ‘substantial’ answers (like “Because from the 12th move on, White was not able to prevent the promotion of Black’s free pawn”) may not yet be helpful. 18 Conceptual explanation appears to be a fairly neglected topic. But see Benjamin Schnieder (Schnieder 2006a, esp. 402-411; 2006b, esp. 31-39) for some interesting thoughts.

First Person Authority without Glamorous Self-Knowledge

417

rect. More than that, it is especially relevant for our considerations because the reversal of (3), namely (R3) Ceteris paribus: If S is able to say authoritatively, (in L), what he thinks, then he is able to think out loud (in L), is also a conceptual truth. That is to say that the two abilities at issue here are conceptually sufficient for one another – on the assumption that S is a normal speaker. (This qualification is crucial for the argument below.) In order now to answer this point let me first of all specify the conditions which have to hold, in general, for an ability A1 to be (in regard to a certain group G) conceptually explanatory of an ability A2: namely, (i) for members of G, having A1 is conceptually sufficient for having A2, (ii) A1 and A2 are distinct abilities, (iii) A1 can be acquired without acquiring A2, but (iv) not vice versa. For our concerns, the group in question consists of the normal speakers of L. Condition (ii) is satisfied in spite of the mutual conceptual sufficiency of the two abilities at issue which we have just noted. The ability to think out loud is distinct from the ability to authoritatively say what one thinks. One can have the one without the other. This is true although a normal speaker indeed cannot fail to have both of them (in tandem, as it were). A child who has not yet fully acquired the English language may nevertheless have mastered enough of it to do some thinking out loud (simple things like, e.g., “Daddy is home”). In particular, she may have acquired this ability without having yet mastered the uses of the word “think”. To see that condition (iii) is satisfied, consider a very crude three-step aetiological picture: Teach a child many things about the use of L, but not yet the use of “to think that” (or of other linguistic devices which serve similar purposes); then teach or encourage her, if needs be, to soliloquize in your presence (so she learns to think out loud); and in the third step teach her the knack about “I am thinking that …”, in the sense of this phrase which is pertinent for our topic. This third step amounts basically to the following: Teach her which parts of her speech productions are such that the result of her prefixing them by “I am thinking that …” is something true. As soon as the child has learnt this, she is able to say, authoritatively, what she thinks. – What is crucial for our present purposes is not that this simple model comes close to the truth about how the ability to correctly use “I’m thinking that …” is actually acquired. It suffices that this model is

418

Andreas Kemmerling

coherent and therefore shows a way how, in principle, the ability to authoritatively say what one thinks can be developed by exploiting an antecedent ability to think out loud (and without presupposing an independent ability to ‘find out’, or ‘come to know’ what one is thinking). As to condition (iv), no equally coherent and persuasive model suggests itself for illustrating how an antecedent authority in quid-saying what one thinks may lead to acquiring the ability to think out loud. Hence all four conditions are satisfied. – Moreover, there seems to be no other, independently plausible reason to assume that we could give an explanation ‘the other way round’. This completes my sketch of an argument for the claim that the ability to think out loud is conceptually explanatory of the ability to authoritatively say what one thinks. 5. Genuine self-knowledge of our thinking unglamorous Above I had expressed some of my misgivings about the appropriateness of speaking of knowledge (instead of awareness) to characterize our epistemological position vis-à-vis our conscious thoughts. In order not to create grave misimpressions, let me emphasize here that I do not, of course, deny the plain fact that we can (and often do) have genuine knowledge of our thoughts and of our thinking them. Neither do I deny that this sort of knowledge has several remarkable features not to be found elsewhere. In the usual course of events, we know what we have just thought and that we have thought it. We know it by simply remembering it. Only in special cases our conscious thinking goes by without leading to knowledge of it – e.g., when we are interrupted or distracted. Most of such knowledge is short-term. Generalizing from my own case I’d say that we tend to forget most of what we have just thought fairly soon. But however short-term it is, as long as such knowledge lingers, it has several characteristics which make it very special. First, there is this striking asymmetry between the one subject who knows what he has just thought and everyone else. Only he can have this piece of knowledge first-hand. Only he can know it ‘for sure’. Second, there is this distinctively high reliability. If it seems to him that he has just thought that such-&-such then almost certainly he actually did have that thought. This gives rise to a third noticeable feature: If he tells us what he has just thought and we grant him honesty, then – presuming overall normal conditions– there is almost no toehold for rea-

First Person Authority without Glamorous Self-Knowledge

419

sonably doubting the truth of what he says. Hence when he tells us what he has just thought, he does so with extraordinary authority. Clearly then, this is a very special sort of knowledge. And some of the honorific epithets mentioned above can arguably be applied to it (“arrived at via privileged access”, for example, in some sense of this phrase). But equally clearly, it isn’t epistemologically glamorous in the sense specified above. Most importantly, it isn’t infallible. Furthermore, the extraordinary authority with which one says that he just had this thought isn’t supreme authoritativeness. Even the thinker himself to whom it seems that he just had the thought that Harvey is stupid may find reason to doubt that he actually did have this thought. Well, he could try to enhance his almost perfect certainty (that he gets it exactly right) by employing his ability to think out loud. Here’s the thought. Imagine someone who sometimes feels a tiny bit uncertain whether he really remembers the exact thought he’s just had. But he is convinced that he remembers some things fully accurately if he hears himself vocalizing them (telephone numbers, for example). So occasionally, when he wants to be as certain as can be about what exactly he is thinking, he thinks out loud. If he now were to utter “Harvey is stupid”, in full awareness, as only he can be, of the fact that he is honest in doing so (and not, for example, inwardly thinking other things while outwardly reciting some text he knows by heart), he might feel that there is no doubt left for him about what it was that he has just been thinking. So much is true: He can legitimately dismiss the possibility of his being dishonest in quod-saying what he thinks. Nobody else is in a position to do that. For anyone else the possibility of the speaker’s being dishonest is bound to be a relevant alternative. Still, even the knowledge he could acquire in this way would not be immune to any sort of rational (albeit farfetched) doubt. After all, might he not fear that some nasty device in his surroundings distorts his speech sound, and other things of this kind? So even for him (who cannot doubt his own honesty and now, on hearing himself thinking out loud, trusts his memory unconditionally), there seems to be no way to still all doubts he may reasonably have. Even he would have to presume (but arguably cannot know) that nothing interferes. In brief, genuine self-knowledge of our ‘current’ thoughts is knowledge about the past, however ‘immediate’ the past may be. Therefore, even if supported by additional sensory (audible) evidence, such self-knowledge is not immune to each and every sort of reasonable doubt. It is an extraordi-

420

Andreas Kemmerling

nary sort of knowledge. But after all it is fallible, like all other knowledge we have of contingent matters. – The rest is self-awareness. 6. Stale glamour Let’s take stock. I have tried here to find a way of explaining our ability to say authoritatively what we think, without presupposing that we have self‘knowledge’. The explanation I have sketched depends on several claims to the effect that certain ceteris paribus generalizations are conceptual truths. Such claims are bound to be contentious. But assuming, ‘for the sake of the argument’, that what I have presented here is on the right track, what would it show about the issue of glamorous self-knowledge? It shows at least one thing, namely that the assumption of such knowledge is not needed. It is not needed in order to account for the most remarkable features characteristic of the special authority with which we can say what we think: infallibility and indubitability. Concerning some of our (first-person) sayings we are infallible, and the truth of what we say cannot be doubted.19 This, as we have seen, can be conceptually explained in terms of our ability to think out loud. So, there seems to be no reason to assume that we have glamorous self-knowledge, or just knowledge, in order to understand whence we have first person authority concerning what we currently think or why we can speak with supreme authoritativeness on this issue. The infallibility which is characteristic of saying what we currently think does not require a cognitive achievement. The infallibility at issue is not due to some cognitive condition which enables the speaker to exclude the relevant alternatives. It’s not due to some supremely reliable process of belief-formation concerning the issue what I am thinking right now. The

19

I am not so sure about the applicability of some of the other labels to be found in the literature. But “a priori” definitely is a horrid misnomer. The alleged knowledge is neither about something which is true a priori, nor can it be had independently of experience. For, first, what one says when one says what one currently thinks is not an a priori truth. Second, one couldn’t come to know what one then allegedly knows without the experience of having the thought in question.

First Person Authority without Glamorous Self-Knowledge

421

speaker requires no reason for believing what he says.20 Nothing like that (reasons, beliefs, etc.) is needed for being able to say authoritatively what one thinks. All you have to do is to exercise your ability to think out loud, or to say the very thing you would have said, had you acted on this ability. Concerning the truth of what one says in saying what one thinks, there is, as it were, only one relevant alternative: dishonesty. However, if lack of dishonesty in saying something is all that is needed to render unassailably true what a speaker says, then he displays no knowledge in saying it. Honesty is a fine thing, and given favorable conditions it may suffice to guarantee the truth of what is said. About judgments of the “I’m Ψing that p” variety –where Ψ stands for a suitable verbum dicendi vel declarandi vel intelligendi vel sentiendi vel affectus– Husserl (1921, II/2, 221) says: “they are indeed true or false, but truth here coincides with honesty [Wahrhaftigkeit]”. I am not sure what exactly he means by “coincide” [zusammenfallen], but I think he might have been making the same point: With regard to some judgments, honesty is conceptually sufficient for truth. But truth is one thing, and knowledge another. Even if truth may sometimes be assured by honesty alone, knowledge never is. If some people nevertheless want to speak of self-knowledge in this context, it is most probably because they find it hard not to speak of a person’s knowing that p, when the proposition is contingent and the person’s honestly saying that p conceptually guarantees that p. For those who are fascinated by the sheer existence of an ability to be (almost) infallibly right about contingent matters, the uplifting word “knowledge” may seem a meagre minimum of linguistic appropriateness. The friends of GSK will typically ask: How on earth could such a grandiose ability not involve knowledge, especially when acting on it entails getting things right? The short answer suggested by this paper is this, to repeat: If honesty in saying something is all it takes for getting certain things right without fail, then knowledge is simply not needed (for getting these things right). Anyway, those who still find it irresistible to speak of knowledge in this connection should also recognize some of the less glamorous aspects of 20

In a way, as we have noticed in section 4, the fact that he says what he says, and hears himself saying it, may give him a reason to believe that what he says is what he is thinking.

422

Andreas Kemmerling

one’s ‘knowledge’ about one’s own current thoughts. Let me, in concluding, mention some of the features I have in mind: (a)

(b)

(c)

Such ‘knowledge’ is cheap – in the sense that it is a conceptual free gift which comes with acquiring the status of a normal speaker. (How do you know that you are thinking that Harvey is stupid? – It would be an answer to say “I have learnt English”.) The infallibility of such ‘knowledge’ is extremely ephemeral – the expiry date for this infallibility is the time the next thought occurs. Its glamour lasts exactly as long as the thinking of the thought. Items of such ‘knowledge’ are indeed substantial, if only in the sense that they are about contingent facts. But usually they are not very interesting, at least not for other people. Usually, we are not curious about what people just think and wouldn’t give a penny to be told. (Lovers, biographers and psychologists may be exceptions.) What we, occasionally, are interested in is what they think-and-believe, or at least think-and-take-to-be-conversationally-relevant. Speaking from my own case, most of my conscious sayable thoughts are not worth mentioning. I myself definitely prefer to live in social surroundings in which people are not given to share with me items of their ‘knowledge’ which have this shabby sort of epistemological glamour.

7. Appendant remarks on thought-reports So far, I have argued that GSK is not needed in order to give an account of first person authority. I want now to briefly consider a point which casts doubt on the explanatory effectualness of the GSK account itself. It has to do with simultaneity. I have been pretending so far that it makes sense to speak of having and expressing a conscious thought at one and the same ‘exact moment’. This pretence has been made for the sake of argument. I myself am suspicious about the feasibility of this idea. It’s fictitious, I think, nothing but a product of philosophical reveries of super-precision. Clearly it plays no role in how we talk about what we think and when. If in order to let us share his current thoughts, somebody said “I’m thinking right now that so-&-so”, it certainly would be fairly odd, if not just outright silly, to ask him “Well, you said right now. What do you mean? When exactly was it that you had that thought? When you started speaking, just

First Person Authority without Glamorous Self-Knowledge

423

before you started speaking, or as long as you were saying it?”. – This being said, let me continue to ignore such misgivings. The friends of GSK depend on the idea of such simultaneity in their account of the special authority we have in saying what we currently think. Let me briefly sketch the GSK picture of what goes on when somebody says “I’m thinking that p”. According to it there are three items involved: (a) the thinking that p, (b) the knowing that one is thinking that p, and (c) the act of assertively ascribing to oneself the property of thinking that p. The simultaneity of (a) and (b) is needed for there to be the alleged glamour of the alleged knowledge. For if there were a time lapse, however short, between (a) and (b), memory would become an issue and therefore a gateway for doubts would be opened. The simultaneity of (b) and (c) is needed to secure the special authority the speaker has in making the selfascriptive utterance. For if there were a time lapse, however short, between (b) and (c), memory would again become an issue, allowing doubts to arise. On this picture, first person authority is crucially tied to the glamour of (b), but this glamour is gone as soon as (a) is past. As soon as the act of thinking that p is over, the knowledge of one’s thinking that p has turned, at best, into regular unglamorous knowledge about something in the past. And such regular knowledge would not be explanatory, in the way envisaged by the friends of GSK, of the extraordinary authority with which we can say what we think. We have first person authority in respect of certain types of assertoric present tense self-ascriptions. Given our focus here (on a person’s current thinking) let’s call them thought reports. Such a report is an assertive speech act. And it is this which creates a certain difficulty for the GSK picture just sketched. In performing an assertive speech act, the speaker means what he says and this involves that he has, or entertains, the thought he expresses by saying what he says. At least in a fully clear case, a speaker who asserts that, e.g., Harvey is stupid has the thought that Harvey is stupid. (If his utterance were an utterly thoughtless production of speech noise, it would not be an assertion at all. If he thought of something else while uttering “Harvey is stupid”, his performance would at best be a deficient case of asserting that Harvey is stupid.) So in a non-deficient thought report (“I’m thinking that Harvey is stupid”) the speaker is bound to think that he is thinking that Harvey is stupid. Let us assume –not unreasonably, I think– that at each point in time a nor-

424

Andreas Kemmerling

mal speaker can consciously entertain at most one sayable thought. Now here’s the difficulty. In the thought report, the speaker ascribes to himself the thought that Harvey is stupid. In reporting this, he is to entertain the thought that he’s thinking that Harvey is stupid. Hence the thought of which he reports that he is entertaining it is not the thought he is entertaining. And it is not clear that there is any way for him –or anyone else for that matter– to achieve the simultaneity which would be needed, according to the GSK picture, for first person authority. At least with respect to reporting it ‘in real-time’, the thought I am having right now seems to be an ever-elusive sort of entity. How does a thinking-out-loud account of first person authority fare with this problem? It may seem that it is inherently hostile to the very idea that there are genuine reports about what one is currently thinking (in an artificially strict sense of “currently”). Given the exhibitive nature of thinking out loud, wouldn’t such an account be bound to construe alleged reports of this kind as ‘avowals’, or at least as something which is strictly speaking neither true nor false? No. There are various options for dealing with ‘realtime’ thought reports. Let me just mention two which can be employed in support of my argument for (3) in section 4 above. The first says that the report given by assertively uttering (σ)

I’m thinking that Harvey is stupid

is a hybrid speech act in which the subject actually acts on his ability to think out loud. To get the idea, think of a paratactic paraphrase of the sentence uttered: (Pσ) I’m thinking that. Harvey is stupid. Note that I am not making any claims here about the syntax or semantics of (σ). Considering this paraphrase at this point is just a preparatory step in sketching the structure of a speech act which can be performed by uttering (σ). It is meant to make it clearer what I am driving at. Now here’s the idea. Construe the utterance of the first part of (Pσ) as an announcement of an immediately ensuing act of thinking out loud. Con-

First Person Authority without Glamorous Self-Knowledge

425

strue the utterance of the second part of (Pσ) as the performance of the act announced. And consider the report made in uttering (σ) as the union of these two speech acts. In assertively uttering (σ) the speaker switches, as it were, from the assertive to the exhibitive use of language.21 As to the truthconditions of such thought reports: What the speaker says, in performing this kind of complex, hybrid speech act is true, given normal conditions, iff he is honest in performing the exhibitive part of the speech act. As regards the report’s truth condition in abnormal circumstances we keep silent. That is to say, given normal conditions the report is true (as it should be) iff the speaker then thinks that Harvey is stupid. A second way in which a thinking-out-loud account could be applied to this issue does not view the speaker as actually acting on his ability to think out loud when he utters (σ). It’s enough that he could have done so. The idea is now that what the speaker says in giving the report is true iff he, at the time of he utterance, could just as well have uttered “Harvey is stupid” as a thinking-out-loud. To put it differently: If he had at this moment, instead of uttering (σ), thought out loud, then “Harvey is stupid” is what he would have uttered. – This is not to say that thought reports really are counterfactuals. We are not concerned here with linguistic analysis or the logical form of thought-report sentences. Rather it is to emphasize that the truth condition of such a self-ascriptive present tense thought report can, for normal circumstances, be specified in a way which is grist to the mill of the thinking-out-loud account. – Both suggestions would need some elaboration. *** All these considerations concerned ‘real-time’ thought reports. Philosophers have a fascination with such reports, but as far as I can tell they are rare in real life. A more realistic way of looking at thought reports would 21

Beware of taking “switch” too literally. This word is definitely not meant to suggest any psychological reality. Compare what one does in the course of saying something like “Harvey gets angry with Birgit whenever she tickles him”: he variously switches from using words referentially to using (other) words predicatively. The switch, in uttering (σ), from asserting to exhibiting would have about the same sort of reality as the switches from referring to predicating in uttering a sentence like the one just mentioned.

426

Andreas Kemmerling

be this. When one is asked at time t what one is thinking and answers promptly, at t’, then one standardly tries (and is taken to try) to say what one was thinking at t. One’s answer is usually meant and understood as a report about an event in the immediate past, or in the ‘specious present’. As we have seen in section 5, in such a report, the speaker may bring genuine self-knowledge to bear. However, for the reasons pointed out above, such knowledge is not glamorous.22 References Austin, J.L., 1946: Other Minds, Proceedings of the Aristotelian Society 20, 148-87, quoted after the reprint in Philosophical Papers, ed. J.O. Urmson and G.J. Warnock, Oxford: Oxford University Press 21970, 76-116 Burge, T., 1988: Individualism and Self-Knowledge, Journal of Philosophy 85, 649663 Burge, T., 1996: Our Entitlement to Self-Knowledge, Proceedings of the Aristotelian Society 96, 91-116 Descartes, René, 1897–1910: Œuvres, ed. C. Adam and P. Tannery, Paris: Cerf Dretske, F., 2003a: How do you know you are not a zombie, in: Privileged Access and First Person Authority, ed. B. Gertler, Burlington: Ashgate, 1-13 Dretske, F., 2003b: Externalism and self knowledge, in: New Essays on Semantic Externalism and Self-Knowledge, ed. S. Nuccetelli, Cambridge, Mass.: MIT Press, 131–142 Dretske, F., 2004: Knowing what you think vs knowing that you think it, in: The Externalist Challenge, Vol. 2, ed. Richard Schantz, Berlin / New York: de Gruyter, 389-399 Husserl, E., 1921: Logische Untersuchungen, 2 volumes, 2nd ed., Halle: Niemeyer Kant, I., 1787: Kritik der reinen Vernunft, 2nd ed., Riga: J. F. Hartknoch Kant, I., 1800: Anthropologie in pragmatischer Hinsicht, 2nd ed., Königsberg: F. Nicolovius Kemmerling, A., 2006: Kripke’s Principle of Disquotation and the Epistemology of Belief Ascription, Facta Philosophica 8, 119-143

22

Thanks to Steven Davis, Fred Dretske, Wolfgang Freitag, Sanford Goldberg, Stefanie Grüne, Klaus Host, Adam Leite, Felix Mühlhölzer, Friederike Schmitz and Marcus Willaschek for their help in finally bringing about the final version. Special thanks to Mark Helme.

First Person Authority without Glamorous Self-Knowledge

427

Ryle, G., 1949: The Concept of Mind, London / New York: Hutchinson’s University Library von Savigny, E., 1969: Die Philosophie der normalen Sprache, Frankfurt a.M.: Suhrkamp Schnieder, B., 2006a: A Certain Kind of Trinity: Dependence, Substance, Explanation, Philosophical Studies 129, 393-419 Schnieder, B., 2006b: Truth-making without Truth-makers, Synthese 152, 21-46 Searle, J.R., 1979: Expression and Meaning, Cambridge: Cambridge University Press Warfield, T., 1992: Privileged Self-Knowledge and Externalism Are Compatible, Analysis 52, 232-237

Subjective and Objective Certainty as Regards Knowledge and Action MICHAEL KOBER, University of Freiburg

1. Wittgenstein’s investigations concerning certainty are deviating substantially from classical or traditional epistemology. For traditional epistemology was mainly concerned with propositional knowledge and propositional certainty, that is with propositions (or judgements, assertions, propositionally formed thoughts, or the like) that count as expressing knowledge and/or certainty. For example, Plato’s famous attempted definition of knowledge as justified true belief (Theaitetus, 201c) made philosophers scrutinizing specific propositions such as “There is an apple in front of me” or “This is a hand” whether or not they are true, whether or not they are justified, and whether or not they express a person’s belief. Additionally, Descartes’ definition of certainty as knowledge that cannot reasonably be doubted (in the first of his Meditationes de prima philosophia) is also restricted to certainties that are instantiated by means of propositions, such as “I think” (or “It thinks”, or “Thinking happens to be the case” respectively). Wittgenstein in On Certainty, however, is departing from these traditional accounts in several respects. To begin with, he did not reflect on propositions ‘as such’, he was not looking for apriori-certainties, rather he was investigating knowledge-claims that are made within some of our (discursive) practices, for example by making an assertion. In other words, Wittgenstein was ‘contextualizing’ the phenomena of knowledge and certainty and thus shifted his focus on the circumstances and preconditions in which knowledge-claims (such as assertions, reports, or messages) are made. Wittgenstein was therefore no longer concerned with a philosophical reflection on what knowledge or certainty is, or, answering sceptical doubts, whether there actually is knowledge or certainty possible, rather he tried to describe, so-to-speak in a phenomenological manner, the role of knowledge and certainty in our everyday practices. As regards certainty, and this has been put forward in the literature on On Certainty within the

430

Michael Kober

past twenty years, Wittgenstein realized that there are at least two kinds of certainty involved in a practice, namely subjective and objective certainty. Objective certainty can be described as the constitutive norms of our practices. For these constitutive norms like “This is a hand” define or establish how participants of the respective practice look at some aspects of the world and how they need to use the word “hand” (this has been developed at some length in Kober 1993, 1996, and 2005). As these certainties constitute their respective practices, it is senseless to call them in question within the very practice they constitute. They can be rated objective since they are, as preconditions, already ‘given’ in advance for all those who are participating in such practices. What I have in mind are, for instance, practices in which we need to use our hands or thematize how to use our hands appropriately, for example during the instruction of young persons who want to become piano-players, mechanics, or other kinds of craftsmen. But these considerations do not imply – as the works of Stanley Cavell (1979), Thompson Clarke (1972), or Barry Stroud (1984) proved, and contrary to some sloppy wordings of Wittgenstein in On Certainty –, that it is not at all possible to call the practice-constituting norms or certainties like “This is a hand” in question. Of course, as for example Sextus Empiricus or Descartes have shown us, they can meaningfully be called in question, for example if one is participating in a philosophical practice, and Wittgenstein is accepting this fact explicitly in OC §452. Therefore, anyone who is claiming that Wittgenstein has refuted scepticism needs to be corrected (cf. Kober 1993, 352-377; Kober 2006b, 141-143). In addition to this adumbrated kind of objective certainty, Wittgenstein has also made us aware of subjective certainty. It is difficult to say whether the phenomenon of subjective certainty still belongs to the domain of the theory of knowledge or to the domain of the philosophy of mind. For subjective certainty can be described as being an attitude or a stance of a participant of a practice who claims to know something. Subjective certainty is meant to be something non-propositional, something non-cognitive, as “something animal”, as Wittgenstein put it (OC §359). This needs to be stressed, for the attitude of being certain is not meant to be some kind of tacit or unconscious knowledge. It is not a conviction or belief-content that usually goes ‘without saying’ or ‘without thinking’, such as “This is Mr. Smith, my next-door neighbour since 20 years” while uttering “Hello” to a man just passing by. The epistemic attitude of being certain rather is like being sure that one has toes while one is preparing oneself to cutting one’s

Subjective and Objective Certainty as Regards Knowledge and Action

431

own toenails (cf. OC §430). Another example might be “The earth has existed many years past” while engaging in a discussion on historical or evolutionary matters. Though we can obviously describe our attitude of being certain or being sure by means of propositions, the attitude of being certain or being sure itself does not consist in holding specific propositions to be true (cf. Kober 2005). Therefore, the title “On Certainty”, that was given to the remarks by the editor and translator G.E.M. Anscombe (as the second editor, Georg Henrik von Wright, once confirmed to me in conversation), is a bit a misnomer. Wittgenstein in his remarks did not often apply the words “gewiss” (“certain”) and “Gewissheit” (“certainty”), but frequently uses phrases like “es ist sicher” and “ich bin sicher” (literally “it is sure” or “I am sure”, for which, however, “it is certain” or “I am certain” are indeed more appropriate in many contexts of English). Perhaps “On Being Sure” would have been a better heading. 2. Wittgenstein presumably adopted the expressions “objective certainty” and “subjective certainty” from the introduction to Gottlob Frege’s Basic Laws of Arithmetic. For Frege, objective certainty relates to something being true because its reference is objectively existent, that is something being real independently from anyone who claims its existence (Frege 1893, xviixviii), whereas subjective certainty concerns the aspect of an individual that is taking something to be true (“ein Fürwahrgehaltenwerden”; Frege 1893, xv-xvii), and for Frege, the logician, subjective certainty is merely an irrelevant psychological aspect of knowing that has definitely to be dismissed from the philosophical grounding of logic and arithmetic as well as from the truth of logical, mathematical, and scientific propositions. Frege was interested only in propositions and their truth-value, not at all in the mental attitudes of those who take some propositions to be true. Frege’s anti-psychological dismissal of subjective certainty thus took place in a setting or argumentative context that differs widely from Wittgenstein’s asthonishing re-introduction of something psychological into his own pragmatically shaped theory of knowledge, or, as one should rather say, into his conceptual investigations of knowledge-practices. As already mentioned above, Wittgenstein was not only looking for the truth of propositions, but mainly describing the practices in which knowledge and cer-

432

Michael Kober

tainty are part of. For if making a knowledge-claim is ‘a move in a language-game’, or a speech-act within a linguistically shaped practice, one also needs, in describing such a practice, to thematize as well the psychological states in which an agent that is bringing forward a claim is in. In sum, Wittgenstein in his remarks of On Certainty used the Fregean expressions “subjective” and “objective certainty” in an entirely different meaning. However, instead of using the expression “objective certainty”, “intersubjective certainty” appears to me to be more appropriate term, since it is a certainty or constitutive norm of a practice only for those who participate in that practice. But no one can intellectually, or philosophically, be forced to participate in any practice, as the pyrrhonian sceptics have shown us (cf. Sextus Empiricus’ presentation of pyrrhonian scepticism, as explained e.g. by Barnes 1983). It may be held that Wittgenstein also changed the meaning from Plato’s attempted definition of knowledge as justified true belief. The psychologically relevant feature of belief in this formulation may also indicate that knowledge can only be ascribed to beings that are capable of having beliefs, or to beings that have a sophisticated mental life in general. These beings, human beings in particular, may also need to be in other psychologically relevant states in order to have knowledge, or to show knowledge and competency, and that may be included in a more comprehensive understanding of Plato’s account of knowledge. In other words: Plato’s characterization of knowledge may be read as if it also hints at the attitude of a person being sure or certain (thus, the present paper merely is another footnote to Plato). 3. In addition to G. E. Moore’s and N. Malcolm’s papers on certainty and common sense, two other sources have been detected that were definitely inspiring Wittgenstein to his notes on sureness and certainty: John Henry Newman’s Essay in Aid of a Grammar of Assent (1890) on the one hand and William James’ The Varieties of Religious Experience (1902) on the other. John Henry Newman already coined the expression “beyond the possibility of doubt” (Newman 1890, 294), which is a forerunner of Wittgenstein’s catching phrase “certain beyond all reasonable doubt” (OC §§416, 607). Newman considered beliefs such as “Great Britain is an island”, held by an Englishman of Newman’s times, as expressing something that is absolutely certain, though hardly anyone actually tested it (Newman 1890,

Subjective and Objective Certainty as Regards Knowledge and Action

433

294). I take this as one of the many origins of what we have now accepted as Wittgenstein’s conception of objective certainty. Newman also used the word “certitude” (ibid.) in order to indicate that persons having such beliefs and sharing them with others belonging to the same culture, society, or form of life should not be rated as irrationally holding fast to some peculiar propositions, but to have developed a specific mental attitude towards their common surroundings. On the other hand, the proper idea that there are not only certainties, that is propositions that function as norms of our practices, but that there is also something like an individual’s attitude of being certain, seems to be rooted in William James’ lectures on so-called ‘religious experiences’ (cf. Kober 2005). James in his published book had given numerous examples of people that show different religious attitudes: being depressed, being elated, or being an atheist, an agnostic, or a passionate believer in God. All these possible attitudes, or “faith-states”, as James called them (James 1902, 247), have in common (i) that they are not substantiated by arguments or justifications and (ii) that they do not ask for any reasons – therefore, they are conceived to be indubitable. Wittgenstein analogously insisted that we all show a kind of being sure, or of subjective certainty, that is needed by our hearts or souls, not by our speculative intellect (cf. Culture and Value 1998, 38-39). Of course, these are attitudes that are describable and identifiable by means of propositions, yet they are attitudes and not propositional beliefs or tacit convictions. For example, the utterance “I believe in the Last Judgment” does not need to be a claim that implies the existence of a powerful transcendent being but may express an awoval-like conviction of feeling safe, come what may (cf. Culture and Value 1998, 96-97; cf. Kober 2006, 102). In order to describe these attitudes, one may draw a parallel to feelings or moods (cf. Kober 2005, 246-249). Clearly, one can identify a tickle on one’s back, or one can realize that one is a bit melancholic today (without knowing why one is melancholic today), yet the feelings and moods themselves are not something propositional. Similarly, epistemic attitudes can be identified by means of propositions such as “The earth has existed many years past” or “The events in the world more or less evolve and cease regularly and do not ‘buck’” (cf. OC §§616-619), but the attitude of being sure or being certain is not of a propositional kind. It is simply there, it permeates all our thinking or our mental life in general. Attitudes, just as moods, are simply there, they do not need any justifications.

