Constructing Practical Reasons 0198754329, 9780198754329

Table of contents :
Contents
Acknowledgements
Introduction
1 What is constructivism?
2 Reasoning and reason judgments
3 The nature of reason judgments
4 The truth about reasons
5 Correct reasoning
6 Mind-dependence and objectivity
Concluding remarks
References
Index

Citation preview

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 3/7/2020, SPi

Constructing Practical Reasons Some things are reasons for us to perform certain actions. That it will spare you great pain in the future, for example, is a reason for you to go to the dentist now, and that you are already late for work is a reason for you not to read the next article in the morning paper. Why are such considerations reasons for or against certain actions? Constructivism offers an intriguing answer to this question. Its basic idea is often encapsulated in the slogan that reasons are not discovered but made by us. Andreas Müller elaborates this idea into a fully fledged account of practical reasons, makes its theoretical commitments explicit, and defends it against some well-known objections. Constructing Practical Reasons begins with an examination of the distinctive role that reason judgements play in the process of practical reasoning. This provides the resources for an anti-representationalist conception of the nature of those judgements, according to which they are true, if they are true, not because they accurately represent certain normative facts, but because of their role in sound reasoning. On the resulting view, a consideration owes its status as a reason to the truth of the corresponding reason judgement and thus, ultimately, to the soundness of a certain episode of reasoning. Consequently, our practical reasons exhibit a kind of mind-dependence, but this does not force us to deny their objectivity.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 3/7/2020, SPi

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 3/7/2020, SPi

Constructing Practical Reasons ANDREAS MÜLLER

1

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 3/7/2020, SPi

3

Great Clarendon Street, Oxford, OX2 6DP, United Kingdom Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide. Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries © Andreas Müller 2020 The moral rights of the author have been asserted First Edition published in 2020 Impression: 1 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by licence or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this work in any other form and you must impose this same condition on any acquirer Published in the United States of America by Oxford University Press 198 Madison Avenue, New York, NY 10016, United States of America British Library Cataloguing in Publication Data Data available Library of Congress Control Number: 2020935466 ISBN 978–0–19–875432–9 Printed and bound in Great Britain by Clays Ltd, Elcograf S.p.A. Links to third party websites are provided by Oxford in good faith and for information only. Oxford disclaims any responsibility for the materials contained in any third party website referenced in this work.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 3/7/2020, SPi

Contents Acknowledgements

vii

Introduction

1

1. What is constructivism? 1.1 1.2 1.3 1.4 1.5 1.6 1.7

The basic idea Constructivism and representation Constructivism and truth Constructivism and mind-dependence Constructivism, explanation, and the grounding relation Putting constructivism on the map Plan of the book

2. Reasoning and reason judgements 2.1 2.2 2.3 2.4 2.5 2.6 2.7

Preliminaries Conditions of adequacy The normative guidance account The inadequacy of alternative accounts Reasoning and conceptual sophistication Regress worries Metareasoning

3. The nature of reason judgements 3.1 3.2 3.3 3.4

Preliminaries Guidance and representation Constructivism and expressivism Extending the account

4. The truth about reasons 4.1 4.2 4.3 4.4 4.5 4.6

Constructivism, correspondence, and deflationism Towards a compatible theory of truth Wright on truth Alethic pluralism Truth and sound reasoning Constructing reasons

6 6 7 13 17 20 26 31

33 33 41 44 52 65 72 77

82 82 89 100 107

110 111 114 116 120 128 137

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 3/7/2020, SPi

vi



5. Correct reasoning 5.1 5.2 5.3 5.4 5.5 5.6 5.7

Correct and incorrect reasoning Correctness and reasons Correctness without reasons The constitutive rules account of correct reasoning Developing the account How to determine the rules of reasoning The status of the rules of reasoning

6. Mind-dependence and objectivity 6.1 6.2 6.3 6.4

The varieties of mind-dependence Siding with Euthyphro Fallibility and universality Modal robustness

Concluding remarks References Index

141 142 145 156 159 163 172 185

189 190 197 200 206

219 223 233

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 3/7/2020, SPi

Acknowledgements This book grew out of my doctoral dissertation, which I submitted to Humboldt-Universität zu Berlin in August 2013. A lot of time has passed since then, and while the position I develop in this book has not changed, some of the arguments and a lot of the text that presents them have. Writing this book and the dissertation on which it is based would not have been possible without the support of a number of persons and institutions. First and foremost, I am greatly indebted to my PhD supervisor at Humboldt, Thomas Schmidt, who has supported and encouraged me not only throughout the time I worked on the dissertation, but also a long time before I had even thought about pursuing a doctorate in philosophy. His insightful comments and suggestions were immensely helpful in shaping my thinking about the topic of this book. During a research visit spent at the Philosophy Department of Princeton University in the spring term of 2012, Michael Smith was very generous with his time and his comments, for which I am very grateful. Our conversations, in Princeton and on various later occasions, helped me to develop a clearer perspective on many of the issues that I discuss in what follows. Many people helped me to improve the material that found its way into this book. I have greatly benefited from conversations with and comments from Maike Albertzart, Hannah Altehenger, Max Barkhausen, Philipp Brüllmann, Anne Burkard, Andreas Cassee, Christopher Cowie, Sinan Dogramaci, Gerhard Ernst, Daan Evers, Christoph Fehige, Simon Gaus, Jan Gertken, Stefan Gosepath, Logi Gunnarsson, Tim Henning, Ulf Hlobil, Rebekka Hufendiek, Benjamin Kiesewetter, Christian Kietzmann, Felix Koch, Errol Lord, David Löwenstein, Barry Maguire, Susanne Mantel, Leo Menges, Cory Nichols, Herlinde Pauer-Studer, Peter Schaber, Oliver Schott, Moritz Schröder, Sharon Street, Christine Tiefensee, Barbara Vetter, Jay Wallace, Ralph Wedgwood, and Jack Woods, as well as two anonymous

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 3/7/2020, SPi

viii



readers for OUP. I would like to express my gratitude to all of these people, and to those I have forgotten to mention. I am also much obliged to Timo Junger, who was a great help in preparing the final manuscript. For valuable discussions and comments, I would like to thank various audiences in Berlin, Essen, Frankfurt, Munich, Münster, Konstanz, Leipzig, Saarbrücken, Tübingen, and Vienna, where I had the opportunity to present earlier versions of some of the material included here. I am also grateful to Dominik Perler and his Leibnitz-Preis-Projekt, to the Graduate School of Princeton University, to Humboldt-Universität zu Berlin and the Sonderforschungsbereich 644, to the Studienstiftung des deutschen Volkes, and to the Centre for Advanced Study in Bioethics in Münster for supporting me and my work, as well as to the University of Bern and my colleagues here, for providing me with the opportunity to continue doing philosophy in such an ideal environment. Finally, I owe a special debt of gratitude to my parents, whose unconditional support has helped me on every step of the way, and to Moritz Schröder, who has listened to my thoughts about this book and its contents with unfailing patience and encouragement. Chapter 2 contains material previously published in A. Müller (2019). Reasoning and normative beliefs: not too sophisticated. Philosophical Explorations 22(1), 2–15, which is reprinted here by permission of the publisher (Taylor & Francis Ltd).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 1/7/2020, SPi

Introduction Agents have reasons: some things are reasons for them to perform or refrain from certain actions. That it will spare her great pain in the future, for example, is a reason for an agent to go to the dentist now, that she promised her friend to help him move is a reason for her to do so, and that she is already late for work is a reason for her not to read the next article in the morning paper. What makes either of these considerations a reason for or against the respective action? Why do we have such reasons? In some cases, an answer is easy to come by. You might have a reason to buy a train ticket simply because you have a reason to get from Berlin to Hamburg. Here, your reason to buy the ticket can be explained in terms of another reason from which it derives. But not all reasons can be derivative reasons, so this kind of answer will not always be available. Can we give a more general explanation why something is a reason for an agent (not) to perform a certain action, one that applies to all such reasons? Some philosophers, including Thomas Scanlon (1998, ch. 1; 2014) and Derek Parfit (2011a; 2011b), are sceptical about the feasibility of this explanatory project. They deny that such a general account of our practical reasons—that is, our reasons for or against performing certain actions—can be provided. We might be able to explain, in each case, why the reason-giving fact holds, such as the fact that going to the dentist now will spare someone great pain in the future. But the reason fact itself, the fact that this fact about the consequences of going to the dentist is a reason for the agent to go, cannot always be explained in an informative way: at least some reason facts are primitive, fundamental facts, facts at which our explanations come to an end. Others, like Michael Smith (1994; 2013) and Mark Schroeder (2007), are more optimistic about the prospects of giving a general account of practical reasons. They emphasize that what there is reason to do for an agent is closely related Constructing Practical Reasons. Andreas Müller, Oxford University Press (2020). © Andreas Müller. DOI: 10.1093/oso/9780198754329.001.0001

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 1/7/2020, SPi

2



to what that agent is motivated to do, or what she would be motivated to do if certain conditions were met. This relation, they claim, suggests that facts about what is a reason for doing what can be explained in terms of facts about the relevant agent’s actual or hypothetical motivational states, such as her desires. This book explores another attempt to provide a general explanation of practical reasons, one that is offered by a constructivist account of those reasons. Constructivism shares the optimism of the second kind of view mentioned; it, too, purports to give a general and informative answer to the question why certain reason facts obtain. But it suggests a different approach to developing such an account: to understand why some things are reasons for certain actions, we have to start by thinking about the relation between those reason facts and an agent who acknowledges them by forming the corresponding reason judgements. The basic idea of constructivism is that reasons are not discovered, but made. That is, the relation between a reason fact and the agent’s corresponding judgement is not one of discovery; instead, those facts must be considered the result of some activity in which agents are engaged. To explain why something is a reason for an agent to perform a certain action, we have to explain how they are made, what activity they result from, and what role reason judgements play in that activity. As a view that is closely associated with the work of John Rawls and Christine Korsgaard, constructivism has attracted considerable attention over the last few decades. Much of that attention, however, has been critical. The idea that constructivism about practical reasons might offer an interesting and plausible alternative to the more established views in the meta-ethical debate is often met with scepticism (see, e.g., ShaferLandau 2003, ch. 2; Hussain and Shah 2006; Enoch 2009; Wallace 2004; 2012; and Ridge 2012). On the other hand, there is little agreement as to what exactly a constructivist account of practical reasons consists in and what its central claims are, over and above the basic idea sketched in the previous paragraph. Hence, before we can properly assess its merits, we need a better understanding of what constructivism actually amounts to as an account of practical reasons. My main aim in this book is thus to elaborate the basic constructivist idea into a fully fledged account of practical reasons, to make its

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 1/7/2020, SPi



3

theoretical commitments explicit, and to defend it against some wellknown objections. Doing so will require me to address a variety of different topics.¹ Given the limited space available, it would be impossible to discuss all of these topics as extensively as they deserve, and to address the many positions that one could take with respect to each of them. I have thus chosen to focus on those aspects that are most relevant for elaborating the details of the constructivist position and to address only those problems and opposing views that pose an immediate threat to the overall coherence of that position. As a consequence, my overall argument is primarily constructive: what I hope to establish is not that constructivism must be adopted or that it offers a better account of practical reasons than its competitors, but that it can be adopted and that it offers a coherent and prima facie attractive account that avoids many of the problems and concerns with which it is confronted and hence deserves further discussion. Above all, this book is meant to facilitate such a discussion. Before we begin, there are a few clarificatory remarks that I would like to make in advance so as to avoid potential misunderstandings. Firstly, constructivism about practical reasons is a meta-ethical view that must be distinguished from views in moral or political philosophy that also go by the name ‘constructivism’, such as John Rawls’s (1971) position on the fundamental principles of justice or Thomas Scanlon’s (1998, part II) position on moral wrongness. The difference between constructivism as a meta-ethical position and constructivism as a first-order normative position lies in the questions each position is supposed to answer. Firstorder normative views answer questions like these: when is the distribution of a scarce amount of goods just? Is it sometimes permissible to break a promise? Is happiness an intrinsic value? Why is it wrong to lie to someone? Is there a reason to develop one’s talents? Meta-ethical positions, on the other hand, are concerned with the metaphysical status of normative facts, such as facts about what is a reason for what, and they might offer an explanation for why the things that a first-order normative theory identifies as reasons for a certain kind ¹ I present the main tenets of a constructivist account of practical reasons and give an overview of how this book proceeds with elaborating that account in Chapter 1 below.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 1/7/2020, SPi

4



of action are reasons for that action. They are also concerned with the nature of normative judgements, with whether and in what sense they can be said to be true or false, with the meaning of terms like ‘just’, ‘wrong’, or ‘reason’, etc. So what distinguishes a form of meta-ethical constructivism from, say, Scanlon’s moral constructivism is not that the latter is ‘merely’ about moral wrongness, whereas the former is concerned with practical reasons in general, for a first-order normative position might share this wider scope. The difference lies in what questions they address about wrongness or reasons and our judgements about either. That is not to say, however, that meta-ethical and first-order normative positions are entirely independent of each other, or that one can remain entirely neutral on any substantive normative issues while defending a fully specified meta-ethical view. Giving certain answers to some of the questions with which meta-ethical views are concerned will commit you to accepting or denying certain answers to first-order normative questions. A meta-ethical view that explains an agent’s reasons entirely in terms of her actual desires, for example, will face significant difficulties in accommodating the substantive normative claim that some reasons are shared by all agents regardless of any differences in their motivational dispositions. Hence, that meta-ethical and first-order normative positions must be distinguished because they answer different kinds of questions does not mean that one can be defended without an eye to the other. Secondly, the constructivist position that I am concerned with here is an account of practical reasons. More precisely, it is an account of facts about what is a reason for whom to do what (and not of the ordinary first-order facts that give the agent a reason). It thus has a wider scope than an account of, for example, moral facts, but it does not purport to cover all normative phenomena. How best to understand the term ‘normative’ here is itself a controversial matter. Some use it only for facts and judgements that are about—or closely connected to facts and judgements about—reasons, while others use it more broadly so as to include, for example, facts and judgements about correctness, rationality, virtue, etc., even if those are not to be understood in terms of reasons (cf. Thomson 2008). But even if ‘normative’ is used narrowly to apply

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 1/7/2020, SPi



5

only to the realm of reasons, the constructivist position under consideration targets only a subset of all normative facts, because it only offers an account of practical reasons, putting other kinds of reasons, such as epistemic reasons—reasons for belief—to one side. There is no general reason to suspect that the constructivist account of practical reasons that will emerge in the following chapters could not be extended, with some adaptions, to the case of epistemic reasons. Still, I think that it is not a trivial question whether epistemic and practical reasons should receive the same theoretical treatment—at least the mere fact that we talk about ‘reasons’ in both cases does not suffice to show that we are dealing with a homogeneous phenomenon that requires a unified account. Hence, the plausibility of a constructivist account of epistemic reasons must be assessed separately and on another occasion.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

1 What is constructivism? 1.1 The basic idea Constructivism, as I understand it here, is first and foremost a view about the relation between thought and its object. To be a constructivist about a certain domain is to take the facts of that domain to be an upshot, rather than the measure, of correct judgement in that domain. A constructivist about practical reasons thus denies that, when an agent comes to realize that something is a reason for a certain action, she does so in response to the corresponding normative fact, that is, a fact about what is a reason for doing what. Instead, such facts are somehow explicable in terms of our reason judgements. That does not mean that the facts are whatever we think they are. Constructivists do not deny that we can be mistaken about what is a reason and what is not. But they do suggest that normative facts are ultimately grounded in our mental states or activities. In particular, according to the constructivist account of practical reasons that I will present in this book, facts about what is a reason for doing what obtain in virtue of the activity of practical reasoning and the role that the corresponding reason judgements play in it. More specifically, such a fact obtains because the corresponding reason judgement is true, and the judgement is true because the episode of reasoning that it is apt to guide is sound. In this first chapter, I want to start by presenting the main aspects of the constructivist account I have just sketched in a bit more detail. I do this, firstly, in order to give you an overview and to motivate the remaining chapters of this book, where these aspects and the questions to which they give rise will be discussed at much greater length. But I also do this because the term ‘constructivism’ has been used to pick out a variety of rather different positions in meta-ethics, both by advocates and by opponents of the views so labelled. Hence, I want to be as clear as possible about how and why I use that term, to avoid adding to the Constructing Practical Reasons. Andreas Müller, Oxford University Press (2020). © Andreas Müller. DOI: 10.1093/oso/9780198754329.001.0001

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

7

confusion. Let me be clear, though, that I do not claim that my proposed understanding of constructivism somehow synthesizes or subsumes all of the various conceptions of constructivism that can be found in the literature.¹ The view that I put forward adopts many important insights from those other conceptions (as I will point out along the way). In fact, little of what I have to say in what follows has never been suggested before. But my aim here is not to identify a common denominator; I am sceptical that one could be found. The basic idea of constructivism is essentially a metaphor: some things are constructed from certain other things; they are made, rather than discovered, as the well-known slogan has it. Such metaphors can be spelled out in different ways, none of which is, as such, the right one. Nevertheless, some ways might be more useful or more interesting than others. The position that I present in this book is meant to be true to the idea encapsulated in that metaphor, and it takes up some important thoughts of paradigmatic constructivists like John Rawls and Christine Korsgaard, as we will see in the following sections.² But most importantly, it is meant to offer a distinct and prima facie attractive account of practical reasons.

1.2 Constructivism and representation Let me now elaborate on the main aspects of the constructivist position that I sketched in section 1.1. I will start with its account of the nature of reason judgements in this section, and then address its understanding of truth as well as the mind-dependent status it assigns to reason facts in sections 1.3 and 1.4. Constructivism rejects a familiar conception of the kind of attitude that is involved in thinking about reasons. Consider the following

¹ Advocates of constructivist views in meta-ethics include John Rawls (1980; 1993), Christine Korsgaard (1996; 2003; 2009a; 2009b), Carla Bagnoli (2002; 2012; 2013), Aaron James (2007; 2012), and Sharon Street (2008; 2010; 2012). For a comprehensive overview, see Bagnoli (2017). ² Constructivism is often considered to have a Kantian pedigree, and both Rawls and Korsgaard’s views are certainly inspired by Kant. However, whether Kant is best interpreted as a constructivist is controversial among Kant scholars, and I will not address this issue here. For a helpful discussion of meta-ethical constructivism that pays special attention to its roots in Kant’s philosophy, see Schafer (2015a; 2015b).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

8

  ?

example. Lara and her friends are out for dinner. They are about to order from the menu, which, this being a very small restaurant, offers only two options for the main course: a stuffed squash and duck breast marinated in hoisin sauce. Lara is unsure which to choose, so she confers with her friends, who have been here before. They tell her that the squash’s stuffing is vegetarian, and that the duck breast comes with rice. Lara also learns that hoisin sauce contains soya beans, and this she takes to be a reason against having the duck breast (she is allergic to soya beans). This reason judgement settles her choice, and she proceeds to order the squash. What role did Lara’s judgement that the fact that hoisin sauce contains soya beans is a reason against ordering the duck breast play in the process that lead to her decision? What kind of attitude do we describe an agent as having when we attribute such a reason judgement to her? The familiar conception answers those questions like this: to believe that the duck breast comes with rice and that hoisin sauce contains soya beans is to represent certain facts as obtaining, and when those facts really do obtain, as we shall assume, then those representations are accurate. Hence, by forming those beliefs, Lara acquires new information about the options in a choice problem she currently faces, information that she can apply in making a decision. The familiar conception extends this account of the cognitive role of ordinary beliefs to the case of reason judgements. Her reason judgement,³ if true, also provides Lara with information about the options in her choice problem, information that will help her to come to a decision. Of course, the beliefs and the reason judgements differ in the kind of information they provide. The former provide non-normative information, that is, they accurately represent certain non-normative facts about the options in question, such as the fact that the duck breast comes with rice, or that hoisin sauce contains soya beans. To form a reason judgement, on the other hand, is to acquire

³ I refer to these attitudes as ‘reason judgements’ rather than, e.g., ‘beliefs about reasons’, because the latter is often associated with a representationalist account of the function of the attitudes in question. As will become clear shortly, constructivism rejects such an account of the attitudes’ cognitive function. So, in order to avoid prejudging the issue, I am using what I take to be a less committing term. ‘Judgement’ is thus meant to pick out a certain kind of mental state, not the event of forming such a state.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

9

normative information about the choice situation and to represent some normative fact, some fact about what is a reason for what, as obtaining. But underlying such differences in content—in what kinds of facts they are about—is a fundamental similarity in the role those two attitudes play in the cognitive economy of an agent who faces a choice. Just like ordinary beliefs about non-normative matters, reason judgements are representational states which, if accurate, provide the subject with information that she can apply in making a decision. The details of this conception can be filled in in numerous different ways, particularly when it comes to characterizing the facts that reason judgements purport to represent. Are they sui generis normative facts, or can they somehow be reduced to other, non-normative facts? Are they causally efficacious? How can we gain knowledge of them? These are just some of the questions on which different version of this conception can disagree, and they are certainly important questions. Nevertheless, we can put them to one side for the moment, because the fundamental aspects of the conception, the ones that will be most relevant in what follows, are not affected by how these questions are answered. These aspects concern the relation between those reason facts on the one hand and our reason judgements on the other, as well as the role reason judgements play in the cognitive processes of agents like Lara. According to the familiar conception, judgements about what is a reason for what relate to certain facts in the same way that the belief that hoisin sauce contains soya beans relates to the fact that hoisin sauce contains soya beans: they purport to represent those facts. Hence, they provide an agent with information that is relevant to her choice, if they represent accurately. This conception of the nature of reason judgements is quite popular in contemporary meta-ethics. The idea that reason judgements purport to represent certain facts and, if accurate, provide agents with information about the normative aspects of her situation seems to underlie, in one of its guises, the positions developed by David Enoch (2007; 2011a), Derek Parfit (2011b), Thomas Scanlon (2003; 2014), Mark Schroeder (2007), Michael Smith (1994), and many others. It is often not so much articulated explicitly as hidden in the assumption that reason judgements and ordinary beliefs about non-normative matters are the same kind of attitude towards different contents. If you take the main difference

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

10   ? between the two to consist in the facts they represent, it is understandable why you will focus on giving an account of the facts our reason judgements are about, while saying little about the kind of attitude you take those judgements to be—there is no need to, if it is the same as our ordinary beliefs. Constructivism offers an alternative to this familiar conception. It suggests that the cognitive function of reason judgements differs fundamentally from the cognitive function of ordinary beliefs about nonnormative matters. According to the constructivist, reason judgements play a distinct role in the psychology of a deliberating agent: the point of those judgements is not to provide an agent with additional information that she can apply in making her decision. Instead, reason judgements guide her in making a decision on the basis of the information she has. To use a computational analogy: while ordinary beliefs can provide the input for decision-making processes, reason judgements provide the instructions that such processes follow in translating those inputs into outputs, that is, into decisions. Thus, when Lara forms the belief that the hoisin sauce contains soya beans, she acquires new information about one of her options, but when she forms the judgement that this is a reason not to order the duck breast, she does not acquire more information. Instead, this judgement will allow her to apply the information she already has to the choice she faces. Information is not enough to come to a decision; we also need to have an idea of how to bring any such information to bear on the problem of choosing one of the available options. This is the distinct contribution reason judgements make to the process of reasoned decision-making. To be more specific about the main point of disagreement between constructivism and the familiar conception of the nature of reason judgements outlined above, let me introduce the following thesis about what kind of attitude reason judgements are: Representationalism Reason judgements are representational states. Their cognitive function is to represent certain facts, and if they do so accurately, they provide the agent with normative information by indicating that a certain fact concerning what is a reason for what obtains.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

11

Proponents of the familiar conception accept Representationalism, or one particular version of it, while constructivists reject it altogether and take reason judgements to play a fundamentally different cognitive role from, for example, beliefs about the composition of some condiment. Constructivism is thus a form of anti-representationalism about the nature of reason judgements.⁴ Constructivists are not the only ones who reject Representationalism. They share this negative part of their view with expressivists such as Simon Blackburn (1998) and Allan Gibbard (1990; 2003). The two views differ, however, in their positive accounts of the kind of mental states they take normative judgements to be. Expressivists typically conceive of them as motivational or desire-like states, that is, states that are characterized in terms of their role in the explanation of an agent’s actions. Constructivists, on the other hand, think that reason judgements are a distinct kind of mental state, the nature of which can only be understood by attending to its role in guiding the agent’s practical reasoning. In other words, whereas expressivism characterizes reason judgements as a kind of pro-attitude, constructivism identifies them in terms of their role in a process that leads to the formation of such attitudes. The two views thus disagree substantially on the cognitive role of the attitudes in question.⁵ Representationalism is also rejected by global expressivists such as Robert Brandom (1994) and Huw Price (2011; 2013). Moreover, Brandom’s claim that the function of normative vocabulary is to make explicit ‘an attitude that otherwise could be implicit only in what is done—namely, the endorsement of a pattern of practical reasoning’ (1994, 271) seems to attribute to reason judgements a cognitive role similar to the one described in Chapter 2 below. The constructivist theory presented in this book indeed bears important similarities to the views of Brandom and Price. But there are also significant differences. For one, constructivism about practical reasons is a local theory that ⁴ Carla Bagnoli (2002, 125) agrees that constructivists do not consider moral judgement to involve the ‘aspiration to correctly represent a moral reality’. ⁵ Sharon Street (2008; 2010; 2012) also argues that the attitudes that provide the basis for her constructivist account of practical reasons must be distinguished from ordinary desires. But she nevertheless characterizes them as attitudes that are ‘by their nature motivating’ (2008, 230), which seems to make her account of normative judgements more difficult to distinguish from the expressivist’s than the reasoning-based account that I propose in this book.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

12   ? contrasts the cognitive function of reason judgements with that of, e.g., beliefs about one’s environment—it does not reject Representationalism across the board. Constructivists also do not share Brandom’s and Price’s commitment to deflationism (cf. Chapter 4). John Rawls, probably the most influential proponent of constructivism in the twentieth century and the author responsible for introducing that label into the meta-ethical debate,⁶ agrees that constructivists reject the idea that normative judgements resemble ordinary beliefs in playing a representational role in an agent’s psychology. He contrasts constructivism with a view he calls ‘rational intuitionism’.⁷ This view holds that ‘moral first principles and judgements, when correct, are true statements about an independent order of moral values’ and they are ‘known by theoretical reason’, which means that their epistemology is structurally similar to that of empirical judgements: ‘moral knowledge is gained in part by a kind of perception and intuition’ (1993, 91–92). The rational intuitionist’s conception of the relation between moral judgements and moral facts is thus very similar to Representationalism. According to Rawls, constructivism rejects this conception, along with the ‘sparse conception of the person as . . . a knower’ (1993, 92) that goes with it. Rawls’s characterization of constructivism thus puts the view in fundamental opposition to Representationalism. Christine Korsgaard is the second major proponent of constructivism in the last decades. Especially in her more recent writings, she puts the contrast between constructivism and realism, which she considers to be its main competitor, in terms of the function or purpose of moral and other normative concepts: [T]he difference between a realist and a constructivist theory rests in the way the two views understand the function of concepts . . . . A realist

⁶ It is important to note that Rawls’s views on constructivism have changed over time, or at least shifted in focus. Whereas his Dewey lectures (1980) are simply concerned with ‘Kantian constructvisim’, he later distinguishes between ‘political constructivism’ and ‘Kant’s moral constructivism’ (1993, 89–129). Only the latter of these two is what he calls a ‘comprehensive moral doctrine’, that is, a view that also addresses metaphysical and epistemological questions and thus qualifies as a meta-ethical view. Hence, I will focus on Rawls’s characterization of constructivism as such a comprehensive doctrine. ⁷ Cf. Rawls (1980, 557–64; 1993, 91–101). He identifies Samuel Clarke, Richard Price, Henry Sidgwick, G. E. Moore, and David Ross as paradigmatic advocates of rational intuitionism.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

13

believes that the function of concepts is to describe the world, to mark out the entities we find there, while a constructivist believes that the function of (at least some) concepts is to mark out, in a schematic way, the solution to some problem that we face. (Korsgaard 2008, 22)

Constructivism thus rejects the idea that the cognitive function of normative judgements is to represent reality. That is not what normative concepts are for; their ‘cognitive job’ is not a descriptive one (2003, 105). Korsgaard thus clearly rejects a representationalist account of normative thought. Like Rawls, she understands constructivism as a view that is meant to offer an alternative to Representationalism.⁸

1.3 Constructivism and truth Constructivism offers an explanation why some things are reasons for certain actions. It explains such reason facts, in the first instance, in terms of the truth of the corresponding reason judgement: they obtain because that judgement is true.⁹ Constructivists thus do not deny that there are facts about what is a reason for doing what. Nor do they deny that those facts correspond to true reason judgements, as long as that claim is understood as a simple biconditional: Correspondence The judgement that p is true if and only if it is a fact that p.

⁸ Her alternative proposal for the function of normative concepts is admittedly less clear. On her preferred way of formulating that alternative, those concepts denote, in a ‘formal’ or ‘schematic’ way, the solution to some problem we face in so far as we are agents (1996, 113). While this does not strike me as being in conflict with the view that I develop in this book, I am not confident enough in my understanding of her proposal to make the case for any substantive agreement. ⁹ Note that what explains a reason fact according to this proposal is the truth of the corresponding reason judgement, not the true judgement. That is, no actual judgement— understood as some specific psychological reality—is necessary for a reason fact to obtain. What is required, rather, is that the conditions that ground the truth of that judgement are actually met.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

14   ? This claim is symmetrical and thus leaves open the question of explanatory priority between its left- and right-hand sides. It is on this latter issue that constructivists disagree with their opponents. They maintain that the order of determination in Correspondence goes from left to right: they treat reason facts as the explanandum rather than the explanans.¹⁰ What motivates this reversal is the constructivist’s rejection of Representationalism as an account of the cognitive function of reason judgements. Truth, it seems, is closely connected to successful functioning. So, if the cognitive function of reason judgements is not to represent certain facts, then their truth will not be grounded in the accurate representation of such facts either. Denying that reason judgements are true, if they are true, because they accurately represent certain reason facts, allows constructivists to employ Correspondence as part of an explanation of its right-hand side, rather than its left-hand side. For them, a reason fact is not the ground of a reason judgement’s truth, but rather its ontological shadow.¹¹ For their explanatory project to succeed, constructivists must offer an alternative account of what it is for reason judgements to be true. Here, the threat of circularity looms. To avoid it, the constructivists’ account of the truth of reason judgements must not presuppose the prior existence of any reason facts. Moreover, their account must understand truth in terms that are substantive enough to bear the explanatory weight that the

¹⁰ Compare Bagnoli (2013, 167–8), who writes that constructivists reject the claim ‘that moral judgments are truth-evaluable because they correspond to or represent some portion of reality’ and that, rather than holding ‘that [moral] facts are truth-makers’, they hold ‘that such a domain [of facts; A. M.] is the result of the activity of practical reasoning’. ¹¹ This way of drawing the contrast between constructivism and its alternatives takes cues from Kit Fine’s distinction between ‘antifactualist’ and ‘factualist’ accounts of a certain practice: The factualist’s account must in this sense be representational: it must link up the practice with the underlying facts or subject matter, while the antifactualist’s account will be nonrepresentational. In the one case, the practice must be seen as engaging with the possible facts and it must be understood—at least, in part—in terms of how it engages with those facts. In the other case, the practice is taken to be disengaged from the facts; and rather than understanding the practice in terms of how it represents the possible facts, the facts themselves should be understood in terms of how they are ‘projected’ by the practice. (Fine 2001, 24) Note, however, that despite the similarities between ‘shadows’ and ‘projections’, constructivism differs importantly from Simon Blackburn’s expressivist projectivism, which has projected facts correspond to whatever normative judgement an agent actually endorses; see Blackburn (1988b) and the introduction to Blackburn (1993).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

15

constructivists’ use of Correspondence puts on them. Providing such an account is one of the main challenges that advocates of constructivism have to meet in order to deliver on the promise that their theory of practical reasons holds. As I will argue in Chapters 4 and 5, the best way to meet this challenge is for constructivists to build on their account of the nature of reason judgements as an attitude that guides practical reasoning and propose that the truth of those judgements is to be understood in terms of the soundness of the episodes of reasoning that they are apt to guide.¹² Offering a constructivist account of some domain thus has consequences for understanding what the truth of judgements in that domain amounts to. Here, too, I am in agreement with how both Rawls and Korsgaard characterize constructivism. Rawls acknowledges that because they reject the rational intuitionists’ account of moral judgements, constructivists cannot ‘conceiv[e] of truth in a traditional way by viewing moral judgements as true when they are both about and accurate to the independent order of moral values’ (1993, 92). In other words, constructivists cannot understand truth in terms of accurate representation. Rawls seems unsure what to make of this consequence. Sometimes he recommends that constructivists should not think of moral judgements or principles as being true but rather as being the most reasonable ones, highlighting that this allows us to avoid any commitments to a theory of truth that deviates from the standard representationalist account that is most plausible in the case of ordinary empirical judgements (1980, 554). At other times, however, Rawls seems to suggest that constructivism puts forward an alternative conception of the truth of moral principles and judgements, one that understands their truth in terms of the notion of reasonableness and which allows constructivists to affirm the genuine truth-aptness of those judgements, despite their rejection of the intuitionist’s representationalist account of them (1980, 569; 1993, 126 n. 34).

¹² Showing how a constructivist account of practical reasons can avoid circularity is also a major concern in James (2007). James argues that this commits constructivists to showing that the norms of practical reasoning do not derive their authority from the fact that they lead to reason judgements that are true on independent grounds (2007, 308). The account of correct practical reasoning presented in Chapter 5 allows constructivists to discharge that commitment.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

16   ? Korsgaard is less ambivalent when it comes to the constructivist’s commitments in the theory of truth. Her constructivism and what she calls substantive realism do not disagree on whether or not moral judgements can be true, but on what makes them true, if they are.¹³ The substantive realist claims that there are correct answers to moral questions ‘because there are moral facts . . . that those questions ask about’ (1996, 35; her italics).¹⁴ In other words, an answer to a moral question is correct because it gets the moral facts right. Korsgaard’s constructivism, by contrast, is meant to give an account of what makes answers to moral questions correct that does not refer to any antecedent moral facts. She rejects the idea that the procedures which lead us to correct answers to moral questions are procedures that ‘track’ certain ‘facts which exist independently of those procedures’ and thereby provide ‘ways of finding out about a certain part of the world, the normative part’ (1996, 36–37).¹⁵ This does not prompt her to be sceptical about the truth-aptness of moral judgements, though. All that is required for truth, Korsgaard claims, is that there is a standard for the correct use of the relevant concepts: ‘when a concept is applied correctly, what we get is truth’ (2003, 117). Such a standard is provided by what she calls ‘the correct conception’ of the concept in question. But what makes a conception correct, in the case of moral and other normative concepts, is ‘that it solves the problem,¹⁶ not that it describes some piece of external reality’ (2003, 117).

¹³ Hence, they are both versions of realism in a broader sense, which includes anyone who believes that there are correct answers to moral (and other normative) questions and thus thinks that ‘ethics isn’t hopeless’ (1996, 34–5). Korsgaard calls this broad version ‘procedural realism’, presumably because she thinks that the availability of procedures that will lead us to correct answers to moral questions is a necessary condition, if not for there being such answers, then at least for the pursuit of those answers to not be a hopeless endeavour. ¹⁴ Korsgaard counts Samuel Clarke, Richard Price, G. E. Moore, David Ross, H. A. Pritchard, Peter Railton, David Brink, Thomas Nagel, and Derek Parfit, among others, as proponents of substantive realism (1996, 19; 2003, 101). Note the similarity between this list and the list of philosophers Rawls identifies as ‘rational intuitionists’ (see n. 7 above). ¹⁵ See also Korsgaard (2003; 2008, 22–3). ¹⁶ That is, the practical problem schematically denoted by the concept in question; see also n. 8 above.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  -

17

1.4 Constructivism and mind-dependence According to constructivism, that some things are reasons for certain actions can be explained, initially, in terms of the truth of certain reason judgements and, ultimately, in terms of the soundness of certain episodes of practical reasoning. Constructivists thus deny that reason facts are entirely independent of our mental states and activities. They are committed to the following claim: Mind-Dependence Facts about our mental states or activities figure ineliminably in an account of why some consideration is a reason for an agent to do something. Note that this is a rather weak understanding of mind-dependence. It does not entail, for example, that an agent’s reasons co-vary with her mental states or with how she engages in the relevant mental activities. Hence, constructivists do not suggest that practical reasons are subjective in any such straightforward sense. But they are committed to denying that reason judgements are objectively true in the sense that the measure of their truth is provided by some realm of independent objects. In so far as they want to allow for objectivity in matters of practical reasons, constructivists thus need to offer a different understanding of what that objectivity consists in, one that is compatible with their commitment to mind-dependence.¹⁷ Some authors, including both sympathizers and critics of constructivist views, have characterized them solely in terms of their commitment to the mind-dependence of the relevant facts. Russ Shafer-Landau (2003, 13), for example, introduces moral constructivism as a view which, like the realism he prefers, ‘endorse[s] the idea that there is a moral reality that people are trying to represent when they issue judgements about what is

¹⁷ Aaron James (2012) calls this kind of objectivity ‘protagorean objectivity’. Carla Bagnoli (2002, 131), too, notes that constructivists are committed to an alternative understanding of objectivity that does not presuppose the accurate representation of ‘a (special) sector of reality’.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

18   ? right and wrong’. What distinguishes constructivism, he continues, is how it conceives of this reality which moral judgements are trying to represent: it explains this reality as the ‘output’ of a ‘constructive function’ that takes ‘attitudes, actions, responses, or outlooks of persons, possibly under idealized conditions’ as its inputs, whereas realists endorse the mind- or ‘stance-independence of moral reality’ (2003, 14–15). Shafer-Landau thus portrays constructivists as accepting Representationalism and taking the facts represented by moral judgements to be mind-dependent in some sense. Sharon Street, an advocate of constructivism, also focuses on the issue of mind-dependence in characterizing her position, drawing upon Shafer-Landau’s notion of ‘stance independence’ (Street 2006, 110–11). Characterizing constructivism exclusively in terms of its commitment to mind-dependence results in a much broader understanding of the position than the one I have suggested.¹⁸ In fact, the position which Shafer-Landau and those who adopt his characterization of constructivism seem to pick out with that label is a version of Representationalism. More specifically it is a version of Representationalism that endorses a certain conception of the kind of facts represented by our judgements: what we are ‘trying to represent’ is a certain mind-dependent part of reality. It is thus a position that conflicts with the constructivist position presented in the present book, which rejects Representationalism altogether. Now, as I mentioned before, everyone is free to define and use such labels as they like. The positions characterized by Shafer-Landau and myself can be understood as two alternative ways of spelling out the constructivist metaphor that reasons are not discovered but made. They emphasize different aspects of that metaphor, focusing either on the suggestion that reasons are some sort of artefact or product that is the output of a process that takes mental states or activities as an input or on the metaphor’s implications for the relation between those reasons and our judgements about them, which is claimed not to be one of

¹⁸ On Shafer-Landau’s characterization, dispositionalist views of value like the ones developed by David Lewis (1989) and Michael Smith (1994), as well as Mark Schroeder’s (2007) Humean reductivism about reasons, seem to qualify as forms of constructivism.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  -

19

‘discovery’. Nevertheless, it would be foolish to deny that both ways of characterizing constructivism are legitimate attempts to translate the constructivist metaphor into a set of more precise meta-ethical theses. Still, I submit that characterizing constructivism as a position that accepts mind-dependence and rejects Representationalism carves out a more promising position that offers a genuine and fundamental alternative to the more established meta-ethical positions. Moreover, as we saw in the previous sections, this way of understanding constructivism takes up core elements of the views developed by Rawls and Korsgaard and is thus hardly revisionist. Nevertheless, Rawls and Korsgaard also agree that adopting a constructivist account for a certain domain entails that the facts of that domain are mind-dependent in some sense, and that this has consequences for how we must understand the idea that truth in that domain is nevertheless an objective matter. In the case of Rawls, this can be illustrated with the following passage from his Dewey lectures: Kantian constructivism holds that moral objectivity is to be understood in terms of a suitably constructed social point of view that all can accept. Apart from the procedure of constructing the principles of justice, there are no moral facts. Whether certain facts are to be recognised as reasons of right and justice, or how much they are to count, can be ascertained only from within the constructive procedure, that is, from the undertakings of rational agents of construction when suitably represented as free and equal moral persons. (Rawls 1980, 519)

This passage is often referred to when characterizing how Rawls understands constructivism,¹⁹ and it makes clear that he takes moral facts to depend on our mental states. Rawls also emphasizes that constructivism ¹⁹ See, e.g., Darwall et al. (1992, 138) and Street (2010, 365). Note, though, that while this passage provides a pithy statement of some important aspects of Rawls’ constructivism, it leaves others entirely unmentioned, in particular the rejection of the rational intuitionist’s representationalist account of moral judgements. Focusing solely on this passage thus runs the risk of ignoring that constructivism—understood as a comprehensive moral doctrine—is committed not only to the mind-dependence of moral facts but also, and primarily, to an antirepresentationalist account of moral judgements.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

20   ? and rational intuitionism come with different conceptions of objectivity. Constructivism, he stresses, is not a form of subjectivism; it allows for fallibility by providing a standard of correctness to which individual judgements can be held and which they can fail to meet. But it denies that this standard of correctness is provided by a prior and ‘independent order of moral values’ (1993, 112; cf. 2000, 243–7). Instead, constructivism takes a moral judgement to be correct if it ‘meets all the relevant criteria of reasonableness and rationality’ which are incorporated into a constructivist procedure (1993, 114). Korsgaard, too, acknowledges her view’s commitment to the minddependence of normative facts. As we saw earlier, she denies that a normative judgement is true because it accurately describes any normative facts that are part of some external reality. But she does not deny that there are normative facts, or that these facts correspond to our true normative judgements. She takes them to form a ‘constructed reality’, a reality constructed by the correct use of normative concepts (2003, 117). Hence, normative facts are not discovered in practical thinking, that is, thinking that is trying to solve the problems associated with those concepts. Here, Korsgaard reverses the order of explanation. According to her, normative facts can be explained in terms of practical thinking that is done correctly, that is, in accordance with the principles provided by the correct conception of practical reasons: ‘When those principles are applied to facts and cases, they pick out the substantive considerations that we then regard as reasons’ (2008, 3). That is why, on her account, reasons are ‘mind-dependent’, where this means that ‘the body of facts in question would not exist were it not necessary for human beings to conceptualize the world in a certain way, where the aim of that mode of conceptualization is not simply one of describing the way things are’ (2009a, 28).

1.5 Constructivism, explanation, and the grounding relation Constructivists, I have said repeatedly, want to explain reasons in terms of true reason judgements, rather than vice versa; they claim that reason

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

, ,     21 facts ultimately obtain because of or in virtue of the soundness of certain episodes of practical reasoning. To be able to fully understand the constructivist position and compare it to the various other accounts of practical reasons on offer, more needs to be said about what these phrases are meant to convey. What kind of explanation does the constructivist seek to provide? Reason facts are not a causal consequence of any of our judgements, so the kind of explanation at issue is not the one provided by saying that a bridge collapsed because it was hit by a lorry, for example. Instead, constructivists aim to give a metaphysical explanation of reason facts. This is the kind of explanation we might be offering when we say that the ball is red because it is scarlet, or that Peter is in pain because his c-fibres are firing, or that the vase is fragile because it is made of thin glass. We often use the phrase ‘in virtue of ’ to refer to the relation that underwrites such explanatory claims: the ball is red in virtue of being scarlet; he is in pain in virtue of his c-fibre activity; the vase is fragile in virtue of being made of thin glass. This phrase, when it is used in this way, refers to a particular ontological relation between facts: the grounding relation. Facts can be grounded in other facts, and if they are, we can give a special kind of explanation of the former in terms of the latter. The grounding relation underlies this metaphysical explanation in the same way that the relation between cause and effect underlies causal explanations. It is necessary for the corresponding explanation to be available, but it is not sufficient. As Peter Lipton points out, ‘[t]he big bang is part of the causal history of every event, but explains only a few’ (1990, 249). In general, whether or not a cause makes for a good explanation is highly sensitive to our interests and our background knowledge. Similarly, the grounding relation is necessary for the availability of a distinctive noncausal kind of explanation, but whether or not giving the ground of a fact makes for a good explanation might depend on additional factors. The grounding relation is studied and discussed extensively in contemporary metaphysics.²⁰ What motivates a large part of this attention is a shift in our conception of ontology, away from the traditional Quinean ²⁰ See e.g. Fine (2001), Rosen (2010), Schaffer (2009), and the contributions in Correia and Schnieder (2012). For an overview, see Bliss and Trogdon (2016).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

22   ? view, according to which ontology is solely concerned with what there is and aims at drawing up a large list of everything that exists, an inventory of the world, as it were. As Jonathan Schaffer puts it, Quinean ontology conceives of its object as flat, asking only whether or not it includes a particular entity, whereas his alternative ‘neo-Aristotelian’ conception of ontology takes reality to be structured by a relation that orders things from more to less fundamental (Schaffer 2009, 345–56). This structuring relation is the grounding relation. In addition to answering Quine’s ‘What is there?’ ontologists thus have to answer the question ‘What grounds what?’ Like ‘existence’, the notion of grounding is a primitive metaphysical notion, one that cannot be defined in any other terms. To give an informative account of the grounding relation, we can, however, give illustrative examples, characterize its formal features, and compare it to other philosophically important relations. I have already mentioned some examples of what appears to be claims about what grounds what earlier in this section. Further examples of grounding claims include ‘This is a triangle because it has three sides,’ ‘Anne is married to Claire because they were pronounced spouses by a suitably authorized person,’ and ‘There is a table in the room because there are particles arranged tablewise in the room.’ In general, the relation is often taken to hold between facts concerning dispositional and categorical properties (the vase is fragile because it is made of thin glass), determinable and determinate properties (the ball is red because it is scarlet), macro- and microphysical objects (there is a table in the room because there are particles arranged tablewise in the room), sets and their members (there is a football team on the field because there are several players on the field), etc.²¹ In all of these cases, there is an intuitive ‘directedness’ in the relation between the facts in question, which makes it asymmetric—if Peter is in pain because his c-fibres are firing, then it is not the case that his c-fibres are firing because he is in pain (unless we equivocate on ‘because’ and ²¹ I do not claim that the grounding relation actually holds in all of these cases, but merely that these are typically understood as instances of that relation. Of course, some of them might turn out to be false examples. As long as they help to make clear what they are supposed to be examples for, they have done their job here.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

, ,     23 interpret it as indicating a grounding relation on the first occurrence and an evidential relation on the second). This is true no matter which phrase we use to indicate grounding: if the ball is red in virtue of being scarlet, then it cannot be scarlet in virtue of being red, and, to use another such phrase, if being scarlet makes the ball red, then being red does not make the ball scarlet. Hence, the most salient formal feature of the grounding relation is its asymmetry. Moreover, the relation is irreflexive (it makes no sense to claim that the ball is red because it is red) and transitive (if the ball is coloured because it is red, and it is red because it is scarlet, then the ball is coloured because it is scarlet).²² How does this grounding relation compare to other philosophically important relations? One such relation is reducibility. A stated goal of many theories in all areas of philosophy, including, of course, metaethics, is to reduce certain facts to certain other facts. Is asking what grounds what just a different way of asking what is reducible to what? That depends on how exactly we understand the reductive project. Firstly, establishing the reducibility of, say, pain to c-fibre activity should be distinguished from providing a conceptual analysis in the narrow sense of defining ‘pain’ in terms c-fibre activity and claiming that this conceptual relation is accessible a priori to all competent speakers. Reductivists are ultimately interested in the relations between facts or properties, not those between words or concepts (even though, of course, the latter can be a guide to the former—if ‘pain’ can be defined as ‘c-fibre activity’, then pain is also reducible to c-fibre activity). The classic example of reducibility in this sense is the relation between (facts about) water and (facts about) H₂O. Note, though, that such reduction claims are often understood as establishing the identity of the properties or facts picked out by different concepts: water just is H₂O, so ‘this glass is filled with water’ and ‘this glass is filled with H₂O’ correspond to the same fact. If the aim of a reductivist project is to establish an identity claim, then grounding differs significantly from reducibility, since the grounding ²² Many metaphysicians also introduce the notion of partial grounding: a partial ground is a proper part of a set of facts that (fully) grounds a fact, and while the full ground of a fact necessitates it, a partial ground does not (Audi 2012, 698). Here and in what follows, I am concerned with full grounding.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

24   ? relation is asymmetric, irreflexive, and holds between distinct facts, neither of which is true of identity. I suspect, though, that very often and despite assertions to the contrary, reductivists are not actually interested in establishing identities. After all, they are usually trying to reduce some facts or properties to some ‘more basic’ or ‘more fundamental’ facts or properties.²³ Note also that just like ‘because’ and ‘in virtue of ’, ‘is reducible to’ has a clear connotation of directedness, which suggests that what is meant by reducibility is an asymmetric relation between facts or properties that differ in their fundamentality. Grounding is just such a relation. Thus, talk of reduction is at best ambiguous and at worst misleading. To reduce A to B might simply amount to establishing a symmetric identity claim, as (perhaps) in the case of H₂O and water. But it might also turn out to mean that A-facts are grounded in B-facts, in which case we might as well say so. Another important relation that might bear some similarity to the grounding relation is supervenience. First and foremost, supervenience is a modal relation between two sets or kinds of properties: to say that the A-properties supervene on the B-properties is to say, roughly, that two things that do not differ in their B-properties cannot differ in their A-properties either. There are several more precise specifications of this relation in the literature (see McLaughlin 1995 and McLaughlin and Bennett 2011, secs. 3.5; 4.1). But they all characterize supervenience as a reflexive²⁴ and non-symmetric²⁵ relation. Thus, the fact that certain A-properties supervene on certain B-properties does not suffice to establish the sort of directedness that is distinctive of the grounding relation, and hence does not entail that the former are grounded in the latter. But if, as is plausible, the grounding relation holds with (metaphysical)

²³ Sometimes, that might mean only that the same facts are better understood under another description that is in some sense more basic. In other cases, however, reduction claims are clearly intended to assert an asymmetric relation between properties or facts picked out using different concepts—this will be the case whenever someone who purports to have reduced A to B denies that B is reducible to A as well. ²⁴ Trivially, no two things can differ in their A-properties without differing in their A-properties. ²⁵ A binary relation is non-symmetric if and only if it is neither symmetric nor asymmetric. That is, if R is a non-symmetric relation, then aRb entails neither bRa nor its negation, but is consistent with either. To see that two kinds of properties can be mutually supervenient, consider for example the radius and the surface of perfect spheres.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

, ,     25 necessity,²⁶ then that relation might, conversely, offer a good explanation for why a supervenience claim holds for certain properties. After all, supervenience claims merely describe a certain pattern among properties; they tell us nothing about why those properties exhibit that pattern. Hence, supervenience claims are usually in need of an explanation (see McLaughlin and Bennett 2011, sec. 3.7 and Horgan 1993). One possible explanation is that the two kinds of properties are actually identical. Yet many philosophically interesting supervenience claims are made while insisting that the two kinds of properties and hence the facts about their instantiation are numerically distinct. An account of the A-properties that takes them to supervene on but be nevertheless distinct from the B-properties is often labelled a ‘nonreductive’ account of A-properties in terms of B-properties (here ‘reduction’ is understood as establishing identity). But it seems that establishing such a modal connection between distinct properties does not itself explain anything or solve any problems; it rather states a phenomenon that it is the task of a genuinely informative account of the properties in question to make sense of (see Kim 1993, 165–9). The grounding relation offers a way to do so.²⁷ The preceding paragraphs, I hope, provide an understanding of the grounding relation and the sort of metaphysical explanation it affords that is sufficient for the use they will be put to in what follows. Like ‘existence’, ‘grounding’ is a primitive metaphysical notion; it denotes a relation that we are familiar with from a number of contexts and examples and that we often refer to by using ‘because’ or ‘in virtue of ’, a relation that is asymmetric, irreflexive, transitive, and holds between distinct facts. Thus, claims about what grounds what must be distinguished from identity claims as well as claims about what supervenes on what, but they might provide an explanation for the latter. Moreover, grounding allows us to talk less ambiguously about the asymmetric, explanatorily potent dependence relation we often appear to have in mind when we think about reducibility. This notion of grounding allows

²⁶ For arguments to the effect that it does, see Fine (2012, 38) and Rosen (2010, 118). ²⁷ See Väyrynen (2017) for a helpful overview of the difficulties that the challenge to explain supervenience raises for various other meta-ethical views.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

26   ? us to formulate constructivism more clearly as a metaphysical view about the place of reason facts in an ontology that is structured by the grounding relation. Constructivism is a view about what grounds reason facts.

