Epistemology of Modality and Philosophical Methodology [1 ed.] 036743167X, 9780367431679, 9781032425962, 9781003002192

This book collects original essays on the epistemology of modality and related issues in modal metaphysics and philosoph

237 35 4MB

English Pages 392 [393] Year 2023

Report DMCA / Copyright

DOWNLOAD PDF FILE

Recommend Papers

Epistemology of Modality and Philosophical Methodology [1 ed.]
 036743167X, 9780367431679, 9781032425962, 9781003002192

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

Epistemology of Modality and Philosophical Methodology

“This is an excellent collection of cutting-edge research in the ­epistemology of modality and philosophical methodology. It takes the next step forward in an exciting and fundamental research programme in philosophy.” – David Efird, University of York, UK

This book collects original essays on the epistemology of modality and related issues in modal metaphysics and philosophical methodology. The contributors utilize both the newer “metaphysics-first” and the more traditional “epistemology-first” approaches to these issues. The chapters on the epistemology of modality mostly focus on the problem of how we can gain knowledge of possibilities, which have never been actualized, or necessities which are not provable either by logico-mathematical reasoning or by linguistic competence alone. These issues are closely related to some of the central issues in philosophical methodology, notably to what ­extent is the armchair methodology of philosophy a reliable guide for the formation of beliefs about what is possible and necessary. This question also relates to the nature of thought experiments that are extensively used in science and philosophy. Epistemology of Modality and Philosophical Methodology will be of ­interest to researchers and advanced students working on the epistemology and metaphysics of modality, as well as those whose work is concerned with philosophical methodology more generally. Anand Jayprakash Vaidya is a Professor of Philosophy at San Jose State University, USA. His research focuses on epistemology and philosophy of mind. For 25 years, he has worked on the epistemology of modality and how it relates to problems in the philosophy of mind, such as the nature of consciousness. Duško Prelević is a Professor of Philosophy at the Faculty of Philosophy, University of Belgrade, Serbia. His research focuses on the epistemology of modality, philosophy of mind, and metaphysics. For 13 years, he has critically examined various aspects of the physicalist research program and modal rationalist account of our modal knowledge.

Routledge Studies in Epistemology Edited by Kevin McCain, University of Alabama at Birmingham, USA and Scott Stapleford, St. Thomas University, Canada

Intellectual Dependability A Virtue Theory of the Epistemic and Educational Ideal T. Ryan Byerly Skeptical Invariantism Reconsidered Edited by Christos Kyriacou and Kevin Wallbridge Epistemic Autonomy Edited by Jonathan Matheson and Kirk Lougheed Epistemic Dilemmas New Arguments, New Angles Edited by Kevin McCain, Scott Stapleford and Matthias Steup Proposition and Doxastic Justification New Essays on Their Nature and Significance Edited by Paul Silva Jr. and Luis R.G. Oliveira Epistemic Instrumentalism Explained Nathaniel Sharadin New Perspectives on Epistemic Closure Edited by Duncan Pritchard and Matthew Jope Epistemic Care Vulnerability, Inquiry, and Social Epistemology Casey Rebecca Johnson Epistemology of Modality and Philosophical Methodology Edited by Anand Jayprakash Vaidya and Duško Prelević For more information about this series, please visit: https://www.­ routledge.com/Routledge-Studies-in-Epistemology/book-series/RSIE

Epistemology of Modality and Philosophical Methodology

Edited by Anand Jayprakash Vaidya and Duško Prelević

First published 2023 by Routledge 605 Third Avenue, New York, NY 10158 and by Routledge 4 Park Square, Milton Park, Abingdon, Oxon, OX14 4RN Routledge is an imprint of the Taylor & Francis Group, an informa business © 2023 Taylor & Francis The right of Anand Jayprakash Vaidya and Duško Prelević to be identified as the authors of the editorial material, and of the authors for their individual chapters, has been asserted in accordance with sections 77 and 78 of the Copyright, Designs and Patents Act 1988. All rights reserved. No part of this book may be reprinted or reproduced or utilised in any form or by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers. Trademark notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to infringe. ISBN: 978-0-367-43167-9 (hbk) ISBN: 978-1-032-42596-2 (pbk) ISBN: 978-1-003-00219-2 (ebk) DOI: 10.4324/9781003002192 Typeset in Times New Roman by codeMantra

For Manjula and Ivana who made actual, what would have remained merely possible

Contents

Introduction

1

A NA N D JAY PR A K A SH VA I DYA

The Central Questions  1 Summary of Articles

14

DUŠKO PR E L E V IĆ

1 Modality, Worlds, Essence, and Modal Knowledge

19

B OR I S K M E N T

Necessity, Possible Worlds, Essence, and Counterfactuals  19 A Theory of Necessity  25 Metaphysical Laws, Grounding, and Causal Explanation  30 Modal Knowledge  34 2 An Agency-Based Epistemology of Modality

44

BA R BA R A V E T T E R

Introduction 44 Modality in Action  47 Toward Mere Possibility  52 Modality Beyond Action  55 Modality in Philosophy  58 3 The Price of Sensitivity R E BE C CA H A N R A H A N

Jenkins on Concepts  71 Churches and Q-Churches  73 Collapsing Possibilities  76 Accepting Insensitivity  79 Conclusion 82

70

viii Contents 4 Modal Epistemology for Modalists

88

O TÁV IO BU E NO A N D S C O T T A . SH A L KOWSK I

Introduction 88 Modal Rationalism: Intuitions  89 Modalism and Modal Epistemology  96 Conclusion 105 Acknowledgments 105 5 How (Meta-)Semantics Defuses Modal Pessimism

108

CHRISTIAN NIMTZ

Introduction: Modal Pessimism and Why We Should Care  108 Modal Pessimism and the Argument from Autonomy  110 Modal Knowledge from Science and Semantics I: Identities  113 Paradigm Terms Semantics  116 From Paradigm Terms to Proto-Necessities  119 Modal Knowledge from Science and Semantics II: Beyond Identities  121 Conclusions and Objections  122 6 How Things Have to Be

128

NAT H A N SA L MON

Putnam on Natural-Kind Terms  128 Non-Rigid General Terms  131 Modal-Essentialism in the Putnam Theory  134 Reconfiguring the Putnam Program  137 The Failure of the Reconfigured Putnam Program  144 7 In Search of a Structurally Complete Epistemology of Essence

150

M IC H A E L WA L LN E R

Introduction 150 Lowe’s Epistemology of Essence  151 Hale’s Epistemology of Essence  152 The Road Map  157 Husserl’s Epistemology of Essence  161 Conclusion 169 Acknowledgments 170 8 Morals and Modals: Puzzling about the Dual Use of Modal Verbs 176 A R I N DA M C H A K R A BA RT I A N D I A N N IC OL AY

“Should” vs. “Should”: Modality and the Dual-Use Subjunctive Verbs  176

Contents  ix “May” and “Could”: Syāt vis-à-vis ◇ 180 “If ” and “Would”: Syāt vis-à-vis □ 182 “Ought” and “Should”: Syāt vis-à-vis O 189 Injunctive Modality and Our Knowledge of It  193 9 The Explanatory Power of Modal Rationalism

199

DUŠKO PR E L E V IĆ

Chalmers’s Modal Rationalism and Roca-Royes’s Essentialist Challenge  199 Modal Rationalism and the Essentiality of Kind 203 Modal Rationalism and the Essentiality of Origin 210 10 Conceivability: Still Not Enough: A Response to Prelević

218

S ON I A RO CA- ROY E S

Introduction and Scope  218 Non-Epistemic Conceivability and De Re Modal Knowledge  219 Exegetical Reconstruction of Prelević’s Rescuing Strategy  223 Still Unexplained Friction  227 11 Reviving the Modal Account of Essence

231

R E BE C CA C H A N

Introduction 231 Two Conceptions of Essence  233 The (Alleged) Virtues of Definition  235 The Revised Modal Account  238 Two Theoretical Advantages  243 Concluding Thoughts  244 12 A Neo-Aristotelian Reply to A Modalist

247

SA N NA M AT T I L A

Introduction 247 The Neo-Aristotelian Account  248 The Determinacy Modification and the Need to Individuate  251 The Sparseness Modification and Borderline Cases  256 Conclusion 260 13 Semantic Rules, Modal Knowledge, and Analyticity A N T ON E L L A M A L L OZ Z I

Introduction 265 What Is Modal Normativism?  266

265

x Contents Semantic Mastery vs. Metaphysical Modal Knowledge  268 Independence Conditionals for Knowledge of Metaphysical Modality  270 A Possible Way-Out: Modal Monism  275 Modal Normativism and Knowledge of Analyticity  277 14 Modal Knowledge and Modal Methodology

284

T H E OD OR E L O C K E A N D A M I E T HOM A S S ON

Introduction 284 Worries about Modal Knowledge  285 The Normativist Account of Modal Knowledge  286 A Normativist Account of Essences  288 Worries Surrounding Independence Conditionals  290 Modal Monism vs. Modal Dualism  294 Worries about Knowledge of Conceptual Truths  295 Conclusion 297 15 Gettier’s Thought Experiments

302

JOAC H I M HORVAT H

Introduction 302 A Suppositional Reconstruction of Gettier’s Thought Experiments  303 Williamson’s Counterfactual Reconstruction  316 Conclusion 320 16 Horvath on Gettier’s Thought Experiments

327

T I MO T H Y W I L L I A M S ON

Introduction 327 Hermeneutic Difficulties  327 Theories of Counterfactuals  329 Horvath on Suppositional Truth  332 Supposition and Conditionals  335 17 Challenges for an Experimentalist’s Skepticism about Cases M A RG O T S T ROH M I NGE R A N D J U H A N I Y L I -VA K KU R I

Introduction 338 What Is the Method of Cases?  340 Does PPB Apply the Method of Cases Appropriately by Its Own Criteria?  342

338

Contents  xi Epistemic Peerhood, Disagreement, and X-Phi 345 What Does the Method of Cases Have to Do with Modal Immodesty? And How Much Modally Immodest Philosophy Is There?  346 Is PPB Modally Immodest?  349 Prescriptions for Reform  350 18 In Defense of Modest Modal Skepticism

359

E D OUA R D M AC H E RY

The Method of Cases and Modal Epistemology  360 Are Arguments against the Method of Cases Self-Defeating?  362 Is Modal Skepticism Coherent?  364 Conceptual Analysis and Conceptual Engineering  366 Conclusion 370 Notes on Contributors Index

373 375

Introduction Anand Jayprakash Vaidya

The Central Questions The contributions and exchanges in this volume aim to address two main questions: a

The epistemology of modality: How can we come to know, or be justified in believing, that something is necessary, possible, contingent, essential, or accidental? b The methodology of philosophy: What is philosophical knowledge and how can we acquire it? In particular, what are the distinctions between a priori, a posteriori, armchair, and experimental methods and the prospects of these for the acquisition of philosophical knowledge?

In order to situate the reader, this introduction will only present the major approaches to the epistemology of modality, the experimental critique of rationalist approaches to philosophical methodology, and recent developments in the epistemology of modality. Kant Kant is my point of departure, since it is within his work that one finds discussion of both questions. It is standard to interpret him as holding that necessity is the same as universality. Simplifying a bit, I will interpret this as the view that necessity is truth in all possible worlds; and that contingency is truth in some possible worlds, but not all possible worlds. In his Critique of Pure Reason, Kant drew three distinctions between: (i) the epistemic notions of a priori and a posteriori knowledge, (ii) the metaphysical notions of necessity and contingency, and (iii) the semantic notions of analytic and synthetic truth. A primary concern of the Critique was the question: how is metaphysics possible as a discipline, if it is partly based on reason and partly based on experience? Simply put, Kant wondered: how can we acquire informative knowledge about the world, which is nevertheless necessary?

DOI: 10.4324/9781003002192-1

2 Introduction He held that for a statement to be known a priori is for it to be capable of being known independently of sense experience; and for a statement to be known a posteriori is for it to be knowable only on the basis of sense experience. For example, while mathematical statements are knowable a priori, external world knowledge is essentially a posteriori. Kant’s distinction between analytic and synthetic statements is drawn through the notion of conceptual containment. An analytic statement is a statement in which the concept expressed by the predicate term is contained in the concept expressed by the subject term. Using these distinctions, Kant held that: i ii iii iv

S is a priori iff S is necessary. S is a posteriori iff S is contingent. If S is analytic, then S is a priori. If S is synthetic, then either S is a priori or S is a posteriori.

Kant’s account of philosophical methodology is grounded in synthetic a priori cognition. Kant identified philosophy, in particular metaphysics, with the search for synthetic a priori knowledge. He held that metaphysics, like mathematics, is a synthetic a priori discipline. To arrive at philosophical knowledge, one must gain synthetic a priori knowledge— knowledge that is knowable independent of sense experience, which is yet necessary and informative. In virtue of the connections he drew between the a priori and the a posteriori and the necessary and the contingent, he established a connection between the methodology of philosophy and modal knowledge. In particular, he thought that we acquire philosophical knowledge through rationality and intuition. Kant is often seen as being the first philosopher to set up a partial semantic approach to the epistemology of modality because of the connection he drew between analytic truths and a priori truths. Kripke In his ‘Identity and Necessity’ (1971) and Naming and Necessity (1980), Kripke challenged Kant’s claim that necessity coincides with a priori knowledge and contingency coincides with a posteriori knowledge. One of Kripke’s counterexamples to Kant’s (i) is the identity statement ‘Hesperus = Phosphorus’. 1. ‘Hesperus = Phosphorus’ is true and knowable only a posteriori. 2. ‘Hesperus’ and ‘Phosphorus’ are both rigid designators, where a term is a rigid designator only if it picks out the same entity in every world in which it has reference. 3. ‘Hesperus = Phosphorus’ is necessarily true. ∴ 4. ‘Hesperus = Phosphorus’ is necessary and a posteriori.

Introduction  3 One of Kripke’s counterexamples to Kant’s (ii) is the statement ‘I exist’. 1. ‘I exist’ when asserted by a person is true of and known by that person a priori. 2. No human person necessarily exists. Each human person has a contingent existence. ∴ 3. ‘I exist’ is contingent and known a priori. Although Kripke himself does not advance a thesis about the nature of how philosophy should be done, he does offer a template for the epistemology of necessity. Where ‘□’ stands for the necessity operator, and ‘→’ for the material conditional, Kripke argues that the following is a model for knowledge of some necessary truths (AP)   P → □(P) (E)     P ∴ (C)   □(P) The basic idea is that through a priori philosophical analysis, we can come to know certain conditionals that relate non-modal (probably essentialist) truths to modal truths. For example, Kripke (1980: 114, fn. 56) argues that given that an object o originates from a certain particular material origin m, it is essential that o originate from m. It is an a priori philosophical matter whether the origin of an object is an essential property of the object. However, once we accept that origin is essential, we are in a position to accept a conditional in the form of (AP). Using a conditional in the form of (AP) as our background metaphysical principle, in addition to empirical truths, such as that a certain table T originates from M, one can deduce that it is metaphysically necessary that T originate from M. 1. If T originates from M, then it is metaphysically necessary that T originates from M. 2. T originates from M. ∴ 3. It is metaphysically necessary that T originates from M. (3) serves as an instance of philosophical knowledge that is known a posteriori, which is nevertheless necessary and informative. Thus, contradicting Kant’s general thesis that philosophical knowledge is a priori, necessary, non-analytic, and informative. Kripke’s position on the necessary a posteriori presents an obstacle for substantive philosophical knowledge independent of contributions from science. The reason why is that often enough we can think something is

4 Introduction possible merely because we lack knowledge of what is necessary or ­essential. Consider the following line of reasoning found in the work of Yablo (1993). 1. S finds p conceivable, because S can construct a situation in which p seems true because there are no obvious contradictions in the scenario that seems to show that p is true. 2. However, Q is true, necessary, only knowable a posteriori, and □Q rules out that p is possible. ∴ 3. A priori conceiving of the possibility of p is, in general, always open to being undermined by a necessary truth that can only be known a posteriori. Rationalist Accounts In the wake of Kripke’s work, many philosophers sought to rebuild an account of the nature of philosophical knowledge that is a priori and rationalistic. The following are the main contributions to this enterprise: Bealer’s (2002) Christopher Peacocke’s (1999) Being Known, George ­ ‘Modal Epistemology and The Rationalist Renaissance’, and David Chalmers’ (2002) ‘Does Conceivability Entail Possibility?’ Peacocke (1999) argues that while it is true that some instances of necessity are known a posteriori as Kripke shows, we can factor those instances of knowledge into an a priori component and an a posteriori component as Kripke himself does. Peacocke defends three theses; thesis II is most relevant for moderate rationalism. Thesis (II): In every case in which a content containing a metaphysical modality is known, any modal premises in the ultimate justification which underwrites the status of the belief as knowledge are a priori premises. Chalmers (2002) argues for weak modal rationalism (WMR). (WMR): Primary Positive Ideal Conceivability entails Primary Possibility. Central to WMR is the distinction between primary and secondary intensions, which for our purposes I will skip over. The important result that Chalmers defends is that a posteriori necessities don’t rule out an a priori rational route to possibility and necessity because there are two methods by which we can evaluate a modal claim. Consider (T). (T): It is possible that water contains carbon and oxygen only and no hydrogen.

Introduction  5 On the one hand, (T) can be evaluated as a hypothesis about how our world is, an actual world evaluation. On the other hand, (T) can be evaluated counterfactually with respect to what water could be at another world, given what it is at the actual world. On the counterfactual world evaluation, (T) is false taking our world, Earth, as the actual world. Given that water contains hydrogen on Earth, no counterfactual situation relative to Earth contains water, if it fails to contain hydrogen. On the actual world evaluation, however, (T) is true at some scenario, if the description associated with water at that scenario rationally leads one to the conclusion that water contains only carbon and oxygen. For example, if all the lakes and rivers and oceans, and what falls from clouds, and what people use to quench thirst contains only carbon and oxygen, then it is rational to conclude that (T) is true at the scenario. It is furthermore a mistake to hold that water must contain hydrogen from an epistemic point of view because when such a situation is considered as actual, we are rationally led to the conclusion that water doesn’t contain hydrogen. Bealer (1987, 1999) argues for the autonomy of philosophical knowledge based on the claim that philosophical truths must be necessary truths. He states: In being interested in such things as the nature of mind, intelligence, the virtues, and life, philosophers do not want to know what those things just happen to be, but rather what those things must be. Bealer (1987, 289) Bealer’s theory of modal reliabilism offers an account of how we have modal knowledge of what is necessary so as to give us access to philosophical knowledge. Bealer’s account rests on a theory of determinate understanding of concepts and the notion that intuitions are evidence. The core idea is that we have reliable modal knowledge about what a concept can apply to and what it cannot apply to, based on our determinate understanding of the concept. When our intuitions about concept application are not truth-tracking, it is best explained by the fact that either we were not attentive or we lack determinate understanding. Thus, the autonomy of philosophy from science is grounded in necessary truths and our access to these necessary truths via intuitions based on determinate understanding. Counterfactual Armchair Accounts Williamson (2005, 2007a,b) offers a counterfactual theory of modal knowledge that makes an explicit connection between the twin questions of modality and methodology. The key theses of Williamson’s counterfactual theory are: Logical Equivalence: Metaphysical possibility and necessity can be proven to be logically equivalent to counterfactual conditionals.