434

Michael Kober

4. Admittedly, most of what I have presented so far has already been published, for example by myself, by Danièle Moyal-Sharrock, or by Wolfgang Kienzler (cf. the list of references). The moderate aim of the present paper is to show that the considerations as regards objective and subjective certainty do not only apply to explicit or tacit knowledge-claims within specific, e.g. discursive practices, but to acting in general (remember, making a claim or an assertion is also a kind of acting). Wittgenstein has therefore made many hints on acting in his remarks of On Certainty, and I will try to develop his hints more systematically. It should perhaps be added that John Henry Newman in his Grammar of Assent also stressed several times that some certainties underpin all our judging and acting in general (cf. Kienzler 2006, 131). In reflecting on our intuitive understanding of what it means to be an agent, one may acknowledge the following crude characterization: An act is an intentional and meaningful behaviour in a specific context by a person that has a motive (that is a set of beliefs, emotions, and wishes) and that freely decides in view of one or more reasons to carry out some bodily movements in order to achieve one or more specific goals; the person moreover considers his - or herself responsible of the act and its consequences. Obviously, this account is extremely rationalistic, perhaps hopelessly rationalistic, since we usually are not aware of all our beliefs, emotions, wishes, intentions, and reasons, presumably not even of the most of them. I therefore agree that the adumbrated account of action is so rationalistic that it is not describing our actual acting, but still, under philosophical or juridical circumstances, we may reconstruct an action along these lines. In what follows, I will merely focus on the notion of reason anyway. I will not scrutinize the perplexing issues as regards good or bad reasons for acting (cf. Searle 2001, 111-113). But it seems to be uncontroversial that reasons for acting can be given by means of propositions (cf. Searle 2001, 100-104). However, we usually give our reasons in an incomplete manner. For example, I intend to go out for a walk and pick up an umbrella before I open the front door of my house. My wife asks me: “Why do you take the umbrella?”, and my answer is: “It is likely to rain soon.” This, as a matter of fact, is a correct justification of my action, it is a perfect reason for my

Subjective and Objective Certainty as Regards Knowledge and Action

435

picking up an umbrella (under normal circumstances). Yet, a philosophical nerd may ask me what the ‘logical’ relation between my walk, the rain and my picking up an umbrella is. A full reconstruction of my justification therefore needs to develop what John Searle calls in his theory of action a “total reason” (cf. Searle 2001, 115; Kober/Michel 2011, 104-105). That is to say: In order to provide a philosophically acceptable argument for my picking up an umbrella, one needs to develop at least the following reasoning (“ETRU” means: Excerpt of a Total Reason for using an Umbrella): ETRU: It is more convenient to stay dry than to get wet while taking a walk. Rain may cause that a person who is out for a walk gets wet. There is a convenient way to stay dry while having a walk during a rain shower, and that is using an umbrella in an appropriate way. Therefore, it is a good reason to take along an umbrella if one aims at going out for a walk while having the belief that it may rain soon. Of course, no ordinary person would state his or her reason this way (under normal circumstances) [actually, the greater part of the audience in Kirchberg started to laugh after listening to ETRU], but this is no good objection against the reconstruction just given of why someone is picking up an umbrella. One may rather say that the explicit outline of a total reason of why I was picking up an umbrella made us conscious of many issues and logical relationships that are usually tacit. The whole story was made up in order to show that a proposition like “It is more convenient to stay dry than to get wet while taking a walk” may have a meaningful use. To be sure, it is a prima facie awkward proposition that is hardly uttered outside a philosophical context, yet it is an indispensable part of the reconstruction of the total reason for the action in question. Giving reasons is a possible move in a discursive practice. Reasons are therefore embedded in a common practice. Hence, reasons are reasons for all those who could participate competently in such a practice. One may thus say: There is an objective reason for my picking up the umbrella. This reason belongs to the possibility of the common practice of having a joyful walk in rainy weather, and it is therefore part of the foundations of the practice of being out for a walk. In this sense, the awkward proposition “It is more convenient to stay dry than to get wet while taking a walk” as well as ETRU in toto are examples of an objective certainty concerning the practice of taking a walk. They are accepted by all those who participate –

436

Michael Kober

tacitly, usually unconsciously, but we are allowed to reconstruct them as suggested. Taking Herbert Paul Grice’s account of communicative strategies into consideration, one may even justify why we usually do not say things like “It is more convenient to stay dry than to get wet while taking a walk” [that is: I want to explain why the Kirchberg-audience was right to laugh at ETRU]. In his seminal paper “Logic and Conversation”, Grice established the Cooperative Principle that demands: “Make your conversational contribution such as it is required, at the stage at which it occurs, by the accepted purpose or direction of the talk exchange in which you are engaged” (Grice 1975, 26). From this follows, among other things: “Do not make your contribution more informative than is required” (Grice 1975, 26). As all who are acquainted with what the practice of having a walk under rainy weather conditions amounts to, it is not necessary to make an objective certainty such as “It is more convenient to stay dry than to get wet while taking a walk” as part of the total reason explicit. Therefore, that is, for communicative reasons under normal circumstances, we do not need to mention it when we are asked to justify our picking up an umbrella. 5. Finally, subjective certainty can also be found in action, though one has to look for examples that are not taken from Wittgenstein’s writings. As already mentioned, John Henry Newman described “certitude”, or the attitude of being sure, as being part of anyone who engages in conversation or action. It was not Wittgenstein, but Newman who pointed out that the ways we act show what we are certain of or sure about, though we are not relying on any appropriate proves. As already mentioned, Newman’s example was that English persons of his times who were not convinced that Great Britain is an island would, so-to-speak, cease to be English. Newman indicates: the British act as inhabitants of an island (the wording is by Kienzler 2006, 131). Newman probably had in mind that the British were, without further justification or reasoning, always looking for ship-connections when they were planing to leave their country. Correspondingly, we, who are all competent in the practice of how a walk in the rain may look and feel like, are acting as persons who are trying to avoid getting wet (under normal circumstances). This is our attitude of being sure, shown by the

Subjective and Objective Certainty as Regards Knowledge and Action

437

way how we dress up with specific clothing that is protecting us from rain, or that we pick up an umbrella. Similarly, there is also an attitude of being sure that a bridge will not collapse when we are trying to cross it. As a matter of fact (that is, a conceptual fact), it is not knowledge. For the bridge could indeed collapse, and then the respective proposition were not true, and that is why it is not knowledge, since knowledge implies truth (cf. OC §§600, 603; cf. PI I §§466, 469). No thing, no fact, no person could warrant the ‘absolute’ stability of any bridge, rather it is our attitude that is part of our acting while we are crossing a bridge. Wittgenstein expresses this idea of an attitude that is at the bottom of any action as follows: “[...] even if [ ... a person] never uses the words ‘I know ...’, his conduct exhibits the thing we are concerned with” (OC §427) – and ‘this thing’, I suggest, is the attitude of being sure or certain about something. However, this attitude of being sure is in ordinary language not called “certainty” or “certitude” – these are philosopher’s terms, so-to-speak –, but trust. In other words: The ‘subjective’ attitude of certainty in action is the trust in the regularity of things, events, or persons (cf. OC §§125, 150, 337, 509). It is also shown by the unspoken expectation that facts do not ‘buck’. Under normal circumstances, acting indicates another attitude of being certain which is also of great importance. It is the attitude of “I can do it”, and this is a conviction that we are hardly be conscious of. But any kind of actually carried out acting rests on it – our life shews it, as Wittgenstein put it (cf. OC §7). Again, we should not call this subjective attitude of being sure in acting “certainty”, or even “sureness”. “Confidence” appears to me to be more appropriate. certainty epistemic certainty subjective attitude or stance of being sure about s.th.

objective constitutive norms (rules) of practices

certainty in action subjective a) confidence of oneself b) trust in things, events, persons

objective elements of a total reason

438

Michael Kober

References Barnes, J., 1983: The Beliefs of a Pyrrhonist, Elenchos 4, 5-43 Cavell, S., 1979: The Claim of Reason. Wittgenstein, Skepticism, Morality, and Tragedy, Oxford: Oxford University Press Clarke, T. 1972: The Legacy of Skepticism, Journal of Philosophy 69, 754-769 Frege, G., 1893: The Basic Laws of Arithmetic, Vol. 1, ed. and trans. M. Furth, Berkeley / Los Angeles, University of California Press 1967 (references are made to the original pagination) Grice, H.P., 1975: Logic and Conversation, in: Studies in the Way of Words, ed. H.P. Grice, Cambridge, Mass.: Harvard University Press 1989, 22-40 James, W., 1902: The Varieties of Religious Experience, ed. M.E. Marty, Harmondsworth: Penguin 1982 Kienzler, W., 2006: Wittgenstein and John Henry Newman on Certainty, in: Deepening our Understanding of Wittgenstein, Grazer Philosophische Studien 71, ed. M. Kober, 117-138 Kober, M., 1993: Gewißheit als Norm. Wittgensteins erkenntnistheoretische Untersuchungen in Über Gewißheit, Berlin: de Gruyter Kober, M., 1996: Certainties of a World-Picture. The Epistemological Investigations of On Certainty”, in: The Cambridge Companion to Wittgenstein, ed. H. Sluga and D. Stern, Cambridge: Cambridge University Press, 411-441 Kober, M., 2005: ’In the beginning was the Deed’: Wittgenstein on Knowledge and Religion, in: Readings of Wittgenstein’s On Certainty, ed. D. Moyal-Sharrock and W. Brenner, Houndmills etc.: Palgrave Macmillan, 225-250 Kober, M. (ed.), 2006: Deepening our Understanding of Wittgenstein, Grazer Philosophische Studien 71, 2006 Kober, M., 2006a: Wittgenstein and Religion, in: Deepening our Understanding of Wittgenstein, Grazer Philosophische Studien 71, ed. M. Kober, 87-116 Kober, M., 2006b: Review of Understanding Wittgenstein’s On Certainty by D. MoyalSharrock, in: International Journal of Philosophical Studies 14, 136-143 Kober, M., and J. Michel, 2011: John Searle, Paderborn: Mentis Moyal-Sharrock, D., 2004: Understanding Wittgenstein’s On Certainty, Basingstoke: Palgrave Macmillan Newman, J.H., 1890: Essay in Aid of a Grammar of Assent, London: Longman Green Searle, J.R., 2001: Rationality in Action, Cambridge, Mass.: MIT Press Sextus Empiricus: Outlines of Pyrrhonism, for example in: Benson Mates: The Skeptic Way, Oxford: Oxford University Press 1996 Stroud, B., 1984: The Significance of Philosophical Scepticism, Oxford: Clarendon Press

Zu einigen Bemerkungen Wittgensteins über die Seele RICHARD RAATZSCH, EBS Universität

Abstract In chapter iv of the so-called Part II of his Philosophical Investigations, Wittgenstein makes some rather well-known remarks about the soul. In the paper, some hints will be given to which extent these remarks are more complex than they seem to be. In order to do this, some connections to some of Wittgenstein’s early remarks will be considered. Among the topics discussed in II/iv are the questions whether, or in which sense, there is just one soul as well as the question as to the nature of the concept of the soul. 1. Vorbemerkungen Gegenstand der folgenden Überlegungen ist das iv. Kapitel des sog. Teil II der Philosophischen Untersuchungen (fortan: II/iv), genauer: ein Teil desselben. Enger kann die Textgrundlage kaum sein, die für die Frage nach Wittgensteins „Philosophie der Seele“ heranzuziehen wäre. Denn man könnte durchaus sagen, etwa unter Verweis auf das letzte Kapitel von Teil II der Untersuchungen (siehe: Savigny 1994, S. 10) und auch unabhängig davon, ob man in den PU die Begründung von Thesen sieht oder nicht, man könnte also durchaus sagen, die gesamten PU gehörten direkt oder indirekt zu dieser Textgrundlage. Allerdings sollen im Folgenden auch einige sehr frühe Bemerkungen Wittgensteins herangezogen werden. Dies geschieht jedoch nur, um ein Licht auf die betrachteten Ausführungen aus II/iv zu werfen. Zu diesem Zweck wird hier auch so getan werden, als spiele es keine Rolle, ob Wittgenstein das Wort „Seele“ oder „Geist“ verwendet. Tatsächlich macht Wittgenstein manchmal einen Unterschied zwischen Seele und Geist. Über einen häufigen Diskussionspartner notiert er zum Beispiel: „Er hatte einen hässlichen Geist. Aber keine hässliche Seele.“ (PPO 14ff.) Als Zeichen dafür, dass er keine hässliche Seele hatte, werden dann sein Genuss und

440

Richard Raatzsch

Verständnis von Musik ebenso aufgezählt, wie die Tatsache, dass die Musik stark auf ihn wirkte. Aber bereits zwei Tage nach dieser Notiz heißt es anlässlich einer Kritik einer Aufführung von Bruckners 4. Symphonie, in der dieser neben anderen beschimpft wurde: „In gewissem Sinne fühle ich mich berührt (seltsamerweise) wenn ich denke dass der Geist nie verstanden wird.“ (PPO 18) Man könnte nun meinen, es brauche eine (schöne) Seele, um den Geist verstehen und genießen zu können. Aber diese beiden Beispiele, die leicht um weitere ergänzt werden könnte, legen doch eher Zeugnis dafür ab, dass Wittgenstein die Wörter „Seele“ und „Geist“ so verwendet, wie sie gewöhnlich verwendet werden: in manchen Kontexten sind beide Wörter füreinander einsetzbar, manchmal jedoch werden sie auch verwendet, um einen Unterschied zu machen. Diesen Unterschied kann man dann begrifflich fixieren, aber er kann auch auf die Situation beschränkt bleiben. – Der Zusammenhang von Gleichheit und Verschiedenheit der Bedeutung findet, im Übrigen, seinen Niederschlag auch in philosophischen Texten nicht selten darin, dass systematisch beide Wörter – „Seele“ und „Geist“ – verwendet werden. – Wie auch immer, im Allgemeinen macht Wittgenstein keinen kategorialen Unterschied zwischen Geist und Seele. Der Satz aus II/iv, der hier im Zentrum stehen soll, gehört zu den bekannteren Sätzen der Untersuchungen. 2. „Meine Einstellung zu ihm ist eine Einstellung zur Seele.“ … heißt es ungefähr in der Mitte des kurzen vierten Kapitels. Welcher Gegensatz hiermit angesprochen ist, und damit auch, was der zitierte Satz sagen soll, scheint der nächste Satz des gleichen Abschnitts des Kapitels anzudeuten: Ich habe nicht die Meinung, dass er eine Seele hat.

Hier möchte man im Stillen ergänzen: „Und natürlich glaube ich auch nicht, dass er keine Seele hat.“ Auch ist seine Einstellung zu mir ebenso eine Einstellung zur Seele, wie auch er nicht die Meinung hat, dass ich eine, oder auch keine, Seele habe. – Es geht hier, so gesehen, um den Unterschied zwischen Einstellung zur Seele und Meinung über das Haben einer Seele. Aber so einfach kann es nicht sein. Denn wenn ich nicht der Meinung bin, dass er (k)eine Seele hat, warum sollte meine Einstellung zu ihm dann

Zu einigen Bemerkungen Wittgensteins über die Seele

441

eine Einstellung zur Seele sein? Ähnelte dies nicht dem, dass ich ihm vertraue und – resp.: aber, siehe unten – nicht glaube, dass ihm zu trauen ist? Vor diesem Hintergrund wird es dann bemerkenswert, dass im Text nicht nur Einstellung und Meinung einander gegenübergestellt werden, sondern auch der Artikel wechselt: von der Seele zu einer Seele. Das kann nicht einfach daran liegen, dass Wittgenstein die Wendung „Einstellung zur Seele“ als kürzere Form von „Einstellung zu seiner Seele“ verwendet. G.E.M. Anscombes englische Übersetzung bestätigt dies partiell: My attitude towards him is an attitude towards a soul. I am not of the opinion that he has a soul.

Anscombe übersetzt also nicht mit: „… an attitude towards his soul„. Würde Wittgenstein die Wendung „Einstellung zur Seele“ als kürzere Form von „Einstellung zu seiner Seele“ verwenden, dann wäre unser gegenwärtiges Problem nur noch dringender. Wenn wir den Unterschied zwischen bestimmtem und unbestimmten Artikel dagegen ernst nehmen, haben Einstellung und Meinung nicht unbedingt den gleichen Inhalt, um es so zu auszudrücken. Das aber rückt dann wieder die Betonung – „... nicht die Meinung, dass er eine Seele hat„ – in ein schiefes Licht. Es sieht also danach aus, dass Wittgenstein zwar den Kontrast von Meinung und Einstellung betont, ihn aber zugleich mit dem Kontrast von einer und der Seele verbindet. Das aber scheint den ersten Kontrast partiell wieder aufzuheben. (Diese und ähnliche Phänomene finden sich beim späten Wittgenstein recht häufig, siehe z. B. PU 1 und 3.) Wenden wir uns zunächst dem ersten Kontrast zu, soweit dies ohne Rücksichtnahme auf den zweiten möglich ist. 3. Erster Kontrast: Einstellung und Meinung Die Gegenüberstellung von Einstellung zur Seele und Meinung, dass er eine Seele hat, bedeutet nicht, dass Einstellung und Meinung einfach unabhängig voneinander wären. Das liegt nicht etwa dran, dass das Haben einer Meinung zuweilen auch „Einstellung“ genannt wird. Wittgenstein selbst meint z. B., man könne etwa sagen, das Mitleid sei „eine Form der Überzeugung, dass ein Andrer Schmerzen hat“ (PU 287). Aber mit dieser Bemerkung wird das Mitleid zugleich als eine Form von Überzeugung gerade von anderen Formen des Glaubens unterschieden. Man könnte nun einwenden wollen, dass man doch auch glauben kann, dass jemand leidet,

442

Richard Raatzsch

ohne dass man Mitleid empfindet. Aber daraus lässt sich nicht einfach der Schluss ziehen, dass wir es, wenn jemand Mitleid empfindet, gewöhnlich mit zwei Formen von Überzeugungen zu tun haben: jener Form, welche im Mitleid besteht, und dieser jetzt, die nicht darin besteht. Denn wenn dem so wäre, sollte es nicht seltsam sein, wenn jemand auf die Bemerkung eines Dritten hin, dass er Mitleid mit N.N. empfindet, mit der Frage reagiert, ob er auch glaube, der Andere leide. Wir sehen, die Bemerkung über das Mitleid als Form der Überzeugung ist u.a. ein Hinweis darauf, dass Mitleid nicht die (m.o.w. angemessene) Folge der Überzeugung sei, dass der Bemitleidete leidet. In diesem Lichte ist Selbstmitleid besonders aufschlussreich. Dass Selbstmitleid ein Mangel ist, wäre erstaunlich, wenn Mitleid mit N.N. einfach die angemessene Form der Reaktion auf die Überzeugung wäre, dass N.N. leidet. Wenn man dagegen sagt, dass Selbstmitleid eine Form der Überzeugung ist, dass man selber leidet, macht es die ausdrückliche Anerkennung der Formbestimmtheit der Überzeugung leichter, den Einstellungscharakter des Mitleids anzuerkennen. – Das ist insofern kein entscheidendes Argument, als die Ansicht, Selbstmitleid sei ein Mangel, sich im Rahmen der anderen Auffassung dadurch retten ließe, dass man sagt, der Fall des eigenen Leids sei von dem Fall des Leidens eines Andern in der Weise verschieden, dass die Angemessenheit des Mitleids als Reaktion auf die Überzeugung, dass man leidet, verloren geht. – Der Punkt ist, dass man die Ansicht retten muss. Dass jemand kein Mitleid empfindet, obwohl es nicht der Fall ist, dass er nicht glaubt, dass der Andere leidet, besteht dann nicht einfach darin, dass etwas ausbleibt, was gewöhnlich oder jedenfalls bei dem, der Mitleid empfindet, zu seiner sozusagen „reinen“ Überzeugung hinzukommt. Sondern dieses Ausbleiben von Mitleid steht auf einer Stufe mit dem Mitleid – wir nennen es etwa „Gleichgültigkeit“ oder „Gefühlskälte“. Als solche ist sie auch eine „Form der Überzeugung, dass ein Andrer Schmerzen hat“, wenn auch eine Fehlform. Wenn in II/iv also Einstellung und Meinung einander gegenübergestellt werden, dann ist meine Einstellung zu ihm als Einstellung zur Seele, wenn überhaupt, dann keine Überzeugung von der Form, welche etwa die Überzeugung besitzt, dass er blaue Augen hat. Das ist in II/iv nicht mit „Einstellung“ gemeint. Meine Einstellung zu ihm soll ja gerade keine (solche) Meinung über ihn sein. Der Betrachtung des Kontrastes von Meinung und Einstellung vor dem Hintergrund der Rede von Formen von Überzeugungen korrespondiert die

Zu einigen Bemerkungen Wittgensteins über die Seele

443

Betrachtung des folgenden Satzes, auf den oben schon kurz angespielt wurde: Meine Einstellung zu ihm ist eine Einstellung zur Seele, und ich glaube nicht, dass er eine Seele hat.

Das klingt auf eine Weise schief, dass man versucht ist, folgende Substitution vorzunehmen: Meine Einstellung zu ihm ist eine Einstellung zur Seele, obwohl ich nicht der Meinung bin, dass er eine Seele hat.

Oder, noch ausdrücklicher, so: Zwar glaube ich nicht, dass er eine Seele hat, aber meine Einstellung zu ihm ist trotzdem eine Einstellung zur Seele.

Das heißt, die Tatsache, dass ich nicht den Glauben habe, dass er eine Seele hat, erscheint als erklärungsbedürftig, wenn meine Einstellung zu ihm eine Einstellung zur Seele ist; und umgekehrt. Also hängen beide Sätze auch zusammen. Die Möglichkeit, dass Mitleid – eine Einstellung – eine Form von Überzeugung sei, macht diesen Zusammenhang schon terminologisch deutlich. Halten wir nun die beiden letzten betrachteten Sätze oben vor den folgenden: Zwar hat er kein Kind, jedoch ist er trotzdem Vater.

so spüren wir, dass es sich bei jenen Sätzen oben nicht einfach um Widersprüche handelt. (In der Formsprache: Kein-Vater-sein ist entweder keine Form von Vaterschaft – wie die biologische, rechtliche, akademische usw. – oder eine andersartige Form.) Ebenso wenig gleichen jene beiden Sätze formal dem Satz: Er geriet in den Regen, wurde aber dennoch nicht nass. (Er hatte einen Schirm dabei.)

Die Beziehung zwischen einer Einstellung zur Seele und der Überzeugung, dass er eine Seele hat, scheint insofern zwischen einem logischen Widerspruch und, um es sprachlich etwas anzugleichen, einem empirischen Gegensatz angesiedelt, ähnlich dem Satz „Es regnet, aber ich glaube es nicht“. Bevor wir dem näher nachgehen, lohnt sich ein Blick auf den zweiten Kontrast. Dieser verbindet unsere beiden Sätze dann auch deutlicher mit frühen Überlegungen Wittgensteins.

444

Richard Raatzsch

4. Zweiter Kontrast: die Seele und eine Seele Meine Einstellung zu ihm als Einstellung zur Seele ist eine Einstellung zur einen und einzigen Seele. Meine Einstellung zu ihr wäre also insofern nicht nur der Art nach die gleiche wie meine Einstellung zu ihm, als sie nicht die Einstellung zu einer anderen Seele wäre, so wie sie eine andere Person ist als er, oder ich ihr vertraue und ihm misstraue. Man könnte nun meinen, es sei nicht weiter entscheidend, dass es sowohl mit als auch ohne „nicht“ sprachwidrig ist zu sagen: Ich habe (nicht) die Meinung, dass er die Seele hat.

Denn diesen Satz könnte man einfach umformen zu: Ich habe nicht die Meinung, dass er an der Seele teilhat.

Aber der Satz, dass ich nicht der Meinung bin, dass er eine Seele hat, legt natürlich nahe, dass jeder seine Seele hat, er also eine andere als sie und ich wieder eine andere, eben meine Seele – wie man auch sagen könnte, dass (meine Überzeugung, dass) er den und den Charakter hat, nichts darüber sagt, ob (ich auch glaube, dass) sie diesen Charakter hat, oder wie wir zwischen meinem, ihrem und seinem Körper unterscheiden. Diese Freiheit besteht nicht mehr in Bezug auf den Ausdruck „Einstellung zur Seele“. Insofern wäre der Übergang von „Ich habe nicht die Meinung, dass er eine Seele hat“ zu „Ich habe nicht die Meinung, dass er an der Seele teilhat“ so wenig philosophisch unschuldig, wie Lichtenberg eine philosophisch bedeutsame Bemerkung macht, indem er meint, man solle statt „Ich denke“ sagen „Es denkt“ – wie man auch sagt „Es blitzt“. (Lichtenberg, Bemerkung K 76, S. 412.) Und soweit die Übersetzung korrekt ist, erklärt sich auch, wieso es oben wie ein Widerspruch aussieht. Aber dann kann man die Übersetzung eben auch umdrehen. Und: woran hat man dann Teil? (Wenn Wittgenstein über seinen Diskussionspartner sagt, er habe keine hässliche Seele, wenn auch … (s.o), dann scheint er eine Meinung auszudrücken – und diese scheint eine weitere Meinung vorauszusetzen, dass der Andere eine Seele hatte! Diese aber muss dann wohl verschieden sein von der hässlichen Seele eines Dritten.) Ein Indiz, mehr ist es zunächst nicht, ein Indiz also dafür, den zweiten Kontrast – die Seele vs. eine Seele – nicht über die Möglichkeit jener Umformung von „eine Seele haben“ zu „an der Seele teilhaben“ zugunsten des ersten – Einstellung vs. Meinung – beiseite zu lassen, ist die Tatsache, dass

Zu einigen Bemerkungen Wittgensteins über die Seele

445

Wittgenstein in einigen frühen Bemerkungen den zweiten Kontrast in einer Weise behandelt, die ein Licht auf wirft auf den Satz: „Meine Einstellung zu ihm ist eine Einstellung zur Seele.“. 5. Wittgenstein am 15. Oktober 1916 Hier ist der vierte Satz aus einer Reihe von Überlegungen in einem Tagebucheintrag mit diesem Datum: Eine Auffassung: Wie ich aus meiner Physiognomie auf meinen Geist (Charakter, Willen) schließen kann, so aus der Physiognomie jedes Dings auf seinen Geist (Willen).

Diese „Auffassung“ ist nun nicht einfach Wittgensteins, sondern vielmehr eine (von ihm) zu betrachtende „Auffassung“. Der erste Punkt, auf den es hier ankommt, betrifft das vermeintliche Schließen. Soll der Schluss nach dem Muster von „Es regnet, also ist die Straße nass“ erfolgen, wäre es 1916 noch kein Schluss, weil der „Kausalnexus gar kein Nexus ist“, wie es am 15.10. vier Sätze weiter unten heißt. Aber es handelt sich beim Schluss von der Physiognomie auf den Geist auch nicht einfach um einen Übergang von einem zu einem anderen Wort mit der gleichen Bedeutung, und schon gar nicht um einen Schluss von der Art „Wenn er rote Ohren hat, dann hat er Ohren“, wenn man das noch „Schluss“ nennen kann. Dass wir von einem Schluss von der Physiognomie auf den Geist reden, hängt vielmehr damit zusammen, dass mein Körper der „interne Ausdruck von etwas“ ist, dass das böse Gesicht z. B. „an sich böse“ ist, und nicht „bloß, weil es empirisch mit böser Laune verbunden ist“. In der Tat, und um es paradox auszudrücken, wir sagen nicht jedes Mal, wenn jemand ein böses Gesicht macht, dass er wahrscheinlich ein böses Gesicht macht. Das heißt nicht schon, dass ein böses Gesicht sozusagen als solches, im Unterschied zu: an sich, böse wäre, also böse, was immer sonst der Fall sein mag. („Ein lächelnder Mund“, heißt es in PU 583, „lächelt nur in einem menschlichen Gesicht“. In einem nichtmenschlichen Gesicht oder ohne jedes Gesicht lächelt ein lächelnder Mund also nicht – was schon andeutet, dass auch der nichtnegierte Satz vielleicht nicht geradeheraus zu verstehen ist.) Als Ausdruck böser Laune ist das böse Gesicht natürlich weder mit der Laune empirisch verbunden, noch ist es mit ihr identisch. Und in der Tat, ich kann vergessen haben, wie ihr Gesicht aussah, aber mich doch daran

446

Richard Raatzsch

erinnern, dass sie böser Laune war. Nur, wenn ich nun etwa zeichnen sollte, wie sie ausgesehen haben könnte, als sie böser Laune war – würde ich ihr Gesicht da nicht so zeichnen (sollen), dass ein Andrer erkennen könnte, dass sie ein böses Gesicht macht? Um zu sehen, wie Physiognomie und Geist zueinander stehen, hilft es daher, daran zu denken, dass ein Gesicht nicht nur rund oder spitz sein und verschiedene Farben haben kann usw., sondern auch neugierig, überrascht, wütend oder eben böse. Wenn nun jemand ein böses Gesicht macht, dann weist sein Gesicht nicht zusätzlich zu, oder neben, den Merkmalen, die es schon hat, ein weiteres auf: das Merkmal des Bösen. Denn was heißt hier „neben“, „zusätzlich“ und „aufweisen“? Es hilft auch nichts, diese Wendungen als metaphorisch zurückzuweisen. Denn was bedeutet „und“, wenn man sagt, es weise diese und jene Merkmale auf? Lehrt Ryles „Sie kam in einer Sänfte und in einer Flut von Tränen“ uns nicht, dass man nicht Beliebiges sinnvoll mittels „und“ verbinden kann? Und was bedeutet „Merkmal“? Was in Ryles Beispiel offensichtlich ist, ist es im Fall des Redens von Merkmalen, oder auch Eigenschaften, eher nicht. Aber durch die Fragen nach der Bedeutung von „zusätzlich zu“, „neben“ und „und“ wird es mehr und mehr ans Licht gebracht. Die Schwierigkeit beginnt also nicht erst, nachdem man etwa gesagt hat: wenn man die Rundheit eines Gesichts, die Ausgeprägtheit seiner Züge, die Art, wie die Augen liegen und geformt sind usw. „die natürlichen Eigenschaften des Gesichts“ nennt, ist das Böse eines Gesichts keine natürliche Eigenschaft, sondern eine nichtnatürliche Eigenschaft, um es mit Moore zu sagen. Die eigentliche Schwierigkeit liegt nicht darin, zu klären, wie sich diese beiden Arten von Eigenschaften zueinander verhalten, sondern darin, sich zu fragen, mit welchem Recht man in beiden Fällen überhaupt von Eigenschaften spricht. Von natürlichen und nichtnatürlichen Eigenschaften zu reden, bedeutet ja nicht, eine etablierte Unterscheidung zu benutzen, die eine Hinsicht der Unterscheidung und damit Möglichkeiten der Beziehung beider Arten von Eigenschaften mit sich brächte. Sondern der Ausdruck „nichtnatürliche Eigenschaft“ wurde eingeführt, um einen Unterschied zu benennen – oder wie wir jetzt genauer sagen müssten: um etwas zu benennen, was uns wie ein Unterschied vorkam, denn warum soll für das Wort „Unterschied“ etwas anderes gelten als für die Wörter „Eigenschaft“ oder „Merkmal“? Es ist also auch kein Wunder, dass unklar ist, was „neben“, „zusätzlich“ und „aufweisen“ bedeuten, wenn man sagt, etwas habe zusätzlich zu, oder neben, seinen natürlichen auch noch nichtnatürliche Eigen-

Zu einigen Bemerkungen Wittgensteins über die Seele

447

schaften. Das ist ebenso wenig erstaunlich wie die Tatsache, dass wir, wenn wir fragen, wie sich natürliche zu nichtnatürlichen Eigenschaften verhalten, den entscheidenden Schritt schon gemacht haben. Dass wir trotz der Probleme, die sich dann einstellen, so reden, wird zwar verständlich, wenn man sich vor Augen hält, dass etwas neben dieser oder jener Form, oder zusätzlich zu ihr, auch noch diese oder jene Färbung aufweisen kann, dass also etwas diese neben jener Eigenschaft haben kann, und dass man in diesem Sinne auch von Arten von Eigenschaften reden kann, etwa wenn man sagt, etwas habe neben allgemein bekannten auch noch einige wenig bekannte Eigenschaften. Aber was in Bezug auf Form und Farbe und Arten von Eigenschaften wie die genannten klar ist, ist es in Hinsicht auf „natürliche versus nichtnatürliche Eigenschaften“ gerade nicht. Es würde auch nicht helfen, zu sagen, wir wären, wenn wir ein böses Gesicht sehen oder auch, wie eine Person sich freut, insofern schon über das hinausgegangen, was wir (wirklich) vor uns haben, als dies unter eine „Minimalbeschreibung der Farben, Formen und Hautverschiebungen fällt“, wie Noel Fleming es ausdrückt (Fleming 1998, S. 245, die folgenden Zitate von S. 251). Über eine „Minimalbeschreibung“ hinauszugehen, bedeutet, dass dasjenige, was in der „Minimalbeschreibung“ vorkommt, in dem, was über sie hinausgeht, enthalten ist. Aber wenn ich sage, dass er sich freut, ist nicht nur weder von Farben noch von Formen oder gar Hautverschiebungen die Rede, sondern es wäre auch ganz unklar, was es bedeuten würde, wenn wir in diesem Zusammenhang von Formen, Farben usw. reden würden. (Wenn wir sagen „Er war ganz rot vor Wut“ beschreiben wir den Grad seiner Wut oder sagen, warum sein Gesicht rot war resp. was diese Röte bedeutet, und sagen weder einfach, dass er ein wütendes Gesicht gemacht hat, noch dass sein Gesicht rot war, auch wenn der Satz natürlich falsch sein würde, wäre sein Gesicht nicht rot.) Vielleicht soll die Idee einer „Minimalbeschreibung“ lediglich darauf hinweisen, dass irgendetwas Physisches, Körperliches geschieht, wenn jemand böse ist oder sich freut (siehe oben). Das kann im besonderen Fall nur das Äußern eines Lautes sein, selbst wenn es gewöhnlich auch noch ein Verziehen des Gesichtes und ein Ändern der Körperhaltung ist. – Aber die Frage ist ja gerade, in welchem Sinn man über eine solche „Minimalbeschreibung“ hinausgeht. In welchem Sinn gehe ich denn über einen mathematischen Satz hinaus, wenn ich sage: „2+2=4 ist ein ödes Beispiel“? Natürlich kann jemandes Freude in seinem Gesicht oder an seiner Haltung zu sehen sein, so dass man meinen könnte, es müsse schon feststehen,

448

Richard Raatzsch

dass es sich um ein Gesicht handelt, und ein Gesicht sei nun einmal rund oder spitz oder was auch immer. Aber soweit das ein wichtiger Punkt ist, ist es eben auch die Tatsache, dass wir hier nicht fragen: Wo im Gesicht, oder an der Haltung, sehen wir die Freude? – wie wir etwa „Wo?“ fragen können, wenn jemand sagt, er habe den und den in der Menge stehen sehen, einen Farbfleck an jemandes Jacke oder eben einen Leberfleck in seinem Gesicht bemerkt. Zu sagen, dass „(j)eder physikalistischen oder nichtmentalistischen Beschreibung … das Leben und der Geist, den wir an der Haltung und am Verhalten des Körpers – der Person – wahrnehmen, ... entgehen (wird)“, ist, soweit die Sache nicht schon im Wort „nichtmentalistisch“ steckt, kaum mehr als eine verneinende Geste in Richtung des Physikalismus. Denn der besagt ja gerade, dass es neben jenem Nichtmentalen einfach nichts gibt, also auch nichts, was ihm entgehen könnte. (Dies ist freilich auch nur eine Geste, die auszuführen das Gegenteil dessen bedeutet, was sie, angeblich, bestreitet.) Sollte dagegen gar nicht der Physikalismus gemeint sein, sondern etwa die Physik, dann wird dieser unterstellt, etwas nicht zu können, was zu können sie gar nicht beansprucht. Wittgensteins Redeweise vom Körper, der „selbst der interne Ausdruck von etwas“ ist, oder vom bösen Gesicht als „an sich böse“ ist zwar von gleicher logischer Multiplizität wie diejenige von natürlichen und nichtnatürlichen Eigenschaften oder die von „Minimalbeschreibung“ und dem, was über sie hinausgeht. Sie fällt aber mit keiner der beiden zusammen. Denn wenn mein Körper und mein Gesicht „selbst der interne Ausdruck von etwas“ sind, so gehört es bereits zu meinem Körper und zu meinem Gesicht, wie sie sind, dies oder das auszudrücken. (Diesen Punkt teilt auch Fleming.) Ja, es gehört auch zu meinem Gesicht, nichts ausdrücken zu können, nichtssagend zu sein, während es schwerlich jede geometrische Form verlieren kann. – Und wäre nun ein ausdrucksloses Gesicht etwas, was einem Physiker sozusagen nicht entgehen kann? Sollten aber nicht vielmehr „natürliche (und nichtnatürliche) Eigenschaften“ resp. „Minimalbeschreibung“ und „mentalistische Beschreibung“ auseinanderfallen und zumindest die ersten Glieder beider Paare für sich sein können? (Wenn das Gute (Böse, der Geist) historisch (kausal) entstanden ist, dann gab es eine Zeit ohne Böses (Gutes, den Geist). Hier gab es nur „natürliche Eigenschaften“, denen eine „Minimalbeschreibung“ gerecht würde. – Nur, ist es auch richtig, so von der Welt, dem Bösen und dem Geist zu reden? Oder gehört diese Sprechweise eher zum Material, welches philosophisch zu betrachten ist?) Dagegen wäre es ungereimt, zu sagen, dass etwas der Fall

