Wilfrid Sellars and Twentieth-Century Philosophy 2019031972, 2019031973, 9780815384991, 9781351202756

Table of contents :
Cover
Half Title
Series Page
Title
Copyright
Contents
Acknowledgments
List of Abbreviations
Introduction
PART I Influences
1 Cook Wilson, Sellars, and the Explication of Language
2 Sellars’s Twist on Carnap’s Syntax
3 Ajdukiewicz and Sellars on World-Perspectives
4 Sellars and Wittgenstein on Following a Rule
PART II Sellars and the Analytic Tradition
5 Wilfrid Sellars as an Analytic Philosopher
6 How Pragmatist was Sellars? Reflections on an Analytic Pragmatism
7 Transcendental Principles and Perceptual Warrant: A Case Study in Analytic Kantianism
PART III Learning From Sellars
8 Sellars on Inference
9 Sellars, Truth Pluralism, and Truth Relativism
10 Some Remarks on Sellars’s Theory of Experience
11 Sellars on Self-Knowledge
List of Contributors
Index

Citation preview

Wilfrid Sellars and TwentiethCentury Philosophy

This collection features 11 original essays, divided into three thematic sections, which explore the work of Wilfrid Sellars in relation to other twentieth-century thinkers. Section I analyzes Sellars’s thought in light of some of his influential predecessors, specifically Ludwig Wittgenstein, Rudolf Carnap, John Cook Wilson, and Kazimierz Ajdukiewicz. The second group of essays explores from different perspectives Sellars’s place within the analytic tradition, including his relation with analytic Kantianism and analytic pragmatism. The book’s final section extracts some of the most significant lessons Sellars’s work has to offer for contemporary philosophy. These chapters address his views on inference, his views on truth, and its connection to recent discussions about truthrelativism and truth-pluralism, his conception of self-knowledge, and his theory of perceptual experience. Stefan Brandt is Assistant Professor at the Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany. His work has been published in the British Journal for the History of Philosophy and Philosophical Investigations. Anke Breunig is Assistant Professor at the University of Halle-Wittenberg, Germany. She has published a book, Von Grund zu Grund (mentis, 2019), on Wilfrid Sellars and the Myth of the Given.

Routledge Studies in American Philosophy Edited by Willem deVries University of New Hampshire, USA Henry Jackman York University, Canada

Community and Loyalty in American Philosophy Royce, Sellars, and Rorty Steven A. Miller Sellars and the History of Modern Philosophy Edited by Luca Corti and Antonio M. Nunziante The Ethics of Wilfrid Sellars Jeremy Randel Koons Wilfrid Sellars and Buddhist Philosophy Freedom from Foundations Edited by Jay L. Garfield The Network Self Relation, Process, and Personal Identity Kathleen Wallace Deweyan Experimentalism and the Problem of Method in Political Philosophy Joshua Forstenzer Pragmatic Perspectives Constructivism beyond Truth and Realism Robert Schwartz Wilfrid Sellars and Twentieth-Century Philosophy Edited by Stefan Brandt and Anke Breunig For more information about this series, please visit: www.routledge.com/ Routledge-Studies-in-American-Philosophy/book-series/RSAP

Wilfrid Sellars and TwentiethCentury Philosophy

Edited by Stefan Brandt and Anke Breunig

First published 2020 by Routledge 52 Vanderbilt Avenue, New York, NY 10017 and by Routledge 2 Park Square, Milton Park, Abingdon, Oxon OX14 4RN Routledge is an imprint of the Taylor & Francis Group, an informa business © 2020 Taylor & Francis The right of the editors to be identified as the authors of the editorial material, and of the authors for their individual chapters, has been asserted in accordance with sections 77 and 78 of the Copyright, Designs and Patents Act 1988. All rights reserved. No part of this book may be reprinted or reproduced or utilised in any form or by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers. Trademark notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to infringe. Library of Congress Cataloging in Publication Data Names: Brandt, Stefan Leonhard, editor. | Breunig, Anke, editor. Title: Wilfrid Sellars and twentieth-century philosophy / edited by Stefan Brandt and Anke Breunig. Description: New York, New York : Taylor & Francis, 2019. | Series: Routledge studies in American philosophy; 21 | Includes bibliographical references and index. Identifiers: LCCN 2019031972 (print) | LCCN 2019031973 (ebook) | ISBN 9780815384991 (hardback) | ISBN 9781351202756 (ebook) Subjects: LCSH: Sellars, Wilfrid. | Philosophy, American—20th century. Classification: LCC B945.S444 W5485 2019 (print) | LCC B945. S444 (ebook) | DDC 191—dc23 LC record available at https://lccn.loc.gov/2019031972 LC ebook record available at https://lccn.loc.gov/2019031973 ISBN: 978-0-8153-8499-1 (hbk) ISBN: 978-1-351-20275-6 (ebk) Typeset in Sabon by Apex CoVantage, LLC

Contents

Acknowledgments List of Abbreviations Introduction

vii viii 1

ANKE BREUNIG AND STEFAN BRANDT

PART I

Influences 1 Cook Wilson, Sellars, and the Explication of Language

9 11

BORIS BRANDHOFF

2 Sellars’s Twist on Carnap’s Syntax

29

ANKE BREUNIG

3 Ajdukiewicz and Sellars on World-Perspectives

55

PETER OLEN

4 Sellars and Wittgenstein on Following a Rule

75

STEFAN BRANDT

PART II

Sellars and the Analytic Tradition 5 Wilfrid Sellars as an Analytic Philosopher

93 95

TADEUSZ SZUBKA

6 How Pragmatist was Sellars? Reflections on an Analytic Pragmatism JAMES R. O’SHEA

110

vi

Contents

7 Transcendental Principles and Perceptual Warrant: A Case Study in Analytic Kantianism

130

JOHANNES HAAG

PART III

Learning From Sellars

151

8 Sellars on Inference

153

JOHANNES HÜBNER

9 Sellars, Truth Pluralism, and Truth Relativism

174

LIONEL SHAPIRO

10 Some Remarks on Sellars’s Theory of Experience

207

WILLEM A. DEVRIES

11 Sellars on Self-Knowledge

221

FRANZ KNAPPIK

List of Contributors Index

240 241

Acknowledgments

The idea for this volume dates back to a conference in Erlangen, Germany, in June 2015. Without the generous financial support provided by Gerhard Ernst, this conference would not have been possible. Willem A. deVries suggested that our book would fit well with the Routledge Studies in American Philosophy, and we are very grateful for his advice. We would also like to thank Johannes Hübner for providing further resources that were very helpful in the final stages of the editing process. For extremely helpful suggestions on how to improve the style of writing we are grateful to Marc Hiatt. We are also indebted to an anonymous referee for providing detailed comments on some of the chapters. But most of all, we would like to thank the contributors to this volume for all their effort and dedication to this project.

Abbreviations

The following abbreviations have become the standard way of referring to Sellars’s writings and are used throughout the volume. The bibliographical information in this list includes first editions as well as reprints. It contains the writings mentioned in this volume, together with a reference to the chapters where these are discussed. AAE

AE AMI AR

BBK

CC

CDCM

CIL

EAE

1973. “Actions and Events.” Nous 7: 179–202. (Paper presented at the University of North Carolina, November, 1969.) [Chs. 4, 8] 1963. “Abstract Entities.” Review of Metaphysics 16: 627–71. Reprinted in PPME, 49–89, and SR, 163–205. [Introduction] 1967. “Aristotle’s Metaphysics: An Interpretation.” In PP. Reprinted in PPHP, 73–124. [Introduction] 1975. “Autobiographical Reflections (February, 1973).” In Action, Knowledge and Reality: Critical Studies in Honor of Wilfrid Sellars, edited by Hector-Neri Castañeda, 277–93. Indianapolis: Bobbs-Merrill. [Chs. 1, 5, 6] 1960. “Being and Being Known.” Proceedings of the American Catholic Philosophical Association 34: 28–49. Reprinted in SPR, 41–59, and SR, 209–228. [Chs. 1, 4] 1973. “Conceptual Change.” In Conceptual Change, edited by Glenn Pearce and Patrick Maynard, 77–93. Dordrecht: D. Reidel. Reprinted in EPH, 172–88. [Ch. 9] 1957. “Counterfactuals, Dispositions, and the Causal Modalities.” In Minnesota Studies in the Philosophy of Science, Vol. 2, edited by Herbert Feigl, Michael Scriven, and Grover Maxwell, 225–308. Minneapolis: University of Minnesota Press. Reprinted in part in Causation and Conditionals, edited by Ernest Sosa, 126–46. Oxford: Oxford University Press, 1975. [Chs. 1, 9] 1948. “Concepts as Involving Laws and Inconceivable without Them.” Philosophy of Science 15 (4): 287–315. Reprinted in PPPW, 86–114. [Chs. 1, 3, 6] 1963. “Empiricism and Abstract Entities.” In The Philosophy of Rudolf Carnap, edited by Paul Schilpp, 431–68. LaSalle, IL:

Abbreviations

ENWW

EPH EPM

FD

FMPP

GEC

I

IIO IKTE

ILE

ix

Open Court. Reprinted in EPH, 245–86. [Introduction, Chs. 2, 5, 9] 1947. “Epistemology and the New Way of Words.” The Journal of Philosophy 44 (24): 645–60. Reprinted in PPPW, 28– 40. [Chs.1, 3, 5] 1974. Essays in Philosophy and its History. Dordrecht, Holland: D. Reidel. 1956. “Empiricism and the Philosophy of Mind.” In Minnesota Studies in the Philosophy of Science, Vol. I, edited by Herbert Feigl and Michael Scriven, 253–329. Minneapolis: University of Minnesota Press. Reprinted in SPR, 127–96, with additional footnotes. Published separately as Empiricism and the Philosophy of Mind: with an Introduction by Richard Rorty and a Study Guide by Robert Brandom, edited by Robert Brandom. Cambridge, MA: Harvard University Press, 1997. Also reprinted in W. deVries and T. Triplett, Knowledge, Mind, and the Given: A Reading of Sellars’ “Empiricism and the Philosophy of Mind.” Cambridge, MA: Hackett Publishing, 2000. (Presented at the University of London Special Lectures in Philosophy for 1956 under the title “The Myth of the Given: Three Lectures on Empiricism and the Philosophy of Mind.”) [Introduction, Chs. 2, 4, 5, 7, 9, 10, 11] 1966. “Fatalism and Determinism.” In Freedom and Determinism, edited by Keith Lehrer, 141–74. New York: Random House. [Introduction, Chs. 8, 9] 1981. “Foundations for a Metaphysics of Pure Process.” (The Carus Lectures.) The Monist 64: 3–90. First lecture “The Lever of Archimedes” also in SR, 229–57. [Chs. 1, 7, 10] 1973. “Givenness and Explanatory Coherence.” The Journal of Philosophy 70: 612–22. (Presented at a symposium on Foundations of Knowledge at the 1973 meeting of the American Philosophical Association, Eastern Division.) [Ch. 7] 1972. “…this I or he or it (the thing) which thinks.” Proceedings of the American Philosophical Association 44: 5–31. Reprinted in EPH, 62–90, KTM, 341–62, and SR, 411–36. (The presidential address, American Philosophical Association, Eastern Division, for 1970.) [Ch. 1] 1956. “Imperatives, Intentions, and the Logic of ‘Ought.’” Methodos 8: 228–68. [Ch. 1] 1978. “The Role of Imagination in Kant’s Theory of Experience.” In Categories. A Colloquium, edited by Henry W. Johnstone, Jr., 231–45. University Park, PA: Pennsylvania State University Press. Reprinted in KTM, 419–30, and SR, 454–66. (The Dotterer Lecture 1978.) [Chs. 7, 10] 1950. “The Identity of Linguistic Expressions and the Paradox of Analysis.” Philosophical Studies 1 (2): 24–31. [Ch. 1]

x Abbreviations IM IRH

ITSA

KTE

KTI KTM

LRB

LSPO

LT

LTC

1953. “Inference and Meaning.” Mind 62: 313–38. Reprinted in PPPW, 218–37, and SR, 3–27. [Chs. 2, 4, 9] 1966. “The Intentional Realism of Everett Hall.” In The Southern Journal of Philosophy 4: 103–15. Reprinted in PPME, 29–48. (Part of a series of lectures in honor of Everett Hall, University of North Carolina.) [Ch. 5] 1953. “Is There a Synthetic A Priori?” Philosophy of Science 20: 121–38. Reprinted in a revised version in American Philosophers at Work, edited by Sidney Hook, 135–59. New York: Criterion Books, 1956. Revised version in SPR, 298– 320. [Chs. 2, 3, 6, 9] 1967. “Some Remarks on Kant’s Theory of Experience.” The Journal of Philosophy 64: 633–47. Reprinted in EPH, 44-61, KTM, 269–82, and SR, 437–53. (Paper given at the meeting of the American Philosophical Association, Eastern Division, December 1967.) [Chs. 7, 10] 1976. “Kant’s Transcendental Idealism.” Collections of Philosophy 6: 165–81. Reprinted in KTM, 403–17. [Ch. 7] 2002 Kant’s Transcendental Metaphysics: Sellars’ Cassirer Lectures Notes and Other Essays, edited with introduction by Jeffrey F. Sicha. Atascadero, CA: Ridgeview Publishing Co. [Chs. 7, 10] 1949. “Language, Rules and Behavior.” In John Dewey: Philosopher of Science and Freedom, edited by Sidney Hook, 289–315. New York: The Dial Press. Reprinted in PPPW, 186–214. [Chs. 1, 4, 6, 8] 1957. “Logical Subjects and Physical Objects.” Philosophy and Phenomenological Research 17 (4): 458–72. (Contribution to a symposium with Peter Strawson held at Duke University, November, 1955.) [Introduction, Ch. 1] 1961. “The Language of Theories.” In Current Issues in the Philosophy of Science, edited by Herbert Feigl and Grover Maxwell, 57–77. New York (Holt, Rinehart, and Winston). Reprinted in SPR, 106–26, and in The Problem of Scientific Realism, edited by Edward A. MacKinnon. New York (Appleton-Century-Crofts). (Presented in 1959 at the meeting of the American Association for the Advancement of Science.) [Introduction] 1969. “Language as Thought and as Communication.” Philosophy and Phenomenological Research 29: 506–27. Reprinted in Language and Human Nature, edited by Paul Kurtz, with commentaries by Mikel Dufrenne, Edouard Morot-Sir, Joseph Margolis, and Edward S. Casey. St. Louis: Warren H. Green, 1971. Reprinted in EPH, 93–117, and SR, 57–80. [Chs. 4, 7, 8, 9]

Abbreviations ME

MEV

MFC

MGEC

NAO

NI

NS

P PH

xi

1989. The Metaphysics of Epistemology: Lectures by Wilfrid Sellars, edited by Pedro V. Amaral. Atascadero, CA: Ridgeview. (Lectures given during the fall of 1975.) [Ch. 7] 1981. “Mental Events.” Philosophical Studies 39: 325–45. Reprinted in SR, 282–300. (Paper given at the meeting of the American Philosophical Association, Western Division, in Detroit, April 1980.) [Chs. 6, 7, 8] 1974. “Meaning as Functional Classification: A Perspective on the Relation of Syntax to Semantics.” Synthese 27 (3–4): 417–37, with commentaries by Daniel Dennett, Hilary Putnam, Sellars’s Reply, and a General Discussion of Sellars’s Paper: 439–70. Reprinted in Intentionality, Language and Translation, edited by J.G. Troyer and S.C. Wheeler, III. Dordrecht, Holland: D. Reidel, 1974. Reprinted in SR, 81–100. (Paper given at the conference “Language, Intentionality, and Translation Theory” at the University of Connecticut, March 1973.) [Introduction, Chs. 1, 2, 4, 8, 9, 11] 1979. “More on Givenness and Explanatory Coherence.” In Justification and Knowledge, edited by George Pappas, 169–82. Dordrecht, Holland: D. Reidel. Reprinted in Perceptual Knowledge, edited by Jonathan Dancy, 177–91. Oxford: Oxford University Press, 1988. [Ch. 7] 1979. Naturalism and Ontology: The John Dewey Lectures for 1974. Reseda, CA: Ridgeview Publishing Company. (Reprinted with corrections in 1996.) [Chs. 3, 5, 6, 9] 1964. “Notes on Intentionality.” Journal of Philosophy 61 (21): 655–65. Reprinted in PPME, 128–40. Reprinted in Intentionality, Mind and Language, edited by Ausonio Marras. Chicago: University of Illinois Press, 1972. Reprinted in New Readings in Philosophical Analysis, edited by Herbert Feigl, Wilfrid Sellars, and Keith Lehrer. New York: Appleton-Century-Crofts, 1972. (Presented at a meeting of the American Philosophical Association, Eastern Division, December, 1964.) [Chs. 5, 9] 1962. “Naming and Saying.” Philosophy of Science 29: 7–26. Reprinted in SPR, 225–46. (Presented at a meeting of American Philosophical Association, Western Division, in 1961, in a symposium on reference and use.) [Introduction, Ch. 4] 1952. “Particulars.” Philosophy and Phenomenological Research 13: 184–99. Reprinted in SPR, 282–97. [Chs. 6, 9] 1967. “Phenomenalism.” In Intentionality, Minds and Perception, edited by Hector-Neri Castañeda, 215–74, with comments by Bruce Aune, and a reply by Sellars. Detroit: Wayne State University Press. (An abbreviated version of PHM.)

xii Abbreviations PHM

PP PPE PPHP

PPME

PPPW

PR PRE

PSIM

QMSP

RAL

RC

RDP

1963. “Phenomenalism.” In SPR, 60–105. (A paper presented at the Wayne State University Symposium in the Philosophy of Mind, December 1962.) Reprinted in SR, 303–49. [Chs. 7, 11] 1967. Philosophical Perspectives. Springfield, IL: Charles C. Thomas. 1947. “Pure Pragmatics and Epistemology.” Philosophy of Science 14 (3): 181–202. Reprinted in PPPW, 4–25. [Chs. 1, 2, 5, 9] 1977. Philosophical Perspectives: History of Philosophy. Atascadero, CA: Ridgeview Publishing Company. (A reprint of part I of PP.) 1977. Philosophical Perspectives: Metaphysics and Epistemology. Atascadero, CA: Ridgeview Publishing Company. (A reprint of part II of PP.) Reissued in 1996 with an updated bibliography. 1980. Pure Pragmatics and Possible Worlds: The Early Essays of Wilfrid Sellars, edited and introduced by Jeffrey F. Sicha. Atascadero, CA: Ridgeview Publishing Company. New edition published by Ridgeview Publishing Company in 2005. 1954. “Physical Realism.” Philosophy and Phenomenological Research 15: 13–32. Reprinted in PPME, 5–28. [Ch. 6] 1954. “Presupposing.” The Philosophical Review 63 (2): 197– 215. Reprinted in Essays on Bertrand Russell, edited by E.D. Klemke, 173–89. Chicago: University of Illinois Press, 1970. [Ch. 1] 1962. “Philosophy and the Scientific Image of Man.” In Frontiers of Science and Philosophy, edited by Robert Colodny, 35–78. Pittsburgh: University of Pittsburgh Press. Reprinted in SPR, 1–40, and SR, 369–408. (Paper given as two lectures at the University of Pittsburgh, December 1960.) [Introduction, Ch. 1, 4, 5, 10, 11] 1950. “Quotation Marks, Sentences and Propositions.” Philosophy and Phenomenological Research 10 (4): 515–25. Reprinted in PPPW, 170–79. [Ch. 1] 1973. “Reason and the Art of Living in Plato.” In Phenomenology and Natural Existence: Essays in Honor of Marvin Farber, edited by Dale Riepe, 353–77. New York: The University of New York Press. Reprinted in EPH, 3–26. (Presented in a conference of Greece: The Critical Spirit, 450–350 B.C. at Ohio State University, 1968.) [Introduction] 1948−49. “Review of Language and Myth by Ernst Cassirer.” Philosophy and Phenomenological Research 9 (2): 326–29. [Ch. 5] 1974. “Reply” (to Putnam and Dennett). Synthese 27 (3–4, special issue: Intentionality, Language and Translation, edited by J.G. Troyer and S.C. Wheeler): 457–66. [Ch. 9]

Abbreviations RM RNWW

RP

RPH

RQ SFA

SK

SM

SPR

SR

SRI

SRLG

xiii

1973. “Reply to Marras.” Canadian Journal of Philosophy 2: 485–93. Reprinted in EPH, 118–28. [Ch. 8] 1948. “Realism and the New Way of Words.” Philosophy and Phenomenological Research 8 (4): 601–34. Reprinted in Readings in Philosophical Analysis, edited by Herbert Feigl and Wilfrid Sellars, 424–56. New York: Appleton-CenturyCrofts, 1949. Reprinted in PPPW, 46–78. [Chs. 1, 3, 5] 1950. “Review of Elements of Analytic Philosophy by Arthur Pap.” Philosophy and Phenomenological Research 11 (1): 104– 9. [Ch. 5] 1966. “The Refutation of Phenomenalism: Prolegomena to a Defense of Scientific Realism.” In Mind, Matter, and Method: Essays in Philosophy of Science in Honor of Herbert Feigl, edited by Paul K. Feyerabend and Grover Maxwell, 198–214. Minneapolis: University of Minnesota Press. [Ch. 5] [1973. “Reply to Quine.” Synthese 26: 122–45. Reprinted in EPH, 148–71. [Ch. 7] 1957. “Substance and Form in Aristotle.” The Journal of Philosophy 54: 688–99. Reprinted in PPHP, 125–36. (Presented as a paper at a meeting of the American Philosophical Association, December, 1957.) [Introduction] 1975. “The Structure of Knowledge: (1) Perception; (2) Minds; (3) Epistemic Principles.” In Action, Knowledge and Reality: Critical Studies in Honor of Wilfrid Sellars, edited by HectorNeri Castañeda, 295–347. Indianapolis: Bobbs-Merrill. (Presented as The Matchette Foundation Lectures for 1971 at the University of Texas.) [Introduction, Chs. 1, 5, 7, 10] 1968. Science and Metaphysics: Variations on Kantian Themes. London: Routledge and Kegan Paul. Reissued by Ridgeview Publishing Company in 1992. (The John Locke Lectures for 1965–66.) [Chs. 1, 4, 6, 7, 8, 9] 1963. Science, Perception and Reality, London: Routledge and Kegan and Paul. Reissued by Ridgeview Publishing Company in 1991. 2007. In the Space of Reasons: Selected Essays of Wilfrid Sellars, edited by Kevin Scharp and Robert B. Brandom. Cambridge, MA: Harvard University Press. 1965. “Scientific Realism or Irenic Instrumentalism: A Critique of Nagel and Feyerabend on Theoretical Explanation.” Boston Studies in the Philosophy of Science, Vol. II, edited by Robert Cohen and Max Wartofsky, 171–204. New York: Humanities Press, 1965. Reprinted in PPME, 337–69. [Ch. 9] 1954. “Some Reflections on Language Games.” Philosophy of Science 21 (3): 204–28. Reprinted in SR, 28–56. Expanded Version from 1955 in SPR, 321–58. [Introduction, Chs. 2, 4, 6, 8, 9]

xiv Abbreviations SRPC

1978. “Some Reflections on Perceptual Consciousness.” In Crosscurrents in Philosophy: Selected Studies in Phenomenology and Existential Philosophy 7, edited by R. Bruzina and B. Wilshire, 169–85. The Hague and Boston: Martinus Nijhoff. Reprinted in KTM, 431–41. [Chs. 7, 10] SRTT 1967. “Some Reflections on Thoughts and Things.” Noûs 1: 97–121. Reprinted as Ch. 3 of SM, 56–83. Reprinted in New Readings in Philosophical Analysis, edited by Herbert Feigl, Wilfrid Sellars, and Keith Lehrer. New York: Appleton-Century—-Crofts, 1972. [Ch. 5] SSMB 1953. “A Semantical Solution of the Mind-Body Problem.” Methodos 5: 45–82. Reprinted in PPPW, 186–214. [Introduction] SSOP 1982. “Sensa or Sensings: Reflections on the Ontology of Perception.” Philosophical Studies 41 (Essays in Honor of James Cornman): 83–111. (Presented at a Colloquium in Honor of James Cornman at the University of North Carolina, October 1976.) [Chs. 7, 10] TA 1966. “Thought and Action.” In Freedom and Determinism, edited by Keith Lehrer, 105–40. New York: Random House. [Ch. 4] TC 1962. “Truth and ‘Correspondence.’” Journal of Philosophy 59: 29–56. Revised version in SPR, 197–224. Reprinted in New Readings in Philosophical Analysis, edited by Herbert Feigl, Wilfrid Sellars, and Keith Lehrer. New York: AppletonCentury-Crofts, 1972. [Chs. 1, 3, 5, 6, 8, 9] TTP 1983. “Towards a Theory of Predication.” In How Things Are, edited by James Bogen and James E. McGuire, 281–318. Dordrecht, Holland: D. Reidel. (Presented at a conference on predication at Pitzer College, April 1981.) [Chs. 6, 9] VTM 1955. “Vlastos and ‘The Third Man.’” Philosophical Review 64 (3): 405–37. Reprinted in PPHP, 23–54. [Introduction, Ch. 1] WNDSL 2015. Notre Dame Lectures 1969–1986. Revised edition, edited by Pedro Amaral and David Landy. Atascadero, CA: Ridgeview Publishing Company. [Chs. 1, 5, 6, 7, 9]

Writings Without Abbreviation Wilfrid S. Sellars Papers, 1899–1990. Collection number ASP.1991.01. Archives of Scientific Philosophy, Special Collections Department, University of Pittsburgh, online at https://digital.library.pitt.edu/ collection/wilfrid-s-sellars-papers. [Chs. 1, 4, 5, 9] Sellars-Harman Correspondence (1970). Edited by A. Chrucky. URL: www.ditext.com/sellars/sh-corr.html [Ch. 9]

Abbreviations

xv

Sellars-Marras Correspondence (1972–76). Edited by A. Chrucky. URL: www.ditext.com/sellars/csm.html [Ch. 9] Sellars, Wilfrid and Castañeda, Hector-Neri, 2006. Correspondence on Philosophy of Mind, edited by Andrew Chrucky. URL: www. ditext.com/sellars/corr.html

Introduction Anke Breunig and Stefan Brandt

Having had his most productive years from the late 1940s to the early 1970s Wilfrid Sellars belongs, along with such figures as W.V. Quine, Donald Davidson, Peter Strawson, and Elizabeth Anscombe, to a golden generation of analytic philosophers who have decisively influenced the course of anglophone philosophy, not only through their views but also through the topics they addressed and the broadly analytical approach they all embraced. Sellars substantially contributed to the development of anglophone philosophy in all these dimensions. First, he played an influential role in bringing analytic philosophy to America. Together with Herbert Feigl he published Readings in Philosophical Analysis (Feigl and Sellars 1949), which served to introduce many American graduate students to the new analytic movement,1 and in 1950 they founded the journal Philosophical Studies, which was explicitly devoted to “philosophy in the analytic tradition”. Along with Quine, Sellars was among the first major American philosophers to engage deeply and productively with the logical positivism of the Vienna Circle. But while Quine moved from logical empiricism to a more radical empiricism, which ultimately eschewed all a priori knowledge, Sellars took up the remaining rationalist elements of positivism to move into a more Kantian direction. He is reported by Robert Brandom to have said in conversation that his hope was to move analytic philosophy from its Humean to its Kantian phase (cf. Brandom 2015, 4–5). More specifically, he adopted and refined the linguistic conceptions of necessity and apriority of the positivists to develop a deeply Kantian epistemology with a linguistic twist.2 Second, Sellars not only addressed a large number of perennial philosophical subjects, such as the nature of free will (cf., e.g., FD), the foundations of knowledge (cf., e.g., EPM, SK), or the nature of universals (cf., e.g., AE), but also a number of topics especially characteristic of current and recent debates, such as the status of the theoretical posits of scientific theories (cf., e.g., LT), the place of mental and intentional phenomena in an ontology that is determined by the natural sciences (cf., e.g., SSMB, EPM, PSIM), or the nature of meaning and linguistic rule-following (cf.,

2 Anke Breunig and Stefan Brandt e.g., MFC, SRLG). Third, he was often the first to formulate views on these issues that have by now become standard positions, such as a functionalist theory of mental acts and a version of the theory-theory of their attribution,3 while at the same time defending positions that have only fairly recently received substantial attention. Sellars, for instance, stressed the normative character of intentional phenomena long before that became a “hot topic”, and he defended a normativist-expressivist account of modal discourse that has only very recently become a focus of intense debate.4 Sellars’s ability to contribute to a large variety of debates was matched by a remarkable ability to integrate insights from a wide range of philosophers and philosophical traditions. He wrote on thinkers from almost all periods of the history of philosophy, from Plato and Aristotle (cf., e.g., VTM, RAL, AMI, SFA) to Wittgenstein and Carnap (cf., e.g., NS, EAE), and he tried to incorporate insights from a diverse range of at his time contemporary or recent philosophical traditions, such as pragmatism, ordinary language philosophy, and logical positivism.5 However, despite an increasing interest in Sellars’s work in recent years,6 his historical role in twentieth-century philosophy has so far received very little attention.7 This is in stark contrast to his most important contemporaries working in the same areas of philosophy, Quine and Davidson, whose work has already received a substantial amount of attention from historians of analytical philosophy.8 In Scott Soames’s influential history of analytical philosophy (Soames 2003), for instance, there are four chapters on Quine, two on Davidson, and none on Sellars. This is somewhat ironic, since Sellars would have been the first to welcome a growing historical self-awareness within the analytic tradition. The aim of the current volume is to contribute to changing this situation by exploring some of the twentieth-century influences on Sellars’s philosophy, relating his work to the traditions in which he worked and drawing some lessons from it for current debates. Accordingly, the book is divided into three thematic blocks. In the first part, different influences on Sellars’s philosophy are explored. In the second part, Sellars’s thought is related to different philosophical traditions in the twentieth century. In the third part, lessons from Sellars’s thought for contemporary debates are derived. We hope that combining both historical and systematic approaches in this way will illustrate the fruitfulness of addressing one and the same or similar topics from different viewpoints (see, e.g., Chapters 4 and 8; Chapters 6 and 9). The first part of the book begins with a chapter by Boris Brandhoff that explores Sellars’s relation to the mostly forgotten Oxford philosopher John Cook Wilson (1849–1915). Brandhoff takes as his point of departure a rather surprising remark Sellars makes in his 1957 paper “Logical Subjects and Physical Objects”. Sellars writes: “I can say in all seriousness that 20 years ago I regarded Wilson’s Statement and Inference as the philosophical book of the century” (LSPO §1: 458). Brandhoff

Introduction

3

shows that Sellars must have become familiar with Cook Wilson’s opus magnum Statement and Inference (Cook Wilson 1926) during his student days at Oxford in the 1930s, where Cook Wilson was at the time a widely read and influential author. Brandhoff argues, in particular, that Cook Wilson’s conception of the nature of philosophy had a lasting influence on Sellars, just as it had on Oxford ordinary language philosophers such as Ryle and Strawson. According to Cook Wilson a central task of philosophy is to make explicit the system of rules implicit in ordinary linguistic usage. This view was taken up by ordinary language philosophers and also by Sellars, who thought it was a central task of philosophy to make explicit the linguistic rules that make up our “manifest” or commonsense picture of the world. But unlike his contemporaries in the ordinary language tradition, Sellars thought it was also the task of philosophy to make explicit the conceptual scheme of the emerging scientific picture of the world and investigate the challenges it poses to the commonsense picture. In this respect, Brandhoff contends, Sellars was arguably closer to Cook Wilson than the latter’s Oxonian descendants. For Cook Wilson, too, tried to extend his investigations to the foundations of nineteenthcentury science. In Chapter 2, Anke Breunig discusses from a Sellarsian perspective three problems for Carnap’s syntactical conception of language. The first problem is that Carnap lacks the resources to distinguish structures of inscriptions or sounds that are or can be languages from ones that cannot function as a language. Sellars offers a solution to this problem: only structures that are autonomous, i.e., that could be a person’s only language, should count as a language. The second problem is a dilemma that arises out of Carnap’s sharp distinction between logical meaning, i.e., the meaning of logical and mathematical expressions, and empirical meaning, i.e., the meaning of expressions referring to empirical phenomena. Breunig argues that Carnap is either forced to accept some kind of reductive naturalism for empirical meaning, which would make the relation between empirical meaning and logical meaning, which is according to Carnap irreducible, mysterious; or he will be pushed towards a kind of concept empiricism that is according to Sellars an instance of the Myth of the Given. Finally, Breunig argues that Carnap’s failure to give a unified account of logical and empirical meaning leaves him with a deficient account of logical meaning. Breunig suggests that the latter two problems can be avoided by adopting the kind of unified normative account of meaning that Sellars recommends. In his contribution “Ajdukiewicz and Sellars on World-Perspectives”, Peter Olen traces another influence of early European analytic philosophy on Sellars. In his paper “The Scientific World-Perspective”, which was translated from German into English by Sellars, the Polish logician and philosopher Kazimierz Ajdukiewicz (1890–1963) argues that every meaningful language contains what he calls a “world-perspective”. This

4 Anke Breunig and Stefan Brandt thought is picked up and further developed by Sellars in his earliest publications, where he argues that meaningful languages have to contain a “world-story”. In later publications, Olen argues, this thought is not given up. It reappears in the form of the picture theory Sellars derives from Wittgenstein’s Tractatus. According to the later Sellars, languages do not contain a world-story but the physical vehicles of any empirically meaningful language form pictures of the world the language is about. While Sellars’s picture theory is explicitly derived from Wittgenstein’s Tractatus, the influence of Wittgenstein’s later work on Sellars remains mostly under the surface. Stefan Brandt argues in his contribution that there are nevertheless clear lines of influence and significant points of agreement. He shows that one point of agreement is the central role rules of language play for both Sellars and the later Wittgenstein. Still, Brandt argues, it is also an area where differences in their outlooks reveal themselves. Sellars’s conception of what it takes to follow a rule of language shows his fundamental commitment to an event-causal explanation of rational action, whereas no such commitment can be found in Wittgenstein. The second part of the book comprises three chapters. In Chapter 5, Tadeusz Szubka explores Sellars’s place within the analytic tradition. He first puts together various details from Sellars’s life that provide a glimpse into his exposure to different facets of analytic philosophy and his role in shaping it. Subsequently, he discusses Sellars’s conception of philosophy and the sense in which it embodies characteristically analytical ideas. Szubka claims that Sellars’s place within the analytic tradition is rather unique. Though he was tempted in his earliest writings to focus exclusively on analysis, narrowly conceived, and to confine philosophy to the formal theory of language, he later developed a much broader notion of analysis while putting equal stress on the importance of synthesis. This metaphilosophical shift allowed him to describe the ultimate aim of philosophical inquiry as giving a unified and comprehensive synthesis of the scientific and commonsense images of the world. Szubka’s conclusion fits well with Brandhoff’s description of Sellars’s philosophical methodology in Chapter 1. In Chapter 6, James R. O’Shea traces another important influence on Sellars’s philosophy, that of American pragmatism. O’Shea first explores classical pragmatist influences on Sellars’s views by discussing aspects of the intellectual background in which those views developed. He then outlines more abstractly some of the enduring pragmatist themes in Sellars’s philosophy, including his conceptions of the Myth of the Given, the space of reasons, and his normative-inferentialist theory of meaning. He concludes with a discussion of Sellars’s views on truth and “picturing”, which present a complex and controversial case for the question of how pragmatist Sellars’s philosophy actually is. Sellars’s views on truth are also explored in detail by Lionel Shapiro in Chapter 9. Johannes Haag discusses in Chapter 7 a central aspect of Sellars’s Kantianism: the role and nature of the conceptual capacities exercised in

Introduction

5

perception. He does so by discussing a criticism of Sellars’s conception of our warrant for perceptual judgments recently raised by John McDowell (McDowell 2016). According to McDowell, Sellars held a deeply flawed view of perceptual warrant because he failed to even recognize the alternative conception McDowell himself endorses, namely, that a perceptual belief is conclusively warranted by having a state of affairs conceptually present to one. Haag argues that McDowell’s criticism rests on a misinterpretation of Sellars’s account of trans-level justification and argues that the main difference between Sellars’s and McDowell’s accounts is that Sellars is committed to the idea that a perceptual belief is warranted by having a state of affairs conceptually present to one in a cognitive act that itself is already a believing, while McDowell needs to insist that the perceptual belief is warranted by a cognitive act that is more basic than any believing. Haag concludes that while Sellars never intended to defend the kind of view advocated by McDowell, he can be interpreted as defending a conception of perceptual warrant closely akin to it, a conception that McDowell in turn fails to have properly in view. The third part of the book comprises four chapters that in some way or other systematically engage with Sellars’s work by either criticizing or building on it. In Chapter 8, Johannes Hübner develops in great detail Sellars’s account of inference as a kind of rule-following and thereby touches, from a different perspective, on a number of issues also discussed by Brandt in Chapter 4. Hübner shows that Sellars’s distinction between “rules of action” and “rules of criticism” plays a central role within his account of inference, as it allows him to conceive of inferential judgments as mental acts that are not intentional actions while still being guided by inferential rules. This makes it possible for Sellars’s account to avoid problems that face other accounts, such as that of Paul Boghossian, which conceive of inference as an instance of rule-following but of rulefollowing as a form of intentional action. In Chapter 9, Lionel Shapiro discusses Sellars’s views on truth and relates them to current debates. He argues, in particular, that Sellars endorses two doctrines that have recently become the focus of much discussion: first, truth pluralism, the view that while the predicate “true” is used univocally across diverse domains of discourse, truth still behaves significantly different in those different domains; second, truth relativism, i.e., the idea that truth is in some sense relative to a perspective or a context. Shapiro argues that the key to Sellars’s truth pluralism is his analysis of truth as “semantic assertability”, which is motivated by the same consideration that leads him to adopt truth relativism. This is his interest in the application of the notion of truth to conceptual schemes that are substantially different from ours. Shapiro argues that despite their common motivation Sellars’s truth pluralism and his truth relativism are not only independent but also in tension with each other. In Chapter 10, Willem A. deVries engages critically with three aspects of Sellars’s theory of experience. He first criticizes his use of the phrase

6 Anke Breunig and Stefan Brandt “immediate experience” to refer to sense impressions in “Empiricism and the Philosophy of Mind”. Second he criticizes a claim Sellars makes in a number of relatively late publications, namely, that our perceptual judgments are almost universally wrong in that we typically take features of our consciousness to be properties of the objects we perceive. This claim, deVries argues, may well fit Sellars’s ontology, but it is difficult to reconcile with his epistemology. Third, deVries argues Sellars’s focus on the visual in his discussions of perception leads him to make confused claims about how we perceive causal properties. DeVries concludes with a critical question about Sellars’s distinction between imagining and sensing. In the final chapter, Franz Knappik offers a detailed interpretation and critical evaluation of Sellars’s account of the knowledge we have of our own occurrent thoughts. He argues that Sellars’s account, which is closely modeled on his theory of perception, undermines influential arguments by Tyler Burge and Sydney Shoemaker against perceptual accounts of self-knowledge. Knappik concludes by criticizing Sellars for rejecting any role for phenomenal consciousness in the explanation of the knowledge we have of our own occurrent thoughts.

Notes 1. Donald Davidson remarked that he “got through graduate school reading Feigl and Sellars” (Davidson 1980, 261). Cf. also Hacker (1996, 184). 2. For this interpretation of the relation between Sellars and Quine, cf. Brandt (2017). 3. For Sellars’s relation to the theory-theory, see Rosenberg (2007), chapter 8, and O’Shea (2012). 4. Sellars’s views are related to the latter debate, for instance, in Brandom (2008), chapter 4, and Price (2017). 5. For Sellars’s relation to pragmatism, see also the contribution by O’Shea to this volume, for his relation to ordinary language philosophy the contribution by Brandhoff, and for logical positivism the contributions by Olen and Breunig. 6. There has been a noticeable increase in the number of books devoted to Sellars’s work in the last ten years. See, for instance, deVries (2005), O’Shea (2007), deVries (2009), Seibt (2007), Lance and Wolf (2006), Rosenberg (2007), Brandom (2015), O’Shea (2016), Corti and Nunziante (2018), Pereplyotchik and Barnbaum (2017), Reider (2016), Olen (2016), and Gironi (2017). 7. An exception are Corti and Nunziante (2018) and Olen (2016). 8. Besides the volumes by Soames, mentioned later, see for instance the historically well-informed treatments of Quine and Davidson in Glock (2003), Gibson (2004), Hylton (2007), Harman and Lepore (2014), Lepore and Ludwig (2013), Lepore and Ludwig (2005), and Lepore and Ludwig (2007).

References Brandom, Robert. 2008. Between Saying and Doing: Towards an Analytic Pragmatism. Oxford: Oxford University Press. ———. 2015. From Empiricism to Expressivism: Brandom Reads Sellars. Cambridge, MA: Harvard University Press.

Introduction

7

Brandt, Stefan. 2017. “Sellars and Quine on Empiricism and Conceptual Truth”. British Journal for the History of Philosophy 26 (1): 108–32. Cook Wilson, John. 1926. Statement and Inference: With Other Philosophical Papers. Oxford: Oxford University Press. Corti, Luca and Antonio M. Nunziante, eds. 2018. Sellars and the History of Modern Philosophy. New York: Routledge. Davidson, Donald. 1980. Essays on Actions and Events. Oxford: Oxford University Press. deVries, Willem A. 2005. Wilfrid Sellars. Chesham: Acumen. ———, ed. 2009. Empiricism, Perceptual Knowledge, Normativity, and Realism. Oxford: Oxford University Press. Feigl, Herbert and Wilfrid Sellars, eds. 1949. Readings in Philosophical Analysis. New York: Appleton-Century-Crofts. Gibson, Roger F. 2004. The Cambridge Companion to Quine. Cambridge: Cambridge University Press. Gironi, Fabio, ed. 2017. The Legacy of Kant in Sellars and Meillassoux: Analytic and Continental Kantianism. New York: Routledge. Glock, Hans-Johann. 2003. Quine and Davidson on Language, Thought and Reality. Cambridge: Cambridge University Press. Hacker, Peter M. S. 1996. Wittgenstein’s Place in Twentieth-Century Analytic Philosophy. Oxford: Blackwell. Harman, Gilbert and Ernest Lepore, eds. 2014. A Companion to W.V.O. Quine. Oxford: Blackwell. Hylton, Peter. 2007. Quine. London: Routledge. Lance, Mark N. and Michael P. Wolf, eds. 2006. The Self-Correcting Enterprise: Essays on Wilfrid Sellars. New York: Rodopi. Lepore, Ernest and Kirk Ludwig. 2005. Donald Davidson: Meaning, Truth, Language and Reality. Oxford: Clarendon Press. ———. 2007. Donald Davidson’s Truth-Theoretic Semantics. Oxford: Clarendon Press. ———, eds. 2013. A Companion to Donald Davidson. Oxford: Blackwell. McDowell, John. 2016. “A Sellarsian Blind Spot”. In Sellars and His Legacy, edited by James R. O’Shea, 100–16. Oxford: Oxford University Press. Olen, Peter. 2016. Wilfrid Sellars and the Foundations of Normativity. London: Palgrave Macmillan. O’Shea, James R. 2007. Wilfrid Sellars: Naturalism with a Normative Turn. Cambridge: Polity Press. ———. 2012. “The ‘Theory Theory’ of Mind and the Aims of Sellars’ Original Myth of Jones”. Phenomenology and the Cognitive Sciences 11 (2): 175–204. ———, ed. 2016. Sellars and His Legacy. Oxford: Oxford University Press. Pereplyotchik, David and Deborah R. Barnbaum, eds. 2017. Sellars and Contemporary Philosophy. New York: Routledge. Price, Huw. 2017. “Wilfrid Sellars Meets Cambridge Pragmatism”. In Pereplyotchik and Barnbaum (2017), 123–40. Reider, Patrick J. 2016. Wilfrid Sellars, Idealism, and Realism: Understanding Psychological Nominalism. London: Bloomsbury. Rosenberg, Jay F. 2007. Wilfrid Sellars: Fusing the Images. Oxford: Oxford University Press. Seibt, Johanna. 2007. Wilfrid Sellars. Paderborn: Mentis.

8 Anke Breunig and Stefan Brandt Sellars, Wilfrid. 1953 [SSMB]. “A Semantical Solution of the Mind-Body Problem”. Methodos 5: 45–82. ———. (1954) 1991 [SRLG]. “Some Reflections on Language Games”. Reprinted in an expanded version. in Wilfrid Sellars. 1963. Science, Perception and Reality, 321–58. London: Routledge and Kegan and Paul. Reissued by Ridgeview Publishing Company in 1991. Hereafter SPR. ———. 1955 [VTM]. “Vlastos and ‘the Third Man’”. Philosophical Review 64 (3): 405–37. ———. (1956) 1991 [EPM]. “Empiricism and the Philosophy of Mind”. In SPR, 127–96. ———. 1957 [LSPO]. “Logical Subjects and Physical Objects”. Philosophy and Phenomenological Research 17 (4): 458–72. ———. 1957 [SFA]. “Substance and Form in Aristotle”. The Journal of Philosophy 54: 688–99. ———. (1961) 1991 [LT]. “The Language of Theories”. In SPR, 106–26. ———. (1962) 1991 [NS]. “Naming and Saying”. In SPR, 225–46. ———. (1962) 1991 [PSIM]. “Philosophy and the Scientific Image of Man”. In SPR, 1–40. ———. (1963) 2007 [AE]. In Wilfrid Sellars. 2007. In the Space of Reasons: Selected Essays of Wilfrid Sellars, edited by Kevin Scharp and Robert B. Brandom, 163–205. Cambridge, MA: Harvard University Press. Hereafter SR. ———. 1963 [EAE]. “Empiricism and Abstract Entities”. In The Philosophy of Rudolf Carnap, edited by Paul Schilpp, 431–68. LaSalle, IL: Open Court. ———. 1966 [FD]. “Fatalism and Determinism”. In Freedom and Determinism, edited by Keith Lehrer, 141–74. New York: Random House. ———. (1967) 1977 [AMI]. “Aristotle’s Metaphysics: An Interpretation”. In Wilfrid Sellars. 1977. Philosophical Perspectives: History of Philosophy, 73–124. Atascadero, CA: Ridgeview Publishing Company. ———. (1973) 1974 [RAL]. “Reason and the Art of Living in Plato”. In Wilfrid Sellars. 1974. Essays in Philosophy and Its History, 3–26. Dordrecht, Holland: D. Reidel. ———. (1974) 2007 [MFC]. “Meaning as Functional Classification: A Perspective on the Relation of Syntax to Semantics”. In SR, 81–100. ———. 1975. [SK] “The Structure of Knowledge: (1) Perception; (2) Minds; (3) Epistemic Principles”. In Action, Knowledge and Reality: Critical Studies in Honor of Wilfrid Sellars, edited by Hector-Neri Castañeda, 295–347. Indianapolis: Bobbs-Merrill. Soames, Scott. 2003. Philosophical Analysis in the Twentieth Century, vols. 1 & 2. Princeton: Princeton University Press.

Part I

Influences

1

Cook Wilson, Sellars, and the Explication of Language Boris Brandhoff

1 Introduction Wilfrid Sellars was truly a synthesizer of philosophical traditions. In his works, quite different threads running through the history of philosophy are woven together. That it was Sellars’s ambition to get twentiethcentury Analytical Philosophy back into dialogue with, on the one hand, the great tradition of Western philosophy, especially Kant, and, on the other, American traditions such as Critical Realism or Pragmatism, had always been clear. In the last few years, detail-oriented, meticulous studies added new facets to our picture of the “historical Sellars”. As a result, we now understand far better how Sellars’s thought took shape under influences as diverse as the philosophy of his father Roy Wood Sellars (Gironi 2017), Gustav Bergmann and the Logical Positivists at the Iowa Department (Olen 2016a, 2016b, 2017), or the French sociologist Émile Durkheim (Olen and Turner 2015). In his article “Logical Subjects and Physical Objects”, published in 1957, Sellars himself pointed his readers to quite another line of influence: “I can say in all seriousness that 20 years ago I regarded Wilson’s Statement and Inference as the philosophical book of the century” (LSPO §1: 458). In this passage, Sellars is referring to the central work of the British philosopher John Cook Wilson (1849–1915), the founding figure of Oxford Realism. “Logical Subjects and Physical Objects” was Sellars’s contribution to a symposium that was held at Duke University in December 1955 in honor of Oxford philosopher Peter Strawson, at that time a rising star at the philosophical firmament. Besides Strawson, not many in the audience will have been familiar with John Cook Wilson or with Statement and Inference, arguably not even by name. According to the manuscript of his talk, Sellars came even more to the fore when he gave his presentation: “I am indeed indebted to the sponsors of this symposium and to Mr. Strawson for sending me back to my well worn copy of this book, which shaped so much of my early thinking on philosophical matters” (Wilfrid S. Sellars Papers, box 33, folder 11). It is beyond dispute that Sellars did indeed engage with Cook Wilson’s philosophy when he was a graduate student at Oxford in the mid-1930s

12

Boris Brandhoff

(see AR 284). Still, what Sellars said about Statement and Inference at the Duke symposium implies more than mere familiarity and appreciation. Was his endorsement just a matter of courtesy, perhaps a rhetorical gesture in order to find common ground with Strawson, the guest from abroad? Or was it the acknowledgment of a sincerely felt intellectual debt? And is it at all likely that of all books it was indeed Cook Wilson’s little-read and somewhat convoluted Statement and Inference that “shaped so much of [Sellars’s] early thinking on philosophical matters”? My chapter is an attempt to explore the largely unknown line of reception that connects Sellars’s thought to the Oxford tradition of realism and ordinary language philosophy. Its question is whether the distinctive understanding of philosophy and philosophical methodology that is the hallmark of Statement and Inference is likely to have influenced Sellars’s own philosophical writing. I’m going to present my answer in three steps: I shall begin by reviewing what we know about how Sellars came into contact with Oxford philosophy and, in particular, with Statement and Inference (Section 2). In Section 3, I take a close look at Cook Wilson’s metaphilosophical and methodological commitments as expressed in this book. My focus is on his idea that philosophy involves the explication of concepts that are embedded in our everyday ways of thinking and speaking about the world we live in. Finally, in Section 4, I address the question whether there is a sense in which Sellars follows up on this idea in his philosophy. To this end, I shall reconstruct his reflections on “explication” and “clarification” in his first essays in a way that brings them into contact with larger concerns of his philosophical project.

2 Cook Wilson and Sellars: Points of Contact That Oxford was an important philosophical point of reference for Sellars, is, of course, not news for any serious student of his work. Sellars’s long-standing engagement with the philosophy of Oxford professor Gilbert Ryle culminated, after all, in the appearance of a mythical community of “Ryleans” at a very prominent position in “Empiricism and the Philosophy of Mind” (see EPM §§ 48–50: 178ff.). Wherever Sellars, in his early works, sets out to present his own account of the logic of “ought”, he takes ethical intuitionism as his point of departure—and he leaves no doubt that it is the variant skillfully defended by Oxford philosophers D.W. Ross and H.A. Prichard that he has in mind (see, e.g., LRB 294, or IIO 227–8). Strawson, finally, is Sellars’s most important contemporary informant for a philosophically refined articulation of ordinary experience, or, to use Sellars’s term, the “manifest image”.1 In all of these cases, Sellars takes a distinctively “Oxonian” view as a dialectical foil for developing the details of his own position—carefully appraising its insights, while correcting what he regards as its flaws. However, things are different with Cook Wilson. The opening paragraphs of “Logical Subjects and Physical Objects” are the only place in

Cook Wilson, Sellars, Explication of Language

13

Sellars’s published work where he explicitly draws on Cook Wilson and Statement and Inference in a systematic manner. Besides that, we find fleeting references to Cook Wilson in a few other articles from the early and the later phase of Sellars’s work.2 But these instances are merely tips of the hat, acknowledging that Sellars is adopting a distinction or a piece of philosophical terminology that was borrowed from Cook Wilson, such as his notion of “thinking without question” (see, e.g., CDCM §77: 281). Taken as it stands, Sellars’s overt engagement with Cook Wilson in his published writings doesn’t prove much more than the uncontroversial claim that Sellars was, in any case, at least cursory familiar with Cook Wilson’s philosophy. But, of course, sometimes even substantial influences may remain largely under the surface. In order to understand how Statement and Inference may have had a formative influence on Sellars’s early thought, we will have to take a closer look at his student days in Oxford. Wilfrid Sellars came to Oxford in the fall of 1934 as a Rhodes scholar. At the age of 24, he was certainly not an impressionable freshman anymore. That summer, he had submitted an ambitious master’s thesis on Substance, Change and Event at the University of Buffalo. The year before, he had obtained a bachelor’s degree from the University of Michigan. At Oxford University, he enrolled for an undergraduate degree and was awarded an A.B. with first class honors in philosophy, politics, and economics in 1936. Sellars remained at the University of Oxford for another year to work on a doctoral thesis on Kant, before transferring to Harvard in order to continue his studies. Thus, Sellars stayed in Oxford for three academic years, from fall 1934 to fall 1937. As Sellars tells us in his “Autobiographical Reflections”, “[he] had already come to think of [himself] as having a system” (AR 284) when he came to Oxford. In the previous years, he had grappled with Kant, Husserl, Cambridge Analysis, and, of course, with the Evolutionary Naturalism and Critical Realism of his father, Roy Wood Sellars. In Oxford, he now got into contact with a school of thought that had, at that time, very little resonance elsewhere in the academic world: When I arrived on the Oxford scene, what struck me with truly revolutionary force was not the surging undercurrent of radical empiricism and positivistic analysis, but rather—and some of you may be surprised to learn this—the framework of philosophical ideas developed by John Cook Wilson and his students, particularly H.A. Prichard. (LSPO §1: 458) Cook Wilson was, in 1934, already dead for almost 20 years but certainly not forgotten. As Fellow of Oriel and, since 1889, as Wykeham Professor of Logic, he had lived and taught at the University of Oxford for more than 40 years. Although he hadn’t published much during his lifetime, he

14

Boris Brandhoff

became, through his lectures as well as through personal communication and correspondence, “the centre of gravity of the school that came to be known as Oxford Realism” (Marion 2000, 305). Oxford Realism combines an anti-representationalist realism in the philosophy of perception and cognition with a distinctive attitude towards metaphilosophical and methodological questions that provided the point of origin for the version of Ordinary Language Philosophy championed by Ryle and Austin.3 Its “central plank” (Marion 2015), however, was its view on the nature of knowledge. Following Cook Wilson, Oxford Realists tended to regard knowledge as an unanalyzable, factive state of mind, coordinate with or more basic than belief (see Marion 2015). In Oxford, Sellars met with an academic environment where the ideas of Cook Wilson were very much alive. Several of his academic teachers saw themselves in his tradition and were, each in his own way, committed to his philosophical project. H.A. Prichard, at that time the White’s Professor of Moral Philosophy, was a particularly clear case in point. A direct student of Cook Wilson, he was the most distinguished exponent of the Oxford Realists in the mid-1930s. Today, Prichard is best known for his defense of intuitionism in moral philosophy, but his studies in the philosophy of knowledge and perception were equally influential at the time. It is in these fields that Prichard took up and developed insights from Cook Wilson and helped to spread Oxford Realism within the Oxford philosophical community. In his “Autobiographical Reflections”, Sellars acknowledges that Prichard had been an important influence on him in his Oxford days and that it was due to him that he turned to Cook Wilson (see AR 284). Besides Prichard, Sellars studied with other prominent Oxford Realists as well. H.H. Price, a student of Prichard who had already done important work in the philosophy of perception himself, led him through another read of Kant (see AR 285); John Austin gave a seminar on C.I. Lewis that Sellars attended (see AR 287). A notebook in Sellars’s Nachlass suggests that he also took at least one course with Ryle.4 During his time in Oxford, then, Sellars was in close contact with philosophers who would constantly point him towards the ideas of Cook Wilson. That he did indeed intensively engage with Cook Wilson’s philosophy, is, to some extent, reflected in his surviving Nachlass as well. There, we find a 17-page typewritten paper by Sellars on Cook Wilson and Prichard that was written during his time as a graduate student in Harvard.5 Moreover, Sellars had annotated copies of both volumes of Statement and Inference and of Richard Robinson’s 1931 study The Province of Logic, the only monograph ever written on Cook Wilson’s book, in his library.6 Thus, our review of circumstantial evidence is complete: Sellars had means, motive, and opportunity for copious engagement with Cook Wilson’s Statement and Inference—and we even have more than one written confession. What remains to be established is whether

Cook Wilson, Sellars, Explication of Language

15

Sellars’s early exposition had any lasting impact on his own philosophical writing. To this end, I shall, in the next section, take a closer look at the distinctive understanding of philosophy and its methodology that is the hallmark of Statement and Inference.

3 Ordinary Usage and the Task of Explication in Statement and Inference The two volumes entitled Statement and Inference: With Other Philosophical Papers certainly are a rather unusual book. Edited posthumously by A.S.L. Farquharson, a former student of Cook Wilson, they include various materials related to the philosophical work and life of his academic teacher. Besides a memoir by Farquharson and a list of Cook Wilson’s publications, we find a compilation of acknowledgments of Cook Wilson by other authors; a selection of Cook Wilson’s private correspondence, ranging from a homesick schoolboy’s letters to his parents to a senior academic’s unsolicited advice to the War Office on military matters in the first months of World War I; and a number of “tentative investigations” on diverse philosophical issues, partly in the form of correspondence with other Oxford dons such as Bernard Bosanquet or H.A. Prichard. But at the center of Farquharson’s edition is a compilation of the logic lectures Cook Wilson had given for almost 40 years at the University of Oxford, at first as Fellow of Oriel, later as Wykeham Professor of Logic. Farquharson reconstructs an ideal text version of these lectures on the basis of source materials covering the four decades from 1874 to 1914. These sources vary widely in scope and quality; they include handwritten and typewritten manuscripts by Cook Wilson as well as lecture notes by students (see Cook Wilson 1926, 886–9). A substantial part of it, however, existed in the form of three volumes of typewritten texts, referred to as Dictata, that were printed by Cook Wilson himself between 1913 and 1914 for private circulation (see Cook Wilson 1926, 877). Farquharson’s reconstruction of Cook Wilson’s logic lectures is presented in his edition under the title Statement and Inference.7 Statement and Inference, in this narrower sense, makes for a surprising read itself. If the reader takes just a cursory look at the table of its contents, the overall arrangement of the material and the titles of its sections seem to suggest a treatise on logic in textbook style. As soon as one begins to read through the volumes, it becomes clear that this book is most certainly not a textbook, and that it deals with many other things besides logic. Statement and Inference is a collection of at times highly original but also somewhat idiosyncratic investigations into topics covering almost the whole range of theoretical philosophy. It is the sum of Cook Wilson’s lifelong struggles with issues from logic, the theory of knowledge, the philosophy of mind and language, and metaphysics, organized around the section headings of a nineteenth-century logic lecture.

16

Boris Brandhoff

One of the most striking features of Statement and Inference is Cook Wilson’s willingness to take on long-standing philosophical problems head on and grapple with them, as Sellars might have said, in a “catch as catch can” manner. In doing so, Cook Wilson takes ordinary linguistic usage as his point of departure, carefully exploring the distinctions inherent in everyday language. As he sees it, there is a systematic reason why philosophers should pay closer attention to common parlance than to philosophical jargon: The authority of language is too often forgotten in philosophy, with serious results. Distinctions made or applied in ordinary language are more likely to be right than wrong. Developed, as they have been, in what may be called the natural course of thinking, under the influence of experience and the apprehension of particular truths, whether of everyday life or of science, they are not due to any preconceived theory. (Cook Wilson 1926, 874) The distinctions drawn by philosophers, on the other hand, do not have to prove their usefulness in the same way. What is more, there is always the risk that a particular philosophical distinction is the result of an erroneous assumption elsewhere in the theory. As a consequence, it is “prima facie more likely to be wrong than what is called a popular distinction” (Cook Wilson 1926, 874). Although Cook Wilson does not use this expression himself, his references to the “natural course of thinking” and the “influence of experience” seem to point to an evolutionary argument that speaks in favor of ordinary linguistic usage. Over the centuries and millennia, our everyday speech has been molded into a more and more finely tuned means for capturing particular facts and situations. In the domain of its original application, i.e., in the world of persons and things we interact with, it is, therefore, prone to be a more reliable guide than abstract philosophical theory (see Cook Wilson 1926, 875). In the end, considerations along these lines are the reason why, for him, “distinctions current in language can never be safely neglected” (Cook Wilson 1926, 46). Cook Wilson strives to put these metaphilosophical insights into practice throughout Statement and Inference. Admittedly, there are longish passages where he treats of philosophical theory, rather than observe details of linguistic usage—especially when he discusses the Classics such as Aristotle. But there are also those passages that Gilbert Ryle, a philosophical grandchild of his, saw as descriptive analyses on a par with those of Husserl (Ryle 2009, 183, 211n.). At the beginning of Statement and Inference we find a particularly memorable, and historically influential, example.8 Here, Cook Wilson attempts to clarify the meaning of “thought”, or “thinking”. Instead of resorting to one of the well-established philosophical doctrines on thought, he turns to what speakers of English

Cook Wilson, Sellars, Explication of Language

17

actually mean when they talk about “thought” or “thinking”, and points out that there is a close connection, according to ordinary usage, between thinking and knowing (Cook Wilson 1926, 34–5). In what follows, this connection is meticulously elaborated. Cook Wilson draws distinctions with regard to different variants of thinking and expounds their relation to knowledge, while at the same time broadening our understanding of “knowing” (Cook Wilson 1926, 35–7). Thus, he arrives at the conclusion that the different forms of thinking, such as reasoning, opining, wondering, or planning, are, in fact, either a particular way of knowing (i.e., reasoning) or depend on that form of knowing (i.e., the latter three). “Thinking”, then, doesn’t refer to a clear-cut compartment in the ontology of mind but rather to a network of quite diverse mental activities clustered around reasoning as the central form of “thinking which is knowing” (Cook Wilson 1926, 38). Cook Wilson understands his analysis of the relationship between thinking and knowing, at least in part, as an inquiry into the meaning of the associated terms, although he stresses that in such investigations we are never far from examining “the nature of a thing” as well (Cook Wilson 1926, 40–1). As he notes in passing, philosophers have indeed always been interested in analyzing the meaning of certain crucial terms (Cook Wilson 1926, 41). For Cook Wilson, any such attempt to arrive at a philosophically adequate meaning analysis should, of course, follow “actual linguistic usage” (Cook Wilson 1926, 40). Doesn’t this turn philosophical inquiries into the empirical investigation of the linguistic behavior of competent speakers? After all, as Cook Wilson points out, “the application of the word to certain things [is] a fact, which we simply recognize as existing” (Cook Wilson 1926, 40). At least insofar as the philosopher is concerned with meaning analysis, her task would then appear to be, essentially, to collect sufficient data about “actual linguistic usage”, either by exercising her own linguistic abilities or by observing the verbal behavior of other speakers, and to formulate convincing hypotheses as to the meaning of the terms she is interested in on this basis. A “philosophical” meaning analysis would be as purely descriptive as the ones to be found in a dictionary, with the only possible difference, perhaps, that the terms are especially grand and the analysis is, with luck, unusually refined. Our philosopher would be able to record and reconstruct the way in which certain words are used, but she would be unable to provide a standard of correctness against which actual linguistic behavior can be put to the test. Cook Wilson realizes this problem quite clearly: It seems to remain that we can only take as data the actual application of the names: if so, we are at the mercy of usage. We cannot criticize it: even if we found in it anything apparently contradictory we should be helpless to decide. (Cook Wilson 1926, 41)

18

Boris Brandhoff

Cook Wilson works out a solution for this problem by adopting a method for which Socrates was his model. In Plato’s early dialogues, Socrates is looking for universally valid definitions for various virtues—e.g., in the Politeia, for the virtue of justice. As Cook Wilson observes, here, as elsewhere, there is not much disagreement among Socrates’s interlocutors as to whether given acts are just or unjust (see Cook Wilson 1926, 28). The real challenge for Socrates consists in reaching agreement about what particular acts of justice have in common. Socrates’s skill is directed at finding the universal that unifies the particulars (see Cook Wilson 1926, 28). In the case of the particular ways that manifest the meanings of the terms we use, the unifying factor is operative within the acting individuals themselves. According to Cook Wilson, what we think and do is governed by “certain principles which exist implicitly in our minds” (Cook Wilson 1926, 42). These principles become palpable in how we respond to particular situations, either in thought or in action. They are, however, not at all transparent to us: We realize them at first only in particular cases; not as definite general or universal rules, of which we are clearly conscious and by which we estimate the particular cases. On the contrary, there is no such formulation to precede the particular cases: the principle lives only in the particulars. (Cook Wilson 1926, 42–3) Similar to Socrates who tried to find the universal of a virtue exemplified in particular virtuous acts, it is the task of the philosopher to make the principles explicit that manifest themselves in our doing and thinking. Since “the principle lives only in the particulars”, the explicative project of philosophy has to start off with particular instances in which it is operative. To find the principle that governs the use of a word that she is interested in, the philosopher will have to carefully look into how she and other speakers actually apply that word in context. She will have to carefully work through particular situations, consider counterexamples, and observe verbal behavior. In all this, it is her own linguistic competence that provides the touchstone. As Cook Wilson puts it, “in every such step we rely upon the rightness of our use of the principle in particular ways” (Cook Wilson 1926, 44). As the principle is implicitly at work in the philosopher’s own mind as well, it makes itself known in “a certain feeling of affinity between particular cases” (Cook Wilson 1926, 42). It is this feeling of affinity that will guide her in determining the class of particular instances that is relevant to the question she is investigating. Just as Socrates needed to rely on his own and his interlocutors’ ability to discern between just and unjust acts, she needs to rely on her competence as a speaker to decide whether, e.g., the word she is examining is used correctly in a given sentence. Else she would run the risk of attempting

Cook Wilson, Sellars, Explication of Language

19

to reconstruct the principle on the basis of instances where it is, in fact, violated. At the outset, then, it should be her aim to get a good overview of the actual usage of the word in particular situations. On this basis, she can begin, at first tentatively, to explicate the principle as a universal rule. Once the philosopher feels that she has a good grip on the rule in question, her descriptive project may get a “normative twist”. She is able to resort to the rule in terms of which she reconstructed the principle as a standard of correctness: “We may now correct some of our applications of the name because we see that some instances do not really possess the quality which corresponds to what we now understand the principle to be” (Cook Wilson 1926, 45). Thus, philosophy regains its “sting” as a critical enterprise. It would give a false impression, however, to suggest that, for Cook Wilson, philosophy would be, essentially, a critique of ordinary language. Cook Wilson saw himself as a logician, even though, by today’s standards, in a rather wide sense of the term. For him, logic is “a study of thinking” (Cook Wilson 1926, 78), and it is the principles that govern the various activities that fall under this rubric that he is after. In this pursuit, the task for the philosopher goes beyond that of describing patterns or regularities in thinking: “Logic doesn’t merely analyse: it makes explicit what was implicit” (Cook Wilson 1926, 49). But Cook Wilson insists that the route that philosophy must take in order to arrive at an adequate explication of the principles behind any form of thinking goes through a careful examination of particular cases. His openness for the details and intricacies of particular instances of linguistic usage leads him to a staunchly non-reductionist attitude towards the diversity of linguistic and mental activities. Instead of taking the “easy way” of assimilating divergent modes of linguistic expression to one or two well-understood “toy models”, he acknowledges that different kinds of activity follow different principles— and have to be investigated in a way that does justice to the special features that set them apart. The guiding idea of Cook Wilson’s non-reductionism is summed up quite accurately in the following remarks on “judgement”: The notion of judgement is unique, it cannot be reduced to any other denomination. We must simply recognize it in its universal character through instances in which we exercise it. (Cook Wilson 1926, 277) For Gilbert Ryle, it was precisely this streak in Statement and Inference that connected Cook Wilson’s approach to the questions of philosophy to the movement of phenomenology that began to flourish on the continent at the turn of the century (see Ryle 2009, 211n.). Cook Wilson did not take any notice of the parallels that existed between his philosophical project and that of Edmund Husserl and his students. But it is not unlikely that Sellars, just as Ryle, did.

20

Boris Brandhoff

4 Philosophical Explication in Sellars’s First Essays When Sellars came to Oxford in 1934, the methodological and metaphilosophical stance he found in Statement and Inference must have geared in well with topics he had just recently been wrestling with. While he was a graduate student at the University of Buffalo, Sellars had studied with Marvin Farber, arguably the most distinguished expert on Husserl and phenomenology in America at that time.9 Through Farber, Sellars got acquainted with Husserl and phenomenology (see AR 283). Thus, Cook Wilson’s philosophical “master strategy”, the recognition of universal structures through a meticulous study of particular instances, will have felt familiar and perhaps even congenial to him. It is not unlikely that Cook Wilson’s interest in linguistic phenomena may have been more appealing to him than Husserl’s investigations of the intentionality of consciousness within the conceptual framework of transcendental phenomenology. After all, Sellars had cherished similar tendencies within Cambridge Analysis and, in particular, in the writings of G.E. Moore, but ultimately found them wanting: he felt that this approach was inadequate, as it failed to live up to the full complexity of the problems at hand (see AR 284). Cook Wilson, on the other hand, combined the phenomenologists’ respect for the particular and the concrete with an emphasis on the careful study of linguistic forms that was also characteristic of Cambridge Analysis. During his student days in Oxford, Sellars will have found in Statement and Inference significant puzzle pieces for what a phenomenological approach may look like after the “linguistic turn”. But did he follow up along these lines in his subsequent work? In this section, I shall argue that, in a sense, he did. To see how, we will have to take a closer look at the earliest phase of Sellars’s philosophical writing. Sellars began his career as a published author in 1947 and 1948 with a series of three essays— "Pure Pragmatics and Epistemology”, “Epistemology and the New Way of Words”, and “Realism and the New Way of Words”—that are overwhelmingly dense and full of remarkable insights. When we take in the bird’s eye view, the large contours present themselves roughly as follows: Sellars’s earliest three essays revolve around two fundamental concerns. On the one hand, they argue for a particular understanding of philosophy and its methodology—an understanding that Sellars depicts as the guiding insight of analytic philosophy. Against the background of this conception of philosophy, they furthermore address the task of constructing a metatheoretical framework for clarifying the notion of an “empirically meaningful language”, i.e., a language “that is used in the world it is about” (ENWW 654). In doing so, PPE, ENWW, and RNWW constitute an attempt to bring out what structure a language must have in order to embody empirical knowledge. Thus, we can discern a metaphilosophical concern and a metaepistemological concern in Sellars’s earliest essays.10

Cook Wilson, Sellars, Explication of Language

21

The metaphilosophical project of Sellars’s earliest essays starts out from the old conundrum of philosophy’s relation to the empirical sciences. In Sellars’s view, analytic philosophy is characterized by a commitment to the idea that philosophy is not science (see PPE 181). As he sees it, it is by pointing out that philosophical statements are logically different from the statements employed by the empirical sciences, that a clearcut demarcation line between philosophy and science can be established (see PPE 181, 202). Following Sellars, we shall refer to the latter class of statements as “factual” statements (see ENWW 646). Within the context of Sellars’s earliest three essays, the term “factual statement” has a clearly determined meaning. Factual statements are tokens of sentences that are subject to the standard of factual truth. As such, these sentences are supposed to appear in what Sellars calls a “world-story”, very roughly a coherent and exhaustive set of elementary sentences that enables a language to represent the world in which it is used (see ENWW 647–8; RNWW 609–10). When tokened, i.e., as statements, they are supposed to function as constituents of a dynamic picture of the language user in her environment (see RNWW 610). Philosophical sentences, however, are not subject to the standard of factual truth (see PPE 202). Consequently, philosophical statements are not supposed to function as constituents of a dynamic picture of the language user in its environment. Let us refer to these purely negative claims about philosophical sentences and statements as metaphilosophical non-factualism. In his earliest three essays, Sellars’s metaphilosophical non-factualism is backed up by a positive account of the function performed by philosophical statements, or rather, by two such accounts—one neat and clearcut, the other somewhat confusing. From the external point of view of an empirical scientist who studies the behavior of language-using animals, Sellars characterizes this function as being related to “the ‘clarification’ or ‘unpacking’ of linguistic phenomena” (RNWW 625). For the empirical scientist, the clarification of linguistic phenomena that is the task of philosophy is achieved by a “confrontation of habit by habit”: The philosopher develops a higher-level system of linguistic habits that regulates his own lower-level habits in order to correct their inconsistencies and “confusions” (see PPE 202). The point of the “explicative metalinguistic activity” (CIL 314) of the philosopher, seen as a natural phenomenon, is to enforce that lower-level linguistic activities conform to certain causal patterns. When Sellars speaks as a philosopher, i.e., when he adopts the “internal” point of view of a philosopher who engages in distinctively philosophical linguistic activity, he uses the exact same terms to characterize the function of what he does. Here are two examples: Returning, then, to formal considerations .  .  ., we must come to a final reckoning with naive realism. This reckoning will consist in

22

Boris Brandhoff a clarification of the relation between an object-language and “the world it is about”. (PPE 197–8) In order to explicate the grammar of epistemological predicates, we need to consider them in use, that is to say, as applied to the expressions of an object language. (ENWW 648, n. 5)

Evidence for this manner of speaking is ample throughout Sellars’s earliest essays. Given his external account as to what it takes for philosophical activity, seen as a natural phenomenon, to achieve the clarity it is after, one would expect philosophical statements to set out, clearly and distinctly, standards to which lower-level linguistic activity, according to the philosopher, ought to conform. If philosophical statements were indeed statements of rules for lower-level linguistic activity, this would give us an idea why, by stating them, the philosopher is able to govern her own lower-level linguistic activity in such a way as to make sure that it conforms to the patterns she appreciates. Sellars tells us that philosophical statements do indeed set out such a standard: they are supposed to contribute to the formulation of an ideally coherent system of norms for language use (see PPE 192; RNWW 608–9, 633–4; Brandhoff 2017, 57–8). This ideally coherent system of norms provides the standard against which lower-level linguistic activity can be clarified and criticized (see PPE 192; RNWW 622–3). So far, Sellars’s view of the distinctive conceptual role of philosophical statements seems to accord well with his “external” account of philosophical activity. But he also insists, rather confusingly, that philosophy is a “formal science” (PPE 193) or even “pure formalism” (PPE 202), and that “a philosophical concept must be decidable on purely formal grounds” (PPE 191; italics omitted). Peter Olen has presented convincing evidence for the claim that Sellars’s use of the term “formal” in his earliest essays appears to be “radically inconsistent” (Olen 2016a, 37), if it is understood along the usual lines of the analytic tradition as meaning either “structural”, “logical”, or “metalinguistic” (see Olen 2016a, 37–8; Olen 2016b, 12). I have proposed to regard Sellars’s use of “formal” in his earliest essays as a first, searching attempt to indicate the functional dimension of concept-use. Sellars’s apparent commitment to a “formalistic” understanding of philosophy would then appear, in retrospect, as an initial step towards the idea that the philosopher addresses language as a system of functional roles that serves as a standard for linguistic behavior (see Brandhoff 2017, 65). Sellars begins to develop the vocabulary of an overtly functionalist metatheory of conceptual activity just a few years later in “Quotation Marks, Sentences, and Propositions” and “The Identity of Linguistic Expressions and the Paradox of Analysis”, two papers published in 1950.