6 Introduction Epistemic Pathway: Counterfactual reasoning in imagination through the method of counterfactual development can provide one with justified beliefs or knowledge about metaphysical possibility and necessity. Consider the following example from Williamson: Suppose that you are in the mountains. As the sun melts the ice, rocks embedded in it are loosened and crash down the slope. You notice one rock slide into a bush. You wonder where it would have ended if the bush had not been there. A natural way to answer the question is by visualizing the rock sliding without the bush there, then bouncing down the slope into the lake at the bottom. Under suitable background conditions, you thereby come to know the counterfactual: If the bush had not been there, the rock would have ended in the lake. (2007b: 142) Simply put, we are justified in asserting that A is possible when a robust and good counterfactual development of the supposition that A does not yield a contradiction. We are justified in denying that A is possible when a robust and good counterfactual development of A yields a contradiction. Williamson’s account of the connection between modality and methodology derives from his commentary on the traditional distinction between a priori and a posteriori knowledge. Many contemporary theorists maintain that what separates the a priori from the a posteriori is that in the former case, experience only plays an enabling role—a role in enabling possession of a concept for an individual thinker—while in the latter case, experience plays not only an enabling role, but an evidential role—the justification for a claim involving the concept requires appeal to experience by the thinker making the claim. However, Williamson (2009) maintains that several instances of counterfactual knowledge (the route by which we acquire modal knowledge) will be neither a priori nor a posteriori but would fall into the category of what he calls armchair knowledge. Williamson’s armchair knowledge is neither strictly a priori nor a posteriori. Hybrid Essentialist Deduction Accounts E.J. Lowe (2008a, 2012) and Bob Hale (2013) provide a picture of our knowledge of modality that contrasts with accounts that take conceivability (such as Chalmers), intuition (such as Bealer), or counterfactuals (such as Williamson), to be our fundamental source of justification for believing metaphysically modal truths. Lowe and Hale have independently developed accounts of the epistemology of modality based on metaphysical essentialism. The two core theses of metaphysical essentialism are:

Introduction  7 (i) entities have essential properties or essences that are not merely dependent on language, and (ii) not all necessary truths capture an essential truth or the essence of an entity. Although their views differ at crucial points in the epistemic landscape, the program they share maintains the following: Metaphysical Grounding: The essential properties or essences of entities are the metaphysical ground of metaphysical modality. When we look for an explanation of why something is metaphysically possible or necessary we ultimately look to the essential properties or essences of the entities involved. Epistemic Guide: The fundamental pathway to acquiring knowledge of metaphysical modality derives from knowledge of essential properties or essences of the entities involved. When we look for an explanation of how we can know metaphysical modality, we ultimately look to our knowledge of essential properties or essences as the basis upon which we make inferences to metaphysical modality. Following Fine (1994), who is building off Aristotle’s work on essence, Lowe and Hale share the view that the essential properties of an entity are distinct from the mere metaphysical necessities that are true of the entity, given that essential properties are more fine-grained than necessary properties. As a consequence, we cannot simply take essential properties or essences to be what an object has in every possible world in which it exists. While Lowe sees a strong connection between the twin questions of modality and methodology where he uses knowledge of essence arrived at through intuition and understanding as central to philosophical methodology, Hale is more or less silent on the issue of methodology. Hale is a necessity-first theorist in the epistemology of modality. A necessity-first approach holds that we first arrive at knowledge of necessary truths, and then derive knowledge of possibility through compatibility with knowledge of necessity. So, he is interested primarily in giving an account of our knowledge of necessity. In addition, Hale is not merely concerned with showing that there are necessary truths that can be known a posteriori, but rather with showing how the simple inferential model can be used in a great variety of cases. For example, the real definition of a circle is that it is a set of points in a plane equidistant from a given point. Thus, the essence of the kind circle is that anything that is a circle is a set of points in a plane equidistant from a given point. The property of being a circle is incompatible with the property of being a rectangle. Thus, given the essence of circles, it is metaphysically impossible for anything that is a circle to be a rectangle at the same time. Note that the discussion here is not about whether a given circle c could have been a rectangle. Rather, the point here is that

8 Introduction the kind circle is incompatible with the kind rectangle. Where ‘ES’ is an essentialist operator (read: it is essential that), we move from essence to necessity to knowledge of possibility via the following route. 1. 2. 3. ∴ 4. 5. ∴ 6.

P → ES(P), premise ES(P) → □(P), premise P, premise □(P), from 1-3 by hypothetical syllogism □(P) → ¬◊(Q), premise ¬◊(Q), from 4-5 by modus ponens

Hale’s aims to offer a robust inferentialist account of modal knowledge that is necessity-first and can handle both a priori cases, such as geometric objects, and a posteriori cases of natural kinds, such as water, light, and heat. Empiricist Accounts Work on the epistemic value of intuitions is part of the base for rationalism, especially in the work of Bealer. However, in the first part of the 2000s, new work on intuitions was used to challenge their epistemic value. Experimental philosophers, such as Jonathan Weinberg, Shaun Nichols, and Stephen Stich (WNS) argued that empirical research challenges the idea that we have reliable intuitions about the application of concepts to cases. They argued that intuitions vary across cultures. Experimental philosophers also argued that one’s intuitions, for example, about whether a given case was a case of knowledge, are sensitive to the order in which the cases are presented. The period also produced work that argued for empiricist approaches to the epistemology of modality. Empiricists working in the epistemology of modality argued that there were inadequacies in the rationalist program. Sonia Roca-Royes (2010, 2011) argued that rationalist approaches, such as those forwarded by Peacocke (1999) and Chalmers (2002) implicitly rested on knowledge of essence or constitutive principles. In the case of Peacocke (1999), Roca-Royes (2010) draws attention to the fact that on Peacocke’s epistemology of modality, our knowledge of modality is parasitic on our knowledge of constitutive principles, whether these principles are implicitly or explicitly known. For example, we determine that something is possible or necessary for an entity in part through our knowledge of what is constitutive of the entity. If we know that being human is a constitutive property of a given human, such as Tom, then we can come to know that it is impossible for Tom to be a zebra, but that it is possible for Tom to be born somewhat later than he was actually born. As a consequence, a comprehensive account of modal knowledge is incomplete

Introduction  9 without a picture of how we come to know the relevant essentialist or constitutive principles involved in modal knowledge. Empirically inclined modal epistemologists, like Fischer (2017) and later Mallozzi (2018, 2020, 2021) and Wirling (2020) noticed the relevance of proposing non-uniform accounts of the epistemology of modality. The basic idea of non-uniformism is that modal knowledge can be acquired in different ways depending on what kind of thing one is aiming to get modal knowledge of. Perceptualists, such as Legg (2012), Strohminger (2015), and Legg and Franklin (2017), advanced the non-uniformist approach by challenging the idea that modal knowledge cannot be gained via perception. Strohminger (2015) challenges Kant’s notion that perception can only tell us what is or is not the case, but not what must be the case or what merely could be the case, by focusing on a class of cases where one gains modal knowledge through perception. For example, consider the case of a cup C on a table T at location L. Arguably, one can see that C could have been at L*. ­Non-uniformism is also a driving force in two of the most prominent empiricist approaches to modal knowledge: inductive and abductive accounts. Sonia Roca-Royes and Bob Fischer both take a modest approach to the epistemology of modality. Roca-Royes (2017, 2018) divides the space of epistemological investigation via the ontological distinction between concrete and abstract entities, thus offering a non-uniform account. Fischer (2017) divides the space of epistemological investigation via a topical distinction between ordinary and extraordinary claims. Fischer’s basic idea is that we are justified in believing an extraordinary modal claim, m, only if we are justified in believing a theory T from which m follows. For example, we are justified in believing that mind-body dualism is metaphysically possible only if we are justified in believing a theory T from which mind-body dualism follows. Fischer holds that abductive methods for theory choice, such as using theoretical virtues like simplicity, are central to being justified in believing a theory. If the theory T from which one would be justified in believing that mind-body dualism is metaphysically possible is not the simplest theory, all else being equal, one would not be justified in believing it, and thus not be justified in believing that mind-body dualism is metaphysically possible. Finally, while Legg and Roca-Royes defend empiricist accounts of modal knowledge, they remain silent over the proper methodology of philosophy. In contrast, Fischer is interested both in the epistemology of modality and in the proper methodology of philosophy, favoring a theory-based approach. Beyond the Standard Model Contemporary work in the epistemology of modality and philosophical methodology aims to go beyond the foundational questions laid out in

10 Introduction the work of Yablo’s (1993) paper on conceivability and possibility, Van Inwagen’s (1998) skepticism about modal knowledge, and WNS’s (2001) skepticism about the reliability of intuitions about cases. Here are some recent developments. Otávio Bueno and Scott Shalkowski in a series of papers (2009, 2013) have developed a modalist account of modality. They argue that modality is fundamental and cannot be analyzed in terms of quantification over worlds. Simply put, modality cannot be reduced to anything that is completely non-modal. The challenge for their view is to develop an account of the epistemology of modality that makes sense of it. How does one know that something is possible when possibility is irreducibly modal? Boris Kment (2006a,b, 2014, 2021) offers an account of modality that focuses on comparative possibility. How easily something could have been the case in comparison to something else. Linguists have long focused on comparative modal claims, while philosophers for the most part have stayed away from them. Although Lewis offers an analysis of modality on which the notion of a closest possible world plays a central role, until Kment’s work most philosophers had not thought to develop an account of the metaphysics and epistemology of modality that takes the idea of spheres of possibility where something is more easily possible relative to other things. Amie Thomasson (2020) offers a normative account of modality. On this account, metaphysical modal discourse is not descriptive. Instead, modal vocabulary serves the function of signaling (in the object language) constitutive semantic and conceptual rules. This is, she contends, compatible with the meaningfulness of modal vocabulary and, as such, modal claims are truth-apt and mind-independent: even if no one ever existed, seals would necessarily (still) be mammals. Barbara Vetter (2015, 2020) continues in the empiricist tradition by offering a perceptual and agency-based account of modal knowledge. She is a thorough-going naturalist. And she joins Timothy Williamson (2009) in being an anti-exceptionalist about philosophical methodology—neither holds the view that philosophy is discontinuous with the sciences because it has a special methodology. However, she differs from Williamson in that she takes potentiality to be a non-reductive basis for modality because modality is explained by potentiality and abilities. And, in addition to Kment (2014), Vetter has also contributed to the development of comparative modality. She has articulated a view where modality is closely tied to potentiality in a way that allows for a graded view of modality. She has argued that necessity is a kind of maximal potentiality. A potentiality that cannot fail to manifest. As a consequence, her view also shows that there are grades of possibilities. Antonella Mallozzi has recently articulated three important moves in the epistemology of modality. In Mallozzi (2018), she argues for a metaphysics-first approach to an account of the

Introduction  11 epistemology of modality. This means that she favors epistemic accounts that have solid foundations in the metaphysics of modality. The leading idea is that it doesn’t make sense to think that modality can be known if we don’t have a strong enough grasp of what the entities are that we know about. Second, Mallozzi (2018) has defended an account of essences that are superexplanatory. The leading idea here is, on the one hand, to demystify essences, and, on the other hand, to show how ordinary scientific explanation can be used to give a serviceable account of what essences are. Finally, in Mallozzi (2020), she applies her account of essences as superexplanatory properties to criticize and improve Williamson’s (2009) counterfactual approach, via imagination, to the epistemology of modality. Anand Vaidya and Michael Wallner (2021) have articulated and defended, what they call, The Problem of Modal Epistemic Friction. Take imagination as the candidate for the mental capacity we use in our pursuit of modal knowledge. In order for us to be properly guided to modal knowledge, we cannot apply imagination in a completely unrestricted manner. If no restrictions apply, we could imagine all sorts of impossible things like water without hydrogen, transparent iron, etc. Hence, some restrictions have to be in place. Thus, a central problem in the contemporary study of the epistemology of modality is how to solve the problem of modal epistemic friction in a way that captures the freedom of imagination to generate new information, while at the same time restricting it enough so that the imagination has normative epistemic force.

References Bealer, George. (1987). “Philosophical Limits of Scientific Essentialism”. ­Philosophical Perspectives, 1: 289–365. Bealer, George. (1999). “A Theory of the A Priori”. Philosophical Perspectives, 13: 29–55. Bealer, George. (2002). “Modal Epistemology and the Rationalist Renaissance”. In T. Gendler and J. Hawthorne (Eds.), Conceivability and Possibility. Oxford University Press: 71–125. Bueno, Otávio & Scott Shalkowski. (2009). “Modalism and Logical Pluralism”, Mind, 118: 295–321. Bueno, Otávio & Scott Shalkowski. (2013). “Logical Constants: A Modalist Approach”, Nous, 47: 1–24. Chalmers, David. (2002). “Does Conceivability Entail Possibility”. In J. ­Hawthorne & T. Gendler (Eds.), Conceivability and Possibility. Oxford: Oxford University Press: 145–200. Fischer, Robert. (2017). Modal Justification via Theories. Dordrecht: Springer Publishing. Godman, Marion, Antonella Mallozzi, & David Papineau, (2020). “Essential Properties Are Super-Explanatory: Taming Metaphysical Modality”. Journal of the American Philosophical Association, 6(3): 316–334.

12 Introduction Hale, Bob. (2003). “Knowledge of Possibility and Necessity”, Proceedings of the Aristotelian Society, 103: 1–20. Hale, Bob. (2013). Necessary Beings: An Essay on Ontology, Modality, and the Relations between Them. Oxford: Oxford University Press. Kment, Boris. (2006a). “Counterfactuals and Explanation”, Mind, 115(458): 261–310. Kment, Boris. (2006b). “Counterfactuals and the Analysis of Necessity”, ­Philosophical Perspectives, 20: 237–302. Kment, Boris. (2014). Modality and Explanatory Reasoning. Oxford: Oxford ­University Press. Kment, Boris. (2021). “Essence and Modal Knowledge”, Synthese 198(8): 1957–1979. Kripke, Saul. (1971). “Identity and Necessity”. In M.K. Munitz (Ed.), Identity and Individuation. New York: New York University Press: 135–164. Kripke, Saul. (1980). Naming and Necessity. Cambridge, MA: Harvard University Press. Legg, Catherine. (2012). “The Hardness of the Iconic Must: Can Peirce’s Existential Graphs Assist Modal Epistemology?”, Philosophia Mathematica, 20(1): 1–24. Legg, Catherine & James Franklin. (2017). “Perceiving Necessity”, Pacific Philosophical Quarterly, 98(3): 320–343. Lowe, E. J. (2008a). “Two Notions of Being: Entity and Essence”, Royal Institute of Philosophy Supplements, 83(62): 23–48. Lowe, E. J. (2008b). “Essentialism, Metaphysical Realism, and the Errors of Conceptualism”, Philosophia Scientiæ, 12(1): 9–33. Lowe, E. J. (2012). “What Is the Source of Our Knowledge of Modal Truths”, Mind, 121(484): 919–950. Mallozzi, Antonella. (2018). “Putting Modal Metaphysics First”, Synthese 198(8): 1937–1956. Mallozzi, Antonella. (2020). “Superexplanations for Counterfactual Knowledge”, Philosophical Studies 178(4): 1315–1337. Mallozzi, Antonella, Anand Vaidya, & Michael Wallner. (2021). “The Epistemology of Modality”, In The Stanford Encyclopedia of Philosophy (Fall 2021 Edition), Edward N. Zalta (Ed.), URL: https://plato.stanford.edu/archives/ fall2021/entries/modality-epistemology/. Roca-Royes, Sònia. (2010). “Modal Epistemology, Modal Concepts and the Integration Challenge”, Dialectica, 64(3): 335–361. Roca-Royes, Sònia. (2011). “Conceivability and De Re Modal Knowledge”, Noûs, 45(1): 22–49. Roca-Royes, Sònia. (2017). “Similarity and Possibility: An Epistemology of De Re Possibility for Concrete Entities”, In Modal Epistemology After Rationalism, Fischer and Leon Dorderecht: Springer Publishing: 221–245. Roca-Royes, Sònia. (2018). “Rethinking the Epistemology of Modality for Abstracta”, in Ivette Fred-Rivera and Jessica Leech (Eds.), Being Necessary. Themes of Ontology and Modality from the Work of Bob Hale, Oxford: Oxford University Press: 245–265. Strohminger, Margot. (2015). “Perceptual Knowledge of Nonactual Possibilities”, Philosophical Perspectives, 29(1): 363–375.

Introduction  13 Swain, Stacey, Joshua Alexander, & Jonathan M. Weinberg. (2008). “The ­Instability of Philosophical Intuitions: Running Hot and Cold on Truetemp”, Philosophy and Phenomenological Research, 76(1): 138–155. Weinberg, Jonathan M., Shaun Nichols, & Stephen Stich. (2001). “Normativity and Epistemic Intuitions”, Philosophical Topics, 29(1/2): 429–460. Thomasson, Amie. (2020). Norms and Necessity. Oxford: Oxford University Press. Vaidya, Anand Jayprakash, & Michael Wallner. (2021). “The Epistemology of Modality and the Problem of Modal Epistemic Friction”, Synthese, 198(S8): S1909–S1935. Van Inwagen, Peter. (1998). “Modal Epistemology”, Philosophical Studies 92: 67–84. Vetter, Barbara. (2015). Potentiality: From Dispositions to Modality, Oxford: Oxford University Press. Vetter, Barbara. (2020). “Perceiving Potentiality: A Metaphysics for ­A ffordances”, Topoi, 39(5): 1177–1191. Wirling, Ylwa Sjölin. (2020). “Non-Uniformism about the Epistemology of Modality: Strong and Weak”, Analytic Philosophy 61(2): 152–173. Williamson, Timothy. (2005). “Armchair Philosophy: Metaphysical Modality and Counterfactual Thinking”, Proceedings of the Aristotelian Society, 105(1): 1–23. Williamson, Timothy. (2007a). “Philosophical Knowledge and Knowledge of Counterfactuals”, Grazer Philosophiche Studien, 74: 89–123. Williamson, Timothy. (2007b). The Philosophy of Philosophy, Malden, MA: Wiley-Blackwell. Yablo, Stephen. (1993). “Is Conceivability a Guide to Possibility?” Philosophy and Phenomenological Research, 53: 1–42.