Zu einigen Bemerkungen Wittgensteins über die Seele

449

sein kann, ohne dass zugleich seiner interner Ausdruck vorliegt, wenn es einen solchen hat. Anders gesagt, es kann Dinge geben, die keinen internen Ausdruck haben. Aber wenn etwas einen hat, dann muss er auch da sein, wenn das, was ihn hat, da ist. Interne Ausdrücke sind nicht wie Inhalte von Teekannen. Man könnte einwenden, dass man doch z. B. jemandem, der erfolglos versucht, ein böses Gesicht zu machen, u.a. dadurch auf die richtige Spur setzen kann, dass man ihm sagt, er solle die Lippen zusammenkneifen. Wenn er sie nun zusammenkneift, dann gleichen sie eher einem Strich als vorher; und was könnte „minimaler“ sein als diese Beschreibung? Aber soweit man so sprechen will, kann ein Gesicht auch dadurch böser werden, dass die Lippen weniger strichartig werden; und dass sie strichartiger werden, kann ein Gesicht auch freundlicher machen. Auch erkennt man an der Strichartigkeit nicht, dass Lippen zusammengekniffen sind, unabhängig davon, dass ein Zusammenkneifen wiederum Ausdruck von Freude sein kann (man verkneift sich das Lachen). Usw. Begriffsbildungen wie „(nicht)natürliche Eigenschaft“ oder „Minimalbeschreibung“ lenken die Aufmerksamkeit also in eine falsche Richtung. Und das Bestehen von Unterschieden zu erklären, setzt natürlich voraus, dass die Unterschiede selbst klar sind. Worin genau aber ein „interner Ausdruck“ besteht, ist damit noch nicht geklärt. Aber erinnern wir uns des Unterschiedes oben zwischen dem, dass ein böses Gesicht an sich böse ist, und dem, dass es als solches böse ist. Hier knüpft an, was Wittgenstein einen „springende(n) Punkt“ nennt: ob sich tatsächlich, „mein Charakter nach der psychophysischen Auffassung nur im Bau meines Körpers oder meines Gehirns und nicht ebenso im Bau der ganzen übrigen Welt ausdrückt“, so dass der „Parallelismus ... also eigentlich zwischen meinem Geiste, i.e. dem Geist, und der Welt (besteht).“ – Der Übergang von „meinem Geiste“ zu „dem Geist“ soll dem geschuldet sei, dass „der Geist der Schlange, des Löwen ... (mein) Geist“ ist. Denn: „nur von ... (mir) her ... (kenne ich) überhaupt den Geist.“ Der psychophysische Parallelismus würde nun antworten: „Wenn ich so aussähe wie die Schlange und das täte, was sie tut, so wäre ich so und so.“ – Aber das ist ein ganz seltsamer Schluss! Denn entweder es bleibt etwas übrig, wenn man mein Aussehen und mein Verhalten von dem abzieht, was (oder: wer?, siehe unten) ich bin, oder es bleibt nichts übrig. Wenn nichts übrig bleibt, was hieße es dann, dass ich so aussehe und mich verhalte wie die Schlange, wenn nicht, dass ich die Schlange bin. Es bedeutet als ent-

450

Richard Raatzsch

weder gar nichts, zu sagen, dass ich so aussehe und mich so verhalte wie die Schlange, oder dies ist eine Kontradiktion, und dann bedeutet es wieder nichts. Wenn ich aber mehr (oder gar anderes) bin als mein Aussehen und mein Verhalten, warum sollte ich dann so sein wie die Schlange, wenn ich so aussehe und mich so verhalte wie sie? Was aber, wovon Wittgenstein glaubt, dass es gewiss so sei, wenn „mein Körper mit dem der Wespe und der Schlange auf einer Stufe steht“? Dann fällt die Klausel „Wenn ich so aussähe wie die Schlange und das täte, was sie tut“ als irrelevant weg. Die Klausel benennt, zumindest in der hier relevanten Hinsicht, keinen Unterschied mehr. Es gäbe hier kein Schließen mehr. Allerdings könnte man sagen, die Klausel deute in die richtige Richtung. Was der psychophysische Parallelismus für eine empirische Möglichkeit hält, erwiese sich sozusagen als eine metaphysische Tatsache. Das führt Wittgenstein zu der Frage: Ist das die Lösung des Rätsels, warum die Menschen immer glaubten, ein Geist sei der ganzen Welt gemein?

Denn wenn das die Lösung ist, dann ist dieser eine Geist „freilich auch den unbelebten Dingen gemeinsam“. Sonst stünden eben nicht alle Körper auf einer Stufe. Wenn ich von vornherein den Geist kenne, wenn ich meinen Geist kenne, dann schließe ich natürlich auch nicht von meinem Geist auf deinen oder den Geist der Schlange ... Denn dann rede ich eo ipso schon von deinem Geist oder dem der Schlange ..., wenn ich von meinem rede. Darum kann ich hier nicht fälschlich identifizieren, identifiziere also gar nicht oder nur scheinbar. Den Geist haben wir alle nur in dem Sinn, in dem die Tatsache, dass etwas von mir gehabt wird, für das, was da gehabt wird, unwesentlich ist. Die Frage, wer den Geist hat, wenn ihn jemand hat, ist belanglos. Genau das aber legt auch die Sprechweise von der Teilhabe am Geist nahe. Es kann sein, dass es den Geist nicht gibt, wenn niemand an ihm teilhat; aber es reicht, dass irgendwer an ihm teilhat. Für jemanden, dem die Mystik fremd ist, müssen diese Überlegungen unverständlich oder atemberaubend sein (cf. Silesius 1979, Weininger 1903, Zemach 1964, Raatzsch 2001). Denn soweit es in Bezug auf meine Identität auf meinen Geist ankommt – und würde man nicht zunächst sagen wollen, dass es für die Frage, was (ein Mensch, und kein Tier, keine Pflanze, Zahl, Radnabe …) und wer (R.R. und nicht P.P.) ich bin, genau darauf ankommt? – also soweit es auf meinen Geist ankommt, bin ich Eins mit den Andern, am Ende mit allem.

Zu einigen Bemerkungen Wittgensteins über die Seele

451

Wittgenstein fasst seinen Gedankenweg in folgender Weise zusammen: Der Idealismus scheidet aus der Welt als unik die Menschen aus, der Solipsismus scheidet mich allein aus, und endlich sehe ich, dass auch ich zur übrigen Welt gehöre, auf der einen Seite bleibt also nichts übrig, auf der anderen als unik die Welt. So führt der Idealismus streng durchdacht zum Realismus. (Siehe auch: A 5.64)

Gewiss, wenn man beim Idealismus anfängt und dann, mittels strengen Durchdenkens desselben, bei seinem Gegenteil, dem Realismus, ankommt, muss natürlich etwas schief sein an der Sache. (Zur Erinnerung, es gibt nur logische Schlüsse.) Wichtig ist hier aber für uns vor allem, dass der Verlauf des Gedankengangs in gewisser Weise auch in seinem Ergebnis erhalten bleibt. Zwar kann ich „in diesem Sinne … von einem der ganzen Welt gemeinsamen Willen sprechen“, aber „dieser Wille ist in einem höheren Sinne mein Wille.“ Es handelt sich dabei um einen höheren Sinn, insofern das strenge Durchdenken des Idealismus mir, wie man zunächst sagen könnte, eine besondere Stellung unter den Menschen einräumt. Anders gesagt, indem der Solipsismus streng durchdachter Idealismus ist, ist die Idee der Menschen als unik der Welt gegenüberstehend mangelhaft. Der Idealismus übersieht eben, dass ich nur von mir her überhaupt den Geist kenne (s. o.). Diesen Mangel des Idealismus bringt der Solipsismus ans Licht. Allerdings leidet er an einem ähnlichen Mangel: er wird der Tatsache nicht gerecht, dass mein Körper nicht in der Weise von anderen Körpern verschieden sein kann, auf die er verschieden sein müsste, damit man sagen dürfte, was der Solipsist sagen will. (Hier hat der Idealist also einen Punkt.) Die Überwindung dieses Mangels besteht daher darin, dass ich „endlich“ sehe, dass auch ich zur übrigen Welt gehöre. Meine Sonderstellung unter den Menschen erweist sich als vermeintliche, wenn auch nicht im herkömmlichen Sinn. Im Grunde hätte man die Gleichstellung zwar sofort sehen müssen, denn das Ich ist ja nicht eines, wenn auch ein hervorgehobenes, unter vielen. Und von meinem Ich im Unterschied etwa zu deinem zu reden, würde das Problem nur wieder verschleiern oder verschieben. (Was wäre hier mit „mein“ und „dein“ gemeint?) Statt dessen wird Wittgenstein dem, was wir hier sagen wollen, durch die Rede vom philosophischen Ich, dem metaphysischen Subjekt, als Grenze der Welt gerecht. Denn wie zuvor ist auch hier die logische Multiplizität dieselbe. Es sind zwei Dinge im Spiel: wie vorher mein Ich und dein Ich, so jetzt Ich und Welt. Es wäre ungereimt, zu sagen, dass es eigentlich um (mindestens) drei Dinge ginge: mein Ich, dein

452

Richard Raatzsch

Ich und die Welt. (Schon dass man hier „mein“ und „dein“ verwenden muss, zeigt das Problem.) Dies wäre eine schiefe Klassifizierung. In dem Moment, in dem von mir und der Welt die Rede wäre, gäbe es auch schon keinen Platz mehr für Dich als etwas für sich Seiendem. (Wenn ich von mir, dir und der Welt rede, dann ja nicht so, wie ich von meinem Haus, meinem Auto und meinem Boot rede. Zum Beispiel kann die Welt nicht einfach wegfallen und alles andere bleiben, wie es ist.) Du wärst, soweit überhaupt etwas, ja schon Teil der Welt. Nur bin ich, in dem Sinn, in dem du Teil der Welt bist, selbst Teil der Welt – oder eben ihre Grenze. Nicht: eine ihrer Grenzen, denn in diesem („höheren“) Sinn, in dem jetzt von der Grenze der Welt die Rede ist, hat die Welt nur eine Grenze; und soweit sie mehrere hat, sind dies alle meine Grenzen. Die logische Multiplizität ist also in jedem Fall die einer Zweiheit – oder, wenn man so will, einer Einheit, die in sich zweifach bestimmt ist: sei es, indem die Menschen („als unik“) der Welt gegenüber gestellt werden, sei es, dass wir von Grenze und Begrenztem reden. Jener „höhere Sinn“ nun verdankt sich dem, dass es, wie man sagen könnte, an sich gleichgültig ist, ob von mir oder dir oder einem Stein die Rede ist, aber nicht für mich: „Ich habe die Welt zu beurteilen, die Dinge zu messen.“ (Vgl. TB 2.9.16). Diese Messung der Dinge ist nicht die gewöhnliche. In welcher Hinsicht denn, nach welchem Maßstab? Sondern diese Messung ist, wie auch jene Beurteilung, die/eine philosophische Messung resp. Beurteilung. Das heißt, es geht nicht darum, wie die Dinge sind, oder wie es um die Welt steht, sondern um das Wesen der Dinge und der Welt. Indem Messung und Beurteilung aber nicht die gewöhnliche Messung und Beurteilung sind, ist auch derjenige, welcher misst und beurteilt nicht wie gewöhnlich aufzufassen, sondern philosophisch. Das bedeutet, „(d)as philosophische Ich ist nicht der Mensch, nicht der menschliche Körper oder die menschliche Seele mit den psychologischen Eigenschaften, sondern das metaphysische Subjekt, die Grenze (nicht ein Teil) der Welt.“ (TB 2.9.16; vgl. auch § 29 (S. 237f.) von Schopenhauer 1818.) Das lässt den Menschen, seinen Körper, die menschliche Seele usw. nicht verschwinden; es macht sie vielmehr gleich, eben zu „Teil(en) der Welt unter anderen Teilen der Welt.“ (Ebd.) Sie sind da, sind auch verschieden – aber in einem höheren Sinn zählt der Unterschied nicht. Ich könnte also sagen, dass ich zwar glaube, dass er eine Seele hat, aber dass dabei von mir in einem niederen Sinn die Rede ist, nicht philosophisch.

Zu einigen Bemerkungen Wittgensteins über die Seele

453

Wittgenstein meint nun, derjenige, der das verstanden hat, wird seinem „Körper oder dem menschlichen Körper nicht eine bevorzugte Stelle in der Welt einräumen wollen“, sondern „Menschen und Tiere ganz naiv als ähnliche und zusammengehörige Dinge betrachten.“ (Ebd.) Aber wir können auch, an die Mystiker denkend, anders herum, und damit in richtiger Ordnung, sagen: wer seinem Körper keine bevorzugte Stellung in der Welt einräumt und Menschen und Tiere ganz naiv als ähnliche und zusammengehörige Dinge betrachtet, braucht die Logisch-philosophische Abhandlung nicht. Das ist die richtige Ordnung, weil, wenn man seinem Körper usw. ganz naiv keine bevorzugte Stellung einräumt, es einem genau an dem fehlen wird, was man „die philosophische Einstellung“ nennen könnte. Denn dass man sie einnimmt, bedeutet nicht einfach, dass man aufhört, naiv zu sein. Sondern es bedeutet zunächst, dass es einen Impuls zum Philosophieren gibt, der die Naivität aufhebt. Denn als ungetrübt, als rein, führt die Naivität nicht über sich hinaus. Man darf, anderes gesagt, mit sich und der Welt nicht völlig im Reinen sein. Der Ursprung eines Philosophierens, wie es der frühe Wittgenstein exemplarisch vorführt, ist also eine gewisse Entfremdung eines Einzelnen der Welt gegenüber. Diese Entfremdung ist insofern praktischer Art, als sie darin besteht, sich nicht so in die Welt einzufügen, dass es nicht zu dem kommt, was man, wenn man es hat, „philosophisches Problem“ nennt. So gesehen, ist die Philosophie eine Krankheit, und schon die Abhandlung, nicht erst die Untersuchungen, sollen ihre Therapie sein. 6. Philosophische Untersuchungen Dass die Therapie in beiden Fällen verschieden ausfällt, ist offensichtlich. Dies hängt, wie ebenfalls unstrittig ist, mit der Verschiedenheit der Diagnosen zusammen. Beginnen wir mit einem Punkt, der für mich von einiger Bedeutung ist. Zwar heißt es in der Abhandlung: (In der Philosophie führt die Frage „wozu gebrauchen wir eigentlich jenes Wort, jenen Satz“ immer wieder zu wertvollen Einsichten.) (6.211)

Aber in gewissem Sinne wird hier genau das Gegenteil von dem gesagt, was die Untersuchungen kennzeichnet, auch wenn dort manches ganz ähnlich klingt. Denn sehen wir uns die Sequenz an, zu der diese Bemerkung gehört:

454

Richard Raatzsch

6.2

Die Mathematik ist eine logische Methode. Die Sätze der Mathematik sind Gleichungen, also Scheinsätze. 6.21 Der Satz der Mathematik drückt keinen Gedanken aus. 6.211 Im Leben ist es ja nie der mathematische Satz, den wir brauchen, sondern wir benützen den mathematischen Satz nur, um aus Sätzen, welche nicht der Mathematik angehören, auf andere zu schließen, welche gleichfalls nicht der Mathematik angehören. Man sieht hier, inwiefern jene wertvollen Einsichten, zu denen die Frage „wozu gebrauchen wir eigentlich jenes Wort, jenen Satz“ in der Philosophie immer wieder führen, ihr Maß nicht an sich selbst haben. Denn selbst wenn uns die Frage, wozu wir den mathematischen Satz gebrauchen, zeigt, dass wir ihn nie im Leben brauchen, ist diese Einsicht deshalb wertvoll, weil sie uns auf die Idee führt, dass eigentliche Sätze Gedanken ausdrücken, Bilder der Tatsachen sind usw. Das heißt die Frage nach dem Gebrauch ist in der Philosophie heuristisch wertvoll, nicht an sich. Sie liefert sozusagen Indizien, keine Kriterien. Denn wenn wir davon ausgehen, wie wir das Wort „Satz“ gebrauchen, dann ist ein mathematischer Satz ... eben entweder ein Satz, wie auch der Satz „Ich werde um 5 Uhr da sein“ oder er ist einfach das, was er ist. Der erste Fall bedeutet nicht, dass zwischen beiden Sätzen nicht Unterschiede bestehen könnten, die es erlauben würden, sie verschiedenen Formen von Sätzen zu zuordnen. Aber von diesem Schritt hin zu dem, den mathematischen Satz zu einem Scheinsatz zu erklären, ist es noch einmal ein Schritt – und dieser Schritt wird einem von keiner Antwort auf die Frage aufgedrängt, wozu man einen Satz eigentlich gebraucht. Soweit wir diese Frage stellen, finden wir nicht Höheres, aber soweit diese Frage entscheidend ist, finden wir freilich auch nichts Niederes. Dieser Unterschied fällt weg. Insofern ist Wittgensteins spätere Philosophie dann konsequenter als seine frühere, wenn die Frage, wie man nun dazu kommt, eine Form des Satzes zum Muster für alle Formen zu machen, dadurch zu beantworten ist, dass man sich von bestimmten Satzformen beeindrucken lässt. (Dann ist es, nebenbei gesagt, auch nicht erstaunlich, dass Argumente in der Philosophie oft nicht wirklich oder endgültig überzeugen.) Denn als Formen stehen alle Satzformen ja auf gleicher Stufe. Das aber bedeutet, dass die Abhandlung weniger fehlerhaft ist, als vielmehr verzerrt, oder eben fehlerhaft im Sinne von verzerrt.

Zu einigen Bemerkungen Wittgensteins über die Seele

455

Wie auch immer, ein Punkt an dem oben geschilderten frühen Gedankengang fällt aus der Sicht der Philosophischen Untersuchungen besonders leicht auf: dass ich „nur von ... (mir) her ... überhaupt den Geist (kenne)“. Zunächst könnte man dem entgegenhalten wollen, dass ich gerade dann, wenn ich überhaupt von einem anderen Geist reden darf, sei es der Geist der Schlange oder dein Geist, auch schon nicht mehr sagen kann, dass ich nur von mir her „den Geist“ (meine Hervorhebung, R. R.) kenne. Denn dann würde ich eben auch nur meinen Geist kennen. Wie aber könnte ich diesen dann mit etwas identifizieren, was die Schlange oder du haben? – Ja, wie wir vom Käfer in der Schachtel wissen, führt dies zu der Frage: Woher weiß ich, dass sie, und auch ich, überhaupt etwas haben? (Siehe PU 293; auch das folgende Zitat von dort; zum Ganzen siehe auch: Philipp 1998.) So gesehen, besteht das Problem im frühen Tagebuch nicht einfach darin, „den einen Fall in so unverantwortlicher Weise (zu) verallgemeinern“, sondern darin, von einem Gegenstand der Bezeichnung zu sprechen, den man angeblich nur vom eigenen Fall her kennen kann. Wenn dieser Teil des Argumentes wegfällt, dann können in den Untersuchungen bestimmte Themen keine Rolle mehr spielen. Dass „(i)ch ... die Welt zu beurteilen, die Dinge zu messen (habe)“ – dies ist nun nicht mehr eine Aussage, die uns, streng durchdacht, zum Subjekt als Grenze der Welt führt. Was aber wird dann aus jener Lösung – des Rätsels, „warum die Menschen immer glaubten, ein Geist sei der ganzen Welt gemein?“ In seinen Tagebüchern liefert Wittgenstein so etwas wie eine Erklärung, warum die Menschen diesen Glauben hatten. Es war nur „so etwas wie eine Erklärung“, und nicht einfach „eine Erklärung“, weil es, wie wir wissen, Unsinn sein soll. Was genau es heißt, dass etwas Unsinn ist, wenn die Erklärung, dass es Unsinn ist, in dem Text steht, der es zu Unsinn erklärt – wenn es also im radikalsten Fall Unsinn ist, dass es Unsinn ist – ist gewiss nicht leicht zu erklären. Für unsern Zweck reicht es hoffentlich, zu sehen, dass auch der späte Wittgenstein so etwas wie eine Erklärung liefert für den Glauben an den Weltgeist, wenn man so will: für den populären Pantheismus, die nicht, oder nicht in gleicher Weise, der Frage nach ihrem Status ausgesetzt ist wie die frühe Erklärung. Das heißt, ihr Status ist ein anderer. Sie ist aber auch keine psychologische Erklärung. Man könnte sie „eine ästhetische Erklärung“ nennen, denn sie besteht in der Betrachtung des Gebrauchs der Sprache derart, dass man sieht, wie bestimmte ihrer Formen es nahe legen, sie misszuverstehen. Der Witz dieser Betrachtung

456

Richard Raatzsch

ist dann natürlich, dass, indem man dies sieht, das Missverständnis auch schon aufgehoben ist – ähnlich dem, dass man ein Vexierbild auflöst. Wie der Käfer in der Schachtel ebenfalls zeigt, ist dies natürlich nicht die einzige Methode. Denn der Käfer ist ein Muster einer reductio ad absurdum. Bei Wittgenstein wird diese Argumentform jedoch komplementiert durch jene ästhetische Erklärung. So wird der Eindruck, den eine sprachliche Tatsache macht, aufgeweicht. In unserm Fall bedeutet dies, dass wir nicht nur sehen, dass die Idee, man kenne den Geist nur vom eigenen Fall, streng durchdacht, sich selbst aufhebt, sondern auch, welche Rolle ein Satz wie „Ich muss ja wohl wissen, ob ich Schmerzen habe“ tatsächlich spielt. Das geht damit einher, dass wir sehen: Es kann nicht, ein einziges Mal nur, eine Mitteilung gemacht, ein Befehl gegeben, oder verstanden worden sein, etc. – Einer Regel folgen, eine Mitteilung machen, einen Befehl geben, eine Schachpartie spielen sind Gepflogenheiten (Gebräuche, Institutionen). (PU 199)

Es gibt keine „Gebräuche, Institutionen“, ohne dass sich Menschen so und so verhalten. Aber es muss nicht gerade ich sein, der sich so verhält. Für sich betrachtet, findet jeder Einzelne sein Mass an der Gruppe, auch wenn die Gruppe keine selbständige Existenz hat oder, besser, nicht zusätzlich zu, oder neben, dem existiert, was Einzelne tun. Insofern gibt es die Seele, oder den Geist, beim „vereinzelten Einzelnen“, um einen Ausdruck von Marx zu verwenden, nicht oder nur als Ausnahme von der Regel. Man sieht dies bei Wittgenstein sehr schön dort, wo er trotzdem von einer einzelnen Seele spricht, wie etwa im Braunen Buch, in dem es (S. 215) in Bezug auf eine Idee, welche einen philosophisch umtreibt, heißt, sie bestünde darin, „dass Du in dem geheimnisvollen Vorgang des Meinens, der Intention, alle Übergänge irgendwie schon gemacht hast, ohne sie wirklich zu machen. Deine Seele fliegt gleichsam voran und macht die Übergänge, ehe Dein Körper noch dort angelangt ist.“ Was dem Anderen hier in den Mund gelegt wird, ist natürlich nichts Biographisches, oder jedenfalls nichts, was ihn im Unterschied zu anderen kennzeichnet. Es kennzeichnet uns alle, aber wiederum nicht so, wie es uns alle kennzeichnet, dass wir zu so und so viel Prozent aus Wasser bestehen. Die Seele ist nicht in uns, aber auch nicht außer uns. Insofern hat jeder, der eine Seele hat, diese Seele insofern, als er Teil hat an der Seele, die nur dadurch gegeben ist, dass welche an ihr Teil haben.

Zu einigen Bemerkungen Wittgensteins über die Seele

457

Ist diese Seele nun auch allem Unbelebten gemein? – Nun, wir unterliegen zwar einer „Täuschung“, wenn uns „(d)as Denken, die Sprache, ... als das einzigartige Korrelat, Bild, der Welt (erscheint)“ (PU 96), wenn für uns das Denken mit einem „Nimbus“ umgeben ist, wenn wir sein „Wesen, die Logik“ für eine Darstellung der „Ordnung a priori der Welt, d.i. ... (der) Ordnung der Möglichkeiten“ halten, als vor aller Erfahrung liegend und sich durch alle Erfahrung hindurchziehend usw. (PU 97) Richtig an der Idee oder dem Bild einer Harmonie von Gedanke und Wirklichkeit ist dagegen, dass, ... ... wenn ich fälschlich sage, etwas sei rot, es doch immerhin nicht rot ist (also nicht etwa: nicht rund, widerspenstig oder was auch immer, R.R.) Und wenn ich jemandem das Wort „rot“ im Satze „Das ist nicht rot“ erklären will, ich dazu auf etwas Rotes zeige. (PU 429)

Es gibt also eine Verbindung, aber diese ist keine Übereinstimmung a priori: 430. „Lege einen Maßstab an diesen Körper an; er sagt nicht, dass der Körper so lang ist. Vielmehr ist er an sich – ich möchte sagen – tot, und leistet nichts von dem, was der Gedanke leistet.“ – Es ist, als hätten wir uns eingebildet, das Wesentliche am lebenden Menschen sei die äußere Gestalt, und hätten nun einen Holzblock von dieser Gestalt hergestellt und sähen mit Beschämung den toten Klotz, der auch keine Ähnlichkeit mit einem Lebewesen hat.

Genau genommen, stimmt das Letzte nicht: der Holzblock hat, der Voraussetzung nach, eine ähnliche Gestalt wie ein Mensch. Hier findet also eine begriffliche Bewegung statt: das Wesentliche am lebenden Menschen ist die Gestalt, sagen wir, dann stellen wir eine Figur her – und jetzt sagen wir: der Klotz (nicht: die Figur!) habe aber auch gar keine Ähnlichkeit. Deshalb ist es wichtig, dass Wittgenstein sagt: wir sehen es mit Beschämung, und nicht nur: wir sehen es. So landen wir schließlich bei dem Gedanken: das einzig Wesentliche am lebenden Menschen sei der Geist, die Seele, sein Denken und Wollen usw.! Und das ist auch in Ordnung so, wenn es nicht mehr heißen soll, als dass wir, wenn wir, fälschlich oder nicht, zum Beispiel sagen, etwas sei rot, dies gewöhnlich ein bedachtes Tun ist, nicht sozusagen automatisch geschieht. Literatur (mit Siglen) Anscombe, G. E. M., 1985: Has mankind one soul – an angel distributed through many bodies? In: Human Life, Action and Ethics, hg. v. M. Geach und L. Gormally, St. Andrews: Imprint Academic 2005, 17-25

458

Richard Raatzsch

Fleming, N., 1998: Seeing the Soul, Philosophy 53, 1978, S. 33-50; zitiert nach der deutschen Übersetzung J. Schultes: Blick auf die Seele in: Ludwig Wittgenstein. Philosophische Untersuchungen, hg. v. E. v. Savigny, Berlin: Akademie Verlag, 239-268 Lichtenberg, G. Chr., 1971: Schriften und Briefe. Zweiter Band. Sudelbücher II. Materialhefte, Tagebücher, München: Hanser Moore, G. E., 1903, Principia ethica, Cambridge: Cambridge University Press Philipp, P., 1998: Philosophical Investigations 293: Private vs. Public Beetles, in: ders., Logisch-philosophische Untersuchungen, Berlin und New York: de Gruyter, 393-404 Raatzsch, R., 2001: Wittgenstein as a Mystic, in: R. Haller und K. Puhl (Hg.), Wittgenstein und die Zukunft der Philosophie. Akten des 24. Internationalen Wittgenstein-Symposiums Kirchberg a. W./Österreich 2001, Hölder-PichlerTempsky: Wien 2002, 366-379 Savigny, E. v., 1994, Wittgensteins “Philosophische Untersuchungen“. Ein Kommentar für Leser, Band 1: Abschnitte 1-315, Frankfurt a.M.: Klostermann Schopenhauer, A., 1818/1844: Die Welt als Wille und Vorstellung, Leipzig: Brockhaus, 1818 (Band I) und 1844 (Band II), zitiert nach der Ausgabe von W. Frhr. V. Löhneysen, Insel: Leipzig und Frankfurt a. M. 1996 Silesius, A., 1979: Der Cherubinische Wandersmann, Geistreiche Sinn- und Schlußreime, ed. E. Brock, Zürich: Diogenes Weininger, O., 1904: Die Kultur und ihr Verhältnis zu Glauben, Fürchten und Wissen, in: ders., Über die letzten Dinge, München: Matthes & Seitz: München 1997, S. 141-182, Original: Wien 1904 Wittgenstein, L., 1935/36: Eine Philosophische Betrachtung (Das braune Buch), in: ders., Werkausgabe Band 5, Frankfurt a. M.: Suhrkamp 1984, S. 117-282; zitiert mit Titel und Seitenzahl. Wittgenstein, L., 1918: Logisch-Philosophische Abhandlung. Tractatus Logico-philosophicus. Kritische Edition, hg. v. B. McGuinness und J. Schulte, Frankfurt a. M.: Suhrkamp 1998, zitiert mit Sigle „A“ und Nummer des Satzes Wittgenstein, L., 1953: Philosophische Untersuchungen. Kritisch-genetische Edition, hg. v. J. Schulte, in Zusammenarbeit mit H. Nyman, E. v. Savigny und G.H. v. Wright, Frankfurt a. M.: Suhrkamp 2001; auf Teil I erfolgt die Bezugnahme mit Sigle „PU“ und der Nummer der Bemerkung resp. Seitenzahl (für Teil II) Wittgenstein, L., 2003: Public and Private Occasions, hg. v. J.C. Klagge und A. Nordmann, Lanham: Rowman & Littlefield, zitiert mit Sigle „PPO“ und Seitenzahl Wittgenstein, L. 1914-16: Tagebücher 1914-1916, in: ders., Werke, Band 1, Frankfurt a. M.: Suhrkamp 1984, 87-187, zitiert mit Datum der Eintragung Zemach, E., 1964: Wittgenstein’s Philosophy of the Mystical, in: Review of Metaphysics, 18 (1964), 38-57

Conjecture, Proof, and Sense in Wittgenstein’s Philosophy of Mathematics SEVERIN SCHROEDER, Reading

Abstract One of the key tenets in Wittgenstein’s philosophy of mathematics is that a mathematical proposition gets its meaning from its proof. This seems to have the paradoxical consequence that a mathematical conjecture has no meaning, or at least not the same meaning that it will have once a proof has been found. Hence, it would appear that a conjecture can never be proven true: for what is proven true must ipso facto be a different proposition from what was only conjectured. Moreover, it would appear impossible that the same mathematical proposition be proven in different ways. – I will consider some of Wittgenstein’s remarks on these issues, and attempt to reconstruct his position in a way that makes it appear less paradoxical. 1. The most prominent idea in Wittgenstein’s philosophy of mathematics is that mathematics is a part of syntax, or grammar. Mathematical propositions are rules of grammar (PR 130, 143, 216-217; RFM 99, 163). This immediately provokes the question, first asked by Georg Kreisel (in his notoriously unsympathetic review of Wittgenstein’s Remarks on the Foundations of Mathematics): why in that case proofs should be needed, ‘since a rule of language, as ordinarily understood, is a matter of simple decision’ (Kreisel 1959, 140). – Well, this is a bit like asking: ‘If tennis is a ball game, then why can’t one score any goals?’ The answer is: It’s not analytic that ball games involve goals. It may be that the most popular ball games do; but there are others that don’t. Similarly, it’s not analytic that a grammatical rule must be arbitrarily chosen. Perhaps most of them are; but, on Wittgenstein’s account, there is at least one kind of grammatical rule (in what we call mathematics) that is not arbitrarily chosen, but introduced by proof. 2. It is of course true that elsewhere Wittgenstein calls grammar ‘arbitrary’ (PG 184; cf. PI §372). But what he means by that is that it’s essentially a

460

Severin Schroeder

human artefact that cannot be assessed as true or false to nature.1 That, however, does not mean that anybody will be allowed to introduce any kind of grammatical rule as the whim takes him. There may well be certain grammatical rules which – although we make them – we make only according to certain rule-governed procedures, which we call proofs. 3. Mathematics is a part of our grammar that essentially involves proofs or calculations. We can of course imagine grammatical rules about amounts or numbers without any calculations; but we would hardly call them ‘mathematics’. Imagine a tribe of people who can count, but do not calculate. Then one day their high priest introduces the rule that 57 things and 33 things are 90 things altogether. (Perhaps these are quantities that occur frequently in their lives, so that this unique addition rule is useful to them.) Is that a piece of mathematics? – Not really. It may acquire the status of a grammatical rule about quantities and in that respect be very much like our mathematical equations. But there is a crucial difference: It is not the result of a calculation or proof according to certain rules. Yet it is an essential feature of mathematics that its sentences are introduced and justified in a certain way. This is part of their grammar or use (just as it is part of the grammar or use of religious statements that they are not introduced and treated like scientific hypotheses). Mathematical sentences, if they are not definitions giving meaning to new symbols, are neither verified inductively, by repeated tests, nor adopted on authority. They have to be deduced by proof or calculation. That is, they are second-order grammatical rules: produced by applying other grammatical rules. – Hence, without an arithmetical calculus, the equation ‘57 + 33 = 90’ would not be a mathematical proposition.