Cook Wilson, Sellars, Explication of Language

23

There, he introduces the notion of a functional role, although his terminology is still shifting. In the latter paper, he illustrates what he means by drawing on an analogy to the game of chess: If one were asked “What is a knight in the game of chess?” one might be tempted to reply by describing the shape of a knight in one’s chess set. This, however, would clearly be a mistake. . . . Fundamentally, then, chessmen are roles and the roles are specified by the rules of chess. Just as chessmen are roles specified by rules, so the words of a language are roles specified by rules. (ILE 27) Functional roles are shared by linguistic items insofar as they are subject to the same, or relevantly similar, linguistic rules. I think it is helpful to think of them as classes of rules that govern the use of particular classes of linguistic items. If we look back at what we have said earlier about Sellars’s view of the function of philosophical statements in his earliest essays, we can now introduce the distinction between “explication” and “clarification” in the following way: from the internal point of view of the philosopher who engages in distinctively philosophical activity, philosophical statements explicate functional roles, in the sense of specifying the individual rules that belong to the class in question. In doing so, the philosopher contributes to an increasingly complex account of language as a system of functional roles. It goes without saying that the fineness and precision in which the philosopher specifies the rules she is interested in is, in each case, subject to pragmatic considerations. Our philosopher has achieved a philosophical clarification when she succeeds in bringing it about that her own linguistic activity conforms to the functional roles she has previously explicated. The philosopher has cleared up a “confusion” within her own linguistic activity by replacing a deficient set of linguistic habits with one that is in line with the rules she has set out. It is obvious that, in fact, explication often fails to lead to clarification, but this is not a philosophical matter.

5 Conclusion: A Lasting Inheritance? When the early Sellars began to think of the role of philosophical statements as that of making explicit the norms that implicitly guide the doing and saying of language users, he picked up the predominant theme in Cook Wilson’s reflections on philosophy and its methodology. Although his stance on metaphilosophical questions did undergo, without any doubt, some change and development throughout four decades of philosophical writing, I do not think that Sellars ever let go of the insight that the task of the philosopher involves the explication of functional roles that provide a norm for conceptual activity. At a much later stage of his

24

Boris Brandhoff

philosophical development, in the opening paragraphs of his Matchette Foundation Lectures “The Structure of Knowledge” delivered at the University of Texas in 1971, Sellars characterized “the ideal aim of philosophizing” as that of becoming “reflectively at home in the full complexity of the multi-dimensional conceptual system in terms of which we suffer, think, and act” (SK I §3: 295). Whatever else it may take to count as being “reflectively at home” in the intricacies of conceptual activity, becoming aware of, and critically endorsing, the principles that govern our use of concepts will be an important part of the process. Cook Wilson and Sellars agree in what I would propose to refer to as an explicative understanding of philosophy: The idea that to philosophize is to make explicit, and hence available for critical discussion and scrutiny, the norms that structure the way in which we conceive ourselves and the world we live in and interact with. But agreement with regard to the aim of philosophy does not necessarily imply agreement as to its proper modus operandi. In contrast to Cook Wilson and his Oxonian successors, Sellars took full measure of the idea that modern science constantly challenges, and transforms, our picture of both ourselves and the world. For Sellars, scientific discourse is part and parcel of our conceptual system. Consequently, the pressure science exerts on “the manifest image of man-in-the-world” (PSIM) is nothing alien or external, but rather an aspect of the internal dynamics of conceptual change. Sellars’s commitment to naturalism informs his take on the explicative project of philosophy on two levels. As a philosopher of science, he regards scientific concepts, theories, and methods as an important subject matter for philosophical explication. But there is more to it. For Sellars, science is supposed to provide the understanding of the world in the light of which philosophy would, ideally, carry out its enterprise. After all, functional roles are, essentially, ought-to-be uniformities that regiment the way in which linguistic and mental episodes cohere with one another, with modes of behavior, and with relevant circumstances in the environment. And it is science that tells us what these things really are. Thus, “the task of explicating the concept of a rational animal” becomes, for Sellars, that of explicating the concept of “a language-using organism whose language is about the world in which it is used” (SK III §46: 346). Interestingly, however, these tendencies appear to set Sellars apart more from Ryle and Strawson than from Cook Wilson himself who did extend his method to the logical foundations of nineteenth-century science (see Cook Wilson 1926, parts III and IV).11 In some of the most compelling passages in his philosophical writing, Sellars arrives at systematic positions by carefully examining particular cases. Think of his “super-inscriber” in “Truth and ‘Correspondence’” (see TC 219–24); the anthropoid robot in “Being and Being Known” (see BBK §§36–44: 51–4); John, the necktie salesman, in “Empiricism and the Philosophy of Mind” (see EPM §§14–15: 142–4), or the various

Cook Wilson, Sellars, Explication of Language

25

appearances of Jones—in all these cases, Sellars does not merely provide illustrations, but he constructs models that are supposed to bear an important part of the burden of his argument. In his own way, he takes heed of Cook Wilson’s insight that “the principle lives only in the particulars” (Cook Wilson 1926, 43). Where Cook Wilson carefully looks into instances of ordinary linguistic usage, Sellars tells stories that synthesize abstract scientific findings and philosophical insights into a concrete, vivid scenario. Both are, arguably, strongest when they reason “through the particulars” in this way. It should not surprise us, then, that even Sellars’s famous Jones may have been an Oxonian by birth (see Cook Wilson 1926, 115–17 and 123–6). Is it really so unlikely, all things considered, that Statement and Inference might indeed have struck young Sellars as “the philosophical book of the century” back then when he studied in Oxford? I’m not even sure if I should find such a judgment imperceptive or naïve, for there is, without any doubt, something forceful and original in these two, idiosyncratic volumes. But in 1934, the twentieth century was still young. Sellars took up leads towards an explicative understanding of philosophy he found, among others, in Cook Wilson, Cambridge Analysis, and Farber’s reading of Kant and Husserl and transformed them into his very own framework of metatheoretical functionalism. And although Cook Wilson’s book may never have been a serious contender for philosophy’s hall of fame, it did exert some influence on its course, and was at the source of some of the finest philosophical writing done in the past century—Sellars’s included.

Notes 1. Sellars devoted two articles to the task of criticizing positions championed by Strawson (i.e., PRE and LSPO). Besides, there are numerous references to Strawson throughout Sellars’s published work (see, e.g., SM or I). Strawson is actually one of the most cited contemporary authors in Sellars’s writings. 2. Namely, in VTM (423), CDCM (§77: 281), GEC (616), FMPP (§119: 26), and in WSNDL (330). 3. See Marion (2000) for a detailed survey of the development of Oxford Realism, with special emphasis on knowledge and the theory of perception and cognition, Marion (2009) for an account of the broader context in the history of British philosophy in which Oxford Realism evolved, and Marion (2011) for a study of Cook Wilson’s influence on Ordinary Language Philosophy. 4. See Wilfrid S. Sellars Papers, box 9, folder 3. 5. See Wilfrid S. Sellars Papers, box 7, folder 12. 6. See Wilfrid S. Sellars Papers, Series V: Library, Subseries 1: Annotated Books, items 29, 30 and 709. 7. I was unable to find conclusive evidence as to whether this title can be traced back to Cook Wilson or whether it was an invention of Farquharson. 8. The discussed passage is the point of departure for a complex line of argument that leads to the conclusion that knowledge is more basic than belief. This idea was, in recent years, taken up, developed, and eloquently defended by, e.g., John McDowell and Timothy Williamson, both Oxford philosophers.

26

Boris Brandhoff

9. Farber had been a direct student of Husserl in the 1920s and became subsequently very influential in spreading the ideas of phenomenology in America. He founded Philosophy and Phenomenological Research, served as president of the International Phenomenological Society for 40 years, and was the author of influential studies on phenomenological issues and Husserl in particular, such as Farber (1943) (see Cho and Rose 1981). 10. Sellars characterizes the philosophical project of his earliest essays as an attempt to bring “Kant’s Copernican Revolution” to “its non-psychologistic fruition” (PPE 185). For a discussion of the transcendental dimension of Sellars’s epistemology, see Haag (2019), this volume, section II. 11. In doing so, however, Cook Wilson was strikingly out of step with cuttingedge developments in then-contemporary science. Thus, he argued that the notion of a non-Euclidean space is self-contradictory (see Cook Wilson 1926, 561–68).

References Brandhoff, Boris. 2017. “Pure Pragmatics and the Idea of a Metatheoretical Functionalism”. In Sellars and Contemporary Philosophy, edited by David Pereplyotchik and Deborah R. Barnbaum, 55–66. New York and London: Routledge. Cho, Kah Kyung and Lynn E. Rose. 1981. “Marvin Farber (1901–1980)”. Philosophy and Phenomenological Research 42 (1): 1–4. Cook Wilson, John. 1926. Statement and Inference with Other Philosophical Papers, 2 vols, edited by A. S. L. Farquharson. London: Oxford University Press. Farber, Marvin. 1943. The Foundation of Phenomenology. Cambridge: Harvard University Press. Gironi, Fabio. 2017. “A Kantian Disagreement between Father and Son: Roy Wood Sellars and Wilfrid Sellars on the Categories”. Journal of the History of Philosophy 55 (3): 513–36. Haag, Johannes. 2019. “Transcendental Principles and Perceptual Warrant: A Case Study in Analytic Kantianism”. In Wilfrid Sellars and Twentieth-Century Philosophy, edited by Stefan Brandt and Anke Breunig, 130–50. London: Routledge. Marion, Mathieu. 2000. “Oxford Realism: Knowledge and Perception”. British Journal for the History of Philosophy 8: 299–338 and 485–519. ———. 2009. “Theory of Knowledge in Britain from 1860 to 1950: A NonRevolutionary Account”. Baltic International Yearbook of Cognition, Logic and Communication 4: 1–34. ———. 2011. “Cook Wilson and Austin on Knowledge, Wittgenstein, and the Rise of Ordinary Language Philosophy”. In John L. Austin et la philosophie du langage ordinaire, edited by Christophe Al-Saleh and Sandra Laugier, 81–105. Hildesheim: Georg Olms. ———. 2015. “John Cook Wilson”. In The Stanford Encyclopedia of Philosophy, Spring 2016 edition, edited by Edward N. Zalta, online at https://plato.stan ford.edu/archives/spr2016/entries/wilson. Olen, Peter. 2016a. “From Formalism to Psychology: Metaphilosophical Shifts in Wilfrid Sellars’s Early Works”. HOPOS: The Journal of the International Society for the History of the Philosophy of Science 6: 24–63.

Cook Wilson, Sellars, Explication of Language

27

———. 2016b. Wilfrid Sellars and the Foundations of Normativity. London: Palgrave Macmillan. ———. 2017. “A Forgotten Strand of Reception History: Understanding Pure Semantics”. Synthese 194 (1): 121–41. Olen, Peter and Stephen Turner. 2015. “Durkheim, Sellars, and the Origins of Collective Intentionality”. British Journal for the History of Philosophy 23: 954–75. Ryle, Gilbert. 2009. Critical Essays: Vol. 1 of Collected Papers, 2nd edition. London: Routledge. Sellars, Wilfrid. Wilfrid S. Sellars Papers, 1899–1990, ASP.1991.01, Archives of Scientific Philosophy, Special Collections Department, University of Pittsburgh. URL: https://digital.library.pitt.edu/collection/wilfrid-s-sellars-papers. Sellars, Wilfrid. 1947 [ENWW]. “Epistemology and the New Way of Words”. The Journal of Philosophy 44 (24): 645–60. ———. 1947 [PPE]. “Pure Pragmatics and Epistemology”. Philosophy of Science 14 (3): 181–202. ———. 1948 [CIL]. “Concepts as Involving Laws and Inconceivable without Them”. Philosophy of Science 15 (4): 287–315. ———. 1948 [RNWW]. “Realism and the New Way of Words”. Philosophy and Phenomenological Research 8 (4): 601–34. ———. 1949 [LRB]. “Language, Rules and Behavior”. In John Dewey: Philosopher of Science and Freedom, edited by Sidney Hook, 496–505. New York: The Dial Press. ———. 1950 [ILE]. “The Identity of Linguistic Expressions and the Paradox of Analysis”. Philosophical Studies 1 (2): 24–31. ———. 1950 [QMSP]. “Quotation Marks, Sentences and Propositions”. Philosophy and Phenomenological Research 10 (4): 515–25. ———. 1954 [PRE]. “Presupposing”. The Philosophical Review 63 (2): 197–215. ———. 1955 [VTM]. “Vlastos and ‘the Third Man’”. Philosophical Review 64 (3): 405–37. ———. (1956) 1991 [EPM]. “Empiricism and the Philosophy of Mind”. In Science, Perception and Reality, 2nd edition, edited by Wilfrid Sellars, 127–196. Atascadero: Ridgeview. ———. 1956 [IIO]. “Imperatives, Intentions, and the Logic of ‘Ought’”. Methodos 8: 228–68. ———. 1957 [CDCM]. “Counterfactuals, Dispositions, and the Causal Modalities”. In Concepts, Theories, and the Mind-Body Problem, edited by Herbert Feigl, Michael Scriven, and Grover Maxwell, 225–308, vol. 2 of Minnesota Studies in the Philosophy of Science. Minneapolis: University of Minnesota Press. ———. 1957 [LSPO]. “Logical Subjects and Physical Objects”. Philosophy and Phenomenological Research 17 (4): 458–72. ———. (1960) 1991 [BBK]. “Being and Being Known”. In Science, Perception and Reality, 2nd edition, edited by Wilfrid Sellars, 41–59. Atascadero: Ridgeview. ———. (1962) 1991 [PSIM]. “Philosophy and the Scientific Image of Man”. In Science, Perception and Reality, 2nd edition, edited by Wilfrid Sellars, 1–40. Atascadero: Ridgeview. ———. (1962) 1991 [TC]. “Truth and ‘Correspondence’”. In Science, Perception and Reality, 2nd edition, edited by Wilfrid Sellars, 197–224. Atascadero: Ridgeview.

28

Boris Brandhoff

———. (1967) 1992 [SM]. Science and Metaphysics: Variations on Kantian Themes, 2nd edition. Atascadero: Ridgeview. ———. 1972 [I]. “. . . This I or He or It (The Thing) Which Thinks”. Proceedings of the American Philosophical Association 44: 5–31. ———. 1973 [GEC]. “Givenness and Explanatory Coherence”. The Journal of Philosophy 70: 612–24. ———. 1974 [MFC]. “Meaning as Functional Classification: A Perspective on the Relation of Syntax to Semantics”. Synthese 27 (3–4): 417–37. ———. 1975 [AR]. “Autobiographical Reflections”. In Action, Knowledge and Reality: Critical Studies in Honor of Wilfrid Sellars, edited by Hector-Neri Castaneda, 277–93. Indianapolis: Bobbs-Merrill. ———. 1975 [SK]. “The Structure of Knowledge”. In Action, Knowledge and Reality: Critical Studies in Honor of Wilfrid Sellars, edited by Hector-Neri Castaneda, 295–347. Indianapolis: Bobbs-Merrill. ———. 1981 [FMPP]. “Foundations for a Metaphysics of Pure Process”. The Monist 64: 3–90. ———. 2009 [WSNDL]. Notre Dame Lectures: 1969–1986: The Bootleg Version, edited by Pedro Amaral, online at www.ditext.com/amaral/wsndl.pdf.

2

Sellars’s Twist on Carnap’s Syntax Anke Breunig

When we explore the philosophical relationship between Wilfrid Sellars and Rudolf Carnap, we encounter a number of misunderstandings. The first is part of the common philosophical lore about the history of twentieth-century philosophy. According to this story, logical empiricism was rendered obsolete by the joint efforts of Quine and Sellars.1 In particular, it is thought that Sellars completely dismantled Carnap’s view, presumably the one from the Aufbau (Carnap [1928] 1967), as a paradigmatic case of the Myth of the Given. However, it is not at all clear that Carnap ever held any view that can be straightforwardly classified as givennist,2 or that Sellars thought that he did. In fact, Sellars rarely comments on Carnap’s Aufbau at all. Leaving this debate on one side, this chapter focuses on a second misunderstanding, which we can find in a rather disappointing written exchange between Sellars and Carnap (see EAE; Carnap 1963). This is a case of the two philosophers misunderstanding each other. Here3 as elsewhere, Sellars expresses great enthusiasm for Carnap’s Logical Syntax of Language: Today, for the first time, the naturalistic-empiricist tradition has the fundamentals of an adequate philosophy of mind. To the creation of this truly revolutionary situation, which is just beginning to make itself felt, Carnap’s Logical Syntax of Language and Introduction to Semantics have contributed at least as much as any other single source. (EAE 468) While very sympathetic to his general approach, Sellars sharply criticizes some parts of Carnap’s view. His main criticism of Carnap’s syntax,4 I will argue, is that it lacks an adequate conception of rules.5 For Sellars, rules are norms for doing something or behaving in certain ways. They always contain a prescriptive component. Rules of language specify what our linguistic behavior should be. Carnap, on the other hand, uses the term “rule” in a more contrived sense. Take, for example, rules of inference or so-called transformation rules. According to Carnap, such rules

30 Anke Breunig assert structural relations between sentences and collectively constitute the definition of “direct consequence in S”. Sellars objects that “direct consequence” is an implicitly normative notion, meaning that if we are entitled to assert some sentences, we also have the license to assert some other sentences as well. In a nutshell, he thinks that Carnap’s definition must be inadequate because it attempts to define a normative concept in non-normative terms. Sellars regards the failure to account for the normativity of linguistic rules as a grave mistake. Trying to present the most charitable reading, however, he looks for hints that Carnap might be willing to acknowledge the “normative flavour” of rules. Here is a passage from “Inference and Meaning” (IM), where Sellars makes much the same point as in EAE: [W]hen one turns to Carnap’s thesis that transformation rules might be formulated as definitions of “direct consequence in S”, one finds no such [normative] flavour. The term “direct consequence” has the same sort of feel as “next to” or “between”. This is not true of the predicate i.e., “derivable”. . . . The term “derivable” is one of those “able” words which connotes “may be done” in the sense of . . . “is permissible”.  .  . . Now it is my impression that when Carnap was looking for another word to share the burden of transformation rules formulated as definitions with “directly derivable”, he failed to bear in mind that what he needed was another word with the same rulish force. (IM 18) In his answer to EAE, Carnap rejects this suggestion as a misinterpretation: Above all, I wish to emphasize that not only pure syntax and pure semantics but also descriptive syntax and descriptive semantics, as I understand them and intend to construct them, do not contain any kind of prescriptive components. . . . Sellars’ belief that my descriptive syntax and descriptive semantics contained prescriptive components is perhaps due to the fact that I used the word “rule” both in syntax and semantics. Perhaps he understood this term in its everyday sense, i.e., as referring to prescriptive rules, prohibitions, or permissions. However, I use the word “rule” in this field only in order to conform to the customary usage in logic. (Carnap 1963, 923) Clearly Carnap is not willing to acknowledge that syntactic and semantic rules are norms. But his answer is less than satisfactory because he shows no sensitivity to the worries that led Sellars to insist upon the normative character of language. Indeed, Sellars thinks that any theory of meaning that ignores this point will be inadequate.

Sellars’s Twist on Carnap’s Syntax

31

Recently, commentators have increasingly come to recognize both Sellars’s esteem for Carnap’s views and the positive influence Carnap had on Sellars’s thinking. Some have also held that Sellars misinterpreted and even distorted Carnap’s views in trying to make them his own.6 As will emerge in the course of this chapter, I don’t think this is true. In particular, Sellars’s suggestion that Carnap should recognize the normative character of linguistic rules is not simply a misreading. Indeed Sellars’s objection should have worried Carnap much more than it did. We will see that there are problems internal to Carnap’s philosophy that can be solved by adopting a normative conception of language. This is also the road from Carnap’s syntax to Sellars’s inferentialist semantics. In what follows, I trace the initial steps leading to this road.

1 Carnap’s Insight In his book Logical Syntax of Language, dating from 1934,7 Carnap develops a theory with which to describe syntactic and logical relations between expressions of a language in an exact manner. In formulating a syntactic system, the role of the logician, according to Carnap, is to devise combinatorial rules and to investigate which consequences these rules entail.8 In order to show how a logician is supposed to proceed, Carnap defines two formal languages that serve as examples of the method. Carnap distinguishes syntax from semantics and pragmatics. The subject matter of syntax is confined to the relations holding between linguistic expressions. Rules of syntax determine the structure of a language.9 To give an account of the relation between linguistic expressions and their objects of reference falls within the task of semantics. Pragmatics also takes into consideration the use of language. To describe a language comprehensively, we must describe its syntactic, semantic, and pragmatic aspects. But if our focus is on the syntactic part of a linguistic theory only, we must abstract from the semantic and pragmatic features of a language. Since syntax doesn’t presuppose, but is presupposed by semantics and pragmatics, it is more fundamental than any of these.10 In spite of the fact that a language, understood in the usual sense, contains features that cannot be covered by syntax,11 Carnap tends to identify a language with a system of syntactic rules, i.e., a calculus.12 Logical syntax contains two kinds of rules, formation and transformation rules. Formation rules specify for a certain language L how expressions may be combined so that the result is a sentence. Together, the formation rules amount to a definition of “sentence in L”.13 Traditionally, syntax was seen as grammar, which is why it was taken to contain formation rules only. But Carnap saw that rules of inference could be treated in parallel fashion, namely, as rules of transformation licensing transitions from certain sentences of a language L to other ones in the same language.14 As mentioned, the system of transformation rules would make

32 Anke Breunig for a definition of “direct consequence in L”.15 It would also determine all logical properties of (such as “. . . is analytic/synthetic/contradictory”) and relations between sentences of L (e.g., “. . . contradicts . . .”),16 as well as fix the meaning of the logical signs of L. In this way, Carnap was hoping to make logic part of syntax, calling the result “logical syntax”. Because syntax deals only with linguistic expressions and their relations to each other, it is a formal theory.17 Carnap writes: By the logical syntax of a language, we mean the formal theory of the linguistic forms of that language. . . . A theory, a rule, a definition or the like is to be called formal when no reference is made in it either to the meaning of the symbols (for example, the words) or to the sense of the expressions (e.g. the sentences), but simply and solely to the kinds and order of the symbols from which the expressions are constructed. (Carnap [1934] 1937, 1) Carnap is thus suggesting that syntactic investigations do not presuppose knowledge of meaning. As we will see, this claim is extremely important for Sellars’s reception of Carnap. In his syntactic period, Carnap identifies philosophy with the study of the logical syntax of language.18 This he does, first, by limiting philosophy in general to philosophy of language, and second, by limiting philosophy of language to logical syntax. Here, I want to focus on the second thesis. We have seen that Carnap acknowledges the fact that what we ordinarily call languages have properties that cannot be investigated within syntax. Having limited philosophical to syntactic investigations, Carnap holds that these other properties are also of no philosophical relevance. It is the task of the empirical sciences to investigate all nonsyntactic properties of a language. Thus it follows that philosophy, more specifically, philosophy of language, is not concerned with the meanings of symbols and the sense of expressions. However, it is a bit misleading to say that for Carnap before his semantic turn the investigation of meaning was no part of philosophy of language.19 In syntax, we can indeed give the meaning of an expression if it is determined by the logical relations between expressions alone. This is so in the case of the logical expressions of a language.20 In what follows, I will call this kind of meaning the logical meaning of an expression. Only where meaning is not sufficiently determined by relations between linguistic expressions does it become intractable for syntax and thus philosophy in general. Such is the case with empirical meaning. Nor is it entirely true that empirical meaning is intractable for syntax. We can say something about the meaning of empirical sentences. For this, Carnap makes use of the notion of content. Content can be seen as a

Sellars’s Twist on Carnap’s Syntax

33

syntactic property of sentences. Carnap gives the following formal definition of content21 for a sentence S1 or class of sentences K1: By the content of K1 (or of S1 .  .  .) in S, we understand the class of the non-valid sentences which are consequences of K1 (or S1, respectively). (Carnap [1934] 1937, 175)22 Carnap here uses the notion of consequence to define content. For languages containing only logical rules of transformation, the non-valid sentences are the non-analytic sentences of the language. For languages containing also physical or, in Sellars’s terms, material rules, the nonvalid sentences are those that are neither logically nor physically valid.23 Note that it follows from this definition that only non-valid sentences or classes of sentences that contain at least one non-valid sentence have content. Carnap thinks that the valid sentences don’t have content since they affirm nothing about reality. Therefore, they are also not part of the content of another sentence.24 The definition brings to mind Sellars’s view of meaning as an expression’s functional role, where this essentially includes its inferential role.25 It is also worth noting that Carnap’s concept of content is clearly holistic. Because it is not possible to determine the empirical meaning of non-valid sentences within logical syntax, however, Carnap’s definition of content is no substitute for a non-syntactic theory of empirical meaning.26 Sellars emphatically agrees with Carnap that in philosophy of language we should proceed formally, i.e., we should not presuppose a prior notion of meaning. For Sellars, there is no meaning outside of relations between symbols. In particular, we should not assume that we have something like a pre-linguistic awareness of ourselves or the world, either as whole or in parts. Invoking a prior notion of meaning typically involves the thought that pre-linguistic awareness is the source of linguistic meaning. If such awareness is conceived atomistically, it is a clear case of the Myth of the Given.27 Let me take the opportunity to drop a few hints about what I take the Myth of the Given to consist in.28 Clearly, in his attack on the Given, Sellars means to criticize certain theories of justification. Indeed, I take it that all versions of foundationalism committed to internalism make use of the Given. But besides trying to dismantle a theory of justification, Sellars also worries about intentionality. To think that there exist privileged intentional states that possess intrinsic intentionality is also a version of the Myth of the Given. I will say that a state is intrinsically intentional if the fact that it is intentional is independent of any relations it may have to other intentional states of the same kind. The structure of foundationalism about intentionality is parallel to that of foundationalism

34 Anke Breunig about epistemic justification. This time, it is aboutness which is explained atomistically. This can only be done by claiming that the content of some mental states is given.29 It is the latter version of the Myth of the Given that is most relevant to our current concerns. I will call the principle that we should proceed formally “Carnap’s insight”. This principle is very important for Sellars, because it rules out making use of the Myth of the Given when giving a philosophical account of meaning. Insofar as meaning figures in syntax, it meets the following criterion: syntactically tractable meaning results from an interplay of elements that do not possess meaning when considered in isolation. We must respect Carnap’s insight in one way or the other if we aim to offer a philosophical theory of language that is to be free of the Myth of the Given.30 If we follow Carnap, we can ask, however, whether it is necessary to resort to the Myth of the Given in order to deal with those questions that Carnap excluded from the philosophy of language. To answer this question, we must investigate which account of empirical meaning is compatible with Carnap’s position. This is what I want to do in Sections 3 to 5. As I will show, every account of empirical meaning that respects the constraints Carnap has imposed turns out to be deeply problematic. Before turning to the question of empirical meaning, however, I will take a closer look at Carnap’s account of logico-mathematical relations in the next section and formulate the first problem that Carnap faces.

2 Avoiding Psychologism Carnap holds, as most others do, too, that logico-mathematical relations31 have certain modal properties. For example, there is a relation of necessity between premises and conclusion in a valid inference, since the conclusion must be true if the premises are true. The modality in question may be called logical modality.32 Logical modality demarcates the subject matter of logic. Carnap strictly opposes psychologism. He shares this opposition with Frege and other logicians of his time.33 According to psychologism, logical relations are nothing but psychological dispositions. Being not just an opponent of psychologism but also a conventionalist, Carnap thinks that the logical properties of an object language depend on our choice of rules. These rules are formulated in the metalanguage. The dependency of properties of the object language on metalinguistic rules, however, doesn’t matter much as long as we stick to one set of rules. Relative to a specific set of rules, the modal properties of the object language are determined. For Carnap, pure syntax is analytic and a priori. The pairing of an object language with a set of rules presents no problem for a purely syntactic investigation. In pure syntax the object language is created at the same moment the rules are formulated. It does not exist independently,

Sellars’s Twist on Carnap’s Syntax

35

but owes its existence to a logician’s activity in constructing languages. The logician does not describe existing regularities but devises rules or definitions. By formulating a set of rules in the metalanguage, she defines the object language into existence.34 If asked why logical relations have logically modal properties, Carnap would possibly refer to the special subject matter of logic, which he contrasts with the subject matter of the empirical sciences. Logic deals with stipulations, not facts, and it is concerned with language, not with the world. As Carnap points out, all we do in pure syntax is to give definitions and to investigate which consequences these definitions entail. The rules or laws of logic determine abstract relations between signs. In pure syntax, we don’t ask whether there is someone who actually uses or has used the language that we have defined by some set of rules, either in spoken or written form. We also cannot falsify a law of logic by pointing to a chain of sign tokens that does not conform to the law. A law of logic becomes obsolete only if we choose to give it up or to substitute it with a different law. It is not amenable to empirical confirmation or disconfirmation. It thus seems that we are justified in claiming a special status for logically modal properties, and in attributing logically modal properties to logical relations, insofar as a clear distinction between logical and empirical investigations can be drawn.35 But if we make such a distinction the question arises: how do logically modal properties pertain to a language in actual use? Carnap tends to contrast languages designed as calculi with languages that have developed historically. He calls the latter kind natural languages. Because historical word languages are usually much more complicated than symbolic languages that have been deliberately constructed, it is very difficult to come up with explicit rules for them and, ultimately, to formulate a syntactic system that the language will conform to. A language designed as a calculus, however, must be distinguished not only from languages that have emerged historically but also from artificial languages in use.36 It is, so to say, in the fictional rubric “Suppose a language L . . .” (Sellars, EAE 457). I will thus contrast calculi with languages in use. It does not matter much whether the latter are historical languages or artificially constructed ones, nor if their elements are words or symbols. If we regard a language as a calculus, we abstract from all possible use. A language in use is not to be identified with a language considered as a calculus. This is so even in the extreme case where we deal with the same symbol system in both cases.37 But now we must ask what relation there is between the system of logical rules, a calculus, and a language in use. In distinguishing between pure syntax and descriptive syntax, Carnap addresses this question. In pure syntax, we set up the rules in the metalanguage, thereby also defining the object language. In descriptive syntax, we investigate (chains of) sign tokens (typically, inscriptions or sounds) which are spatio-temporally

36 Anke Breunig located. To be able to determine which logical relations hold between chains of sign tokens, we must first choose suitable rules. The task of setting up rules and definitions in the metalanguage belongs to pure syntax. This time, however, our choice of rules and definitions is to fit the actual structures we have encountered empirically. There does not have to be a perfect fit between the syntactic rules and the corresponding empirical structure, because there is always a possibility of error. Thus even if we let ourselves be guided by actual empirical encounters, we have some latitude in how we formulate the rules in a system of pure syntax. To use a system of rules belonging to pure syntax to describe a particular empirical structure, we also need what have been called correlative definitions. The rules tell us which kinds of sounds, marks, and the like belong to which syntactic category.38 With the help of these rules of correlation we can then engage in descriptive syntax, i.e., we can investigate which logical relations hold between tokens of signs that we encounter somewhere.39 For Carnap, descriptive syntax is an empirical science.40 However, descriptive syntax is dependent on pure syntax, and the latter is clearly not an empirical science. The dependence holds because we cannot engage in descriptive syntax without having set up a system of rules in pure syntax. There is a passage where Carnap himself seems to acknowledge this dependency: Descriptive syntax is related to pure syntax as physical geometry to pure mathematical geometry; it is concerned with the syntactical properties and relations of empirically given expressions (for example, with the sentences of a particular book). (Carnap [1934] 1937, 7)41 There is then a difference between different parts of linguistics. The whole of linguistics is an empirical science. But within linguistics some subdisciplines are dependent on another discipline which is pure and which therefore does not belong to empirical science. Other subdisciplines lack a pure counterpart. Given this difference, it seems necessary to distinguish within linguistics between those subdisciplines that are merely empirical and those that are empirical, but not merely empirical. Pragmatics would be an example of the former, whereas descriptive syntax would belong to the latter.42 (Before discovering pure semantics, Carnap thought semantics also had no pure counterpart.) Because descriptive syntax is dependent on pure syntax, we can also say that the logical relations we ascribe to empirical structures have logically modal properties. The logical properties of languages in use inherit their modal status from the logical properties of languages as calculi that are devised in pure syntax.43 The object language of descriptive syntax is, so to say, a projection of the object language which is the subject matter

Sellars’s Twist on Carnap’s Syntax

37

of pure syntax. It is this dependence that allows Carnap to avoid psychologism when dealing with languages in use.44 There is, however, a fundamental objection to the view I have here developed and ascribed to Carnap. For Carnap also seems committed to saying that we could give a syntactic description of any empirical structure we encounter, as long as it is remotely structurally similar to the calculi we know or can make up. The empirical structure in question does not have to be what we would usually call a language. Carnap is extremely tolerant when it comes to applying syntactic rules to empirically given serial orders.45 But this attitude of tolerance does not seem justified. It is just false to attribute logical properties and a special modal status to any empirical structure upon which we happen to stumble. Sellars thinks that an empirical structure only exhibits logical properties if it itself is or belongs to an encompassing symbolic system that can be called autonomous. A symbolic system is autonomous if it is such that we can explain a being’s logical, or, more broadly, intellectual capacities in terms of its capacity to use the system, thereby producing empirical structures of the kind in question. As Brandom puts the idea, what we need to consider is an “autonomous fragment of language—one that might be understood though no others are”.46 Similarly, a language game is autonomous if it is a game we could play even if we played no others. It is a game such that one who has mastered the game’s complexity can for that reason be called intelligent.47 An autonomous symbol system is the system underlying an autonomous language game. According to what I will call the autonomy requirement, it is crucial for an autonomous symbolic system to have a certain degree of complexity. Among many other things, this means that the system must contain different syntactic functions. It is beyond the scope of this chapter to discuss which list of criteria would be sufficient for an autonomous symbol system. But it seems reasonable to hold that syntactic differentiation is among the necessary criteria. And we might also say that truth-functional completeness is a necessary requirement. Carnap allows us to construct calculi that lack truth-functional completeness. A calculus might be extremely simple. This follows from his far-reaching principle of tolerance.48 We might admit that the expressions of such a calculus have logically modal properties. But it would be strange to say that an empirical structure that is a manifestation of such a simple calculus, but not a part of a more encompassing empirical structure, also has the relevant modal properties. Even if we refuse to attribute logically modal properties to an empirical structure on the basis of its relation to a non-autonomous symbolic system, we still have another option. We can always try to correlate a given empirical structure with a calculus that fulfills the autonomy requirement. It might, after all, be the case that some very simple empirical structure is the only manifestation of a very complex abstract symbolic system.

38 Anke Breunig Indeed, every empirical structure is such that it could be explained by seeing it as the result of an autonomous language game.49 However, where an easier explanation as the result of a mechanism that is not an autonomous language game is also available, we should resist this move. I propose that we only correlate an empirical structure with a calculus that is an autonomous symbolic system if this is the best way to explain how the structure came about. Only then are we justified in attributing logically modal properties to a given empirical structure. The objection against Carnap, which I will call the “objection from autonomy”, can now be stated fully. Carnap blurs the boundary between empirical structures that don’t possess any logically modal properties and empirical structures that do. Only the latter are properly called languages in use. We can restore this crucial distinction by specifying what counts as an autonomous symbol system, and what it means for something to be a manifestation of such a system. However, Carnap shows no inclination to make any such distinction.

3 Thought and Language Carnap was one of the main protagonists of the linguistic turn. Exactly what the linguistic turn requires, however, is not always clear. Among the various advocates of what Sellars calls “the new way of words”, we find quite different views on the role of language in philosophy. As I have mentioned previously, Carnap wants to limit philosophy to logical syntax of language. This rather radical view is his way to propagate linguistic philosophy. We have also seen that Carnap defends a new conception of logic, having made logic a part of the syntax of language. Traditionally, it was held that logic investigates the laws of thought, not of language. Carnap expresses his rejection of this tradition as follows: But the development of logic during the past ten years has shown clearly that it can only be studied with any degree of accuracy when it is based, not on judgments (thoughts, or the contents of thoughts) but rather on linguistic expressions, of which sentences are the most important, because only for them is it possible to lay down sharply defined rules. (Carnap [1934] 1937, 1) Although Carnap hardly discusses what relation there is between thought and language, he would most likely agree with the view that we can learn something about thought by investigating language. We usually ascribe logical properties not only to sentences and their utterances but also to thoughts, judgments, and beliefs. If this practice is justified, we should expect of philosophy, even in Carnap’s restricted sense, some elucidation

Sellars’s Twist on Carnap’s Syntax

39

of the nature of such properties of thought. Thus, it cannot be quite right that we must turn to the empirical sciences, most notably psychology, if we are interested in a theory of thought, as Carnap tends to claim. The philosopher must partake in this task, side by side with the psychologist. To understand what we can learn from philosophy of language concerning the logical properties of thought, we must turn once more to Carnap’s treatment of the relation between pure and descriptive syntax. Let us assume that Carnap might have introduced a descriptive syntax of thought in addition to the descriptive syntax of language. On the one hand, such a descriptive syntax of thought would be dependent on empirical descriptions of states of the brain given by psychology or the neurosciences.50 On the other hand, it would also be dependent on the pure syntax of language. The relation between descriptive syntax of thought and pure syntax of language would be analogous to the relation between descriptive and pure syntax of language. As before, we would need correlation rules specifying which behavioral dispositions or brain states correspond to which categories of pure syntax.51 As we have seen, it is Carnap’s view that only by first constructing and then investigating artificial symbolic languages in pure syntax can we achieve an exact understanding of logical concepts. This view implies that pure syntax is fundamental from a methodological point of view both for descriptive syntax of language and an envisaged descriptive syntax of thought. The latter thesis is closely related to a methodological version of the thesis of the primacy of language over thought: (PLM) An understanding of the syntactic, semantic and epistemic properties of thought depends upon an understanding of the syntactic, semantic and epistemic properties of language.52 In the syntactic period, Carnap is committed to PLM only with regard to syntax, not with regard to semantics or epistemology. As I have already stressed, no theory of empirical meaning can be given within the syntax of language as Carnap conceives of it. We encounter the same limitation within the envisaged descriptive syntax of thought. If we look for a theory of empirical meaning we have no other choice but to turn to the empirical sciences. But if scientists in purely empirical investigations are supposed to adhere to a methodological principle grounded in PLM, Carnap doesn’t say so (perhaps wisely). We might expect philosophers in the wake of the linguistic turn defending a principle like PLM to resist the Myth of the Given. This is because the plausibility of PLM rests on the assumption that all the relevant properties of thought and of language are analogous. But someone who claims that some of our thoughts or mental representations possess intentionality intrinsically, independent of any relations that may happen to hold between such a privileged state and any other mental representations,

40 Anke Breunig cannot admit that thought and language are analogous throughout, i.e., through all their syntactic, semantic, and epistemological properties. It would be absurd to say that some of our linguistic expressions are also intrinsically intentional. The absurdity of such a claim results from the conventionality of linguistic expressions. Because linguistic expressions are mere sign figures or types of sound event only conventionally imbued with meaning it is clear that they don’t exhibit intrinsic intentionality. Now to say that some states are intrinsically intentional implies that the content of these states is given.53 Therefore, PLM is not compatible with this version of the Myth of the Given.54 Sellars himself hints at this incompatibility when he writes: Once a radical innovation, the notion that thought is a “symbolic process” has become a commonplace, almost a truism. Unfortunately, as is the case with many contentions that have become truisms, its implications are no longer as passionately scrutinized as they were when it was new, and it is often combined with modes of theorizing with which it is radically incompatible. (ITSA 310) To conceive of thought as a symbolic process is to conceive of it in analogy to and on the model of our use of language. This in turn means being committed to PL in either its methodological or its substantial form. Now Sellars does not say explicitly what he takes to be incompatible with this view, but it is a good guess that among the “modes of theorizing” he has in mind are those relying on the Myth of the Given. The hope in the background of the present considerations was that it would be possible to show that Carnap, being one of the leading figures of what has been called the linguistic turn, is not entitled to resort to the Myth of the Given. But, so far, this hope has been frustrated. I have attributed PLM to Carnap, but only with respect to syntax. Carnap’s syntax indeed seems to be free of the Myth of the Given, but this is just what was to be expected, given the fact that syntax does not provide us with an account of empirical meaning. Since it falls to the empirical sciences to develop a theory of empirical meaning, and since these are free to reject PLM, I haven’t yet shown why one must avoid givenness in one’s account of empirical meaning. I will come back to these issues in Section 5 by reconsidering the claim that PLM can be disregarded in dealing with empirical meaning.

4 Physical Rules I have expounded Carnap’s view that we can learn something about meaning from syntax and therefore from philosophy only insofar as we are focusing on logical meaning. But I haven’t mentioned yet that Carnap also seems to make room for a syntactic treatment of empirical meaning.

Sellars’s Twist on Carnap’s Syntax

41

Such a treatment seems possible because Carnap allows us not only to formulate logical rules of inference (L-rules), but holds that we can also introduce physical rules of inference (P-rules) for languages containing descriptive vocabulary. A valid inference is logically valid either if the premises and the conclusion do not contain any descriptive terms, or if one can replace all the descriptive terms by any other terms of the same logical category without making the inference invalid. A valid inference is physically valid if this condition is not satisfied.55 To express a law of nature would be a typical case where we might want to introduce a P-rule. Suppose there is a law of nature saying “(Necessarily) All F are G”. If we introduce a P-rule corresponding to this law, the claim “a is G” would be a direct consequence of the claim “a is F”, and the claim “All F are G” a conceptual truth. The introduction of P-rules is expedient for the purpose of developing languages suitable for science. According to Carnap, philosophy is first and foremost logic of science. Making use of P-rules, the logician of science can see to it that the axiomatic method, as employed, e.g., in physics, is also reflected in the form of the physical language. However, Carnap holds that the form of the language we use should always be adapted to our current purposes. This is all the more so for the various languages of science. An attitude of tolerance regarding the form of a language is central to Carnap’s whole approach. We are free to construct a language in almost any way we want, although prudence recommends that the choice of rules be guided by our goals. Tolerance also prevails when it comes to the question whether to set up P-rules for a language with descriptive vocabulary at all. When we devise P-rules for a particular language, the meaning of those descriptive terms occurring in one of the P-rules is partly determined by the P-rule. If our language contains a P-rule corresponding to the expression “Water is H2O”, then it is part of the meaning of “water” that whatever we call water consists of hydrogen and oxygen. Thus, clearly P-rules are relevant to empirical meaning. Availing ourselves of P-rules we can at least partly integrate an account of empirical meaning into the syntactic approach. However, P-rules can only partly determine an expression’s empirical meaning. A complete account of empirical meaning requires that at least some descriptive terms are used in records of observation. How we use expressions in response to observation is not a question of pure syntax. For Carnap, it is not a philosophical question at all, because it must be answered by purely empirical means.

5 Carnap’s Dilemma My question was whether anyone giving an account of empirical meaning compatible with Carnap’s syntax (although not a part of it) is pushed

42 Anke Breunig towards the Myth of the Given. To answer this, we must investigate what features an account of empirical meaning belonging to psychology or empirical linguistics might have. It is Carnap’s view that we must somehow come to terms with the relations holding between signs and their referents in any account of empirical meaning. For this it seems advisable to focus on statements of observation. Whatever we say of the relevant relations must amount to a purely empirical description. To describe the relation between a statement of observation and the situation referred to we might try the following: a particular sentence is assigned to the situation momentarily observed by the speaker who utters the sentence. The capacity to make the relevant assignment depends on knowing that certain sounds regularly occur in a certain type of situation. A field linguist or a behavioristically minded psychologist will be able to observe such patterns.56 The suggestion here is that empirical meaning consists in correlations between sounds and observed states of affairs. Referring to Price, Sellars calls it the thermometer view,57 holding that it is clearly inadequate as a theory of meaning. For Sellars, knowledge of meaning involves more than the capacity to react reliably to a certain stimulus. Traditionally, philosophers have held that meaning is not a purely empirical relation at all. With this Sellars agrees. One reason to deny the reducibility of meaning to empirical properties is that knowing how an expression is actually used does not entail knowing how it is to be used correctly. We can see why, according to Sellars, meaning is normative. Reductionism opposes the traditional view of meaning as a non-empirical (or not purely empirical) property. Meaning is to have a place in nature. According to the reductionist there is only one way to ensure this, namely to reduce the property of meaning to purely empirical properties. The thermometer view offers just such a reduction. Being a naturalist, the thorough reductionist rejects the traditional concept of meaning and uses the word “meaning” in another sense. But the Myth of the Given presupposes just that traditional concept. Therefore the strict reductionist has no use for the Myth of the Given. Reductionism is, however, in danger of turning into a position of a wholly different kind. This happens when the traditional and the reductionist senses of meaning are conflated. If the alleged reductionist claims to have offered an explanation of meaning in its traditional sense, she has tacitly committed herself to the Myth of the Given. Two questions regarding Carnap arise out of these considerations. First, we must ask whether Carnap envisages a reductionist view of empirical meaning at all. If we answer this question in the affirmative, we can ask, secondly, if Carnap is aware of all the consequences of the reductionist view, because if he isn’t, we should expect him sooner or later to resort to the Myth of the Given. Although Carnap does not dwell on it much, the answer to the first question is affirmative. Carnap does not develop a theory of empirical

Sellars’s Twist on Carnap’s Syntax

43

meaning himself, because in his view, this is not the task for a philosopher. But the fact that he hands the task over to empirical sciences like psychology or linguistics is itself telling. Carnap is under no illusion as to the kinds of properties investigated by an empirical science. Being also a physicalist, Carnap thinks they must all be reducible to physical properties in the end.58 Among these, empirical meaning in its traditional sense has no place. It is most likely that for Carnap, the assumption that there is a non-reductive meaning property would have to be discarded along with the rest of metaphysics. An answer to the second question, however, is more difficult. It is somewhat surprising that Carnap should opt for a reductionist view when it comes to empirical meaning, while at the same time decidedly opposing reductionism (in the guise of psychologism) with regard to logical meaning. The irreducibility of logical relations has to do with their special modal properties. A non-reductionist like Carnap needs to insist on the strict separation of logical and empirical investigations. The disparity between his views on logical and empirical meaning seems odd, since there is evidently a connection between the two kinds of meaning. Every expression of a language necessarily has logical and, more broadly, syntactic properties, because these are what account for something’s being part of a language at all. Besides, an expression containing descriptive terms also has empirical meaning. Its meaning is therefore made up of two components, a logical and an empirical part. With two radically different kinds of meaning, it is hard to see how they can come together in a single expression’s meaning. There is another argument for the incompatibility of a reductionist view of empirical meaning with Carnap’s non-reductionist view of logical meaning. As we have seen in the preceding section, Carnap allows us to construct languages containing P-rules. P-rules partly determine an expression’s empirical meaning. This turns out to be problematic, because the concept of empirical meaning now also becomes ambiguous. Insofar as empirical meaning is determined by P-rules, it is a syntactic, nonempirical property as befits the traditional approach to meaning. Insofar as empirical meaning is not determined by merely optional P-rules, however, it is an empirical property according with the reductionist position. But such a Janus-faced view seems hopeless.59 Sellars criticizes proponents of concept empiricism, a position wedded to the Myth of the Given, for recognizing two very different species of meaning. These two species I have called empirical and logical meaning. Sellars writes: Unqualified concept empiricism is patently incapable of accounting for many of our most familiar concepts, among others those of logic and mathematics. To remedy this defect, the theory is usually modified by introducing a radical dualism into its account of concepts

44 Anke Breunig and concept formation. The theory now recognizes a second mode of concept formation, namely the learning to use symbols in accordance with rules of logical syntax. The concepts of logic and mathematics are held to be symbols which gain meaning in this second way, rather than by association with empirical phenomena. (ITSA 312) Originally, Sellars here attacks those in favor of the Myth of the Given. But we can see that this line of attack is even more forceful with respect to our current concerns. In the position I have ascribed to Carnap, the two kinds of meaning are not even part of a common genus. The concept of meaning has turned out to be merely ambiguous. Therefore, it is wholly mysterious how the two kinds of meaning, meaningempirical and meaninglogical, hang together. To deal with this problem one might try to invoke the Myth of the Given.60 Instead of saying that the signs or sounds of a language are directly associated with the things outside of us, we might say that this association is mediated by certain mental states. There are different ways to further describe the mediating mental states, not all of them invoking the Myth of the Given. But if we say that the states in question themselves possess meaning or intentionality and describe them as states of awareness, we have clearly committed ourselves to the Myth.61 And this seems to be just what we have to do if we want to stick to the traditional, nonreductionist approach to empirical meaning. To say that the meaning or content of a mental state is given is not to give an empirical description of the state.62 What is given is always grounded in a natural and thus empirically describable property of a mental state. But that natural property is now to account for the state’s having an intentional property, i.e., for possessing content, which in turn is not a merely natural property. However, there can be no purely empirical evidence for identifying a natural property with an intentional one, or, alternatively, for assuming a one-to-one correlation between these properties. Intentional properties are empirically intractable. It is just because of this that one can escape reductionism by invoking the Myth, thus hoping to answer the traditional question of empirical meaning after all. Someone taking this route is not entitled to defend the primacy of language over thought in the form stated earlier (PLM or PLS). As I have shown, PLM is not compatible with an account of intentionality that relies on the Myth of the Given. This consequence cannot be avoided here by pointing out that PLM holds only for philosophical theorizing about thought and language, not for empirical approaches to that topic. To limit PLM in this way is something I have granted Carnap. But the limitation cannot help avoid the ambiguity of the concept of meaning, because, as I have argued, someone relying on the Myth of the Given to do so is clearly not proposing a merely empirical account of meaning but

Sellars’s Twist on Carnap’s Syntax

45

a philosophical account thereof. Vice versa, being committed to the primacy of language rules out any chance of solving our problem by invoking the Myth of the Given. Carnap is thus faced with a dilemma. Either he stays true to the claim that we must not attempt to give an account of empirical meaning within philosophy. Then a reductionist account of empirical meaning is the only remaining option. But such an account is at odds with Carnap’s antireductionist view of logical meaning, which is at the heart of his project in Logical Syntax of Language. Or Carnap invokes the Myth of the Given to solve the problem of the unity of the concept of meaning. But then he not only breaks with the principle that as philosophers we must refrain from theorizing about empirical meaning. He also gives up PLM, a thesis he seems strongly committed to. Let me call this problem Carnap’s dilemma. It is hard to say which horn of the dilemma Carnap would choose, since he mostly just circumvents the whole issue of empirical meaning. I think the first, reductionist horn fits somewhat better with his overall approach. But maybe he is torn between both sides, without recognizing that he can’t have his cake and eat it too. The sparseness of his comments on these thorny questions might help to mask such incompatible tendencies. Should he think that an empirical psychologist might one day discover mental states exhibiting intrinsic intentionality with which to explain how it is that merely conventional signs have empirical meaning, he would be mistaking the Myth of the Given for a purely empirical theory. Sellars, at least, is suspicious that Carnap does not firmly reject the Myth after all, as can be seen from the following passage: Metaphysicus is not alone in suspecting that if Carnap had once been asked, How is the acceptance of the framework of sense-data to be justified? He would, in effect, have replied “because colors, sounds, etc. are given”, where further questioning would have made it clear that this givenness did not involve a covert use of a (however rudimentary) symbolic framework of sense-data. I think it is obvious that many empiricists have taken this line. Whether or not Carnap still does (or has ever done) it is not easy to say. Certainly he has, on many occasions, availed himself of the philosophical jargon of givenness, and nowhere he has explicitly discussed and rejected the epistemological views it embodies. (EAE 435)

6 Autonomous Languages I now have to come back once more to the objection from autonomy developed earlier in Section 2. There, the objection arose from considering Carnap’s view of logical meaning alone. By taking into account the

46 Anke Breunig relation between logical and empirical meaning, we can now sharpen the objection. I will call the new version the “extended objection from autonomy”. It arises out of asking whether any empirical structure corresponding to one of the syntactic systems constructible along Carnap’s lines meets the demands for an autonomous symbolic system. So far, I have assumed that at least some of the calculi constructed by Carnap, when actually adopted and used, are indeed “languages”, i.e., (parts of) autonomous symbol systems, where something is a language only if could be used by someone as her only language. But this assumption turns out to be unwarranted. When I speak of adopting and subsequently using a symbolic system I mean to imply that our actual behavior corresponds more or less closely to the syntactic rules of the system. But we should not assume that our use of the system is therefore also integrated with our perceptions and actions. Because of this lack of integration with perception and action the calculus still lacks empirical meaning.63 I haven’t specified the criteria something must meet to count as an autonomous symbol system. But it seems very plausible to say that having empirical meaning is among these criteria. At least, it is clear that Sellars thinks so.64 Carnap gives examples both of formal symbolic systems that don’t include descriptive terms and of some that do. The former obviously lack empirical meaning, but so do the latter. The descriptive terms’ meaning is left unspecified by Carnap’s syntactic theory. As we have seen, Carnap does not develop a unified account of empirical and logical meaning. If we think that a language proper must contain not just logical but also empirical meaning, Carnap’s syntactic theory cannot explain what makes something a language, even if we assume that one of his symbol systems was actually adopted and used. We look in vain for a system that contains all the rules we need in order to play an autonomous language game.

7 Endorsing Normativity I have developed three problems for Carnap’s view of meaning from the syntactic period. I have also claimed that these problems stem from tensions internal to Carnap’s system. These tensions come to the fore if one confronts Carnap’s view with certain questions that he seeks to avoid. Let me now come back to what I have called Sellars’s main objection to Carnap, an objection that, as we have seen, Carnap brusquely rejects. The objection was that Carnap neglects the normative character of linguistic rules. By focusing on our actual use of language, and by recognizing that such usage is implicitly norm-governed, we can overcome all the problems I have formulated here. This is why I think the normativity objection is indeed Sellars’s main objection.

Sellars’s Twist on Carnap’s Syntax

47

The first problem was that Carnap does not give us any criteria with which to decide which empirical structures genuinely exhibit logical relations and thus can appropriately be interpreted as syntactic systems. I have called this the objection from autonomy, because we should say that only those empirical structures that belong to an autonomous language system genuinely exhibit logical relations. Now turning to norms, there is an analogous problem; that is, how to decide when we should say that a practice and the empirical structures it produces are implicitly governed by norms and therefore can be explained by them, as opposed to saying that behavioral patterns merely conform to norms that are external to them. Sellars’s account of normativity solves this problem in a holistic way, by specifying the criteria for what counts as an autonomous language game. The idea is that only such behavior as participates in an autonomous language game is indeed governed by norms. The second problem I have called Carnap’s dilemma. It arises when one asks how empirical and logical meaning fit together. I have argued that Carnap has two options with regard to empirical meaning, both of which are problematic. First, he might give a reductionist account of empirical meaning in spite of being committed to a non-reductionist account of logical meaning, thus sacrificing the unity of meaning. Second, he might make use of the Myth of the Given to account for empirical meaning. Doing so, however, violates the thesis of the primacy of language, a thesis he holds dear. With a normative theory of meaning we can escape the dilemma, because it makes available a third option besides the reductionist and givennist accounts of empirical meaning. And we can restore the unity of meaning by accepting not just a normative theory of empirical meaning but also of logical meaning. In this way, we can uphold commitments to both anti-psychologism and the primacy of language. The third problem for Carnap, the so-called extended objection from autonomy, arose out of the first one, because we realized that the autonomy requirement could never be fulfilled by an empirical system that does not also have empirical meaning. Lacking a unified theory of meaning, we would not only be unable to ascribe empirical meaning to empirical structures, we would also be unable to ascribe to them logical meaning. But the answer to the second problem shows how a normative approach can overcome this obstacle, so that the first and third problems can now also be solved. I should point out that it is not the case that any normative account of meaning will succeed in solving all three problems. In particular, to avoid both reductionism and the Myth of the Given is not an easy task. I won’t be able to show how Sellars’s account achieves this goal here. But maybe I have said enough to make it plausible that anyone who rejects the normative stance is either committed to reductionism or to the Myth of the Given.65

48 Anke Breunig

Notes 1. Plausibly this widespread view can be traced back to Rorty. As Rorty sees it, Sellars and Quine launch a joint attack on empiricism (see Rorty 1979, 170). Since Quine directly attacks Carnap, one could think that Carnap is also one of Sellars’s main targets. But Rorty’s view of the relation between Sellars and Quine is not convincing, as has been argued by Brandt (2017). See also Breunig (2019a). 2. This question has, for example, been examined thoroughly in Coffa (1991) and Richardson (1998). 3. In EAE, Sellars focuses on Carnap ([1950] 1956), but his [1934] 1937 and 1942 are clearly in the background. 4. Besides being a short name for Carnap’s second major work, “syntax”, just like its cousin “semantics”, is also the name of the discipline Carnap seeks to establish. 5. Sellars’s view of Carnap’s semantics is more complicated and less sympathetic. I will not try to disentangle it here. The complaint that Carnap does not give an adequate treatment of rules applies to his semantic as well as his syntactical phase. 6. See Carus (2004). Olen (2016) argues that Sellars inherited a misreading of Carnap’s semantics in his earliest writings, but later came to recognize his mistake (see chapter 2, in particular). Olen also claims that there is a deep rift that separates Sellars’s earliest papers from the bulk of his work. If one accepts this latter claim, what I am going to say will be compatible with Olen’s reading, since I am not dealing with Sellars’s earliest writings. 7. An expanded version appeared in English translation in 1937. References are to the latter. 8. Cf. Carnap ([1934] 1937, 1, 7). 9. Cf. Carnap ([1934] 1937, 5f). 10. Cf. Carnap (1942, 9). 11. Cf. Carnap ([1934] 1937, 5), and (1942, 3). 12. “Thus the system of a language, when only the formal structure is considered, is a calculus” (Carnap [1934] 1937, 4). “When we say that the objects of logical syntax are languages, the word ‘language’ is to be understood as the system of the rules of speaking, as distinguished from the acts of speaking” (Carnap 1935, 5). Carnap also characterizes syntax in the following way: “The syntax of a language, or of any other calculus, is concerned, in general, with the structures of possible serial orders (of a definite kind) of any elements whatsoever” (Carnap [1934] 1937, 6). On the relation between logical syntax and calculi, Carnap writes: “In the widest sense, logical syntax is the same thing as the construction and manipulation of a calculus” (Carnap [1934] 1937, 5). 13. Cf. Carnap ([1934] 1937, 169) and (1935, 42). 14. Cf. Carnap ([1934] 1937, 1f). David Hilbert’s metamathematics was an important inspiration for Carnap and the development of logical syntax. See Carnap (1935, 40f). 15. Cf. Carnap ([1934] 1937, 169) and (1935, 44). 16. Carnap ([1934] 1937, 2). 17. Carnap (1942, 232) distinguishes different meanings of the term “formal”. In Carnap’s preferred usage, “formal” means “in abstraction from meaning”, “without reference to designata” or “syntactical”. A cause for confusion might be that “formal” sometimes also means “logical” or “necessary”. 18. Carnap ([1934] 1937, 280). In his 1942 Carnap turns to semantics. Influenced by Tarski, Carnap comes to see that pure semantics is possible, after all. Pure

Sellars’s Twist on Carnap’s Syntax

19. 20.

21. 22. 23. 24. 25.

26.

27.

28. 29. 30.

49

semantics is meant to complement pure syntax. Although pure syntax continues to be relevant, it comes to play a different role in the semantic period. After the semantic turn, Carnap ceases to identify philosophy with logical syntax. This chapter only deals with Carnap’s syntactic period. The objections I develop are not immediately relevant to the semantic period, although it might be possible to raise similar criticisms against Carnap’s later views. I focus on the syntactic period because it is the one that most influenced Sellars. Indeed, for Sellars, Carnap’s semantic turn was, in a way, a wrong turn. Cf. Carnap (1935, 56ff.). Logical expressions differ from descriptive expressions in that only the meaning of the former is determined by logical relations between expressions alone (cf. Carnap [1934] 1937, 177f.). The idea that the meaning of logical vocabulary is determined solely in virtue of the relations between expressions of a language stems from Wittgenstein. Clearly, form and content are not opposites here. Since the text was translated into English from German, Carnap continued to use the German abbreviations, e.g. “S” for language (“Sprache”) instead of “L”. See Carnap ([1934] 1937, xii). Cf. Carnap ([1934] 1937, 185). For the notion of physical rules see also Section 4. Wittgenstein’s Tractatus ([1921] 1961) seems to be the source of the view that valid sentences don’t have content, because they don’t picture. Just like Sellars’s, Carnap’s theory is one of functional roles. This can be seen from Carnap’s remark that two languages are isomorphic if they share a formal structure, and the claim that syntax is only concerned with this formal structure, not the shape or physical properties of the signs (see Carnap [1934] 1937, 6). For Sellars’s theory of meaning as functional role, see MFC 89ff. Carnap thinks that although we can have knowledge about the relations of consequence holding between (classes of) sentences in the absence of links to observation, knowledge of the meaning of any empirical sentence is impossible without such links. I do not want to rule out the possibility that such pre-linguistic awareness might itself be structured holistically. Such awareness would then have to be pre-linguistic but not pre-conceptual. If this possibility is indeed ruled out, as Sellars seems to suggest, this would have to be argued for separately. I tackle this question in much more detail in Breunig (2019b). I think it is a further question whether a belief that the given content is true is intrinsically justified. Peter Olen (2016) points out that in the attempt to develop a “pure pragmatics” in his earliest writings, Sellars adopts a formalist approach. According to Olen, Sellars’s formalism is motivated by metaphilosophical considerations concerning philosophy’s relation to the sciences (2016, 11). It is characterized by the claim that philosophy must not involve any factual considerations. Olen also argues that Sellars nowhere defines what he means by formal, and indeed makes use of several different notions, which are not consistent with one another (2016, 12). According to Olen, formalism is incompatible with recognizing that some philosophically relevant phenomena are normative, which is why Sellars abandoned formalism early on (2016, 8). If Olen is right about this, the position I attribute to Sellars cannot be of a piece with his earlier formalism. But see Brandhoff (2017) for a challenge to Olen’s reading. I think Sellars’s view of meaning is “broadly formal” in being “functional”, “broadly syntactical”, or “inferentialist”. Sellars’s theory of meaning essentially involves normative aspects. It also has room for factual considerations, but must re-describe them in normative terms.

50 Anke Breunig 31. For the sake of brevity, from now on I will not mention mathematical relations explicitly, but speak of logical relations only. 32. I will talk of “logical modality” and “logically modal properties” instead of “logical necessity”, which is more common, to make room for the view that logical possibility also has a role to play. The view is, of course, Sellars’s, who says “a full discussion would refer to may-be’s (or permitteds) as well as ought-to-be’s” (LTC 62, n4). 33. For an explicit rejection of psychologism, see Carnap (1950, 37ff). Frege famously criticizes psychologism in his [1884] 1974 and the preface to his [1893/1903] 2013. 34. Because syntax only investigates the structural properties of a language, the rules formulated in the metalanguage can be realized by different but isomorphic object languages (see Carnap [1934] 1937, 5f.). 35. For now I will grant that Carnap’s explanation of logical modality in relation to languages as calculi is successful. Sellars does not make that concession. He insists that only languages in use have logical properties (see EAE 455ff.). Regarding semantic properties, Sellars makes a parallel point: “Now Carnap is, of course, aware that a pure semantical theory is a semantical theory only if it relates its vocabulary to semantical expressions in actual usage” (EAE 462). Carnap, however, rejects Sellars’s interpretation: “As Sellars aptly expresses it, pure semantics is nothing but combinatorics of sign designs and extra-linguistic entities” (Carnap 1963, 926). 36. Carnap ([1934] 1937, 2f.) also distinguishes between artificial word languages, such as, e.g., Esperanto, and constructed symbolic languages. This distinction is not relevant for our discussion. 37. It might be thought that a calculus can also be used. For lack of a better term, however, I stipulate that a language as calculus is not used. There are a number of differences between languages in use and languages as calculi. A language in use only exists if it is actually used, whereas a language as calculus exists already when its syntactical rules have been formulated. Also, a language in use can contain malformed expressions, as a result of rules being applied incorrectly, whereas the expressions of a language as calculus are necessarily formed correctly. 38. Cf. Carnap ([1934] 1937, 7). 39. Carnap explains: “Sentences of descriptive syntax may, for instance, state that the forth and the seventh sentences of a particular treatise contradict one another; or that the second sentence in a treatise is not syntactically correct” (Carnap [1934] 1937, 7). 40. “Descriptive syntax and semantics deal with certain features of languages investigated empirically. Even here, the statements about these features are descriptive; what Sellars calls ‘rule-bound words’ do not occur” (Carnap 1963, 923). 41. In EAE 462, Sellars cites a very similar passage from Carnap (1942, 12). 42. In his earliest writings, Sellars defends the possibility of pure pragmatics. For discussion, see Olen (2016). 43. In the introduction I cited a passage where Carnap denies that descriptive syntax contains any prescriptive components. This claim is compatible with acknowledging that descriptive syntax describes modal properties of languages in use. Only if one thinks with Sellars that “the language of modality is .  .  . a ‘transposed’ language of norms” (IM 21) do these claims become incompatible. 44. Ricketts (2003) presents a similar reading of Carnap: “Via the correlation of formal expressions of the calculus with utterance types, the consequence

Sellars’s Twist on Carnap’s Syntax

45. 46. 47. 48.

49.

50.

51. 52.

51

relation of the calculus and attendant syntactic distinctions are projected onto the statements of individuals whose speech habits, with respect to this correlation, agree with the calculus in the loose fashion just described” (262). Ricketts takes Carnap to adopt a behavioristic approach to languages in use. He emphasizes that only because of the projection of logical categories onto a language in use can we attribute logical properties to particular utterances of speakers (263). Ricketts is unconvincingly criticized by Carus (2007, 274ff.) and Wagner (2012). See, for instance, Carnap’s remark that a card-index system should count as a language (cf. Carnap [1934] 1937, 6). Brandom (1994, 90). By this condition I want to rule out that someone might be said to play a game if they can only make one move in the game. Analogously, we would not say that someone plays chess if they can only move the pawns. Carnap first formulates the principle of tolerance in [1934] 1937, 51, writing “It is not our business to set up prohibitions, but to arrive at conventions”. He goes on to explain “Everyone is at liberty to build up his own logic, i.e. his own form of language, as he wishes” (ibid., 52). The principle is stated again in [1950] 1956, 221. It plays a vital role in his thinking. For example, we might be tempted to hold that a parrot politely “saying” “Good morning” makes a move in a language game. But we should resist this urge, since the parrot has mastered only such a small part of the game. (This example is somewhat special, though, because the parrot was trained by a competent speaker. Therefore his behavior is not just incidentally related to an autonomous language.) Instead, a psychologist might adopt a behavioristic approach, identifying thoughts with dispositions for linguistic and non-linguistic behavior. Carnap thinks that the behavioristic analysis of thoughts in terms of behavioral dispositions will eventually be replaced by a physiological description of the relevant states of our central nervous system. See Carnap ([1932/1933] 1959, 172, 175–6). Whether suitable correlation rules can be found is, of course, at least partly an empirical matter. Here is a substantial version of the thesis: (PLS ) Syntactic, semantic and epistemic properties of thought are dependent upon syntactic, semantic and epistemic properties of language.

53. As I have said in Section 1, to assume that some states possess intentionality intrinsically already means to invoke a version of the Myth of the Given, even if one refrains from saying that a belief with a corresponding content would be directly justified. 54. It might, however, be compatible with other versions of the Myth. 55. For a similar, but more complicated definition, see Carnap ([1934] 1937, 181). 56. In his reply to Zilsel and Duncker (1932/1933), Carnap gets surprisingly specific about how children acquire empirical vocabulary. He writes: “Initial language learning evidently does not proceed according to explicit instructions, but by exerting influence through practical means. Just as you teach a child to spit cherry pits by example and gestures, you teach him (as a ‘conditioned response’, as the behaviorists say) the habit of saying, in certain circumstances, ‘This house is big’, or ‘I have a toothache’. If the child linguistically reacts as we would (i.e., by saying ‘The house is red’ if the house is red), we say that he is ‘speaking correctly’” (ibid., 182, my translation). I want to thank Kenneth Westphal for recommending this paper.

52 Anke Breunig 57. See EPM §31: 162, §35: 167. 58. See Carnap ([1932/1933] 1959, 165). 59. Reflecting on the status of P-rules in IM, Sellars comes to a similar conclusion. However, my objection and Sellars’s own are not identical. 60. As Sellars points out in the quoted passage from ITSA, one does not get a unified theory of meaning by relying on the Myth of the Given. But invoking the Myth promises some improvement, since it is then possible to regard empirical meaning and logical meaning as two species of a common genus. 61. Sellars makes this point in EPM §29: 160. 62. This is a paraphrase of Sellars’s famous remark that “in characterizing an episode or a state as that of knowing, we are not giving an empirical description of that episode or state; we are placing it in the logical space of reasons, of justifying and being able to justify what one says” (EPM §36: 169). I take it that the reductionist would reject the first part of the quote, whereas someone invoking the Myth of the Given would have to agree with Sellars’s negative statement. His mistake is that he does not account for the content of a mental state in terms of the role it plays in the logical space of reasons. 63. This conception of language as a pure transformative mechanism is the opposite of the “thermometer view”. Both accounts separate intra-linguistic “moves” or transformations on the one hand, from “language entry transitions” and “language departure transitions” on the other, and focus only on one part of an autonomous language. For this terminology, see SRLG §§22–3: 329. 64. McDowell (1994, 3f). McDowell also attributes this view to Kant, referring to Kant ([1781/1787] 1998, A51/B75). 65. I would like to thank Stefan Brandt, Marc Hiatt, Paulus Esterhazy, and participants both of the Erlangen conference “Sellars’s Place in 20th Century Philosophy” and Michael Beaney’s colloquium in Berlin for helpful comments and discussion.

References Brandhoff, Boris. 2017. “Pure Pragmatics and the Idea of a Metatheoretical Functionalism”. In Sellars and Contemporary Philosophy, edited by David Pereplyotchik and Deborah R. Barnbaum, 55–66. New York and London: Routledge. Brandom, Robert. 1994. Making it Explicit. Cambridge, MA: Harvard University Press. Brandt, Stefan. 2017. “Sellars and Quine on Empiricism and Conceptual Truth”. British Journal for the History of Philosophy 25 (1): 108–32. Breunig, Anke. 2019a. “Historische Stationen: 20. Jahrhundert”. In Handbuch Erkenntnistheorie, edited by Martin Grajner and Guido Melchior, 50–7. Stuttgart: Metzler. ———. 2019b. Von Grund zu Grund: Zum Zusammenhang von Denken und Wissen bei Wilfrid Sellars. Paderborn: Mentis. Carnap, Rudolf. (1928) 1967. The Logical Structure of the World: Pseudoproblems in Philosophy. Berkeley: University of California Press. ———. 1932/1933. “Erwiderung auf die vorstehenden Aufsätze von E. Zilsel und K. Duncker”. Erkenntnis 3: 177–88. ———. (1932/1933) 1959. “Psychology in Physical Language”. In Logical Positivism, edited by A.J. Ayer, 165–98. Glencoe, IL: The Free Press. ———. (1934) 1937. The Logical Syntax of Language. London: Kegan Paul, Trench, Trubner & Co.