Summary of Articles Duško Prelević

In “Modality, Worlds, Essence, and Modal Knowledge,” Boris Kment develops his theory of modality and offers an account of how modal knowledge is possible. Kment first offers an analysis of metaphysical modality aimed at explaining the familiar links between modality and related notions, such as possible worlds, counterfactuals and essence. His account entails that a proposition is metaphysically possible if and only if it is logically consistent (in a non-modal sense) both with the metaphysical laws and with the truths about which propositions are metaphysical laws. He goes on to argue that modal knowledge can be gained on the basis of knowledge about the metaphysical laws. Finally, Kment offers an explanation of how knowledge of the metaphysical laws is possible by appealing to inference to the best explanation and to our conceptual or linguistic competence. In “An Agency-Based Epistemology of Modality,” Barbara Vetter develops an agency-based modal epistemology that runs in parallel with her earlier dispositional theory of the metaphysics of modality. She notices that as a matter of fact, agents in the world do possess pertinent modal knowledge of what objects in their environment can do, and of what those agents can do with them. Vetter then goes a step further trying to explain that our knowledge of unactualized metaphysical possibilities is attainable, including the knowledge of extraordinary possibilities that are present in philosophical debates. She argues that such knowledge can be attained once we realize in which sense our ordinary modal knowledge is contextually restricted. Vetter also argues that there are natural boundaries as to metaphysical modality that are well suited to some common views in the metaphysics of modality. In “The Price of Sensitivity,” Rebecca Hanrahan considers the costs associated with accepting or rejecting the sensitivity counterfactual. This counterfactual holds that if the modal status of a proposition changed, one’s experiences would change as well. Hanrahan argues that if one accepts this counterfactual, there would be a collapse of the modal, which means that all conceptual possibilities would be physical and hence metaphysical possibilities, and vice versa. If one rejects this counterfactual and accepts the insensitivity counterfactual instead, one would have to hold conceptual possibilities to be distinct from either physical or metaphysical possibilities. DOI: 10.4324/9781003002192-2

Summary of Articles  15 The first option leads one to having to countenance metaphysically inert properties, while the other implies accepting metaphysically strong possibilities. Hanrahan opts for accepting strong metaphysical possibilities, and believes that doing so accords more with our modal intuitions. In “Modal Epistemology for Modalists” Otávio Bueno and Scott ­Shalkowski scrutinize various forms of rationalist and n ­ on-rationalist modal epistemologies, trying to provide the middle ground between them. They offer an account of our modal knowledge in terms of ­non-idealized reasoning from certain principles, which are, according to them, local and dependent on the particularities in each domain. Relatedly, they hold that properties and empirical regularities are implicitly modal, which in turn enables us to build up a modest account of modal knowledge anchored in the actual world. In “How (Meta-)Semantics Defuses Modal Pessimism,” Christian Nimtz develops a semantic answer to modal pessimism (that is, moderate modal skepticism), a view that modal knowledge is characteristically limited. Nimtz argues that science (which delivers true statements of empirical fact) and semantics (which allows identifying predicates he calls “paradigm terms”) enable us to have modal knowledge that extends beyond the limits posed by modal skeptics, as well as beyond the Kripkean template of identity statements and rigid designators. In “How Things Have to Be,” Nathan Salmon develops his view defended in Reference and Essence by criticizing Penelope Mackie’s and Scott Soames’s arguments for Hilary Putnam’s thesis that the ­direct-reference theory of natural-kind terms, together with empirical premises, supports non-trivial essentialism. Salmon defends a distinction between rigid and non-rigid general terms. He also casts doubt on Mackie’s and Soames’s reformulation of Putnam’s ostensive definition of natural-kind term “water,” arguing that, on the one hand, it does not correctly explain the meaning of “water” as a substance designator, while, on the other hand, it supports only trivial modal essentialism concerning things like explanatory kinds, and not non-trivial modal essentialism concerning things like natural compounds. In “In Search of a Structurally Complete Epistemology of Essence,” Michael Wallner tries to show, on the one hand, that Bob Hale’s and Jonathan Lowe’s essence-based accounts of modal knowledge are incomplete, while, on the other hand, Husserl’s essence-based account, which appeals to eidetic variation, can be defended against criticism. Still, Wallner holds that eidetic variation alone cannot provide a structurally complete epistemology of essence, given that imagination involved in it is not considered a directly justifying mental capacity. However, he argues that once imagination in eidetic variation is restricted properly, a structurally complete epistemology of essence is available to Husserl’s account. In “Morals and Modals: Puzzling about the Dual Use of Modal Verbs,” Arindam Chakrabarti and Ian Nicolay try to build up an epistemology of

16  Duško Prelević modality by using insights from classical Indian epistemology, logic, grammar and ­semantics. In doing so, they analyze key terms involved in modal discourse that can be found in classical Indian philosophy, in particular, the ambiguity of the optative verb syāt. Chakrabarti and Nicolay emphasize that counterfactual thinking was present in classical Indian philosophy, but unlike in contemporary debates, in which the main focus is on the meaning and truth of counterfactual conditionals, more attention was paid on how those conditionals are used, that is, on their conversational pragmatics and the ethics of belief. They hold that within this philosophical tradition deontic modality was considered primary, whereas alethic modality was considered derivative, which has some commonalities with contemporary views like that of Amie Thomasson’s modal normativism. In “The Explanatory Power of Modal Rationalism,” Duško Prelević argues that a Chalmersian conceivability-based modal rationalism can explain in a non-question-begging way non-trivial essentialist de re modal knowledge, in particular, the knowledge of the Essentiality of Origin and of the Essentiality of Kind, contrary to Sonia Roca-Royes’s influential criticism. By using pluralism in geometry as illustration, Prelević draws the distinction between absolute and relative conceivability and applies these insights into the metaphysical theories of objecthood in order to show that conceivability (or inconceivability) might depend on a previously accepted axiomatization (or a set of beliefs). He argues that when it comes to the Essentiality of Kind, had Roca-Royes’s objection tacitly assumed the Aristotelian view of objects, this would exclude the negation of the very principle and render the entire enterprise self-defeating. Prelević also points out that our knowledge of the Essentiality of Origin can be explained by using a similar line of reasoning. In “Conceivability: Still Not Enough: A Response to Prelević,” Sonia Roca-Royes offers a clarification of her argument against non-epistemic conceivability-based accounts of modal knowledge in order to handle Duško Prelević’s criticism voiced in this volume. Although Roca-Royes allows that Prelević’s distinction between absolute and relative conceivability might be useful to conceivability-based accounts that do not rely upon the two-dimensional semantic framework (like the ones proposed by Timothy Williamson and Stephen Yablo), she thinks that more should be done in clarifying the very distinction. Roca-Royes also thinks that her critique of Chalmers’s account was hypothetical to the effect that she is not committed to the Aristotelian conception of objecthood, as well as that conceivability alone is still not enough for elucidating our essentialist de re modal knowledge. In “Reviving the Modal Account of Essence,” Rebecca Chan offers an amended version of the modal account of essence by appealing to determinacy, according to which determinateness of properties (and their essentialness thereof) comes in degrees, depending on how narrow or wide is the scope of what a thing that has that property could be, and

Summary of Articles  17 sparseness, according to which necessarily coextensive properties are the same property. She argues that her fully reductive modal account can handle Kit Fine’s famous counterexamples to the modal account, as well as that it should be preferred over other accounts that can be found in the literature due to its simplicity and non-arbitrariness. In “A Neo-Aristotelian Reply to a Modalist,” Sanna Mattila d ­ iscusses Rebecca Chan’s modalist account proposed in this volume and defends neo-Aristotelian essentialism against it. She extends the discussion between modalists and neo-Aristotelians beyond Kit Fine’s counterexamples, and stresses other significant differences between these two views, arguing that Chan’s own account suffers from problems that neo-Aristotelians can avoid. According to Mattila, essences have important explanatory roles in the neo-Aristotelian framework, but the modalist account is not suitable to explain them in a satisfactory way. In “Semantic Rules, Modal Knowledge, and Analyticity,” Antonella Mallozzi argues against modal normativism, a view that modal knowledge can be attained by conceptual mastery, reasoning abilities and corresponding empirical information, and argues that conceptual mastery is not enough, but should be accompanied by corresponding essentialist knowledge (knowledge of general essentialist-bridge principles and facts). She allows that modal normativists can explain our knowledge of logical-conceptual modality or analyticity (without relying upon modal monism), yet denies that they can do the same when it comes to genuine metaphysical modality. In order to show that, Mallozzi offers a counterexample to modal normativism, by envisaging a situation in which a trained chemist, who mastered the rules for using the term “water” and knows relevant empirical information about water, still wonders whether it is metaphysically possible for water to contain carbon. She holds that modal normativism has to deem the chemist as incompetent, or irrational. In “Modal Knowledge and Modal Methodology,” Theodore Locke and Amie Thomasson defend modal normativism, and argue that this view demystifies our knowledge of metaphysical modality. They respond to Antonella Mallozzi’s criticism voiced in this volume, by claiming that her purported example does not show what she claims it does, namely, that some grasp of essentialist bridge-principles is required for knowledge of metaphysical modality. Independently of that, they argue that modal normativists could accept, for the sake of scientific utility, talk of essences and that they could provide a satisfactory epistemology of essence, again, by appealing to conceptual mastery, reasoning abilities and empirical knowledge. In “Gettier’s Thought Experiments,” Joachim Horvath proposes a suppositional reconstruction of Gettier’s thought experiments and raises some worries about Timothy Williamson’s competing counterfactual account. He recalls criticisms against Williamson’s account, according to which it faces the problem of deviant realizations (the existence of

18  Duško Prelević scenarios that makes a thought experiment under consideration ineffective). Horvath thinks that Gettier’s purported examples should be understood as metaphysical possibilities presented in the form of hypothetical suppositions, in which certain facts about the actual world are irrelevant to the assessment of their plausibility. He resorts to what he calls the “substitution gambit” (the substitution of actual names of people for fictional names used by Gettier) in order to avoid the problems related to the semantics of fictional names. Horvath also underscores commonalities between his and Williamsons’s accounts, but nonetheless holds that his own account is more in accordance with textual evidence that we can find in Gettier’s seminal paper. In “Horvath on Gettier’s Thought Experiments,” Timothy Williamson argues that his reconstruction of Gettier’s thought experiments is simpler than Joachim Horvath’s reconstruction, as well as that the latter reconstruction departs from the textual evidence we can find in Gettier’s paper. Williamson reminds us that he has revised his views of counterfactuals (the Lewisian account has been replaced with a view that “would” is considered as a necessity operator, contextually restricted), which enables him to avoid the problem of deviant realizations of thought experimental scenarios, stressed by Horvath. Williamson also holds that Horvath’s substitution gambit is ad hoc, not supported by textual and historical evidence when it comes to Gettier’s own paper and the time when it was written, and argues that Horvath remained silent on whether his account is committed to a non-standard metalogical framework. In “Challenges for an Experimentalist’s Skepticism about Cases,” Margot Strohminger and Juhani Yli-Vakkuri assess Edouard M ­ achery’s critique of the method of cases (the usage of thought experiments in philosophy) in view of his research in experimental philosophy. They argue that Machery’s criticism itself is modally immodest, and therefore self-defeating. Also, they point out that modal modesty or immodesty does not depend on the usage of the method of cases, given that certain philosophical views are modally modest, yet counterintuitive and do not rely upon the method of cases. In “In Defense of Modest Modal Skepticism,” Edouard Machery responds to Strohminger and Yli-Vakkuri and clarifies his views voiced in Philosophy within Its Proper Bounds. Machery argues that his critique of the method of cases is not self-defeating despite the fact that the very method has been used in his book more than once, given that, according to him, his critique of that method does not rely essentially upon its validity. Machery also claims that his view was mischaracterized to some extent by Strohminger and Yli-Vakkuri, given that his critique was directed against unrestricted claims about metaphysical modality, not to the restricted ones, like that of nomological possibility (or necessity), and the like. In the rest of his paper, Machery stresses the importance of conceptual analysis and prescriptive conceptual engineering.

1 Modality, Worlds, Essence, and Modal Knowledge Boris Kment

A theory of what it is for a proposition to be metaphysically necessary should satisfy a number of desiderata. Among other things, it should account for the characteristic connections between metaphysical modality and various other philosophically important notions, and it should explain how it is possible for us to possess the modal knowledge we have. I will describe the links between modal concepts and three other concepts (possible worlds, counterfactuals, and essence) in the first section. It is tempting to try to exploit the connections between modality and these other notions to give an illuminating account of necessity. ­However, existing attempts to do so face significant limitations. I will draw on previous work (Kment 2014) to propose a theory of modality that can shed light on the conceptual connections (Section “A Theory of Necessity”). Finally, after discussing some preliminaries in Section “Metaphysical Laws, Grounding, and Causal Explanation,” I will argue (Section “Modal Knowledge”) that the account outlined in Section “A Theory of Necessity” also yields an attractive explanation of how modal knowledge is possible.

Necessity, Possible Worlds, Essence, and Counterfactuals Let us begin by considering the connections that exist between metaphysical necessity and three other notions that are often invoked to clarify the nature of modality. Possible worlds. This connection is expressed by the following familiar biconditionals: (1) (a) A proposition is metaphysically necessary iff it is true (not only as things actually are but) at all possible worlds. (b) A proposition is metaphysically possible iff it is true at some possible world. These biconditionals are widely applied in philosophy. They can be useful when we try to decide modal questions. For example, when we are wondering whether a certain scenario S is possible, it is often helpful

DOI: 10.4324/9781003002192-3

20  Boris Kment to reformulate the question in terms of possible worlds: is there some ­possible world where S obtains? David Lewis famously used the two biconditionals in (1) in his analysis of modality. In his view, possible worlds are (to simplify somewhat) spatio-temporally extended objects like the universe we inhabit, existing in mutual spatio-temporal isolation.1 This account defines possible worlds in non-modal terms, thereby making it possible to give a reductive analysis of necessity as truth at all possible worlds.2 The approach has not gained a wide following, however. Not only is it hard to believe in the plurality of universes postulated by Lewis, but many philosophers find it implausible that Lewisian worlds form the subject matter of modal claims even on the assumption that they exist.3 The best-known alternative to Lewis-style realism identifies possible situations with entities whose existence seems easier to accept than that of Lewisian worlds, such as stories or other representations that could have correctly and exhaustively described reality, maximal states of affairs that could have obtained, or maximally strong properties the universe could have had.4 In all of these views, the property of being a possible world is itself seemingly modal, and (1) therefore does not look like a promising starting point for an illuminating theory of the nature of modality. These views seem to entail that (1), far from being an informative account of what necessity is, merely articulates a fairly trivial connection between two modal properties. But that leaves us without a satisfying explanation of why possible worlds have played such a central role in so many philosophical accounts of modality. Counterfactuals. Necessity is connected to counterfactuals in important ways. (a) Necessary truths are counterfactually robust. One characteristic feature of the necessary truths can be described, as a first shot, as follows: (2) If a proposition is necessary, then it would have held no matter what (else had been the case). Let us say that the proposition that P counterfactually implies the proposition that Q iff, had it been that P, it would have been that Q. Then, (2) can be restated as follows (“p” and “q” are propositional variables): (3) ∀q (□q → ∀p ( p □→ q)) A necessary truth is counterfactually implied by every proposition. It is controversial whether this principle is true, however. Some philosophers have defended it (Williamson 2007, 171–175, 2017, 2018, 2020; also see Lewis 1973a; Stalnaker 1968, 1996). Others, including myself, have rejected the claim that every necessary truth is counterfactually implied by every impossible proposition, including its own negation (see Nolan

Modality, Worlds, Essence, and Modal Knowledge  21 1997; Vander Laan 2004; Kment 2006, 2014; Brogaard and Salerno 2007a,b, 2013; Lange 2009, section 2.7; Berto et al. 2018). For example, it is necessary that whales are mammals. And yet, to many philosophers, it seems false to say: if whales had not been mammals, then whales would have been mammals. This suggests that the quantifier in “no matter what else had been the case” in (2) should be understood as restricted to metaphysically possible scenarios. So interpreted, (2) can be spelled out as the following principle. (4) ∀q (□q → ∀p (◊p → (p □→ q)))5 A necessary proposition is counterfactually implied by every possible proposition. Given classical logic, the idea underlying (4) can be reformulated as the claim that the following inference rule is valid (i.e., truth-preserving under any interpretation of the schematic sentence letters “P” and “Q”). (5) □Q, ◊P ⊢ P □→ Q (b) Counterfactual implication transmits possibility. We can capture an important connection between counterfactuals and possibility by saying that the following inference rule is valid. (6) ◊P, P □→ Q ⊢ ◊Q Assuming that □ and ◊ are duals, (6) can be derived from (5) with the help of (7).6 (7) ◊P ⊢ ¬((P □→ Q) & (P □→ ¬Q)) Since it is plausible that (7) is valid, it is natural to view the validity of (6) as a corollary of the validity of (5). (6) is commonly used in showing that a proposition is metaphysically possible. For example, a philosopher might say the following to convince an interlocutor that gold could have failed to be yellow: “There could have been a metal with the same microstructure as gold in the actual world that was not yellow. If there had been such a metal, then (that metal would have been gold and so) gold would not have been yellow.” This line of reasoning employs an instance of (6) (with P in (6) being instantiated with “There is a metal with the same microstructure as gold in the actual world that is not yellow” and Q with “Gold is not yellow”).7 The truth-conditions of ordinary-language counterfactuals are commonly held to be context-dependent. I do not claim that (5) and (6) are valid for such counterfactuals and for metaphysical modality in every context. Instead, I hold that they are valid in a specific class of contexts that characterizes most uses of counterfactuals in explanatory reasoning and

22  Boris Kment practical deliberation (Kment 2014, section 2.7, Kment forthcoming). P □→ Q in (5) and (6) is to be understood as having the truth-conditions that an ordinary-­language counterfactual with the same antecedent and consequent would have in such contexts. A detailed account of these truth-conditions is given in Kment (2014), chapters 8 and 9. (a) and (b) are not distinctive of metaphysical modality but are true for a larger class of modal properties. Consider transmission of possibility across counterfactual implication. There seem to be readings of (6) in which “◊” expresses a form of possibility different from metaphysical possibility and which make (6) valid. Note that the following argument would fit perfectly naturally into an everyday conversation, and it appears to be valid: (8) Team A could have won (the soccer tournament). For they could have hired Jones as a trainer, and if they had done that, they would have won. (8) is an application of (6) if “◊” in (6) is interpreted as expressing the same form of possibility as “could” in (8). Now, in most realistic scenarios in which (8) is uttered, “could” expresses a kind of possibility much stronger than metaphysical possibility. (Understood as a claim about metaphysical possibility, the conclusion of (8), i.e., the first sentence, is trivial.) Moreover, it is easy to imagine a long series of contexts c1, c2, … such that (i) “could” expresses a form of possibility stronger than metaphysical possibility in all of c1, c2, …, and (ii) “could” expresses a stronger form of possibility in ci+1 than in ci (i.e., for “Team X could have won” to be true in ci+1, Team X needs to be better than it has to be for the same sentence to be true in ci), and (iii) (8) seems to be a valid argument in all of c1, c2, …. This seems to show that there is a whole range of possibility properties that are transmitted across counterfactual implication. There are equally strong reasons to deny that counterfactual robustness is distinctive of metaphysical modality. First, many philosophers believe that there is a form of necessity (called “nomic,” “physical,” or “natural” necessity) that is weaker than metaphysical necessity, and which attaches to all propositions metaphysically necessitated by the natural laws. A high degree of counterfactual robustness, i.e., the power to support a wide range of counterfactuals, is usually thought to be the hallmark of this kind of necessity. Second, there are readings of (5) in which ◊ and □ express a form of modality different from both metaphysical and nomic modality and which make (5) valid. Suppose that we are watching the semi-finals of a sports tournament in which Team A loses after their goalkeeper commits an avoidable blunder. After the game, I say: (9) (i) Team A could have made it to the finals, though (ii) they could not have won the tournament.