1.6 Putting constructivism on the map Let me summarize the view as it has been outlined so far. Constructivism about practical reasons rejects a representationalist account of the nature of reason judgements. Their cognitive function is not to provide an agent with normative information by representing facts about what is a reason for doing what. Instead, they play the distinctive role of guiding the agent’s practical reasoning. Accordingly, reason judgements are not true in virtue of accurately representing any reason facts. That does not mean that they cannot be true at all. But according to the constructivist, what it is for a reason judgement to be true should rather be understood in terms of the soundness of the episode of reasoning that it is apt to guide. Constructivists thus deny that true reason judgements are true because certain reason facts obtain. But they do not deny that those judgements correspond to such facts, that is, they do not deny that a reason fact obtains whenever a reason judgement is true. What is crucial about constructivism, however, is that it explains the former in terms of the latter: some considerations are reasons for certain actions because the corresponding reason judgement is true, and thus, ultimately, because a particular episode of practical reasoning is sound. The constructivist’s picture thus includes reason facts, but their existence is not independent of us. Instead, they are mind-dependent: facts about our mental states or activities figure ineliminably in the explanation of why they obtain. How does this view relate to other, more established meta-ethical theories? It has been argued that constructivism does not constitute a meta-ethical theory at all (Darwall et al. 1992; Hussain and Shah 2006). It should be clear, however, that this scepticism is unfounded when it comes to the view presented here. It might be difficult to say exactly what conditions have to be met in order for a theory to qualify as metaethical. But I take a sufficient criterion to be that the theory answers (some of) the same questions that more paradigmatic meta-ethical

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

    

27

theories (i.e. those whose credentials as meta-ethical are not called into question) are characteristically concerned with. Surely questions on the nature of reason judgements, on whether they can be true and what their truth consists in, and on what grounds reason facts are among those characteristically meta-ethical questions. But what kind of meta-ethical theory is constructivism? In particular, is it a form of realism or anti-realism? I am not sure how useful this distinction is any more, because it can mean so many things. If being a realist about practical reasons just means to acknowledge that we can be correct or mistaken in our judgements about them and that, if we are, these judgements are true or false in a substantive sense, then constructivists are realists. They also agree with realists that our reason judgements sometimes are true and hence that there really are practical reasons. Typically, however, realism is meant to be more demanding: reason facts do not only have to exist, but they have to be especially ontologically robust, which is often spelled out in terms of mind- or stance-dependence (cf. Shafer-Landau 2003). A related way of drawing the distinction takes inspiration from Plato’s Euthyphro and characterizes realism and antirealism as disagreeing about whether there are reasons ultimately because we think there are, or vice versa. Understood in this way, constructivism is a version of anti-realism, it embraces mind-dependence, and sides with Euthyphro (cf. James 2007; 2012 and Street 2010). However, many people—particularly in contemporary metaphysics— seem to think that the question whether some particular kind of fact exists (i.e. is part of our ontological inventory) is not the only interesting, and perhaps not even the most interesting ontological question to ask. We should also be concerned with ontological structure and the grounding relation, that is, we should ask not only ‘What is there?’ but also ‘What grounds what?’ (Schaffer 2009). If you agree, as I do, then the classical contrast between realism and anti-realism becomes less important, and the contrast between views that do and views that do not take reason facts to be grounded in certain other facts becomes more relevant. We can distinguish three positions on this issue. Firstly, one might take reason facts to be fundamental and not grounded in any other kind of facts; they are facts sui generis and make up a distinct part of reality. Furthermore, our reason judgements purport to represent those facts,

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

28   ? and they are true, if they are true, because the relevant facts exist and they hence represent accurately. Thus, nothing grounds reason facts, and reason facts ground the truth of our reason judgements. This seems to best describe the position taken by reason fundamentalists like Derek Parfit (2011b) and Thomas Scanlon (2014).²⁸ Secondly, one might deny that reason facts are fundamental and take them to be grounded in some other kind of facts, e.g. facts about our desires, but retain the fundamentalist’s claim about the relation between those facts and our reason judgements. That is, reason facts are themselves grounded in certain other facts, but they ground the truth of our reason judgements when they are accurately represented by them. This position is taken, for instance, by Humeans like Mark Schroeder (2007), and I will call it reason reductivism.²⁹ Finally, one might take reason facts to be grounded in our true reason judgements, so that the fact that some consideration is a reason for an agent to do something can be explained by the fact that the corresponding reason judgement is true, rather than vice versa. Of course, to avoid circularity, such a position must take the truth of our reason judgements to be grounded in something other than the obtaining of certain reason facts, and thus to consist in something other than their accurate representation. This is the constructivist view as I have introduced it. Note that Humean and other versions of reductivism that take reason facts to be grounded in psychological facts, for instance about the agent’s desires, also entail that reasons depend in some way on our mental states or activities. Focusing solely on a position’s stance on mind-dependence will thus lead you to assimilate constructivism to reductivism and to ignore the important difference that distinguishes constructivism from

²⁸ Note that reason fundamentalism—the claim that reason facts are neither grounded in nor identical to any other facts—must be distinguished from primitivism about the concept of a reason, which claims that this concept cannot be analysed in terms of any other concepts. Scanlon and Parfit are reason primitivists in this sense too, but, in general, those two theses can come apart. That is, one might deny the analysability of the concept of a reason while proposing a metaphysical ‘reduction’ of reason facts (Schroeder 2007 is an example of such a position). ²⁹ I choose this term despite the ambiguity noted in section 1.5 because it is so widely established in the debate. As I said above, I suspect that many of those who are considered—or consider themselves—reductivists about reasons actually intend to argue for a grounding claim. For the sake of completeness, however, I will classify anyone who argues for the strict identity between facts about reasons and certain other facts as a reductivist as well.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

    

29

both fundamentalism and reductivism: constructivists reject the idea that the truth of a reason judgement is grounded in the existence of a fact that it represents as obtaining. This is why Representationalism is at the heart of the issue between constructivism and its alternatives. This brings us to a worry about constructivism that has been raised repeatedly in the literature: is constructivism really a distinct metaethical view, or does it, upon closer inspection, collapse into one of the more familiar views?³⁰ Those who think of constructivism primarily as a view that rejects Representationalism often question whether it offers a genuine alternative to expressivism. As I have outlined in section 1.2, however, while constructivism and expressivism share a commitment to rejecting Representationalism, they differ substantially in how they conceive of the role that reason judgements play in an agent’s psychology and, consequently, in their positive accounts of the nature of those judgements. Those who think of constructivism primarily as a view that embraces mind-dependence, on the other hand, tend to question whether it offers a genuine alternative to familiar reductive accounts. David Enoch (2009, 328–9), for example, notes that dispositionalist views like David Lewis’ (1989) qualify as a version of constructivism as he characterizes it. But as we just saw, if we understand constructivism along the lines suggested here, it differs significantly from those reductive accounts, because it reverses the order of explanation between the truth of reason judgements and the obtaining of the facts to which they correspond. In addition to this ‘someone already said that’-version of the distinctness worry, Enoch also presents a more general version of the worry: It seems to be possible to divide up the logical space of metaethical positions rather neatly, using a series of yes-no questions. . . . In such a way, we seem to get an exhaustive classification of all logically possible metaethical positions. But when the details of this little exercise are filled in . . . , no room remains for a distinctively constructivist position. (Enoch 2009, 329) ³⁰ Versions of this worry are raised, e.g., by David Enoch (2009), James Lenman (2012), and Michael Ridge (2012).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

30   ? His idea is that by asking constructivists to answer a series of yes-or-no questions, one can usher them along the forking paths of a taxonomic tree, all of which eventually lead to boxes that already have different and familiar names on them (such as ‘Dispositionalism’). We can call this the argument from taxonomic exhaustion. Does it establish that constructivism is not a distinct meta-ethical view? It does not. Firstly, the most we can and should say about meta-ethical views that end up in the same box of such a taxonomy is that their disagreement on one or more of the many issues discussed in meta-ethics is sufficiently small to allow them to answer certain questions in the same forcedly non-nuanced way. Whether or not this should prompt us to classify them as versions of the same position rather than as distinct alternatives, it seems to me, is not a particularly sensible question to ask, as long as there is some substantive element of disagreement. Moreover, as we saw in the case of ‘realism’ earlier in this section, such classifications are always prone to lead to merely verbal disputes. They derive the little importance they have from the much more important goal of finding out which answers to the underlying questions are correct, and their adequacy is highly sensitive to the context and the background assumptions of the discussions in which they are used. This is not to deny that taxonomies can be very useful, say, for giving an overview of an area or for making certain dialectical situations explicit. But we should keep in mind that what questions we use in generating them will depend not only on abstract systematic considerations, but on what questions we deem important enough to give different names to different answers, which in turn depends on what problems we deem worth solving and what general concerns we take meta-ethical theories to be in the business of addressing. Secondly, even if we put these general concerns aside, Enoch’s argument still fails to establish that there is ‘no room . . . for a distinctively constructivist position’ in meta-ethics (2009, 329). This is because the taxonomy he offers is not exhaustive; it conflates important distinctions in the meta-ethical realm. In particular, it moves from answering ‘yes’ to the question ‘Are any moral sentences strictly speaking true?’ directly to the question ‘Are moral properties (or facts) naturalistically reducible?’ (2009, 329) Thereby, it ignores precisely those questions concerning the

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

   

31

proper understanding of truth and the relation between facts and true judgements on which constructivists and their opponents disagree. Moreover, it does not include any questions on the nature of the normative judgements in question. Rather than demonstrating that there is no room for constructivism in meta-ethics, Enoch’s argument thus illustrates how a one-sided conception of the issues that meta-ethicists should be concerned with can make us blind to the very possibility of a distinctively constructivist account of practical reasons.

1.7 Plan of the book In the remainder of this book, I will develop the constructivist account of practical reasons that I outlined in the preceding sections into a fully fledged meta-ethical view, make its theoretical commitments explicit, and respond to the most pressing worries to which it gives rise. My aim in doing so is not to convince you that constructivism is the best metaethical view, or that it is superior to any of its alternatives. In fact, I will have little to say about those alternatives, except where I need to distinguish constructivism from them. My aim is a more modest one: I want to show that constructivism offers a distinct, coherent, and prima facie attractive account of practical reasons, and to make as clear as possible what accepting that account would involve. This, I hope, will facilitate a more focused and in-depth debate about its merits. To that end, the following chapters will elaborate the aspects of the constructivist view that I outlined in sections 1.2–1.4 and address the two worries that have already been mentioned—concerning the view’s distinctness and circularity—in much greater detail. Chapters 2 and 3 will develop the constructivist’s alternative to Representationalism. Chapter 2 argues that reason judgements play an important and essential role in practical reasoning: they guide the reasoning process. Chapter 3 then shows how this role can be used to formulate an anti-representationalist account of the nature of reason judgements. It also discusses how that account differs from how the expressivists, who also reject Representationalism, understand of the nature of such judgements.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

32   ? Chapter 4 presents an account of the truth of reason judgements that does not presuppose the prior existence of reason facts but nevertheless understands truth to be a substantive property that can bear the explanatory weight that the constructivist account of practical reasons puts on it. In particular, it suggests that constructivists can formulate such an account by adopting a theory of truth that has been developed most prominently by Crispin Wright (1992) and Michael Lynch (2009). According to the account of truth presented in Chapter 4, the truth of a reason judgement is ultimately grounded in the soundness of the episode of practical reasoning that it is apt to guide. Most importantly, such an episode is sound only if it is an instance of correct reasoning. That is why Chapter 5 addresses the question of what it is for practical reasoning to be correct. To avoid circularity, constructivists must answer this question without presupposing any truths about practical reasons. I show that this can be accomplished by taking the standard of correctness for practical reasoning to be provided by the constitutive rules of that activity. Like the rules of chess, these rules necessarily govern the activity even though there need not be any reason to comply with them. Finally, Chapter 6 distinguishes the particular version of minddependence that constructivists endorse from the version to which reductivists like Mark Schroeder (2007) are committed. These versions differ in the resources they offer for accommodating the idea that practical reasons are objective in some sense. I thus discuss how constructivism and its alternatives fare in accommodating the fallibility of our reason judgements, the fact that some reasons are universally shared, as well as the modal robustness of our reasons, which are three rather different issues associated with the ambiguous notion of objectivity.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

2 Reasoning and reason judgements This chapter takes the first step in developing a constructivist theory of practical reasons. It does so by giving an account of the cognitive role that judgements about those reasons play in an agent’s psychology. In particular, it presents an account of practical reasoning according to which such reasoning always involves a reason judgement that guides it. The thesis that reason judgements guide episodes of practical reasoning will be taken up in Chapter 3, where I will argue that constructivists can formulate an anti-representationalist account of the nature of reason judgements by characterizing their cognitive function in terms of this guiding role.

2.1 Preliminaries 2.1.1 The reasoning process Let me start by clarifying the concepts that will be central to this chapter. Practical reasoning, the kind of reasoning on which I will focus here, is thinking about what to do in a situation in which one has more than one option and hence faces a choice between different courses of action. That is not to say that agents encounter situations in which their options are already carved out for them and then begin to reason about which to choose. Framing one’s situation—determining the available options as well as specifying and distinguishing them at a suitable level of generality—is often part of the deliberative process. Another part of this process is bringing to mind potentially relevant considerations: inquiring into the details of the options and one’s circumstances, imagining the potential consequences of one’s actions, searching one’s memory for relevant information, goals, or previous encounters with similar Constructing Practical Reasons. Andreas Müller, Oxford University Press (2020). © Andreas Müller. DOI: 10.1093/oso/9780198754329.001.0001

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

34     situations, etc. If the reasoning is successful, it will eventually lead to a decision, that is, the formation of an intention to follow one of the available courses of action.¹ Not all reasoning is successful. The process of thinking about what to do in a certain situation can be interrupted by an occurrence that requires the agent’s immediate attention, or the agent might be distracted by an unrelated thought that keeps bumping back into the forefront of her mind. Sometimes we also ‘get stuck’ in reasoning, thinking about how all the aspects of the different choices bear on our decision until the opportunity to make a choice has passed. Such cases of deliberative failure in which we ponder our options without making any decision should be distinguished from reasoning that leads to the result that no decision is necessary (yet), where no intention to perform any of the initially considered options is formed either. The latter kind of reasoning is, nevertheless, successful, because it concludes with a (higher-order) decision not to decide among the available options, in which case this (higher-order) option—not to decide between the other options—is simply added to the range of options the agent takes to be available in the situation at hand. Of course, not everything we do, not all of our choices are the result of reasoning on our part. Very often, we perform actions without first spending any time thinking about or consciously deciding to perform them—we switch on the light when we enter our dark living room, we pick up the pen we have just dropped, we raise our hand to greet a friend we have just seen on the other side of the road. Reasoning, as I understand it in this chapter, is a conscious mental process by which a subject forms a new attitude on the basis of some of the attitudes she already holds.² Reasoning in this narrow sense must be distinguished from other, unconscious and automatic processes, which are also sometimes called ‘reasoning’. This distinction is widely accepted in the litera-

¹ In speaking of intentions as a distinctive mental state, I follow Michael Bratman’s (1987, 20) influential account of intentions and intentional actions, whose characterization of those states as ‘conduct-controlling pro-attitudes’ I assume to be by and large correct in what follows. ² Reasoning can also result in dropping an attitude (cf. Harman 1986, ch. 1). However, like Broome (2013, 221), Boghossian (2014, 3), and many others I will focus on reasoning that results in the formation of a new attitude.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi



35

ture. John Broome calls the former ‘active’ and contrasts it with ‘passive’ reasoning (2013, 221). Paul Boghossian (2014) picks out the kind of reasoning he is interested in by characterizing it as a cognitive process that involves the activity of what cognitive psychologists call ‘System 2’. Daniel Kahneman, a leading proponent of the underlying ‘dual process theory’ of thinking, describes System 2 as the one that ‘allocates attention to the effortful mental activities that demand it’ and that is ‘often associated with the subjective experience of agency, choice, and concentration’ (2011, 21). Boghossian calls such reasoning ‘person-level, conscious and voluntary’, and distinguishes it from ‘sub-personal, subconscious and automatic’ processes (2014, 3). Like Broome and Boghossian, I am concerned only with conscious, active reasoning. Here is a quite ordinary example of such reasoning: suppose that in the morning, you are about to finish your coffee and head out to go to work. You know that you could take either the train or your car, and you plan to be at work by 9 a.m. because there is an important meeting. You also remember that there is going to be extensive construction work on the road you would have to take into the city today, which is sure to cause a traffic jam that would significantly prolong your commute time. Realizing that the train is your only option to get to work in time, you decide to take the train and get on your way to the station. For any particular episode of reasoning, we can identify a set of attitudes—e.g. beliefs and intentions—from which the process of reasoning starts and on which its results are based. We can call these the premise-attitudes of that reasoning. If it is reasoning that leads to the formation of a new attitude, we can call this newly formed attitude the conclusion-attitude of that episode of reasoning.³ Being a premiseattitude or a conclusion-attitude are thus two different roles an attitude can play in the process of reasoning. Reasoning from a certain set of premise-attitudes to a particular conclusion-attitude can be classified as correct or incorrect. Whether or not a particular episode of reasoning is correct depends on the content of the premise- and conclusion-attitudes. Reasoning should thus be understood to operate on the content of these attitudes, where this ³ I adopt the terminology of premise- and conclusion-attitudes from Broome (2013, 221).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

36     means that the semantic features of those attitudes play an essential role in the reasoning process (cf. Broome 2013, 231–2).⁴ But the relevant notion of content here is not simply that of propositional content. Different kinds of attitudes can share the same propositional content, but the kind of attitude they are is still relevant for whether an episode of reasoning in which they are involved is correct. The belief that it is raining and the hope that it is raining share the same propositional content, but while reasoning from the belief that it is raining and the belief that if it is raining, then the streets are wet, to the belief that the streets are wet is correct, reasoning from the hope that it is raining together with the belief that if it is raining, then the streets are wet, to the belief that the streets are wet is not.⁵ Hence, if we want to characterize the role our attitudes play in the process of reasoning in terms of their content, we need a richer notion of content that adds a marker for the kind of attitude to the bare propositional content.⁶ Importantly, though, to say that an episode of reasoning starts from certain premise-attitudes does not mean that it involves thinking about these attitudes. In our example, you are thinking about the meeting at work and the traffic jam on the way there, not about your beliefs concerning these issues. So, when an episode of reasoning starts from a certain attitude, your attention is directed at what this attitude is

⁴ Some philosophers argue that there is an important difference between reasoning that concludes with beliefs and reasoning that concludes with an intention, as I assume practical reasoning does: while the former kind of reasoning is inferential, the latter, they insist, is not (see e.g. Dancy 2004, 101–8). Whether or not this is true depends on how we understand the contrast between inferential and non-inferential reasoning, a distinction that has proven to be quite difficult to explicate (see Streumer 2007a for an overview and a suggestion). One suggestion that Dancy and others seem to have in mind is that reasoning is inferential if it presupposes an evidential relation between the contents of the premise- and conclusionattitudes. If we understand their claim in this sense, the thesis that all reasoning operates on the contents of these attitudes is perfectly compatible with the thesis that not all reasoning is inferential. The claim that the reasoning process is sensitive to the content of the attitudes is significantly weaker than the claim that it is sensitive only to the evidential relations between those contents. ⁵ The fact that an attitude’s propositional content is insufficient to determine its significance for the process of reasoning has also been highlighted by David Velleman (1992, 9) and John Broome (2013, 251). ⁶ Broome (2013, ch. 14.1) addresses this issue by outlining a theory of such ‘marked contents’ and our means of expressing them in ordinary language. For the purposes of this chapter, I will simply adopt Broome’s proposal, assuming that it is on the right track.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi



37

about—you are reasoning with certain attitudes, but you are reasoning about their contents, not about them.

2.1.2 Practical reasoning Practical reasoning, as it was introduced in section 2.1.1, concludes with the formation of an intention. It contrasts with theoretical reasoning, which concludes with a change in our beliefs. The idea that that there is a kind of reasoning that concludes with the formation of an intention, however, has been criticized from two directions. Some authors, including Aristotle (on some interpretations; cf. Kenny 1966) and those inspired by him, like G. E. M. Anscombe (1957, 57–62), think that the conclusion of practical reasoning is an action. But action requires more than reasoning: it requires the cooperation of the world, as it were, including our bodies. Not acting on an intention you have acquired through practical reasoning is not a fault in your reasoning but a failure to act on your reasoning.⁷ Such a failure to act on your reasoning might be simply due to the fact that, unbeknownst to you, some necessary means is not (or no longer) available: perhaps in the very moment you decided to wave to your friend, your arm became paralysed, making it impossible for you to act on your intention to greet her. Moreover, even if you do act on an intention acquired through practical reasoning, the reasoning process plausibly concludes with the intention and does not extend to the action, for you might now acquire the intention to perform some action next year, in which case it is a stretch to say that you have not concluded your reasoning before next year (see Streumer 2010, 248). Other authors, most notably Robert Audi (2006, 70–2) and Joseph Raz (2011, 131–7), have argued that practical reasoning does not even take us to intentions. Instead, they claim, it concludes with a normative judgement to the effect that one has decisive reason to perform some action, or

⁷ Bart Streumer (2010) attributes this helpful way of putting it to Joseph Raz. Broome makes a related point: ‘Making a decision is as close to acting as reasoning can possibly get you. Reasoning could not actually get you to act, because acting requires more than reasoning ability’ (1999, 407).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

38     that doing so would be best.⁸ Their case for this position parallels the arguments just given against the action-as-conclusion view. They claim that the failure to form the intention to act on the judgement that one has decisive reason to perform some action is not a failure in one’s reasoning; practical reasoning is successfully concluded once one has formed that judgement. Both Audi and Raz support this claim by pointing to the familiar phenomenon of weakness of will, that is, to cases in which an agent reasons to the conclusion that she has decisive reason to do something but then does something else, or nothing at all. The possibility of such akratic actions, however, does little to undermine the thesis that there is a kind of reasoning that concludes in an intention. It shows that some episodes of reasoning take us no further than to the judgement that we have most reason to, say, stop eating meat. But that does not entail that the akratic’s intention to order a steak is not also the conclusion of an episode of reasoning. Why deny that there can be two processes of reasoning involved in akratic action: one that concludes with a normative judgement in support of one action, and one that concludes with the intention to perform another? In that case, only the latter qualifies as an instance of practical reasoning, whereas the former is more akin to theoretical reasoning because it concludes with a judgement, albeit one that is practically relevant. Of course, in such cases the practical reasoning is less than optimal in that it fails to exhibit a certain sensitivity to the result of the reasoning about what there is most reason to do (this is what turns the resulting action into an akratic one), but that alone does not entail that it is not an instance of reasoning.⁹

⁸ I will use the notions of ‘decisive’ and ‘sufficient’ reason as follows: An agent has decisive (or simply: most) reason to do A if and only if there is one or more reasons for her to do A and any reasons not to do A are either outweighed by those reasons or disabled by other considerations. If an agent has decisive reason to do something, we might also say that she ought to do it, where ‘ought’ is used in the practical sense in which it occurs in the question ‘What ought I to do?’ (which is thus equivalent to ‘What do I have decisive reason to do?’) and in which an agent is criticizable for not doing what he ought to do. An agent has sufficient reason to do A if there are one or more reasons for her to do A and those reasons are neither outweighed by reasons against doing A nor disabled by other considerations. Hence, an agent can have sufficient reason both to do A and not to do A, but having decisive reason to do A excludes having sufficient reason not to do A. ⁹ I will get back to the relation between these two kinds of reasoning in section 2.7 below.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi



39

Raz, however, insists that all reasoning that results in intention or action proceeds via the conclusion that one has decisive reason for the action in question. According to him, such a judgement ‘is essential to all reasoning about what to do’ (2011, 126). Reasoning can, nevertheless, result in intention and action because an agent who arrives at such a conclusion will also form the corresponding intention, as long as she is rational. But, Raz insists, taking the enkratic step from the normative judgement to the intention is not a matter of reasoning (2011, 136). This account of reasoning and its relation to action faces a number of problems. Firstly, recall that reasoning is a conscious, person-level activity that operates on the contents of its premise- and conclusion attitudes. According to Raz’ account, practical reasoning always involves operating on the content of a normative judgement. It thus necessarily involves thinking about what one has conclusive reason to. But that overintellectualizes much of practical reasoning.¹⁰ Secondly, the account entails that akratic actions cannot directly result from practical reasoning in the same way that ordinary actions can. To see that this is implausible, suppose that you are on a diet and think that you ought not to eat any ice cream. Yet, watching your favourite TV show after a long week of work, you nevertheless form the intention to eat some ice cream. If there was a bowl of ice cream in front of you and you simply started eating it, this would be a paradigmatic example of an akratic action. But you have no ice cream in your home. Hence, you realize that in order to eat ice cream now, you first have to buy some at the corner shop. So, you form the intention to go to the shop and walk out the door. This intention and the ensuing action seem to be the direct result of an episode of reasoning, just as your intention to take the train was in our earlier example. But it is not the result of taking the enkratic step from the judgement that you ought to buy ice cream at the corner shop, because you do not have that judgement. You continue to think that you ought not to eat any ice cream and that you ought not to buy any either. Going to the corner shop is thus an akratic action, just as eating ice cream would have been, had there been any in front of you. But it is ¹⁰ A similar concern regarding Raz’s position is raised by Jonathan Dancy (2014, 10–11).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

40     an akratic action that results from what we can call akratic reasoning. Now, Raz might insist that in addition to your judgement that you ought not to buy ice cream, you also think that you ought to buy ice cream, and that this latter judgement is the conclusion-attitude of your reasoning. That would allow him to explain your intention to buy ice cream as the result of taking the enkratic step from this judgement, and hence as the result of reasoning. But assuming that you also have that judgement seems entirely unjustified. Firstly, it would attribute a quite different and a more fundamental mistake to you by taking you to have overtly inconsistent judgements. Moreover, it is difficult to see from what premises you could have reasoned to that conclusion, given your overall commitment to your diet. If Raz was right and reasoning could take you to an action only by way of the enkratic step, then akratic actions like this one could not be the result of reasoning. But it seems clear that they can. A third problem is that Raz’s argument in favour of his account does not support the conclusion he draws from it. Here is what Raz writes: First, the premises of a reasoning that would yield an intention to Φ as its conclusion would also entail, or at any rate warrant, the conclusion that one must Φ. Second, unless one reasons to the conclusion that one must Φ one would not be justified in forming an intention to Φ as a result of the reasoning. That is, though the intention may be justified it cannot be regarded as a valid result of the reasoning. The existence of reasons to Φ is not sufficient to justify an intention to Φ. The intention is justified only if the reasons are not defeated. So if the intention is the conclusion of reasoning then the reasoning must include, as an intermediate conclusion, either that the reasons to Φ are not defeated by the conflicting reasons, or, that the reasons to Φ are conclusive. But if the intermediate step is permissive, that is, merely that the reasons to Φ are undefeated, then the intention is not warranted by the reasoning . . . . So if the intention is warranted by the reasoning the intermediate step must be that the reasons to Φ are conclusive. (2011, 135)

This suggests that Raz subscribes to a strong kind of internalism about justification via reasoning. If I understand his line of thought in the quoted passage correctly, Raz assumes that reasoning is a way of arriving

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

41

at a justified intention, and then argues that an intention could not be justified as a result of one’s reasoning unless two conditions are met: firstly, the reasoner must have conclusive reason for the intended action, and secondly, the reasoner must be aware that he has conclusive reason for the intended action. The second condition is hardly uncontroversial, as the discussion about the problems of inferential internalism shows (cf. Boghossian 2003a, 229–36). More importantly, however, that kind of internalism only supports the claim that the reasoner must have the reason judgement, not that it must be the conclusion-attitude of the reasoning. Thus, accounts of practical reasoning that allow reason judgements to play a different role from that of a premise- or conclusion-attitude—an option that Raz does not appear to consider—could also accommodate such an internalist commitment. Notably, this includes accounts that take the conclusion-attitude of practical reasoning to be an intention.¹¹

2.2 Conditions of adequacy Before we look at specific theories of practical reasoning, it will be helpful to reflect on the criteria for assessing the merits of different proposals. What do we expect from a theory of practical reasoning? In this section, I present three conditions of adequacy for such theories, that is, three conditions they have to meet in order to be acceptable: They should allow for episodes of incorrect reasoning to count as reasoning, distinguish practical reasoning from other processes of attitude formation, and accommodate the fact that reasoners cannot consistently disapprove of their own reasoning. Firstly, not all reasoning is correct reasoning. This is true of theoretical as well as of practical reasoning. Just as we sometimes form a belief that p by reasoning from the belief that p entails q and the belief that q

¹¹ The account that I will propose later in this chapter is one that takes reason judgements (albeit not ones about conclusive reasons) to play a role in practical reasoning that differs from that of a premise- or conclusion-attitude.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

42     (i.e. by affirming the consequent), we sometimes form an intention to do A by reasoning from the intention to do B and the belief that doing B is a necessary means for doing A (i.e. by reasoning from means to ends). Hence, we must distinguish between the theory of reasoning, which is concerned with the nature of that process and its necessary and sufficient conditions, and the theory of good or correct reasoning, which is concerned with the conditions an episode of reasoning has to meet in order to qualify as correct.¹² Since incorrect reasoning is still reasoning, these latter conditions must be carefully distinguished from the conditions identified by the former kind of theory. In particular, a theory of practical reasoning must not set conditions for a mental process to qualify as reasoning that episodes of incorrect reasoning cannot meet, that is, it must allow for the possibility of incorrect reasoning. Secondly, not every mental process that takes you from a certain initial set of attitudes to the formation of a new attitude is an episode of reasoning. A person’s attitudes might cause her to form other attitudes without the latter being the conclusion of reasoning from the former. Suppose Aki intends to visit his mother and believes that in order to do so, he must buy a plane ticket. This causes his blood pressure to rise (he is terribly afraid of flying), which causes him to feel dizzy, which causes him to intend to sit down. Hence, the attitudes caused him to form the intention to sit down, but we would not say that they did so by way of reasoning. Note that this process too is sensitive to the content of the attitudes involved, so that condition does not suffice to distinguish such cases of ‘mere’ causation from cases of genuine reasoning. In fact, two mental processes can involve transitions from the same set of initial attitudes to the formation of the same new attitude, and yet only one of them qualifies as reasoning. A theory of practical reasoning should allow us to distinguish episodes of reasoning from other processes of attitude formation even in those cases. As we will see later in this chapter, doing so is especially challenging if those transitions do not conform to the standards of correct reasoning. Here is an example to illustrate this challenge: the pilot of a military plane intends to bomb an ammunition

¹² The theory of correct reasoning will be the topic of Chapter 5.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

43

factory. Upon approaching the site, she realizes that bombing the factory will also destroy the nearby palace of the country’s evil dictator. As a result, she forms the intention to destroy the palace. This transition might be an instance of a purely emotion-driven process (she hates the dictator). But it might also be an instance of reasoning from the intention to perform a certain action to intending a known side effect of that action, which, I assume, is generally not a form of correct practical reasoning.¹³ An adequate theory of practical reasoning should provide us with a criterion to distinguish these two cases; it should tell us what sets the pilot who is engaging in incorrect practical reasoning apart from the pilot who arrives at the same intention by way of a merely causal process of attitude formation. Thirdly, reasoners cannot just take any stance towards their own reasoning. Focusing on the case of theoretical reasoning, or inference, Ulf Hlobil points out that it is ‘impossible or seriously irrational’ to make an inference and ‘to judge, at the same time, that [it] is not a good inference’ (2014, 420). He calls this the ‘Inferential Moorean Phenomenon (IMP)’ (2014, 420). David Enoch highlights a similar phenomenon for the case of practical deliberation. Such deliberation, he points out, differs from simply picking a course of action in the following way: ‘We can just pick in the face of a known (or believed) absence of reasons. But we cannot, it seems, deliberate in the face of a believed absence of reasons’ (2011a, 74). John Broome too thinks that part of what makes a mental process an episode of reasoning is that it ‘seems right’ and is, in that sense, personally endorsed by the reasoner (2013, 238). Thus, while they differ on various details, all of these authors suggest that reasoning is incompatible with a stance that denies it any normative backing, that is, with denying that it is a good inference, or believing that it is not supported by any reasons, or failing to endorse it. To put their point in more general terms: reasoners cannot disapprove of an episode of reasoning while they are engaging in it. This is another characteristic feature of reasoning that any plausible account must accommodate.

¹³ Those who do not share this assumption are invited to substitute what they take to be an example of incorrect practical reasoning.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

44    

2.3 The normative guidance account In the remainder of this chapter, I will present and defend an account of practical reasoning that I call the Normative Guidance Account. This account comprises two main theses. The first thesis states a necessary condition for a process of attitude formation to qualify as an episode of practical reasoning: every such episode involves a reason judgement in support of its conclusion. The second thesis specifies the role that this reason judgement plays in the reasoning process: that judgement guides the episode of reasoning. Such an account has considerable intuitive appeal. After all, should reasoning not involve taking something to be a reason? Moreover, accounts that require reasoners to take their conclusions to be normatively supported by the considerations from which they reason easily meet the conditions of adequacy introduced in section 2.2. As we will see in section 2.4, this is a significant advantage over alternative accounts, for which those conditions raise significant problems. However, the claim that reasoning necessarily involves a normative judgement also faces some serious objections. Does it not over-intellectualize reasoning and require a level of conceptual sophistication that some reasoners—such as small children—do not yet possess? Moreover, the claim that episodes of practical reasoning are guided by those reason judgements has raised the worry of leading into an infinite regress. I will discuss these objections against the Normative Guidance Account in sections 2.5 and 2.6, respectively, and address some additional questions in section 2.7.

2.3.1 The normative judgement condition According to the first main thesis of the Normative Guidance Account, every episode of practical reasoning involves a reason judgement. To start out, let me give a more explicit formulation of this thesis: The Normative Judgement Condition (NJC) If a subject S forms an intention to do A by reasoning from certain considerations, then S judges that these considerations provide a reason to do A.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

   

45

Two aspects of the required judgement bear highlighting. Firstly, the judgement is concerned with the normative relation between the starting point (the content of the premise-attitudes) and the end point (the content of the conclusion-attitude) of the reasoning—the former normatively supporting, or providing a reason for, the latter. Secondly, the reason judgement is specific to the particular episode of reasoning in question, because its content specifies the relata of the support relation. Episodes of reasoning that differ in their premise- or conclusionepisodes involve different reason judgements, even if they are instances of the same kind of reasoning (e.g., instrumental reasoning). Claims that resemble NJC have a number of supporters, at least for the case of theoretical reasoning. Frege, for example, writes that ‘To make a judgement because we are cognisant of other truths as providing a justification for it is known as inferring’ (1979, 3). In a more recent discussion of the nature of inference, Paul Boghossian suggests that inference is subject to what he calls the ‘Taking Condition’, according to which it ‘necessarily involves the thinker taking his premises to support his conclusion’ (2014, 5).¹⁴ Similar claims have been accepted by many other authors.¹⁵ Besides the fact that they are concerned with theoretical reasoning, there are some other important differences between NJC and the conditions endorsed by Frege and Boghossian. In particular, they differ with regard to the kind of attitude they require. Frege’s attitude appears to be doubly factive: it seems to entail both that the beliefs from which one infers are in fact true (being cognizant of other truths) and that they do in fact provide a justification for the conclusion (being cognizant of them as providing a justification).¹⁶ This is too strong, because we can reason from false premises and in ways in which the premises would not justify the conclusion even if they were true—not all reasoning is correct reasoning. Boghossian acknowledges this problem and proposes ‘taking’ as a way to avoid it. But this notion can be understood in a number of ¹⁴ This is only the first half of the Taking Condition; the second half states that the thinker draws the conclusion because he takes it to be supported by the premises (Boghossian 2014, 5). ¹⁵ See, e.g., Audi (1993), Leite (2008), Neta (2013), Pauer-Studer (2014), Thomson (1965), and Valaris (2014). ¹⁶ I understand ‘being cognizant of p’ as entailing that p, similar to ‘being aware of p’.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

46     ways. The most natural interpretation is that ‘taking’ is a belief or a judgement, perhaps an implicit one: to take Robin to be the strongest child in the class, it seems, is simply to believe, though perhaps not occurrently, that she is the strongest child in the class. Boghossian, however, rejects such a doxastic understanding of his Taking Condition (2014, 6–7). Instead, he thinks that it is best understood as a sui generis state that is involved in following a rule but that cannot be further analysed (2014, 17).¹⁷ It is also not clear whether the notion of ‘support’ in the formulation of Boghossian’s Taking Condition that I quoted above is a normative one. Frege’s ‘providing a justification’ is less ambiguous in this respect. Later in the same paper, however, Boghossian writes that what he insists is essential to reasoning is ‘taking a circumstance to be a reason for believing something’ (2014, 12; my emphasis). Both Frege and Boghossian thus appear to agree with NJC (or rather, its counterpart for theoretical reasoning) that reasoning involves an attitude with normative content. However, NJC demands less than Frege in that it does not take that attitude to be factive, but it does not share Boghossian’s reluctance to interpret it as a belief or a judgement either.

2.3.2 The guiding role The second main thesis of the Normative Guidance Account of practical reasoning concerns the role which the reason judgements that the Normative Judgement Condition requires play in practical reasoning. A good way to start characterizing that role is to ask what kind of process practical reasoning is. So far, I have specified its starting point (the premise-attitudes) and its end point (the conclusion-attitude). But little has been said about the transition between the two. How does reasoning get from the premise-attitudes to the conclusion-attitude? I think that the formation of the conclusion-attitude is best understood as the manifestation of a certain disposition (cf. Wedgwood 2006). Like most ¹⁷ I will discuss Boghossian’s characterization of taking as a sui generis state and other nondoxastic accounts in section 2.4 below.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

   

47

authors writing on reasoning, including Broome (2013, 206–7) and Wedgwood (2006), I also think that reasoning is or involves a causal process—in some way or another, the formation of the conclusionattitude is caused by the premise-attitudes.¹⁸ Perhaps the manifestation of a disposition is not necessarily a causal process, but it certainly is sometimes: if the fragility of a vase manifests when it is struck, for example, the shattering is certainly caused by its being struck. Let us focus on the stimulus of the disposition that manifests in an episode of reasoning. Of course, this stimulus includes the premiseattitudes. But, I want to suggest, it also includes the reason judgement required by NJC. Both the premise-attitudes and the reason judgement are what we might call stimulus-attitudes of the reasoning disposition.¹⁹ This might seem problematic: if both the premise-attitudes and the normative judgement are part of the stimulus and hence of the cause of the formation of the conclusion-attitude, then does that not mean that they play the same role in the reasoning process?²⁰ Not necessarily. First, notice that different elements of a disposition’s stimulus can contribute to its manifestation in different ways. Suppose an octopus has the disposition to adapt its skin colour to the colour of its surroundings when approached by a predator. The stimulus of this disposition has two parts that play different roles. The approach of a predator determines whether or not the disposition manifests at all— without it, the octopus would not suddenly change its colour. The particular colour of its surroundings determines how the disposition manifests—without it, the octopus would not change its skin to this colour. The premise-attitudes and the reason judgement contribute in different ways to the manifestation of a disposition too. In a nutshell, while the

¹⁸ See Valaris (2014) for a critique of this orthodoxy. ¹⁹ Thus, in so far as the manifestation of a disposition can be said to occur because of its stimulus, I also accept the second part of Boghossian’s Taking Condition, according to which the conclusion-attitude is formed because the reasoner takes it to be supported by the premises (2014, 5). ²⁰ An objection like this is raised by Valaris (2014, 107–8), who argues that the (alleged) inability to distinguish the role of the required normative judgement from that of a premiseattitude forces accounts of reasoning like the one presented here into an infinite regress. Rather than rejecting NJC, he suggests giving up the causal conception of reasoning.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

48     premise-attitudes provide the starting point of the reasoning process, the reason judgement guides that process to a specific end point. Let me illustrate this contrast with a spatial analogy. Suppose you are exploring a new city. You have no idea what the interesting sights are, nor are you familiar with the city’s layout. Unfortunately, you do not have access to a map either. You do, however, have a small leaflet that describes the way from the railway station to a famous church (‘Walk down Elm Street until it crosses Washington Road; then turn left’, etc.). As it happens, you are standing in front of the railway station, so you are following the instructions in the leaflet and arrive at the church. In this scenario, the leaflet acts as your guide: its instructions take you from a specific starting point to a specific destination. The leaflet hence determines where walking from that starting point will lead you. A different leaflet would either be useless to you, because it would begin at a starting point that differs from your current location, or it would take you to a different destination, because it contains a different set of instructions but begins at the same starting point. Of course, manoeuvring through mental space is very different from walking through a city. For one, it does not involve any intentional movement at all. Nevertheless, the way a reason judgement guides you when you are reasoning shares an important feature with the way the leaflet guides you through the city. Like the leaflet, the judgement determines where (that is, to the formation of which conclusion-attitude) the reasoning process takes you from a specific starting point (that is, a specific set of premise-attitudes). Note that the reason judgements required by NJC have exactly the right structure to provide such guidance. Their content normatively connects the premise-attitudes to the conclusion-attitude (see section 2.1). If your current mental states do not include the premise-attitudes specified by a certain reason judgement (if the specified starting point does not match your current ‘mental location’), a reasoning process guided by that judgement could not get started. If your current mental states include a certain starting point, then different reason judgements that share this starting point could lead you to the formation of different conclusion-attitudes. For example, if you believe that it is going to rain, then the normative belief ‘That it is going to rain is a reason to take an umbrella’ can guide your reasoning to

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

   

49

the intention to take an umbrella. The belief ‘That it is going to rain is a reason to wear a raincoat’, on the other hand, can guide your reasoning to the intention to put on a raincoat. The role the reason judgement plays in bringing about the formation of the conclusion-attitude thus differs in a crucial respect from the role of the other stimulus-attitudes (i.e. the premise-attitudes). Changing the reason judgement while holding all the other stimulus-attitudes fixed can redirect the reasoning process, as it were, and result in the formation of a different conclusion-attitude. It does so only if changed in the right way, i.e. by altering the second relatum (what is normatively supported) but not the first one (what provides the support). Changing one of the premise-attitudes while holding all other stimulus-attitudes fixed, on the other hand, is never sufficient to result in the formation of a different conclusion-attitude. All it can do is make the reason judgement inapplicable, so that no reasoning based on these attitudes will occur. A different starting point always requires a different guide. Thus, only the reason judgement required by NJC meets the following condition: Guidance An attitude guides an episode of reasoning if and only if (i) it is a stimulus-attitude of the reasoning disposition and thus contributes to causing the formation of the conclusion-attitude, and (ii) changing its content in a suitable way while leaving all other stimulus-attitudes unaltered can be sufficient for the disposition to manifest in the formation of a different conclusion-attitude. It might seem that some premise-attitudes meet this condition too. Take an episode of instrumental reasoning from the premise-intention to do A and the premise-belief that doing B is a necessary means for doing A to the conclusion-intention to do B. Would replacing the premise-belief with the belief that doing C is a necessary means for doing A not result in reasoning to the conclusion-intention to do C, for example? Only if we also change the normative belief to reflect the change in the starting point. Recall that the reason judgement required by NJC is specific to the premise- and conclusion-attitudes of the particular episode of reasoning

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

50     in question. Two instances of instrumental reasoning that differ in their premise-beliefs thus necessarily involve different reason judgements too. Of course, in such cases, we can expect a reasoner who is familiar with that kind of reasoning to have both reason judgements, and hence to be ready to engage in both episodes of reasoning. But the two episodes differ not just in one premise-attitude, but also in the reason judgement they involve. The two episodes of reasoning in the umbrella vs raincoat example, on the other hand, differ only in one stimulus-attitude: the reason judgement. The Guidance condition thus allows us to distinguish, among the stimulus-attitudes of an episode of reasoning, between its premise-attitudes and the reason judgement that the Normative Judgement Condition requires. The latter is the stimulus-attitude that meets the condition, whereas premise-attitudes are stimulus-attitudes that do not meet the condition. They provide the starting point of the reasoning process, whereas the reason judgement guides it to a particular conclusion-attitude. Both contribute to the manifestation of the reasoning disposition, but they do so in different ways.

2.3.3 The account’s adequacy According to the Normative Guidance Account, an episode of practical reasoning necessarily involves a reason judgement that guides the reasoning process. Why accept this account? My case for it comes in two main parts. First, I will show that the account easily satisfies the conditions of adequacy that were introduced in section 2.2, and that all alternative accounts struggle with at least one of them. This will be the tasks for the present section and section 2.3.4, respectively. Secondly, I will address various objections and concerns that have been raised for similar accounts and argue that none of them poses a threat to the Normative Guidance Account as it has been developed here. This will be done in sections 2.5–2.7. The first condition of adequacy for an account of practical reasoning is to accommodate the possibility of incorrect reasoning. The Normative Guidance Account can do so easily, because it does not require the reason judgement that guides the reasoning to be true. Thus, it can

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

   

51

allow for a subject to reason from a set of premise-attitudes to the intention to perform a certain action even though the former do not provide any normative support for the latter. That means that the reason judgement required by NJC will be false, but, of course, reasoners can be mistaken about their reasons. Nothing in the account of guidance presented in section 2.3.2 requires the guiding attitude to be true either. The second condition of adequacy requires an account of practical reasoning to allow us to distinguish episodes of reasoning from other mental processes, particularly ones that lead from the same initial set of attitudes to the formation of the same new attitude. This will turn out to be particularly challenging for some accounts in cases where the transition does not conform to the standards of correct reasoning, as in the example of the pilot of the military plane. The Normative Guidance Account, however, provides us with a criterion for distinguishing the two cases. To identify the case of reasoning, we need to ask whether the pilot’s process of attitude formation is one in which a reason judgement plays the characteristic guiding role that it plays in reasoning. More precisely, the process that leads to the formation of the new attitude qualifies as reasoning only if it is the manifestation of the reasoning disposition, that is, the disposition whose stimulus-attitudes include a reason judgement whose role meets the Guidance condition. Thus, the pilot only qualifies as reasoning (incorrectly) from the intention to bomb the ammunition factory and the belief that this will lead to destroying the palace to the intention to destroy the dictator’s palace, if she judges the former to provide a reason for the latter and if the process by which she forms the new intention is guided by that judgement. And, indeed, asking whether she forms the new intention because she takes it to be normatively supported by those other considerations seems to be an independently plausible way of distinguishing the case of reasoning from the case of emotion-driven attitude formation. Finally, according to the third condition of adequacy, an account of practical reasoning must be able to explain why reasoners cannot take a stance of disapproval towards an episode of reasoning while they are engaging in it. It should explain why it is ‘impossible or seriously

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

52     irrational’ to make an inference and ‘to judge, at the same time, that [it] is not a good inference’, as Hlobil (2014, 420) puts it. The Normative Guidance Account provides a straightforward explanation of this. It is irrational or even impossible to engage in an episode of reasoning while judging that there are no reasons supporting its conclusion, because that judgement is inconsistent with the reason judgement required by NJC. On the assumption that in a bad inference, the conclusion is not normatively supported by the considerations from which it is inferred, the Normative Guidance Account thus entails that we cannot consistently judge an inference to be bad at the moment we draw it, as Hlobil points out.

2.4 The inadequacy of alternative accounts Alternative accounts of practical reasoning, those that deny NJC, do significantly less well in meeting our three criteria of adequacy. In this section, I will discuss various alternative accounts—most of which have been defended in the recent literature—and argue that each of them fails with respect to at least one of the criteria.

2.4.1 Pure dispositionalism The first alternative account agrees with the Normative Guidance Account that reasoning is best understood as the manifestation of a certain disposition on the part of the reasoner. It denies, however, that reason judgements play any role in that process. According to what I will call pure dispositionalism, the disposition’s stimulus-attitudes comprise solely the premise-attitudes, and the manifestation consists solely in the formation of the conclusion-attitude of the reasoning. To evaluate this account, we first need to fill in a number of details. Dispositions in general are typically characterized in terms of a conditional that connects a manifestation to a stimulus. For example, a soluble sugar cube will dissolve when immersed in water. Here, the disposition to be characterized is (water) solubility, the manifestation is dissolving, and

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

    

53

the stimulus is being immersed in water.²¹ However, not all dispositional properties can be characterized by a single conditional. Take electrical charge: an electron with a specific charge will experience a certain force F1 when placed at a distance D1 to another charge C1, a force F2 when placed at a distance D2 to another charge C2, and so on. In this case, the disposition must be characterized by a series of conditionals that connect different manifestations to different stimuli in a systematic way. Such complex dispositions, which cannot be adequately characterized by a single conditional, are usually called multi-track dispositions (Vetter 2013, 334; cf. Bird 2007, 21). If practical reasoning can be characterized in terms of a disposition, it must be a multi-track disposition, since reasoning will lead to different manifestations (i.e. conclusion-attitudes) in the presence of different stimuli (i.e. premise-attitudes). These multitrack dispositions can be understood as bundles of simpler dispositions that correspond to each of the conditionals used to characterize them.²² Crucially, the pure dispositional account of practical reasoning needs to specify which particular (multi-track) disposition is manifested in such reasoning. Not just any disposition to form certain attitudes in the presence of certain other attitudes will suffice for a mental process to qualify as reasoning. A plausible way to identify the relevant disposition is to require that its modal profile (i.e. which attitudes it would produce if which other attitudes were present) must accord sufficiently well with the standards of correct practical reasoning. The accordance must be sufficiently good, but it need not be perfect. Otherwise, the account would exclude the possibility of incorrect reasoning and thus violate the first criterion of adequacy. This, however, makes it difficult for the account to meet the second criterion of adequacy, that is, to distinguish cases of incorrect reasoning from other mental processes. A mental transition from certain attitudes to the formation of a new attitude could be either an instance of incorrect reasoning or of some other process. Recall the example of the pilot of the military plane from section 2.2. Her transition from the intention to ²¹ The conditional analysis of dispositions has not remained unchallenged. For an overview of different versions of the analysis and the problems they face, see Choi and Fara (2018, § 1). ²² For the idea that capacities (or abilities) can be understood as bundles of dispositions, see Vihvelin (2004, 431).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

54     bomb the ammunition factory and the belief that doing so will also destroy the dictator’s palace to the intention to destroy the palace might be an instance of a purely emotion-driven process, or an instance of incorrect practical reasoning. How can the pure dispositional account distinguish between these two cases? What condition is violated in the first case but satisfied in the second, so that the latter qualifies as a case of reasoning, but the former does not? For a particular transition to qualify as an episode of reasoning, the pure dispositional account requires it to be (i) the manifestation of a (simple) disposition to form an intention to destroy the dictator’s palace in the presence of the intention to bomb an ammunition factory and the belief that doing so will also destroy the palace, which (ii) is part of the bundle of dispositions that makes up the complex, multi-track disposition to reason. But in the non-reasoning case, the transition might be the manifestation of a disposition too. Surely reasoning is not the only mental process that involves the manifestation of a disposition, and given her enduring hatred of the dictator, the pilot might well be reliably disposed to form an intention to destroy his palace whenever she becomes aware of an opportunity to do so. Moreover, we can assume that in both cases the pilot (also) has a multi-track disposition that accords sufficiently well with the standards of correct reasoning, that is, even in the non-reasoning case, the pilot is generally able to engage in reasoning—she has the capacity to reason, even though she does not exercise it in this particular moment. So, the crucial issue for the pure dispositional account is whether the particular disposition that is manifested by the pilot’s transition is part of the bundle of dispositions that constitutes the pilot’s reasoning capacity. If it is, then the transition is an instance of (albeit incorrect) reasoning; if it is not, then the transition is an instance of some other process. But how do we decide whether some simple disposition is part of a particular multi-track disposition? What criterion holds that bundle of dispositions together? Neither of the ways this question is answered elsewhere in the literature on multi-track dispositions is feasible in the case of the pure dispositional account. Sometimes, the bundle of dispositions is unified by a scientific law. In the case of electrical charge, for example, the way the various conditionals connect quantitatively

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

    

55

different stimuli to quantitatively different manifestations can be summarized by Coulomb’s Law. In the case of reasoning, there are certain regularities as well, which correspond to the standards of correct reasoning. But unlike scientific laws like Coulomb’s Law, these rules can be violated. Not all reasoning is correct reasoning, and the pilot example is one of the exceptions. Therefore, requiring all members of the bundle to meet the standards of correct reasoning would not help advocates of the pure dispositional account. In other cases, the various parts of a multi-track disposition are taken to be unified because they intuitively belong to the same phenomenon. Some authors argue that fragility must be understood as a multi-track disposition too, because a fragile vase will shatter when struck, but it will also crack when twisted, for example—more than one conditional is needed to fully characterize the property (Bird 2007, 21; Vetter 2013, 335). Here, the simple dispositions that correspond to the two conditionals are taken to be part of the fragility-bundle because they both capture one aspect of what we intuitively understand as fragility. But again, this criterion is unavailable to an advocate of the pure dispositional account, because whether or not the particular disposition that is manifested by the pilot’s mental transition belongs to the same phenomenon as the (other) members of the reasoning-bundle is precisely what is at issue. Distinguishing the reasoning from the non-reasoning scenario by pointing out that in one but not the other the manifested disposition captures some aspect of what we intuitively understand as reasoning is simply begging the question. How else might we decide whether the manifested disposition is part of the reasoning-bundle? It seems that the only way to identify another criterion is to look, as it were, into the psychological process involved in the manifestation of other, uncontroversial members and try to find some similarity there. This is already moving away from the basic idea underlying the pure dispositional account, which was meant to characterize reasoning solely in terms of the modal connections between the premise- and the conclusion-attitudes. Adding further elements to it would either turn it into the Normative Guidance Account or something akin to it, if the addition is another attitude. Alternatively, one might try

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

56     to distinguish the reasoning disposition by addition a non-attitudinal element. This is what Ralph Wedgwood does in his account of reasoning.

2.4.2 Wedgwood’s account As in the pure dispositional account, Wedgwood denies that the stimulus of the disposition that one manifests in reasoning includes a normative judgement. Instead, he argues that the normative fact itself—the fact that the premise-attitudes ‘rationalise’ the formation of the conclusionattitude, as he puts it (2006, 672)—is part of the stimulus. This commits him to defending the controversial claim that normative facts can be causally efficacious (2006, 677–82). However, that a mental disposition’s stimulus includes such a normative fact cannot be a necessary condition for its manifestation to qualify as reasoning. Facts can play a causal role and trigger a disposition only if they obtain. But the premise-attitudes of an episode of reasoning do not always rationalize its conclusion-attitude. In cases of incorrect reasoning, the premise-attitudes do not actually rationalize (or provide reason for) the conclusion-attitude. Wedgwood explicitly sets aside such cases and considers only correct reasoning (2006, 662). But this is not an innocent ‘idealizing assumption’ (2006, 662), for by doing so he significantly limits the scope of his account. Wedgwood does not deny that there are cases of incorrect reasoning; he is content with offering an account that is limited to correct reasoning (2006, 663). That means that in order to qualify as the kind of account that we are looking for in this chapter, Wedgwood’s proposal would have to be amended so as to accommodate incorrect reasoning. But however that is done, it will turn out that correct and incorrect reasoning involve causal processes with a significantly different structure: an important causal factor in the formation of the conclusionattitude in the case of correct reasoning will be absent in the case of incorrect reasoning. Considering how similar correct and incorrect reasoning otherwise appear to be, that certainly detracts from the account’s overall plausibility. A unified account that covers both correct and incorrect reasoning would be preferable.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

    

57

In addition, Wedgwood’s account does not meet the third condition of adequacy either, even when it is limited to correct reasoning. If the normative fact can play its causal role in the process of reasoning without being recognized by the reasoner, then a reasoner should be able not only to be ignorant of the relevant normative fact, but also to explicitly deny it. As Hlobil has pointed out, however, doing so is ‘impossible or seriously irrational’ (2014, 420). I do not see how Wedgwood’s account could explain this important aspect of reasoning, without reintroducing some condition to the effect that reasoners must be ‘cognisant of ’ the relevant normative fact, as Frege (1979, 3) has it.

2.4.3 The treating account Another alternative to the Normative Guidance Account that has been discussed in the recent literature suggests that what is distinctive of reasoning is that the reasoner treats the considerations from which he reasons as a reason for the conclusion-attitude. Of course, the Normative Guidance Account does not deny that reasoning involves treating something as a reason. But, advocates of the alternative account emphasize, treating something as a reason does not require any reason judgement; it does not even require having normative concepts (Sylvan 2015, 598). Their proposal is thus that reasoning requires only to treat something as a reason, but not any judgement about its being reason (or any other propositional attitude to that effect). What, then, does treating something as a reason consist in? One sophisticated conception of what it is to treat something as a reason has been proposed by Kurt Sylvan. He suggests that ‘to treat P like an objective normative reason to f is to be disposed to respond to the appearance that P in all or at least most of the ways that would be favored if P were an objective normative reason to f’ (2015, 601). Unfortunately, Sylvan does not tell us what those ways involve. One thing that would be favoured if P were an objective normative reason to f is, of course, f-ing. But if being disposed to f in response to the appearance that P is all there is to treating P as a reason to f, then the suggestion that reasoning involves treating the considerations from which one reasons as a reason

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

58     for the conclusion-attitude boils down to the suggestion that such reasoning involves (or manifests) a disposition to form the conclusionattitude in the presence of the premise-attitude(s). As we saw in section 2.4.1, however, that does not suffice to distinguish reasoning from other mental processes. But maybe ‘the ways that would be favored if P were an objective normative reason to f’ involve more than just f-ing. One natural thing to suggest here is that such responses include giving consideration to P in thinking about whether to f, that is, giving it a certain role in one’s reasoning concerning f -ing (cf. Lord 2018, 25). But, of course, in the present context, this suggestion begs the question. Our aim here is to identify reasoning and distinguish it from other mental processes. But if we try to identify reasoning as a mental process that involves treating something as a reason, we cannot then suggest that treating something as a reason involves considering it in one’s reasoning. Thus, it is not clear whether the manifestation of the disposition that is involved in treating something as a reason (‘respond to the appearance that P in all or at least most of the ways that would be favored if P were an objective normative reason to f’) can be spelled out in a way that neither begs the question in the present context nor collapses the account into a version of pure dispositionalism. Perhaps, though, the relevant disposition can be further specified in terms other than its stimulus and manifestation. Indeed, Sylvan argues that the disposition involved in treating something as a reason should be understood as a competence, more specifically, a competence ‘to conform to some pattern of thought’ (2016, 382). Such competences, he explains, are ‘a special case of dispositions to succeed in appropriate situations and when in appropriate “shape”, relative to some standard of success’ (2016, 382). In the case of treating, that standard consists in conformity to the relevant pattern. Beyond that, he takes the notion of competence as primitive. Does that help the treating account to distinguish reasoning from other mental processes? It does allow its advocates to suggest that in the example of the pilot of the military plane her mental transition is a manifestation of competence only in the reasoning case, but not in the case where the transition is driven by her emotions. But I find it difficult to assess the plausibility of that claim. After all, as we saw in section 2.4.1,

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

    

59

the emotion-driven transition might also be the manifestation of a disposition, and if the emotion-driven pilot were to encounter the same or similar situations repeatedly while in the same ‘shape’, then her thoughts regarding the ammunition factory and the dictator’s palace are likely to conform to the same pattern of thought. So why not say that the emotion-driven transition is a manifestation of competence too? Maybe the answer is that, in this case, the disposition is not of the relevant ‘special kind’. But that answer strikes me as not satisfactory, especially if providing any further specification of this kind of disposition is refused with reference to the primitiveness of the notion of competence. Moreover, even if employing an unanalysed notion of competence were to allow advocates of the treating account of reasoning to meet the second criterion of adequacy and distinguish reasoning from other mental processes, the account still falls short when it comes to the third criterion. Why is it that reasoners must not disapprove of their reasoning, such that making an inference while judging it to not be a good inference is ‘impossible or seriously irrational’ (Hlobil 2014, 420). Insisting that reasoning is the manifestation of a competence does not suffice to answer that question. In general, agents need not endorse their behaviour as a good exercise of a competence in order to count as manifesting that competence. I can manifest my competence to drive a car while acknowledging that I drive really badly today, and I can manifest my competence to play tennis even though, while serving, I think that this is a really bad serve. Hence, the treating account does not meet all of the criteria for an adequate account of practical reasoning.