Sellars’s Twist on Carnap’s Syntax

53

———. 1935. Philosophy and Logical Syntax. London: Kegan Paul. ———. 1942. Introduction to Semantics. Cambridge, MA: Harvard University Press. ———. 1950. Logical Foundations of Probability. Chicago: University of Chicago Press. ———. (1950) 1956. “Empiricism, Semantics, and Ontology”. In Meaning and Necessity: A Study in Semantics and Modal Logic, 205–21. Chicago: University of Chicago Press. ———. 1963. “Wilfrid Sellars on Abstract Entities and Semantics”. In The Philosophy of Rudolf Carnap, edited by Paul Schilpp, 923–27. La Salle, IL: Open Court. Carus, André. 2004. “Sellars, Carnap, and the Logical Space of Reason”. In Carnap Brought Home, edited by Steve Awodey and Carsten Klein, 317–55. Chicago: Open Court. ———. 2007. Carnap and Twentieth-Century Thought: Explication as Enlightenment. Cambridge: Cambridge University Press. Coffa, Alberto. 1991. The Semantic Tradition from Kant to Carnap: To the Vienna Station. Cambridge: Cambridge University Press. Frege, Gottlob. (1884) 1974. The Foundations of Arithmetic: A Logico-Mathematical Enquiry Into the Concept of Number. Oxford: Blackwell. ———. (1893/1903) 2013. Basic Laws of Arithmetic: Derived Using ConceptScript. Oxford: Oxford University Press. Kant, Immanuel. (1781/1787) 1998. Critique of Pure Reason. Edited by Paul Guyer and Allen Wood. Cambridge: Cambridge University Press. McDowell, John. 1994. Mind and World. Cambridge, MA: Harvard University Press. Olen, Peter. 2016. Wilfrid Sellars and the Foundations of Normativity. London: Palgrave Macmillan. Richardson, Alan. 1998. Carnap’s Construction of the World: The Aufbau and the Emergence of Logical Empiricism. Cambridge: Cambridge University Press. Ricketts, Thomas. 2003. “Languages and Calculi”. In Logical Empiricism in North America, edited by Gary Hardcastle and Alan Richardson, 257–80. Minneapolis: University of Minnesota Press. Rorty, Richard. 1979. Philosophy and the Mirror of Nature. Princeton: Princeton University Press. Sellars, Wilfrid. (1953) 2007 [IM]. “Inference and Meaning”. In Wilfrid Sellars. 1963. In the Space of Reasons: Selected Essays of Wilfrid Sellars, edited by Kevin Scharp and Robert Brandom, 3–27. Cambridge, MA: Harvard University Press. Hereafter SR. ———. (1953) 1991 [ITSA]. “Is There a Synthetic A Priori”. In Sellars, Wilfrid. 1963. Science, Perception, and Reality, 298–320. London: Routledge and Kegan Paul. Reissued by Ridgeview Publishing Co. in 1991. References are to the 1991 edition. Hereafter SPR. ———. (1954) 1991 [SRLG]. “Some Reflections on Language Games”. In SPR, 321–58. ———. (1956) 1991 [EPM]. “Empiricism and the Philosophy of Mind”. In SPR, 127–96. ———. 1963 [EAE]. “Empiricism and Abstract Entities”. In The Philosophy of Rudolf Carnap, edited by Paul Schilpp, 431–68. La Salle, IL: Open Court.

54 Anke Breunig ———. (1974) 2007 [MFC]. “Meaning as Functional Classification”. In SR, 81–100. Wagner, Pierre. 2012. “Natural Languages, Formal Languages, and Explication”. In Carnap’s Ideal of Explication and Naturalism, edited by Pierre Wagner, 175– 89. Houndmills, Basingstoke: Palgrave Macmillan. Wittgenstein, Ludwig. (1921) 1961. Tractatus Logico-Philosophicus. Translated by D. F. Pears and B.F. McGuinness. New York: Humanities Press.

3

Ajdukiewicz and Sellars on World-Perspectives Peter Olen

1 Introduction When exploring Wilfrid Sellars’s philosophical development, attention should be given to his 1949 anthology, co-edited with Herbert Feigl, Readings in Philosophical Analysis, which was intended to gather articles1 representing a “decisive turn” in philosophy. The introduction of the volume clearly indicates how Feigl and Sellars understood the then-recent developments in philosophy: The conception of philosophical analysis underlying our selections springs from two major traditions in recent thought, the Cambridge movement deriving from Moore and Russell, and the Logical Positivism of the Vienna Circle (Wittgenstein, Schlick, Carnap) together with the Scientific Empiricism of the Berlin group (led by Reichenbach). These, together with related developments in America stemming from Realism and Pragmatism, and the relatively independent contributions of the Polish logicians have increasingly merged to create an approach to philosophical problems which we frankly consider a decisive turn in the history of philosophy. (Feigl and Sellars 1949, VI) Most of the philosophical groups and individuals listed are familiar; there is no question about the influence and historical prominence of Wittgenstein, the Vienna Circle, Reichenbach, and others. What differs, though, is the appearance of the “relatively independent contributions of the Polish logicians”. Here, Feigl and Sellars are referencing the Lvov-Warsaw School of Logic, which consisted of (most famously) Alfred Tarski, Kazimierz Twardowski, Kazimierz Ajdukiewicz, Stanisław Leśniewski, Jan Łukasiewicz, and a number of others. Although not exactly obscure, these references should give pause when situated amongst clearly influential articles by Bertrand Russell, Gottlob Frege, Rudolf Carnap, and others. In particular, the appearance of Kazimierz Ajdukiewicz’s “The Scientific World-Perspective” is doubly surprising. As the only other Lvov-Warsaw school member included in the volume (apart from Tarski

56

Peter Olen

himself), Ajdukiewicz’s work would rightly be seen (then and now) as fairly obscure. Why would Feigl and Sellars include Ajdukiewicz’s chapter alongside more well-known and influential contributors? In what follows I trace the historical relationship between Sellars and Ajdukiewicz as it concerns the latter’s notion of a “world-perspective”. Specifically, I argue that Sellars appropriated Ajdukiewicz’s notion of a world-perspective in his early works, seeing it as a crucial aspect in offering a purely philosophical (as opposed to empirical) account of language. Although this line of influence is largely inspirational (Sellars and Ajdukiewicz have different conceptual axes to grind in discussing language), one can see shared themes and commitments across their projects. Ajdukiewicz’s essays are therefore especially important for understanding some of Sellars’s early concepts. Initially published in 1932, the original English translation of “The Scientific World-Perspective” appears in Feigl and Sellars’s collection. More surprisingly, the article was translated from German by Sellars himself and remains the only English translation currently available. In addition to making the historical case for Sellars’s early appropriation of Ajdukiewicz’s work, I argue that exploring Sellars’s work through this particular lens allows us to examine conceptual changes in his overall philosophy. Specifically, looking at the relationship between Ajdukiewicz and Sellars allows us to see how Sellars’s appropriation of Ajdukiewicz’s “world-perspectives” underpinned shared similar commitments (and, thus, inherited potential problems surrounding a kind of linguistic idealism), how these commitments fit into Sellars’s later introduction of the concept of picturing, and how these philosophical commitments change when he rejects his earlier metaphilosophical commitment to formalism.

2 Ajdukiewicz and World-Perspectives Since many of the central concepts found in “The Scientific WorldPerspective” presuppose background knowledge of Ajdukiewicz’s related works (which presumably Sellars had),2 I shall explain the concept of a world-perspective in terms of Ajdukiewicz’s “radical conventionalism” and linguistic pluralism, which he introduces as part of his account of an intersubjective conception of meaning, as opposed to a psychological, subjective notion of meaning that would vary between individuals (Ajdukiewicz 1934a, 35). On this conception, meaning is defined by the coordination of expressions with specific meanings and ideas, as well as rules that govern their use, in addition to designating expressions. Ajdukiewicz, like Sellars, avoids a naïve realism about meaning where one would see meaningful expressions as only those that hold an explicit connection to their designata. Since many expressions simply don’t denote a specific object, Ajdukiewicz argues that designation fails to capture the

Ajdukiewicz and Sellars on World-Perspectives

57

full scope of linguistic meaning (Ajdukiewicz 1934a, 39). The challenge for Ajdukiewicz, then, is to construct an intersubjective sense of meaning that captures the different roles played by expressions and sentences without retreating into subjectivism. Ajdukiewicz’s conception of conventionalism, despite its epistemological status, also plays an important role in his conception of language. His “radical conventionalism” is the view that “of all the judgments which we accept and which accordingly constitute our entire world-picture, none is unambiguously determined by experiential data; every one of them depends on the conceptual apparatus we choose to use in representing experiential data” (Ajdukiewicz 1934b, 67). Initially, one might think this form of conventionalism is fairly standard. Under a traditional conventionalism (what Ajdukiewicz calls “ordinary conventionalism”) one would claim that “some problems cannot be solved by appeals to experience without the introduction of a convention” (1934b, 67; emphasis added). What makes Ajdukiewicz’s position “radical” is his denial that any problem can be solved (or any knowledge can be obtained) from experiential data alone. The difference between positions, in part, simply comes from taking a more traditional conventionalism to its logical extremes. Instead of understanding some aspect of our conceptual choices as voluntaristic—conceptual choices through which we interpret our experiences—Ajdukiewicz argues such choice is present in all aspects of interpreting or reporting our experiences, making judgments, or obtaining knowledge.3 This kind of radical conventionalism underwrites Ajdukiewicz’s linguistic pluralism.4 Here, a brief contrast with Rudolf Carnap’s principle of tolerance is instructive. For Carnap, the idea of tolerance towards language choice is the idea that “In logic, there are no morals. Everyone is at liberty to build up his own logic, i.e. his own form of language, as he wishes” (Carnap 1937, 52; emphasis original). Yet Carnap’s “tolerance” is primarily concerned with the formally constructed languages. Unlike Carnap, Ajdukiewicz’s remarks apply equally to scientific, formal, and naturally occurring languages; his linguistic pluralism therefore stretches beyond Carnap’s principle of tolerance: Suppose two individuals speak English without violating either English phonetics or English syntax. Suppose their meaning-specifications agree respecting all words, with one exception: by the word “star”, one of them understands fixed stars only while the other understands both fixed stars and planets. Would it be said of these individuals that one of them speaks English and the other does not? Hardly. If two people use the same sounds with different meaning, we would say they did not speak the same “language” only when their respective meaning-specifications become decidedly different. (Ajdukiewicz 1934a, 63)

58

Peter Olen

This line of reasoning leads to a fairly broad pluralism: “we cannot say there is one English language. We would assert, rather, that there are many English languages. . . . Indeed, we can enumerate many English languages in common use which (apart from different dialects and historical phases) employ the same words and sounds” (Ajdukiewicz 1934a, 64). In this case, being a pluralist about language is being committed to the idea that difference in meaning between terms is not simply indicative of different conceptions of a given term, but of different languages. Ajdukiewicz’s pluralism is extreme in that “A single ambiguous word points to the existence of two languages whose sounds and words are the same and whose coordinations between word and meaning differ at one point only” (Ajdukiewicz 1934a, 64). So, there can be an almost infinite number of differing English languages (in this example) but, in many instances, those differences will only be a matter of degree (only one or two differing meaning-specifications for a number of terms). Most English speakers technically speak different languages (in Ajdukiewicz’s sense), but most of these differing languages concern variations at only a small number of points such that one may not notice a practical difference. This sounds like a fairly radical, if not confusing, position until one realizes that Ajdukiewicz draws a distinction between our everyday conception of a language and a strict conception of language. If we are using a “strict” conception of language, one that is unambiguously determined by meaning-specifications and linguistic rules, then we must conclude that languages can be demarcated due to even one difference in meaning-specification for an expression (Ajdukiewicz 1934a, 64). Why should we hold such a strict conception of language and meaning? From Ajdukiewicz’s standpoint, philosophically reconstructed languages are unambiguously determined in ways that natural languages obviously are not. Thus, such explicit disagreement in meanings point to, technically speaking, different languages. Yet, as Ajdukiewicz points out, it would be absurd to claim that—practically speaking—everyone is speaking a different language simply because of one point of disagreement. Instead of narrowly focusing on languages qua formally characterized deductive systems, Ajdukiewicz directly confronts the problematic practice of treating natural languages as if they were formal languages.5 The recognition that the concepts we apply to natural language only approximate linguistic practices allows Ajdukiewicz to focus on natural language while employing idealized concepts. Consequently, the linguistic pluralism envisioned by Ajdukiewicz is, on the whole, substantially less radical than one might assume. Again, if the difference between languages can be determined on pragmatic grounds (if “decidedly different” means something like “the difference in meaning-specifications, while real differences in meaning, make no difference until one undergoes a breakdown in communication”), then Ajdukiewicz can make the philosophical point that, technically speaking, there is a vast plurality of English languages without

Ajdukiewicz and Sellars on World-Perspectives

59

claiming that, practically speaking, everyone is using a wholly different language simply because of one disagreement in meaning-specification. Additionally, linguistic rules grounding the use of sentences are, for Ajdukiewicz, split between axiomatic, deductive, and empirical meaningrules. Axiomatic meaning-rules are rules that demand “an unconditional readiness to accept certain sentences”, while deductive meaning-rules “demand a readiness to accept certain sentences, not unconditionally, but only on the supposition that certain other sentences are accepted” (Ajdukiewicz 1935, 112), whereas empirical meaning-rules “demand the readiness to accept certain sentences in the presence of certain data of experience” (Ajdukiewicz 1935, 112). In addition to the sentences one must accept when faced with certain experiential data, a language remains open to the inclusion of sentences that are entailed by these sentences (such as sentences accepted on the basis of inductive inference) (Ajdukiewicz 1935, 113). This conception of language pushes to the front one immediate question: if languages are taken to be radically pluralistic, how could one determine if a language is a more or less adequate representational language? Not only is the problem pressing for representational questions (e.g., how do our languages more or less accurately represent the world?), but the problem is exacerbated by the sheer number of languages that exist under Ajdukiewicz’s strict conception of a language. Ajdukiewicz denies that “the” correct representational structure exists—we simply have a multiplicity of linguistic structures that are chosen based on practical factors unrelated to experiential data.6 Ajdukiewicz’s pluralism cannot be overstated, as he denies any neutral ground to evaluate truthclaims outside of their occurrence in a specific language. Ajdukiewicz is not skeptical of truth itself per se, but of the idea that one could determine the truth or falsehood of some claim outside of a specific language. Philosophy’s role, then, is conceived as articulating “the changes which occur in the conceptual apparatus of science and in the corresponding world-perspectives, and should seek to ascertain the motives which bring these changes about” (Ajdukiewicz 1935, 117). Thus, the philosopher’s job is to articulate the variety of conceptual structures within or between languages, changes from within those structures (and the corresponding theoretical motives that lead to those changes), but not to evaluate the truth or falsehood of such structures.7 One can reconstruct Ajdukiewicz’s solution to this representational issue as positing a given perspective of the world embodied by, or in, a language. Ajdukiewicz explicitly defines what he calls a world-perspective as “the totality of all sentences” of a language “combined with certain data of experience” (Ajdukiewicz 1935, 112), which functions as “the indubitable component of the knowledge for anyone who makes use of that language” (Ajdukiewicz 1935, 112–13). Such a perspective can be understood as exhibiting the total picture disclosed by a given language,

60

Peter Olen

containing all accepted sentences, including those accepted unconditionally due to the axiomatic rules of a language, and the sentences directly or indirectly entailed by these sentences. Thus, the three kinds of linguistic rules not only govern the use of all sentences within a given worldperspective but also dictate how certain sentences are correlated or “combined” with certain data of experience (i.e., empirical meaning rules would cover the use of those sentences that are licensed in the presence of certain experiences). Thus, such a perspective functions as the “knowledge of anyone who makes use of that language” because the concept is that of what a given language says about the world in which it is used (Ajdukiewicz 1935, 113). Though world-perspectives are the world disclosed by a language, such worlds are only a world. That is, such perspectives contained with a given language function as one world amongst an almost (technically speaking) infinite number of worlds disclosed by any number of languages. This is where Ajdukiewicz’s pluralism and rejection of a neutral evaluative perspective is most informative. Given that there are a seemingly infinite number of languages (given how, technically speaking, Ajdukiewicz demarcates between languages), and given that such languages are not matched up directly against the world in order to determine which is more or less true, every language contains a set of sentences that depicts its world, its particular perspective on the world (i.e., on the “material of experience” and the sentences issued as conceptual interpretations of such experiences). Such perspectives are constraining in the sense that the sentences of a given world-perspective dictate how we would conceptualize experiential data (given the correlation of empirical meaning rules and certain experiential data). “Constraining” should not be read too strictly here; insofar as one is choosing to use one language over another, one “obtains only an excerpt of the entire world-picture”, hence a “world-perspective” (Ajdukiewicz 1935, 116). The world disclosed by a given language constrains how we interpret the material of experience insofar as we choose to adopt that language (and adopt the specific world-perspective entailed by, or “contained in”, a language). Thus, a world-perspective functions as a kind of map that is part of the chosen language. Since we do not interpret our experiences without some conceptual structure, world-perspectives offer a slice of the overall picture of the world. From Ajdukiewicz’s perspective, there is a kind of voluntarism that underlies the choice to adopt this or that language, to apply this or that concept to a given experience in order to conceive of some experience from some perspective. Anticipating Sellars’s later remark that pragmatic and semantic metalanguages attempt “to mirror, within language itself, the relation of language to the world” (Sellars, CIL 114; emphasis added), world-perspectives account for meaning (the key semantic or pragmatic connection between language and world) by creating the relationship between language and world within language itself.

Ajdukiewicz and Sellars on World-Perspectives

61

Since the world-language relation is internal to some particular language or other, one might worry that we are trapped in a kind of linguistic idealism. If the total picture or perspective of a language user’s individual world is contained within the totality of sentences (as well as some correlation with experiential data and empirical meaning rules that ground the use of such sentences), then what is the connection between language and world? Given that Ajdukiewicz dismisses a more naïve relationship between expressions and their referents, his solution to this problem is essentially to mirror (or otherwise model, exhibit, replicate, etc.) the relationship between expressions and the world in the language itself, though he provides few details as to how this would work. This problem is compounded by Ajdukiewicz’s pluralism and refusal to grant a neutral perspective in order to adjudicate between different languages. Taken to its logical extreme, Ajdukiewicz would seem to have no way of adjudicating differences in meaning between speakers. Given that truth is always relative to some language, and each speaker could—in principle— have their own language (and, thus, their own world-perspective), there would be no possibility of reconciling these differences. Ajdukiewicz does not directly address this problem, but he does gesture towards pragmatic grounds for reconciling differences in language and meaning.8 Nonetheless, what is relevant for our concerns is how the formulation of this problem falls directly out of Ajdukiewicz’s conception of a language. Given his commitment to pluralism, as well as his notion of a world-perspective, one can see how this problem threatens to undercut his conception of language. I take this to be a meaningful issue not only because it problematizes some aspect of Ajdukiewicz’s work, but also because such a problem is inherited by Sellars. Given his use of concepts like Ajdukiewicz’s world-perspective, as well as his adherence to a kind of pluralism almost identical with Ajdukiewicz’s view, one should see the exact same problem transplanted into Sellars’s early work.

3 Sellars’s Appropriation Sellars’s initial appropriation of the concept of a world-perspective, which shows up as “world-story” in his publications, occurs in the midst of an attempted formalization of pragmatics. Although fairly complex, the salient aspects are as follows: Sellars’s early construction of what he called “pure pragmatics” is an attempt to “formalize” a pragmatic reconstruction of language. Sellars is not concerned with characterizing languages as uninterpreted calculi or semantical systems (as one finds in Carnap 1937 or Carnap 1942). Instead, Sellars is concerned with what he calls “empirically meaningful languages” that are “about a world in which they are used” (PPE 22). As will be explained later, although Sellars does refer to a multitude of linguistic structures (empiricist, rationalist, idealist, phenomenalist, naive realist, etc.), his overall concern is with

62

Peter Olen

languages as “used”. This is important because a pragmatic account of language means that representation or designation cannot depend on the simple inclusion of rules of designation, but require looking at the conditions under which such concepts are used. Like Ajdukiewicz, Sellars must credit meaningful roles to sentences and expressions not exhausted by straightforward referential relationships. Sellars’s early goal, in part, is to show how pragmatic accounts of language best capture the representational character of natural language while avoiding a steadfast commitment to a one-to-one correspondence between language and world.9 The specifically pragmatic dimension of this analysis turns on Sellars’s inclusion of what he calls a formal predicate, “co-ex”, which is modeled on the psychological concept “is-present-to-consciousness-along-with” (PPE 10). This concept is intended to capture the role language users play by representing the direct connection between pieces of language (e.g., heard, spoken, or thought linguistic items) and a language user’s experience. Such an inclusion allows Sellars to claim that he is offering a pragmatic analysis of language because his co-ex concept allows for the structural representation of, and reference to, an individual language user’s “immediate experience” in conjunction with linguistic phenomena (whether uttered, heard, or otherwise “experienced”) (RNWW 54). This is an early attempt to offer a “pragmatic” account of language in the sense that such a “formal” concept ostensibly accounts for the agential role individuals play in producing and employing a language. This is not to say that such a theoretical move works (“works” in the sense that such a concept is needed to understand meaning, for example, or that such a concept needs specifically formal articulation), but that Sellars himself saw this concept as the missing link between the syntactical and semantic phase in then-recent changes to analytic philosophy.10 World-stories are presented as “representational maps” (Sellars, NAO 128–9) of language through which a language user knows her world (RNWW 55). The “world” in “world-story” is a “collective term for the designata of a world-story” (ENWW 38), whereas a “world-story” serves as “the body of logically simple (atomic) sentences which constitute the story of the universe in which” a language user lives (ENWW 30). Sellars divides the sentences of a world-story between “verified” and “confirmed” sentences. A verified sentence is one that confronts the designata of its designating terms (and, thus, “is-present-to-consciousness-along-with” the experiential data of a language user), whereas confirmed sentences do not receive experiential confrontation with their referents (RNWW 55–6). So, a world-story functions as the meaning-base of a language that constitutes a language user’s knowledge of her world (i.e., the world depicted through the use of a given language). As Sellars puts it: “Thus, to understand the notion of different worlds, we must understand those of different stories and different languages” (ENWW 33). The possibility of formulating almost infinite world-stories, especially largely false ones,

Ajdukiewicz and Sellars on World-Perspectives

63

is what allows for natural language to exhibit the flexibility it does, while retaining some connection to the world. How, exactly, this gives us a conception of language that maintains a connection with the world remains unclear. As with Ajdukiewicz’s conception of a world-perspective, Sellars’s understanding of a world-story, especially as the concept allows for infinite formulations of different stories, would seemingly suffer from the exact same form of linguistic idealism. Just as this issue is a direct result of Ajdukiewicz’s commitment to pluralism and the idea of a world-perspective, one can see the exact same issues in Sellars’s early conception of language. This concern does not follow simply because Sellars coincidentally used a similar, if not the same, notion as Ajdukiewicz, but rather because linguistic idealism arises specifically because of Sellars’s inheritance from Ajdukiewicz. In these early publications linguistic rules play the usual role of constituting the structure of a language. In addition to the usual rules of formation and transformation found in pure accounts of language (at the time), Sellars introduces the notion of conformation rules (which, given his historical context, shadow Carnap’s conception of material rules of inference) as a way to “unify” the verified and confirmed sentences of a world-story: In order for a world-story to contain sentences which are confirmed but not verified, the atomic sentences which constitute the story must have a unity over and above that of satisfying the syntactic requirements (formation rules) of the language. The status of being confirmed but not verified requires a criterion of togetherness in one sentence structure; conformation rules as well as formation rules. (RNWW 57) Conformation rules allow for formally reconstructed languages to exhibit the kind of unity found in natural languages by restricting the combinations of sentences possible under a given conception of language. For Sellars, conformation rules provide this unity, in one instance, in the distinction between “verified” and “confirmed” sentences. They do so by stipulating that the individual constants of such sentences “must participate in relational as well as non-relational sentences” (RNWW 57), which in turn (depending how one might flesh out a given rule) restricts the possible combinations of individual constants and sentences themselves. Although offering a significantly more developed account of linguistic rules than one finds in Ajdukiewicz’s 1930s papers, Sellars’s conformation rules play the same demarcational role between languages one finds in Ajdukiewicz: same rules, same languages; different rules, different languages (and, thus, different worlds). Notice the remarkable similarity between Sellars’s understanding of verified and confirmed sentences and Ajdukiewicz’s distinction between sentences falling under empirical

64

Peter Olen

meaning-rules (that are correlated with experiential data) and sentences entailed by, but not directly falling under, such rules (and, thus, not directly correlated with experiential data): Knowledge, however, need not be restricted to those sentences already gained in accordance with the meaning rules of the language; for it can also include sentences which are accepted, although, on the basis of the hitherto achieved portion of the world-perspective and the hitherto enjoyed data of experience, they are not demanded (nor forbidden) by any meaning-rule of the language. . . . With its help, we attempt to piece together by way of anticipation the as yet unknown parts of the world-perspective. (Ajdukiewicz 1935, 113) Additionally, Sellars’s definition of a world-story plays the exact same role as Ajdukiewicz’s world-perspective. Both concepts play a role in the meaning base of a language, are partially correlated with experiential data, yet include sentences that are not correlated with such data and are relative to a given language. Yet even if Sellars is employing the same notion found in Ajdukiewicz’s philosophy, does this necessarily commit him to a form of radical conventionalism or pluralism about languages? Ajdukiewicz’s pluralism consists in the idea that there are infinitely many world-stories and infinitely many empirical language forms. In an almost identical line of reasoning, Sellars remarks: The pure theory of empirical languages as formally defined systems which are about worlds in which they are used, has no place for THE world; but only for the world designated by the story which is the meaning base of a language. A given set of conformation rules defines a family of empirical languages, or, which is the same thing, a family of possible worlds which have the same laws. An understanding of the completely non-factual character of epistemological statements rests on the insight that not even the predicates “verified” and “confirmed” have an intrinsic tie with any single world, with “the REAL world”. They are purely formal predicates and no properly constructed world-story stands in a privileged position with respect to them. This principle of indifference could be discarded only if something akin to an ontological argument could be formulated in the pure theory of empirical languages; if it could be shown, for example, that only one set of conformation rules is possible which enables a story to be constructed in the language form of which they are the rules; and if only one story could be constructed in that language form. (RNWW 65)

Ajdukiewicz and Sellars on World-Perspectives

65

Sellars is committed to the dual ideas that there are a multitude of empirically meaningful languages, yet none of the world-stories of such languages represent the world, but simply a world. That is, even though every language reflects some kind of representational structure or “map”, this does not mean that an empirically meaningful language is accurate to the world (i.e., every empirically meaningful language simply reflects the totality of verified and confirmed sentences contained within its meaning-base). What Sellars calls the “principle of indifference” just is Ajdukiewicz’s rejection of “neutral ground” for the philosophical or epistemological evaluation of the truth of such languages as a whole. The points of agreement here are fairly clear; both Sellars and Ajdukiewicz think of languages as containing a multitude of kinds of linguistic rules that, despite terminological shifts, seem to correspond with each other. Any variation of such rules signals the use of a different language, not simply a disagreement in meaning. Sellars’s “principle of indifference” and Ajdukiewicz’s lack of neutral perspective radicalize Carnap’s principle of tolerance. Whereas Carnap’s principle theoretically extends to any subject that is reconstructible in formalist terms (e.g., mathematical and scientific languages), Sellars and Ajdukiewicz are casting wider nets. Both philosophers are not making claims about the reconstruction of scientific languages per se, but about any language that counts as “empirically meaningful” in Sellars’s terminology. From a philosophical, as opposed to empirical, perspective, Sellars’s commitment to what I have been calling linguistic pluralism is doubly joined with Ajdukiewicz’s adherence to the denial of neutral ground for evaluating languages. Invoking the exact same language as Ajdukiewicz, Sellars argues that it is not the epistemologist’s job to evaluate the accuracy or usefulness of linguistic structures (ENWW 38). Instead, “Epistemology, or the pure theory of empirically meaningful languages, can develop the formal properties of languages with different conformation rules, but cannot ‘choose’ the story or the language. Epistemology can show, or expose the formal confusion that underlies attempts to show, that one or other type of story or language is internally inconsistent” (ENWW 38). Invoking the same phrasing as Ajdukiewicz, Sellars is fairly clear that “epistemology as the pure theory of languages can develop the formal properties of languages with different conformation rules and can compare realistic with non-realistic languages: but as a purely formal discipline cannot choose THE confirmation rules or THE language. It is a mistake to look for a formal (epistemological) justification of ‘Realism’ or ‘Idealism,’ etc” (RNWW 69). Both philosophers give us a conception of language opposed to psychological treatments of language, and both philosophers argue for the necessity of world-perspectives or world-stories that present an overall picture of meaning and knowledge relative to an individual’s adopted language. That being said, I am not arguing for the truth or necessity of

66

Peter Olen

these claims. The point of this chapter is to clarify the historical relationship between Ajdukiewicz and Sellars, as well as present some critical perspective on their conceptually and historically related ideas. Whether Ajdukiewicz’s or Sellars’s concepts are “correct” or true is largely beside the point in this instance. What I have tried to make explicit in this section is simply the relationship between Ajdukiewicz’s and Sellars’s work. Specifically, it is the shared commitments to pluralism and the need for some world-perspective or story that, in part, raises issues of idealism (amongst others). More important, both philosophers use specific ideas (e.g., world-perspectives, the distinction between confirmed and verified sentences) in the exact same way. Insofar as I am concerned to establish a historical connection between both philosophers, I think it is clear that some of Sellars’s early concepts and commitments follow Ajdukiewicz’s approach to the exact same issues and ideas.

4 Problems and Extensions Despite originating out of different conceptual pressures, Sellars clearly appropriated the concept of a world-perspective directly from Ajdukiewicz’s work. Additionally, some of the critical concerns one might have about Ajdukiewicz’s philosophy of language can be found in Sellars’s early work as well. This is, as I have argued earlier, a direct result of the historical relationship between the two philosophers. The preceding cited passages, combined with Sellars’s role as Ajdukiewicz’s translator, makes the veracity of this claim unassailable. Nonetheless, this is not enough to guarantee anything philosophically interesting exists in light of this connection. Even if the historical connection is there, exploring the relationship between Ajdukiewicz and Sellars may clarify nothing. This would be history without a purpose; an account of philosophers or philosophy that doesn’t address the “so what?” question fails in separating mere historical bookkeeping from explanations that clarify and extend our previous knowledge. Of course, I do not think exploring the relationship between Ajdukiewicz and Sellars amounts to mere bookkeeping. I want to propose three critical directions for how we might find the relationship between Sellars and Ajdukiewicz philosophically compelling—one internal to Sellars’s philosophy, one concerning the role of idealizations, and one largely historical. My initial suggestion concerns the internal consistency of Sellars’s philosophy. To be clear, though, by “internal consistency” I do not mean the overall coherence and consistency of Sellars’s philosophy. While often touted as a systematic philosopher, I have argued at length (Olen 2016) that such views mistakenly assume Sellars’s philosophy did not undergo significant changes throughout his career. What we can do is see if Sellars’s use of world-stories makes sense in light of differences in his earlier and later work.11 As noted in Section 3, world-stories are initially

Ajdukiewicz and Sellars on World-Perspectives

67

introduced under the guise of being a formal device—one that is properly at place in philosophy qua formal reconstruction of language.12 When re-introduced in Sellars’s later work, the formalist framework that underpinned Sellars’s early philosophy has disappeared. Does this metaphilosophical change make any difference to the employment of world-stories? At first pass, it does not seem that Sellars’s shifting metaphilosophy makes much difference here. Whether world-stories are considered formal devices or not would seem to make little difference to their overall explanatory power. Although the status of formal concepts as formal concepts13 is frequently used as a justificatory device in his early publications, there is nothing about a language presupposing a representational map that requires it be a formal concept. That being said, Sellars does reference his earliest publications, much in their same form: “In introducing the topic of maps in the previous section I referred to the view, which I have held since my earliest publications that the representational features of an empirical language require the presence in the language of a schematic world story” (NAO 108). Despite substantial changes to his metaphilosophy, the concept of a world-story is still presented in the exact same fashion one finds in Sellars’s earliest publications. The change that matters here is in the change in target language. That is, when evaluating differences between the early and later Sellars, we might see one of the biggest differences as the language under investigation. Whereas the early Sellars is explicitly concerned with developing formally constructed languages, while drawing a rigid distinction between formal and empirical investigations of language, the later Sellars is clearly concerned with natural languages. This is a departure from Sellars’s earliest philosophy, where the investigation of natural language concepts were generally left to psychology or anthropology. Philosophy’s job, instead, was to construct formal accounts of language that potentially kept one eye on practices (though it need not). Once we shift away from a formal metaphilosophy into squarely working within natural languages, the constraints on concept formation are different. When standing in the formal domain of analysis, how our concepts match the world is, according to the early Sellars, not an issue for philosophical analysis. That is, rather, an empirical question of linguistics or descriptive inquiries. The later Sellars, though, is clearly concerned (as described later) with offering something approaching the fundamental structure of natural language and thought (albeit approximations of the actual concepts we employ in navigating the world). Within his later metaphilosophy, exploring how philosophical concepts match “things or practices” in the world plays an active role in our understanding of natural language and thought. As noted elsewhere (see Olen 2016), the shift from a formalist to nonformalist metaphilosophy must been seen as making a meaningful difference both in general and specific to the early Sellars’s overall project. Ajdukiewicz attempts, at least in gesture, to separate psychological from

68

Peter Olen

logical aspects of language, but this does not carry the same strict connotations as one finds in the early Sellars. Ajdukiewicz points out that, insofar as we are concerned with the logical conception of language, then we would not be concerned with an agent endorsing or rejecting a given sentence. Logical depictions of language can portray the negation of a given sentence, for example, but not the psychological act of endorsing one. Again, there is an interesting historical issue in the distinction between psychological and logical accounts of language. How can meaning go beyond subjective association of words and sounds? How can the use of language depict more than individual meanings? Both Ajdukiewicz and Sellars are explicitly concerned with these issues, and their distinction between psychological and logical understandings of language protects a role for specifically philosophical accounts of language. Although Ajdukiewicz gestures towards this issue, it is one of Sellars’s main preoccupations with language. While there are clear differences between the early and later Sellars on this point, both Sellars and Ajdukiewicz remain somewhat ambiguous as to how idealized concepts relate to natural languages. Ajdukiewicz, more so than Sellars, attempts to deal with the question directly: In meeting this objection, let us remark at the outset that almost all the sciences exhibit this “tendency toward idealization”. Physics, e.g., sets up theses about ideal gases, although it is well known that no such gases exist. Again, in mechanics, the concern is with the motions which run their course under conditions never actually realized. Physics does this, we submit, because it is the only possible way for knowledge to approach reality. First of all, one sets up laws which hold rigorously only for ideal gases; for real gases, they hold only to within a rather larger error of approximation. To begin by instantly demanding absolute agreement between the laws and reality is to set much too difficult a problem for ourselves. We begin by considering an ideal case which agrees only approximately with reality. (Ajdukiewicz 1934b, 88) Ajdukiewicz is responding to the objection that philosophical notions of a language amount to far-fetched fictions. Insofar as we treat idealizations about language (e.g., stable representational maps, formal rules of language, exact coordination between linguistic rules and experiences) as approximations that leave room for the messiness of natural language, then good philosophy is simply mimicking good science. Our concepts work (in terms of describing and explaining the world) insofar as we treat them as approximations of the concepts operative in natural language and thought. Sellars, especially in his later work, shows no uneasiness with taking idealizations as somehow related to natural languages. As I have argued

Ajdukiewicz and Sellars on World-Perspectives

69

elsewhere,14 I think it is far from obvious that this practice is not problematic. When discussing the inference rules of logic, for example, Sellars claims that: Now I submit that the logician’s concepts of formation rule, transformation rule, and rule permitting the substitution of one expression for another, have legitimate applications to natural languages. By this I mean not that it is possible for the logician to construct such rules for natural languages, but rather that rules of these types are embedded in natural languages themselves without any help from the logician. That the vague, fluctuating, and ambiguous character of ordinary usage extends to these rules is, indeed, granted. But does not the same hold true with respect to the logician’s concept of a sentence? Or of a predicate? Yet we do not hesitate to discuss natural languages in these terms. I see no reason in the Heracleitean character of ordinary usage to reject what would seem to be the obvious implication of the fact that natural languages can be illuminated by confronting them with artificial languages obeying explicitly formulated rules of transformation and synonymity. (ITSA 301–2) This is a confusing passage, at least in terms of getting clear on Sellars’s thoughts about idealizations. While he initially seems to be claiming that the concepts of formal logic are “embedded”15 in natural languages, he nevertheless also treats such formal concepts as mere approximations of their natural language counterparts. I take it that the important aspect of these claims is the idea that natural languages can be clarified by being held against artificial languages. Here, as elsewhere, his view is similar to Ajdukiewicz’s view. I do not think this somehow problematizes Sellars’s use of worldstories in his later work. This is largely because the link between agent, language, and world shifts between his earlier and later work as well. The co-ex predicate—what Sellars calls the defining predicate of pure pragmatics—is taken to be a formal representation of a coordination between an agent’s experiences and linguistic practices, which serves as a connection between language and the world (much like with Ajdukiewicz’s inclusion of experiential data in his conception of a world-perspective), and picturing fulfills a similar (if not the same) role in Sellars’s later work. In fact, Sellars’s later discussion of world-stories replaces the role of the co-ex predicate (something that permanently disappears from Sellars’s philosophy along with his early formalist commitments) with a discussion of picturing.16 Where these changes matter most, at least in terms of worrying about the relationship between Ajdukiewicz and Sellars, might be with the threat of a kind of linguistic idealism. As mentioned earlier, one critique

70

Peter Olen

of Ajdukiewicz’s position might be to claim that his conventionalism and pluralism are too radical and, thus, do not allow for a clear connection between language and world. All we get out of radical conventionalism is a kind of language-to-language relationship that never achieves any critical or evaluative relationship to the world. Ajdukiewicz’s response would be to point out the correlated connection between concepts and experiential data in order to reinforce the link between language and world, as well as gesture towards [ultimately] pragmatic grounds of confrontation between different languages. However, given the fact that radical conventionalism denies the existence of experiential data independent of our concepts or chosen language, it is difficult—if not impossible—to see how such experiences could function as the connection between language and world. Ajdukiewicz is clear that one never “meets” experiential data without some “conceptual apparatus” at play, yet such a conceptual apparatus is seemingly required prior to “meeting” said experiential data. Perhaps he would say that such rules are always tied to some language or conceptual apparatus, thus how we interpret a given experience is always bound to some language or conceptual apparatus. Empirical meaning rules, consequently, would never give us a direct connection to the referents of terms, for example, but would only constrain our linguistic usage relative to some set of sentences or concepts. So, even my assent to given sentences in the presence of experiential data would be conditioned on the acceptance of some set of concepts or a given language. Ajdukiewicz’s pluralism goes all the way down. Picturing, ostensibly, can solve this problem in ways unavailable to the formal concepts. The co-ex predicate, in addition to being absent from Sellars’s later philosophy, arguably shares the same downside as Ajdukiewicz’s discussion of coordinating concepts and experiential data. Even as the formal correlate of the psychological (i.e., descriptive) notion of sensations and pieces of language “being-present-to-consciousnessalong-with”, it is fairly difficult to see how this concept fits with the kind of conventionalism residing in Ajdukiewicz’s and Sellars’s work. We might see Sellars’s repositioning of world-stories away from the co-ex predicate and towards picturing along much of the same lines. This is not to say the later Sellars should be read as moving away from the distinction between the world and a world in our representational systems; he insists that this distinction is an integral starting point for a correct account of language and thought (NAO 108–9). Instead, I think the move away from the co-ex predicate (arguably a conventionalist solution to the dilemma of connecting language and world)17 to picturing allows Sellars to keep the idea of world-stories qua maps, a connection between language and world, while quietly dropping the formalist metaphilosophy that marks his earliest publications.18 While the concept of a world-story stays intact and unchanged, the metaphilosophy (and the target of such concepts) changes. In Sellars’s later philosophy, world-stories can be seen

Ajdukiewicz and Sellars on World-Perspectives

71

as a transcendental requirement for a language to be empirically meaningful, whereas Sellars’s early formalism dictates that such a concept be purely structural (at best) within the context of pure pragmatics. Sellars’s later work can target, and incorporate, facts about linguistic usage that (despite working under the guide of “pragmatics”) were generally ignored as factualist delusions in his early philosophy. So, we do see some change and transition in Sellars’s overall view, but not in the concept of a world-story itself.19 I do not think, of course, this change comes directly from Sellars’s exploration of Ajdukiewicz’s work, but I do think these changes become explicit when thinking through Sellars’s use of world-stories as it appears in his earliest essays and re-appears in his later discussions of language. Understanding Sellars’s work in light of his connection to Ajdukiewicz not only helps solve a small, historical puzzle but also brings attention to relatively neglected aspects of Sellars’s philosophy. That being said, there are other areas of influence between Ajdukiewicz and Sellars that seem quite clear, but may not be philosophically significant. Ajdukiewicz’s use of the language of “acceptance”, “endorsement”, and “rejection” when depicting individuals as full-blown language users and knowers is eerily predictive of Sellars’s own employment of these terms in “Empiricism and the Philosophy of Mind”.20 Yet, aside from describing a possible line of influence, I am not sure why this connection would be particularly informative. The language used by both philosophers is strikingly similar and they are, all things considered, making remarkably similar points about the difference between reporting and expression roles.21 But in this instance there really is no difference that makes a difference. One might also see the connection between Sellars and Ajdukiewicz as one nuance of an increasingly complex narrative of logical positivism’s reception in North America.22 While this project began over a decade ago (best encapsulated by Hardcastle and Richardson 2003), the fruits of developing a more nuanced picture of logical positivism have not developed within scholarship surrounding Sellars’s work. Carnap, frequently seen as the positivist, is not the only influence from this tradition to be found in Sellars’s work. References to Gustav Bergmann, Hans Reichenbach, Herbert Feigl, Philipp Frank, and others abound throughout Sellars’s publications.23 This is not to say historical connections between Sellars and other philosophers have been ignored by the scholarly community, but the influence of logical positivism on Sellars’s thought has generally been boiled down to Carnap’s or W.V.O. Quine’s influence on Sellars. This is unfortunate, mainly because there are many ways in which a variety of positivists lurk in the background of Sellars’s thought.24 Thinking through the relationship between Ajdukiewicz and Sellars not only elaborates our understanding of the history of American philosophy but also helps locate Sellars’s place within it.

72

Peter Olen

Notes 1. Although Feigl and Sellars held editorial control, the composition of this anthology was—at least in spirit—a group effort: “The project in preliminary form was presented by circular letters to about ninety teachers of philosophy in this country and in England. We asked for responses to our proposed selections, that is, endorsements or rejections of titles contained in a list of about 130 items” (Feigl and Sellars 1949, V). 2. That being said, there is no concrete evidence for this claim; Sellars does not cite Ajdukiewicz in any of his work and, so far as I can tell, there is no surviving correspondence between Ajdukiewicz and Sellars. Yet the two essays in question (Ajdukiewicz 1934a, 1934b) were both published in Erkenntnis, one of which even appearing in the same issue as Ajdukiewicz’s world-perspective piece. It seems reasonable to assume that Sellars was familiar with both pieces. 3. One could see this, for example, as a “radicalization” of Henri Poincare’s conventionalism. For an extended discussion of Ajdukiewicz’s conventionalism, see Wolenski (1996). 4. The observation that Ajdukiewicz’s conventionalism is related to his linguistic pluralism is, in part, also made by Wolenski. See Wolenski (1996, 109–10). 5. This is—among other things—just one difference between some of the LvovWarsaw philosophers and the logical positivists (especially Carnap). Where Carnap might bemoan the complexity of applying formal concepts to natural language (see the appendix of Carnap 1942), Ajdukiewicz is comfortable with the rough approximation with the idealized character of such notions. 6. Ajdukiewicz’s formulation of this point predates Carnap’s “Empiricism, Semantics, and Ontology” by 15 years. Although both philosophers have differing conceptions of what Carnap would describe as internal and external questions, there is an open question of influence I believe has not been addressed in the secondary literature. 7. See Ajdukiewicz (1934b, 81–4) for a lengthy discussion of his position on “neutral standpoints”. 8. For a fuller treatment of this point, see pages 81–84 in Ajdukiewicz (1934b). 9. One fact that complicates Sellars’s early account is his insistence that such pragmatic accounts be seen as formal (as opposed to empirical). This is drastically out of step with then-current discussions of pragmatics as a descriptive and empirical study of language or linguistic habits (as championed, for example, by Carnap and Charles Morris). See Olen (2016) (especially chapters I and II) for a discussion of this point. 10. See Sellars PPE, 4–5. 11. I am taking for granted that the disappearance and reappearance of “worldstories” in Sellars’s work is not accidental. Given that Sellars spent years touching on the same topics in the philosophy of language without invoking the concept of a world-story, one assumes Sellars’s later re-introduction of the term really should be understood as a re-appropriation of its earlier appearance in his “pure pragmatics” papers. 12. What Sellars means by “formal” in his early writings is far from clear. What’s more, he frequently employs inconsistent uses of the term, only making the matter a bit more vexing. For a longer account of this problem, see Olen (2016) (especially chapters I and II). 13. In his earliest publications, Sellars tends to assert the formal status of concepts as a justification for their necessity (over and above empirical considerations). This is, at best, confusing and, at worst, downright incoherent. 14. For further discussion on this point, see Olen (2016, 90–6).

Ajdukiewicz and Sellars on World-Perspectives

73

15. A paragraph after this Sellars raises the idea that such rules could be the reason behind one’s behavior, despite our inability to formulate said rules. This is, perhaps, foreshadowing his discussion of rules of criticism and rules of conduct, but Sellars (again) leaves it ambiguous as to the status of idealizations and natural language. 16. This begins on page 113 in NAO, which is a reprint of his discussion of picturing in TC. 17. Determining the “right” understanding of the connection between language and world (or terms and their referents) functions as a motivating problem for the early Sellars. This problem, in turn, arises out of a misreading of Carnap. See Olen (2016) and Olen (2017) for explanations of this issue. 18. For more on picturing, see Sellars, TC 211–24. For more recent work on Sellars’s conception of picturing see Sachs (2018). 19. One could argue Sellars’s early and late insistence on discussing languages “about a world in which they are used” signals, perhaps, a greater consistency between his early and later work than I’m granting in the preceding argument. Although Sellars discusses languages as “empirically meaningful” in various stages of his philosophy, I think reading a uniform narrative into his philosophical development is too narrow. Sellars, at least initially, is an unapologetic formalist about what constitutes a “proper” philosophical analysis of language and what kind of language is under investigation. The difference is one of kind, not degree. 20. See Ajdukiewicz (1934a, 42–3). 21. See Section III of “Empiricism and the Philosophy of Mind” (“The Logic of Looks”) for Sellars’s discussion of endorsement and reporting roles. Sellars and Ajdukiewicz even use the exact same “bent stick” example. 22. This is not to say that Ajdukiewicz should be seen as a logical positivist per se, but that the initial appearance of his work in Erkenntnis in German (especially after Carnap’s time with the Polish group) during the early 1930s would place him alongside the positivists (especially, one imagines, in Sellars’s mind). The Lvov-Warsaw school’s continued interest in “traditional” problems of philosophy, as well as traditional commentaries on these problems (see Ajdukiewicz’s discussion of “the Copernican idea of Kant”, as just one example, in Ajdukiewicz 1934b, 86–8), sets them apart from the more narrow, revolutionary project of logical positivism. 23. For examples of why some of these figures matter when reading Sellars’s work, see Olen (2016) and Olen (2017). 24. One of the more interesting and fruitful unexplored connections might be that of the relationship between Hans Reichenbach and Sellars. Although currently unexplored, a discussion of Reichenbach’s Experience and Prediction in relation to Sellars’s epistemological views on givenness could be promising.

References Ajdukiewicz, Kazimierz. 1934a. “Sprache und Sinn”. Erkenntnis 4: 100–38. Reprinted as “Language and Meaning.” In Kazimierz Ajdukiewicz. 1978. Kazimierz Ajdukiewicz: The Scientific World-Perspective and Other Essays 1931–1963, edited by Jerzy Giedymin, 35–65. Boston: D. Reidel Publishing Company. Hereafter Ajdukiewicz 1978. ———. 1934b. “Das Weltbild und die Begriffsapparatur”. Erkenntnis 4: 259–87. Reprinted as “The World-Picture and the Conceptual Apparatus.” In Ajdukiewicz 1978, 67–89.

74

Peter Olen

———. 1935. “Die Wissenschaftliche Weltperspektive”. Erkenntnis 5: 22–30. Reprinted as “The Scientific World-Perspective”. In Feigl and Sellars 1949, 182–8, and in Ajdukiewicz 1978, 111–17. Carnap, Rudolf. 1937. The Logical Syntax of Language. London: Routledge & Kegan Paul LTD. ———. 1942. Introduction to Semantics. Cambridge: Harvard University Press. Feigl, Herbert and Wilfrid Sellars, eds. 1949. Readings in Philosophical Analysis. New York: Appleton-Century-Crofts. Hardcastle, Gary and Alan Richardson, eds. 2003. Logical Empiricism in North America. Minneapolis: University of Minnesota Press. Olen, Peter. 2016. Wilfrid Sellars and the Foundations of Normativity. London: Palgrave Macmillan. ———. 2017. “A Forgotten Strand of Reception History: Understanding Pure Semantics”. Synthese 194: 121–41. Sachs, Carl. 2018. “‘We Pragmatists Mourn Sellars as a Lost Leader’: Sellars’s Pragmatist Distinction between Signifying and Picturing”. In Sellars and the History of Modern Philosophy, edited by Antonio Nunziante and Luca Corti, 157–76. New York: Routledge. Sellars, Wilfrid. (1947) 2005 [ENWW]. “Epistemology and the New Way of Words”. In Wilfrid Sellars. 2005. Pure Pragmatics and Possible Worlds: The Early Essays of Wilfrid Sellars, edited by Jeffrey F. Sicha, 28–40. Atascadero: Ridgeview Publishing Company. Hereafter PPPW. ———. (1947) 2005 [PPE]. “Pure Pragmatics and Epistemology”. In PPPW, 4–25. ———. (1948) 2005 [CIL]. “Concepts as Involving Laws and Inconceivable without Them”. In PPPW, 86–114. ———. (1948) 2005 [RNWW]. “Realism and the New Way of Words”. In PPPW, 46–78. ———. (1953) 1991 [ITSA]. “Is There a Synthetic a Priori?” In Wilfrid Sellars. (1963) 1991. Science, Perception, and Reality, 298–320. Atascadero: Ridgeview Publishing Company. Hereafter SPR. ———. (1962) 1991 [TC]. “Truth and ‘Correspondence’”. In SPR, 197–224. ———. 1979 [NAO]. Naturalism and Ontology. Reseda, CA: Ridgeview Publishing Company. Wolenski, Jan. 1996. “Radical Conventionalism and Empiricism”. Philosophia Scientiae 1: 107–17.

4

Sellars and Wittgenstein on Following a Rule Stefan Brandt

1 Introduction As far as I know, the only views of his that Sellars explicitly derives from Wittgenstein are his theory of linguistic picturing and his related theory that predicate-expressions are ultimately dispensable parts of language.1 Both these theories are derived from the Tractatus. Wittgenstein’s later work, on the other hand, is occasionally referenced but, to my knowledge, never explicitly acknowledged as the source of Sellars’s views.2 Still, in his “Philosophy and the Scientific Image of Man” (PSIM) Sellars expresses considerable appreciation of Wittgenstein and Wittgensteinian philosophy. He writes: I think it is correct to say that the so-called “analytic” tradition in recent British and American philosophy, particularly under the influence of the later Wittgenstein, has done increasing justice to the manifest image, and has increasingly succeeded in isolating it in something like its pure form. (PSIM 15)3 Given this opinion one should expect substantial similarities between Wittgenstein’s later views and Sellars’s philosophy, especially where the latter is concerned with delineating the “manifest” or commonsense picture of the world. I think that this is indeed the case and that at least some of these similarities are likely to be due to Wittgenstein’s influence. The latter suspicion is strengthened by the fact we now know that Sellars was aware of some of Wittgenstein’s later views long before the publication of Philosophical Investigations in 1953. For the Sellars Archive at the University of Pittsburgh contains a copy of Wittgenstein’s Blue and Brown Books, which is signed and dated: “Wilfrid Sellars, Harvard 1937” (Wilfrid S. Sellars Papers, box 9, folders 12–13). One area, in which I strongly suspect an influence of Wittgenstein’s later writings on Sellars, is the topic of rules and rule-following. There are several reasons for this. First, Sellars’s first and most detailed development of his mature views on this subject in “Some Reflection on

76

Stefan Brandt

Language Games” (SRLG) refers quite clearly to Wittgenstein in its title; although only implicitly and Wittgenstein is not mentioned in the paper itself.4 Second, in a notebook entry dated “11 Aug 1953” Sellars writes: It is about time that I attempted once again to think through the problem about linguistic rules which I ran into while writing “Language Rules and Behavior” and which I discussed rather cavalierly in a footnote.5 (Wilfrid S. Sellars Papers, box 25, folder 01–02, 66–7)6 This remark is followed by an outline and subsequent discussion of a puzzle about rule-following that opens SRLG. So, it is reasonable to assume that August 1953 marks the start of Sellars’s work on that paper, which was published in July 1954. He may very well have been stimulated to do this by his reading of Wittgenstein’s Philosophical Investigations; for the latter was published in 1953 and Sellars’s copy is signed and dated “May 15 1953”, i.e., almost exactly three months before he returned to the topic of rule-following in his notebooks. Furthermore, the sections on rule-following in Sellars’s copy of the Investigations are the most heavily underlined and annotated parts of the book. So, it is likely that Sellars’s thinking on rule-following is one part of his philosophy that is directly influenced by Wittgenstein’s later writings. Furthermore, it is an area in which both similarities and differences in their views come particularly clearly into view. I shall, therefore, now turn to a comparison of their views. I shall begin with a brief discussion of the place of rules of language in their philosophy quite generally. I shall then turn to Sellars’s conception of following rules and subsequently compare it to Wittgenstein’s views. Finally, I will locate the differences in their views as resulting from Sellars’s commitment to an event-causal conception of acting for a reason; a commitment that is not shared by Wittgenstein.

2 The Place of Rules in Wittgenstein’s and Sellars’s Thinking The notion of a rule of language plays an important role in both Wittgenstein’s and Sellars’s thinking. This comes out in a number of claims about linguistic rules they appear to agree on: first, they both identify the meaning of a word with its rule-governed use in a linguistic practice.7 A conception that is well illustrated by Wittgenstein’s notion of a “language game”, which Sellars frequently employs. Second, they therefore also agree that when we use a word with a specific meaning we follow certain rules for its use. Third, they agree that linguistic practices are essentially normative, i.e., that rules of language are genuine norms for the use of words and not mere generalizations about how words are actually used. Sellars repeatedly stresses the essentially normative character of language

Sellars and Wittgenstein on Following a Rule

77

and intentional phenomena quite generally (cf., e.g., IM, 24; EPM §5: 131, §36: 169) and Wittgenstein writes, for example: “Following according to the rule is FUNDAMENTAL to our language game. It characterizes what we call description” (Wittgenstein 1978, VI-28). Fourth, along with a shared conception of meaning as rule-governed use come similar conceptions of how we specify the meaning of linguistic expressions. Prima facie, though, the way Wittgenstein and Sellars talk about specifying the meaning of words seems very different. Wittgenstein typically talks about explanations of meaning and tends to stress that they can be informal and, as it were, hand-waving, without therefore being in any way deficient (cf., e.g., Wittgenstein 1958, 1ff.; Wittgenstein [1953] 2009, §§68–80). A characteristic example is his discussion of the explanation of the meaning of “game” in Investigations §69: he stresses that an explanation by listing a number of games and saying “This and similar things are called ‘games’” is perfectly in order. Sellars, on the other hand, typically uses his technical apparatus of dot-quotes to give the logical form of statements specifying the meaning of a word, phrase, or sentence. In particular, he argues that sentences of the form (i) x means y, such as (ii) “dreieckig” means triangular, have the logical form: (iii) “x”s are •y•s, or, in the case of our example: (iv) “dreieckig”s are •triangular•s. Dot-quotes turn expressions into what Sellars calls “illustrating sortals” (MFC, 95), which serve to classify expressions as playing the same rulegoverned role within their language as the dot-quoted (“illustrating”) expression plays in the metalanguage. Hence, (iv) classifies “dreieckig” as playing the rule-governed role in its language, i.e., German, that is played by “triangular” in English. And since (iv) is an English sentence, it not merely says that “dreieckig” and “triangular” play equivalent roles in their respective languages, it also tells us what this role is. Despite these differences, Sellars’s and Wittgenstein’s remarks about specifications of meaning share two important similarities: they both stress that explanations of meaning do not relate expressions to some non-linguistic items that are their meaning, but rather give a

78

Stefan Brandt

(Wittgenstein) or the (Sellars) rules for its use. And both emphasize that specifications of meaning remain within language. Sellars stresses that they are non-relational since “means”, as his analysis shows, is a specialized form of the copula (cf. MFC, 95). And while Wittgenstein does not make this particular claim, he stresses that explanations of meaning do not relate expressions to items outside of language. This becomes particularly clear in his discussion of ostensive definitions. Contrary to appearances (cf. Wittgenstein 1958, 1) ostensive definitions do not relate an expression to an item that is outside language and that is the meaning of the expression. They rather give a rule for the use of an expression by means of a sample. And such a sample, as Wittgenstein puts it, is a “means of representation” and a “part of language” (Wittgenstein [1953] 2009, §50). Fifth, Wittgenstein and Sellars share the view that necessary truths are, in one sense or another, expressions of linguistic rules. Sellars on one occasion says that necessary truths “correspond” to rules of language (SRLG §27: 330f.) and on another claims that the “language of modalities is . . . a transposed language of norms” (IM, 21). His idea is that an object-language statement, which is necessary, corresponds to a metalanguage statement specifying a rule for the use of words occurring in the object-language statement. For example, “Dogs are animals” is necessary because it corresponds to the metalanguage statement: “One may infer ‘X is an animal’ from ‘X is a dog’”. Similarly, Wittgenstein typically takes strong necessity as a sign that a statement expresses what he calls a “grammatical rule” or “norm of expression”. He writes for instance: “Whenever we say that something must be the case we are using a norm of expression” (Wittgenstein 2001, 16; cf. also Wittgenstein 1978, VI-8).8 Sixth, both Wittgenstein and Sellars frequently compare language to games and rules of language to the rules that constitute games. Hence, the rules of language that are given in specifications of meaning and expressed in necessary truths are for both constitutive of the meaning of expressions. Finally, they also share a view that seems to follow quite naturally from the preceding points: since treating language as meaningful implies treating its expressions as subject to certain rules and since necessary statements correspond to these rules, the meaningful use of language implies the acceptance of certain statements as necessary truths. Sellars expresses this idea rather pithily in the title of one of his earliest papers: “Concepts as Involving Laws and Inconceivable without Them”, and Wittgenstein writes about “the deep need for the convention” (Wittgenstein 1978, I-74). Wittgenstein’s idea here is parallel to Sellars’s: since the meaning of expressions is determined by grammatical rules—or conventions—the meaningful use of language requires the acceptance of certain grammatical rules and hence of the corresponding necessary truths.9

Sellars and Wittgenstein on Following a Rule

79

3 Sellars on Rule-Following All these parallels will almost certainly have to be qualified in numerous ways. Still, they constitute important similarities in Wittgenstein’s and Sellars’s thinking. Here, I wish to focus on my second point, the claim that meaning something by one’s use of words means following rules for their use. I believe that there are important parallels but also significant differences in how Sellars and Wittgenstein understand this claim. I will argue that the latter reveal different conceptions of the causal role reasons play in the explanation of action. I start with a regress-argument (which I shall call SR1) that opens Sellars’s SRLG and which at a first glance seems very Wittgensteinian. Sellars writes: (SR1) Thesis. Learning to use a language (L) is learning to obey the rules of L. But, a rule which enjoins the doing of an action (A) is a sentence in a language which contains an expression for A. Hence, a rule which enjoins the using of a linguistic expression (E) is a sentence in a language which contains an expression for E—in other words a sentence in a metalanguage. Consequently, learning to obey the rules for L presupposes the ability to use the metalanguage (ML) in which the rules for L are formulated. So that learning to use a language (L) presupposes having learned to use a metalanguage (ML). And by the same token, having learned to use ML presupposes having learned to use a meta-metalanguage (MML) and so on. But this is impossible (a vicious regress). Therefore, the thesis is absurd and must be rejected. (SRLG §2: 321) In formulating this regress Sellars assumes that learning to obey a rule of language presupposes the ability to use a metalanguage in which the rule can be formulated. He does not question this assumption. He even assumes that obeying a rule requires explicitly thinking of it, although he admits that this assumption is somewhat artificial. He writes: [It] should be emphasized that the phrase “rule obeying behavior” is not restricted in its application to behaviour in which one makes moves in a game via making moves in its rule metagame. There is a sense in which it is quite legitimate to say that Jones is obeying the rules of chess, even though he is not actually making moves in the rule language. . . . For there are many true subjunctive statements we could make about Jones and the rule language. . . . In this chapter,

80

Stefan Brandt however, I have limited my discussion of rule-obeying to the more pedestrian cases, oversimplifying in order to focus attention on the fundamentals. For a sensitive and illuminating account of the complex logical devices built into ordinary language about human behaviour, the reader is referred to Gilbert Ryle’s The Concept of Mind. (SRLG §65: 349–50)

Sellars is saying that there is a form of rule-obeying behavior which merely requires the ability to think of and express the rules one is following—that is, I take it, what the reference to subjunctive statements and to Ryle’s The Concept of Mind suggests. But, he claims, focusing “on the fundamentals” requires treating rule-obeying behavior as involving explicit thoughts of the rules one is following. This invites the question of why a focus “on the fundamentals” should require these specific limitations. The reason, I believe, lies in Sellars’s philosophy of action. He characterizes rule-obeying as action done “with the intention of fulfilling the demands of an envisaged system of rules” (SRLG §12: 325). And an intention is, according to him, a special kind of practical attitude, which issues in a volition once the time for action has come. A volition, in turn, is an occurrent thought to do something “here and now” that causes the behavior intended.10 Hence, rule-obeying requires for Sellars “making a move in the rule metagame” since it requires a volition “to fulfill the demands of an envisaged system of rules”. Still, Sellars thinks he can avoid the aforementioned regress argument. For according to him there is a way of following a rule of language, which does not consist in obeying it and, therefore, neither presupposes the ability to use a metalanguage nor its actual use in thought or speech. Thus, Sellars qualifies the initial thesis in a way that makes it false to say that following the rules of L “presupposes the ability to use the metalanguage (ML) in which the rules for L are formulated”. This kind of rulefollowing is exhibited in what Sellars calls “pattern-governed behavior”: The general concept of pattern-governed behavior is a familiar one. Roughly it is the concept of behaviour which exhibits a pattern, not because it is brought about by the intention that it exhibit this pattern, but because the propensity to emit behaviour of the pattern has been selectively reinforced, and the propensity to emit behaviour which does not conform to this pattern selectively extinguished. (MFC 86–7) According to Sellars, we become language users by being trained to exhibit pattern-governed behavior in conformity with the rules of our language. However, pattern-governed behavior is more than mere ruleconforming behavior; it is rule-governed behavior, for there is still a sense

Sellars and Wittgenstein on Following a Rule

81

in which we act the way we do because of the rules. And this is so because our language teachers intended us to comply with the rules of our language. But while this implies for Sellars that the language teachers explicitly thought of these rules,11 language learners can follow them without explicitly thinking of them or even being able to formulate them (cf. MFC 86–7). And since the regress, as Sellars states it, depends on the assumption that whenever we follow a rule of language we have to be able to formulate it, it is stopped. In publications succeeding SRLG Sellars explicates the distinction between pattern-governed behavior and rule-obeying behavior by suggesting that these two types of behavior are governed by two different types of rules.12 Obeying a rule is a matter of following what Sellars calls a “rule of action” or “ought-to-do” rule. In his “Language as Thought and as Communication” (LTC) he characterizes these rules as follows: If one is in [condition] C, one ought to do [action] A. (LTC 58) Following these rules requires significant conceptual capacities. Sellars writes: [The] “subjects” to which these rules apply must have the concepts of doing A and being in C. . . . Furthermore, for the rule itself to play a role in bringing about the conformity of “is” to “ought”, the agents in question must conceive of action A as what ought to be done in circumstances C. This requires that they have the concept of what it is for an action to be called for by a certain kind of circumstance. (LTC 59) Since following a rule of action requires conceiving of the action one is executing as being called for by the rule one is following and the circumstances one finds oneself in, it presupposes the application of a variety of concepts. But applying a concept is for Sellars also a matter of following rules for its employment. So, if all linguistic rules were rules of action, following them would require observing further rules for the use of the concepts presupposed by the former rules, and observing these further rules would require following still further rules, and so on ad infinitum. Hence, not all rules of language can be rules of action or ought-to-do rules. Pattern-governed behavior, on the other hand, is guided by rules, which can be followed by someone who does not have the concepts required for formulating them or conceptually identifying their conditions of application. It is governed by what Sellars calls “rules of criticism” or “ought-tobe rules”. They have the form: “Xs ought to be in state Φ, whenever such

82

Stefan Brandt

and such is the case” (LTC 59). Or, to put it more adequately for rules that guide an activity: It ought to be the case that Xs do A when in circumstances C. Someone can follow this rule, i.e., do A when in C, without possessing the concepts required for formulating this rule or identifying the circumstances as C. Still, pattern-governed linguistic behavior is properly characterized as rule-following behavior because it has been brought about by language trainers who reasoned from the fact that pattern-governed behavior of a certain kind ought to be exhibited by speakers, to the conclusion that they, the trainers, ought to do certain things to bring it about. Because trainers follow the ought-to-dos implied by the ought-to-bes governing ordinary language use, we can properly say that the latter occurs because of the ought-to-bes, where this “because” is both causal and rational (cf. MFC 87). Sellars’s solution to the regress argument can now be spelled out in terms of his distinction between ought-to-do and ought-to-be rules. Instead of the original thesis: learning to use a language (L) is learning to obey the rules of L, we can now say: learning to use a language (L) is learning to follow the ought-to-bes (rules of criticism) governing L. The regress is stopped because learning to follow an ought-to-be rule does not depend on the ability to formulate that rule, let alone a mental act of thinking of it. Still, Sellars insists, in order to become a full-fledged speaker, it is not sufficient to merely exhibit pattern-governed behavior; one also has to grasp the rules one is following. One has to be able to not merely follow ought-to-bes but also to grasp the corresponding ought-to-dos (cf. MFC 86–87). For the ability to think about and rationally change the rules of one’s language is for Sellars an integral part of the advance of science and knowledge. This does not mean, however, that full-fledged speakers generally obey rules. According to Sellars, “[the] linguistic activities which are perceptual takings, inferences and volitions never become obeyings of ought-todo rules” (MFC 88). As far as I can see there are three main reasons for this. First, we have already seen that if all language use were a matter of obeying rules of action, there would occur a regress about language use similar to the one about language learning that opens SRLG. Here it is again, now put in terms of obeying rules of language:

Sellars and Wittgenstein on Following a Rule

83

(SR2) Using a language L consists in obeying the rules of L. Obeying the rules of L requires being able to use (and as we have seen, according to Sellars, actually using) a metalanguage ML in which the rules of L are formulated. Using ML requires being able to use (and actually using) a meta-metalanguage MML in which the rules of ML are formulated and so on ad infinitum. Second, Sellars suggest on several occasions that a mistaken conception of rule-following leads to the Myth of the Given (cf. EPM §38: 169–70, LRB §24: 127, SRLG §34: 333–4). The idea seems to be the following: if in the application of observational predicates to our environment we obeyed ought-to-do rules, then we would have to recognize that the conditions for the application of these predicates obtained before we applied them. But this would require an awareness of one’s environment that is prior to the use of language. And that is for Sellars an instance of the Myth of the Given. Finally, as we have seen, Sellars believes that obeying a rule is an intentional action. For Sellars, however, language use serves as a model for thinking (cf., e.g., EPM §§56–9: 186–9). And it is extremely implausible, as Sellars stresses on numerous occasions, to conceive of occurrent thoughts as consisting of intentional actions (cf., e.g., EPM §58: 188, MFC, 84–5). He illustrates this with a nice example: looking into the next room is an intentional action, thinking there is a burglar in that room is not (MFC, 84).13 Mainly for these reasons Sellars believes that pattern-governed behavior is at the root of all language use. Whenever we obey a rule or follow an ought-to-do, we, as it were, have to make a transition from the rule to the action. This transition is itself rule-governed, but on pain of regress, it cannot itself be a case of obeying an ought-to-do. So it must be a bit of pattern-governed behavior, regulated by an ought-to-be or rule of criticism.

4 Wittgenstein on Following a Rule In the famous rule-following sections 143 to 242 of Philosophical Investigations Wittgenstein examines how a rule can determine which actions accord with it and which do not. This is part of a more general investigation into the question of how the correct use of linguistic expressions is determined. One natural suggestion is that a rule, or rather rule-formulation, only determines which actions accord with it if it is interpreted; and similarly that the correct use of a linguistic expression is determined by its interpretation. However, this view can lead to paradox, as Wittgenstein illustrates with a well-known example (cf. Wittgenstein 2009: §§143, 185).

84

Stefan Brandt

Imagine a pupil being taught to write down different series of natural numbers and being tested up to 1,000. After some training, he is asked to write down the series of even numbers. He proceeds correctly until he reaches 1,000 but then continues “1,004, 1,008, 1,012” etc. When asked to continue with the next number but one, he insists that that is what he is doing. It seems that regardless of what one might tell the pupil, i.e., what interpretation of the original order one might give him, he could agree on the interpretation but still continue the series in an abnormal way. Furthermore, whatever he does can be interpreted as being in accord with the rule. After all, every application of the rule “add 2” is in some sense new—either because a new number is being reached or a sequence of the series of even integers is reproduced on a new occasion. So, the pupil can always provide an interpretation of the rule justifying his actual application. Ultimately, this seems to mean that an interpretation cannot determine how a rule is to be followed. And since a rule is essentially the kind of thing that tells us how it is to be followed, this means that an interpretation cannot give a rule any determinate content. This, Wittgenstein suggests, seems paradoxical: This was our paradox: no course of action could be determined by a rule, because every course of action can be brought into accord with the rule. The answer was: if every course of action can be brought into accord with the rule, then it can also be brought into conflict with it. And so there would be neither accord nor conflict here. That there is a misunderstanding here is shown by the mere fact that in this chain of reasoning we place one interpretation behind another, as if each one contented us at least for a moment, until we thought of yet another lying behind it. For what we thereby show is that there is a way of grasping a rule which is not an interpretation, but which, from case to case of application, is exhibited in what we call “following the rule” and “going against it”. That’s why there is an inclination to say: every action according to a rule is an interpretation. But one should speak of interpretation only when one expression of a rule is substituted for another. That’s why “following a rule” is a practice. And to think one is following a rule is not to follow a rule. (Wittgenstein 2009, §§201–2) Wittgenstein is saying at least four things. First, the idea that a rule determines its applications through an interpretation is a mistake or rests on a misunderstanding, since by means of an interpretation any action can be made to accord with a rule. That is what his thought-experiment of the deviant pupil shows. Second, consequently there must be a way of understanding a rule that is not an interpretation but that shows itself in

Sellars and Wittgenstein on Following a Rule

85

the kind of actions that we call “following a rule”. Third, interpreting a rule is best understood as replacing one formulation of a rule by another one. Fourth, he concludes from all this that following a rule is a practice and that, hence, thinking one is following a rule is different from following a rule. It is tempting to read Sellars as providing an elaboration of the claim “that there is a way of grasping a rule which is not an interpretation”, which simultaneously explains why following a rule is a practice. Equipped with Sellars’s distinction between rule-obeying and pattern-governed behavior we might say that grasping a rule in a way that is not interpretation is a matter of exhibiting a piece of pattern-governed behavior. When we make a transition from, say uttering a rule or thinking of a rule, to an action that accords with it, we exhibit a piece of behavior that we have been trained to exhibit and that is governed by an ought-to-be. And since pattern-governed behavior is a kind of practice, this also shows why rulefollowing is a practice. This is, for instance, roughly the way Robert Brandom reads Wittgenstein (cf. Brandom 1994, 18–26). According to Brandom, both Wittgenstein’s paradox and Sellars’s argument SR1 point to a need for what he calls “norms implicit in practice” and which he contrasts with “norms as explicit rules”. And although Brandom does not explicitly employ Sellars’s terminology, I take it that Sellars’s distinction between ought-to-bes and ought-to-dos and between pattern-governed and rule-obeying behavior is one way of spelling out his distinction.14 Furthermore, this interpretation fits well with several remarks Wittgenstein makes about the training that is involved in learning to follow a rule. He writes, for instance: “So is whatever I do compatible with the rule?”—Let me ask this: what has the expression of a rule—say a signpost—got to do with my actions? What sort of connection obtains here?—Well, this one, for example: I have been trained to react in a particular way to this sign, and now I do so react to it. (Wittgenstein 2009, §198) This remarks seems very similar to Sellars’s claim that in order to avoid the regress generated by the idea that all rule-following is matter of obeying ought-to-dos we have to assume a level of pattern-governed behavior at which speakers follow ought-to-bes. However, I believe these parallels between Wittgenstein and Sellars are deceptive. Apart from the somewhat odd formulation “that there is a way of grasping a rule which is not an interpretation, but which, from case to case of application, is exhibited in what we call ‘following the rule’ and ‘going against it’” there is no indication in the Investigations that Wittgenstein believes there to be two different ways of following

86

Stefan Brandt

a rule: obeying it and exhibiting pattern-governed behavior. Nor does he assume two different types of rules corresponding to these two types of behavior. Hence, I believe we should rather read Wittgenstein as claiming in §201 that understanding a rule is a capacity that can be exercised in interpreting it and in following it, but that the ultimate criterion for whether someone has understood a rule—or any linguistic expression for that matter—is what he does, not what he thinks (cf. also Wittgenstein 2009, §146). After all, a person’s actions are the kind of things that comply with or go against a rule. Furthermore, this reading fits much better with the general purport of Wittgenstein’s remarks on understanding (roughly §143 to §197), which argue that understanding is a kind of ability (cf. §150) and that the only criterion for whether someone has understood something are his actions not what kind of mental state he is in (cf., e.g., §§149, 157) or what he is consciously thinking of (cf. §148).15 Moreover, the invocation of practice by Wittgenstein plays a different role in his discussion of rule-following than Sellars’s invocation of pattern-governed behavior. This is related to the fact that Wittgenstein and Sellars are primarily concerned with two different questions. Sellars’s primary question seems to be: How can I act because of a certain rule despite the fact that this rule does not explicitly figure in my thinking? In SRLG he introduces the notion of pattern-governed behavior in order to show how there can be a “necessary” but by the rule-follower himself “unintended relation” between a rule and the corresponding conduct (cf. SRLG §12f, 325ff.). Wittgenstein, on the other hand, is principally concerned with the question of how a rule can determine its applications (cf., e.g., Wittgenstein 2009, §186) or—more generally—how the correct use of linguistic signs is determined. Wittgenstein’s point in invoking the notion of practice in response to this latter question is not to say that in order for something like an ought-to-do rule or, more generally, a rule I explicitly think of, to guide my action there has to be some further ought-to-be rule or “norm implicit in practice” to connect the first rule to my action. He rather makes the much more minimal point that in order for a rule, or rather an expression of a rule, to determine any action at all there has to be a practice of using the expressions in which the rule is formulated. He writes, for example: It is not possible that there should have been only one occasion on which only one person followed a rule. It is not possible that there should have been only one occasion on which a report was made, an order given or understood, and so on.—To follow a rule, to make a report, to give an order, to play a game of chess, are customs (usages, institutions). (Wittgenstein 2009, §199)

Sellars and Wittgenstein on Following a Rule

87

The point is typically Wittgensteinian. He does not speak about practices, customs, usages and institutions to explain anything, such as how a “norm explicit as rule” relates to its application through a “norm implicit in practice”. He rather reminds his readers of something that should be obvious: Expressions for rules only have a determinate content if they are part of an established linguistic practice. Sellars’s distinctions between rule-obeying and pattern-governed behavior and between rules of action and rules of criticism have no counterpart in Wittgenstein’s discussion of his rule-following paradox. Does Sellars’s question, how we can follow a rule that does not explicitly figure in our thinking, play any role in the Investigations? It does. In the following passage Wittgenstein addresses Sellars’s worry: Just think of the kinds of case where we say that a game is played according to a particular rule. The rule may be an aid in teaching the game. The learner is told it and given practice in applying it. Or it is a tool of the game itself. Or a rule is employed neither in the teaching nor in the game itself; nor is it set down in a list of rules. One learns the game by watching how others play it. But we say that it is played according to such-and-such rules because an observer can read these rules off from the way the game is played like a natural law governing the play.—But how does the observer distinguish in this case between players’ mistakes and correct play? There are characteristic signs of it in the players’ behaviour. Think of the behaviour characteristic of someone correcting a slip of the tongue. It would be possible to recognize that someone was doing so even without knowing his language. (Wittgenstein 2009, §54) Wittgenstein here distinguishes three kinds of cases in which we would say that a game is played according to a particular rule. First, the rule may be used in teaching the game; in the preceding section Wittgenstein gives as an example the use of color charts in teaching color words. Second, the rule may be a “tool of the game itself”. Again we might think of the use of color charts in selecting, say, a particular shade of red. Finally, we might recognize certain reasonably complex forms of behavior as rule-governed, although no rule is explicitly employed. In this case the difference between rule-governed behavior and mere regularities of conduct lies in the presence of behavior typical for correcting a mistake. The first kind of case is similar to the teaching Sellars thinks is necessary for linguistic rule-following. But the other examples make clear that Wittgenstein does not think that being taught to follow certain rules of language by people who have those rules explicitly in mind is a necessary condition for following a rule. Nor does he suggest that there are different kinds of rules corresponding to different ways of following them. All

88

Stefan Brandt

he suggests in the preceding remark is that someone who follows a rule should have some conception of what counts as a correct application of the rule and what does not. But such a conception must, of course, not be actually exercised, although its exercise is the only criterion for whether someone has it.