Modality, Worlds, Essence, and Modal Knowledge  23 Clearly, the occurrences of “could” in this utterance express a form of possibility stronger than metaphysical possibility. (Otherwise, (9)(i) would be trivial and (9)(ii) would be obviously false.) You would take (9) to commit me to: (10) If Team A had made it to the finals, they would (still) not have won the tournament. Thus, where ◊ represents the notion of possibility expressed by “could” in (9), it seems that ◊P and ¬◊Q entail P□→¬Q. Assuming that □ is the dual of ◊ and that ¬◊Q is therefore equivalent to □¬Q, it seems that ◊P and □¬Q entail P□→¬Q, and more generally, that ◊P and □Q entail P□→Q. As before, we can imagine a series of contexts in which “could” in (9) expresses increasingly strong notions of possibility and the inference from (9) to (10) remains valid. (5) therefore seems to be valid for a broad range of forms of modality. A central element of our conception of necessity seems to be the idea that the truth of a necessary proposition is unconditional, or secure in some non-epistemic sense, while that of a contingent proposition is contingent on what else is the case (and is therefore not unconditional). The thesis that the necessary truths are particularly counterfactually robust seems to provide a good way of capturing this idea (Kment 2006, 2014; Lange 2009). The foregoing reflections suggest that this connection to unconditionality or security is not distinctive of metaphysical modality but characteristic of a broader family of modal properties. A theory of modality should accommodate this finding. Timothy Williamson and Marc Lange have offered accounts of necessity in terms of counterfactual robustness. Williamson (2005, 2007, chapter 5) holds that it is the hallmark of the metaphysical necessities that they are counterfactually implied by all propositions whatsoever—a feature captured by rule (3). (Also see McFetridge 1990; Hill 2006.) In particular, metaphysical necessities are the only propositions that are counterfactually implied even by their own negations. This account can accommodate the centrality of counterfactual robustness to the notion of necessity. However, as mentioned above, it is highly controversial whether (3) holds for metaphysical necessity. The account I will propose stays clear of this controversial commitment. Lange’s theory (1999, 2000, 2005, 2009) focuses on logical and natural necessity. On his account, the natural necessities are, in all contexts, counterfactually implied by all propositions consistent with them, and the same is true of the logical necessities. Moreover, the natural necessities and the logical necessities are the only deductively closed collections of truths of which this is true. This view avoids commitment to (3). (On Lange’s account, (3) is false—see Lange (2009, section 2.7).) However, Lange focuses on a property that he takes to be exemplified only

24  Boris Kment by the nomic and logical necessities. Consequently, his account does not ­illuminate the aforementioned broader family of modal concepts characterized by counterfactual robustness and the transmission of possibility across counterfactual implication (i.e., the family that includes the modal properties expressed by “could” in typical utterances of (8) and (9)). Essence. Metaphysical necessity is often believed to be intimately connected to essence. The essential truths about a articulate part of what it is to be a, or part of what makes something a. For example, it is an essential truth about gold that it is a metal, since being a metal is part of what it is to be gold, or part of what makes a substance gold. By contrast, it is merely an accidental truth about gold that it fills the cavities and adorns the necks of the rich, since that is not part of what makes a stuff gold. It used to be common to define essentialist notions in modal terms. For example, Kripke (1981) thought that a property was essential to an entity iff the entity could not have existed without having the property. However, Kit Fine (1994) argued that such modal definitions are unpromising. It is arguably necessary that, if an object exists, then its singleton set exists as well. Hence, Socrates could not have existed without being a member of {Socrates}. And yet, being a member of {Socrates} is not part of what it is to be Socrates, or of what makes something Socrates. (By contrast, it may very well be part of what it is for something to be {Socrates} to have Socrates as a member.) Fine argues in detail—convincingly in my view—that there is no modal condition that is both necessary and sufficient for it to be essential to a that P. Nevertheless, many philosophers who are friendly toward a hyperintensional notion of essence accept that there are two important connections between essence and necessity. First, while the truth of (11) is not sufficient for the truth of (12), the converse does hold: if it is essential to a that P, then it is necessary that (if a exists, then) P. ( 11) It is necessary that (if a exists, then) P. (12) It is essential to a that P. Second, it seems that modal facts are often explained by facts about essence. If the question is raised why water could not have (existed but) failed to have the chemical structure H2O, it seems natural to reply: because that is part of what it is to be water. Is it possible to use the notion of essence to give an account of what it is for a proposition to be necessary? We may try out a view along the following lines (see Fine 1994; also see Lowe 2006, 2007, 2012; Hale 2013, chapter 11; Teitel 2019; Ditter 2020): (13) For a proposition to be necessary is for it to be true in virtue of the essences of things.

Modality, Worlds, Essence, and Modal Knowledge  25 Further work would be required to turn this into a full-fledged account of necessity. But irrespective of how the details are worked out, there are reasons to doubt that any theory like (13) can give us a correct and complete account of what it is for a proposition to be metaphysically necessary. Such an account should not merely yield the right extension in all possible worlds, but should tell us what makes a proposition necessary. As we saw before, it seems plausible that what makes a proposition necessary has something to do with the fact that it is especially unconditional and secure in some metaphysical sense. (13) does not capture this thought (or at least it does not do so unless it is supplemented with an account according to which the essential truths themselves enjoy a special kind of unconditionality and security). (See Kment 2014, section 7.4.) Moreover, (13) sheds no light on the aforementioned broader family of modal concepts, since it is true only of metaphysical necessity and it is unclear how to generalize it.

A Theory of Necessity I will outline an account of what necessity is (which is developed in detail in Kment 2014) that seems particularly well-suited to explain the data surveyed in the preceding section. Let me begin by describing some of the basic elements of the view. Impossible Worlds. Following Daniel Nolan (1997, 2013) and others, I will employ a theoretical framework that allows for impossible worlds as well as possible ones. One way to bring out the motivation for this approach is to consider counterpossibles, i.e., counterfactual conditionals with metaphysically impossible antecedents. On the standard account of counterfactuals (Stalnaker 1968; Lewis 1973a, 1986), Q □→ P iff at the worlds closest to actuality where Q, P. On the assumption that all worlds are metaphysically possible, this account entails that all counterpossibles are vacuously true (since there are no antecedent-worlds), irrespective of the contents of their antecedents and consequents. But to many philosophers, that seems very implausible. We already encountered one apparent counterexample above: the counterfactual “If whales had not been mammals, they would still have been mammals” sounds false, despite the fact that its antecedent is metaphysically impossible. In many other cases, the fact that the antecedent of a counterfactual is metaphysically impossible does not seem to settle the truthvalue of the conditional. It is metaphysically impossible for me to be the son of Hillary and Bill Clinton. But that leaves it an open question in what ways I would be different if I were their son. Similarly, it is metaphysically impossible for there to be no numbers, but that does not seem to answer the question whether physical events would have unfolded differently if numbers had not existed. Since this problem arises from disallowing worlds

26  Boris Kment where impossible propositions are true, the obvious remedy—suggested and developed by a number of philosophers (Routley 1989; Mares 1997; Nolan 1997; Zalta 1997; Vander Laan 2004; Kment 2006, 2014, chapters 2–5, 7, and 8, in particular chapter 4)—is to lift this restriction. Instead of appealing to possible worlds, we can formulate the account in terms of worlds more generally, including both possible and impossible worlds. Worlds are simply ways for reality to be, and they include both ways reality could have been and ways reality could not have been.8 I think of worlds as collections of structured Russellian propositions (which are constructed from the individuals, properties, etc., that they are about). I argue in Kment (2014, chapter 4) that we can introduce nonmodal notions of logical entailment and logical consistency for Russellian propositions, to be explained in terms of the logical structures of these propositions. To count as a world, a collection of propositions w needs to answer every question, in the sense that (to simplify slightly) either P or the negation of P is logically entailed by w for every Russellian proposition P. Hence: (14) w is a world iff w is a collection of propositions and, for every proposition P, either P or P’s negation is true at w. Moreover: (15) P is true at w iff P is a proposition that is logically entailed (in the non-modal sense) by the propositions in w. Unlike the notion of a possible world, the concept of a world is non-modal. Gradability of modal properties.9 Although it is often assumed that necessity and possibility are all-or-nothing matters, I think that there is some linguistic evidence to the contrary. Just as we can say that suchand-such could have been the case, we can say that this could more easily have been the case than that. On the face of it, that sounds like a comparison of degrees of possibility. I propose that we take this appearance at face value: possibility and necessity come in degrees (see also Lewis 1973b). Proposition P has a higher degree of possibility than proposition Q just in case P could more easily have been true than Q. Similarly, a true proposition P has a higher degree of necessity than another true proposition Q just in case Q could more easily have been false than P. To get a better handle on claims about how easily something could have been the case, it is useful to consider how we ordinarily support such a claim. When talking about a soccer game, we may say: “The game ended in a draw, but our team could easily have won. If the goalkeeper had stood two inches further to the right a minute before the end, the other team would not have scored their goal.” In less favorable circumstances, we may say instead: “Our team couldn’t easily have won. They would

Modality, Worlds, Essence, and Modal Knowledge  27 have beaten their opponents only if Mary had been on the team, Katie had been sober, Bob had known the rules, and so forth.” How easily our team could have won depends on how great a departure from actuality is required for them to win. If they win in some situations that are only minimally different from the way things in fact are, then we can say that they could easily have won, or that their winning had a high degree of possibility. We can say the opposite if all situations where they win depart very significantly from actuality. Similarly, for any true proposition P, how easily P could have failed to be true depends on how great a departure from actuality is required for P not to be true. The greater the departure required, the higher P’s degree of necessity. Talk about degrees of possibility is ubiquitous in ordinary life, but the locutions we use are not always overtly modal. Often (but not invariably), we employ idioms that involve metaphors of proximity, fragility, or security. You are running to catch the train, but the doors close on you before you can jump in, causing you to sigh in frustration “I nearly made it.” Your utterance expresses the thought that you could easily have caught the train: a minimal departure from actuality—the doors closing half a second later—is all that was necessary. We say that something was a close call, that an event almost happened, or that somebody came within a hair’s breadth of disaster, to communicate that various situations could easily have obtained. The peace between two nations during some period in history can be called fragile or secure depending on how easily their tensions could have escalated into war. The theoretical framework of (possible and impossible) worlds can be used to sharpen the account of modality just sketched. Let us say that: (16) w is actualized iff w is a world that entails all and only the true propositions. There is exactly one actualized world (“the actual world” or “actuality”)—a world that provides a wholly accurate and complete description of reality. Other worlds depart from actuality to varying degrees. The degree of possibility of a proposition P is determined by how close the closest P-worlds are to actuality: the closer these worlds, the more easily P could have been true. I will say that: (17) S is a sphere iff S is a class of worlds and every world in S is closer to the world that is actualized than any world not in S.10 The ordering of unactualized worlds by their closeness to actuality generates a system of nested spheres. For each sphere, there is a grade of necessity that attaches to just those propositions that are true at every world in that sphere, as well as a grade of possibility attaching to all propositions that are true at some world in the sphere. The larger the sphere, the

28  Boris Kment greater the associated grade of necessity. One sphere (described in more detail below) is associated with metaphysical modality: a world counts as metaphysically possible just in case it is in that sphere and a proposition is metaphysically necessary just in case it is true at every world in that sphere.11 I call the relations of comparative possibility and necessity and the modal properties that relate to specific grades of possibility and necessity (such as metaphysical necessity) ontic modal properties/relations, to distinguish them from various other kinds of modal properties, like those of epistemic or deontic modality. The words “could” and “must” involve quantification over worlds. In most contexts, the quantification is restricted to worlds that meet specific conditions, and it is well known that the restrictions vary greatly between contexts (Kratzer 1991, 2012). “Could” expresses a grade of ontic possibility in a given context just in case the quantification over worlds is restricted to the worlds that are within a certain sphere around actuality and there are no other restrictions on these quantifiers. To complete the analysis of modality, we need an account of the rules that determine the ordering of worlds by their closeness to actuality. Such an account is developed in detail in Kment (2014, chapters 8 and 9). On this occasion, I will confine myself to sketching its most relevant parts. The weightiest criterion of closeness to actuality is logical consistency, followed by similarity to actuality with regard to the laws.12 Not all similarities regarding the laws carry the same weight, however. We need to distinguish between natural laws and what I call “metaphysical laws” (Kment 2014, chapter 6). The metaphysical laws include a range of metaphysical principles, such as essential truths and certain ontological principles, that play a distinctive explanatory role (which will be described in Section “Metaphysical Laws, Grounding, and Causal Explanation”). Let us say that a world matches actuality with respect to the metaphysical laws just in case it has the same metaphysical laws as actuality and it perfectly conforms to these laws. Match with respect to the metaphysical laws has greater weight in determining the closeness ordering than conformity to the natural laws of actuality. Match in matters of particular fact is also relevant to the closeness ordering but carries less weight than law-related similarities. On this account, the logically consistent worlds are closer to actuality than any other worlds—they form a sphere around actuality. The consistent worlds that match actuality with respect to the metaphysical laws form a second sphere inside the first one, and the consistent worlds that match actuality with respect to the metaphysical laws and also conform to the actual natural laws form a third sphere inside the second. There are many other spheres as well—some larger than the three mentioned, some intermediate between them, and some smaller than the three. Each sphere corresponds to a grade of necessity that attaches to just those propositions that are true at every world in that

Modality, Worlds, Essence, and Modal Knowledge  29 sphere. Metaphysical necessity is the grade corresponding to the sphere of consistent worlds that match actuality with respect to the metaphysical laws. Hence: (18) P is metaphysically possible (necessary) iff P is a proposition that is true at some (all) worlds in the sphere of consistent worlds that match the world that is actualized with respect to the metaphysical laws. This account can easily explain the connections of modality to the other notions that were discussed in Section “Necessity, Possible Worlds, Essence, and Counterfactuals”. Let us consider them, in turn. Possible Worlds. The theory I sketched explains why it seems illuminating to explain the possibility and necessity of individual propositions in terms of possible worlds. Such an explanation highlights an important aspect of ontic modality, namely that it is by its very nature holistic (Kment 2014, Section 2.5; also see Adams 1974, 225): a proposition’s degree of ontic possibility is determined by how great a departure from the complete way things are is required for P to be true, i.e. by how much we need to change in all of reality to allow P to be true. The apparatus of worlds allows us to state this idea in less metaphorical terms. We start with a true and complete description of reality and then consider the changes we need to make to it to turn it into a complete description of reality that entails P. The true description we start from is the actual world. A complete description of reality that entails P is a P-world. P-worlds can be ordered by how great a change to the actual world is required to obtain them from the actual world. The smaller the changes that are required to obtain a P-world, the greater P’s degree of possibility. Counterfactuals. As we saw in Section “Necessity, Possible Worlds, Essence, and Counterfactuals,” there is evidence for thinking that metaphysical necessity belongs to a broader family of modal properties that have two characteristic connections to counterfactuals: they conform to the principle that necessary propositions are counterfactually robust (which is encapsulated in the inference rule (5)) and to the principle that possibility is transmitted across counterfactual implication (enshrined in principle (6)). The account outlined in this section captures this idea. It entails that there is a scale of different grades of necessity that correspond to spheres of different sizes around the actual world and that metaphysical necessity is one of these grades. Moreover, rules (5) and (6) are valid for all of these grades of modality. Let S be any sphere around actuality, and let necessityS and possibilityS be (respectively) the grade of necessity and of possibility corresponding to S. Now, suppose that P is possibleS and Q is necessaryS. Then, P is true at some worlds in S and Q is true at all of these worlds. Moreover, since S is a sphere, the P-worlds in S are closer to actuality than any P-worlds outside of S. Therefore, Q holds

30  Boris Kment at all the closest P-worlds, so that P □→ Q is true. This shows that (5) is valid for possibilityS and necessityS. Next, suppose that P is possibleS and that P □→ Q is true. Then, P is true at some worlds in S. Given that S is a sphere, the P-worlds in S include the closest P-worlds. Moreover, since P □→ Q is true, Q holds at the closest P-worlds. It follows that Q holds at some worlds in S and is therefore possibleS. So, (6) is valid for possibilityS. Essence. As discussed in Section “Necessity, Possible Worlds, Essence, and Counterfactuals,” friends of essence often hold that there appear to be two important connections between metaphysical necessity and essence: (i) essential truths are metaphysically necessary, and (ii) for any essential truth E, the fact that E is metaphysically necessary is explained by the fact that E is an essential truth. The account I proposed explains both of these facts. It immediately entails (i). Moreover, if E is an essential truth, then that explains the fact that E holds at every world in the sphere of worlds that match actuality with respect to the metaphysical laws, which, in turn, explains the fact that P is metaphysically necessary. In the remainder of this paper, I will argue that the theory outlined in this section also allows us to explain how it is possible for us to gain the modal knowledge we seem to have. This will require an account of how we can acquire knowledge about the metaphysical laws. Before offering such an account, some preliminary discussion of the notion of a metaphysical law and of its role in explanation will be required.

Metaphysical Laws, Grounding, and Causal Explanation Since I take the notion of a metaphysical law to be defined by its theoretical role in a specific conception of grounding, I will start by briefly sketching this conception (for a more detailed statement of the account, see Kment 2014, chapter 6, 2015). Grounding is the form of explanation described in statements like the following: What makes 28 a perfect number is the fact that it is a positive integer equal to the sum of its proper positive divisors. This particle is a hydrogen atom because (in virtue of the fact that) it is composed of one proton and one electron in such-and-such configuration. Grounding relationships connect metaphysically non-fundamental facts to the more fundamental facts that give rise to them.13 I think that the metaphysical laws play a role in grounding very similar to that of the natural laws in deterministic causation. Metaphysically more

Modality, Worlds, Essence, and Modal Knowledge  31 f­ undamental facts typically give rise to less fundamental ones in accordance with the metaphysical laws, just as in a deterministic universe, earlier events bring about later ones in accordance with the laws of nature. The metaphysical laws are, as it were, the covering laws of such grounding relationships.14 The metaphysical laws include the essential truths, which relate to the question of what it takes to be a certain entity or to have a certain property.15 For example, the essential truths about a non-fundamental property F lay down necessary and sufficient conditions for having F that are formulated in terms of more fundamental entities. An essential truth that lays down such conditions can be called a “real definition” of F.16 When F’s real definition lays down that meeting condition c is necessary and sufficient for having F, we can express this by saying that to have F is to meet condition c, or that F-ness is the property of meeting condition c. To illustrate, the following might be a real definition of the property of being a hydrogen atom: (19) x is a hydrogen atom iff x is composed of one proton and one electron that stand to each other in the H configuration. The claim that (19) is a real definition of the property of being a hydrogen atom can be stated more briefly by saying that (what it is) to be a hydrogen atom is to be composed of one proton and one electron that stand in the H configuration, or that hydrogen-atom-hood is the property of being so composed (Kment 2014, section 6.1.2).17 We have already encountered a number of real definitions in Section “A Theory of Necessity,” for I think that (14), (15), (16), (17), and (18) are real definitions of worldhood, truth-at, actualization, spherehood, and metaphysical necessity, respectively. For P to be metaphysically necessary is for P to be a proposition that is true at every world in the sphere of consistent worlds that match the world that is actualized with respect to the metaphysical laws. For w to be a world is for w to be a collection of propositions such that, for every proposition P, either P or P’s negation is true at w. And so forth. If the real definition of F lays down that condition c is necessary and sufficient for having F, then facts about which entities have F are at least partly grounded in facts about which entities satisfy c. For example, (20) and (21) are grounded in (22) and (23), respectively. ( 20) Hydra is a hydrogen atom. (21) You are not a hydrogen atom. (22) Hydra is composed of one proton and one electron that stand in the H configuration. (23) You are not composed of one proton and one electron that stand in the H configuration.

32  Boris Kment (19) functions as a covering law for both instances of grounding. The fact that (19) is essential to the property of being a hydrogen atom explains the fact that (22) grounds (20) and the fact that (23) grounds (21). Similarly, the fact that the proposition that 2 + 2 = 4 is metaphysically necessary is grounded in the fact that this proposition is true at all worlds in the sphere of consistent worlds that match the world that is actualized with respect to the metaphysical laws. The fact that the proposition that 2 + 2 = 5 is not metaphysically necessary is grounded in the fact that this proposition is not true at all worlds in that sphere. These grounding relationships are explained by the fact that (18) is essential to metaphysical necessity. (It is a crucial assumption of this account that (20) and (22) state different facts—after all, the fact stated by (22) is supposed to explain the fact stated by (20) and this is not meant to be a case of self-explanation! I think that this assumption is independently plausible. (20) and (22) are not even about the same entities. (20) ascribes to Hydra the property of being a hydrogen atom, and this property is not mentioned, nor ascribed to anything, in (22). By contrast, (22) says that Hydra stands in the composition relation to two other objects that are configured in a certain way and instantiate the properties of protonhood and electronhood, respectively. (20) does not say any of this—it says nothing about composition, protonhood, electronhood, or the H configuration. Mutatis mutandis, the same observations are true of (21) and (23).) In addition to real definitions that lay down necessary and sufficient conditions for having a certain non-fundamental property, there may be others that lay down necessary and sufficient conditions for being a certain non-fundamental entity (specified in terms of more fundamental entities). For example, where “N” is a directly referential name for the singleton of the number 2, the following might be a real definition of this set: x is N iff x is the set whose sole member is 2.18 We can express this by saying that to be N is to be the set that has 2 as its sole member. Many facts about N are grounded in facts that are not about N but are instead partly about the more fundamental entities in terms of which N is defined (2, membership, sethood, etc.). For instance, the fact that N figures in many philosophical examples is grounded in the fact that the set whose sole member is 2 figures in many such examples. The real definition of N stated above covers this instance of grounding. The metaphysical laws may include principles other than essential truths as well, for example, certain ontological laws. These are principles that tell us that under certain conditions, there exists something of a certain kind. Laws of mereological composition are one example. They might include principles to the effect that whenever there are things meeting a certain condition c, there is something composed of exactly these things. Another example is principles of plenitude for properties, like the one stated by schema (24).