1

NB: Even though grammatical rules as such are not true or false, they may well be based on, or inspired by, factual observations. Thus Wittgenstein holds that mathematical propositions are empirical observations ‘hardened into a rule’ (RFM 324). Of course it is a fact (and not a convention) that when we add two apples to two apples we almost always count four in all. Yet, the mathematical proposition ‘2 + 2 = 4’ is not just a generalisation of such empirical propositions, but has acquired the status of a grammatical rule, which is no longer exposed to empirical confirmation or disconfirmation. (Cf. Steiner 2000)

Conjecture, Proof, and Sense in Wittgenstein’s Philosophy of Mathematics

461

4. Note that the idea of a practice with second-order rules that are not stipulated from the beginning, but can be, or have to be, introduced later according to certain rule-governed procedures is not that unheard of. A common example is a legal system, which consists not only of a set of laws, but also of procedural rules for legally introducing new laws. 5. So, according to Wittgenstein, mathematical propositions have two key features: (i) They are given the status of grammatical rules, norms of representation. (ii) Definitions (and axioms) apart, they have to be legitimised by proof. 6. If mathematical truth means that something has the status of a grammatical rule, then it would appear that there cannot be any unknown mathematical truths. For the idea of a grammatical rule that nobody knows to be valid is nonsense: ‘A law I’m unaware of isn’t a law’ (PR 176). For something to have a normative function it must be known to have a normative function; it cannot fulfil this function secretly. 7. However, this needs to be qualified. A system of rules may be so complicated that it’s not always immediately clear whether a given move is correct or not. Legal systems are often so complicated that it takes an expert lawyer to ascertain how a given legal question is to be decided: what in this case would be in accordance with the law. However, for the legal system to function as such it must generally be possible, at least in principle, to work out what is and what is not in accordance with the law. Similarly, where a system of grammatical rules (e.g. for the use of addition, multiplication &c.) allows for the generation of second-order grammatical rules (e.g. equations), it is only to be expected that their correctness may not always be self-evident. The crucial point is that it must always be possible to work out, to ascertain according to accepted rules, whether a given proposition is a grammatical rule or not. 8. If a given proposition is to be regarded as part of a system of norms, it must be such that we know how to check its status. Hence, Wittgenstein says of mathematical propositions: ‘Every proposition is the instruction [Anweisung] for a verification’ (PR 174).2 In other words, a mathematical 2

The published translation (‘signpost’) is inaccurate.

462

Severin Schroeder

proposition (such as an arithmetical equation) comes with certain methods of proof, which belong to – and determine – the proposition’s mathematical sense (PG 375; RFM 162; 295f.). An equation, such as ‘25  25 = 625’, is not an isolated proposition, but part of a calculus (PG 376b). Hence, the equation is not just a substitution rule (WVC 158; PG 347) – as even the isolated rule ‘57 + 33 = 90’ of our imaginary tribe above (§3) was – it also says that I get 625 if I apply the rules of multiplication to ‘25  25’. That is to say, the sign ‘=’ has essentially two aspects. It means both ‘can be replaced by’ and ‘yields according to the rules of arithmetic’ (PG 377-378). In this way an equation refers to the calculation from which it results, and its full analysis would spell out this implicit reference: it would provide the proof (PR 179). 9. Indeed, since the proof is just the fully analysed mathematical proposition, Wittgenstein even says that the proof is the actual mathematical proposition, of which ordinary mathematical propositions are just abbreviations (PR 192; 184; WVC 33; RFM 296). 10. However, there is not only elementary school mathematics, the calculus of basic arithmetic everybody learns at school to use in everyday life, there is also mathematics as a discipline of research (whose results may later find an application in the sciences). Mathematical research is a matter of extending existing calculi: developing new mathematical concepts that are (more or less) natural continuations of existing techniques. For instance, unrestricted use of subtraction leads to the introduction of negative numbers; unrestricted use of division leads to fractions. These are extensions of the calculus. In the original calculus of addition and subtraction with natural numbers, ‘5 – 7’ is meaningless; like an illegal move in chess. 11. So, we have two types of mathematics: (A) Everyday mathematics, which is static: calculations within a given calculus; (B) Mathematical research, which is concerned with problems that cannot be solved, methodically, within the existing calculus: the solution cannot simply be worked out using a given set of rules. Mathematical research is essentially dynamic: it does not just handle an existing calculus, but extends it. 12. In (B) mathematical research, we ask questions of which we do not yet know how to find their answer, and we advance conjectures, would-be mathematical propositions, for which we do not have a proof, nor a method

Conjecture, Proof, and Sense in Wittgenstein’s Philosophy of Mathematics

463

of producing a proof. But if the meaning of a mathematical proposition is determined by its proof, then it would appear that (P1) a mathematical conjecture has, for the time being, no determinate meaning. At any rate, a conjecture cannot have the same meaning that it will have once a proof has been found. Hence, it would appear that (P2) a conjecture can never be proven true: for what is proven true must ipso facto be a different proposition from what was only conjectured (PR 183, 191; RFM 366). Moreover, it would appear, very implausibly, that (P3) there cannot be more than one proof for a given mathematical proposition (PR 184b; WVC 109). 13. Of course Wittgenstein wouldn’t hold that mathematical conjectures are entirely meaningless; that they are just pieces of nonsense (PR 170). He only suggests that they have no mathematical sense: they are not as yet part of a mathematical calculus and hence should not be regarded as mathematical propositions proper. That is not to say that they are altogether meaningless. One suggestion is that their meaning lies in their heuristic value: they give us new impulses (PR 190; PG 359).3 A mathematical question is a challenge. And we might say: it makes sense, if it spurs us on to some mathematical activity … if it stimulates the mathematical imagination. [Z §§696-7; cf. WVC 144]

Mathematical problems invite or challenge us to find a certain kind of extension of our existing calculi, as illustrated by the following analogy: Imagine someone set himself the following problem. He is to discover a game played on a chessboard, in which each player is to have 8 pieces; the two white ones which are in the outermost files at the beginning of the game (the “consuls”) are to be given some special status by the rules so that they have a greater freedom of movement than the other pieces; one of the black pieces (the “general”) is to have a special status; a white piece takes a black one by being put in its place (and vice versa); the whole game is to have a certain analogy with the Punic wars. Those are the conditions that the game is to satisfy. – There is no doubt that that is a problem, a problem not at all like the problem of finding out how under certain conditions white can win in chess. – But now imagine the problem: “How can white win in 20 moves in the war-game whose rules we don’t yet know precisely?” – That

3

In Philosophical Remarks he toyed with the idea that even for new, difficult problems mathematicians already have a method of solving them, only not in writing, but in psychological symbols ‘in their heads’ (PR 176). But he soon discarded this idea, which is evidently implausible in many cases.

464

Severin Schroeder

problem would be quite analogous to the problems of mathematics (other than problems of calculation). [PG 363]

In other words, a serious mathematical problem is like a problem in a game whose rules haven’t been fixed yet! And the corresponding conjecture would be like the suggestion that in such a game in a certain position it would be possible for white to win in 20 moves. 14. Is that a plausible view? Probably not. The mathematical situation Wittgenstein seems to have in mind here is one where a question leads us to define new mathematical symbols, comparable to new pieces introduced in a board game (‘consuls’ and ‘general’). For example, for someone who has learnt only calculations with natural numbers the question ‘What is 5 – 7?’ has no answer. For an answer to become possible, he first needs to introduce new symbols in his calculus: negative numbers. – However, mathematical problems are not always like that. For instance, the question whether there is a greatest prime number is (or was) a real mathematical problem, not just a routine calculation. And yet its solution doesn’t require the definition of any new mathematical concepts or the introduction of any new mathematical techniques.4 It is much more like solving a problem in an existing game. After all, even the solution of a problem within a given framework of fixed rules may require a considerable amount of imagination. Moreover, it doesn’t seem right to say that the sense of a mathematical conjecture consists merely in its heuristic value: that ‘it trains your thought on a particular region’ of mathematics (PR 190). Wittgenstein himself, considering Fermat’s last theorem, came to have doubts about this view (RFM 314). After all, the concepts involved in this conjecture are all welldefined and used in what appears to be a straightforward and understandable way. So why should we not be able to attach a clear meaning to it? 4

Here’s Euclid’s reductio ad absurdum proof: Suppose Pn is the greatest prime, then multiply all the prime numbers up to Pn and add 1: (P1  P2  P3  …  Pn) + 1 The resulting number is either prime, in which case Pn isn’t the greatest prime after all; or, if it isn’t, it must be divisible by a prime number, but it isn’t divisible by any prime number up to Pn (for there will always be rest 1), so it would have to be divisible by a greater prime number, in which case, again, there is a greater prime number than Pn. So there cannot be any greatest prime number Pn.

Conjecture, Proof, and Sense in Wittgenstein’s Philosophy of Mathematics

465

15. In some places Wittgenstein suggests a different, and more persuasive, account of the meaning of mathematical conjectures (or problems). They are not merely heuristic stimulants; they can have a fairly clear sense – but it’s not a genuinely mathematical sense. In as much as we understand their content, we take them as empirical propositions, corresponding to, but crucially different from, the mathematical proposition we would like to establish by proof. For example, we are under the impression that we understand the idea of the construction of a heptagon with ruler and compass (which is impossible). But that is only because we have a clear empirical idea of a heptagon, that is, we can easily think of a 7-sided figure whose sides and angles when measured come out as all the same. So we are inclined to understand the problem as that of drawing such a figure. But in fact that is not the mathematical problem. The mathematical problem is that of finding a mathematical construction of a heptagon, analogous to the way one can give a mathematical construction of, say, a pentagon. The result of such a construction would of course also fulfil the empirical criteria (that measurement shows 7 sides and angles, all roughly equal), but that is not enough. As a solution to a geometrical problem, it is essential that the figure be arrived at, step by step, in a regular, repeatable and teachable way, using only ruler and compass. We are looking not just for a shape, but for a very specific way of producing it. Yet this specific way of producing such a shape is something we are unable to describe. We have no idea of such a geometrical construction; and therefore, our talk of such a construction – the conjecture of such a construction – has no clear mathematical sense; even though it has a very straightforward empirical sense, derived from empirical measurements of drawn figures (cf. LSP 572). 16. As another example, consider Goldbach’s conjecture that every even number is the sum of two primes. Don’t we understand that? – Again, Wittgenstein’s response is that without a proof we have of course some understanding of it, but only as an empirical generalisation; meaning that for any even numbers we will ever consider we will be able to find two primes adding up to that number. That is an empirical hypothesis inductively supported by our evidence to date; but not a mathematical proposition (cf. RFM 280f). Infinity in mathematics is always the endless applicability of a law (cf. PR 313-14; RFM 290b). Hence, where (as yet) we have no law, no mathe-

466

Severin Schroeder

matical rules that can be understood to have an endless applicability, we cannot meaningfully speak of mathematical infinity. So we cannot as yet make sense of the infinite scope of Goldbach’s conjecture. 17. What we’re looking for when trying to find a solution to a mathematical problem is a proof, which of course we cannot really describe before we’ve actually found it. For an accurate description of a proof is the proof itself. This brings out the difference between looking for some physical object and looking for something in mathematics: In the former case I can know exactly what I’m looking for. I can, for example, give you a perfectly accurate description of the spectacles I’ve mislaid. Not so in mathematics, where the thing looked for consists only in its accurate description (LSP 574). It follows that in mathematics we can never know exactly what we’ve been looking for till we have it. Hence a mathematical conjecture can never have a clear and precise mathematical sense, although, to be sure, we may be able to attach a fairly clear empirical sense to it, and some vague idea as to, roughly, what kind of proof strategy we expect might be successful. 18. Wittgenstein compares a mathematical proof to a jigsaw puzzle (MS 122, 49v). Indeed, sometimes he regards actual jigsaw puzzles as mathematical problems (RFM 55ff., LFM 53-5). In such a case, the conjecture to begin with would be something like: ‘These 200 pieces can be assembled to form a rectangular picture of a mountain’. Here it is obvious that the proof – putting all the pieces together in the right way – would do more than establish the truth of the conjecture. It would not only convince us that the pieces can be put together to form a picture of a mountain, it would show us how they fit together (cf. RFM 301; 308). Thus the proof does not only verify a proposition, one can say that it gives us a much fuller understanding of it, showing us what exactly that proposition means. 19. As explained above, mathematical propositions are characterised by a necessity that must be established by a demonstration. If that is correct, then the proposition that (1) there is no greatest prime should be rendered more appropriately as: (2)

There can’t be a greatest prime.

Indicating the necessity we attribute to a proposition when we take it as a piece of mathematics. Then, of course, the meaning of the modal verb in

Conjecture, Proof, and Sense in Wittgenstein’s Philosophy of Mathematics

467

(2) needs to be understood. One is entitled to ask: ‘What do you mean by “can’t”?’ And the answer that gives meaning to the ‘can’t’ is that it follows from such-&-such considerations – the proof – that there is no greatest prime. In this way, the mathematical proposition, when taken as such: as a demonstrably necessary truth, refers us to its proof (cf. RFM 309). 20. Finally, consider that, for all we know, a mathematical conjecture could be proven false (RFM 314d), that is, shown up to be inconsistent. Yet if something is inconsistent, or contradictory, it doesn’t make sense: it cannot be understood: there is nothing to be understood. But then, given that we cannot even know whether a mathematical conjecture is at all understandable (and not nonsense), then a fortiori we cannot claim to understand it. A sentence that as far as we know may be inconsistent, i.e. nonsense, can hardly be said to have a clear sense for us. This again leads to the conclusion that a conjecture, as such, has not as yet a clear mathematical sense. Or shall we say that it may well have a clear mathematical sense, only we don’t yet know it? – Well, as far as the contents of the proposition are concerned we could perhaps put it that way. (Wittgenstein recognizes that the expression ‘sense of a mathematical proposition’ is not sharply defined and can be construed in different ways (MS 122, 113).) After all, when at a later time a proof is found, it is accepted as a proof of that conjecture (e.g. Andrew Wiles’s proof became famous as a proof of Fermat’s Last Theorem). We take the proof to show that the conjecture had been correct, i.e. provable, and hence meaningful, all along. – However, another aspect of our concept of the sense of a linguistic expression is that it is the internal object of correct linguistic understanding. Normally, the sense of a linguistic expression is what a competent speaker understands by it. Hence, in a case where for the time being even the most competent speakers (mathematicians) do not fully understand a would-be mathematical proposition (nor know how to come by such an understanding), we can also with some propriety say that so far it has no clear sense for us (or anybody). 21. More importantly, for a proposition to have mathematical sense it must not only have the contents, but also the normative status that characterises mathematics (RFM 425): it must be acknowledged as a grammatical rule, which obviously an unproven conjecture is not. Nothing unknown can fulfil a normative function (PR 143; 176). Therefore, even if we assume that it

468

Severin Schroeder

is possible to find a proof for Goldbach’s conjecture – that the potential for such a proof is already there –, until it has actually been produced Goldbach’s conjecture will not be accorded the status of a grammatical rule. That is, until then it cannot be accorded the status of a mathematical proposition. Consider again the legal analogy mentioned above: A country has a complicated piecemeal system of laws, bye-laws, statutes and regulations. Moreover, there are lawyers that have developed an equally complex system of rules as to how apparent conflicts between different pieces of legislation are to be resolved. One day a lawyer produces a legal proof that the President is to give the Annual Opening Speech in Parliament. Henceforth it is regarded as a legal obligation that the President is to give the Annual Opening Speech in Parliament. Of course, if the proof is correct, then it could have been given before. In a sense, the laws haven’t changed; so implicitly they would have required the President to fulfil that duty all along. But then, as long as that implication wasn’t acknowledged as such, it could not have had any normative force: it did not have the status of a law. Similarly, a mathematical conjecture, even if provable all along, must first be proven, before it can attain the status of a mathematical proposition. 22. Above (§14) I rejected the idea that solving a mathematical problem (or proving a mathematical conjecture) always involves introducing new symbols and concepts (e.g. signed integers). Even a proof that makes use only of existing concepts can be extremely difficult to find. And now, I hope, it can be seen that the change of meaning from a conjecture to a theorem can be accounted for without the claim that a proof introduces new symbols and concepts. It follows from Wittgenstein’s key idea that mathematical propositions are grammatical rules. As such they cannot exist unbeknown. Hence even if (unlike Wittgenstein) we want to say that the connections developed by a proof were already implicit in the rules, they need to be drawn out and acknowledged before they can become part of our system of mathematics. 23. What about sums whose correctness we haven’t checked yet: would we not be forced to say that for all we know they may have no mathematical sense either? – That was not Wittgenstein’s view in the early 30s when he held that sums and homework questions are meaningful in virtue of the

Conjecture, Proof, and Sense in Wittgenstein’s Philosophy of Mathematics

469

well-known method of calculation by which we check or answer them (AL 197-8). Thus, even an incorrect sum is meaningful insofar as it reflects an understandable ‘mistake in calculation’ (AL 200). We can understand it as a failed attempt to apply a well-known method. – Some of Wittgenstein’s later remarks, however, point in a different direction (RFM 76-8). Although it is true that our familiar techniques of calculation give a straightforward use and meaning to any sum of the form ‘a  b = c’, in the case of a mistaken sum there is no genuinely mathematical content. A false sum, such as ‘16  16 = 169’, is meaningful insofar as it can be applied outside mathematics (probably leading to an empirical error), and insofar as it can be checked by a calculation. When I mistakenly believe that 16  16 = 169, I do not believe a mathematical proposition; rather, I mistakenly believe that ‘16  16 = 169’ is a mathematical proposition, a rule of our mathematical grammar – which it isn’t. It has no meaning in arithmetic, just as moving a pawn backwards is not a move in chess, not even a bad one (cf. RFM 78). So, we should indeed say that an incorrect sum has no mathematical sense, and hence, that an arithmetical equation we haven’t checked yet may turn out to have no mathematical sense, even though it has a straightforward use: Someone mistaking it for a mathematical proposition will apply it accordingly, and someone sceptical about its correctness will know how to check it. 24. Finally, let us turn to the third paradox (P3), namely the claim that since the meaning of a mathematical proposition is determined by its proof, two different proofs will yield two different propositions. In other words: a mathematical proposition cannot have two different proofs. Wittgenstein acknowledges that this is contrary to what we say in mathematics (RFM 189). In fact it is quite common for one mathematical proposition to have more than one proof. For instance, at least six different proofs can be cited for the theorem that there is an infinity of primes (Aigner & Ziegler 2001, ch. 1). 25. Let us note first that Wittgenstein’s main reason for insisting on a difference in sense between a conjecture and the corresponding theorem does not apply in this case. The point is that without an endorsement by proof a conjecture cannot have the normative status of a mathematical proposition. By contrast, two different proofs will still agree in conferring this norma-

470

Severin Schroeder

tive status on their conclusions. Consequently, where (what appears to be) the same mathematical proposition has been proven in two different ways, both proofs will confer on it the status of a grammatical rule and thus qualify it to be used in a certain way, especially for applications outside mathematics: That these proofs prove the same proposition means, e.g.: both demonstrate it as a suitable instrument for the same purpose. And the purpose is an allusion to something outside mathematics. [RFM 367]

26. The question to what extent mathematical proofs are to be regarded as constitutive of the sense of their conclusions may be compared with analogous verificationist issues outside mathematics. Given that meaning is use, it is very plausible to insist on an important difference in meaning between, on the one hand, propositions that are susceptible of, and intended for, empirical confirmation or disconfirmation, and, on the other hand, propositions that do not allow, or are never intended to be exposed to, any empirical testing (e.g. articles of faith, moral principles). Thus Wittgenstein emphasised that the most prominent confusion in the philosophy of religion was due to neglecting this crucial distinction: treating expressions of religious belief as if they were cosmological hypotheses (LC 53ff.). But Wittgenstein was inclined to go further. He would claim that even among the class of verifiable propositions differences in meaning resulted from different methods of verification. He held, for example, that statements of length had a different sense depending on whether they were to be established by using a tape measure or by astronomical calculations (RFM 146-7; BT 80v). Or again, the meaning of a proposition about the time of death would partly depend on the methods by which the time of death is ascertained (MS 122, 113v-114r). This further claim is, of course, very convincing where different methods introduce different criteria that can lead to different results in the same case. Thus, depending on whether a diagnosis of death is based on cessation of heart action or on failure of brain activity, the proposition that the patient died at 5.30pm may indeed have different meanings. It is, however, doubtful whether this line of reasoning can ever be applied to mathematics, where different and inconsistent definitions would hardly be tolerated. Thus, different proofs of the infinity of primes would surely not involve different and inconsistent definitions of a prime number.

Conjecture, Proof, and Sense in Wittgenstein’s Philosophy of Mathematics

471

27. Indeed even with regard to a mathematical proposition’s role inside mathematics, some of Wittgenstein’s later remarks suggest that different proofs need not be seen to give the proposition different meanings. The idea now is that a proof alone is anyway not sufficient to endow a mathematical proposition with meaning. Rather, its meaning depends on a whole system of rules and techniques, only some of which are made explicit in the proof: The proof of a proposition certainly does not mention, certainly does not describe, the whole system of calculation that stands behind the proposition and gives it its sense. [RFM 313; cf. 367d]

A proof locates a proposition in our system of mathematics; and there may be different proofs, that is, different ways of reaching the same position. Yet the proposition’s sense would be its position in the system, not the path by which it is reached: A proof of the proposition [that there are infinitely many prime numbers] locates it in the whole system of calculations. And its position therein can now be described in more than one way, as of course the whole complicated system in its background is presupposed. [RFM 313]

Nonetheless, a proof can be said to determine the sense of a proposition, since it determines its location in the system. Moreover, every new proof enriches our system of mathematics, by adding new conceptual links to it. Hence it could still be maintained that by finding a new (second or third) proof for an already established theorem we add to its meaning, or at any rate to our understanding of it. (That is especially plausible when after a non-constructive proof, a constructive one is found (RFM 282d; 308b).) 28. In conclusion, we need not attribute to Wittgenstein (P1) the paradoxical view that a mathematical conjecture has, for the time being, no clear meaning. Conjectures are surely meaningful; it’s only that they’re still waiting to be given a precise mathematical content and the normative status of a mathematical proposition. Does that mean that (P2) a conjecture can never be proven true: for what is proven true must ipso facto be a different proposition from what was only conjectured? Yes and no. In one important sense, what is proven is indeed a different proposition; but in another sense, it can obviously be recognised as having the same contents, albeit seen through a glass darkly. Finally, Wittgenstein’s considered view

472

Severin Schroeder

does not seem to commit him to (P3): on balance he would not insist that the same mathematical proposition cannot have more than one proof.5 References Ludwig Wittgenstein: AL

Wittgenstein’s Lectures, Cambridge, 1932-1935, ed. A. Ambrose, Oxford: Blackwell 1979

BT

The Big Typescript: TS 213, ed. and tr. C.G. Luckhardt and M.A.E. Aue, Oxford: Blackwell 2005

LC

Lectures and Conversations on Aesthetics, Psychology and Religious Belief, ed. C. Barrett, Oxford: Blackwell 1978

LFM

Wittgenstein’s Lectures on the Foundations of Mathematics Cambridge, 1939, ed. C. Diamond, Hassocks, Sussex: Harvester Press 1976

LSP

Logik, Sprache, Philosophie, by Friedrich Waismann [based on dictations by Wittgenstein], Stuttgart: Reclam 1976

MS

Manuscript in Wittgenstein’s Nachlass: The Bergen Electronic Edition, Oxford: Oxford University Press 2000

PG

Philosophical Grammar, ed. R. Rhees, tr. A.J.P. Kenny, Oxford: Blackwell 1974

PI

Philosophical Investigations, ed. P.M.S. Hacker and J. Schulte, tr. G.E.M. Anscombe P.M.S. Hacker, and J. Schulte, Oxford: Wiley-Blackwell 2009

PR

Philosophical Remarks, ed. R. Rhees, tr. R. Hargreaves and R. White, Oxford: Blackwell 1975

RFM

Remarks on the Foundations of Mathematics, ed. G.H. von Wright, R. Rhees, and G.E.M. Anscombe, tr. G.E.M. Anscombe, rev. ed., Oxford: Blackwell 1978

WVC

Ludwig Wittgenstein and the Vienna Circle, Conversations recorded by Friedrich Waismann, ed. B. McGuinness, tr. J. Schulte and B. McGuinness, Oxford: Blackwell 1979

Z

Zettel, ed. G.E.M. Anscombe and G.H. von Wright, tr. G.E.M. Anscombe, Oxford: Blackwell 1967

Aigner, M., and G. Ziegler, 2001: Proofs from the Book, Berlin: Springer

5

I am indebted to Hanoch Ben-Yami and David Dolby for comments on earlier versions of this paper.

Conjecture, Proof, and Sense in Wittgenstein’s Philosophy of Mathematics

473

Kreisel, G., 1959: Wittgenstein’s Remarks on the Foundations of Mathematics, in: British Journal of the Philosophy of Science 9, 135-158 Steiner, M., 2000: Mathematical Intuition and Physical Intuition in Wittgenstein’s Later Philosophy, in Synthese 125, 333-340

Through Pictures to Problems: Cognitive Epistemology and Therapeutic Philosophy EUGEN FISCHER, University of East Anglia

In his most extensive meta-philosophical discussion, in the Big Typescript (BT), and in contemporaneous and related sources, Wittgenstein moots two ideas about the genesis of philosophical problems, which have major consequences for philosophical method: ‘We encounter [philosophical problems] only when we are guided not by practical purpose in forming our sentences, but by certain analogies within language’ (BT 427). Many a ‘false analogy’ has been ‘accepted into language’ (BT 409). ‘To use [psychoanalysis’] way of putting things … [such] a simile [is] at work in the unconscious’ (VW 69). ‘A simile that has been absorbed into the forms of our language produces a false appearance, and this disquiets us. “But this isn’t how it is!” – we say. “Yet this is how it has to be!”’ (PI 112)1

That is: (i) In philosophical reflection, we are frequently guided by ‘analogies in language’, some of which are ‘false’, without being aware of being guided by anything of the sort. (ii) Where these analogies are ‘false’, conflicts arise between philosophical intuitions and familiar facts, and such conflicts give rise to a characteristic kind of philosophical problems. This paper provides a précis of the book Philosophical Delusion and its Therapy (Fischer 2011a, henceforward PDT) which develops these two ideas so as to vindicate a third. With the help of concepts and findings from cognitive linguistics and cognitive psychology, it turns the first idea – (i) – into a well-motivated empirical hypothesis. It develops the second idea – (ii) – through historical case studies on structure and genesis of some influential problems about the mind and perception. The book then shows, third, that these two ideas together vindicate and facilitate a comprehensive reorientation of philosophical work, namely, the practice of philosophy as a kind of (cognitive) therapy – which proceeds by strictly rational means and 1

The BT (TS 213) was compiled in 1933. VW 69 is from the ‘Dictation for Schlick’ probably dating from the same year. PI 112 derives from passages in MS157b (1937) which re-use material from the BT.

476

Eugen Fischer

fully respects the intellectual integrity and prowess of the thinkers addressed. This paper will roughly outline some of the main arguments for these three core points. The aim is not so much to engage in Wittgenstein exegesis – though one chapter of PDT is devoted to such exegesis – as to develop ideas mooted by Wittgenstein, with concepts and findings from cognitive and clinical psychology not yet at his disposal. 1. Analogies and Pictures Many fundamental philosophical intuitions have been accepted without independent argument by the philosophers who had them. Where intuitions conflict with common-sense convictions or other beliefs, their acceptance requires justification. At any rate where philosophers accept paradoxical intuitions in the absence of independent argument, they have warrant for accepting them only to the extent to which the mere fact that competent thinkers like them have the intuitions at issue speaks for their truth. A number of current philosophers – including the champions of experimental philosophy – seek to develop psychological explanations of philosophically relevant intuitions which allow us to assess this evidentiary value (Knobe and Nichols 2008). These efforts typically focus on conceptual, classificatory, or modal intuitions about specific situations, such as specific Gettier-cases (overview: Alexander and Weinberg 2007). As the case studies of PDT reveal, however, general factual intuitions which are neither situation-specific nor modal play at least as important a role in philosophical reflection. PDT develops the, to my knowledge, first assessmentfacilitating explanation for intuitions of this important but neglected kind. To do so, it seizes on experimental work which has given fresh content to the time-honoured distinction between conscious thought and unconscious cognition: Some cognitive processes are non-intentional, typically unconscious, and place low demands on attentional resources (i.e., their performance is not impaired by multi-tasking); such automatic processes are distinguished from processes that are potentially controlled, typically conscious, and require significant attention (Moors and De Houwer 2006). Processes of these different kinds may interact (Evans and Frankish 2009). Intuitive judgements are often explained as the result of such interaction (e.g. Kahneman and Frederick 2005): Automatic inferences generate conclusions which are subject to spontaneous – and typically insufficient – correction or ‘adjustment’ through potentially controlled and conscious

Through Pictures to Problems

477

processes. PDT develops an explanation of roughly this kind by building on the first of the Wittgensteinian ideas above: In philosophical reflection we are frequently guided by ‘analogies in language’ which are ‘at work in the unconscious’.2 One way in which analogies may enter language is through metaphorical extension: Speakers extend the use of whole families of related terms from more concrete (source-) to more abstract (target-) domains, in a way that preserves the inferential relations between the terms involved. Thus English has come to employ a host of terms from the more concrete domain of sight, in talk about intellectual activities and achievements, which are more abstract. Take the intellectual achievement of understanding actions, about which we may say, e.g.: ‘It is clear or obscure to me why you did what you did, according to whether or not I manage to see any reasons for acting that way. I may look for reasons where these are hidden or be blind to reasons in plain view. An illuminating explanation throws new light on your action and lets me see reasons I had previously overlooked. It may reveal threats in whose light your action looks more reasonable than it did at first sight. At least, it will let me catch some glimpse of them, where I was previously completely in the dark.’ This extension preserves the inferential relations between the terms involved: Regardless of whether I am looking for my keys or your reasons, I cannot see what is lying in the dark, and something that sheds light may help me to see it. Such extensions are motivated by conceptual metaphors which map properties and especially relations from the source-domain (here: visual perception) to the target-domain (here: intellectual effort): (1) (2)

2

S sees x  S knows x I see your point. S does not see x  S does not know x I am in the dark about his plans.

For proper sourcing of following claims, to linguistic, psychological, and philosophical literature, please refer to PDT.

478

(3) (4) (5)

(6) (7) (8)

Eugen Fischer

It is possible to see x / x is visible (for S)  It is possible (for S) to know x The implication was clear. It is not possible (for S) to see x  It is not possible (for S) to know x His point was obscure. X makes it possible (for S) to see y  X makes it possible (for S) to know y His illuminating explanation shed light on the causes of our misfortune. X makes it impossible (for S) to see y  X makes it impossible (for S) to know y Their lawyer buried the key conditions under layers of jargon. S tries to get to see x  S tries to get to know x I am looking for a solution. S overlooks x  S doesn’t know x, though it is possible for him to know x I managed to overlook the most obvious solution.

Such mappings forge structural or second-order analogies (e.g. between intellectual effort and visual perception): The individual elements in the different domains (e.g. implications and physical objects, respectively) do not have the same properties or stand in the same relations. So there is no first-order similarity and no first-order analogy. (Easter eggs and implications are not hidden in the same sense. And I don’t stand in the same relation to the ball and to the implication I see.) But properties and relations from one domain can be, and are, mapped onto properties and relations from the other domain in such a way that the correlated properties and relations stand in some of the same higher-order (causal or inferential) relations among each other: ‘X is hidden’ implies ‘S does not see x’, regardless of whether x is an Easter egg or an implication. Therefore a second-order analogy obtains. While the second-order analogies forged by conceptual metaphors actually obtain, they facilitate automatic inferences which presuppose the ‘false analogies’ to which Wittgenstein adverts (and to which we shall presently turn). Much of our reasoning is by analogy. So is much of our automatic cognition: When something new or only partially known is structurally analogous to a more familiar or better known model, we automatically align the

Through Pictures to Problems

479

two and spontaneously project features from the model onto the corresponding part of the other, the target-domain, without being aware of doing so. The occurrence of such non-intentional analogical reasoning has been established through text-comprehension experiments (e.g. Day and Gentner 2007): When given structurally similar texts with varying degrees of detail, subjects spontaneously fill in gaps and ambiguities in the less detailed text, in analogy to the corresponding points in the similar, more detailed text. And, crucially, they are not aware of making any analogical inference or referring back to any model. In these and similar experiments, subjects are not pursuing any particular or practical goal when reading the texts – they have just been told to ‘read these texts, to answer some questions about them later’. They lack knowledge of context and detail about the target (in this case the scenario of the less detailed target story). And they are provided with a structurally similar model with which they are more familiar. These conditions frequently obtain in philosophical reflection: Such reflection is often not geared towards particular, let alone practical goals, and involves swift generalisation or general reasoning without detailed reference to specific examples, which leaves thinkers adrift without much awareness of actually relevant context and detail. They are then ready to unwittingly seize on any structurally analogous and familiar model offered up by the language they use, and will spontaneously project features from the more concrete (say, sight) to the more abstract (say, thought). Automatic cognition often provides us with a host of rough-and-ready conclusions that are then winnowed through potentially controlled reasoning. Where structural analogies facilitate analogical inferences, thinkers are prone to spontaneously project further relations and properties, from source to target – and to correct the conclusions only where these are obviously wrong. When thinking about the abstract domain of reflection, thinkers may thus, e.g., unwittingly presuppose that just as when we see something, it must be around to be seen, and seen with an organ of sense (the eyes), so something must be around, and an organ of sense employed, when we ‘look at’ a problem or think about something: R

To think about something is to perceive some thing that is located at some place, with an organ of sense.