2.4.4 Reasoning and rule-following Some authors take a different and somewhat indirect approach to giving an account of practical reasoning (or reasoning in general). They propose understanding reasoning as a rule-following activity and provide a more general account of rule-following in order to illuminate the nature of reasoning. The Normative Guidance Account as such is not inconsistent with the claim that reasoning is a kind of rule-following, but it is not

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

60     committed to that claim either. However, the two most prominent accounts of reasoning as rule-following offer accounts of the latter that do render their conception of reasoning incompatible with the Normative Guidance Account. The first of these accounts is proposed by John Broome. He characterizes ‘active reasoning’, the conscious, person-level reasoning that I am also interested in in this chapter, as follows: Active reasoning is a particular sort of process by which conscious premise-attitudes cause you to acquire a conclusion-attitude. The process is that you operate on the contents of your premise-attitudes following a rule, to construct the conclusion, which is the content of a new attitude of yours that you acquire in the process. Briefly: reasoning is a rule-governed operation on the contents of your conscious attitudes. (2013, 234)

What is distinctive of reasoning is thus that, in reasoning, you follow a rule that governs your activity. As is well known, the notion of following a rule has gained some notoriety in philosophy. Some authors, most notably Saul Kripke (1982), have argued that the notion is fraught with fundamental problems and can ultimately not be made sense of. Broome is more optimistic. He argues that to follow a rule is to be disposed to behave in a way that conforms to the rule and for that behaviour to seem right to you relative to that rule (2013, 237; 2014, 625). That is how the rule guides you: it provides the standard relative to which your behaviour seems right to you. What does it mean that your behaviour seems right to you relative to a certain rule? One straightforward way to understand this would be for the subject to have a representation of the rule in mind (as the content of some propositional attitude) and to evaluate the behaviour in the light of that representation. Some of what Broome writes about rule-following suggests such a picture, for example his claim that following a rule ‘involves intending to follow a particular rule’ (2016, 3438–9) or his explanation of the kind of guidance a rule provides in terms of what he calls ‘intentional guidance’: ‘When you follow a rule, you . . . are always

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

    

61

intentionally guided; you follow a rule because you intend to comply with it’ (2014, 631). However, Broome denies that rule-following (and reasoning in particular) requires an explicit representation of the rule. What he—perhaps somewhat idiosyncratically—means by ‘intention’ is just ‘a disposition to behave in a particular way and for this to seem right’ (2014, 632). According to Broome, then, seeming right does not involve a comparison to an explicitly represented standard of rightness or correctness, nor is it necessarily associated with a phenomenal state (2013, 238). Instead, it is simply another disposition, a disposition ‘to think there is a right answer, and to think you have it’ (2016, 3437). In particular, it involves openness to the possibility of correction, that is, the possibility that the answer no longer seems right to you after ‘checking’ (2016, 3437; see also 2013, 238). But, of course, what seems right need not be right. So what provides the relevant standard of correctness, if not some explicit representation of the rule? According to Broome, ‘for a process to be right is for you to have a steady disposition for it to seem right’ (2013, 239). This steady disposition ‘encodes the rule you intend to follow’, even though it does not explicitly represent it (2016, 3439). The idea that being right can be explicated in terms of stably seeming right even though the latter has no distinctive phenomenal character has been met with criticism (cf. Cullity 2016), and I am puzzled by that aspect of Broome’s account too. But I would like to set that aside and focus on two other problems of the account. Firstly, it seems to me that the account of rule-following that Broome provides is quite thin. That might make it less prone to certain objections that have been raised against more demanding accounts. But it also makes me question whether it provides enough substance to distinguish reasoning from other mental processes by characterizing it as a rule-following activity. Consider a person whose ‘gut feeling’ tells her that black men are dangerous. When she comes to believe that a man approaching her on the pavement is dangerous, she might well be open to the possibility that she is wrong about this particular man (that is what she would say if you asked her). Indeed, she might even check her initial reaction by taking a second look and seeing what her gut leads her to believe this time. Suppose, though, that this yields the same result, and suppose further

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

62     that this is how she responds in all or nearly all cases of encountering black men. Does that mean that this person is following a rule in classifying those men as dangerous? After all, she has a steady disposition to come to believe that they are dangerous upon encountering them, and she is also open to the possibility that she might be wrong. Clearly, though, that person is not engaging in reasoning. Her beliefs are the result of an associative process that is driven by an unconscious racial bias. But it seems that the process that takes her from the belief that the man in front of her is black to the belief that this man is dangerous meets Broome’s characterization of a rule-following activity. (It also meets his other conditions for reasoning: the formation of the second belief is caused by the first one, and, being an associative process, it operates on the contents of the attitudes in question.) So, either characterizing reasoning as a rule-following activity does not suffice to distinguish reasoning from other mental processes—in that case, Broome’s account of reasoning would fail to meet the second criterion of adequacy—or a process’s qualifying as a rule-following activity requires more than the process’s seeming right, on Broome’s dispositional understanding of it. In that case, his characterization of rule-following— and hence his account of reasoning—is incomplete. A second problem of Broome’s account of reasoning concerns its ability to meet the third criterion of adequacy. In general, that x seems F to me does not commit me to x’s being F: it is not inconsistent with believing that x is not F. For example, that a stick partly immersed in water seems bent to me does not commit me to its being bent. Believing at the same time that the stick is straight does not make me irrational. Similarly, that it seems to me that there are more integers than there are odd integers does not commit me to there being more integers than odd integers. It is perfectly rational for me to believe, at the same time, that the set of all integers and the set of all odd integers have the same cardinality.²³ But if seemings do not generate a commitment, then taking reasoning to be a rule-following activity as Broome characterizes it does not suffice to explain why it is irrational to engage in an episode of reasoning while judging that it is bad or incorrect reasoning (i.e. reason²³ Hlobil (2014, 427) makes a similar point.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

    

63

ing that fails to be right). Thus, Broome’s proposal to understand reasoning as a rule-following activity falls short of meeting our criteria of adequacy for an account of practical reasoning. Paul Boghossian (2014) also thinks that the best approach to understanding the nature of reasoning is to characterize it as a rule-following activity. This, he suggests, allows us to explain why reasoning is subject to the ‘Taking Condition’, that is, why reasoning necessarily ‘involves the thinker taking his premises to support his conclusion’ (2014, 5). This taking, however, must not be understood as a belief about a relation of normative support between the former and the latter. Boghossian rejects such a doxastic understanding of the Taking Condition for fear that it would require too much conceptual sophistication of ordinary reasoners (2014, 6–7). Like Broome, Boghossian thus denies that reasoning is subject to the Normative Judgement Condition. He suggests that characterizing reasoning as a rule-following activity allows us to ‘effortlessly accommodate the Taking Condition’ (2014, 11), because following a rule essentially involves treating something as a reason for a certain response (2014, 12). Unfortunately, Boghossian tells us not much more about what is involved in following a rule. He does have something to say about what rule-following does not involve. He rejects what he calls the intentional view of rule-following, according to which such activity can be explained by the agent’s being aware of what a rule requires via some ‘intentional state that represents the rule’ and her applying it to the particular case at hand (2014, 13). The problem is that, according to this view, rule-following involves an inference (i.e. an episode of reasoning), which would set us off on a vicious regress (Boghossian 2014, 14; 2008). This problem is avoided by a view of rule-following that explicates it solely in terms of an agent’s dispositions to conform to the rule under suitable circumstances. But Boghossian (2014, 14–15) rejects this ‘dispositional view’ as well, because such dispositions are to be explained by, rather than identical to, our following the rule in question, and because it fails to accommodate the Taking Condition. Boghossian’s (tentative) positive proposal is to take the notion of rulefollowing as an ‘unanalyzable primitive’ (2014, 17). But then, how much of an account of reasoning does his proposal to understand it as a rule-

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

64     following activity actually provide? Boghossian (2014, 11) claims that this proposal allows us to ‘effortlessly accommodate the Taking Condition’. Yet, in the end, it does so only in so far as it is intuitively plausible—‘seems right to say’, as Boghossian (2014, 12) puts it—that something much akin to the Taking Condition is true of rule-following. But that the mental activity in question is subject to something like the Taking Condition is intuitively plausible anyway, whether or not we characterize it as rule-following. Part of what we are looking for in an account of reasoning is an explanation of why the activity is subject to such a condition. Boghossian’s proposal does not offer an informative answer to that question. Moreover, as Hlobil (2014, 425) points out, it is not clear whether the understanding of the Taking Condition that Boghossian ends up with is robust enough to accommodate our third condition of adequacy. After all, Boghossian’s reasons for rejecting a doxastic interpretation of the Taking Condition—in particular his worries about conceptual sophistication—commit him to denying that ‘taking’ amounts to a propositional attitude at all. That makes it difficult to see how the irrationality of making an inference and judging, at the same time, that it is not a good inference could be explained by reference to a consistency constraint that is violated by the ‘taking’ involved in the reasoning and the judgement that it is not good reasoning, because such constraints appear to cover only attitudes with propositional content. This concludes our discussion of accounts of reasoning that, in one way or another, offer an alternative to the Normative Guidance Account advocated in this chapter. We saw that each of the accounts fell short with respect to at least one of the criteria of adequacy that we established for an account of practical reasoning. Of course, all this shows is that none of the alternative accounts shares all of the advantages of the Normative Guidance Account. That does not suffice to establish that the latter is preferable overall, because it might also have disadvantages that the others lack. Indeed, many authors who discuss the Normative Guidance Account (or at least NJC) and end up rejecting it do so because they think it succumbs to a decisive objection, despite its various advantages. The two most influential objections were already touched upon in the previous sections: they concern the level of conceptual sophistication that an account that endorses NJC requires of reasoners, and the worry

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

   

65

that such an account leads into a vicious regress. In sections 2.5 and 2.6, I will discuss these objections and argue that neither of them poses a significant problem for the Normative Guidance Account.

2.5 Reasoning and conceptual sophistication The first objection that I will discuss is the sophistication objection. It is based on the worry that, by taking a mental state with normative content to be essential for reasoning, accounts that endorse NJC require too much conceptual sophistication and thus exclude certain reasoners. The objection is usually advanced by presenting examples of reasoners who (allegedly) lack the conceptual resources to entertain the normative judgements required by NJC. Consider the following scenario, which is based on an example Boghossian repeatedly uses (2001, 25, 2014, 6): Two small children, Lisa and Laura, play hide-and-seek. Lisa sees Laura’s bicycle leaning against the tree. She also believes both that if Laura were hiding behind that tree, she would not have left her bicycle there, and that she must be hiding either behind the tree or behind the hedge. As a result, Lisa forms the belief that Laura must be hiding behind the hedge. Back at Laura’s home, Lisa is thirsty for some juice. She remembers that Laura’s mother keeps the juice in the refrigerator. As a result, Lisa forms the intention to open the refrigerator.

Proponents of the objection take these to be perfectly ordinary cases of reasoning. Lisa first engages in an episode of theoretical (more specifically, of disjunctive) reasoning, and then in an episode of practical (specifically, of instrumental) reasoning. But, they suggest, children like Lisa do not yet have sophisticated concepts like that of a reason or of normative support. Without possessing such normative concepts, however, they are not able to entertain the kinds of judgement that are required for reasoning according to NJC. Therefore, they conclude, NJC must be false. It requires too much conceptual sophistication from

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

66     reasoners such as Lisa. To put it more explicitly, proponents of the objection argue as follows: The Sophistication Objection (P1) There are individuals who do not have the conceptual resources to entertain the normative judgements required by NJC. (P2) Some of those individuals can nevertheless engage in reasoning. (C) Therefore, NJC is false. Versions of this objection have been raised by a number of authors.²⁴ As I will show, however, it is highly questionable whether Lisa or any other person satisfies both of the objection’s premises. Why think that Lisa does not have the conceptual resources to entertain the normative judgements required by NJC? Firstly, we should be careful not to put too much argumentative weight on a particular terminology. Describing a person as thinking that some consideration ‘is a reason for’ or ‘normatively supports’ an action or a belief uses philosophers’ jargon in order to ascribe a certain attitude to that person. A more common (though perhaps less precise) way to ascribe the same attitude would be to say that she thinks that the consideration ‘counts in favour of ’ the belief or action (cf. Scanlon 1998, 17). Someone can have that attitude without being able to express it using particular terms like ‘reason’. Hence, the plausible assumption that Lisa is not familiar with such terms does not suffice to establish that she lacks the ability to entertain those attitudes.²⁵ More importantly, the attitudes in question are among the simplest and most primitive normative attitudes. It is difficult to see how someone who lacks the necessary resources to entertain them could engage in any normative thought at all. In any case, the lack of those resources would have serious consequences that proponents of the objection seem to ignore. A basic competence with the normative concepts in question is ²⁴ See, e.g., Winters (1983, 209), Audi (1993, 241), Wedgwood (2006, 675), Broome (2013, 229), Boghossian (2014, 6–7), and McHugh and Way (2016a, 317). ²⁵ The same is true for the ‘members of primitive cultures’ who feature in Winters’ other example (Winters 1983, 209).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

   

67

a precondition of some of our most fundamental agential capacities. In particular, it is essential to participating in the practice of justification. If you lack the resources to entertain the normative judgements required by NJC, you are not able to justify your beliefs or actions. This is because to justify a belief or an action simply is to give what you consider to be a reason for it. Anyone who is able to answer requests for justification such as ‘Why do you think that p?’ or ‘Why did you do A?’ thinks that something counts in favour of believing that p or doing A. They do so even if they cannot express this by using the term ‘reason’, or indeed in any other way than by answering that question. It thus takes very little to prove that you have the conceptual resources to think that some considerations are reasons for (or count in favour of) certain beliefs or actions: being competent with the justificatory use of ‘because’ in answering such ‘why’ questions is sufficient. Conversely, without those resources you would not be able to answer such a request for justification. In fact, you would not even be able to understand it, for that would involve understanding that you are requested to adduce something that you take to be—and that your interlocutor can accept as—a reason for what is to be justified. Now imagine Lisa being asked why she thinks that Laura is not hiding behind the tree, or why she opened the refrigerator. Boghossian does not tell us how old the children in his example are, but I assume the children imagined by most readers will be able to respond to those questions. So suppose that Lisa answers ‘Because then she would not have left her bicycle there’ or ‘Because the juice is in there’. Her ability to understand and respond to those requests for justification, then, shows that she actually has the conceptual resources necessary to entertain the normative judgements required by NJC. Hence, if Lisa is able to respond to a request for justification, as most readers will have taken her to be when imagining the scenario, the first premise of the sophistication objection does not apply to her.²⁶

²⁶ Note that her ability to understand and answer such requests also shows that she has at least a rudimentary understanding of what it is to believe and do something, i.e. of the kind of things that NJC requires reasoners to judge themselves to have reasons for.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

68     Of course, proponents of the objection can simply stipulate that their example involves children who are unable to understand or comply with requests for justification. Such children might well satisfy the objection’s first premise. But it is hardly obvious that the cognitive processes by which they form the new belief or intention are instances of genuine reasoning. Recall that what we are concerned with here is conscious, active reasoning, not the automatic, subconscious processes of attitude formation that are sometimes called ‘reasoning’ when this is understood in a much wider sense. Advocates of NJC can and should agree that these latter processes do not require the possession of normative concepts, as they might well occur in some non-human animals too. But NJC is a claim about a distinctive and more advanced process that differs from those automatic processes. Proponents of the sophistication objection like Broome and Boghossian do not dispute this either. As we saw in section 2.1, they are concerned with active, person-level reasoning too. The objection’s second premise, which claims that the individual in question engages in reasoning of the relevant kind, is thus a substantive claim that requires support. Simply stipulating that some children who do not yet possess any normative concepts are nevertheless able to engage in this kind of reasoning would beg the question against NJC. Coming up with a more plausible example, however, is much more difficult than advocates of the objection seem to think. Empirical research in developmental psychology shows that the ability to differentiate between moral and conventional rules develops between the ages of 26 and 42 months, and that this development is correlated with the development of language comprehension (Smetana 1981; Smetana and Braeges 1990; cf. Turiel 2007) This is tested, in part, by asking the children how bad a rule transgression is, and whether the transgression would be permissible if no authority (such as a teacher) saw it. It is reasonable to assume that children who are able to understand and answer such questions are beginning to develop a basic competence with the normative concepts under consideration, at least for the practical sphere. That research thus suggests that the mental development of the kind of individual described by the objection’s first premise does not exceed that of a typical 3-year-old. NJC commits us to denying such individuals the ability to engage in genuine reasoning. But this is hardly a

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

   

69

disadvantage, unless there is a compelling reason to think that they have this ability. It is important to note that this is not just a quarrel about particular cases. Its proponents advance the sophistication objection by presenting (often in not too many words) a putative counterexample to NJC. This warrants NJC’s rejection only if it is either obvious that the example validates both of the objection’s premises, or if its proponents bring forward some positive argument that it does. Those who have pressed the objection in the recent literature appear confident that the cases they describe are obvious counterexamples. What we have seen so far is not just that their cases are ill-chosen, but that once it has been made clear what has to be true of an individual to qualify as a candidate for satisfying both premises, it is highly questionable whether there are any cases that obviously constitute a counterexample. Such an individual would have to be incapable of even understanding requests for justification, and her cognitive abilities would have to be comparable to those of a typical 3-year-old. I doubt that it will be obvious that any such individual has engaged in the conscious, active, person-level reasoning that both advocates and opponents of NJC are interested in. Thus, proponents of the sophistication objection have to argue for the claim that a person whose cognitive abilities are sufficiently undeveloped to satisfy the first premise nevertheless satisfies the second one. To do so without begging the question against NJC, they must offer some independently plausible criterion for identifying reasoners. But not any such criterion will do. Having the ability to justify one’s beliefs or actions, for example, might well suffice to qualify as a reasoner. But that criterion is not available to proponents of the objection, because meeting it contradicts their first premise. To suit their purposes, the criterion must require enough sophistication to exclude those who can form attitudes solely via automatic, subconscious processes, but not so much sophistication that only those who have some grasp of basic normative concepts can meet it. That is a difficult balance to strike, and I am sceptical that it can be accomplished. Since I have not shown that it cannot be done, however, this does not establish that it is impossible to formulate a compelling version of the sophistication objection. The burden of proof is on its proponents,

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

70     though. Until they come up with such a criterion and present a case that does validate both premises of their argument, those who are drawn to NJC by its benefits need not be discouraged by the fact that it requires reasoners to possess basic normative concepts. But shifting the burden of proof concerning the objection back unto their opponents is not all that advocates of the Normative Guidance Account of reasoning can do to rebut it. I think that the objection appears compelling to many because it seems be supported by a certain observation about the phenomenology of reasoning. Upon closer inspection, however, it turns out that this observation is perfectly compatible with the Normative Guidance Account. By showing this, we can undermine much of the motivation for raising the sophistication objection in the first place. Here is the observation I have in mind: reasoners are often not conscious of any normative judgements. Normative thoughts do not always cross our minds during reasoning. This explains why NJC might not seem particularly plausible as long as we focus on the phenomenology of reasoning as we experience it first-personally. Broome, for example, approaches the subject from this perspective (cf. 2013, 223 et passim). Wedgwood too rejects NJC partly for this reason: [M]ost of us are engaged in simple reasoning for much of our waking lives; but we rarely spend much time thinking about our own mental states and forming higher-order beliefs about whether those mental states rationalize certain new beliefs or intentions. (2006, 675)

This, he argues, poses a serious problem for any view that requires the reasoning process to involve a normative judgement in support of its conclusion, a problem that he takes the sophistication objection, which he raises immediately after the quoted passage, to get to the essence of (2006, 674–5). But, in fact, there is no such problem. That reasoners are not aware of any normative judgements does not entail that they do not have any. Those judgements might be implicit ones that remain unconscious and in the background of the reasoning process. This is nothing unusual. We often attribute implicit attitudes and grant them a role in explaining a subject’s other beliefs or actions. For example, some people

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

   

71

believe that women are generally less trustworthy than men without being conscious of this belief. Such sexist and other biased beliefs are often implicit and remain invisible from the person’s own perspective. Nevertheless, they can play an important role in explaining, for example, the subject’s hiring behaviour or his assessment of an employee’s performance. Similarly, an implicit normative judgement could play an important role in explaining why a reasoner forms a particular attitude while remaining invisible from the reasoner’s first-person perspective. However, while Wedgwood’s observation is not inconsistent with NJC, it does pose a challenge for its proponents. They must give some account of the role the required normative judgements play in the process of reasoning, and that account must not require those judgements to be conscious and in the front of an agent’s mind during reasoning. Doing so would make their account vulnerable to a different kind of sophistication objection, that of over-intellectualizing ordinary reasoning by requiring it to be reflective in a way that our everyday experience tells us it often is not. Avoiding this is particularly challenging because, as I have emphasized, we are concerned with active, personlevel reasoning that is itself a conscious mental process. Moreover, as we saw in section 2.1, that process operates on the contents of the premise- and conclusion-attitudes, which means that they have to be conscious too. Thus, Wedgwood’s observation would pose a problem for advocates of NJC if the only role that a normative judgement can play in the process of reasoning was that of a premise- or conclusionattitude. But, as I argued in section 2.3.2, that is not the case. The reason judgements required by NJC play a different role: they guide the reasoning process. Hence, the Normative Guidance Account does not over-intellectualize ordinary reasoning. There is no reason to suspect that reasoners are necessarily aware of the attitudes that guide their reasoning. Reasoning operates on the contents of the premise- and conclusion-attitudes, which is why those attitudes are necessarily conscious to the reasoner. The guiding attitude merely provides the instructions for that operation, which it can do while remaining in the back of our mind. We often follow instructions without being consciously aware of them, for example, when we form grammatically correct sentences. The Normative Guidance

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

72     Account can thus accommodate the fact that normative thoughts do not always cross our mind during reasoning.

2.6 Regress worries Another objection that is often raised against accounts that accept NJC is that this requirement leads into a vicious regress. This objection has been discussed more prominently in the debate about the nature of theoretical reasoning or inference, but it is just as relevant in the case of practical reasoning.²⁷ I will address this objection by distinguishing three versions of it and arguing that none of them succeeds in refuting the account presented here. The first version of the regress objection is often illustrated with Lewis Carroll’s (1895) famous story, in which the Tortoise challenges Achilles to what might be called a race of reasoning.²⁸ The conclusion Achilles is challenged to reach by way of reasoning is this one: (Z) The two sides of this Triangle are equal to each other. His starting point is the following two premises: (A) Things that are equal to the same are equal to each other. (B) The two sides of this Triangle are things that are equal to the same. Accepting these two claims, however, is not sufficient, the Tortoise claims: Achilles can only reach the conclusion Z from his beliefs in A and B if he also accepts that ‘if A and B are true, Z must be true’ (1895, 279). This prompts Achilles to add another claim to his list of premises: (C) If A and B are true, Z must be true.

²⁷ Regress worries are discussed, e.g., by Boghossian (2001; 2003a; 2014), Railton (2004), Wedgwood (2006), Wright (2014), and Valaris (2014). ²⁸ Carroll’s story is concerned with theoretical reasoning, but a similar challenge can be formulated for the case of reasoning to an intention.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

 

73

This step is then repeated with ‘If A and B and C are true, Z must be true’, and so on. Those who think that a similar regress looms if we accept NJC suspect that, like the Tortoise, the condition forces us to add more and more premise-attitudes to any episode of reasoning (cf. Railton 2004, 184–5; Valaris 2014, 107). They think that if, in order to qualify as reasoning, a mental process that takes you from a given set of premise-attitudes to the formation of a conclusion-attitude must involve a reason judgement that connects the former to the latter, we get a new potential episode of reasoning, whose premise-attitudes now include the premise-attitudes of the initial episode plus the reason judgement. This new episode of reasoning would then also require a new reason judgement, one that connects the extended set of premise-attitudes to the conclusionattitude, and so on. But, of course, this regress only gets started if we add that the required reason judgement is another premise-attitude. However, according to the Normative Guidance Account, the judgement required by NJC must not be understood as another premise-attitude—it plays the guiding role in the episode of reasoning. Hence, this version of the regress objection does not get off the ground for the account of reasoning advocated in this chapter. The second version of the objection is presented especially clearly by Ralph Wedgwood. Wedgwood considers a view that requires every episode of reasoning to involve a belief to the effect that the premiseattitudes ‘rationalize’ the conclusion-attitude, but the objection could just as easily be formulated in terms of reason judgements. Wedgwood argues that the required normative belief: would itself appear to be irrational unless those beliefs are themselves formed through reasoning. But rational reasoning cannot essentially depend on irrational beliefs. So the account that this [view] gives of rational reasoning suffers from a vicious regress: to form any belief through rational reasoning, one would first have to do a further piece of rational reasoning, to form the belief that one is in mental states that rationalize forming that first belief; and so on ad infinitum. (2006, 675)

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

74     Here is a more explicit formulation of the objection: 1. Premise: Rational reasoning cannot essentially depend on irrational beliefs. 2. Premise: The required normative belief is irrational unless it is itself formed through rational reasoning. Conclusion: If reasoning necessarily involves a normative belief, then every episode of rational reasoning presupposes an infinite series of episodes of rational reasoning to be completed beforehand. By ‘rational reasoning’, Wedgwood (2006, 662) means reasoning that is not fallacious, reasoning that can make it rational for the reasoner to form the conclusion-attitude.²⁹ We should thus note, first, that the objection only targets the Normative Guidance Account in so far it is applied to correct reasoning. Still, if the account gives rise to a vicious regress for such reasoning, this would be a sufficient reason to reject it. However, I think that neither of the objection’s two premises is compelling. The first premise states effectively that an episode of reasoning confers warrant on its conclusion-attitude only if any beliefs that it ‘essentially depends on’ are also warranted. But while that might well be true for any premise-beliefs, it strikes me as more plausible to require that the normative judgement that guides an episode of reasoning is true in order for that reasoning to qualify as ‘rational’ in Wedgwood’s sense, but not that it is rational. Consider the widely shared idea that such rational reasoning transmits warrant from the premise- to the conclusion-attitudes. On that assumption, it makes perfect sense that the premise-attitudes have to be warranted (because otherwise there is nothing to transmit) but that the guiding attitude has to be true (because otherwise the premise-attitudes are not connected to the conclusionattitude in the right way for the transmission to work). Thus, it is neither inconsistent nor inadmissibly ad hoc for advocates of the Normative

²⁹ Boghossian (2001; 2003a) also discusses versions of the regress objection that are concerned with reasoning as a means of arriving at justified conclusions.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

 

75

Guidance Account to deny that the objection’s first premise applies to the required reason judgement. The second premise is even less compelling. Why should we accept that reasoning is the only way to arrive at a normative judgement that is warranted? This would exclude, for example, testimony and intuition as sources of warranted reason judgements, which hardly seems plausible. In fact, the claim that reasoning is the only source of warrant for normative judgements faces its own regress objection. For if we add the plausible assumption that any rational episode of reasoning that succeeds in conferring warrant on a normative judgement must also have a normative judgement among its premise-attitudes,³⁰ and accept all premise-attitudes of such an episode of rational reasoning have to be warranted,³¹ then for any normative judgement that is warranted there would have to be an infinite series of warranted normative judgements that support it. To stop that regress, we would need exactly what the objection’s second premise denies: a normative judgement that is warranted even though it has not been formed through rational reasoning. Hence, since neither of the objection’s premises is compelling on its own, and since an advocate of the Normative Guidance Account can consistently deny them, this version of the regress objection does not pose a problem for the Normative Guidance Account either. The third version of the regress objection is not concerned with justification, but with the guiding role that the reason judgement plays in the process of reasoning. The worry here is that such guidance involves its own episode of reasoning. If that is the case, then every episode of reasoning is guided by a reason judgement through another episode of reasoning, which sets off a different kind of regress in which the episodes of reasoning do not form a sequence but where each episode is nested within the previous one. But that would make reasoning impossible for creatures with finite cognitive resources like us. Crispin Wright (2014, 31–2) takes this version of the objection to apply to any account of reasoning (or ‘inference’, in his terminology) that requires an

³⁰ Which seems to be entailed by what is widely known as ‘Hume’s Law’: that no ‘ought’judgement can be correctly inferred from mere ‘is’-judgements. ³¹ As the objection’s first premise entails.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

76     episode of reasoning to be guided or ‘controlled’ by an attitude that ‘registers’ a support relation between the premise- and conclusionattitudes. He presents it in the form of a dilemma. Either the content of that attitude is general, that is, it pertains to a certain kind of reasoning, such as instrumental reasoning, of which there can be various instances that differ in the content of their premise- or conclusionattitudes, or the content of that attitude is specific to the particular episode that it guides. If the attitude’s content is general, then its guidance of the transition involved in a particular episode of reasoning will have to be understood ‘as mediated by an “appreciation” that [this transition] comes within the ambit of the [attitude]’, which amounts to an episode of reasoning on its own (Wright 2014, 31). If the attitude’s content is specific, we must be in a position to have a potentially infinite number of them, since we are in a position to engage in a potentially infinite number of different episodes of reasoning. But, Wright (2014, 32) argues, the only way to explain our ability to have all those potential guiding attitudes is to take them to be the result of reasoning from a finite set of more general attitudes. This is a forceful objection, but I do not think that it refutes the account of reasoning presented in this chapter. Since the reason judgements that guide episodes of reasoning according to the Normative Guidance Account are specific to a particular episode, the account must tackle the second horn of Wright’s dilemma. The challenge here is to explain how reasoners can form a potentially infinite multitude of specific reason judgements without deriving them by reasoning from more general principles. But I think that challenge does not pose a problem for the Normative Guidance Account. First, there are various other processes of attitude formation, such as intuition, accepting testimony, mimicking one’s peers, etc. More importantly, Wright’s notion of reasoning or inference seems to be much wider than what I am concerned with in this chapter, which is only conscious, person-level reasoning of which the reasoner is aware (section 2.1). Wright, however, seems to include any process in which ‘general information states’ are applied to particular cases (2014, 32). But such processes need not involve reasoning in my sense; they will often be automatic and subconscious. In fact, Wright himself compares the ability

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi



77

to form a potentially infinite multitude of specific guiding attitudes to our ‘ability to parse no end of novel utterances in our mother tongue’ (2014, 32). That ability might well involve the application of a finite set of general information states that constitutes our understanding of that language. But, of course, parsing any particular novel utterance need not involve a conscious episode of reasoning, or any other conscious process.³² Similarly, forming a particular reason judgement based on one’s overall appreciation of the standards of correct reasoning (to maintain the analogy to the language case) would not need to involve any conscious reasoning. So, while it is true that every reason judgement that guides a particular episode of reasoning must have been formed in a way that is consistent with our finite cognitive resources, that way need not involve any further reasoning. Instead, the judgement might be the result of various automatic and subconscious processes—which include not only those that bring states with a more general content to bear on a particular case, but also various maintenance processes that ensure coherence, etc.—or they might have been acquired by testimony. Therefore, neither the way in which a reason judgement guides a particular episode of reasoning (first horn) nor the way in which that reason judgement has been acquired (second horn) needs to involve any further reasoning. Taking practical reasoning to be guided by a reason judgement thus does not lead into any problematic kind of regress.

2.7 Metareasoning So far, we have focused on the guiding role that reason judgements play in the process of reasoning. But, of course, that is not the only role they can play in an episode of reasoning. Sometimes, we reason about what

³² For the record, I do not think that this process is adequately described as an instance of reasoning or inference even on a wide understanding of that notion either. Not every systematic cognitive activity amounts to inference or reasoning.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

78     reasons we have, and in these cases reason judgements are among the premise- and conclusion-attitudes. Here is a simple example: (1) That it ends sentient life is a reason not to kill cows for food. (Reason judgement) (2) Fish are sentient living beings too. (Belief) (3) So, that it ends sentient life is a reason not to kill fish for food too. (Reason judgement) Reasoning like this can lead us to form new reason judgements, or to drop ones we already have. It thus affects the attitudes that (potentially) guide other instances of reasoning. We can call such reasoning about our reasons metareasoning.³³ Engaging in metareasoning like (1)–(3) allows us to influence our first-order reasoning: by acquiring a new reason judgement, for instance, we extend the range of conclusion-attitudes to which we can reason from a given set of premise-attitudes. For example, acquiring (3) enables the agent to engage in the following episode of reasoning, which is guided by (3): (4) Killing fish ends sentient life. (Belief) (5) So, I shall not kill fish for food. (Intention) Thus, metareasoning can have a significant effect on what we decide to do and how we come to that decision. But note that it is not itself a form of practical reasoning; it does not conclude with an intention, but with a reason judgement, and is hence more akin to theoretical reasoning. Metareasoning is ubiquitous, and it serves a number of important functions. For example, it enables reasoners to increase coherence in their reasoning (as in (1)–(3)), and it can make reasoners aware of reason judgements they already hold implicitly, which will make it more difficult for impeding factors like haste, distractions, or a lack of concentration to keep the agent from actually reasoning in the way that corresponds to those judgements.

³³ I adopt this term from John Broome (2013, 245).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi



79

Another important function of metareasoning is that it allows agents to resolve conflicts between different reason judgements through a conscious mental process. In most choice situations we face, more than one reason judgement is relevant to our decision, and in many of those cases, the different judgements pull our reasoning in different directions. Consider the following example: you visit your grandmother, and she serves you a piece of chocolate cake. Your options are to either eat it or (politely) refuse to do so. You believe that eating the piece of cake would make your grandmother happy, and you judge this to be a reason for eating it. But you also believe that eating this piece of rich, highcholesterol buttercream cake is bad for your health, and you judge this to be a reason against eating it. Hence, your reason judgements, together with your beliefs about the piece of cake and the consequences of eating it, enable you to reason to conflicting conclusions. Often, such conflicts are resolved by psychological processes below the level of consciousness, which do not themselves qualify as the kind of active reasoning that we are concerned with here. But agents can also address them explicitly: they can bring the conflicting reason judgements to their attention and resolve the conflict by reasoning to a judgement about what they have sufficient or decisive reason to do in the particular situation. Forming such an overall judgement about their reasons will help agents to come to a decision in cases of conflict. If, in our example, you take the healthrelated reason to outweigh the happiness-related reason and conclude that you have decisive reason not to eat the piece of cake, this judgement will allow you to put the reason judgement in favour of eating it on hold, as it were, and decide not to eat the piece of cake on the basis of its impact on your health. Such conflict-resolving metareasoning that concludes with an overall judgement about one’s reasons may be called ‘weighing reasoning’, because it appears to resolve such conflicts by weighing one reason against another. But note that we do not have to understand such metareasoning as simply ‘comparing’ the ‘weight’ of one reason with that of another. This ‘kitchen scale model’ (Dancy 2004, 105) of resolving such conflicts might well be too simplistic. Moreover, the possibility of such weighing does not presuppose that reasons come with some sort of fixed weight, or that reason judgements always involve an attribution of

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

80     such weight, which then serves as an input into the process of weighing reasoning. Instead of such a bottom-up approach, in which individual reasons are assigned a certain weight that is then compared in a weighing process, one might rather take a top-down approach.³⁴ Such an approach starts from the fact that an agent would take one reason to outweigh or ‘overpower’ another reason in conflict-resolving metareasoning and understands the ‘weight’ such an agent can be taken to assign to those reasons as a numeric representation of his propensity to resolve the conflict in that way.³⁵ While metareasoning to an overall judgement about one’s reasons is often helpful in coming to a decision in cases of conflict, not all practical reasoning is backed up by such a judgement. As mentioned in section 2.5, a child can reason from the intention to drink some juice and the belief that her mother keeps the juice in the refrigerator to the intention to open the refrigerator without being able to think or talk explicitly about what is a reason for what. Such a child will hence not be able to engage in the kind of conflict-resolving metareasoning that was just discussed.³⁶ We can call first-order reasoning that is accompanied by metareasoning concerning the reason judgement that guides it critical reasoning and reasoning whose guiding attitude remains implicit and unscrutinized uncritical reasoning.³⁷ A child who is not sufficiently competent with the concept of a reason to think explicitly about reasons is then capable of uncritical but not of critical reasoning. Such uncritical reasoning provides an example of practical reasoning that is not accompanied by any judgement about what one has sufficient or decisive reason to do in a particular situation. In addition, an agent can decide to perform an action as a result of (first-order) reasoning even ³⁴ I borrow the helpful distinction between ‘top-down’ and ‘bottom-up’ approaches to the weight of reasons and reason judgements from Thomas Scanlon (2014, 8). ³⁵ Note that while conflicts between opposing reason judgements can often be resolved through weighing (meta)reasoning, it is not obvious that this is always possible. Some reasons might well be incommensurable, in which case conflicts between them could not be resolved, or at least not by way of reasoning. This possibility puts pressure on the idea that reasons necessarily come with a ‘weight’ that allows us to place them on a single scale. ³⁶ This does not mean that the child is unable to resolve conflicts between different reason judgements we can ascribe to it. As already mentioned, such conflicts can also be resolved by unconscious automatic processes that do not require thinking explicitly about one’s reasons as reasons. ³⁷ This helpful terminology too is adopted from John Broome’s recent work (2013, 235–7).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi



81

though she has come to the conclusion that she has decisive reason not to perform that action. Consider again our example: it is possible that you deliberately decide to eat the piece of cake on the ground that it makes your grandmother happy, despite having concluded that you have decisive reason not to eat it. If you do so, your reasoning to the intention to eat the piece of cake is guided by your judgement that the fact that eating it makes your grandmother happy is a reason to do so, even though you also concluded that this reason is outweighed by another reason against it. We already discussed a case of such akratic reasoning in section 2.1.2. Its possibility shows that first-order practical reasoning might be accompanied by an overall judgement that fails to support it. Together with the case of uncritical reasoning, this illustrates that practical reasoning does not require a concurring judgement about the overall balance of one’s reasons on the part of the agent. Thus, while metareasoning serves a number of important functions for an agent, it is not a necessary element of an agent’s practical thinking.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

3 The nature of reason judgements In Chapter 2, I argued for an account of practical reasoning according to which episodes of such reasoning are guided by reason judgements. This chapter builds on that account. It suggests that attending to the role that reason judgements play in the process of reasoning is essential for understanding what kind of mental state those judgements are. In particular, it provides us with the resources to formulate an account of the nature of reason judgements that takes guiding our reasoning, rather than representing certain facts, to be their characteristic cognitive function. This provides constructivists with the alternative to Representationalism that lies at the heart of their theory of practical reasons (Chapter 1.2). After some preliminary clarifications (section 3.1), I formulate and defend a non-representationalist account of the nature of reason judgements in section 3.2. The similarities and differences between that account and the conception of normative judgements underlying expressivist views in meta-ethics is discussed in section 3.3. The final section 3.4 outlines how the account presented here can be extended beyond basic first-person reason judgements.

3.1 Preliminaries This section prepares the ground for the anti-representationalist account to be presented in the following section. It outlines the functionalist approach to characterizing the nature of mental attitudes adopted by the argument, specifies the view it targets—Representationalism about reason judgements—and some of its main proponents in a bit more detail, and it clarifies the relation between rejecting Representationalism and the popular distinction of two directions of fit, to avoid potential misunderstandings. Constructing Practical Reasons. Andreas Müller, Oxford University Press (2020). © Andreas Müller. DOI: 10.1093/oso/9780198754329.001.0001

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi



83

3.1.1 The background: functionalism As I introduced those views in Chapter 1, Representationalism and its alternatives disagree about the cognitive function of reason judgements. In framing the dispute in that way, I adopt a functionalist approach to characterizing particular types of mental states. This approach is widely shared both in the philosophy of mind and in contemporary meta-ethics. Versions of it are explicitly endorsed, for example, by Allan Gibbard (1990, ch. 4), Michael Smith (1994, 113–16), Simon Blackburn (1998, 56–9), and Michael Ridge (2014, 10), and it is implicitly accepted by many more. According to functionalism, what makes something a mental state of a certain type is the role it plays in the system of which it is a part (rather than, for example, its internal constitution; see Levin 2018). Whether something is a desire or a pain depends on how that state relates to other elements of the cognitive system, as well as to sensory stimulations and behaviour. These (typically causal) relations are described by a general theory of an agent’s psychology. Particular types of mental state can thus be identified in terms of their place in such a theory. In the case of reason judgements, I think that we can gain a better understanding of the type of mental state they are by focusing on their role in the theory of practical reasoning. That is not to say that their role in this theory—which was the topic of Chapter 2—fully characterizes the nature of reason judgements. Treating something as a reason in practical reasoning is not all there is to having a certain reason judgement. But it is an essential element of their cognitive role, and it is the element that is crucial for distinguishing them from mental states that meet the representationalist’s characterization. The canonical method for constructing functionalist characterizations of particular types of mental states uses the so-called Ramsey sentence of the theory that describes their cognitive role (see Lewis 1972, Field 1978, and Levin 2018, § 4.1). This sentence is formed by conjoining all general claims that theory makes about the various types of mental states and their relations into one giant conjunction. Then, all names of different types of mental states that occur in this sentence are replaced with different variables (x, y, z, etc.), which are existentially quantified. Very schematically, the result looks something like this:

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

84      ∃x ∃y ∃z . . . (F(x, sensory stimuli) & G(x, y, behaviour) & H(x, y, z) & . . . )¹

(1)

Importantly, this sentence no longer contains any names for any types of mental states, only variables that range over mental states, but it describes the properties and relations of such states. For example, ‘S(x, sensory stimuli)’ might stand for the claim that a certain mental state (x) is reliably caused by certain sensory stimuli and ‘M(x, y, behaviour)’ might stand for the claim that this mental state (x) and a different mental state (y) together cause the subject to exhibit a certain behaviour. Such Ramsey sentences can then be used to provide a non-circular definition of particular types of mental states. For example, we might offer the following (overly simplistic and schematic) definition of what it is for a mental state to be a desire: To be a desire is to be the y such that ∃x ∃z . . . (S(x, sensory stimuli) & M(x, y, behaviour) & H(x, y, z) & . . . ). (2) This statement defines desires partly in terms of their role in the causal regulation of behaviour (M(x, y, behaviour)) and partly in terms of some other relation (H) that it bears to two other types of mental states (x, z). Chapter 2’s account of practical reasoning and the role that reason judgements play in it is a (small) part of a comprehensive theory of an agent’s psychology. Its claims will thus be a part of that theory’s Ramsey sentence, which will therefore include a variable for the kind of mental state that guides such reasoning. Of course, this variable will also occur in other parts of the sentence—as I said at the beginning of this section, playing the guiding role is not all there is to being a reason judgement. A complete functionalist characterization of the nature of reason judgements can then be given, in the same way that (2) characterizes the nature of desires, by saying that to be a reason judgement is to be the type ¹ Strictly speaking, Ramsey sentences also contain a uniqueness clause, which I omit here and in the following for reasons of brevity. The uniqueness clause for (1) would read as follows: 8x` 8y` 8z` ((F(x`, sensory stimuli) & G(x`, y`, behaviour) & H(x`, y`, z`) & . . . ) ⟺ x`= x & y`= y & z`= z).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi



85

of mental state that satisfies the complex role which the Ramsey sentence ascribes to the variable that is assigned the guiding role in the part concerning practical reasoning.

3.1.2 The target: representationalism Representationalism provides a functional characterization of reason judgements. It claims that their cognitive role is to indicate that a certain fact obtains. That is, reason judgements bear a characteristic relation to certain non-mental entities, a relation that also holds between, say, a subject’s perceptual beliefs and certain facts concerning her immediate surroundings. Of course, representationalists do not claim that this is all there is to say about the nature of reason judgements. Using the terminology introduced in section 3.1.1, we can specify their view as follows: a comprehensive theory of an agent’s psychology includes an account of mental representation, that is, an account of how certain mental states indicate (or ‘track’, as it is often put) whether or not a certain fact obtains. The variables in that theory’s Ramsey sentence to which this representational role is assigned include the variable that is used to characterize the nature of reason judgements. This formulation allows for the possibility that there is more than one type of mental state that plays a representational role in an agent’s psychology. Reason judgements might, for example, differ from perceptual beliefs in some other, fundamental respect that suffices to render them distinct types of mental states. Typically, however, representationalists will simply take reason judgements to be the same type of mental state as ‘ordinary’ beliefs and then explicitly or implicitly endorse a representationalist account of the cognitive role of beliefs more generally. The idea that reason judgements are representational attitudes is a popular one in contemporary meta-ethics. A number of authors have made their commitment to a representationalist position explicit. In his influential defence of non-naturalistic moral realism, Russ ShaferLandau subscribes to cognitivism about moral judgements, which he understands as a view that allows for a central class of judgements within a domain to count as beliefs, capable of being true or false in virtue of their more or less accurate representation of the facts within the domain. (2003, 17)

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

86      David Enoch proclaims on the first page of his book-length defence of Robust Realism about normativity that he ‘believe[s] that there are irreducibly normative truths and facts’ and that ‘our thinking and talking about them amounts not just to an expression of any practical attitudes, but to a representation of these normative truths and facts’ (2011a, 1). Throughout the book, Enoch leaves no doubt that he takes ‘normative judgements [to express] fully representational beliefs about responseindependent pieces of reality’ (2011a, 249). Another advocate is Bart Streumer, who argues in favour of the view that ‘[n]ormative judgements aim to represent the world’ (2013, 452). Many others implicitly endorse such a view by rejecting positions like expressivism or other forms of noncognitivism, which they characterize as denying Representationalism (see, e.g., Smith 1994, 16). Or their endorsement is hidden in the assumption that reason judgements (or ‘normative beliefs’) differ from non-normative beliefs only in the kind of facts they purport to represent, which is why they see no need to characterize the attitude and move straight to a discussion of the metaphysical status of those facts. All of these authors subscribe to an account of the nature of reason judgements that the constructivist rejects and to which the account presented in section 3.2 offers an alternative.

3.1.3 Representationalism and directions of fit It is important, however, not to overstate the consequences of rejecting a representationalist account of reason judgements. Firstly, to deny such an account is not to deny that those judgements have propositional or, as it is sometimes put, representational content, where this simply means that they can be characterized as attitudes that relate a subject to a proposition that can be true or false and that is true if and only if the corresponding fact obtains. Many attitudes that do not have a representational function have content in this sense: you can hope, imagine, or desire that p, or assume it for the sake of an argument, for example. More importantly, denying that the ‘cognitive job’ of reason judgements is to indicate or track that certain facts obtain does not entail that those judgements have a different direction of fit from that of ordinary

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi



87

beliefs and must be understood as a kind of motivational state like desires either. The distinction between mental states with different directions of fit is, by now, an established part of the terminological inventory of philosophical psychology, and it plays a central role in a number of influential arguments in that area.² In his well-known defence of a Humean account of motivation, which employs the distinction between mental states with different directions of fit at a crucial place in the argument, Michael Smith suggests that ‘the difference between beliefs and desires in terms of direction of fit can be seen to amount to a difference in the functional role of belief and desire’ and he proceeds to sketch an account of the respective functional roles (1994, 115). He then argues that, because being motivated necessarily involves being in a state with the direction of fit of a desire, merely having certain beliefs cannot be sufficient for an agent to be motivated, so that motivation also requires having a certain desire (1994, 116–25). Smith’s explication of the notion of direction of fit in terms of a mental state’s functional role is unobjectionable as far as it goes, but we have to be careful when considering what is and is not entailed by this way of cashing out the metaphor. In particular, we need to distinguish between claiming that for some mental states their direction of fit is an essential part of their functional role and that these directions of fit are mutually exclusive and the much stronger claim that a mental state is fully characterized by its direction of fit. With respect to desires, Smith accepts this latter, stronger claim: ‘all that there is to being a desire is being a state with the appropriate direction of fit’ (1994, 116). Again, this may be perfectly acceptable in the context of his argument and the role the notion of a desire plays in it, but it invites the misunderstanding that there are really only two kinds of mental states with propositional content, beliefs, and desires, both of which are exhaustively characterized by their respective direction of fit. At least when ‘belief ’ is understood in the way it is commonly used by ordinary speakers, this seems clearly false, for there are propositional attitudes that differ from beliefs but do not share the direction of fit of a ² Some, however, have criticized the distinction as being fundamentally misguided: see Frost 2014.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

88      desire. If you imagine something in order to see how it affects you emotionally, or assume something merely for the sake of an argument, you can do so without believing it to be true, and yet you will show no inclination whatsoever to bring about the truth of the proposition in question. Thus, while the distinction between different directions of fit marks an important difference between belief and desire, the same difference appears to separate assumption and imagination from desire too. It is thus plausible to say that beliefs, assumptions, and imaginations share the same direction of fit with respect to their content: they all involve entertaining their content as something that is true, rather than as something that is to be made true. This is what distinguishes all of them from desires understood in the broad sense suggested by Smith. If we understand the direction of fit of states like beliefs along these lines, then those states and their functions are not fully characterized by their direction of fit. There are different functional roles that involve entertaining something as true. One such role is the representational role played, for example, by perceptual beliefs. In their case, a content is entertained as true in order to indicate that the world is a certain way, i.e. that the fact it presents as obtaining actually obtains. If we have such a state even though the content we thereby entertain as being true is actually false, then we are mistaken. Other states, like imaginations, have a different function: if you imagine something, the fact that what you imagine does not correspond to the way the world is does not testify to any mistake on your part. In this case, what you entertain as true is not meant to reflect what is actually the case. The same holds for assumptions too. Characterizing states with a belief ’s direction of fit in this way must thus be clearly distinguished from a characterization of such states as ‘aiming to track’ the world. That would be a much narrower conception of what is involved in a belief ’s direction of fit, one that excludes many kinds of mental state whose functional roles differ fundamentally from that of desires. I do not intend to argue that such a more demanding conception of direction of fit is inadequate; that might depend on the use to which it is put in the context of a particular argument. My point here is rather that rejecting a representationalist account of a certain class of mental states is incompatible with taking them to have a belief ’s

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

89

‘direction of fit’ only on this latter, more demanding conception. But on this conception, the incompatibility leaves plenty of alternative functions for those states and puts them in the company of states like imaginations or assumptions. In particular, it does not establish that those states bear any significant functional similarities to desires and other motivational states, because on this conception, being a state with the representational ‘direction of fit’ of perceptual beliefs and being a state with the motivational ‘direction of fit’ of desires are no longer the only options. Hence, rejecting a representationalist account of judgements about practical reasons does not already commit us to understanding those judgements as motivational states similar to desires.

3.2 Guidance and representation Like Representationalism, the account of reason judgements proposed in this chapter provides a functionalist characterization of the nature of those judgements: to be a reason judgement is to play a certain role in an agent’s psychology, as described by a comprehensive theory of that psychology. But the role it assigns to reason judgements differs fundamentally from the role suggested by the representationalist. It is not found in the part of the psychological theory that describes how some of the agent’s mental states indicate whether or not certain extra-mental facts obtain. Instead, the characteristic role of reason judgements is described in the part of the theory that is concerned with practical reasoning. In Chapter 2, I argued that episodes of practical reasoning are guided by reason judgements. What I suggest now is that playing that guiding role is part of the nature of those attitudes. To be a reason judgement is to be the kind of attitude that guides practical reasoning. Using the functionalist framework introduced in section 3.1.1, we can put this more precisely by saying that the theory of practical reasoning ascribes a complex property—that of playing the guiding role, symbolized by G—to a certain class of mental states (x*). The proposed account of reason judgements uses that part of the corresponding Ramsey sentence to give a definition of them:

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

90      To be a reason judgement is to be the x* such that ∃y₁ ∃y₂ . . . (S(y₁, sensory stimuli) & M(y₁, y₂, behaviour) & G(x*, episode of practical reasoning) & . . . )

According to this statement, for a mental state to be a reason judgement is, inter alia, to be the type of mental state that plays the guiding role in episodes of practical reasoning, along the lines of the Normative Guidance Account of such reasoning that was presented in Chapter 2 (see especially Chapter 2.3.2).³ To show that this is indeed an alternative to Representationalism, we have to answer two questions. Firstly, we have to ask whether the guiding role differs from the role assigned by the representationalist or whether it involves some representational element. It might already be clear from Chapter 2 that this role differs fundamentally from a representational one. However, to make this difference more vivid, I will illustrate it in section 3.2.1 with the help of an analogy. Secondly, the definition proposed above is, of course, incomplete: it contains an ellipsis. So even if the guiding role itself does not involve any representational element, we have to ask whether there is any reason to add an element that ascribes a representational role to reason judgements over and above the guiding role. I will address this question in section 3.2.2 below.

3.2.1 An analogy Consider a computer program that controls a machine which autonomously performs a certain task, say, a robotic vacuum cleaner that navigates a certain area and vacuums it, adapting its brushes to the flooring material. Such a program will include a number of variables, including ones for the detected flooring material, the distance to an object, the direction of movement, and the status of the brushes. The program will also include certain procedures for manipulating the values of some variables on the basis of the values of certain other variables. For such a procedure, the values of certain variables thus serve as an input, ³ The other parts included here characterize other types of mental states (denoted by y₁, y₂, . . . ).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

91

while the values of others are its output. To simplify a great deal, the program might control the robot on a particular occasion in the following way: suppose one of the sensors with which the robot is equipped causes the value of the flooring variable to change to ‘carpet’. The program then ‘calls’ (i.e. activates) one of its procedures, which, when applied to the new value of the flooring variable that serves as its input, changes the value of the brushes variable to ‘brushes in’. This in turn causes some mechanism to retract the robot’s brushes. Let us say that a robot whose computing hardware runs such a program has different computational states, which correspond to the different variables’ values as well as to the procedures the program might call. We can think of these states as the realizations of the respective parts of the program in the robot’s memory. In the specific scenario just described, several of these states control the robot’s behaviour by interacting in a certain way. The robot’s sensors cause the formation of a new state (corresponding to the flooring variable’s value ‘carpet’), which then serves as the input to a computational process that is determined by another state (corresponding to the called procedure). That process leads to the formation or alteration of a third state (corresponding to the brushes variable’s value ‘brushes in’) as its output, which in turn causes the robot to retract its brushes. There are important parallels between the way the computer program controls the robot’s behaviour and an agent’s practical reasoning. Just like practical reasoning, the controlling process that was just described involves states that play three distinct roles: input states, which parallel an agent’s premise-attitudes in providing the starting point of the process, output states, which parallel an agent’s conclusion-states in being the immediate result of the controlling process, and what we might call procedure states, which parallel the attitudes that guide an agent’s reasoning in mediating the transition from the process’s starting point to its result. This is not to say that the process that controls the robot’s behaviour qualifies as reasoning. The robot lacks a number of features that are necessary to be a reasoner, including the ability to be reflectively aware of its computational states or its behaviour as well as the capacity to have genuinely cognitive states, like beliefs, which involve the tokening of concepts. Nor is this analogy meant to entail a commitment to a

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

92      computational theory of mind. Reasoning is not understood as a merely syntactic operation but as sensitive to the contents of the attitudes involved (cf. Fodor 1975). The analogy illustrates, however, that reasoning is computational in the wider sense of being a process that generates outputs from certain inputs by following a certain general pattern or algorithm.⁴ The states that are involved in the robot’s control process differ in their functional roles. In our example, the robot’s output state is causally connected to its motor system (e.g. the mechanism that moves its brushes) in order to effect a certain change in the robot’s behaviour which, mediately, leads to a change in its environment. In this, it resembles an agent’s intentions and their motivational function. The input state, on the other hand, is causally connected to the robot’s sensory system so as to indicate certain aspects of the robot’s environment (e.g. that the local flooring is carpet). In this, it resembles an agent’s perceptual beliefs and their representational function. Note that in general, the input and output states of the robot’s computational processes need not be states of these particular kinds, that is, the input states of the robot’s computational processes need not all be states that indicate aspects of its environment, and the output states need not be states that effect certain behaviour.⁵ Things are different when it comes to the robot’s procedure state. Firstly, the role of that state differs from those of the other two states mentioned: it neither serves to indicate an aspect of the robot’s environment nor is set up to effect a change in the robot’s bodily state. The role of the procedure state is not to connect the robot to its body or environment at all; it rather connects some of the robot’s computational states to others. Of course, if the output state of a process is a behavioureffecting state, then the procedure state that mediates the process will indirectly affect the robot’s behaviour too. But that is not an essential ⁴ John Broome agrees that reasoning is computational in this sense (2009, 72; 2013, 232), see also Philip Pettit’s account of reasoning (2007, 498–502). ⁵ A control process that takes states with an indicating function as an input and yields behaviour-effecting states as its output, however, allows for the resulting behaviour to be responsive to those aspects of its environment that the input states indicate. This is an important parallel to the way practical reasoning allows an agent to act in a way that is sensitive to her beliefs about her environment.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

93

aspect of procedure states, because the output state could just as well be a different kind of state. Secondly, it is essential that the states that mediate the computational process (by providing the algorithm for it) are states with this distinctive functional role. Mediating a process that produces output states relative to certain input states requires a state that connects two or more of the robot’s internal states, rather than one that connects the robot to its environment. The same holds in the more sophisticated case of practical reasoning too. What we say about an attitude if we characterize it as playing the guiding role in such reasoning is simply very different from what we say about an attitude when we characterize it, along representationalist lines, as indicating whether certain facts obtain. The robot analogy helps us to see this. Moreover, it illustrates the more basic point, already discussed in section 3.1.3, that a mental (or computational) state can have a function that differs both from the representational (indicating) function of, for example, perceptual beliefs and from the motivational (behavioureffecting) function of, for example, desires or intentions.