Modality, Worlds, Essence, and Modal Knowledge  33 (24) There is a unique property of individuals F such that to have F is to be an individual x such that A(x), or in other words, there is such a thing as the property of being an individual x such that A(x). Like essential truths, ontological laws are instantiated in certain cases of grounding. For example, the fact that (i) there is an object composed of the as might be grounded in the fact that (ii) the as exist and meet condition c. The law of mereological composition is the covering law for this instance of grounding, and the fact that it is a law explains the fact that (ii) grounds (i). Grounding and causal explanation are closely intertwined (Kment 2014, section 1.2.1, 2015, section 1.1; Schaffer 2016). In many cases, X causally explains Z by causally explaining some other fact Y that, in turn, grounds Z. Let me run through a couple of examples. (a) The fact that you rubbed your hands together causally explains the fact that they are hotter than they were previously. It does so by causally explaining the fact that the kinetic energy of your hands’ surface molecules increased, which, in turn, grounds the fact that your hands are hotter than before. The laws that connect the rubbing fact to the increase in temperature include various natural laws that are instantiated in the process by which the rubbing raises the kinetic energy. But they also include the real definition of being-hotter-than: x is hotter at t than y is at t* iff the mean kinetic energy of x’s molecules at t is higher than that of y’s molecules at t*. (b) The fact that you mixed certain ingredients in the right proportions and baked the resulting batter at a specific temperature for a suitable amount of time causally explains the fact that there is a Bundt cake on the kitchen counter. The facts about your actions are connected to the later existence of a Bundt cake by specific laws of nature (for example, those instantiated in the various processes that happen during the baking of the cake) and certain metaphysical laws, including the real definition of being a Bundt cake: something is a Bundt cake iff it is made by such-and-such a process from such-and-such ingredients, combined in such-and-such proportions. (c) Suppose, for the sake of illustration, that the physical facts ground the mental facts by giving rise to them in accordance with the laws of metaphysics. You take a sip of coffee, which brings it about that such-and-such part of your brain is in a certain state, which (possibly in combination with certain background facts) grounds the fact that you have a specific taste sensation. The fact that you sipped your coffee causally explains the fact that you have the sensation via an explanatory chain that features both links of causal explanation (for example, between the sipping fact and the facts about your brain state) and grounding links (for instance, between brain-state facts and sensation facts). Moreover, the laws involved include both natural laws (such as those instantiated by the relevant brain processes) and metaphysical laws (including those in accordance with which the facts about your brain state ground the facts about the sensation).

34  Boris Kment These examples illustrate how both natural and metaphysical laws can be instantiated in the chain of explanatory relationships that connect a fact f to the facts that causally explain f. Elsewhere, I have argued (Kment 2014, sections 1.2.1 and 6.4, 2015, section 1.1) that that is a pervasive phenomenon. Only a small set of properties and relations—the “physically fundamental” ones—are mentioned in the (fundamental) natural laws. If a fact f involves physically non-fundamental properties or relations and f is causally explained by other facts, then both natural and metaphysical laws figure as covering laws in the chain of explanatory connections that lead from f’s causal explainers to f. The natural laws alone do not suffice to forge the required connections.

Modal Knowledge Knowledge of Modal Facts and Knowledge about the Metaphysical Laws One test for a metaphysical account of modality is whether it enables us to explain how it is possible for us to have the modal knowledge we do. In the remainder of this paper, I will aim to show that the account outlined in Section “A Theory of Necessity” passes this test. Note that the account entails the following principle: (25) For every proposition P, P is metaphysically necessary (possible) iff the metaphysical laws and the truths about which propositions are metaphysical laws jointly logically entail (are jointly logically consistent with) P. Moreover, given basic logical competence, modal knowledge can be gained from ( i) knowledge that (25) is true, and (ii) knowledge about metaphysical laws. For example, if you know that (25) holds and that There are flying pigs is logically consistent with all metaphysical laws and truths about which propositions are metaphysical laws, you can conclude that there could have been flying pigs. Similarly, suppose you know both that (25) is true and that it is essential to gold that something is gold iff it is the element with atomic number 79. You can then conclude that it is necessary that, if gold exists, then gold is the element with atomic number 79. We can therefore explain the possibility of modal knowledge consistently with the account of Section “A Theory of Necessity” if we can explain the possibility of the two kinds of knowledge specified in (i) and (ii). Moreover, since (25) can be known on the basis of knowledge of real definitions such

Modality, Worlds, Essence, and Modal Knowledge  35 as (14)–(18), and since real definitions are metaphysical laws, an explanation of how knowledge about the metaphysical laws is possible also yields an explanation of how we can know that (25) is true. The remaining task, therefore, is to explain how we can gain knowledge about the metaphysical laws. Let us turn to this question next. Knowledge of the Metaphysical Laws I will discuss two methods by which we may be able to gain knowledge about the metaphysical laws.19 I do not claim that they are the only ones. The first method proceeds by inference to the best explanation (IBE). As discussed in Section “Metaphysical Laws, Grounding, and Causal Explanation”, metaphysical laws figure as covering laws in the causal and grounding explanations of non-fundamental facts. We can apply our standards of theory evaluation to find the best account of how various non-fundamental facts are grounded, or the best account of what causally explains these facts (an account of their grounds will be part of a complete account of what causally explains them, as discussed in Section “Metaphysical Laws, Grounding, and Causal Explanation”). When made fully explicit in the right way, the complete version of such an account will involve certain assumptions about the metaphysical laws that cover the grounding relationships involved, and the abductive inference that establishes the account will support these assumptions about the metaphysical laws as well. For example, according to the best account of what grounds facts about temperature, these facts are grounded in facts about molecular kinetic energy. What makes it so that the sun is hotter than the earth, for instance, is that the sun’s mean molecular kinetic energy is higher than the earth’s. This account requires the assumption that there is a certain metaphysical covering law that is instantiated in the instance of grounding described. The relevant law is the real definition of being-hotter-than stated above. When we support our theory of what grounds facts about temperature by an IBE, we support the hypothesis about the real definition of being-­ hotter-than at the same time. The theory of what grounds facts about temperature is also part of a fully explicit version of the best account of how facts about temperature are causally explained. Suppose that you have just rubbed your hands and that they are now hotter than they were before. According to the best account of what causally explains the fact that your hands heated up, the fact that you rubbed your hands led to an increase in the mean kinetic energy of the molecules that make up your hands, and the fact that their kinetic energy increased grounds the fact that your hands are hotter than before. This account requires the aforementioned hypothesis about what the real definition of being-hotter-than is, and the abductive inference that establishes the account also confirms the hypothesis.

36  Boris Kment In the examples just considered, the grounding relationships require an essential truth as a covering law. In other cases, an ontological law plays this role. For example, on the best account of facts about macroscopic objects, they are grounded in facts about the arrangement of particles. The metaphysical laws covering these instances of grounding include certain laws of composition. In some cases, the IBE may start from truths that are known a priori and may not require recourse to sense experience. For example, mathematicians had a good grasp of the notion of the limit value of a function long before they knew a precise definition. Eventually, the Bolzano-Weierstrass definition gained widespread acceptance. Perhaps one way to rationalize that acceptance is to regard it as the result of an abductive inference. Mathematicians knew many facts of the form The limit of f(x), as x approaches a, is y. According to the best account of what grounds these facts, the fact that the limit of f(x), as x approaches a, equals y is grounded in the fact that, for every real number ε > 0, there is a real number δ such that, for all x∈(a − δ, a + δ), |f(x) − y| < ε. This claim about what grounds the former fact requires the assumption that the Bolzano-Weierstrass definition specifies the real definition of the relation of being-a-limit-of, and an abductive inference that supports the grounding claim also supports the claim about the real definition of the relation. The conclusion of an abductive inference will often give us no more than a partial account of what explains the explanandum. It might be that all it tells us is that the factors that explain the explanandum are of suchand-such general kind. Picking a fruit from the apple tree in your garden, I notice that it is covered with ugly splotches and smells bad. I look at apples from other nearby trees and the same is true of them. Given this and similar evidence (and a modest amount of background knowledge), I can conclude that the trees in your garden suffer from some disease. This is an IBE—I conclude that the best account of what is responsible for the appearance of your apples involves the assumption that your trees are afflicted by disease. This conclusion falls far short of a complete account of what explains my data—I do not know what disease your trees have, or why that disease causes the symptoms I have observed. Similarly, an IBE may establish some truths about the essence of X without giving us a complete account of its essence. Suppose that I am able to recognize a Bundt cake by its look, taste, and smell, but I do not know what it is to be a Bundt cake. One day, I watch you make a dish that I recognize to be a Bundt cake. I know that the best account of what causally explains the presence of a Bundt cake on the kitchen counter appeals to specific assumptions about what it is to be a Bundt cake, and I can consequently support certain assumptions about the essence of Bundtcake-hood abductively. But my evidence may not be sufficient to arrive at a complete account of what it is to a Bundt cake. Perhaps I have watched you from a distance and noticed that you mixed flour, milk, and eggs, and then added some other ingredients. But I couldn’t quite tell what those

Modality, Worlds, Essence, and Modal Knowledge  37 additional ingredients were, nor do I know precisely in what proportions the ingredients were added to the batter or how a cake was made from the batter. In that case, all that I can conclude from my observations is that being a Bundt cake involves being made from ingredients that include flour, milk, and eggs. But that falls short of complete knowledge of the essence of Bundt-cake-hood. The second way of gaining knowledge about the metaphysical laws relies on our conceptual or linguistic competence. It seems plausible that at least in some cases, competence with a concept or linguistic term E constitutively requires (at least implicit) knowledge of a specific condition that an entity needs to satisfy to be picked out or expressed by E. In some cases, this condition might be both necessary and sufficient for an entity to be the one picked out or expressed by E, while in other cases the condition might be merely necessary. Sometimes, the condition might be that of having a certain (kind of) essence. The notion of being a vixen might illustrate this. Full mastery of this concept arguably constitutively requires knowing (at least implicitly) that it picks out a given property F only if F’s extension includes exactly the female foxes. Moreover, it arguably also constitutively requires knowing the stronger claim that the concept picks out a property F iff to have F is to be female and a fox, i.e., iff F is the property of being female and a fox. To see this, suppose that Fred knows that all and only female foxes are vixens. However, he is not sure whether being female and a fox is what makes something a vixen, or whether what makes something a vixen is the fact that it satisfies some other condition that merely happens, as a matter of accidental fact, to be met by all and only female foxes. (Perhaps he is wondering why it is the case that no non-foxes and no non-female foxes have ever been found to be vixens.) It seems plausible that Fred cannot be credited with a full grasp of the concept of being a vixen. Contrast this example with the case of the concept of water. A specific person’s mastery of that concept might be partly constituted by the person’s knowledge that the concept picks out x only if x is some stuff that plays a certain role R in our lives. However, it is not partly constituted by the speaker’s knowledge that the concept picks out x only if (or if and only if) being a stuff that plays role R is (part of) what it is to be x. The speaker does not know the latter proposition, since that proposition is false. (Playing role R is not part of what it is to be water.) Assume that the account of the possession conditions of the concept of being a vixen that I sketched in the preceding paragraph is correct. Then, competence with the notion of being a vixen might put a speaker in a position to know the following: (26) If the property of being a vixen exists, then being a vixen is the property of being female and a fox. (27) If the property of being female and a fox exists, then it is the property of being a vixen.

38  Boris Kment Now suppose that the speaker also knows that femaleness and foxhood exist and that she knows, perhaps on the basis of IBE, that (24) is valid. Then, she can infer that there is such a thing as the property of being female and a fox. Given her knowledge of (27), she can conclude that this property is the property of being a vixen, and hence that to be a vixen is to be a female fox. Even when competence with a concept does not constitutively require knowing any proposition of the form Concept C picks out F only if F’s real definition is P,20 it might still require knowing (at least implicitly) some proposition of the form C picks out F only if the essential truths about F meet such-and-such conditions. For example, it might require knowing that a property F is picked out by C only if the following holds: it is essential to F that an entity x has F only if x meets condition c. (The proposition that F meets this condition does not entail that it is also essential to F that an entity x has F if x meets condition c; i.e., it does not entail that it is essential to F that c is sufficient for having F.) To take a concrete example, competence with the concept of propositional knowledge might require knowing (at least implicitly) that a relation R is picked out by this concept only if R meets the following condition: it is essential to R that x stands in R to a proposition P only if P is true and x believes P. (This constrains, but does not determine, which relation the concept picks out.) Someone who has mastered the concept of propositional knowledge might then be in a position to know that, if the relation of knowledge exists, then it is essential to knowledge that it relates a thinker to a proposition only if the proposition is true and the thinker believes the proposition. (While these conditions might be necessary for knowledge, they are not sufficient.)

Notes 1 Lewis (1986, chapter 1). 2 Not everyone agrees that Lewis’s account is truly reductive. See Lycan (1988), Shalkowski (1994), and Divers and Melia (2002). See Cameron (2012) for an overview of and contribution to this debate. 3 See Lewis’s discussion of the objection from actualism in Lewis (1986, section 2.1). 4 Examples of such views include the accounts in Plantinga (1974, 1976), Adams (1981), Stalnaker (1968, 1996, 2011), and Kment (2014, chapters 4 and  5), among many others. 5 Some philosophers doubt even the qualified principle (Nolan 1997; Vander Laan 2004; Bernstein 2016). For defenses, see Mares (1997) and Jago (2014). Also see Berto et al. (2018). 6 Suppose that ◊P and P □→ Q hold and assume for reductio that ◊Q is false. Then □¬Q is true. By (5), we can infer P □→ ¬Q. Hence, (P □→ Q) & (P □→ ¬Q) is true. But from ◊P we can infer ¬((P □→ Q) & (P □→ ¬Q)) by (7). 7 For other accounts that place emphasis on counterfactuals in thought experiments that are intended to establish modal conclusions, see Jackson (1998, 2010), Chalmers (2002), and Williamson (2007). Also see Lange (2009). 8 For more on impossible worlds, see Nolan (1997, 2013). See Stalnaker (1996) for some arguments against impossible worlds.

Modality, Worlds, Essence, and Modal Knowledge  39 9 See also Lewis (1973b), Kratzer (1991, 2012), and Lange (1999). 10 Note that the description “the world that is actualized” is a non-rigid designator. Roughly speaking, when (17) is evaluated at a possible world w, the phrase picks out w. (For complications, see Kment 2014, sections 4.7−4.8.) So understood, (17) is true at every possible world. The same observation applies to principle (18). 11 The graded form of modality I described is similar to what Kratzer (1991, sections 3.3 and 5) calls “circumstantial modality,” which she defines in terms of a closeness ordering. However, there is an important difference in the order of explanation and a closely related further difference between the non-reductive character of her account and the reductive nature of my own. I start with a non-modal concept of a world and a notion of comparative closeness between worlds that is defined non-modally, and I use them to give a non-modal definition of a graded notion of possibility. The graded notion, in turn, is used to define metaphysical possibility (as well as other specific grades of possibility, such as nomic possibility): the metaphysically possible propositions are those that have at least a certain degree of possibility, i.e., those that are true at some world within a certain distance from actuality. By contrast, Kratzer starts with an unanalyzed modal notion of a possible world, and then defines the closeness ordering and the graded notion of possibility in terms of it. Her view is non-reductive. 12 See, e.g., Lewis (1986) and Bennett (1984, 2001, 2003), among many others. 13 For recent discussions of grounding and metaphysical explanation, see Schaffer (2009, 2016), Rosen (2010), Jenkins (2011), Bennett (2011, 2017), Koslicki (2012), Audi (2012a,b), Fine (2012a, b), and Kment (2014, Chapter 6, 2015, 2021). For some skeptical voices, see Hofweber (2009), Sider (2011), chapter 8; Daly (2012), and Wilson (2014, 2018). 14 Versions of the covering-law conception of grounding and metaphysical explanation are proposed and defended in Kment (2014, chapter 6, 2015), Wilsch (2015a,b), Glazier (2016), Rosen (2017), Schaffer (2017). 15 See Fine (1994, 1995). 16 For a slightly different conception of real definition, see Rosen (2015, 2017). 17 For other accounts of idioms like “To be F is to be G,” see Rayo (2013) and Dorr (2016). 18 A real definition of an individual is a singular proposition about that individual, and stating such a real definition requires a directly referential name. The most common expressions for {2}, such as “the singleton of two” and “{2},” are not obviously directly referential, which is why I needed to introduce the term “N.” 19 Both methods are discussed in somewhat more detail in Kment (2021). For other discussions of (the possibility of) knowledge of essences, see Lowe 2012; Hale 2013, chapter 11; Tahko 2018; Mallozzi 2021. 20 See Putnam (1975) and Johnston and Leslie (2012) for arguments for the claim that competence with most expressions does not require knowledge of necessary and sufficient application conditions like the ones my present discussion is concerned with.

References Adams, Robert. 1981. “Actualism and Thisness.” Synthese 49: 3–41. Audi, Paul. 2012a. “A Clarification and Defence of the Notion of Grounding.” In Metaphysical Grounding: Understanding the Structure of Reality, edited by Fabrice Correia and Benjamin Schnieder, 101–121. Cambridge: Cambridge University Press.