480

Eugen Fischer

When a thinker systematically presupposes such a ‘false analogy’ as (R) in philosophical reflection, he assimilates an abstract target to a concrete model in ways in which he knows the two to be different. PDT employs a Wittgensteinian term to characterise this phenomenon: Those presupposing (R) are ‘under the spell of the perceptual picture of reflection’. More generally: A thinker is under the spell of a philosophical picture iff he systematically makes non-intentional analogical inferences which assimilate targets to source-models of a conceptual metaphor, in ways in which the thinker knows them to be different (picture-driven inferences). When automatic inferences lead to conflicts with explicit knowledge, their conclusions are often spontaneously corrected or adjusted – but typically insufficiently so. These two moves can jointly explain a wide range of otherwise perplexing philosophical intuitions, including the intuitions constitutive of the early modern conception of the mind as a space and organ of inner perception. For example, conventional metaphorical uses of these perception-terms let me say that I think about or ‘consider’ an issue ‘put before me’, and ‘look at’ it ‘from all sides’ and ‘with eagle-eyes’. But of course none of these intellectual efforts essentially involve the literal use of our eyes, or objects literally in our environment: T1

To think about something is not to perceive any thing that is located in our environment, and is not to use one’s eyes – or any other external sense-organ.

Thinkers under the spell of the perceptual picture of reflection will leap from this truism to a conclusion it entails in conjunction, and only in conjunction, with (R): C1

To think about something is to perceive something located (if not around us, then) within us, by using (if not an external, then) an internal organ of sense.

Locke, for instance, takes this claim for granted, in distinguishing two principal actions of the mind: ‘volition, or willing’ and ‘perception, or thinking’ (EHU II.vi.2), with the implication that, just as when somebody performs an act of volition when he ‘wills something’, so somebody who thinks performs an act of perception. Anyone can get clear on the latter ‘by

Through Pictures to Problems

481

reflecting on what he does himself, when he sees, hears, feels, etc. or thinks’ (II.ix.2). Such perception only occurs when an idea is in the mind (II.ix.2-4). Hence: To think about something is to perceive an idea within us, in the mind. Analogous inferences lead to analogous conclusions about other intellectual activities and achievements. Together, they introduce the early modern concepts of ‘the mind’ and ‘ideas’: D1 D2

‘the mind’ = def.: (a) the inner space in which we perceive the things we do not perceive with our five senses; (b) the inner organ of sense with which we perceive the objects in that space. ‘idea’= def.: whatever is perceived in the mind, with the mind. (Cp. EHU II.viii.8)

This much takes us to the conception of a ‘mind’ literally located in us, which Locke employs, e.g., when envisaging how ideas are ‘convey[ed] ... to their audience in the brain, the mind’s presence-room’ (II.iii.1). The first conclusion appears to conflict with the common-sense conviction that T2

When we think we do not perceive anything in our bodies and need not employ any bodily organ of sense.

A common response to such conflicts is spontaneous reinterpretation of the conclusion, which ‘sublimates’ the conclusion by papering over the conflicts, thus appearing to render the intuitively compelling conclusion intellectually acceptable. Many early modern thinkers tell us that (C1) is not to be taken literally: C2

To think about something is to perceive something in an ‘inner space’ that is not literally, physically, in us, with an ‘inner sense’ that is not to be confused with any bodily organ of sense.

Locke refers to this common explanation but cannot decide whether to adopt it himself (EHU IV.iii.6): ‘the mind is thought to take up no space, to have no extension’ (II.ix.10). Crucially, this sublimating explanation does not take us back to the innocuous starting-point: It does not tell us that to think about things is to ‘perceive’ them merely in the well-established metaphorical sense of some perception-verbs (like ‘look at’ in ‘looking at a question from all sides’). It does not retract the picture-driven conclusion

482

Eugen Fischer

that thinking is a kind of perception along with seeing, etc., and with proper objects and organs which must be somehow internal if they are not external. It only tells us that these objects, spaces, and organs of perception do not compete for space in our body, as the organs of the other senses do. An automatic inference and insufficient conscious adjustment can thus account for early modern intuitions that posited in us ‘minds’ with contents, which need a place to stay but do not readily fit into the physical world. These two moves may interact with experimentally well established belief-bias effects which are no more under our direct control than the automatic inferences just discussed: Regardless of their content, our prior beliefs, intuitive and other, shape our perception and reasoning. Among other things, we accept argument, sound and unsound, more readily when it leads to a conclusion we believe true, and interpret facts, new and familiar, in the light of prior beliefs. This may lead us to conflate substantive claims with truisms interpreted in their light. Thus, the moment we believe that – C1/2 – to think is to perceive an idea in a – material or immaterial – mind, we will leap from the actual truism that we almost continually think of or about things, and have thoughts, to the conclusion that we almost continually perceive ideas in our minds, and regard this too as a truism. This in turn will have us regard the existence of ideas and minds that house them as patently obvious, like Locke: ‘I suppose it will be easily granted me that there are such ideas in men’s minds; everyone is conscious of them in himself’ (EHU I.i.8). Belief bias can also give rise to a mismatch between implicit beliefs presupposed in reasoning, and explicit beliefs. Such mismatch arises for instance when a picture-driven conclusion gets presupposed in its ‘raw’ or pre-sublimated shape even once it has been explicitly rejected through a sublimating explanation. By developing the account thus outlined, PDT transforms the first idea from Wittgenstein’s Big Typescript into a substantive empirical hypothesis: In automatic cognition, philosophers unwittingly but systematically assimilate the abstract target-domains of conceptual metaphors to their more concrete source-domains, in ways in which they know the two to be different. This hypothesis can be supported by studying the work of particular thinkers and showing that it provides the best available explanation for phenomena we can observe in their texts, in particular for intuitive judgements they accept without further argument, even though they would seem to conflict with other convictions, and for intuitive leaps of thought which

Through Pictures to Problems

483

make for otherwise salient gaps in philosophical arguments. This explanation can be compared with competing explanations of different kinds, derived from work in cognitive science or from the exegetical literature on the authors studied. The latter will typically seek to account for intuitive judgements and leaps of thought by showing that views the author explicitly maintains elsewhere can be used to motivate those judgements or fill those gaps in the argument. Explanations derived from work in cognitive science can then complement this familiar approach where it does not work: where a thinker’s intuitions are at odds with his explicit claims and the chasms in the arguments he finds blindingly obvious cannot be bridged by invoking only his explicit views. PDT supports the hypothesis through case-studies on the genesis of two lastingly influential conceptions, in 17th century work in the empiricist tradition. With a focus on Galileo, Boyle, Locke, and Berkeley, and a good look at Berkeley’s heir in 20th century analytic philosophy, A.J. Ayer, the book examines the genesis of the conception of the mind as a space of inner perception and of sense-datum doctrines of perception relying on this conception. PDT systematically traces the influence on those thinkers of two kinds of philosophical pictures: of perceptual pictures of intellection and sensation, and of the mereological picture of qualities which assimilates the relationship between qualities and the things that have them to that of physical parts or pieces to the wholes to which they belong. The book shows that the process outlined can account for consequential intuitions including: I1 I2 I3 I4

There is in us some sort of space and organ of perception (‘mind’) and things (‘ideas’) which we perceive in and with it. Others cannot possibly perceive the things I perceive in my mind. (The mind is a private realm.) We know we have a certain thought or belief when we have it, and vice versa. (The mind is a transparent medium.) To see colours, hear sounds, smell odours, taste tastes, and feel heat or cold is to perceive ideas in the mind, with the mind.

The hypothesised process can also explain striking gaps in arguments like Galileo’s notorious ‘argument from conceivability’ which, with bold immediacy, leaps from the assumptions that the determinable qualities of colour, odour, taste, and sound are neither essential nor a priori properties

484

Eugen Fischer

of physical bodies, to the mind-boggling conclusion that colours and their ilk really exist only as sensations in our own sensitive bodies. 2. Ill-motivated Problems The hypothesised process can systematically generate philosophical problems of a characteristic kind, aptly summed up by the second of Wittgenstein’s initially quoted claims: ‘A simile that has been absorbed into the forms of our language produces a false appearance, and this disquiets us. “But this isn’t how it is!” – we say. “Yet this is how it has to be!” (PI 112) Picture-driven inferences take us to conclusions that conflict – like C1 above – with common-sense convictions and get, all the same, presupposed in further reasoning. This systematically engenders tensions between intuitively compelling conclusions and other convictions: They jointly appear to entail a contradiction or otherwise unacceptable consequence. The first of these tensions get papered over by sublimating explanations. But the moment such tensions become salient in or through further reasoning, they give rise to the formulation of philosophical problems. Where we reject the intuitive conclusion, we conceptualise these problems as paradoxes. Where we accept both the common-sense conviction that p and the intuitively compelling claim that q which appears to be at odds with it (“But this isn’t how it is!”), and insist upon our intuition (“But this is how it has to be!”), we wish to reconcile the two and formulate reconciliation problems, frequently (though not invariably) articulated by questions of the form: How is it (possible) that p (given that q)? The intuitions and leaps of thought examined in PDT all raise such problems. E.g.: How is it possible to say that physical bodies are coloured (given that colours really exist only as sensations in the mind)? Or: How is it possible to perceive physical objects (given that all we are immediately aware of in perception are ideas in the mind)? The common approach to paradoxes is to try to resolve them by exposing mistakes in the – often largely tacit – reasoning that leads to the paradoxical conclusion that q. By contrast, the common approach to reconciliation problems consists in the construction of philosophical theories which are to honour the theorists’ intuitions (including q), to accommodate the relevant facts (including p), and to explain how these facts can obtain, namely, in the light of their apparent clashes with those ‘intuitive insights’. Many philosophical theories thus are to solve various related reconciliation

Through Pictures to Problems

485

problems – and are required to accept the very intuitions that raise these problems. Where such problems are raised only by intuitions we have no right to accept, however, these problems are ill-motivated and call not for a theory that solves them (and is built on those unwarranted intuitions) but for the kind of quest for mistakes we commonly engage in to overcome paradoxes, a quest for hidden mistakes that may dissolve an ill-motivated reconciliation problem by showing that it is ill-motivated, i.e., that the thinkers raising it lack all warrant for maintaining at least one of the parties to the apparent conflict that creates the impression of a difficulty. PDT studies reconciliation problems that are ultimately raised by paradoxical intuitions whose acceptance is prima facie puzzling: Their champions typically maintain them without any explicit argument or defend them with arguments that already presuppose the intuitions at issue. When they are thus in need of justification but not justified by anything else, we are entitled to accept intuitions only if the mere fact that we have them already speaks for their truth. This is the case, for example, when they are due to the exercise of cognitive or linguistic competencies. But it is not the case when they are due to a cognitive process which is unsound and relies on automatic inferences the thinker knows to be fallacious, or on tacit assumptions he knows to be false. We can hence assess what warrant thinkers have for accepting intuitions unsupported by independent argument, by explaining why they have these intuitions: by tracing the intuitions back, e.g., to cognitive competencies or to such unsound processes. Where such processes generate unwarranted intuitions that in turn give rise to ill-motivated problems, we can dissolve these problems, namely, show that they are illmotivated, that their proponents have no reason to believe there is any such difficulty about the familiar fact that p as they imagine: We can show this by explaining the intuitions from which the problems arise, and establishing that these intuitions are due to cognitive processes which are systematically unsound – for instance, to picture-driven reasoning. This approach involves a significant reorientation of philosophical work at the strategic level: In order to overcome some philosophical questions, we turn from them and their topics (e.g. colour or sense-perception) to the meta-philosophical question of how thinkers come to raise those questions and, more generally, think about that topic when engaged in philosophical reflection. Let’s call this the meta-philosophical turn. PDT vindicates this turn and explores one helpful way of taking it: We can sometimes dissolve philosophical problems by getting clear on how and why we come to raise

486

Eugen Fischer

them, in particular by hunting for particularly decisive mistakes that are made already at the level of automatic cognition of which we are largely unaware. This kind of research can be characterised as ‘cognitive epistemology of philosophy’. Such a cognitive epistemology studies how philosophical reflection is shaped by automatic cognition of which we are typically unaware but which we can study through experiments. It seeks to explain intuitive judgements and leaps of thought to be found in philosophical texts as the result of cognitive processes for which there is independent experimental evidence, and uses these explanations to assess the soundness of those judgments and inferences. To the extent to which it reveals that judgements and inferences essential for raising philosophical problems are unsound, such a cognitive epistemology of philosophy will put us into a position to dissolve those problems. The proof of this pudding lies in the eating. But some problems stand out from the start as candidates for such dissolution: According to Plato, philosophical reflection is initially motivated by a sense of wonder in the face of familiar facts (Theaetetus 155). Where we find a familiar fact that p wonderful or puzzling and want to ask how it is as much as possible that p, prior to deliberate reflection, our philosophical question is likely to be motivated by a paradoxical intuition we accept without independent argument. In contrast with other kinds of problems raised by philosophers, these puzzles are not taken up by other disciplines. Such peculiarly philosophical ‘Platonic’ problems warrant an examination of the processes that generate the intuitions from which they arise, an examination which may well reveal that they are ill-motivated. 3. Therapeutic Philosophy The present account of the genesis of some philosophical intuitions and the nature of some philosophical problems lends substance to a therapeutic conception of philosophy of the kind for which Wittgenstein is notorious. To see this, we need to get clear on the notion of therapy which is frequently used very loosely. The term ‘therapy’ implies that there is some ‘illness’ or ‘disease’ the therapy is to ‘cure’. This implication cannot be cancelled simply by deciding to use the term metaphorically. As we have seen, the metaphorical extension of established terms escapes vacuity by preserving core inferential relations to others (if only upon their meta-

Through Pictures to Problems

487

phorical extension). Any substantive – non-vacuous – talk of ‘therapy’ hence has to indicate some kind of ‘illness’ it is to ‘cure’. People qualify as ‘ill’ when they fall significantly short of an ideal of health. The ideal of mental health that informs most talk of ‘illness’ and ‘therapy’ in nonsomatic contexts is first developed in Plato’s Republic (Bk.4, culminating in 443c-444e). It is the ideal of a balanced, rational agent who can master his feelings and impulses sufficiently well to be rationally autonomous. Non-somatic ‘illness’ is constituted by a serious lack of rational autonomy. We lack such autonomy when we have emotional or behavioural problems, as mental health professionals call them: A subject has emotional problems iff he experiences unwarranted and distressing or otherwise disabling emotions which he is unable to control. A subject has behavioural problems to the extent to which his behaviour is unreasonable and fails to be autonomous. A subject’s behaviour fails to be autonomous in case he refrains from doing things that he would want to do if not in a certain disabling condition that is not under his direct control or does things he would not want to do but for that condition. Think, for example, of the overwhelming feelings of anxiety that may prevent a phobic from boarding a plane, which he would want to take but for his phobic condition; or of a paranoid man who ceases to go out for fear of being attacked by – imaginary – conspirators. To cure a non-somatic illness is to solve such emotional or behavioural problems, i.e., to put an end to, and prevent the recurrence of, emotions or behaviours which betray a lack of autonomy. Philosophical efforts will qualify as therapeutic in a substantive sense in case they are guided by the paramount aim of solving such problems. This can happen in two very different ways (Fischer 2011b): Ancient Stoics, Epicureans, and Pyrrhonic sceptics, and some modern philosophical counsellors, primarily seek to solve emotional or behavioural problems that arise in ‘real life’, prior to and independently from philosophical reflection. Let’s say they practise philosophical therapy. By contrast, therapeutic philosophy addresses emotional and behavioural problems that arise only in and through philosophical reflection. A case in point is the author of the Big Typescript. Here, Wittgenstein examines the ‘disquiet’ that intellectual problems can cause a

488

Eugen Fischer

thinker (BT 409, 415, 416, 421, 431), distinguishes the sort of disquiet that scientific or technical problems can cause from the sort that philosophical questions bring about (BT 414-15), and declares that the removal of the latter kind of disquiet is the paramount aim of his own philosophical work: ‘As I do philosophy, its entire [!] task is to shape expression in such a way that certain disquietudes [German original: Beunruhigungen] disappear’ (BT 421). At first sight, this may look plainly mad. When an intellectual problem disquiets a thinker struggling to solve it, he can put a lasting end to this disquiet only by solving the problem. In such a case there is typically another and stronger motivation for wanting to solve the problem than the desire to put an end to one’s own disquiet. (Indeed, typically only somebody who has some such stronger motivation will find the problem seriously disquieting, in the first place.) And the disquiet caused by the problem will disappear on its own, the moment one gets to know the solution. So why should it be reasonable to adopt the removal of the disquiet as a distinct and, indeed, as the paramount aim of an intellectual endeavour? One answer is this: Where we are worried by a real problem, we need to solve the problem. But where we are worried by an illusory problem we have no right to believe confronts us, the only thing we can reasonably do is to get ourselves to stop unreasonably worrying about it. (Compare: When a child is afraid that monsters might creep in at night, there is no real security problem; we do not have to guard the room. But we face, all the same, a real problem of another kind: We need to liberate the child from the unwarranted fear it cannot control, and thus solve an emotional problem.) Hence whenever we imagine a problem where there is none, due to conditions that are not under our direct control, the only genuine problems we face are emotional and behavioural problems. The emotional problems consist in our unreasonable worries or disquietudes; the behavioural problems are constituted by theoretical or other attempts we make to solve those illusory problems – efforts we would reject as intellectually pointless if we were not subject to those conditions beyond our direct control. Where philosophers struggle with ill-motivated questions, and such emotional or behavioural problems are the only genuine problems they face, they can reasonably adopt the solution of such problems as the primary aim of their endeavours. Let’s call the reorientation of philosophical work through adoption of this new guiding aim the ‘therapeutic turn’.

Through Pictures to Problems

489

If therapeutic philosophy is defensible at all, it will have to be vindicated by a certain account of nature and genesis of the philosophical problems targeted: an account which (i) (ii)

shows these ‘problems’ to be ill-motivated and traces their formulation back to conditions or processes which are not under our direct control.

PDT develops an account of some philosophical intuitions and problems which does just that: The account, roughly outlined above, shows that the problems at issue are raised only by intuitions which are unsound, and traces these intuitions – and thus the problems they raise – back to picturedriven reasoning, i.e., to a certain kind of unsound automatic reasoning of which we are typically unaware and which is never under our direct control. In this way, the account identifies a pertinent ‘disabling’ condition: being under the spell of philosophical pictures. It thus vindicates the at first sight wild therapeutic turn, in a part of our subject. Indeed, the account not only motivates the adoption of new aims but also provides us with much of what we need to pursue them. To practise therapeutic philosophy, we can adapt methods of cognitive therapy, which build on its central insight. This well-established form of psychotherapy (e.g. for depression) proceeds mainly by reasoned argument and collaborative assessment of evidence, in Socratic dialogue (Beck 1995). Apart from the identification and modification of psychological motivations, such therapy turns on helping the patient acquire ‘meta-cognitive insight’, that is, general insight into the mainly automatic psychological processes, cognitive and other, which have him maintain troubling beliefs in the absence of warrant (Nelson 2005, 108-28). The central finding is that such insight weakens those beliefs and attendant emotions, and enables subjects not to act on them, thus alleviating their emotional, and solving their behavioural problems. We can try to replicate this success in philosophy. Philosophically relevant meta-cognitive insight can be gained by engaging in cognitive epistemology: by developing meta-philosophical accounts like the account of picture-driven reasoning outlined above, and using them to gain a shared understanding of how sound and unsound processes of automatic cognition shape intuitive judgements and leaps of thought of the thinker addressed (who may be the therapist himself, in self-therapy), with a focus on the

490

Eugen Fischer

intuitions and leaps that raise the philosophical problems that puzzle him or her. We thus arrive at a fresh vision of – a significant part of – philosophy: At any rate where philosophers raise their problems under the spell of philosophical pictures, philosophy is best practised as a kind of cognitive therapy which enhances meta-cognitive insight, guided by the findings and theories of a novel meta-philosophy – a cognitive epistemology of philosophy – in something like the way in which cognitive psychotherapy is guided by the findings and theories of different branches of psychology. Further pursuit of this vision is consistent with the Wittgensteinian precept that to resolve philosophical problems we should not seek to establish ‘new truths about the subject of the investigation’ (BT 416): We come to resolve problems by discovering truths not about this subject but about how philosophers think about it and came to raise those problems. This approach does, however, take us beyond other notorious Wittgensteinian strictures (some of which Wittgenstein himself does not seem to consistently respect): It certainly goes far beyond ‘assembling reminders’ of what we already know, and has us frame hypotheses that are potentially controversial; indeed, it even has us construct theories, namely about how people think when engaged in abstract reflection. But while we have seen good reason to be wary of philosophical theories, whose raison d’être all too often is the reconciliation of familiar facts with unwarranted intuitions we should not honour but abandon, we have every reason to construct metaphilosophical theories which can guide therapeutic efforts. The research programme summed up by the present ‘vision’ is certainly not without precedent or pedigree: PDT argues that, independently from each other, both Wittgenstein and J.L. Austin at least intermittently took the therapeutic as well as the meta-philosophical turn. In at least some of their work, they focused on what were subsequently called ‘emotional’ or ‘behavioural problems’, and sought to solve such problems by getting clear on how philosophical reflection is shaped by cognitive processes of which we are unaware: Wittgenstein sought to achieve ‘thoughts at peace’ (CV 50) and to ‘cure in himself many diseases of the understanding’ (ibid.), e.g., by getting clear on how ‘similes work in the unconscious’ (VW 69) (cp. Fischer 2004, 2008). Austin sought to ‘dissolve philosophical worries’ by ‘expos[ing] concealed motives’ of which we are commonly unaware (Austin 1962, 5). While thus building on two pillars of an important strand of 20th century analytic philosophy, the present proposals take us beyond its linguistic paradigm: We retain the idea that many philosophical problems

Through Pictures to Problems

491

are not to be solved but dissolved, though by showing them ill-motivated rather than necessarily meaningless. But the knowledge that allows us to do so, and to put an end to unwarranted toil and worry, is not, first and foremost, conceptual or linguistic knowledge (which may help) but selfknowledge of a kind we can acquire by turning to experimental psychology. References Alexander, J., and J. Weinberg, 2007: Analytic Epistemology and Experimental Philosophy, Philosophy Compass 2, 56-80 Austin, J.L., 1962: Sense and Sensibilia, Oxford: Oxford University Press Beck, J., 1995: Cognitive Therapy: Basics and Beyond, New York: Guildford Day, S., and D. Gentner, 2007: Non-intentional Analogical Inference in Text-Comprehension, Memory & Cognition 35, 39-49 Evans, J.S.B.T., and K. Frankish (eds.), 2009: In Two Minds: Dual Processes and Beyond, Oxford: Oxford University Press Fischer, E., 2011a: Philosophical Delusion and its Therapy. Outline of a Philosophical Revolution, New York: Routledge Fischer, E., 2011b: How to Practise Philosophy as Therapy: Philosophical Therapy and Therapeutic Philosophy, Metaphilosophy 42, 49-82 Fischer, E., 2008: Wittgenstein’s Non-Cognitivism – Explained and Vindicated, Synthese 162, 53-84 Fischer, E., 2004: A Cognitive Self-Therapy – Philosophical Investigations §§138-97, in: Wittgenstein at Work. Method in the “Philosophical Investigations”, ed. E. Ammereller and E. Fischer, London: Routledge, 86-126 Kahneman, D., and S. Frederick, 2005: A Model of Heuristic Judgement, in: The Cambridge Handbook of Thinking and Reasoning, ed. K. Holyoak and R. Morrison, Cambridge: Cambridge University Press, 267-293 Knobe, J., and S. Nichols, 2008: An Experimental Philosophy Manifesto, in: Experimental Philosophy, ed. J. Knobe and S. Nichols, Oxford: Oxford University Press, 3-14 Locke, J., 1700/1975 (EHU): An Essay Concerning Human Understanding, 4 th ed., ed. P. Nidditch, Oxford: Clarendon Moors, A. and J. De Houwer, 2006: Automaticity: a theoretical and conceptual analysis, Psychological Bulletin 132, 297-326 Nelson, H., 2005: Cognitive-Behavioural Therapy with Delusions and Hallucinations, 2nd ed., Cheltenham: Nelson Thornes

492

Eugen Fischer

Wittgenstein, L., 1933/2005 (BT): The Big Typescript: TS213, Oxford: Blackwell Wittgenstein, L., 1953/2001 (PI): Philosophical Investigations, Oxford: Blackwell Wittgenstein, L., 1998 (CV): Culture and Value, rev. ed., Oxford: Blackwell Wittgenstein, L. and F. Waismann, 2003 (VW): The Voices of Wittgenstein, London: Routledge

Werte und Vernunft in Wittgensteins Spätwerk ANDREAS KORITENSKY, Theologische Fakultät Paderborn

Problemstellung Wittgensteins Werk wird von einem starken Strang aus Bemerkungen zur Lebenssinn- und Wertproblematik durchzogen, der im Kontext der analytischen Tradition sehr auffällig ist. Hans-Johann Glock hat vorgeschlagen, die Spannungen, die zwischen dem sprachphilosophischen Programm, verstanden als eine linguistisch zugespitzte Form des kantischen Kritizismus, und den kulturkritischen, wissenschaftsfeindlichen und moralischen Äußerungen bestehen, durch eine scharfe Trennung der beiden Fragestellungen aufzulösen (Glock 2001, 19). Da Letztere irrationale, mystizistische Tendenzen erkennen lassen, können durch die Trennung die sprachphilosophischen Analysen als rationale Unternehmungen sichergestellt werden. Auch wenn er sie als durchgängige Merkmale von Wittgensteins Denken versteht, so begründet Glock seine beiden Thesen vor allem durch die Texte aus dem Umfeld des Tractatus. Im Tractatus lassen sich die sprachtheoretischen Teile tatsächlich sowohl genetisch als auch logisch-argumentativ als unabhängige Einheit verstehen (Glock 2001, 202). Erst in einem zweiten Schritt werden sie von Wittgenstein zum Mittel der Suche nach dem „unsagbaren Grund“ deklariert (TLP Vorwort u. 6.52). Für die Zeit nach 1929 wird eine solche klare Abgrenzung schwierig, so dass auch eine selektive, rein sprachanalytische Lektüre problematisch wird. Denn in den späteren Jahren rücken die existentiellen Bemerkungen zwar an den Rand, gleichzeitig wird aber die gesamte Sprachuntersuchung selbst von einer ethischen Fragestellung durchdrungen. Außerdem tritt die Sinnfrage nun aus dem Schatten der „mystischen“ Unaussagbarkeit heraus, so dass Sinndiskurse prinzipiell möglich werden. Für eine Bewertung der Funktion der Problematik von Werten und Sinn in Wittgensteins späteren Werken sind daher drei Fragen zu beantworten:

494

Andreas Koritensky

(1) Wie hat Wittgenstein die Synthese von Sprachanalyse und Sinnfrage konzipiert? (2) Welche Auswirkung hat diese Synthese auf die Rationalität seiner Philosophie? (3) Warum bleibt Wittgenstein trotz der Bedeutung der Sinnfrage für sein Denken so zurückhaltend, wenn es darum geht, in explizite Sinndiskurse einzutreten? Wittgensteins Philosophiekonzept und seine Probleme (1) Ziele: Realismus und Eudaimonia. Zwei Leistungen schätzt Wittgenstein an Beethovens Musik (MS 183, 72): (a) Sie sei erstens realistisch in ihrem Blick auf die Wirklichkeit (Welt und Leben). (b) Dieser fiktionsfreie, „undichterische“ Realismus wird zweitens zur Voraussetzung für so etwas wie eudaimonia – Wittgenstein selbst verwendet die religiös eingefärbte Terminologie von „Trost“ und „Erlösung“ – unter den Bedingungen einer kontingenten Existenz. Diese Passage beschreibt ein Ideal, das Wittgenstein offenkundig auch für seine Philosophie anstrebt (MS 109, 202, 205f.). Diese doppelte Zielsetzung von Realismus und eudaimonia machen die ethische Formierung seines Denkens aus. (2) Aufklärung oder Konservatismus? In der Zielbestimmung seines Philosophierens lassen sich bei Wittgenstein zwei gegenläufige Bewegungen ausmachen, eine, die sich als Aufklärungsprojekt beschreiben lässt, und ein Zug zum kulturellen Konservatismus. Letzterer ist wohl einer der Gründe dafür, warum die existentiellen Passagen in Wittgensteins Werk manchem suspekt sind. Diese Spannung lässt sich ausgehend von einem klassischen Text zum Philosophiekonzept (PU 109) erläutern: „[...] Und wir dürfen keinerlei Theorie aufstellen. Es darf nichts Hypothetisches in unseren Betrachtungen sein. Alle Erklärung muß fort, und nur Beschreibung an ihre Stelle treten. Und diese Beschreibung empfängt ihr Licht, d.i. ihren Zweck, von den philosophischen Problemen. Diese sind freilich keine empirischen, sondern sie werden durch eine Einsicht in das Arbeiten unserer Sprache gelöst, und zwar so, daß dieses erkannt wird: entgegen einem Trieb, es mißzuverstehen. Die Probleme werden gelöst, nicht durch das Beibringen neuer Erfahrung, sondern durch Zusam-

Werte und Vernunft in Wittgensteins Spätwerk

495

menstellung des längst Bekannten. Die Philosophie ist ein Kampf gegen die Verhexung unseres Verstandes durch die Mittel unserer Sprache.“ (PU 109)

(a) Es ist eine klassische Aufklärungsszene, die uns Wittgenstein vor Augen stellt. Sie lässt sich durch einen Vergleich mit Motiven aus Platons Werken erläutern, deren Lektüre sich in Wittgensteins Notizbüchern seit 1931 nachweisen lässt (MS 111, 13, 31). Wie Platons Höhlenbewohner sind die Menschen in ihren allgemein geteilten, falschen Denkgewohnheiten gefangen. Der Ausbruch eines Einzelnen vollzieht sich wie bei Platon durch Zwang (Pol. 515c). Das Motiv des Herausreißens erscheint explizit in einer früheren Version der Befreiungsmetapher (MS 113, 45). Das Herausgerissenwerden ist nur möglich, wenn der Mensch bereits in „instinktiver Auflehnung gegen die Sprache“ lebt (ebd.). Die kulturelle Entfremdung, die Wittgenstein in den Entwürfen Zu einem Vorwort von 1930 zum Ausdruck bringt, ist damit eine wichtige Voraussetzung für die Philosophie, wie sie Wittgenstein betreibt, und nicht ein idiosynkratrisches Anhängsel (MS 109, 204-207). Platons Mythenkritik hat ihre Analogie in Wittgensteins Forderung nach einer „undichterischen“ Weltsicht, die auf Fiktionen verzichtet (MS 183, 72, MS 110, 210f.). So nimmt Wittgenstein auf eine Passage aus Renans Peuple d‘ Israel Bezug, in der die nüchterne Auffassung des Lebensgrundes als Lebensatem in der Bibel den reich ausgestalteten Mythologien anderer Völker von der Seele und deren Nachleben gegenübergestellt wird (Renan 1887, 41f.), und erklärt die gleiche Nüchternheit zum Charakteristikum seiner Philosophie (MS 109, 202). Die philosophischen Beschreibungen gewähren Einsicht in die Funktionsweise der Sprache. Aber diese Einsicht ist nicht Selbstzweck, sondern dient der Einübung und Aufrechterhaltung der neuen, richtigen Sichtweise und – im besten Fall – der Bekehrung weiterer Höhlenbewohner. (b) Das platonische anamnêsis-Motiv klingt in der Forderung an, dass die Beschreibung nur Bekanntes zusammenstelle (PU 109). Beschreibung setzt daher einen Akt der Erinnerung (MS 111, 164, MS 112, 235), der Rückbesinnung auf den Sprachgebrauch voraus. Dieses Motiv der Rückbesinnung auf Bekanntes steht bereits in einer eigentümlichen Spannung zur Idee einer Sprache, die von einer Neigung geformt wurde, die die falschen Denkgewohnheiten kreiert (MS 113, 45). Hier treffen nämlich das Aufklärungsprogramm und ein kulturkonservativer Zug in Wittgensteins Philosophie aufeinander. Die Rückbesinnung macht nur Sinn, wenn es einmal – analog zur objektiven rationalen Wirklichkeit bei Platon – einen richtigen

496

Andreas Koritensky

Sprachgebrauch gegeben hat, der sich auf diese Weise wieder entdecken lässt („Zurück auf den rauen Boden“, PU 107). (3) Deutungsprobleme. Das Aufklärungsprogramm fordert, einen nüchternen, fiktionsfreien Blick auf die Wirklichkeit einzunehmen, der Störfaktoren des Denkens erkennt und so unschädlich macht. Wahrheit und Klarheit wären dann Selbstzweck, aber auch praxisrelevant, als Voraussetzung eines Lebens, das hinderliche Illusionen überwunden hat. Ähnlich formuliert es das Manifest des Wiener Kreises: „Die wissenschaftliche Weltauffassung dient dem Leben und das Leben nimmt sie auf“ (Carnap u.a. 1999, 145). Einmal angenommen, ein Leben ohne Illusionen sei glücklicher sein als eines mit ihnen, so ist damit allerdings noch nicht notwendigerweise jene eudaimonia erreicht, die Wittgenstein anstrebt. Das konservative Programm geht einen Schritt weiter, insofern es eine Rückkehr zur ursprünglichen Sprachverwendung nahelegt und damit unterschwellig diese als Norm proklamiert. Nicht jede Lebensform, die sich in funktionstüchtigen Sprachformen widerspiegelt, ist aber lebens- und damit erstrebenswert. Sie garantiert also ebenfalls nicht die gesuchte eudaimonia. Eine präzisere Verortung der Wertproblematik (1) Wittgenstein hat diese Problematik durchaus gesehen und reflektiert sie am Beispiel des „Problems des Lebens“ in einer Passage von 1937. Dort verwirft er den „konservativen“ Vorschlag, dass dieses Problem durch eine in die „Form des Lebens“ passende Existenzweise zum Verschwinden gebracht werden kann. Wittgenstein vermutet stattdessen, „[…] daß wer richtig lebt, das Problem nicht als Traurigkeit, also doch nicht als problematisch, empfindet, sondern vielmehr als Freude; also gleichsam als einen lichten Äther um sein Leben, nicht als einen fraglichen Hintergrund.“ (MS 118, 17)

Eine Variante dieses Gedankens findet sich auch in MS 110, im Kontext der sogenannten Bemerkungen zu Frazers „Golden Bough“. Hier werden Riten gedeutet, die sich an Gegenständen festmachen, die für das Leben von großer Wichtigkeit sind: „Man könnte sagen nicht ihre Vereinigung (von Eiche und Mensch) hat zu diesen Riten die Veranlassung gegeben, sondern vielleicht ihre Trennung [sondern, in gewissem Sinne, ihre Trennung]

Werte und Vernunft in Wittgensteins Spätwerk

497

Denn das Erwachen des Intellekts geht meiner Trennung von dem ursprünglichen Boden der ursprünglichen Grundlagen des Lebens vor sich. (Die Entstehung der Wahl.) (Die Form des erwachenden Geistes ist die Verehrung.)“ (MS 110, 298f.)