5 The Source of the Difference Between Wittgenstein and Sellars Did Wittgenstein overlook the need for a distinction that Sellars has made explicit? Do we have to distinguish between something like patterngoverned and rule-obeying behavior and between rules of action and rules of criticism in order to make linguistic rule-following intelligible? I do not think so. It is rather that Sellars makes these distinctions because he has specific commitments in the theory of action. He believes that when we act with the intention of realizing a specific aim, in the case we are interested in, to act in accordance with a specific rule of language, this aim must be part of the content of a mental event causing the action. However, this mental event is itself a “position in the game of reasoning” (cf. SRLG §10, 324) and, hence, an act that is governed by rules. On pain of regress this act cannot itself be an action of intentionally obeying a rule, which is caused by a further mental event whose content specifies a further set of rules. Hence, there must be a way of following rules, which does not require an intention to follow them. However, even in these cases Sellars wants to hold on to two fundamental commitments. First, a performance will only count as being guided by some rule R, rather than as merely being in accord with it, if R plays some role in its causal etiology (cf., e.g., SRLG §§12ff.: 325ff.). Second, this causal role must ultimately be spelled out in terms of some mental event with R as its content. Because for Sellars the only way in which an abstract entity, such as a rule, can have a causal impact is by occurring as the content of a mental event (cf., e.g., AAE §§35–52: 200–6). The purpose of Sellars’s notions of ought-to-be rules and pattern-governed behavior is to show how these two fundamental commitments can be honored while at the same time allowing for a form of rule-following that does not consist in acting with the explicit intention of realizing the rule one is following. However, as Sellars himself admits when he refers to Ryle’s The Concept of Mind and the possibility of explaining obeying a rule in terms of the abilities and dispositions of the rule-follower, there is, at least prima facie, a way of explaining acting on a rule that does not require reference to mental events. Sellars dismisses the option as not “fundamental” because he is ultimately committed to an event causal conception of acting for a reason. Wittgenstein, on the other hand, is simply not committed to this conception. More specifically, he believes a speaker already counts

Sellars and Wittgenstein on Following a Rule

89

as following a specific rule of language if he has certain specific abilities and dispositions. He writes, for example: Is it correct for someone to say: “When I gave you this rule, I meant that in this case you should . . . ”? Even if he did not think of this case at all as he gave the rule? Of course it is correct. For “to mean it” just did not mean: to think of it. But now the question is: How are we to judge whether someone meant such-and-such?—That he has, for example, mastered a particular technique in arithmetic and algebra, and taught someone else the expansion of a series in the usual way, is such a criterion. (Wittgenstein 2009, §692) What we mean by a word, what rules for its use we follow, is not determined by what goes through our or our teacher’s minds while using it, but by the kinds of dispositions we have with regard to its use, i.e., what techniques for its use we have mastered. It is one of the major themes running through the Investigations that the occurrent mental goings-on that accompany language use are irrelevant to determining meaning. Now, one might object that Wittgenstein still needs something like Sellars’s distinction between rule-obeying and pattern-governed behavior or between ought-to-dos and ought-to-bes. For, according to Wittgenstein, following rules of language requires some conception of how the words one uses are correctly used. And this conception requires that one is able to use a metalanguage, although it might not require that one actually use it. But once we admit this, we are off on something like Sellars’s regresses SR1 and SR2: in order to follow a rule of language we have to be able to speak a metalanguage, and speaking a metalanguage requires being able to speak a meta-metalanguage, and so on ad infinitum. In conclusion I would like to make three remarks regarding this argument. First, this cannot be exactly the argument Sellars wants to give. For he, too, thinks that a fully competent speaker must be able to specify the rules of his language (cf. MFC 86). So the distinction between ought-tobes and ought-to-dos and between rule-obeying and pattern-governed behavior cannot have been introduced to solve this last regress. Second, there is a sense in which we do have at least the resources to speak infinitely many metalanguages once we speak one. Once we have acquired the ability to use quotes and to refer to words and sentences, we have the resources to construct infinitely many metalanguages simply by the reiterated use of quotation; although there is, obviously, a practical limit to the number of metalanguages we could even potentially construct. At some point we will inevitably get confused or run out of time. Finally, and most importantly, the fact that we need some conception of the correct use of the words we are employing does not imply that we have to be able

90

Stefan Brandt

to give an endless series of justifications for our use of a word. At some point, as Wittgenstein suggests, justifications give out: “How am I able to follow a rule?”—If this is not a question about causes, then it is about the justification for my acting in this way in complying with the rule. Once I have exhausted the justifications, I have reached bedrock, and my spade is turned. Then I am inclined to say: “This is simply what I do”. (Wittgenstein 2009, §217) Wittgenstein appears to say that it is not a problem if, at some point, we cannot go on justifying our use of words. As he puts it a little later in the Investigations: “To use a word without a justification does not mean to use it wrongfully” (Wittgenstein 2009, §289). Ultimately, the regress SR1, with which Sellars opens SRLG, is misleadingly framed. There is no real problem with claiming that following rules of language requires being able to use a metalanguage in which these rules are formulated. What makes the regress truly vicious is the assumption that obeying a rule of language requires an actual act of thinking of these rules. It is this assumption that forces Sellars to introduce a new type of rule-following conduct, i.e., pattern-governed behavior, and a corresponding new type of rule, i.e., ought-to-dos or rules of criticism. But since Wittgenstein does not share the initial assumption he has no use for Sellars’s distinctions. However, if one shares Sellars’s event-causal conception of acting for a reason, Sellars’s distinctions remain useful.16,17

Notes 1. See, for example, NS, TC, SM V, BBK. 2. For explicit references, see, for example, LTC 64, MFC 88, EPM §59: 189, SM I §39: 16. Sellars also occasionally alludes to Wittgenstein without mentioning his name; for example, when he writes in EPM about picturing a “child—or a carrier of slabs—learning his first language” (EPM §30: 161, emphasis added) and, a little later, mentions “Augustinian” theories of language (EPM §30: 162). 3. I believe Sellars is particularly thinking of Peter Strawson here, whom he tends to treat as the main contemporaneous philosopher of the manifest image (see, e.g., SM I §43: 17; III §27: 72; III §71: 88; VI §55: 171). 4. Still, the paper has been reprinted in the sixth volume of John Canfield’s 15-volume anthology of papers on Wittgenstein (cf. Canfield 1986, 2–26). 5. Sellars is referring to footnote 5 (LRB §22: 126). 6. I am grateful to Lionel Shapiro for making me aware of this remark. 7. See, e.g., Sellars’s SRLG and MFC and Wittgenstein ([1953] 2009), especially §§43, 199. It should be noted that Sellars accuses Wittgenstein of being careless with the semantic-pragmatic distinction in his characterization of meaning as use. He writes: “The Verbal Behaviorist agrees with Wittgenstein that the meaning of an utterance is its use, but, unlike Wittgenstein, he is careful

Sellars and Wittgenstein on Following a Rule

8.

9. 10. 11.

12.

13. 14. 15. 16. 17.

91

to distinguish ‘use’ in the instrumental sense (in which utterances are construed as Austinian performances) from use as function” (AAE §27: 197–8). Sellars’s point here is that the meaning of a linguistic expression is determined by its role within a linguistic practice and not by the communicative intentions with which it is used in speech acts and that Wittgenstein did not pay sufficient heed to this distinction. I think this criticism of Wittgenstein is misplaced. In Philosophical Grammar, for instance, Wittgenstein explicitly writes that the purposes, with which words are used, do not play a role in the explanation of their meaning: “[E]xplaining the purpose or effect is not what we call explaining the meaning” (Wittgenstein 1974, 69). Sellars’s remark, therefore, does not register a genuine disagreement but is merely a sign of misunderstanding. It should be noted, though, that Wittgenstein only takes necessity that is stronger than causal necessity (“harder than the causal must” (Wittgenstein 1978, I-121)) as indicative of grammatical statements (cf. ibid. I-118 to I-122; Wittgenstein 2001, 16). Sellars, on the other hand, also thinks of natural laws as expressions of linguistic rules (cf., e.g., SRLG §§75ff.: 352ff.). For this interpretation of Wittgenstein, see also Glock (1996), 217–24. For a very similar interpretation of Sellars’s views, see my Brandt (2017). Cf., e.g., TA, 105–10. Presumably it suffices that our language trainers have some general intention to the effect that we follow the rules of the language we are being taught. It is simply too implausible to assume that for every rule we are following either we or our trainers must have had the explicit intention that we follow this rule. As far as I know these two types of rules are first introduced in LTC, which was first published in 1969, 15 years after the first publication of SRLG and six years after the publication of a revised version of the latter in SPR in 1963. For this last point, see also the contribution of Johannes Hübner to this volume. For a, to my mind apposite, criticism of Brandom’s reading of Wittgenstein, see McDowell (2002). I defend this reading of Wittgenstein’s remarks on understanding in Brandt (2014); see also Baker and Hacker (2005), ch. XVII. For an illustration of this usefulness, see, for instance, the contribution by Johannes Hübner in this volume. I would like to thank Anke Breunig and Erasmus Mayr as well as the participants of a conference on Sellars at the University of Erlangen-Nürnberg in the spring of 2015 for helpful comments on earlier versions of this chapter.

References Baker, G.P., and P.M.S. Hacker. 2005. Wittgenstein: Understanding and Meaning— Part I: Essays, 2nd edition. Oxford: Blackwell. Brandom, Robert. 1994. Making It Explicit. Cambridge, MA: Harvard University Press. Brandt, Stefan. 2014. “How Not to Read Philosophical Investigations: McDowell and Goldfarb on Wittgenstein on Understanding”. Philosophical Investigations 37 (4): 289–311. ———. 2017. “Sellars and Quine on Empiricism and Conceptual Truth”. British Journal for the History of Philosophy 26 (1): 108–32.

92

Stefan Brandt

Canfield, John V. 1986. The Philosophy of Wittgenstein, Vol. VI: Meaning. New York: Garland. Glock, Hans-Johann. 1996. “Necessity and Normativity”. In The Cambridge Companion to Wittgenstein, edited by Hans Sluga and David. G. Stern, 198– 225. New York: Cambridge University Press. McDowell, John. 2002. “How Not to Read Philosophical Investigations: Brandom’s Wittgenstein”. In Wittgenstein and the Future of Philosophy: A Reassessment after Fifty Years, edited by R. Haller and K. Puhl, 245–56. Vienna: Hölder-Pichler-Tempsky. Sellars, Wilfrid. Wilfrid S. Sellars Papers, 1899–1990, ASP.1991.01, Archives of Scientific Philosophy, Special Collections Department, University of Pittsburgh. URL: https://digital.library.pitt.edu/collection/wilfrid-s-sellars-papers. ———. (1949) 2005 [LRB]. “Language, Rules and Behavior”. In Pure Pragmatics and Possible Worlds: The Early Essays of Wilfrid Sellars, edited by Jeffrey F. Sicha, 186–214. Atascadero, CA: Ridgeview Publishing. ———. (1953) 2007 [IM]. “Inference and Meaning”. In Wilfrid Sellars. 2007. In the Space of Reasons: Selected Essays of Wilfrid Sellars, edited by Kevin Scharp and Robert B. Brandom, 3–27. Cambridge, MA: Harvard University Press. Hereafter SR. ———. (1954) 1991 [SRLG]. “Some Reflections on Language Games”. In SPR, 321–58. ———. (1956) 1991 [EPM]. “Empiricism and the Philosophy of Mind”. In SPR, 127–196. ———. (1960) 1991 [BBK]. “Being and Being Known”. In SPR, 41–59. ———. (1962) 1991 [NS]. “Naming and Saying”. In SPR, 225–46. ———. (1962) 1991 [PSIM]. “Philosophy and the Scientific Image of Man”. In SPR, 1–40. ———. (1962) 1991 [TC]. “Truth and ‘Correspondence’”. In SPR, 197–224. ———. (1963) 1991 [SPR]. Science, Perception and Reality. Atascadero, CA: Ridgeview Publishing. ———. 1966 [TA]. “Thought and Action”. In Freedom and Determinism, edited by Keith Lehrer, 105–40. New York: Random House. ———. (1968) 1992 [SM]. Science and Metaphysics: Variations on Kantian Themes. Atascadero, CA: Ridgeview Publishing. ———. (1969) 2007 [LTC]. “Language as Thought and as Communication”. In SR, 57–80. ———. (1973) 1974 [AAE]. “Actions and Events.” In Wilfrid Sellars. 1974. Essays in Philosophy and Its History, 189–213. Dordrecht, Holland: D. Reidel. Hereafter EPH. ———. (1974) 2007 [MFC]. “Meaning as Functional Classification”. In SR, 81–100. Wittgenstein, Ludwig. (1953) 2009. Philosophical Investigations. Oxford: Blackwell. ———. 1958. The Blue and Brown Books. Oxford: Blackwell. ———. 1974. Philosophical Grammar. Oxford: Blackwell. ———. 1978. Remarks on the Foundations of Mathematics. Oxford: Blackwell. ———. 2001. Wittgenstein’s Lectures: Cambridge, 1932–1935, edited by Alice Ambrose. New York: Prometheus Books.

Part II

Sellars and the Analytic Tradition

5

Wilfrid Sellars as an Analytic Philosopher Tadeusz Szubka

Wilfrid Sellars was a highly original systematic philosopher of the past century, and for this reason alone has a unique place in contemporary philosophy. In his philosophy he drew on various historical ideas and did not confine himself to a particular movement or approach. However, the most natural setting for his views is the analytic tradition, which provides an illuminating framework for understanding those views and seeing them in proper perspective. In what follows, I shall make an attempt to justify this approach by putting together various details from Wilfrid Sellars’s life that provide a glimpse into his exposure to different facets of analytic philosophy and his role in shaping it.1 Subsequently, I shall describe the Sellarsian conception of philosophy and the sense in which it embodies characteristically analytical ideas.

Some Biographical Episodes In his autobiographical reminiscences Sellars claims that he started serious work in philosophy while taking a course on British empiricists by Cooper Harold Langford at the University of Michigan in Ann Arbor. Langford was a psychologist by education and an accomplished mathematical logician who was also well acquainted with Cambridge analytic philosophy (he spent a considerable amount of time at the University of Cambridge on long-term fellowships). Thus his course was, as Sellars remembers, “at least as much on G.E. Moore and Cambridge Analysis as it was on the Empiricists” (AR 281). Sellars was impressed by the rigorous analytical way of doing philosophy and by the use of logical systems and techniques for various philosophical purposes, while at the same being well aware that this methodology had severe limitations and had to be supplemented by more traditional means. His metaphilosophical convictions along these lines were not weakened but rather strengthened during his time at the University of Oxford as a Rhodes scholar. He expresses his attitude in this matter by making a comparison between G.E. Moore and the more traditionally oriented C.D. Broad: My experience with Moore . . . had convinced me that clarity and distinctness can be achieved at the expense, to borrow a Spinozistic

96 Tadeusz Szubka term, of adequacy, and that one should give one’s muddiest intuitions the fullest benefits of the doubt. I had long felt that, although C.D. Broad might not be clearer than Moore, nevertheless he had a more adequate grasp of the problems they shared. I now think this can be traced to Broad’s awareness of, and technical competence in, the scientific background of these problems. (AR 284) If one takes these words at face value, Sellars came very early to the conclusion that philosophical analysis must be supplemented by a systematic and synoptic vision, but as we shall see later, there is no evidence of this emphasis on system and synopsis in Sellars’s early writings. Sellars was also heavily influenced by the strong realist tendencies of the analytic movement, not only through its Cambridge founders G.E. Moore and B. Russell, but perhaps even more by Oxford Realists H.A. Prichard and J. Cook Wilson, since—as he claims—one may find there “a clearly articulated approach to philosophical issues which undercut the dialectic, rooted in Descartes, which led to both Hume and 19th Century Idealism” (AR 284). Sellars’s analytic realism was reinforced during his second stay in Oxford, undertaken with the aim (ultimately unattained) of completing a doctoral dissertation on Kant. During that period he was attending a seminar on C.I. Lewis’s theory of knowledge, run by J.L. Austin together with I. Berlin. Here is Berlin’s vivid, if perhaps somewhat embellished, description of the seminar: I may be mistaken about this, but I think that this was the first class or seminar on a contemporary thinker ever held in Oxford. Austin’s reputation as a teacher was by this time considerable, and a relatively large number of undergraduates came once a week to our class in All Souls. I had no notion how joint classes were held, and assumed that their holders would begin by a dialogue on points provided by the text, in which they would show each other the almost exaggerated respect which was then common form at philosophical debates among dons. Austin opened by inviting me to expound a thesis. I selected Lewis’s doctrine of a specific, sensible characteristics— what Lewis called qualia, and said what I thought. Austin glared at me sternly and said, “Would you mind saying that again”. I did so. “It seems to me”, said Austin, speaking slowly, “that what you have just said is complete nonsense”. I then realized that this was to be no polite shadow-fencing, but war to the death—my death, that is. There is no doubt that Austin’s performance at our class had a profound and lasting effect upon some, at any rate, of those who attended it. Some of them later became eminent professional philosophers and have testified to the extraordinary force and fertility of Austin’s performance. (Berlin 1973, 7–8)

Wilfrid Sellars as an Analytic Philosopher

97

Among those who “later became eminent professional philosophers” was Wilfrid Sellars, and certainly his anti-phenomenalism and realist stance in epistemology was bolstered by Austin. After his return from Oxford to America, Sellars took some philosophical courses at Harvard University and had a unique occasion to encounter logical positivism through the lenses of W.V. Quine. In general he was suspicious of the anti-metaphysical rhetoric of that movement and thought it led nowhere. Several years later, while reviewing Elements of Analytic Philosophy by Arthur Pap, a book that had inherited a lot of that naïve and somewhat irritating anti-metaphysical rhetoric, Sellars criticized Pap for inventing a series of straw men, described as metaphysicians, supposedly guilty of various blunders and absurdities. Sellars found this strategy misleading to the extreme, since it kept readers from realizing that “most of the revisions of Analytic Philosophy in the past two decades have been in the direction of re-absorbing insights which were cast aside as metaphysical in the salad days of positivism” (RP 105). The course Quine taught on logical positivism, which Sellars audited in the late 1930s, was mainly focused on Rudolf Carnap’s substantial ideas from Der logische Aufbau der Welt and Logische Syntax der Sprache. If at that time Sellars did not find Carnap’s ideas too appealing, this was because of Sellars’s anti-phenomenalism and his belief that “while a rigorous account of syntax was clearly a desideratum, as far as its philosophical content was concerned, Carnap was putting the cart before the horse” (AR 287–88). And though with the passing of time Sellars would take a more favorable stance towards Carnap’s ideas, his overall opinion of logical positivism as an essentially anti-metaphysical movement, or—to be more precise—a movement endorsing a bad metaphysics in disguise, remained rather low. While discussing Carnap’s views he was disparagingly referring to “the Procrustean convictions of early logical positivism, when the modalities were in eclipse, and emotivism rampant” (EAE 285). And looking back in the 1970s on his philosophical development, Sellars strongly emphasized this critical tendency: “After striking out on my own, I spent my early years fighting in the war against Positivism—the last of the great metaphysical systems; always a realist, flirting with Oxford Aristotelianism, Platonism, Intuitionism, but somehow convinced, at the back of my mind, that something very much like Critical Realism and Evolutionary Naturalism was true” (NAO 9). Nevertheless, this assessment of logical positivism did not prevent Sellars from establishing a close collaboration early in his academic career with one of its enthusiastic practitioners, namely Herbert Feigl. Together they embarked on the arduous task of editing a collection of readings in analytic philosophy. Feigl and Sellars did not rely merely on their own judgments in choosing material for the anthology, but discussed their selections with numerous teachers of philosophy in the United States and England.2 The crucial criteria for making their final choice were related primarily to the “didactic effectiveness” of the selected pieces, and only

98 Tadeusz Szubka secondarily to their authorship. That is to say, they conceived of their collection as a selection of exemplary and interconnected essays in philosophical analysis, and not as a loose bundle of pieces by famous thinkers. Feigl and Sellars thought of analytic philosophy as consisting mainly in the Cambridge movement initiated by Moore and Russell and the logical positivism of the Vienna Circle and the Berlin Group. In the preface to the collection, published in 1949 by Appleton-Century-Crofts as Readings in Philosophical Analysis, they declared that these two movements “together with related developments in America stemming from Realism and Pragmatism, and the relatively independent contributions of the Polish logicians have increasingly merged to create an approach to philosophical problems which we frankly consider a decisive turn in the history of philosophy” (Feigl and Sellars 1949, vi). It is wiser to leave posterity to decide whether analytic philosophy has indeed been “a decisive turn in the history of philosophy”. However, given its development in the second half of the twentieth century, there are good reasons to claim that the anthology edited by Feigl and Sellars made a very important contribution to the movement’s consolidation and to establishing its canon. It was “for decades a staple in the classrooms of analytic philosophers” (Soames 2008, 455).3 Encouraged by the wide and frequent use of their anthology, Feigl and Sellars, with substantial and extensive cooperation from Keith Lehrer,4 edited and published over two decades later New Readings in Philosophical Analysis, conceived as a sequel to the original collection (Feigl, Sellars and Lehrer 1972). It was certainly a useful volume, but by that time it was one that had to compete with other similar publications. There was no overlap between these two anthologies, so the latter could not be taken (contrary to Beaney 2013, 101) as the second edition of the former. The editors claimed that their intention was “that the two volumes should be complementary, so that the reading of one will enhance one’s appreciation of the other” (Feigl, Sellars and Lehrer 1972, vii). The first anthology includes a revised version of Sellars’s early paper “Realism and the New Way of Words”, while the second one reprints three essays from his middle period: “Truth and ‘Correspondence’”, “Notes on Intentionality”, and “Some Reflections on Thoughts and Things”. Feigl and Sellars also launched a new journal entitled Philosophical Studies, dedicated to philosophy in the analytic tradition. In a letter to the publisher of Readings in Philosophical Analysis, dated 22 September 1949, Sellars wrote: “As you see from the letterhead, Feigl and I have talked the University [of Minnesota] Press into backing a new journal. It will carry forward the work in Philosophical Analysis of which our Readings in Philosophical Analysis is a record” (Wilfrid S. Sellars Papers, University of Pittsburgh, box 158, folder 3). The journal was unambiguously modeled on the British periodical Analysis and, like the latter, originally preferred short pieces and discussion notes. By the end of its first year, Feigl and Sellars would announce that the “size and rapid growth

Wilfrid Sellars as an Analytic Philosopher

99

of our circulation, as well as the number of manuscripts which are being submitted, prove that our efforts have met with success and aroused the good wishes and cooperation of analytic philosophers not only in this country but also in many other parts of the globe” (The Editors 1950, 94). Feigl and Sellars successfully edited Philosophical Studies for a good number of years before handing it over to philosophers of a younger generation. Feigl was the first to withdraw from the editorship of the journal when in 1972 it was greatly expanded and taken over by D. Reidel Publishing Company. Sellars continued to edit Philosophical Studies with the assistance of a new editorial board for a short time, but in 1974 handed over this burden to Keith Lehrer.5 In 1947 a steady succession of Sellars’s own publications had begun, a process which would continue even well after his death in 1989. These show him in an ongoing dialogue both with the great thinkers of the distant past and with his contemporaries, especially in the analytic tradition (including B. Russell, L. Wittgenstein, W.V. Quine, P.F. Strawson, and many others). However, in the course of this dialogue, and while developing his own highly original ideas, Sellars never fully elaborated his own conception of philosophy. This does not mean that he despised metaphilosophical considerations. Indeed, in the opening section of his essay on Everett Hall’s philosophy of mind, Sellars claimed that any fully articulated philosophy should be self-conscious and self-referential, that is, it should include as an important part “a theory of the philosophical enterprise, a theory which faces up to the ultimate challenge which any systematic philosophy must face: What is the status of your philosophical claims, and what are the criteria by which you distinguish them as true from the false and unacceptable claims made by rival philosophies?” (IRH 29). It also does not mean that Sellars lacked suitable promptings and opportunities to expound his metaphilosophy. Quite the contrary! When Richard Rorty was working on a collection of essays in philosophical methodology, eventually published in 1967 as The Linguistic Turn (Rorty [1967] 1992), he tried to encourage Sellars to write a special piece for it. In a letter dated 24 June 1965, Rorty wrote: There was a matter which I vaguely remember discussing with you once, but which (if my memory is correct) we left up in the air. Let me now raise it again. Would you have any interest in writing an article on metaphilosophy which I could include in an anthology of articles thereon which I’m doing for the U[niversity] of Chicago Press? . . . I’m not asking anybody else to write something for the anthology, since there’s an embarrass de richesse in print. But I would very much like to have something of yours in it, while nothing you’ve so far written seems suitable. If, as you were meditating, you do your Locke Lectures on what’s living and what’s dead in ordinary-language analysis, then one of those lectures might be just what I need. . . . Possibly

100 Tadeusz Szubka you regard metaphilosophy as an unmanageable subject, but, as I say, I’d be delighted if you cared to air your views. (Wilfrid S. Sellars Papers, University of Pittsburgh, box 161, folder 17) Sellars’s reply to this invitation has not been so far located, and it is presumed to be lost (Gross 2008, 180). Thus we do not know whether he accepted the invitation and attempted to write the required piece. Of course, we know that in the end he did not write it, and we know for sure that his Locke Lectures were not on ordinary-language analysis, but were a sustained attempt to give an overall picture of Sellars’s substantial philosophical views, presented as variations on Kantian themes. For this reason if one wants to give an account of Sellarsian metaphilosophy, one has no choice but to comment upon various scattered statements and programmatic remarks made in passing, very often in the introductory sections of his lectures and essays.

Sellarsian Metaphilosophy A careful reader of Sellars’s early writings very quickly realizes that he presented himself there as a paradigmatically analytic philosopher of his times, who conceived this discipline as a theoretical activity focused mostly on language, where questions, arguments, and arguing matter more than results. While Sellars was composing and publishing his early papers, analytic philosophy was in the second phase of its development, which lasted, roughly speaking, from the late 1920s till the end of the 1950s. The metaphilosophy of this period was dominated mainly by (i) the Wittgensteinian idea of philosophy as the activity of clarifying our thoughts expressed in language and dissolving or dismantling various puzzles of traditional metaphysics or epistemology, and by (ii) the doctrine of logical positivism that the only legitimate form of philosophy is logical and metalinguistic analysis of linguistic aspects of our knowledge and science. Although Wittgensteinians and logical positivists agreed that philosophy is a rational activity and consists in formulating arguments, it leads us only to a better understanding and clarification of our conceptual structures and languages, and not the discovery of new facts about the world. In his first two published papers, Sellars claims: The analytic movement in philosophy has gradually moved towards the conclusion that the defining characteristic of philosophical concepts is that they are formal concepts relating to the formation and transformation rules of symbol structures called languages. Philosophy, in other words, tends to be conceived of as the formal theory of languages. (PPE 4)

Wilfrid Sellars as an Analytic Philosopher

101

Philosophy is pure formalism; pure theory of language. The recommendation of formalisms for their utility is not philosophy. (PPE 103) I shall argue that philosophy is properly conceived as the pure theory of empirically meaningful languages, and that pure semantics, as it now exists, is but a fragment of such a theory. (ENWW 28) At that time Sellars was not only an ardent advocate of the linguistic turn taken by analytic philosophy, but took great pains to show that philosophers, unlike scientists, are not interested in factual matters and do not proceed inductively in their investigations. In the course of their philosophical inquiries they should deal with no other symbolic or linguistic forms “than those analyzed in pure semantics and perhaps . . . pure pragmatics” (RC 328). For Sellars at that time the main enemy of philosophy conceived as a pure or formal theory of language was psychologism, defined as “the mistake of identifying philosophical categories with those of psychology, whether introspective or behavioristic” (PPE 4). Psychologism so construed was a common form of a more general error which may be described as metaphilosophical factualism, consisting in portraying philosophy as a factual or descriptive discipline. In a letter of 14 August 1947 to Gustav Bergmann, Sellars insisted that one of the aims of his early work was to show “the implications of the non-factual character of philosophical assertions, while avoiding the Wittgensteinian notion that there are no philosophical propositions” (Olen 2016, 178). Sellars argues that one can find a clear example of this general error of metaphilosophical factualism in Hume’s philosophy. Hume’s scepticism was a consequence of his mistake in supposing that the philosophical questions he asked in the study were sweeping questions of fact, and that therefore outside the study he took an unquestioning attitude towards factual propositions questioned in the study. The truth of the matter, and I speak in the tradition of Hume, is very opposite. There are no factual statements which become philosophical in the study (though there are non-factual statements which are philosophical outside the study); and in philosophy scepticism is a self-contradictory position. (PPE 25) It is a difficult task to make these metaphilosophical pronouncements from Sellars’s early essays fully intelligible. It seems, as the passage on Hume makes fairly clear, that Sellars was at that time ready to disregard a large group of traditional philosophical problems (including the problem of skepticism) as in principle non-philosophical at all, or as based on confusions and therefore negligible. Presumably the best way to make sense

102 Tadeusz Szubka of Sellars’s early metaphilosophy is to read it as a kind of (mis)interpretation and modification of Rudolf Carnap’s account of philosophy.6 Sellars seemed eager to pursue the idea of philosophy as an analysis of language, although he tried to do it within the limits of constraints imposed on this program by Carnap in his syntactical phase. Consequently, Sellars’s account of pure semantics and pure pragmatics was hardly in agreement with Carnap’s later conception of semantics and pragmatics. In addition, for Sellars an adequate pure theory of language had to do justice to the pervasive normative dimension of language and its rules, and that was the dimension which Carnap tended to disregard or at least misrepresent.7 In another early publication, Sellars emphasized that philosophy, unlike the empirical sciences, is not in the business of discovering substantial new truths about the world. Instead it is constantly clarifying its own questions, to which the answers are rather obvious. He writes: This conception of philosophy as a quest of which the goal is the obvious is, I believe, a sound one. It is the problems and not the answers that are difficult; and a genuine advance is constituted by the replacement of a confused by a less confused question, where the two are in some sense the same. (RNWW 46) Moreover, in his review of Arthur Pap’s Elements of Analytic Philosophy, Sellars insisted that when properly conducted, philosophy is not in the business of establishing incontrovertible results, and confessed: “I, for one, have far greater confidence in the power of the spirit of analysis to dispel illusion, than in that of any learned results however sound” (RP 105). To summarize Sellars’s early metaphilosophical pronouncements: philosophy is an activity focused on clarification of our thoughts; it aims at removing various confusions, and not at establishing positive results. Its constructive dimension is confined to the pure and formal theory of languages. However, in one of his best and most well-known essays, “Philosophy and the Scientific Image of Man”, published several years later in 1962 (and given first as a set of two public lectures at the University of Pittsburgh in December 1960), Sellars proposed a more traditional vision of philosophy, implicitly repudiating his early metaphilosophy.8 In subsequent years he would never make any attempt at a comprehensive revision of this more traditional metaphilosophy, but would merely refine its various details. Perhaps this is one of those major continuities which, as he puts it in his “Autobiographical Reflections”, run through his writings “like shared traits of character” (AR 277). Sellars begins his account of philosophy in “Philosophy and the Scientific Image of Man” with the following statement, which is presumably the most quoted sentence from his work: “The aim of philosophy,

Wilfrid Sellars as an Analytic Philosopher

103

abstractly formulated, is to understand how things in the broadest possible sense of the term hang together in the broadest possible sense of the term” (PSIM 1). The emphasis in this statement should be placed on understanding, which for Sellars is a kind of practical ability “to know one’s way around”, a kind of knowledge how. He insists, however, that—at least at the level of characteristically human activities—practical knowledge how presupposes some propositional knowledge that. Hence “it is obvious that the reflective knowing one’s way around in the scheme of things, which is the aim of philosophy, presupposes a great deal of reflective knowledge of truths” (PSIM 2). Of course, this does not mean that knowledge how may be fully reduced to knowledge that. Nevertheless, one should not assume that this reflective knowledge of truths is knowledge of philosophical truths. The reason why this assumption is incorrect consists in the fact that, as Sellars emphasizes: Philosophy in an important sense has no special subject-matter which stands to it as other subject-matters stand to other special disciplines. If philosophers did have such a special subject-matter, they could turn it over to a new group of specialists as they have turned other special subject-matters to non-philosophers over the past 2500 years, first with mathematics, more recently psychology and sociology, and, currently, certain aspects of theoretical linguistics. What is characteristic of philosophy is not a special subject-matter, but the aim of knowing one’s way around with respect to the subject-matter of all the special disciplines. (PSIM 2) Perhaps one might take this passage as an expression of the radical view, endorsed from time to time, that philosophy has no subject matter whatsoever. But this is not what Sellars wants to claim. He seems to be airing the much more moderate view that philosophy has a subject matter quite different from those of the special disciplines: it is a general framework within which all particular subject matters may be placed. As Sellars puts it more poetically in lectures given on various occasions, in various places: “The ideal aim of philosophizing is to become reflectively at home in the full complexity of the multi-dimensional conceptual system in terms of which we suffer, think, and act” (SK I §3: 295; cf. WSNDL 85). Sellars believes that it is to some extent unfortunate to describe philosophical understanding or knowledge as analysis. He elaborates his worries as follows: It has often been said in recent years that the aim of the philosopher is not to discover new truths, but to “analyse” what we already know. But while the term “analysis” was helpful in its implication that philosophy as such makes no substantive contribution to what

104 Tadeusz Szubka we know, and is concerned in some way to improve the manner in which we know it, it is most misleading by its contrast to “synthesis”. For by virtue of this contrast these statements suggest that philosophy is ever more myopic, tracing parts within parts, losing each in turn from sight as new parts come into view. One is tempted, therefore, to contrast the analytic conception of philosophy as myopia with the synoptic vision of true philosophy. And it must be admitted that if the contrast between “analysis” and “synthesis” were the operative connotation in the metaphor, then a purely analytic philosophy would be a contradiction in terms. (PSIM 3–4) Sellars holds then that to call philosophy analytic or analytical is useful merely as a way to emphasize that it does not give us any new substantive knowledge, does not discover any new facts. However, using this adjective to suggest that philosophy should avoid synthesis would be totally inappropriate. Philosophy by its own nature has to be synthetic, since it is “the ‘eye on the whole’ which distinguishes the philosophical enterprise” (PSIM 3). The sense of the whole is obtained by a philosopher from reflective specialists in other disciplines, and “from the pre-reflective orientation which is our common heritage” (PSIM 3). Having the eye on the whole also requires a certain amount of reflection upon the nature of philosophy itself. As Sellars puts it: It is this reflection on the place of philosophy itself, in the scheme of things which is the distinctive trait of the philosopher as contrasted with the reflective specialist; and in the absence of this critical reflection on the philosophical enterprise, one is at best but a potential philosopher. (PSIM 3) Thus, there is no fully fledged philosophy without metaphilosophy. Given this account of philosophical aims, analysis should no longer be conceived along classical lines, that is, as the definition of a single term or concept, but rather as “the clarification of the logical structure— in the broadest sense—of discourse, and discourse no longer appears as one plane parallel to another, but as a tangle of intersecting dimensions whose relations with one another and with extra-linguistic fact conform to no single or simple pattern” (EPM §40: 171). Sellars expressed in this way his dissatisfaction with the traditional conception of analysis, propounded at the beginning of analytic philosophy by G.E. Moore. Sellars’s favored idea of philosophical analysis was in consonance with Gilbert Ryle’s conception of analysis as logical geography of concepts (Ryle [1949] 2009), and in some important respects Sellars may be taken as a

Wilfrid Sellars as an Analytic Philosopher

105

forerunner of P.F. Strawson’s model of connective analysis. For Strawson, philosophical analysis is an elucidation or elaboration of the “network or system of connected items such that the function of each item could be properly understood, in our sense of understanding, only by grasping its connections with the others, its place in the system—or, perhaps better, the picture of a set of interlocking systems of this kind” (Strawson 1973, 816).9 Differences between these two accounts of analysis are largely terminological: Sellars’s broadly conceived logical structure of discourse is basically the same as Strawson’s network or system of interconnected items. Both these philosophers are also well aware that one is ultimately confronted not with one unified discourse or network but with a group of discourses with intersecting dimensions or a set of interlocking systems. Strawson seems to be satisfied with simple recognition of this diversity and abandonment of further efforts to unify these discourses or conceptual systems. Sellars is more ambitious and wants to see them somehow merged in the light of his metaphysical principles. However, while stressing all these reservations about philosophy as analysis and delineating a new model of philosophical analysis, Sellars appears sometimes to be too ready to accept the idea, characteristic of analytical philosophy of his time, that philosophy makes no substantive contribution to our knowledge. After all, he holds that the reflective activity of understanding or knowing one’s way around within the scheme of things is not purely practical and presupposes some theoretical knowledge of truths. Of course, most of those truths will not be philosophical truths, but in addition to them there will be very general and purely philosophical truths about relationships between various disciplines and their subject matters. Hence to say in a bygone analytical vein that “philosophy as such makes no substantive contribution to what we know” is undue modesty, presumably motivated by a very restrictive idea of what a “substantive contribution” to knowledge can be. Moreover, if reflectively knowing one’s way around should lead one to a synoptic vision of “how things in the broadest possible sense of the term hang together in the broadest possible sense of the term”, it is unlikely that one might achieve this aim just by improving the manner in which we know various things. Presumably this vision would require a great deal of theory construction that would necessarily involve putting forward wide-ranging hypotheses, such as, for example, metaphysical naturalism or ontological nominalism. This is exactly what Sellars constantly does in his mature and later writings. Especially his commitment to a scientifically oriented metaphysical naturalism and its influence on the direction of a wide range of his philosophical arguments is of crucial importance in this regard. This commitment was famously expressed by Sellars in the statement “that in the dimension of describing and explaining the world, science is the measure of all things, of what is that it is,

106 Tadeusz Szubka and of what is not that it is not” (EPM 173). It was also endorsed as the superiority of the scientific ontology over the commonsensical worldview as in the following passage: The world of common-sense objects is a phenomenon bene fundatum, a conceptual projection of what there really is, which is of immediate practical utility, but which reveals its inadequacy when pressed into the service of more demanding questions. That Dinge an sich are knowable objects of scientific theory rather than unknowable objects of metaphysical theory enables the transformation of Kant’s critical idealism into realism, both critical and scientific, in which the current tension between analysis and reconstruction finds its proper resolution. (RPH 214) Of course, these hypotheses could not be taken as pieces of our established knowledge in the strict sense. But they are presented as factual statements and may be considered knowledge claims. Nevertheless, at the same time they are highly contentious and speculative, since they transcend the particular data that are available. That means that in order to achieve a synoptic vision in philosophy a certain amount of modestly speculative thought is required. Sellars ought to have explicitly acknowledged that this is both true and unsurprising, since when C.D. Broad was introducing and describing a distinction between critical and speculative philosophy, in his programmatic and important essay for the early analytical philosophy, he emphasized: “I do not assert that either can be wholly separated from the other. The second quite certainly presupposes the first, and it is probable that in the first we tacitly assume some things that belong to the second” (Broad [1924] 1965, 82). That is to say, critical or analytic philosophy and more traditional speculative philosophy are interconnected and interdependent.

Conclusion If one takes into account Sellars’s philosophical education, the major influences on his views, and his argumentative style of writing, he fits squarely into the analytic tradition. However, Sellars’s place within this tradition is rather unique. Though he was tempted in his early writings to embrace a narrowly critical and unsystematic approach to philosophical questions, as well as to confine philosophy to the formal theory of languages, he later changed his metaphilosophical views and consistently defended during his middle and later period a systematic conception of philosophy, grounded on the holistic analysis of our discourse and its structure (or better, our discourses and their structures), but having as its ultimate aim a unified and comprehensive synthesis of our images of the

Wilfrid Sellars as an Analytic Philosopher

107

world. So it is perhaps not an overstatement that “Sellars has left us not more nor less than an instance of a new kind of philosophy: both analytic and synoptic, aimed at an integral understanding of a world common to scientists and people” (Seibt 1990, 8).10

Notes 1. Peter Olen rightly observes that Sellars’s role in shaping contemporary American analytic philosophy is very often neglected by historians of the analytic tradition (Olen 2016, 83). 2. Written evidence of those discussions has not been preserved, or at least it is not included in the archival collection of the Wilfrid Sellars Papers, in which the most relevant item documenting the production of the anthology is box 158, folder 3. 3. The importance of Feigl’s and Sellars’s anthology is also emphasized in various recent surveys of analytic philosophy in which it is described as the first reader in modern philosophical analysis (Beaney 2013, 101) and “a key teaching text for several decades” (Dainton and Robinson 2014, 555). 4. Keith Lehrer was a student of Sellars at the University of Minnesota, and was strongly influenced by him. A succinct account of this influence is contained in Lehrer’s intellectual autobiography (Lehrer 1981, 4–6). 5. Presumably, Feigl and Sellars were no longer interested in editing the journal that they founded, since after its heavy expansion it required a daunting amount of work. Moreover, both of them were to some extent dissatisfied with the way in which current analytic philosophy was developing. Sellars expresses his dissatisfaction in a letter to Feigl dated 18 January 1979: “It is hard to make headway against the currents that have been dominating the waters during these past few years. The basic issues which we explored together are disappearing from view under a dust of technical details and set-theoretical machinery. Although quite different in spirit, the overall effect reminds me of the deadening quality of Oxbridge word-mongering” (Wilfrid S. Sellars Papers, University of Pittsburgh, box 159, folder 14). 6. The reading of this kind is convincingly presented by Peter Olen who argues that “Carnap’s philosophy, especially the meta-philosophy found in his Logical Syntax of Language, plays a monumental role in shaping Sellars’ initial attempts to formulate concise philosophical and meta-philosophical positions” (Olen 2016, 5). 7. This point is developed in greater detail by Anke Breunig in her contribution to this volume (Breunig 2019). 8. This discontinuity between the early and mature philosophy of Wilfrid Sellars is emphasized by Olen 2016. In his later period Sellars openly presented himself as an advocate of the classical conception of philosophy. For instance, in his Notre Dame lectures he stated: “For me, philosophy is the crowning not only of science but, of course, of ethics and all the dimensions of experience. . . . Philosophy is the perfection of all these enterprises and if that isn’t a classical notion, I don’t know what is” (WSNDL 105–6). 9. The term “connective analysis” does not figure in the paper from which this quotation is taken. Strawson introduced it explicitly in his 1983 Woodbridge lectures (Strawson 1985, 25) and alluded to it frequently in his later writings. 10. I am very grateful to the editors of this volume for their encouragement and patience. I am also indebted to the publisher’s anonymous referee who pressed me to clarify, elaborate, and substantiate a number of points made in the chapter.

108 Tadeusz Szubka

References Beaney, Michael, ed. 2013. The Oxford Handbook of the History of Analytic Philosophy. Oxford: Oxford University Press. Berlin, Isaiah. 1973. “Austin and the Early Beginnings of Oxford Philosophy”. In Essays on J. L. Austin, 1–16. Oxford: Clarendon Press. Breunig, Anke. 2019. “Sellars’s Twist on Carnap’s Syntax”. In Wilfrid Sellars and Twentieth-Century Philosophy, edited by Stefan Brandt and Anke Breunig, 29–54. London: Routledge. Broad, C.D. (1924) 1965. “Critical and Speculative Philosophy”. In Contemporary British Philosophy: Personal Statements, edited by J.H. Muirhead, 75–100. London: George Allen & Unwin Ltd. Dainton, Barry and Howard Robinson, eds. 2014. The Bloomsbury Companion to Analytic Philosophy. London: Bloomsbury. The Editors. 1950. “A Message to Our Readers”. Philosophical Studies 1 (6): 94. Feigl, Herbert and Wilfrid Sellars, eds. 1949. Readings in Philosophical Analysis. New York: Appleton-Century-Crofts, Inc. Feigl, Herbert, Wilfrid Sellars, and Keith Lehrer, eds. 1972. New Readings in Philosophical Analysis. New York: Appleton-Century-Crofts, Educational Division, Meredith Corporation. Gross, Neil. 2008. Richard Rorty: The Making of an American Philosopher. Chicago: The University of Chicago Press. Lehrer, Keith. 1981. “Self-Profile.” In Keith Lehrer, edited by Radu J. Bogdan, 3–104. Dordrecht: D. Reidel Publishing Company. Olen, Peter. 2016. Wilfrid Sellars and the Foundations of Normativity. London: Palgrave Macmillan. Rorty, Richard M., ed. (1967) 1992. The Linguistic Turn. Chicago: The University of Chicago Press. Ryle, Gilbert. (1949) 2009. The Concept of Mind. London: Routledge. Seibt, Johanna. 1990. “Analysis without Synopsis Must Be Blind: Obituary for Wilfrid Sellars”. Erkenntnis 33 (1): 5−8. Sellars, Wilfrid S. (1947) 2005 [ENWW]. “Epistemology and the New Way of Words”. In Wilfrid Sellars. (1980) 2005. Pure Pragmatics and Possible Worlds: The Early Essays of Wilfrid Sellars, edited and introduced by Jeffrey F. Sicha, 28–40. Atascadero, CA: Ridgeview Publishing Company. Hereafter PPPW. ———. (1947) 2005 [PPE]. “Pure Pragmatics and Epistemology”. In PPPW, 4–25. ———. (1948) 2005 [RNWW]. “Realism and the New Way of Words”. In PPPW, 46–78. Reprinted in a revised version in Feigl and Sellars 1949, 424–56. ———. 1948−1949 [RC]. “Review of Language and Myth by Ernst Cassirer”. Philosophy and Phenomenological Research 9 (2): 326–9. ———. 1950 [RP]. “Review of Elements of Analytic Philosophy by Arthur Pap”. Philosophy and Phenomenological Research 11 (1): 104–9. ———. (1956) 1991 [EPM]. “Empiricism and the Philosophy of Mind”. In Wilfrid Sellars. (1963) 1991. Science, Perception and Reality, 127–96. Atascadero, CA: Ridgeview Publishing Company. Hereafter SPR. ———. 1962 [TC]. “Truth and ‘Correspondence’”. The Journal of Philosophy 59(2): 29–56. Revised version in SPR, 197–224. Reprinted in Feigl, Sellars, and Lehrer 1972, 214–31. ———. (1962) 1991 [PSIM]. “Philosophy and the Scientific Image of Man”. In SPR, 1–40.

Wilfrid Sellars as an Analytic Philosopher

109

———. (1963) 1974 [EAE]. “Empiricism and Abstract Entities”. In Wilfrid Sellars, Essays in Philosophy and Its History, 245–86. Dordrecht: D. Reidel Publishing Company. ———. 1964 [NI]. “Notes on Intentionality”. The Journal of Philosophy 61 (21): 655–65. Reprinted in Feigl, Sellars, and Lehrer 1972, 430–6. ———. (1966) 1977 [IRH]. “The Intentional Realism of Everett Hall”. In Wilfrid Sellars, Philosophical Perspectives: Metaphysics and Epistemology, 29–48. Atascadero, CA: Ridgeview Publishing Company. ———. 1966 [RPH]. “The Refutation of Phenomenalism: Prolegomena to a Defense of Scientific Realism”. In Mind, Matter, and Method: Essays in Philosophy of Science in Honor of Herbert Feigl, edited by Paul K. Feyerabend and Grover Maxwell, 198–214. Minneapolis: University of Minnesota Press. ———. 1967 [SRTT]. “Some Reflections on Thoughts and Things”. Noûs 1(2): 97–121. Reprinted as Ch. 3 of Wilfrid Sellars. (1967) 1992. Science and Metaphysics: Variations on Kantian Themes, 60–90. Atascadero, CA: Ridgeview Publishing Company. Reprinted in Feigl, Sellars, and Lehrer 1972, 444–57. ———. 1975 [AR]. “Autobiographical Reflections”. In Hector-Neri Castañeda, ed. 1975. Action, Knowledge, and Reality: Critical Studies in Honor of Wilfrid Sellars, 277–93. Indianapolis: The Bobbs-Merrill Company, Inc. Hereafter Castañeda 1975. ———. 1975 [SK]. “The Structure of Knowledge”. In Castañeda 1975, 295–347. ———. (1979) 1996 [NAO]. Naturalism and Ontology: The John Dewey Lectures for 1974. Atascadero, CA: Ridgeview Publishing Company. ———. (2015) 2018 [WSNDL]. Wilfrid Sellars Notre Dame Lectures 1969– 1986, revised edition edited by David Landy and Pedro V. Amaral. Atascadero, CA: Ridgeview Publishing Company. ———. Wilfrid S. Sellars Papers, 1899–1990. Collection number ASP.1991.01. Archives of Scientific Philosophy, Special Collections Department, University of Pittsburgh, online at https://digital.library.pitt.edu/collection/wilfrid-s-sellarspapers. Soames, Scott. 2008. “Analytic Philosophy in America”. In The Oxford Handbook of American Philosophy, edited by Cheryl Misak, 449–81. Oxford: Oxford University Press. Strawson, P. F. 1973. “Different Conceptions of Analytical Philosophy”. Tijdschrift voor Filosofie 35 (4): 800−34. Strawson, P. F. 1985. Skepticism and Naturalism: Some Varieties. The Woodbridge Lectures 1983. New York: Columbia University Press.

6

How Pragmatist was Sellars? Reflections on an Analytic Pragmatism James R. O’Shea

How pragmatist was Wilfrid Sellars’s philosophy? Both pragmatism and Sellars present moving targets on this question: pragmatism because it is impossible to characterize that polymorphic philosophical tradition without drawing well-informed fire, and Sellars because his relationship to pragmatism was itself a dynamic field of forces of attraction and repulsion. In this chapter, I argue that Sellars’s philosophy overall was deeply pragmatist both in its motivation and in its content, whether considered conceptually, historically, or in his own estimation, and that this is the case even in the important respects in which his views differ from most pragmatists.1 However, this overall assessment has been rejected by many recent pragmatists of both the “classical American” and “analytic neopragmatist” varieties (if one still wants to insist on that overwrought distinction, which I do not emphasize).2 Many “classicalist” pragmatists have found regrettable the dominance of formal-linguistic paradigms of inquiry characteristic of much analytic philosophy in general, and also the downplaying of “experience” by Sellarsinspired neo-pragmatists such as Richard Rorty and Robert Brandom (cf. Brandom 2011, 197, and the recent critique of Brandom in relation to classical pragmatism on this point in Levine 2012; I discuss the issue in O’Shea 2014). Furthermore, the analytic neo-pragmatists themselves have pointed out many respects in which Sellars’s philosophy arguably runs against the grain of pragmatism. Rorty remarked in 1979 that in light of Sellars’s representationalist theory of “picturing” in relation to matterof-factual empirical discourse, “we pragmatists mourn Sellars as a lost leader” (1979b, 91, cited by Sachs 2018a; cf. Rorty 1979a, Ch. 6, §5).3 Brandom, too, has recently claimed that “Sellars never explicitly identified himself with pragmatism” (Brandom 2015, 5). For good reason, however, Rorty, Brandom, Hookway, Misak, Williams, Bernstein, and others have done much to highlight the fundamentally pragmatist character not only of Sellars’s famous rejection of the “Myth of the Given”, but also of the Sellarsian normative-inferentialist conceptions of meaning and justification within the “logical space of reasons” that underwrite that rejection.

Reflections on an Analytic Pragmatism

111

Furthermore, as we shall see, Sellars’s own attitude toward pragmatism was more welcoming than Brandom’s preceding remarks suggest, a result that Brandom would welcome given his own insightful and sustained emphasis on “the essential role that pragmatism plays in working out [Sellars’s] metalinguistic form of neo-Kantianism” (Brandom 2015, 5). In what follows, Section 1 explores the classical pragmatist influences on the development of Sellars’s philosophy, with reference to aspects of the intellectual background in which those views formed. Section 2 then outlines more abstractly some of the enduring pragmatist themes in Sellars’s philosophy, including his conceptions of the Myth of the Given, the space of reasons, and his normative inferentialist theory of meaning. I conclude in Section 3 with Sellars’s views on truth and picturing, which present a complex case for the questions of “how pragmatist” Sellars’s views both were and should be.

1 Sellars and Classical Pragmatism: Historical and Motivational Connections Both Olen (2015, §6)4 and Brandom (2015, 5–6) support the idea that, as Brandom puts it, Sellars “never thought of himself as a pragmatist”, in part by highlighting the following retrospective paragraphs from Sellars’s Naturalism and Ontology (NAO, based on his 1974 John Dewey lectures at Chicago): When I was coming to philosophical consciousness, the great battles between the systems which began the Twentieth Century were drawing to a close, although the lightning and the thunder were still impressive. I cut my teeth on issues dividing Idealist and Realist and, indeed, the various competing forms of upstart Realism. I saw them at the beginning through my father’s eyes, and perhaps for that reason never got into Pragmatism. He regarded it as shifty, ambiguous, and indecisive. One thinks in this connection of Lovejoy’s “thirteen varieties”, though that, my father thought, would make too tidy a picture. . . . Pragmatism seemed all method and no results. After striking out on my own, I spent my early years fighting in the war against Positivism—the last of the great metaphysical systems; always a realist, flirting with Oxford Aristotelianism, Platonism, Intuitionism, but somehow convinced, at the back of my mind, that something very much like Critical Realism and Evolutionary Naturalism was true. Thus it wasn’t until my thought began to crystallize that I really encountered Dewey and began to study him. . . . He caught me at a time when I was moving away from “the Myth of the Given” (antecedent reality?) and rediscovering the coherence theory of meaning.

112

James R. O’Shea Thus it was Dewey’s Idealistic background which intrigued me the most. I found similar themes in Royce and later in Peirce. I was astonished at what I had missed. . . . As for Naturalism. That, too, had negative overtones at home. It was as wishy-washy and ambiguous as Pragmatism. (NAO 1–2)

Sellars is not saying in this passage that he never got into pragmatism— exactly the opposite. He is saying that he never got into pragmatism during his early exposure to philosophy at home under the influence of his father, but that he did so after striking out on his own and eventually getting into Dewey’s pragmatism in a serious way just when his own thought first began to come together in two of its most crucial and lasting respects. These were his rejection of the Myth of the Given and his embrace of a holistic, coherence theory of meaning, both of which were enduring themes in both the classical pragmatism of Peirce, Royce, and Dewey and in the Hegelian idealist background that was absorbed by those thinkers. Sellars’s “flirting with Oxford Aristotelianism, Platonism, Intuitionism” after first striking out on his own describes well the sorts of influences that Sellars in his later “Autobiographical Reflections” (AR) ascribed to the years of his studies at Oxford in his early twenties (1934– 1936), set of course against the enduring background of the critical realism and physical realism that he maintained and adapted from his father (as was already evident in Sellars’s earlier MA thesis at Buffalo in 1933). The period when he “really encountered and began to study Dewey” has to be before 1949, since already by 1949 in “Language, Rules, and Behavior” (published in a volume on Dewey’s philosophy) we find Sellars writing with confident knowledge of Dewey’s rejection of “the cognitive given-ness of sense-data”, as follows: Here we must pay our respects to John Dewey, who has so clearly seen that the conception of the cognitive given-ness of sense-data is both the last stand and the entering wedge of rationalism. Thus, since anything which can be called cognition involves classification, the conception of the cognitive given-ness of sense-data involves as a necessary condition the given-ness of universals—even if only sense-universals. (1949 LRB §25: 127) In this passage, Sellars anticipates the two most fundamental conceptions that will drive his own famous rejection of the myth of the givenness of sense-data (et al.) in “Empiricism and the Philosophy of Mind” (EPM) only five years later, in 1956. There the idea that all cognition properly speaking involves classification—for the later Sellars of 1981, this will include both logico-conceptual, linguistic classification, and the

Reflections on an Analytic Pragmatism

113

more basic ur-conceptual “animal representational” classification (cf. MEV)—is marshalled against the idea, for example, that the sensing of sense-data by itself provides direct knowledge by acquaintance (cf. EPM I §§6–7). And the resulting diagnosis in EPM, as above, is that traditional empiricist defenses of the given in Locke, Berkeley, and Hume (and likewise, for Sellars, in C.I. Lewis’s unfortunate defense of the given as “qualia”) inevitably presuppose that we have an allegedly unproblematic and incorrigible direct “recognition” of sensory shareables or “sense repeatables” (cf. EPM VI). The holistic “coherence theory of meaning” from the Hegelian idealist and classical pragmatist traditions bolstered the insight that becomes, in Sellars’s hands in 1956, the idea that any item or state of knowledge or intentionality in general has its conceptual significance only in virtue of that item or state’s having an acquired normative standing or placing “in the logical space of reasons, of justifying and being able to justify what one says” (EPM VIII §36). The development of these central Sellarsian tenets must be understood, again as Olen rightly emphasizes, against the background of his father R.W. Sellars’s critical physical realist view of interpretive meanings as enabling justified judgments that are directly “about” physical objects themselves, while nonetheless denying any non-natural “cognitive relation” or direct intuition or literal “presence to mind” of those physical objects (e.g., R.W. Sellars [1922] 1969, Ch. III; compare W. Sellars’s characteristic view that “meaning is not a relation” to the world, though it of course presupposes various appropriate causal relations; cf. O’Shea 2007, Chs. 3–4). But it also seems clear that the classical pragmatist tradition was likewise crucial to the distinctive development, from very early on, of these two most important and lasting pillars of Sellars’s philosophy, i.e., the idea of the Myth of the Given and of the logical space of reasons. These have also been the two main themes highlighted by all of the Sellars-influenced neo-pragmatists mentioned earlier. What both R.W. Sellars and W. Sellars did hold throughout their writings, correctly in my view (O’Shea 2014, 2015), is that the pragmatist tradition in general has an in-built tendency, which with varying success it has also sought to resist, to slide from realism, via the pragmatic maxim’s otherwise laudable aim of clarifying conceptual meaning in terms of possible verifiable practical effects in experience, into various forms of idealism and phenomenalism. The frequent result, as R.W. Sellars puts this tendency in pragmatism, is that “experience is . . . equated with reality” ([1922] 1969, 77). What R.W. Sellars by contrast admires in pragmatism is that it has served as a corrective to what he saw as the mysteriously direct or naïve realism of the “new realists”, praising pragmatism’s “more empirical idea of the structure of thought as a process” involving behavioral, conceptual, and experiential criteria ([1922] 1969, 77).5 But pragmatism, unlike critical realism, has failed to distinguish adequately between such experiential content or practical criteria on the

114

James R. O’Shea

one hand, and the independent physical object of judgment itself on the other (which often includes various scientifically posited “imperceptible” processes; [1922] 1969, 42–3). The result in both Sellars’s eyes is that the pragmatists from James to Lewis have tended to slip from their official epistemological realism into idealist or phenomenalist temptations (cf. Sellars on R.W. Sellars and C.I. Lewis in PR, passim). As R.W. Sellars concludes: Because of his idealistic antecedents, the pragmatist still thinks of knowledge as an intra-experiential affair. Pragmatism seems to me to be at present in unstable equilibrium. The American branch, at least, is naturalistic and realistic in its tendencies. I am inclined to believe that the blanket term “experience” still hides the genuine problem of knowledge from these would-be realists’ eyes. (R.W. Sellars [1922] 1969, 56)6 A structurally similar diagnosis of the pragmatists’ strengths and weaknesses can be found in his son Sellars’s sympathetic criticisms and adaptations of C.I. Lewis’s “conceptual pragmatism” (1929, 1946). Sellars recognizes Lewis’s overall realist intentions, but he criticizes (again in my view, correctly) his subtle account of “the given element in experience” as falling afoul of the Myth of the Given and as marred by phenomenalist temptations, despite Lewis’s official denials (Lewis 1929, Ch 2; O’Shea 2015, forthcoming; Sachs 2014, Ch. 3). For example, Sellars argues that while Lewis writes “of the interpretation of the given by means of concepts whose implications transcend the given, he also holds that the sensible appearances of things do wear their hearts on their sleeves, and that we do have a cognitive vision of these hearts which is direct, unlearned, and incapable of error”. By contrast Sellars holds that “we must extend to all classificatory consciousness whatever”—i.e., including Lewis’s supposed direct recognition of repeatable qualia-givens—“the striking language in which Lewis describes our consciousness of objects” (ITSA 310–11, italics added). Sellars agrees with Lewis’s modified Kantian and deeply pragmatist insight that our consciousness of persisting objects is constituted by the predictive conceptualization of an objective lawfulness in empirical reality; and furthermore Sellars agrees with Lewis’s view that such conceptualization takes place within a pragmatic conception of the a priori, which involves the replaceability in principle of any given conceptual framework. In this latter case ITSA undertakes to correct Lewis’s purely analytic conception of the a priori with Sellars’s own “material” inferentialist view (cf. O’Shea 2018). So with respect to each of the two fundamental components of Lewis’s conceptual pragmatism—his conception of the given on the one hand and his analytic a priori view of the conceptual domain on the other— Sellars certainly offered fundamental critiques. But in each case this was against the background of fundamental agreement with Lewis’s overall

Reflections on an Analytic Pragmatism

115

pragmatist view of our properly conceptual knowledge of objects and laws, including the idea of fundamental conceptual change based on pragmatic and explanatory considerations.7 Even his father R.W. Sellars’s hostility to pragmatism, from which we started earlier in the chapter, is tempered significantly in his published writings, where one finds that he is equally concerned to emphasize how much the classical pragmatists got right, and how, in effect, a pragmatism properly corrected on the aforementioned key epistemological point would otherwise be fully in the spirit of his own critical physical realism. A glance at R.W. Sellars’s preface to his 1922 Evolutionary Naturalism, for example, makes clear that despite his hostility within Wilfrid’s earshot at home to “Pragmatism” and “Naturalism” as “wishy-washy”, and notwithstanding his own important criticisms of pragmatism’s quasiidealist epistemology, he places his own Evolutionary Naturalism in the positive context “of pragmatism, genetic psychology, behaviorism, electronic physics, social ethics and epistemological realism”, noting that “American pragmatism is” fortunately “strongly biological and naturalistic in its outlook” ([1922] 1969, i, viii). There is much more to be said about the deep historical and attitudinal connections between Sellars and classical pragmatism. However, I want to turn now in Section 2 to a more conceptual and thematic analysis of some of the enduring pragmatist themes in Sellars’s philosophy. There can be no doubt, I think, that Sellars himself connected his most influential philosophical views to those of the classical pragmatists, and I will close these historical remarks with some apposite comments Sellars made during a question and answer session after delivering his 1969 Epistemology lectures at Notre Dame: So, in effect, I am agreeing, with Dewey and Peirce, that justification always occurs within the context of beliefs about the world. . . . The view I am recommending is what I think is in the spirit of Peirce when Peirce denied that there is any intuitive knowledge. . . . I think the same is true of Dewey. Dewey also emphasizes that any particular pattern of cognitive justification occurs in the context of other beliefs which are not themselves questioned at the time. I think this is true. What I want to emphasize is the different pattern of justification that comes in for the case of what we call perceptual or non-inferential knowledge. . . . As I said, I think that this is essentially pragmatic and Peircean in its general line. (WSNDL, 195–6, 200 = Lec. III §§185–7, 198)

2 A Synoptic View of Some Key Pragmatist Themes in Sellars’s Philosophy We have seen that Sellars regarded his rejection of the Myth of the Given as the continuation of a classical pragmatist theme, one that was in part

116

James R. O’Shea

derived also from the Hegelian idealist background to pragmatism as well as from Sellars’s own inherited critical realism and physical realism. We also saw, however, that the pragmatists’ criterial emphasis on verifiable experience was regarded by both R.W. Sellars and Sellars as presenting a continual danger of sliding into idealism or phenomenalism despite the official epistemological realism characteristic of classical pragmatism, C.I. Lewis being the key case in point, but Peirce, James, and Dewey all arguably exhibiting this general tension, too. The key to avoiding that slide, for Sellars, was the other pillar of his philosophy that we saw to be derived primarily from the classical pragmatist and Hegelian traditions, namely, “the coherence theory of meaning”. However, already by the time of his first publications in the late 1940s on what he called “the New Way of Words”, which showed the strong influence of Carnap and the early Wittgenstein (as well as “the Iowa School”; see Olen 2016), and then from 1949 onwards with the increasing influence on Sellars of Ryle’s and Wittgenstein’s views on shared sociallinguistic norms and implicitly rule-governed practices (e.g., Sellars’s 1949 “Language, Rules, and Behavior” and 1954 “Some Reflections on Language Games”), this coherence dimension had taken a more explicitly linguistic form in Sellars’s thinking than it had in either the pragmatist or German idealist traditions (though the influence on Sellars of the semiotics of both Peirce and Charles Morris would be an interesting topic for further historical investigation). Sellars’s invokes Peirce and Dewey as rejecting the Myth of the Given due to their holism: Dewey in relation to his conception of the dynamically changing holistic background of beliefs and practices that is necessary for any particular successful context of intelligent inquiry and experience, and Peirce’s conception of the background of inferences, norms, and habits that make possible what is mistakenly taken by the Cartesian and atomistic empiricist traditions to be purely “unmediated” knowledge by “direct intuition”. This certainly did find a more “analytic” linguistic development in Sellars’s hands reflecting the enrichment of American philosophy from British and continental European influences. In my view, however, Sellars’s development in terms of social-linguistic practices of the “coherence theory of meaning” that he cites as a key pragmatist and idealist theme in the development of his own thinking did not represent a sharp break away from the classical pragmatist tradition of Peirce, James, Dewey, and Lewis. Sellars was right to see his own early philosophical development in terms of such a continuity of influences, stretching in this key respect across German idealism, classical pragmatism, and the “new way of words” in “analytic” philosophy. The crucial conceptual revolution in Sellars’s eyes—speaking now in more abstract thematic terms, but still in the general historical terms that Sellars’s himself continuously used—traces back to Kant on the cognition of empirical objects as made possible by the application of concepts understood

Reflections on an Analytic Pragmatism

117

as involving the prescription of law, rules, and principles, rather than as consisting at bottom, or ideally, in the direct, clear and distinct apprehension of “ideas” as understood in either the classical empiricist or rationalist traditions. The idea is that the concept of any object—of a dog, an electron, or whatever the case may be—is constituted by our implicit understanding of what Sellars often described as the ways of objects: how such objects would, could, or must behave in various circumstances, independently of, but necessarily conceived in possible relation to, our various modes of perspectival encounter with them in one space and one time-order. To have the concept of any mind-independent empirical object just is to be employing lawful, modal, or implicitly rule-constituted generalizations of this kind, and the “categories”, as Sellars interprets and develops Kant’s view, represent our most general ways of functionally classifying the most fundamental types of such concepts as used in experience or inquiry (substance, cause, quality, and so on). Intentionality or the having “in mind” of any type of object, whether real or imagined, is on this view not understood in terms of the model of the “eye of the mind” directly gazing upon or representing the object by means of images or ideas, but rather in terms of the counterfactual-sustaining lawfulness or rule-governed functioning of this object in actual and possible experience: how a thing of this type could, would, or must behave, conceived in terms of further possible experiences. Concepts do what they do by how they function to constrain our possible experience in terms of our inherited and acquired ways of understanding the “ways of things” in the environment (cf. SRLG §50: 340).8 Sellars, developing in his own material-inferentialist way C.I. Lewis’s pragmatic conception of a priori yet replaceable conceptual frameworks, as well as Carnap’s metalinguistic conception of “abstract entities” and categories, sought to broaden Kant’s conception of concepts as functions in judgment in a manner that was thereby deeply informed by the revolutionary logical and scientific developments taking place after Kant’s time, and to offer a Peircean pragmatist, regulative conception of the scientific development of conceptual frameworks “in the long run” of our explanatory inquiries. But the most important aspect of these developments philosophically, at least for Sellars, is how they transform our understanding of the place of mind and meaning in nature, which are thereby no longer understood on the models either of acquaintance with images or ideas, or of “grasping abstract entities” understood in broadly platonic or quasiplatonic terms, but rather on the model of functioning or knowing how to go on in an acquired norm-governed way within a practice, including in particular our social-linguistic practices. This anti-atomist and functional role outlook on the nature of conceptual content and meaning can be traced as a central theme across the classical and neo-pragmatist traditions: in the pragmatic maxim in Peirce and

118

James R. O’Shea

James, as well as Peirce’s emphasis on inferential cognition and James’s account of the concrete function of conceptual or symbolic cognition in terms of practical “leadings” within experience from words, thoughts, or images to the objects themselves; through C.I. Lewis and Sellars to Brandom’s pragmatic inferentialist semantics and his “Kant-Sellars thesis” concerning the causal modalities. I will not repeat that story here, but this holistic pragmatist outlook on conceptual meaning, when bolstered by aspects of the linguistic turn noted earlier, was clearly central to Sellars’s particular way of mobilizing a “coherence theory of meaning” and the “logical space of reasons” to reject the foundationalist myth of “immediate knowledge” or the given in both its empiricist and rationalist forms. When one looks more closely at Sellars’s actual working out of these pragmatist themes, however, we do find him offering further important correctives to these general lines of thinking of the pragmatists, on points other than the tendency of the classical pragmatists to slip back into the presumption of experiential givenness that I have just focused on (not to mention Sellars’s physical realist desire to avoid the idealism of the German idealist tradition). I will close with a brief examination of these further important issues.