3.2.2 Why not both? The robot analogy supports the claim that, in so far as we characterize a kind of mental state as playing the guiding role, we are not characterizing it as a representational state: what we learn about the nature of a mental state if we learn that it guides practical reasoning is very different from what the representationalist tells us about the nature of a mental state. As I said before, however, this does not yet suffice for formulating an alternative to Representationalism. Its advocates may insist that the project of characterizing the nature of reason judgements is not complete without adding a representationalist element. They need not deny that reason judgements play the guiding role, as long as those judgements are also characterized as playing a representational role in an agent’s psychology. Why, they might ask, should we not characterize reason judgements in terms of both roles? At this point, it is important to keep the dialectical situation of this chapter and of the book as a whole in mind. Recall that my overall aim is not to refute Representationalism but to formulate a constructivist

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

94      alternative that is philosophically viable. Hence, I need not show that it would be false to also characterize reason judgements in representationalist terms, or that adding the representational role to the guiding role turns out to be incoherent. To establish that characterizing reason judgements in terms of their guiding role in practical reasoning offers a genuine alternative to Representationalism, it is sufficient to show that not adding a representationalist element to that characterization does not result in an account that is unstable or obviously incomplete. In particular, it must be shown that there are no other important features of reason judgements—over and above their role in practical reasoning— that the account cannot accommodate while remaining antirepresentationalist. If there are no such reasons for adding a representationalist element, then the aim of formulating a genuine alternative to the representationalist’s account of reason judgements has been achieved. Hence, the question is not why we should not characterize reason judgements in terms of both roles, but why we should: is there any good reason to think that the mental states that play the guiding role in practical reasoning are states that indicate whether or not certain facts obtain?

3.2.3 Enoch on the deliberative indispensability of normative truths To answer this question, it is instructive to discuss an argument that has been put forward by David Enoch (2011a). Enoch defends a metanormative position he calls Robust Realism, which includes a commitment to Representationalism (see section 3.1.2). Unlike many other advocates of such a view, Enoch’s ambition is to go beyond defending it against various objections and provide a positive argument for his Robust Realism. What is interesting for our purposes is that his argument takes the phenomenology of practical reasoning—or ‘deliberation’, as Enoch prefers to put it—as its starting point. The main idea underlying the argument is that ‘objective, irreducibly normative facts are indispensable . . . for deliberation, and that this indispensability suffices to justify belief in their existence’ (2011a, 9). The core thesis of Robust Realism, that there are mind-independent normative facts that our

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

95

normative judgements aim to represent accurately, is thus claimed to gain its support from the same source that the constructivist alternative developed here relies on: the nature of practical reasoning. If it is compelling, Enoch’s argument would thus force the constructivist to add a representationalist element to his account of reason judgements precisely because he wants to build it on a theory of practical reasoning, which amounts to abandoning the aim of formulating an alternative to Representationalism. On closer inspection, however, Enoch’s argument does not pose a threat to the constructivist’s aims. He presents the argument as an indispensability argument, similar to indispensability arguments for the existence of numbers or sets in the philosophy of mathematics, as well as to inferences to the best explanation in the sciences. Enoch understands the latter as a special kind of indispensability argument, where the existence of certain entities is indispensable to the explanation of certain phenomena (2011a, 54–6). But, he argues, there are other kinds, where the existence of certain entities is indispensable to a different project from that of explanation, such as the project of deliberation. Why is the existence of independent normative facts indispensable to that project? Enoch’s answer points to the phenomenology of deliberation: when you are thinking about what to do, or what ‘makes best sense’ for you to do: it doesn’t feel like just trying to make an arbitrary choice. This is just not what it is like to deliberate. Rather, it feels like trying to make the right choice. . . . Making the decision is up to you. But which decision is the one it makes most sense for you to make is not. This is something you are trying to discover, not create. Or so, at the very least, it feels like when deliberating. (2011a, 72–3)

In particular, Enoch continues, there is an important difference between deliberation and merely picking a course of action: We can just pick in the face of a known (or believed) absence of reasons. But we cannot, it seems, deliberate in the face of a believed absence of reasons. Knowing that there is no decision such that it makes most sense for us to make it, we cannot—not consistently,

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

96      anyway, in a perfectly commonsensical sense of ‘consistently’—try to make the decision it makes most sense for us to make. . . . Thus, in deliberating, you commit yourself to there being (normative) reasons relevant to your deliberation. (2011a, 74)

According to Enoch, practical reasoning necessarily involves a commitment to accepting the truth of certain reason judgements. This is why it is inconsistent—‘in a perfectly commonsensical sense’—to engage in an episode of reasoning and to judge, at the same time, that there are no reasons in support of the action in question: the contents of this latter judgement and the judgement whose truth one is committed to accepting are incompatible. So far, I am in agreement with Enoch (see 2.2). However, he then proceeds by arguing that because the truth of certain reason judgements is, in this sense, indispensable to the project of deliberation, and because that project is itself ‘rationally non-optional’ for us (2011a, 70), we are warranted in accepting the existence of the those ‘objective, irreducibly normative facts’ that Robust Realism postulates. This, however, is where his argument moves too quickly. We can distinguish two steps in this argument, both of which have to be convincing for it to make a compelling case for Robust Realism. The first step is to argue that if reasoners cannot help but accept the truth of certain reason judgements—if those truths are indispensable for the project of practical reasoning—and if practical reasoning is itself rationally non-optional, then those judgements are justified. The judgements’ indispensability to a rationally non-optional project warrants their truth. The judgements whose truth is warranted by this first step of the argument are ordinary normative judgements about what is a reason to do what. But, of course, defending the truth of such judgements is not distinctive of Robust Realism. That view holds that there are certain irreducible and mind-independent normative facts, and that normative judgements are true because they accurately represent such facts. Thus, for the argument to support Robust Realism, a second step is required, according to which the truth of the reason judgement entails the existence of an irreducible and mind-independent normative fact that it accurately represents. Only then does the fact that accepting the truth of certain reason judgements is indispensable to the rationally non-

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

97

optional project of practical reasoning also warrant the ontological claim about the existence of independent normative facts that is at the heart of Robust Realism. These two steps can be distinguished for other indispensability arguments too, specifically for inferences to the best explanation, that is, arguments that are concerned with what is indispensable to the project of explanation. Take Enoch’s example of a cloud chamber in which a physicist observes a vapour trail (2011a, 51). Believing that this observation is best explained by the fact that a proton passed through the chamber, the physicist infers that this is what happened. Here, the indispensability of believing that a proton passed through the chamber to the (also rationally non-optional) project of explaining observable phenomena warrants that belief. But this belief must again be distinguished from what the equivalent of Robust Realism for physics claims, which is that there are independent facts about protons and their movements that such beliefs accurately represent. Only if the physicist’s belief entails this latter claim does the indispensability of believing that a proton passed through the chamber for the rationally non-optional project of explaining the observed vapour trail also warrant the robust ontological claim about protons. In both cases, the second step of the argument presupposes that the beliefs or judgements whose truth is warranted by the first step are ontologically committing, and that this commitment must be understood along the lines of Robust Realism. Enoch does not distinguish these two steps explicitly in his argument, but the distinction is crucial for our purposes. The constructivist can happily accept the claim that engaging in practical reasoning commits us to the truth of certain reason judgements, as long as this means no more than what Enoch (2011a, 73) quotes from Thomas Nagel in support of his account of the phenomenology of deliberation, which is that ‘[i]n deliberation we are trying to arrive at conclusions that are correct in virtue of something independent of our arriving at them’ (Nagel 1986, 149). This supports the idea that we take the reason judgements that we accept when engaging in practical reasoning to be correct in virtue of conforming to a standard that is not up to us and that we are not guaranteed to meet. But it does not entail that this standard is provided by a realm of mind-independent,

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

98      irreducibly normative facts. Nagel is right in characterizing the commitments of reasoners in more careful terms: the ontological status of normative facts is not revealed in the phenomenology of reasoning. This is why the second step is required for the argument to support Robust Realism, and it is this step’s presuppositions concerning the ontological commitment of the truth of reason judgements with which the constructivist disagrees, that is, constructivists agree that reasoners are committed to the truth of reason judgements, and that some of these judgements are in fact true.⁶ But they deny that those judgements are representational states that are true because thy accurately represent independent normative facts. As we will see in Chapter 4, they offer an alternative account of the truth of those judgements. This allows them to accommodate the phenomenology of practical reasoning while denying Enoch’s Robust Realism.⁷

3.2.4 Guidance, truth, and objectivity Let me recapitulate. Constructivism provides an anti-representationalist account of the nature of reason judgements by characterizing them in terms of the guiding role they play in practical reasoning. This role itself differs from the role in terms of which representationalists characterize those judgements. But the constructivist’s account qualifies as a genuine alternative to Representationalism only if it is not forced to add a representationalist element in order to accommodate any other important features of reason judgements. We thus asked whether there is any good reason to think that the mental states that play the guiding role in

⁶ They also agree that if such judgements are true, then the corresponding reason fact exists (see Chapter 1.3). What they deny is that those judgements are true because the corresponding facts exist. ⁷ Enoch does acknowledge that, for his argument to be successful, he has to show that Robust Realism does a better job at accommodating the phenomenology of deliberation than alternative views. But he assumes that those views deny that deliberation commits us to accepting the truth of certain normative claims and trying to get by with less (2011a, 79). At least in the case of constructivism, however, that is not correct. Enoch does not take the constructivist option seriously here because he fails to see that constructivism offers a genuine alternative to the established meta-ethical views (see Chapter 1.6).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

99

practical reasoning are states that indicate whether or not certain facts obtain. As the discussion in section 3.2.3 has shown, the phenomenology of practical reasoning does not provide such a reason. Constructivists who argue, as I did in Chapter 2, that every episode of practical reasoning is guided by a reason judgement will agree with Enoch that reasoners are committed to accepting certain reason judgements. They can also agree that, in doing so, we take those judgements to be true, where this means, in the first instance, that we take them to conform to a standard of correctness that is not up to us and that they are not guaranteed to meet. But a commitment to the truth of reason judgements must be distinguished from a commitment to taking those judgements to be representational states that are true because they accurately indicate that certain facts obtain. While the phenomenology of reasoning does not support adding a representationalist element to the characterization of reason judgements, the discussion of Enoch’s argument has made two features of reason judgements salient that a plausible account of those judgements should be able to accommodate: they can be true or false, and their truth is objective (at least) in the sense that the truth of someone’s reason judgement is not guaranteed by the mere fact that he has formed it. The constructivist’s account of the nature of these judgements must thus accommodate those features without resorting to representationalist resources. That means, in particular, that constructivists must provide an account of the truth of reason judgements that does not explicate their truth in terms of the accurate representation of a realm of mindindependent normative facts. As I explained in Chapter 1.3, however, a constructivist theory of practical reasons must include such an account anyway, because the truth-aptness of reason judgements is not just a feature that a plausible account of their nature must accommodate, but a resource that the constructivist theory relies on in its account of reason facts (which, recall, it takes to be grounded in the truth of certain reason judgements). In Chapters 4 and 5, I will present an account of the truth of reason judgements that draws only on their guiding role and that is thus compatible with the constructivist’s overall theory. In Chapter 6, I will show that this allows for the truth or falsity of our reason judgements to be an objective matter.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

100     

3.3 Constructivism and expressivism Showing that the constructivist’s account of the nature of reason judgements is not a representationalist one, however, does not suffice to set it apart as a distinct meta-ethical view. Other, more established positions also reject Representationalism. Expressivism in particular is closely associated with an anti-representationalist approach in meta-ethics. Thus, in order to show that constructivism offers an alternative to the established meta-ethical views, we have to show that it differs from expressivism. In fact, the suspicion that, upon closer inspection, metaethical constructivism turns out just to be another version of expressivism is one of the main reasons why many are sceptical of the constructivist project.⁸ In this section, I will show that the constructivist theory presented in this book differs in significant respects from standard versions of expressivism. This allows us to reject the objection that, because it collapses into expressivism once its details are spelled out, constructivism does not merit consideration as a distinct meta-ethical view.

3.3.1 Two kinds of anti-representationalism Discussing the similarities and differences between constructivism and expressivism is difficult not least because expressivism is somewhat of a moving target. Particularly over the last fifteen years, proposals as to what exactly should be considered the defining characteristic of an expressivist theory have multiplied (cf. Schroeder 2008; 2010; Chrisman 2008; 2011; Ridge 2014). Along with most meta-ethicists, I take an expressivist theory to be one that offers a certain kind of account of the meaning of normative language, e.g. of sentences like: (1) ‘That it causes cats great pain is a reason not to torture them.’

⁸ This suspicion has been articulated by, among others, David Enoch (2009), James Lenman (2012), Michael Ridge (2012), and Jay Wallace (2012).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

101

According to those accounts, what we need to focus on in order to explain the meaning of such sentences is the mental states a speaker expresses by uttering them. Thus, expressivism is chiefly a position in the philosophy of language. But since its account of the meaning of normative language places mental states centre stage, expressivism also comes with an account of the nature of the states expressed by the utterance of sentences like (1), that is, an account of the nature of reason judgements. Here, expressivists deny that those mental states are representational ones, which is why their account of the meaning of normative language must draw only on features of the expressed states that are not exclusive to representational states. Thus, expressivism and constructivism both reject Representationalism. But this no more makes them the same view than the shared rejection of the thesis that an equal distribution of property is intrinsically desirable makes Robert Nozick and Derek Parfit proponents of the same view in political philosophy.⁹ And while expressivism and constructivism are alike in rejecting Representationalism, they differ substantially in their positive accounts of the nature of reason judgements. In general, expressivists characterize those judgements—that is, the mental states they take to be expressed by the utterance of sentences like (1)—as desire-like pro-attitudes.¹⁰ Such attitudes are taken to play a special role in the explanation of intentional behaviour. Roughly, an agent who has a pro-attitude towards a certain state of affairs is disposed to behave so as to bring about that state of affairs. Having a pro-attitude thus essentially involves being motivated, and it is this motivational role on which expressivists typically base their positive accounts of the attitudes they take to be expressed by the relevant utterances. This is witnessed by one of the main arguments put forward in favour of expressivism: the argument from motivational internalism.¹¹ Motivational internalism is the thesis that there is a close connection between normative judgements and being motivated to act in accordance with them. If one accepts a strong version of this thesis, according to

⁹ Cf. Nozick 1974 and Parfit 1997. ¹⁰ Cf. Schroeder 2008, 3. ¹¹ Cf. Schroeder (2010, 8–12) and van Roojen (2018, sec. 3.3) on the role of this motivational argument in motivating expressivist positions.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

102      which normative judgements are necessarily accompanied by the corresponding motivation, then the best explanation of this connection appears to be that those normative judgements are themselves motivational states. Taking this argument to support expressivism thus entails that the expressivist’s account of the nature of normative judgements characterizes them as pro-attitudes in the above sense. Since most expressivists also accept the Humean thesis that representational states (such as ordinary beliefs) alone cannot motivate, they take this argument to support their rejection of Representationalism too. To highlight the emphasis that such an account of the nature of normative judgements places on the (alleged) motivational role of those judgements, we can call this a conativist version of anti-representationalism. Constructivists, on the other hand, characterize reason judgements in terms of the guiding role they play in practical reasoning. They thus opt for what we might call rationalist anti-representationalism. This is an important difference, for it locates the cognitive role of reason judgements in different parts of the theory of an agent’s psychology. According to expressivists, it is the part that is concerned with intentional behaviour. According to the constructivist, it is the part concerned with practical reasoning. In other words, constructivism characterizes reason judgements in terms of their role in the process that leads to the formation of the kinds of pro-attitudes that figure in the explanation of action, which are the attitudes that expressivists focus on. Moreover, the activity that is involved in practical reasoning is not the same kind of activity as intentional action, and it does not lend itself to the same explanatory model. We do not have the kind of control over our reasoning that we have over our actions: we cannot engage in a particular episode of reasoning at will as we can raise our arm at will. We can, of course, intentionally avoid reasoning at all, for example by distracting ourselves, and we can also intentionally bring ourselves to reason about a certain issue, for example by sitting down in a quiet room and bringing that issue to mind. But we cannot intentionally reason from a specific set of our present attitudes to the formation of an arbitrary new attitude as we can make an arbitrary bodily movement. To see this, suppose that someone offers you a large amount of money if you manage to reason from the intention to call a taxi and the belief that calling a taxi is a

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

103

necessary means to getting the next flight to Paris to the intention to get the next flight to Paris. You will not be able to form that intention by way of the demanded episode of reasoning. Now, you might point out that this is because the demanded reasoning is incorrect and that engaging in an episode you consider correct would be much easier. It is true that the reasoning is incorrect, but why should that make a difference if reasoning can be explained in terms of desirelike pro-attitudes, just like any (other) kind of intentional behaviour? After all, you can certainly perform an ordinary action that you consider incorrect according to some standard. And even if the behaviour in question is more similar to reasoning in that it comes with its own, constitutive standard of correctness, such as playing a certain piece on the piano, it is nevertheless possible to intentionally engage in that behaviour incorrectly, for example in order to entertain an audience or simply to illustrate a common mistake in the context of teaching.¹² But a teacher cannot illustrate incorrect reasoning by engaging in it: she has to describe it or point to unwitting examples. If intentionally engaging in the activity in a way that you judge incorrect is not possible in the case of reasoning, this is because that judgement excludes the reason judgement needed to guide the reasoning. You cannot engage in the demanded episode of reasoning because you lack the required guiding attitude. But that shows that the guiding attitude is not the expressivist’s pro-attitude, because those attitudes are generally not excluded by taking their object to be incorrect.

3.3.2 Expressivism as conativist anti-representationalism So, the motivational role in intentional behaviour and the guiding role in practical reasoning must be kept apart, and attitudes that are identified in terms of either role are not the same kind of attitude. The distinction between conativist and rationalist anti-representationalism therefore marks a real difference between meta-ethical views, and whereas

¹² On this, see the discussion in Chapter 5.5 below.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

104      constructivism falls on one side of this distinction, most expressivist accounts fall on the other. Most, or all? Simon Blackburn, one of the most influential proponents of expressivism, has given an account of the relation between reasoning and reason judgements which might appear to take the constructivist’s side. Blackburn writes that what distinguishes the ‘movement of the mind’ involved in reasoning from other mental processes is that the transition from one mental state to another is a ‘guided’ one (2010, 6–7). Concerning reason judgements, he then writes that ‘we say . . . that the fact of x is a reason for doing y when we think it is good to . . . be moved towards doing y upon apprehending x’ (2010, 9). By making such a judgement, we ‘commend or endorse the kind of guidance of the mind that it indicates’ (2010, 9). Superficially, that seems quite similar to the account presented here. But there are crucial differences. Most importantly, Blackburn does not suggest that reasoning is guided by a reason judgement. Instead, he takes reasoning to be ‘guided’ by the mental state from which it starts (2010, 6). According to him, reason judgements are simply pro-attitudes towards a particular kind of conduct, that is, ‘movements of the mind’. They involve approval of such a movement in the same way that the judgement that it is good to offer your seat on the bus to an elderly person involves approval of such conduct (cf. 2010, 12). Blackburn thus conflates the distinction between reasoning and ordinary intentional behaviour that was drawn in section 3.3.1. Consequently, he assimilates the relation between reason judgements and the process of reasoning to (what he takes to be) the relation between evaluative judgements and intentional behaviour. For him, reason judgements thus belong in the same category as evaluative judgements: they are conative states. Considering the Humean background of Blackburn’s approach to moral psychology, I suspect that this is the only non-representationalist category he is willing to admit. But apart from that, it is important to note that because the account of reasoning that Blackburn outlines does not require the transition from the premise-attitudes to the conclusionattitudes to be guided by an additional attitude, his view does not offer the resources to characterize any type of mental state in terms of such a guiding role. Thus, his view should not be understood as a prior instance

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

105

of the rationalist anti-representationalism that is distinctive of the constructivist position presented in this book. There are, however, versions of expressivism whose accounts of the nature of reason judgements come closer to the one proposed in this chapter. The one that comes closest is probably the version presented by Allan Gibbard (1990), who gives an account of normative thought in terms of the psychological state of ‘accepting a norm’ (1990, 55). This state is then characterized in terms of its role in what Gibbard calls the ‘normative control system’, which he distinguishes from another part of our motivational system, which he calls the ‘animal control system’ (1990, 56). The normative control system is ‘linguistically infused’ (1990, 57), which means, roughly, that its motivational tendencies can be both expressed and influenced through the use of language, specifically normative language. This provides an evolutionary advantage because it facilitates interpersonal coordination and thus enables the members of a group to implement a mutually advantageous strategy for certain recurring bargaining problems (1990, 64–8). It is difficult to assess how similar this is to the constructivist’s account presented here. Gibbard does not tell us much about how the normative control system works, how it influences an agent’s behaviour. It is thus unclear whether the process of practical reasoning as it has been characterized in Chapter 2 is best understood as the (or one) activity of the system as Gibbard understands it, or whether Chapter 2 describes a competing account of normative control.¹³ Gibbard does not characterize norm acceptance as an attitude that guides the normative control system—what he says about the workings of this system does not suggest that it requires guidance by an attitude. Instead, he identifies that state mainly through the way in which it manifests in normative discussion (1990, 73–5). But, as Gibbard himself points out, his account of normative thought does not simply classify it as a conative phenomenon either: ‘the psychic states and processes I have depicted—the acceptance of norms, and normative control of action and emotion—do not fit neatly

¹³ Gibbard acknowledges that there are important parallels between his account of normative control and traditional conceptions of reason, but he also insists that ‘it would be misleading to call the normative control system the faculty of reason’ (1990, 80–1).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

106      into the traditional psychologists’ categories of cognition, affect, and conation’ (1990, 80). To be honest, it is simply not clear to me whether the account of normative thought presented by Gibbard (1990) should be classified as a version of conativist or rationalist anti-representationalism, and whether the account of the nature of reason judgements presented here is best understood as a way of elaborating or amending that account, or as a genuine alternative to it. Be that as it may, Gibbard’s overall view differs significantly from the constructivist’s: he does not take the normative control system to be subject to a robust (that is, not simply conventional) standard of correctness, nor does he take normative judgements to be capable of being true or false in any substantive sense. As we will see in Chapters 4 and 5, constructivists take a different stance on both of these issues. Hence, the constructivist view presented in this book differs significantly from Gibbard’s (1990) view, even if the contrast between them does not follow the line between rationalist and conativist versions of anti-representationalism. I should also note that I do not take this line to coincide with the widespread distinction between cognitivism and non-cognitivism. These terms are used in a variety of senses in the debate. As I understand them, cognitivism holds that normative judgements have substantial truth conditions that allow for their classification as true or false in a nondeflationary sense, while non-cognitivism denies this (cf. van Roojen 2018). As I argue in Chapter 4, reason judgements can be true or false in a substantial sense even though they are not representational states. Hence, cognitivism is not committed to Representationalism. Can there also be cognitivist versions of conativist anti-representationalism? Is it possible to come up with a conception of desires or intentions that identifies them in terms of their motivational role but nevertheless assigns substantial truth conditions to them? I am not sure. In any case, these distinctions should not simply be conflated: the contrast between cognitivism and noncognitivism concerns the truth-aptness of the mental states in question, whereas both the distinction between Representationalism and AntiRepresentationalism and that between conativist and rationalist versions of the latter concern their functional role. These questions are related, but they are, nevertheless, different questions.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

107

There are meta-ethical views that are labelled as versions of expressivism and that claim that the mental states expressed in normative discourse are intrinsically motivational and capable of being true or false in a substantial sense. But these are hybrid theories: very roughly, they take those states to comprise a pair of attitudes, one of which is an ordinary, representational belief, the other a desire-like state (cf. in particular Ridge 2014). Those theories resemble constructivism in the advantages they try to combine and the commitments they try to avoid, but they do so in a very different way. They claim that normative judgements are (or involve) both representational and conative states, while constructivists claim that they are neither. Contemporary hybrid theories are extremely sophisticated, and adequately presenting their details or discussing their merits would go well beyond the scope of this book (for an overview, see Fletcher and Ridge 2014). Remember, though, that my aim here is not to establish that constructivism is preferable to other meta-ethical views, but only that it offers a genuine alternative to them. As we have seen in this section, this is also true when it comes to expressivism. Once the differences in their accounts of the nature of normative judgements are acknowledged, the widespread suspicion that constructivism is simply another version of meta-ethical expressivism can be laid to rest.

3.4 Extending the account What has been said so far about the nature of reason judgements applies primarily to basic first-person judgements, in which an agent takes herself to have a reason to perform a certain action. These reason judgements can readily guide her in her reasoning. But what about other reason judgements—conditional ones or those that ascribe reasons to other people, for example? In this section, I outline how the present account can be extended to those other kinds of reason judgements. Sometimes, we think about what we would have reason to do in circumstances that we do not currently face. I might think, for example, that if I were attacked by a wolf, the sensitivity of its nose would be a reason to kick it there. I do not currently have to decide whether or not to

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

108      kick an attacking wolf, but that does not mean that this judgement cannot currently guide any reasoning of mine. We should acknowledge both the ubiquity and the importance of what can be called hypothetical reasoning: reasoning about what to do in a scenario that one considers possible but not actual. Such reasoning might be thought of as occurring in an offline module of the mind and as starting from suppositions rather than beliefs. Its conclusion-attitude is best characterized as a conditional intention to perform a certain action if one encounters a certain situation. But whatever the details of hypothetical reasoning, it is clear that it is often very useful because it allows us to prepare for circumstances in which real-time reasoning is unfeasible. Acting on the conditional intention to kick a wolf in the nose if you are attacked by one will require considerably less time and attention than forming the intention to kick it in the nose by reasoning from your beliefs about the sensitivity of the wolf ’s body parts. Moreover, if we anticipate encountering such circumstances at some point in the future, deciding now what to do later allows us to coordinate other choices we make in the meantime. Conditional reason judgements, I submit, guide such hypothetical reasoning. Another interesting case is judgements about other people’s reasons. Why should I be concerned with what you have reason to do? After all, my reason judgements could hardly guide your reasoning. In response, we can start by pointing out that judgements about our own and judgements about other people’s reasons are related in a certain way: if I judge that something is a reason for you to do something, then I must be prepared to judge that it would also be a reason for me to do it if I faced circumstances that were relevantly similar to yours—if I was in your shoes, as it were. This universalizability of reason judgements—i.e. the thesis that for there to be a difference in the reasons two persons have, they have to differ in some relevant non-normative respect—is one of the few theses that are accepted by virtually everyone in the debate.¹⁴ ¹⁴ Some have claimed that the universalizability of reasons is incompatible with particularism. Indeed, Jonathan Dancy, the chief proponent of particularism, explicitly rejects a universalizability thesis (1993, 79–81), and Roger Crisp, after defending universalizability against some criticisms, argues that this incompatibility is a serious disadvantage of the particularist’s position (2000a, 40–2). However, the thesis that Dancy rejects (and that is indeed incompatible with his particularism) is a different and much stronger thesis than the universalizability thesis considered here, which, as Schroth (2003) shows, is perfectly compatible with particularism and

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

109

Second- and third-person reason judgements can thus be understood along the lines of the conditional reason judgements we have just discussed. To judge that you (currently) have a reason to do something can play the same cognitive role as judging that I would have a reason to do it if I were to face your (current) circumstances: they both can guide an episode of hypothetical reasoning on my part. This, however, leaves us with the question why we make and express such second- or third-person reason judgements as often as we do. For all that has been said, a judgement about the reasons we ourselves would have in certain circumstances would play the same cognitive role. But while guiding one’s reasoning about what to do in actual or hypothetical situations is the characteristic role those attitudes play in an agent’s psychology, it is not the only one. As we saw in Chapter 2.5, the attitudes that guide an agent’s reasoning also play an important role in how she will respond when requested to justify an action. That suggests that they will also affect what justifications by others she will be inclined to accept (by granting certain second-person reason judgements) and what justifications on behalf of others she may propose (by submitting certain third-person reason judgements). In addition to guiding our reasoning, reason judgements also play an important role in the related interpersonal practice of justification.

is implicitly acknowledged by Dancy himself in his discussion of what coherence in moral judgement involves (see 1993, 63–4).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

4 The truth about reasons Constructivism needs an account of what it is for a reason judgement to be true. For one, it seems obvious to most people that reason judgements can be assessed as true or false, and constructivism, like any other theory of practical reasons, would do well to accommodate that. More importantly, however, the truth of reason judgements plays a particularly important role in the constructivist’s theory of practical reasons. According to that theory, reason facts obtain because of the truth of the corresponding reason judgements. An account of what that truth consists in is thus needed because, for constructivists, it amounts to an account of what grounds practical reasons. Moreover, some of the more popular options in the debate on the nature of truth turn out to be incompatible with this approach. It is thus vital to the constructivist’s project to identify options that are compatible it. This chapter explores constructivism’s commitments in the theory of truth. It starts by considering in more detail what an account of the truth of reason judgements that is compatible with constructivism would have to look like (sections 4.1 and 4.2). I then present an account of truth that has been developed by Crispin Wright and others, which meets these requirements and fits well with the constructivist’s overall approach (sections 4.3 and 4.5). The account of practical reasons that combining this account of truth with the constructivist approach yields will be discussed in section 4.6.

Constructing Practical Reasons. Andreas Müller, Oxford University Press (2020). © Andreas Müller. DOI: 10.1093/oso/9780198754329.001.0001

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

, ,  

111

4.1 Constructivism, correspondence, and deflationism What theory of the truth of reason judgements should a proponent of constructivism adopt?¹ What criteria does such a theory have to meet in order to be compatible with the constructivist’s account of practical reasons? Firstly, constructivists must avoid any theory of truth that presupposes the prior existence of the reason facts that correspond to true reason judgements. Adopting such an account of truth would render their account of practical reasons viciously circular, because then what they offer as the explanatory ground of facts about practical reasons would actually presuppose the prior existence of those facts. This does not mean that constructivism must deny Correspondence, that is, the claim that that reason judgements are true if and only if things are as they say they are. As we saw in Chapter 1.3, that claim is symmetric and thus compatible with attributing explanatory priority to either side. But it does exclude a popular theory of truth that explicates truth in terms of accurate representation. According to this theory, a true judgement is true because it represents things as being a certain way and things really are that way (cf. Künne 2003, ch. 3). The theory thus attributes explanatory priority to the right-hand side of Correspondence, which renders it incompatible with constructivism. I will refer to this theory as the representationalist theory of truth. It is more commonly known as the correspondence theory, but because it is not the only theory that is

¹ Throughout this chapter, I assume that mental states like reason judgements or ordinary beliefs are appropriately construed as truthbearers, that is, they are the kind of thing to which truth can be ascribed. In the various debates about truth, many other things have been proposed as truthbearers: sentences (both types and tokens), statements, utterances, propositions, etc. (for an overview see Kirkham 1992, 54–8). Indeed, in the discussion of some alternatives to the conception of truth that I advocate, I will join their proponents in taking some of those other things to be truthbearers too. Despite the debates surrounding these issues, I do not think that this is a problem, for I do not think that there is only one sort of thing that can really be true or false. Instead, I agree with Richard Kirkham that the best attitude here is one of tolerance: [T]here is no ‘correct’ answer to the question of what kind of thing can possess truth values. The matter is one of choice, not discovery. . . . For a given philosophical program it will be more useful to regard this sort of entity rather than that one as the bearer of truth values. For a different program it may turn out that the reverse decision is the wiser one. (1992, 59)

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

112     compatible with the Correspondence claim, that might be misleading in the present context. Secondly, constructivism requires truth to play a significant explanatory role: it is the truth of certain reason judgement that is supposed to provide the ground for reason facts and thus explain why some considerations provide reason for certain actions. A theory of truth that is compatible with constructivism thus needs to characterize truth as a property that is substantive enough to bear that explanatory significance. This criterion excludes another popular theory of truth: deflationism.² Deflationists approach the issue of truth by asking what we do when we say of something that it is true. Roughly put, their idea is that we use ‘is true’ and related expressions to assent to the statement to which we apply it: to say ‘p is true’ is just another (perhaps more emphatic) way of assenting to p, which one might as well do by simply saying ‘p’. That does not mean that ‘true’ and its cognates are redundant, however, because we can use them to assent to statements whose content we do not know but which we can pick out by using a definite description (‘What the Pope says about transubstantiation is true’), or to an indefinite number of judgements we do not want to or cannot enumerate (‘All the Pope’s statements are true’). But according to deflationism, all we need to know about truth in order to explain the use of its attributions in such indirect or generalized forms of assent is that these attributions obey the following schema: (T) ‘p’ is true if and only if p. Hence, all you are committed to by claiming that some reason judgement (or the statement used to express it) is true is the content of that very reason judgement. In particular, such a claim does not entail that the judgement (or statement) in question successfully represents some fact. Consequently, it does not presuppose the prior existence of such a fact. Deflationism thus meets the first criterion introduced above. But it does not meet the second one. A deflated notion of truth is too ² Important recent advocates of deflationism include Paul Horwich (1990) and Hartry Field (1994).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

, ,  

113

lightweight, as it were, to play the role that it is assigned in the constructivist account of practical reasons. Deflationists view truth as a purely expressive notion. They do not think that understanding truth requires us to identify a substantive property of judgements—such as a certain relation between their content and the facts—that all and only true judgements have. Accordingly, truth ascriptions do not pull any explanatory weight of their own, that is, over and above the explanatory weight of the content of the statements they are used to assent to. If deflationism is correct, then ‘The judgement that F is a reason to do A is true’ has the same explanatory significance as ‘F is a reason to do A’. But then a constructivist account of practical reasons would amount to the claim that the fact that F is a reason to do A obtains because F is a reason to do A, which is tautological. To offer an informative account of practical reasons, constructivism requires a substantive conception of the truth of reason judgements, one according to which attributing truth to such a judgement adds something explanatorily significant to the content of that judgement. A related worry is that by adopting a deflationist conception of truth, constructivists would deprive themselves of the theoretical resources to distinguish their view from alternative accounts of practical reasons. Deflationism often does not stop with the notion of truth, and the suspicion is that it cannot stop there either. For once the notion of truth has been accepted to be exhaustively characterized by the Tschema, it seems that similar schematic accounts can be given for notions like belief, fact, property, representation, and so on. This is what James Dreier aptly calls the problem of creeping minimalism:³ ‘Minimalism sucks the substance out of heavy-duty metaphysical concepts’ and thereby threatens to make different meta-ethical positions ‘indistinguishable’ (2004, 26).⁴ As we saw in Chapter 2, one important point of disagreement between constructivism and its alternatives concerns the ³ Following Horwich (1990), Dreier and others use ‘minimalism’ as a name for (a certain kind of) deflationism about truth. Since I use that term for the non-deflationist account of truth proposed by Crispin Wright, I will stick with ‘deflationism’ for what Dreier calls minimalism. ⁴ Dreier is specifically concerned with the consequences of this creeping minimalism for distinguishing expressivism and classical forms of realism, but it is clear that the general problem he describes would affect constructivism and its contrast to other meta-ethical views as well.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

114     nature of reason judgements and whether those judgements are adequately characterized by Representationalism. But if, via the path described by Dreier (2004, 25–32), a deflationist conception of truth creeps all the way to the notion of representation, we should be sceptical whether the remaining deflated notion of representation is apt to mark any substantive difference between views that take reason judgements to be representational states and views that do not. Adopting a deflationist conception of truth thus threatens to obliterate the differences between constructivism and other accounts of practical reasons.⁵

4.2 Towards a compatible theory of truth So far, we can conclude that constructivists should not opt for a representationalist theory of truth, since that would render their account of reasons circular, nor should they adopt a deflationist approach to truth, since that would deprive them of the theoretical resources to formulate their account of practical reasons and distinguish it from alternative accounts. What, then, should constructivists say about the truth of reason judgements? A theory of truth that is compatible with constructivism must characterize truth as a substantive property with explanatory significance, but it must not presuppose the prior existence of the reason facts that correspond to true reason judgements, so the property in question must not involve the accurate representation, fitting, matching, or any similar relation to those facts. This leaves us with at least two well-developed options. One is the coherence theory of truth. Roughly put, this theory takes truth to consist in the truthbearer’s coherence with a certain set of beliefs.⁶ Versions of this theory can differ in the way they characterize the coherence relation, as well as in how they identify the relevant set of beliefs. What unifies

⁵ Of course, some advocates of deflationism will be happy to accept that consequence because they are motivated by a broad anti-metaphysical sentiment, which is likely to prompt them to be sceptical about the entire meta-ethical project constructivists and their opponents engage in. I do not share that scepticism. ⁶ A classic articulation and defence of the coherence theory of truth can be found in Blanshard (1939). For a comprehensive discussion of the theory’s merits, see Walker (1989).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

     

115

them is the idea that truth must be understood in terms of the truthbearer’s relation to other truthbearers, rather than to some realm of facts. This is why the coherence theory meets the conditions for compatibility with the constructivist’s project I just introduced. The second theory that meets these conditions is the pragmatist theory of truth. Very roughly, this theory holds that being true is to be understood in terms of having a certain status in an activity that is subject to certain rules or norms.⁷ One of the most prominent versions of this theory is due to Charles S. Peirce, who claims that ‘[t]he opinion which is fated to be ultimately agreed to by all who investigate, is what we mean by truth’ (1878, 300). Elsewhere, he writes that the truth is ‘the predestined result to which sufficient inquiry would ultimately lead’ (1907, 419). In Peirce’s version of the pragmatist theory, the relevant activity is that of investigation or inquiry, and the relevant status is that of being ultimately accepted or agreed upon. Since this activity is regimented by certain norms of justification, such that a judgement’s being accepted in inquiry entails that an inquirer is justified in accepting it, Peirce is best understood as explicating truth in epistemological terms.⁸ Pragmatist approaches to truth like this meet the conditions for compatibility with constructivism, as long as the relevant activity and the norms governing it do not presuppose the prior existence of the corresponding facts. In the case of Peirce, it is clear that they do not: in the famous first passage quoted above, he continues: ‘the object represented in [the opinion fated to be ultimately agreed to] is the real. That is the way I would explain reality’ (1878, 300; my emphasis). Peirce acknowledges that this entails that reality is mind-dependent in some way, but he emphasizes that it does not depend on the opinion of any person or finite group of persons ⁷ Note that this broad characterization of pragmatist theories includes views that characterize the status in question in terms of justifiability or verifiability. Such views are also often classified as epistemic or verificationist theories of truth. ⁸ Other well-known proponents of a pragmatist theory of truth are William James (1907) and Hillary Putnam (1981). Note that the distinction between the pragmatist and the coherence theory of truth is sometimes blurry: if inquiry is governed by norms that ensure coherence among one’s beliefs, for example, a pragmatist who takes truth to consist in being accepted at an ideal end point of such inquiry will effectively agree with a coherentist who takes the truth-fixing set of beliefs to be the one that would be held at that ideal end point. This is why positions like the one defended in Putnam 1981 are sometimes classified as versions of coherentism, e.g. by Young (2018, § 1).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

116     (1878, 300). Like the constructivist, Peirce thus reverses the representationalist’s order of explanation between truth and reality, explaining what exists in terms of what is true. In a recent series of papers, Dale Dorsey defends a coherence theory of truth for the domain of ethics (2006; 2010) and argues that adopting such a theory of truth would help the constructivist avoid a thorny problem (2012). In what follows, I will pursue a different path and present a version of the pragmatist theory of truth that I think the constructivist should adopt. This is not because I take Dorsey’s proposal to be mistaken, but because I think that a pragmatist approach to truth and reality is more congenial to constructivism and that the theory I will present fits better with the constructivist’s account of the nature of reason judgements presented in Chapter 3. Moreover, as we will see, the pragmatist theory shares those features of Dorsey’s coherentism that help avoid the problem Dorsey (2012) identifies. In any case, as far as assessing the prospects of the constructivist account of practical reasons is concerned, establishing that there are two rather than just one viable option for the constructivist’s account of truth is certainly a relevant result on its own.

4.3 Wright on truth The classical pragmatist theory of truth, particularly the version proposed by Peirce, faces a number of serious problems.⁹ Firstly, Peirce’s definition of the truth as ‘[t]he opinion which is fated to be ultimately agreed to by all who investigate’ (1878, 300) entails that, at some point, there will be sufficient justification available to inquirers for accepting every true judgement. But it seems that there might well be true claims that are forever beyond the reach of our epistemic capacities, such as truths concerning events in the distant past. Hence, the epistemic constraint that Peirce’s definition imposes on the truth of a judgement is implausible for some classes of judgements. A second problem concerns the very intelligibility of the account. In particular, the idea of ideal ⁹ For critical discussions of Peirce’s pragmatism about truth, see, e.g., Kirkham (1992, 79–87) and Künne (2003, 393–9).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

117

conditions—such as those at the supposed end of inquiry—that ensure both the completeness and the infallibility of the judgements accepted under those conditions¹⁰ is of questionable coherence: applying the account to the judgement that the ideal conditions are met will always yield that it is true. In other words, the account entails that we have already reached the end of inquiry (cf. Wright 2000, 314–42; and Plantinga 1982). These problems pose a significant challenge for Peirce’s proposal, but they do not undermine the pragmatist approach to truth in general. This is witnessed by the neo-pragmatist conception of truth developed by Crispin Wright in Truth and Objectivity (1992) and elsewhere.¹¹ On this conception, a judgement’s truth does not depend on whether it would be justified under certain ideal conditions, but rather on whether justification will at some point be available and whether some of that justification will remain undefeated upon arbitrarily close and extensive scrutiny: Rather than ask whether a statement would be justified at the limit of ideal empirical investigation, or under ideal empirical circumstances, whatever they are, we can ask whether an ordinary carefully controlled investigation, in advance of attaining any mythical limit, justifies the statement, and whether, once justified, that statement continues to be so no matter how much further information is accumulated. (1992, 47)

¹⁰ That is, the ideal conditions ensure that the judgements made under those conditions include all and only true judgements. ¹¹ Prior to Wright, Hilary Putnam has proposed an account of truth that takes cues from Peirce’s account but attempts to steer clear of its problems. Putnam suggests that truth should be understood as ‘an idealization of rational acceptability’, but instead of the Peircean idea that a judgement is true if and only if it would be rationally acceptable under certain unique ideal circumstances that are identical for all potentially true judgements, he suggests that it only has to be rationally acceptable if epistemic conditions were sufficiently good for appraising that particular judgement (see 1981, 55; and 1990, vii–viii). By relativizing the ideal conditions to the judgement whose truth is in question, Putnam’s proposal thus avoids a commitment to the dubious idea of circumstances that ensure completeness and infallibility in one’s judgements, which gave rise to the second problem presented in the main text. But it still faces the first problem because it presupposes that no truth ‘totally outruns the possibility of justification’ (1990, ix)—a restriction Putnam himself later came to reject (see 1995, 293–5).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

118     Thus, Wright’s proposal is continuous with earlier versions of pragmatism about truth in that it understands a judgement’s (or a statement’s) being true in terms of having a certain justificatory status, but he idealizes that status by requiring it to be especially robust, rather than by requiring it to be present in certain counterfactual circumstances. Wright calls this particularly robust form of justification ‘superassertibility’ and characterizes it more explicitly as ‘the property of being justified by some (in principle accessible) state of information, and then remaining justified no matter how that state of information might be enlarged upon or improved’ (1992, 47).¹² Lynch (2009, 38) uses the term ‘superwarrant’ for this property, which strikes me as more useful and which I will therefore adopt in what follows. How does this neo-pragmatist conception of truth fare with respect to the two problems raised at the beginning of this section? Since Wright dispenses with the idea of an ‘end of inquiry’ or any other circumstances that guarantee the completeness and infallibility of one’s judgements, he avoids the second problem. But what about the first one? It is clear, and Wright does not deny, that understanding a judgement’s truth in terms of a particularly robust form of justification still entails that its truth is epistemically constrained—such truths can obviously not be beyond the reach of our capacities to gain justification for them. Wright, however, suggests that we can steer clear of the problems associated with such a constraint by limiting the scope of this conception of truth to classes of judgements where such a constraint is plausible on independent grounds. Wright’s example of choice is discourse about the comic: ‘there seems no sense to be attached to the idea that the comedy of a situation might elude the appreciation even of the most fortunately situated judge’ (1992, 58; see also 1995, 217). Limiting the scope of the superwarrant conception of truth in this way is possible for Wright because it is only a part of his overall account of truth. That account consists of an ‘analytic theory’ of the concept of truth which applies globally, that is, independently of what kind of judgement that concept is applied to (1999, 272). Such a theory can be formulated ¹² Of course, to say that a judgement is justified in a certain situation is not meant to entail that someone actually makes that judgement, but only that justification is available for it.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

119

by providing what Michael Smith calls a ‘network analysis of the concept of truth’. This analysis proceeds by accumulating various platitudes concerning truth, which form a network that connects truth to various other properties, and then defines truth as the property that simultaneously satisfies all (or sufficiently many) of these platitudes (Smith 1994, 44–7; see also Lewis 1970; and Jackson 1998). The platitudes about truth that Wright mentions include the following: (i) To assert is to present as true. (ii) Any truth-apt content has a significant negation which is likewise truth-apt. (iii) To be true is to correspond to the facts. (iv) A statement may be justified without being true, and vice versa.¹³ The theory of truth provided by this analysis is minimal¹⁴ in the sense that it does not include, as part of the concept of truth, that truth is necessarily to be understood in terms of accurate representation (when this is distinguished from the correspondence platitude (iii)) or, for that matter, in terms of superwarrant. Neither the representationalist theory of truth nor a pragmatist one can be established by an analysis of the concept of truth. Crucially, the global characterization of truth provided by this analytic theory can be met by different properties, and a conception of truth (such as the neo-pragmatist one) tells us which property (such as superwarrant) plays the truth-role described by the analytic theory for a particular domain of judgements (such as judgements about the comic). Wright is thus a pluralist concerning the properties that the concept of truth picks out in different domains of judgement. Consequently, different conceptions of

¹³ These platitudes are listed in Wright (1992, 34). Slightly extended lists can be found in Wright (1999, 271–2) and Wright (2001, 760), which also include Timelessness (‘if a proposition is ever true, then it always is, so that whatever may, at any particular time, be truly asserted may—perhaps by appropriate transformations of mood, or tense—be truly asserted at any time’) and Absoluteness (‘there is, strictly, no such thing as a proposition’s being more or less true; propositions are completely true if true at all’). ¹⁴ Note that Wright uses ‘minimalism’ to refer to a different theory of truth from, for example, Horwich (see n. 3 above). In particular, Wright’s minimalism about truth is not meant to be a form of deflationism (cf. 1992, ch. 1; 1999, 248–61).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

120     truth can hold for different classes of judgements. If defensible, this alethic pluralism allows Wright to avoid the first problem raised for Peirce’s pragmatist theory at the beginning of this section. Wright’s approach to truth thus promises to be an attractive option for constructivists: accepting it would allow them to adopt a pragmatist conception of the truth of reason judgements which meets the two criteria for compatibility with the constructivist’s project we identified in section 4.1, while avoiding the problems of other versions of pragmatism. However, the pluralist framework that Wright suggests is hardly uncontroversial and comes with its own set of problems. These will be discussed in section 4.4, where I will suggest that avoiding them requires a slightly different version of alethic pluralism from the one Wright proposes. Moreover, the conception of truth in terms of superwarrant presented above will require some amending in order to be applicable to the case of reason judgements. This will be the subject of section 4.5.

4.4 Alethic pluralism Before discussing alethic pluralism, there is a dialectical issue that needs to be addressed. In section 4.3, I argued that some of the problems faced by traditional pragmatist conceptions of truth can only be avoided by accepting such a conception as part of a pluralist approach to truth. But that approach is itself controversial. So why, one might ask, should constructivists opt for a pragmatist conception of the truth of reason judgements and incur the additional theoretical costs associated with alethic pluralism, instead of choosing an alternative account, such as the coherence theory of truth mentioned in section 4.2? The answer to this question is that whatever costs are associated with alethic pluralism, they are not additional costs for constructivists, because constructivists are committed to alethic pluralism anyway. After all, constructivism is meant to be a local theory that contrasts the nature of reason judgements with that of ordinary beliefs and the status of facts about what is a reason for doing what with that of, for example, facts about the number of columns of the Brandenburg Gate or about the weather in Scotland. The constructivists’ motivation for rejecting Representationalism and the

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

 

121

associated theory of truth is only concerned with a specific domain of judgement. They do not deny the adequacy of the representationalist picture for scientific judgements or those about the perceivable world.¹⁵ Since the dispute between constructivism and Representationalism involves a disagreement on the appropriate conception of truth, constructivists must thus endorse alethic pluralism in order to be able to confine their opposition to Representationalism to a specific domain. This is why Dorsey too proposes that constructivists accept the coherence theory only as part of a pluralist approach to truth (2012, 109).

4.4.1 Some problems of alethic pluralism After that clarification, we can turn to the pluralist position itself. The basic idea of alethic pluralism is that the question a pragmatist (or any other) conception of truth is supposed to answer admits of different answers in different domains of judgement. Different versions of alethic pluralism, however, disagree on how exactly that question should be understood. Wright’s overall account of truth comprises an analytic theory of the (unique) concept of truth and a potential multitude of conceptions of truth that tell us which property (like superwarrant) is picked out by that concept in a particular domain.¹⁶ On this version of alethic pluralism, ‘true’ functions similarly to a definite description like ‘the brightest object in the sky’.¹⁷ There is a constant element of meaning associated with that expression when it is used on different occasions; yet it might denote different objects—the moon, the sun, Venus, a meteor— depending on the context of utterance. Just as ‘the brightest object in the sky’ picks out different objects in different contexts, the account suggests, ‘true’ picks out different properties for different domains of judgement.¹⁸ ¹⁵ This distinguishes constructivism from the global anti-representationalism advocated, e.g., by Robert Brandom and Huw Price (see Chapter 1.2). ¹⁶ For more detailed presentations and defences of his pluralist account of truth, see Wright (1992, 1999, 2001, and 2013). ¹⁷ This helpful analogy is due to Michael Lynch (2009, 61). ¹⁸ Wright’s version of alethic pluralism should not be confused with what Michael Lynch (2009, 54) calls ‘simple alethic pluralism’, according to which the word ‘true’ has no unique meaning but is systematically ambiguous in the way ‘bank’ or ‘slip’ is, that is, according to which

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

122     This ‘one concept, many properties’-pluralism, however, faces several problems. The first has been raised forcefully by Christine Tappolet (1997). Consider the following inference: (1) Wet cats are funny. (2) This cat is wet. (3) Therefore, this cat is funny.¹⁹ It seems obvious that this is a valid inference. But according to a very common way of understanding validity, valid inferences are truthpreserving (cf. Beall et. al 2019). The two premises of this inference, however, belong to domains for which, according to Wright, ‘true’ picks out different properties. But then, what feature is preserved by the inference from (1) and (2) to (3)? According to ‘one concept, many properties’-pluralism, the only thing that is preserved by such ‘mixed’ inferences is the applicability of the concept of truth. But is validity helpfully understood as preservation of falling under the concept of truth? To many, I take it, this will seem like an awkwardly indirect account of validity.²⁰ Hence, Wright’s version of alethic pluralism appears to conflict with ordinary accounts of what it is for a mixed inference to be valid. This is the problem of mixed inferences. A second and related problem is known as the problem of mixed compounds (Williamson 1994; Tappolet 2000). Consider, for example, the conjunction: (4) This cat is funny and it is wet. Its conjuncts belong to domains for which ‘true’ picks out different properties. But what property does ‘true’ refer to when it is applied to the whole conjunction? Does it refer to superwarrant, as it does when

it means different things on different occasions. Simple alethic pluralism faces all the problems raised for Wright’s version of pluralism as well. ¹⁹ This example is from Tappolet (1997, 209). ²⁰ Wright himself is prepared to bite that bullet and accept such an account of validity (2013, 133).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

 

123

applied to the left conjunct, does it refer to representational accuracy, as it does when applied to the right conjunct, or does it refer to an entirely different property? In general, if the truth of one conjunct does not consist in the accurate representation of some fact (but perhaps in superwarrant), then it is implausible that the truth of the whole conjunction consists in representational accuracy. But the other conjunct, whose truth does consist in accurate representation, might fail to be superwarranted; there might not even be any warrant available for its truth at all. In that case, the conjunctive claim will not be superwarranted either, since a conjunctive claim is only warranted if both conjuncts are. Hence, in addition to the validity of mixed inferences, Wright’s version of alethic pluralism faces significant difficulties in accommodating the idea that a conjunction owes its truth to the truth of its conjuncts. A third problem for Wright’s view arises from the fact that when we are considering the truth of various judgements, we can easily generalize over judgements from different domains. Suppose, for example, that the Pope has opinions on what is funny as well as on past and present events and the colour of his shoes, and consider: (5) All of the Pope’s judgements are true. If ‘true’ picks out a different property in those domains, it is not clear what property we ascribe by such general endorsements. But on the face of it, we have no problem understanding them as straightforward and unambiguous claims about a range of different judgements.