(2) Eine „Urgeschichte“ von Sprache und Sinnfrage. Wenn wir das aufgeklärte und das konservative Modell von Sprache mit diesen Aussagen zusammenfügen, entsteht folgendes Bild: Erste Sprachformen entstehen zunächst unreflektiert im Kontext sozialer Praktiken. Die Ablösung aus einem determinierten, „instinktiven“ Gebrauch der Kommunikationsformen vollzieht sich zu einem sehr frühen Zeitpunkt der Sprachentwicklung. Sie äußert sich zum Beispiel in der Möglichkeit, alte Sprachformen in neuen Kontexten zu verwenden (MS 110, 206, 256). Auf diese Weise wird es unter anderem möglich, „mythische“, bildhafte Ausdrucksformen zu entwickeln. Die Ablösung aus dem Instinktiven hat folgende Konsequenzen: (a) Sie bewirkt ein graduelles Erwachen des Intellekts (MS 144, 22, MS 167, 32f., MS 137, 117), das sich in einem bewussten Umgang mit den Mitteln der Sprache ausdrückt. Die Fähigkeit zur Einsparung der Vielfalt bei den Ausdrucksmitteln ist nur dann unproblematisch, solange nicht versucht wird, die multifunktionale Verwendung zu vereinheitlichen und zu systematisieren. Genau dies geschieht aber in fiktiven Mythologien und in der Metaphysik, die Produkte des „unaufgeklärten“, d.h. unreflektierten Intellekts sind. (b) Zum anderen wirft die Ablösung vom instinktiven Agieren die Sinnfrage und damit das Problem der Ethik auf. Denn die Menschen müssen nun ihr Leben durch Entscheidungen (Wahl) gestalten. (c) Dies setzt die Erkenntnis von Entscheidungsprinzipien voraus. Im Vortrag über Ethik scheint Wittgenstein davon auszugehen, dass diese Wertkriterien Gegenstand einer besonderen Art von Werterfahrung sind. Im oben zitierten Text aus MS 118 wird mit der Transformation der Sinnfrage in „Freude“ zumindest ein Erfolgskriterium für „richtiges Leben“ genannt. In MS 110 beschreibt er sie als geistige Haltung („Verehrung“, wohl im Sinne von Ehrfurcht), die offenbar mehrere Aspekte hat. (i) Zum einen hat sie eine kognitive Komponente, insofern sie eine Einsicht in das impliziert, was für unser Leben wichtig ist (Lebensgrundlagen). Genauer gesagt, in diesem Kontext bilden sich Begriffe wie „Wichtigkeit“, „Wert“ und „Sinn“ in dieser emphatischen Bedeutung erstmals heraus. Es entstehen Diskurse, in denen „gutes“ und „schlechtes Leben“ eine Rolle spielen. Es ist wohl nicht abwegig davon

498

Andreas Koritensky

auszugehen, dass sie uns „vom Leben“ auf die gleiche Weise aufgezwungen werden, wie Wittgenstein das in einer sehr späten Bemerkung vom Gottesbegriff annimmt (MS 174, 1f.). (ii) Zum anderen drückt die Haltung offenbar so etwas wie Respekt aus, der sich in der Aufrechterhaltung der Distanz (Sinnoffenheit als Sinnfähigkeit) widerspiegelt, die die Voraussetzung für den reflektierenden bewussten Umgang ist. Im Respekt schwingt auch jene Forderung nach „Gerechtigkeit gegenüber den Tatsachen“ mit, die Wittgenstein mit seiner Philosophie erstrebt (MS 110, 184, PU 131). Hier kommt der alte philosophische Topos von der Relevanz bestimmter ethischer Charakterzüge für die menschliche Erkenntnisfähigkeit zum Tragen. (d) Das Wählen setzt die intellektuelle Fähigkeit voraus, Werthorizonte zu erkennen und in den konkreten Umständen anzuwenden (praktische Rationalität). (3) Konsequenzen. Trotz ihrer Kürze lässt diese Skizze einige Rückschlüsse darauf zu, wie Wittgenstein sich die Sinnfrage und ihr Verhältnis zu seinem philosophischen Programm vorgestellt hat. Die realistische Betrachtungsweise der Welt ist einmal eine dianoetische Haltung, insofern sie ein richtiges Erfassen der Wirklichkeit ermöglicht. Zugleich hält sie die Neigung in Schach, die Erfahrung der Sinnoffenheit durch fiktive mythische Erzählungen und uniforme metaphysische Systeme zu überwinden, die ihren Zweck verfehlen müssen, weil sie das Wesen des Problems verkennen. Die Metaphysik missversteht es als theoretische Fragenstellung, die fiktive Mythologie erfindet irreale Lösungen, die keinen Trost spenden können (MS 183, 72), weil sie aufgrund irriger Annahmen über die Grammatik falsche Folgerungen aus unseren Ausdrucksformen (z.B. über die Verwendung des Wortes „Seele“) ziehen, die sich daher im Leben nicht bewähren können. Aristoteles fordert dort, wo Emotionen das Urteil vernünftiger Überlegung beeinflussen können, den Schutz der Vernunft durch die Ausbildung einer Haltung des rechten Maßes (sôphrosynê, EN 1140b11f.). Da Wittgenstein die Neigung, der Sinnoffenheit auszuweichen, dem Bereich des Gefühls – und damit dem Prädiskursiven – zuordnet (MS 110, 189), setzt der Realismus eine ethische Formierung des Charakters zumindest voraus. Die philosophische Therapie arbeitet am ganzen Menschen (MS 112, 46). Aus dem bisher Gesagten wird auch verständlich, warum die kulturkonservative Lösung nicht in Frage kommt. Der Wunsch nach einer Rückkehr zur instinktiven Geborgenheit ist ebenfalls ein Fluchtversuch aus der Sinnoffenheit, die eine kaum realisierbare Regressi-

Werte und Vernunft in Wittgensteins Spätwerk

499

on des Bewusstseins erforderte. Wie ist dann aber das Wiedererkennungsmotiv, das „Zusammenstellen von längst Bekanntem“ (PU 109), zu interpretieren? Es würde sich nahelegen, hier von einem vorgreifenden Erkennen dessen zu sprechen, was ein gutes, richtiges Leben ausmacht, das am harmonischen Ineinandergreifen von Sprache, Leben und Lebenswelt erkennbar wird. Unter den Bedingungen einer sinnoffenen und kontingenten Existenzform wie der des Menschen kann es zwar keine allgemeingültige korrekte Lebensform geben. Wenn die Sinnoffenheit aber tatsächlich Sinnfähigkeit sein soll, muss es einen nicht willkürlichen Wertmaßstab geben, an dem sich die kluge Entscheidungsfindung in konkreten Situationen, aus denen sich das Leben zusammensetzt, immer wieder neu orientiert. „Längst bekannt“ ist dieser Wertmaßstab, weil das, was als Sinn erfahren werden kann, durch die conditio humana und ihre Bedürfnisse vorgegeben ist. Den realisierten Wert erkennt man an seiner Wirkung (der „Freude“, MS 118,17). Die Notwendigkeit, diese ethisch-existentielle Zielsetzung aus Andeutungen rekonstruieren zu müssen, gibt Anlass zu der Frage, warum Wittgenstein diese Aspekte seines Denkens selbst nicht weiter ausgeführt hat. Um diese Frage beantworten zu können, müssen wir zuvor noch klären, in welcher Weise die so bestimmte Zielsetzung die Rationalität seiner Philosophie affiziert. Die Funktion der Urgeschichte und die Rationalität der Methode der Beschreibung (1) Methode. Es ist wichtig, die Funktion dieser Urgeschichte im Auge zu behalten. Sie ist Teil der Methode der zweckgebundenen Beschreibungen, von der auch in unserem Ausgangstext PU 109 die Rede war. Wittgenstein ist nämlich nicht an einer genetischen Rekonstruktion der Sprache als Ersatz für die verborgenen logischen Formen aus dem Tractatus interessiert. Eine Beschreibung, die mehr ist als nur eine Wiederholung der Sprachformen, muss eine indirekte Form annehmen, indem sie das Eigene im Fremden, wie z.B. in der oben geschilderten „Urszene“, gespiegelt betrachtet. Vor allem die klassische Form des Sprachspiels, die dem Braunen Buch als Strukturprinzip dient, ist als ein solches Vergleichsobjekt konzipiert, das „durch Ähnlichkeit und Unähnlichkeit ein Licht in die Verhältnisse unserer Sprache werfen“ soll (PU 130). Diese Aufgabe kann es nur wahrnehmen, wenn es einfacher und damit übersichtlicher als der Untersuchungsgegen-

500

Andreas Koritensky

stand ist. Daher rührt die „Primitivität“ der Sprachspiele (PU 2). Fiktive Entwicklungshypothesen, bei denen das Vergleichsobjekt in einen genetischen Zusammenhang mit unserer Sprache und Lebensform gestellt wird, sind nicht mehr als eine Methode der Projektion, um Vergleichbarkeit herzustellen (MS 110, 257). Gerade im Kontext der Sinnfrage ist es wichtig, die Primitivität der Vergleichsobjekte wie unserer „Urgeschichte“ nicht auf die Sinndiskurse für unser Leben zu übertragen. Wittgenstein geht es nicht um eine Rückkehr zum geistigen Niveau der „Urmenschen“ mit ihren magisch-mythischen Sinndiskursen, sondern um die Einsicht in einzelne Aspekte der Sinnfrage und ihres Ausdrucks. Welche Aspekte auf diese Weise erläutert werden, hängt von der Anwendung des Vergleichsobjekts ab. Ein solcher Aspekt, den Wittgenstein im Auge hat, ist die Unterscheidung und Abgrenzung der Sinndiskurse von (proto-)naturwissenschaftlicher Theoriebildung. Sie richtet sich gegen die positivistische Interpretation der Religion, wie sie sich zum Beispiel bei Frazer findet, und damit gegen eine Monokultur der Deutung von Sprache und des sich in ihr spiegelnden menschlichen Lebens. Zusammen mit der Primitivität der Beispiele kann dadurch leicht der Eindruck entstehen, Wittgenstein entziehe die Sinnfrage der kritischen rationalen Reflexion. Diese Gefahr hat Wittgenstein selbst erkannt, und daher die urgeschichtliche Skizze vom Erwachen des Intellekts, die die Sinnfrage erste möglich macht, eingefügt, die einen weiteren Einzelaspekt unser realen, gegenwärtigen Lebenssinnproblematik beleuchten soll. Da der erste Aspekt bei Wittgenstein immer eine größere Aufmerksamkeit gefunden hat, wird der zweite Aspekt leicht übersehen. Dass mit diesen zwei Perspektiven die Sinnthematik noch nicht umfassend in den Blick genommen ist, ist offensichtlich. (2) Grenzen der Rationalität. Inwiefern ist Wittgensteins Methode der Betrachtung von Sprachspielen ein rationales Unterfangen? Zunächst fallen drei Beschränkungen ins Auge, die mit der Zweckgebundenheit der Beschreibung einhergehen: (a) Die zweckgebundenen Beschreibungen beanspruchen nicht, eine absolute Perspektive auf ihren Gegenstand zu eröffnen (PU 132). Das Interesse an einer theoretischen Durchdringung der Sprache ist also begrenzt. (b) Die Qualität einer Beschreibung bemisst sich an ihrer Fähigkeit, bei den Adressaten Zustimmung auszulösen. Sie ist also auf deren subjektiven Horizont ausgerichtet. (c) Die intendierte Wirkung wird nicht zuletzt in einer veränderten emotional-appetitiven Haltung sichtbar,

Werte und Vernunft in Wittgensteins Spätwerk

501

der „Resignation“ der metaphysisch-mythischen Neigung (MS 110, 189). Das Ziel der Beschreibung ist also die Formung des Charakters, zu dem erkenntnisrelevante ethische und wohl auch dianoetische Haltungen gehören. (3) Aspekte der Rationalität. Was unterscheidet Wittgensteins therapeutische Philosophie von einer Bekehrung durch rhetorisch geschickte Überredung? Es ist hilfreich, zwei Konzeptionen von Rationalität bei Wittgenstein unterscheiden: (a) Kontingente Rationalität. Wittgenstein ist sich der Kontingenz menschlicher Rationalität bewusst. Die intellektuelle Tätigkeit des Menschen ist formal und inhaltlich von den veränderlichen „Gegebenheiten“ der Lebensformen abhängig. Dazu gehören formal vor allem die charakterlichen („emotionalen“) und die sozialen Faktoren. Es gibt aber offenbar auch inhaltliche kulturelle Vorgaben (Gewissheiten), die einen kontingenten Horizont des argumentativen Diskurses eröffnen, wenn er geteilt wird, und beschränken, wenn das nicht der Fall ist. Wittgenstein exemplifiziert das am Beispiel der imaginären Diskussion, die G.E. Moore mit einem König führen muss, der glaubt, die Welt habe mit seiner Geburt erst begonnen (MS 174, 20). Wer an die menschliche Vernunft appellieren will, muss diese Beschränkungen im Auge behalten. Dieser argumentative Realismus gehört sicher zu den Verdiensten Wittgensteins. (b) Reflektierende Rationalität. Auch wenn Wittgenstein die Kontingenz der begründend-argumentativen Vernunft in Texten wie den späten Aufzeichnungen zum Thema Gewissheit deutlich heraushebt, so heißt das nicht, dass die Menschen Gefangene dieser Strukturen bleiben müssen. Das gilt selbst für eingefleischte Denkgewohnheiten, die so tief in der kollektiven Vergangenheit der Sprechergemeinschaft eingebettet sind, dass sie die grundlegenden Sprachstrukturen durchdringen (MS 113, 45f.). Denn selbst diese Denkgewohnheiten sind Schritte im Prozess der Ablösung aus instinktiven Vorgaben und damit zumindest potentielle Objekte einer reflektierenden Betrachtung. Es sind gerade die Skizzen zur existentiellen Zielsetzung seiner Philosophie, die deutlich machen, dass eine Reflexion dieser Strukturen prinzipiell möglich ist: Die Ablösung aus den instinktiven Gewissheiten der Lebensform geht mit dem Erwachen des Geistes bzw. Intellekts einher (MS 110, 299, MS 144, 22, u.a.). Die Sinnoffenheit impliziert ein Moment der Freiheit, das konstitutiv für eine Vernunft ist, die mehr

502

Andreas Koritensky

umfasst als die kontingent begründende Rationalität. Diesen Freiheitsspielraum nutzt Wittgenstein in seiner Methode überlegt und zielgerichtet – und insofern rational. Zur Sprachuntersuchung gehören daher durch Erfahrungsprozesse gesteuerte Geschicklichkeit bei der Findung (inventio) von Vergleichserzählungen und eine Urteilskraft, die Ähnlichkeiten und Differenzen mit der faktischen Sprache erkennt. Die Bekehrung des König, der glaubt, die Welt habe erst mit seiner Geburt begonnen (MS 174, 20), muss daher kein willkürlicher, irrationaler Akt der Überredung sein. Der Gesprächspartner des imaginären Königs könnte ihn auch mit den Mitteln der philosophischen Methoden Wittgensteins durch vernünftige Reflexion zu einer Modifikation seiner Weltsicht bewegen. Die Grenzen des Wertdiskurses Damit können wir uns der Frage zuwenden, warum der Wertdiskurs in Wittgensteins Aufzeichnungen nicht offener und deutlicher thematisiert wird. (1) Gefahrenquellen. Wittgenstein benennt drei potentielle Gefahrenquellen für Wertdiskurse: (a) Aufgrund der Kontingenz der menschlichen Reflexionsfähigkeit ist es denkbar, dass es zu einer Devolution der Lebensform kommt. Insbesondere in den Jahren nach dem Zweiten Weltkrieg äußert sich Wittgenstein verstärkt pessimistisch. So notiert er 1946 die Befürchtung, dass der Lebensstil in der modernen Gesellschaft die Fähigkeit zum reflektierenden Umgang mit dem Leben (Wahl) verliert. Wobei nicht mehr der Instinkt, sondern die Maschine den Takt vorgibt (MS 131, 186f.). (b) Wertdiskurse können unfruchtbar werden, weil sie in der Form von „metaphysischen“ Abstraktionen und illusionären mythologischen Bildwelten ausgedrückt werden. Diese überlagern gewissermaßen den Raum der angemessenen Sinndiskurse (c) Wenn solche irrigen Ausdrucksformen enttarnt und überwunden sind, müsste nicht nur die Sinnoffenheit freigelegt, sondern auch die Möglichkeit von Sinndiskursen geschaffen sein. Es ist aber denkbar, dass diese Möglichkeit dennoch nicht realisiert wird, weil es einer Sprachgemeinschaft nicht gelingt, geeignete Ausdrucksformen zu schaffen. Auf diese Weise versteht Wittgenstein seine eigene Situation Anfang der 1930er Jahre. Die kulturelle Gemeinschaft, der er angehört, befindet sich in einem tiefgreifenden Umbruch. Alte Ausdrucks-

Werte und Vernunft in Wittgensteins Spätwerk

503

formen der Werte erscheinen auf einmal leer, es sind aber noch keine neuen an ihre Stelle getreten. Was bleibt ist die Erfahrung eines Sinnanspruchs, der offengehalten werden muss, aber nicht mehr expliziert werden kann (MS 109, 204-206, MS 110, 12f.). (2) Voraussetzungen für Wertdiskurse. Wertdiskurse sind also prinzipiell möglich, ihre Realisierung ist aber von der Erfüllung einiger Voraussetzungen abhängig. (a) Wertdiskurse sind ihrem Wesen nach notwendig prekär. In ihrer Mitte steht das Problem, die Aufgabe, das Leben zu gestalten. Das Bemühen um eudaimonia erfordert nicht nur eine prinzipielle Ausrichtung des Lebens, sondern auch die Umsetzung in den Schritten des Alltagslebens. Wertorientiertes Leben erfordert daher sowohl eine Vorstellung vom Guten als auch die Fähigkeit, dieses durch kluges Handeln umzusetzen (phronêsis). Wertdiskurse umfassen daher in der Regel einmal große Sinnerzählungen (etwa die der Religion), aber auch erfahrungsbasierte Regeln. Sie haben Gleichnis- oder Beispielcharakter, entsprechen also Wittgensteins Vergleichsobjekten. Aristoteles weist bereits darauf hin, dass sich kluge Entscheidungsfindung nur unzureichend explizieren lässt (EN 1143b11-14). Die Explizierung von Werten hat daher immer einen approximativen Charakter. (b) Sinn, wie er von Wittgenstein aufgefasst wird, ist in einer konkreten Lebensführung zu realisieren. Es gibt nun keine abstrakten, „absoluten“ guten Lebensentscheidungen, weil es kein abstraktes, absolutes menschliches Leben gibt. Menschliche Existenz ist immer in soziale und kulturelle Zusammenhänge eingebunden. Die Lebensform beschränkt die Sinnfähigkeit und die Wahl, sie macht aber auch Alternativen als Wahlmöglichkeiten innerhalb einer Lebensform erst möglich. Zahl und Art der Alternativen zur sinnvollen Lebensführung, die Ignatius von Loyola seinen Exerzitanten im 16. Jahrhundert zur Entscheidung vorlegt, differieren von denen eines zeitgenössischen Europäers. Gleichzeitig wären die Alternativen, zwischen denen sich der Exerzitant zu entscheiden hat, ohne den spezifischen kulturellen Kontext des frühneuzeitlichen Katholizismus mit seinem Wertesystem nicht existent. Und dieses Wertesystem verfügt über ein elaboriertes System von Ausdrucksformen verschiedenster Art und Kriterien für die Entscheidungsfindung. Sinndiskurse sind daher immer traditionsgebunden, d. h. zum einen Traditionen weiterführend, zum anderen aber auch traditionskritisch. Denn die Sinnfrage setzt gerade die Ablösung aus den antrai-

504

Andreas Koritensky

nierten Lebens- und Sprachformen voraus (MS 110, 298). Wittgenstein nimmt daher an, dass sich das Wertesystem einer Kultur am ehesten im Moment von deren Niedergang ausdrücken lässt (MS 110, 12f.). Die reflexive Distanz ist dann bei noch gegebener Einbindung in diesen kulturellen Kontext am größten. Sinn ist ohne diese Krise der überkommenen Lebenspraxis gar nicht denkbar. Den Regeln der Sinndiskurse folgen Menschen nicht blind. Daher können die Wertdiskurse sowohl den Veränderungen der Lebensformen folgen als auch selbst aufgrund ihrer reflektierenden Freiheit die Lebensform weiter gestalten. Aufgrund ihrer reflektierenden Funktion sind genuine Wertdiskurse immer von einem Moment der Fragilität gekennzeichnet. Hier seien nur zwei Gründe für mögliche Krisen des Ausdrucks von Werten genannt: (i) Das „Erwachen“ und die Reflexion auf die Werte können in einer Sprachgemeinschaft ungleichzeitig ablaufen, so dass die Ausdrucksformen nicht mehr kommunikationstauglich sind. Der Preis für die Pluralisierung kann der Sprachverlust sein. (ii) Lebensformen werden nicht nur durch Reflexion modifiziert, sondern auch durch andere Faktoren. Für Wittgenstein und seine Zeit dürften hier vor allem die einschneidenden Kriegserfahrungen und die technische Entwicklung eine solche Rolle gespielt haben. Bewährte Diskurse und Entscheidungskriterien können auf diese Weise funktionsuntüchtig werden. Die reflektierende Aufarbeitung solcher Umbrüche nimmt viel Zeit in Anspruch. (3) Fazit. Wittgenstein negiert also nicht die Möglichkeit an sich, dass Sinn und Werte zum Ausdruck gebracht werden können. Prägend ist für ihn aber eine doppelte Erfahrung, die der Krise der überkommenen Sinnausdruckssysteme und die der Sinnoffenheit, die auch ohne Ausdrucksformen fortbesteht. Diese Krise ist nun ein kontingentes, prinzipiell überwindbares Phänomen. Wittgensteins späte Philosophie wird von dieser Problemstellung geformt. Die asketische Beschränkung auf den Aspekt des fiktionslosen Realismus ist zum einen ein Notbehelf, eine Minimalethik in Zeiten der Krise. Zum anderen schafft Wittgenstein damit aber auch zwei Voraussetzungen für die Wiedergewinnung von konkretisiertem und explizitem Sinn. Erstens wird so die Erfahrung der Sinnoffenheit ins Bewusstsein gehoben und vor der Überschreibung durch inadäquate Lösungsversuche bewahrt. Zweitens erlaubt es Wittgensteins Methode der Sprachuntersuchung, die Funktionsweise und die Irreduzibilität der alten Sinnausdrucksformen in den Blick zu nehmen. Wittgensteins Bemerkungen zur Religion dürfen

Werte und Vernunft in Wittgensteins Spätwerk

505

daher nicht als eklektische, irrationale Idiosynkrasien gelesen werden. Sie sind mehr oder minder glücklich gewählte einfache Modelle des Vergleichs (Sprachspiele), die es ermöglichen sollen, ein Verständnis für Wertdiskurse unter den Bedingungen der gegenwärtigen Gegebenheiten zu entwickeln. Leitbild für Rationalität ist dann nicht die theoretische, sondern die praktische Vernunft nach Art einer Klugheit (phronêsis), die das konkrete Leben in einer kontingenten Umwelt unter Berücksichtigung des Aspekts des Guten gestalten kann. Literatur Aristoteles, EN: Ethica Nicomachea, ed. L. Bywater, Oxford: Clarendon 1894 Carnap, R. u. a., 1999: Wissenschaftliche Weltauffassung – Der Wiener Kreis, in: Österreichische Philosophie von Brentano bis Wittgenstein, ed. K. R. Fischer, Wien: UTB, 125-171 Glock H.-J., 2001: Wittgenstein and Reason, in: Wittgenstein. Biography and Philosophy, ed. J. Klagge, Cambridge: Cambridge University Press, 195-220 Platon Pol.: Politeia (Opera IV), ed. J. Burnet, Oxford: Clarendon 1902 Renan, E., 1887: Histoire du Peuple d’Israel I, Paris: Gallmann Lévy Wittgenstein, L., TLP, PU: Tractatus Logico-Philosophicus. Tagebücher. Philosophische Untersuchungen (Werkausgabe 1) Frankfurt: Suhrkamp 91993 Wittgenstein, L., MS: Wittgenstein’s Nachlass. Bergen Electronic Edition, Oxford: Oxford University Press 1998-2000

Wittgenstein über Gedankenexperimente JOACHIM BROMAND, Universität Bonn

1. Einleitung Während Wittgenstein selbst seine Kritik an der traditionellen Philosophie als eines seiner Hauptanliegen erachtet („Wenn mein Name fortleben wird, dann nur als der Terminus ad quem der großen abendländischen Philosophie. Gleichsam wie der Name dessen, der die Alexandrinische Bibliothek verbrannt hat.“ (Wittgenstein 1999, 39)), werden seine diesbezüglichen Überlegungen oft als „Werbesprüche ohne argumentative Stützung“1 abgetan. Im Folgenden soll zunächst untersucht werden, ob und gegebenenfalls wie weit Teile Wittgensteins späterer Philosophiekritik argumentativ untermauert sind. Im Anschluss daran soll dann auf die Implikationen von Wittgensteins Überlegungen für die Methoden der Begriffsanalyse und insbesondere des Gedankenexperiments eingegangen werden. Ebenfalls soll gezeigt werden, dass Wittgensteins philosophiekritische bzw. metaphilosophische Überlegungen sich auch auf semantischer Ebene bewegen und sich nicht lediglich in psychologischen Erwägungen erschöpfen, wie sie in der Debatte um den therapeutischen Charakter von Wittgensteins Philosophie derzeit vornehmlich diskutiert werden. 2. Begründungsversuche von Wittgensteins Philosophiekritik Eine wohlwollende Interpretation Wittgensteins kommt angesichts der Bedeutung, die Wittgenstein selbst seinen philosophiekritischen Bemerkungen zumaß, nicht umhin, in diesen Bemerkungen wohlbegründete Positionen und eben nicht nur die besagten „Werbesprüche ohne argumentative Stützung“ zu vermuten. Entsprechend wurden in der Fachliteratur auch verschiedene Begründungsversuche von Wittgensteins Philosophiekritik vorgeschlagen. Bevor im nächsten Abschnitt ein eigener Begründungsversuch vorgelegt wird, soll zunächst kritisch auf einige der bislang vorge1

Zitiert nach Glock (2000, 266), der aber nicht dieser Meinung ist.

508

Joachim Bromand

schlagenen Versuche eingegangen werden. So unterbreitet etwa Paul Horwich (2004, 102-103) den folgenden ‚psychologischen‘ Deutungsansatz von Wittgensteins Philosophiekritik. Dieser Ansatz beruht im Wesentlichen auf den folgenden beiden Thesen: (1) (2)

Es kann keine philosophischen Erklärungen, Theorien oder Entdeckungen geben. Philosophische Verwirrung besteht in der Tendenz, Analogien im Wortgebrauch übermäßig zu strapazieren und unnötigerweise von den sich ergebenden begrifflichen Spannungen verblüfft zu sein.

Wesentlich ist an diesem Ansatz, dass nicht der Versuch unternommen wird, Wittgensteins Philosophiekritik ausgehend von seinen sprachtheoretischen Überlegungen zu begründen. Vielmehr wird lediglich darauf hingewiesen, dass unsere Motive, bestimmte philosophische Fragen zu stellen, auf sprachlichen Missverständnissen beruhen. Horwich hat hier als Beispiel etwa Fragen nach der Ontologie mathematischer Entitäten im Auge, die wir dem obigen Ansatz zufolge aufgrund des Missverständnisses stellen, die in der Mathematik verwandten singulären Terme wie „0“ seien in Bezug auf ihre Semantik analog zu Eigennamen der Umgangssprache zu verstehen (Horwich 2004, 104). Das Problem dieser und ähnlicher Begründungsversuche liegt aber auf der Hand: Ein Begründungsversuch wie der von Horwich vorgelegte erklärt nämlich bestenfalls, welche subjektiven bzw. psychologischen Ursachen dazu führen, dass wir bestimmte Fragen stellen, und zeigt somit bestenfalls, dass unsere Motive, bestimmte Fragen zu stellen, nicht adäquat begründet sind. Natürlich ist aber nicht auszuschließen, dass man auch auf ‚schlechtem‘ Wege zu ‚guten‘ Fragen gelangen kann. Um also zu zeigen, dass die Beschäftigung mit solchen Fragen vergeblich ist, reicht es nicht aus zu zeigen, dass wir uns mit den Fragen aus den ‚falschen‘ Motiven beschäftigen. Vielmehr wäre es vonnöten zu zeigen, dass die Fragestellung selbst verfehlt ist, und hierzu benötigte man eine Theorie über die tatsächliche Funktionsweise von Zahlzeichen, deren Möglichkeit (1) gerade ausschließt. Diese Kritik an solchen psychologischen Begründungsversuchen legt nahe, nach anderen Ansätzen Ausschau zu halten, die zu zeigen versuchen, dass die kritisierten philosophischen Fragen selbst in einem Sinne problematisch sind und nicht lediglich unsere Motive, solche Fragen zu stellen.

Wittgenstein über Gedankenexperimente

509

Horwich diskutiert ebenfalls einen sprachtheoretischen Begründungsversuch von Wittgensteins Philosophiekritik (Horwich 2004, 106), der aufzeigen soll, was an den kritisierten Fragen selbst problematisch ist, und damit die genannten Schwierigkeiten des psychologischen Ansatzes umgehen könnte. Dieser zweite Begründungsversuch setzt an bei Wittgensteins ‚Gebrauchstheorie‘ der Bedeutung, die einer weit verbreiteten Interpretation zufolge die Bedeutung eines Wortes mit dessen Gebrauch in der Sprache identifiziert. Die zentralen Schritte dieses Ansatzes sind die folgenden: (3) (4) (5) (6) (7)

Bedeutung = Gebrauch Daher ist jede Verwendung eines Worts abweichend von seiner üblichen Gebrauchsweise bedeutungslos. Beim Philosophieren verwenden wir aber Wörter nicht in ihrer üblichen Gebrauchsweise. Daher sind die beim Philosophieren resultierenden Theorien bedeutungslos. Wir müssen uns daher damit begnügen, unseren Drang zum Philosophieren zu beseitigen.