3 Truth as Correspondence: Correcting the Pragmatist Tradition? In an important article published late in his career in 1981, “Mental Events”, Sellars writes of his long-standing normative-functional role view of mind and language that this “functionalist theory of meaning and intentionality is but the prologue to a naturalistic philosophy of mind. It prepares the way for, but does not provide, a demystification of the place of mind in nature” (MEV §37: 288). Earlier I characterized Sellars’s coherentist, inferential role account as an enduring pragmatist theme that was crucial to Sellars’s rejection of the Myth of the Given and to his companion epistemic conception of the logical space of reasons. But what does he mean by the preceding comment to the effect that this pragmatist account of mind and meaning is necessary but not sufficient for the naturalistic demystification of the place of mind in nature? We saw earlier that naturalism and physical realism were also central themes for both R.W. Sellars and his son, and that they both viewed pragmatism—as indeed American pragmatism has always viewed itself—as making crucial contributions to the possibility of a coherent philosophical naturalism, despite the recurrent slides into idealism, phenomenalism, and various more subtle givenist temptations. The functionalist and pragmatist account of conceptual meaning, by rejecting any primitive meaning relations or intentional relations between mental events and reality (MEV §35: 288), “prepares the way” for a naturalistic demystification

Reflections on an Analytic Pragmatism

119

of mind and meaning by avoiding problematic conceptions of “intrinsic intentionality” (MEV §23: 285), of mysterious psychological relations to abstract entities, and of any alleged “immediate knowledge” of qualia, sense-data, or any other putatively unmediated givens. So again, what does Sellars mean by suggesting that his practice-based, world-involving coherence account of mind and meaning prepares the way for, but does not by itself provide, an adequate account of the place of mind in nature? In the rest of the “Mental Events” article Sellars proceeds to argue that his pragmatist inferential role account of the nature of thought and meaning has to be supplemented with a naturalistic account of what he calls “animal representational systems” or “RSs”. This is Sellars’s naturalistic “picturing” conception of how both human linguistically structured cognitive systems of representation (Sellars dubs them “Aristotelian” or syllogizing RSs), which literally incorporate a culturally evolved logical space of reasons capable of explicitly representing logical relationships and generalizations, and the more basic and comprehensive non-logical, non-linguistic representational cognitive systems and learning mechanisms involved in the environmentally engaged cognition and agency of both humans and other animals. The biologically evolved and learned patterns of “inference” of the latter “Humean” or associative RSs “ape reason”, to use the historical phrase, in the sense that such animals, while lacking in their RSs the logical operators essential to an “Aristotelian” RS, nonetheless by their nature form sequences of representations in ways that, for the most part, reflect the sorts of “thunder now, lightning shortly” associative inferences that are capable of being logically formulated and assessed by Aristotelian RSs. Sellars argues in MEV that such Humean (animal) RSs as a result have the capacity to form quasi-propositional, proto-conceptual “cognitive maps” of objects in their environment as having various properties and standing in various relations. The theme of a naturalized Tractarian “picturing” correspondence between language/mind and the world was fundamental to Sellars’s philosophy, expressed in varying terminology and undergoing substantial developments and modifications, from early on until the end of his career.9 Sellars’s conception of ground-level empirical or (as he calls it) “matterof-factual” truth (SM V §§1–7) as a correspondence to, or “mirroring” of, the world has generally been held, for understandable reasons, to be a problematic and deeply non-pragmatist theme in Sellars’s philosophy, one that runs against the grain of both classical pragmatism and neopragmatism. This was the source of Rorty’s remark that “we pragmatists mourn Sellars as a lost leader”, and it is reflected also in the widespread rejection of the picturing dimension of Sellars’ philosophy by Rorty, Brandom, Williams, Kukla and Lance, and many other Sellars-inspired pragmatist philosophers (or “left-wing Sellarsians”, as they have been called; for more on the distinction, see O’Shea 2016). This is a complex

120

James R. O’Shea

issue in itself, in the interpretation of Sellars, and in recent debates about the nature and role of truth in pragmatism, old and new. Here in closing I can only highlight some of the key issues in this important debate.10 Broadly put, the pragmatic maxim in Peirce and James had explicated the meaning of a concept in terms of the practical effects its application or functioning in experience has or would have, both on the whole and in the long run, so that, as Peirce put it in 1878 in “How to Make Our Ideas Clear”, “there is no distinction of meaning so fine as to consist in anything but a possible difference of practice” (Peirce 1992, 131). This holds for the concept of truth as for any other concept, so the pramatist looks, for example, to the function or purpose of our practices of asserting or denying that a belief is true. Peirce famously argued that on any given matter of inquiry, the “opinion which is fated to be ultimately agreed to by all who investigate, is what we mean by the truth, and the object represented in this opinion is the real” (1992, 139). Cheryl Misak has argued that this does not mean that Peirce holds that truth is communal agreement, ideal or otherwise, but rather that “a belief is true if it would be ‘indefeasible’ . . . or would forever meet the challenges of reasons, argument, and evidence. In this sense, a true belief is the belief we would come to, were we to inquire as far as we could on the matter” (Misak 2015, 265; cf. Misak 2004). James [1907] 1978 extended the maxim by arguing that those beliefs are true that “agree” with reality by functioning usefully “either intellectually or practically” in a sense that ranged quite widely to include such evidentially open but practically “momentous” and (James argues) long-run beneficial beliefs as the religious hypothesis. Both James and Peirce agree, however, in arguing that the traditional correspondence or representationalist conception of truth—roughly, the claim that a belief or proposition is true if and only if it “corresponds to”, “accurately represents”, or “agrees with” reality— is a merely empty metaphysical or perhaps nominal definition of truth (compare “p” is true if and only p) that tells us nothing about the nature of truth or about how the concept of truth functions in relation to our actual and evolving beliefs, practices, and inquiries. Sellars offered a multifaceted and original account of truth that resembles the classical pragmatists’ accounts in several respects, but differs from them in others. The key elements in Sellars’s account include (for an overview, see O’Shea 2007, Ch. 6; for a penetrating analysis and critique, see Williams 2016): (i) A distinctive, purely intra-linguistic, metalinguistic interpretation of the significance of the “Carnap-Tarski ‘semantic’ definition of truth”, understood in terms of the translational aspects of Sellars’s own normative-pragmatic functional role theory of meaning and abstract entities (e.g., SM IV §§24–9: 92–4; SM V §58: 122f., §81: 128; TC 197–207, 224; and NAO IV §§88–98: 85–7). Roughly, the upshot of his

Reflections on an Analytic Pragmatism

121

analysis is that, in an instance of the Tarskian schema in our language such as “‘Snow is white’ is true (in English) if and only if snow is white”, the function or meaning of “. . . is true” is not to assert that any property or relation obtains at all, whether of correspondence, coherence, or “successful working”. Rather, it has the sense (leaving implicit here Sellars’s metalinguistic conceptual role analyses of propositions and facts, and stating it in terms of the relevant English sentence): “‘Snow is white’ (in English) is correctly assertible” (SM IV §26: 92; cf. NAO IV §§93–4: 86). Sellars believes that the “axiomatizations of semantical theory” are insightful if seen as capturing certain functional forms exhibited by and emerging from our normative inferential practices, but that unfortunately “technical semantics” often obscures the “primacy” of the latter roles (such as “the truth performance”), and in such cases “when it has not been philosophically barren, has spawned new forms of outworn metaphysics” (SM IV §60: 104). This is, in fact, one of the most pervasive and characteristic pragmatist themes running throughout Sellars’s philosophy. (ii) Sellars understands truth as “correct assertibility” in terms of his ground-level account of our socially norm-governed and evolving linguistic practices or conceptual frameworks, i.e., his functional role or “meaning as use” semantics, the richly pragmatist nature of which I have emphasized in Section 1.11 “‘True’, then, means semantically assertible (‘S-assertible’) and the varieties of truth correspond to the relevant varieties of semantical rule” (SM IV §26: 92). Sellars is in this sense a pluralist about truth (compare Lynch 2009): across all types of inferential practices, truth is univocally understood as correct assertibility, but “as a generic concept it takes specific forms which are functions of the semantical rules which govern these different types of propositions” (SM V §1: 106), so there are “essential features of different varieties of truths” (TC 198). For example, “in the case of logical and mathematical propositions . . . S-assertibility means provability” (SM IV §62: 105), while ethical truths must be understood, inter alia, within the context of the shared community intentions and the conditions for human welfare that underwrite the objectivity of the ethical “ought”, on Sellars’s view (SM VII: see Koons 2019 for the bases of Sellars’s ethics). (iii) However, in the specific case of basic matter-of-factual truths such as (within the manifest image) that this cup is black, Sellars argued that at bottom “the primary concept of factual truth” is “truth as correct picture”, i.e., the correct linguistic or mental representation of objects or events in the world (SM V §9: 108). This for Sellars is the sense “in which empirical truths correspond to objects or events in the world” (TC 198). So broadly put, while our discourse about what is true, or about what means what, or what refers to what, on Sellars’s view do not directly concern or specify or say anything about “relations between language and the world”, but rather have the function of normatively classifying

122

James R. O’Shea

the various conceptual-linguistic role-players that constitute us as thinkers and rational agents; nonetheless, nonetheless those normative practices both generate and depend upon (and thus imply) specific causal and representational “picturing” relationships to the world (cf. O’Shea 2007, passim): We have seen that the concept of a [representational] world story is an epistemic concept, the concept of a story which is generated by (is) and required by (ought) the rule governed involvement of a language in the world it is about. (NAO V §66: 110) As Sellars sees it, “whatever else language does, its central and essential function, the sine qua non of all others, is to enable us to picture the world in which we live” (TC 213).12 More on picturing and pragmatism in a moment. (iv) Sellars embedded the preceding account of empirical truth as norm-governed assertibility, which by its nature thereby continuously generates the underlying representational correspondences and theoreticalmathematical structures that on Sellars’s account form “cognitive maps” or “pictures” of the ongoing events and processes that make up the world, within an account of the logic of conceptual change and continuity across fundamental framework replacements over time (SM V, NAO IV §§129–38: 94–96). We have already noted the debts to C.I. Lewis’s pragmatic a priori as far as conceptual change is concerned (see also Sellars’s early essay LRB, §§36–40: 131–3 on conceptual change, and his concluding remark (§43: 133f.) that the “kinship of my views with the more sophisticated forms of pragmatism is obvious”). A crucial additional pragmatist element in Sellars’s account of conceptual change begins with the recognition that “truth as S-assertibility raises the question: assertible by whom?” where the answer for Sellars is, “S-assertible by us” but then argues from the nature of conceptual change for what he calls “the Peirceian framework”, the idea of truth as assertibility in the framework that would result from ideally exhaustive inquiry, i.e., “the framework which is the regulative ideal which defines our concepts of ideal truth and reality” (SM V §95: 132, italics added). The merits of and problems for the Peircean conception of truth have been and continue to be vigorously debated among neo-pragmatist philosophers (for references to critics and a defense of Peirce’s view, see Misak 2004, 2015; Rosenberg 2007 for a Sellarsian defense of Peircean convergence). At any rate, this Peircean theme in Sellars is clearly both a classical and neopragmatist theme, in sharp contrast to Sellars’s naturalistic account of the essential “picturing”-correspondence or causal-representationalist aspects of empirical truth and reference, which have generally been thought to run against the grain of the anti-representationalism (anti-“copy theory”) that certainly has been an enduring hallmark of pragmatism across its classical and analytic proponents. Michael Williams (2016), in the

Reflections on an Analytic Pragmatism

123

anti-representationalist and inferentialist neo-pragmatist spirit of Rorty and Brandom, has recently argued in detail that both the epistemic and picturing dimensions of Sellars’s views on truth are deeply problematic. By contrast, he argues that Sellars’s normative-pragmatic inferentialism about meaning-as-use has in fact made possible an across the board deflationary understanding of truth as primarily having a useful expressive function (e.g., enabling us to generalize over endorsed assertions of various kinds). Accordingly, disquotation by means of Tarski’s “equivalence schema itself fixes the meaning of ‘true’”, without the need for any of the more substantive assertibility, Peircean, or representationalist views of truth, with all of the problems they bring in their wake. Williams argues, furthermore, that Sellars’s distinction between “matter-of-factual empirical” discourse on the one hand, and the various other modes of human discourse on the other (mathematical, ethical, modal, etc.), which is the anti-pragmatist dichotomy or bifurcation that really motivated Sellars’s representationalist theory of picturing, can in fact be accounted for entirely within the resources of a pragmatic inferentialism about meaning, for example by appealing to the different kinds of evidential and other practices associated with the different kinds of discourses. “So far as Sellars needs to explain how language can be about the world in which it is used, the theory of picturing is a fifth wheel” (Williams 2016, 254). The challenges to Sellars’s representationalist theory of picturing by otherwise Sellars-inspired neo-pragmatists such as Rorty, Brandom, Kukla and Lance, and Williams leave us with several options on this last important matter. Perhaps such neo-pragmatist criticisms are simply correct. Alternatively, perhaps Sellars’s pragmatism and normative inferentialism both can and ought to be integrated with something similar to Sellars’s naturalistic theory of cognitive representation (deVries 2016; O’Shea 2014; Price 2013; Rosenberg 2007 Ch. 5; Sachs 2018a); or perhaps the social-normative and inferentialist dimensions of Sellars’s thought were not as fundamental to the enduring insights of Sellars’s philosophy as the scientific naturalist dimensions (Churchland 1979; Millikan 2016; Rosenthal 2016; Seibt 2016). Notwithstanding these last currently much discussed issues, however, I hope to have clarified many of the ways in which Sellars’s philosophy was both historically and thematically influenced in deep and enduring ways by the pragmatist tradition since its start.

Notes 1. Stefanie Dach’s recent PhD dissertation at the University of West Bohemia is a good example of the extent to which even Sellars’s ostensibly reductive scientific naturalism can in fact be read consistently with his pragmatism: Dach, The Ontological Privilege of Science: Wilfrid Sellars, Pragmatism and Scientific Realism, 2018, https://zcu.academia.edu/StefanieDach. Also relating pragmatism to other aspects of Sellars’s views is Dach (2016).

124

James R. O’Shea

2. For a comprehensive example of the sharp distinction classicalist pragmatists have often drawn between what is presented as genuine “American Philosophy”, represented by the classical pragmatists and their descendants, and the merely geographic catch-all grouping, “Philosophy in America”, portrayed as dominated by British analytic and European positivist influences in ways generally hostile to classical pragmatism, see The Blackwell Guide to American Philosophy (Marsoobian and Ryder 2004) and further references there. For a sampling of more positive recent assessments of the contributions of analytic philosophy to the development of the pragmatist tradition throughout the twentieth century, see Bernstein (2010), Brandom (2011), Misak (2013), O’Shea (2008), and Talisse and Aiken (2011, Introduction). An interesting case study is presented by Sellars and his father Roy Wood Sellars on the question of the significance of the rise of analytic philosophy for the development of philosophy in America: cf. O’Shea (2008, 231–5), and in greater depth, Gironi (2017). See also Olen (2015), who presents the important historical and conceptual grounds for placing Wilfrid Sellars in the scientific and critical realist tradition that traces back to Roy Wood Sellars and which was often hostile to classical pragmatism and in particular to C.I. Lewis’s “conceptual pragmatism”, rather than viewing him through the currently influential lens of post-Rorty/Brandom analytic pragmatism. In this chapter, however, I argue that the pragmatist and realist dimensions were combined more thoroughly in both the motivation and content of Sellars’s thought than is often thought, despite the divergences of both kinds. 3. Sachs (2018a) argues compellingly that Sellars’s distinction between normative “signifying” in discourse and tracking or “picturing” the environment in cognitive systems was an attempt by Sellars, inter alia, to harmonize pragmatist themes in C.I. Lewis on conceptual meaning and Dewey on ecological cognition. I argue for a structurally similar rapprochement in O’Shea (2012), in this case between Ruth Millikan’s Sellars-inspired defense (against Rorty’s and Brandom’s rejection) of picturing or mapping-correspondence and Brandom’s Sellars-inspired defense (as opposed to Millikan’s rejection) of normative inferentialism about meaning or signification. Sachs also has a more detailed paper on this topic that has just been published online in Phenomenology and the Cognitive Sciences (see his 2018b). 4. Olen (2015) contains valuable historical considerations directly relevant to our topic, but with the aim of arguing against the view, which I defend here, that “classical pragmatism . . . played an influential role in the development of [Sellars’s] philosophy”. However, I find that most of Olen’s claims are actually agreeable to my argument in what follows—such as the idea that “any claims of W. Sellars’ place in the pragmatist tradition need to be balanced with the realist and positivist dimensions inherited from a variety of differing sources” (§38), or in general Olen’s overall argument “that, historically speaking, thinking of W. Sellars as ‘essentially pragmatist’ is too narrow of a reading to be historically accurate” (§34). It is only Olen’s stronger characterizations that I would reject, as when he remarks that “despite all” the historical considerations brought forward in the paper, “one might still think that there is a pragmatic dimension to Sellars’ philosophy”, while in fact, Olen argues, such a dimension (in this case he is discussing Misak’s pragmatist praise for Sellars on “truth as assertibility”) “could only be placed within the pragmatist tradition if one ignores the realist streak in W. Sellars’ thought” (§§37, 38). Against the latter claim, my contention will be, in harmony with all of the historical considerations that Olen presents and with his emphasis on Sellars’s critical realism and physical realism, that Sellars

Reflections on an Analytic Pragmatism

5. 6.

7.

8. 9.

125

from early on in his career (at least as far back as LRB in 1949, since I do not need to take issue here with the thesis of Olen’s 2016 book) saw himself as reconciling classical pragmatism and realism, both of which he took to be fundamental. I certainly do agree with Olen, therefore, that Sellars from early on saw himself as correcting pragmatism in certain key respects to be discussed in what follows. I discuss the early twentieth-century new realists and critical realists, and their relationship to pragmatism and to R.W. Sellars and Wilfrid Sellars, in O’Shea (2008). Rorty and Brandom offer frequent complaints about what they contend are the counterproductive and vague uses of “experience” in classical pragmatism. For example: “So central is the concept to Dewey’s thought that sometimes in reading these works it is difficult to overcome the impression that he is, as Richard Rorty once put it, ‘using the term “experience” just as an incantatory device to blur every conceivable distinction’” (Brandom 2011, 6). I argue for a Sellarsian-pragmatist accommodation of the concept of experience, while recognizing the tensions inherent in the pragmatist tradition in this regard, in O’Shea (2014). For a couple of other instances of Sellars’s general agreement with Lewis’s “pragmatic empiricism” while seeking to correct it in fundamental ways, see for example CIL, 287, on Lewis’s defense of “real connections”: “I am in complete agreement with this thesis, if not with Lewis’ explication of it, and I have a general sympathy with the epistemology he builds upon it. . . . On the other hand, where I do disagree, the sources of my disagreement strike deep”. Or again, see P, 293–4, on Sellars’s own conception of alternative frameworks and the framework-relative a priori: “This, I believe, is a pragmatic conception of the a priori akin to that developed under this heading by C.I. Lewis in his Mind and the World Order, though I should reject the phenomenalism in which he clothes his formulation”. Again, I am aware that both Lewis and many of his recent interpreters have denied that his epistemology is phenomenalist, but notwithstanding I agree with Sellars’s criticisms of Lewis on the given (O’Shea 2015, forthcoming). That is, §50 in the expanded version of SRLG in SPR, but §34 in the shorter 1954 version of SRLG reprinted in SR. Sellars’s (1979) Naturalism and Ontology, Ch. 5 had two years earlier discussed again the conception of picturing that Sellars had explored in depth in various writings, often with varying terminology and subject to substantial development over time, from the late 1950s onwards. The question of the origin and development of Sellars’s theory of picturing is a complex one, in my view. For example, Sellars returns to the terminology of “world stories” as “mirroring the world” that he had used for not unrelated purposes in his earliest “New Way of Words” writings in the late 1940s, but without developing the theory of predication that is essential to his later account of linguistic picturing. Sellars himself seems to have thought that his account of picturing was in or implicit in his earliest essays (see NAO V §§43–59). In Sellars’s late 1983 article, “Toward a Theory of Predication” (TTP), it becomes clear that in many contexts (e.g., in SM and MEV, too) Sellars uses the term “representation”, e.g., “linguistic representation”, to refer to what he calls “picturing”; and it also becomes clear that picturing or the theory of linguistic representation is supposed to play the role in Sellars’s philosophy of a kind of normatively constrained causal theory of reference and characterization, including a theory of animal cognition (cf. O’Shea 2007, Ch. 6). Of course, in most other contexts involving “reference” Sellars is not discussing picturing but rather the normative-classificatory dimensions of meaning.

126

James R. O’Shea

10. For an in-depth discussion of Sellars’s views on truth, see Lionel Shapiro’s contribution to this volume, “Sellars, Truth Pluralism, and Truth Relativism”. 11. Truth as assertibility was of course a central theme in both Dewey’s pragmatism and (in certain phases of) Putnam’s neo-pragmatism, but I cannot pursue those topics here. Sellars made several attempts over the years to specify more carefully the relevant species of normativity involved in truth as “correct semantic assertibility”. For example: “Obviously, ‘true’ does not mean the same as either ‘known to be true’ or ‘probably true’. Thus ‘correctly assertible’ does not mean the same as ‘assertible with good reason or warrant’. Semantic and epistemic oughts must be handled with the same care as that involved in distinguishing (and relating) the various senses in which an action can be said to be morally suitable” (NAO IV §95: 86f.). 12. For explanations of what Sellars means by his modified Tractarian account of “picturing”, see, for example, O’Shea (2007, Ch 6), or Rosenberg (2007, Ch. 5).

References Bernstein, Richard. 2010. The Pragmatic Turn. Cambridge: Polity Press. Brandom, Robert B. 2011. Perspectives on Pragmatism: Classical, Recent, and Contemporary. Cambridge, MA: Harvard University Press. ———.2015. From Empiricism to Expressivism: Brandom Reads Sellars. Cambridge and London: Harvard University Press. Churchland, Paul M. 1979. Scientific Realism and the Plasticity of Mind. Cambridge: Cambridge University Press. Dach, Stefanie. 2016. “Wilfrid Sellars and Pragmatist Aspects of Wittgenstein’s Tractatus”. Beiträge der Österreichischen Ludwig Wittgenstein Gesellschaft 29: 39–41. deVries, Willem A. 2016. “Images, Descriptions, and Pictures: Personhood and the Clash”. In O’Shea 2016: 47–59. Gironi, Fabio. 2017. “A Kantian Disagreement between Father and Son: Roy Wood Sellars and Wilfrid Sellars on the Categories”. Journal of the History of Philosophy 55 (3): 513–36. Hookway, Christopher. 2008. “Pragmatism and the Given: C.I. Lewis, Quine, and Peirce”. In The Oxford Handbook of American Philosophy, edited by Cheryl Misak, 269–89. Oxford: Oxford University Press. James, William. (1907) 1978. Pragmatism: A New Name for Some Old Ways of Thinking and The Meaning of Truth: A Sequel to “Pragmatism”. Cambridge, MA: Harvard University Press. Koons, Jeremy Randel. 2019. The Ethics of Wilfrid Sellars. London: Routledge. Kukla, Rebecca and Mark Lance. 2009. ‘Yo!’ and ‘Lo!’ The Pragmatic Topography of the Space of Reasons. Cambridge, MA: Harvard University Press. Levine, Steven. 2012. “Brandom’s Pragmatism”. Transactions of the Charles S. Peirce Society 48(2): 125–40. Lewis, C.I. 1929. Mind and the World-Order: Outline of a Theory of Knowledge. New York: Dover. Lewis, C.I. 1946. An Analysis of Knowledge and Valuation. La Salle, IL: Open Court. Lynch, Michael P. 2009. Truth as One and Many. Oxford: Oxford University Press.

Reflections on an Analytic Pragmatism

127

Marsoobian, Armen T. and John Ryder. 2004. The Blackwell Guide to American Philosophy. Oxford: Blackwell. Millikan, Ruth B. 2016. “Confessions of a Renegade Daughter”. In O’Shea 2016, 117–29. Misak, Cheryl. 2004. Truth and the End of Inquiry: A Peircean Account of Truth, 2nd edition. Oxford: Oxford University Press. ———. 2013. The American Pragmatists. Oxford: Oxford University Press. ———. 2015. “Pragmatism and the Function of Truth.” In Meaning Without Representation Essays on Truth, Expression, Normativity, and Naturalism, edited by Steven Gross, Nicholas Tebben, and Michael Williams, 262–78. Oxford: Oxford University Press. Olen, Peter. 2015. “The Realist Challenge to Conceptual Pragmatism.” European Journal of Pragmatism and American Philosophy 7: 152–67. online at https:// journals.openedition.org/ejpap/413. ———. 2016. Wilfrid Sellars and the Foundations of Normativity. London: Palgrave Macmilllan. O’Shea, James R. 2007. Wilfrid Sellars: Naturalism with a Normative Turn. Cambridge: Polity Press and Wiley. ———. 2008. “American Philosophy in the Twentieth Century”. In The Routledge Companion to Twentieth Century Philosophy, edited by Dermot Moran, 204–53. London and New York: Routledge. ———. 2012. “Prospects for a Synoptic Vision of Our Thinking Nature: On Sellars, Brandom, and Millikan”. Humana Mente: A Journal of Philosophical Studies 5 (21)1: 149–72. (Special Issue: Between Two Images: The Manifest and the Scientific Understanding of Man, 50 Years On.) online at www.human amente.eu/index.php/HM/issue/view/33. ———. 2014. “A Tension in Pragmatist and Neo-Pragmatist Conceptions of Meaning and Experience”. European Journal of Pragmatism and American Philosophy 6 (2), doi: 10.4000/ejpap.297, online at https://journals.openedition. org/ejpap/297. ———. 2015. “Concepts of Objects as Prescribing Laws: A Kantian and Pragmatist Line of Thought”. In Pragmatism, Kant, and Transcendental Philosophy, edited by Robert Stern and Gabriele Gava, 196–216. London and New York: Routledge. ———, ed. 2016. Sellars and His Legacy. Oxford: Oxford University Press. ———. 2018. “The Analytic Pragmatist Conception of the a Priori: C.I. Lewis and Wilfrid Sellars”. In Pragmatism and the European Traditions: Encounters with Analytic Philosophy and Phenomenology before the Great Divide, edited by Maria Baghramian and Sarin Marchetti, 203–27. London and New York: Routledge. ———. 2019, forthcoming. “Relocating the Myth of the Given in Lewis and Sellars”. In C.I. Lewis’s Conceptual Pragmatism: The a Priori and the Given, edited by Quentin Kammer, Jean-Philippe Narboux, Timur Uçan, and Henri Wagner. London and New York: Routledge. Peirce, C.S. 1992. The Essential Peirce, Volume 1, edited by N. Houser and C. Kloesel. Bloomington, IN: Indiana University Press. Price, Huw. 2013. Expressivism, Pragmatism, and Representationalism. Cambridge: Cambridge University Press. Rorty, Richard. 1979a. Philosophy and the Mirror of Nature. Princeton, NJ: Princeton University Press.

128

James R. O’Shea

———. 1979b. “Transcendental Arguments, Self-Reference, and Pragmatism”. In Transcendental Arguments and Science, edited by Peter Bieri, Rolf Horstmann, and Lorenz Krüger, 77–103. Boston: D. Reidel Publishing Company. Rosenberg, Jay F. 2007. Wilfrid Sellars: Fusing the Images. Oxford: Oxford University Press. Rosenthal, David. 2016. “Quality Spaces, Relocation, and Grain.” In O’Shea 2016, 149–85. Sachs, Carl B. 2014. Intentionality and the Myths of the Given: Between Pragmatism and Phenomenology. London: Pickering and Chatto. ———. 2018a. “‘We Pragmatists Mourn Sellars as a Lost Leader’: Sellars’ Pragmatist Distinction between Signifying and Picturing”. In Sellars and the History of Philosophy, edited by Luca Corti and Antonio M. Nunziante, 157–77. London and New York: Routledge. ———. 2018b. “In Defense of Picturing: Sellars’s Philosophy of Mind and Cognitive Neuroscience”. Phenomenology and the Cognitive Sciences. doi: 10.1007/ s11097-018-9598-3. Seibt, Johanna. 2016. “How to Naturalize Sensory Conscousness and Intentionality with a Process Monism with Normativity Gradient: A Reading of Sellars”. In O’Shea 2016, 186–222. Sellars, Roy Wood. (1922) 1969. Evolutionary Naturalism. New York: Russell & Russell. Sellars, Wilfrid. 1948 [CIL]. “Concepts as Involving Laws and Inconceivable without Them”. Philosophy of Science 15: 287–315. ———. 1949 [LRB]. “Language, Rules and Behavior”. In John Dewey: Philosopher of Science and Freedom, edited by Sidney Hook, 289–315. New York: The Dial Press. Reprinted in PPPW, 289–315. Citations refer to Wilfrid Sellars. (1980) 2005. Pure Pragmatics and Possible Worlds: The Early Essays of Wilfrid Sellars, edited and introduced by Jeffrey F. Sicha. Atascadero, CA: Ridgeview Publishing Company. Hereafter PPPW. ———. (1952) 1963 [P]. “Particulars”. In SPR, 282–97. ———. 1953 (1963) [ITSA]. “Is There a Synthetic a Priori?” In SPR, 298–320. ———. 1954 [PR]. “Physical Realism”. Philosophy and Phenomenological Research 15: 13–32. ———. (1954) 1963 [SRLG]. “Some Reflections on Language Games”. Expanded Version of 1955. In SPR, 321–58. ———. (1962) 1963 [TC]. “Truth and ‘Correspondence’”. In SPR, 197–224. ———. 1963 [SPR]. Science, Perception and Reality. London: Routledge and Kegan and Paul. Reissued by Ridgeview Publishing Company in 1991. ———. 1968 [SM]. Science and Metaphysics: Variations on Kantian Themes, The John Locke Lectures for 1965–66. London: Routledge and Kegan Paul. Reissued by Ridgeview Publishing Company in 1992. ———. 1975 [AR]. “Autobiographical Reflections: (February, 1973)”. In Action, Knowledge and Reality, edited by Hector-Neri Castañeda, 277–93. New York: Bobbs-Merrill. ———. 1979 [NAO]. Naturalism and Ontology: The John Dewey Lectures for 1974. Atascadero, CA: Ridgeview Publishing Company. ———. (1981) 2007 [MEV]. “Mental Events”. In SR, 282–300. ———. 1983 [TTP]. “Towards a Theory of Predication”. In How Things Are, edited by James Bogen and James McGuire, 281–318. Dordrecht, Holland: D. Reidel.

Reflections on an Analytic Pragmatism

129

———. 2007 [SR]. In the Space of Reasons: Selected Essays of Wilfrid Sellars, edited by Kevin Scharp and Robert Brandom. Cambridge, MA: Harvard University Press. ———. 2015 (1969) [WSNDL]. Wilfrid Sellars Notre Dame Lectures 1969– 1986, edited and introduced by Pedro V. Amaral. Atascadero, CA: Ridgeview Publishing Company. Shapiro, Lionel. 2019. “Sellars, Truth Pluralism, and Truth Relativism”. In Wilfrid Sellars and Twentieth-Century Philosophy, edited by Stefan Brandt and Anke Breunig, 174–206. New York: Routledge. Talisse, Robert and Scott Aikin, eds. 2011. The Pragmatism Reader: From Peirce through the Present. Princeton and Oxford: Princeton University Press. Williams, Michael. 2016. “Pragmatism, Sellars, and Truth”. In O’Shea 2016, 223–60.

7

Transcendental Principles and Perceptual Warrant A Case Study in Analytic Kantianism Johannes Haag

John McDowell is certainly one of the most prominent and most subtle representatives of Analytic Kantianism today. In many of his writings he engages with the work of Wilfrid Sellars who himself can, together with Peter Strawson, be counted among the founders of the deep analytic engagement with Kant. In one of his most recent takes on Sellars’s thought, McDowell offers a new and, if successful, very serious criticism of Sellars. This criticism had not been foreshadowed in any of McDowell’s earlier writings, though he does connect it to some of those earlier discussions of Sellars’s work. McDowell claims that in the light of his re-reading some of Sellars’s late papers he had to decisively change his views on what, in Sellars’s view, the warrant of perceptual knowledge (henceforth: perceptual warrant) ultimately consists in. His revised view of Sellars’s conception sees it as ultimately resting on Sellars’s exclusion of alternative views. Proceeding in this way, Sellars neglects, as McDowell claims, a crucial option—namely, the one McDowell wants to defend. Sellars, according to McDowell, never so much as envisages this alternative—hence the title of McDowell’s paper: “A Sellarsian Blind Spot” (2016).1 This blind spot, McDowell argues, is particularly harmful since it prevents Sellars from considering a theoretical option that is philosophically preferable to the one he in fact develops. In the following case study on Analytic Kantianism, I want to show that McDowell seriously misunderstands Sellars’s views on perceptual warrant. I will suggest that McDowell’s interpretation of Sellars’s views on perceptual warrant rests on an understanding of Sellars’s conception of the trans-level credibility and justification involved in assessing this warrant that does not do justice to this important part of Sellars’s thought. Moreover, what McDowell seems to not take sufficiently into consideration is exactly the transcendental principles that Sellars ultimately rests this conception on. In other words, McDowell underestimates the conceptual resources that Sellars in his own peculiar version of Analytic Kantianism is able to draw on. McDowell mostly takes his clue to understanding the Sellarsian perceptual warrant from a couple of late papers: “Epistemic Principles”, the

Transcendental Principles, Perceptual Warrant

131

third lecture from “The Structure of Knowledge” (SK 1975, III), and “More on Givenness and Explanatory Coherence” (MGEC 1979).2 From these papers he concludes that his original interpretation of EPM (that placed Sellars much closer to his own conception of perceptual warrant) needs substantial revision. We will see, however, that in his earlier reading McDowell was much closer to Sellars than he now seems prepared to believe. Still, to a certain extent McDowell is right: Sellars never wanted to defend a conception of conceptual warrant that mirrors McDowell’s own. However, I want to show that he was deeply and persistently committed to a conception that belongs in the same family as McDowell’s.

1 Probable Perceptual Warrant?—On an Alleged Blind Spot in Sellars Let me start by very briefly (and tentatively) characterizing the three fundamental alternatives that McDowell envisages with respect to perceptual warrant: (1) Mythical option: Perceptual belief, while essentially conceptual, is (conclusively) warranted by direct, nonconceptual apprehension of a state of affairs. (2) Inferential option: Perceptual belief qua essentially conceptual is (with high probability) warranted by a (trans-level) inference from the occurrence of a certain (inter alia) conceptual episode to the corresponding perceptual belief.3 (3) McDowellian option: Perceptual belief qua essentially conceptual is (conclusively) warranted by having a state of affairs conceptually present to one.4 Option (1) is an option that both McDowell and Sellars reject on grounds of its being a version of the Myth of the Given. However, it figures prominently in Sellars’s discussion of these matters as a foil against which he assesses conceptions like those put forward by Roderick Chisholm and Roderick Firth. McDowell’s criticism of Sellars can then be pinpointed as the claim that Sellars is unaware of option (3) and, consequently, cannot but endorse option (2). However, option (2), as McDowell construes it, has to be rejected first and foremost because it cannot afford perceptual knowledge the warrant necessary for playing its foundational role (a role that Sellars, after all, wants to defend),5 since the warrant in question is never conclusive: On Sellars’ conception the warrant an experience provides for a perceptual belief . . . is never conclusive. . . . But no one has ever offered a satisfactory answer to this question: if one acknowledges that one’s

132

Johannes Haag warrant for believing something leaves open a possibility that is not true, how does that differ from acknowledging that for all one knows—for all one’s warrant warrants one in believing—things are not as one believes them to be? (McDowell 2016, 106; emphasis JH)

If we accept McDowell’s plea for conclusiveness in perceptual warrant (as I do for the sake of this chapter), our approach to the assessment of Sellars’s conception of perceptual warrant must center on the question whether there is such a thing as conclusive perceptual warrant on Sellars’s view. I would like to suggest that we can reply to McDowell’s criticism on Sellars’s behalf by arguing along the following lines: (i) For Sellars perceptual beliefs are conclusively warranted until challenged. (ii) If they are challenged, however, the reasons we can provide for them can afford us warrant-claims of only high probability. An important part of this reply will be to show that Sellars, in subscribing to this take on perceptual warrant, is not committed to the following absurd principle: Perceptual belief P is conclusively warranted, because it is justifiable (by the subject) with high probability. In a more perspicuous form this principle reads like this: Perceptual belief P is justified conclusively, because it is justifiable (by the subject) with high probability. And that, of course, is a nonstarter we certainly should not ascribe to Sellars. That we have to account for it at all is due to the fact that, in our sketch just now of a Sellarsian reply to McDowell, we addressed the possibility of a challenge to a perpetual belief being raised. So it is natural to ask why we should include the possibility of a challenge in an analysis of perceptual warrant in the first place. The reason is that perceptual beliefs, though generally true, can be wrong after all. As McDowell would put it: the perceptual capacity is fallible—a fact we have to take into account, however, only if our perceptual beliefs are challenged.6 Including the possibility of challenge in (i) implies that the warrant of a perceptual belief, for Sellars, is conclusive, as it were, until further notice. This possibility (due to a possible dysfunction of the epistemic capacity involved) does not, however, endanger the original conclusiveness of the warrant.

Transcendental Principles, Perceptual Warrant

133

Nevertheless, we cannot make the warrant reflectively explicit by way of a higher-order judgment without explicitly taking into account this fallibility—a fallibility that ordinarily, i.e., if unchallenged, we do indeed “take in stride” (McDowell 2016, 112). The burden of proof lies completely with the challenger. That is why the subject of the perceptual claim only needs to point to the (in all probability) normal functioning of her capacity. The challenger must therefore give concrete evidence that counts against the prima facie evidence of the fact that the perceptual belief was correctly formed. We will see that evidence resulting from general skeptical reflections will not be sufficient due to the transcendental philosophical embedding of the underlying epistemic principles. This will be part of the elaboration of Sellars’s reason for why we can and must “take this fallibility in stride”. It should be clear that, if we can indeed find in Sellars an answer along the lines sketched earlier, this answer would be very similar to what I have called the McDowellian option (3): Essentially conceptual perceptual belief is conclusively warranted by having a state of affairs conceptually present to one. However, this leaves open the possibility of a deep philosophical difference with respect to how our two protagonists conceive of “having a state of affairs conceptually present to one”. For one thing, a question must be raised about what exactly the nature of the conceptual presence of a state of affairs is. Must it be different from a belief or can it be the belief itself? We will see that this difference is not in fact negligible. Rather it has serious consequences for McDowell’s criticism and the viability of Sellars’s approach to perceptual warrant. Moreover, one has to ask about the nature of the state of affairs that is thus conceptually present to us. In fact, I would claim that it is the latter question that demarcates the real and deep difference between the two versions of Analytic Kantianism. This difference concerns Sellars’s peculiar kind of phenomenalism (and McDowell’s rejection thereof). But since it does not directly affect the conception of perceptual warrant, I will have to leave it to one side.7

2 Epistemic Principles and Trans-Level Inference In order to focus our discussion, it will be helpful to start with the epistemic principle that McDowell finds so revealing in Sellars’s late work on epistemic warrant—and not in a positive way. It is a principle that results from Sellars’s discussion of a corresponding principle by Chisholm. Let us call it the modified Chisholm-principle: (S) If there is a certain sensible characteristic F such that S believes that he perceives something to be F, then it is evident to S that there is something that is F and, hence, that he is perceiving something to be F. (SK III §41: 344)

134

Johannes Haag

Sellars goes on to elucidate: Notice that I have reversed the relative position of the two clauses in the consequent as they appear in Chisholm’s principle [instead of starting with seeing and proceeding to being from there]. This is because, on my interpretation, the core of the principle is (S1) If I ostensibly see there to be an F object here, then it is highly reasonable for me (to believe) that there is an F object here. And the move to (S2) If I ostensibly see there to be an F object here, then it is highly reasonable for me (to believe) that I see there to be an F object here is justified by the conceptual tie between “ostensibly see”, “see”, and truth. (SK III §42: 344–5) There is, of course, a lot to unpack here. For the purpose at hand probably the most important concept in need of clarification is the concept of ostensibly seeing there to be an F object that Sellars substitutes for the phrase “a certain sensible characteristic F such that S believes that he perceives something to be F” from the modified Chisholm-principle. Before coming to that, however, we must first clarify what an epistemic principle is for Sellars. Of course, epistemic principles are not themselves statements of justification. Epistemic principles are metajudgments8 that formulate general practical rules for the epistemic assessment of first-order judgments or, more generally, beliefs.9 If valid, they are part and parcel of “a conceptual framework which defines what it is to be a finite knower in a world one never made” (MGEC §73: 179).10 The close proximity of this last characterization to Sellars’s description of what is the essence of the Kantian critical endeavor is remarkable: Kant is not seeking to prove that there is empirical knowledge, but only to show that the concept is a coherent one and that it is such as to rule out the possibility that there could be empirical knowledge not implicitly of the form “such and such a state of affairs belongs to a coherent system of states of affairs of which my perceptual experience is a part”. (KTE 635) This, I take it, is the best characterization (and, albeit indirectly, one of Sellars’s clearest embraces) of transcendental philosophy. With this in mind we can read the quote from MGEC as an endorsement of this Kantian approach.

Transcendental Principles, Perceptual Warrant

135

Nevertheless, Sellars makes it clear that he is not completely content with an understanding of this method if it commits one to justifications “of the ‘this or nothing’ kind familiar to the Kantian tradition” (SK III §43: 345). He wants to do more, namely, to at least attempt to justify our acceptance of the framework of epistemic principles.11 And as the deeper reason for our acceptance of this framework, Sellars offers us “the necessary connection between being in a framework of epistemic evaluation and being agents” (MGEC §80: 180): Since agency, to be effective, involves having reliable cognitive maps of ourselves and our environment, the concept of effective agency involves that of our IPM [i.e. introspective, perceptual, memory] judgments being likely to be true, i.e., to be correct mappings of ourselves and our environment. (MGEC §82: 180) The final twist of this consideration with its reference to correct mappings of the environment would have clearly pointed to Sellars’s underlying naturalism even without him adding, in a note: “May I call them pictures?” (MGEC 182, n.18). And yet, this “naturalistic turn”, while an asset from Sellars’s own perspective, is not, I presume, something conceptually inherent in the connection between epistemic principles and agency. And while the naturalistic turn is reinforced by corresponding remarks towards the very end of SK III,12 we there find Sellars highlighting a more classical transcendental philosophical perspective in which he embeds his naturalistic proclivities. Emphasizing that the authority of the epistemic principles, according to his account, must ultimately be “construed in terms of the nature of concept formation and of the acquisition of relevant linguistic skills” (SK III §44: 345), he flags an apparent problem for such a view: Must we not conclude that any such account as I give of the principle that perceptual beliefs occurring in perceptual contexts are likely to be true is circular? It must, indeed, be granted that principles pertaining to the epistemic authority of perceptual and memory beliefs are not the sort of thing which could be arrived at by inductive reasoning from perceptual belief. (SK III §45: 345) His solution, again, points to the transcendental philosophical approach about a necessary “framework for a finite knower in a world one never made”—that on this occasion is thus directly applied to the problem of perceptual warrant: But the best way to make this point is positive. We have to be in this framework to be thinking and perceiving beings at all. (Ibid.)

136

Johannes Haag

Hence the point is a conceptual one: This conceptual framework defines what the capacity to perceive consists in and thereby shows that this capacity can only receive its authority through “the nature of concept formation and the acquisition of the relevant linguistic skills”.13 Let me next introduce the justification judgments that result from the epistemic meta-judgments expressed by epistemic principles (S) and (S1). The canonical form in Sellars is often given in the framework of Verbal Behaviourism (VB) and this gives rise to some exegetical problems for McDowell’s reading that I will address later on.14 For now, let us simply accept this framework of overt thought-episodes and silent propensities to these episodes.15 A judgment of justification for a given perceptual claim in this framework has the form exemplified by the following example: S just thought-out-loud “Lo! Here is a red apple” (no countervailing conditions obtain); So, there is good reason to believe that there is a red apple in front of S.16 Importantly, this is the form justification judgments have not only in the case of a judgment from a third-person perspective but also from the firstperson perspective. Here is Sellars’s own example: I just thought-out-loud “Lo! Here is a red apple” (no countervailing conditions obtain); So, there is good reason to believe that there is a red apple in front of me. (SK II §40: 325) The subject of these judgments of justification “must know these same facts about what is involved in learning to use perceptual sentences in perceptual contexts” (ibid.)—hence the explicit reference to possible countervailing conditions (and, as we will see, the implicit reference to the peculiar way that the thought occurs as an involuntary or spontaneous response to a visual stimulus). On Sellars’s account, the fact that even from the first-person perspective we need to engage in this justification in order to explicitly justify our own perceptual claims reflects the fact that the authority of a claim to perceptual knowledge must indeed be “construed in terms of the nature of concept formation and of the acquisition of relevant linguistic skills”. By becoming part of a conceptual framework that is passed on by way of linguistic training by our linguistic peers, i.e., by becoming competent speakers, we acquire a capacity that has the required authority. We have seen that in the case at hand it is the capacity to perceive that, on Sellars’s account, must be conceived as being acquired through “concept

Transcendental Principles, Perceptual Warrant

137

formation and the acquisition of the relevant linguistic skills”. Hence, the authority in question is the authority of a perceptual capacity that is itself grounded in the acquisition of the conceptual framework. (This conception of authority of an epistemic capacity, of course, is a reflection in Sellars’s later work of what in the original account in EPM was described by means of the “two hurdles” that need to be jumped in order to lend epistemic authority to perceptual claims.)17 We must keep in mind that these justifications are clearly higher-order statements. The perceptual claim is not used either in the premise or the conclusion. It is mentioned in both claims, though obviously not in the same way. In the premise it is referred to “as a propositional event in a context to which the basic features of language learning are relevant” (SK II §39: 325) without already being epistemically evaluated,18 while in the conclusion it is the object of epistemic evaluation. This inference is a trans-level inference (and the resulting credibility or epistemic warrant a trans-level credibility19 or trans-level warrant) not because we go from object-level to meta-level—a level we are already at in the premise—but because we move from a non-evaluative meta-level of stating the occurrence of a belief under certain circumstances to the evaluative one that states its epistemic warrant.20 The object of this meta-level reasoning is clearly the perceptual experience that in our example is encapsulated by the spontaneous thinkingout-loud “Lo! Here is a red apple!” The proposition that is mentioned (twice, in different ways) in the trans-level-inference to its perceptual warrant, is used only in the context of the original perceptual experience, the thinking-out-loud. Sellars conceives of the experience itself as a case of direct, non-inferential (perceptual) knowledge.21 It is to this “original perceptual experience” (SK II §41: 325) that we now must turn our attention, since it is both a decisive piece in the puzzle that is Sellars’s conception of perceptual warrant and is, furthermore, closely connected with Sellars’s use of the technical concept of ostensible seeing, which figures so prominently in his analysis of the modified Chisholm-principle (S) with which we started our discussion.

3 Ostensible Seeing and Perceptual Taking Sellars uses the concept of ostensible seeing as early as EPM,22 but it only comes to play a major role in his presentations of the problem from the late 1960s onwards, more or less coinciding with the papers on the epistemology and metaphysics of perception that shaped our discussion of perceptual warrant.23 “To ostensibly see” is introduced in these papers as a technical term that has as its ordinary language counterpart “seems to see”.24 Ostensibly seeing an F object, furthermore, involves the presence in some way of something that is in some way F other than as thought of,

138

Johannes Haag

as Sellars canonically puts it in his late writings on perception in order to pinpoint the nonconceptual element of perceptual experience.25 For our present purpose, however, we need to concentrate on the conceptual aspect of the ostensible seeing of an F object. First of all, ostensibly seeing an object to be F involves, on Sellars’s account, a proposition that we can in first approximation characterize using the example of a thinking-out-loud that we encountered earlier: “Lo! Here is a red apple!” Ostensible seeings, hence, involve not only a nonconceptual presence of the object ostensibly seen but a conceptual presence of this same object as well. However, when we explicitly classify something as an ostensible seeing we do more: we signal that we do not at the same time commit ourselves to the truth (or falsity) of the proposition expressed by the thinking-outloud. This is Sellars’s reason for equating the phrase S ostensibly sees an object to be F and, indeed, ostensibly sees its very F-ness with the phrase an object looks F to S.26 In this way, the introduction of the concept of ostensible seeing is closely tied to Sellars’s analysis of looks-talk that, as he emphasizes repeatedly,27 has not changed from the position elaborated in EPM: by using lookstalk we are already in the business of epistemic evaluation.28 It is, as Sellars put it succinctly in EPM, a way of indicating that “the question ‘to endorse or not to endorse’” (EPM §16(2): 145) has come up. In particular, by using looks-talk we are explicitly withholding judgment. And the same holds for the use of the concept of an ostensible seeing that is its equivalent: by classifying a propositional event we already have moved from the object-level to a higher-order level; and by classifying it as an ostensible seeing we signal that we have moved to the peculiar higher-order level characteristic of statements of justification. Hence, the use of the term “ostensible” in an epistemic principle like the one under discussion is clearly a use that falls entirely under the category of epistemic evaluation. Talk of ostensible seeings is closely related to another technical term Sellars uses to describe what goes on in perception, namely perceptual taking. However, the connection is not as immediate as one might think. At first glance, it might seem that what is perceptually taken for Sellars is simply the propositional content of ostensible seeings. This is true at a certain level of analysis, but it is not Sellars’s final word on the matter. It is true, insofar as classifying a propositional event as an ostensible seeing is a way of epistemically classifying spontaneous conceptual

Transcendental Principles, Perceptual Warrant

139

responses to external stimuli, a way of taking an epistemic stance with respect to it. A consequence of this is that the conceptual responses that are the objects of these endorsements (or their withholding), must themselves already have the propositional form that is under epistemic scrutiny in our principle. They are themselves a special case of believings.29 Thus, when S (ostensibly) sees a red apple in front of him, he conceptually responds with an experience [that] contains as its conceptual core the thought (which (S) is caused to have) Here is a red apple. (SK II §35: 324) Notice, that this is a (Mentalese) sentence with fully developed propositional form. It is a spontaneous propositional response to a causal, in our case visual, stimulus.30 These responses are “pulled out of (us) by nature” (WSNDL 243). The ascription of the thought to subject S (again in the oversimplified framework of VB) goes like this: S thinks-out-loud: “Lo! Here is a red apple”.31 Sellars describes this as “the paradigm case of a non-inferential belief” (SK II §37: 324). The non-inferentiality of this belief obviously contrasts sharply with the inferentiality of the justifications that result from applying the modified Chisholm-principle. Indeed, this contrast is designed to achieve the explicit Sellarsian goal of accounting for the non-inferentiality of direct knowledge (perceptual or other) by means of the coherentist, inferentialist framework he develops in order to account for conceptual content in general. Right now, it is the classification of the ostensible seeing as a belief that I would like to draw attention to: it is a belief that is already “involved in his original perceptual experience” (SK II §41: 325; again in SK III §34: 342). As such, it is not a belief or a thinking that “involves thinking in the question-answering sense of this term, as contrasted with thinking-that-p as a conceptual response to a stimulus” (FMPP III 89 n. 11). The belief in question thus is elicited as an involuntary reaction as opposed to, say, a carefully crafted answer to a challenge in the game of giving and asking for reasons. Nevertheless, it is important to keep in mind that, according to Sellars, even if it is a special kind of belief, it still is a belief. I want to point out that the content of this belief (i.e., of this ostensible seeing), which is sometimes classified by Sellars as a perceptual taking in its own right, is, in the final analysis, a perceptual taking only in a broader sense of the term. It inherits this classification, as it were, from the perceptual taking in the proper technical sense of the term that

140

Johannes Haag

we can tentatively identify with (Sellars’s conception of) Kantian intuitions.32 Intuitions, Sellars points out, in the Kantian framework have the important “role of bringing a particular object before the mind for its consideration” (IKTE §48: 465). And that is exactly the task of a perceptual taking in the technical sense (henceforth: PT). As such, neither intuitions nor PTs have full-fledged propositional form, i.e., subject-predicate structure. They are best expressed by complex demonstrative phrases that function as the subjects of perceptual beliefs— be they perceptual responses or thinkings “in the question-answering sense”. In our example, the distinction between the content of the perceptual experience qua belief and the content of the PT would be illustrated by contrasting Lo! Here is a red apple with This red apple where in the latter formulation the belief has, as it were, shrunk into a complex demonstrative phrase.33 Nevertheless, Sellars insists on classifying even this shrunken form as a case of belief or believing. It is, however, not a believing that, but a believing in: In a perceptual taking one believes in this cat on this mat, and may believe, for example, that this cat on this mat is a Siamese. Thus construed, perceptual takings are in many respects the counterparts of Kant’s “intuitions of manifolds”. They represent this-suches; and it is worth noting that although they are not explicitly propositional in form, they obviously contain propositional form in the sense in which “that green table is broken” contains “that table is green”. (KTI §19: 408)34 PTs, in the final analysis, are believings-in, rather than believings-that. Nevertheless, even they are believings that “obviously contain propositional form”. That is why, at one point towards the end of his first Carus Lecture, Sellars makes it clear that, ultimately, both ways of describing the propositional content of perceptual experiences—as perceptual takings in the broader sense and as PTs—capture important ways of taking up the sensory content that we are presented with in perceptual experience (i.e., the sensing of red we take up in the case of a perceptual taking of the red apple):35 we can take this content as being available “either for a logically distinct act of direct apprehension (i.e., an apprehension of it as a case

Transcendental Principles, Perceptual Warrant

141

of [red]), or . . . for a logically distinct act of believing it to be a case of [red]” (FMPP I §159: 256). While both are cases of believing, in the first case we classify the taking as a PT or an intuition (believing-in), whereas in the second case we classify it as a perceptual taking in the broader sense of being a conceptual response that would be a believing-that or a perceptual judgment that is not yet formed in the question-answering mode.36

4 Whose Blind Spot?—McDowell on Basic Conceptual States With these distinctions in mind let us finally look at McDowell’s take on the passage in SK III, §§ 41–2. He draws attention to Sellars’s claim that he has “reversed the relative position of the two clauses in the consequent as they appear in Chisholm’s principle” (SK III §42: 344): What is Sellars’s point in reversing the order of the clauses? For him, that things are a certain way in the environment is prior in the order of justification to someone’s seeing them to be that way. That would make it incoherent to suppose one could be warranted in believing things to be a certain way by seeing them to be that way, as in the conception I described. (McDowell 2016, 104) This diagnosis is puzzling, to say the least. As our analysis has shown, the if-clauses in the principles S1 and S2, by employing the concept of ostensibly seeing, point to a taking of things to be a certain way. The believings in question are spontaneously formed conceptual responses to certain stimuli that can normally be taken at face value without further ado. (That is how we take the fallibility of the involved perceptual capacity in our stride.) When, considering “the core of the principle”, Sellars qualifies the seeing as an ostensible seeing, this only reflects the fact that in formulating an epistemic principle we are already in the higher-order mode of reflection: since we are formulating the principle explicitly we have to take into account the possibility of a challenge to the perceptual taking in question. Sellars, as we have seen, emphasizes the importance of the absence of any epistemic evaluation in the first premise of the trans-level inference of the justification this principle aims at capturing. This, again, is reflected in the antecedent of the “core of the principle” itself, which by employing the concept of ostensible seeing likewise withholds any epistemic evaluation. Nevertheless, the possibility of challenge must be built into such principles in order to factor in the fallibility of the first-order beliefs in question—and by the use of the noncommittal concept of ostensible seeing this possibility is integrated from the very beginning.

142

Johannes Haag

With respect to the premises of the concrete inferences resulting from the modified Chisholm-principle, this means that by invoking the facts (i) that the thought in question occurred to one in a special way (i.e., as a taking that was “pulled out from me by nature”) and (ii) that no countervailing conditions obtain, one brings into view the fallibility of a the capacity, which provides conclusive evidence provided that it functions normally. This does not imply that we do not start with the implicitly validated assumption of its being a proper seeing in the premise of a judgment of justification that would subject first-order beliefs to epistemic evaluation according to the principle. It only means that this assumption may be challenged—as it certainly may, since for various reasons even “self-presenting states of affairs need not obtain” (FMPP I §139: 253).37 The kind of self-presentingness Sellars accepts is defined by him in the following way: A self-presenting state of affairs is one which is such that if the relevant person at the relevant time were to believe it to obtain, the belief would be noninferentially warranted or self-warranting. (FMPP I §138: 253) Consequently, the fact that within the then-clause of the modified Chisholmprinciple Sellars has reversed the order of seeing that a state of affairs obtains and the obtaining of the state of affairs so that the inference proceeds directly from (ostensible) seeing to the evident obtaining of the state of affairs (and not first to our seeing that it obtains) should, if anything, be in line with what McDowell himself wants to say. Given that our capacity functions normally, the claim about the obtaining of the state of affairs thus presenting itself to us is warranted by the very employment of that capacity: “The central feature of this alternative is that it is the belief in the state of affairs which has epistemic authority” (GEC 618).38 But that is clearly not how McDowell reads Sellars. He takes the passage to be a clear indication of Sellars’s commitment to a view that “does not single out experiences that provide conclusive warrant for beliefs” (McDowell 2016, 103) as a seeing (as opposed to a merely ostensible seeing). Instead, he takes Sellars to refer only to ostensible seeings that provide merely probable warrant for the belief that things are as they seem to be. For Sellars, according to McDowell, seeing can be warranted as such only by supplementing seeming to be with how things really are “in the environment”: If the experience that warrants a belief is a seeing, its being describable like that is derivative from the fact that things are as it makes them seem to be, and that fact is not guaranteed by its warranting the belief in the way it does. (McDowell 2016, 103)

Transcendental Principles, Perceptual Warrant

143

Consequently, the warrant of an experiential belief “consists in” the belief’s “being the conclusion of a ‘trans-level inference’” (McDowell 2016, 107; emphasis JH)—and it would therefore be “only in a Pickwickian sense that this approach finds application for the idea of self-presenting states of affairs” (ibid.). Still, if Sellars seems to be quite serious in making a case for self-presentingness (for instance in GEC and FMPP)39 in the nonmythical version he calls the “belief interpretation of self-presentingness” (FMPP I §161: 256), then this should give McDowell pause. I have suggested a reading that contrasts with this interpretation in taking seriously the idea that the ostensible seeing in question includes a propositional event that is itself, ceteris paribus, a perceptual taking (in any of the two senses distinguished earlier) and, hence, a belief. Why does McDowell not even consider this reading? One way to diagnose what seems to go wrong in McDowell’s reading would be to point to the way it misrepresents Sellars’s remarks that explain the relation between the two principles (S1) and (S2) and the reversal of order with which McDowell finds fault. When Sellars emphasizes the “conceptual tie between ‘ostensibly see’, ‘see’, and truth” (SK III §42) in order to facilitate the transition from (S1) to (S2) he indeed refers to the conceptual fact that seeing is true ostensible seeing. Nevertheless, one should not neglect the fact that the core of (S) according to Sellars is (S1), not (S2)—and that (S1) at the very least can be read as thematizing a warrant that is merely probable only if challenged, i.e., as formulating a principle that presupposes the default perceptual warrant as conclusive. However, there is another, more illuminating way of diagnosing what goes wrong in McDowell’s take on Sellars’s conception of perceptual warrant. In order to see this, note that the interpretation I have suggested works only if we take what is thematized in the if-clause of either (S) or (S1) to be a belief in its own right (as I have suggested earlier in the analysis of ostensible seeing). Otherwise, McDowell would certainly be right: we would then indeed start with something like a conceptually informed experience that is not itself a perceptual belief and infer directly to the obtaining of a state of affairs without classifying the perceptual experience as a seeing, i.e., without classifying it as a perceptual belief or claim—and a fortiori without the possibility of ascribing to it the conclusiveness that characterizes seeings proper. As a consequence, the transition from (merely highly probable) being to (absolutely conclusive) seeing that this state of affairs obtains could indeed not be warranted! Consequently, McDowell has to read Sellars as claiming that in the ifclause there cannot be anything sufficiently akin to a full-blown belief or judgment. And that indeed seems to be what McDowell believes he can find in Sellars. He presupposes it in the exposition of the warrant-problem in his characterization of what I have called the inferential option. According to this option, as we have seen, perceptual belief is warranted

144

Johannes Haag

by a (trans-level) inference from the occurrence of a certain (inter alia) conceptual episode to the corresponding claim. Still in the same context, McDowell writes even more circumspectly about an “occurrence of a thinking that figures in the experience on which the [observation] report is grounded” (McDowell 2016, 102). What is strikingly missing here are terms like “judgment” or even “belief”. McDowell gives us a clue that this choice of words may be important when he adds in a note: It is a complication that Sellars there sets out the inference that displays the warrant for observational knowledge as starting from the occurrence of the thinking-out-loud that is his model for the thinking that an experience is or involves. . . . I have stated the idea in terms of what thinking-out-loud is a model for [emphasis JH]. (McDowell 2016, 102, n. 7) We might wonder how this is supposed to constitute a complication, given the fact that the theory that takes the thinking-out-louds of VB as its model not only transfers the structural properties to inner conceptual episodes but furthermore still allows for the classification of the verbal episodes it takes as its model (i.e., the same thinkings-out-loud) as thought-episodes. McDowell seems to imply that on Sellars’s approach there is a quite dramatic change from the thinking-out-loud that is the model to the thinking in foro interno it is a model for. We find this confirmed in a later section of McDowell 2016 in another note, one that starts off innocuously enough: In Sellars’s approach, the cognitive act that warrants a perceptual believing is the occurrence of a thinking in the manner in which thinkings occur in ostensible perceivings. (McDowell 2016, 109 n. 18) This seems to be in line with what I have said about ostensible seeings. Yet, McDowell proceeds: “That is not a believing” (ibid., emphasis JH). And in what I take to be an elucidation of the supposed complication addressed in the earlier note (McDowell 2016, 102 n. 7), he gives the following explanation: This fact .  .  . is sometimes obscured by Sellars’s habit of explaining the warrant for perceptual believings in terms of inferences that starting from the thinking-out-loud—which is indeed expressive of a believing—that is his model for the thinking-not-out-loud that is an element in an experiencing. And he may be letting that confuse him here, into a suggestion that he needs to think all cognitive acts are believing. That would be wrong by what ought to be his own lights. (Ibid.)

Transcendental Principles, Perceptual Warrant

145

Given the importance of the whole complex apparatus of the interaction of (conceptually informed) image-models that clearly are not believings, and its relation to intuitions that Sellars developed roughly contemporaneously with the late papers on epistemology, I think it is safe to dispose of the claim that Sellars is confused about such an elementary distinction as an unwarranted suspicion.40 Let us instead concentrate on the claim that “by his own lights” Sellars is committed to the perceptual experiences in question not being beliefs. In the context of the note just cited, McDowell uses this claim to assimilate Sellars’s position to what for him is an important aspect of his own account—that not all cognitive acts are believings. Somewhat ironically, we can now see that this doctrine is of pivotal importance if we wish to separate the two approaches, since it is ultimately what facilitates McDowell’s “blind spot” analysis. It does so by suggesting that the ifclause of the epistemic principles (S) and (S1) cannot refer to something that is a belief (even an involuntary, spontaneous one) in its own right— and neither can the premise of the justification judgment resulting from an application of this principle. McDowell’s reason for this assumption is in line with his reading of (S) and (S2): as principles that allow us to infer probable perceptual warrant of a belief from an antecedent perceptual experience that is itself not a belief, they would indeed imply that the experience in question cannot supply the relevant warrant on its own. But Sellars is simply not committed to the claim that these perceptual experiences are not beliefs: it is a central feature of the full-blown theory of mental episodes into which he develops VB by way of the Jonesean theory that the objects the inner episodes are modeled after (i.e., spontaneous thinking-out-louds) are themselves part of the resulting theory: a mental episode either remains an inner episode or it culminates as the episode that it is in the thinking-out-loud it is modeled after.41 Consequently, if the thinking-out-loud can be classified as a belief then so must the thinking in foro interno that is modeled after this thinking-out-loud and may culminate in it. Hence, it is not a warrant that a conceptual experience (that, by implication is not a seeing) provides for a (perceptual) belief. Given that the capacity functions normally, the “experience” itself is the seeing and thus includes the (spontaneous) belief—contrary to what McDowell says.42 But this opens up a new line of interpreting Sellars’s conception of trans-level justification and perceptual warrant—an option that necessarily must have escaped McDowell’s attention. We can pinpoint it by first making the option McDowell subscribes to more explicit: (3a) McDowellian option (explicit): Perceptual belief qua essentially conceptual is (conclusively) warranted by having a state of affairs conceptually present to one in a cognitive act that is more basic than any believing.43

146

Johannes Haag

The Sellarsian option that is not in McDowell’s view is the following: (3b) Sellarsian option: Perceptual belief qua essentially conceptual is (conclusively) warranted by having a state of affairs conceptually present to one in a cognitive act that is itself a believing. And we have already seen how we can square this conception with the trans-level inference account of justification licensed by the epistemic principles (S) or (S1). For, ultimately both accounts have an implicit reference to the normal functioning of the relevant epistemic capacities built into them, simply by acknowledging fallibility due to malfunction. Sellars in (S) and (S1) as well as in the premises of his trans-level judgments of justification makes this explicit by using the noncommittal concept of ostensible seeing and the reference to countervailing conditions, respectively, together with the qualification “highly reasonable” in both cases. These qualifications, as we have seen, are qualifications that are necessary only at the meta-level that is characteristic both for the justification statements and the corresponding epistemic principle. There is nothing comparable to this on the level of the original perceptual experience, which nevertheless is a perceptual belief in its own right—and as such, i.e., as the application of a capacity, it is a belief that has conclusive warrant. For Sellars perceptual beliefs are conclusively warranted until the question “to endorse or not to endorse” comes up. (Of course the challenge that brings this question up need not come from a source external to the subject.) As soon as this question comes up, however, the explicit reasons we can provide for them can afford us warrantclaims of only high probability since we have to take into account the fallibility of a capacity. With this in mind let me return, by way of conclusion, to the worry I introduced when I first outlined my strategy for an answer to McDowell on Sellars’s behalf. One might be tempted to ask: does this strategy not commit Sellars to the absurd claim that Proposition P is justified absolutely, because it is justifiable (by the subject) with high reasonableness? I called this claim a nonstarter since it would license an unjustified leap from high reasonableness to conclusiveness. It should by now be easy to see why Sellars is not committed to such an unwarranted leap: his account would include a leap of this kind only if the trans-level inference in the justification started from the original perceptual experience or belief. And we have seen that this is not the case: this belief is merely referred to in the premise of the justification. It is, as Sellars puts it, not used, but mentioned. The intrinsic conclusive authority or warrant of the original perceptual belief is, as it were, put on hold for further

Transcendental Principles, Perceptual Warrant

147

consideration as soon as we engage with it in the question-answering mode of believing and endorsing. Only then does the possibility of a challenge have to be taken into account—and the warrant can no longer be conclusive sans phrase. However, as our reflection on the transcendental status of the epistemic principles has shown, the two capacities—the capacity to spontaneously respond with a perceptual taking broadly conceived to the environmental stimulus and the capacity to justify perceptual claims—are mutually dependent. The original perceptual belief would not have the conclusive authority it has by default without our ability to justify the claim with a high degree of reasonableness. This reflects Sellars’s deep commitment to coherentism and the attempt to unite it with the non-inferential warrant of perceptual takings that motivates much of the epistemic apparatus introduced earlier. Both capacities are thus reflections of our “concept formation and of the acquisition of relevant linguistic skills”. And both are part and parcel of the “framework we have to be in to be thinking and perceiving beings at all”. Consequently, what Sellars indeed seems to be committed to is not the absurd claim (1) but rather the highly defensible claim that: Proposition P is justified absolutely (if unchallenged), because the subject has the capacity to justify P with high reasonableness (if challenged). Importantly, the object of the because-clause is not the proposition P (as it was in (1)); it is the capacity to justify P. This capacity is rooted in “the fact that Jones has learned how to use the relevant words in perceptual situations” (SK II §37: 324). That is how Sellars avoids the apparent leap from high reasonableness to absolute justification. The resulting picture of perceptual warrant, I would suggest, presents us with a convincing Sellarsian alternative to McDowell’s account, to which I take it to be closely related, nevertheless. McDowell seems either to not take this alternative seriously or to completely overlook it. In any case, this interpretation highlights Sellars’s deep commitment to the reconciliation of “the claims of those who stress warrantedness grounded in explanatory coherence (among whom I count myself) with the claims of those who stress the non-inferential warrantedness of certain empirical statements (among whom I also count myself)” (MGEC §42: 174). And this, after all, is a truly Kantian commitment.44

Notes 1. Cf., e.g., McDowell (2016, 100, 106, 109). 2. Sellars develops the theme of perceptual warrant and related subjects in a number of different papers that fall roughly in the same period of his writing.

148

3.

4. 5. 6.

7. 8. 9. 10. 11. 12. 13.

14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27.

Johannes Haag Among others these are GEC, SRPC, SSOP, FMPP. Furthermore, there are the lecture-transcripts from this period that are published as Notre Dame Lectures (WSNDL) and The Metaphysics of Epistemology (ME). Many of these expositions are very similar, with smaller and even larger parts of the texts completely overlapping. In addition, these topics importantly relate to some of the themes discussed in his papers on Kant, in particular KTE, KTI and IKTE. In what follows, I will draw freely on these sources. Cf. McDowell (2016, 102). We will return later on to the details of McDowell’s characterization. I have condensed it here in order to make the contrast to the other options more obvious. The trans-level character of the inference does not play a role at this early stage of McDowell’s argument. It is, however, introduced a few pages later. Cf. ibid., 105. For a formulation that is very close to this one cf. McDowell (2016, 108). Cf. EPM 170. Sellars ends both SK III (§47: 346) and MGEC (§87: 180–1) with a reference to that passage. Cf. McDowell (2016, 112). Sellars, of course, does not use the capacity-talk that has recently undergone a renaissance. While I do not believe it to be particularly helpful, I will frequently use it in what follows in order to bring to the fore the proximity of the two accounts of perceptual warrant, since it figures prominently in McDowell’s conception. I addressed some of these issues in Haag (2014) and Haag (2017). Cf. MGEC §46: 175. Cf. MGEC §68: 178. Similarly, in GEC 622 and 624, with an emphasis on the coherence of epistemic principles. Cf. MGEC §78: 180. Cf. SK III §§44: 345 & 46: 346. Even if Sellars feels entitled to a naturalistic interpretation of the “correct mappings” arguably built into this transcendental account, it is worth noting that (at least in the context of SK and MGEC) he adds no philosophical support for this further and much stronger claim. What he says here is completely compatible with a purely transcendental interpretation of the concept of a correct mapping resulting from the fact that we “have to be in this framework to be thinking and perceiving beings at all”. What the correct mapping is a mapping of is, at this level of abstraction, dependent only on the underlying completely indeterminate concept of reality. Cf. below p. 144 f. For an exposition of VB cf., e.g., SK II §§7–48: 318–27. For an analogous example, cf. SK II §39: 325. Cf. EPM VIII. The mention of countervailing conditions likewise merely reflects nonevaluative facts about the context of the perception. Cf. PHM 88. Cf. WSNDL 242–43 & 250. Notice the importance that Sellars ascribes to this distinction in his criticism of Chisholm in SK III §§36–43: 342–5. Cf., e.g., PHM 88. Cf. EPM §7: 133. In his Notre Dame Lectures he attributes the concept to Gustav Bergmann. Cf. WSNDL 180. Cf. FMPP I §71: 241. Cf., for instance, SK I §55: 309–10. Cf. FMPP I §76: 242–43. Cf. SK I §54: 309 and WSNDL 239 f.

Transcendental Principles, Perceptual Warrant 28. 29. 30. 31. 32.

33.

34. 35. 36. 37. 38. 39. 40.

41.

42. 43. 44.

149

Cf. FMPP I 241 n .5. Cf., for example, FMPP I §124: 251. Cf. SK II §38: 324. Cf. SK II, §36: 324. Cf. Haag (2017) 30 ff. I discuss McDowell’s take on these distinctions in this paper and more extensively in Haag (2014). McDowell points to this distinction in McDowell (2016, 101), only to deem it irrelevant for his own position. If that is true, he should be comfortable with my Sellarsian option (3b). Sellars uses this form in order to avoid the overly psychologistic and intellectualistic framing of perceptual takings as “thinking without question”. Cf. WSNDL 330–35. He associates this framing with the conception of “the followers of Cook Wilson” (FMPP I §119: 250). And, as we have seen, he is not at all reluctant to use this “psychologist” version himself in other contexts, p. 139. Cf. SM §§46–7: 18 f.; more explicitly in WSNDL 336 f. Cf. FMPP I §149: 254. On this conception of intuitions, cf. McDowell (2009b, 33 f.). For the purposes of this chapter I ignore the adverbial form this sensing would have to have in the framework of the manifest image. Cf. Sellars’s related insistence that these beliefs are not actions or performances in any sense of those terms in SK III §6: 334. Cf. GEC 617 and MGEC §27: 172. A belief in that might be either a PT or a perceptual taking in the sense of a belief formed not in the question-answering mode. Cf. in particular GEC 617–188 and 622 ff. and FMPP I §137–40: 253 and, importantly, §161: 256. McDowell tends to neglect this dimension of Sellars’s thinking with the important exception of McDowell (2009a). Cf. in particular McDowell (2009a) 114 f. I have discussed McDowell’s take on image-models and its relation to intuitions in some detail in Haag (2014) and Haag (2017). Cf. EPM 188, SM 159. To say the contrary is, in effect, to commit the mistake Sellars beautifully captures in MEV: “I find that I am often construed as holding that mental events in the sense of thoughts, as contrasted with aches and pains, are linguistic events. This is a misunderstanding. What I have held is that the members of a certain class of linguistic events are thoughts. The misunderstanding is simply a case of illicit conversion, the move from ‘All A is B’ to ‘All B is A’” (MEV §1: 282). It is still, however, an actualization, not an action. Cf., e.g., LTC 66 f. and 75 f. and RQ §24: 155. Cf. McDowell (2016, 108–9). I would like to thank Lionel Shapiro for his careful reading of, and extremely helpful comments on, an earlier version of this chapter.

References Haag, Johannes. 2014. “McDowells Kant und McDowells Sellars”. In Die Philosophie John McDowells, edited by Christian Barth and David Lauer, 179–202. Paderborn: Mentis. ———. 2017. “Analytic Kantianism: Sellars and McDowell on Sensory Consciousness”. Contextos Kantianos 6: 18–41. McDowell, John. 2009a. “Sensory Consciousness in Kant and Sellars”. In Having the World in View: Essays in Kant, Hegel, and Sellars, 108–26. Cambridge, MA: Harvard University Press.