40  Boris Kment Audi, Paul. 2012b. “Grounding: Toward a Theory of the in-Virtue-of Relation.” The Journal of Philosophy 109: 685–711. Bennett, Jonathan. 1984. “Counterfactuals and Temporal Direction.” ­Philosophical Review 93: 57–91. Bennett, Jonathan. 2001. “On Forward and Backward Counterfactual ­Conditionals.” In Reality and Humean Supervenience, edited by Gerhard Preyer and Frank Siebelt, 177–202. Lanham, MD: Rowman and Littlefield. Bennett, Jonathan. 2003. A Philosophical Guide to Conditionals. Oxford: ­Clarendon Press. Bennett, Karen. 2011. “Construction Area (No Hard Hat Required).” Philosophical Studies 154: 79–104.S Bennett, Karen. 2017. Making Things Up. Oxford: Oxford University Press. Bernstein, Sara. 2016. “Omission Impossible.” Philosophical Studies 173: 2575–2589. Berto, Francesco, French, Rohan, Priest, Graham and Ripley, David. 2018. “Williamson on Counterpossibles.” Journal of Philosophical Logic 47: 693–713. Brogaard, Berit and Salerno, Joe. 2007a. “Why Counterpossibles are Non-­ Trivial.” The Reasoner 1(1): 5–6. Brogaard, Berit and Salerno, Joe. 2007b. “Williamson on Counterpossibles.” The Reasoner 1(3): 5–6. Brogaard, Berit and Salerno, Joe. 2013. “Remarks on Counterpossibles.” ­Synthese 190: 639–660. Cameron, Ross. 2012. “Why Lewis’s Analysis of Modality Succeeds in its ­Reductive Ambitions.” Philosophers’ Imprint 12: 1–21. Chalmers, David. 2002. “Does Conceivability Entail Possibility?” In Conceivability and Possibility, edited by Tamar Szabo Gendler and John Hawthorne, 145–200. Oxford: Oxford University Press. Daly, Chris. 2012. “Skepticism about Grounding.” In Metaphysical Grounding: Understanding the Structure of Reality, edited by Fabrice Correia and Benjamin Schnieder, 81–100. Cambridge: Cambridge University Press. Ditter, Andreas. 2020. “The Reduction of Necessity to Essence.” Mind 129: 351−380. Divers, John and Melia, Joseph. 2002. “The Analytic Limit of Genuine Modal Realism.” Mind 111: 15–36. Dorr, Cian. 2016. “To be F is to be G.” Philosophical Perspectives 30: 39–134. Fine, Kit. 1994. “Essence and Modality.” Philosophical Perspectives 8: 1–16. Fine, Kit. 1995. “The Logic of Essence.” Journal of Philosophical Logic 24: 241–273. Fine, Kit. 2012a. “Guide to Ground.” In Metaphysical Grounding: Understanding the Structure of Reality, edited by Fabrice Correia and Benjamin Schnieder, 37–80. Cambridge: Cambridge University Press. Fine, Kit. 2012b. “The Pure Logic of Ground.” Review of Symbolic Logic 5: 1–25. Glazier, Martin. 2016. “Laws and the Completeness of the Fundamental.” In Reality Making, edited by Mark. Jago, 11–37. Oxford: Oxford University Press. Hale, Bob. 2013. Necessary Beings: An Essay on Ontology, Modality, and the Relations between Them. Oxford: Oxford University Press. Hill, Christopher. 2006. “Modality, Modal Epistemology, and the ­Metaphysics of Consciousness,” In The Architecture of the Imagination: New Essays on Pretense, Possibility, and Fiction, edited by Shaun Nichols, 205–235. Oxford: Oxford University Press. Hofweber, Thomas. 2009. “Ambitious, Yet Modest, Metaphysics.” In ­Metametaphysics: New Essays on the Foundations of Ontology, edited by David

Modality, Worlds, Essence, and Modal Knowledge  41 Chalmers, David Manley, and Ryan Wasserman, 260–289. Oxford: Clarendon Press. Jackson, Frank. 1998. From Metaphysics to Ethics: A Defence of Conceptual Analysis. Oxford: Oxford University Press. Jackson, Frank. 2010. Language, Names and Information. Walden, MA: Wiley-Blackwell. Jago, Mark. 2014. The Impossible: An Essay on Hyperintensionality. Oxford: Oxford University Press. Jenkins, Carrie. 2011. “Is Metaphysical Dependence Irreflexive?” The Monist 94: 267–276. Johnston, Mark, and Leslie, Sarah-Jane. 2012. “Concepts, Analysis, Generics and the Canberra Plan.” Philosophical Perspectives 26: 113–171. Kment, Boris. 2006. “Counterfactuals and the Analysis of Necessity.” Philosophical Perspectives 20: 237−302. Kment, Boris. 2014. Modality and Explanatory Reasoning. Oxford: Oxford ­University Press. Kment, Boris. 2015. “Modality, Metaphysics, and Method.” In The Palgrave Handbook of Philosophical Methods, edited by Chris Daly, 179–207. Basingstoke: Palgrave Macmillan. Kment, Boris. 2021. “Essence and Modal Knowledge.” Synthese 198: 1957−1979. Kment, Boris. Forthcoming. “Decision, Causality, and Predetermination.” ­Philosophy and Phenomenological Research. Koslicki, Kathrin. 2012. “Essence, Necessity and Explanation.” In Contemporary Aristotelian Metaphysics, edited by Tuomas Tahko, 187–206. Cambridge: Cambridge University Press. Kratzer, Angelika. 1991. “Modality.” In Semantics: An International Handbook of Contemporary Research, edited by Arnim von Stechow and Dieter Wunderlich, 639–650. Berlin: Walter de Gruyter. Kratzer, Angelika. 2012. Modals and Conditionals: New and Revised Perspectives. Oxford: Oxford University Press. Kripke, Saul. 1981. Naming and Necessity. Oxford: Blackwell. Lange, Marc. 1999. “Laws, Counterfactuals, Stability, and Degrees of ­Lawhood.” Philosophy of Science 66: 243–267. Lange, Marc. 2000. Natural Laws in Scientific Practice, New York: Oxford University Press. Lange, Marc. 2005. “A Counterfactual Analysis of the Concepts of Logical Truth and Necessity,” Philosophical Studies 125: 277–303. Lange, Marc. 2009. Laws and Lawmakers. Oxford: Oxford University Press. Lewis, David. 1973a. Counterfactuals. Cambridge, MA: Harvard University Press. Lewis, David. 1973. “Counterfactuals and Comparative Possibility.” Journal of Philosophical Logic 2: 418–446. Lewis, David. 1986, “Counterfactual Dependence and Time’s Arrow,” In Philosophical Papers, vol. 2, edited by David Lewis, 32–66. New York and Oxford: Oxford University Press. Lowe, Jonathan. 2006. The Four-Category Ontology. Oxford: Clarendon Press. Lowe, Jonathan. 2007. “A Problem for A Posteriori Essentialism Concerning Natural Kinds.” Analysis 67: 286–292. Lowe, Jonathan. 2012. “What is the Source of Our Knowledge of Modal Truths?” Mind 121: 919–950.

42  Boris Kment Lycan William G. 1988. “Review of on the Plurality of Worlds.” Journal of Philosophy 85: 42−47. Mallozzi, Antonella. 2021. “Putting Modal Metaphysics First.” Synthese 198: 1937–1956. Mares, Edwin. 1997. “Who’s Afraid of Impossible Worlds?” Notre Dame Journal of Formal Logic 38: 516–526. McFetridge, Ian. 1990. “Logical Necessity: Some Issues.” In Logical Necessity, and Other Essays, edited by John Haldane and Roger Scruton, 135−154. ­London: Aristotelian Society. Nolan, Daniel. 1997. “Impossible Worlds: A Modest Approach,” Notre Dame Journal of Formal Logic 38: 535–573. Nolan, Daniel. 2013. “Impossible Worlds.” Philosophy Compass 8: 360−372. Plantinga, Alvin. 1974. The Nature of Necessity. Oxford: Clarendon Press. Plantinga, Alvin. 1976. “Actualism and Possible Worlds.” Theoria 42: 139–160. Putnam, Hilary. 1975. “The Meaning of ‘Meaning’.” In Mind, Language, and Reality. Philosophical Papers, vol. 2, edited by Hilary Putnam, 215−271. Cambridge: Cambridge University Press. Rayo, Augustin. 2013. The Construction of Logical Space. Oxford: Oxford ­University Press. Rosen, Gideon. 2010. “Metaphysical Dependence: Grounding and Reduction.” In Modality: Metaphysics, Logic, and Epistemology, edited by Bob Hale and Aviv Hoffmann: 109–136. Oxford: Oxford University Press. Rosen, Gideon. 2015. “Real Definition.” Analytic Philosophy 56: 189–209. Rosen, Gideon. 2017. “Ground by Law.” Philosophical Issues 27: 279–301. Routley, Richard. 1989. “Philosophical and Linguistic Inroads: Multiply ­Intensional Relevant Logics.” In Directions in Relevant Logic, edited by Jean Norman and Richard Sylvan, 269–304. Dordrecht: Kluwer. Schaffer, Jonathan. 2009. “On What Grounds What.” In Metametaphysics, edited by David Chalmers, David Manley, and Ryan Wasserman, 347–383. Oxford: Oxford University Press. Schaffer, Jonathan. 2016. “Grounding in the Image of Causation.” Philosophical Studies 173: 49–100. Schaffer, Jonathan. 2017. “Laws for Metaphysical Explanation.” Philosophical Issues 27: 302–321. Shalkowski, Scott A. 1994. “The Ontological Ground of the Alethic Modality.” Philosophical Review 103: 669–688. Sider, Theodore. 2011. Writing the Book of the World. Oxford: Oxford University Press. Stalnaker, Robert. 1968. “A Theory of Conditionals.” In Studies in Logical Theory, edited by Nicholas Rescher, 98–112. Oxford: Blackwell. Stalnaker, Robert. 1996. “Impossibilities.” Philosophical Topics 24: 193–204. Stalnaker, Robert. 2011. Mere Possibilities: Metaphysical Foundations of Modal Semantics. Princeton: Princeton University Press. Tahko, Tuomas. 2018. “The Epistemology of Essence.” In Ontology, Modality, Mind: Themes from the Metaphysics of E. J. Lowe, edited by Alexander Carruth, Sophie Gibb, and John Heil, 93–110. Oxford: Oxford University Press. Teitel, Trevor. 2019. “Contingent Existence and the Reduction of Modality to Essence.” Mind 128: 39–68.

Modality, Worlds, Essence, and Modal Knowledge  43 Vander Laan, David. 2004. “Counterpossibles and Similarity.” In Lewisian Themes: The Philosophy of David K. Lewis, edited by Frank Jackson and Graham Priest, 258–275. Oxford: Oxford University Press. Williamson, Timothy. 2005. “Armchair Philosophy, Metaphysical Modality and Counterfactual Thinking.” Proceedings of the Aristotelian Society 105: 1–23. Williamson, Timothy. 2007. The Philosophy of Philosophy. Malden, MA: Wiley-Blackwell. Williamson, Timothy. 2017. “Counterpossibles in Semantics and Metaphysics.” Argumenta 2.2: 195−226. Williamson, Timothy. 2018. “Counterpossibles.” Topoi 37: 357–68. Williamson, Timothy. 2020. Suppose and Tell: The Semantics and Heuristics of Conditionals. Oxford: Oxford University Press. Wilsch, Tobias. 2015a. “The Nomological Account of Ground.” Philosophical Studies 172: 3293–3312. Wilsch, Tobias. 2015b. “The Deductive-Nomological Account of Metaphysical Explanation.” Australasian Journal of Philosophy 94: 1–23. Wilson, Jessica. 2014. “No Work for a Theory of Grounding.” Inquiry 57: 535–579. Wilson, Jessica. 2018. “Grounding-Based Formulations of Physicalism.” Topoi 37: 495–512. Zalta, Edward. 1997. “A Classically-Based Theory of Impossible Worlds.” Notre Dame Journal of Formal Logic 38: 640–660.

2 An Agency-Based Epistemology of Modality Barbara Vetter

Introduction My aim in this paper is to sketch, with a broad brush and in bare o ­ utlines, an approach to modal epistemology that is characterized by three distinctive features. First, the approach is agency-based: it locates the roots of our modal thought and knowledge in our experience of our own agency. “Modality in Action” will argue that simply by being agents in the world, we possess a wealth of modal knowledge. More specifically, we possess knowledge about our own abilities and dispositions, as well as the dispositions and affordances of objects around us. Second, the approach is ambitious in that it takes the experience of certain modal properties in agency to be the sole distinctive feature of specifically modal thought and knowledge; everything that we know about modality beyond the experience of agency is a matter of applying standard methods of inquiry such as deduction, induction, and abductive methods for choosing between theories.1 “Modality Beyond Action” will sketch how this works for our ordinary modal thought. “Modality in Philosophy” will turn to modal thought and knowledge in philosophical contexts. A third feature of the account arises naturally out of the first two. Given that modal thought and knowledge start with modal properties, and that our methods for expanding modal knowledge beyond the context of agency do not make a distinctive addition to it, it is natural that modal thought and knowledge in general are, first and foremost, about modal properties that are sufficiently like those encountered in agency. I call such modal properties “potentialities” and have provided a ­non-reductive metaphysics of them elsewhere (Vetter 2015), but for present purposes, we can think of them as de re possibilities. A complete version of the account will have to specify how from there we get to knowledge of de dicto modal truths; in this paper, I will only be able to sketch an epistemology for de re modality, but unlike Roca-Royes (2017), I am optimistic that it can be extended. What are the attractions of such an account? I take it that if the basic claim is correct and we do have a route to modal thought and knowledge

DOI: 10.4324/9781003002192-4

An Agency-Based Epistemology of Modality  45 in our experience of agency, then this constitutes a natural starting point for an empiricist account of modal thought and knowledge. Why make the account ambitious and not allow for other distinctive sources of our modal thought and knowledge? Simply for reasons of theoretical parsimony: if we can do without other sources, then we should. But the present paper should be read as an exploration, rather than as a defense, of this ambitious approach. If it fails, then we will still have learned something about the role that the experience of agency can, and the roles that it can’t, play in modal epistemology. The remainder of this introductory section aims to locate the account to be explored in the landscape of modal epistemology more generally. Modal epistemologies can be categorized along a number of dimensions; I will choose those which I believe best help to situate the present account. First, exceptionalism vs anti-exceptionalism. An exceptionalist modal epistemology claims that our knowledge of metaphysical modality has an exceptional source, a source whose specific function is insight into questions of metaphysical necessity and possibility. An a­ nti-exceptionalist claims the opposite: our modal knowledge arises, in some way or another, from cognitive abilities that we use for other more mundane purposes, such as planning our actions. Following Williamson (2007), anti-exceptionalists will typically claim that an exceptionalist theory would have to endow us with cognitive faculties for which there is no plausible evolutionary explanation. I will here simply assume that an anti-exceptionalist account is to be preferred if we can get one. In fact, it is not clear that anyone in the literature is a self-described exceptionalist; exceptionalism is generally attributed to a view by its opponents. But an account can be more or less obviously anti-exceptionalist. The view I will explore is very obviously an anti-exceptionalist one—what could be less exceptional than our experience as agents in the world? Second, monism vs dualism. This distinction concerns the relation between metaphysical and some kind of epistemic modality, in particular, conceivability.2 Monists believe that there is but one phenomenon, modality, of which metaphysical and epistemic modalities are merely aspects; there is only one “space of possible worlds”. This is particularly vivid in two-dimensional semantics, on which an epistemic kind of possibility (“primary possibility”) and metaphysical possibility (“secondary possibility”) are characterized by two perspectives on the one modal realm—by whether a given possible world is considered as actual or counterfactual (see Chalmers 2010, 2012). Dualists hold that there is a deep distinction between any epistemic modality on the one hand, and metaphysical modality on the other: the one concerns the extent of our knowledge; the other concerns objective features of reality itself. Dualists will, accordingly, develop their accounts of metaphysical modality with little or no regard to such epistemically modal phenomena.

46  Barbara Vetter In principle, these alternatives can be freely combined. In practice, those who put much weight on their modal anti-exceptionalism are typically dualists (Williamson 2007, 2016b), but an anti-exceptionalist can be a monist as well. The clearest statement that I know of an anti-exceptionalist monism is Ichikawa (2016). Ichikawa’s argument for monism starts from modal semantics, where objective modality (of which metaphysical modality is a species) and epistemic modality are typically considered merely different sets of restrictions on one space of possible worlds. For reasons given in Viebahn and Vetter (2016), I doubt that this is the best account of our modal language: modals, I believe, are polysemous between epistemic and objective readings. Thus, I take modal dualism to be encoded in modal semantics already (which is not, of course, proof of its truth, but merely a rejoinder to semantic arguments for its falsity). Hence, my account will follow the more widespread combination of anti-exceptionalism and modal dualism. A third distinction is between modal empiricism and modal rationalism: while empiricism takes our modal knowledge to have its source in our empirical knowledge of the world (using perception, induction,and so forth), modal rationalism takes it to be non-empirical—be it a priori insight, conceptual knowledge, or linguistic understanding. The distinction is not exclusive: there are a variety of mixed views, which hold that our modal knowledge has both empiricist and rationalist sources. Full-blown modal empiricists are, of course, committed to anti-exceptionalism (what could be a less exceptional source of knowledge than empirical methods?), but not vice versa; an anti-exceptionalist rationalist or defender of a mixed view might, for instance, locate the source of some or all of our modal knowledge in our ordinary understanding of language (Thomasson 2018). It goes almost without saying that my proposed account will be empiricist, but as we shall see below, it leaves room for the possibility of innate modal knowledge (albeit not of the kind that the rationalist typically appeals to). Fourth, a modal epistemology may be symmetric, giving equal weight to our knowledge of different modalities (necessity, possibility, and perhaps the counterfactual conditional) or asymmetric, starting out with one kind of modality and treating our knowledge of others as derivative (Hale 2003). In the latter case, depending on which kind of modality it starts with, an account may be possibility-based, necessity-based, or counterfactual-based. The most prominent account which, like mine, combines ­anti-exceptionalism, empiricism, and modal dualism is no doubt the counterfactual-based account provided by Williamson (2007). My account differs from his in being possibility-based. Like Roca-Royes (2017), I take de re possibility to be a more natural starting point for thought about metaphysical modality (see also Vetter 2016). In the recent literature, a distinction is also sometimes drawn between uniformism and non-uniformism (Vaidya 2015; Wirling 2020), where the

An Agency-Based Epistemology of Modality  47 former holds and the latter denies that “there is only one single route to modal knowledge at the most fundamental level of explanation” (Vaidya 2015). In identifying the source of modal knowledge in agency alone, my account would seem to be a uniform one. But I am not committed to our using a single method or faculty in acquiring that modal knowledge; as we shall see, perception as well as introspective phenomenology may play a role, and some of the modal knowledge I argue for in “Preconditions for Action” might, in principle, even be innate. Hence my preference for the label “ambitious” over “uniform”. Having situated and motivated the view, let me now start to develop it. The view, as I have said in the beginning, will be painted with a broad brush; a detailed development would require more space than I have here (but I hope to give it in future work). I will not here argue that the view is superior to all or indeed any of its competitors. My aim is to lay it out as a serious albeit underdeveloped option in the epistemology of modality.

Modality in Action Preconditions for Action When we perform an action—such as raising a hand, moving a cup on the table, or reading a paper—there are typically both internal and external preconditions for performing that action. On the internal side (internal, that is, to the agent), there is the ability to perform the action in question: to raise one’s hand, move a cup, or read a paper. I am using the term “ability” here in the sense of a stable, intrinsic property of the agent, what is sometimes called a “general ability” (Maier 2010) or a “narrow ability” (Vihvelin 2013, 11). On the external side—intuitively, outside the agent, in their environment—there are (inter alia) the corresponding properties of the objects on which the agent acts, or with which the agent interacts: the cup’s being such that it can be moved, the paper’s being such that it can be read. Those latter properties are often discussed under the label affordances (Gibson 1986): the cup affords moving; the paper affords reading. Affordances are generally taken to be relative to an agent and her abilities. Thus, the cup is movable for me but not for a newborn; this paper is readable for you but not for someone not fluent in English. Affordances will generally have a basis both in the intrinsic properties of the object that has them (the cup’s size, shape, and weight) and in the relation in which it stands to an agent (the size relative to the grip of my hand, the weight relative to my bodily strength, the distance from me and its relation to my arm length). I will use the term “affordances” for such properties, but without thereby committing either to a full-blown metaphysics of them or to the role they play in so-called ecological psychology, following Gibson. Affordances are simply part of what constitutes an opportunity for an agent to act.