Während dieser Begründungsversuch nicht an der falschen Stelle ansetzt und somit von Grund auf verfehlt ist wie der obige psychologische Versuch, hängt seine Überzeugungskraft jedoch wesentlich von den herangezogenen Prämissen ab. Das Problem hier ist nun, dass eine ‚Gebrauchstheorie‘ der Bedeutung, die wie im Falle von (3) einfach in der Identifikation von Bedeutung und Gebrauch besteht, absurd ist. Dies wird etwa am folgenden Beispiel klar: Im Mittelalter wurde die Erde als Scheibe bezeichnet. Nach der Entdeckung der Kugelform der Erde bezeichnete man sie nicht mehr als Scheibe. Somit liegt hier eine Änderung des Sprachgebrauchs vor. Demgegenüber würden wir aber wohl bestreiten wollen, dass sich hier auch die Bedeutung des Wortes „Scheibe“ (bzw. die Bedeutung des entsprechenden mittelalterlichen Ausdrucks) geändert habe. Gibt es aber solche Änderungen im Gebrauch eines Worts, ohne dass sich seine Bedeutung ändert, können Bedeutung und Gebrauch nicht identisch sein. Das Problem ist hier, dass jede Bedeutungstheorie auch der Möglichkeit eines weitverbreiteten fehlerhaften Sprachgebrauchs – auch über längere Zeit hinweg – gerecht werden muss. Tatsächlich identifiziert auch Wittgenstein nicht einfach Bedeutung und Gebrauch, sondern allenfalls Bedeutung und regelkonformen Gebrauch:

510

Joachim Bromand

Eine Bedeutung eines Wortes ist eine Art seiner Verwendung. Denn sie ist das, was wir erlernen, wenn das Wort zuerst unserer Sprache einverleibt wird. Darum besteht eine Entsprechung zwischen den Begriffen ‚Bedeutung‘ und ‚Regel‘. (ÜG §§61–62)

Aufgrund dieser Überlegungen kann das folgende Zwischenfazit gezogen werden: Eine adäquate Begründung von Wittgensteins Philosophiekritik sollte aufzeigen, wieso und in welchem Sinne die von Wittgenstein kritisierten Fragen selbst problematisch sind, weshalb es nahe liegt, eine solche Begründung in Wittgensteins sprachphilosophischen Überlegungen zu suchen; ein psychologischer Ansatz wie der obige scheint demgegenüber aus den besagten Gründen von Grund auf verfehlt zu sein. Versucht man, Wittgensteins Philosophiekritik ausgehend von seiner sprachphilosophischen Position zu begründen, wird sich die Überzeugungskraft eines solchen Versuches daran bemessen, wie überzeugend die sprachphilosophischen Positionen sind, die Wittgenstein dabei zugeschrieben werden. Ein Problem eines sprachtheoretischen Begründungsversuchs von Wittgensteins Philosophiekritik könnte aber werkimmanenter Natur sein: So scheint Wittgenstein in §109 der Philosophischen Untersuchungen eine feindliche Haltung gegen Sprachtheorien im Allgemeinen einzunehmen: [W]ir dürfen keinerlei Theorie aufstellen. Es darf nichts Hypothetisches in unsern Betrachtungen sein. Alle Erklärung muß fort, und nur Beschreibung an ihre Stelle treten. (PU §109)

Im Folgenden soll dennoch versucht werden, Wittgensteins Philosophiekritik ausgehend von seiner Sprach-‚theorie‘ zu begründen. Dazu sollen im nächsten Abschnitt zunächst die relevanten Überlegungen Wittgensteins kurz vorgestellt werden. Im Anschluss daran wird dann darauf einzugehen sein, ob diese Überlegungen hinreichend überzeugend sind, um Wittgensteins Philosophiekritik argumentativ untermauern zu können. Ebenfalls wird zu fragen sein, ob bzw. wie diese Überlegungen vereinbar mit Wittgensteins vermeintlicher (sprach-)theoriefeindlichen Einstellung sind. 3. Einige sprachtheoretische Überlegungen Wittgensteins Wittgenstein entwickelt im Rahmen seiner Philosophischen Untersuchungen eine Auffassung von Begriffen, die dem nahekommt, was gegenwärtig unter der Bezeichnung der partiell definierten Prädikate diskutiert wird. Unter einem partiell definierten Prädikat versteht man dabei einen prädika-

Wittgenstein über Gedankenexperimente

511

tiven Ausdruck F, bei dem es zumindest möglich ist, dass es Objekte gibt, die weder zur Extension noch zur Antiextension von F gehören.2 In seinen Philosophischen Untersuchungen legt Wittgenstein nahe, dass viele umgangssprachliche Begriffswörter partiell definierte Prädikate sind (ohne diese Terminologie zu verwenden). Ein dort besonders intensiv diskutiertes Beispiel ist der Begriff Spiel. Dass dieses Begriffswort ein partiell definiertes Prädikat ist, erklärt Wittgenstein damit, dass dessen Bedeutung, genauer gesagt dessen Extension und Antiextension, durch Regeln bestimmt wird, die festlegen, wann das Begriffswort auf ein gegebenes Objekt korrekt angewendet werden kann und wann das Begriffswort einem solchen Objekt regelkonform abgesprochen werden kann. Somit wird festgelegt, wann ein Objekt zur Extension des Prädikats gehört und wann es zu dessen Antiextension zählt. Im Falle des Begriffs Spiel sind wir, um die Regeln explizieren zu können, welche die Extension dieses Begriffsworts festlegen, Wittgenstein zufolge auf die Präsentation paradigmatischer Beispiele sowie eine vage Ähnlichkeitsrelation angewiesen, die Wittgenstein unter der Bezeichnung der Familienähnlichkeit diskutiert: So beschreiben wir nach Wittgenstein die Extension des Begriffswortes Spiel vollständig dadurch, dass wir paradigmatische Beispiele präsentieren und dann fortfahren: „das, und Ähnliches, nennt man ‚Spiele‘“ (vgl. PU §69). Die Extension von F wird somit gewissermaßen rekursiv definiert: In einem ‚Basisschritt‘ werden paradigmatische Beispiele genannt und es wird dann in einem ‚rekursiven Schritt‘ angegeben, wie man ausgehend von den paradigmatischen Beispielen zu weiteren Objekten gelangt, die zur Extension des Begriffs zählen; im Falle des Begriffs Spiel handelt es sich dabei um diejenigen Objekte, die den bereits zur Extension von F zählenden Objekten ähneln. In analoger Weise wird die Antiextension von F rekursiv festgelegt: Zu ihr zählen die paradigmatischen Gegenbeispiele zu F sowie alle Objekte, die den bereits zur Antiextension von F zählenden Objekten ähneln (vgl. PU §75). Allerdings dürften nicht bei jedem Begriff die Regeln, die die Extension des Begriffs bestimmen, rekursiv durch die Aufzählung von paradigmatischen (Gegen-)Beispielen und einer Ähnlichkeitsrelation angegeben werden. Im Falle (nach Putnam so genannter) monokriterialer Begriffe wie 2

Die Antiextension eines Prädikats F ist dabei die Extension des prädikativen Ausdrucks nicht-F. Zu partiell definierten Prädikaten siehe etwa Soames 1999, insbesondere Kap. 6, oder Tappenden 1999, Abschnitt 3.

512

Joachim Bromand

Junggeselle etwa könnte erwartet werden, dass die den Gebrauch des Begriffs leitende Regel einfach in der Erklärung Ein Junggeselle ist ein unverheirateter Mann besteht. Ebenfalls nicht durch die Exemplifizierung anhand paradigmatischer Beispiele erläutert Wittgenstein die Regel, die ihm zufolge den Gebrauch des Wahrheitsbegriffs leitet: „Was heißt denn, ein Satz ‚ist wahr‘? ‚p‘ ist wahr = p. (Dies ist die Antwort.)“ (BGM I, Anhang III-6; vgl. auch PU §136).3 Da wie erwähnt die Überzeugungskraft einer sprachtheoretischen Begründung von Wittgensteins philosophiekritischen Thesen von der Überzeugungskraft seiner Sprachtheorie abhängt, soll hier kurz auf einen entscheidenden Zug des soeben vorgestellten Interpretationsansatzes hingewiesen werden: Wittgenstein wird oftmals als Vertreter der sog. Bündeltheorie erachtet, die zu den klassischen Beschreibungstheorien zählt. Letztere gelten jedoch aufgrund der Kritik durch den semantischen Externalismus Kripkes und Putnams als weitgehend überholt. Beruhte Wittgensteins Philosophiekritik tatsächlich auf einem solchen überholten Ansatz, wäre es um ihre Überzeugungskraft freilich nicht gut bestellt. Dem obigen Interpretationsansatz zufolge handelt es sich bei Wittgensteins Überlegungen aufgrund der zentralen Rolle, die der Exemplifizierung durch paradigmatische Beispiele zukommt (vgl. etwa PU §§51, 71), jedoch keineswegs um eine der klassischen deskriptivistischen Theorien. Vielmehr ähnelt die Begriffsexplikation durch Exemplifikation den Taufakten in der kausalen Referenztheorie Kripkes und Putnams, weshalb Wittgensteins Ansatz eher Züge einer semantisch externalistischen Position besitzt. Eine abschließende Bewertung der Tragfähigkeit der oben skizzierten begriffstheoretischen Überlegungen kann in diesem Rahmen freilich nicht erfolgen; es ist aber festzuhalten, dass Wittgensteins sprachtheoretischer Ansatz und damit seine Philosophiekritik nicht vorschnell dadurch abgetan werden können, dass man Wittgenstein dem Lager des klassischen Deskriptivismus zuordnet, dessen Position durch die Kritik von Seiten der semantischen Externalisten überholt sei. Für Wittgensteins philosophiekritische Überlegungen ist zentral, dass es im Falle der rekursiv durch Exemplifikation bestimmten Begriffe möglich 3

Die hier verfolgte Interpretation von Wittgensteins begriffs-‚theoretischen‘ Überlegungen findet sich ansatzweise etwa bei Williamson (1994, 86-87) und wird ausführlicher dargelegt in Bromand 2009.

Wittgenstein über Gedankenexperimente

513

ist, dass die entsprechenden Regeln unterbestimmt sein können in dem Sinne, dass es möglich ist, dass es Objekte gibt, für welche die Regeln nicht festlegen, ob sie zur Extension oder zur Antiextension des Begriffsworts zählen. Die Unterbestimmtheit von Regeln ist wiederum dadurch möglich, dass die Regeln nicht in trennscharfen notwendigen und zusammen genommen hinreichenden Merkmalen bestehen (wie etwa im Rahmen eines deskriptivistischen Ansatzes im Sinne Freges). So ist es nach Wittgenstein möglich, dass es Objekte gibt, die weder hinreichend Gegenständen ähneln, die berechtigterweise zu den Dingen gezählt werden, die F (z.B. Spiele) sind, noch hinreichend Gegenständen ähneln, die berechtigterweise zu den Dingen gezählt werden, die nicht-F sind. Die Unterbestimmtheit eines Prädikats resultiert somit in der Möglichkeit von Grenzfällen, denen das Prädikat weder eindeutig zu- noch abgesprochen werden kann (wobei dies nicht in einem Wissensdefizit hinsichtlich des Objektes begründet ist). Die Existenz solcher Grenzfälle wird oftmals als Zeichen der Vagheit des betroffenen Prädikats interpretiert, so dass die Unterbestimmtheit eines Prädikats dessen Vagheit nach sich zieht. Wie kommt es aber, dass wir Begriffe verwenden, deren Anwendungsregeln unterbestimmt sind? Begriffe sind nach Wittgenstein sprachliche Werkzeuge, die für bestimmte praktische Anwendungsfälle im Rahmen bestimmter Kontexte, Wittgenstein spricht auch von Lebensformen, ‚gedacht‘ bzw. konzipiert sind. Ihre Anwendung ist dabei nur für Fälle geregelt bzw. bestimmt, die in diesen Kontexten üblicherweise auftreten. Außerhalb dieses Anwendungsbereichs ist die korrekte Verwendung der Begriffe nicht geregelt (PU §142), was gerade die Unterbestimmtheit ausmacht. Ebenfalls – und deutlich überraschender – ist es möglich, dass es Objekte gibt, die sowohl regelkonform zu den Dingen gezählt werden, die F sind, als auch regelkonform zu den Dingen gezählt werden, die nicht-F sind. In diesem Sinne können Begriffe Wittgenstein zufolge auch überbestimmt sein. Sowohl die Unter- als auch die Überbestimmtheit von philosophisch zentralen Begriffen können nach Wittgenstein philosophische Probleme nach sich ziehen: (i) (ii)

Einige philosophisch zentrale Begriffe (deren Regeln unterbestimmt sind) besitzen keine scharf begrenzten Extensionen bzw. keine exakten Definitionen. Im Falle einiger Begriffe (mit unterbestimmten Regeln) können Gedankenexperimente scheitern.

514

(iii)

Joachim Bromand

Einige philosophisch zentrale Begriffe (deren Regeln überbestimmt sind) besitzen keine konsistenten Explikationen (evtl. Wahrheit, Erfüllung, Referenz, Wissen).

Bevor ab dem nächsten Abschnitt detailliert auf (ii) eingegangen wird, soll zuvor kurz ein Blick auf (i) geworfen werden; das Phänomen der Überbestimmtheit und die entsprechende Konsequenz (iii) können an dieser Stelle nicht weiter verfolgt werden.4 Aus Punkt (i) zieht Wittgenstein die Konsequenz, dass die Suche nach exakten Definitionen in Fällen von unscharf begrenzten Begriffen problematisch ist. Als Beispiele solcher Begriffe erörtert Wittgenstein etwa folgende: Wenn aber im Original die Farben ohne die Spur einer Grenze ineinanderfließen, – wird es dann nicht eine hoffnungslose Aufgabe werden, ein dem verschwommenen entsprechendes scharfes Bild zu zeichnen? […] Und in dieser Lage befindet sich z.B. der, der in der Ästhetik oder Ethik nach Definitionen sucht, die unseren Begriffen entsprechen. (PU §77)

Ebenfalls betroffen von dieser Analyse dürften weitere philosophisch relevante Begriffe wie Mensch oder Person sein. Eine Konsequenz dieser Überlegungen könnte so etwa in der Vergeblichkeit des Versuchs bestehen, nur mit Hilfe einer Begriffsanalyse klären zu wollen, wo genau die Grenzen der Begriffe Mensch oder Person verlaufen, um zu bestimmen, wo das Menschsein anfängt und wo es aufhört. So erlaubt uns der unscharf begrenzte umgangssprachliche Begriff Mensch zwar zumeist, Menschen von anderen Lebewesen oder Objekten zu differenzieren. Der Wittgensteinschen Auffassung zufolge wäre es aber vergebens, etwa im Rahmen von bioethischen Debatten um die Zulässigkeit von Abtreibungen, trennscharfe Kriterien für das Menschsein aufgrund einer Analyse des Begriffs Mensch finden zu wollen. Allerdings sollten die Folgen dieser Überlegungen auch nicht überschätzt werden. So schließen Wittgensteins Überlegungen nicht aus, dass ein vager Begriff durch Kriterien expliziert werden kann, die genauso vage sind wie der Begriff selbst (so dürfte es sich etwa bei der Explikation des Begriffs Schimmel durch weißes Pferd verhalten). Wenn es auch keine scharfe Kopie eines unscharfen Bildes geben mag, spricht doch nichts gegen eine unscharfe Kopie eines unscharfen Bildes. 4

Eine ausführlichere Behandlung beider Phänomene findet sich in Bromand 2009.

Wittgenstein über Gedankenexperimente

515

Bevor auf die philosophiekritischen Konsequenzen von Wittgensteins sprachphilosophischen Überzeugungen für die Methode des Gedankenexperiments eingegangen wird, soll zuvor eine etwaige werkimmanente Schwierigkeit des hier vorgeschlagenen Interpretationsansatzes ausgeräumt werden, auf die am Ende des letzten Abschnitts hingewiesen wurde. Dort ging es um die vermeintlich theoriefeindliche Haltung Wittgensteins, die er insbesondere in §109 der Philosophischen Untersuchungen formuliert („[W]ir dürfen keinerlei Theorie aufstellen.“, s. o.). Ist diese Haltung vereinbar mit sprachtheoretischen Feststellungen, wie sie in diesem Absatz getroffen wurden? Dem ist so. Wittgenstein fährt nämlich im Anschluss an die soeben zitierte Passage fort: „Alle Erklärung muß fort, und nur Beschreibung an ihre Stelle treten.“ Offenbar hat Wittgenstein somit keine Einwände gegen sprach-‚theoretische‘ Feststellungen im Sinne von Beschreibungen einzelner Fallbeispiele. Vielmehr scheint er lediglich theoretische Verallgemeinerungen abzulehnen. Tatsächlich behauptet Wittgenstein nicht, dass die Semantik eines jeden Begriffs der des Spielbegriffs gleicht. Wittgenstein entwickelt somit keine allgemeine Begriffstheorie. Auch wenn aufgrund von Wittgensteins Beobachtungen somit keine allgemeinen Aussagen über alle Begriffe getroffen werden können, gelten Entsprechungen von Wittgensteins Feststellungen über den Spielbegriff jedoch für alle Begriffe, deren Semantik der des Spielbegriffs hinlänglich ähnelt (etwa in Hinblick auf die Art und Weise, wie wir den Begriff erwerben). Wenn Wittgensteins Beobachtungen zum Spielbegriff somit auch nicht als allgemeine Begriffstheorie verstanden werden sollten, spricht doch nichts dagegen, seine Bemerkungen als eine Art Argumentationsschema zu verstehen, das auf jeden Begriff anwendbar ist, der sich hinsichtlich seiner Semantik nicht wesentlich vom Spielbegriff unterscheidet. 4. Wittgenstein und Gedankenexperimente Der in Hinblick auf Gedankenexperimente relevante Punkt (ii) ergibt sich aus Wittgensteins Überlegungen zur Unterbestimmtheit, denen zufolge es nicht auf jede philosophische Frage eine Antwort gibt: Wendet man nämlich einen unterbestimmten Begriff F auf einen Fall a an, der nicht zu den Objekten gehört, für den die Regeln festlegen, ob F oder nicht-F gilt, besitzt die Frage „F(a)?“ nach Wittgenstein keine richtige Antwort: Es gilt weder F(a) noch nicht-F(a), so dass aufgrund von Wittgensteins deflationärem Wahrheitsverständnis (BGM I, Anhang III-6; PU §136) der Satz

516

Joachim Bromand

F(a) weder wahr noch falsch ist. Wichtig ist hier, dass es Wittgenstein lediglich auf extensionaler Ebene um die Frage der regelkonformen Anwendbarkeit von Begriffen bzw. der Wahrheit der resultierenden Sätze geht und dass die obigen Überlegungen keinen Wiederbelebungsversuch des verifikationistischen Sinnkriteriums des Wiener Kreises darstellen. Es geht also nicht darum, entsprechenden Prädikationen auf intensionaler Ebene den Sinn abzusprechen, vielmehr soll lediglich auf extensionaler Ebene die Wahrheitswertlosigkeit solcher Aussagen bzw. die Nichtbeantwortbarkeit entsprechender Fragen behauptet werden. Von der Nichtbeantwortbarkeit betroffen sind dabei eben solche Fragen, in denen es um die Anwendbarkeit eines unterbestimmten Prädikats auf entsprechende Grenzfälle geht. Insbesondere Gedankenexperimente, in denen die Frage nach der Anwendbarkeit eines solchen Begriffs unter radikal veränderten Rahmenbedingungen bzw. um die Anwendbarkeit auf gänzlich untypische Anwendungsfälle aufgeworfen wird, für die der Begriff nicht ‚gedacht‘ bzw. nicht hinreichend bestimmt (d. h. unterbestimmt) ist, sind hiervon betroffen. Wittgenstein verdeutlicht dies am Beispiel des folgenden Gedankenexperiments: Ich sage: „Dort steht ein Sessel.“ Wie, wenn ich hingehe und ihn holen will, und er entschwindet plötzlich meinem Blick? – „Also war es kein Sessel, sondern irgendeine Täuschung.“ – Aber in ein paar Sekunden sehen wir ihn wieder und können ihn angreifen, etc. – „Also war der Sessel doch da und sein Verschwinden war irgend eine Täuschung.“ – Aber nimm an, nach einer Zeit verschwindet er wieder, – oder scheint zu verschwinden. Was sollen wir nun sagen? Hast du für solche Fälle Regeln bereit, – die sagen, ob man so etwas noch „Sessel“ nennen darf? Aber gehen sie uns beim Gebrauch des Wortes „Sessel“ ab; und sollen wir sagen, daß wir mit diesem Wort eigentlich keine Bedeutung verbinden, da wir nicht für alle Möglichkeiten seiner Anwendung mit Regeln ausgerüstet sind? (PU §80; vgl. Z §350)

Bereits im Blauen Buch macht Wittgenstein Begriffe, deren Anwendung nicht für alle möglichen Fälle geregelt ist, für das Entstehen philosophischer Probleme verantwortlich (BB, 74). Diese Position behält er auch im Rahmen seiner Philosophischen Untersuchungen bei. In diesem Sinne stellt er etwa fest: „[D]ie philosophischen Probleme entstehen, wenn die Sprache feiert“ bzw. außerhalb ihrer üblichen, ‚werktäglichen‘ Verwendungsweise gebraucht wird (PU §38 vgl. §§116, 119, 132–133). Welche weiteren philosophischen Fragen könnte Wittgensteins Kritik nun betreffen? Sainsbury (1990) erörtert entsprechende Beispiele in Wittgensteins Sinne. So erlauben uns etwa die Begriffe Erdbeere und Himbeere jeweils zwischen Erdbeeren bzw. Himbeeren und anderen Früchten zu dif-

Wittgenstein über Gedankenexperimente

517

ferenzieren, solange wir die Begriffe in alltäglichen Kontexten verwenden (wie etwa auf dem Markt), in denen nur die gängigen Sorten vorzufinden sind. Im Rahmen eines solchen begrenzten Gegenstandsbereichs funktionieren die Begriffe wie (etwa von Frege) erwünscht und unterteilen den Gegenstandsbereich in zwei ‚Hälften‘. Diese Leistung wird Wittgenstein zufolge nicht dadurch geschmälert, dass die Begriffe nicht mehr unbedingt in dieser Weise funktionieren, wenn der Gegenstandsbereich erweitert wird. So ist es zumindest denkbar, mit gentechnischen Mitteln eine Reihe von Pflanzen zu erzeugen, von denen die erste eine Erdbeerpflanze ist, die letzte ein Himbeerstrauch, so dass deren Zwischenstufen sich hinsichtlich ihrer Früchte paarweise jeweils nur geringfügig unterscheiden. Wendet man die Begriffe Erdbeere und Himbeere auf einen Gegenstandsbereich an, der auch solche gentechnisch manipulierten Früchte enthält, werden sich für beide Begriffe Grenzfälle finden, so dass der Begriff einem entsprechenden Zweifelsfall weder regelkonform zu- noch abgesprochen werden kann. Die umgangssprachlichen Begriffe unterteilen einen derart erweiterten Gegenstandsbereich nicht disjunkt und sind sozusagen nicht für solche Anwendungsfälle ‚gedacht‘. Dies schließt nicht aus, dass die Gebrauchsregeln für die Anwendung auf Grenzfälle erweitert werden können. Aus einer solchen Erweiterung resultierte dabei freilich ein anderer, in der fraglichen Hinsicht weniger vager Begriff. Ob ein unterbestimmtes Prädikat auf einen Grenzfall zutrifft, kann somit aber nur willkürlich festgelegt und nicht durch eine Begriffsanalyse des ursprünglichen Begriffs (vor der Erweiterung seiner Verwendungsregeln) eruiert werden. Anwärter für eingehend diskutierte philosophische Fragen, die sich aus Wittgensteins Perspektive als Scheinprobleme erweisen könnten, sind etwa auch Fragen nach diachroner Identität bzw. Persistenz. Ein berühmtes Beispiel eines Gedankenexperiments, das im Rahmen entsprechender philosophischer Diskussionen immer wieder herangezogen wurde, dreht sich um das Schiff des Theseus, bei dem auf hoher See alle Planken durch neue ersetzt, die alten aber verwahrt werden, um daraus wieder ein Schiff nach dem ursprünglichen Bauplan zu bauen. Auch in diesem Falle könnte man vermuten, dass die Regeln, welche die Bedeutung bzw. die korrekte Anwendung von identisch auf zeitlich ausgedehnte Objekte festlegen, nicht hinreichend spezifiziert sind, um die Frage zu beantworten, welches der beiden resultierenden Schiffe identisch mit dem ursprünglichen ist. Dieser Fall soll im Folgenden eingehender analysiert werden.

518

Joachim Bromand

5. Skizze eines Fallbeispiels Wie genau lässt sich nun mit Wittgensteins Verständnis vager Ausdrücke der Fall von Theseus’ Schiff erklären? Eine Möglichkeit, wie Wittgensteins allgemeine Überlegungen auf diesen speziellen Fall angewendet werden könnten, soll im Folgenden angedeutet werden, wobei es bei einer Skizze der wichtigsten Argumentationsschritte bleiben muss. Zunächst einmal handelt es sich beim Ausdruck x ist dasselbe Schiff wie y um einen Relationsausdruck, der nicht auf einzelne Gegenstände, sondern auf Paare von Gegenständen zutrifft. Typische Beispiele für Paare, die unter den Relationsausdruck ist dasselbe Schiff wie fallen, könnten Paare von Schiffen sein, die (1) oder (2) klarerweise erfüllen: (1) (2)

x entsteht aus y durch die Ersetzung höchstens einer begrenzten Anzahl von Teilen (oder umgekehrt) x und y bestehen zum selben Zeitpunkt aus denselben Teilen

Typische Gegenbeispiele könnten etwa Paare sein, die keine der beiden schwächeren Bedingungen (1') und (2') erfüllen, (1') (2')

zwischen x und y besteht eine raumzeitliche Kontinuität x und y bestehen aus denselben Teilen

also Paare, die weder zu einem Zeitpunkt aus denselben Teilen bestehen noch raumzeitlich kontinuierlich sind. Betrachten wir das Paar bestehend aus Theseus’ ursprünglichem Schiff und dem Schiff mit den erneuerten Planken. Dieses Paar ähnelt den typischen Beispielen für Paare, die unter die Relation fallen, insofern nicht, als es kein klares Beispiel für (1) ist. Klare Beispiele für diese Bedingung sind Beispiele komplexer Gegenstände mit vielen Bestandteilen, von denen nur sehr wenige Bestandteile etwa zu Reparaturzwecken ausgetauscht wurden. Dies trifft in diesem Falle, bei dem im Vergleich zu Theseus’ ursprünglichem Schiff alle und somit sehr viele Bestandteile ausgetauscht werden, sicherlich nicht zu. Bei diesem Paar handelt es sich aber auch nicht um ein typisches Gegenbeispiel, da es in diesem Falle sehr wohl eine raumzeitliche Kontinuität zwischen dem ursprünglichen Schiff des Theseus und dem Schiff mit den erneuerten Planken gibt. Ähnlich kann auch im Falle von Theseus’ ursprünglichem Schiff und dem aus den alten Planken rekonstruierten Schiff gezeigt wer-

Wittgenstein über Gedankenexperimente

519

den, dass sie weder den typischen Beispielen noch den typischen Gegenbeispielen von Paaren ähneln, die unter die Relation x ist dasselbe Schiff wie y fallen. Beide Paare bilden somit Grenzfälle für die besagte Relation, in denen eben nicht geregelt ist, ob der Relationsausdruck auf die Paare korrekterweise angewandt werden kann oder nicht. 6. Ausblick: Wie weit reichen die Konsequenzen von Wittgensteins Überlegungen? Wie das Beispiel der diachronen Identität zeigt, gelten Wittgensteins Überlegungen dabei nicht nur für (einstellige) Prädikate, sondern lassen sich auch auf (mehrstellige) Relationsausdrücke übertragen. Ebenso unbestimmt wie der Wahrheitswert eines Satzes, der in der Anwendung eines vagen Prädikats auf einen entsprechenden Grenzfall besteht, scheint aber auch der Referent von funktionalen Ausdrücken und singulären Kennzeichnungen sein zu können. Auch Funktionsausdrücke wie der Vater von scheinen nur dann unproblematisch auf eine Person zu referieren, wenn sie auf typische Fälle angewandt werden; auch hier sind aber Grenzfälle von Personen zumindest denkbar, in denen die Regeln für den Begriff Vater bzw. für den Funktionsausdruck der Vater von nicht festlegen, wer die betreffende Person ist. Vorstellbar ist so beispielsweise, dass etwa die Hälfte des genetischen Materials einer Samenzelle durch das entsprechende genetische Material einer anderen Person ausgetauscht wird. Im Falle eines mit einer solchen Samenzelle gezeugten Kindes wäre unbestimmt, wer ‚der Vater‘ dieser Person wäre. Ein ähnliches Beispiel einer Kennzeichnung, deren Referent unbestimmt ist, könnte etwa auch der Ausdruck der Sinn des Lebens darstellen. Auch hier liegt nahe, dass der Funktionsausdruck der Sinn von in einem Kontext verwendet wird, für den seine Verwendung nicht geregelt wurde. Ist dem so, ähnelte die Frage nach dem Sinn des Lebens der Frage nach dem Resultat der Division einer Zahl durch 0. Dabei ist unsere Schwierigkeit, mit solchen Fragen umzugehen, nicht darin begründet, dass etwa unsere mathematischen Fähigkeiten zu beschränkt wären, um das korrekte Resultat einer solchen Division durch 0 zu berechnen – vielmehr gibt es in diesem Fall keine Tatsache bzw. kein Ergebnis der Division, das wir im Rahmen einer richtigen Antwort auf die Frage angeben könnten. Somit zeigen solche Fragen auch keine Begrenzung unseres Wissens auf, da es hier eben keine Tatsachen gibt, von denen wir wissen könnten. Der Wittgenstein-

520

Joachim Bromand

schen Position zufolge handelt es sich hier lediglich um Scheinprobleme, also um Fragen, die lediglich den Anschein haben, beantwortet werden zu können, nicht aber um Probleme, die tatsächlich lösbar wären. Der so verstandene Wittgenstein ist mit seiner Diagnose der Vergeblichkeit bestimmter philosophischer Gedankenexperimente nicht allein. Auch Derek Parfit erachtet einige der oben als unbeantwortbar bezeichneten Fragen als leer (Parfit 1984, 214). Eine ähnliche Diagnose wie im Falle von Theseus’ Schiff liegt Parfit zufolge (Parfit 1984, 213f.) auch im Falle vieler Gedankenexperimente zur personalen Identität nahe, insbesondere in sog. Fission-Fällen, wie sie bisweilen in der Philosophie des Geistes diskutiert werden. Hier weist auch Parfit auf die Möglichkeit von unbestimmten (wahrheitswertlosen) Behauptungen personaler Identität hin. Auch hier resultierte die Unbestimmtheit des Satzes aus der Vagheit des Identitätsprädikats. Wie im Falle von Wittgensteins Sessel-Gedankenexperiment (PU §80, s. o.) handelte es sich aber auch bei unbestimmten (wahrheitswertlosen) Behauptungen personaler Identität nicht um ein Wissensdefizit unsererseits; vielmehr gibt es im Sinne der Auffassung Wittgensteins nichts, von dem wir wissen könnten. Auch Quine und Rescher (vgl. etwa Rescher 2005, 151ff.) teilen die obige Diagnose von vielen Gedankenexperimenten zu Fragen der personalen Identität: „The method of science fiction has its uses in philosophy, but […] I wonder whether the limits of the method are properly heeded. To seek what is ‘logically required’ for sameness of person under unprecedented circumstances is to suggest that words have some logical force beyond what our past needs have invested them with“ (Quine 1972, 490). Mit den vorgestellten Überlegungen sollen allerdings nicht Gedankenexperimente per se als philosophisch unfruchtbar erwiesen werden. Vielmehr ist zu erwarten, dass Gedankenexperimente, in denen es um Szenarien geht, die denen unserer alltäglichen Lebenswelt ähneln, erfolgreich in dem Sinne sein könnten, dass sie Fragen aufwerfen, die mit Hilfe unserer bisherigen Begrifflichkeit beantwortet werden können. Beispiele solcher alltagsnahen Gedankenexperimente stellen etwa die berühmten GettierBeispiele oder das so genannte Gödel-Schmidt-Gedankenexperiment Kripkes dar. Auch schließen Wittgensteins Überlegungen nicht die Möglichkeit erfolgreicher Science-Fiction-Gedankenexperimente wie etwa das von Putnams Zwillingserde aus. Wichtig ist hier, dass dieses Gedankenexperiment nicht in irreduzibler Weise von Science-Fiction-Szenarien Gebrauch macht: Die Science-Fiction-Szenarien hier sind im Wesentlichen

Wittgenstein über Gedankenexperimente

521

verzichtbar und nahezu alle Aspekte des Zwillingserde-Gedankenexperiments können von entsprechenden Gedankenexperimenten ohne ScienceFiction-Elemente erfasst werden. Selbst irreduzible Science-FictionSzenarien, die unseren augenblicklichen Lebensumständen in Hinblick auf den Sprachgebrauch noch hinreichend ähneln, wären aus der obigen Perspektive Wittgensteins unproblematisch. Problematisch sind lediglich solche Gedankenexperimente, in denen radikal neue Szenarien für die Anwendung vager Begriffe und somit eventuell Grenzfälle für solche Begriffe konstruiert werden und dann die Frage aufgeworfen wird, ob das fragliche vage Prädikat auf den im Gedankenexperiment konstruierten Grenzfall zutrifft. Unsere Schwierigkeit, solche Fragen zu beantworten, ist dabei nicht darin begründet, dass unsere kognitiven Möglichkeiten zu begrenzt sind, um die Antworten in Erfahrung zu bringen – vielmehr gibt es in solchen Fällen keine richtigen Antworten, von denen wir wissen könnten. Der Wittgenstein’schen Position zufolge handelt es sich hier lediglich um Scheinprobleme, die durch sprachtheoretische Überlegungen aufzulösen sind, worin nach Wittgenstein auch das hauptsächliche Anliegen der Philosophie bestehen sollte (vgl. PU §§109, 133). 7. Literatur Bromand, J., 2009: Grenzen des Wissens, Paderborn: Mentis Glock, H.-J., 2000: Wittgenstein Lexikon, Darmstadt: Wissenschaftliche Buchgesellschaft Horwich, P., 2004: Wittgenstein’s metaphilosophical development, in: Wittgenstein’s Lasting Significance, hg. v. M. Kölbel and B. Weiss, London / New York: Routledge, 100-109 Parfit, D., 1984: Reasons and Persons, Oxford: Oxford University Press Quine, W.V., 1972: Review of Milton K. Munitz (Ed.), ‘Identity and Individuation’, The Journal of Philosophy 69, 488-497 Rescher, N., 2005: What If? Thought Experimentation in Philosophy, New Brunswick / London: Transaction Publishers Sainsbury, R.M., 1990: Concepts without Boundaries (Inaugural Lecture, King’s College London), wiederabgedruckt in: Vagueness, hg. v. R. Keefe and P. Smith, Cambridge, Mass. / London: MIT-Press 1997, 251-264 Soames, S., 1999: Understanding Truth, Oxford: Oxford University Press

522

Joachim Bromand

Tappenden, J., 1999: Negation, Denial, and Language Change in Philosophical Logic, in: What is Negation? Hg. v. D.M. Gabbay and H. Wansing, Dordrecht: Kluwer, 261-298 Williamson, T., 1994: Vagueness, London / New York: Routledge Wittgenstein, L., 1999: Denkbewegungen. Tagebücher 1930–1932, 1936–1937, hg. v. I. Somavilla, Frankfurt a. M.: Fischer

Wittgenstein, Ethics and Therapy EDWARD HARCOURT, University of Oxford

This paper tries to bring together some thoughts about the ethical meaning of Wittgenstein’s work and the nature of psychotherapy. It thus also touches, though indirectly, on the idea, prominent in recent Wittgenstein commentary, that philosophy is ‘therapeutic’. The approach taken to Wittgenstein’s work may be thought of as literary as well as philosophical, for two reasons: first, it relies in its understanding of Wittgenstein’s work partly on ideas – the ideas of allegory and symbol – which have no place in the work itself. Secondly, insofar as it relies on a distinction between the ethical meaning Wittgenstein’s work had for Wittgenstein and the ethical meaning it may or may not have simpliciter, it sees his work not simply as a set of thoughts or arguments but also as the self-expression of a person. 1. Notwithstanding the ‘Lecture on Ethics’ (Wittgenstein 1929 (1993a)), Wittgenstein’s work is almost never about ethics: ethics is mentioned only rarely in the Tractatus, and in Philosophical Investigations not at all. Nonetheless it is hard to escape the sense that Wittgenstein’s claim about the Tractatus that ‘the point of the work is an ethical one’ (1969, 35) is no less true of the later work than it is of the Tractatus itself. Saying why it’s true thus requires some care. The ethical significance of Wittgenstein’s work (from whatever period) relates in the first place not to the work’s content, but to the fact that it exemplifies the practice of philosophy, no matter which particular problems it is addressing: to put it dogmatically for the moment, Wittgenstein’s philosophy had ethical significance at least to him because he saw its practice as a spiritual exercise, that is, as a discipline whose aim is to transform the practitioner’s moral being for the better.1 Locating the work’s ethical sig-

1

‘As strolling, walking and running are bodily exercises, so every way of preparing and disposing the soul to rid itself of all the disordered tendencies, and, after it is rid, to seek and find the Divine Will as to the management of one’s life for the salvation of

524

Edward Harcourt

nificance primarily not in what it says but in the practice in which its sayings result is consistent with what little Wittgenstein does say about ethics,2 because of course what he most famously says is that the ethical is something about which nothing can be said (Wittgenstein 1922 (1974), 6.421). That Wittgenstein saw philosophy as a spiritual exercise surfaces at various points in his work, for example in the Big Typescript, where we find the idea that ‘work on philosophy is …a kind of work on oneself’ (Wittgenstein 1936 (1993b), 161). But it’s not (I hope) the idea of a spiritual exercise itself, of which there are many different kinds, that needs explaining: what needs explaining is how Wittgenstein could come to see the practice of philosophy as one. The answer follows from two thoughts of Wittgenstein’s: first, that philosophical confusion is a mark of personal badness and, secondly, that the practice of philosophy remedies confusion. It follows that philosophy when practised successfully makes one better – here to be understood not as ‘more healthy’ but as ‘less bad’. The second of these ideas - that philosophy unravels confusion - is widely (and correctly) seen as a constant in Wittgenstein’s work, early and late: ‘Philosophy does not result in “philosophical propositions”, but in the clarification of propositions’, (1922 (1974) 4.112); ‘The problems are solved, not by giving new information, but by arranging what we have always known. Philosophy is a battle against the bewitchment of our intelligence by means of language’ (1958, 98). It’s the first idea therefore that requires more by way of comment. The idea that philosophical confusion is a mark of badness receives different expressions at different phases of Wittgenstein’s career. In the Big Typescript, the thought that philosophy engages the moral (as opposed to the merely intellectual) aspects of oneself is clear because Wittgenstein points to the ‘resistances of the will’ that need to be overcome if it is to succeed (1936 (1993b), 161). In the 1914-16 Notebooks, by contrast, the organizing idea is that of ‘being in agreement with the world’, a state of moral goodness, Wittgenstein makes clear, because one can only enjoy it if one has a clear conscience. ‘When my conscience upsets my equilibrium, then I am not in agreement with Something. But what is this? Is it the world?’ (Wittgenstein 1979, 8.7.16). The same thought is expressed in the soul, is called a Spiritual Exercise.’ St Ignatius of Loyola, Spiritual Exercises 1, tr. E. Mullan SJ, www.nwjesuits.org/JesuitSpirituality/ Exercises/SpEx001_020.html. 2 And not just about ethics: ‘Philosophy is not a body of doctrine but an activity’, Wittgenstein 1922 (1974), 4.112.