150

Johannes Haag

———. 2009b. “The Logical Form of an Intuition”. In Having the World in View: Essays in Kant, Hegel, and Sellars, 23–43. Cambridge, MA: Harvard University Press. ———. 2016. “A Sellarsian Blind Spot”. In Sellars and His Legacy, edited by James R. O’Shea, 100–16. Oxford: Oxford University Press. Sellars, Wilfrid. (1956) 1963 [EPM]. “Empiricism and the Philosophy of Mind”. In Wilfrid Sellars. 1963. Science, Perception, and Reality, 127–96. London: Routledge and Kegan Paul. Hereafter SPR. ———. 1963 [PHM]. “Phenomenalism”. In SPR, 60–105. ———. 1967a [KTE]. “Some Remarks on Kant’s Theory of Experience”. The Journal of Philosophy 64: 633–47. ———. 1967b [SM]. Science and Metaphysics. London: Routledge and Kegan Paul. ———. (1969) 2007 [LTC]. “Language as Thought and as Communication”. In Wilfrid Sellars. 2007. In the Space of Reasons: Selected Essays of Wilfrid Sellars, edited by Kevin Scharp and Robert Brandom, 57–80. Cambridge, MA: Harvard University Press. Hereafter SR. ———. 1973 [GEC]. “Givenness and Explanatory Coherence.” The Journal of Philosophy 70: 612–22. ———. (1973) 1974 [RQ]. “Reply to Quine”. In Essays in Philosophy and its History, edited by Wilfrid Sellars, 148–71. Dordrecht, Holland: D. Reidel. ———. 1975 [SK]. “The Structure of Knowledge: (I) Perception; (II) Minds; (III) Epistemic Principles”. In Action, Knowledge and Reality: Studies in Honor of Wilfrid Sellars, edited by Hector-Neri Castañeda, 295–347. New York: Bobbs-Merrill. ———. (1976) 2002 [KTI]. “Kant’s Transcendental Idealism”. In Wilfrid Sellars. 2002. Kant’s Transcendental Metaphysics: Sellars’ Cassirer Lectures Notes and Other Essays, edited by Jeffrey F. Sicha, 403–17. Atascadero, CA: Ridgeview. Hereafter KTM. ———. (1978) 2002 [SRPC]. “Some Reflections on Perceptual Consciousness”. In KTM, 431–41. ———. (1978) 2007 [IKTE]. “The Role of Imagination in Kant’s Theory of Experience”. In SR, 454–66. ———. 1979 [MGEC]. “More on Givenness and Explanatory Coherence”. In Justification and Knowledge, edited by George Pappas, 169–82. Dordrecht, Holland: D. Reidel. ———. 1981 [FMPP]. “Foundations for a Metaphysics of Pure Process”. (The Carus Lectures). The Monist 64: 3–90. First lecture “The Lever of Archimedes” also in SR, 229–57. (References to FMPP I according to SR). ———. (1981) 2007 [MEV]. “Mental Events”. In SR, 282–300. ———. 1982 [SSOP]. “Sensa or Sensings: Reflections on the Ontology of Perception”. Philosophical Studies 41: 83–111. ———. 1989 [ME]. The Metaphysics of Epistemology: Lectures by Wilfrid Sellars, edited by Pedro V. Amaral. Atascadero, CA: Ridgeview. ———. 2015 [WSNDL]. Wilfrid Sellars: Notre Dame Lectures 1969–1986, edited with an introduction by Pedro Amaral. N.p. [Independently published].

Part III

Learning From Sellars

8

Sellars on Inference Johannes Hübner

Many of the theses Sellars defends are so deeply intertwined with one another that it is impossible to accept one without also committing oneself to many others. But there is one distinction Sellars makes of which this is not true, his distinction between two kinds of rules he calls “rules of action” and “rules of criticism”. It is possible to embrace this distinction without becoming a Sellarsian. Moreover, the distinction offers the key to answering a question that is currently the subject of intense discussion, the question, namely, of how to formulate an adequate conception of inference. Considering the distinction is thus a particularly good way to demonstrate Sellars’s present relevance. In what follows I will attempt to do this, while paying particular attention to the nature of inference. I proceed thus: the first three sections introduce and motivate Sellars’s distinction between two types of rules. The fourth section is intended to show how this distinction can help us avoid difficulties that, as Paul Boghossian has shown, confront the project of developing a conception of inference. In the fifth section, I show how Sellars develops a satisfying conception of inference on the basis of his distinction.

1 Actions and Rules of Action Sellars’s distinction is indispensable for any position that regards linguistic and conceptual activity as essentially rule-based. Sellars introduces it to avoid regress problems (RM 485f.). Let us suppose that having a concept of F consists in using the expression “F” correctly and that using “F” correctly entails the ability to follow a rule specifying the use of “F”. If following a rule means doing an action with the intention of conforming to the rule, the intention to satisfy the rule would presuppose a representation of the rule and, therefore, of the concept of the expression “F”. The having of this concept would presuppose mastery of the concept of the expression “expression ‘F’”, and so on. To avoid this sort of problem, Sellars introduces rules that are not rules of action, and a kind of rulefollowing that does not consist in doing actions.

154

Johannes Hübner

Since rules of action concern actions, and rules of criticism, by contrast, acts or states that are not actions, it is first of all necessary to discuss the concept of action. Sellars has a traditional notion of action, according to which an action is something that is or can be done voluntarily. His criterion is this: “An action is the sort of thing one can decide to do—though, of course, in particular cases one may do it without deciding, as one may salute an acquaintance” (SM III §33: 74; cf. FD 153). The relevant modality is conceptual. According to Sellars, whether or not φ is a kind of action depends on whether it is conceptually possible for a person to decide in favor of φ. The rough idea is clear: someone can, for example, decide to inspect their car, but they can’t (reasonably) decide to arrive at a perceptual judgment that the car has a dent. Rather, they arrive at a perceptual judgment by reacting to what they perceive. Someone can decide not to let any annoyance they may feel show, but they can’t (reasonably) decide not to feel annoyed. Thus in each case we have on the one hand something that is a kind of action and, on the other, something that is not. But let us make this a little more precise. (i) If a person is not fully competent in the use of some concepts, such as the concept of judging, it might happen that they decide to do something that is not an action proper, such as making a particular perceptual judgment. But in that case the person will not be able to implement their decision successfully. To avoid the undesirable consequence that the perceptual judgment is an action, Sellars’s criterion should be supplemented by the condition that a person must be able to implement their decision successfully. (ii) The criterion does not always help us to decide the status of particular acts. Consider coughing, for example; some cases of coughing are intentional actions, others are involuntary—they happen to a person. An involuntary cough is not an action, even if coughing is the kind of thing one can decide to do. Sellars’s criterion is not capable of excluding an involuntary cough from the class of actions. (iii) This problem can be solved by taking account of the fact that actions can be described in multiple ways. To do this is appropriate in any case. You cannot, for example, decide to inadvertently trigger an alarm, but if you choose to do something, such as touching a painting, which leads you to inadvertently trigger an alarm, the triggering is an unintentional action. The reason it is possible to describe the triggering in this way is that there is a correct description under which it is intentional. Taking these points together results in an improved formulation of the criterion: [Sellars’s criterion for actions] φ is an action iff φ is intentional under a description “d” that describes an action. An expression “d” describes an action iff it is conceptually possible for a

Sellars on Inference

155

person to decide in favor of d and to succeed in implementing the decision. Habitual actions fulfill the criterion, since an action done out of habit is intentional; for example an act under the description “greeting somebody” is intentional insofar as one greets people as a matter of habit. In what follows, this formulation of the criterion will be assumed. Sellars applies his criterion to the mental in the same way as to public behavior. Mental actions may include things like “deliberating, turning one’s attention to a problem, searching one’s memory” (SM III §33: 74). On the other hand, there are many mental acts, i.e., datable occurrences, that are not actions, e.g., sensations of pain or upsurges of emotion. Some rational acts are not actions either. In particular, judging that p or deciding to φ are not actions because neither the decision to judge that p nor the decision to decide to φ is one that can be made and implemented.1 A helpful test question is whether one could (assuming the absence of obstacles) implement the request to φ. Clearly one could not make a certain judgment or a certain decision on command. Judgments and decisions are typically conscious; they can be made in the context of deliberations or consultations that are themselves actions; and they can be rational insofar as they are based on reasons. These things mark relevant differences to other kinds of mental acts, but do not make judgments and decisions into actions.2 Thus, like all mental acts that aren’t actions, they aren’t subject to rules of action.3 Rules of action determine which actions one is obliged, allowed, or forbidden to do under given circumstances. For present purposes it is enough to restrict ourselves to obligation. Rules of action of this kind are expressed in ought-sentences of the form: If one is in circumstances C, one ought to φ. In this formula, verbs of action are to be substituted for “φ”. Let us designate every behavior that is governed by rules as rule-following, and the species of rule-following that consists in actions as rule-obeying.4 If a person follows a rule of action, they infer via modus ponens that under given circumstances they ought to do something in particular and do it on that account. Four steps are involved. First, a diagnosis of the circumstances: (1) P judges that (i) they are in C (e.g., that they are approaching red traffic lights). Second, an awareness of the rule: (2) P judges that (ii) one ought to φ in C (e.g., that one ought to stop at red traffic lights).

156

Johannes Hübner

Third, a practical inference from the two premises (i) and (ii): (3) P infers that (iii) they ought to φ now (e.g., that they ought to stop now). Fourth, an action as a response to the inferred practical judgment (iii): (4) P φ-s (e.g., they stop). Sellars has a story to tell about the relation between practical “ought” judgments and the forming of intentions or decisions. I will simplify somewhat and pretend that actions are immediate reactions to “ought” judgments. The reader will have noticed that actions that consist in following rules of action are necessarily things one can do on the basis of decisions, and thus always fulfill Sellars’s criterion for actions. The obeying of rules of action is an advanced achievement. In cases such as stopping at traffic lights information concerning the circumstances is delivered by perception. In these cases, rule-obeying involves transitions of exactly those three kinds that Sellars deems essential for intentionality, namely, language entry transitions, intra-linguistic moves, and language departure transitions (cf., e.g., MFC 87f.; SM IV §61: 114). •

Language entry transitions are spontaneous reactions to perceptual objects or inner states and consist in perceptual or introspective judgments.

Typically, premises about the circumstances in which rules are applied are language entry transitions. •

Intra-linguistic moves are inferences. In inferring, one moves from one thought to another, sometimes according to suitable patterns of formal inference.

Inferences may be practical (as in cases where a rule is obeyed) or theoretical. •

Language departure transitions are actions. One moves from the intention to φ to φ-ing.

In following a rule, the intention or the practical “ought” judgment is the result of a practical inference. When someone obeys a rule, they infer via modus ponens and thereby acquire a belief about what they ought to do. So rule-obeying involves not just the drawing of inferences but more specifically the inferential acquisition of beliefs. Obeying a rule implies the acquisition by inference of a belief.5

Sellars on Inference

157

2 What if All Rules Were Rules of Action? As far as I can see, at the time he published “Some Reflections on Language Games”, Sellars held the assumption that all rules are what he would later call rules of action (SRLG §34: 333f.).6 [Rules of action only] All rules are rules of action. Any rule-governed activity is the obeying of rules of action. Given the further assumption that intentional activity is always rulegoverned activity, the Rules of action only assumption leads to a regress. In his early writings, Sellars is aware of the danger of a regress here (LRB §23: 126; SRLG §2: 321). He initially counters it by avoiding the assumption that every kind of intentional activity is rule-governed (SRLG §38: 335).7 By contrast, in his later writings Sellars makes precisely this assumption (LTC 61). In particular, he assumes that transitions of the three types already mentioned are rule-governed, committing himself to the thesis that inferring is rule-governed. In what follows, I would like to exhibit the regress problem with the help of an example of inferring. I will call Sellars’s just-mentioned thesis about inferring his: [Rule assumption about inferring] Inferring is rule-governed activity. This thesis is often espoused, though not unanimously. It appears to be very plausible.8 One consideration that supports it, among others, is that one can infer correctly and with error.9 Every conception of inferring should do justice to this fact. A conception that understands inferring as rule-following is able to do this because rules represent a standard of correctness, and can be checked for appropriateness in their turn. Correct inferences are based on conformity with appropriate rules of inference, while incorrect inferences are based on misapplication of the rules or inappropriate rules of inference. I will focus on a special case of inferring where one acquires a belief that p by inferring p according to modus ponens from premises one believes to be true. There are several things that make this case special. For one thing, not every inferring of p is the formation of a belief p. It is possible for somebody to infer p in the course of a hypothetical train of thought without acquiring the belief p, for example, if they ask themselves what follows from assumptions whose truth they are merely investigating; or if they wish to refute by reductio an assumption which they consider wrong (cf. Wright 2014, 28f.). Furthermore, it is not the case that for every correct inference there exists a valid form of argument; material inferences (e.g., from “the car is black” to “the car is not white”) and inductive inferences (e.g., from “the clouds are dark” to “it will rain soon”) do not correspond to such forms.

158

Johannes Hübner

One more remark on the nature of inferences: inferences can be drawn quickly and automatically or slowly and attentively, such as when somebody articulates the steps of a train of thought out loud. Inferences should be assigned to the personal rather than the sub-personal level, because even in those cases where somebody draws an inference automatically, they can become conscious of it by means of reflection. Which is not to say that it should be presupposed that inferences are necessarily actions, simply because they are conscious and personal—although this presupposition can be found in the contemporary debate over inferring. For example, Paul Boghossian (2014, 2) describes inferring as “person-level, conscious, voluntary mental action”. But this is problematic, for reasons which will be presented in Section 4. At the very least it should not be presupposed that inferences are voluntary actions, since there is no intersubjectively shared datum according to which inferring is experienced as voluntary action. Strawson (2003), for one, has pointed out that he experiences himself as passive in relation to the contents of his thinking and inferring. The rule relevant to the case of an inference drawn by modus ponens may be spelled out something like this: [MP] If one is entitled to an assumption of the form “p and (if p, then q)”, one ought to judge that q (in contexts where the question whether q arises).10 According to the Rule assumption about inferring is rule-governed activity. According to Rules of action only one has to obey a rule of action in order to infer. This rule would be MP. So in order to acquire a belief by inferring, one would have to obey the rule of action MP. So we get the result that acquiring a belief by inferring implies obeying a rule of action, and we have already seen that obeying a rule implies acquiring a belief by inferring. If acquiring a belief by inferring implies obeying a rule and if obeying a rule implies acquiring a belief by inferring, we run either into a regress or into a circle. As it happens, we are faced with a regress because the belief to be acquired by rule-obeying is not identical with the belief implied by rule-obeying. This becomes clear when one looks at the details of what would be required in order to acquire a belief by drawing a first order inference. The person would have to move to the meta-level and, first, register their entitlements: (1′) P judges that (i) they are entitled to an assumption of the form “p and (if p, then q)” e.g., that P is entitled to the assumption that it rained last night and that the streets are wet if it rained last night. Second, P would have to acknowledge the relevant rule:

Sellars on Inference

159

(2′) P judges that (ii) one ought to judge that q when one is entitled to an assumption of the form “p and (if p, then q)” e.g., that one ought to judge that the streets are wet when one is entitled to the assumption that it rained last night and that the streets are wet if it rained last night. Third, P would have to make a second order inference about what they should judge: (3′) P infers from (i) and (ii) that (iii) they should judge that q e.g., that they should judge that the streets are wet. Finally, in response to the second order judgment (iii) they would have to make the first order judgment that q: (4′) P judges that q e.g., P judges that the streets are wet. If this had to take place whenever somebody drew an inference, drawing an inference would be impossible. In order to draw the first order inference that q, P would have to draw a second order inference to the effect that they should judge that q. This second order inference would (according to Rules of action only) be a further case of obeying a rule of action. So in order to draw the second order inference, a third order inference to the judgment that they should judge that they should judge that q would be necessary. More generally, in order to make the n-th order inference that q, one has to draw an inference of order n + 1 to the judgment that one should judge that q. The further inferences would bring P further and further back from the first order conclusion that q. Given that rule-obeying implies inferring, there is a conflict between the Rule assumption about inferring and Rules of action only. One of them has to go. Sellars drops Rules of action only. Analogous problems result from analogous assumptions for the other two kinds of transition involved in rule-following, Sellars’s language entry and language departure transitions. The problems arise if one retains Rules of action only and, by analogy with the Rule assumption about inferring, assumes that these transitions, too, are rule-governed. Let’s consider judgments of perception. In this case a circle threatens (cf. SRLG §34: 333f.; Wright 2007, 495). Let’s assume that judgments of perception are rule-guided acts. We can formulate the relevant rules as rules of action, like this (cf. LTC 63): If, in a perceptual situation, one is confronted with something red, one is entitled to say that the thing is red. In order to obey the rule, somebody would have to judge that under the given circumstances the condition of the rule’s application is met. And in

160

Johannes Hübner

order to do that, they would have to reach the perceptual judgment that the thing is red. Thus they would have to reach precisely the judgment that is supposed to arise through following the rule. In order to arrive at a perceptual judgment, they would have to follow a rule of action, and in order to follow a rule of action, they would have to arrive at the perceptual judgment—which is circular. Problems of the kind just considered motivate Sellars’s assumption that not all rules are rules of action and that rule-governed activity is not always rule-obeying. The other rules are rules of criticism.

3 Rules of Criticism Rules of criticism specify how things ought to be. For this reason, Sellars also calls them “ought-to-bes”.11 Rules of criticism can apply to things and beings which do not possess the concepts necessary for knowing that they are subject to them. Rules of criticism are expressed by propositions of the form: x ought to be in state φ, whenever such and such is the case.12 It is important to note that rules of criticism may require one to do something, but of course not in the action sense of “doing” but in the broad sense of “doing” in which snoring and sweating are doings (cf. FD 153). Inferring and judging count as doings in this broad sense. A sentence of the form “x ought to do φ, whenever such and such is the case” might express a rule of criticism. This means that one cannot read the kind of rule off from the grammatical form of a sentence that expresses a rule. Instead, one has to consider whether φ-ing can be an action. Rules of criticism specify the roles which certain things ought to play. We refer to rules of action in order to determine which actions are required; we refer to rules of criticism in order to check whether someone (possibly oneself) is behaving in the right way. If a person does not behave in the required way, correction is due. Sellars illustrates rules of criticism with rules pertaining to emotional reactions, e.g., “one ought to feel sympathy for bereaved people” (LTC 60) and “one ought to feel gratitude for benefits received” (SM III §38: 76). The point is that emotional reactions are obviously not actions even though on occasion it is natural to say that one ought to display them. There is a conceptual connection between rules of criticism and rules of action, roughly, that “ought-to-bes imply ought-to-dos” and vice versa (LTC 60; cf. SM III §37: 76). There would be no rules of criticism if there were no rules of action. The rule of action implied by a rule of criticism says that one ought to bring it about that the subjects of the rule of criticism conform to the rule. If, for example, a rule of criticism says that a person about to have their photo taken ought not to have a sweaty, shiny

Sellars on Inference

161

face, then the corresponding rule of action will say that one ought to take suitable measures to make sure that a person about to have her photo taken does not have a sweaty, shiny face. Central to Sellars’s theory of intentionality are rules of criticism pertaining to perceptual judgments, inferences, and decisions, i.e., to the three kinds of transition elaborated in Section 1. A rule of criticism for the application of “red” in perceptual situations could be formulated thus: If, in a perceptual situation, one is confronted with something red, one is entitled to judge that the thing is red. In order to follow this rule, a person must be able to react with either an inner tokening of “that is red” or one spoken aloud. This reaction constitutes the judgment that the thing is red and does not presuppose it. Things would be different if the rule were a rule of action (see the end of Section 2). In general, in order to satisfy the rule of criticism that one should φ in C, a person does not have to meet cognitively demanding requirements. Basically, it is enough if one has been trained to conform to the rule. It is neither necessary to judge that C obtains, nor to have knowledge of the rule, nor to infer that one should φ now. If this were otherwise, conforming to rules of criticism would presuppose intentionality and could not possibly be constitutive of it. Problems of regress would return (for more on this, see Section 5).13 Sellars assumes a division of labor between generations of trainers and trainees. Trainees are subject to rules of criticism because trainers, wanting trainees to conform to the rules, see to it, via positive and negative reinforcement, that they do so. Successful training generates behavioral regularities that Sellars calls “pattern governed behavior” (MFC 86f.). By the time the trainees learn to criticize themselves and others they have learned to make rules of criticism explicit. In this way they become apt addressees for reproach or blame. Yet, as Sellars emphasizes, this does not mean that they learn to obey the rules of criticism (MFC 88). It is as impossible to obey a rule of criticism as it is to turn a non-action into an action. So there are two kinds of rule-governed behavior: obedience to rules and pattern-governed behavior. I would like to avoid a possible misunderstanding. It would be a mistake to think that states and acts governed by rules of criticism are necessarily sub-personal. To see why, consider perceptual judgments. A perceptual judgment is a paradigmatic case of an act governed by rules of criticism. It is made spontaneously, without prior deliberation. Yet perceptual judgments are personal acts because they are typically conscious and are correctly ascribed to persons and not to parts of persons, e.g., to certain regions of the brain. A person might be blamed for a mistaken perceptual judgment if the mistake is due to lack of care and attention. In the second half of this chapter, I wish to discuss inference in order to show why Sellars’s distinction between two types of rules is indispensable.

162

Johannes Hübner

4 Boghossian on Inference If the point is to avoid intellectual dead ends, then Sellars’s distinction proves fertile.14 For an example of one such dead end I turn to the attempt by Paul Boghossian (2014) to explain what an inference is. By “inference” Boghossian means “reasoning with beliefs”, i.e., reasoning that starts from beliefs and possibly affects beliefs (2014, 2). Boghossian concentrates on the more particular question—introduced in Section 2—concerning what it would mean to gain a belief by inferring it, by modus ponens, from premises taken to be true. To infer is to move from one thought to another. The question is: what kind of transition is needed, exactly? According to the causal proposal, a necessary condition is that one thought (i.e., one act of thinking) causes another. Of course, this condition is not sufficient, since mere association is a causal but not an inferential connection. If one adheres to the causal idea, then the task is to figure out the right causal connections. While Sellars accepts this idea (MEV §59: 292), Boghossian regards it as hopeless. He rules this strategy out by way of a simple example: A habitual depressive’s judging “I am having so much fun” may routinely cause and explain his judging “Yet there is so much suffering in the world”, as directly as you please, without this being a case in which he is inferring the latter thought from the earlier one. (Boghossian 2014, 4) Boghossian drops the causal strategy without further discussion. According to him, the depressive’s judging about suffering fails to be inferred because what he calls the “Taking Condition” is not satisfied. The Taking Condition says that [Taking Condition] Inferring necessarily involves the thinker taking his premises to support his conclusion and drawing his conclusion because of that fact. (Boghossian 2014, 5; cf. 2008, 268) The Taking Condition has two clauses. First, in inferring, the person has to take their premises as a reason for believing the conclusion (cf. 2014, 12); second, their taking has to control the transition to the conclusion in some way (“control” is taken from Wright 2014, 31). The first clause is a rationality condition insofar as it would not be rational for a person to infer p and at the same time to think that their premises do not support the conclusion.15 The second clause is a guidance condition. Boghossian regards it as a condition of adequacy for an account of inference that it accommodates this condition. The Taking Condition, however, is problematic. What it says exactly is unclear, because “taking” can be interpreted in different ways. One

Sellars on Inference

163

possibility is to read it in the sense of “treating”, and to read the Taking Condition in such a way that somebody who infers treats their premises as supporting the conclusion (cf. McHugh and Way 2018, 170f.). But this would say no more than that one draws the conclusion out of the premises; it would not represent an additional condition but simply redescribe the phenomenon that is to be explained. Another possibility is to conceive of taking as a kind of representing, for example as a judgment or intuition. On this reading, the Taking Condition is not obviously true. One obvious worry concerns the resulting over-intellectualization. In order to fulfill the Taking Condition read in this way, someone who infers must possess concepts like premise, inference, and support or reason. While it is plausible to attribute to children the capacity to infer, it is not plausible to impute to them the possession of these concepts. It would be completely unclear what content these concepts would have to have in order for children to be able to master them.16 Besides, it is unclear why one should accept the Taking Condition as an adequacy condition. What does speak in its favor? Boghossian thinks that the following considerations are enough to establish the Taking Condition: In the relevant sense, reasoning is something we do, not just something that happens to us. And it is something we do, not just something that is done by sub-personal bits of us. And it is something that we do with an aim—that of figuring out what follows or is supported by other things one believes. (Boghossian 2014, 5) The last proposition in the quotation is the decisive one. We saw in Section 2 that Boghossian understands inference to be a case of voluntary action. Here he describes it as action in the service of a particular aim. This assumption is indeed sufficient for the Taking Condition. If someone does an action in the service of a certain aim, then first they consider the action to be suited to the aim and second they do it on just that account. This corresponds to the two clauses of the Taking Condition. If inference is action with the aim Boghossian indicates, then a person who infers takes the conclusion to be something that is supported by the premises and draws the conclusion on just that account. The Taking Condition would then be satisfied. We can abstract from the question whether inferring, if it did indeed amount to action directed at some aim, would have just the aim that Boghossian indicates. What is more important is the basic question whether inferring can ever be said to be action in the service of an aim. In light of Sellars’s criterion for actions, it is clear that this is not the case. Let us try to decide to draw a certain conclusion from given premises and implement that decision. There are exactly three ways we might evaluate the situation before making the attempt: we can judge that the

164

Johannes Hübner

conclusion can indeed be drawn from the premises; we can judge that the conclusion cannot be drawn from the premises; or we may make no judgment concerning the question whether the conclusion can be drawn. In the first two cases, we can at best decide to utter or think a sentence of the form “p, therefore q” and then implement that decision. In the first case, we will take the sentence to be true, in the second false, but in neither case will we infer q from p. In the third case we will certainly be able to make a decision to reflect on whether q follows from p, and implement that decision. It is possible that in the course of the reflection we will in fact infer q from p. But if we do then infer q from p, that does not mean that the inference is an action. The reflection is admittedly a mental action, but the acts in which it consists are not themselves actions. Since actions are made up of component acts, they must ultimately consist in acts that are not actions. Otherwise every action would be made up of infinitely many actions. Thus in none of the three possible cases is inferring an action. The upshot of these considerations is that inferences are not actions and thus that Boghossian’s argument for the Taking Condition fails.17 I do not think that there are any good reasons for accepting this condition. Even though Boghossian takes it for granted, his own considerations effectively undermine it. He tries out several interpretations that might suffice to accommodate it and finds devastating problems for every single suggestion. This may be viewed as a reductio ad absurdum of the Taking Condition. I will confine myself to the rule-following proposal because it is widely held and also accepted by Sellars. The rule-following proposal has it that the transition from one thought to another is an inference if and only if it is guided by inference rules. Boghossian (2014, 12) argues that if one accepts the rule-following proposal, one takes account of the Taking Condition. This seems correct. If someone follows a rule that entitles to do φ under circumstances C, then according to Boghossian, they take the circumstances as a reason to do φ, and do φ for that reason. In cases where the rule is a rule of inference, they take certain premises as a reason to draw a certain conclusion and draw that conclusion for that reason. That is exactly what the Taking Condition demands. Boghossian discusses two versions of the rule-following proposal. The intentional view of rule-following (2014, 12–14) postulates the existence of an intentional state that represents the rule. The dispositional view says that to follow a rule is to be disposed to conform to the rule (under appropriate conditions) (2014, 14). Let us consider the intentional view. (I will return to the dispositional view at the end of my chapter.) The problem Boghossian develops for the intentional view of rule-following is identical with the regress problem that arises, as we have seen, on the assumption that all rules, including inferential rules, are rules of action.

Sellars on Inference

165

Boghossian calls it the “Inference Problem for Intentional Views of rulefollowing” and puts it thus: On the one hand we have the Intentional View of rule-following, according to which applying a rule always involves inference. On the other hand, we have the Rule-Following picture of inference according to which inference is always a form of rule-following. These two views, however, can’t be true together. (Boghossian 2014, 14) Since he is unable to solve this problem (and since he thinks that the other proposals for meeting the Taking Condition fail), Boghossian is “firmly” inclined “to take the notion of following a rule as an unanalyzable primitive” (2014, 17). Sellars’s distinction between two kinds of rules rescues us from this unfortunate situation. The intentional view of rule-following corresponds to Rules of action only and the rule-following picture of inference corresponds to the Rule assumption about inferring. The Sellarsian way out should be clear. Boghossian’s claims should be modified in light of Sellars’s distinction; Boghossian should say: Applying a rule of action always involves inference; inference is always a form of rule-governed activity, viz. of being guided by a rule of criticism. In this way no regress ensues. This result was to be expected; after all, the whole point of Sellars’s distinction was to avoid regresses and vicious circles.18 The result emphasizes how important it is to take Sellars’s distinction to heart and to avoid conceiving of inference as intentional action.19 Taken by itself, however, it is not yet an answer to the question: what is inference? Sellars’s answer to this question will be presented in the following section.

5 The Sellarsian Account of Inference Let us take a look back. Why is the depressive’s association of fun with suffering not a case of inference? According to Sellars’s diagnosis, the associative disposition of the depressive person is not due to a rule of inference. Sellars develops his own account of inference most fully in “Actions and Events”. Still, it is rather sketchy. Some well-known points are presented in detail, while the crucial points are stated only briefly. Sellars combines a causal-dispositional analysis of inference with a rulefollowing proposal. His strategy is to look at the origin of inferential dispositions. The crucial point is that the depressive’s associative disposition

166

Johannes Hübner

is not due to an inferential rule (SRLG §53: 341; MFC 87). This point combines two Sellarsian thoughts, the first of which I will call, for want of a better name, the [Genetic Thought] Whether the actualization of a disposition to move from one thought to another is a case of inference depends on the genesis of the disposition. The Genetic Thought does not impose requirements on what is going on within the mind when a person infers. Inferences can, but need not be associated with a feeling of constraint, for example, or with a higher order judgment about the logical rigor of one’s own thinking. Nor does the rule of inference need to become an object of thought; making it into one would mean assuming an additional premise. Instead, Sellars emphasizes the history of the person’s learning. What matters is how they acquire their dispositions to move from one thought to another. Of course, the Genetic Thought is not a self-evident truth. One may regard it as a lesson from the reflections we have undertaken above. On pain of regress, one should not postulate the presence of further premises in order to explain inference (cf. AAE §18: 183f.). The second thought is that there must be an appropriate connection to an inferential rule. An activity is a case of inference only if it is guided by an inferential rule. The following criterion of adequacy for rule-guidance is neutral enough to be uncontroversial: If P φ’s in C and is guided by the rule “in C, one ought to φ”, it must be correct (in some way) that P φ’s in C because the rule says that one ought to φ in C. If P infers that q and is guided by an inferential rule, it must be true that P moves to the thought that q because the rule says that one ought to do so. At this general level Boghossian agrees. He thinks that a certain reaction is a case of rule-following only if the person reacts in that way “because of some appropriate relation” they bear to the rule (Boghossian 2008, 118). The tricky question is which relation is appropriate. If one puts the two thoughts together, one gets the Sellarsian answer: the actualization of a disposition counts as inference only if its genesis is traceable to an inferential rule. The crucial point is that rules of inference implement inferential dispositions. Sellars makes this implementation claim with regard to material inferences. He says that “it is because that-p entails that-q, that a thought that-p is, ceteris paribus, followed by a thought that-q” (AAE §17: 183). This “because” has to be spelled out. As Sellars says, one has “to understand how, in the case of inference, the causal relationship [between premise thought and conclusion thought] involves the fact that the premise entails the conclusion—or even the

Sellars on Inference

167

belief that the premise entails the conclusion” (AAE §22: 185). If one succeeds in doing so, one has an account of inferring. Sellars’s conception can be spelled out in three steps. (I) Sellars regards statements of the form “that-p implies that-q” as normative statements (AAE §30: 188). In his view, they express rules of criticism. For this reason, his implementation claim should be understood as a statement to the effect that a situation in which rules of inference have validity leads to sequences of thought that stand in conformity with the rules. One need not share Sellars’s identification of statements about logical relations with rules of inference. Even if one does not make this identification, one can assume a close relationship between the two: since a certain pattern of logical relationships obtains, and since in a community attitudes that stand in such relationships are supported, in that community validity is possessed by rules of inference that demand agreement with the logical patterns. If we apply the implementation claim to our particular case, modus ponens, it says the following: Because someone’s thought that “p and (if p then q)” ought to be accompanied by their thought that q, thoughts “p and (if p then q)” are usually accompanied by thoughts that q. Because someone’s thought that “p and (if p then q)” ought not to be accompanied by their thought that not q, thoughts “p and (if p then q)” are usually not accompanied by thoughts that not q. More generally: because such and such ought to be the case, such and such is usually the case. The implementation claim is a special case of a well-known Sellarsian thesis: “Espousal of principles is reflected in uniformities of performance” (TC 216; cf. LRB §17: 123). This thesis is very plausible for the espousal of rules of meaning in general and inferential rules in particular. MP, the inference rule corresponding to modus ponens, is espoused within a community if the community’s language contains an expression with the meaning which “if . . . then” has in English. The meaning of “if . . . then” is tied to its function within the speakers’ talk and thought. It is a matter of meaning that utterances or thoughts of the form “p, and (if p, then q)” tend to be followed by utterances or thoughts that q, and tend not to be followed by utterances or thoughts that not q. Otherwise, “if .  .  . then” would have a different meaning than the one it has. Therefore, Sellars claims, it is a conceptual truth that there are regular transitions according to modus ponens and other rules (AAE §16: 183; §29: 187; §50: 194). It remains to be explained how inferential habits are established.

168

Johannes Hübner

(II) Sellars’s answer, of course, refers to language acquisition. Trainees learn to use “if . . . then” because trainers see to it that public reasoning of trainees conforms to MP (MFC 87).20 The trainers follow the rule of action that they ought to correct deviations and reinforce conformity. They either recall the relevant rule of criticism (RM 490) or they are at least sensitive to infringements even if they can’t make the rules explicit. In the beginning of the history of an individual’s learning, explicit acknowledgment of the rules and mere conformity with them in practice is distributed across different individuals. Via the activity of the trainers, MP genetically explains why a thinker has the disposition to infer in accordance with modus ponens, and it thereby explains why a thinker infers in that way in a particular situation. (III) Later on, trainees learn to make explicit the rules with which they and others ought to conform when they infer, or they at least learn to identify as errors their own and others’ divergences from the rules. They acquire the readiness to listen to criticism in cases of suspected infractions of the rules and to either accept or reject it.21 This gives rise to a rational explanation for the maintenance of inferential dispositions. Reflective thinkers criticize each other and monitor their own inferential behavior. Good patterns of inference are confirmed while bad patterns are eliminated (RM 493; cf. AAE §32: 188). The standard of criticism is whether conformity with a given pattern of inference leads to sequences of thought in which the truth of their premises does in fact argue the truth of their conclusions. Since MP corresponds to a valid logical pattern, there is no need for it to be corrected. It is not necessarily the case that every inferential rule that is in force in a community is also correct (cf. Broome 2014; Boghossian 2008, 118). Invalid reasoning can certainly take place, for example according to the pattern “if p, then q; q; therefore p”. Yet it will not be possible to go beyond a certain proportion of invalid inferences while the sequences of thought in question remain identifiable as inferences. Let me sum up: The Sellarsian account of inferring combines a rulefollowing proposal with a subtle dispositional analysis. It states: If a person P moves from the thought that p to the thought that q, they infer q from p iff their transition from p to q is guided by a rule of inference. The transition is guided by a rule of inference if it is the actualization of a disposition to behave in a way that conforms to the rule of inference; and if this disposition has been acquired through linguistic training and strengthened or at least not eliminated by criticism. The Sellarsian account meets the uncontroversial condition of adequacy for rule-guidedness, according to which a move is guided by a rule only

Sellars on Inference

169

if it occurs because a rule says that it ought to occur. The content of the inferential rule explains the formation of a corresponding inferential disposition because the linguistic trainers, keeping the rules in view, have trained their trainees in behavior conforming to them; and it explains the maintenance of the disposition because the disposition and its actualizations are criticized by reference to the rules. The acquired disposition, in its turn, explains why persons draw inferences in certain ways in particular situations. Thus the Sellarsian account enables us to say that an inference is drawn because an inferential rule says that it ought to be drawn. We can make a concluding point in favor of the Sellarsian account if we consider an objection that Boghossian lodges against a simplistic dispositional conception of rule-following. According to the simplistic view, rule-following consists in the disposition to act in accordance with a rule under suitable conditions. According to the objection: “the rules we follow explain why we have the dispositions we have. . . . No such thing could be true on a Dispositional View, though, since our dispositions can’t explain themselves” (2014, 14; cf. 2008, 132). The explanatory condition, that rules should explain dispositions, can certainly be accepted. The relevant point is that the Sellarsian account meets it. From Sellars’s perspective, we simply need to differentiate the “we” and to take account of the fact that different members of our community play different roles in the formation of inferential dispositions. The rules by which those who teach the language orient themselves explain why the trainees have the inferential dispositions that they have—without the need for anything to explain itself. The Sellarsian account of inferring is typical of Sellars in that it steers a middle course between a simple dispositional view and an intellectualistic view as expressed by Boghossian’s Taking Condition. The content of rules is involved in explaining how a person is guided by inferential rules, but only indirectly and not directly via a representation of the content that guides the inference.22 (Translation by Marc Hiatt)

Notes 1. By contrast, Pink (2009, 116f.) conceives of decisions and intentions as goaloriented actions. According to him, the decision to take an action relates to the action, as the means does to the goal. However, it seems difficult to me to understand the relationship between the decision to do something and the doing of it as a relationship between means and goal. A means is selected because of its suitability with regard to a given goal. On the other hand, somebody determines the goal in the first place by making a decision. 2. To give a terminological emphasis to the peculiarity of rational mental acts, one could introduce a broad sense of “action” by stating that φ is an action if one can have reasons for φ. Considerations such as those put forward in Burge’s (1998, 250f.) elaboration of a “broad notion of agency” could make such a choice seem sensible. Actions in Sellars’s sense would then be a species

170

Johannes Hübner

of actions in this broad sense. Taking such a route wouldn’t change anything with regard to the following important point: whether mental or not, actions that are cases of rule-following must be actions in the sense of Sellars’s criterion. 3. One could object that many more mental acts are actions than Sellars admits, and argue that actions can be described with reference to their consequences. For example, if a person A flicks a switch and in consequence a light turns on and person B wakes up, then A has turned on the light and woken B. The turning on of the light and the waking are A’s action. That suggests that the following formula is correct: If P carries out an action that leads to P φ-ing, then φ is an action done by P.

4. 5. 6.

7.

8. 9.

If a φ that is brought about by an action is mental, then according to the formula it is also a mental action. This has an effect on the classification of those mental acts that are not actions according to Sellars. Since a person can bring it about by action that they perceive something particular or make a judgment or a decision, perceptions, judgments, and decisions can be actions according to the formula.—The answer to this objection is that the formula is not correct. Someone can perform an action that results in them doing φ without φ being an action. To borrow an example from Mele (2009, 18f.): if someone counts sheep in order to fall asleep, and then actually falls asleep, then falling asleep is not an action. Only bringing the falling asleep about is an action. Thus bringing the doing of φ about can be an action even if φ is not an action. This holds in the case of perceptions, judgments, and decisions. Following Sellars, I use “obey a rule” without the implication that obeying a rule ensures conformity to the rule. In this sense, a person might make a mistake in obeying a rule. This point has also been made by Boghossian (2014, 13) and Wright (2007, 490f.), though in a different terminology. As I see it, SRLG points towards the distinction between two types of rules but does not yet make it. What SRLG does make is a distinction between pattern-governed behavior and rule-obeying behavior. Within SRLG, no form of pattern-governed behavior is classified as a kind of rule-governed behavior although it is recognized that behavior may non-accidentally be conforming to rules without being the obeying of rules (SRLG §12: 325). At this time of his writing, Sellars seems to assume that pattern-governed behavior could be rule-governed only if it was rule-obeying. He says: “Pattern governed behaviour of the kind we should call ‘linguistic’ involves ‘positions’ and ‘moves’ of the sort that would be specified by ‘formation’ and ‘transformation’ rules in its metagame if it were rule obeying behaviour” (SRLG §17: 327). By Sellars’s later lights this is wrong. As he points out most clearly in MFC (86–9), the formation and transformation rules of SRLG are rules of criticism pertaining exclusively to pattern-governed behavior. Therefore, if linguistic behavior of any kind was rule-obeying behavior, it could not be covered by formation and transformation rules. Sellars began working on SRLG three months after having received a copy of Wittgenstein’s Philosophical Investigations (thanks to Stefan Brandt for this note). It is also possible that the well-known regress arguments presented by Ryle (e.g., [1949] 2000, 67), Carroll (1895), and Quine ([1935] 1976, 103f.) sensitized Sellars to regress problems. For a dissenting voice, cf. Valaris (2017, 2012–16). For further arguments, cf. Boghossian (2014, 11f.).

Sellars on Inference

171

10. In what follows, I will drop the bracketed condition.—An alternative formulation instead of MP is: [MP*] If one is entitled to an assumption of the form “p and (if p, then q)”, one is entitled to judge that q.

11.

12. 13.

14.

15. 16. 17.

18.

I will stick to MP. For my present purposes the formulation does not matter because I am concerned with the problem of regress and a regress arises either way. Of his distinction, Sellars (SM III §37: 76) says that it “is essentially akin to that which has been drawn between ‘ought-to-do’ and ‘ought-to-be’”. With this remark, Sellars is presumably referring to several authors. Moore ([1903] 1993, 198) describes “rules of action” as “statements not of what ought to be, but of what we ought to do”. In an essay reprinted in a volume co-edited by Sellars, Pritchard ([1912] 1952, 152) explains: “The word ‘ought’ refers to actions and to actions alone. The proper language is never ‘So and so ought to be’, but ‘I ought to do so and so’”. Besides these passages, we can also be reminded of the development of deontic logic. In von Wright’s (1951) early deontic system the deontic operator “O” is connected with act-predicates and can be glossed as “one ought to” or “it is obligatory to”. By contrast, in what is today known as Standard Deontic Logic the deontic operator “O” is used as a sentential operator that can be glossed as “it ought to be the case that” or “it is obligatory that”. As Sellars notes, only “an important subclass” of rules of criticism are formulated in this way (LTC 59). Further rules of criticism concern permissions and constraints (LTC 64 n. 4; MFC 86). Following Wittgenstein, Wright (2007, 497) speaks of a primitive form of rule-following that is “blind” insofar as it takes place “without reason”. In so doing, Wright approaches the Sellarsian thesis that there is a form of rulefollowing that consists in something other than obedience to rules. However, Wright fails to take the decisive step, since he maintains that even this primitive kind of rule-following is intentional action (2007, 497f.; 2014, 36). Sellars’s distinction has not, on the whole, been intensively studied and applied. One exception is, e.g., Chrisman (2008). Chrisman interprets “ought to believe” as “ought to be” in Sellars’s sense and thus side-steps the debate between doxastic involuntarists and epistemic deontologists. However, a negative condition would suffice to secure this point: the person doesn’t believe that the premises don’t support the conclusion. Cf. Hlobil (2014) for discussion. For this and other criticisms of the Taking Condition, cf. McHugh and Way (2016). Valaris (2014, 102) defends the following version of the Taking Condition: “If one believes p by reasoning from R, one believes that p follows from R”. But his argument (112–18) suffices only to support a weaker dispositional condition: “If one believes p by reasoning from R, one is disposed to believe that p follows from R”. Valaris himself effectively admits this (2014, 116 n. 20). The weaker condition is acceptable for sufficiently sophisticated, conceptually competent thinkers. However, it is too weak to ground Valaris’s constitution thesis, according to which the forming of a belief p by drawing it as a conclusion from reasons R consists in the assumption that R and that p follows from R (2014, 121). The approach taken by Valaris (2014) avoids the regress in a different way. Valaris defends the Taking Condition. He maintains that someone who derives a belief p by inferring it from R believes that p follows from R.

172

19.

20. 21. 22.

Johannes Hübner According to Valaris, this higher order belief is not an additional assumption. Instead it constitutes the inference: “Explicitly reasoning from R to p just is, in part, believing that p follows from R” (118). In my view, this is to assume that inference is more sophisticated than in fact it is. From the standpoint of this result, the conception of reasoning defended by McHugh and Way appears to leave the issue undecided. On the one hand, they understand reasoning as intentional action, namely as an activity by which thinkers pursue the goal of procuring for themselves “fitting attitudes” and which is regulated by this goal (2018, 177–80). On the other hand, the goal is supposed to guide inference without any necessity for the goal or its implementation to be an explicit topic of concrete cases of reasoning. Guidance by the goal is said to ensue via rules of inference that one follows while reasoning (180–2), and this rule-following is, in its turn, somehow supposed to be understood dispositionally. But McHugh and Way do not clarify what, exactly, it means to follow a rule. Following widespread usage, I make no difference in meaning between “reasoning” and “inferring”. For the thought that rule-following involves the possibility of correction, cf. Wright (2007, 498), Broome (2014, 22). I would like to thank the editors and the participants of the Erlangen Sellars Conference for helpful comments.

References Boghossian, Paul. 2008. Content and Justification: Philosophical Papers. Oxford: Oxford University Press. ———. 2014. “What Is Inference?” Philosophical Studies 169: 1–18. Broome, John. 2014. “Comments on Boghossian”. Philosophical Studies 169: 19–25. Burge, Tyler. 1998. “Reason and the First Person”. In Knowing Our Own Minds, edited by Crispin Wright, Barry C. Smith, and Cynthia Macdonald, 243–70. Oxford: Oxford University Press. Carroll, Lewis. 1895. “What the Tortoise Said to Achilles”. Mind 4: 278–80. Chrisman, Matthew. 2008. “Ought to Believe”. Journal of Philosophy 105: 346–70. Hlobil, Ulf. 2014. “Against Boghossian, Wright and Broome on Inference”. Philosophical Studies 167: 419–29. McHugh, Conor and Jonathan Way. 2016. “Against the Taking Condition”. Philosophical Issues 26: 314–31. ———. 2018. “What Is Reasoning?” Mind 127: 167–96. Mele, Alfred. 2009. “Mental Action: A Case Study”. In Lucy O’Brien and Matthew Soteriou, eds. 2009. Mental Actions, 17–37. Oxford: Oxford University Press. Hereafter O’Brien and Soteriou 2009. Moore, George E. (1903) 1993. Principia Ethica, revised edition, edited by Thomas Baldwin. Cambridge: Cambridge University Press. Pink, Thomas. 2009. “Reason, Voluntariness, and Moral Responsibility”. In O’Brien and Soteriou 2009, 95–120. Pritchard, Harold. (1912) 1952. “Does Moral Philosophy Rest on a Mistake?” In Readings in Ethical Theory, edited by Wilfrid Sellars and John Hospers, 149–62. New York: Appleton-Century-Crofts. Quine, Willard Van Orman. (1935) 1976. “Truth by Convention”. In The Ways of Paradox and Other Essays, revised and enlarged edition, 77–106. Cambridge, MA: Harvard University Press.

Sellars on Inference

173

Ryle, Gilbert. (1949) 2000. The Concept of Mind, with an Introduction by Daniel C. Dennett. Chicago: University of Chicago Press. Sellars, Wilfrid. (1949) 2005 [LRB]. “Language, Rules and Behavior”. In Pure Pragmatics and Possible Worlds: The Early Essays of Wilfrid Sellars, edited by Jeffrey F. Sicha, 186–214. Atascadero, CA: Ridgeview Publishing. ———. (1954) 1991 [SRLG]. “Some Reflections on Language Games”. In Wilfrid Sellars. (1963) 1991. Science, Perception and Reality, 321–58. Atascadero, CA: Ridgeview Publishing. Hereafter SPR. ———. (1962) 1991 [TC]. “Truth and ‘Correspondence’”. In SPR, 197–224. ———. 1966 [FD]. “Fatalism and Determinism”. In Freedom and Determinism, edited by Keith Lehrer, 141–74. New York: Random House. ———. (1967) 1992 [SM]. Science and Metaphysics: Variations on Kantian Themes. Atascadero, CA: Ridgeview Publishing. ———. (1969) 2007 [LTC]. “Language as Thought and as Communication”. In Wilfrid Sellars. 2007. In the Space of Reasons: Selected Essays of Wilfrid Sellars, edited by Kevin Scharp and Robert B. Brandom, 57–80. Cambridge, MA: Harvard University Press. Hereafter SR. ———. 1973 [AAE]. “Actions and Events”. Nous 7: 179–202. ———. 1973 [RM]. “Reply to Marras”. Canadian Journal of Philosophy 2: 485–93. ———. (1974) 2007 [MFC]. “Meaning as Functional Classification”. In SR, 81–100. ———. (1981) 2007 [MEV]. “Mental Events”. In SR, 282–300. Strawson, Galen. 2003. “Mental Ballistics or the Involuntariness of Spontaneity”. Proceedings of the Aristotelian Society 103: 227–56. Valaris, Markos. 2014. “Reasoning and Regress”. Mind 123: 101–27. ———. 2017. “What Reasoning Might Be”. Synthese 194: 2007–24. von Wright, Georg Henrik. 1951. “Deontic Logic”. Mind 60: 1–15. Wright, Crispin. 2007. “Rule-Following without Reasons: Wittgenstein’s Quietism and the Constitutive Question”. Ratio 20: 481–502. ———. 2014. “Comment on Paul Boghossian, ‘What Is Inference’”. Philosophical Studies 169: 27–37.

9

Sellars, Truth Pluralism, and Truth Relativism Lionel Shapiro

Wilfrid Sellars wrote about truth throughout his career, but his views appear to have had little impact.1 In part, this may be due to the implausibility of the positions he has been represented as defending. For example, Richard Rorty includes him among those who “identify ‘true’ with ‘warranted[ly] assertible by us’” (1979, 296), while Hilary Putnam describes him as “identifying truth . . . with warranted assertibility in the ideal limit of scientific investigation” (1978, 36). In fact, Sellars tried to ward off such interpretations: It might be thought that I am offering something like the “warranted assertability” theory of truth. . . . But to make this move is to confuse truth with probability, for presumably to be warranted is to be warranted by evidence. (NI 664; cf. NAO 4 §95: 85) It won’t help to invoke an ideal limit: as I’ll argue later, Sellars doesn’t explain truth in terms of evidence at all.2 A related reason for the lack of attention to Sellars’s views may be the impression, endorsed by Donald Davidson ([1969] 1984, 50–1), that he failed to grasp the importance of Alfred Tarski’s work.3 According to Davidson, what he missed is how Tarski ([1933] 1983) vindicated a “correspondence theory” of truth by explaining truth in terms of the relation of satisfaction between open sentences and sequences of objects (Davidson [1969] 1984, 48). There’s irony to this rebuke, since the claim that Tarski thus vindicated a correspondence theory is one Davidson would come to firmly repudiate ([1977] 1984, 218, 223–4).4 Though Sellars’s approach to truth hasn’t been influential, it bears closely on topics of current interest. This chapter reexamines his work through the lens of recent proposals, with the dual aim of bringing into focus aspects of his position and broadening our view of the terrain by adding his perspective. As we’ll see, Sellars adopts versions of both truth pluralism and truth relativism. Moreover, he shares the commitment that motivates most of their current advocates, a commitment to recognizing

Sellars, Truth Pluralism, and Truth Relativism

175

heterogeneity within our truth-apt discourse.5 However, what accounts for his own adoption of these doctrines turns out to be something very different: his interest in “conceptual structures” other than ours. That motivation yields an interesting and distinctive form of truth relativism, but I’ll argue that Sellars goes wrong when he takes his pluralism about conceptual structures to pave the way for truth pluralism. Contemporary truth pluralism emerged in Crispin Wright’s Truth and Objectivity (1992), where it’s introduced to explain how some discourses can be truth-apt yet display failures of objectivity. Contemporary truth relativism likewise originated as a way of accommodating subjective discourses: one harbinger was Max Kölbel’s Truth Without Objectivity (2002). In Sections 1–3, I examine the evolving ways in which Sellars’s own interest in the heterogeneity of discourses shows up in his views about truth. He eventually embraces a version of truth pluralism that overcomes a challenge confronting pluralists, albeit at a serious cost.6 That’s because the key to his pluralism is his unsatisfactory analysis of truth as “semantical assertibility”. Section 4 argues that Sellars adopts that analysis in order to formulate a semantic notion that remains applicable to discourse, including “matter-of-factual discourse”, even when it embodies conceptual structures other than ours. In Section 5, I explain how the same motivation leads him to truth relativism. Section 6 argues that despite their common motivation, Sellars’s relativism is independent of, and indeed stands in tension with, the analysis of truth that underlies his pluralism. (Two appendices are included: the first puts Sellars’s relativism to use in making sense of his rejection of bivalence; the second addresses the relation between Sellars’s and W.V. Quine’s views of truth.)

1 Truth Pluralism: Points of Contact Recent truth pluralists share three core commitments with Sellars: (i) “True” applies univocally to propositions expressed in all domains of discourse. (ii) Still, truth behaves differently in different domains. (iii) One domain features propositions whose truth requires a relation of “representation” or “correspondence” to hold between linguistic or mental items and items in the world. Admittedly, commitment (i) is merely implicit in Wright (1992). But Wright has since clarified that on the proposal he intended, “true” expresses a “single concept” (2013, 128). Likewise, Sellars claims that “all true statements of whatever kind are true in the same sense of ‘true’” (TC 223; SM V §1: 116, IV §30: 102). He isn’t equivocating when he says some of his claims “apply . . . to all truths, whether empirical or mathematical or metaphysical or moral” (TC 206–7).

176

Lionel Shapiro

As for (ii), Wright holds that despite this univocity “one wants to think differently about the import of ‘true’” as it applies to discourse about the comic and about mathematics (2013, 123). Similarly, Michael Lynch explains that “the pluralist about truth begins with the intuition that there is more than one way for propositions to be true” (2009, 82). Sellars shares this intuition that truth “takes specific forms” or “modes” in different discourses (SM V§1: 116, §9: 119; Sellars-Harman Correspondence, Feb. 26, 1970). He speaks of the different “essential features of different varieties of truths” (TC 198), and later of “the varieties of truth” (SM IV §26: 101, §31: 102; NAO 4 §94: 85). But it’s with respect to (iii) that Sellars most strikingly anticipates Wright. Wright argues that in some domains, truth involves a “correspondence” more substantial than anything that can be extracted from the claim that a truth bearer is true just in case things are the way it says they are. He concedes that this “correspondence platitude” is captured by the “equivalence schema” according to which the proposition that p is true just in case p (Wright 1992, 24–5). Anyone who recognizes that schema as essential to truth will have no difficulty accommodating intuitions about the relationship between truth and correspondence so long as doing so is held to require no more than demonstrating a right to the phrases by which those intuitions are characteristically expressed. (Ibid., 27) But Wright adds that the equivalence schema fails to supply an “intended further substantive content” of correspondence theories. There’s something “more substantial there, struggling for an outlet” (ibid., 143). All this could have been written by Sellars 30 years earlier. According to Sellars, the philosophical content of Tarski’s “semantical account” of truth is exhausted by the equivalence schema. Since that schema is common ground to all theorists, it fails to vindicate the correspondence theorist: But what comfort can the correspondence theorist take in a victory which, to all appearances, reduces his claim to a formula which . . . has, according to his erstwhile antagonists, nothing to do with the philosophical problem of truth? (TC 198) This raises the central question of Sellars’s paper: Is there a place for a correspondence theory of truth (some truth, at least) in addition to the semantical account? Or, at least, for a theory that finds a nontrivial job for a “correspondence” other than

Sellars, Truth Pluralism, and Truth Relativism

177

the correspondence which finds its expression in the familiar equivalences of semantic theory? (TC 198) Wright and Sellars both answer affirmatively. According to Wright, the “intended substance” of correspondence theories is the idea that . . . talk of “representation of the facts” is not just admissible phrasing, a harmless gloss on talk of truth, but incorporates a philosophically correct—as we might say, seriously dyadic— perspective on the truth predicate (at least for discourses where realism is appropriate). (Wright 1992, 83) For such domains, there’s “real substance in the idea that . . . we function as representational systems, responsive to states of affairs which, when we are successful, our beliefs and statements serve to map” (ibid., 147). Sellars agrees. In his view, “the most likely domain for a substantive correspondence theory of truth . . . is the domain of empirical or matterof-factual truth” (TC 198). Solving “the problem of empirical truth” requires going beyond the equivalence schema, which doesn’t point to a “relational property of statements” (TC 206). Doing so requires recognizing a relational property of “correct picturing”, defined in terms of an isomorphism between a certain “structure of natural-linguistic objects” (true sentence tokenings, considered in abstraction from their semantic features) and a “structure of nonlinguistic objects” (TC 222). Sellars comes to describe this relational perspective in the very language later used by Wright, in terms of “representational systems” and what they “map” (NAO 5 §§37, 42: 104–5).

2 Varieties of Truth or Varieties of Truths? Despite these strong points of contact with Wright’s truth pluralism, Sellars isn’t clear in “Truth and ‘Correspondence’” (1962) about whether he’s proposing a genuine truth pluralism. His comments are consistent with a weaker and a stronger version of commitment (ii). On the stronger version, truth amounts to something different in different discourses— there are differences between, e.g., moral and empirical truth. This is a natural way to understand his suggestion that there’s “a place for a correspondence theory of . . . some truth”. Yet he concedes it would suffice to “at least” find some “job” for correspondence, even if it doesn’t count as a kind of truth. This suggests a weaker version of (ii), according to which there’s only one kind of truth, but different kinds of truths (true statements). In particular, true statements of empirical discourse enjoy a relational property of correspondence, whereas those of moral discourse

178

Lionel Shapiro

don’t. Sellars’s talk of “different essential features of different varieties of truths” (TC 198), and of correspondence as characteristic of “a significant variety of truths” (TC 197), likewise suggests the weaker view (note his use here of “truths” rather than “truth”). Summarizing the paper’s conclusions, he speaks of “finding a mode of ‘correspondence’ other than truth that accompanies truth in the case of empirical statements” (TC 222). Hence there’s good reason to deny that Sellars is distinguishing varieties of truth, as opposed to just claiming that different properties accompany truth in different discourses. Compare: triangles come in different varieties, e.g., isosceles and scalene, with different characteristic properties. This doesn’t show that being a triangle amounts to something different for isosceles and for scalene triangles; it doesn’t yet show that there are different varieties of triangularity. As Michael Williams aptly observes (2016, 232), “acknowledging that there are different kinds of truths is no reason to introduce domain-specific concepts of truth”, or (I would add) domain-specific truth properties.7 The same ambiguity attaches to Wright’s initial discussion of “a pluralist view of truth” (1992, 23n, 25, 38). Subsequently, Wright resolves it in favor of the stronger position. He describes his “pluralistic conception of truth” as one that “allows us to think of truth as constituted differently in different areas” (1999, 225), i.e., as “variably constituted” (1996, 926). Pluralism denies “that truth everywhere must possess a uniform constitution: that the truth of any true proposition always consists in the same sort of thing” (1999, 224; 1996, 924). Pluralists following Wright hold that there’s a single property of truth, but each true proposition instantiates it in virtue of instantiating some further property.8 No one such property is available to perform this work for propositions in discourses as diverse as morality, mathematics, and empirical science. To understand where truth pluralism goes beyond Sellars’s conclusion in 1962, let’s look at what Wright and Sellars say about the concept and property of truth. We find two parallels. First, both hold that the concept is implicitly defined by the equivalence schema. Wright explains: [I]f we are dealing with a range of genuine contents . . ., then nothing can stand in the way of the definitional introduction of a predicate . . . which is subject to the equivalence schema: that p is φ if and only if p. (Wright 1999, 230n; 1992, 74, 96) Such a predicate will express “the concept of truth” (1999, 226, 229–30). Similarly, Sellars concludes that “the word ‘true’ gets its sense from” the “principle of inference” that governs the inference from “that p is true” to “p” and its converse, and thereby explains the apriority of “the necessary equivalences highlighted by the semantic theory” of Tarski (TC 206). Second, both hold that whenever a concept applies to an object, a correlative property is instantiated. Wright employs an “abundant metaphysics of

Sellars, Truth Pluralism, and Truth Relativism

179

properties” on which “properties are essentially tied to well-determined satisfaction-conditions of predicates” (2013, 150, 131–3). Sellars too recognizes such a conception: “one use of the terms ‘property’ and ‘relation’ is such that it is correct to say of any meaningful expression which has the grammatical characteristics of a predicate that it means a quality [sic] or relation” (EAE 451).9 Where Wright goes beyond Sellars is in holding that instantiating truth always “consists in” instantiating some further “truth property”. Wright and followers invoke an “analytic functionalism” inspired by David Lewis (1972).10 On this approach, applied to a predicate “F”, one first identifies a set of “platitudes” expressed using “F” and uses these to specify a “role” playable by a property. Thus Wright’s “correspondence platitude” for “true” specifies a role: a property x plays this role just in case an item y exemplifies x when and only when things are the way y says they are.11 Next, one identifies some property that plays the role—or some properties, each of which plays the role over a different domain of entities. Finally, something’s instantiating Fness is said to consist (perhaps in a local domain) in its instantiating the relevant property. This is far removed from anything in Sellars.12 Moreover, as debates between pluralists show, it’s not clear how the functionalist approach makes sense of truth being variably constituted (Wright 2013). How are we to understand the relation between truth and its domain-specific constituting properties? Pluralists sometimes analyze truth as a “secondorder role property”, that of having some first-order property that plays the truth role over the relevant domain (Lynch 2004). This yields a clear sense in which each true proposition is true in virtue of having some further property. However, it can seem an ad hoc way to ensure truth counts as variably constituted. Why not instead recognize truth as a firstorder property that plays the truth role and enjoys different accompanying properties across different domains?13 Lynch has since proposed that truth is a first-order property “manifested by” domain-specific properties (2009, 74–7). A property Fness other than truth is said to manifest truth for some domain provided it’s a priori that being F plays the truth role over that domain. The problem now is that this seems indistinguishable from a monism on which truth has different a priori accompanying properties across different domains (cf. Caputo 2012, 860–1).

3 Sellars’s Analysis of Truth and his Truth Pluralism Despite sharing key commitments of truth pluralists, we’ve seen, Sellars doesn’t advocate truth pluralism in 1962. However, he soon changes his mind and embraces a distinctive pluralism, one that meets the challenge just raised. Sellars’s truth pluralism results from a new account of the concept of truth. In an endnote added in 1963 to “Truth and ‘Correspondence’”

180

Lionel Shapiro

(TC 224), he retracts his claim there that this concept is constituted by a “principle of inference” that ensures the apriority of the equivalence schema’s instances (L. Shapiro 2014). His new account has two components, elaborated respectively in Chapters III and IV of Science and Metaphysics. First, talk of the proposition that p is analyzed as a way of talking generically about concrete items that mean p. Here meaning p is playing a “linguistic role”; roles are individuated by “semantical rules” governing items that play them. Second, the claim that the proposition that p is true is analyzed as the claim that it’s correct, by the rules of a language containing items that mean p, to assert such items. To make explicit the way an assertion that it’s true that p licenses an assertion that  p, Sellars no longer thinks we need to state the principle of inference “It’s correct to infer that p from the proposition that it is true that p”. Rather, since it’s a claim about correct assertibility, an assertion that it’s true that p already makes explicit such licensing (TC 224; SM IV §28: 101–2). The analysis of truth as “semantical assertibility” (“S-assertibility”) thus furnishes a new explanation of why instances of the equivalence schema are “conceptually necessary”. The account of this conceptual necessity I wish to recommend is that these equivalences “follow” from the “definition” of truth in that for a proposition to be true is for it to be . . . correctly assertible; assertible, that is, in accordance with the relevant semantical rules, and on the basis of such additional, though unspecified, information as these rules may require. (SM IV §26: 100–1)14 This requires elucidation. First, the “relevant semantical rules” apply to transitions: “intra-linguistic” transitions exemplified in inference as well as “world→language” and “language→world” transitions exemplified, respectively, in perception and action (SM IV §61: 114). As Rosenberg (2007, 199) complains, it isn’t clear what correctness according to these rules amounts to. Sellars says little, but the rough idea might be captured by the following reconstruction. A proposition is correctly assertible according to a system of semantic rules provided it’s the last member of some sequence of propositions that starts with one or more propositions whose assertion is permissible in light of a world→language rule, and where each additional proposition in the sequence is such that, in light of some intra-linguistic rule, it’s impermissible to reject it while asserting all the previous propositions.15 Second, to say that a proposition is assertible in accordance with relevant semantic rules and on the basis of relevant “additional information” isn’t to say that someone could justify asserting the proposition by citing those rules and that information. In particular, Sellars isn’t thinking of the “additional information” as information possessed by the asserter,

Sellars, Truth Pluralism, and Truth Relativism

181

even hypothetically. Rather, it’s information an assessor would need in order to determine whether the agent’s assertion is correct in light of the rules. That is to say, the information concerns whether an externalist condition is satisfied (O’Shea 2007, 218 n3; contrast Williams 2016, 239–45). But semantic assertibility isn’t an externalist epistemic notion either: it isn’t an epistemic notion at all. By this I mean that the truth of someone’s assertion isn’t a matter of the asserter’s actual, potential, or counterfactual epistemic standing. To be sure, if one were to assert a true proposition as a result of (i) inferring by the intra-linguistic rules and (ii) exhibiting permissible world→language transitions while recognizing one’s own tendency not to exhibit impermissible ones, then one would enjoy good epistemic standing (EPM §35: 167–8). But this subjunctive conditional doesn’t serve as Sellars’s explanation of what the proposition’s truth consists in. Rather, he explains truth entirely in terms of his externalistically formulated semantic rules. Williams’s construal of Sellarsian truth in epistemic terms, as “justifiability in epistemically ideal conditions” (2016, 239), is responsible for a revealing error. For Sellars, he claims, “an important reason for identifying truth with semantic assertibility is to avoid identifying a proposition’s being true with its being assertible here and now, when perhaps not all the relevant evidence is on hand” (2016, 238). On the contrary, for Sellars a proposition’s truth is its assertibility here and now—that’s why “the cash value of S-assertibility is assertion by us hic et nunc” (SM V §53: 134). But assertibility here and now isn’t a matter of the evidence available here and now.16 Williams rightly points out that Sellars takes his analysis to give rise to a truth pluralism (2013, 131 n1; 2016, 232; cf. also L. Shapiro 2014, 805 n27). Thus Sellars immediately proceeds to introduce “varieties of truth” that we attribute using “specific forms” of the “generic concept”: “True”, then, means semantically assertible (“S-assertible”) and the varieties of truth correspond to the relevant varieties of semantical rule. . . . [V]arieties of semantical correctnesses . . . correlate with the varieties of truth. (SM IV §26: 101, §31: 102) [T]he concept of truth as S-assertibility is universal in its scope, applying to propositions of the most divergent types. On the other hand, as a generic concept it takes specific forms which are functions of the semantical rules which govern these different types of propositions. (SM V §1: 116, SM vii, ix; also NAO 4 §95: 85) As I understand Sellars here, the claim is that a proposition’s truth is a matter of its being assertible according to whatever system of semantic rules governs propositions of its type. Accordingly, each specific truth

182

Lionel Shapiro

property is the property of being assertible according to some particular system of rules. To illustrate his pluralism, Sellars concentrates on two cases. For mathematical propositions, the relevant rules are those of an “axiomatic framework”, whence truth consists in provability (SM IV §62: 115, V §55: 135; see Appendix 1 at the end of the chapter). The rules for atomic propositions of empirical discourse are different. For one thing, they involve world→language and language→world correctnesses. But Sellars’s key claim is that propositions whose assertion these rules license enjoy a special property: items expressing them, considered “as items in the order of causes and effects”, count as “correctly picturing” some configuration of objects (SM V §§8–10: 118–19, §§56–9: 135–7). For such “matter-of-factual discourse” truth consists in assertibility according to rules whose being in force is explained by the fact that items assertible according to them are correct as pictures. The pluralism made possible by Sellars’s new analysis of truth avoids the difficulty raised at the end of Section 2 about the sense in which truth is variably constituted. A proposition’s being true is its having the following property: being assertible by the rules “relevant” to the class of propositions it belongs to. Compare owning the house one lives in. A person has that property in virtue of having the property of owning some particular house. In an unproblematic sense, what it consists in for Alice to own the house she lives in may differ from what it consists in for Brian to own the house he lives in. Likewise, what it consists in for a chess move to be permissible by the rules of the game it’s made in differs from what it consists in for a checkers move to have that same property. In the same way, a proposition will be true in virtue of having some specific truth property.17 While Sellars’s pluralism overcomes one obstacle, it faces another. The problem is his entitlement to the equivalence schema’s instances, e.g., “That snow is white is true ↔ snow is white” (SM IV §§24–9: 100–2).18 These biconditionals must still in some sense “follow from” his account of the concept of truth, though the account is no longer based on a “principle of inference” that derives the biconditionals. The biconditionals are now justified by pragmatic considerations that involve no such semantic rule for “true” (L. Shapiro 2014, 803–4). The left-to-right direction of “That snow is white is true ↔ snow is white” allegedly follows from the fact that “the assertion of the right-hand side of the implication statement is a performance of the kind authorized by the truth statement on the left”. Pressed by Gilbert Harman regarding the right-to-left direction, Sellars claims a “pragmatic implication” arising from the Moorean “incoherence” of asserting a proposition while refusing to assert that it’s correctly assertible (Sellars-Harman Correspondence, Feb. 26, 1970; cf. NI 664–5). Harman’s rejoinder is devastating. He notes a crucial difference between pragmatic implications and principles of inference: the former don’t underwrite conditionals.

Sellars, Truth Pluralism, and Truth Relativism

183

There is a pragmatic implication in either direction between “I believe that p” and “p”. Notice, however, that it is not true that I believe that p if and only if p. Pragmatic implications do not in general support conditional statements. (Ibid., March 24, 1970) In his reply to Harman, Sellars ignores this objection, though there are notes that suggest it did bother him.19 In fact, then, Sellars’s analysis of truth won’t vindicate the equivalence schema. This carries a cost. As Hartry Field argues (2008, 209–10), some of our uses of a truth predicate depend on intersubstitutivity strong enough to yield the schema’s biconditionals. In the conditional “If anything he said is true, someone will have to resign”, the antecedent allows us to simulate an existential generalization over “He said that p, and p” without using quantification into sentence position. But it can do this only because “He said that p, and p” is intersubstitutable with “He said that p, and that p is true” in the antecedent of a conditional. And that’s what the pragmatic implication claimed by Sellars fails to ensure.

4 Alternative Conceptual Structures We’ve now seen how the analysis of truth Sellars endorses starting in 1963 yields a distinctive but problematic pluralism. He no longer merely recognizes different “varieties of truths”; truths of different discourses are now said to exemplify different “varieties of truth”. What explains his change of view from the position of “Truth and ‘Correspondence’”? I’ll argue that his chief motivation has nothing to do with discourses beyond the “matter-of-factual”.20 To identify this motivation, we must return to his earlier work. The main thesis of “Inference and Meaning” (1953) is that “material rules of inference”, which Sellars equates with Rudolf Carnap’s nonlogical “P-rules” (Carnap [1934] 1937 §51: 180), “determine the descriptive meaning of the expressions of a language” (IM 25; also ITSA 316).21 He draws a consequence: In other words, where “ψa” is P-derivable from “φa” . . . it is . . . correct to say that “φaψa” is true by virtue of the meanings of “φ” and “ψ”. Concept-constitutive inference rules yield cases of conceptual truth, “truth ex vi terminorum” (ITSA 317–18; see Brandt 2017). Before long, Sellars qualifies this conclusion. In a footnote added in 1958 to another 1953 paper, he cautions: Note that, strictly speaking, one can only say that a sentence of L is true ex vi terminorum, as one can only say that a sentence of L is true simpliciter, if one’s own language contains a translation of these

184

Lionel Shapiro sentences, which will not be the case if expressions occurring in these sentences conform to different P-rules from those obeyed by their closest counterparts in one’s own language. (ITSA 317n)

If we can only apply “true” to sentences translatable into our own language, this poses a difficulty for Sellars. One of his key claims is that “there are an indefinite number of possible conceptual structures (languages) or systems of formal and material rules” (IM 26; P 293; ITSA 318). In each case, conditionals corresponding to the material rules state “laws of nature” (P 293; SRLG §29: 331). These languages may be so “radically different” as not to be “mutually translatable” (SRLG §77: 353–4). As he later says, such languages don’t embody the same conceptual structure (SM V §49: 132). Where a language isn’t translatable into ours, we “must abandon indirect discourse” in describing it: we can’t describe any sentence as meaning p (SRLG §79: 354). Sellars infers that we can’t describe the “lawlike sentences” of such a language as true, let alone true ex vi terminorum (SRLG §77–8: 354).22 By contrast, it remains “appropriate” to call such sentences “unconditionally assertable”. A journal entry makes the point succinctly: [T]o say of S that it is true is, in effect, to say “S means p and p” where “p” is an expression in one’s language—that is, the language one uses. . . . In short, one can correctly say S is true ex vi terminorum only when one uses a translation of S (as German is a translation of French). But one can say that S is unconditionally assertable by virtue of the linguistic rules governing S, without using a translation of S. (Dec. 24, 1953, Wilfrid S. Sellars Papers, box 25, folder 2: 119) Most likely, Sellars doesn’t mean that correctly applying the concept of truth to someone’s utterance requires actually being in a position to ascribe a content. For he’s not interested in cases where what precludes indirectdiscourse ascription is ignorance. Rather, he thinks there are cases where making correct indirect-discourse ascriptions to other speakers’ utterances would require changing our conceptual structure. It’s when faced with alien conceptual structures that we can’t correctly apply the concept of truth. What he appears to have in mind is a language otherwise much like ours, but in which “All swans are white” or “Combustion requires release of phlogiston” are “unconditionally assertable by virtue of the linguistic rules”. He thinks it reasonable for speakers of such a language to modify their conceptual structure, on empirical grounds, in a way that makes them abandon these sentences. Yet he worries that nothing can make it “reasonable to abandon true sentences” (SRLG §77: 354). His solution to this puzzle is that we shouldn’t call such “lawlike sentences” true.23 Later writings suggest cases where we might more plausibly think truth

Sellars, Truth Pluralism, and Truth Relativism

185

can’t be used to evaluate speakers of a language embodying another conceptual structure. Imagine a proponent of non-Euclidean physical geometry evaluating a Euclidean’s assertion of “The angles of any triangle in physical space sum to 180°”, or an intuitionist evaluating a classical mathematician’s assertion of “Every statement is true or not true” (SM V §39: 128–9; CC §§49–52: 89–90 [NAO 4 §§132–6: 94–5]). I’ll return to his treatment of such cases later. Suffice it to say here that Sellars, unlike Davidson ([1974] 1984, 194–5), doesn’t think the unavailability of truth ascriptions to propositions of an alien conceptual structure calls into question the very idea of such a conceptual structure. Even where truth is inapplicable, we can attribute assertibility by the rules of the language or conceptual structure. But Sellars isn’t content to forswear truth talk in describing alien conceptual structures. A parenthetical insertion in SRLG’s 1955 revision already proposes a way we can speak of truth: we can “‘relativise’ the notion of truth as true in W [the world of L], true in W’ [the world of L’], etc.—as opposed to true of this world” (SRLG §78: 354, brackets in original).24 Subsequently, this socalled relativized notion takes priority in Sellars’s thought. He concludes that “the fundamental form of ‘true’ is ‘true in conceptual structure CSi’” (SM V §50: 132–3). (1) A proposition belonging to a conceptual structure CSi is “true in CSi” just in case it’s semantically assertible according to the rules of CSi. Only when the “truth” we’re attributing is truth in our own conceptual structure CSO, i.e., assertibility according to the semantic rules of CSO, does correct attribution of truth require the in-principle availability of indirect-discourse ascriptions. To summarize: Sellars is committed to restricting attributions of truth to propositions of our conceptual structure, while recognizing a truthlike notion of assertibility for propositions of arbitrary conceptual structures. The key step he takes after 1963 is to unify these two commitments by identifying truth as semantic assertibility in CSO. It is this step that leads to his truth pluralism.