4.4.2 The two-level approach These problems show that the fundamental challenge for any plausible version of alethic pluralism is to offer a sufficient degree of unity in its account of truth in order to accommodate the intuition that despite all differences, truths from different domains have something in common. Moreover, they suggest that this unity is not merely conceptual: there is a property shared by all true judgements, a property that we ascribe in generalizations like (5), and that is preserved by valid inferences

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

124     like (1)–(3). A more plausible version of alethic pluralism will hence not only claim that ‘true’ univocally expresses a single concept but also admit that this concept picks out a single property, independently of the kind of judgement to which it is applied. The pluralism only comes in at a second, lower level: underlying the single property shared by true judgements of different domains is another property, and this property might vary from domain to domain. The main question such a two-level approach must answer concerns the relation between the two levels. What does it mean to say that a property like superwarrant or representational accuracy underlies the truth of judgements in a certain domain? Several answers to this question are being discussed in the literature on alethic pluralism.²¹ The version of two-level alethic pluralism that I favour has been developed by Douglas Edwards (2011; 2013), who calls it simple determination pluralism. Edwards suggests that while there is only one property that satisfies the platitudes that form the analytic theory of the concept of truth, there are different ways of being true, depending on the domain in question, and these ways include being representationally accurate and being superwarranted. More precisely, a careful examination of a certain domain will allow us to formulate biconditionals that instantiate the following schema: (cf. Edwards 2013, 116–18). Depending on the domain, F might stand for properties like representational accuracy or superwarrant. Importantly, according to Edwards these biconditionals have an order of determination: their left-hand side depends on their right-hand side in that the judgements in the domain in question are true because they have the property mentioned on the right-hand side (see 2011, 43–4; and 2013, 118), that is, properties like representational accuracy or superwarrant ‘explain how propositions in the domains in question come to be true—they are true because they possess the relevant truth-determining property for their domain’ (2011, 45). It seems to me that the determination relation between a judgement’s truth and a particular way of being true that Edwards has in mind here is best understood as an instance of ²¹ For an overview, see Pedersen (2012), Pedersen and Wright (2018), and the papers collected in Pedersen and Wright (2013).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

 

125

the grounding relation introduced in Chapter 1.5: to say that a certain judgement is true because it is representationally accurate, or that its truth is determined by its representational accuracy, is to say that the fact that this judgement is true is grounded in the fact that it is representationally accurate. This version of alethic pluralism can be helpfully illustrated by considering the analogy between truth and winning, that is, between the property of being true and that of being a winner.²² On the one hand, there is a wide variety of activities—games and contests—that a person can win, and the winner will be determined in very different ways in the various activities. Winning a game of chess requires you to achieve something very different from what is required for winning a 100-metre race. A person wins the race if and only if and because she crosses the finishing line first, but she wins the game of chess if and because she checkmates her opponent. On the other hand, the property of being a winner does not appear to differ from activity to activity. There is something that the winner of a game of chess and the winner of a race have in common, something over and above whatever is required for winning in the respective activities. If they meet the day after they each competed in a tournament and one tells the other that she won yesterday, the other may straightforwardly reply ‘Me too!’ without causing any puzzlement. Hence, being a winner is a distinct and invariable property across all activities. The only thing that varies is what it takes to be a winner in the respective activity, that is, the facts that ground and explain the fact that this person rather than someone else has won in a particular activity. Similarly, truth is the same distinctive property in different domains, but what it takes to be true, what grounds the fact that certain judgements are true and explains why they are true, may differ significantly across different domains. Just as there are different ways of winning in different games and contests, there are different ways of being true in different domains of judgement.

²² The idea that we can learn something about truth by comparing it to winning goes back at least to Dummett (1959, 142); it also is a core thesis of Edwards (2013).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

126    

4.4.3 Distinguishing domains of judgement This version of alethic pluralism provides us with a plausible account of how truth or its grounds can vary across different domains of judgement while avoiding the problems faced by Wright’s ‘one concept, many properties’ account and other approaches to alethic pluralism that deny that there is a single, invariable property of truth. There is, however, an important issue that advocates of this (and any other) version of alethic pluralism have yet to address: how exactly do we distinguish different domains of judgement? How do we decide which domain a particular judgement belongs to and hence which property it is required to have in order to be true? One option is to distinguish the domains in terms of their subject matter. Then, a judgement belongs to a certain domain in virtue of its content. Consequently, what grounds a judgement’s truth too depends on its content.²³ This answer, however, threatens the account’s status as a pluralistic conception of truth. The worry is this: judgements from the same domain obviously differ in content too. Some perceptual beliefs are about trees, while others are about chairs, for example. These differences in content, however, do not suffice to establish the kind of variability in how their truth is grounded that the pluralist claims is present in the case of judgements from different domains. Hence, we need an account of what distinguishes (a) differences in content that merely separate different judgements of the same domain from (b) differences in content that mark different domains of judgement and hence result in the truth of two judgements being grounded in different properties (like representational accuracy and superwarrant). Without such a distinction, the differences in what grounds truth—that is, the differences on the lower level of our two-level conception of truth—might simply dissolve into ordinary differences in content.²⁴ To put the point differently: if both individual judgements and domains of judgements are distinguished in ²³ This seems to me to be what Michael Lynch suggests when he claims that domains can be individuated in terms of ‘the kind of concepts’ that compose the content of the judgements in that domain (2009, 79–80). ²⁴ A similar objection can be found in Frank Jackson’s (1994) and Philip Pettit’s (1996) discussions of Wright’s pluralism about truth.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

 

127

terms of content, then why do we not end up either with as many domains as there are judgements with different content or with one domain that includes judgements with all kinds of content, neither of which seems to amount to an interesting pluralism about truth that needs to be distinguished from the obvious pluralism about content? The results of Chapter 3 provide us with better resources to demarcate the different domains of judgement. There, we distinguished reason judgements from, for example, perceptual beliefs in terms of their functional role in an agent’s psychology. Whereas the latter are the paradigm example for mental states with a representational function, I argued that reason judgements should be characterized in terms of a different function in order to accommodate their role in practical reasoning. These functional differences offer a better criterion for distinguishing domains of judgement, one that does not face the problem raised by a contentbased approach. Moreover, a function-based criterion also helps us to explain why the truth of judgements in these different domains is grounded in different underlying properties. There seems to be a close connection between the truth of a judgement and the fact that it successfully performs its distinctive function.²⁵ This is why the representationalist theory of truth is most plausible for judgements whose cognitive function is best captured by a representationalist account, such as perceptual beliefs: accurate representation simply is successful representation, that is, it is the successful performance of the cognitive function that is distinctive of mental states like perceptual beliefs. So, if different kinds of judgement can differ in their cognitive function and hence in what it takes for them to perform that function successfully, we should expect those differences to translate into a difference in what it takes for those different kinds of judgement to be true. To conclude, truth is a single property whose instantiations can be multiply grounded in different domains of judgement, depending on the cognitive function of the judgements in those domains. The pluralist approach to the cognitive functions of different kinds of judgement

²⁵ Note that this connection can at most establish a necessary condition for truth but not a sufficient one: many attitudes have a distinctive function that they can successfully perform without even qualifying as truth-apt.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

128     defended by the constructivist and a pluralist approach to the grounds of the truth of those judgements thus go hand in hand.

4.5 Truth and sound reasoning Let me take stock. Constructivists must provide an account of the truth of reason judgements, both because their view should accommodate the truth-aptness of those judgements and because they propose to explain reason facts in terms of the truth of the corresponding reason judgements. Given the role truth plays in their overall picture, the account they provide must characterize it as a substantive property, but not as one that presupposes the prior existence of the reason facts that correspond to true reason judgements. This excludes a representationalist theory as well as a deflationist approach to truth. Among the remaining options that meet the constructivist’s criteria, I suggested that they should follow Crispin Wright’s proposal, both in adopting a pluralist approach to truth and in characterizing the truth of reason judgements in terms of superwarrant. As we will see in this section, however, ensuring that such an account of truth is compatible with the constructivist’s overall approach requires some additional clarifications and amendments. The account of truth that I will ultimately propose on behalf of the constructivist thus goes beyond what Wright has put forward in his writings, but it remains thoroughly Wrightean in spirit.

4.5.1 From superwarrant to sound reasoning As we saw in section 4.1, an important requirement for a theory of truth to be compatible with constructivism is to avoid introducing a vicious circularity into its account of practical reasons. One way of introducing such circularity would be for constructivists to understand the truth of reason judgements in terms of accurate representation. But there are two more ways of introducing a problematic circularity. Firstly, one might worry that the proposal to ground truth in something like superwarrant

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

   

129

is itself circular. That is because, on a quite natural understanding of the term, one is warranted in asserting a statement or accepting a claim when one has evidence for it. But if we understand evidence as something that indicates that the claim is likely to be true, we ultimately explain a judgement’s truth in terms that again refer to its truth.²⁶ Secondly, even if the notion of superwarrant need not itself be spelled out in terms of truth, it might turn out that it does presuppose the prior existence of reason facts when it is applied to reason judgements and the specific norms that govern their acceptance. The resultant circle in the constructivist’s account of practical reason would then be wider, but no less vicious. To put these worries to rest, constructivists need to say more about what is involved in a reason judgement’s being superwarranted, that is, about the status in which they want to ground the truth about practical reasons. In general, being superwarranted requires that warrant for a judgement is available, and that this warrant will remain undefeated upon (arbitrarily close) scrutiny. To determine whether a reason judgement meets these requirements, it is helpful to look at the episode of reasoning that the judgement in question is apt to guide. This is in line with the pragmatist approach underlying the Wrightean conception of truth, which aims to characterize truth in terms of the truthbearer’s status in a certain activity, traditionally its (ultimate or persistent) acceptability in inquiry. Now, if the constructivist account of the nature of reason judgements presented in Chapters 2 and 3 is on the right track, then reason judgements play a crucial role in a very particular kind of activity, namely practical reasoning. This suggests that the standards of their acceptability derive from the standards governing that activity. Indeed, discovering that the episode of reasoning that a particular reason judgement is apt to guide would be incorrect is sufficient to undermine any warrant for that reason judgement that one might otherwise have. Imagine a military pilot who is about to reason from the intention to bomb an ammunition factory and the belief that bombing the factory will also destroy the dictator’s palace to the intention to ²⁶ Richard Kirkham (1992, 50–1) raised a similar objection against Peircean Pragmatism and other proposals to understand truth in epistemological terms.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

130     destroy the palace. Once she realizes that reasoning in this way would be incorrect, she is no longer warranted in taking the fact that bombing the factory will also result in destroying the palace to be a reason to destroy the palace. Thus, one necessary condition for a reason judgement to be superwarranted is that the episode of reasoning it is apt to guide would be correct. But an episode of reasoning might also be flawed in other ways that undermine one’s warrant in the guiding reason judgement. Whereas correctness is concerned with the transition from the premiseattitudes to the conclusion-attitude, those flaws concern the premiseattitudes themselves. One such flaw is the falsity of a premise-belief. Reasoning from the belief that Jack needs your help to the intention to help him might well be correct. Nevertheless, realizing that Jack does not actually need your help undermines any warrant you might have for thinking that Jack’s need for help provides you with a reason to help him. Arguably, premise-intentions can also be flawed in a way that undermines warrant. Consider the following example: Poison Linda intends to kill her flatmate without leaving any clues. She also believes that the only way for her to do so is to put poison in his food. Hence, instrumental reasoning allows Linda to form the intention to put poison in her flatmate’s food. We can suppose that Linda’s reasoning is correct and that her premisebelief is true. Nevertheless, it appears that her reasoning is flawed because it starts from an intention that, we may assume, is impermissible. Linda has no reason to kill her flatmate, and plenty of reason not to do it. Further, suppose that Linda comes to realize this. That too seems to undermine her warrant for the guiding reason judgement in support of poisoning her flatmate. Thus, scrutinizing the episode of reasoning that a reason judgement is apt to guide allows us to identify various defeaters of our warrant for it. One the other hand, if we know that an episode of reasoning is correct and that none of its premise-attitudes exhibits any of these flaws—that is, if we know that the episode of reasoning is sound, as I will put it from

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

   

131

now on—then our warrant for the guiding reason judgement is immune to defeat. That is not to say that there are no other sources of defeat than the ones just mentioned. But these other sources succeed in defeating our warrant in a reason judgement only if they succeed in undermining our knowledge that the episode of reasoning is sound. Moreover, that the episode of reasoning that a reason judgement is apt to guide would be correct is also a positive source of warrant for that judgement. Suppose you intend to stay dry on your way to work. That reasoning from this intention and the belief that the only way to stay dry is to take an umbrella to the intention to take an umbrella would be correct provides you with (albeit defeasible) warrant for the guiding reason judgement in favour of taking an umbrella. That you can reason correctly to an intention warrants you in judging that you have a reason in favour of it. Recall that for a reason judgement to be superwarranted is for it to meet two conditions: (i) warrant for that judgement is available, and (ii) this warrant will remain undefeated upon scrutiny. We have just seen that both of these conditions can be verified by attending to the episode of reasoning that the judgement is apt to guide. Determining that this reasoning would be correct provides us with the warrant required by (i), and determining that, in addition, the reasoning’s premise-attitudes do not exhibit any flaws either, allows us to conclude that condition (ii) is met too. This allows constructivists to give a more informative account of what superwarrant amounts to in the case of reason judgements: for a reason judgement to be superwarranted is for the episode of reasoning it is apt to guide to be sound. In other words, they can characterize the ground of a reason judgement’s truth in terms of sound reasoning.

4.5.2 Avoiding circularity What does that entail for the constructivists’ prospects for dispelling the worries about circularity? That depends on whether they can give an account of what it is for an episode of reasoning to be sound that does not presuppose the prior existence of any facts about practical reasons. Doing so requires, first, an account of what it is for an episode of reasoning to be correct that does not rely on any such prior facts. It

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

132     must not, for example, explain the correctness of such an episode in terms of one’s reasons for or against reasoning in that way. Nor, of course, should constructivists propose that an episode of reasoning is correct because the content of the premise-attitudes provides a reason for the conclusion-attitude. In Chapter 5, I present an account of correct reasoning that meets these requirements. In a nutshell, I propose that reasoning is an activity that is governed by constitutive rules similar to the rules of games like chess, that these rules are not normative in the sense of presupposing any reasons to comply, and that for an episode of reasoning to be correct is for it to conform to these rules. An elaboration and defence of these ideas, however, have to wait until Chapter 5. Even if correct reasoning can be explicated without reference to reasons, however, the constructivist’s account avoids circularity only if the conditions for the premise-episodes to qualify as flawless do not ultimately depend on any truths about practical reasons either. I take it to be obvious that, except in special cases, the truth of premise-beliefs does not depend on any truths about practical reasons. The flawlessness of premiseintentions, on the other hand, is a less clear case. In section 4.5.1, I suggested that a reasoner’s premise-intention, such as Linda’s intention to kill her flatmate, might qualify as flawed if she has no reason to have it. But then, for a premise-intention to qualify as flawless, there has to be some reason in support of it. In that case, the soundness of an episode of reasoning and hence—according to constructivism—the truth of the guiding reason judgement depends on the truth of another reason judgement. Now, it should not surprise us that the truth of one reason judgement sometimes depends on the truth of another one. This will be the case whenever one reason is derivative of another (e.g. via some means-end relation). What constructivists must ensure, though, is that such a dependency does not hold for all reason judgements. Otherwise, their conception of the truth of reason judgements leads them into an infinite regress. Whether it does depends on whether or not there are forms of correct practical reasoning whose premise-attitudes do not include any intentions. If there are, then any potential regress could be stopped.²⁷ But what ²⁷ To see this, suppose that the episode of reasoning that the reason judgement r₁ is apt to guide is one whose premise-attitudes include an intention, i₁. On the assumption that a premise-

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

   

133

if all correct forms of reasoning have an intention among their premiseattitudes? In that case, I think we have good reason to reconsider our assessment of cases like Linda’s. The suggestion that sound reasoning requires premise-intentions that are supported by reasons is driven by the intuition that agents like Linda do not have a reason to take the means to their unsupported ends. On this picture, the fact that an agent actually pursues an end does not generate a reason to take the means. Rather, what generates such a reason is the fact that the agent has a reason to pursue to end. But if an agent’s reasons are independent of her actual intentions in this way, then our capacity to reason correctly would be systematically insufficient to enable us to comply with our reasons if it was limited to forms of reasoning whose premise-attitudes necessarily include intentions. For if all correct practical reasoning had to start from an intention, then an agent could not come to comply with a reason that is entirely independent of her actual intentions by reasoning correctly from her acknowledgment of the reason-giving considerations. But, it seems plausible to assume, it must at least in principle be possible for an agent to comply with her reasons by way of correct reasoning (we will get back to this condition in section 4.5.3). Hence, either an agent’s actual intentions at least sometimes make a difference to her reasons or correct practical reasoning is not limited to episodes whose premise-attitudes include an intention. In other words, the rationale underlying our initial assessment of cases like Linda’s does not square well with the presupposition that all forms of correct reasoning include a premise-intention. If we resolve that conflict by dropping the presupposition, the regress will eventually stop. If, on the other hand, we resolve that conflict by dropping the requirement that premise-intentions must be supported by reasons in order to be flawless, the regress does not even get started. In either case, there is no cause for concern for the constructivist. intention is flawless only if it is supported by some reason, the truth of r₁ then requires the truth of another reason judgement, r₂, in support of i₁. What about the episode of reasoning that r₂ is apt to guide? That episode is either another episode with a premise-intention, in which case the truth of r₂ requires the truth of some r₃. Or it is an episode whose premise-intentions include only beliefs. Then, the truth of r₂ does not require the truth of any other reason judgement. And if the regress does not stop with r₂, then it might stop with r₃, or r₄, etc. As long as there are episodes of deliberative reasoning that do not require premise-intentions, such regresses need not be infinite, so there is no cause for concern.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

134    

4.5.3 The deliberative constraint Let us grant, for now, that a Wrightean conception of truth can be spelled out and integrated into the constructivist’s overall account in a way that avoids any problematic circularity. Still, it provides a suitable conception of the truth of reason judgements only if that domain of judgement meets conditions which ensure that the truth in that domain cannot outrun our abilities to obtain warrant (see section 4.3). Wright argues for the applicability of his account to discourse about the comic by pointing out that ‘there seems no sense to be attached to the idea that the comedy of a situation might elude the appreciation even of the most fortunately situated judge’ (1992, 58; see also 1995, 217). What we need to show next is that the domain of judgements about practical reasons is subject to a similar constraint. Recall the connection between a reason judgement’s warrant and sound reasoning established in section 4.5.1. In light of that connection, the requirement that the truth of reason judgements does not outrun our ability to obtain warrant translates into the requirement that the truth of reason judgements does not outrun our ability to reason soundly. That truths about practical reasons are constrained in this way, however, seems to be a truism on a par with Wright’s claim about comedy. Indeed, the idea of a reason that reasoners could not recognize as reason-giving is as nonsensical as the idea of an undetectably funny joke (cf. Lillehammer 2003). Thus, consider: The Deliberative Constraint If some consideration F is a reason for an agent S to do A, then it is possible for S to reason soundly from the acknowledgment of the reason-giving consideration F to the intention to do A. This constraint on what can be a practical reason seems to capture something utterly obvious and trivial: reasons are for reasoning with, that is, they are the things that we think about when we are reasoning well. It is thus no surprise that a constraint like this is widely accepted in

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

   

135

the literature.²⁸ Like any philosophical claim, however, it has not remained unchallenged. In particular, one might worry that the constraint excludes reasons that the agent cannot respond to by way of reasoning. This might be the case, firstly, if the circumstances that provide the reason also make it unfeasible for the agent to engage in any reasoning, for example because there is not enough time or because the circumstances impair her ability to reason correctly by causing her acute distress. But note that the Deliberative Constraint only requires that it is possible for the agent to engage in the reasoning, not that she is able to do so whenever she is confronted with that reason. Moreover, the sound reasoning required by the constraint can be hypothetical reasoning (see Chapter 3.4). That is, if the fact that an agent is in circumstances C gives her a reason to do A, but being in circumstances C also excludes engaging in reasoning to the intention to do A, the agent might nevertheless come to do A in C by acting on an intention that she has acquired as the conclusion of an episode of sound reasoning. For before she actually faced those circumstances—in less hectic circumstances and with a cooler mind, for example—she might have engaged in hypothetical reasoning about what to do in circumstances C, and she might have concluded that reasoning by forming a standing intention to do A should circumstances C arise.²⁹ A more serious challenge to the Deliberative Constraint is posed by cases in which the consideration that seems to provide an agent with a reason does so only as long as the agent is not aware of that consideration. Mark Schroeder (2007, 33) illustrates this with the case of Nate, who loves successful surprise parties. While he is at work, his friends ²⁸ See, e.g., Williams (1980), Darwall (1983), Korsgaard (1986), Smith (1995), and Wallace (1999). Moreover, Hieronymi (2005), Schroeder (2007), Setiya (2007, 2014), Sinclair (2012), Silverstein (2016), Way (2017), McHugh and Way (2016b), Asarnow (2017), and Snedegar (2018) accept what Way calls the ‘reasoning view’, which identifies being a reason with being a consideration that figures as a premise in sound reasoning and which thus entails (but is significantly stronger than) the Deliberative Constraint. ²⁹ Note that putting such a conditional intention into action once the circumstances have arisen does not require another episode of conscious reasoning; it often occurs through an automatic and subconscious process. Even if the agent does not manage to act on that intention in the relevant circumstances, however, it will still be true that she might have acquired the intention as a result of sound reasoning from her acknowledgement of the reason-giving considerations as part of the (hypothetical) circumstances in question. This is sufficient for the Deliberative Constraint to be met.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

136     have prepared a surprise party for him at home. It seems that this provides him with a reason to go home. But that would not be a reason that Nate could consider in his reasoning, because once he becomes aware of the reason-giving consideration, it ceases to be reason-giving (he hates unsuccessful surprise parties).³⁰ Is Nate’s case a compelling counterexample to the Deliberative Constraint? I do not think so. It does not strike me as plausible that the fact that there is a surprise party waiting at home really is a reason for Nate to go home. Nate does not make a mistake if he decides not to home. Nor can he justify his decision to go home by adducing the enjoyable party, once he has been surprised by it. Doing so would be disingenuous. It is true that the party gives him a reason to be glad if he goes home, and to be dismayed if he does not (Setiya 2009, 538). Moreover, it would be good for Nate to go home, and so it makes sense for a friend to advise him to do so (Sinclair 2012, 657). But we can acknowledge all of this without claiming that the party also gives Nate a reason to go home. Considering the overall plausibility of the Deliberative Constraint, I thus think that this is the best account of what is going on in the case of Nate: the party does give him a reason to be glad or dismayed, depending on his decision, and it does give his friends a reason to advise him to go home because it explains why it would be good for him to do so. But it does not give Nate a reason to go home, and our initial impression that it does is best explained by our mistaking the former reasons for the latter one. Examples like Nate’s do not undermine the Deliberative Constraint. Hence, we can assume with confidence that the truth about practical reasons does not outrun our ability to engage in sound reasoning. This supports the suggestion that the conception of truth developed in the previous sections is indeed a suitable conception for the domain of practical reasons.

³⁰ Note that the possibility of hypothetical reasoning does not help here, because putting the conditional intention to go home if there is a surprise party waiting into action requires awareness of the waiting party too.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

 

137

4.6 Constructing reasons Let me assemble the pieces of the constructivist account of practical reasons that we have gathered so far. Reason facts—facts about what is a reason for doing what—do not exist independently of us and our mental capacities. Rather, they are an upshot of our practice of practical reasoning. More precisely, a reason fact obtains if and only if, and because, the corresponding reason judgement is true (Chapter 1.3). The function of such judgements in an agent’s psychology is not to indicate that a certain fact obtains; their cognitive role is to guide her practical reasoning (Chapters 2 and 3). That is why their truth is not grounded in representational accuracy, but in the soundness of the episode of reasoning that they are apt to guide (section 4.5). Taken together, this yields that something is a reason for an agent to perform a certain action if and only if, and because, the judgement that it is a reason for the agent to perform that action is apt to guide an episode of practical reasoning that is sound. Thus, according to constructivism, a consideration is a reason in virtue of being the starting point of sound reasoning.³¹ The idea that there is such a close connection between reasons and sound reasoning is widely shared; it has become known as the ‘reasoning view’.³² What is distinctive of constructivism, however, is how it arrives at that view, how it

³¹ Some argue that there are considerations that bear on whether there is a reason to perform a certain action but that do not themselves provide a reason for (or against) that action. Enablers or disablers, as Jonathan Dancy (2004, ch. 3) defines them, might be a case in point (though see Raz (2006) for a sceptical stance). Since those considerations would play a role in sound reasoning too, one might wonder how a proponent of the constructivist account can distinguish between a consideration that provides a reason for a certain action and a consideration that merely serves as an enabler. I think that this distinction can be drawn as follows. R is a reasongiving consideration if and only if a belief in R is among the premise-attitudes of a sound episode of reasoning. E is an enabler of R if and only if adding a belief in E to a set of premiseattitudes that already includes a belief in R does not undermine but is also not necessary for the soundness of the reasoning, whereas adding a belief in E’s negation does undermine soundness. In other words, what distinguishes enablers from reason-giving features is that sound reasoning may but need not consider them, whereas it is undermined by a belief in their negation. ³² That name is due to Jonathan Way (2017). Other advocates of the view are Pamela Hieronymi (2005), Kieran Setiya (2007; 2014), Matthew Silverstein (2016), and Samuel Asarnow (2017). Of course, not all of them share the account of reasoning developed in Chapter 2, nor do they all agree with the account of what correct reasoning consists in that I will propose in Chapter 5.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

138     conceives of the relations between reasons, sound reasoning, and true reason judgements. It is particularly important to be clear about what roles reason facts do and do not play in the account proposed here. Reason facts are what reason judgements are about. Those judgements say that certain considerations provide reasons for some agents to perform certain actions, and when they are true, then things are as they say they are, that is, the considerations then really do provide reasons for the agents to perform the actions. Potential facts concerning what is a reason for what thus form the content of reason judgements, and whether or not they actually obtain indicates whether or not the corresponding judgement is true. In this sense, they are a condition for the truth of those judgements. In other words, a judgement of the form ‘F is a reason for S to do A’ is true if and only if the consideration, the agent, and the action that are substituted for F, S, and A respectively stand in the relation that the concept of a reason picks out. But, and this is crucial, the judgement is not true because that condition of its truth is met, that is, the judgement is not true because the consideration, agent, and action stand in the reason-relation. Rather, it is the other way around: the consideration, agent, and action denoted by the respective elements of our reason judgements stand in the reasonrelation because that judgement is true. That means that this account reverses the usual order of explanation between truth and successful reference with which we are familiar from domains of judgement where a representationalist account is apt. On the reversed account, the extension of the concept of a reason is fixed by the truth or falsity of the reason judgements in which it occurs: a particular combination of consideration, agent, and action falls in the extension of that concept because the corresponding judgement is true, rather than vice versa. Michael Lynch agrees that this is a consequence of adopting a nonrepresentationalist conception of truth such as Wright’s for a certain domain: An essential difference between domains where truth is manifested by superwarrant and domains where it is manifested by something like [representational accuracy] is the order of explanation with regard to

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

 

139

denotation/reference and truth. . . . In short, in domains where truth is manifested by superwarrant, denotation is explained by truth. (2009, 147)

This reversal of explanatory orders (relative to a representationalist account, that is) has another important consequence: reason judgements are not about the facts that explain their truth. More precisely, the facts that form the content of (true) reason judgements come apart from the facts that ground and thus explain their truth, the facts that make them true, as it were. Reason judgements are about reason facts, they say that certain reason facts obtain, but they are not made true by those facts in virtue of accurately representing them.³³ Rather, their truth is grounded in the fact that the episodes of reasoning those judgements are apt to guide are sound. Hence, we must distinguish between the truth condition of a reason judgement, that is, the condition of which the judgement says that it is met and which thus gives us the content of the judgement, and the truthmaker of a reason judgement, that is, the fact or facts that ground its truth and that can hence provide an explanation for it.³⁴ This peculiarity of reason judgements is a consequence of their distinctive function. In section 4.4.3, I suggested that there is a close connection between a judgement’s truth and the fact that it successfully performs its cognitive function. Given such a connection, it is plausible that if the cognitive function of a certain kind of judgement is accurate

³³ Does that mean that reason facts cannot be the truthmaker of any representational attitude? After all, constructivists do not deny their existence, so why should it not be possible to accurately represent them, once they have been constructed? On the constructivist picture, the concepts and judgements that figure in our thinking about reasons do not serve a representational function, and we cannot simply add concepts or attitudes with particular functions to our mental equipment. So reason facts do not play the role of a truthmaker in our ordinary normative thinking. Nevertheless, I do not think that constructivists should deny that we can represent those facts indirectly, as it were, by using descriptions like ‘the facts whose existence is grounded in our reason judgements’. That is, reason facts might well be the truthmakers of certain metaphysical or meta-ethical claims. But while our reason judgements are also about those facts, they are not made true by them. ³⁴ Like Dorsey’s coherence theory, the pragmatist theory of truth thus allows constructivists to reject what Dorsey calls the ‘semantic theory of truth’, which holds that ‘a truth-bearer is true if and only if that truth-bearer’s meaning bears “the right relation” to that bit of the world, or state of affairs, that would make it true’ (2012, 101). As I explain in the text, a reason judgement’s meaning need not correspond to or mirror (which are two of Dorsey’s examples for what the ‘right relation’ might consist in) its truthmaker in order for it to be true.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

140     representation, that is, if its function is to say that things are a certain way whenever things really are that way, then the content or truth condition of such a judgement—the fact that it says obtains—also serves as its truthmaker (if it does obtain). But if Representationalism is false and reason judgements are mental states with a non-representational function, then the connection between truth and successful functioning suggests that the truth of those judgements is not grounded in the facts of which they say that they obtain either.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

5 Correct reasoning This chapter presents an account of what it is for an episode of practical reasoning to be correct. It thus complements the theory of practical reasoning that was presented in Chapter 2. More importantly, however, it provides an indispensable building block for the constructivist account of practical reasons developed in Chapter 4. According to that account, practical reasons are grounded, via the truth of the corresponding reason judgement, in the soundness of the relevant episode of reasoning. For an episode of practical reasoning to be sound, its premise-attitudes have to be flawless (Chapter 4.5.1). But soundness also requires that the reasoning be correct, that is, that the transition from the premise-attitudes to the conclusion attitude meets the standards of correctness that govern such reasoning. The concept of correct reasoning is thus a crucial component of the constructivists’ overall account. Moreover, grounding practical reasons in correct practical reasoning invites the worry that the resulting account of practical reasons is circular. This is because many share the assumption that ‘correctness’ is itself a notion that must be understood in terms of reasons. Constructivists, however, need to provide a conception of correct reasoning that does not itself presuppose any facts about what is a reason for doing what. The main aim of the first part of this chapter (sections 5.2 and 5.3) is to address this worry and to show that at least when it is applied to reasoning, the notion of correctness need not be understood in terms of reasons, and that the account of correct reasoning such an understanding would yield faces substantial problems. In the second part (sections 5.4–5.7), I will turn to the positive task of presenting an account of correct reasoning that does not presuppose the notion of a reason and is hence compatible with the constructivist project. That account characterizes correct reasoning as reasoning that complies with the constitutive rules of that activity, similar to the way in which the movement of a Constructing Practical Reasons. Andreas Müller, Oxford University Press (2020). © Andreas Müller. DOI: 10.1093/oso/9780198754329.001.0001

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

142   bishop in a game of chess is correct or incorrect depending on whether it complies with the constitutive rules of that game. I will elaborate the core elements of this account of correct reasoning and discuss the consequences of adopting it. Note, however, that my aim is not to argue that constructivists are committed to this particular way of understanding correct reasoning, but to show that there is at least one viable option that does not render a constructivist account of practical reasons circular. Hence, I will not spend much time comparing the account to potential alternatives.

5.1 Correct and incorrect reasoning Reasoning, as it was introduced in Chapter 2, is a conscious psychological activity by which an agent modifies her mental attitudes, such as her beliefs or intentions. Reasoning can lead her to drop certain attitudes or to form new ones. In the latter case, an episode of reasoning starts with a set of premise-attitudes and results in the formation of a conclusionattitude. You might, reason, for example, from the belief that if there is a red sky tonight, then there will be fair weather tomorrow, and the belief that there is a red sky tonight, to the belief that there will be fair weather tomorrow. Or you might reason from the belief that your friend John told you that it is going to rain tomorrow to the belief that it is going to rain tomorrow. Both of these, I take it, are examples of correct reasoning.¹ But episodes of reasoning are not always correct: people sometimes reason from the belief that if p then q and the belief that q to the belief that p, for example. This kind of reasoning by affirming the consequent is an example of incorrect reasoning. Practical reasoning can be correct or incorrect too. Giving examples, however, is less straightforward in this case, as the nature and form of correct reasoning are generally more contested in the practical than in the theoretical realm. I take it that instrumental reasoning—like reasoning from the intention to get to work before 9 a.m. and the belief that getting there before 9 a.m. ¹ Though only the modus ponens reasoning corresponds to a deductively valid argument, of course.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

   

143

requires taking the train to the intention to take the train—is the least controversial example of correct practical reasoning.² Other plausible examples include reasoning from the belief that you promised to do something to the intention to do it, and reasoning from the belief that doing something will cause you severe pain to the intention not to do it. For an example of incorrect practical reasoning, consider someone who reasons from the intention to take the train and the belief that taking the train is a necessary means for getting to work before 9 a.m. to the intention to get to work before 9 a.m. I hope you will agree that such reasoning to the facilitated end, as we might call it, is quite obviously incorrect—it gets instrumental reasoning backwards, much as affirming the consequent gets modus ponens reasoning backwards. Note that what is assessed as correct or incorrect in these examples is the transition of reasoning from the premise-attitudes to the conclusionattitude, not the result of that transition, that is, the conclusion-attitude. In general, the correctness of an episode of reasoning can be distinguished from the correctness of its result. That some episode of reasoning is correct or incorrect does not simply come down to its conclusion-attitude being true or false, justified or unjustified, consistent or inconsistent with the agent’s other attitudes, or whatever correctness or incorrectness might amount to when applied to beliefs or intentions. Of course, to give an informative account of correct reasoning, one has to go beyond listing particular examples of correct or incorrect reasoning. In particular, such an account should meet at least the following two criteria. First, it should tell us which kinds of reasoning are correct and incorrect, that is, it should allow us to determine for a particular episode of reasoning whether it is correct. In doing so, it should accommodate our pre-theoretical convictions about what counts as correct reasoning,

² If you think that instrumental reasoning is not a case of correct reasoning, then you can simply substitute it with one you do consider correct. The arguments in what follows do not depend on the choice of examples. Some people will accept that what we ordinarily call instrumental reasoning is a case of correct reasoning, but they will take issue with the gloss I gave of its structure and insist that such reasoning (along with any other kind of practical reasoning) involves attitudes with normative contents (e.g. judgements about what one ought or has reason to do) among its premise- or conclusion-attitudes. I have discussed and rejected such an account of the role of normative judgements in practical reasoning in Chapter 2.2.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

144   for instance concerning the examples given above, or give a plausible error theory of those convictions that it cannot accommodate. In other words, the criterion for correct reasoning such an account provides must be extensionally adequate. Secondly, the account should not just give us a criterion for deciding whether or not some episode of reasoning is correct but tell us something about why it is correct. That is, a satisfactory account of correct reasoning goes beyond identifying conditions that allow us to check particular episodes of reasoning for their correctness: it also tells us what correct episodes of reasoning have in common that makes them correct, what their correctness consists or is grounded in. Hence, a comprehensive account of correct reasoning has to be explanatorily adequate too. Importantly, giving an account of correct reasoning must be distinguished from what logicians are typically concerned with. Logic, as Gilbert Harman (1984; 1986, ch. 1) has pointed out repeatedly, is not the theory of reasoning. Logic is concerned with the validity of arguments and the question whether certain (sets of) propositions or sentences entail other propositions or sentences. In the logician’s sense, a valid argument is simply an ordered set of sentences where (at least) the last sentence is entailed by the ones preceding it. Reasoning, on the other hand, is a psychological process that involves the causal interaction of attitudes like beliefs and intentions. Logic is certainly relevant for the theory of correct reasoning, but that relation is not as straightforward as one might initially suspect either (see Field 2009; Harman 1984; 1986, ch. 2; MacFarlane 2004). Note that correct reasoning, even if it is concerned solely with beliefs, need not be based on a logically valid argument. For one thing, some forms of reasoning do not lead to the formation of a new attitude but to dropping an old one, in which case there is no conclusion-attitude whose content could be entailed by the content of the premise-attitudes. But even if an episode of reasoning leads to the formation of a new attitude, it is not necessary that such an entailment relation holds for the reasoning to be correct. After all, not all correct reasoning is deductive; there are certainly correct forms of inductive and abductive reasoning. Hence, the conditions for an episode of reasoning to qualify as correct cannot be simply read off from the principles of entailment which are described by logicians.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

145

The aim of this chapter is to present the outlines of a comprehensive account of correct reasoning. In keeping with the overall focus of this book, much of the discussion will be restricted to the case of correct practical reasoning, that is, reasoning that concludes with the formation of a new intention. As I said before, this restriction is not meant to suggest that theoretical and practical reasoning (or the corresponding reason judgements) require substantively different accounts. Indeed, it seems to me that most of the considerations in the following sections can be easily adapted to the case of theoretical reasoning. Still, I do not intend to take an official stance on these issues here, so I will not pursue them in what follows.

5.2 Correctness and reasons 5.2.1 The buck-passing account of correct reasoning What is it for an episode of practical reasoning to be correct? One way of approaching this question is to point out that reasoning is not the only thing that we assess in terms of correctness. Pronouncing a word, answering a question, reciting a poem, moving a chess piece, buttoning a shirt, jumping an axel—all of this, and much more, can be done correctly or incorrectly. It thus seems natural to try to give a general account of what correctness consists in and then apply it to the case of practical reasoning. One such general account of correctness has been proposed by Mark Schroeder. Schroeder starts from the assumption that ‘correctness is clearly a normative property’ (2007, 134). Like many other philosophers in the recent literature, he also believes that all normative properties can be analysed in terms of reasons. This leads Schroder to proposes that for it to be correct to do A is for the reasons in favour of doing A to outweigh the reasons against doing A (2007, 134). The general line of thought that leads Schroeder to his proposal—that correctness is a normative concept and must thus be understood in terms of reasons—has been influential in a number of contexts. Many philosophers of language, for example, argue that because, for an expression to be meaningful is for it to have

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

146   conditions of correct use, an adequate theory of meaning must be concerned with how speakers have reason to use the expressions of their language.³ This ‘argument from correctness’ (Glüer and Wikforss 2009, 34) presupposes the above connection between correctness and reasons. A popular position in the theory of value draws on that connection too. Its advocates argue that for something to be valuable is for it to be correct (or fitting) to value it, that is, to respond to it with a certain favourable attitude, like admiration. They then explicate this correctness relation between the attitude and its object in terms of reasons. It obtains if certain features of the object provide reasons in favour of the attitude. This leads them to propose what is widely known as the ‘buck-passing account’ of value, according to which an object is valuable in virtue of having other features that give us reasons to value it.⁴ Because of its kinship to this buck-passing account of value, I will call the general account of correctness that Schroeder proposes the buckpassing account of correctness: BPC For it to be correct to do A is for it to be the case that there is (at least) sufficient reason of the right kind⁵ to do A.⁶ What do we get if we apply BPC to the case of correct reasoning? Let me introduce some notation that allows for a more succinct statement of such an account of correct reasoning. We can refer to reasoning from the set of premise-attitudes Π, which includes the individual premiseattitudes π₁, . . . , πn, to the conclusion-attitude κ as ℛ(Π, κ).

³ Versions of this sort of argument can be found in Boghossian (2003b), Glock (2005), and Whiting (2007). For critical discussions, see Glüer and Wikforss (2009), Hattiangadi (2006; 2007), and Wikforss (2001). ⁴ The name is due to Scanlon (1998, 96), who is also its most influential proponent. ⁵ We will get back to the restriction to reasons ‘of the right kind’ in section 5.2.2. ⁶ In fact, Schroeder’s formulations above suggest that he endorses the stronger claim that there has to be decisive reason to do A in order for it to be correct (because the reasons in favour have to outweigh those against). I think that this is implausible, for it seems possible that there are situations in which both doing A and not doing A are correct. In any case, since I will reject Schroeder’s general account of correctness, at least when it is applied to the case of deliberative reasoning, it is certainly legitimate to attribute this weaker claim to him.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

147

‘ℛ(Π, κ)’ thus refers to a certain kind of reasoning that is individuated by its premise- and conclusion-attitudes and that can be instantiated by different reasoners at different times. An episode of reasoning is the psychological process of an agent who engages in reasoning of a certain kind on a particular occasion and in particular circumstances. I will refer to particular episodes of reasoning by using the somewhat cumbersome construction ‘to engage in ℛ(Π, κ)’ as an abbreviation for ‘to reason from the premise-attitudes π₁, . . . , πn to the conclusionattitude κ’. An account of correct reasoning is concerned with the correctness or otherwise of such episodes of reasoning. Hence, if correctness is connected to reasons in the way BPC suggests, then the correctness of engaging in ℛ(Π, κ) will depend on the reasons for or against engaging in that kind of reasoning. Remember, though, that the idea of a certain episode of reasoning being correct should be understood as establishing a permission, not a requirement—in general, an agent is not making a mistake, and thus not violating a requirement, if he fails to reason in a way that would be correct. Hence, the only requirements issued by an account of correct reasoning appear to be negative: reasoners should avoid reasoning incorrectly, and they make a mistake if they do not. This suggests that an account of correct reasoning that is based on BPR is best formulated as a condition for incorrect reasoning, taking it to be reasoning that an agent lacks sufficient reason to engage in and thus has decisive reason to avoid. That leaves open whether an agent ever has decisive (and not merely sufficient) reason to reason in a way that is correct. Such an account tells us that for an episode of reasoning to be incorrect is for it to be the case that there is decisive reason (of the right kind) against it. More precisely, and using the notation introduced above, an account of correct reasoning that is based on BPR, which I will call a buck-passing account of correct reasoning (BPCR) can be put as follows: BPCR Engaging in ℛ(Π, κ) is incorrect if and only if and because there is decisive reason (of the right kind) not to engage in ℛ(Π, κ). An episode of reasoning is correct iff it is not incorrect.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

148   This account of correct reasoning shifts the burden of determining the correctness of an episode of reasoning to an account of the reasons for or against engaging in such reasoning, just as a buck-passing account of value shifts the burden of determining whether an object is valuable to an account of the reasons for or against valuing that object. The account offers both a criterion (‘ . . . if and only if . . . ’) and an explanation (‘ . . . and because . . . ’) of the correctness or incorrectness of an episode of reasoning; it hence promises to meet the desiderata introduced at the beginning of this chapter. As I noted before, constructivists cannot accept BPCR, because the explanation it offers for the correctness of an episode of reasoning would render the constructivist account of reasons viciously circular—it explains reasons in terms of correct reasoning, which in turn would be explained in terms of reasons. The commitment to rejecting BPCR is not a disadvantage of constructivism, though. Both BPCR and BPC, the more general account of correctness underlying it, face a number of significant problems.

5.2.2 The wrong kind of reasons problem Like its cousin in the theory of value, the buck-passing account of correct reasoning faces what is known as the wrong kind of reasons problem.⁷ The problem is that in order to avoid implausible classifications of episodes of reasoning as correct or incorrect, advocates of the buckpassing account have to exclude some reasons for and against engaging in certain kinds of reasoning as reasons of the wrong kind. The source of this problem lies in the fact that what reasons an agent has for or against reasoning in a certain way depend on aspects of her circumstances that are irrelevant for the correctness or incorrectness of that reasoning. To illustrate the problem, consider the following case: ⁷ The wrong kind of reasons problem for buck-passing accounts of value was made prominent by Rabinowicz and Rønnow-Rasmussen (2004); earlier formulations of the problem are due to Crisp (2000b) and D’Arms and Jacobson (2000a). It has since been a topic of intense discussion: see, e.g., Olson (2004), Rabinowicz and Rønnow-Rasmussen (2006), Stratton-Lake (2005). For a helpful overview, see Gertken and Kiesewetter (2017).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

149

Evil Demon An evil and powerful demon threatens to inflict great pain on John if John reasons from his present intention to buy some bananas and his present belief that in order to buy bananas he has to go to the supermarket to the intention to go to the supermarket (an intention John does not yet have). It seems that the demon’s threat turns John’s potential reasoning to the conclusion-intention to go to the supermarket into an episode of reasoning that he has overall decisive reason to avoid. Yet, as an instance of instrumental reasoning, that episode of reasoning would, nevertheless, be correct.⁸ To accommodate such cases, buck-passers have to exclude reasons like those provided by the demon’s threat from affecting their account’s prediction of the correctness of the reasoning in question. This is what the restriction to reasons ‘of the right kind’ in the buck-passing account (which was taken over from Schroeder’s Correctness, who includes it for that very reason) is meant to accomplish: while John’s overall reasons might speak decisively against engaging in that episode of instrumental reasoning, his reasons of the right kind do not, because they do not include the reasons to avoid the pain the demon threatens to inflict on him. Therefore, the reasoning would still be correct, even though it might not be advisable for John to engage in it. Of course, for this solution to work, we need a specification of what distinguishes reasons of the right kind from reasons of the wrong kind that reliably excludes problematic reasons like those provided by the evil demon. Such a specification must not be circular—as taking the right kind of reasons to be those that favour correct and disfavour incorrect reasoning would be. Nor should it be entirely ad hoc, that is, specified solely to escape particular counterexamples, for instance by excluding reasons that are provided by threats of powerful demons. How, then, might advocates of BPCR characterize these reasons of the right kind? A popular suggestion for solving the analogous problem for a buckpassing account of value draws on a distinction between object-given and ⁸ Of course, the scenario could easily be adapted to other examples of correct reasoning; it does not hinge on the episode being one of instrumental reasoning.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

150   state-given reasons.⁹ Roughly, a buck-passing account of value states that an object is valuable if and only if its features give us decisive reasons of the right kind to value it. It thus has to exclude reasons provided by an evil demon who forces you to value something that we take to be worthless, like an arbitrary lump of mud. Now, we can distinguish between reasons to value something that are provided by features of the object of that attitude—e.g. when we value something that gives us pleasure—and reasons to value something that are provided by features of the attitude itself—e.g. when having the attitude allows us to avoid the pain an evil demon might otherwise inflict on us. On the proposal in question, these latter, state-given reasons are excluded as reasons of the wrong kind for valuing something; their presence is taken to be irrelevant for whether or not the object of that valuing is valuable. This offers a promising way to avoid the wrong kind of reasons problem for a buckpassing account of value.¹⁰ This solution, however, is not available for advocates of BPCR. The distinction between state-given and object-given reasons only makes sense in the case of reasons for attitudes like valuing, because those attitudes have objects. But reasons (not) to engage in ℛ(Π, κ) are reasons for or against certain acts, not attitudes. Granted, reasoning differs significantly from other acts we perform. We do not have the same degree of control over our reasoning as we have over our bodily movements, for example. In this respect, reasoning is more similar to attitudes like belief than to ordinary actions. Still, reasoning is something we do, a mental act, if you will, and such acts do not have an object in the sense in which attitudes like desiring, valuing, admiring, etc. have one. Hence, the distinction between state-given and object-given reasons is not applicable in the case of activities like reasoning. Mark Schroeder acknowledges this limitation of the standard solution to the wrong kind of reasons problem (2007, 133 n. 11). Since his main concern is the correctness of certain acts, he proposes a more general way of distinguishing reasons of the right kind from reasons of the wrong ⁹ This distinction is originally due to Parfit (2001). ¹⁰ This distinction is frequently employed in defending of a buck-passing account of value, e.g. by Parfit (2001) and Olson (2004). For a critical discussion of this defence, see Rabinowicz and Rønnow-Rasmussen (2004).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

151

kind that is also applicable to reasons to do something (see Schroeder 2007, 133–6). Instead of drawing a distinction between the kinds of fact that provide the reasons in question, Schroeder suggests that we focus on who has these reasons: reasons of the right kind are ‘the ones that the people involved in that activity have, because they are engaged in that activity’ (2007, 135). The basic idea is this: many paradigmatic reasons of the wrong kind, including the one provided by the demon’s threat in our example, are only reasons for some of the agents who are engaged in the activity whose correctness is in question: someone who is not threatened by a demon but whose situation is otherwise similar to John’s does not have those reasons not to engage in the relevant episode of instrumental reasoning. Other reasons of the wrong kind are too general, that is, they apply to agents independently of whether or not they are engaged in the activity in question. For example, there might be an overriding moral reason not to be amused by some joke even though the joke is funny, so that a buck-passing account of being funny would require that there are decisive reasons of the right kind to be amused by it (Schroeder 2007, 133; cf. D’Arms and Jacobson 2000b). Thus, Schroeder suggests that in order to determine the right kind of reasons to engage in some activity, we have to identify those reasons that are specific to the activity in question: The right kind of reasons to do A are reasons that are shared by everyone engaged in the activity of doing A, such that the fact that they are engaged in doing A is sufficient to explain why these are reasons for them. (2007, 135)

Does this characterization of the right kind of reasons help advocates of BPCR? That depends on whether there are reasons that are specific to the activity of reasoning in this sense, and whether these reasoning-specific reasons yield the right predictions for the correctness of specific episodes of reasoning when combined with the BPCR. Both of these conditions are questionable. Reasoning is a very broad activity, unlike, for example, playing chess, which is another activity with correctness conditions that Schroeder

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

152   discusses. In fact, reasoning arguably is an activity that one is necessarily engaged in, or disposed to be engaged in, in so far as one is an agent.¹¹ That is part of what it is to be an agent. But then it seems that any reason that an agent has simply because she is an agent—that is, any reason such that being an agent is sufficient for explaining why someone shares that reason—is reasoning-specific in the required sense. This arguably includes moral reasons. Surely, if there are any reasons specific to the activity of reasoning, moral reasons are among them. Those, however, appear to be reasons of the wrong kind when it comes to the correctness of an episode of reasoning. To see this, take some action that it would be morally wrong for an agent to perform and that the agent hence has decisive reason not to perform. Let’s assume, for example, that it would be wrong for Linda to poison her flatmate’s food, that she has decisive moral reasons not to do it. Suppose also that Linda presently intends to kill her flatmate without leaving clues, and that she believes that the only way to do so is to put poison in his food. This puts Linda in a position to reason from an intention-belief pair to the intention to put poison in her flatmate’s food, and as an instance of instrumental reasoning, this episode of reasoning would be correct. But if Linda has decisive reasons not to put poison in her flatmate’s food, she also has decisive reasons not to engage in an episode of reasoning that concludes with the intention to do so. This derivative moral reason not to engage in an episode of reasoning, however, must be a reason of the wrong kind for BPCR, since the episode in question would nevertheless be correct. Can Schroeder’s criterion accommodate this result, or does Linda’s derived reason pass his test for reasons of the right kind? Well, it seems implausible to deny that the source reason, i.e. the reason not to poison one’s flatmate, is a reason that is necessarily shared by anyone who is engaging in reasoning, and it also seems plausible that they share the reason simply because they are agents and thus beings that are constitutively engaged in the activity of reasoning. But suppose Schroeder denies ¹¹ Schroeder acknowledges this generality in the case of ‘the activity of placing weight on reasons’, which is ‘simply the activity that every agent is engaged in’, and exploits it to argue that the reasons in question are agent-neutral and that the weight of a reason is thus stable for all agents (2007, 141–4).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

153

this, or suppose he denies that the above derivation of the reason not to engage in the episode of reasoning yields a reason of the right kind, even if the source reason does qualify as one. Such responses, it seems to me, would make it more and more difficult for the present solution to the wrong kind of reasons problem to meet the other condition I mentioned. For the more reasons we exclude from being reasons of the right kind, the more mysterious those alleged reasoning-specific reasons become. Note that nothing has been said yet about what those reasons actually are. This is a general problem for advocates of BPCR: on the one hand, they have to systematically restrict the kinds of reasons that are relevant for the correctness of an episode of reasoning in order to avoid the wrong kind of reasons problem; but on the other hand, they also have to make plausible that there are enough reasons of the right kind left for their account of correct reasoning to be extensionally adequate. To sum up, advocates of BPCR have to specify which reasons qualify as reasons of the right kind. Because these reasons are not reasons for attitudes, the distinction between object- and state-given reasons, which is popular among proponents of a buck-passing account of value, is not applicable in this case. Schroeder has suggested a more general criterion for distinguishing reasons of the right kind from reasons of the wrong kind, which relies on the idea that certain reasons are specific to the activity whose standard of correctness is in question. Yet we saw that applying this proposal to BPCR reasoning is problematic because it seems that the class of reasons that are specific to the activity of reasoning includes moral reasons, which would be reasons of the wrong kind, as the case of Linda illustrates.

5.2.3 Correctness and the possibility of compliance A second problem for BPCR arises from acknowledging that reasons are restricted by the agent’s ability to comply with them. Claims about what an agent has some, sufficient, or decisive reason to do are, just like claims about what an agent ought to do, subject to the ‘ought implies can’ principle (or rather an analogous ‘reason implies can’ principle; cf. Streumer 2007b). It does not make sense to ask an agent to justify—

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

154   that is, give her reasons—for not doing something that she cannot do. That she is unable to do it removes it from the realm of things for which we can ask or give reasons. Moreover, the ‘can’ that is relevant here requires not only that it is possible that you do A, but also that it is within your power as an agent to do so; it is not the ‘can’ of bare possibility but the ‘can’ of ability, as Schroeder acknowledges (2011, 9–10). It makes no sense to advise your friend to smile spontaneously when meeting his date (as opposed to advising him to take certain steps that might cause him to do so) or to blame the croupier for dealing you another queen in blackjack. Hence, if it can be shown that acts within a certain activity can be assessed as correct or incorrect even if it is not possible for a participant to comply with those assessments on a particular occasion, then we cannot understand those assessments in terms of reasons to perform (or avoid) those acts. On occasions where compliance is not possible for the agent in question, the relevant claim about reasons would be undermined by the impossibility of compliance, but the assessment of the act’s correctness would not. The activity of reasoning is such a case, that is, whether or not an episode of reasoning is correct does not depend on whether it is possible for the agent to reason correctly on that occasion. Now, since BPCR was formulated as an account of incorrect reasoning, illustrating this is complicated by the fact that the reason claims it entails are claims about reasons not to reason in a certain way. Yet it seems that in ordinary circumstances agents can always refrain from engaging in a particular episode of reasoning. Finding a suitable test case—that is, a case in which an agent cannot avoid reasoning in a certain way—thus requires us to stipulate that the agent be in quite unusual circumstances. With this in mind, consider the following scenario: The Evil Philosophy Lecturer An evil philosophy lecturer convinces his students that reasoning from means to end is correct whenever the means would be necessary to achieve the end in question. In addition to his sophistry, he hypnotizes his students so as to follow this form of reasoning whenever possible. The effects of his teaching and those of his hypnosis will wear off after a while, considering the consequences

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

155

such reasoning is likely to have. But right at the end of the lecture, he has his students take a small test, which prompts Lucy, one of the students, to reason in the following way: (1) I shall look up the year of birth of Pythagoras. (Intention) (2) Looking up the year of birth of Pythagoras is a necessary means for learning all the biographical data on Pythagoras. (Belief) (3) So, I shall learn all the biographical data on Pythagoras. (Intention) As a result of the lecturer’s teaching and the hypnosis, Lucy could not have avoided reasoning in this way at that particular moment. At the time of the test, it was psychologically necessary for her to draw those conclusions. Is Lucy’s reasoning in these circumstances incorrect? I take it to be obvious that the answer to this question is yes. If you do not think that the form of reasoning she engages in is incorrect in general, adapt the example so as to include an episode of reasoning that you consider incorrect in ordinary circumstances. What matters is whether the particular circumstances of Lucy’s reasoning are relevant for the assessment of her reasoning as correct or incorrect. My claim is that they are not. This is not unusual: there are many activities where the incorrectness of a particular performance does not presuppose the participant’s ability to perform correctly: a pianist plays a piece incorrectly if she hits the wrong notes, even if she is too untalented or too nervous to play it in accordance with the score, and a figure skater performs an axel jump incorrectly when he lands on both feet, whether or not his physical abilities would have allowed him to land it only on one. But does Lucy have a reason not to engage in the episode of reasoning in our example? Initially, one might be inclined to ascribe such a reason to her. Once it is realized that it would be psychologically impossible for her to comply with that reason in the circumstances at hand, however, it is clear that such a claim would have to be retracted, since any reason she might otherwise have had is undermined by her current inability to comply with them. Hence, the incorrectness of Lucy’s reasoning cannot

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

156   be explained by reference to any reasons not to engage in that kind of reasoning. The problem this case illustrates is a fundamental one. BPCR takes the incorrectness of an episode of reasoning to be co-extensional with (and explicable in terms of) the agent’s having sufficient reason of a certain kind not to engage in it. But, quite generally, what an agent has reason (not) to do depends on a variety of factors that are entirely irrelevant for the correctness of that reasoning. These factors can provide additional reasons for or against engaging in the reasoning and thus tip the balance of what the agent has decisive reason to do. To prevent this from affecting the extensional adequacy of his account, the buck-passer has to exclude the reasons provided by those factors. This is the wrong kind of reasons problem discussed in section 5.2.2. But those factors can also affect what the agent has decisive reason to do by undermining certain reasons. In order to prevent such interference from affecting the extensional adequacy of his account, the buck-passer would have to limit the relevant, correctness-grounding reasons to reasons that cannot be undermined by such factors. But some aspects of an agent’s circumstances undermine reasons universally, that is, they undermine all reasons for (or against) a particular action. The agent’s inability to perform (or refrain from) the action is a prime example for such a factor. This is what leads to the problems discussed in the preceding paragraphs, a problem that advocates of BPCR cannot avoid by limiting the reasons mentioned on the right-hand side of their account to reasons of a specific kind, however that kind is individuated.