Wittgenstein, Ethics and Therapy

525

Engelmann’s report on a conversation with Wittgenstein from the same year: ‘If I am unhappy and know that my unhappiness reflects a gross discrepancy between myself and life as it is, I have solved nothing; I shall be on the wrong track and I shall never find a way out of the chaos of my emotions and thoughts so long as I have not achieved the supreme and crucial insight that that discrepancy is not the fault of life as it is, but of myself as I am’ (Engelmann 1967, 76-7, my italics). But via its unravelling of misunderstandings - which replaces the complication of the confused philosopher’s consciousness with a simplicity borrowed from the world that it’s a now unconfused consciousness of - ‘agreement with the world’ is precisely what the activity of philosophy brings about. ‘All the propositions of ordinary language, just as they stand, are in perfect logical order’ (Wittgenstein 1922 (1974) 5.5563); ‘[philosophy’s] results must be simple, but its activity is as complicated as the knots it unravels’ (1936 (1993b), 183); ‘It is the business of philosophy, not to resolve a contradiction by means of a mathematical or logico-mathematical discovery, but to make it possible for us to get a clear view of the state of mathematics that troubles us: the state of affairs before the contradiction is resolved. … The aspects of things that are most important for us are hidden because of their simplicity and familiarity’ (1958, 125, 129). The unfavourable comparison between the simplicity of the world and the complexity of the confused philosophical consciousness connects with the influence on Wittgenstein – which Wittgenstein acknowledged (1980, 16) - of the architect Adolf Loos. In his 1908 essay ‘Ornament and Crime’ (Loos 1931, 79-91), Loos had argued that – subject to certain conditions – decorative elements with no structural role were a crime in architecture. To work in an undecorated idiom is therefore a distinctively moral requirement. But, allowing for the differences between philosophy and architecture, working in an undecorated idiom is just what Wittgenstein aspired to do: the theme makes itself felt in the Investigations in various ways (most clearly at Investigations 217, ‘the definition a kind of ornamental coping that supports nothing’, but also in disparaging references to redundant parts of mechanisms, e.g. ‘a wheel that can be turned though nothing else moves with it’, (1958, 271)), and is highly insistent in the Tractatus. For example, expressing identity of object by identity of sign rather than by a sign of identity (5.53) is just one example of the way a perspicuous notation is also an undecorated one (in the example, the decoration is ‘=’). That language ‘is in perfect order as it is’, and thus that constructing a perspicuous (undecorated) notation is the key to exposing philosophical problems as

526

Edward Harcourt

pseudo-problems is, on its own, nothing to do with Loos, or with ethics. But the Loosian moralization of plainness enables Wittgenstein to read every detail of his philosophizing as satisfying a moral imperative. Clearly much of the evidence for this understanding of the ethical significance – as Wittgenstein saw it – of his work comes from the philosophical work itself. But the understanding can be reinforced by appeal to biographical material, for here there is abundant evidence that Wittgenstein thought of himself as bad or not ‘decent’; thought of his badness as enforcing his separation from a better and simpler and more wholesome world; and craved ‘redemption’ understood as readmission to such a world. Evidence for the first point is Malcolm’s recollection that ‘during his service in the First World War, and frequently thereafter, Wittgenstein expressed his need to become ‘a different man’, ‘a decent human being’ (Malcolm 1984, xix);3 Wittgenstein himself speaks of his ‘impurity’ (1980, 35; cp. 1975, Foreword: ‘[the author] cannot free [the book] of these impurities [specified as ‘vanity etc.’] further than he himself is free of them’). Evidence for his craving for readmission to a better and simpler world is surely his abandonment of philosophy in the 1920s to become a gardener and then a village schoolteacher, ‘his decision’, as his sister Hermine put it, ‘to choose a completely ordinary occupation’ (H. Wittgenstein, 1984, 4). But the following circa 1944 passage is especially telling: Someone to whom it is given in [‘ultimate’] distress to open his heart instead of contracting it, absorbs the remedy into his heart. Someone who in this way opens his heart to God in remorseful confession opens it for others too. He thereby loses his dignity as someone special & so becomes like a child. That means without office, dignity & aloofness from others. You can open yourself to others only out of a particular kind of love. Which acknowledges as it were that we are all wicked children. It might also be said: hate between human beings comes from our cutting

3

Cp. letter to Engelmann, 16.1.1918, ‘I am now slightly more decent. By this I only mean that I am slightly clearer in my own mind about my lack of decency’ (Engelmann 1967, 12). Consider also Wittgenstein’s ‘confessions’ to clear ‘his oppressive burden of guilt’ (Pascal 1984, 34ff). I regard as more significant still Pascal’s rhetorical question: ‘Did he [Wittgenstein] know how much he inhibited others, while his dearest wish was for them to behave in a natural way?’ (1984, 29). I strongly suspect the answer is ‘yes’, and Wittgenstein’s feeling of not being ‘decent’ stemmed in part from his awareness that it was he himself who got in the way of his being able to relate to others in the ‘ordinary’ way he so idealized.

Wittgenstein, Ethics and Therapy

527

ourselves off from each other. Because we don't want anyone else to see inside us, since it's not a pretty sight in there (1980, 52-3; my italics).4

The passage makes it clear that, to Wittgenstein, what confessing one’s own badness specifically remedies is the state of being cut off from others. Now for all that has so far been said, philosophy could have the kind of ethical significance I’ve argued it had for Wittgenstein no matter which particular philosophical problems it addressed itself to. But to leave things there would be to miss out something important. A leading topic of Wittgenstein’s work is mind’s place in the world, and a leading ‘thesis’ of Wittgenstein’s that mentality is an aspect of the totality of our lived relations with the world, and so as ‘external’ as any part of the world it has as its object - not something inner whose relations to the world need to be puzzled over. Another prominent Wittgensteinian topic is the relation between language and world. Here, the ‘thesis’ is that the Wortsprache – something that can be sandwiched between the covers of a dictionary and grammar book – is an abstraction from a complex set of verbal and non-verbal interactions of humans with their environments and with each other, and only if we mistakenly identify the abstraction with language itself will we see language as something whose relation to the world is problematic (1958, 435 ventriloquizes the wrong-headed question: ‘“How do sentences manage to represent?”’). Yet another topic is ‘other minds’: only by mistakenly abstracting ‘mind’ and ‘body’ from the totality of embodied human experience can we get so far as to worry that there might be nothing but myself plus automata. The ethical significance Wittgenstein’s work had for him extends, I would argue, to his very choice of these topics. As Cavell says, ‘[T]he correct relation between inner and outer, between the soul and its society, is the theme of the Investigations as a whole … and also its moral’ (1979, 329). A comparison may be helpful in order to spell this connection out further. In his essay ‘The Effectiveness of Symbols’ (Lévi-Strauss 1977, 186205), Lévi-Strauss describes the way the shamans of the Panamanian Cuna 4

For more on redemption see 1980, 38: ‘What inclines even me to believe in Christ's resurrection? … If he did not rise from the dead, then he decomposed in the grave like every human being … & we are once more orphaned & alone. And have to make do with wisdom & speculation. … But if I am to be REALLY redeemed, – I need certainty – not wisdom, dreams, speculation – and this certainty is faith. And faith is faith in what my heart, my soul, needs, not my speculative intellect. For my soul, with its passions, as it were with its flesh & blood, must be redeemed, not my abstract mind.’

528

Edward Harcourt

Indians use a ‘long incantation’ to help women experiencing difficulties in childbirth. The poem describes a contest between on the one hand the shaman and his spirit helpers and, on the other, ‘Muu’, meaning the spirit responsible for the baby’s getting stuck but also - in line with the poem’s increasingly ‘rapid oscillation between mythical and physiological themes, as if to abolish in the mind of the sick woman the distinction’ between them - the womb itself (1977, 193). The shaman and his helpers enter the woman’s body, fighting their way past all manner of obstacles, calling on reinforcements – ‘the “clearers of the way”, Lords-of-the–burrowinganimals, such as the armadillo’ (1977, 196) – and eventually coming back out again (in the poem) with the baby. The goings-on described in the poem are of course an allegory of labour. Reciting it is not an act of magic, but by the labouring mother’s imaginative engagement with the labouring spirits and with the poem’s imaginary geography of her own body, her labour is made easier, and a baby emerges not just in the poem but in real life. So it is, approximately, with Wittgenstein. His labour is the substantive self-directed moral project, or spiritual exercise, of becoming better. But we can explain our sense that philosophy’s ethical significance for Wittgenstein reaches to his very choice of themes if we understand his leading themes as symbols of this labour and of its desired outcome, chosen therefore precisely so that his way to it can be eased by his imaginative engagement with them. There is thus a match between the goal of the practice as I have outlined it – returning the cut-off self to a state of ‘agreement’ with its world by making it philosophically unpuzzled – and the content of the thoughts winning one’s way to which would constitute success in the practice. Wittgenstein becomes (once again) a part of the world by becoming philosophically unpuzzled, but at the same time what he (and anyone who follows him) thinks when he is once again philosophically unpuzzled is that he is part of it. To win through to a ‘clear view’ of the ‘workings of our language’ (Wittgenstein 1958, 109, 122) in such a way that we can ‘set our faces against the picture of the “inner process” [here, of remembering]’ (1958, 305) is a symbolic ‘opening of one’s heart for others’, but thereby a real remedy for ‘aloofness’, for the thought that ‘we don’t want anyone else to see inside us’. The movement of thought from confusion to clarity about the mind’s place in the world, or the continuity between the linguistic and the non-linguistic, or the body as ‘a picture of the human soul’ (1958 II.iv, 179), is an allegory of the (ethical, personal) transformation the thinker undergoes in making it - a picture of the process that is constituted

Wittgenstein, Ethics and Therapy

529

by the thinking of it – and (as with the Cuna Indians) the thinker’s step-bystep following of the allegory is essential to effecting the transformation. 2. Wittgenstein made various remarks about philosophy and therapy, including that the treatment of a philosophical problem is ‘like the treatment of an illness’ (1958, 255), but we cannot infer automatically that it was psychotherapy that he had in mind. Moreover, if philosophy is a kind of spiritual exercise, can it really be a species of psychotherapy as well? A longstanding part of the self-image of psychotherapy – and the one that influenced an earlier generation of Wittgensteinians, such as Alice Ambrose and Morris Lazerowitz - has been a medical one, so surely it addresses certain kinds of illness, and doesn’t mental illness begin just where questions of moral good and bad leave off? On this view the very thing that gives Wittgenstein’s practice of philosophy ethical meaning would rule it out as psychotherapeutic. Now the story of psychotherapy’s confused understanding of its relation to the ethical would take too long to unravel here. But one reason why psychotherapy has been so slow to make anything of its own ethical character is the fact that it has got itself stuck, at least until quite recently, with a very limited idea of what the ethical is about: roughly, the study of the commanded and the forbidden, i.e. a (perhaps) secularized version of an already very narrowly understood religious morality (Anscombe 1981). So even when psychotherapy has been explicitly concerned with problems of how to live or with ideals of character, it has fought shy of describing its concerns as ethical. Nonetheless the idea that it does address such problems, i.e. that it’s guided by ethical notions, has become familiar in recent years – D.W. Winnicott (Winnicott 1965), Peter Lomas (Lomas 1973, 1999) and (from within philosophy) Ilham Dilman (Dilman 1983) are important figures here – and this is so especially in the British Isles, where optimistic Christian or post-Christian notions about man’s potential for good have been influential alongside Freudian cynicism: it has been said that the British Left owes ‘more to Methodism than it does to Marxism’, and something rather similar could be said about British psychotherapy, British expressions of Christianity, and Freud. More specifically, one of the ideals by which thinking in psychotherapy has been guided for some years has been an ideal of relationship to others: ‘real relating, which implies separateness from the object’, as Betty Joseph puts it (2004, 161) – the intended contrast is with thinking of others merely as

530

Edward Harcourt

extensions of oneself, or in a way clouded by fantasy. When we leave out the talk of ‘pathologies’, we can see a kinship between the ethical ideal of relationship to others which animates much psychotherapy and the preoccupation with relationship to world and to others in which the ethical meaning which philosophy had for Wittgenstein makes itself felt in the grain of his work. So there is after all at least one sense in which philosophy as Wittgenstein conceived it is ‘like psychotherapy’. There is at least one difference, however. Wittgenstein’s thoughts of his own badness often involve the idea of being on the wrong side of a judge, and the ‘coercive power of an absolute judge’ features in one of the very few explicitly ethical discussions in his philosophical work (1929 (1993a), 40). So one might infer that getting better means coming to be on the right side of the same judge – thus Wittgenstein’s ‘getting better’ would be quite close to a Christian sense of ‘redemption’ as being cleansed of sin. Wittgenstein once quoted JS Bach to capture a thought about his own work: ‘Bach wrote on the title page of his Orgelbüchlein “To the glory of the most high God, and that my neighbor may be benefited thereby”. That is what I would have liked to say about my work’ (Drury 1984, 168). To capture this aspect of his ethical sensibility, then, perhaps I can borrow the following: Mein Wille trachtet nur nach Bösen. Der Geist zwar spricht: ach! wer wird mich erlösen? … Rechne nicht die Missetat, Die dich, Herr, erzürnet hat!5

As far as I know, Wittgenstein never questions or criticizes the idea of the judge itself. But it is very black and white: Wittgenstein’s is an ethical sensibility that seems incapable of ambivalence.6 It’s thus not the fact that philosophy, in Wittgenstein’s view, is an ethical enterprise that sets it apart 5

Bach BWV 78. His veering between idealization and utter condemnation is nicely exemplified by his dissatisfaction with Maria Schutz am Semmering – where he was sent as part of his training as a schoolteacher – on the grounds that the place wasn’t rustic enough (‘Hier gibt es Park mit Springbrunnen – ich will aber ganz ländliche Verhältnisse!’ (Monk 1990, 193)), and his subsequent despair at the degraded condition of the inhabitants of the (genuinely ländlich) Trattenbach: ‘Es ist wahr, daß die Menschen im Durchschnitt nirgends sehr viel wert sind; aber hier sind sie viel mehr als anderswo nichtsnutzig und unverantwortlich’, Wittgenstein 2004, letter to Russell 23.10.1921. 6

Wittgenstein, Ethics and Therapy

531

from psychotherapy: on the contrary, that is something they have in common. But now remember the Lecture on Ethics, with its insistence on commandment, obedience, absolute authority. If anything sets psychotherapy apart from philosophy as Wittgenstein conceived it, it’s rather the fact that the restrictive view of the ethical – as revolving around the ideas of the prohibited and the commanded – which made psychotherapy so slow to become aware of its own ethical character is also Wittgenstein’s own view. 3. In conclusion I want to touch briefly on the third theme I announced at the beginning, the sense in which in recent commentary Wittgenstein’s works have been argued to be ‘therapeutic’. Minimally, I take the claim that Wittgenstein’s philosophy is therapeutic to mean that it conceives of philosophical problems as arising out of misunderstandings, and aims to do away with them. So understood, I think it’s unarguable that Wittgenstein’s philosophy is therapeutic. The question is how much this unarguable claim rules in or out. (For example, whether it implies anything as to the methods – for instance reductio arguments versus a mixed toolkit of comparisons, jokes and so on - to be used in getting rid of misunderstandings.) But I am not going to address that question here. There is also the question whether Wittgenstein’s philosophy is merely therapeutic in this sense: that is, whether having got rid of misunderstandings it just stops, or whether it tries to put anything else in their place. This is the territory marked out by the claim that we don’t ‘advance theses’ (Wittgenstein 1958, 128) in philosophy. What I do want to do is to relate the claim either that Wittgenstein’s philosophy is therapeutic or (more strongly) merely therapeutic, as just explained, to the claim I explored in section 2 that it is (to the extent that it is) psychotherapeutic, and to the account of Wittgenstein’s conviction as to its ethical significance which I defended in section 1. I want to argue that however strongly we read the claim that Wittgenstein’s philosophical method is ‘therapeutic’ – that is, even if it’s true that it contains no arguments and establishes no ‘theses’ – it is in a sense an accident that Wittgenstein’s philosophical practice both was ‘therapeutic’ (in whatever the best sense of that term is to be derived from recent commentary) and possessed the ethical meaning for him which it did. In one sense of course, this conjunction is no accident at all. More precisely, what’s no accident is that Wittgenstein’s philosophical practice pos-

532

Edward Harcourt

sessed the ethical meaning for him which it did and was regarded by him as ‘therapeutic’, and indeed as merely so (no arguments, or at least, a mixed toolkit of ‘therapies’ (PI 133) which includes argument as just one technique among many; and certainly no ‘theses’). To understand why this is no accident, we need to supplement the familiar Wittgensteinian thought that the practice of philosophy remedies confusion – which I have already discussed - with the further thought that there’s no result or product of thought that is at once free from confusion and genuinely philosophical. It follows that to get rid of confusion is also to get rid of philosophy. This is certainly a familiar Wittgensteinian idea, both from his philosophical writing (‘the real discovery … is the one that gives philosophy peace’, PI 133) and from his life (in which a leading ambition of his, though one he constantly failed to live up to, was to live without philosophy). But if – as I argued in section 1 – confusion is the mark of badness, then to have become good, and thereby to have come to live ‘in agreement with the world’, must involve having ceased to philosophize. Now to say that philosophy is merely therapeutic – no arguments, no theses - is a way of saying that it is self-effacing, that a condition of its success is that it write itself out of the scene. But if to remedy philosophical confusion is to strive to be better and there’s no such thing as unconfused philosophy, then Wittgenstein’s conception of the ethical significance of philosophy’s practice required him to see its method not only as therapeutic but as merely so. This requirement, however, is internal to Wittgenstein’s own imaginative construction of the meaning of philosophy. Granted his view of the ethical significance of its practice, Wittgenstein was required to think of his method as merely therapeutic. But it doesn’t follow that his method actually had to be merely therapeutic, and that’s one way in which the connection between (merely) therapeutic method and ethical meaning is an accident. Indeed for those who – myself included – can’t help finding ‘theses’ all over the place in Wittgenstein, it’s a puzzle that Wittgenstein seems so steadfastly to deny that there can be any, that is, that he seems so unselfknowing in his description of his own practice. Tracing the denial to the ethical meaning which his philosophical practice had for him solves the puzzle. But even if the method which Wittgenstein’s conception of the ethical meaning of philosophical practice forced him to regard as his own really was his own, the connection between ethical meaning and ‘merely therapeutic’ method is an accident for a further reason. What forges the link in Wittgenstein’s own thought is his idea that philosophical confusion is a

Wittgenstein, Ethics and Therapy

533

mark of personal badness, or that philosophical consciousness per se places us in ‘disagreement with the world’. These are powerful and readily intelligible ideas, and they surely help to explain the appeal of Wittgenstein as a writer of philosophy. They also exemplify the way in which themes or topics whose content is not apparently anything to do with ethics can have ethical meaning. But they are nonetheless idiosyncractic ideas. It is no strain on the imagination to think of someone to whom philosophical confusion is merely an intellectual irritant, and who thus might help themselves without misunderstanding to Wittgenstein’s treatment of the philosophical questions, without seeing their engagement with them as having anything like the significance in their own life that Wittgenstein took them to have in his. Nor is it any strain to think of someone whose moral practice upon himself takes place in a medium other than philosophy. (Drury, for instance, tried to follow a Wittgensteinian ethic – living simply, learning a skill that can be used to benefit others – while barely touching the philosophical questions (1984).) There thus really might be no such thing as a product of thought that is both unconfused and genuinely philosophical, and so it might really be that philosophy properly conceived must strive to write itself out of the scene, and yet the practice of philosophy so conceived, to a practitioner other than Wittgenstein, have no ethical significance at all. References Anscombe, G.E.M., 1981: Modern Moral Philosophy, in: Ethics, Religion and Politics: Collected Philosophical Papers Vol. III, Oxford: Blackwell Cavell, S., 1979: The Claim of Reason, Oxford / New York: Oxford University Press Dilman, I., 1983: Freud and Human Nature, Oxford: Blackwell Drury, M. O’C., 1984: Some Notes on Conversations with Wittgenstein, in: Recollections of Wittgenstein, ed. R. Rhees, Oxford / New York: Oxford University Press, 76-96 Drury, M. O’C., 1984: Conversations with Wittgenstein, in: Recollections of Wittgenstein, ed. R. Rhees, Oxford / New York: Oxford University Press Engelmann, P., 1967: Letters from Ludwig Wittgenstein with a Memoir, ed. B.F. McGuinness, Oxford: Blackwell Joseph, B., 2004: ‘Where there is no vision’: from Sexualization to Sexuality, in: Reason and Passion: A Celebration of the Work of Hanna Segal, ed. D. Bell, London / New York: Karnac Lévi-Strauss, C., 1977: Structural Anthropology, Penguin: Harmondsworth

534

Edward Harcourt

Lomas, P., 1999: Doing Good? Psychotherapy Out of Its Depth, Oxford: Oxford University Press Lomas, P., 1973: True and False Experience, London: Allen Lane Loos, A., 1931: Ornament und Verbrechen, in: Trotzdem: Die Schriften von Adolf Loos in Zwei Bänden. Zweiter Band, Innsbruck: Brenner-Verlag, 79-91 Loyola, St Ignatius of, Spiritual Exercises 1, tr. E. Mullan SJ, www.nwjesuits.org/ JesuitSpirituality/Exercises/SpEx001_020.html Malcolm, N., 1984: Introduction, in: Recollections of Wittgenstein, ed. R. Rhees, Oxford / New York: Oxford University Press Monk, R., 1990: Ludwig Wittgenstein: The Duty of Genius, London: Cape Pascal, F., 1984: A Personal Memoir, in: Recollections of Wittgenstein, ed. R. Rhees, Oxford / New York: Oxford University Press Winnicott, D.W., 1965: The Maturational Processes and the Facilitating Environment, London: Hogarth Press / Institute of Psychoanalysis Wittgenstein, H., 1984: My Brother Ludwig, in: Recollections of Wittgenstein, ed. R. Rhees, Oxford / New York: Oxford University Press Wittgenstein, L., 1922 (1974): Tractatus Logico-Philosophicus, tr. D.F. Pears and B.F. McGuinness, London: Routledge & Kegan Paul Wittgenstein, L., 1929 (1993a): Lecture on Ethics, in: Ludwig Wittgenstein: Philosophical Occasions 1912-1951, ed. J. Klagge and A. Nordmann, Indianapolis / Cambridge: Hackett, 37-44 Wittgenstein, L., 1936 (1993b): Philosophy [from ‘The Big Typescript’], in: Ludwig Wittgenstein: Philosophical Occasions 1912-1951, ed. J. Klagge and A. Nordmann, Indianapolis / Cambridge: Hackett, 158-199 Wittgenstein, L., 1958: Philosophical Investigations, 2nd ed., tr. G.E.M. Anscombe, Oxford: Blackwell. Wittgenstein, L., 1969: Briefe an Ludwig von Ficker, ed. G.H. von Wright and W. Methlagl, Salzburg: Otto Müller Verlag Wittgenstein. L., 1975: Philosophical Remarks, ed. R. Rhees, Oxford: Blackwell Wittgenstein. L., 1979: Notebooks 1914-16, ed. G.H. von Wright and G.E.M. Anscombe, Oxford: Blackwell Wittgenstein, L., 1980: Culture and Value, ed. G.H. von Wright and H. Nyman, Oxford: Blackwell Wittgenstein, L., 2004: Ludwig Wittgenstein: Gesamtbriefwechsel, Innsbrucker Elektronische Ausgabe

Publiccations of the Auustrian Ludwigg Wittgeensteinn Societty. Neew Seriies Friedrich Stadleer and Michael Stöltzner (Eds..) TTime and Histo ory P Proceedings of the t 28. Internattional Ludwig W Wittgenstein Sym mposium, Kirchberg am Wechhsel, Austria 20005 ISBN N: 978-3-9387933-17-6 621 ppp Hardcover € 79.00 Alois Pichhler, Simo Säättelä (Eds.) Wittgenstein: W TThe Philosopher and his Work ISBN N: 978-3-9387933-28-2 461 ppp Hardcover € 98.00 Christian C Kanziaan, Edmund Ruunggaldier (Edss.) Cultures. C Conflict-Analyssis-Dialogue Proceedinns of the 29. Intternational Ludw wig Wittgenstein Symposium in Kirchberg, Auustria 2006 ISBN N: 978-3-9387933-66-4 431 ppp Hardcover € 59.00 Geeorg Gasser (E Ed.) How Succcessful is Natturalism? ISBN N: 978-3-9387933-67-1 300 ppp Hardcover € 69.00 Chrristian Kanzian,, Muhammad Legenhausen (E Eds.) Subsstance and Attribute ISBN N: 978-3-9387933-68-8 248 ppp Hardcover € 69.00 Alois Pichle r, Herbert Hracchovec (Eds.) Wittgenstein andd the Philosop phy of Informattion Proceeedings of the 300th Internationaal Ludwig Wittgenstein-Symposium, 2007, Voolume 1 ISBN N 978-3-86838-001-9 356 ppp Hardcover€ 79,00 Herbert Hraachovec, Alois Pichler P (Eds.) Philosophy of the Informaation Society Proceeedings of the 300th Internationaal Ludwig Wittgenstein-Symposium, 2007, Voolume 2 ISBN N 978-3-86838-002-6 326 ppp Hardcover € 79,00

P.O O. Box 1541 ● D-63133 D Heuseenstamm bei Frrankfurt www w.ontosverlag.ccom ● [email protected] Tel.. ++49-6104-6665733 ● Fax +++49-61046657334

Jesúss Padilla Gálvezz (Ed.) Phenom menology as Grammar G ISBN N 978-3-9387933-91-6 224 ppp Hardcover € 59,00 Christiian Kanzian, Voolker Munz, Sasscha Windholz (Hrsg.) „W Wir hofften, jedes Jahr noch eein weiteres Symposium ma achen zu könnnen“ erg am Wechseel Zum 30. Internationalen Wittg enstein Sympoosium in Kirchbe ISBN N 978-3-9387933-61-9 191 ppp Paperback € 14,90 W Wulf Kellerwesssel Witttgensteins Sprrachphilosophhie in den „Phiilosophischen Untersuchunggen“ Eine komm mentierende Erssteinführung ISBN N 978-3-86838-032-3 330 ppp Paperback € 39,90 Joohn Edelman (E Ed.) Seense and Reallity Essaays out of Swansea ISBN N 978-3-86838--41-5 235 ppp Hardcover € 89,00 Alexander H Hieke, Hannes Leitgeb L (Eds.) n – Analysis Reduction – Abstraction Proceedings of the 31th Internationall Ludwig Wittgeenstein-Sympossium in Kirchberrg, 2008 ISBN N 978-3-86838-047-7 414 ppp Hardcover € 79,00 Alexander H Hieke, Hannes Leitgeb L (Eds.) Reduction Betweenn the Mind and the Brain ISBN N 978-3-86838-046-0 216 ppp Hardcover € 69,00 E Elena Tatievskaya Wittgensttein über das Verstehen V ISBN N 978-3-86838-053-8 546 ppp Hardcover € 119,00

O. Box 1541 ● D-63133 D Heuseenstamm bei Frrankfurt P.O www w.ontosverlag.ccom ● [email protected] Tel.. ++49-6104-6665733 ● Fax +++49-61046657334

Volker Munzz, Klaus Puhl, Joseph J Wang Languagge and World. Part One Essays on thhe Philosophy of o Wittgenstein ISBN N 978-3-86838-079-8 399 ppp Hardcover € 89,00 Volker Munzz, Klaus Puhl, Joseph J Wang Languagge and World. Part Two Signss, Minds and Actions ISBN N 978-3-86838-080-4 336 ppp Hardcover € 79,00 Richard Heeinrich, Elisabeeth Nemeth, Wolfram Picchler, David Waagner (Eds.) Image and Im maging in Phillosophy, Scien nce and the Arrts. Volume 1 Proceedings of the 33rd Internationall Ludwig Wittgeenstein-Sympossium in Kirchberrg, 2010 ISBN N 978-3-86838-115-3 414 ppp Hardcover € 98,00 Richard Heeinrich, Elisabeeth Nemeth, Wolfram Picchler, David Waagner (Eds.) maging in Phillosophy, Scien nce and the Arrts. Volume 2 Image and Im Proceedings of the 33rd Internationall Ludwig Wittgeenstein-Sympossium in Kirchberrg, 2010 ISBN N 978-3-86838-116-0 390 ppp Hardcover € 98,00 Henrik Voß Greenzen grammaatischer Willkü ür bei Wittgens stein ISBN N 978-3-86838-134-4 291 ppp Hardcover € 98,00

O. Box 1541 ● D-63133 D Heuseenstamm bei Frrankfurt P.O www w.ontosverlag.ccom ● [email protected] Tel.. ++49-6104-6665733 ● Fax +++49-61046657334