48  Barbara Vetter In this section, I want to argue that not only are there such abilities and affordances when we act, but qua agents we have rather basic knowledge of them. (See Strohminger 2015 for a similar point.) Note that I do not here aim to answer the question how we come to have such knowledge; some suggestions on answering that question will be made in “The phenomenology of action”, but I hope to have more to say on this in future work. Suppose that I want to drink some tea, and I know that the mug in front of me contains tea. It is the easiest thing in the world for me to extend my hand, grasp the mug, move it to my mouth, and drink the tea contained in it. But now suppose that I had no knowledge (or otherwise true representations) of my own abilities and opportunities—of the options available to me. How would I know what to do to fulfill my desire for tea, given the mug of tea in front of me—how would I know to extend my hand and grasp the mug, rather than try to extend my lips or to move the mug telepathically toward my mouth? How would I know to grasp the mug while lifting it rather than try to balance it on a finger, or merely touch it and pull it through adhesion? Of course, I could try to do those things and, seeing that they fail, try to extend my hand instead. But that, clearly, is not what we do. If it were, our simplest actions would be chaotic guesswork, not to mention a lot more time-consuming. In performing a simple action such as drinking a cup of tea, we are already guided by our awareness of our own abilities (I am able to extend my hand and grasp the mug, but not to extend my lips or move things telepathically) and of the opportunities that things offer to us (the mug can be grasped, but not pulled by merely touching it).3 From an evolutionary standpoint, too, it is unsurprising that our perception is geared toward properties that are relevant for our actions— what we can do to the things around us, what they can do to us, and how we can interact with our environment (Nanay 2011, 319). Thus, we should know, without much time-consuming calculation, that we can eat the apple, that the tiger can eat us, that we can fit into the cave ahead while the pursuing tiger cannot, or that our friend can help us reach the apple on the high tree. It is an interesting—and largely empirical—question exactly how such knowledge is acquired. I cannot go into this question here. What matters is that there is clear evidence, both empirical and reflective, that we do have access to the properties I have described— properties that concern what the objects in our environment can do, and what we can do to and with them—and that this access is related to our being agents in the world. Moreover, the processes that we use to acquire such access to particular properties, be it perceptual, inferential, or anything else, seem at least in some aspects hard-wired; we all learn about such properties, no matter the differences in the environments that we are brought up in (see Gibson and Pick 2000, 178). If we did not, then acting in the world would be practically (albeit perhaps not metaphysically) impossible.

An Agency-Based Epistemology of Modality  49 Let me stress again that my considerations in this section do not show how we have the modal knowledge required for agency. For all I am saying here, that knowledge—or a tendency to acquire it on the basis of very little evidence, very early on—may be innate. Rather than answering the question, “how do we get any modal knowledge?”, the considerations of this section seek to dispel the idea that often underlies this question: that modal knowledge is something very surprising and difficult, something that must come on top of our more basic non-modal knowledge. If I am right, then we always (or, at any rate, at a very basic stage) already have (some) modal knowledge. The question of how we get it is an interesting one, but not one to be answered from the armchair. (As Nolan 2017 recommends, we must go naturalist and look to developmental psychology for answers.) Another important question remains, however: how do we get modal knowledge beyond that which is required for agency? The next section suggests that agency can play a role in answering that question too. The Phenomenology of Action We have seen, so far, that we need to have some knowledge of both our own abilities and the affordances of the objects around us, in order to act in the purposive, coordinated way that we do. In this section, I want to focus on slightly higher-level ways of gaining knowledge about our own abilities and opportunities, which are not preconditions for, but rather arise from, our agency in the world. I want to suggest that the very phenomenology of agency provides further routes of access to modal properties, such as abilities and dispositions. My first case concerns the experience of our own dispositions; the second case concerns the experience of our own skill; the third case concerns the experience of resistance and the associated practical impossibility.4 First: the experience of our own dispositions. This comes in two varieties: the experience of exercising a disposition and the experience of resisting a disposition. The first is a way of accessing a modal property through its manifestation; the second provides access also to not or not fully manifested modal properties. Character traits seem to involve, even if perhaps they are more than just, dispositions (Alvarez 2017). In addition, agents possess various dispositions which may be too specific to qualify as character traits: the disposition to eat too much chocolate, to forget birthdays, to get angry, to whistle when nervous. Such dispositions, I suggest, can be experienced in two ways: as we exercise them, and as we resist their exercise. Take any such disposition—say, loquaciousness, the disposition to talk. When this disposition begins to manifest, there are two basic scenarios (and many gradations in between): the agent may “go along” with it, let it manifest, and start chattering along; or else, he may resist it (perhaps for fear of coming across as a chatterbox, or a mansplainer) and resist the

50  Barbara Vetter urge to talk. Likewise for the disposition to eat too much chocolate, to get angry, to whistle, and so forth: in each of these cases, a self-reflective agent may notice the beginning of its exercise and go along with it, or else exhibit self-control and resist the full-blown manifestation.5 In each case, there is again a distinctive phenomenology: the ease of going along with our dispositions is very much distinct and distinguishable from the effort of resisting them. Both ease and effort can be seen as an experience of the “pull” of the disposition toward one (kind of) behavior; in the first case, one is being pulled along, while in the second one is opposing the pull. It may be objected that these experiences of ease and effort, of going along and resisting, need not be experiences of dispositions. They might represent something actual—say, a desire or an exertion of willpower. Even if that represented event is in fact importantly related to the agent’s dispositions, it need not be represented as such. My response is most straightforward in the case of resisting a disposition: there seems to be a clear experience of the directedness of the disposition; its pull is a pull toward an (unactualized) manifestation. Desire may provide such directedness, but it is not always involved (think of the “pull” of irascibility); what I am after is the experience that is common to cases with and without desire. How else should the experience be understood than in modal terms? In the case of “going along” with a disposition, the disposition is, of course, typically manifested (unless something interferes). But that is not so from the start. The “going along” is also experienced as directed toward a manifestation which is not always desired, and not present from the start, and hence must be understood in modal terms. In short: it is the experience of directedness, which seems implicit in feeling the “pull” of a disposition, that supplies the modal content. The second case I want to consider is the phenomenology of skilled action. Compare, to start with, the following two scenarios, which most readers are likely to have experienced for themselves. Scenario 1: at the hospital, a physician tests your reflexes by knocking your right knee gently; your lower right leg kicks forward. Scenario 2: sitting on a bench, you idly kick your lower right leg forward. The two scenarios are clearly distinct even if the overt movement is identical: the second, but not the first, constitutes an action, an intentional movement of your leg. The two scenarios also clearly differ in phenomenology, even disregarding the knock of the physician’s instrument: the second feels like a movement that is controlled and guided by you; the first does not.6 Now think of more complex cases of skilled action: playing a piece on the piano, swimming or cycling, drawing a picture, giving a talk. When we perform these actions with skill (and do not suffer from distorting amounts of self-doubt), there is a distinctive phenomenology to them: a feeling of having things under control, of knowing what we are doing, of being in the driving seat. Like before, we can compare the skilled

An Agency-Based Epistemology of Modality  51 action to its unskilled counterpart, though in this case the counterpart will involve intentional action too: hitting keys on the piano randomly or simply drawing lines on a paper can occasionally result in the same outcome as a skilled performance of piano-playing or drawing (this is perhaps more likely when the skilled performance intentionally imitates the unskilled one), but, self-doubts or hubris aside, the performances differ not only in such features as their repeatability. They differ in phenomenology as well, in much the same way as the two scenarios considered above. I suggest that we have here, in the experience of skilled action, an experience of our own abilities in the course of their exercise. Note that this experience differs from that which we have seen at work with dispositions. In the case of an ability, there need be no “pull”, since we can have abilities without having the least tendency to exercise them; hence, what we experience is not related to giving in to, or resisting, a pull or urge. Instead, there is the experience of control, of selecting among different options, of guiding our own action toward an envisaged goal. (The phenomenology of skilled agency is a special case of what is sometimes called the “sense of agency”; see, e.g., Marcel 2002. It is also closely related to what Siegel (2005) calls the “phenomenology of efficacy”, which Siegel speculates may be experienced as an exercise of a power.) Again, it may be objected that what we experience here is not the ­ability—a modal property—but merely a complex of actual goings-on. These actual processes may constitute the exercise of an ability, but the question is not whether they do, but whether they are experienced as such. Do we experience skilled action as the exercise of a modal property, or merely as a particular and complex kind of behavior? In response, we may again point to the modal element of directedness. In the case of dispositions, directedness comes in the form of a “pull” toward a certain outcome; in the case of abilities, it comes in the form of guiding our own movements, or the movements of things we interact with, toward a certain outcome. In both cases, the outcome is not yet realized. In experiencing the guiding directedness of our behavior toward it, we are therefore plausibly experiencing a modal element of skilled behavior. And I suspect that there is a great deal of other modal facts that we experience in exercising our skill: the various options (i.e., things that we can do) among which we select what to do; the dispositions and lower-level abilities that are drawn on, exercised or stopped from exercising, in skillful performance; and so on. Exactly how these are experienced, and how the relevant modal elements are characterized, deserves much more detailed consideration than I can give here. But I hope to have made it plausible that the exercise of skills provides us with one route of access to our own abilities. My third and last case concerns not possibility, but impossibility, and relies on Schrenk (2014). (Note that Schrenk is concerned with the

52  Barbara Vetter experience of causation, not modality, but causation, as he points out, requires a modal element.) Consider, again, my desire to drink some tea, and the mug in front of me. It looks like I can lift it and drink the tea in it. But when I grasp the mug and try to lift it, I fail: it is glued to the table. Or imagine trying to lift a heavy weight, and failing; trying to run the last kilometer of a marathon when you are completely exhausted, and collapsing; pushing against a wall which won’t move. In each of these cases, it is the world (including our own body) that resists our intentions and our attempts to act in it in particular ways. In these cases, we do not experience a disposition pushing or pulling in a particular direction; the direction is given by our intentions, but thwarted by the objects involved. What we seem to experience here is not what can be done, but rather what (in the situation) cannot be done: I cannot lift the mug, we cannot lift the heavy weight, run the last kilometer, or move the wall. This “cannot” is not metaphysical impossibility, of course; it is situational and relative to our own abilities, just like affordances are. We might think of it as the absence of an affordance, as the absence of an ability (an inability), or instead as something more positive: as a disposition not to be lifted, moved, or used for running further, that is so strong as to amount to a practical necessity. The three cases I have sketched here provide intro- and extrospective access to our own abilities and dispositions as well as those of objects around us that arise in, rather than being a precondition for, acting. Together with the, perhaps more basic, knowledge that “Preconditions for Action” argued we must have, this provides a rich basis for modal thought and knowledge in our experience as agents. I believe that this is a highly promising starting point for modal epistemology, but it is no more than a starting point. A first potential problem is that the properties that we have looked at especially in this section seem to be much richer than mere possibilities. How then do they provide access to just that: mere possibility? That is the issue to be addressed in the next section.

Toward Mere Possibility In the previous section, I have argued that our experience as agents provides us with quite straightforward and easy access to certain modal properties: our own abilities and dispositions, as well as the affordances (and lacks thereof) of objects in the world around us. But does it thereby provide us with access to modality itself? The properties that I have pointed to are, of course, modal properties: they involve unactualized possibilities or, in the terms of my preferred metaphysics, potentialities. But they also involve more than mere possibility, or potentiality. To have an ability is not just to have a potential; it is to have a certain amount of control over its exercise. To have a disposition, too, is not just to have a

An Agency-Based Epistemology of Modality  53 potentiality; it is to have a certain tendency to exercise that potentiality. (The case of affordances is less clear, and I will take it up in a moment.) In the philosophical literature, both dispositions and abilities are standardly linked with something much stronger than a possibility: a counterfactual conditional. To have a disposition, in this view, is not just to possibly behave in a certain way, but to be such that one would behave in such a way if certain conditions held. To have an ability, on a similar view, is not just to possibly do something, but rather to be such that one would do that thing if one intended, chose, or tried to do it. A similar analysis has been extended to affordances (Scarantino 2003). I have argued against all three counterfactual views in different places (see Vetter 2013 and Vetter 2014 for dispositions, Vetter 2019 and Jaster and Vetter 2017 for abilities, and Vetter 2020 for affordances). But even if, as I have suggested, we think of these modal properties as more possibility-like than counterfactual-like, the problem remains that they seem to involve more than mere possibility; and that, for this reason, they do not conform to some basic principles that philosophers have taken to be constitutive of possibility. One such principle, which is often referred to in the epistemology of modality (Roca-Royes 2017; Hanrahan 2017), is the inference from actuality to possibility, known as axiom (T) in modal logic. But neither abilities nor dispositions seem to validate this inference. For abilities, the case has been argued by Kenny (1976). As Kenny points out, “a hopeless darts player may, once in a lifetime, hit the bull, but be unable to repeat the performance because he does not have the ability to hit the bull” (Kenny 1976, 214). A one-off success does not entail ability, since ability requires a certain sort of control, which typically results in the reliability and repeatability of performances. And so, actuality does not entail ability. With dispositions, the issue is not control but rather strength: a disposition is not just a matter of mere possibility, but of a certain positive probability or tendency. Thus, things can break without being fragile, i.e., disposed to break: a sturdy bridge may break when coming under enormous stress, thus falsifying the inference from actuality to disposition. Is this a problem for taking abilities, dispositions, and affordances to provide our entry point to metaphysical modality? I think not, for the story I have told in “Modality in Action” does put other modal properties at our epistemic disposal, properties that are akin to mere (albeit, of course, restricted) possibilities. These include some affordances, especially those related to risk: to see (or otherwise recognize) that a cliff “affords falling off” seems to involve the recognition of no more than a possibility, albeit restricted and situational, of one’s falling off that cliff. And affordances that are related to opportunity rather than risk may exhibit the same structure: even an incompetent darts player will be able to recognize that a dart board affords scoring a bull’s eye to her, while the wall to which it is attached does not; and that recognition seems

54  Barbara Vetter veridical. This may seem to suggest that affordances are not as closely related to abilities as I made it look above in “Preconditions for Action”. But in fact, we can question whether the observations concerning abilities and dispositions really hold with full generality. There is certainly a sense in which even an incompetent darts player has the ability to hit the bull’s eye, while a dolphin, say, does not. (This point is well-rehearsed; see Lewis 1976.) An ability in this more liberal sense plausibly satisfies the T axiom, and it is as important for action as the more demanding kind of ability invoked in Kenny’s counterexample: we often find ourselves in situations where we attempt to perform actions without having reliable skill (not least in the course of practicing to acquire the skill, or because there are no better options), and even there it will be useful to recognize what we can (in principle) and what we can’t do. Likewise, there are certainly some cases that we may classify as dispositions, and which do not come with a strong tendency to manifest. Contrast “x is fragile” with “x is breakable”: the former is not, but the latter arguably is, entailed by “x breaks” (though we may have to adjust tenses). Is breakability, then, not a disposition? I am not sure that we have circumscribed the extension of “disposition”, a philosophical term of art, with sufficient precision to answer this question. What is more important is that such properties are relevant for action: a sturdy bridge’s breakability, for instance, may well be relevant to the actions of the workers tasked with its demolition, and recognized by them as such for the reasons given above in “Preconditions for Action”. Whatever exactly we want to say about abilities and dispositions, then, the story I have told so far does allow for properties that seem to correspond to mere, albeit of course restricted or relativized, possibilities. Some of the modal properties that we have access to qua agents involve more than mere possibility; and it was these additional features, the element of activity or control in (some) abilities, and the element of “pull” toward a manifestation in (some) dispositions, that I have appealed to above in “The phenomenology of Action”. The considerations of “Preconditions for Action” made no such appeal, however, and we have now seen why—since no such additional elements need to be present. A question that remains is how it is that we manage to subsume all these different modal properties under a common concept of possibility (or potentiality) expressed by, say, “can”. I suspect that this is a matter of recognizing the similarity between the mere-possibility cases that we are familiar with (as I have just argued), and the more complex cases encountered in the phenomenology of agency, and exercising our general capability for recognizing common genera among our concepts—abstracting away, in this case, from their differentiating features such as the element of control, the pull toward manifestation, and our activity or passivity in their exercise, and recognizing the shared element of (potentially)

An Agency-Based Epistemology of Modality  55 unmanifested possibility. But here, again, I venture into territory that will need a more empirical basis. What I hope to have shown is that there are no obvious principled philosophical obstacles to a route such as the one I have envisaged, from our experience in agency to a concept of mere possibility or potentiality.

Modality Beyond Action Projection In the previous section, I have addressed one dimension of generalizing from the results of “Modality in Action”, by abstracting from the more specific features of abilities, dispositions, etc., to their common status as possibilities, or potentialities, thus also allowing for the inference from actuality to possibility. In this section, we will generalize in a different and more obvious dimension. The modal judgments that can be accounted for as either preconditions for, or implicit in the phenomenology of, action form an interesting but clearly narrow subclass of our modal judgments in general. Starting as I have claimed we do from modal knowledge in the context of agency, how do we gain knowledge of possibilities that are not of direct relevance to our actions? My general response to this question will be: in just the same ways in which we gain knowledge of other properties that go beyond the context of our direct interaction with objects. There is no special method to modal knowledge. What makes it above knowledge is its starting point, as sketched in “Modality in Action”. To use terminology from Vaidya and Wallner (2021), once we have entered modal space in agency, we navigate it in the same way in which we navigate any epistemic space. So much for the general gist of my response; now let me be a little more specific. A first point to notice is that the access to modality that has been sketched in “Modality in Action” is not, of course, limited to contexts in above we do act. With the exception of those modal judgments that arise from experience of one’s own action, the knowledge of modal properties that was argued for in “Preconditions for action” will be present whether or not we act on it; one function of such modal knowledge is to provide us with knowledge of the options among which we can choose in our actions. A second more wide-ranging point is that we naturally tend to project our experience of objects beyond the context of our interaction with them, i.e., to perform what can be rationalized as inductive reasoning. This is a familiar point that requires no argument in the case of perceptual knowledge. Having looked at an apple and seen that it is red, we will project this judgment beyond the situation in which we made the sensory

56  Barbara Vetter experience (asked about the apple’s color an hour later, I will respond without looking that it is (now) red) and beyond the particular object of which we had sensory experience (having seen many red apples, I will take apples in general to be red). My point is simply that this same tendency naturally applies to our attribution of modal properties to objects as well. Having looked at a mug and recognized (perceptually or otherwise) that it can break, I will project this judgment beyond the situation in which I had the experience (asked about the mug an hour later, I will respond without looking that it can (now) break) and beyond the particular object of which I had sensory experiences (having recognized many mugs as breakable, I will take mugs in general to be breakable). Such projections are presumably basic to our psychology, and most of the time happen unconsciously. Epistemologically, they will count as justified and in the best cases as knowledge, if the ways in which we arrive at them correspond, explicitly or (more often) implicitly, to sound reasoning. Sound reasoning, in these cases, will tyically be good inductive and abductive reasoning; reasoning by similarity, analogous reasoning, and the use of the imagination (to be discussed in more detail in a moment) will certainly also play a role. Moreover, once we go beyond the context of our own actions, firsthand experience is of course not required for modal knowledge. We can learn what things can do by testimony, just as we can learn what things actually do by testimony. Children acquire a great deal of modal knowledge in this way: they learn that the stove can burn them (no matter how harmless it looks), or that drinking the washing-up liquid can make them sick (no matter how drinkable it looks), all— ideally—­w ithout any firsthand experience of the respective properties. Imagination So far, I claim, there is nothing special about modal knowledge. We project properties from our experience of, and interaction with, objects to those and other suitably related objects in situations where we do not experience and interact with them. But I have so far said nothing about what may well be the most prominent way of gaining modal knowledge, among rationalists and empiricists alike: the imagination. According to Tim Williamson, “[a]t least metaphorically, one might regard imagination as a form of attention to possibilities” (Williamson 2016a, 115). Surely, imagining is important to our gaining knowledge of possibilities. But the imagination has epistemic functions other than alerting us to possibilities.7 We arguably use it to gain knowledge of other people’s state of mind (see, e.g., Church 2016)—whether they are actually, not just potentially, happy or sad, jealous or nostalgic, anxious or angry.