5.3 Correctness without reasons The preceding sections raised problems for the buck-passing account of correct reasoning. But they also suggest that the idea that correctness must in general be understood in terms of reasons is much less plausible than it might initially seem. This becomes clear, I think, once we consider the breadth of activities that we assess as correct or incorrect in everyday life. I have already mentioned the example of a figure skater who performs an axel jump incorrectly, but who is unable to jump it correctly

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

157

and thus cannot be said to have a reason to do so. But there are many other examples: you can pronounce the word ‘nuclear’ incorrectly as ‘nucular’ while having every reason to do so (e.g. because you are mimicking George W. Bush or voice-acting Homer Simpson), an answer to a question can be incorrect even though you have no reason to give the correct answer (as when a killer asks you for the location of his victim), a description of some event is correct or incorrect independently of any reasons for giving it; you can dance Gangnam style incorrectly, use the cutlery incorrectly, make an incorrect turn at a junction, all without having any reason for not doing so. It is thus not surprising that the assumption that correctness is a normative notion that must be understood in terms of reasons has occasionally come under attack in the literature. Judith Thomson, for example, rejects ‘the widespread tendency to take it that the appropriateness of the word “correct” or “incorrect” in a context is, by itself, a conclusive sign that there is normativity at work in that context’, which she thinks is due to the mistaken idea that ‘the concepts “must”, “obligation”, “correct”, and “ought” come to pretty much the same’ (2008, 108). And in a discussion of how broadly we should understand the notion of normativity, Gideon Rosen points out that judging X to be correct does not always involve ‘acknowledg[ing] a reason to perform some X-involving act’ (2001, 620). He illustrates this with the following example: Your daughter is practicing the piano, but she’s getting frustrated. So, you decide to cheer her up with a laughable performance of her piece. You start in on Mozart’s C-major sonata, but you play a C# where a B should be as the first note of the second measure. Now you knew all along that to play the C# . . . would be to play the piece incorrectly. Did this judgement provide you with a reason not to play the note? To the contrary, it seems plausible that your judgement of correctness was practically neutral in itself. (2001, 620–1)¹² ¹² This leads him to answer the question of the normativity of correctness as follows: Correctness thus differs from the paradigmatically normative features in this respect: From the standpoint of practical reason, the judgement of correctness is like a ‘factual’ judgement: inert in itself, reason-giving only in conjunction with something else. If ‘intrinsic practical valence’ is the sine qua non of normativity, then it would appear that correctness is not a normative notion. (Rosen 2001, 621)

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

158   It seems that, in general, all that is required for something to be adequately characterized as correct or incorrect is some standard relative to which that assessment can be made. Such a standard can be provided by a specific set of reasons, but it does not have to be backed up by any reasons, nor does it always give rise to some reason. It can also be provided by conventions (as in the case of using cutlery) or by the intentions of a certain person or group and the documents that express them (such as the score and the underlying intentions of the composer), for example. Standards like this often license a certain kind of evaluative judgement too. A performance of Mozart’s sonata that does not meet the standards of correctness set up by the score is a bad performance of the sonata (even though it might be a good way to cheer up your daughter), and someone who pronounces ‘cinquecento’ in accordance with the standards accepted by the Italian linguistic community pronounces it well. Note, however, that evaluative terms like ‘good’ and ‘bad’ are used attributively (rather than predicatively) on such occasions: that your piano playing was a good or bad performance of Mozart’s sonata does not entail that it was a good or bad way to spend your time or that it was good or bad, full stop. As our examples show, this kind of evaluative judgement does not entail any claims about reasons: if all you have a reason to do at the moment is cheering up your daughter, and you can do so only by playing the sonata badly, then you have a reason to play it badly and no reason to frustrate her any further by playing it masterfully (cf. Rosen 2001, 621). This opens up the possibility of giving an account of correct reasoning that does not presuppose any truths about reasons. In what follows, I will present such an account, one that I think fits particularly well with the constructivist’s overall approach. That does not mean, however, that the account to be presented is the only one that does not presuppose reasons and is thus constructivism-compatible.¹³ Nor is BPCR the only option ¹³ McHugh and Way (2018), for example, have recently proposed understanding correct reasoning as reasoning that preserves the fittingness of the attitudes involved. I worry, however, that, at least in the case of intentions, the concept of fittingness cannot be understood without reference to reasons, which would make their account unsuitable for constructivists (though see McHugh and Way (2016b) on this).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

       159 for explaining the correctness of an episode of reasoning in terms of reasons.¹⁴ Hence, I have not shown that such accounts are problematic in general. My goal in this chapter, however, is merely to show that there is a viable option for constructivists when it comes to giving an account of correct reasoning.

5.4 The constitutive rules account of correct reasoning To start formulating a constructivism-compatible account of correct reasoning, it will in fact be helpful to go back to an example Mark Schroeder uses to illustrate his conception of activity-specific reasons. Here is what Schroeder writes about the correctness of moves in chess: [T]here are correct and incorrect moves to make in chess. The incorrect moves are ruled out . . . by reasons to follow the rules of the game. Who has those reasons? Anyone who is playing chess. No one is playing chess . . . unless she has some minimal level of desire to be following the rules of the game. And so anyone who is playing chess is guaranteed to have some reason not to, for example, castle out of check, because that is against the rules. (2007, 135)

Schroeder’s rationale in this passage can be reconstructed as follows: (1) Having a desire to follow the rules of the game is necessary for playing chess. (2) Therefore, anyone who is playing chess has some activity-specific reason to follow the rules of the game. (3) The rules of the game prohibit castling out of check.

¹⁴ One alternative, which Joseph Raz (2011, 90) appears to endorse, is that correct reasoning preserves reasons, that is, it ensures that there is a reason for the conclusion-attitude if the premise-attitudes are supported by reasons. But that seems to be implausible in the case of justified but false premise-attitudes. If the glass that I believe to be filled with gin is in fact filled with petrol, then I do not have a reason to drink its contents, even if I am justified in my false belief (cf. McHugh and Way 2018, 162).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

160   (4) Therefore, anyone who is playing chess has some activity-specific reason not to castle out of check. (5) For a chess move to be incorrect is for it to be the case that there is decisive activity-specific reason not to make it. (6) Therefore, castling out of check is an incorrect move in chess. Is this a plausible explanation of why castling out of check is an incorrect move in chess? Several of its steps give rise to serious reservations. The validity of the first step, for example, is certainly far from being obvious. To get from (1) to (2), we have to accept Schroeder’s controversial account of reasons, according to which there is a reason for an agent to do something whenever doing it would promote the satisfaction of some actual desire of the agent (see Schroeder 2007, 59). Of course, Schroeder spends much of his book providing a sophisticated defence of such a desire-based theory of reasons, but the fact that it depends on such a contentious account of practical reasons nevertheless makes his explanation of the incorrectness of particular chess moves much less appealing. Secondly, to get from (4) and (5)—the latter of which is simply the application of BPC, combined with Schroeder’s account of the right kind of reasons, to the activity of chess—to (6), we have to assume that the activity-specific reason mentioned in (4) is decisive and thus not outweighed by some other activity-specific reason. Are there any other activity-specific reasons in chess? Admittedly, it seems intuitively quite clear that no other activityspecific reason could outweigh the reason to follow the rules and turn the move into a correct one. But the explanation for this is as simple as it is problematic for Schroeder’s proposal: we know that no additional reason of the right kind—however we spell this out—could falsify (6) because (3) alone is already sufficient to establish its truth. That is to say, in order to find out whether a chess move is correct, we only have to consult the rules of the game—the standard of correctness for this activity is set by the rules themselves, not by any reasons the participants might have for following them. Once this option has been pointed out, it is obvious, I think, that Schroeder’s account of what it is for castling out of check to be an incorrect chess move takes an unnecessary detour.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

       161 This is an interesting result. Firstly, it provides another counterexample to BPC: the correctness of chess moves need not be understood in terms of reasons. Secondly, and more importantly, it illustrates another approach to explaining the correctness or incorrectness of acts that belong to a certain activity. The source of the standard of correctness that applies to those acts can be a set of rules that govern the activity in question. To be a correct act or move within such an activity is simply to comply with those rules, and to be an incorrect move is to violate them. Understanding correctness in terms of rules provides an attractive option for the case of practical reasoning too. The correctness or otherwise of particular episodes of reasoning, it seems to me, can also be explained in terms of certain rules that govern the activity of reasoning. This is the alternative to BPCR that I will outline in the remainder of this chapter. It claims that an episode of practical reasoning is correct if and only if and because it complies with certain rules that govern the activity of reasoning. Those rules, rather than any particular kind of reasons, provide the standard of correctness for practical reasoning. That proposal, however, gives rise to an immediate worry. It seems to suggest that the standards of correct reasoning are on equal footing with, for example, the standards of etiquette, which also allow us to assess the correctness of certain actions relative to a set of rules. But there is an important difference between the standards of etiquette and the standards of correct reasoning. The former seem to be entirely contingent and might just as well have been quite different, requiring us to greet business partners by high-fiving them or to sip red wine from a ceramic bowl, for example. Moreover, an agent can simply ignore the standards of etiquette while engaging in the activities they apply to; she might not even be aware of them. The standards of correct reasoning, on the other hand, do not appear to be a matter of contingency, nor can a reasoner be ignorant of the fact that reasoning is subject to those standards. There is, in other words, an element of arbitrariness in the standards of etiquette and an agent’s compliance with them that is lacking in the case of reasoning. How can a rule-based account of correct reasoning accommodate this difference? The key to answering this question lies in the fact that unlike, for example, the rules of etiquette, the rules that provide the standards of

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

162   correct reasoning are constitutive rules.¹⁵ A paradigmatic example for such rules are the rules for the movement of the pieces in chess. A bishop, for example, may be moved to any square on the two diagonals it is currently on, unless there is another piece on a square between that target square and its current position. If there is an opposing piece on a square the bishop could otherwise reach, it can be moved to that square and captures the opposing piece. These rules define the notion of a bishop in chess. If a particular piece may be moved in a different way, then it is not a bishop, and if you do not consider what you are doing with the various pieces to be subject to these rules at all, then you are simply not playing chess. Assuming that the rules of reasoning are constitutive of the activity of reasoning in the same way that the rules of chess are constitutive of the activity of chess allows us to explain why the standards of correct reasoning are not contingent like the standards of etiquette, and why reasoners cannot ignore them: that certain ways of moving a bishop are correct and others incorrect is not a contingent matter either (a board game in which the ‘bishop’ can move horizontally and vertically as well as diagonally would not be chess), nor is it something a chess player can be ignorant of. Adding this specification of the kind of rules in question to a rule-based account of correct reasoning thus allows us to accommodate the differences between the standards of etiquette and the standards of correct reasoning, and thereby to dispel the worry raised above. This, then, is the account of correct reasoning proposed here: The Constitutive Rules Account An episode of reasoning is incorrect if and only if and because it violates the constitutive rules of reasoning. An episode of reasoning is correct iff it is not incorrect. In section 5.5, I will flesh out this proposal by elaborating the notion of a constitutive rule and explaining in more detail why standards of

¹⁵ The idea that the standards of correct reasoning are constitutive of agency plays a central role in Korsgaard (2009b). James (2012) too takes it to be a main tenet of constructivism.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

163

correctness that are based on such rules are not contingent and cannot be ignored.

5.5 Developing the account The notion of constitutive rules was made prominent by John Searle, who distinguishes them from what he calls (merely) regulative rules. Here is how he introduces those two kinds of rules: Some rules regulate antecedently existing forms of behavior. For example, the rules of polite table behavior regulate eating, but eating exists independently of these rules. Some rules, on the other hand, do not merely regulate but create or define new forms of behavior: the rules of chess, for example, do not merely regulate an antecedently existing activity called playing chess; they, as it were, create the possibility of or define that activity. The activity of playing chess is constituted by action in accordance with these rules. Chess has no existence apart from these rules . . . . Regulative rules regulate activities whose existence is independent of the rules; constitutive rules constitute (and also regulate) forms of activity whose existence is logically dependent on the rules. (1964, 55)

Without using the terminology of constitutive rules, John Rawls already made a very similar distinction in an earlier paper, where he distinguishes a conception of rules according to which they ‘define a practice’ and are hence ‘logically prior to particular cases’ (1955, 24–5) from another conception that takes rules to apply to cases that can be identified on independent grounds (1955, 22). According to both Searle and Rawls, the crucial features of the special kind of rules that Searle calls constitutive is that they define the activity (or ‘practice’) to which they apply, and that the rules are thus logically prior to that practice and bring it into existence. Merely regulative rules, on the other hand, apply to an activity that exists and can be identified independently of those rules. To elaborate these features and their significance for the Constitutive Rules Account’s ability to accommodate

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

164   the differences between the rules of reasoning and the rules of etiquette, it will be helpful to start with a few remarks on the notion of a rule more generally.

5.5.1 Rules in general A rule, as I understand it here, is a general prescription that applies to an indefinite number of particular cases. Rules are often formulated as imperatives or by using modal auxiliaries like ‘must’ or ‘may’, but they need not be put in such terms: the same rule of chess can be expressed by ‘If you are playing white, make the first move!’, by ‘White must make the first move,’ and by using a simple declarative sentence like ‘White moves first.’ Prescriptive rules like those of chess are sometimes distinguished from descriptive rules like those governing the movement of subatomic particles or the forces between two masses, but I will reserve the term ‘rule’ for the prescriptive case. We can identify three core elements of every prescriptive rule (cf. von Wright 1963, ch. 5; Raz 1975, 50). Firstly, a rule has a certain deontic character, that is, it requires, permits, or prohibits a certain kind of action. Secondly, a rule has a particular content, which is the kind of action it requires, permits, or prohibits. And thirdly, a rule has a condition of application: that condition might be trivial—e.g. when a rule requires a certain kind of action on the condition that an opportunity to perform such an action arises (see von Wright 1963, § 5.6)—, but it might also impose a genuine limitation on the occasions on which the rule should be followed (relative to the occasions on which it could be followed). Consider the rule ‘If you want to use the train, you must buy a ticket.’ As the ‘must’ signals, the deontic character of that rule is that of a requirement. The content, i.e. the action required, is that of buying a ticket. And the rule’s prescription is genuinely limited by its condition of application. Not everyone who could buy a ticket is required to do so; only those who want to use the train are. Together, these three elements allow us to determine whether or not an agent’s performance (or omission) of a particular action conforms to a certain rule. It does if it is an instance of the kind of action specified by

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

165

the content of the rule, the rule’s condition of application is met, and the performance or omission accords with the deontic status (that is, the agent performs the action and it is required or permitted by the rule, or the agent does not perform it and it is merely permitted or prohibited by the rule). The action does not conform to the rule if it is an instance of the kind of action specified by the content of the rule, the rule’s condition of application is met, but the performance or omission does not accord with the deontic status (that is, the agent performs an action that is prohibited by the rule or fails to perform an action that is required). If the action is not an instance of the kind of action specified by the rule’s content or if the condition of application is not met, the rule does not apply to the particular action and the question of conformity does not arise.¹⁶ Note, though, that not every case in which an action conforms to a certain rule is also a case of compliance. Depending on the agent’s awareness of and attitude towards the rule in question, conformity and compliance can come apart. Suppose, for example, that you are taking a walk in the park, and that much to the entertainment of a group of children watching you, you are conforming to the rule not to step on the lines between the paving slabs, a rule those children just made up. Since you are not even aware of the rule, however, the conformity of your behaviour is entirely accidental. Now compare the accidental conformity of your walking to the conformity of one of the children walking by her friends on the walkway. Her behaviour is determined by the rule which she and her friends decided to adopt. If her walking conforms to the rule, she acts in compliance with it, and if she steps onto a line, she makes a mistake by her own lights because she violates a rule that she is trying to follow. If you were to step on a line, however, you would not be making a mistake in that sense. The difference between you and the girl lies, firstly, in your awareness of the rule: you cannot follow and thus cannot comply with a rule that you are not even aware of. But awareness of the fact that the rule applies

¹⁶ The possibility that a rule applies to a case even though the case does not conform to it is an important difference between rules and mere regularities or descriptive laws. The latter would be disproved by or would have to be adapted to such counterexamples.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

166   to your walking is not all that is missing for your behaviour to qualify as compliance. You can acknowledge that a rule applies to something you do while being entirely indifferent to that fact, so that it does not play any role in explaining your behaviour, even if that behaviour happens to conform to the rule. If a historian finds an ancient document that specifies a rule to the effect that a man’s hair must not cover his ears, he might, upon his next visit to the barber, acknowledge that his request to have his hair cut to half an inch at the sides conforms to that rule, but his decision might be based on entirely different considerations, in which case his choice of hairstyle will not be a case of compliance (but merely of conformity). What distinguishes the girl’s behaviour from yours and turns it into a case of compliance is thus not only her awareness of the rule, but her attitude towards it. For a person’s behaviour to qualify as compliance with a rule, she must have some favourable attitude towards conforming to that rule. We can call this attitude her acceptance of the rule in question. I will not attempt to define this attitude any further—it simply is the kind of attitude towards a rule that is required for complying with it. I take it that the example makes it clear that some such attitude is indeed required to distinguish compliance from mere conformity, and that our familiarity with rules and their role in determining our behaviour affords us some intuitive grasp of that attitude. It is because you do not accept the rule not to step on the lines between the paving slabs (and cannot accept it, given that you are not even aware of it) that your conformity to that rule is not a case of compliance. For the same reason, you would not be mistaken by your own lights if you did step on a line.

5.5.2 Constitutive vs merely regulative rules With this terminological framework in place, we can come back to the distinction between constitutive and merely regulative rules. The crucial difference concerns the way the activity to which those rules apply, that is, the activity that figures in the content of those rules, is identified. In the case of merely regulative rules, that activity can be identified independently of those rules, so that whether or not a particular act is an

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

167

instance of that activity can be ascertained without considering whether or not it conforms to the rule in question. This is the case for the children’s rule considered in section 5.5.1: whether or not you step on a line between two paving slabs depends simply on whether or not your act meets a certain description, one that does not mention the children’s rule. The same holds for the activities to which the rules of etiquette apply. Whether or not you are eating, using your utensils, entering a room before or after someone else, etc.—all that can be ascertained without referring to the rules of etiquette. Those activities ‘exist independently’ of these rules, as Searle puts it. Consequently, you can intentionally engage in those activities without even being aware of the rules of etiquette or the fact that they apply to those activities. This is why it is easily possible for you to ignore those rules while engaging in the activities, and it is also why it is a matter of contingency what the rules of etiquette require of those who are eating, using their utensils, entering a room, etc. If the rules were different, you could still engage in the very same activities, because the rules do not play any role in identifying those activities. Neither of these is true for constitutive rules. A rule is constitutive of the activity to which it applies if and only if that activity is defined as what an agent does when her behaviour is determined by the acceptance of that rule. Moving a bishop, for example, is defined as what an agent does with a certain piece on the board if his behaviour is determined by his acceptance of the rules of chess. In such cases, you cannot ascertain whether a certain act is an instance of the activity without referring to the rule in question. This is the sense in which constitutive rules are ‘logically prior to’ (Rawls 1955) and ‘create’ the relevant activity. You can move around wooden figures on a checquered board any way you want, but what you do counts as moving a bishop, and as playing chess more generally, only if what you are doing is determined by your acceptance of the rules of chess. By accepting those rules, you turn what you are doing into a game of chess. Hence, moving a bishop, castling, putting your opponent’s king check, etc. are kinds of action that are only available to—only exist for—agents who accept certain rules. Consequently, agents cannot participate in an activity that is constituted by certain rules without being aware of the fact that those rules

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

168   apply to it, that is, they cannot be ignorant of those rules. Moreover, an activity that is constituted by different rules would necessarily be a different activity. Hence, you could not engage in the same activity if the rules in question were different ones. That is what distinguishes constitutive rules from merely regulative rules like those of etiquette.

5.5.3 The violability of constitutive rules This does not mean, however, that one cannot fail to comply with constitutive rules. An agent can accept certain rules and engage in the activity they constitute even if his behaviour does not conform to those rules in every instance. For one thing, the agent might be mistaken about the precise content of a certain rule, for example, when both players in a game of chess falsely think that the king is moved three (instead of two) squares when castling with the queenside rook. As long as those players, nevertheless, accept the rules of chess—which they can identify despite the incomplete knowledge of their content, e.g. by referring to the rules accepted by Garry Kasparov or written down in a certain book—and would consider the relevant moves as incorrect if it were pointed out to them that they do not conform to those rules, the players are still playing chess, albeit incorrectly. Moreover, agents can—at least sometimes—violate constitutive rules whose content they are fully aware of. This might be difficult to imagine in the case of chess,¹⁷ but consider another game that is constituted by certain rules, such as Monopoly. It is certainly possible that a player accidentally moves her piece eight steps forward even though the dice indicate nine, without thereby ceasing to play Monopoly. And a player who does the same thing knowingly need not be mistaken about the game she is playing; she might simply be cheating. A player who cheats ¹⁷ This is due both to the fact that the movement rules of chess are comparatively simple and easy to follow, so that a violation is easily recognized, and to the fact that the general rules of the game determine that any violation immediately terminates the game. Nevertheless, it does seem possible that a player who knows how to castle moves her king three instead of two squares towards her queenside rook accidentally and without her opponent noticing it, thus violating a rule of chess while continuing to play the game. It seems possible too that she does so intentionally rather than accidentally in order to cheat.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

169

still accepts the rules of Monopoly and her behaviour is still informed by her acceptance of those rules; she is just not complying with them on that occasion. Instead, she makes a move that she herself considers to be incorrect but to her own advantage. Or, to take up an example discussed in section 5.3, an agent might play Mozart’s C-major sonata incorrectly, be it by accident or to cheer up his daughter, even though the rules for playing that piece of music, which are laid down in the score by the composer, are constitutive for that particular activity (what someone is doing at the piano counts as playing or trying to play that sonata only if he accepts the rules laid down in the score). There is, however, a limit to how far our actual behaviour can deviate from what the constitutive rules of some activity prescribe while still counting as engaging in that activity. Both accidental and intentional violations of such rules can be made sense of only against a background of compliance—compliance is, as it were, the default response to the rules one accepts for what one is doing. If the participants in an activity violate its constitutive rules too often or too readily, the activity degenerates: if, for example, the people playing on a Monopoly board start to develop their properties before owning all properties in that colour group, waive the rent owed by other players who land on their properties at will, or gift their properties to other players,¹⁸ it becomes less and less plausible to characterize them as engaging in a game of Monopoly, rather than doing something else. Another way to put this is that constitutive rules impose a constraint on interpretation. For an agent’s (or a group of agents’) behaviour to be interpretable as an instance of an activity that is constituted by certain rules, any case of nonconformity to those rules— that is, anything that would amount to an incorrect move within that activity—requires some sort of explanation as a fluke, a misapplication of the rules, a case of cheating, etc. Without such an explanation, the divergence between the rules’ prescriptions and the actual behaviour is difficult to square with the agent’s acceptance of the rules. But, as we saw earlier, such acceptance is a precondition for engaging in the activity in question.

¹⁸ None of this is in conformity with the rules of Monopoly.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

170   This aspect of constitutive rules lends further support to the Constitutive Rules Account of correct reasoning. For while incorrect reasoning is certainly possible, the standards of correct reasoning appear to impose a constraint on interpreting someone as reasoning, or as a reasoner, just as an account of those standards in terms of constitutive rules would predict. Suppose someone systematically fails to engage in all kinds of correct reasoning but regularly does things like forming an intention to pursue the end for which he believes some other action, which he already intends to perform, to be a necessary means (i.e. he seems to be engaging in what I have called ‘reasoning to the facilitated end’ in section 5.1). Without some other explanation, it will be less and less plausible to interpret such a person as reasoning, rather than as someone who is engaged in some other, possibly pathological psychological activity. There is hence a limit to how far a person’s psychological processes can deviate from what the standards of correct reasoning require while still being interpretable as (incorrect) episodes of reasoning. The Constitutive Rules Account provides a straightforward explanation for this.

5.5.4 Rules of reasoning and reason judgements The Constitutive Rules Account is an account of correct reasoning. By characterizing reasoning as an activity that is governed by constitutive rules, however, it also tells us something about the nature of reasoning. Applying the results of the previous sections to the case of reasoning yields that an agent counts as engaging in reasoning only if what he does is determined by his acceptance of certain rules of reasoning. How does this fit with the conception of reasoning developed in Chapter 2? There, I argued that every episode of practical reasoning is guided by a reason judgement that connects the contents of its premise- and conclusionattitudes. To see how these ideas fit together, we need to clarify the relation between this reason judgement, which guides the reasoning, and the attitude of accepting a rule of reasoning, which determines it. Accepting a rule of reasoning is a much more general attitude than the reason judgement that guides a particular episode of reasoning. Recall

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

171

that the reason judgement is specific to the premise- and conclusionattitudes of the particular episode of reasoning in question. Two instances of instrumental reasoning that differ in their premise-attitudes thus necessarily involve different reason judgements too. But they will both be determined by the agent’s acceptance of the same rule of instrumental reasoning. Compare again the case of chess. Accepting that bishops only move diagonally is accepting a rule of chess, whereas the judgement that you may move your bishop from field C1 to field A3 is a much more specific attitude. Like a reason judgement, it can only guide a particular move in the activity. Moreover, one can accept a certain rule without being fully and explicitly aware of its content. I already mentioned this possibility in the case of chess, but it is even more obvious in the case of the rules of grammar. You might be able to confidently and reliably identify correct and incorrect uses of the gerund form, for example, without resorting to (or even being able to articulate) an explicit representation of the rules that govern the use of that form. This suggests that accepting a rule, understood as the attitude that is required for being able to comply with it, is not a propositional attitude at all. Instead, it is best thought of as a syndrome of dispositions that manifest both in how one engages in the activity and how one responds to other people’s as well as one’s own moves in the activity. In the case of reasoning, accepting a rule of reasoning thus involves a disposition to form certain reason judgements, which can then guide episodes of reasoning that conform to the rule, as well as a disposition to respond in certain favourable or critical ways to particular episodes of reasoning (one’s own as well as others’). An agent’s acceptance of a rule can thus be said to determine her reasoning in so far as it manifests in the formation of a certain reason judgement, which then guides a particular episode of reasoning on her part.¹⁹ This is only a

¹⁹ The rules of reasoning that a reasoner accepts will thus be much less numerous than the reason judgements that she entertains over time. In fact, the constructivist account developed in this book is compatible with the view that instrumental reasons are the only kind of practical reason there is. In that case, there would still be innumerable different reasons (deriving from our various ends), which correspond to innumerable different reason judgements. But there would have to be only one constitutive rule of reasoning: the one permitting instrumental reasoning.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

172   rough sketch, but it should give you an idea of how these two pieces of the constructivist picture fit together.

5.6 How to determine the rules of reasoning The Constitutive Rules Account provides a general criterion for the correctness or incorrectness of practical reasoning: an episode of reasoning is incorrect if and only if it violates the constitutive rules of reasoning. In order to determine whether or not particular episodes of reasoning are correct or incorrect, however, the account requires additional input from a substantive theory of the rules that constitute the activity of reasoning.²⁰ After outlining the Constitutive Rules Account of correct reasoning and elaborating its central concepts, we are thus left to ask: what are the constitutive rules that set up the standards of correctness for practical reasoning? This is, no doubt, an important question. Instead of answering it, however, I suggest that we step back and consider the role that an answer to that question plays in the overall account of practical reasons proposed by the constructivist.

5.6.1 Constructivism and normative theory Recall that the basic idea of the constructivist account of practical reasons is that facts about what is a reason to do what are not something that agents discover in practical reasoning. Instead, those facts must be understood as grounded in the truth of our reason judgements, where the truth of those judgements is spelled out in terms of their role in correct practical reasoning. The details of this connection between correct reasoning and the truth of reason judgements were presented in

²⁰ The same, of course, is true for the buck-passing account of correct reasoning we dismissed in section 5.2. That account provides a general criterion for the correctness of episodes of reasoning in terms of certain reasons. But in order to yield definite results for individual cases, that account too requires additional input from a substantive theory of our reasons for and against reasoning in certain ways. In this respect, then, the two accounts are on an equal footing.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

      

173

Chapter 4. What is crucial, however, is that on such a constructivist account, facts about what is a reason for what are ultimately grounded in and therefore determined by facts about what qualifies as correct reasoning. Given this role of the Constitutive Rules Account in the constructivist’s overall account of practical reasons, a theory of the constitutive rules of reasoning—that is, the kind of theory that provides the additional input necessary to ascertain whether particular episodes of reasoning are correct or incorrect—entails substantive normative claims about what is a reason for doing what. Should the constructivist be expected to provide such a substantive theory? On the one hand, that depends on the ambitions of the constructivist project. In this regard, it is helpful to distinguish between what Aaron James calls ‘specificatory’ and ‘explanatory constructivism’. Whereas the former aims to ‘justify particular first-order judgements’ about reasons, e.g. by showing that their truth is grounded in a substantive account of correct reasoning, the aim of the latter ‘is to explain how there could be truths or facts of the matter of a certain kind’ (2007, 310; italics added). Hence, if all the constructivist aspires to is to explain how there could be truths or facts about practical reasons, it seems that all he has to do is show that we can understand what it means for an episode of reasoning to be correct or incorrect independently of any assumptions about what is a reason for what, so that the former could provide the explanatory basis for the latter without thereby closing a vicious explanatory circle. That much, I believe, is established by the Constitutive Rules Account of correct reasoning. What truths about reasons there actually are, however, is not a question that explanatory constructivism is intended to answer. Explanatory constructivism is thus compatible with a (partial or complete) error theory of normative discourse. It merely establishes that for there to be any truths about reasons, there have to be certain constitutive rules of the activity of reasoning. If it turned out that there are no such rules at all, or rules that uphold only a subset of the reason judgements we ordinarily take to be true, then any discourse that presupposes that there are (additional) truths about reasons would be subject to a fundamental mistake. On the other hand, the overall plausibility of a constructivist account of practical reasons will depend on whether or not it is capable of

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

174   accommodating our most strongly held convictions about what there is reason to do. If, for example, a constructivist account would yield that we never have a reason to avoid pain or to keep our promises, this would probably suffice for many to reject such an account of practical reasons. But without the addition of a substantive theory of the rules of reasoning, the normative implications of a constructivist account, and hence a major factor for deciding whether or not to accept such an account, cannot be evaluated. This suggests that a constructivist, even if his primary aim is merely to explain how reason judgements could be true or false in virtue of their role in correct reasoning, should provide enough of an outline of a substantive theory of the rules of reasoning to assure us that accepting his explanatory proposal does not commit us to an implausibly revisionist stance towards our everyday judgements about reasons.

5.6.2 The order of explanation and the order of justification Constructivists thus need to show that their account of correct reasoning provides the resources to avoid such revisionism and to accommodate our convictions about reasons. It is quite a different matter, however, how they can or must argue for this. It is tempting to think that because this account is supposed to ground the truths about reasons and hence must not itself depend on any of those truths, the constructivist must be able to defend such a theory without referring to particular reason judgements in support of his proposal. According to the constructivist, practical reasons can be explained in terms of true reason judgements, which in turn are explained in terms of what qualifies as correct practical reasoning. Hence, it might seem that he must first provide an account of correct reasoning, including a substantive theory of the rules of reasoning if he opts for the Constitutive Rules Account, then combine this with an account of the truth of reason judgements, and then show that this affords us an adequate understanding of what reasons we have, without at any step relying on results to be established at a step further up the explanatory ladder. That view of how a constructivist must—and, more

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

      

175

importantly, must not—argue, however, is mistaken: the order of justification within the different elements of the constructivist’s picture need not strictly follow the order of explanation. This is not a distinctively constructivist thesis; it can be illustrated by considering the dialectics of a representationalist account of the relation between facts and judgements about reasons. According to such an account, a particular reason judgement is true, because it accurately represents a certain fact about reasons, that is, because the fact it purports to represent actually exists. Hence, the truth of the judgement is grounded in the existence of a certain fact, and the ontology of reason facts, that is, the general theory of what reason facts there are and are not, determines which reason judgements turn out to be true or false on such a representationalist account. In a way, then, the role an ontology of reason facts plays in a representationalist account is similar to the role a theory of correct reasoning plays in a constructivist account: both determine the extension of ‘is true’ for the domain of judgements about reasons.²¹ But now, how does a proponent of Representationalism decide whether or not the ontology of reasons includes a particular reason fact, that is, whether it holds the resources to accommodate the truth of a particular reason judgement? That depends on whether he takes those facts to be fundamental, or whether he takes them to be reducible to certain other, non-normative facts. A fundamentalist will probably answer this question by referring to the results of first-order normative theorizing and suggest that we ‘read off ’ the existence of certain reason facts from the reason judgements we hold to be true as a part of our best normative theory. In other words, the fundamentalist will suggest that the best way to answer the ontological question is simply to answer the

²¹ Of course, they do so in very different ways. The representationalist takes judgements to be true iff and because they accurately represent the facts that exist according to the ontology, whereas on the constructivist account, the judgements are not true because they accurately represent anything (including facts to the effect that certain kinds of reasoning are correct), but because of their role in correct episodes of practical reasoning (Chapter 4.5). Hence, while both the ontology of reason facts and the theory of correct reasoning provide the necessary input into an account that then tells us which particular judgements are true, the way in which these accounts explain the truth of those judgements in terms of their respective input differs significantly between constructivism and Representationalism.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

176   substantive normative question.²² Note that on this proposal, the explanans for the truth of a particular reason judgement is justified in terms of its explanandum and not established on independent grounds. On a reductionist version of their account, on the other hand, representationalists have two options. They can either establish the existence of certain reason facts on independent grounds, that is, by establishing directly that the facts to which they can be reduced obtain. Or they can determine whether some consideration is a reason for doing something by considering both how the fact to which this reason fact would have to be reduced fits with an overall account of the reduction base and how crucial that reason fact is to accommodating their entrenched normative convictions, trying to bring these two considerations into a (wide) reflective equilibrium (cf. Rawls 1971; Daniels 1979).²³ If reductivists proceed in this second way, the explanans for the truth of particular reason judgements is again (partly) justified by the plausibility of some of those judgements. There is nothing wrong with justifying the more fundamental parts of a theory in this way. That the direction of explanation and the direction of justification come apart is not unusual; it is a characteristic aspect of inferences to the best explanation and hence of one of the most common modes of justification not just in philosophy and everyday life, but also in the sciences. If, for example, you justify the claim that there was a bear at the cabin last night by inferring it from the observation that there are fresh tracks in the snow in front of the cabin, or if you infer, as nineteenth-century astronomers did, from an unexpected deviation in the orbit of Uranus that there is another, as yet unobserved

²² He will probably take this method of determining the ontology of reasons to be fallible, that is, he will insist that even the judgements that we hold as part of our best normative theory might be false. This does not, however, undermine the point about the reversal of the direction of explanation relative to the direction of justification. ²³ This second option will be more plausible for some reductive accounts than for others. If reason facts are reduced to certain ordinary, empirically accessible facts (e.g. about behavioural dispositions), the idea that an empirical theory of this reduction base has to be brought into reflective equilibrium with our normative convictions will probably appear misguided. If, on the other hand, the reduction base consists of facts about the psychology of an ideally coherent agent, this reflective equilibrium approach might be more plausible.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

      

177

planet in the solar system, then you are justifying the claim that some fact holds based on evidence that would be explained by the fact in question. In other words, you are establishing an explanatorily more fundamental fact based on the facts that it explains—that is the point of an inference to the best explanation. Similarly, the representationalist’s order of explanation (from fact to truth) need not be followed by his order of justification, which might proceed from the conviction that some reason judgement is true to the claim that a certain reason fact obtains (and perhaps further to a claim about the reduction base of those reason facts).

5.6.3 Constructivism and reflective equilibrium Once we accept that in the theory of practical reasons, as well as elsewhere, explanation and justification can proceed in opposite directions, it is difficult to see why this option should not also be available to the constructivist when it comes to justifying his account of correct reasoning. The constructivist might hence justify the claim that a certain kind of reasoning is correct or incorrect in terms of the plausibility of its normative output—that is, in terms of the reason judgements that would turn out to be true when that claim is combined with his account of the truth of such judgements. At least, that should not be taken to conflict with his explaining why these reason judgements are true in terms of the correctness of those kinds of reasoning. Of course, any claim about the correctness of a certain kind of reasoning will have to fit into a general account of correct reasoning. In particular, if the Constitutive Rules Account is on the right track, accepting such a claim will have consequences for what we take to be a constitutive rule of practical reasoning, and its acceptability will depend on the acceptability of those consequences as part of a substantive theory of those rules. Take reasoning from the belief that you will experience severe pain unless you perform a certain action to the intention to perform that action. The claim that this reasoning is correct certainly gains plausibility from its normative output: combining it with the constructivist account of reasons yields that you have a reason to perform such a pain-avoiding

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

178   action.²⁴ But the idea that a rule which permits such reasoning is among the constitutive rules of practical reasoning also sits well with some compelling ideas about correct reasoning in general. For example, it seems very plausible that such reasoning can be thought of as enabling an agent to navigate her surroundings, to maintain her agency, and to coordinate her behaviour with others. Though admittedly quite vague, these claims also support the claim that the pain-avoiding reasoning above is correct. They also suggest that, say, reasoning from the belief that doing A is impossible to the intention to do A is not correct. This, in turn, has plausible normative output: that something is impossible to do is never a reason for doing it. Hence, the constructivist does not simply propose to read off the correctness of certain episodes of reasoning from the reason judgements he wants his account to accommodate. Understanding those assessments as the result of a general theory of correct reasoning, and in particular as flowing from an account of the constitutive rules of reasoning, provides genuine restrictions on what claims about the correctness of certain kinds of reasoning the constructivist can justify in terms of their normative output. In other words, he will have to bring (i) our convictions about reasons, (ii) what we take to be a plausible theory of the constitutive rules of reasoning and hence a plausible account of correct reasoning, as well as (iii) the other elements of his constructivist account, including those that connect the truth of reason judgements to the correctness of certain kinds of reasoning, into a wide reflective equilibrium in which all of these elements pull their weight and none is immune to revision. Moreover, without actually trying to establish such a reflective equilibrium—which includes formulating a substantive theory of practical reasons as well as a theory of correct reasoning—there is no guarantee that this project can succeed: it might turn out that our convictions about reasons and a plausible theory of correct reasoning cannot be brought to harmonize in the way a constructivist account ²⁴ Note that the account only entails that you have a reason for the action, not that this is what you have most reason to do. As most people know, e.g., from their interactions with dentists, a reason to avoid pain can be overridden by a stronger reason to endure it. Similarly, that the corresponding episode of reasoning would be correct does not entail that it is the episode of reasoning that you should engage in all things considered (cf. section 5.3).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

      

179

requires. On the other hand, if one acknowledges that pointing out its compatibility (or lack thereof) with certain plausible reason judgements is a legitimate move in debating the merits of a particular account of correct reasoning, there is little reason to suspect ex ante that the constructivist’s project must fail on first-order normative grounds either.

5.6.4 The project of ultimate justification There is a more ambitious version of the constructivist project. Constructivists can try to justify a theory of correct reasoning entirely on independent grounds, that is, without considering its normative output. If successful, this would allow them to provide an ultimate justification for our reason judgements: they could answer the question why a certain consideration is a reason to do something by pointing out that certain kinds of reasoning are correct, and then justify this claim in a way that does not appeal to one’s acceptance of any claims about reasons. Providing an ultimate justification for a substantive, first-order normative theory has been the aim of a number of influential proposals in moral philosophy and the theory of practical reasons, starting perhaps with Kant’s project of providing a deduction of the categorical imperative in section III of the Groundwork (1785) and (arguably) continuing with Apel (1973), Gewirth (1978), Gauthier (1986), Korsgaard (1996b), as well as, more recently, Smith (2012). The fact that this list includes what many take to be paradigmatic versions of constructivism about morality or practical reasons might explain why it is often assumed that any constructivist account must be committed to that project. Given this assumption, it is indeed appropriate to expect a constructivist account of practical reasons to defend a complete account of correct reasoning without any appeal to what might plausibly be taken to be a reason for what, and then to see whether the normative output of such a theory lines up with our convictions or whether its gross implausibility forces us to shelve the entire constructivist project. It is not my intention to denounce this more ambitious project, and I certainly acknowledge its theoretical appeal and the promise it holds for

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

180   the advancement of first-order normative theorizing. As far as my ambitions here are concerned, however, I will remain agnostic as to the prospects of providing an ultimate justification for a substantive theory of practical reasons based on an independently justified theory of the constitutive rules of reasoning. The point I wish to highlight is rather that this more ambitious aim is optional for proponents of a constructivist account of practical reasons. Hence, if one accepts that the constructivist project is best understood as defending an account of what grounds reason facts by proposing a certain order of explanation between those facts, the truth of our reason judgements, and correct practical reasoning, then one should also accept that constructivism is not committed to providing an ultimate justification for our reason judgements. For it might turn out that even though the truth of those judgements is ultimately grounded in (and hence to be explained by) the constitutive rules of reasoning, ascertaining whether a particular rule is constitutive of the activity of reasoning requires us to take the coherence of its normative output with our substantive normative convictions into account.²⁵ In that case, a complete account of correct reasoning could only be defended as part of a wide reflective equilibrium and hence jointly with a substantive theory of normative reasons. That would undermine the ambition to provide an ultimate justification of the latter theory in terms of the former, but it would not undermine the explanatory project that forms the heart of the constructivist account of practical reasons.

²⁵ Interestingly, there is a footnote in the preface to the Critique of Practical Reason (1788) where Kant appears to acknowledge a similar point and to relinquish his earlier ambition (in section III of the Groundwork) to provide a deduction of the moral law by establishing the reality of our freedom, claiming now that whereas freedom is indeed the ratio essendi of the moral law, the moral law is the ratio cognoscendi of freedom. For, had not the moral law already been distinctly thought in our reason, we should never consider ourselves justified in assuming such a thing as freedom. . . . But were there no freedom, the moral law would not be encountered at all in ourselves. (C2 5: 4n) Picking up Kant’s terminology here, we can put the methodological point made in the main text by saying that even though the activity of reasoning and its constitutive rules are the ratio essendi of practical reasons, our convictions concerning these reasons are part of the ratio cognoscendi of those rules.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

      

181

5.6.5 Reasoning, shmeasoning? Because it ultimately grounds practical reasons in the constitutive rules of our practical reasoning, the constructivist account presented here is also a version of what has become known as constitutivism. Constitutivist accounts of practical reasons generally comprise two claims. One is a claim about the relation between reason facts and some class of facts about agents or agency. It takes the former to depend on and be explicable in terms of the latter. According to the second claim, some of these latter facts are constitutive of being an agent or an action. In the case of constructivism, this class of facts is facts about the rules that govern practical reasoning and provide its standard of correctness. Other versions of constitutivism take different facts to play that role.²⁶ David Velleman (1996; 2009), for example, has argued that reasons are considerations which indicate that an action meets a standard of correctness provided by an aim, and that agents constitutively aim at acting intelligibly. Constitutivism is best thought of as a strategy that proponents of certain meta-ethical positions pursue, rather than as a meta-ethical position in its own right. As such, it has been prominently criticized by David Enoch, who argues that the strategy is hopeless because there might be ‘shmagents’, beings who are very similar to agents but to whom the facts that the constitutivist takes to be constitutive of agency do not apply (2006, 178–80; 2011b, 209). To assess Enoch’s objection, it is necessary to first say more about what constitutivists seek to achieve by adopting that strategy. In particular, we can distinguish two goals. The first goal is to accommodate objective reasons, reasons that are universally shared by all agents. If aims generate reasons, for example, and some aim is necessarily shared by all agents, then the reasons generated by that aim are shared by all agents too. The second goal is to answer the normative sceptic. This sceptic questions the authority of any claim about his reasons. Confronted with a call to give to charity, for example, he will ask why he should accept that he has reason to give to charity. The constitutivist

²⁶ See Katsafanas (2018) for a helpful overview.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

182   might respond to the sceptic’s challenge by answering that a certain fact necessarily applies to him qua agent, and that, in virtue of this fact, he has a reason to give to charity. For this response to succeed in answering the challenge, the constitutivist has to assume that both his dependence claim and his claim about what is constitutive of agency can be ascertained independently of the reason facts that they are meant to explain. Otherwise, the above response could not provide a justification for the reason judgement in question and hence could not succeed in persuading the sceptic (or at least in persuading us that the sceptic’s doubt is unfounded). Distinguishing these two goals allows us to distinguish two versions of constitutivism. A proponent of modest constitutivism pursues only the first goal and is thus not committed to assuming the availability of independent epistemic access to the constitutive facts. A proponent of ambitious constitutivism, by contrast, seeks to (also) achieve the second goal and must thus accept the independent access assumption. While some constructivists may subscribe to the ambitious version of constitutivism in attempting to offer a justification of our reason judgements that can refute the normative sceptic, we saw in section 5.6.4 that proponents of constructivism are not in general committed to it. They can be modest constitutivists and merely rely on the constitutivist strategy to accommodate objective reasons.²⁷ With this in mind, let us return to Enoch’s objection. Enoch claims that constitutivism fails because being an agent or engaging in reasoning—in general, being the kind of creature or engaging in the kind of activity that is constituted by the facts on which having certain reasons depends—is optional in some sense or another. In fact, we can distinguish three versions of this objection, none which poses a serious challenge to constructivism. The first version of the objection insists that being a reasoner—that is, engaging in the activity that is constituted by the rules of reasoning—is metaphysically optional. Now, of course it is possible that there are creatures that do not engage in any reasoning: various non-human animals illustrate that. Enoch’s point, however, is that there might be creatures that do not engage in the activity that is constitutively governed ²⁷ We will discuss their prospects in doing so in Chapter 6.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

      

183

by the rules of reasoning, but who are otherwise just like us. They would be ‘shmeasoners’, if you will. According to the constructivist account presented here, such creatures would not share any of our practical reasons (they might be subject to ‘shmeasons’, though). Nor could we convince them (or us) that their scepticism with respect to such reasons is unfounded. Hence, neither of the constitutivists’ goals can be achieved when it comes to such shmeasoners. But that does not strike me as much of a problem. How much like us a shmeasoner can be is very much up for debate. Either they do not engage in any rule-governed mental processes at all. Or if they do, those rules will differ substantially from our rules of reasoning, and so that process will not be interpretable as reasoning by us. In both cases, they will appear deeply alien, and any attempt at coordination or exchange will be severely impeded by the impossibility of making sense of their thought processes. If we cannot interpret or treat them as fellow reasoners, why insist that they nevertheless have certain reasons? Denying that they have reasons does not mean that we cannot think about what we would have reason to do if we were in their circumstances. Nor does it prevent us from reasoning about how we may treat them. That, I think, is a plausible stance to take towards such hypothetical shmeasoners.²⁸ The second version of the objection suggests that being a reasoner is epistemically optional, that is, it suggests that we might not be able to conclusively show that some individual is a reasoner rather than a shmeasoner (even if he actually is a reasoner). This is tantamount to denying the ambitious constitutivist’s thesis that we have independent epistemic access to the constitutive facts. Such independent access would be gained by deriving those facts—in our case, the rules of reasoning— from some other, uncontested features of agency or reasoning. If such a derivation is available, then it is possible to establish that even the

²⁸ Depending on how many constitutive rules of reasoning there turn out to be, there might be borderline cases in which not all but some or even most of the rules of shmeasoning coincide with our rules of reasoning. In such cases, the shmeasoners’ judgements on what there is shmeason to do might often align with our judgements on what there is reason to do, and coordination might pose less of a difficulty. However, the greater the similarities, the more plausible it will be to interpret them as disagreeing with us on the constitutive rules of a shared practice that we all engage in, rather than being engaged in a separate practice (see also section 5.5.3).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

184   normative sceptic is subject to the rules of reasoning by pointing out that he too exhibits the features in question. Hence, without further argument, this version of the objection is begging the question against the ambitious constitutivist. The modest constitutivist, on the other hand, need not deny the objection’s premise. He can admit that it might not be possible to convince everyone that they are reasoners, whatever normative claims they happen to accept or reject, because his goal is not to refute the normative sceptic. Finally, according to the third way of understanding Enoch’s objection, being a reasoner is normatively optional. That is, even if it turns out that being a reasoner is inescapable—so that it is impossible for an agent not to be a reasoner—and even if this can be conclusively demonstrated, it might still be the case that the agent does not have any reason to be a reasoner and comply with the rules of reasoning. Much of what Enoch writes suggests that this is what he really has in mind. This is most evident in the following passage: even if you find yourself engaging in a kind of activity, and, indeed, even if you find yourself inescapably engaging in it . . . , and even if that activity is constitutively governed by some norm or is constitutively directed at some aim, this does not suffice for you to have a reason to obey that norm or set your sights on that aim. Rather, what is also needed is that you have a reason to engage in that activity. (Enoch 2011b, 210–11)

While this might raise a problem for some of the constitutivist theories that Enoch discusses, it does not challenge the constructivist account presented here. Both ambitious and modest advocates of that account can agree that reasoning’s being constitutively governed by certain rules does not suffice for anyone to have a reason to obey these rules. Constructivists do not claim that reasoners necessarily have a reason to obey these rules; in fact, they insist that complying with them is correct even if you have no reason to comply (section 5.3). This last version of Enoch’s objection should only worry constitutivists who want to derive the reason to, for example, avoid pain from a reason to comply with the constitutive rules of reasoning, or from a reason to be a reasoner. But that

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

      

185

is not what constructivists do. Their answer to questions like ‘Why should I accept that I have a reason to avoid pain?’ is that the corresponding reasoning complies with the constitutive rules of reasoning and is thus correct. Establishing this connection between reasons and the rules of correct reasoning requires a lot of argument and commits the constructivist to various controversial claims. But once it is established— or is granted for the sake of the argument, as Enoch (2006, 178) does before raising his objection—no further claims to the effect that there is a reason to comply with those rules, or a reason to engage in the activity that is governed by them, are required to derive a truth about reasons from a truth about the rules of reasoning. Of course, the constructivist’s interlocutor might deny that the relevant kind of reasoning would be correct. That can be understood either as opening a discussion about the right conception of the rules of reasoning, in which the constructivist can happily participate. Or it can be understood as the interlocutor’s claim that his reasoning is not governed by the constructivist’s set of rules, but by a different one (‘Reasoning in that way might be correct for you, but it is not for me.’). That, however, cannot be literally true, because an activity that is governed by a different set of rules would not be reasoning (compare ‘I’m playing chess according to different rules from you.’). What this response would amount to is thus the claim that the interlocutor is actually a shmeasoner, not a reasoner. That might be true, but it takes us back to the first version of the objection, which, as I argued above, does not pose a serious problem for the constructivist.

5.7 The status of the rules of reasoning So far in this chapter, little has been said about the nature or the status of the rules of reasoning. What does it mean to say that reasoning is governed by a certain rule? If the constructivist account of practical reasons presented here is correct, such facts are what ground facts about what is a reason for doing what. Asking about the status of these rules thus amounts to asking what the most fundamental grounds of our practical reasons are, what those reasons ultimately bottom out in.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

186   Without answering this question, some might object, little has been won by offering a constructivist account of practical reasons. Now, it is true that many of the questions that meta-ethicists ask about practical reasons can also be asked about the rules of reasoning. But does that mean that we cannot make headway on answering the former if we do not also answer the latter? I do not think that this would be a fair assessment of the constructivist account as it has been presented so far. How much is gained by it depends on the aims you pursue in offering it, what questions you seek to answer. If your aim is to determine whether our commitment to certain truths about practical reasons can be reconciled with a naturalistic picture of the world, then you will not be satisfied with an account that grounds those truths in certain facts about the rules of reasoning, unless it is shown how they can be reconciled with that picture. But allaying naturalistic worries about practical reasons is not the only reason for offering a meta-ethical account of those reasons. Your aim might also just be to understand better what it is for some consideration to be a reason for an agent to do something, and to find a general and informative account of when this is the case, one that does simply not refer to another practical reason from which the one in question can somehow be derived. If this is your aim, the constructivist account has something substantive to offer, even if what it says about the rules of reasoning does not go beyond showing that they need not be understood in terms of practical reasons. Compare the buck-passing account of value, according to which for something to be valuable is for it to have other (non-evaluative) properties that provide reasons for valuing it, that is, for having certain positive attitudes towards it or behaving in certain ways with regard to it (Scanlon 1998, 96; cf. Suikkanen 2009). That account surely offers an understanding of what it is for something to be valuable, even if it is not combined with any comprehensive account of the nature of reasons over and above showing that reasons need not in turn be spelled out in terms of values. On its own, it does not suffice to reconcile the existence of value with a naturalistic picture of the world—that would also require a naturalistic account of reasons. But it certainly provides a substantive account of the nature of value, one that is compatible with naturalistic accounts of reasons, like Schroeder’s (2007), as well as non-naturalistic ones, like Scanlon’s own (1998; 2014).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

      

187

Similarly, the constructivist about practical reasons has options when it comes to accounts of the nature of rules of reasoning. Of course, which of these options the constructivist chooses has consequences for what his overall picture of normativity and our practical lives looks like, and it may be decisive for the acceptability of that picture from a naturalistic point of view. But unless the constructivist chooses an account of those rules that explicates them in terms of practical reasons,²⁹ that choice does not undermine his position’s status as a coherent, substantive and distinct account of practical reasons. With that in mind, let me illustrate some of the constructivist’s options by very briefly outlining two accounts of the nature of the rules of reasoning.³⁰ According to the first one, those rules simply capture certain basic truths about what it is to reason well. These truths are part of a broader account of good psychological dispositions—or virtuous character traits, if you will (cf. Setiya 2007, part two). On this account, identifying the constitutive rules of reasoning amounts to providing an evaluative theory of an important part of practical thought, and the facts that the constructivist takes to ground reason facts turn out to be a certain class of evaluative facts. The constructivist who opts for such an account thus offers an explanation of practical reasons in terms of a certain kind of value. This version of constructivism will not be particularly attractive to the naturalist, who is no less suspicious of values than he is of reasons. But it does provide an informative account of what it is for something to be a reason for a certain action, and it offers an interesting alternative to views like Scanlon’s (2014) fundamentalism and Schroeder’s (2007) reductivism, as well as to other value-based account of reasons (Maguire 2016). The second account that constructivists might opt for differs significantly from the first. It starts by addressing the following question: what is it for an activity to be governed by a certain set of rules? Its answer refers to the attitudes of the participants in that activity, specifically their acceptance of those rules, which involves dispositions to comply with ²⁹ Which they need not do, as we saw in section 5.3. ³⁰ See also James (2012) for a helpful discussion of different ways for constructivists to characterize the status of the norms of correct practical reasoning, and for a specific proposal according to which those norms are fixed by our reasoning practice.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

188   them and dispositions to respond in certain ways to particular moves in that activity, those of others as well as one’s own. This leaves many questions open, but it already allows us to see what is crucial about this option: it takes rule acceptance—understood as a complex syndrome of dispositions rather than a propositional attitude (cf. section 5.5.4)—to be explanatorily prior to rule governance. This turns the identification of the rules of an activity like reasoning at least partly into an empirical endeavour, one that is concerned with the psychological reality of its participants. At least for other rule-governed activities like playing chess or using language, this strikes me as a very plausible approach. Now, while the rules of chess are constitutive of that activity, it is a matter of contingency that we are playing a game with those rules (i.e. that we are playing chess rather than shmess). Does the same hold for reasoning, according to this option? That depends on how its advocates answer another question: why are we engaged in an activity that is governed by those rules? In the case of chess, social conventions will play an important part in answering that question. But an answer might also point to a specific function or functions (evolutionary or otherwise) of the activity in question which favour or even necessitate certain rules.³¹ In that case, it will not be an accident but a robust anthropological fact that we are engaged in reasoning rather than shmeasoning.³² This second account of the nature of the rules of reasoning characterizes them in broadly empirical, non-normative terms. Adopting it would thus allow the constructivist to reconcile our commitment to certain truths about practical reasons with a naturalistic picture of the world. Both accounts have been sketched here only in their roughest outlines. And while my sympathies tend towards the second one, I will not develop either of them any further in this book, as doing so would go beyond its scope. Here, the two options only serve as examples. Constructivists are not committed to either of them.