5 Sellars’s Truth Relativism We’ve now seen how Sellars’s analysis of truth is motivated by his interest in describing alternative conceptual structures. Next, I’ll explain how the same interest leads him to truth relativism. The first thing to appreciate is that despite Sellars’s use of “relativise”, there’s really nothing relativized about truth or falsity in a conceptual structure. A proposition can be true or false only in the conceptual structure it belongs to, whence a proposition can never be true in one conceptual structure yet false in another.

186

Lionel Shapiro

Stated using the terminology of Quine (1970, 19), truth in CSi is an “immanent” rather than “transcendent” notion: only propositions belonging to CSi can be true in CSi (or false in CSi, i.e., have negations that are true in CSi). In being immanent, Sellars’s notion resembles its ancestor, Carnap’s notion of being “valid in [language] S” ([1934] 1937 §51: 180), as well as the notion of “truth relative to a scheme” repudiated by Davidson ([1974] 1984: 198). To see why the immanent notion expressed by “x is true in CSi” shouldn’t be called relativized, compare the notion expressed by “x is the national capital of y”. Once one knows that the only country Melbourne belongs to is Australia, it would be absurd to reflect thus: “Melbourne isn’t the capital of Australia, but might it be the capital of some other country?” Being the capital of a country is a relational notion rather than a monadic one (it’s expressed by a two-place predicate), but it isn’t a relativized notion. Contrast the paradigmatic relativized notion expressed by “x is a national capital in the year y”. No prior fact about a city fixes one specific year as the only year during which it can be a national capital. Relativized notions can also be monadic, such as the notions of being a national capital in 1920 (which applies to Melbourne) and being a national capital in 1930 (which doesn’t). When Sellars says that “true in CSi” is the “fundamental form of ‘true’”, that’s because he uses it to define a number of derivative notions (SM V §§49–53: 132–4). Among these will be a genuine notion of relative truth. Specifically, he introduces a relativized predicate “true with respect to CSi” in terms of the non-relativized “true in CSi”. We can streamline Sellars’s explanations, some of which are needlessly confusing, as follows.25 For now, I’ll just provide definitions, where terminology in quotation marks is Sellars’s own. Later, when I turn to the question of why Sellars introduces the defined notions, I’ll apply the definitions to several examples he uses. As Figure 9.1 displays, there are really three contrasts in play: between immanent and transcendent notions, between relational and monadic notions, and between relativized and non-relativized notions. First among the derivative notions of truth, we have the immanent notion expressed by the monadic predicate obtained by replacing the variable “CSi” in Sellars’s fundamental predicate “true in CSi” by the name “CSO” of our own conceptual structure. (2) A proposition belonging to CSO is “absolutely true” provided it’s true in CSO. Absolute truth plays the role that was played in Sellars’s prior thought by truth. It’s said to obey the equivalence schema, and can’t apply to propositions in alien conceptual structures. Here “absolute” isn’t intended to mark a contrast with relativized truth. The contrast is instead with other non-relativized immanent notions, such as the notions of being a true sentence of English or being a true sentence of French.26

Sellars, Truth Pluralism, and Truth Relativism

187

Additionally, there are transcendent notions of truth, notions that can apply to propositions in any conceptual structure. The first of these, which is again monadic and non-relativized, isn’t explicitly marked by Sellars. But it’s worth singling out, since it plays the role we saw played in his prior thought by assertibility. (3) A proposition is intrinsically true provided it’s true in the conceptual structure it belongs to. (That is to say, it’s semantically assertible by whatever rules constitute that structure.) The remaining transcendent notions are relativized. To introduce them, Sellars uses his idea that a proposition belonging to one conceptual structure can have a “counterpart” in another. This lets him define a relational and relativized notion, “a sense in which a proposition in one conceptual structure can be true . . . with respect to any suitably related conceptual structure”. Here a conceptual structure is suitably related when it contains a counterpart of the proposition in question. (4) A proposition is true with respect to CSi (“true quoad CSi”) provided it has a “counterpart” in CSi that’s true in CSi. (Equivalently: provided it has a counterpart in CSi that’s intrinsically true.)27 By replacing the variable “CSi” with the name “CSO”, Sellars defines another transcendent notion, which he later calls “truth simpliciter” (SM V §96: 148). While monadic, this notion is still relativized. (5) A proposition is simply true (“true simpliciter”) provided it’s true with respect to CSO. For propositions belonging to CSO, intrinsic and simple truth amount to the same as absolute truth. However, propositions belonging to other conceptual structures are ineligible for absolute truth, and in their case intrinsic and simple truth can come apart. Why does Sellars introduce relativized truth? One of his motivations resembles that of relativists such as Kölbel (2002) and John MacFarlane (2014), who point to cases where the following seems to occur. One speaker affirms the very proposition another rejects, yet due to their divergent perspectives, both meet the relevant criteria of “correctness”. Examples often involve “subjective” claims, e.g., that some food is tasty or some joke funny. Sellars, by contrast, is most interested in “matter-offactual discourse”, though mathematics is in view too (SM V §55: 135; see Appendix 1). In the matter-of-factual case, the phenomenon arises in some cases of scientific progress. Sellars wants to say that a community of speakers can come to correctly reject the very proposition their predecessors correctly affirmed.28 Here, too, the explanation is supposed to lie

188

Lionel Shapiro

in the fact that the communities have divergent perspectives: they differ with regard to their “conceptual structure (which is necessarily the point of view from which we view the universe)” (SM V §63: 138). This idea already surfaces in Sellars’s 1958 discussion of “thing-kind words” such as “gold” and “salt”: [I]t can be reasonable to say “That wasn’t really gold” in spite of the fact that the object in question was correctly called gold according to the criteria used at the time the claim that is being disputed was made. And the statement that the object really wasn’t gold is not to be construed as a queer way of saying that the word “gold” no longer means exactly what it did. (CDCM §48: 261) It could be that an object that isn’t gold was correctly “called gold”. And this isn’t just to say that the English word “gold” would have been correctly applied to it, given the meaning this word had at that time. Rather, the idea is that the object really was correctly called gold, i.e., said to be gold. Sellars revisits the example in 1974: There is obviously no puzzle about saying that some of the stuffs which people once correctly called gold weren’t really gold. Here we recognize both the kinship and the difference between their use of “gold” and ours. (RDP 462) According to Sellars, there’s no incoherence in imagining a prehistoric speaker who correctly asserted, of a given nugget of fool’s gold, that it is gold. Here it’s crucial that “correctly” doesn’t just mean warrantedly. We can explain what he has in mind using intrinsic and relative truth. On the one hand, in view of the kinship in use, we can characterize the prehistoric speaker as expressing the concept gold. This is to describe them as expressing a proposition belonging to our conceptual structure CSO. We might then elaborate by characterizing their expression as belonging to a 30,000 BCE-ish sub-variety within the class of words meaning gold (SM V §45: 130–1; CC §§50–1: 89–90 [NAO 4 §§135–6: 94–5]; MFC 95 n14 [NAO 4 §63: 78 n19]). On the other hand, in view of the difference in use, we can also describe the speaker as employing a distinct structure CS30K. In this case we’ll say they express a counterpart in CS30K of our concept gold. We can thus give either of the following descriptions of our speaker: (a) They expressed a proposition belonging to CSO, to the effect that the nugget is gold. This proposition is intrinsically false, yet true with respect to CS30K.

Sellars, Truth Pluralism, and Truth Relativism

189

(b) They expressed a different proposition, belonging to CS30K. This proposition is intrinsically true, yet false with respect to CSO.29 Each description accommodates Sellars’s sense that what the speaker said is correct from their perspective but incorrect from ours. Crucially, each does so using the relativized notion of truth with respect to a conceptual structure. Furthermore, regardless which proposition we describe our speaker as expressing, Sellars thinks we may reasonably hold that the proposition is false with respect to CSP, where CSP is a “Peirceish” structure that achieves ideally adequate picturing. That’s how he comes to explain what he means when he speaks of something not being “really” gold. Completing Figure 9.1, we have a further transcendent notion (SM V §§73–5: 141–2, §96: 148; SM ix). (6) A proposition in CSi is “ideally true” provided it’s true with respect to CSP. This explains how even something we ourselves are correct to call gold might not “really” be gold: But does it make sense to say that some of the stuff we correctly call gold really isn’t gold? Yet is not this possibility implied by the very idea of what gold really is? Is not the concept of the ideal denotations of “gold” and “water” correlative with the concept of the ideal meanings of these sortals? (RDP 462)

non-relativized immanent

transcendent

relativized

relational

monadic

(1) truth in CSi

(2) absolute truth (truth in CSO) (5) simple (3) intrinsic truth (4) truth with respect to truth (truth in the CSi (truth with proposition’s respect to own (having a CSO) conceptual counterpart structure) in CSi that’s (6) ideal truth (truth with intrinsically respect to true) CSP)

Figure 9.1 Six Notions of Propositional Truth

relational

monadic

190

Lionel Shapiro

When we say the stuff is gold, the proposition we express is absolutely true, yet it may be ideally false: that is to say, the proposition may have a counterpart belonging to a structure of “ideal meanings” that is intrinsically false.30 On Sellars’s view, whether description (a) or (b) is more appropriate “shift[s] with context and purpose” (MFC 95 n14 [NAO 4 §63: 78 n19]; cf. SM V §47: 131). In some contexts, it’s useful to affirm sameness of meaning. In others it’s useful to deny it. Here he closely anticipates what Stewart Shapiro has written about logical vocabulary: Whether we say that the logical terms have the same meaning, or different meanings, in the different [mathematical] structures or theories, depends on what is salient in a conversation comparing the structures or theories. For some purposes—in some conversational situations— it makes sense to say that the classical connectives and quantifiers have different meanings than their counterparts in intuitionistic . . . systems. In other situations, it makes sense to say that the meaning of the logical terminology is the same in the different systems. (S. Shapiro 2014, 5, 127ff.) Sellars makes the same point with the same example. He says features of “context” account for our “willingness”, on some occasions but not others, to affirm that “classical negation and intuitionistic negation are varieties of negation” (CC §52: 90 [NAO 4 §136: 95])—i.e., that classical and intuitionistic mathematicians use the English “not” or German “nicht” to mean not. Thus in a given context This “nicht” is a •not• [LS: this is Sellars’s regimentation of the ascription “This token of ‘nicht’ means not”] will be true or false depending upon whether the criteria for being a •not• include or do not include the consequence relations involved in the principle of excluded middle. (TTP §55: 296)31 Accordingly, as in the case of “gold”, context will affect how we use attributions of intrinsic and relative truth in assessing claims expressed by speakers whose usage is both akin to and different from ours. In this regard too, Sellars anticipates Stewart Shapiro, whose own discussion invokes a form of truth relativism (2014, 154–5).32

6 Conclusion: A Tension in Sellars’s View I’ve now argued that Sellars’s truth pluralism and his truth relativism are both motivated by the desideratum of having semantics take into account

Sellars, Truth Pluralism, and Truth Relativism

191

a plurality of conceptual structures. Huw Price (1992, 389–90) distinguishes a “vertical pluralism” of functionally heterogeneous domains of discourse (e.g., empirical, ethical, mathematical) from a “horizontal pluralism” of alternative worldviews. Sellars’s truth relativism results from a horizontal pluralism of conceptual structures, unlike those present-day versions of relativism concerned with “subjective” domains of our own conceptual structure. On the surface, his truth pluralism would seem to result from a vertical pluralism of domains. However, while it’s applied vertically, I’ve argued that his truth pluralism results from considerations of horizontal, cross-structure variation. That’s because such considerations underlie his analysis of truth as semantic assertibility, which is what leads him to formulate his approach to heterogeneous domains as a truth pluralism. Whereas Sellars’s truth pluralism can’t be understood apart from his problematic analysis of truth, his truth relativism is independent of that analysis. In fact, as I’ll explain in closing, that analysis is responsible for a tension in his relativism. The tension is revealed by a contrast between Sellars and present-day truth relativists. As we saw, Sellars defines “absolute” truth, the ordinary monadic notion said to obey the equivalence schema, using his “fundamental” notion of truth in a conceptual structure. Present-day relativists don’t use their semantic primitives to define ordinary monadic truth. They use their primitives to specify the conditions for an assertion’s semantic correctness. But they don’t equate such correctness with the asserted proposition’s truth. MacFarlane is especially clear: rather than define ordinary truth in terms of truth relative to a context, he uses the latter to specify the semantics of a monadic predicate “true”. This is done in a way that ensures that “The proposition that p is true if and only if p” expresses a proposition that’s true relative to all contexts, and hence correctly assertible in all contexts (MacFarlane 2014, 93–4; Cappelen and Hawthorne 2009, 12–14). For MacFarlane’s purposes, it’s important not to follow Sellars and equate the ordinary truth of the propositions we express with their truth relative to our context. Suppose my interlocutor denies that the food I’m savoring is tasty. Then she should also deny that the proposition that the food is tasty is true. However, on MacFarlane’s theory, she should concede that this proposition is true relative to my present context, given the standard this context supplies. This explains why it would be a mistake to identify the truth of the proposition with its truth relative to my present context. The point here isn’t that on the relativist view it’s possible for someone to disagree with me about the (absolute) truth of the proposition while agreeing with me about its truth relative to my standard of taste. That could simply be a matter of their being ignorant that (absolute) truth is truth relative to my standard. The point is rather that according to relativists, my interlocutor would be correct in doing so. In that case (absolute) truth can’t be truth relative to my standard.

192

Lionel Shapiro

Turning to Sellars’s own relativism, suppose we’re classical mathematicians evaluating an intuitionist, and that our “context and purpose” are such that we describe the intuitionist as rejecting the very proposition we express by an instance of excluded middle. If we accept the relativist description of what’s going on, we’ll presumably claim that the intuitionist ought not regard the excluded middle proposition as true, in the ordinary sense of “true” that (as Sellars insists) satisfies the equivalence schema. Yet, we’ll say that the intuitonist ought to agree that the proposition is S-assertible by the rules of our classical CSO. Consequently, there’s at least a tension between two uses Sellars makes of assertibility by the rules of a conceptual structure. On the one hand, we have his truth relativism, according to which it would seem that our intuitionist correctly rejects the proposition in question, and thus ought not regard it as true. On the other hand, we have his analysis of (absolute) truth, according to which, still on the assumption that CSO embodies classical logic, the intuitionist ought to regard the proposition as true.33 Could Sellars have relativized the correct assertibility of propositions to a conceptual structure without adopting his analysis of (absolute) truth? Had he instead continued to elucidate propositional truth using the equivalence schema, as present-day truth relativists do, he would also have sidestepped the signal shortcoming in his treatment of truth: the failure of his pragmatic justification of the claim that the proposition that p is correctly assertible if and only if p.34 Without the analysis of truth as S-assertibility, we have seen, he would have lacked the grounds that led him to adopt truth pluralism. Still, he could have retained the more moderate position of “Truth and ‘Correspondence’”, according to which the heterogeneity of discourses is reflected in varieties of truths rather than in varieties of truth.35

Acknowledgments For feedback on successive drafts of the chapter, I thank audiences and discussants in Erlangen (2015), Hradec Králové (2016), Potsdam (2017), and Halle (2018). I’m especially indebted to Anke Breunig, Willem A. deVries, Stefanie Grüne, Johannes Haag, Johannes Hübner, and Jaroslav Peregrin for their very helpful comments, and to Michael Lynch, Huw Price, Marcus Rossberg, and Michael Williams for conversations. Part of this work was done while I was supported by a fellowship from the Alexander von Humboldt Foundation. ***

Appendix 1 Relativism and Failures of Bivalence

In Science and Metaphysics, Sellars rejects the law of bivalence. After sketching the semantic rules for a language that suffices for rudimentary logical, descriptive, and practical discourse, he remarks: “A realistic account of these semantical rules precludes that every proposition of the relevant variety is either true or false” (SM IV §62: 115; Williams 2016, 244). He is most naturally understood as denying that every proposition expressible in the language is either true (i.e., assertible in accordance with the language’s rules) or false (i.e., has a negation that’s so assertible). One might worry that this rejection of bivalence conflicts with Sellars’s endorsement of the equivalence schema for truth. Recall that he affirms the “conceptual necessity” of each instance of “The proposition that p is true if and only if p” (SM IV §26: 100). Given modest logical assumptions, this is incompatible with counterexamples to bivalence (Dummett 1991, 64; Holton 2000, 149–51). Suppose the proposition that p is a counterexample to bivalence, i.e., a proposition that’s neither true nor false. Then it will be untrue that p, and also untrue that not-p. By modus tollens, these claims together with the equivalence schema for truth yield that not-p and also that not-not-p. If he asserts that there are counterexamples to bivalence, Sellars is thus committed to a contradiction. In this appendix, I argue that Sellars’s truth relativism lets him escape the objection: it gives him a way to recognize failures of bivalence while avoiding contradiction. There are at least three contexts in which Sellars discusses bivalence, and in each he may seem to countenance counterexamples to bivalence. The first is the preceding quoted remark, where he seems to say that not every proposition under discussion is true or false. Using any logic that validates the De Morgan equivalences for quantifiers, it would follow that some proposition under discussion is neither true nor false. Sellars gives no indication that he intends to reject the relevant De Morgan equivalence, e.g., by employing intuitionistic logic. To resolve this problem, the first thing to notice is that it only arises for languages that embody our conceptual structure (CSO). Instances of the equivalence schema “The proposition that p is true if and only if p” can

194

Lionel Shapiro

only be stated for propositions in CSO, since these are the only propositions we can refer to using instances of “the proposition that p”. Consider now a formulation of bivalence concerning an arbitrary language L. (BVL) Every proposition expressed in L is assertible, or has a negation that is assertible, according to the semantic rules of the conceptual structure L embodies. There would be no problem had Sellars written that it would be unrealistic to include (BVL) in an account of a language L as expressive as those under discussion. The problem concerns whether such an account can “preclude” bivalence, in the special case where L is a language embodying CSO, such as English (or E). Can our account of English include (BVE)’s negation? I’ll now explain how Sellars is in a good position to assert (BVL)’s negation for other languages L while refusing to assert or deny (BVE)’s negation. The relevant fact about English is that Sellars assumes it’s an example of a language L that embodies a conceptual structure CSL whose users express the notion of being true in CSL (since he assumes English expresses absolute truth, the notion of being true in CSO).36 For some languages L of this sort, the CSO-proposition we express by (BVL)’s negation will be absolutely true. Yet speakers of CSL can’t themselves, without contradiction, assert that proposition’s CSL-counterpart, a proposition that’s nonetheless true with respect to CSO. In the same way, the CSO-proposition we express using (BVE)’s negation is one we can’t assert without contradiction, though we may be confident that it’s true with respect to a more adequate conceptual structure such as the Peirceish CSP. There’s a sense, then, in which Sellars can consistently recognize the failure of bivalence for our own language. He can say that the claim that our language expresses propositions that are neither true nor false is ideally true. Furthermore, there is textual support for connecting Sellars’s approach to bivalence with his truth relativism. Immediately following his remark about empirical languages, he turns to mathematical truth, a second context in which he discusses bivalence. Indeed, in the case of logical and mathematical propositions, where S-assertability means provability, the law of bivalence can be defended only by defining a sense in which propositions in one axiomatic framework can be “identified” with propositions in a more inclusive axiomatic framework. (SM IV §62: 115) His reason is Gödel’s first incompleteness theorem. When he revisits the topic of arithmetic after introducing relative truth, Sellars

Sellars, Truth Pluralism, and Truth Relativism

195

goes beyond questioning the defensibility of bivalence. He now asserts counterexamples. In the case of arithmetic .  .  . the concept of truth (S-assertibility) coincides with that of provability. It follows, of course, from Goedel’s results that, with respect to the conceptual structure (in the sense of axiomatics) to which it belongs, not every arithmetical proposition is either true or false. (SM V §55: 135) Sellars’s claim concerns a proposition’s truth “with respect to the conceptual structure . . . to which it belongs”, which is equivalent to what I’ve called its intrinsic truth. Since he justifies the claim by citing Gödel’s theorem, it’s clear he holds the following. (G) If Ai is a recursive axiomatic system containing basic arithmetic, and Ai is consistent, then there’s a proposition belonging to Ai that’s neither intrinsically true (true in Ai) nor intrinsically false (false in Ai).37 Provided we can assert that Ai is consistent, we can thus assert that there are counterexamples to bivalence for truth in Ai. Invoking relative truth, Sellars adds that such an undecidable proposition can be said to be true in a derivative sense, if its counterpart in a richer axiomatics, Aj, which is also, in a sense difficult to define, an axiomatics of arithmetic, is provable in Aj. Thus, a proposition in Ai can be said to be true quoad Aj. In the case of arithmetic there is no end to the series of “more adequate” axiomatic systems. (Ibid.) This should explain the attenuated sense in which Sellars thinks “the law of bivalence can be defended” for arithmetical propositions. The defensible thesis may be that for every proposition in Ai, there is a more adequate Aj with respect to which it is either true or false, and with respect to whose more adequate successors it retains its truth value. What if we apply (G) to an axiomatic system that captures arithmetic in CSO? This yields no conflict with the equivalence schema. Gödel’s methods can’t be used to prove, concerning such a system, the existence of a sentence that’s neither provable nor disprovable by that system’s own rules. To do that, it won’t suffice to prove the respective instance of (G). We would also need to prove the consistency of that system, an arithmetically formalizable claim which Gödel’s second incompleteness theorem tells us the system, if consistent, won’t prove. Hence Gödel’s results, together with Sellars’s view of mathematical truth, don’t commit Sellars to asserting that our mathematical framework contains propositions

196

Lionel Shapiro

that are neither true nor false. To recognize failures of bivalence, he can instead say that the arithmetical proposition formalizing the claim that there are undecidable arithmetical propositions will be true with respect to a more adequate successor axiomatic system. Still, Gödel’s results do point to a genuine problem for Sellars’s account of mathematical truth. Sellars’s embrace of the equivalence schema turns out to conflict with his claim that (absolute) arithmetical truth consists in provability in a recursive axiomatic system within CSO, say AO. These views jointly commit him to the conditional (*) If “1=0” is provable in AO, then 1=0. and thus to the truth, and hence the provability in AO, of (*)’s arithmetical formalization. But if (*)’s formalization is provable in AO, then so is the sentence formalizing the claim that AO is consistent (Gödel [1933] 1995; Artemov 2007, 4). By Gödel’s second incompleteness theorem, that sentence will be provable in AO only if AO is inconsistent. So even though Sellars’s embrace of the equivalence schema needn’t clash with his rejection of bivalence, it does clash with his account of mathematical truth. A third context in which Sellars discusses bivalence pertains once again to matter-of-factual propositions. He states that “[t]he law of bivalence analytically holds for matter-of-factual propositions in CSP, but it need not hold for matter-of-factual propositions in less developed conceptual frameworks”. This is offered as commentary on a principle he formulates using the notion of truth with respect to a conceptual structure. If a proposition in CSP is true quoad CSP its counterparts in such frameworks (CSi) as contain a counterpart are true quoad CSP, but not necessarily true quoad CSi, though not false quoad CSi. (SM V §74: 142) If we unpack this principle using the definition of “truth quoad CSi” for CSP-propositions, we find Sellars saying that an ideally true CSi-proposition may be neither intrinsically true nor intrinsically false.38 This poses a problem when CSi is CSO, as it would imply that CSO-propositions may be neither absolutely true nor absolutely false. Again, though, he should refuse to assert that ideally true CSO-propositions must be absolutely true or absolutely false, without asserting that they may be neither absolutely true nor absolutely false. Instead, he can allow that it may be ideally true, concerning an ideally true CSO-proposition, that it’s neither absolutely true nor absolutely false. Such a proposition will be ideally true, but the claim that the proposition is absolutely true won’t be ideally true. On the other hand, I argued in Section 6 that Sellars would have done better to avoid paradoxicalsounding consequences like this one by giving up his identification of absolute truth with semantic assertibility according to the rules of CSO.

Appendix 2 Sellars, Quine, and “De-Quoting”

In a much-cited passage from his Philosophy of Logic (1970), W.V. Quine gives Sellars credit for having formulated a claim about truth Quine there endorses. And the attribution might seem to find support in Sellars’s remark about Quine in a recently published lecture. The purpose of this appendix is to set the record straight. Introducing his own view that “the truth predicate is a device of disquotation”, Quine writes: “So long as we are speaking only of the truth of singly given sentences, the perfect theory of truth is what Wilfrid Sellars called the disappearance theory of truth” (1970, 11). This theory holds that the truth predicate is unnecessary in such cases: rather than say that “Snow is white” is true, we could just say that snow is white. To be sure, Quine (who provides no citation to Sellars) doesn’t claim that Sellars advocates a disappearance theory of truth ascriptions to quoted sentences. Still, Quine appears to be attributing the phrase “the disappearance theory of truth” to Sellars, and to be including Sellars among those who at least formulated a component of disquotationalism.39 Furthermore, in writings predating and contemporaneous with Quine’s remark, Sellars does explain truth in terms of “de-quoting” (SM IV §51: 110; Sellars-Harman Correspondence). Subsequently, in a 1975 letter, he even speaks of his “interpretation of truth in terms of de-quoting” (Sellars-Marras Correspondence). Most intriguingly, in a 1969 lecture, Sellars appears to claim precedence for disquotationalism: “I’ve been interested to know that Quine is now coming around to the position that ‘true’ functions essentially as a de-quoting device” (WSNDL II §57: 79–80). Taken together, Quine’s and Sellars’s remarks raise the question of how Sellars’s discussions of truth relate to Quine’s disquotationalism. In fact, Quine’s reference to Sellars and Sellars’s reference to Quine are equally misleading. In the published work that would appear most relevant to Quine’s attribution, Sellars’s contribution to the Library of Living Philosophers volume on Carnap, Sellars does include a remark on “‘definitions by disappearance’ of semantical concepts” (EAE 464). But here he’s quoting P.F. Strawson, and endorsing Strawson’s skepticism about the idea of a “definition-by-disappearance” of a predicate (Strawson

198

Lionel Shapiro

1949, 88). More importantly, when Sellars later appeals to “de-quoting” himself, he’s advocating something very different from Quine’s disquotationalism, namely his own analysis of “true” as “semantically assertible”. That analysis is profoundly un-Quinean in two respects. First, as Williams (2016, 236) stresses, it holds that truth ascriptions are normative claims, claims about the “correctness of a certain performance—roughly the dequoting of a quoted expression” (SM IV §51: 110). Second, it presupposes talk of meaning and propositions. The reason Sellars says truth is only “roughly” a matter of correctness of de-quoting is that for him, truth ascriptions don’t apply to sentences considered with respect to the syntactical characteristics singled out by Quine’s quotation marks. Rather, truth ascriptions apply primarily to propositions, where proposition talk is construed as talk of sentences considered with respect to their meaning. Sellars introduces his own use of quotation marks, dot-quotation, to classify sentences according to their meaning. Thus, his “interpretation of truth in terms of de-quoting” is really the view that truth is correctness of de-dot-quoting. In no sense did Quine come around to that position. Sellars’s discussion of disquotation raises a further historical question. After presenting his idiosyncratic use of the idea, according to which “the force of ‘true’ [is] to say that the quotation marks which precede it, together with itself, may be erased” (SM IV §29: 102), Sellars adds that this fact “has often been noted”. Which discussions might he have in mind? One very likely candidate is a claim by Peter Geach. In [“‘More than twenty people came to Smith’s party’ is a true statement”], the use of “is true” or “is a true statement” is superfluous; we could cancel out these words along with the pair of opening and closing quotes preceding them. . . . [W]e write a statement in quotes in order to mention that statement; if we now write after it “is (a) true (statement)”, this cancels the quotes, and it is as though we simply used the quoted statement. (Geach 1957, 96) This claim appears in a context that had already caught Sellars’s attention, namely, Geach’s independent defense of the characteristically Sellarsian “general thesis that language about thoughts is an analogical development of language about language” (Geach 1957, 98; Sellars notes the agreement in a paper published in 1966, FD 150). Despite the similarity in wording to Sellars’s remark in Science and Metaphysics, what Geach advocates here resembles Quine’s view of truth ascriptions to quoted sentences, not Sellars’s very different view of truth as correct assertibility. Geach’s point is that a truth ascription to a quoted sentence has the effect of canceling the quotation marks; it has what Quine later calls “cancellatory force” (1970, 12). Sellars, by contrast, claims that the truth ascription has cancellatory sense, i.e., that truth ascriptions to dot-quoted sentences “say that” the quotation marks may be removed.

Sellars, Truth Pluralism, and Truth Relativism

199

Notes 1. They’ve also rarely been discussed in any detail, other than in sympathetic expositions of his thought (Bonjour 1973; Seibt 1990; deVries 2005; O’Shea 2007). Exceptions include Harman (1970), Armour-Garb and Woodbridge (2012), Price (2013, 160–70), L. Shapiro (2014), and especially Williams (2016). I lack space here to fully address Williams’s rich paper. See also the chapter by O’Shea in this volume. 2. Rorty’s and Putnam’s claims likely result from misunderstanding two notions Sellars uses in his theory of truth: correct assertibility and (for “matter-offactual” claims) correct assertibility with respect to an ideal conceptual structure. These will be explained in Sections 3 and 5. As we’ll see, neither is an epistemic notion. Sellars doesn’t give an “epistemic account of truth” (Williams 2016, 238) or “confuse the theory of evidence with the theory of truth” (Harman 1970, 417). 3. Davidson comments with remarkable condescension: “There is less excuse for the widespread misunderstanding of .  .  . the semantical [i.e. Tarski’s] approach. The following example [the second paragraph of Sellars’s TC] is no worse than many that could be quoted”. Harman (1970, 413–14) too surmises Sellars “has not fully appreciated Tarski’s theory”. Sellars responds, addressing Davidson’s paper, in the Sellars-Harman Correspondence. 4. Moreover, his reasons are ones Sellars had long embraced. Davidson notes that a Tarskian list-like specification of referents for primitive expressions “does not explain or analyse the concept of reference”. He also claims the truth of sentences, not the reference or satisfaction of their constituents, is the “place where there is direct contact between linguistic theory and events, actions, or objects described in non-linguistic terms” (Davidson [1977] 1984, 218–23). Sellars stresses both points in the Sellars-Harman Correspondence (L. Shapiro 2013, 307–8), but they already appear in the paper Davidson criticized. 5. This commitment of his is highlighted by Brandom (2013, 89) and Price (2013, 151–2, 160–1). 6. The pluralist theme in Sellars’s discussion of truth has been noted by Williams (2013, 131 n1; 2016, 232) and myself (L. Shapiro 2014, 805 n27). 7. To be sure, Sellars in later writings discusses two senses in which we can speak of different “varieties of triangularity”. First, he’s willing to recognize an uninteresting sense in which species of triangle do correspond to species of triangularity, e.g. “isosceles triangularity” and “scalene triangularity”. But the interesting distinction, he argues, is that between (e.g.) Euclidean triangularity and Riemannian triangularity (SM V §§37–41: 128–30; CC §§49–52: 89–90 [NAO 4 §§132–6: 94–5]). This distinction, he argues, is a matter of there being different ways for a word to mean triangular. While we’ll find Sellars recognizing different meanings of the word “true”, he never speaks of there being different ways for a word to mean true. As discussed in the present chapter, truth pluralism is a doctrine according to which there are different ways for a sentence to be true, not different ways for a word to mean true. Hence neither of Sellars’s two versions of triangularity pluralism can serve as a model for a Sellarsian truth pluralism. 8. Here I ignore the fact that Lynch (2009, 90–1) restricts this claim to atomic propositions. 9. This paper was complete by 1954. See also SM IV §32: 103. Admittedly, Sellars insists on giving a “non-metaphysical account of abstract entities” such as properties (EAE 442), but his successive attempts at rejecting “Platonism” while countenancing talk of properties aren’t directly relevant to the present topic. 10. Lynch (2009, 71–2); Wright (1996, 919; 1999, 228; 2013, 128).

200

Lionel Shapiro

11. Unlike Lynch, Wright thinks the equivalence schema yields all platitudes for “true” (1999, 230n). 12. Sellars discusses the “Ramsey sentence approach” on which analytic functionalism is based, but only as a failed proposal for maintaining instrumentalism about talk of theoretical entities (SRI 342–3). 13. This objection needn’t extend to the proposal as applied to psychological predicates. The “pain role” involves causal relations. It might turn out, based on causal facts, that no single property plays this role for all organisms. This would be reason to infer that different organisms are in pain in virtue of exemplifying different properties. 14. Sellars’s scare quotes allude to “Convention T” (Tarski [1933] 1983), according to which the instances of “‘p’ is true if and only if p” must follow from any correct definition of sentential truth for an object language contained in the metalanguage of the definition. Sellars thinks Tarski only achieves this in an extensional sense of “definition”, while he himself only achieves this in a pragmatic sense of “follow”. 15. Sellars explains intra-linguistic rules as “ought-not-to-bes” or “constraints” (e.g., SM V §§5–6: 117–18; MFC 86 [NAO 4 §23: 67–8]). While the deontic modality of his world→language rules is less explicit, they could be understood as stating that assertion of a proposition, e.g., that expressed on an occasion by “This is red” is permitted if a certain extra-linguistic circumstance obtains and prohibited otherwise. They would thus count as both “may-bes” and “ought-not-to-bes” (their former role is suggested by a remark in LTC 512 n4). One limitation of the preceding reconstruction, which grounds semantic assertibility in the existence of derivations that involve no assumptions, is that it makes no room for Sellars’s category (discussed in Section 4) of propositions that are correctly assertible on the basis of intra-linguistic rules alone. See also note 37. 16. Sellars cautions that “‘correctly assertible’ does not mean the same as ‘assertible with good reason or warrant’. Semantic and epistemic oughts must be handled with the same care as that involved in distinguishing (and relating) the various senses in which an action can be said to be morally suitable” (NAO 4 §95: 85). Presumably he has in mind “objective” and “subjective” moral oughts. 17. Edwards (2013) suggests pluralists should understand the relation between truth and domain-specific truth properties on the analogy of that between being a winner and game-specific “win-making” properties. Sellars offers a particularly simple way to substantiate Edwards’s analogy. Just as what it is to be a winner in chess can be spelled out in terms of the rules of chess, what it is to be a truth of empirical discourse can be spelled out in terms of the relevant semantical rules. 18. See L. Shapiro (2014, 805 n28) and Williams (2016, 239–45). I agree with Williams that this obstacle shows Sellars’s definition of truth to be “more trouble than it is worth”, but I reject his reasons for holding that Sellars fails to establish the equivalence schema, since they presuppose that Sellars offers an “epistemic definition of truth”. There isn’t space to examine the idea, suggested by Jaroslav Peregrin, that Sellars might have justified the schema by showing that its instances are themselves assertible. 19. Wilfrid S. Sellars Papers, box 14, folder 30: 2–3. Harman’s objection applies equally to an earlier strategy Sellars tried out for vindicating the equivalence schema. In the 1963 note announcing his new analysis of truth (TC 224), Sellars had reported that the schema’s right-to-left direction requires a “somewhat more complicated argument” than the “straightforward” left-toright one. Williams (2016, 243) observes that Sellars nowhere publishes such

Sellars, Truth Pluralism, and Truth Relativism

20.

21. 22.

23.

24. 25.

26.

27.

28. 29.

201

an argument. But we find one sketched in notes dated March 10, 1963, and marked “SAVE” (Wilfrid S. Sellars Papers, box 17, folder 9: 10–11). Here Sellars argues that on the supposition that p, we can’t coherently say that it’s assertible that ~p. Explicitly assuming bivalence (“assertible that p ∨ assertible that ~p”), he then derives that it’s assertible that p. Harman’s objection already vitiates the first step: there is no pragmatic incoherence in merely supposing that p but that it’s assertible that ~p. Moreover, Sellars himself subsequently rejects bivalence (SM IV §62: 115). See Appendix 1. Previously (L. Shapiro 2014, 805 n27), I suggested Sellars was chiefly motivated by the need to explain how truth has different “essential features” in different domains. Cf. also Williams (2016, 232). Given the better-supported alternative defended here, I no longer find this persuasive. He doesn’t include rules governing world→language or language→world transitions, as he hasn’t yet figured out how such transitions could be rule-governed. More generally, a description of a natural language can only “include semantical statements provided that the language is translatable into the language in which the study is made” (EAE 453). This is Tarski’s position concerning formal languages. Sellars had previously defended the contrary view that it can be “rational to abandon” a sentence that’s “true ex vi terminorum” when modifying one’s concepts (P 293 n3). In his abstract for a 1951 draft of ITSA, he wrote: “Our language is . . . but one of a large number of possible languages each with its own set of truths ex vi terminorum” (Wilfrid S. Sellars Papers, box 25, folder 1: 220). For an earlier discussion distinguishing the “semantic predicate ‘true’” from the “true-in-S” of “pure pragmatics”, see PPE 191 (“S” is a “story” said to designate a world). Cf. Rosenberg (2007, 121–2), who appears to overlook simple truth, mistaking “true simpliciter” for a variant of “absolutely true”. One cause of confusion is that on the page where Sellars distinguishes these notions, he uses the unqualified “true” for both of them (SM V §§50, 52: 133). Another is that when Sellars defines the “truth of a proposition in CS1”, he means the simple truth of a proposition belonging CS1, not its property of being true in CS1. Worse, shortly after reaffirming that “the ‘absolute’ sense of ‘true’ . . . amounts to ‘true proposition belonging to CSO’” (§52: 133), Sellars mischaracterizes it as a sense in which “a proposition in one conceptual structure can be true . . . with respect to our current conceptual structure” (§53: 134). That’s simple truth. Sellars himself places “absolute” in scare quotes; the reference is to Carnap’s (1942, 89–90) use of the term for the truth of propositions, as contrasted with sentences of a given language system. Sellars wishes to distance himself from what he sees as Carnap’s mistake of regarding talk of the proposition that p as talk about “non-linguistic” matters. On Sellars’s view, it’s really talk about “inter-linguistic” matters, i.e., about items in any language that mean p (SM IV §22–23: 99–100, V §48: 131–2; WSNDL II §53: 78–9). Sellars appears to assume that a proposition in one conceptual structure will have at most one counterpart in another conceptual structure. In any case, he nowhere contemplates a proposition’s being both true and false with respect to a conceptual structure (due to having both intrinsically true and intrinsically false counterparts in that structure). It’s well known that Sellars introduces structure-relative truth as a resource for describing scientific progress (O’Shea 2007, 161; Rosenberg 2007, 121–3; Williams 2016, 250), but how he deploys it remains underexplored. According to both (a) and (b), the proposition asserted is simply false.

202

Lionel Shapiro

30. Here Sellars appears to reject a principle he had earlier said “would seem to be valid”, according to which an absolutely true matter-of-factual proposition must also be ideally true (SM V §74: 141–2). In Science and Metaphysics, that is to say, he assumes that if a proposition belonging to CSO is intrinsically true, then its counterpart in CSP will also be intrinsically true (cf. Rosenberg 2007, 123–6). I won’t here be able to adequately explore this discrepancy. But Sellars’s intention in the earlier discussion could be related to his interest in recognizing a “sense in which one proposition can be said to be ‘more true’ than another” (SM V §§54: 134). Perhaps an absolute truth will be more true than a merely ideal truth. If we additionally understand x’s being more true than y as requiring that x possesses any kind of truth that y possesses, then Sellars’s principle would follow. However, that requirement would be tendentious. I might add that his later discussion makes room for a second sense of “more true”, according to which a proposition that’s ideally true is more true than one that’s merely absolutely true. 31. Sellars and Stewart Shapiro take no position on whether this is because the content, or just the extension, of “have the same meaning” or “means not” varies with context (S. Shapiro 2014, 137). Perhaps they can, without generating vicious circularity, simply say that whether “means not” expresses the same content in two contexts will itself vary with our purposes. 32. By contrast, Williams argues that we needn’t “proliferate senses of ‘true’” to enable suitable truth assessments across differences in usage; contextdependence of meaning ascriptions suffices (2016, 250). This doesn’t accommodate Sellars’s relativist intuition that there are contexts in which it makes sense to assess another speaker as having correctly asserted the very claim we deny. 33. This reasoning could be resisted. Perhaps the proposition that it’s true (=S-assertible in CSO) that p and the proposition that p are equivalent, yet speakers whose usage differs sufficiently from ours may affirm the former without commitment to the latter. But it’s at least awkward to describe someone as correctly affirming the truth of a proposition they correctly reject. A second awkward option would be to insist that the intuitionist, even when interpreted as expressing the proposition that p, shouldn’t be interpreted as expressing the proposition that it’s true that p, but at most a counterpart. 34. He might have added that the proposition expressed by an instance of “The proposition that p is true if and only if p” is correctly assertible relative to any conceptual structure that contains a counterpart proposition. 35. It could remain part of the position that truth is accompanied by representational correctness in some domains. That is the view endorsed by Price (2013, 160–70), who has pointed out its resemblance to that of Sellars. 36. Here I won’t be able to address what resources Sellars’s position has for responding to the challenge to this assumption posed by the Liar paradox. 37. Sellars appears to be assuming that an “axiomatic framework” is recursive, i.e., that whether a given sequence of sentences is a proof is mechanically decidable. (I thank Marcus Rossberg for calling my attention to this.) It’s not obvious, though, that being an instance of a Sellarsian inference rule must be recursive. Perhaps there could be an infinitary “ought-not-to-be” rule to the effect that an agent ought not to assert each of propositions P1, P2, . . . while rejecting proposition Q. Such a rule couldn’t be used as a “rule of criticism” (SM V §4: 117; LTC 508), since no finite agent could be criticized for violating it. But might such a rule be in effect nonetheless, where this would make itself felt by rendering propositions “semantically assertible”? That appears to be the position of Peregrin (2014, 156–9, 83, 72), who endorses Sellars’s analysis of truth. In an early paper, Sellars briefly mentions Carnap’s distinction ([1934] 1937 §14: 37–9) between the recursive relation of being

Sellars, Truth Pluralism, and Truth Relativism

203

“directly derivable” and the non-recursive relation of being a “direct consequence”. Alas, he dismisses it as “irrelevant to our problem”, namely the role of material rules of inference in a conceptual structure (IM 19). 38. He also says that such a proposition will never be intrinsically false. In note 30, I suggested that this is a claim he repudiates in later work, when he seems to allow that the claim that some object fails to be gold might be absolutely (whence intrinsically) false yet ideally true. 39. Decades later, again without citation, Quine ([1999] 2008, 163) attributes the phrase “the disappearance theory of truth” to Frank Ramsey instead. This time, however, he claims to “deplore” the use of that label for the “disquotational account”. He calls talk of “disappearance” in this connection the “crowning outrage of my primordial philosophical bugaboo: the confusion of use and mention”. Presumably, he now thinks such talk wrongly suggests that “‘Snow is white’ is true” can be replaced by the name “‘Snow is white’” rather than the sentence “Snow is white”.

References Armour-Garb, Bradley and James Woodbridge. 2012. “Sellars and Pretense on ‘Truth and “Correspondence”’ (With a Detour Through Meaning Attribution)”. Discusiones Filosóficas 13 (21): 33–63. Artemov, Sergei. 2007. “On Two Models of Provability”. In Mathematical Problems from Applied Logic II, edited by Dov Gabbay et al. New York: Springer. Bonjour, Lawrence. 1973. “Sellars on Truth and Picturing”. International Philosophical Quarterly 13 (2): 243–65. Brandom, Robert. 2013. “Global Anti-Representationalism?” In Price 2013, 85–111. Brandt, Stefan. 2017. “Sellars and Quine on Empiricism and Conceptual Truth”. British Journal for the History of Philosophy 25 (1): 108–32. Capellen, Herman and John Hawthorne. 2009. Relativism and Monadic Truth. Oxford: Oxford University Press. Caputo, Stefano. 2012. “Three Dilemmas for Alethic Functionalism”. Philosophical Quarterly 62 (249): 853–61. Carnap, Rudolf. (1934) 1937. The Logical Syntax of Language. Translated by Amethe Smeaton. London: Routledge. ———. 1942. Introduction to Semantics. Cambridge, MA: Harvard University Press. Davidson, Donald. (1969) 1984. “True to the Facts”. In Donald Davidson. 1984. Inquiries into Truth and Interpretation, 37–54. Oxford: Clarendon Press. Hereafter Davidson 1984. ———. (1974) 1984. “On the Very Idea of a Conceptual Scheme”. In Davidson 1984, 183–98. ———. (1977) 1984. “Reality without Reference”. In Davidson 1984, 215–25. deVries, Willem. 2005. Wilfrid Sellars. Montreal: McGill-Queen’s University Press. Dummett, Michael. 1991. The Logical Basis of Metaphysics. Cambridge, MA: Harvard University Press. Edwards, Douglas. 2013. “Truth, Winning, and Simple Determination Pluralism”. In Truth and Pluralism: Current Debates, edited by Nikolaj Pedersen and Cory Wright, 279–91. Oxford: Oxford University Press. Field, Hartry. 2008. Saving Truth from Paradox. Oxford: Oxford University Press.

204

Lionel Shapiro

Geach, Peter. 1957. Mental Acts. London: Routledge and Kegan Paul. Gödel, Kurt. (1933) 1995. “An Interpretation of the Intuitionistic Propositional Calculus”. Translated by John Dawson in Collected Works, vol. 1, edited by Saul Feferman et al., 301–3. Oxford: Oxford University Press. Harman, Gilbert. 1970. “Sellars’ Semantics”. Philosophical Review 79 (3): 404–19. Holton, Richard. 2000. “Minimalism and Truth-Value Gaps”. Philosophical Studies 97 (2): 137–68. Kölbel, Max. 2002. Truth Without Objectivity. London: Routledge. Lewis, David. 1972. “Psychophysical and Theoretical Identifications”. Australasian Journal of Philosophy 50 (3): 249–58. Lynch, Michael. 2004. “Truth and Multiple Realizability”. Australasian Journal of Philosophy 82 (3): 384–408. ———. 2009. Truth as One and Many. Oxford: Oxford University Press. MacFarlane, John. 2014. Assessment Sensitivity: Relative Truth and its Applications. Oxford: Oxford University Press. O’Shea, James. 2007. Wilfrid Sellars. Cambridge: Polity Press. Peregrin, Jaroslav. 2014. Inferentialism: Why Rules Matter. Basingstoke: Palgrave Macmillan. Price, Huw. 1992. “Metaphysical Pluralism”. Journal of Philosophy 89 (8): 387–409. ———. 2013. Expressivism, Pragmatism and Representationalism. Cambridge: Cambridge University Press. Putnam, Hilary. 1978. Meaning and the Moral Sciences. London: Routledge and Kegan Paul. Quine, W.V.O. 1970. Philosophy of Logic. Cambridge, MA: Harvard University Press. ———. (1999) 2008. “Where Do We Disagree?” In Quine in Dialogue, edited by Dagfinn Føllesdal and Douglas B. Quine. Cambridge, MA: Harvard University Press, 159–66. Rorty, Richard. 1979. Philosophy and the Mirror of Nature. Princeton: Princeton University Press. Rosenberg, Jay. 2007. “Sellarsian Picturing”. In Wilfrid Sellars: Fusing the Images. Oxford: Oxford University Press. Seibt, Johanna. 1990. Properties as Processes: A Synoptic Study of Wilfrid Sellars’ Nominalism. Atascadero: Ridgeview. Sellars, Wilfrid. 1947 [PPE]. “Pure Pragmatics and Epistemology”. Philosophy of Science 14 (3): 181–202. ———. (1952) 1963 [P]. “Particulars”. In Wilfrid Sellars. 1963. Science, Perception, and Reality, 282–97. London: Routledge and Kegan Paul. Hereafter SPR. ———. (1953) 1963 [ITSA]. “Is There a Synthetic A Priori?” Revised version of 1958. In SPR, 298–320. ———. (1953) 2007 [IM]. “Inference and Meaning”. In In the Space of Reasons: Selected Essays of Wilfrid Sellars, edited by Kevin Scharp and Robert Brandom, 3–27. Cambridge, MA: Harvard University Press. ———. (1954) 1963 [SRLG]. “Some Reflections on Language Games”. Expanded version of 1955. In SPR, 321–58. ———. (1956) 1963 [EPM]. “Empiricism and the Philosophy of Mind”. In SPR, 127–96. ———. 1957 [CDCM]. “Counterfactuals, Dispositions, and the Causal Modalities”. In Minnesota Studies in the Philosophy of Science, vol. 2, edited by

Sellars, Truth Pluralism, and Truth Relativism

205

Herbert Feigl, Michael Scriven, and Grover Maxwell, 225–308. Minneapolis: University of Minnesota Press. ———. 1962 [TC]. “Truth and ‘Correspondence’”. Revised version. In SPR, 197–224. ———. 1963 [EAE]. “Empiricism and Abstract Entities”. In The Philosophy of Rudolf Carnap, edited by Paul Schilpp, 431–68. LaSalle, IL: Open Court. ———. 1964 [NI]. “Notes on Intentionality”. Journal of Philosophy 61 (21): 655–65. ———. (1965) 1967 [SRI]. “Scientific Realism or Irenic Instrumentalism: A Critique of Nagel and Feyerabend on Theoretical Explanation”. In Philosophical Perspectives: Metaphysics and Epistemology, 337–69. Atascadero, CA: Ridgeview. ———. 1966 [FD]. “Fatalism and Determinism”. In Freedom and Determinism, edited by Keith Lehrer, 141–74. New York: Random House. ———. 1968 [SM]. Science and Metaphysics. London: Routledge and Kegan Paul. ———. 1969 [LTC]. “Language as Thought and as Communication”. Philosophy and Phenomenological Research 29: 506–27. ———. 1973 [CC]. “Conceptual Change”. In Conceptual Change, edited by Glenn Pearce and Patrick Maynard, 77–93. Dordrecht: D. Reidel. ———. (1974) 2007 [MFC]. “Meaning as Functional Classification”. In In the Space of Reasons: Selected Essays of Wilfrid Sellars, edited by Kevin Scharp and Robert Brandom, 81–100. Cambridge, MA: Harvard University Press. ———. 1974 [RDP]. “Reply” (to Putnam and Dennett). Synthese 27 (3–4): 457–66. ———. 1979 [NAO]. Naturalism and Ontology. Atascadero, CA: Ridgeview. Pagination of 1996 printing. ———. 1985 [TTP]. “Towards a Theory of Predication”. In How Things Are, edited by James Bogen and James. E. McGuire, 285–322. Dordrecht: D. Reidel. ———. 2015 [WSNDL]. Notre Dame Lectures 1969–1986, revised edition, edited by Pedro Amaral and David Landy. Atascadero, CA: Ridgeview. ———. Sellars-Harman Correspondence (1970). Edited by A. Chrucky, online at www.ditext.com/sellars/sh-corr.html. ———. Sellars-Marras Correspondence (1972–76). Edited by A. Chrucky, online at www.ditext.com/sellars/csm.html. ———. Wilfrid S. Sellars Papers, 1899–1990, ASP.1991.01, Archives of Scientific Philosophy, Special Collections Department, University of Pittsburgh, online at https://digital.library.pitt.edu/collection/wilfrid-s-sellars-papers. Shapiro, Lionel. 2013. “Intentional Relations and the Sideways-on View: On McDowell’s Critique of Sellars”. European Journal of Philosophy 21 (2): 300–19. ———. 2014. “Sellars on the Function of Semantic Vocabulary”. British Journal for the History of Philosophy 22 (4): 792–811. Shapiro, Stewart. 2014. Varieties of Logic. Oxford: Oxford University Press. Strawson, P.F. 1949. “Truth”. Analysis 9 (6): 83–97. Tarski, Alfred. (1933) 1983. “The Concept of Truth in Formalized Languages”. Translated by J.H. Woodger. In Logic, Semantics, Metamathematics, 2nd edition, edited by John Corcoran, 152–278. Indianapolis: Hackett. Williams, Michael. 2013. “How Pragmatists Can Be Local Expressivists”. In Price 2013, 128–44. ———. 2016. “Pragmatism, Sellars, and Truth”. In Sellars and His Legacy, edited by James O’Shea, 223–60. Oxford: Oxford University Press.

206

Lionel Shapiro

Wright, Crispin. 1992. Truth and Objectivity. Cambridge, MA: Harvard University Press. ———. 1996. “Response to Commentators”. Philosophy and Phenomenological Research 56 (4): 911–41. ———. 1999. “Truth: A Traditional Debate Reviewed”. In Truth, edited by Simon Blackburn and Keith Simmons, 203–38. Oxford: Oxford University Press. ———. 2013. “A Plurality of Pluralisms”. In Truth and Pluralism: Current Debates, edited by Nikolaj Pedersen and Cory Wright, 123–53. Oxford: Oxford University Press.

10 Some Remarks on Sellars’s Theory of Experience Willem A. deVries

I have three points I want to make, all critical of Sellars’s explicit doctrine, though not of his orientation, plus an issue I want to float for discussion, because I don’t know its proper resolution.1

1 My first complaint, briefly stated, is this: Sellars committed a significant blunder in his continued talk in “Empiricism and the Philosophy of Mind” (EPM) of so-called immediate experience. It is understandable why he initially uses the phrase “immediate experience”: he took it to be a phrase in common philosophical parlance that was often used in discussions of the nature of both our sensory acquaintance with objects in the world and the supposed foundations of our cognitive access to the empirical world. Of course, in the empiricist tradition that Sellars was trying to come to terms with, those two roles—expressing sensory acquaintance and providing epistemic foundations—were regularly conflated. And therein lies the problem: depending on the context, “immediate experience” can be a matter of sensory or phenomenal presence, or it can be a matter of non-inferential cognition. But Sellars has only the first of these uses in mind for the phrase. He also uses “sensation”, and, most often in EPM, at least, “sense impressions” for what he intends “immediate experiences” to pick out. Had he made the identification of immediate experiences with sense impressions clear at the outset and then consistently moved to the latter expression, I think a great deal of confusion would have been avoided, because sense impressions are actually not (in Sellars’s lingo) experiences in their own right at all. By continuing to use “immediate experience” Sellars encouraged people to think that sensations, impressions, etc., are experiences, non-inferential cognitions, in their own right. Sellars himself says that it is “unfortunate” that such inner episodes as sensations and feelings have been called “immediate experiences” (EPM §45: 176), so one has to wonder why he continued a practice he considers unfortunate. What Sellars should have said is that there are no immediate experiences— at least no immediate experiencings—in the dual sense intended by the

208 Willem A. deVries tradition he is attacking. That would have been a bold pronouncement and surely would have generated much push back from empiricists, who keep hoping that there is some form of experience that is merely sensory, pre-conceptual, or immediate and yet susceptible to cognitive assessment and capable of playing a direct role in the justification game (i.e., possesses what Sellars calls “epistemic aboutness”—i.e., intentionality—(EPM §29: 161)). In the long run, denying immediate experiencings would have been clearer than the course Sellars did choose, for the relevant form of mediation is conceptuality, which hinges not on being the product of inference or reasoning, but on having a position or role to play in a superordinate network of normatively constituted and licensed statuses and moves. All cognitive experience (and all “epistemic aboutness”)—even immediate (that is, non-inferential) cognition—involves such mediation. The mediatedness of the conceptual should not be modeled on forms of causal mediation in a mechanical chain of causation. This is why Sellars’s talk about the “logical space of reasons” has gained so much uptake. We can see how his continued use of the phrase “immediate experience” might be misinterpreted by just looking at some cases. The phrase shows up for the second time in EPM when Sellars points out that perceptual experiences have something more to them than the spontaneous occurrence to the subject of a propositional claim, even if “wrung” from the subject. The something more is clearly what philosophers have in mind when they speak of “visual impressions” or “immediate visual experiences”. What exactly is the logical status of these “impressions” or “immediate experiences” is a problem which will be with us for the remainder of this argument. (EPM §16bis: 145) Now, when Sellars talks of the logical status of immediate experiences, he does not mean to imply that immediate experiences themselves have logical properties. He means that, just as one must worry about the “logical”, that is, categorial, status of physical objects, of events, and of obligations, etc., the philosopher has to find the appropriate categorial status for impressions, aka “immediate experiences”. But might not someone take Sellars’s turn of phrase to imply precisely that “immediate experiences” have a logical status in the sense that they possess on their own certain logical powers, implying some things, conflicting with others? This could be read into one of the next appearances of the phrase “immediate experience” in EPM: Let me begin by formulating the question: “Is the fact that an object looks to S to be red and triangular, or that there looks to S to be a red and triangular object over there, to be explained in terms of the idea

Sellars’s Theory of Experience

209

that Jones has a sensation—or impression, or immediate experience— of a red triangle?” One point can be made right away, namely that if these expressions are so understood that, say, the immediate experience of a red triangle implies the existence of something—not a physical object—which is red and triangular, and if the redness which this item has is the same as the redness which the physical object looks to have, then the suggestion runs up against the objection that the redness physical objects look to have is the same as the redness physical objects actually do have, so that items, which ex hypothesi are not physical objects and which radically, even categorially, differ from physical objects, would have the same redness as physical objects. (EPM §21: 149–50; bold emphasis mine) Sellars means that an immediate experience implies the existence of something in the same kind of way that wet streets imply the existence of some liquid, but for one leaning toward the view that sense impressions have cognitive properties of their own, the misreading that immediate experiences themselves possess logical properties offers itself easily. I don’t want to go through all his uses of the phrase “immediate experience” in EPM. That might be a good exercise, but slow going. I do want to push a little further to the fourth use of the phrase, in order to correct a misconception I used to have and thus fear might be shared by others. When Sellars lays down his early, broad hint that he’s going to be proposing something outré—that sense impressions or “immediate experiences” are like theoretical entities—he points out that Certainly, those who have thought that qualitative and existential lookings are to be explained in terms of “immediate experiences” thought of the latter as the most untheoretical of entities, indeed, as the observables par excellence. (EPM §22: 150) The view that “immediate experiences” are the observables par excellence is what he then turns to consider, namely, the notion that if we consider lookings or appearances, we simply find in them components that are properly referred to as “the immediate experience of a red triangle”. I used to think that Sellars here was rejecting the notion of finding things in experience altogether, that everything in experience is, to some degree postulational, but I’ve finally realized that his point is the same he later makes by saying that sheer phenomenology shows us that in a perceptual experience of, e.g., a pink ice cube, “something, somehow a cube of pink in physical space is present in the perception other than as merely believed in” (SRPC §35: 437). The point expressed in his later work by saying that “sheer phenomenology or conceptual analysis takes us part of the way, but finally lets us down” (loc. cit.), is expressed in EPM as

210 Willem A. deVries “Thus, the very nature of ‘looks talk’ is such as to raise questions to which it gives no answer” (EPM §22: 152). In that sense something is “found” in experience, but we then have to find (by theorizing) how and where it fits into our broader understanding. Sellars’s continued use of “immediate experience” encourages, it seems to me, the collapsing of the distinction between nonconceptual elements in experience and nonconceptual experience. Again, Sellars believes strongly that there are nonconceptual elements in experience, but denies resolutely that there is any such thing as nonconceptual experience. Failure to notice the difference is sure to generate lots of heat and little illumination.

2 Next, I want to take aim at an interesting but obscure claim that shows up in some fairly late Sellars articles, namely, that perceptual claims are almost universally wrong in that we take what is in fact a sensory state of ourselves, a sensation, to be some aspect of a physical object or event. Our perceptual lives are, then, riven with mistakes, or more precisely, mis-takings. Actually, I think there is a puzzle about this claim. It is presented in “Sensa or Sensings: Reflections on the Ontology of Perception” (SSOP), which was originally a talk given at a colloquium at UNC in honor of James Cornman in 1976, although it wasn’t published until 1982. This makes it contemporaneous with “Some Reflections on Perceptual Consciousness” (SRPC), published in 1977, and thus presumably written also around 1976. But SRPC does not take that line at all. In SRPC Sellars models his conception of perceptual takings, which are elements in all perceptual consciousness, on complex demonstrative phrases, especially ones that include perspectival orientations of and on their purported objects. He then worries about what he calls “saving the reference” of these complex demonstrative phrases and the perceptual takings that are modeled on them. He suggests in SRPC that in the case where the perceptual taking is unsuccessful, and there is no object “out there” to be taken, the reference of the taking can nonetheless be “saved” by saying that the complex demonstrative phrase, the perceptual taking, refers to the sensation that is involved in the perceptual experience. In SRPC it is clear that identifying the reference of the perceptual taking as the sensation involved in the perceptual experience is a last-ditch move. Sellars says, while the referent of the most cautious perceptual taking can be construed as a sensation, we need not conclude that the referent of all perceptual takings is a sensation. (SRPC §61: 441) It is not difficult, in the context of SRPC, to see why Sellars wants to save the reference of perceptual takings, and yet need not try to specify

Sellars’s Theory of Experience

211

a one-size-fits-all kind of referent that is universally referred to in all perceptual takings. The basic motive for saving the reference of perceptual takings is the fact that Sellars is committed to the validity of what I will, for convenience’s sake, call the phenomenological perceptual principle [PPP]. This is the principle, familiar now to all who study Sellars, that mere phenomenology (or conceptual analysis—these are, for Sellars, equivalent), when turned on perceptual experience itself (say, e.g., the visual perception of his famous pink ice cube), suffices to assure us that Something, somehow a cube of pink in physical space is present in the perception other than as merely believed in. (SRPC §35; SSOP §26; FMPP I §91; SK I §55) This principle, in Sellars’s view, makes a serious ontological commitment. As he puts it in “The Lever of Archimedes”, whatever its “true” categorial status, the expanse of red involved in an ostensible seeing of the very redness of an apple has actual existence as contrasted with the intentional in-existence of that which is believed in as believed in. (FMPP I §88: 20–1) So, according to Sellars, whenever we have a perceptual experience (e.g., an ostensible seeing) (i) there is a perceptual taking, and (ii) there is something actual being taken. Since the conception of a perceptual taking is modeled on a complex demonstrative phrase, there has to be something being successfully referred to by the taking. If there were no way to “save the reference” of the taking, PPP would have to be given up. Sellars’s fundamental conception of the nature and structure of perceptual experience would have to be revised radically. But, as I have already pointed out, in SRPC, “saving the referent” of the taking by identifying it as the sensation involved in the perception is a last-ditch effort. Since we generally do a pretty good job of perceiving what’s around us, in most takings the something that is somehow present to us other than as believed in is, in fact, an actual physical object or state of affairs, for example, an actual pink ice cube. An actual pink ice cube is somehow a cube of pink in physical space in the most straightforward way, literally and non-analogically. Thus, in the view of SRPC, what we most often perceptually take to confront us are the medium-sized dry goods that populate the Manifest Image, and this raises no particular problem, because, most of the time, we’re right and the reference in our taking is successful on its own terms. I think it is a non-trivial point that when our perceptual takings are successful on their own terms, they can meld smoothly with the rest of our beliefs about our situation and be taken up into our deliberations concerning action, i.e., how to realize our intentions, given our current

212 Willem A. deVries situation. What is important to realizing my intention to get to class on time is that when I take there to be a clock telling me that there are five minutes to go and I take there to be a doorknob before me as I head out of my office, I’m reliably right. That I have a particular set of proper and common sensibles present to me in consciousness is at best secondary, for the particular sensibles I encounter will vary depending on the illumination in my office (which varies with the seasons, the time of day, and the age of my light bulbs) and on what my exact position in the office is when I start worrying about getting to class. As Sellars himself makes much of in “Phenomenalism”, there is no way to formulate in a sense-datum (or sensum-) specific vocabulary my intentions about getting to class on time or the background and context-specific knowledge that I employ in deliberating about or realizing those intentions. The story we get in SSOP, however, deviates from this fairly significantly. Seeing how it differs is only the wind up to asking why it differs. SSOP is, I find, a pretty turgid piece, sunk deeply into Sellars’s involuted and technical arguments with James Cornman concerning the status of sensations. The basic argument concerns the relative advantages of construing sensory awarenesses, appearances, as relations to an object, a sensum, or instead construing them as objectless sensings that differ adverbially from each other, rather than by relating to different objects. Sellars’s arguments, which I don’t think I even yet fully follow, seem to drive towards the conclusion that, when the analysis proceeds carefully and conscientiously, the sensum and the objectless sensing stories turn out to be so close to each other structurally that there is not really a significant difference between them, especially if one also moves to a process ontology in which sensa are re-categorized in terms of absolute processes. What blocks an easy assimilation of the sensum construal of the sensory to the objectless-sensing construal is, obviously, the significant difference in the fundamental grammar of the attribution/description of these states. That is, sensum theorists are committed to the relational form Sellars formulates as [Sensum]

(∃x) s(red rectangle)x & R(P,x)

where s(red rectangle) expresses the property, not of being a physical red rectangle, but of being red and rectangular in an analogical sense that applies to the objects of one’s direct sensation. “P”, of course, simply stands for a person. Objectless-sensing theorists, in contrast, claim that the fundamental form of attributions/descriptions of sensory states is [Sensing]

P s(red rectangle)s

where s(red rectangle) is a verb expressing the activity of sensing red rectangly.

Sellars’s Theory of Experience

213

Sellars offers that one way to bridge the gap between the sensum and objectless-sensing theorists would be to find some individual, say perhaps, an instance of sensing, attributions of which would permit both relational and verbal transformations, explaining why both sensum and objectless-sensing theories have attractive power. Sellars ultimately doesn’t think this line is “promising”. Says Sellars, But the verbs in question are so closely tied to persons as subjects of which the verbs are predicated that it is difficult to see how the gap between Jones’ sensa which are tied to perceivers by causal ties and Smith’s objectless sensing, which are logically tied to perceivers, can be bridged. (SSOP §78: 105) He mentions as well that this is a problem that plagued Hume’s attempt to articulate and defend a bundle theory of persons. The words “perception” and “thought” are so tied to the context “(person) perceives”, “(person) thinks” that to say that persons consist of perceptions and thoughts has an inescapable air of vicious circularity. (SSOP §79: 105–6) But, in fact, Sellars proposes a very similar move, only he tries to take care of that apparent vicious circularity by moving to an ontology of absolute processes. By assimilating verbs of sensation to subjectless absolute processes, he thinks he can bridge the gap between the two theories of sensation. What’s been standing in the way of a proper view of the sensory has not been a disagreement about the object of sensation, but, really, a stubborn continuing belief in the subject of sensation. So construed, sensory episodes are subjectless absolute processes that occur as constituents of persons. Now, it seems clear that such a proposal is revisionary, though Sellars seems to think it is revisionary within the manifest image (SSOP §82: 107). It is this revisionary view that Sellars then summarizes at the end of SSOP: [W]hen there appears to P to be a physical object with a red and rectangular facing surface: (a) There is a sensum or objectless sensing which is an s(red rectangle) or an s(red rectangle)ing; (b) P is aware of (conceptually refers to) this sensum or sensing; (c) P mistakes this sensum or sensing to be the red and rectangular facing surface of a physical object. (SSOP §94: 111)

214 Willem A. deVries This view is to be compared with the view I’ve found in SRPC in the following ways. In both views, there is, in any ostensible sensing, a sensum or objectless sensing (it doesn’t matter which we choose for these purposes), that is, something actual that is present in the experience. In the SRPC account, though there is reference to (i.e., awareness of) something in the experience, it need not be the sensum or sensing that is referred to, and therefore there need not be any mistaking involved. In the SSOP account our perceptual lives are replete with mis-takings. I am bothered by the Prichardian account in SSOP, according to which our perceptions are almost always mis-takings of our sensations. For one thing, it is far from obvious that, revisionary as it is, it remains within the manifest image, though Sellars claims the analysis in SSOP remains within the manifest image (see as well endnote 2, page 111). For another, it doesn’t comport well with my understanding of Sellars’s anti-givenist, realistic epistemology; it mixes the normative statuses central to conceptuality and epistemology with the causal relations involved in sensibility in ways that I fear threaten the rejection of the given. Let me try to articulate these worries further. There are several reasons for thinking that the view in SSOP is more than a mere revision of the manifest image framework. First, the boundaries between the manifest and the scientific images have to be much fuzzier than made evident in their portrayal in PSIM, particularly since they are in a dynamic, diachronic relation. Consider, for instance, fairly standard treatments of the so-called secondary qualities, treatments that posit that the proper sensibles are mind-dependent in a way in which the common sensibles, the primary qualities, are not. These theses have become relatively commonplace (Sellars himself adopts the view) and thus could be taken now to be elements of the Manifest Image. But it is pretty clear that the origin of the primary/secondary quality distinction is a response to the rise of the new sciences in the early seventeenth century, and Sellars recognizes this (see, for instance, PSIM 29). Indeed, he spent a great deal of time worrying about the problems that arise from drawing such a distinction between different groups of sensible qualities. A major motivation in his treatment of the sensory concerns how best to re-unite color and shape in a way that is responsive to modern scientific theory as well as the basic grammar of descriptive talk concerning physical objects and perceivers. The kinds of pressures that both the sensum theory and the objectlesssensing theory of sensory states respond to, and even more so the move to reconceiving sensations in terms of absolute processes, come more from the sciences than from within the manifest image. Thus, the result of responding to these pressures is as much a step in the construction of the scientific image of humanity-in-the-world as it is a revision of the manifest image. Indeed, it seems to me, the moves made in SSOP are precisely moves that take us out of the manifest image and into the scientific image, which, of course, is still very much under construction.

Sellars’s Theory of Experience

215

To begin with, recall that construing our perceptual takings to be, on the whole, successful helps us understand how perception meshes with our knowledge of the world and our ability to act sensibly, if not always successfully, within it. The view in SSOP makes it much more difficult to see why we have or even how we could have the success in cognition and action that we have. According to SSOP, our perceptual encounters with the world are built on a large and pervasive layer of mis-takings. These are not mis-takes in the sense of noise in the data stream, but in the sense that the data stream pervasively and systematically points us in the wrong direction. It would take a great deal of sophistication to see how to move from a stream of constant, pervasive mistakes to a deep understanding of the fundamental reality of the world. As Sellars himself notes “we were given our perceptual abilities not for the purpose of ontological insight, but to enable us to find our way around in a hostile environment” (SSOP §89: 109). It is, in fact, science that Sellars turns to for deeper “ontological insight” into the structures of the empirical world. For the purposes of a correct third-person view, a scientific view, of what’s going on in perception, the theory articulated in SSOP might suit the bill, but for a framework used in the first-person to engage the world and other people via action it would be, at best, clunky and ill-suited. We can approach the conclusion that the SSOP view abandons the manifest image altogether from another angle as well. In several different ways, Sellars tells us that in the manifest image persons are basic entities. In PSIM he tells us first that “there is an important sense in which the primary objects of the manifest image are persons” (PSIM 9), and follows that up by saying that “it would be incorrect to construe the manifest image in such a way that persons are composite objects” (PSIM 11). Yet Sellars’s efforts in SSOP and FMPP (the Carus Lectures) are aimed, in part, at making room for a genuine bundle theory of persons, a framework in which persons are decidedly not primitive or basic objects, but highly complex structures. Such a framework may be necessary if we are to accommodate the result of scientific investigations into humanity and their place in the world, but how can one say that such a view is a mere revision of the manifest image? If the key to understanding the true place of the sensory in the world is abandoning the stubbornly persisting belief in a subject of perception, which, I take it, would also entail a more general abandonment of the subject and subjectivity, then it seems to me this would count as an utter rejection of the manifest image, not a revision of it.2 As I’ve argued elsewhere, I remain unconvinced that the scientific image could ever replace the manifest image in the wholesale manner that Sellars sometimes seems to promote.3 The scientific image is relentlessly third-personal and objective, but the first-personal, subjectivity-engaging aspects of the manifest image cannot be dispensed with in any actually usable conceptual framework.