³¹ As I suggested in section 5.6.3, candidates for such functions of practical reasoning are that it enables an agent to navigate her surroundings, to maintain her agency, and to coordinate her behaviour with others. ³² Note that because the rules are constitutive of the activity, it is nevertheless true that reasoning is necessarily governed by the rules that govern it, just as chess is necessarily governed by its rules.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

6 Mind-dependence and objectivity Constructivism is usually characterized as a meta-ethical view that endorses the mind-dependence of practical reasons (see, e.g., ShaferLandau 2003, ch. 1; Street 2010; James 2012). What exactly that amounts to, however, can be understood quite differently. According to the definition proposed in Chapter 1.4, an account of practical reasons accepts mind-dependence if it takes facts about our mental states or activities to figure ineliminably in an account of why some consideration is a reason for an agent to do something. This is a rather broad conception of mind-dependence: it does not just include views that take reasons to depend on our actual mental states, but also views that take them to depend on the hypothetical mental states of our idealized counterparts, for example. What is relevant is whether, in explaining why there are practical reasons, we need to refer to certain mental states or activities. By grounding reason facts in facts about practical reasoning and the rules that constitute that activity, the account developed in the previous chapters meets that criterion. But so do many of constructivism’s alternatives, including the views proposed by Michael Smith (1994) and Mark Schroeder (2007). Views that take our practical reasons to be mind-dependent are often met with the objection that they are unable to accommodate the objectivity of those reasons and our judgements about them. The aim of this chapter is to address this objection and to show that at least the constructivist’s version of mind-dependence does not conflict with the objectivity of our practical reasons. My first step will be to show that there are many different options for filling in the details of minddependence as I have defined it. The claim that reasons are minddependent can be understood in a variety of ways, and some of them will be more susceptible to those objections than others. In section 6.1, I will hence distinguish different forms of mind-dependence and identify Constructing Practical Reasons. Andreas Müller, Oxford University Press (2020). © Andreas Müller. DOI: 10.1093/oso/9780198754329.001.0001

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

190 -   the particular version of the thesis to which the constructivist account developed here is committed. Since those who emphasize constructivism’s commitment to mind-dependence often characterize it as taking a Euthyphronic position on practical reasons, I will briefly consider how this characterization relates to the specific version of mind-dependence that is distinctive of constructivism (section 6.2). After these clarifications, I will turn to the worry about objectivity. On a closer look, those pressing that worry are concerned with a number of distinct issues: the fallibility of our reason judgements, the universality of some of our reasons, and the modal robustness of some of those reasons. I will discuss the compatibility of constructivism and its siblings with fallibility and universality in section 6.3; their ability to accommodate modal robustness is the subject of section 6.4.

6.1 The varieties of mind-dependence Particular versions of the claim that practical reasons depend on certain mental states or activities can differ in three respects: firstly, what facts about the psychology of agents do they take reason facts to depend upon? Secondly, what kind of dependence relation do they take to hold between the two? Thirdly, what level of dependence do they endorse?¹ I will address these points of divergence in that order, highlighting what answer the constructivist account developed here gives to each of these questions.

6.1.1 What facts? Many of those who claim that facts about reasons (or values) are minddependent take them to depend on a particular kind of mental state. Those who favour a Humean approach to practical reasons take facts

¹ I do not take this to be the only dimensions along which different versions of minddependence can vary; they simply are the ones that are most relevant to our discussion in this chapter. It will become clearer what I mean by the ‘level of dependence’ in section 6.1.3.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

   -

191

about what an agent has reason to do to depend on facts about that agent’s motivational states, for instance about her desires. Mark Schroeder, for example, defends a view according to which for some consideration to be a reason for an agent to perform a certain action is for that agent to have a desire such that the consideration partly explains why the action is a means for the agent to promote the satisfaction of that desire (2007, 59). David Lewis and Michael Smith too place (higherorder) desires at the centre of their views. Lewis (1989) suggests that something is a value if and only if we would be disposed, under ideal conditions, to desire desiring it, and Smith argues that to say that an agent has a reason to do A in certain circumstances is to say that this agent’s ideally rational counterpart would desire her actual self to do A in those circumstances (1994, 180–1). Constructivism takes practical reasons to be grounded in facts about a particular mental activity: whether something is a reason to perform a certain action ultimately depends on whether a certain episode of practical reasoning would be sound. That dependence is established in two steps: reason facts are grounded in the truth of reason judgements, which is in turn grounded in the soundness of episodes of practical reasoning. Thus, constructivism also takes practical reasons to depend on a certain kind of mental state, though one that differs from the kind that Humeans require for the explanation of an agent’s actions. Importantly, while the Humean alternatives to constructivism just mentioned agree on the kind of mental state they take our reasons to depend on, they disagree on what sort of facts involving those mental states are relevant. Schroeder takes an agent’s reasons to depend on facts about her actual desires, while Lewis and Smith consider the desires an agent would have if certain idealizing conditions were met, that is, they take an agent’s reasons to depend on her hypothetical desires. Similarly, the constructivist account does not explain reasons in terms of the reason judgements anyone actually has, or the episodes of reasoning anyone actually engages in. Rather, we can think of constructivism as taking reasons to depend on the hypothetical mental activities of ideal reasoners, that is, those counterparts of us that are fully informed and fully compliant with the standards of correct reasoning. The reason judgements that guide their reasoning determine what reasons we actually

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

192 -   have, just as our ideally rational counterpart’s desires determine what reasons we actually have on Smith’s (1994) account.

6.1.2 What kind of dependence? Different versions of the mind-dependence thesis can also differ in the kind of dependence relation they take to hold between reason facts and mental facts. Not every kind of dependence will yield an interesting version of mind-dependence. Take, for example, the relation between the fact that there is a glass of water on my desk and the fact that I desire to drink water while I am writing. If I did not have that desire, there would be no glass of water on my desk, because I would not have put one there. It is clear, I take it, that accepting such a causal dependence of some facts on an agent’s motivational attitudes does not make for an interesting philosophical thesis about the status of those facts. Unsurprisingly, the various views just mentioned understand the relation that holds between reasons (or values) and mental states in other ways. Some of them seem to take facts about reasons to be identical to facts about the agent’s psychology. According to Smith (1994), when we talk about our practical reasons, we are really talking about the desires of our fully rational counterparts. Similarly, Peter Railton (1986) identifies facts about value with facts about our (idealized) evaluative attitudes. Schroeder, on the other hand, does not claim to identify facts about reasons with any other facts. Instead, he characterizes his project as an ‘analysis of properties’ that is concerned with establishing ‘constitutive explanations’ (2007, 61–6). Hence, Schroeder can be read as suggesting that the kind of dependence in question is to be understood in terms of the grounding relation introduced in Chapter 1.3. Constructivism too takes reason facts to be grounded in (and hence explicable in terms of) facts about sound reasoning. Note that unlike grounding, identity is a symmetric and reflexive relation, which conflicts with the sort of directedness that talk of ‘dependence’ appears to implicate. Nevertheless, in order to respect the classificatory use of ‘mind-dependence’ in the literature and to have a common term to refer to all of the views just

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

   -

193

mentioned, I will use the notion of dependence widely so as to include both grounding and identity.

6.1.3 What level of dependence? Finally, particular versions of mind-dependence can differ in what they take to be the primary level of mind-dependence: does the connection between reasons and our mental states or activities come into play, as it were, at the conceptual level, at the metaphysical level, or at the level of truth? Let me explain how I understand these distinctions. Some advocates of mind-dependence endorse that claim because they think that the concept of a reason can be analysed in the relevant psychological terms. Smith, for example, argues that the concept of a normative reason can be analysed in terms of the desires we would have if we were ideally rational, such that claims like ‘Sam has a reason to smoke a cigarette’ can be analysed as ‘If Sam were fully rational, he would desire that he smokes a cigarette’ (see 1994, ch. 5). Lewis too proposes his dispositional theory of value as an ‘analytic definition of value’ (1989, 113), and Railton characterizes his account as proposing ‘reforming naturalistic definitions’ of certain normative concepts (1986, 204). On views that endorse mind-dependence because they purport to analyse or define the concept of a reason in psychological terms, the primary level of dependence is the conceptual one: the connection between reasons and mental states already comes into play at the level of the concepts we use in thinking about them. Of course, what such views claim about the concept of a reason also has consequences at the metaphysical level. Analytically equivalent concepts—that is, concepts that can be analysed or defined in terms of each other—are co-extensional, and analytically equivalent judgements are typically taken to describe the same facts.² Hence, if reason ² Some metaphysicians propose individuating facts more finely, but the assumption in the text appears to be implicit in much of twentieth and twenty-first century meta-ethics. After all, one of the main motivations for establishing that moral and other normative judgements can be analysed as, for example, psychological judgements is to show that the former do not commit us to the existence of any (potentially metaphysically queer) facts over and above the facts with

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

194 -   judgements can be analysed in terms of judgements about an agent’s psychology, then reason facts are identical to the corresponding psychological facts. If ‘Sam has a reason to smoke a cigarette in certain circumstances’ can be analysed as ‘If Sam were fully rational, he would desire that he smokes a cigarette in the circumstances in question’, then the fact that Sam has a reason to smoke a cigarette in certain circumstances just is the fact that Sam would desire that he smokes a cigarette in those circumstances if he were fully rational. Hence, dependence on the conceptual level entails dependence at the metaphysical level (though only a borderline case of dependence, i.e. identity). Importantly, however, the reverse does not hold: you can take reason facts to depend on certain psychological facts while denying that the concept of a reason can be analysed in psychological terms. Schroeder, for example, explicitly denies that he offers an analysis of the concept of a reason (2007, 61–72). In general, concepts can be co-extensional without being analytically equivalent. Consequently, it is possible to hold that facts about reasons are identical to certain psychological facts while taking the concept of a reason to be unanalysable. Moreover, if one takes facts about reasons to be grounded in (rather than identical to) certain psychological facts, one is committed to denying that reason judgements and the relevant psychological judgements are analytically equivalent (for that would entail the identity of the corresponding facts). Hence, it is possible to accept mind-dependence at the metaphysical level without accepting it at the conceptual level. It might seem obvious that accepting mind-dependence at the metaphysical level (whether or not one also accepts it at the conceptual level) will require you to take the truth of particular reason judgements to depend on certain psychological facts as well. If reason judgements are about facts that are either identical to or grounded in psychological facts, then the truth of those judgements will depend on those psychological facts too, simply because the facts that the judgements are about depend on them. Both Smith and Schroeder, I suspect, will take this to be trivial. On their views, the truth of reason judgements depends on psychological which the latter are concerned. This seems to presuppose that analytically equivalent judgements are concerned with (and that their truth commits us to the existence of) the same facts.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

   -

195

facts because it depends on the accurate representation of facts that are either identical to or grounded in psychological facts. This explanation, however, makes an assumption about the role that the facts that true reason judgements are about play in an explanation of why those judgements are true that constructivists reject. According to the constructivist’s conception of the truth of reason judgements, the facts that those judgements are about (i.e. the facts of which they say that they obtain) come apart from the facts that explain their truth (Chapter 4.6). Reason judgements are not true because the facts they are about obtain; therefore, their truth does not depend on psychological facts because the facts they are about depend on those facts either. Of course, that does not mean that constructivists deny that the truth of reason judgements depends on certain psychological facts, for according to the conception of truth they favour, the truth of particular reason judgements is grounded in certain facts about the episodes of reasoning these judgements are apt to guide. Constructivists just deny that these facts make reason judgements true because those judgements accurately represent them. Hence, they agree that the truth of reason judgements depends on certain psychological facts, but they take this to be a consequence of what truth consists in for those judgements, rather than of what they are about. Moreover, since constructivism reverses the representationalist’s order of explanation in Correspondence and takes particular reason facts to obtain because the corresponding reason judgement is true, those facts ultimately depend on the psychological facts that ground the truth of that judgement. That is, constructivism too accepts mind-dependence at the metaphysical level, but it does so as a consequence of accepting such dependence at the level of truth.

6.1.4 Conceptual, metaphysical, and alethic mind-dependence While all of the views we considered endorse mind-dependence at the metaphysical level—that is, at the level of the facts our reason judgements are about—, they differ on whether this is the primary level of dependence, as it were, or whether it follows from dependence on

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

196 -   another level. I take Schroeder’s position as an example for the view that the metaphysical level is the primary level of dependence, a view that we can call metaphysical mind-dependence. Those who, like Smith, think that the dependence on the metaphysical level follows from the analysability of the concept of a reason in psychological terms and hence from dependence on the conceptual level, endorse what might be called conceptual mind-dependence. Constructivism, on the other hand, does not take the metaphysical level to be the primary level of dependence, but it does not claim that it follows from dependence on the conceptual level either. Instead, constructivism endorses a distinct version of minddependence that I will call alethic mind-dependence. On this version, the mind-dependence on the level of truth is primary, and the dependence on the metaphysical level follows from it. The contrast between metaphysical and alethic mind-dependence is illustrated in Figure 6.1. The crucial difference can be summarized as follows: according to metaphysical mind-dependence, the relevant psychological facts directly ground (or are identical to) the reason facts, and the relation between the reason facts and our reason judgements is understood along the lines of the representationalist picture: the judgements are about reason facts, and they are true—if they are true— because they accurately represent those facts. Hence, particular reason facts make the judgements that are about them true. According to alethic mind-dependence, however, the psychological facts ground the reason facts only indirectly by making true those reason judgements that are about them. The reason facts are ontological shadows of the truth of those judgements, that is, the facts do not explain the judgements’ truth but are, conversely, themselves explained by it. The possibility of such a version of mind-dependence is opened up by rejecting the Metaphysical MindDependence:

psychological facts

Alethic Mind- psychological facts Dependence:

ground/ are identical to

make true

reason facts

are about make true

reason judgements

Figure 6.1 Metaphysical vs. alethic mind-dependence.

are about ground

reason judgements

reason facts

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

197

representationalist theory of truth—that is, the idea that truth consists in representational accuracy—and adopting a conception of truth such as the Wrightean one developed in Chapter 4. Doing so allows us to distinguish between the facts that make reason judgements true and the facts those judgements are about. This is why, in the part of Figure 6.1 that illustrates alethic mind-dependence, reason facts can be explanatorily downstream from our true reason judgements, while the psychological facts precede those judgements in the order of explanation.

6.2 Siding with Euthyphro The distinction between conceptual, metaphysical, and alethic minddependence brings out a problem with another way of characterizing meta-ethical constructivism as a view that endorses the minddependence of facts about reasons. This characterization portrays constructivists as taking Euthyphro’s side in his famous dispute with Socrates on the pious in Plato’s Euthyphro. At a crucial point in the argument, Socrates and Euthyphro agree that something is loved by the gods if and only if it is pious, but the two appear to disagree on the direction of determination in this biconditional. While Socrates claims that something is god-beloved because it is pious, Euthyphro is taken to maintain that something is pious because it is loved by the gods and hence that being pious is, in some sense, mind-dependent.³ The result of applying a secular version of Euthyphro’s position to facts about reasons or values (rather than about what is pious) is seen by many as a model for constructivism. This includes two of its most recent advocates, as the following quotes by Aaron James and Sharon Street illustrate:

³ As an interpretation of the actual argument in Plato’s dialogue and Euthyphro’s role therein, this is disputable: see Irwin (2006). Nevertheless, the contrast between the two directions of determination in biconditionals of the kind discussed in the main text has become known as the Euthyphro contrast, and the two alternative positions have become associated with the dialogue’s two main characters. It is this contrast that those who characterize constructivism as a Euthyphronic position have in mind. For more general discussions of this contrast and its relevance, see Wright (1992, 108–39) and Johnston (1993, 121–6).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

198 -   [T]he common ground between Euthyphro and Socrates, ‘an act is pious if and only if it is loved by the gods’, allows Euthyphro to claim that an act is pious because it is loved by the gods and Socrates to claim that it is loved by the gods because it is pious. [This] difference in order of determination just is the central contrast between constructivism, on the one hand, and either realism or an interdependence thesis, on the other. (James 2007, 323) [T]he key point at issue between realists and antirealists is the answer to the central question of Plato’s Euthyphro (in rough secular paraphrase), namely whether things are valuable ultimately because we value them (antirealism), or whether we value things ultimately because they possess a value independent of us (realism) . . . . Metaethical constructivism falls squarely on the antirealist side of this divide. (Street 2010, 370–1)

Once we acknowledge the distinction between alethic mind-dependence and its alternatives, however, we see that this characterization of constructivism leaves open an important question and might thus be misleading. It seems that what distinguishes a Euthyphronic position on practical reasons is that it endorses some version of the following Euthyphronic Determination Claim: (EDC) S₁ has a reason to do A if and only if and because S₂ has [attitude, S₁, A].⁴ Now, while advocates of conceptual, metaphysical, and alethic minddependence can all accept some version of (EDC),⁵ they will have very

⁴ The expression ‘[attitude, S₁, A]’ is a placeholder for an attitude that is specific with respect to both S₁ and A, such as a desire that S₁ does A or a disposition to blame S₁ for not doing A, for example. Note also that S₁ and S₂ can be identical. ⁵ Assuming that the advocate of conceptual mind-dependence can make sense of the asymmetrical relation of determination signified by the ‘because’ even though the analytical equivalence of a reason statement and its psychological analysans appears to be a symmetrical relation.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

199

different rationales for doing so. If you endorse the conceptual minddependence of practical reasons, you will accept a version of (EDC) because you think that the concept of a reason that appears on its lefthand side can be analysed in terms of the attitude that you fill in for the placeholder ‘[attitude, S₁, A]’. On Smith’s view, for example, you can substitute ‘a desire that S₁ does A’ in the right-hand side of (EDC) and take S₂ to be the ideally rational counterpart of S₁. If, on the other hand, you endorse metaphysical mind-dependence, you will deny such a close conceptual relation between the two sides of (EDC), but you will nevertheless accept a version of it because you think that the fact picked out by its left-hand side is identical to or grounded in the fact picked out by (your version of) its right-hand side.⁶ Schroeder’s view, for example, appears to entail a version of (EDC) where S₂ just is S₁ and the relevant attitude is any desire whose satisfaction will be promoted by doing A. As we saw in section 6.1.4, however, constructivism endorses alethic mind-dependence. Now, endorsing alethic mind-dependence will also prompt you to accept a version of (EDC), but only because you take the judgement that affirms its left-hand side to be true in virtue of the attitudinal fact on its right-hand side. In other words, constructivists accept the relevant version of (EDC) only because their conception of the truth of reason judgements prompts them to accept both the corresponding version of: (EDC-T) The judgement that S₁ has a reason to do A is true if and only if and because S₂ has [attitude, S₁, A] and the constructivist version of Correspondence (see Chapter 1.3): (C) S₁ has a reason to do A if and only if and because the judgement that S₁ has a reason to do A is true. ⁶ If you take the facts to be identical, then a similar proviso to the one in n. 5 above applies. Since those who argue for such identities nevertheless claim that their account amounts to a ‘reduction’ of practical reasons, they seem intent on meeting that proviso, for ‘is reducible to’ appears to be asymmetric too. As I argued in Chapter 1.3, this suggests that what they actually have in mind when presenting their ‘reduction’ might be a claim about grounding, not identity.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

200 -   Together, (EDC-T) and (C) entail (EDC). Hence, constructivism entails a version of (EDC) that takes the relevant attitude to be the judgement something is a reason for S₁ to do A and specifies S₂ as a counterpart of S₁ that qualifies as an ideal reasoner (section 6.1.1). Thus, understanding constructivism as the position that takes Euthyphro’s side in a dispute on the order of determination between reasons and attitudes yields too wide a characterization. Neither Smith (1994) nor Schroeder (2007) should be classified as constructivist views, for they both appear to accept Representationalism and the representationalist theory of truth. It is true that, like these other views, constructivism is committed to mind-dependence and is, in this sense, a Euthyphronic position. But taking this to be its main characteristic lumps together the versions of mind-dependence that we distinguished in the previous section and thus ignores important differences in the underlying accounts of reason judgements and their truth.

6.3 Fallibility and universality Views that endorse mind-dependence and take practical reasons to depend on certain facts about our mental states or activities are often met with the suspicion that they are incompatible with the objectivity of those reasons. Underlying this suspicion are a number of quite distinct objections that are worth discussing separately. The objectivity-related objections that I want to focus on accuse constructivism and its siblings of being unable to accommodate: (i) the fallibility of our reason judgements, that is, the fact that we can be mistaken about our reasons; (ii) the universality of at least some of our reasons, that is, the fact that they are shared by all agents; or (iii) the modal robustness of our reasons, that is, their insensitivity to counterfactual variations of our psychological make-up.⁷ ⁷ There is also a sense of ‘objectivity’ in which a judgement’s being objectively true or false means that its truth value is fixed by some independent object, which it represents either

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

201

I will address these objections individually, starting with the first two in sections 6. 3.1–6.3.3 and continuing with the third in section 6.4.

6.3.1 Accommodating fallibility Sometimes, we are mistaken about our reasons. That is to say, sometimes we judge some consideration to be a reason to perform a certain action even though it is not actually a reason in favour of that action. What we think we have reason to do and what we actually have reason to do need not coincide. Our reason judgements, in other words, are fallible: Fallibility It is possible that an agent judges that she has a reason to do A even though she does not have a reason to do A. This is one sense in which our reasons are objective: thinking that we have one does not make it so.⁸ A similar claim appears to hold for other attitudes as well. You can desire that you do something, or even intend to do it, even though you actually have no reason to do it. Like our judgements, our desires and intentions can come apart from our reasons too. We can put this as follows: Discrepancy An agent can have attitude α towards doing A even though she does not have a reason to do A. Fallibility is simply the instance of Discrepancy in which α is a (firstperson) reason judgement in favour of the action in question. Whether or not Discrepancy holds for all attitudes an agent might have is more accurately or not. However, insisting that reason judgements are objective in that sense would simply beg the question against constructivism. ⁸ The importance of accommodating the possibility of being mistaken in one’s reason judgements is highlighted by, e.g., Thomas Scanlon (1998, 62–4) and Derek Parfit (2011b, 389).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

202 -   controversial than the truth of Fallibility. Still, I take it that many other instances will be intuitively plausible as well, and that most people, including most philosophers, will accept Discrepancy in general. Hence, being unable to accommodate particular instances would add significantly to the theoretical costs of any account of practical reasons, at least in the absence of a plausible explanation. Are views that endorse mind-dependence compatible with Discrepancy? In section 6.1.1, I distinguished between those among such views that take an agent’s reasons to depend on her actual attitudes, for example, her actual desires, and those that take those reasons to depend on a certain class of ideal attitudes, such as the desires the agent would have in certain hypothetical circumstances. Constructivism is among the latter views: whether or not something is a reason for an agent to perform a certain action depends on whether the corresponding reason judgement is apt to guide a sound episode of reasoning, that is, whether it would guide the reasoning of an ideal reasoner. Thus, those judgements need not coincide with any of the agent’s actual attitudes towards the action, so constructivism need not deny Discrepancy for any of an agent’s actual attitudes, and particularly not for her reason judgements. Versions of mind-dependence that take practical reasons to depend on an agent’s actual attitudes, on the other hand, face significant difficulties in accommodating plausible instances of Discrepancy. On the most extreme version, an agent would have a reason whenever she actually takes herself to have one. In that case, our actual reason judgements would be infallible. But while less extreme versions, which take reasons to depend on other kinds of attitudes, can allow for an agent to be mistaken in her judgements, they have to reject Discrepancy for those other attitudes and thus grant them a status akin to infallibility. Schroeder’s Humean account of reasons, for example, entails that it is impossible for an agent to come to desire doing something without thereby acquiring a reason to do it (or whatever facilitates it).⁹ Moreover, this seems to ⁹ On Schroeder’s account, ‘[f]or R to be a reason for X to do A is for there to be some p such that X has a desire whose object is p, and the truth of R is part of what explains why X’s doing A promotes p’ (2007, 59). If we substitute ‘that X does A’ for ‘p’, we get that anything that explains why X’s doing A promotes that she does A is a reason for her to do A. And since this

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

203

include desires that are based on false beliefs.¹⁰ But the idea that even those desires fail Discrepancy is extremely implausible: if I falsely believe that there is gin in the glass in front of me when in fact it is petroleum, and as a result form the desire to drink its contents, that hardly makes it the case that I have a reason to drink it.¹¹

6.3.2 Accommodating universality: actual attitudeversions of mind-dependence The fallibility of our judgements about practical reasons is just one sense in which they are objective. Another way of understanding the claim that a certain reason is objective is that it is shared by all agents. Many of our reasons may depend on our individual desires, tastes, goals, etc. But some reasons, it seems, are objective in the sense of being universal: they are reasons for everyone, whatever their desires, tastes, or goals are. Moral reasons come to mind here: surely everyone has a reason to help those in need, to avoid inflicting gratuitous harm on others, and to keep their promises. But some other reasons too appear to be universally shared. Derek Parfit claims, for example, that everyone has a reason to avoid future agony, even those who do not want to avoid it and who would not want to avoid it after ideal deliberation (2011a, 73–6). He argues that what he calls ‘subjective theories’ about reasons—which, as I understand him, include constructivism as well as the other views considered in this explanandum (i.e. the fact that X’s doing A promotes that she does A) is quite trivial, there certainly will be something that explains it, perhaps some aspect of the promotion relation. Hence, whenever an agent desires that she does A, she has a reason to do A. ¹⁰ Very roughly and in somewhat simplified form, Schroeder takes a desire whose object is p to be the mental state that grounds an agent’s disposition to perform actions that obviously promote p and to find considerations that help explain why performing those actions promotes p salient (see Schroeder 2007, 156–7). ¹¹ This example is due to Bernard Williams (1980, 102). Schroeder is prepared to bite the bullet that his account of reasons entails that we have a lot more reasons than we usually think we have, but he intends to sweeten it by arguing that many of those reasons will have very little weight, and that our negative existential intuitions about reasons are not trustworthy (2007, ch. 5). Note however, that the objection in the main text does not rely on pitting the implications of his theory for particular cases against our intuitions concerning whether or not an agent has a reason in those cases. My point is that, quite generally, desires that are based on false beliefs satisfy Discrepancy.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

204 -   chapter (see 2011a, 58–65)—are unable to accommodate that we all have that reason, and therefore ought to be rejected. Is Parfit right, or do views that take our reasons to be mind-dependent have the resources to accommodate such universally shared reasons? Again, that depends on the version of mind-dependence such a view endorses. Those who take our reasons to depend on our actual attitudes have two ways of dealing with universal reasons. They can either try to show that the attitudes an agent needs in order to have such a reason are universally shared as well. Perhaps everyone has a reason to avoid pain because everyone has a desire not to be in pain. But Parfit talks about avoiding future agony, and it is much less plausible that everyone has a desire to avoid any future agony (Parfit 2011a, 76). Moreover, it seems overly optimistic to assume that everyone desires to help those in need, for example. As a general strategy to accommodate universally shared reasons, this approach thus seems not very promising. A second way for advocates of actual attitude-versions of minddependence to deal with those reasons is to show that even though they do depend on an agent’s actual attitudes, those reasons do not require everyone to have the same specific attitude. Instead, they might depend on our mental states in a way that makes any of a wide range of attitudes sufficient for having a particular reason, so that it is much more plausible that everyone of us has at least one of them. Schroeder employs this overdetermination strategy in arguing that his account is compatible with the existence of universally shared reasons: on his account, you have a reason to do A whenever you have some desire whose satisfaction would be promoted by your doing A. So if there are actions that promote the satisfaction of a wide variety of desires, then anyone who has at least one of those desires shares that reason, even if there is no single desire that everyone shares (2007, 108–10).¹² Of course, in order to accommodate the particular examples for universal reasons mentioned above,

¹² Indeed, if there are actions that promote the satisfaction of any desire, and having some desire is necessary for being an agent, then Schroeder’s account would entail that it is a necessary fact that every agent has a reason for that action. Schroeder actually employs the overdetermination strategy to show that his Humean account of reasons is compatible with the fact that some reasons have such a robust modal status. We will get back to this third sense in which our reasons can be objective in section 6.4.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

  

205

advocates of this strategy still have to show that the actions those reasons support do in fact promote the satisfaction of at least one of everyone’s actual desires. That is certainly no easy undertaking.

6.3.3 Constructivism and universality Views that take reasons to depend on certain ideal attitudes, on the other hand, merely have to show that the attitudes that are necessary for an agent to have a certain reason meet the relevant conditions to qualify as ideal, whether or not anyone actually has them. On the constructivist account, a particular agent has a certain reason if the judgement that she has that reason is apt to guide an episode of reasoning that is sound. Hence, a particular agent has a reason to avoid future agony if she could form the intention to avoid future agony as the conclusion of sound reasoning. That reason will thus be universal if every agent could form the intention to avoid future agony in that way. This requires that there are correct ways of practical reasoning whose premise-attitudes include only beliefs. If the intention to avoid future agony could only be formed by way of reasoning from a set of premiseattitudes that concludes an intention, then that reasoning will only qualify as sound if the premise-intention is supported by another reason. To avoid an infinite regress, we have to assume that at some point, the required premise-intention can be formed through sound reasoning that does not require another premise-intention (see Chapter 4.5.2). Therefore, constructivism entails that a reason to do A is shared by all agents if and only if correct practical reasoning alone could take an agent from an initial set of premise-attitudes that includes only true beliefs to the intention to do A (be it by a single episode or by a number of successive episodes whose respective premise-attitudes include only true beliefs and intentions formed as conclusions of previous episodes). In other words, there are universally shared reasons to the extent that there are correct ways of practical reasoning whose premise-episodes include only beliefs. This connection is independently plausible. Suppose that the mere fact that someone needs help is supposed to give every agent a reason to help

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

206 -   that person. It would be surprising if, among the agents who are aware of the fact that a person needs help, only some would be in a position to comply with the reason that fact provides by way of correct reasoning. It seems that if the fact that someone needs help is sufficient to provide a reason for an agent to help that person, then the agent’s acknowledgement of that fact should put her in a position to form the intention to help that person by correctly exercising her capacity to reason.¹³ Thus, whether or not the constructivist account can accommodate universally shared reasons depends on the details of the underlying theory of correct reasoning, in particular on whether that theory admits of the possibility of reasoning correctly to a certain intention from premise-attitudes that include only beliefs. This issue was left open in the discussion of correct reasoning in Chapter 5, because deciding it would require us to defend a particular substantive theory of correct reasoning. I argued in Chapter 5.6 that constructivists need not defend such a theory independently of its implications for a normative theory of reasons. Instead, they can justify the claim that a certain kind of reasoning is correct with an eye to the plausibility of its normative output. Hence, if it is plausible that certain reasons are universally shared by all agents, this provides the constructivist with some support for a theory of correct reasoning that offers the resources to accommodate those reasons.

6.4 Modal robustness 6.4.1 Objectivity and modal robustness That our reason judgements (and our attitudes more generally) are fallible and that some reasons are universally shared are important ideas that people often have in mind when they claim that practical reasons are objective. But they are not the only ways of understanding this claim to objectivity. Sometimes we use terms like ‘objective’ or ‘objectively’ to indicate that we take a statement to have a certain ¹³ See also the discussion of the Deliberative Constraint in Chapter 4.5.3.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

 

207

modal status, that is, to indicate that we take a certain counterfactual statement to be true as well. When you say, for example, that S objectively has a reason to avoid future agony, what you might mean is that S would have a reason to avoid future agony even if certain things were different from how they actually are—even if, for example, S did not judge herself to have a reason to avoid future agony, or did not care about her future agony.¹⁴ If we understand the claim that (some) reasons are objective in this way, then they amount to the claim that those reasons are modally robust with respect to certain parameters, that is, that they are insensitive to counterfactual variations of those parameters: Modal Robustness S has a reason to do A, and if it were not the case that p, then S would (still) have a reason to do A. Of course, Modal Robustness will be most interesting for parameters that concern the agent S and her psychology. If we substitute ‘S judges that she has a reason to do A’ for ‘p’, for example, Modal Robustness tells us that an agent’s reasons are not limited to what she takes to be a reason, which complements the claim that our reason judgements are fallible.¹⁵ Modal Robustness also goes beyond the kind of universality we discussed

¹⁴ Ronald Dworkin (1996; 2011, part I) insists that at least in the moral domain, claims about objectivity must be understood as such counterfactual claims: The claim that abortion is objectively wrong seems equivalent, that is, in ordinary discourse, to [another claim]: that abortion would still be wrong even if no one thought it was. (1996, 98) Dworkin thus thinks that it is impossible to deny the objectivity of moral truths from an ‘external’ meta-ethical perspective while remaining completely neutral on substantive moral issues: to deny that it is objective that abortion is wrong is simply to deny a substantive counterfactual moral claim. Hence, according to Dworkin, there is no principled distinction between meta-ethics and normative ethics; one cannot be done independently of the other. ¹⁵ Note that this instance of Modal Robustness does not entail that the agent’s judgement about her reason to do A is fallible. A judgement is fallible if and only if it is possible that an agent accepts it even though it is false, i.e. if and only if: (1) Possibly (S judges that she has a reason to do A and S does not have a reason to do A) But Modal Robustness tells us that an agent would have a reason to do A even if she did not judge that she has that reason, i.e. it entails: (2) Possibly (S does not judge that she has a reason to do A and S has a reason to do A) Neither of these two claims entails the other.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

208 -   in the previous section. Suppose that an advocate of mind-dependence argues that the attitude that an agent must have in order to have a reason to avoid agony is, as a matter of contingent, psychological fact, shared by all human agents. In that case, all of them have this reason. But this is compatible with denying that any of them would still have that reason if they did not have the attitude in question.¹⁶ In that case, the reason would be universal, but it would not be modally robust with respect to that attitude.¹⁷

6.4.2 Modal robustness and mind-dependence The modal robustness of (some of) our reasons is apt to pose a challenge to views that accept some version of mind-dependence. Let us use ‘pMD(S, A)’ to refer to the fact on which a view that endorses minddependence takes a reason for S to do A to depend. Such a view thus entails the following claim: (D) That S has a reason to do A depends on whether pMD(S, A). Of course, different views will suggest different candidates for those psychological facts. On Schroeder’s view, for example, pMD(S, A) can be replaced by ‘S has a desire whose satisfaction she can promote by doing A’, Smith’s view proposes ‘S’s ideally rational counterpart desires her to do A’, and constructivism suggests ‘a reason judgement in support of S’s doing A is apt to guide an episode of sound reasoning’. ¹⁶ Parfit suggests that the reason to avoid future agony is objective in this further sense, i.e. that it is not merely universally shared but also modally robust with respect to whether agents actually desire to avoid future agony or would desire to avoid it if they had carefully considered all the relevant facts (2011a, 76–7). ¹⁷ In another respect, Modal Robustness is weaker than universality: reasons that are not universally shared can nevertheless be modally robust with respect to all of the parameters that we take to be relevant for the reason to be objective in the sense currently under consideration. Suppose, for example, that Sara has an agent-relative reason to foster Kevin’s intellectual development because Kevin is her son. In that case, the reason to foster Kevin’s intellectual development is not universally shared. Yet it seems to be true that Sara would still have a reason to foster Kevin’s intellectual development if she did not take herself to have that reason, or did not care about Kevin, etc. (Of course, her reason would not be modally robust with respect to the fact that Kevin is her son.)

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

 

209

A potential problem for such views arises because (D) seems to conflict with the instance of Modal Robustness in which ‘pMD(S, A)’ is substituted for ‘p’: (R) S has a reason to do A, and if it were not the case that pMD(S, A), then S would (still) have a reason to do A. Hence, if some reasons are plausibly regarded as insensitive to counterfactual variations of the very facts on which a particular version of minddependence takes them to depend, then views that endorse this version of mind-dependence seem unable to accommodate the modal robustness of those reasons. Moreover, many authors suggest that there are reasons that are modally robust with respect to any psychological facts of this kind, including not only facts about an agent’s actual attitudes, but also facts about the attitudes an agent would have in certain ideal conditions. I take it that Parfit, for example, will insist that there are no claims about an agent or her attitudes that we can substitute for ‘p’ so that it will be true that if it was not the case that p, then S would not have a reason to avoid agony. Similarly, at least some moral reasons will appear to many to be modally robust with respect to any facts about an agent’s psychology. This, it seems, is what we often have in mind when we insist that those reasons are completely objective. Are advocates of mind-dependence forced to deny those claims to objectivity? Do they have to insist that all reasons are sensitive to the counterfactual variation of some psychological facts, namely those on which they take those reasons to depend? In what follows, I will discuss two strategies to avoid this undesirable consequence. The first is in principle available to advocates of any of the versions of minddependence that I distinguished in section 6.1, but as we will see, the solution it provides for the challenge posed by such robust reasons is not fully satisfactory. The second strategy offers a better response because it shows that no real challenge arises from the fact that some reasons are insensitive to the counterfactual variations of the very facts on which they depend. That strategy, however, is only available to those who

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

210 -   understand this dependence along the lines of alethic mind-dependence, and hence only to constructivists.

6.4.3 Dependence on necessary facts The first strategy for advocates of mind-dependence to accommodate reasons that are insensitive to the counterfactual variation of any facts about the agent’s psychology is to argue that the psychological fact on which they take that reason to depend obtains with necessity. To illustrate this strategy, I will use Parfit’s reason to avoid future agony as an example for such a reason. In order to accommodate the broad modal robustness of that reason, an account of practical reasons has to be compatible with the following claim, whatever psychological facts we substitute for ‘p’: (1) If it were not the case that p, then S would (still) have a reason to avoid future agony. Advocates of mind-dependence, however, claim that, like any other reason, the agent’s reason to avoid future agony depends on certain psychological facts pMD(S, avoid future agony). This might seem to force them to deny that the reason is modally robust with respect to that fact, that is, to deny the following instance of the second conjunct of (R): (2) If it were not the case that pMD(S, avoid future agony), then S would (still) have a reason to avoid future agony. But if it is necessarily the case that pMD(S, avoid future agony), then (2) might still turn out to be true. Advocates of mind-dependence might argue that, for (2) to be false, there has to be some possible scenario in which its antecedent is true but its consequent is false. But if it is necessarily the case that pMD(S, avoid future agony), then there is no possible scenario in which the antecedent of (2) is true. So it seems that by arguing that the psychological fact on which they take the reason to

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

 

211

avoid future agony to depend obtains with necessity, advocates of minddependence can accommodate the truth of (2) and hence the broad modal robustness of that reason. Of course, for this strategy to be successful, advocates of minddependence still have to argue for the necessity of the corresponding psychological facts. At least in the case of the reason to avoid future agony, however, this does not appear to be a hopeless task. The exact claim whose necessary truth has to be established depends on the particular version of mind-dependence advocated. On Schroeder’s view, it has to be shown that S necessarily has some desire whose satisfaction would be promoted by avoiding future agony. Given a sufficiently weak understanding of the promotion relation, this might well be plausible: agony will be at least a significant distraction in the pursuit of any desire. On Smith’s account of practical reasons, the relevant claim is that S’s ideally rational counterpart would necessarily desire her to avoid future agony. The plausibility of this claim, of course, depends on how the notion of ideal rationality is spelled out. Smith argues that any agent’s ideally rational counterpart would necessarily desire her to help and not interfere with the present and future exercise of her rational capacities (2012, 318). If that argument is on the right track, then the relevant claim about avoiding future agony seems to follow, for such agony would surely threaten to impair an agent’s exercise of her rational capacities. The constructivist account too entails that the relevant psychological fact obtains necessarily, since it is a fact about the correctness of reasoning from the true premise-belief that an action will prevent future agony to the intention to perform that action.¹⁸ According to the Constitutive Rules Account, the correctness of a certain episode of reasoning is a matter of necessity, because it depends only on whether that episode complies with the constitutive rules of reasoning, and since such rules define the activity they govern, reasoning could not be subject to a different set of rules with which that episode might not comply.

¹⁸ I am assuming that the premise-attitudes of the relevant episode of reasoning include only beliefs because I take the reason to avoid future agony to also be universal; see section 6.3.3.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

212 -  

6.4.4 The truth of counterpossible conditionals There is, however, a general problem with accommodating the modal robustness of certain reasons by arguing for the necessity of the psychological facts on which these reasons are taken to depend. This strategy relies on the idea that counterfactual conditionals like (2) are true if their antecedent is necessarily false. But how to deal with such counterpossibles (i.e. counterfactual conditionals with impossible antecedents) is a controversial issue. On the standard semantics for counterfactual conditionals developed by David Lewis (1973) and Robert Stalnaker (1968), counterpossibles turn out to be true for roughly the reasons given in section 6.4.3. Importantly, however, this makes their truth vacuous and uninformative: if counterpossibles are true simply because their antecedent is impossible, then the content of their consequent is irrelevant for their truth. But showing that (2) is vacuously true seems to be an odd way of accommodating the prima facie non-vacuous idea it was meant to express. Moreover, if (2) is true simply because its antecedent is impossible, then so is the conditional that shares its antecedent but negates its consequent: (3) If it were not the case that pMD(S, avoid future agony), then S would not have a reason to avoid future agony. Hence, if counterfactual conditionals are true whenever their antecedents are impossible, then the strategy of arguing for the necessity of the relevant psychological facts allows advocates of mind-dependence to accommodate the modal robustness of certain reasons only at the cost of entailing the claim that denies their modal robustness too. Moreover, that the standard Lewis-Stalnaker semantics entails that all counterpossibles are vacuously true is often regarded as a serious disadvantage. Critics argue that there seem to be clear examples of counterpossibles that are non-vacuously true, such as ‘If Hobbes had squared the circle, sick children in the mountains of South America at the time would not have cared’ (Nolan 1997, 544), as well as examples of counterpossibles that are non-vacuously false, such as ‘If Hobbes had squared the

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

 

213

circle, sick children in the mountains of South America at the time would have been really excited about it.’ In addition, critics point out that much of ordinary philosophical reasoning seems to involve thinking and debating about counterpossibles, for example, about what would be the case if certain metaphysical doctrines were true: ‘If Plato’s theory of recollection were correct, then we would have existed before we were born’ seems true, whereas the conditional we get by adding a negation to the consequent appears to be false (Brogaard and Salerno 2013, 644; see also Nolan 1997, 539–40). If, however, counterpossibles can be non-trivially true or false, depending on the content of their consequent, then only one of (2) and (3) will be true, and advocates of mind-dependence would have to provide an additional argument for why it should be (2), despite their claim that the reason mentioned in the consequent depends on the fact that the antecedent stipulates not to obtain. Hence, arguing for the necessity of the psychological facts on which certain reasons are taken to depend is not a promising strategy for accommodating the broad modal robustness of those reasons, independently of whether we accept the standard Lewis-Stalnaker treatment of counterpossible conditionals.

6.4.5 Alethic mind-dependence and the extension of ‘reason’ Does that mean that advocates of mind-dependence have to deny that reasons like the reason to avoid future agony are modally robust with respect to the facts on which they take them to depend? No, there is a second strategy to accommodate that modal robustness. As we saw in section 6.4.3, this requires them to accommodate the truth of: (2) If it were not the case that pMD(S, avoid future agony), then S would (still) have a reason to avoid future agony. Now, the reason why views that endorse mind-dependence appear forced to deny that claim is that when: (D) That S has a reason to do A depends on whether pMD(S, A).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

214 -   is applied to the reason to avoid future agony, it appears to entail: (3) If it were not the case that pMD(S, avoid future agony), then S would not have a reason to avoid future agony. But (3) seems to be incompatible with (2). Instead of contesting this and arguing that both (2) and (3) are (vacuously) true because their antecedent is necessarily false, advocates of mind-dependence can avoid this problem by showing that the relevant instance of (D) does not actually entail (3), so that their commitment to the former is perfectly compatible with the truth of (2). In the remainder of this section, I will show that this entailment from dependence to counterfactual sensitivity fails, if we understand the dependence claim along the lines of alethic minddependence. Hence, constructivists need not deny modal robustness. According to the constructivist account of practical reasons, reason facts depend on the truth of reason judgements, which in turn depends on the soundness of certain episodes of reasoning. Thus, the fact that explains a reason judgement’s truth is not the fact that it is about (see Figure 6.1 in section 6.1.4). This is a consequence of the constructivist’s conception of the truth of those judgements. As I explained in Chapter 4.6, this means that the constructivist account reverses the usual order of explanation between truth and extension or successful reference. A certain reason judgement is not true because the combination of consideration, agent, and action it mentions falls in the extension of the concept of a reason; rather, that combination falls in the extension because the judgement is true—this is how reason facts are ‘constructed’ from true reason judgements. In other words, the facts that make particular reason judgements true thereby determine the extension of the concept of a reason and, assuming that the term ‘reason’ has its extension in virtue of its relation to that concept, they also determine the extension of that term. Hence, by counterfactually varying (perhaps per impossibile) the psychological facts on which the truth of a particular claim about reasons such as ‘S has a reason to avoid future agony’ depends, we are counterfactually varying the extension of ‘reason’. In particular, if it were the case that S could not form the intention to avoid future agony by way of

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

 

215

sound reasoning—that is, if the constructivist’s pMD(S, avoid future agony) did not obtain—then this would change the extension of ‘reason’ in a way that would ensure that ‘S has a reason to avoid future agony’ would be false and ‘S does not have a reason to avoid future agony’ would, accordingly, be true. Hence, constructivists are committed to: (4) If it were not the case that pMD(S, avoid future agony), then ‘S does not have a reason to avoid future agony’ would be true. But given this explanation of why constructivists are committed to the truth of such metalinguistic claims, those claims do not entail their object-language counterparts in which the claim that the consequent mentions is disquoted, because the term ‘reason’ has a different extension when used in the circumstances stipulated by the antecedent from that which it has when used by us. Therefore, (4) does not entail (3). In (4), we evaluate the truth of ‘S does not have a reason to avoid future agony’ on the basis of the extensions that the terms involved would have if the antecedent was satisfied. In (3), however, we use the term ‘reason’ to make a claim about the reasons an agent would have if the antecedent of the conditional was met. But while those counterfactual circumstances entail that the term ‘reason’ would have a different extension when used in those circumstances (so that ‘S does not have a reason to avoid future agony’ would turn out to be true), they do not affect the extension of that term when it is used by us.

6.4.6 Varying extensions Speakers always use a term with the extension it has in their circumstances, even when they use it to talk about counterfactual circumstances in which that term would be used with a different extension. This is a perfectly general phenomenon that is not peculiar to the case of ‘reason’.¹⁹ For any general term ‘F’ with the extension E(‘F’), it is a contingent fact that ‘F’ has this extension, and there will be some explanation why it has ¹⁹ For a similar rationale in a more general context, see Soames (2003, 345–6).

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

216 -   this extension rather than another one. Such an explanation will be provided by a semantic theory for that term.²⁰ Moreover, due to the contingency of the term’s extension, there will be non-actual circumstances C in which ‘F’ would have a different extension EC(‘F’) from that which it actually has. What these circumstances are depends on the details of the explanation of the term’s extension provided by the semantic theory. Consequently, there will be some item x such that x is an element of E(‘F’) but not of EC(‘F’). Hence, both ‘x is F’ and ‘If C was the case, “x is not F” would be true’ will be true. Nevertheless, ‘If C was the case, then x would not be F’ will be false, because in that statement ‘F’ is not used in the circumstances C but in the actual circumstances. It thus has the extension E(‘F’), and x is an element of that extension.²¹ Let me give an example: suppose the term ‘physician’ has its extension because it is used by English speakers to refer to persons that have a degree in medicine. Suppose also that Angela Merkel does not have a degree in medicine. Hence, ‘Angela Merkel is a physician’ is false. Now, the term ‘physician’ could have a different extension and, given the explanation we stipulated for its actual extension, we can assume that it would have a different extension if English speakers used it to refer to persons who hold a public office. In that case, moreover, Angela Merkel would be a member of the term’s extension, because she holds a public office. Hence, the following claim will be true: (5) If English speakers used ‘physician’ to refer to persons who hold a public office, then ‘Angela Merkel is a physician’ would be true. Nevertheless, the following claim will still be false: (6) If English speakers used ‘physician’ to refer to persons who hold a public office, then Angela Merkel would be a physician. ²⁰ In the case of ‘reason’, this explanation will first relate that term to the concept of a reason and then explain the extension of that concept in terms of the soundness of certain episodes of reasoning. ²¹ For simplicity, I am assuming that while C changes the extension of ‘F’ from E(‘F’) to EC(‘F’) so that x is an element of the former but not the latter, it does not change whatever it is in virtue of which x is an element of E(‘F’). This ensures that ‘If C was the case, x is F’ is true. The example in the following paragraph illustrates a case in which that assumption is met.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

 

217

We assumed that to be a physician, Angela Merkel has to have a degree in medicine. But the fact that English speakers use a certain word differently from how they actually do does not confer such a degree onto her—it would make it the case that ‘Angela Merkel is a physician’ would be true when asserted in those circumstances, but it would not make Angela Merkel a physician. Therefore, (5) does not entail (6).²² Moreover, (5) fails to entail (6) for the same reason that (4) fails to entail (3): the circumstances stipulated in the respective antecedents of the two pairs of conditional statements would change the extension of a term that is mentioned in the consequent of the first member of the pair but used in the consequent of the second. Hence, deriving the second conditional from the first one by disquotation is not permitted, and the two conditionals can differ in truth value. But if, on the constructivist’s account, (4) does not entail (3), then his commitment to (4) does not force him to reject (2). Therefore, the alethic mind-dependence of an agent’s reason to avoid future agony is compatible with that reason’s insensitivity to the counterfactual variation of the very fact on which that reason depends. Constructivists can thus accommodate everything that Parfit and others who emphasize this aspect of objectivity insist on: practical reasons can be modally robust with respect to any facts you like; they, nevertheless, depend on facts about the soundness of certain episodes of practical reasoning.

²² This example illustrates that my argument does not assume ‘reason’ to be a rigid designator (cf. Kripke 1980). A person is an element of the extension of ‘physician’ because she meets a certain description that requires her to have a medical degree. Hence, ‘physician’ does not apply to the same persons in all possible worlds, because the same person might have a medical degree in some of them but not in others. My point is that which description fixes the extension of ‘physician’ across possible worlds is itself a contingent fact that the counterfactual variation of certain circumstances, e.g. the conventions of use in a certain linguistic community, can change. Similarly, while something is an element of the extension of (the rigid designator) ‘water’ not because it (accidentally) meets a certain description but because it has a certain (essential) microphysical structure, i.e. H₂O, which microphysical structure is required to belong in that extension is, again, a contingent fact: ‘water’ might rigidly pick out a different natural kind from that which it actually does, so that ‘Water is not H₂O’ would be true when asserted in some possible worlds, even though ‘If it was the case that p, then water would not be H₂O’ is actually false, no matter what we substitute for ‘p’.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 14/7/2020, SPi

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 3/7/2020, SPi

Concluding remarks The aim of this book was to explore the prospects of a constructivist account of practical reasons by elaborating the basic idea of constructivism into a fully fledged account of practical reasons and making its theoretical commitments explicit, thereby paving the way for a betterfounded assessment of the merits of such an account. The view that has emerged from this can be summarized as follows. The constructivist’s starting point is the situation of an agent who faces the choice between different courses of action and who comes to decide in favour of one of those options by way of practical reasoning. Such reasoning is a psychological process by which the agent forms the intention to perform one of the available courses of action on the basis of her existing beliefs and intentions, which serve as the reasoning’s premise-attitudes. In reasoning from such a set of premise-attitudes to a particular conclusion-intention, the agent takes certain considerations to be a reason for the intended action: every episode of practical reasoning is accompanied by a reason judgement that normatively connects the premise-attitudes to the conclusion-attitude. This reason judgement guides the episode of reasoning; it thus plays a role that must be distinguished from that of a premise- or conclusion-attitude. Identifying reason judgements as those mental states that are apt to play this guiding role throws an important light on what kind of mental state reason judgements are. In particular, it allows the constructivist to characterize reason judgements, not as representational states similar to ordinary beliefs, but as states with a distinctive cognitive function that is captured by the role assigned to them in the theory of reasoning. Hence, constructivism does not conceive of reason judgements as states that aim to represent or track a certain class of facts. Consequently, it does not take those judgements to be true, if they are true, because they accurately represent certain facts, either. On the constructivist’s picture, reason Constructing Practical Reasons. Andreas Müller, Oxford University Press (2020). © Andreas Müller. DOI: 10.1093/oso/9780198754329.001.0001

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 3/7/2020, SPi

220   judgements are true or false in virtue of their role in practical reasoning. An episode of such reasoning can be correct or incorrect, and the standard for assessing its correctness, I suggested, is provided by the rules that constitute the activity of reasoning. If an episode of reasoning meets this standard, and if further scrutiny of its premise-attitudes would not reveal any other flaws, then the reasoning qualifies as sound. This soundness of certain episodes of reasoning provides the ground for the truth of our reason judgements: they are true if and only if, and because, the episode of reasoning that they are apt to guide is sound. Thus, an agent who engaged in that episode of reasoning would not be mistaken in taking the relevant considerations to be a reason for the intended action. Moreover, if an agent would not be mistaken in taking something to be a reason for her to perform a certain action, then it really is a reason for her to perform that action: true reason judgements correspond to reason facts. Those facts, however, are not discovered by the agent—her judgements are not true because they accurately represent them. Rather, facts about what is a reason for doing what obtain because the corresponding judgements are true; they are thus ultimately grounded in the soundness of certain episodes of deliberative reasoning. In this sense, our practical reasons are a construction of that activity. This view, I propose, is the result of fleshing out the main elements of a constructivist position in the most plausible way available. It gives a clear sense to the idea that practical reasons are not discovered but made, and it incorporates the characteristic tenets of such a position that we identified in Chapter 1: it rejects a representationalist conception of reason judgements, offers a substantive but non-representationalist conception of their truth, and endorses the mind-dependence of our reasons. What, then, about its merits as an account of practical reasons? A constructivist position is often considered an attractive middle ground in metaethics, because it does not deny the existence of normative facts, but does not characterize them as inexplicable and sui generis either. Moreover, many people inside and outside academic philosophy share the deeply held intuition that normative truths or facts depend on us in an important sense. The constructivist account developed here can accommodate this intuition, without compromising the idea that practical reasons are an objective matter. It also accommodates and explains

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 3/7/2020, SPi

 

221

the differences between normative and descriptive thought and discourse. This too will be seen as an advantage by everyone who agrees that assimilating the former to the latter is a mistake. Perhaps most importantly, at least to my mind, constructivism does better justice to the role of normative judgements in an agent’s thinking than its competitors. By characterizing them not as states that provide the information that an agent processes in making a decision, but as the states that guide such processing, it acknowledges that our ability to entertain such judgements makes possible a particular kind of processing that is distinctive of rational agents: practical reasoning. I am sure that many of these issues invite further discussion, and there are certainly others that are yet to be addressed. Still, I hope that the results of the preceding chapters are sufficient not only to show what a constructivist account of practical reasons might look like, but also to establish that such an account comes with a number of attractive features as well as fewer problems than is often assumed, and that it thus deserves further scrutiny. A comprehensive assessment of the account’s merits, however, would require, among other things, that the constructivist provide a complete theory of correct reasoning that includes an account of the rules of reasoning and thus allows us to determine the correctness or incorrectness of particular episodes of practical reasoning. Only then will we be able to evaluate the extensional adequacy of the constructivist account. Its advocates, however, do not have to defend such a theory on entirely independent grounds and without drawing on substantive convictions about what is a reason for what. Instead of trying to provide an ultimate justification for the first-order normative position that follows from a particular theory of correct reasoning, constructivists can justify such a theory as part of a wide reflective equilibrium in which the plausibility (or otherwise) of its normative consequences can be legitimately taken into consideration. That means that short of actually establishing such a reflective equilibrium, constructivists cannot guarantee that their account is fully adequate, but it also means that there is little ex ante reason to deny its extensional adequacy either. Thus, a comprehensive assessment of the merits of metaethical constructivism will only be possible in conjunction with a substantive normative theory of practical reasons. This, it seems to me, would be a worthwhile undertaking.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 3/7/2020, SPi

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 3/7/2020, SPi

References Anscombe, G. E. M. (1957). Intention. Cambridge, MA: Harvard University Press [2nd edn, 1963]. Apel, K. O. (1973). Das Apriori der Kommunikationsgemeinschaft. Frankfurt am Main: Suhrkamp. Asarnow, S. (2017). The Reasoning View and Defeasible Practical Reasoning. Philosophy and Phenomenological Research 95(3), 614–36. Audi, R. (1993). Belief, Reason, and Inference. In The Structure of Justification, 233–73. Cambridge: Cambridge University Press. Audi, R. (2006). Practical Reasoning and Ethical Decision. London: Routledge. Audi, P. (2012). Grounding: Toward a Theory of the In-Virtue-Of Relation. Journal of Philosophy 109(12), 685–711. Bagnoli, C. (2002). Moral Constructivism: A Phenomenological Argument. Topoi 21(1), 125–38. Bagnoli, C. (2012). Morality as Practical Knowledge. Analytic Philosophy 53(1), 60–9. Bagnoli, C. (2013). Constructivism about Practical Knowledge. In C. Bagnoli (ed.), Constructivism in Ethics, 153–82. Cambridge: Cambridge University Press. Bagnoli, C. (2017). Constructivism in Metaethics. In E. N. Zalta (ed.), The Stanford Encyclopedia of Philosophy, https://plato.stanford.edu/archives/win2017/entries/ constructivism-metaethics/, accessed 28 March 2020. Beall, J., Restall, G., and Sagi, G. (2019). Logical Consequence. In E. N. Zalta (ed.), The Stanford Encyclopedia of Philosophy, https://plato.stanford.edu/archives/ spr2019/entries/logical-consequence/, accessed 28 March 2020. Bird, A. (2007). Nature’s Metaphysics: Laws and Properties. Oxford: Oxford University Press. Blackburn, S. (1988). How to Be an Ethical Antirealist. Midwest Studies in Philosophy 12(1), 361–75. Blackburn, S. (1993). Essays in Quasi-Realism. Oxford: Oxford University Press. Blackburn, S. (1998). Ruling Passions. Oxford: Oxford University Press. Blackburn, S. (2010). The Majesty of Reason. Philosophy 85(1), 5–27. Blanshard, B. (1939). The Nature of Thought. London: George Allen and Unwin. Bliss, R. and Trogdon, K. (2016). Metaphysical Grounding. In E. N. Zalta (ed.), The Stanford Encyclopedia of Philosophy, https://plato.stanford.edu/archives/win2016/ entries/grounding/, accessed 28 March 2020. Boghossian, P. A. (2001). How Are Objective Epistemic Reasons Possible? Philosophical Studies 106(1), 1–40. Boghossian, P. A. (2003a). Blind Reasoning. Aristotelian Society Supplementary Volume 77(1), 225–48. Boghossian, P. A. (2003b). The Normativity of Content. Philosophical Issues 13(1), 31–45. Boghossian, P. A. (2008). Epistemic rules. The Journal of Philosophy 105(9), 472–500. Boghossian, P. A. (2014). What Is Inference? Philosophical Studies 169(1), 1–18.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 3/7/2020, SPi

224  Brandom, R. (1994). Making It Explicit. Cambridge, MA: Harvard University Press. Bratman, M. (1987). Intention, Plans, and Practical Reason. Cambridge, MA: Harvard University Press. Brogaard, B. and Salerno, J. (2013). Remarks on Counterpossibles. Synthese 190, 1–22. Broome, J. (1999). Normative Requirements. Ratio 12(4), 398–419. Broome, J. (2009). The Unity of Reasoning? In S. Robertson (ed.) Spheres of Reason, 62–92. Oxford: Oxford University Press. Broome, J. (2013). Rationality through Reasoning. Chichester: Wiley-Blackwell. Broome, J. (2014). Normativity in Reasoning. Pacific Philosophical Quarterly 95(4), 622–33. Broome, J. (2016). Responses. Philosophical Studies 173(12), 3431–48. Carroll, L. (1895). What the Tortoise Said to Achilles. Mind 4(14), 278–80. Choi, S. and Fara, M. (2018), Dispositions. In E. N. Zalta (ed.), The Stanford Encyclopedia of Philosophy, https://plato.stanford.edu/archives/fall2018/entries/ dispositions/, accessed 28 March 2020. Chrisman, M. (2008). Expressivism, Inferentialism, and Saving the Debate. Philosophy and Phenomenological Research 77(2), 334–58. Chrisman, M. (2011). Expressivism, Inferentialism and the Theory of Meaning. In M. Brady (ed.), New Waves in Metaethics, 103–25. Basingstoke: Palgrave Macmillan. Correia, F. and Schnieder, B. (eds.) (2012). Metaphysical Grounding: Understanding the Structure of Reality. Cambridge: Cambridge University Press. Crisp, R. (2000a). Particularizing Particularism. In B. Hooker and M. Little (eds.) Moral Particularism, 23–47. Oxford: Oxford University Press. Crisp, R. (2000b). Review of Joel Kupperman, Value and What Follows, Philosophy 75, 458–62. Cullity, G. (2016). Describing Rationality. Philosophical Studies 173(12), 3399–411. Dancy, J. (1993). Moral Reasons. Oxford: Blackwell. Dancy, J. (2004). Ethics without Principles. Oxford: Oxford University Press. Dancy, J. (2014). From Thought to Action. Oxford Studies in Metaethics 9, 1–17. Daniels, N. (1979). Wide Reflective Equilibrium and Theory Acceptance in Ethics. Journal of Philosophy 76, 256–82. D’Arms, J. and Jacobson, D. (2000a). Sentiment and Value. Ethics 110(4), 722–48. D’Arms, J. and Jacobson, D. (2000b). The Moralistic Fallacy: On the ‘Appropriateness’ of Emotions. Philosophical and Phenomenological Research 61(1), 65–90. Darwall, S. (1983). Impartial Reason. Ithaca, NY: Cornell University Press. Darwall, S., Gibbard, A., and Railton, P. (1992). Towards Fin de Siècle Ethics: Some Trends. Philosophical Review 101, 115–89. Dorsey, D. (2006). A Coherence Theory of Truth in Ethics. Philosophical Studies 127(3), 493–523. Dorsey, D. (2010). Truth and Error in Morality. In C. D. Wright and N. J. L. L. Pedersen (eds.), New Waves in Truth, 235–48. London: Palgrave McMillan.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 3/7/2020, SPi



225

Dorsey, D. (2012). A Puzzle for Constructivism and How to Solve It. In J. Lenman and Y. Shemmer (eds.), Constructivism in Practical Philosophy, 99–118. Oxford: Oxford University Press. Dreier, J. (2004). Meta-Ethics and the Problem of Creeping Minimalism. Philosophical Perspectives 18(1), 23–44. Dummett, M. (1959). Truth. Proceedings of The Aristotelian Society 59, 141–62. Dworkin, R. (1996). Objectivity and Truth: You’d Better Believe It. Philosophy & Public Affairs 25(2), 87–139. Dworkin, R. (2011). Justice for Hedgehogs. Cambridge, MA: Harvard University Press. Edwards, D. (2011). Simplifying Alethic Pluralism. The Southern Journal of Philosophy 49(1), 28–48. Edwards, D. (2013). Truth, Winning, and Simple Determination Pluralism. In N. J. L. L. Pedersen and C. D. Wright (eds.), Truth and Pluralism: Current Debates, 113–22. Oxford: Oxford University Press. Enoch, D. (2006). Agency, Shmagency: Why Normativity Won’t Come from What Is Constitutive of Action. Philosophical Review 115(2), 169–98. Enoch, D. (2007). An Outline of an Argument for Robust Metanormative Realism. Oxford Studies in Metaethics 2, 21–50. Enoch, D. (2009). Can There Be a Global, Interesting, Coherent Constructivism about Practical Reason? Philosophical Explorations 12(3), 319–39. Enoch, D. (2011a). Taking Morality Seriously: A Defense of Robust Realism. Oxford: Oxford University Press. Enoch, D. (2011b). Shmagency Revisited. In M. Brady (ed.), New Waves in Metaethics, 208–33. Basingstoke: Palgrave Macmillan. Field, H. (1978). Mental Representation. Erkenntnis 13(1), 9–61. Field, H. (1994). Deflationist Views of Meaning and Content. Mind 103(411), 249–85. Field, H. (2009). What Is the Normative Role of Logic? Aristotelian Society Supplementary Volume 83(1), 251–68. Fine, K. (2001). The Question of Realism. Philosophers’ Imprint 1(1), 1–30. Fine, K. (2012). Guide to Ground. In F. Correia and B. Schnieder (eds.), Metaphysical Grounding: Understanding the Structure of Reality, 37–80. Cambridge: Cambridge University Press. Fletcher, G. and Ridge, M. (eds.). (2014). Having it Both Ways: Hybrid Theories and Modern Metaethics. Oxford: Oxford University Press. Fodor, J. A. (1975). The Language of Thought. Cambridge, MA: Harvard University Press. Frege, G. (1979). Logic. In: Posthumous Writings, 1–8. Chicago: University of Chicago Press. Frost, K. (2014). On the Very Idea of Direction of Fit. Philosophical Review 123(4), 429–84. Gauthier, D. (1986). Morals by Agreement. Oxford: Oxford University Press. Gertken, J. and Kiesewetter, B. (2017). The Right and the Wrong Kind of Reasons. Philosophy Compass 12(5), 1–14.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 3/7/2020, SPi

226  Gewirth, A. (1978). Reason and Morality. Chicago: University of Chicago Press. Gibbard, A. (1990). Wise Choices, Apt Feelings. Cambridge, MA: Harvard University Press. Gibbard, A. (2003). Thinking How to Live. Cambridge, MA: Harvard University Press. Glock, H.-J. (2005). The Normativity of Meaning Made Simple. In C. Nimtz and A. Beckermann (eds.), Philosophy—Science—Scientific Philosophy, 219–41. Paderborn: Mentis. Glüer, K. and Wikforss, Å. (2009). Against Content Normativity. Mind 118(469), 31–70. Harman, G. (1984). Logic and Reasoning. Synthese 60(1), 107–27. Harman, G. (1986). Change in View: Principles of Reasoning. Cambridge, MA: MIT Press. Hattiangadi, A. (2006). Is Meaning Normative? Mind & Language 21(2), 220–40. Hattiangadi, A. (2007). Oughts and Thoughts. Oxford: Oxford University Press. Hieronymi, P. (2005). The Wrong Kind of Reason. The Journal of Philosophy 102(9), 437–57. Hlobil, U. (2014). Against Boghossian, Wright and Broome on Inference. Philosophical Studies 167(2), 419–29. Horgan, T. (1993). From Supervenience to Superdupervenience: Meeting the Demands of a Material World. Mind 102(408), 555–86. Horwich, P. (1990). Truth. Oxford. Oxford University Press [2nd edn, 1998]. Hussain, N. and Shah, N. (2006). Misunderstanding Metaethics: Korsgaard’s Rejection of Realism. Oxford Studies in Metaethics 1, 265–94. Irwin, T. H. (2006). Socrates and Euthyphro: The Argument and its Revival. In L. Judson and V. Karasmanis (eds.), Remembering Socrates, 56–71. Oxford: Oxford University Press. Jackson, F. (1994). Realism, Truth and Truth Aptness. Philosophical Books 35(3), 162–69. Jackson, F. (1998). From Metaphysics to Ethics. Oxford: Oxford University Press. James, A. (2007). Constructivism about Practical Reasons. Philosophy and Phenomenological Research 74(2), 302–25. James, A. (2012). Constructing Protagorean Objectivity. In J. Lenman and Y. Shemmer (eds.), Constructivism in Practical Philosophy, 60–80. Oxford: Oxford University Press. James, W. (1907). Pragmatism. Cambridge, MA: Harvard University Press. Johnston, M. (1993). Objectivity Refigured: Pragmatism without Verificationism. In J. Haldane and C. Wright (eds.), Reality, Representation, and Projection, 85–130. Oxford: Oxford University Press. Kahneman, D. (2011). Thinking, Fast and Slow. London: Pengiun. Kant, I. (1785). Groundwork of the Metaphysics of Morals, ed. and trans. M. Gregor. Cambridge: Cambridge University Press 1997. Kant, I. (1788). Critique of Practical Reason, ed. and trans. M. Gregor. Cambridge: Cambridge University Press 1997.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 3/7/2020, SPi



227

Katsafanas, P. (2018). Constitutivism about Practical Reasons. In D. Star (ed.), The Oxford Handbook of Reasons and Normativity, 367–91. Oxford: Oxford University Press. Kenny, A. J. (1966). Practical Inference. Analysis 26(3), 65–75. Kim, J. (1993). Supervenience and Mind: Selected Philosophical Essays. Cambridge: Cambridge University Press. Kirkham, R. L. (1992). Theories of Truth: A Critical Introduction. Cambridge, MA: MIT Press. Korsgaard, C. M. (1986). Skepticism about Practical Reason. The Journal of Philosophy 83(1), 5–25. Korsgaard, C. M. (1996). The Sources of Normativity. Cambridge: Cambridge University Press. Korsgaard, C. M. (2003). Realism and Constructivism in Twentieth-Century Moral Philosophy. The Journal of Philosophical Research 28 (APA Centennial Supplement), 99–122. Korsgaard, C. M. (2008). The Constitution of Agency. Oxford: Oxford University Press. Korsgaard, C. M. (2009a). The Activity of Reason. Proceedings and Addresses of the APA 83(2), 23–43. Korsgaard, C. M. (2009b). Self-Constitution. Oxford: Oxford University Press. Kripke, S. A. (1980). Naming and Necessity. Cambridge, MA: Harvard University Press. Kripke, S. A. (1982). Wittgenstein on Rules and Private Language. Cambridge, MA: Harvard University Press. Künne, W. (2003). Conceptions of Truth. Oxford: Oxford University Press. Leite, A. (2008). Believing One’s Reasons Are Good. Synthese 161(3), 419–41. Lenman, J. (2012). Expressivism and Constructivism. In J. Lenman and Y. Shemmer (eds.), Constructivism in Practical Philosophy, 213–25. Oxford: Oxford University Press. Levin, J. (2018), Functionalism. In E. N. Zalta (ed.), The Stanford Encyclopedia of Philosophy, https://plato.stanford.edu/archives/fall2018/entries/functionalism/, accessed 28 March 2020. Lewis, D. (1970). General Semantics. Synthese 22(1), 18–67. Lewis, D. (1972). Psychophysical and Theoretical Identifications. Australasian Journal of Philosophy 50(3) 249–58. Lewis, D. (1973). Counterfactuals. Cambridge, MA: Harvard University Press. Lewis, D. (1989). Dispositional Theories of Value. Proceedings of the Aristotelian Society, Supplementary Volumes 63, 113–37. Lillehammer, H. (2003). The Idea of a Normative Reason. In P. Schaber and R. Huntelmann (eds.), Grundlagen der Ethik, 41–65. Berlin: de Gruyter. Lipton, P. (1990). Contrastive Explanation. Royal Institute of Philosophy Supplement 27, 247–66. Lord, E. (2018). The Importance of Being Rational. Oxford: Oxford University Press. Lynch, M. P. (2009). Truth as One and Many. Oxford: Oxford University Press.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 3/7/2020, SPi

228  MacFarlane, J. (2004). In What Sense (If Any) Is Logic Normative for Thought? (Delivered at the American Philosophical Association Central Division Meeting 2004), http://johnmacfarlane.net/normativity_of_logic.pdf, accessed 28 March 2020. McHugh, C. and Way, J. (2016a). Against the Taking Condition. Philosophical Issues 26(1), 314–31. McHugh, C. and Way, J. (2016b). Fittingness First. Ethics 126(3), 575–606. McHugh, C. and Way, J. (2018). What Is Good Reasoning? Philosophy and Phenomenological Research 96(1), 153–74. McLaughlin, B. (1995). Varieties of Supervenience. In E. Savellos and U. Yalcin (eds.), Supervenience: New Essays, 16–59. Cambridge: Cambridge University Press. McLaughlin, B. and Bennett, K. (2011). Supervenience. In E. N. Zalta (ed.), The Stanford Encyclopedia of Philosophy, http://plato.stanford.edu/archives/win2011/ entries/supervenience/, accessed 28 March 2020. Maguire, B. (2016). The Value-Based Theory of Reasons. Ergo 3(9), 233–62. Müller, A. (2019). Reasoning and Normative Beliefs: Not Too Sophisticated. Philosophical Explorations 22(1), 2–15. Nagel, T. (1986). The View from Nowhere. Oxford: Oxford University Press. Neta, R. (2013). What Is an Inference. Philosophical Issues 23(1), 388–407. Nolan, D. (1997). Impossible Worlds: A Modest Approach. Notre Dame Journal of Formal Logic 38(4), 535–72. Nozick, R. (1974). Anarchy, State, and Utopia. New York: Basic Books. Olson, J. (2004). Buck-Passing and the Wrong Kind of Reasons. Philosophical Quarterly 54(215), 295–300. Parfit, D. (1997). Equality and Priority. Ratio 10(3), 202–21. Parfit, D. (2001). Rationality and Reasons. In D. Egonsson, B. Petersson, J. Joselfsson, and T. Rønnow-Rasmussen (eds.), Exploring Practical Philosophy: From Action to Values, 17–39. Aldershot: Ashgate. Parfit, D. (2011a). On What Matters, Vol. 1. Oxford: Oxford University Press. Parfit, D. (2011b). On What Matters, Vol. 2. Oxford: Oxford University Press. Pauer-Studer, H. (2014). Rational Requirements and Reasoning. Economics and Philosophy 30(3), 513–28. Pedersen, N. J. L. L. (2012). Recent Work on Alethic Pluralism. Analysis 72(3), 588–607. Pedersen, N. J. L. L. and Wright, C. D. (eds.) (2013). Truth and Pluralism: Current Debates. Oxford: Oxford University Press. Pedersen, N. J. L. L. and Wright, C. (2018). Pluralist Theories of Truth. In E. N. Zalta (ed.), The Stanford Encyclopedia of Philosophy, https://plato.stanford.edu/ entries/truth-pluralist/, accessed 28 March 2020. Peirce, C. S. (1878). How to Make Our Ideas Clear. Popular Science Monthly 12, 286–302. Peirce, C. S. (1907). Pragmatism. In Peirce Edition Project (ed.), The Essential Peirce, Vol. 2, 398–433. Bloomington, IN: Indiana University Press 1998. Pettit, P. (1996). Realism and Truth: A Comment on Crispin Wright’s Truth and Objectivity. Philosophy and Phenomenological Research 56(4), 883–90.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 3/7/2020, SPi



229

Pettit, P. (2007). Rationality, Reasoning and Group Agency. Dialectica 61(4), 495–519. Plantinga, A. (1982). How to Be an Anti-Realist. Proceedings and Addresses of the American Philosophical Association 56(1), 47–70. Price, H. (2011). Naturalism without Mirrors. Oxford: Oxford University Press. Price, H. (2013). Expressivism, Pragmatism and Representationalism. Cambridge: Cambridge University Press. Putnam, H. (1981). Reason, Truth, and History. Cambridge: Cambridge University Press. Putnam, H. (1990). Realism with a Human Face. Cambridge, MA: Harvard University Press. Putnam, H. (1995). Pragmatism. Proceedings of the Aristotelian Society 95, 291–306. Rabinowicz, W. and Rønnow-Rasmussen, T. (2004). The Strike of the Demon: On Fitting Pro-Attitudes and Value. Ethics 114(3), 391–423. Rabinowicz, W. and Rønnow-Rasmussen, T. (2006). Buck-Passing and the Right Kind of Reasons. Philosophical Quarterly 56(222), 114–20. Railton, P. (1986). Moral Realism. The Philosophical Review 95(2), 163–207. Railton, P. (2004). How to Engage Reason: The Problem of Regress. In R. J. Wallace, P. Pettit, S. Scheffler, and M. Smith (eds.), Reason and Value: Themes from the Moral Philosophy of Joseph Raz, 176–201. Oxford: Oxford University Press. Rawls, J. (1955). Two Concepts of Rules. Philosophical Review 64(1), 3–32. Rawls, J. (1971). A Theory of Justice. Cambridge, MA: Harvard University Press. Rawls, J. (1980). Kantian Constructivism in Moral Theory. Journal of Philosophy 77(9), 515–72. Rawls, J. (1993). Political Liberalism, (exp. edn). New York: Columbia University Press 2005. Raz, J. (1975). Practical Reason and Norms. Oxford: Oxford University Press [2nd edn, 1999]. Raz, J. (2006). Review: The Trouble with Particularism (Dancy’s Version). Mind 115(457), 99–120. Raz, J. (2011). From Normativity to Responsibility. Oxford: Oxford University Press. Ridge, M. (2012). Kantian Constructivism: Something Old, Something New. In J. Lenman and Y. Shemmer (eds.), Constructivism in Practical Philosophy, 138–58. Oxford: Oxford University Press. Ridge, M. (2014). Impassioned Belief. Oxford: Oxford University Press. Rosen, G. (2001). Brandom on Modality, Normativity and Intentionality. Philosophy and Phenomenological Research 63(3), 611–23. Rosen, G. (2010). Metaphysical Dependence: Grounding and Reduction. In B. Hale and A. Hoffmann (eds.), Modality: Metaphysics, Logic, and Epistemology, 109–35. Oxford: Oxford University Press. Scanlon, T. (2014). Being Realistic About Reasons. Oxford: Oxford University Press. Scanlon, T. M. (1998). What We Owe to Each Other. Cambridge, MA: Harvard University Press. Scanlon, T. M. (2003). Metaphysics and Morals. Proceedings and Addresses of the American Philosophical Association 77(2), 7–22.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 3/7/2020, SPi

230  Schafer, K. (2015a). Realism and Constructivism in Kantian Metaethics (1): Realism and Constructivism in a Kantian Context. Philosophy Compass 10(10), 690–701. Schafer, K. (2015b). Realism and Constructivism in Kantian Metaethics (2): The Kantian Conception of Rationality and Rationalist Constructivism. Philosophy Compass 10(10), 702–13. Schaffer, J. (2009). On What Grounds What. In D. Chalmers, D. Manley, and R. Wasserman (eds.), Metametaphysics: New Essays on the Foundations of Ontology, 347–83. Oxford: Oxford University Press. Schroeder, M. (2007). Slaves of the Passions. Oxford: Oxford University Press. Schroeder, M. (2008). Being For. Oxford: Oxford University Press. Schroeder, M. (2010). Noncognitivism in Ethics. London: Routledge. Schroeder, M. (2011). Ought, Agents, and Actions. Philosophical Review 120(1), 1–41. Schroth, J. (2003). Particularism and Universalizability. Journal of Value Inquiry 37 (4), 455–61. Searle, J. R. (1964). How to Derive ‘Ought’ from ‘Is’. Philosophical Review 73(1), 43–58. Setiya, K. (2007). Reasons without Rationalism. Princeton, NJ: Princeton University Press. Setiya, K. (2009). Reply to Bratman and Smith. Analysis Reviews 69(3), 531–40. Setiya, K. (2014). What is a Reason to Act?. Philosophical Studies 167(2), 221–35. Shafer-Landau, R. (2003). Moral Realism. Oxford: Oxford University Press. Silverstein, M. (2016). Reducing Reasons. Journal of Ethics & Social Philosophy 10(1), 1–22. Sinclair, N. (2012). Promotionalism, Motivationalism and Reasons to Perform Physically Impossible Actions. Ethical Theory and Moral Practice 15(5), 647–59. Smetana, J. G. (1981). Preschool Children’s Conceptions of Moral and Social Rules. Child Development 52(4), 1333–6. Smetana, J. G. and Braeges, J. L. (1990). The Development of Toddlers’ Moral and Conventional Judgements. Merrill-Palmer Quarterly 36(3), 329–46. Smith, M. (1994). The Moral Problem. Oxford: Blackwell. Smith, M. (1995). Internal Reasons. Philosophy and Phenomenological Research 55 (1), 109–31. Smith, M. (2012). Agents and Patients, Or: What We Learn about Reasons for Action by Reflecting on Our Choices in Process-Of-Thought Cases. Proceedings of the Aristotelian Society 112(3), 309–31. Smith, M. (2013). A Constitutivist Theory of Reasons: Its Promise and Parts. Law, Ethics and Philosophy 1(1), 9–30. Snedegar, J. (2018). Deliberation, Reasons, and Alternatives. Pacific Philosophical Quarterly, DOI: 10.1111/papq.12262. Soames, S. (2003). Philosophical Analysis in the Twentieth Century, Vol. 2. Princeton, NJ: Princeton University Press. Stalnaker, R. (1968). A Theory of Conditionals. American Philosophical Quarterly, Monograph Series 2, 98–112.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 3/7/2020, SPi



231

Stratton-Lake, P. (2005). How to Deal with Evil Demons: Comment On Rabinowicz and Rønnow-Rasmussen. Ethics 115(4), 788–98. Street, S. (2006). A Darwinian Dilemma for Realist Theories of Value. Philosophical Studies 127(1), 109–66. Street, S. (2008). Constructivism about Reasons. Oxford Studies in Metaethics 3, 207–45. Street, S. (2010). What is Constructivism in Ethics and Metaethics? Philosophy Compass 5(5), 363–84. Street, S. (2012). Coming to Terms with Contingency: Humean Constructivism about Practical Reason. In J. Lenman and Y. Shemmer (eds.), Constructivism in Practical Philosophy, 40–59. Oxford: Oxford University Press. Streumer, B. (2007a). Inferential and Non-Inferential Reasoning. Philosophy and Phenomenological Research 74(1), 1–29. Streumer, B. (2007b). Reasons and Impossibility. Philosophical Studies 136(3), 351–84. Streumer, B. (2010). Practical Reasoning. In T. O’Connor and C. Sandis (eds.), A Companion to The Philosophy of Action, 244–51. Malden, MA: Wiley-Blackwell. Streumer, B. (2013). Do Normative Judgements Aim to Represent the World? Ratio 26(4), 450–70. Suikkanen, J. (2009). Buck-Passing Accounts of Value. Philosophy Compass 4(5), 768–79. Sylvan, K. (2015). What Apparent Reasons Appear to Be. Philosophical Studies 172(3), 587–606. Sylvan, K. (2016). Epistemic Reasons II: Basing. Philosophy Compass 11(7), 377–89. Tappolet, C. (1997). Mixed Inferences: A Problem for Pluralism about Truth Predicates. Analysis 57(3), 209–10. Tappolet, C. (2000). Truth Pluralism and Many-Valued Logics: A Reply to Beall. Philosophical Quarterly 50(200), 382–5. Thomson, J. F. (1965). Reasons and Reasoning. In Max Black (ed.), Philosophy in America, 282–302. Ithaca, NY: Cornell University Press. Thomson, J. J. (2008). Normativity. Chicago: Open Court. Turiel, E. (2007). The Development of Morality. In Nancy Eisenberg (ed.), Handbook of Child Psychology: Vol. 3: Social, Emotional, and Personality Development. 6th edn, 789–857. Hoboken, NJ: Wiley. Valaris, M. (2014). Reasoning and Regress. Mind 123(489): 101–27. van Roojen, M. (2018). Moral Cognitivism vs. Non-Cognitivism. In E. N. Zalta (ed.), The Stanford Encyclopedia of Philosophy, https://plato.stanford.edu/archives/ fall2018/entries/moral-cognitivism/, accessed 28 March 2020. Väyrynen, P. (2017). The Supervenience Challenge to Non-Naturalism. In T. McPherson and D. Plunkett (eds.), The Routledge Handbook of Metaethics, 170–84, New York: Routledge. Velleman, J. D. (1992). The Guise of the Good. Noûs 26(1), 3–26. Velleman, J. D. (1996). The Possibility of Practical Reason. Ethics 106(4), 694–726. Velleman, J. D. (2009). How We Get Along. Cambridge and New York: Cambridge University Press.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 3/7/2020, SPi

232  Vetter, B. (2013). Multi-Track Dispositions. The Philosophical Quarterly 63(251), 330–52. Vihvelin, K. (2004). Free Will Demystified: A Dispositional Account. Philosophical Topics 32(1/2), 427–50. von Wright, G. H. (1963). Norm and Action: A Logical Enquiry. London: Routledge & Kegan Paul. Walker, R. (1989). The Coherence Theory of Truth. London: Routledge. Wallace, R. J. (1999). Three Conceptions of Rational Agency. Ethical Theory and Moral Practice 2(3), 217–42. Wallace, R. J. (2004). Constructing Normativity. Philosophical Topics 32(1,2), 451–76. Wallace, R. J. (2012). Constructivism about Normativity: Some Pitfalls. In J. Lenman and Y. Shemmer (eds.), Constructivism in Practical Philosophy, 18–39. Oxford: Oxford University Press. Way, J. (2017). Reasons as Premises of Good Reasoning. Pacific Philosophical Quarterly 98(2), 251–70. Wedgwood, R. (2006). The Normative Force of Reasoning. Noûs 40(4), 660–86. Whiting, D. (2007). The Normativity of Meaning Defended. Analysis 67(2), 133–40. Wikforss, Å. (2001). Semantic Normativity. Philosophical Studies 102, 203–26. Williams, B. (1980). Internal and External Reasons. Cambridge. In R. Harrison (ed.), Rational Action, Cambridge: Cambridge University Press [repr. in Moral Luck, 101–13. Cambridge: Cambridge University Press (1981)]. Williamson, T. (1994). A Critical Study of Truth and Objectivity. International Journal of Philosophical Studies 30(1), 130–44. Winters, B. (1983). Inferring. Philosophical Studies 44(2), 201–20. Wright, C. (1992). Truth and Objectivity. Cambridge, MA: Harvard University Press. Wright, C. (1995). Truth in Ethics. Ratio 8(3), 209–26. Wright, C. (1999). Truth: A Traditional Debate Reviewed. Canadian Journal of Philosophy, supplementary vol. 24 [repr. in Saving the Differences, 241–87. Cambridge, MA: Harvard University Press (2003)]. Wright, C. (2000). Truth as Sort of Epistemic: Putnam’s Peregrinations. Journal of Philosophy 97(6), 335–64. Wright, C. (2001). Minimalism, Deflationism, Pragmatism, Pluralism. In M. P. Lynch (ed.), The Nature of Truth, 751–87, Cambridge, MA: MIT Press. Wright, C. (2013). A Plurality of Pluralisms. In N. J. L. L. Pedersen and C. D. Wright (eds.), Truth and Pluralism: Current Debates, 123–53. Oxford: Oxford University Press. Wright, C. (2014). Comment on Paul Boghossian, ‘What is Inference.’ Philosophical Studies 169(1), 27–37. Young, J. O. (2018). The Coherence Theory of Truth. In E. N. Zalta (ed.), The Stanford Encyclopedia of Philosophy, https://plato.stanford.edu/archives/fall2018/ entries/truth-coherence/, accessed 28 March 2020.

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 3/7/2020, SPi

Index For the benefit of digital users, indexed terms that span two pages (e.g., 52–53) may, on occasion, appear on only one of those pages. abductive reasoning 144 Absoluteness 119n.13 active reasoning 34–5, 60, 68, 79 agency 34–5, 162n.15, 177–8, 181–4, 188n.31 akratic actions 38–40 akratic reasoning 39–40, 80–1 alethic mind-dependence 192–200, 202–3, 209–10, 213–15, 217 alethic pluralism 119–28, 121n.18 ambitious constitutivism 181–4 Anscombe, G. E. M. 37 antifactualist account 14n.11 anti-realism 27, 198 anti-representationalism 10–11, 31, 82, 93–94, 98, 100–3, 121n.15, see also conativist anti-representationalism, rationalist anti-representationalism Apel, K. O. 179 Aristotle 37 Asarnow, S. 135n.28, 137n.32 assertion 119, 119n.13, 128–9, 217 attitude formation 41–4, 51, 68, 76–7 attitudes 7–12, 8n.3, 14–15, 17–18, 34–7, 34n.2, 41–3, 45–6, 49–51, 53–4, 56, 60–2, 66–7, 69–72, 75–7, 80, 82, 86, 89, 91–3, 101–7, 109, 111n.1, 127n.25, 142–6, 149–50, 153, 158n.13, 165, 166, 170–2, 186, 187–8, 192–3, 198–209, see also conclusionattitudes, premise-attitudes, pro-attitudes, propositional

attitudes, representational attitudes, stimulus-attitudes Audi, R. 23n.22, 37–8, 45n.15, 66n.24 automatic processes 34–5, 68–9, 76–7, 80n.36, 135n.29 Bagnoli 7n.1, 11n.4, 14n.10, 17n.17 beliefs 4–5, 8–12, 8n.3, 35–7, 41–2, 45–6, 51, 53–4, 61–3, 65–75, 78–80, 85–8, 91–5, 97–8, 102–3, 107–8, 111n.1, 113–15, 115n.8, 120–1, 129–31, 137n.31, 142–4, 149–50, 152, 155, 159n.14, 177–8, 202–3, 205–6, 211n.18, 219–20, see also normative beliefs, perceptual beliefs, premise-beliefs, representational beliefs Bennett, K. 24–5 Bird, A. 52–3, 55 Blackburn, S. 11, 14n.11, 83, 104–5 Blanshard, B. 114n.6 Bliss, R. 21n.20 Boghossian, P. A. 28n.29, 34–5, 40–1, 45–6, 47n.19, 63–5, 66n.24, 67–8, 72n.27, 146n.3 Braeges, J. L. 68–9 Brandom, R. 11–12, 121n.15 Bratman, M. 34n.1 Brink, D. 16n.14 Broome, J. 34–6, 35n.3, 37n.7, 43, 46–7, 60–3, 66n.24, 68, 70, 78n.33, 80n.37, 92n.4 buck-passing account of correctness (BPC) 146

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 3/7/2020, SPi

234  buck-passing account of correct reasoning (BPCR) 145–9, 156–7, 172n.20 buck-passing account of value 145–6, 148–50, 148n.7, 153, 186 Carroll, L. 72 chess 32, 125, 132, 141–2, 145, 151–2, 159–64, 167–8, 171, 185, 188 children 44–6, 65–9, 80, 80n.36, 165–7 Choi, S. 53n.21 Chrisman, M. 100 circularity 14–15, 27–8, 31–2, 111–12, 114, 128–9, 131–4, 141–2, 148–9 Clarke, S. 12n.7, 16n.14 cognitive abilities 66–9 activity 77n.32 function 8n.3, 10–14, 26, 33, 82–3, 127–8, 139–40, 219–20 job 13, 86–7 processes 9, 34–5, 68 resources 75–7 role 8–9, 11, 33, 83, 85, 102, 108–9, 137 states 91–2 coherence 3, 77–8, 108n.14, 114–17, 115n.8, 180 coherence theory of truth 114–16, 115n.8, 120–1, 139n.34 competence 58–9, 66–9 compliance 153–6, 161, 165–6, 169 computer program 90–2 conative states 104, 107 conativist antirepresentationalism 101–7 conceptual level 193–6 mind-dependence 195–7, 198n.5 resources 65–7 sophistication 44, 63–72 conclusion-attitudes 35–6, 39–41, 45–9, 52–3, 55–8, 60, 71–8, 104–5, 107–8, 129–32, 141–4, 147, 159n.14, 170–1, 219 conclusion-intentions 49–50, 149, 219

conditions of adequacy 41–4, 50–2, 57, 63–4 conflicts 40, 79–80 conscious active reasoning 34–5, 68 attitudes 60, 71–2 judgements 71 mental processes 34–5, 71, 79 person-level activity 39 person-level reasoning 60, 69, 76–7 premise-attitudes 60 processes 34–5, 76–7 psychological activity 142–3 reasoning 135n.29 constitutive rules 32, 131–2, 141–2, 161–3, 166–70, 172–3, 177–81, 183n.28, 184–5, 187, 211 Constitutive Rules Account 159–64, 170, 172–5, 177, 211 constitutivism 181–4 correctness 4–5, 19–20, 32, 61, 99, 103, 106, 129–32, 141–63, 146n.6, 172, 178–9, 181, 211, 219–21 correct reasoning 32, 35–6, 41–3, 45–6, 50–1, 54–7, 74, 76–7, 103–4, 141–8, 149n.8, 152–3, 156–63, 170, 172–5, 177–80, 184–5, 191–2, 205–6, 221, see also buck-passing account of correct reasoning Correia, F. 21n.20 Correspondence 13–15, 111–14, 195, 199 correspondence theory 111–12 Coulomb’s Law 54–5 counterpossible conditionals 212–13 Crisp, R. 108n.14, 148n.7 Cullity, G. 61 Dancy, J. 36n.4, 39n.10, 79–80, 108n.14, 137n.31 D’Arms, J. 148n.7, 151 Darwall, S. 19n.19, 26–7, 135n.28 decisions 7–10, 33–4, 37n.7, 78–80, 95–6, 111n.1, 136, 165–6, 221

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 3/7/2020, SPi

 decisive reasons 38n.8 deflationism 11–12, 111–14, 113n.3, 119n.14, 128 deliberation 43, 94–8, 98n.7, 203–4 deliberative failure 34 indispensability 94–8 process 33–4 Deliberative Constraint 134–7, 206n.13 dependence 27, 190n.1, 191–5 claim 181–2 level 190, 190n.1, 193–6 on necessary facts 210–12 relation 25–6, 190 see also mind-dependence desires 1–2, 4, 11, 27–9, 83–9, 93, 101–3, 106–7, 159–60, 190–4, 198–9, 198n.4, 201–5, 208, 208n.16, 211 directedness 22–5, 192–3 Discrepancy 201–3 dispositional account of practical reasoning 53 properties 22, 52–3 theory of value 193 understanding 61–2 dispositionalist views 18n.18, 29 dispositions 4, 46–9, 51–9, 61–3, 171–2, 176n.23, 187–8, 198n.4, 203n.10 domains 6, 14n.10, 15, 19, 85, 116, 119–21, 123–8, 134, 136, 138, 175 Dorsey, D. 116, 120–1, 139n.34 Dreier, J. 113–14 dual process theory of thinking 34–5 Dummett, M. 125n.22 Dworkin, R. 207n.14 Edwards, D. 124–5, 125n.22 emotion-driven attitude formation 51 process 42–3, 53–4, 58–9 emotions 58–9, 87–8, 105–6 enabler 137n.31 Enoch, D. 2, 9–10, 29–31, 43, 86, 94–9, 100n.8, 181–5

235

epistemic access 181–4 contraints 116–18 reasons 4–5 theory of truth 115n.7 ethics 16n.13, 116, 207n.14, see also meta-ethical, meta-ethics etiquette 161–2, 167–8 Euthyphro 27, 197–200 Euthyphronic Determination Claim (EDC) 198–200 Euthyphronic position 189–90, 198, 200 evil demon 149–50 evil philosophy lecturer 154 expressivism 11, 14n.11, 29, 84–6, 100–7, 113n.4 factualist account 14n.11 fallibility 20, 32, 176n.22, 189–90, 200–8 Fara, M. 53n.21 Field, H. 83, 112n.2, 144 Fine, K. 14n.11, 21n.20, 25n.26 first-order judgements 173 normative theory/theorizing 3–4, 175–6, 179–80, 221 reasoning 78, 80–1 first-person experience 70 judgements 82, 107 perspective 70–1 Fletcher, G. 107 Fodor, J. A. 91–2 Frege, G. 45–6, 57 function 17–18, 88, 92n.5, 188 of attitudes 8n.3 of beliefs 10, 93 of concepts 12–13, 13n.8 of judgements 10–14, 137, 139–40 of metareasoning 79 of normative vocabulary 11–12 see also cognitive function, motivational function, representational function

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 3/7/2020, SPi

236  functionalism 83–5, 89 fundamentalism 28–9, 175–6, 187 Gauthier, D. 179 Gertken, J. 148n.7 Gewirth, A. 179 Gibbard, A. 11, 83, 105–6 Glock, J.-J. 146n.3 Glüer, K. 145–6 grounding 192–3, 199n.6 claim 22, 28n.29 reasons 141, 143, 189 relation 20–7, 124–5, 192–3 guidance 48–9, 60–1, 75–6, 89–100, 104–6, see also Normative Guidance Account guiding role 33, 46–51, 73, 75–8, 84–5, 89–90, 93–4, 98–9, 102–5, 219–20 Harman, G. 34n.2, 144 Hattiangadi, A. 146n.3 Hieronymi, P. 135n.28, 137n.32 Hlobil, U. 43, 51–2, 57, 59, 62n.23, 63–4 Horgan, T. 25 Horwich, P. 112n.2, 113n.3, 119n.14 Humean account of motivation 86–7 account of reasons 202–3, 204n.12 approach 190–2 background 104–5 reductivism 18n.18, 28–9 thesis 101–2 Humeans 27–8, 191 Hume’s Law 75n.30 Hussain, N. 2, 26–7 hypothetical reasoning 107–9, 134–5, 136n.30 identity 23–5, 28n.29, 192–4 imaginations 87–9 incorrectness 143, 148, 155–6, 160–1, 172, 221 incorrect reasoning 41–3, 53–4, 56, 62–3, 103, 142–5, 147, 149, 154

inductive reasoning 144 infallability 116–18, 117n.11, 202–3 inference 43, 45, 51–2, 59, 63–4, 72, 75–7, 122–4 inferences to the best explanation 95, 97, 176–7 inferential internalism 40–1 Inferential Moorean Phenomenon 43 inferential reasoning 36n.4 inferring 45, 176–7 information 8–10, 26, 33–4, 76–7, 118, 220–1 input 10, 17–19, 79–80, 90–3, 172–3, 175n.21 inquiry 115–18, 129 instrumental reasoning 45, 49–50, 65–6, 75–6, 130, 142–3, 149, 151–2, 170–1, 171n.19 intentional action 34n.1, 102–3 behaviour 101–4 guidance 60–1 view of rule-following 63 violations of rules 169 intentions 33–5, 36n.4, 37–44, 48–51, 53–4, 60–1, 63, 65, 68, 70, 72n.28, 78, 80–1, 92–3, 101–3, 106–8, 129–35, 136n.30, 142–5, 149, 152, 155, 158, 158n.13, 169, 170, 177–8, 180, 201, 205–6, 211, 214–15, 219, see also conclusionintentions, premise-intentions internalism 40–1, 101–2 Irwin, T. H. 197n.3 Jackson, F. 118–19, 126n.24 Jacobson, D. 148n.7, 151 James, A. 7n.1, 15n.12, 17n.17, 27, 162n.15, 173, 187n.30, 189, 197–8 James, W. 115n.8 Johnston, M. 197n.3 justification 40–1, 45–6, 66–9, 75–6, 109, 115–18, 174–7, 176n.22, 179–82, 221

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 3/7/2020, SPi

 Kahneman, D. 34–5 Kant, I. 7n.2, 12n.6, 179, 180n.25 Kantian constructivism 12n.6, 19 Katsafanas, P. 181n.26 Kenny, A. J. 37 Kiesewetter, B. 148n.7 Kim, J. 25 Kirkham, R. L. 111n.1, 116n.9, 129n.26 kitchen scale model 79–80 Korsgaard, C. 2, 6–7, 12–13, 15–16, 19–20, 135n.28, 162n.15, 179 Kripke, S. A. 60, 217n.22 Künne, W. 111–12, 116n.9 Leite, A. 45n.15 Lenman, J. 29n.30, 100n.8 Levin, J. 83 Lewis, D. 18n.18, 29, 83, 118–19, 190–3, 212–13 Lillehammer, H. 134 Lipton, P. 21 logic/logicians 144 Lynch, M. P. 32, 118, 121nn.17,18, 126n.23, 138 MacFarlane, J. 144 Maguire, B. 187 marked contents 36n.6 McHugh, C. 66n.24, 135n.28, 158n.13, 159n.14 McLaughlin, B. 24–5 mental act 150 activities 6, 17, 28–9, 34–5, 63–4, 82, 142–3, 190–3, 200 capacities 137 development 68–9 disposition 56 facts 192 processes 41–3, 51, 53–4, 57–9, 61–2, 71, 73, 79, 104, 182–3 states 6, 8n.3, 11, 17–20, 26, 28–9, 34n.1, 48–9, 65, 70, 73, 82–90, 93–4, 98–9, 101–2, 104–7, 111n.1,

237

127, 139–40, 189–93, 200, 203n.10, 204–5, 219–20 transition 53–5, 58–9 meta-ethical claims 139n.33 constructivism 7n.2, 100, 197–8, 221 debate 2, 12 expressivism 107 perspective 207n.14 positions 3–4, 19, 29, 113–14, 181 theories 26–7, 30–1 views 3–4, 12n.6, 25n.27, 29–31, 98n.7, 100, 103–4, 107, 113n.4, 189 meta-ethics 6–7, 9–10, 23, 30–1, 82–3, 85, 100, 186, 193n.2, 207n.14, 220–1 metaphysical claims 139n.33 concepts 113–14 explanations 21, 25–6 level 193–6 mind-dependence 195–7, 199 necessity 24–5 notions 22, 25–6 questions 12n.6 reduction of reason facts 28n.28 status 86 status of normative facts 3–4 metareasoning 77–81, 80n.35 mind-dependence 17–20, 19n.19, 27–9, 32, 189–97, 190n.1, 203–5, 207–13, 220, see also alethic mind-dependence, conceptual mind-dependence, metaphysical mind-dependence minimalism 113–14, 119n.14 modal profile 53 relation 24–5 robustness 32, 189–90, 200, 206–17 status 204n.12, 206–7 modest constitutivism 181–4 Moore, G. E. 12n.7, 16n.14

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 3/7/2020, SPi

238  moral constructivism 3–4, 12n.6, 17–18 facts 4–5, 12, 14n.10, 16, 19–20, 19n.19, 30–1 judgements 11n.4, 12, 14n.10, 15–18, 19n.19, 85, 108n.14, 193n.2 law 180n.25 philosophy 3, 179 psychology 104–5 reasons 151–3, 203–4, 209 motivational function 92–3 internalism 101–2 role 101–4, 106 states 2, 11, 86–9, 190–1 system 105 multi-track dispositions 52–5 Nagel, T. 16n.14, 97–8 neo-Aristotelian conception of ontology 21–2 neo-pragmatist conception of truth 117–20 Neta, R. 45n.15 Nolan, D. 212–13 non-cognitivism 86, 106 non-normative beliefs 86 facts 8–9, 175–6 normative beliefs 48–50, 73–4, 86 concepts 12–13, 20, 57, 65–70, 145–6, 193 control system 105–6 facts 3–6, 8–9, 20, 56–7, 94–9, 175–6, 220–1 language 100–1, 105 output 177–80, 206 theory/theorizing 3–4, 147, 175–6, 179–80, 206, 221 thought/judgement 13, 66–7, 70–2, 105–6 truths 86, 94–8, 220–1 Normative Guidance Account 44–52, 55–7, 59–60, 64–5, 70–6, 90

Normative Judgement Condition (NJC) 44–52, 63–73 normativity 86, 157, 157n.12, 187 Nozick, R. 101 objectivity 17, 17n.17, 19–20, 32, 98–100, 189–90, 200, 200n.7, 206–10, 207n.14, 217 Olson, J. 148n.7, 150n.10 Parfit, D. 1–2, 9–10, 16n.14, 27–8, 101, 150nn.9,10, 201n.8, 203–4, 208n.16, 209–10, 217 particularism 108n.14 Pauer-Studer, H. 45n.15 Pedersen, N. J. L. L. 124n.21 Peirce, C. S. 115–17, 119–20 perceptual beliefs 85, 88–9, 92–3, 126–7 person-level reasoning 39, 60, 68–9, 76–7 Pettit, P. 92n.4, 126n.24 phenomenology of reasoning/ deliberation 70, 94–5, 97–9, 98n.7 Plantinga, A. 116–17 Plato 27, 197–8, 212–13 pluralism one concept, many properties 122 see also alethic pluralism, simple alethic pluralism practical reasoning 6, 11–12, 14–15, 14n.10, 17, 20–1, 26, 31–4, 36n.4, 37–44, 46–7, 50–4, 59–60, 62–5, 72, 77–8, 80–5, 89–99, 92n.5, 102–6, 127, 129, 132–3, 137, 141–3, 145, 161, 170, 172–5, 175n.21, 177–8, 180–1, 187n.30, 188n.31, 189, 191, 205, 217, 219–21 premise-attitudes 35–7, 39–41, 45–53, 56–8, 60, 71–5, 77–8, 91–2, 104–5, 129–33, 137n.31, 141–4, 147, 159n.14, 170–1, 205–6, 211n.18, 219–20 premise-beliefs 49–50, 74–5, 130, 132, 211

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 3/7/2020, SPi

 premise-intentions 49–50, 129–30, 132–3, 205 Price, H. 11–12, 121n.15 Price, R. 12n.7, 16n.14 Pritchard, H. A. 16n.14 pro-attitudes 11, 34n.1, 101–4 problem of mixed compounds 122 problem of mixed inferences 122 projectivism 14n.11 propositional attitudes 57, 60–1, 63–4, 87–8, 171–2, 187–8 pure dispositional account 54–6 Putnam, H. 115n.8, 117n.11 Quintean ontology 21–2 Rabinowicz, W. 148n.7, 150n.10 Railton, P. 16n.14, 72n.27, 73, 192–3 Ramsey sentences 83–5, 89 rational intuitionism 12, 19–20 rational intuitionists 15, 16n.14, 19n.19 rational reasoning 73–5 rationalist anti-representationalism 102–6 rationality 4–5, 19–20, 211 Rawls, J. 2–3, 6–7, 12–13, 15, 16n.14, 19–20, 163–4, 167, 175–6 Raz, J. 37–41, 137n.31, 159n.14, 164 realism 12, 16–18, 27, 30, 85, 113n.4, 198, see also Robust Realism reality 13, 13n.9, 14n.10, 16–18, 17n.17, 20–2, 27–8, 86, 115–16 reasoning process 31, 33–7, 44, 47–50, 70–1 view 135n.28, 137–8 see also abductive reasoning, active reasoning, akratic reasoning, buck-passing account of correct reasoning, correct reasoning, firstorder reasoning, hypothetical reasoning, incorrect reasoning, inductive reasoning, inferential reasoning, instrumental reasoning, person-level reasoning, phenomenology of reasoning, practical reasoning, rational

239

reasoning, rules of reasoning, sound reasoning reducibility 23–6, 30–1, 175–6, 199n.6 reductivism/reductivists 18n.18, 23–4, 27–9, 28n.29, 32, 175–6, 187 reflective equilibrium 176–80, 221 regress 44, 47n.20, 63–5, 72–7, 132–3, 205 regulative rules 163–4, 166–8 representation 7–15, 17n.17, 27–8, 60–1, 79–80, 85–6, 89–100, 111–14, 119, 122–3, 127–9, 139–40, 171–2, 194–5 representational accuracy 122–7, 137–9, 196–7 beliefs 86, 88–9, 107 function/role 12, 85, 86, 88, 90, 92–3, 93–4, 127, 139n.33 states 8–10, 93, 98–9, 101–2, 106–7, 113–14, 219–20 Representationalism 10–14, 17–19, 28–9, 31, 82–3, 85–90, 93–5, 98–102, 106, 113–14, 120–1, 175–6, 175n.21, 200 Ridge, M. 2, 29n.30, 83, 100, 100n.8, 107 Robust Realism 86, 94–8 robot analogy 90–3 Rønnow-Rasmussen, T. 148n.7, 150n.10 Rosen, G. 21n.20, 25n.26, 157–8, 157n.12 Ross, D. 12n.7, 16n.14 rule-following 59–65 rules of reasoning 162–4, 170–88, 171n.19, 211, 221 Scanlon, T. 1–4, 9–10, 27–8, 66, 80n.34, 146n.4, 186–7, 201n.8 Schafer, K. 7n.2 Schaffer, J. 21–2, 21n.20, 27 Schnieder, B. 21n.20 Schroeder, M. 1–2, 9–10, 18n.18, 27–8, 32, 100, 101n.11, 135–6, 135n.28, 145–6, 146n.6, 149–54, 159–60, 186–7, 189–96, 199–200, 202–5, 208, 211

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 3/7/2020, SPi

240  Schroth, J. 108n.14 second-person reason judgements 109 semantic theory of truth 139n.33 sensory stimuli 83–4, 90 Setiya, K. 135n.28, 136, 137n.32, 187 Shafer-Landau, R. 2, 17–19, 27, 85, 189 Shah, N. 2, 26–7 shmagents 181 shmeasoners/shmeasoning 181–5, 188 Sidgwick, H. 12n.7 Silverstein, M. 135n.28, 137n.32 simple determination pluralism 124 Sinclair, N. 135n.28, 136 Smetana, J. G. 68–9 Smith, M. 1–2, 9–10, 18n.18, 83, 86–8, 135n.28, 179, 189–96, 198–200, 208, 211 Snedegar, J. 135n.28 Socrates 197–8 sophistication objection 65–71 soundness 14–15, 17, 20–1, 26, 32, 132, 137, 137n.31, 141, 191, 214, 216n.20, 217, 219–20 sound reasoning 128–38, 192–3, 205, 208, 214–15 Stalnaker, R. 212–13 stimulus-attitudes 47, 49–52 Stratton-Lake, P. 148n.7 Street, S. 7n.1, 11n.5, 17–18, 19n.19, 27, 189, 197–8 Streumer, B. 36n.4, 37, 86, 153–4 subconscious processes 34–5, 68–9, 76–7, 135n.29 Sufficient reasons 38n.8 Suikkanen, J. 186 superassertibility 118 supervenience 24–5 superwarrant 118–24, 128–31, 138–9 Sylvan, K. 57–8 Taking Condition 45–6, 47n.19, 63–4 Tappolet, C. 122 third-person reason judgements 108–9 Thomson, J. F. 45n.15

Thomson, J. J. 4–5, 157 Timelessness 119n.13 treating account 57–9 Trogdon, K. 21n.20 truth 7, 13–17, 19, 26–8, 30–2, 45–6, 86–8, 94–100, 106, 111–40, 158–60, 173, 177, 184–8, 193, 200–2, 210–15, 217, 220–1 coherence theory 114–16, 115n.8, 120–1, 139n.34 deflationism 11–12, 111–14, 113n.3, 119n.14, 128 epistemic theory 115n.7 moral 207n.14 neo-pragmatist conception 117–20 non-deflationist account 113n.3 normative 86, 94–8, 220–1 of reason judgements 14–15, 28–9, 32, 96–9, 110–14, 120–1, 128–9, 131–2, 134, 141, 147, 172–5, 178–9, 191, 194–7, 199, 214, 219–20 pluralist account/conception 120–1, 121n.16, 126–8 semantic theory 139n.33 verificationist theory 115n.8 Wrightean conception 128–9, 134, 196–7 truthbearer 111n.1, 114–15, 129, 139n.34 truthmaker 14n.10, 139–40 Turiel, E. 68–9 universality 189–90, 200–8, 208n.17 ultimate justification 179–80, 221 Valaris, M. 45n.15, 47nn.18,20, 72n.27, 73 Väyrynen, P. 25n.27 Velleman, J. D. 36n.5, 181 verificationist theory of truth 115n.8 Vetter, B. 52–3, 55 Vihvelin, K. 53n.22 von Wright, G. H. 164 Walker, R. 114n.6 Wallace, R. J. 2, 100n.8, 135n.28

OUP CORRECTED AUTOPAGE PROOFS – FINAL, 3/7/2020, SPi

 Way, J. 66n.24, 135n.28, 158n.13, 188n.32 Wedgwood, R. 46–7, 55–7, 66n.24, 70–1, 72n.27, 73–5 Whiting, D. 146n.3 Wikforss, Å. 145–6 Williams, B. 135n.28, 203n.11 Williamson, T. 122 Winters, B. 66nn.24,25

241

Wright, C. 32, 72n.27, 75–7, 110, 113n.3, 116–23, 119nn.13,14, 124n.21, 126, 126n.24, 128, 134, 138, 164, 197n.3 Wrightean conception of truth 128–9, 134, 196–7 wrong kind of reasons problem 148–53, 156 Young, J. O. 115n.8