216 Willem A. deVries I said that I suspect that there is something epistemologically off about the view that it is correct to say that we almost always mis-take sensa or sensings to be physical objects or their properties. I’m going to leave the detailed epistemological issues for another day, but I will drop some hints to indicate the direction my thoughts take on this issue. I start with the claim that the manifest image is a framework with respect to which there is truth and falsehood (PSIM 14). Second, as anti-foundationalist as he may be, Sellars still insists that “There is clearly some point to the picture of human knowledge as resting on a level of propositions— observation reports—which do not rest on other propositions in the same way as other propositions rest on them” (EPM §38: 170). But to play that supporting role, observation reports and their perceptual beliefs need to be generally true-in-the-framework. If they are built on mis-takings, I don’t see how that could work out. My reasoning is as follows. Judgments such as “The tomato is red” form the basis on which we conclude further judgments such as “The tomato is ripe and ready to eat”, and these latter judgments guide our actions in the world.4 If it turns out that what is really red is not the tomato but a sensum, then, at very least, the inference to the ripeness and edibility of the tomato must be vastly more complicated, if it isn’t simply broken altogether. If what is going on is that a red sensum is being mis-taken to be the red surface of a distal tomato, it would seem that the inference to the ripeness and edibility of the tomato appears sound only to those ignorant of the mistake. If there is a connection between the immediate presence-to-consciousness of a red sensum and the (mediate) presence in one’s environment of a ripe tomato, it would be a far less reliable connection, for epistemic purposes, than the connection between a tomato’s being red and its being ripe. What Sellars really needs to say is something like this: while the manifest image observation report or perceptual belief is true in the manifest framework, its scientific image successor will be a much more complex affair with a far finer grain, and in that successor, there will be a sensum or sensing referred to that is (i) the immediate cause of the conceptual response and (ii) responsible for the phenomenology of the total mental state. There are certainly further epistemological complications when one takes into account the theory/observation distinction, which is only methodological and not ontological, but I am going to leave those for another day.

3 My third and last correction of Sellars is this: Sellars improperly succumbs to the dominance of the visual in his worrying about the presence

Sellars’s Theory of Experience

217

in one’s experience of the pink ice cube of its pinkness and its coolth (e.g., SRPC part III; IKTE §§20ff., §§32–4): 30. Of course, in seeing it as a pink ice cube we are seeing it as having the causal properties characteristic of ice. But do we see of the object these causal properties? The question is, to be sure, an awkward one. But if we consider certain other questions of a similar form, the answer seems to be no. 31. Thus, by contrast, we not only see that the ice cube is pink, and see it as pink, we see the very pinkness of the object; also its very shape—though from a certain point of view. 32. Even more interesting is the fact that in seeing the cube as a cube of ice we are seeing it as cool. But do we see of the cube its coolness? Here we are torn in a familiar manner. On the one hand we want to say that the pinkness and coolness are “phenomenologically speaking” on a par, and are tempted to say that the idea that we don’t see its coolness is a matter not of phenomenology, but of scientific theory. On the other hand, when asked point blank whether we see its coolness, we find an affirmative answer intuitively implausible and are tempted to fall back on the idea that its coolness is believed in, i.e., that it is taken as cool. 33. Obviously we want to say that the ice cube’s very coolness is not merely believed in, even though its very coolness is not seen. It clearly won’t do to say that we feel or imagine a coolness on seeing the cube, what is in question is its coolness. 34. But though this topic is of great intrinsic importance, it must be postponed to another occasion, although I believe that the framework I am about to develop—enriched with an account of synesthesia— provides the key to the answer. (SRPC §§30–34: 436) I don’t see that synesthesia is really relevant here. What I think Sellars should say here is the following: I accept without further ado that, in good conditions “we not only see that the ice cube is pink, and see it as pink, we see the very pinkness of the object; also its very shape—though from a certain point of view” (loc. cit.). The seeming conflict over seeing the cube’s coolth presents itself only because Sellars continues to focus on the visual. If I pick up the ice cube with my bare hand, I feel its very coolth. Sellars seizes on cool as an example of a causal property, but it is at best an impure example, for hot and cold, or heat and coolth, are proper sensibles of touch. Pinkness and coolness are, “phenomenologically speaking”, on a par, but they belong to different sensory modalities, so it sounds odd to say that I see the very coolth of the cube, but at least as odd to say that I feel its very pinkness. Sellars could, and I argue should

218 Willem A. deVries have fleshed out his thought by introducing the notion of a sense-image model discussed in IKTE. The sense-image model we construct in our experience of the ice cube unifies not just the visually sensible properties of the cube but the tactilely sensible properties as well. So I don’t see a problem with saying that coolth is actually present in one’s experience, though if one is not touching the cube, it is present by way of the productive imagination. To pose the question of the presence of causal powers in sensory experience more cleanly, Sellars needs to find a causal power that is not tied directly to any proper or common sensible, something like solubility, or maybe in the case of ice, its refraction index. And then it seems clear that we should deny that we see or feel of a sugar cube its very solubility or of an ice cube its refraction index.

4 I want to leave you with a question to which I do not know the answer, and I’d love to get some help. Sellars’s notion of sense-image-models has recently garnered some increased attention,5 and we’ve just seen some of the reason why: they offer an interesting resource for the analysis of perceptual and imaginative experience. Now, in his discussion of sense-image-models, Sellars tells us that we must distinguish between imagining and imaging, just as we do between perceiving and sensing, and that, indeed, the “distinction to be drawn is essentially the same in both cases” (IKTE §23: 458). Roughly[,] imagining is an intimate blend of imaging and conceptualization, whereas perceiving is an intimate blend of sensing and imaging and conceptualization. (IKTE §23: 458) So, Sellars tells us that imagining a cool, juicy, red apple (as such) is a matter of (a) imaging a unified structure containing as aspects images of a volume of white, surrounded by red, and of mutually pervading volumes of juiciness and coolth, (b) conceptualizing this unified image-structure as a cool juicy red apple. (IKTE §23: 458) Whereas seeing a cool, juicy, red apple involves (a) sensing-cum-imaging a unified structure containing as aspects images of a volume of white, a sensed half-apple shaped shell of red, and an image of a volume of juiciness pervaded by a volume of white; (b) conceptualizing this unified sense-image structure as a cool juicy red apple. (IKTE §24: 459)

Sellars’s Theory of Experience

219

It is evident in all this, then, that imaging and sensing are not the same thing. So my question is: just how do imaging and sensing differ? Sellars drops a hint in the following paragraph (§25) that the difference is Hume’s ubiquitous standby: a difference in vividness. I can’t say I find this very satisfactory. On the one hand, the fact of eidetic images speaks to the possibility that the difference between sense and image may only be quantitative, but if that were really the case, wouldn’t we be much more likely to confuse sensing and imaging than we are? Maybe I suffer from a blighted imagination, but it seems to me that the difference between what I imagine and what I sense is really qualitative and not just quantitative. In any case, the distinction is not something that should be taken to be a matter of course. When the juicy coolth of the apple is present in my perception, though only by way of imagination, just how does that presence differ from the sensed half-apple shaped shell of red?

Notes 1. I proposed my title in tribute to Sellars’s classic article “Some Remarks on Kant’s Theory of Experience” (KTE), but it functions differently for me. Sellars’s article is broad, even synoptic, in its reach, dealing with very general structural features of Kant’s theory. I hear in Sellars’s title an echo of H.J. Paton’s title, “Kant’s Metaphysic of Experience”. I intend to focus much more narrowly on some issues in Sellars’s treatment of sensory experience in particular, and this paper is more “remarkish” in structure. 2. My thinking is that perception and action are tied together (the motive soul and the perceptive soul are one and the same). So abandoning the subject of perception abandons the subject of action, and then there is no need for the “language of individual and community intentions”. That abandons precisely that aspect of the manifest image that Sellars says needs to be retained and joined with the scientific image. 3. See deVries (2012). 4. One might cavil here: isn’t ripeness just as much an observable property as redness, so that we need not base our ripeness judgments on judgments of redness? Perhaps so, but Sellars himself distinguishes between what we see an object as, which can include a fairly broad range of characteristics, and what we see of the object, which is much more tightly restricted to proper and common sensibles. It is then a challenge to the epistemologist to explain how, in cases of sophisticated observation, what we see an object as is grounded in what we see of the object. My example aims to point out that the Prichardian view generates a particularly difficult problem for understanding how what we see of an object can ground or justify what we see it as. And if we cannot explain that connection, the temptation will be to run back the notion that it is given. 5. See, for instance, Coates (2007) and Sachs (2017).

References Coates, Paul. 2007. The Metaphysics of Perception: Wilfrid Sellars, Perceptual Consciousness and Critical Realism. New York: Routledge. deVries, Willem A. 2012. “Ontology and the Completeness of Sellars’s Two Images”. Humana.Mente: Journal of Philosophical Studies 21: 1–18, online at www.humanamente.eu.

220 Willem A. deVries Paton, H.J. 1936. Kant’s Metaphysic of Experience. London: George Allen and Unwin. Sachs, Carl. 2017. “Sentience and Sapience: The Place of Enactive Cognitive Science in Sellarsian Philosophy of Mind”. In Sellars and Contemporary Philosophy, edited by D. Pereplyotchik and D. Barnbaum. New York: Routledge. Sellars, Wilfrid. (1956) 1963 [EPM]. “Empiricism and the Philosophy of Mind”. In Wilfrid Sellars. 1963. Science, Perception and Reality, 127–96. London: Routledge and Kegan Paul. Reissued by Ridgeview Publishing Company in 1991. Hereafter SPR. Reprinted in Willem A. deVries and Timm Triplett. 2000. Knowledge, Mind, and the Given: Reading Wilfrid Sellars’s “Empiricism and the Philosophy of Mind”. Indianapolis and Cambridge, MA: Hackett Publishing. ———. (1962) 1963 [PSIM]. “Philosophy and the Scientific Image of Man”. In SPR, 1–40. ———. 1967 [KTE]. “Some Remarks on Kant’s Theory of Experience”. The Journal of Philosophy 64: 633–47. ———. 1975 [SK]. “The Structure of Knowledge: (1) Perception; (2) Minds; (3) Epistemic Principles”. In Action, Knowledge and Reality: Studies in Honor of Wilfrid Sellars, edited by Hector-Neri Castañeda, 295–347. Indianapolis: Bobbs-Merrill. ———. (1978) 2007 [IKTE]. “The Role of Imagination in Kant’s Theory of Experience”. In Wilfrid Sellars. 2007. In the Space of Reasons: Selected Essays of Wilfrid Sellars, edited by K. Scharp and Robert B. Brandom, 454–66. Cambridge, MA: Harvard University Press. ———. 1978 [SRPC]. “Some Reflections on Perceptual Consciousness”. In Wilfrid Sellars. 2002. Kant’s Transcendental Metaphysics: Sellars’ Cassirer Lectures and Other Essays, edited with introduction by Jeffrey F. Sicha, 431–41. Atascadero, CA: Ridgeview Publishing Co. Hereafter KTM. ———. 1981 [FMPP]. “Foundations for a Metaphysics of Pure Process” (The Carus Lectures). The Monist 64: 3–90. ———. 1982 [SSOP]. “Sensa or Sensings: Reflections on the Ontology of Perception”. Philosophical Studies 41 (Essays in Honor of James Cornman): 83–111.

11 Sellars on Self-Knowledge Franz Knappik

In the last sections of Empiricism and the Philosophy of Mind, Wilfrid Sellars uses the famous Myth of Jones in order to present a novel framework for thinking about the mind. Central to Sellars’s proposal is an original account of the self-knowledge that we have regarding our own thoughts and sense impressions. In a penetrating and thought-provoking correspondence with Hector-Neri Castañeda, Sellars further develops his remarks on self-knowledge about thoughts in EPM into a rich and complex account. This account anticipates in many ways ideas and topics in the more recent literature on self-knowledge. It is my aim in this chapter to reconstruct Sellars’s views on self-knowledge, and to explore some of its relations to the contemporary debate. When discussing self-knowledge of one’s own present thoughts, Sellars builds on our intuitive notion of a “thought” that stands for occurrent mental states with conceptual content (as opposed to the dispositional attitudes that we can take towards such contents, for instance beliefs). Like other mental states and episodes, thoughts are normally considered as being accessible to their subjects in a particular way, which differs from the access we have to other persons’ thoughts. Thus, we seem to have an immediate or non-inferential knowledge of our own thoughts— for instance, we do not have to observe and interpret our own behavior, facial expressions etc. in order to know our thoughts. Moreover, we seem to be in a particularly good epistemic position with regard to our own thoughts. We would expect that our beliefs about what we are currently thinking are very likely true, while our beliefs about other persons’ thoughts may easily be mistaken.1 Sellars shares the view that ordinary self-knowledge about thoughts is special in these ways. Indeed, his entire account of self-knowledge is motivated by the ambition to do justice to our intuitions regarding those special features, while at the same time avoiding the Myth of the Given that, on Sellars’s view, traditional accounts of privileged access have fallen victim to. The chapter falls into three parts. In Section 1, I will reconstruct Sellars’s theory of self-knowledge about thoughts. Section 2 examines how Sellars’s discussion bears on Shoemaker’s and Burge’s influential arguments

222

Franz Knappik

against “perceptual” accounts of self-knowledge. In Section 3, I contrast Sellars’s view with recent work that assigns phenomenal consciousness an important role in the explanation of self-knowledge about thoughts, and I propose to amend his account by combining it with a Higher-Order Thought theory of phenomenal consciousness.

1 Sellars’s Account of Self-Knowledge of One’s Thoughts2 In the last part of EPM, Sellars uses the famous Myth of Jones in order to identify a middle ground between two extremes which he finds equally unattractive. The one extreme is occupied by classical views of the mind in both the rationalist and the empiricist traditions. According to such views, there really are inner episodes and states such as thoughts, perceptions, beliefs, and intentions. All or at least some of these inner phenomena are the object, on the traditional view, of a knowledge with special and perhaps even infallible authority, which can serve as a foundation for all further knowledge. Such views are not acceptable to Sellars, who has argued in the previous parts of EPM that they share a commitment to the Myth of the Given. Arguably, the basic claim of the Myth of the Given is that there can be entities that provide epistemic support to rational attitudes without being themselves epistemically dependent on anything else (deVries and Triplett 2000, xxvi). It is easy to see that the traditional assumption of a kind of self-knowledge that can serve a foundational role fulfills these criteria: on this assumption, the mere occurrence of certain mental states can justify beliefs about those states (which then can justify further beliefs), while the justificatory force of those states is not itself derived from any further factors. Opposed to such views of self-knowledge is the equally radical view of behaviorism. Behaviorism denies that inner states and episodes have any existence in their own right—rather, they are, on this view, reducible to behavioral episodes and dispositions. As a consequence, there can be no substantial epistemological difference between first- and third-personal access to inner states and episodes. Our knowledge about them is always inferred on the basis of observable behavioral evidence, and the firstpersonal case differs from the third-personal case only in that we normally have more such evidence available about ourselves than about other persons.3 Sellars rejects both the ontological and the epistemological claims of behaviorism: he wants to defend both the irreducible existence of inner states and episodes, and the idea that our first-personal access to the inner is special in some ways. In particular, Sellars thinks that we have indeed a noninferential, epistemically very good and distinctively first-personal access to our own thoughts and sense impressions. The last sections of EPM therefore face the task of showing how one can accept such a special first-personal access to really existing inner episodes without falling victim to the Myth of the Given. Sellars’s response

Sellars on Self-Knowledge

223

consists of three major moves. The first move is to interpret the vocabulary of inner episodes such as thoughts and sense impressions as being originally a theoretical vocabulary. Thoughts, in particular, are postulated in the Myth of Jones as theoretical entities that are meant to explain episodes of overt intelligent behavior. As Sellars is eager to emphasize, he does not understand the distinction between theoretical and observable entities as an ontological distinction: by classifying something as a theoretical entity, one does not imply that it does not really exist (EPM §43: 173–4). Rather, one assigns it a different place in one’s theories. Unlike an observable entity, a theoretical entity is not “definable in observational terms” (EPM §58: 187), but this neither entails that it is not part of reality, nor that there is not sufficient reason to believe in its existence (EPM §58: 187). As is often the case with theoretical concepts, Jones’s concept of a thought is formed by analogy with observational concepts—in the case of thoughts, concepts which designate episodes of overt speech (EPM §56: 186). Hence, according to Jones’s theory, all intelligent overt behavior is ultimately caused by inner episodes that are analogous to episodes of overt speech (EPM §56: 186).4 Sellars’s second move in accounting for first-personal access is to show how a first-personal use of Jones’s theoretical concept of a thought can come about. After Jones has taught his Rylean fellows how to explain each other’s behavior in terms of their thoughts, he trains them to apply the concept of a thought to themselves. First, the trainee—whom Sellars calls Dick—learns to use his own behavior as evidence for self-ascriptions of thoughts: Dick is conditioned to respond to such evidence in the same way as an observer would do. This yields a form of access to one’s own thoughts which does not yet display the special features of the firstpersonal perspective. It is not immediate, as Dick has to draw an explanatory inference from his own observed behavior to a thought that he postulates as an explanation for his behavior. And this form of access does not provide a particularly good epistemic situation, as an external observer can have exactly the same access to Dick’s thoughts—indeed, the conditioning that takes place here is precisely a process in which Dick is supposed to take over Jones’s essentially third-personal access to his thoughts. Nevertheless, Sellars thinks that if there is appropriate training, a new situation can arise which does include a distinctively first-personal access. In EPM, Sellars characterizes this new situation only very briefly as follows: And it now turns out—need it have?—that Dick can be trained to give reasonably reliable self-descriptions, using the language of the theory, without having to observe his overt behavior. Jones brings this about, roughly, by applauding utterances by Dick of “I am thinking that p” when the behavioral evidence strongly supports the

224

Franz Knappik theoretical statement “I am thinking that p”; and by frowning on utterances of “I am thinking that p”, when the evidence does not support this theoretical statement. (EPM §59: 189)

If this training succeeds, Sellars points out, the language that first had a purely theoretical use “has gained a reporting role”, and our ancestors “begin to speak of the privileged access each of us has to his own thoughts” (EPM §59: 189). However, the brief discussion of this topic in EPM fails to make it clear how this result can precisely be brought about, and what exactly sets apart the training that Sellars describes here from the first stage of Dick’s training, where he was conditioned to self-interpret his own behavior. Luckily for us, Hector-Neri Castañeda has challenged Sellars in an extensive correspondence to be more explicit about the details of his account of self-knowledge.5 As becomes clearer from that correspondence, Sellars assumes that privileged access comes about if the process of training establishes a causal connection between a subject’s thoughts and his self-ascriptions which does not involve perception of the behavior that had previously served as evidence for the selfascriptions. Thus, when discussing Castañeda’s analogy of a process in which Dick learns to report the presence of a colony of viruses in his kidney that causes determinate symptoms, he describes the following scenario: [1.] We ask Dick to say “I am in state φ” [i.e., in a state in which the viruses are present] every two minutes or so. [2.] Sometimes Dick is in state φ, sometimes he isn’t. [3.] When he is, we know that he is by observing symptoms Ui . . . Uj and using a well confirmed theory. [4.] We put him in a position where he cannot observe these symptoms. [5.] When he says “I am in state φ” and actually is in state φ we reward him; in the contrary case we give him a mild shock. [6.] It turns out that for some states φ (but by no means for all) we can bring about a connection between being in state φ and saying “I am in state φ”. (To Castañeda, Nov. 14, 1969, §4) As this passage shows, Sellars thinks that in the training that brings about immediate self-ascriptions, the trainee has to be precluded from perceptual access to the behavioral symptoms that warrant the trainer’s ascriptions of mental states. As a consequence, the training, if successful, will establish a direct causal link between the trainee’s first-order thoughts and his self-ascriptions which is not mediated by any observation of his own behavior.

Sellars on Self-Knowledge

225

If such a causal link can actually be established, Dick thereby acquires a novel way of self-ascribing his thoughts, which differs from the earlier, evidence-based way in two important respects. First, it is non-inferential: Dick can now self-ascribe his thoughts without having to draw explanatory inferences on the basis of observed behavior. Second, it is specifically first-personal insofar as it is based on a causal mechanism which (at least in worlds that are sufficiently similar to ours) can obtain only among the thoughts of the same person, not among different persons (to Castañeda, Apr. 3, 1961, §15). Once the subject has acquired this new way of making overt selfascriptions of thoughts, it is easy to see how a subject can also acquire an ability for reliable purely mental self-ascriptions, or “meta-thoughts”, as Sellars calls them. For according to Jones’s theory, a statement like “I am thinking that p” has as its immediate cause a thought with the same content. So the mechanism that is established by the conditioning in question will involve the causation of meta-thoughts anyway, and where the subject suppresses the overt self-ascription, a purely mental self-ascription will result. What remains to be done in the third part of Sellars’s account is to explain how precisely our self-knowledge is justified, or how our metathoughts acquire epistemic authority, and in particular how they can acquire the especially strong authority that seems to characterize normal self-knowledge. While he is silent about this part of his theory of selfknowledge in EPM, he develops it in some detail in the correspondence with Castañeda.6 As Sellars points out, “the key to the account I have given [in EPM] of the direct-noninferential knowledge of inner episodes is the apparatus I developed in discussing the status of noninferential perceptual knowledge” (to Castañeda, Apr. 3, 1961, §13). Put very briefly, Sellars thinks that non-inferential perceptual knowledge has an epistemic authority that depends on a background of assumptions and conceptual abilities. For instance, the epistemic authority attached to normal perceptual reports like “Lo! Here is a red apple!” presupposes a background which includes knowledge that such reports are reliable indicators of the actual presence of red apples under standard perceptual conditions, as well as the knowledge that currently, such standard conditions obtain. It is this background which enables competent subjects to draw so-called trans-level inferences like the inference from (Perc-1) I have uttered “Lo! Here is a red apple!” under standard conditions of perception to (Perc-2) There is good reason to believe that there is a red apple in front of me.

226

Franz Knappik

According to Sellars, the availability of such trans-level inferences—a subject’s ability to draw them—justifies the epistemic authority of the normal perceptual reports and beliefs. How can the epistemology of self-knowledge be explained along parallel lines? Sellars describes the trans-level inference for the case of selfknowledge as inference from a premise of the form (1) The thought that I am thinking that p has occurred to me in such and such a manner and context to a conclusion of the form (2) I am thinking that p (to Castañeda, Dec. 8, 1961, Ad II.2).7 In a related context, Sellars clarifies two important points. First, Sellars explains that the meta-thought (MT-1) I am thinking that it is raining is “self-referentially incorrigible”, i.e., it is automatically true insofar as the thought “It is raining” is itself a part of this meta-thought, and is thought by the subject whenever she thinks the meta-thought.8 The same holds, mutatis mutandis, for the third-order thought (MT-2) I am thinking that I am thinking that it is raining. It seems that Sellars wants to set this kind of trivial self-knowledge aside, and therefore prefers the preceding formulation (1) to a formulation like (MT-2) as means to self-ascribe a meta-thought.9 (Of course, Sellars must understand the phrase “has occurred” as referring to a period that started a very short time ago. One might say instead “has just occurred” to highlight this.) Second, Sellars seems to assume that the subject needs to specify a particular “manner and context” in premise (1) only insofar as she needs to rule out conditions of massively impaired rationality, in which an episode that seems to be a thought that p is actually not a thought, but merely “mental noise” (to Castañeda, Mar. 11, 1962, Ad 3c; Ad 6).10 We therefore can replace (1) by the conjunction of the following two premises: (1a) The thought that I am thinking that p has occurred to me (1b) My rationality is not currently massively impaired. Moreover, Sellars seems to assume that knowledge about (1b) is easily had. That our rationality is not presently massively impaired is an

Sellars on Self-Knowledge

227

assumption that is built in into every proper thought: for once again, if rationality would be massively impaired, there would be no real thoughts, but only “mental noise”. Now if a subject S has undergone the conditioning described earlier, it will actually be the case that (3) S is so conditioned that, if her rationality is not massively impaired, the thought that she is thinking that p occurs to her (ceteris paribus) only when she is actually thinking that p. But the truth of (3) does not on its own suffice in order for S to be entitled to make the preceding trans-level inference from (1a) and (1b) to (2). Rather, it is necessary that S also know that she is entitled to make this inference, or that (3) is true. For as in the case of perceptual justification, Sellars thinks that a subject’s belief that (2) is justified only if the subject is able to make the trans-level inference from (1a) and (1b) to (2), and this ability requires a knowledge about the legitimacy of such an inference (cf. EPM §35: 168). So how can a subject know that (3) is true? Sellars does not explicitly address the question, but his exchange with Castañeda does contain the resources for a possible answer. This answer builds on the assumption of an essential connection between rationality and the existence of a reliable causal link between thoughts and meta-thoughts. As Sellars writes: [N]on-inferential autobiographical knowledge of what one is thinking . . . is essential to rationality. In short, the ability to think in the full sense of the term involves the ability to meta-think, and the ability to have noninferential knowledge of what one is thinking. (to Castañeda, Dec. 8, 1961, Ad II.3) Why should it not be possible for a creature to have robust capacities for rational thought and intelligent behavior without having the resources to form meta-thoughts, and the ability for non-inferential knowledge of its thoughts? The idea that there is such a connection seems to derive from Sellars’s more general views on the nature of language use and epistemic activity. For one thing, Sellars holds that one becomes a “full-fledged member of the linguistic community” (MFC 86) only once one is able to think meta-thoughts about linguistic items, for only then is one able to teach one’s language to others (ibid.). In addition, Sellars famously claims that empirical knowledge is a “self-correcting enterprise” (EPM §38: 170). He seems to assume that someone can self-correct, if necessary, his judgments and inferences only if he is able to know in the first place what his judgments and inferences are; and he holds that inferential self-knowledge would not suffice for this purpose. While Sellars does not explicitly argue for this latter point, his motivation for it is possibly

228

Franz Knappik

that inferential self-knowledge of thoughts depends on appropriate evidence for the presence of a thought, that such evidence is not always available, and that we need to be able to exercise rational control also over all the thoughts for whose presence we have no sufficient evidence. Taken together, these points support the idea that rationality requires knowledge that inferences of the above form (from (1a) and (1b) to (2)) are normally legitimate (or that a general proposition like (3) holds). We therefore need to acquire such knowledge as part of the broader web of conceptual and inferential abilities that we are taught in the process of becoming rational beings. As Sellars puts it with respect to the conditioning process described earlier: “the conditioned person comes to conceive of himself as so conditioned” (to Castañeda, Mar. 11, 1962, Ad 6).11 If this is how Sellars understands the justification of our self-knowledge about our thoughts, the following question arises: does not the warrant that we get for second-order thoughts or beliefs with the content “I am thinking that p” through such trans-level inferences itself presuppose knowledge at an even higher order—namely, knowledge that a premise like (1a) is true? And if so, doesn’t Sellars’s account of self-knowledge fall victim to a vicious regress? Castañeda raises this worry in his letter of January 17, 1962 (§10). To this, Sellars replies that he has never claimed that one must actually draw a trans-level inference in order to be justified in believing its conclusion—rather, one only must be able to draw it: “I know a vicious regress when I see it” (to Castañeda, Mar. 11, 1962, Ad 10). Spelt out somewhat more explicitly, Sellars’s reply to the regress objection can be understood as follows. Our entitlement to self-knowledge requires indeed that, on a given occasion, we have knowledge about the first premise, (1a), in the trans-level inference whose availability justifies our self-knowledge. But in order for our belief in the first premise to be justified, it is enough that we are able to draw a further trans-level inference which has that premise as its conclusion. Thus, if challenged, we need to be able to justify the belief “The thought that I am thinking that p has occurred to me” through a further trans-level inference. Of course, the first premise of this further argument could be challenged, too, and the subject has to be able to defend this premise, too, by yet another argument of the same form. But since the availability of those additional arguments is sufficient in order for the subject to be justified in believing their conclusions—and ultimately, in believing the original conclusion “I am thinking that p”—no infinite regress of actual inferences arises. (Of course, Sellars is relying here on his anti-foundationalist view that we “can put any claim in jeopardy, though not all at once” (EPM §38: 170).) Notice, however, that this response to the regress objection works only on a particular understanding of what it means that a subject has justification because she is able to draw some appropriate inference. On one natural understanding of the phrase, this requires that the subject know the premises of the inference. This cannot be the understanding that Sellars

Sellars on Self-Knowledge

229

has in mind. For in that case, the subject would be able to justify the second-order premise (1a) The thought that I am thinking that p has occurred to me through a trans-level inference only if she already knew, and hence also believed, the third-level content (4) The thought has occurred to me that the thought has occurred to me that I am thinking that p. The same would apply to the justification for the subject’s knowledge about this latter, third-level content: that the subject be able to draw yet a further trans-level inference in support of this third-level belief would require her to possess a relevant fourth-level belief, and so on. Justification for self-knowledge about the first-order thought “p” would therefore require an infinite number of higher-order beliefs, and hence be psychologically impossible for finite minds. This consequence can only be avoided if the subject’s ability to draw the requisite trans-level inferences is understood in another way, namely such that it is enough for this ability if the subject is in a position to know the relevant premises. On this view, when I believe that (2) I am thinking that p, I have justification for this belief if I am in a position to know (1a), and am able to use this as a premise in a trans-level inference of the above form. In order to justify my belief in (1a) when challenged, I need to be in a position to know that (4), and be able to use this as a premise in a relevant trans-level inference, and so on. (This presupposes that lower-order thoughts normally trigger corresponding higher-level thoughts whenever a question about the existence of the lower-order thought arises; but it is natural to assume anyway that this is a feature of the causal mechanism that links thoughts to meta-thoughts.)12 So this is how I would reconstruct Sellars’s account of self-knowledge about our thoughts. There are several points in Sellars’s argument that can be challenged quite easily, but these I will mention here only to put them aside. First, the argument rests on a strong claim about the relation between rationality and self-knowledge. Against the idea that rationality requires self-knowledge, it may be argued that substantive forms of rationality can be realized in the absence of an ability for self-knowledge (cf. Peters 2014, 149 on autism). Nevertheless, claims like Sellars’s have been made quite frequently in the contemporary debate (e.g., Shoemaker 1994; Burge 1996; Moran 2001), and I will not challenge Sellars’s position on this front in what follows.

230

Franz Knappik

Furthermore, many philosophers and psychologists today would hold that causal mechanisms like the one that supports non-inferential selfknowledge on Sellars’s account may be innate as the result of evolutionary selection (e.g., Nichols and Stich 2003, 162, about a very similar mechanism). If this view is correct, there is no need to speculate about what kind of training may bring about such mechanisms. Finally, it is natural to object to Sellars’s account that it is overly demanding. In particular, do rationality and self-knowledge really require that one has background knowledge about the functioning of one’s rationality (premise (1b)), and hence also that one possesses a concept of rationality? It may be possible to weaken Sellars’s account in this respect: even by Sellars’s lights, it may suffice for the justification of selfknowledge that (1b) is true even though the subject doesn’t know this. But once again, further discussion of this point would go beyond the limits of this chapter.

2 Sellars and Arguments Against Perceptual Accounts of Self-Knowledge As we have seen, Sellars ascribes a central role in his account of selfknowledge to a causal mechanism that connects our first-order thoughts and our corresponding meta-thoughts, and that we acquire through appropriate training. Hence, Sellars holds a version of what Shoemaker 1994 calls a “broad perceptual model” of self-knowledge (cf. Haag 2001, 319– 24): a view that treats self-knowledge as analogous to perceptual knowledge insofar as it understands it as being based on causal link between a representation and its object. In the last 20 years or so, many authors have criticized perceptual views of self-knowledge (about thoughts, but also about other mental states), and proposed alternative, non-perceptual views instead.13 Probably the most influential arguments against perceptual accounts are the objections that have been raised by Sydney Shoemaker and Tyler Burge (Shoemaker 1994; Burge 1996). Both authors have argued that any adequate explanation of self-knowledge has to take into account the fact that there is a close connection between self-knowledge and rationality. The connection that they have in mind is very similar to the one we have encountered in Sellars’s discussion (although neither Shoemaker nor Burge seem to be aware of Sellars’s remarks on this connection): rationality requires the capacity to engage in self-critical reasoning, to control one’s own thoughts and attitudes in the light of one’s reasons. Yet such self-critical reasoning is possible only for a creature that is able to know what thoughts it is thinking, and what attitudes it has. Now this connection between rationality and self-knowledge, Burge and Shoemaker argue, is incompatible with a view on which self-knowledge is based on a causal link between first- and second-order states or episodes.

Sellars on Self-Knowledge

231

On Shoemaker’s version of the argument (Shoemaker 1994), such a view has to allow for the possibility that the causal link in question breaks down (resulting in a condition of “self-blindness”), so that the subject is rational, but not anymore capable of self-knowledge. Yet this implication is incompatible with the connection that obtains between rationality and self-knowledge, so the assumption of self-knowledge being based on a causal link must be resisted. Burge offers a related argument (Burge 1996). If the link between firstand second-order state were only causal, he argues, both could come apart, and we would have to expect the possibility that our second-order thoughts are mistaken without a fault of our own—the possibility of what Burge calls a “brute error”. In that case, however, detecting through one’s second-order judgment a shortcoming in one’s first-order states or episodes would not immediately give one reason to revise the first-order state or episode: the possibility of a brute error would always offer a possible excuse that releases one from one’s obligation to revise the firstorder state or attitude (Burge 1996, 109–10). Yet it is precisely central to self-knowledge in the context of critical reasoning, Burge argues, that the contents of that self-knowledge can give us immediate reason to revise our first-order states or attitudes. The ability to detect and hence amend shortcomings at the first-order level is exactly what self-knowledge must contribute to rationality. Therefore, Burge concludes, a non-causal, conceptual connection has to obtain between knowledge and its objects in the case of rational self-knowledge. Strictly speaking, Shoemaker and Burge do not exclude that the firstorder and the second-order state are also connected by a causal mechanism. What their arguments are meant to show is that this mechanism cannot play any essential role in the epistemology of self-knowledge. Indeed, Shoemaker’s and Burge’s arguments have convinced many authors that this conclusion is correct. In the wake of their work, various alternative proposals for how else the connection between first- and secondorder state could be understood have been suggested. Several authors, including Shoemaker himself, have argued that a second-order awareness or self-ascription forms a constitutive part of the first-order state: it is of the essence of the first-order state, on this account, to be accompanied by a self-ascribing higher-order state (Shoemaker 1994, 288). In the next section I will mention further views which are meant to offer alternatives to a perceptual account of self-knowledge of one’s thoughts. I believe that one central lesson that the contemporary debate on selfknowledge can draw from Sellars is that the opposition between perceptual and non-perceptual models of self-knowledge is not well motivated, and that Shoemaker’s and Burge’s influential arguments should be seen with more skepticism. For on the one hand, Sellars agrees with Burge and Shoemaker that self-knowledge is required by rationality. On the other hand, Sellars’s account of self-knowledge assigns a crucial epistemological

232

Franz Knappik

role to a causal connection between thoughts and meta-thoughts. It is this causal connection which, after we acquired it through appropriate conditioning, makes the preceding principle (3) true; and we saw that in order to have justification for our self-knowledge, we need to know this principle.14 The reason why Sellars can assume a necessary connection between rationality and self-knowledge, and at the same time, pace Shoemaker and Burge, assign a crucial epistemological role to a causal connection between thoughts and meta-thoughts is the following: in the case of a subject that has acquired the capacity for non-inferential self-knowledge, this causal connection is, on Sellars’s account, embedded in a network of rational connections which confer epistemic authority on meta-thoughts. As Sellars puts it: Now the important difference between a person who has merely been conditioned to respond to his thought that-p by saying “I have the thought that-p” and a person whose statement “I have the thought that-p” expresses direct self-knowledge is not that in the latter case the statement isn’t occurring as a conditioned response. It is. The difference is that in the latter case the conditioning is itself caught up in a conceptual framework. (to Castañeda, Apr. 3, 1961, §13) So on Sellars’s account, our entitlement for self-knowledge is both based on a rational connection between first- and second-order thought (the connection expressed in the background inference), and on a causal connection between both thoughts (since this causal connection underlies a premise of the background inference). This undermines the opposition between “perceptual” and “non-perceptual” accounts of selfknowledge. But what about Shoemaker’s and Burge’s arguments? With regard to Shoemaker’s self-blindness argument, Sellars could simply argue that in the case in which the causal connection in question collapses and selfblindness obtains, rationality collapses, too. (Cf. to Castañeda, Dec. 8, 1961, Ad II.3: “To lose the tendency to have appropriate meta-thoughts is to cease to be rational or lose one’s mind”.)15 By contrast, Sellars’s account can be defended against Burge’s argument quite easily by introducing a further, independently plausible assumption: it is an important feature of our ability for self-correcting reasoning that if our meta-thoughts make it seem the case that there is a rational shortcoming at the level of first-order reasoning, we directly take this as a reason for trying to revise the first-order reasoning accordingly. In other words, in the first-personal case, we normally do not make use of the possibility of a brute error as an excusing condition. This policy might simply be a further part of what we learn in acquiring our capacities for

Sellars on Self-Knowledge

233

self-critical reasoning. It is true that such a policy will lead to cases in which we erroneously attribute to ourselves failures of rationality where what has really gone wrong is the formation of meta-thoughts. But rationality requires anyway, on Sellars’s account, that the mechanism of selfascription works reliably. Therefore, such error cases may be rare enough to not impair the normal exercise of our rational abilities.

3 Self-Knowledge and Consciousness The biggest difference that sets apart Sellars’s account from most of the recent accounts of knowledge about occurrent mental states such as thoughts has to do with the relation between self-knowledge and consciousness. Most recent accounts agree that our ordinary, non-inferential self-knowledge about such occurrent mental states is restricted to states that are conscious—where the relevant form of consciousness is typically thought to be “phenomenal consciousness” (Block 1995), i.e., qualitative experience. (Some of these authors hold that conscious thoughts have a distinctively cognitive species of phenomenal character, which cannot be reduced to sensory phenomenal characters; other authors hold that thoughts are conscious in virtue of sensory phenomenal character,16 e.g., of related inner speech or some other imagery: cf. the contributions in Bayne and Montague 2011). Moreover, these authors assign the fact that those states are conscious an important role in the epistemology of selfknowledge. For example, some authors hold that the mere fact that we are in a conscious mental state gives us defeasible justification for believing that we are in that state (e.g., Smithies 2012). Other authors argue that there are background factors which make it the case that whenever we are in a conscious mental state, we have defeasible justification for believing that we are in that state. Such factors may consist, for example, in an a priori knowable principle according to which self-ascriptions of certain mental states that rationally respond to conscious experience are true (Peacocke 2003, 160); or in the fact that in consciousness-based selfascriptions of present thoughts, we exercise a general rational capacity (McHugh 2012); or, on one version of a “Transparency Theory” about self-knowledge, in an epistemic rule that allows us to move from our experience of inner speech about a certain topic to a self-ascription of a thought about that topic (Byrne 2011). Unlike such “consciousness-based” views, as we may call them, Sellars does not assign any role to phenomenal consciousness of our thoughts in his account of self-knowledge. Instead, he simply denies that thoughts themselves can have phenomenal character (PSIM 33).17 And while he grants that thoughts can be reflected in verbal imagery, which can have phenomenal character, he denies that such imagery (or, presumably, imagery of any other form) is necessary for self-knowledge about our thoughts (EPM §47: 178). We can now see that the distinctive structure of Sellars’s

234

Franz Knappik

account of self-knowledge about thoughts is a direct consequence of this view. For while consciousness-based views typically see the route to selfknowledge as an “ascent” from a conscious first-order state to a higherorder state that amounts to self-knowledge, this option is not available to Sellars. Instead, his account adopts, as we saw, a reverse order of explanation: he understands self-knowledge in terms of a “descent” from the self-ascription of a higher-order state (in premise (1a) of the trans-level inference) to the self-ascription of a first-order state (the conclusion (2) of that inference). In the remainder of this section, I will consider two possible arguments for a consciousness-based view and against the Sellarsian account, and examine how a Sellarsian can react to them. (i) The first argument is the following. When someone is challenged to defend his beliefs about his present occurrent mental states, it seems natural to reply (although not necessarily in these terms) that it is one’s present conscious experience which shows that one has such-and-such mental states. By contrast, a Sellarsian trans-level inference, starting from a premise like (1a), would seem a highly uncommon way to respond to such a challenge. So a consciousness-based account is closer to our actual justificatory practices—or so the argument goes. I think that this is a natural worry to have with Sellars’s account, but not a decisive one. For Sellars could reply that conscious experience can serve as (part of) the justificatory basis for self-knowledge only if some variant of the Myth of the Given is presupposed. More precisely, Sellars could argue as follows. Either conscious experience is itself a propositional, belief-like state. Then it can in principle serve as a reason for a self-ascription of a mental state, but it then stands also itself in need of justification. This is not something that any of the aforementioned views seems to assume. Or conscious experience is not itself a propositional state. But since Sellars holds that only propositional states can provide reasons for something else (cf. deVries and Triplett 2000, 76–7), it cannot be in this case the conscious experience itself which contributes to the justification of the higher-order belief. Rather, one would need to postulate an additional state that detects relevant features of one’s conscious experience, e.g., a report like (5) I am presently having a conscious experience as of a thought that p, or (6) My present conscious experience makes it seem to me that I am thinking that p. This report would have to serve as reason for the self-ascription of the thought that p, and it would have itself to be capable of being justified.

Sellars on Self-Knowledge

235

But in this case, it is not so clear anymore that the justification for the belief “I am thinking that p” proceeds by ascent from the conscious firstorder thought itself. Rather, this justification now seems to start from a propositional state that itself articulates a potential reason for the second-order belief “I am thinking that p”, and to this extent speaks about that second-order content. It is therefore possible for Sellars to argue that a report of a form like (5) or (6) is itself located at the same level of discourse (the third-order level) as Sellars’s own first premise in his trans-level inference, (1a) The thought that I am thinking that p has occurred to me.18 If this line of reasoning is correct, consciousness-based views either need to conceive of the justification for self-knowledge as “descending” from a third-order content to a second-order thought or belief, as does Sellars, or they need to embrace a version of the Myth of the Given on which beliefs can be justified by something else than propositional states. Regardless of whether one thinks that Sellars’s attack on the latter option is successful or not, his position on self-knowledge turns out to be less extravagant than the comparison with consciousness-based views can make it appear at first sight. (ii) The second argument goes as follows. It is a decisive advantage of consciousness-based views over Sellars’s theory, according to this argument, that they capture an intuitively plausible connection between self-knowledge and consciousness, to which Sellars fails to do justice. For consciousness-based views imply that ordinary non-inferential selfknowledge of thoughts is restricted to phenomenally conscious thoughts. By contrast, unconscious thoughts—e.g., steps in unconscious cognitive processes—are not available for such non-inferential self-knowledge. So consciousness-based views agree (although they make this not always explicit) on the following principle: (7) A subject S knows non-inferentially at t that she is thinking that p → S’s thought that p is phenomenally conscious at t (either in virtue of some irreducibly cognitive phenomenal character, or of some related sensory phenomenal character). I think that this principle is eminently plausible. There is only one clear case of thoughts that do not show up in conscious experience (in any of the ways mentioned in (7)):19 namely, thoughts that are part of the unconscious cognitive processes postulated by cognitive psychology.20 Our access to such thoughts is, of course, highly inferential in nature. In order to deny (7), one would need to assume that there are thoughts which are like the unconscious thoughts postulated by cognitive psychology in that they do not show up in the stream of consciousness, but which are at the

236

Franz Knappik

same time also very different in that we can have non-inferential knowledge about them. It seems very hard to identify any cases of thoughts that fit this description. If (7) is true, a theory of self-knowledge about one’s thoughts should be consistent with (7), and be able to explain why (7) is true. But since Sellars assigns no role to phenomenal consciousness in his account of self-knowledge, he fails to account for the truth of (7). And given his further claims that thoughts (a) lack phenomenal character, and (b) can be non-inferentially known independently of any accompanying imagery, he is even committed to denying (7). So if (7) is indeed plausible, this speaks against Sellars’s theory of self-knowledge. Of course, a Sellarsian may respond to this argument by simply denying the intuitions in favor of (7). But there is another and, I think, more interesting option for the Sellarsian: it is possible to supplement Sellars’s account of self-knowledge such that it can actually take into account (7). For Sellars, as we saw, self-knowledge builds on a mechanism through which first-order thoughts trigger corresponding meta-thoughts. Now this idea will have as a consequence a restriction like (7) if it is combined with the view that the occurrence of meta-thoughts which are caused by corresponding first-order thoughts is what makes those first-order thoughts conscious. This is exactly what so-called Higher-Order Thought (HOT) theories of consciousness hold. According to such theories, mental states are phenomenally conscious in virtue of being represented by higher-order thoughts that are caused by the first-order states, and that ascribe those first-order states to the subject.21 If HOT theory is combined with Sellars’s account of self-knowledge, the latter becomes apt to do justice to our preceding observations about the relation between self-knowledge and consciousness. But unlike consciousness-based theories, the resulting view does not see consciousness as a precondition for self-knowledge: rather, consciousness and self-knowledge are both consequences from one and the same factor, namely, the production of higher-order thoughts.22 To be sure, in order to amend her position in this way, the Sellarsian has to abandon Sellars’s claim that thoughts lack phenomenal character. But Sellars’s reason for this claim was that such phenomenal character would resist a physicalist explanation (PSIM 31–33). And since HOT theory is precisely meant to give a physicalist explanation of consciousness, this reason becomes obsolete on the proposed view.23

Notes 1. Cf., e.g., Shoemaker (1994); Burge (1996), and Moran (2001). For skepticism, see Carruthers (2011). For a detailed overview and assessment of the debate, see Gertler (2010). 2. Johannes Haag offers an excellent discussion of Sellars’s account of selfknowledge, from which I have greatly profited, in Haag (2001). I signal some points of disagreement with Haag in footnotes that follow.

Sellars on Self-Knowledge

237

3. The classical statement of this position is Ryle (1949), ch. 6. 4. Sellars’s idea that concepts like that of a thought are originally theoretical concepts has been enormously influential in the philosophy of mind and in psychology. Sellars is often cited as the first author to have proposed a socalled Theory Theory of mindreading (e.g. Lewis 1972, 257; Nichols and Stich 2003, 7). Mindreading is the ability to interpret, explain and predict the behavior of other creatures in terms of mental states, and the Theory Theory of mindreading holds that this ability should be understood as a typically implicit, folk-psychological theory. Cf. O’Shea (2012) for discussion of the relation between Sellars’s views and the Theory Theory of mindreading. 5. Sellars and Castañeda (2006). In references to the Sellars-Castañeda correspondence, I cite dates of letters and paragraph numbers. 6. For the following, cf. Haag (2001, 297–8, 312–13). 7. Haag (2001, 312) argues that in order for the inference to be a trans-level inference, the conclusion should read “There is good reason to believe that I am thinking that p”. But Sellars’s formulation of the inference as it stands matches the way he introduces the term “trans-level inference” at PHM 88.—Sellars oscillates in his characterizations of the resulting self-knowledge between present-directed formulations like (2), and formulations directed to the immediate past (e.g. to Castañeda, Apr. 3, 1961, §17: “I (just) had the thought that-p”). I will bracket this difference in my discussion, and always talk about present-directed self-knowledge. 8. To Castañeda, Mar. 11, 1962, Ad 3c. Sellars anticipates here Burge’s idea of “cogito-like thoughts” (Burge 1996, 92). 9. Elsewhere, Sellars works with the formulation “I am under the impression that I am thinking that p” (to Castañeda, Mar. 11, 1962, Ad 1).—Sellars distinguishes in this context also between mere thoughts and “mental assertions” (cf. Haag 2001, 308–11), where a thought with the content “It is raining” is a mental assertion of “It is raining” “unless it occurs in a larger context (e.g. ‘It is raining or snowing’) which suspends its assertive force” (to Castañeda, Mar. 3, 1962, Ad 3c cf. for example p. 190). Sellars implies that a meta-thought like “I am mentally asserting that p” is not self-reflectively incorrigible. However, when such meta-thoughts are read in the sense “I am hereby mentally asserting that  p”, they do seem to display self-referential incorrigibility. And conversely, meta-thoughts about mere thoughts can lack self-referential incorrigibility if they have the right kind of content—as in (1) and (1a). I therefore bracket the issue of mental assertion in the following. 10. Sellars tends to formulate this condition as self-ascription of a “well-functioning” rationality (e.g. to Castañeda, Mar. 11, 1962, Ad 3c), but cf. his reference to the “more extreme forms” of “madness” (ibid.) in the same context. 11. Notice that according to Sellars, humans are already rational in the preJonesean age, and therefore have already non-inferential self-knowledge of their utterances and their dispositions for utterances at that time: to Castañeda, Dec. 8, 1961, Ad II.3; cf. Haag (2001, 326). 12. At one point of the correspondence with Castañeda, Sellars even suggests that “[o]ne is conditioned to react by having MΘ [i.e., meta-thoughts] when (ceteris paribus) and only when one has Θ [i.e., the corresponding first-order thought]” (to Castañeda, Mar. 11, 1962, Ad 1, emphasis added). 13. For an overview, see Gertler (2010), chs. 6 and 8. (Gertler calls non-perceptual views “rationalist”, and perceptual views “empiricist”.) 14. It is true that principle (3) does not itself contain the term “causality”. But “conditioned” is a causal notion; and Shoemaker and Burge would clearly count a view on which self-knowledge rests on a link between thoughts and meta-thoughts that can be established by mere conditioning as a version of

238

15.

16.

17.

18. 19. 20. 21.

22.

23.

Franz Knappik the perceptual model of self-knowledge, which they aim to refute. (For example, Shoemaker (1994, 288) holds that it is “of the essence” of the first-order state to be accompanied by a higher-order state, and such an essential connection cannot be the result of conditioning.) But see Haag (2001, 324–8), for another view on Sellars vs. Shoemaker. According to Haag, Shoemaker’s argument is undermined by Sellars’s critique of the Myth of the Given, as it assumes that in perceptual knowledge (unlike in self-knowledge), the object of knowledge is independent of the subject’s knowledge of it (Haag 2001, 327). I doubt this point, since the sense in which first-order mental states depend, according to Shoemaker’s constitutivism, on the subject’s knowledge of them is far stronger than the sense in which external objects of our knowledge depend, according to Sellars’s epistemology, on our knowledge of them. Some would insist that in this case, only the imagery, not the thought itself, is conscious. But since nothing hinges on this for our present discussion, I shall use the notion of thoughts being conscious “in virtue of” related imagery in a sufficiently broad way to include the possibility that, strictly speaking, only the imagery is conscious—provided it contains enough information about the thought to support (phenomenally) immediate self-knowledge about it. Sellars’s discussion in this context is framed in terms of “qualitative” or “intrinsic character” (PSIM 33f.). What Sellars denies is that thoughts have a qualitative character that can be introspected (PSIM 33). I understand such introspectively accessible qualitative character to be identical with what is nowadays normally called the “phenomenal character” of a state, i.e., the “what it is like” of consciously experiencing that state.—When Sellars claims that thoughts lack, while “sensations” possess, phenomenal character (PSIM 32f.), he follows a view that dominated analytic philosophy of mind until recently, and that is itself a legacy of behaviorism: cf. Siewert (2011, 238–42). Recall that Sellars uses sometimes “I am under the impression that I am thinking that p”—a formulation that is quite similar to (5) and (6)—as equivalent for (1a). Cf. also the note before the last. Or more precisely, from the Sellarsian viewpoint: by theories in cognitive psychology for which we have not undergone Jonesean training. The most elaborate version of HOT theory is David Rosenthal’s, cf. Rosenthal (2005). Rosenthal himself suggests that his HOT theory can be combined with the framework of Sellars’s philosophy of mind (Rosenthal 2005, 304). But as far as I can see, he nowhere relates HOT theory to the account of self-knowledge that Sellars develops in the correspondence with Castañeda. In order for the premise (1a) in the trans-level inference to be known to the subject, the relevant second-order thought presumably needs to be itself made conscious by a corresponding third-order thought (cf. Rosenthal 2005, 28 on introspection). If that second-order thought is challenged, the third-order thought can be made conscious by a fourth-order thought, so that the corresponding instance of the trans-level inference becomes available, and so on. I am grateful to the participants at the Erlangen conference on Sellars for helpful discussion, and especially to Anke Breunig and Stefan Brandt for their thorough comments on earlier versions of this text.

References Bayne, Timothy and Michelle Montague. 2011. Cognitive Phenomenology. Oxford: Oxford University Press.

Sellars on Self-Knowledge

239

Block, Ned. 1995. “On a Confusion about a Function of Consciousness”. Behavioral and Brain Sciences 18: 227–47. Burge, Tyler. 1996. “Our Entitlement to Self-Knowledge”. Proceedings of the Aristotelian Society 96: 91–116. Byrne, Alex. 2011. “Knowing That I Am Thinking”. In Self-Knowledge, edited by Anthony Hatzimoysis, 105–24. Oxford: Oxford University Press. Carruthers, Peter. 2011. The Opacity of Mind: An Integrative Theory of SelfKnowledge. Oxford: Oxford University Press. deVries, Willem and Timm Triplett. 2000. Knowledge, Mind, and the Given: Reading Wilfrid Sellars’s ‘Empiricism and the Philosophy of Mind’. Indianapolis: Hackett. Gertler, Brie. 2010. Self-Knowledge. London and New York: Routledge. Haag, Johannes. 2001. Der Blick nach innen. Wahrnehmung und Introspektion. Paderborn: Mentis. Lewis, David. 1972. “Psychophysical and Theoretical Identifications”. Australasian Journal of Philosophy 50: 249–58. McHugh, Conor. 2012. “Reasons and Self-Knowledge”. In Self-Knowledge, edited by Annalisa Coliva, 139–63. Oxford: Oxford University Press. Moran, Richard. 2001. Authority and Estrangement: An Essay on Self-Knowledge. Princeton: Princeton University Press. Nichols, Shaun and Stephen Stich. 2003. Mindreading: An Integrated Account of Pretence, Self-Awareness, and Understanding Other Minds. Oxford: Clarendon Press. O’Shea, James. 2012. “The ‘Theory Theory’ of Mind and the Aims of Sellars’s Original Myth of Jones”. Phenomenology and the Cognitive Sciences 11: 175–204. Peacocke, Christopher. 2003. The Realm of Reason. Oxford: Clarendon Press. Peters, Uwe. 2014. “Self-Knowledge and Consciousness of Attitudes”. Journal of Consciousness Studies 21: 139–55. Rosenthal, David. 2005. Consciousness and Mind. Oxford and New York: Oxford University Press. Ryle, Gilbert. 1949. The Concept of Mind. London and New York: Hutchinson. Sellars, Wilfrid. (1956) 1991 [EPM]. “Empiricism and the Philosophy of Mind”. In Wilfrid Sellars. 1963. Science, Perception and Reality, 127–96. London: Routledge and Kegan and Paul. Reissued by Ridgeview Publishing Company in 1991. Hereafter SPR. ———. (1962) 1991 [PSIM]. “Philosophy and the Scientific Image of Man”. In SPR, 1–40. ———. (1963) 1991 [PHM]. “Phenomenalism”. In SPR, 60–105. ———. (1974) 2007 [MFC]. “Meaning as Functional Classification: A Perspective on the Relation of Syntax to Semantics”. In In the Space of Reasons: Selected Essays of Wilfrid Sellars, edited by Kevin Sharp and Robert B. Brandom, 81–100. Cambridge, MA and London: Harvard University Press. Sellars, Wilfrid and Hector-Neri Castañeda. 2006. Correspondence on Philosophy of Mind, edited by Andrew Chrucky, online at www.ditext.com/sellars/ corr.html, accessed February 25, 2018. Shoemaker, Sydney. 1994. “Self-Knowledge and ‘Inner-Sense’”. Philosophy and Phenomenological Research 54: 249–314. Siewert, Charles. 2011. “Phenomenal Thought”. In Bayne and Montague 2011, 236–67. Smithies, Declan. 2012. “A Simple Theory of Introspection”. In Introspection and Consciousness, edited by Declan Smithies and Daniel Stoljar, 260–95. Oxford: Oxford University Press.

Contributors

Boris Brandhoff works for the German Red Cross and is associated with the Philosophy Department at University of Bonn, Germany. Stefan Brandt is Assistant Professor at the Institute of Philosophy at Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany. Anke Breunig is Assistant Professor at the Institute of Philosophy at University of Halle, Germany. Willem A. deVries is Professor of Philosophy at University of New Hampshire, USA. Johannes Haag is Professor of Theoretical Philosophy at Universität Potsdam, Germany. Johannes Hübner is Professor of Theoretical Philosophy at University of Halle, Germany. Franz Knappik is Professor at the Department of Philosophy at Universitetet i Bergen, Norway. James R. O’Shea is Professor at the School of Philosophy at University College Dublin, Ireland. Peter Olen is Assistant Professor of Philosophy at Lake-Sumter State College, USA. Lionel Shapiro is Associate Professor of Philosophy at the University of Connecticut, USA. Tadeusz Szubka is Professor of Philosophy at University of Szczecin, Poland.

Index

acquaintance 117, 207; see also knowledge, by acquaintance action 79, 80, 83, 88, 154, 158, 160, 163, 211–12 Ajdukiewicz, Kazimierz ch. 3 passim analysis 16, 17, 19, 55, 67, 96, 98, 100–107, 209; Cambridge 13, 20, 25, 95; connective 105; of truth (see truth, analysis of) analytic philosophy 1, 2, 11, 20, 75, ch. 5 passim, 110, 116; Analytic Kantianism (see Kantianism, Analytic); analytic pragmatism (see pragmatism, analytic) Anscombe, Elizabeth 1 apprehension 117, 131, 140 assertibility 174, 180, 184, 191; semantic 175, 180–2, 185, 187, 191, 193–6, 198 atomism 33, 116; anti- 117 (see also holism) Austin, J. L. 14, 96 autonomy 37, 45–7 behaviorism 42, 115, 222, 238n17; Verbal 136, 139, 144 (see also thought, thinking-out-loud) Bergmann, Gustav 11, 71, 101 Boghossian, Paul 158, 162–9 Brandom, Robert 1, 37, 110, 118, 123, 124n2, 124n3, 125n6; see also Kant-Sellars thesis Broad, C. D. 95, 106 Burge, Taylor: on self-knowledge 6, 221, 229, 230–3 calculus 12, 35–8, 46, 48n12, 61 Carnap, Rudolf 2, ch. 2 passim, 55, 57, 61, 63, 65, 71, 97, 102, 116, 117, 120, 183, 186, 197; see also

Logical Syntax of Language (book); tolerance, principle of Cartesian 116 category 117, 208, 211, 212 Chisholm, Roderick 131, 133–4, 137, 139, 142 clarification 12, 20–3, 100, 102, 104 classification 112, 114, 122, 138–45 passim, 223; functional 77, 117, 198 co-ex predicate 62, 69 coherence 21, 22, 121, 134, 147; theory of meaning 111, 113, 116, 118 concept: conceptual change 24, 115, 122; conceptualization 60, 114, 218; proto-conceptual 119 conceptual framework see conceptual structure conceptual structure 20, 45, 103, 114, 117, 121, 134, 140, 147, 175, 183–96, 232; alien 184, 186; Peirceish 122, 189, 189, 194, 196; see also image, manifest; image, scientific consciousness 6, 155, 158, 161, 222, 233–6; Higher-Order Thought theories of 222, 236; phenomenal 6, 222, 233–6; “is-present-toconsciousness-along-with” (see co-ex predicate) conventionalism 34, 40, 45, 56, 64, 70, 78 Cook Wilson, John ch. 1 passim, 96; see also realism, Oxford; Statement and Inference Cornman, James 210, 212 Davidson, Donald 1, 2, 6n1, 174, 185, 186 demonstratives: complex demonstrative phrase 140, 210 (see also perceptual taking (PT))

242

Index

Descartes, René 96; see also Cartesian Dewey, John 111, 115 discourse: indirect 184; matter-offactual 110, 121, 123, 175, 182, 183, 187, 196 (see also picturing; truth, matter-of-factual); scientific 24 empiricism 1, 13, 45, 95, 113, 116, 117, 118, 207–8, 222; concept 3, 43; logical 1, 29 (see also positivism, logical); pragmatic 125n7; Scientific 55 epistemology 39, 65, 97, 100, 115, 137, 145, 214, 226, 231, 233 experience: immediate 62, 207–10 (see also Myth of the Given); nonconceptual 210; nonconceptual element in 138, 210; see also model, sense-image; perception; phenomenology; sensing; sensum explication 12, 15, 18–25, 120; explicative understanding of philosophy 24, 25; see also metaphilosophy Farber, Marvin 20, 25 Feigl, Herbert 1, 55, 71, 97 first-person: first-personal access 222–9 (see also Myth of Jones; self-knowledge); perspective 136, 215 (see also image, manifest; thirdperson) Firth, Roderick 131 formalism 22, 32, 33–4, 56, 61, 67, 69–1, 100–102, 106, 110 foundations: anti-foundationalist 216, 228; epistemic 1, 131, 207, 216, 222; foundationalism 33, 118, 222 (see also Myth of the Given) Frege, Gottlob 34, 55 functionalism 2, 22, 25, 118; analytic 179; metatheoretical 25 functional role 22–4, 33, 117, 118, 120 game 23, 77, 78, 79, 86, 182; of giving and asking for reasons 88, 139, 208; language 37, 46, 47, 76; meta- 79 Given: anti-givenist 214 (see also antifoundationalist); Myth of the (see Myth of the Given) Gödel, Kurt 194–6

habits 116, 155, 162; linguistic 21, 23, 167 Hall, Everett 99 Harman, Gilbert 182 Hegel, Georg W. F. 112, 116; see also idealism, German Higher-Order Thought theories of consciousness see consciousness, Higher-Order Thought theories of holism 33, 47, 106, 112, 116, 118 Hume, David 1, 96, 101, 113, 213, 219; see also representational system, Humean Husserl, Edmund 13, 16, 19, 20, 25; see also phenomenology, movement of ice cube: pink 209, 211, 217 (see also model, sense-image; phenomenology) idealism 65, 96, 111–16, 118; critical 106; German 116, 118; linguistic 56, 61, 63, 66, 69 (see also realism) image: imagery 233, 236, 238n16; imaging 218 (see also model, senseimage); manifest 12, 24, 75, 121, 211, 213–16; scientific 214–16 imagination 117, 217–19 inference 30, 41, 69, ch. 8 passim, 182; formal (rule of) 156, 184 (see also inference, logical (rule of)); logical rule of 33, 41; material (rule of) 33, 63, 114, 117, 157, 166, 183; physical rule of 33, 40, 43, 183 (see also inference, material (rule of)); trans-level 5, 130, 131, 133, 136, 141–7, 225–9, 234; see also move, intra-linguistic; thought, movement inferentialism 31, 110, 114, 117, 123, 139 intentionality 20, 33–4, 39–40, 44, 113, 116, 156, 161, 208 internalism 33 intuition 140–1, 145 intuitionism: ethical 12, 14, 97, 111, 112; logical (see logic, intuitionistic) James, William 114, 116, 118, 120 Kant, Immanuel 1, 11, 13, 14, 25, 96, 100, 106, 111, 114, 116, ch. 7 passim Kantianism ch. 7 passim; Analytic 130, 133; neo- 111; see also

Index Kantianism, Analytic; philosophy, transcendental Kant-Sellars thesis 118 knowledge: by acquaintance 113; how 103; immediate 116, 118, 119, 208; perceptual 115, 130–47, 225, 230; self- (see self-knowledge); unmediated (see knowledge, immediate) Kukla, Rebecca 119, 123 Lance, Marc 119 Langford, C. H. 95 language: artificial 35, 69; natural 35, 58, 62, 67–9; primacy of 39, 44, 47 (see also linguistic turn); symbolic 35, 39; in use 35 language entry/exit transition 156, 159, 180 Lehrer, Keith 98 Lewis, C. I. 14, 96, 113, 114, 116, 117, 118, 122 linguistic turn 20, 38, 39, 40, 99, 101, 118; see also new way of words logic: classical 185, 190, 192; intuitionistic 185, 190, 192, 193 logical space of reasons see space of reasons, logical Logical Syntax of Language (book) 29, 31–41, 45, 97 Lynch, Michael 121, 176, 179 McDowell, John ch. 7 passim; see also Kantianism, Analytic MacFarlane, John 187, 191 McHugh, Conor 163, 172n19 map: cognitive 60, 62, 65, 67, 70, 119, 122, 135, 177; see also picturing mathematics 176, 178, 182, 187, 194–6; classical see (logic, classical); intuitionistic (see logic, intuitionistic) matter-of-factual discourse see discourse, matter-of-factual meaning: coherence theory of meaning (see coherence, theory of meaning); empirical 32–4, 38–47; logical 32, 34–8, 40, 43, 45–7 passim; naïve realism about (see realism, naïve realism about meaning) metaphilosophy 12, 14, 16–25, 95, 99, 100–107; metaphilosophical (non-) factualism 21, 101

243

modality 2, 78, 97, 117, 123; causal 118; conceptual 154; logical 34–8, 43 model: sense-image 218 Moore, G. E. 20, 55, 95, 98, 104, 182 Morris, Charles 116 move: intra-linguistic 156, 180; see also language entry/exit transition Myth of the Given 3, 4, 29, 33, 40–7, 83, 110–18, 131, 221, 222, 234; see also atomism, anti-; foundations, anti-foundationalist; Given, antigivenist Myth of Jones 221, 222–5 naturalism 24, 29, 42, 105, 115, 118, 122, 135; Evolutionary 13, 97, 111, 114; see also Sellars, Roy Wood necessity 1, 78; see also modality new way of words 38, 116 normativity 2, 19, 30, 42, 46–7, 76, 102, 110, 113, 118, 120–3, 167, 198, 208, 214 object: physical 113, 208, 210, 213, 216 observation report see report, observation Olen, Peter 11, 22, 48n2, 49n30, 50n42, ch. 3, 101, 111, 113, 116, 124n2, 124n4 ontological commitment 211 ordinary language 15, 16, 19, 25, 69, 80, 82, 99; see also philosophy, Ordinary Language Oxford Realism see realism, Oxford Pap, Arthur 97 Paton, H. J. 219n1 pattern-governed behavior 80–3, 85, 161, 170n6 Peirce, C. S. 112, 115, 116–18, 120, 122; see also conceptual structure, Peirceish, pragmatism perception 14, 46, ch. 7 passim, 154, 156, 159, 180, ch. 10 passim, 222, 225; see also knowledge, perceptual; observation report; seeing, ostensible; self-knowledge, perceptual accounts of; perceptual taking (PT); report; warrant, perceptual perceptual taking (PT) 82, 137–41, 143, 147, 210; perceptual mistaking 214–16

244

Index

phenomenalism; anti- 97, 113, 116, 118 phenomenology 209, 211, 216, 217; movement of 19, 20 philosophical statement 21–3 philosophical truth see truth, philosophical philosophy: analytic (see analytic philosophy); critical 104, 106; explicative understanding of (see explication, explicative understanding of philosophy); Ordinary Language 2, 12, 14; speculative 106; transcendental 20, 71, 130, 133–5, 147 physicalism 43, 236; see also reductionism picturing 56, 69–71, 75, 110, 111, 118–23, 135, 177, 182, 189 pluralism: linguistic 56–61, 63–5, 70; truth 121, 174, 175–83, 185, 190–2 positivism: logical 1, 2, 11, 55, 71, 97, 100; see also empiricism, logical; Vienna Circle pragmatics 31, 102; pragmatic conception of the a priori 114, 117, 122; pragmatic implication 182, 192; pure 61, 69, 71, 101, 102 pragmatism 2, 11, 55, 98, ch. 6 passim; American 110, 115, 118 (see also pragmatism, classical); analytic 110, 116, 123 (see also pragmatism, neo-); classical 110, 111–15, 116–23 passim; conceptual 114; neo- 110, 113, 117, 119, 122; non-pragmatist 119, 122 prescriptive 29 see also normativity Price, H. H., 14, 42 Prichard, H. A. 12, 13, 14, 15, 96, 214 principle: epistemic 133–6, 138 a priori: pragmatic see pragmatics, pragmatic conception of the a priori processes: absolute 212–14; see also image, manifest; image, scientific proposition 120, 137, 146, 147, ch. 9 passim psychologism 34–8, 43, 47, 101 qualia 96, 113, 119 qualities: primary 214 (see also sensibles, common); secondary 214 (see also sensibles, proper) Quine, W. V. O. 1, 2, 29, 71, 97, 186, 197–8

rationalism 1, 112, 117, 118, 222; see also empiricism rationality 226, 229, 230–3; condition 162; rational animal 24 realism 55, 65, 96–8, 106, 111, 113–16, 177, 214; Critical 11, 13, 97, 106, 111–13, 115 (see also Sellars, Roy Wood); naïve 21, 113; naïve realism about meaning 56; Oxford 11, 14; physical 112, 115, 118; scientific 106 reductionism 42–5, 47; non- 19 reference: saving the 210 relativism: truth 61, 174, 185–92, 193–6 report 57, 71, 86, 234; observation 144, 216, 225; reporting role 224 representationalism 110, 120, 122; anti- 14, 122 representational system (RS) 70, 119, 177; animal 119; Aristotelian 119; Humean 119; associative (see representational system, Humean); syllogizing (see representational system, Aristotelian) Rorty, Richard 99, 110, 119, 123, 174 Ross, D. W. 12 Royce, Josiah 112 rule: of action 81, 86, 88–90, 153–60, 165; of criticism 81–2, 86, 88–90, 153, 160–2, 165, 167–8; formation 31, 63, 69, 100; of inference (see inference); of language 18, 22, 19, 34–6, 40–1, 59, 63, 69, ch. 4 passim, 102, 116, 121, 134, ch. 8 passim, ch. 9 passim; L-rule see (inference, logical rule of); P-rule (see inference, physical rule of); rule-following 75–6, 79–88, ch. 8 passim; rule-following vs. ruleobeying 155; semantic 30, 121, 180–2, 185, 193; transformation 29, 31, 33, 63, 69, 100 Russell, Bertrand 55, 96, 98, 99 Ryle, Gilbert 12, 14, 16, 19, 24, 80, 88, 104, 116 Ryleans 12, 223 scientific image see image, scientific seeing: ostensible 137–44, 146, 211, 214 self-blindness 231, 232 self-critical reasoning 230–3 self-knowledge 6, ch. 11 passim; nonperceptual accounts of 230, 231;

Index perceptual accounts of 6, 221, 230–3; recent accounts of 233–6 Sellars, Roy Wood 11, 13, 111–16, 118 Sellars-Castañeda correspondence 221, 224–9, 232, 238n21 Sellars’s earliest essays 12, 13, 20–3, 56, 61–71, 78, 96, 100–102, 106 semantics 31, 39, 102, 118, 121, 190; descriptive 30; pure 30, 36, 48n18, 101; see also assertibility, semantic; rule, semantic sense-data 45, 112, 119, 212; see also sensing, sensum sensibles: common 212, 214, 218; proper 212, 214, 217 sensing 218; objectless 212–16; see also processes, absolute; sensum sensum 212–16; see also sensing Shapiro, Stewart 190 Shoemaker, Sydney: on self-knowledge 6, 221, 229, 230–3 Soames, Scott 2, 98 space of reasons: logical 52n62, 110, 113, 118, 208 (see also game, of giving and asking for reasons) speculative philosophy see philosophy, speculative Statement and Inference 11–20, 25 Strawson, Peter 1, 11, 24, 99, 105, 130, 197 synesthesia 217 synoptic vision 96, 103–7; see also metaphilosophy syntax ch. 2 passim, 57, 97; descriptive 30, 35–7, 39; pure 30, 34–7, 39; see also Logical Syntax of Language Taking Condition 162; see also inference Tarski, Alfred 55, 120, 174, 176, 178; see also truth, equivalence schema for theoretical entities 209, 223 thermometer view 42 thinking see thought third-person: perspective 136, 215 (see also first-person); third-personal access 222

245

thought 6, 16, 19, 38–40, 83, 144, 147, 213, ch. 11 passim; movement 156, 157, 162–9; thinking in foro interno 144; thinking-out-loud 136–9, 144; thinking-in-thequestion-answering sense 139, 140 tolerance: principle of 37, 41, 57, 65 transcendental philosophy see philosophy, transcendental trans-level inference see inference, trans-level truth 59, 65, 102, 105, 111, 118–23, ch. 9 passim, 134, 143, 168, 216; absolute 186–92, 189, 194, 196; analysis of 175, 179–83, 191, 198 (see also assertibility, semantic); conceptual 41, 167; and correspondence 118–23, 174, 175–9; equivalence schema for 121, 123, 176–8, 180, 182, 186, 191, 193–6; ex vi terminorum 183 (see also truth, conceptual); ideal 122, 189, 189, 194, 196; immanent notion of 186, 189; matter-of-factual 21; necessary 78; Peircean conception of 120, 122; philosophical 103, 105; pluralism (see pluralism, truth); relativism (see relativism, truth); relativized notions of 185–9, 189; simple 187, 189; Tarskian schema (see truth, equivalence schema for); transcendent notions of 186, 189, 189; varieties of truth(s) 121, 176–83, 192 Valaris, Markos 171n17&18 Vienna Circle 1, 55, 98; see also Carnap, Rudolf; Feigl, Herbert; positivism, logical visual: dominance of the 216–18 warrant: perceptual ch. 7 passim Way, Jonathan 163, 172n19 Williams, Michael 110, 119, 120, 123, 178, 181, 193, 198 Wittgenstein, Ludwig 2, 55, ch. 4 passim, 99, 100, 116 world-perspective 55–61, 63, 69 world-story 21, 61–4, 67, 70, 122 Wright, Crispin 157, 159, 162, 175–9