An Agency-Based Epistemology of Modality  57 We use it to gain knowledge and understanding of the past: in order to know whether my son had lunch today, it will help to imagine what his day was like and whether there was enough time in it to sit down and eat; in order to understand the French revolution, it may be useful to imagine the life conditions in Paris in the second half of the 18th century; in order to better understand the intention of a text, it may help to imagine the historical situation in which it was written. We use it to gain knowledge and understanding of matters that are spatially remote or occluded from perception: in order to know where the stone that just flew past me has landed, when its destination is occluded from me, it is useful to imagine its trajectory relative to the environment that is being occluded; in order to know what to wear for a hike, it may be useful to imagine the conditions of the hiking path as well as the likely weather conditions. In none of these examples would the knowledge to be gained with the help of the imagination qualify as modal knowledge. None of the examples are unusual. So it is clear that the imagination has epistemic functions that go beyond the attaining of modal knowledge. Before we turn to the use that it unquestionably has in attaining modal knowledge, two points about these non-modal cases bear noting. First: in none of the examples I have given does the imagination do the epistemic work on its own; imagining is merely one among several activities that are used to ascertain the truth or falsity of a proposition (such as that my son has had lunch today, or that the stone landed in the window). Second: in each of the examples I have given, the imagination must be constrained in some way to do the work it does. It will be no help if I imagine my son’s schedule differently from how it is, the living situations in 18th-century France as akin to those today, or the stone as bouncing off objects like a rubber ball. All of these seem to be imaginable, and may well be imagined in other contexts (thinking about improving the schedule at my son’s school, writing an alternative history of France, etc.). Amy Kind and Peter Kung have taken considerations along these lines to give rise to a “puzzle of imaginative use”: “How can the same activity that allows us to fly completely free of reality also teach us something about it?” (Kind and Kung 2016, 1). Kind and Kung suggest a solution to the puzzle which, they argue, has been widely if often implicitly accepted in philosophy: the imagination can teach us about reality if and when it is properly constrained, that is, guided by the right kind of facts (which they spell out in some more detail). How are we to understand that guidance? I suggest that, as a first approximation, x’s imagining is guided or constrained by the fact that p iff the content of x’s imagining either includes or is compatible with p because x knows (explicitly or implicitly) that p. This accounts for my examples above: my imagining that my son had enough time for lunch is guided by the fact that his s­ chedule has a long

58  Barbara Vetter enough lunch break, and that he generally has lunch if at all possible; my imagining of the stone’s trajectory is guided by certain physical facts (of which I have merely implicit, “folk- physical” knowledge); and so on. It is plausible that the epistemic value of imagining in a given case increases with the amount of guiding knowledge that the imagining subject has.8 If this picture is right, then imagination plays a similar role to induction, abduction, etc.: Given prior knowledge of some facts, it will generate new knowledge if all goes well. With all this in place, let us now turn to the role of the imagination in modal knowledge. I see you throw a stone, and I catch it. Holding it in my hand, I wonder whether it could have hit the window, breaking it in the process. My imaginative process seems hardly different than it was in the case above, when I wondered whether the stone did hit the window: I am guided by the stone’s actual trajectory (up to the point where I caught it), size, and weight, as well as the actual position of the window and other objects in its surroundings. Only this time, there is an actual fact—a part of the “world as it is”—that I am ignoring: the fact that the stone did not actually land in the window because I caught it. Instead, what seems to guide my imagination is knowledge of certain modal facts: how things can or must behave when thrown in a certain way. If we may generalize from examples like this, then the use of the imagination in ascertaining possibilities is not very different from its use in ascertaining actual truth, and the imagination is not a distinctive source of modal knowledge (as opposed to non-modal knowledge). Its role in acquiring modal knowledge, as well as non-modal knowledge, seems to be ampliative: like induction, abduction, and so forth, it merely helps us expand our basic modal knowledge beyond the confines of direct experience. And thus the role of the imagination in gaining modal knowledge is not, at least prima facie, a threat to my ambitious, agency-based account.9

Modality in Philosophy Unrestricted Modality So far, this paper has been all about ordinary possibilities: the possibilities we recognize and reason about before (and whether or not) we have done any philosophy. But modal epistemology is done not merely to understand our ordinary modal knowledge, intriguing though it may be. It is typically done as part of philosophical methodology, and as such it is typically concerned not merely with our (ordinary) knowledge of ordinary possibilities, but with our (ordinary or extraordinary) ­k nowledge of unrestricted, metaphysical possibility. What is the approach that I have sketched here to say about that kind of modal knowledge? My approach, as I have said at the outset, is anti-exceptionalist. The possibilities envisaged in philosophical debates may well be extraordinary,

An Agency-Based Epistemology of Modality  59 but our way of knowing about them will not. Metaphysical modal knowledge is merely the continuation of ordinary modal knowledge. Here is a picture of how we get from the more restricted ordinary possibilities (I can reach the mug, you can read this paper, the window can break) to the more “remote” cases that are of interest in much of philosophy. The picture is due to Edgington (2004); I have sketched my version of it already in Vetter (2016). The idea is that we can implicitly recognize that our ordinary possibility claims are contextually restricted in some way (without, of course, thereby committing to a precise view on how the restriction works: by selecting among the possible worlds, as in standard modal semantics, or perhaps among the potentialities, as I would prefer to say). When doing philosophy, we generally want to relax such contextual restrictions: we are interested not just in what qualifies as possible here, now, in this context, but in what qualifies as possible simpliciter. Compare: when we do ontology, we appear to implicitly recognize that our ordinary existence claims are contextually restricted in some way. When doing philosophy, we want to relax such contextual restrictions: we are interested in what there is not just here, now, and saliently for the present context, but in what there is simpliciter. So far, of course, this is just a hypothesis of how we manage to talk about metaphysical possibility: by creating a context, in philosophical conversation, that imposes no restrictions on the possibilities expressed. But how can we know, or have justifying grounds for, any such unrestricted possibility claims? The response to this question, much as my response to the similar question of how we get to ordinary but action-transcending modal knowledge in “Modality Beyond Action”, is very simple: we use whatever methods we have for building and comparing theories. What these methods are or should be is a question that is quite distinct from modal epistemology proper, and so in a sense I must remain somewhat unspecific here. The point, again, is that what’s special about modal judgments is not how we get to the individual, high-level philosophical judgments. It is, rather, how we got “into” the subject-matter in the first place. Vaidya and Wallner (2021) have distinguished between two questions in the epistemology of modality. The “access question” asks: “how is it that we gain access to or acquire epistemic standing for beliefs about ­ allner modality, such as that it is possible for x to be F?” (Vaidya and W 2021, 1909). The “navigation question” asks: “how we can reason with justification from one kind of modality to another, say from logical to metaphysical to physical modality?” (Vaidya and Wallner 2021, 1910). Within the context of the approach sketched here, the relevant navigation question is: how can we reason with justification from ordinary modality (as discussed in previous sections) to metaphysical modality? My answer

60  Barbara Vetter to the access question should be clear: we gain access to modality in our capacity as, and by our experience as, agents in the world. My response to the navigation question is, in a sense, uninteresting: we do whatever we do when we start to think about things in metaphysics. The aim of this section is to further clarify and motivate this uninteresting response. My approach is, in effect, a species of a popular kind of account of metaphysical modality, according to which metaphysical modality differs from ordinary modality by being “absolute”, or the “broadest” kind of objective modality. Like other such views, however, it faces a challenge: how can metaphysical modality be absolute, or the broadest, or the widest, kind of objective modality, when it is more restricted than logical modality? How, in other words, would my “unrestricting” of context not lead to something wider than metaphysical modality? My basis for modal thought and knowledge is very narrow, and we have to generalize a great deal if we are to get from there to metaphysical modality. How should we know to stop at the metaphysical possibilities? Why not generalize further and include metaphysical impossibilities such as water’s being XYZ? (See Clarke-Doane 2017, 2019; Mallozzi 2021b.)10 To begin responding to the challenge, let us reconsider the analogy with ontology. When we do ontology, I said, we drop contextual restrictions and ask what there is simpliciter. This is not to say that anything goes, of course: one can adopt my picture of what we do in ontology while still holding a very restrictive view on what there is simpliciter. We want to drop contextual restrictions, but not change the topic: we want to know what exists, in the same sense of “exists” that we use when we contextually restrict our quantifiers. The same holds for modality: when we think about metaphysical modality, I suggest, we drop all contextual restrictions and ask how things could have been simpliciter. But again, this is not to say that anything goes. We drop contextual restrictions, but we do not change the topic. We want to know how things could have been, in the same sense of “could” that we used when we were speaking with contextual restrictions. We are looking for the widest extension of that which we started from—modal properties of objects, as they were presented to us in agency—not for the widest extension of anything that one might want to call “modality”. In the next section, I want to argue that there are some natural boundaries for the extension of our modal concepts which are relatively (though not entirely) independent of specific theories. As it happens, these boundaries correspond nicely to some common views in the metaphysics of modality. Boundaries In this section, I want to consider three kinds of potential boundaries that we may face in extending our understanding of modality from the paradigmatic cases discussed in the second and third sections.

An Agency-Based Epistemology of Modality  61 First: Meaning and analyticity. Without providing a theory of what it is exactly that we know about in paradigmatic cases, we can recognize that we are dealing with modal properties of some sort, and thus with (what seem to be) objective features of reality—and not a phenomenon that is conceptual, linguistic, or even logical in nature. Modality as we get to know it in agency is about things, not names; about how things are, not how we describe them. And if this is so, then modality should not be sensitive to how things are being described. If it’s possible that this thing is so-and-so, then it’s possible however we name the thing, and however we express being so-and-so. This gives us referential transparency for directly referential expressions, i.e., expressions that contribute nothing but their reference to the truth conditions of a sentence. Proper names such as “Hesperus” and “Phosphorus” are plausibly such directly referential expressions. Thus if it is possible for Hesperus to host life, it is possible for Phosphorus to host life—because the possibility concerns the object referred to, Venus, and not the names used to refer to it. And if Hesperus couldn’t fail to be ­Hesperus, then Hesperus couldn’t fail to be Phosphorus. In other words: it is necessary that Hesperus is Phosphorus; and parallel reasoning leads us to the necessity of identity in other standard cases. The same reasoning holds when we insert natural kind terms, such as “water” and “H 2O”, assuming a Kripkean semantics for them. We also get the necessity of analytic truths (if there are any). If, say, “vixen” just means “female fox”, then any possibility that concerns being a vixen also concerns being a female fox, and vice versa. After all, the possibilities concern the properties ascribed with such expressions (or, more nominalistically speaking, they concern what things are like when they are vixens or female foxes); they do not concern the expressions we use to ascribe such properties. In particular, if nothing can be a vixen without being a vixen, then nothing can be a vixen without being a female fox. In other words: it’s necessary that all vixens are female foxes; and similarly for other textbook analyticities. Second: Essence. Some philosophers believe (and others don’t) that there is such a thing as the essence of an object x, what it is to be x; and, perhaps, that there is such a thing as the essence of a property F, what it is to be F. If that is correct, then essence, too, plausibly imposes certain boundaries on the extension of our modal thought beyond the initial paradigmatic examples. Again, without providing a theory of what it is exactly that we know about in the paradigmatic cases of modal knowledge, we can recognize that we are dealing with modal properties of objects. And the modal properties of an object, x, are plausibly constrained by what it is to be x. If what it is to be Socrates involves being human, then anything Socrates can be or do must be compatible with his being human; otherwise,

62  Barbara Vetter it would not be Socrates being or doing it. How we precisify this informal consideration depends on the metaphysics of the modal properties in question. In possible-worlds terms, we will appeal to Socrates’s identity across worlds; in potentiality terms, to the idea that a thing must manifest its own potentialities (Vetter 2021). Either way, the arguments may not be decisive, but they are highly natural, and that is enough, I take it, to explain our inclination to let the attribution of modal properties be constrained by the thing’s essence, if we are inclined to attribute an essence to it. Likewise, if what it is to be F is to be G (if the essence of being F is that it is being G), then it would seem that any modal property that involves being F must likewise involve being G: for if it did not, it would not involve being F but something else instead. Thus, for instance, if what it is to know that p is to have a justified true belief that p, and if one cannot know p without knowing p, then one cannot have a justified true belief that p without knowing that p. (And if one can, of course, then to know cannot just be to have a justified true belief.) In this way, I suggest, reflection on the paradigmatic cases for our modal thought and knowledge—the modal properties encountered in agency—already yields some of the central philosophical claims about metaphysical modality: the necessity of identity, of textbook analyticities, and of any truths that are essential either objectually or generically. How does this meet the challenge from the previous section? It does so by showing how the phenomenon that we are interested in when we extend from the starting point sketched in earlier sections cannot be extended arbitrarily, but comes with certain natural boundaries. And given certain widespread assumptions, these boundaries can be seen to stop far short of logical or conceptual modality. If we were to drop those boundaries, I submit, we would no longer be generalizing on the phenomenon which we started out with. Rather, we would be changing the topic. Note, however, that the boundaries I have drawn were not entirely motivated by the epistemology of modality alone, and to a certain extent are therefore hypothetical: we can know certain modal truths by being implicitly guided by those constraints, if we can know the contents of those constraints. (That “if” clause again decomposes into at least two elements: if the constraints hold, e.g., if there really are analytic truths or essential properties; and if we have the right kind of epistemic access to them.) Accordingly, being guided by these constraints requires knowledge of essential properties, analytic truths, and so forth. It is beyond the scope of the present proposal to provide the relevant epistemologies here, but of course there is a host of philosophical work on them.11

An Agency-Based Epistemology of Modality  63 Metaphysical Cases and Moderate Modal Skepticism In a recent paper, Amie Thomasson (2018) argues that modal empiricist accounts like the one presented here fail to give a full account of our knowledge of properly metaphysical modal claims. What makes these accounts attractive, she points out, is their integration of modal knowledge into our empirical knowledge of the world. Knowledge of dispositions, for instance, is inductively justified and yields testable predictions—a far cry from the scientifically suspicious methodology of questioning one’s intuitions in the armchair. But, Thomasson points out, “the distinctively metaphysical modal features at issue in characteristic metaphysical debates are cases in which we have the very same empirical information, and same physical laws and properties, and yet come to different modal conclusions” (Thomasson 2018, 7). She concludes that within such empiricist approaches, “there does not seem to be the prospect of explaining how we could come to know these distinctively metaphysical modal properties” (Thomasson 2018, 8). The approach I have developed here, at first look, seems to underwrite Thomasson’s conclusion: starting from abilities, affordances, dispositions, and some generalizations from them certainly do not directly address metaphysical modal claims such as the claim that the lump of clay can, while the statue cannot, survive squashing; that my body can, but I cannot, survive a complete and irreparable memory loss (these are Thomasson’s examples); that there could be a molecule-for-molecule replica of me that lacked my conscious experiences; that the universe could have consisted of nothing but two indistinguishable spheres; and so on. One reaction that is not uncommon among modal empiricists (see Leon 2017; Kung 2017; Hawke 2017; Nichols 2006) is to adopt what van Inwagen (1998) has called “modal skepticism”, a skepticism of a very moderate kind that allows us knowledge of reasonably close and everyday possibilities and necessities, as well as some straightforward cases such as the necessity of logical, conceptual, and mathematical truths, but denies that we have any reliable capacity to acquire knowledge of the more remote metaphysical possibilities, such as those involving philosophical zombies, indistinguishable sphere-universes, and the like. The approach that I have been exploring here is not committed to modal skepticism. It does, however, curtail the role that modal thought and knowledge can play in metaphysics (and other areas of philosophy). The approach is not committed to modal skepticism because it poses no limits on the methods we can use to gain modal knowledge. Basic modal knowledge is empirical (in a broad sense), to be sure, but most modal knowledge is not basic, and can be highly theory-driven. Thus, we can learn about the possibilities mentioned above by utilizing our best philosophical theories, along with the modal repertoire that has arisen

64  Barbara Vetter from our experience in agency. This is what I have already suggested in “Boundaries” that we do in judging cases such as the necessity of identity or of essential properties. In the cases mentioned above, we will have to use other philosophical theories, about artifacts, persons, and the nature of modal properties. If we do so in the right way, there is nothing that would in principle prevent us from gaining knowledge about such cases. For instance, if our best theories of what a person is tell us that a person is essentially tied to their memories, then we can justifiably conclude (by the kind of reasoning sketched in the previous section) that I, a person, could not survive total memory loss. If our best theories of what a person is tell us that a person is not essentially tied to their memories, then we can justifiably conclude that I, a person, could probably survive total memory loss (barring other considerations that would foreclose such a possibility). If our best theories tell us neither of these things, then we had better remain agnostic about the possibility question for the time being. (Here and in the next paragraph, I am very much in agreement with Fischer (2016), though he takes a different route to get to the same conclusion.) Thus, the approach curtails the role that modal thought and knowledge can play in metaphysics (and other areas of philosophy): it makes our modal judgments about the relevant cases not unjustified, but dependent on exactly the kind of theoretical choice that they are often thought to be used for or against. If my judgment on the possibility of my surviving total memory loss depends, and should depend, on what I think is the best theory of personal identity, then that judgment can hardly serve as evidence for such a theory, on pain of circularity. The same will hold for other cases: modal arguments, on my approach, generally get things backward, using the modal judgment to argue for the theoretical judgment which it is and should be based on. Or so it would seem; things don’t look quite so bleak for modal arguments at second blush. For assuming that we implicitly use the kinds of considerations I have sketched in “Boundaries”, we can use our modal judgments as evidence—not so much for the truth of the underwriting theoretical judgment, but for our having already accepted it. Thus, the fact that we all accept Gettier’s counterexamples against the JTB analysis of knowledge is evidence for our already, albeit implicitly, having a different view about the nature of knowledge than the JTB analysis states. If you endorse Gettier’s judgment about his cases constituting justified true belief without knowledge, you cannot then consistently go on to claim that knowledge is justified true belief. In this way, modal judgments can still provide a source of evidence for or against our implicit acceptance of the theoretical judgments that underwrite them. When a modal judgment is accepted unanimously, then it teaches us something about our (often pre-theoretically held) commitments on the subject. But modal judgments cannot then be used in order to forestall revisions to those commitments, as long as we revise the modal judgments along with them.

An Agency-Based Epistemology of Modality  65 Nor are we entitled to think, on the present approach, that there is some sort of direct access to modal truths, perhaps via conceivability, that is theory-independent and can be used, like empirical data, to assess theories. The approach I have sketched does not vindicate the move from the fact that some of us can conceive of their own disembodied survival, or of their own philosophical zombie-twin, to conclusions about personal identity or about the nature of conscious experience, though it does vindicate conclusions about the conceivers’ theoretical commitments, implicit or explicit, about personal identity or the nature of conscious experience (and whatever may follow from their holding those commitments). Thus, the present approach vindicates modal knowledge in metaphysics, while at the same time curtailing its usefulness. This, I take it, is the price we pay for modal empiricism.12

Notes 1 I prefer this term to the better-entrenched “uniform”, for reasons to be discussed below. 2 “Epistemic modality” is ambiguous. In linguistics and the philosophy of language, it is typically used to apply to ordinary modals used to convey, roughly, that something is compatible or not with the contextually salient state of knowledge, as when a detective says “Jones must be the murderer” (Kratzer 1981). In the literature on modal epistemology, it is often used to apply to conceivability—compatibility, we might say, not with our state of knowledge but with any rational state of knowledge. The two phenomena are presumably related, but they are not the same. For present purposes, we can lump them together to contrast with objective modality, which is entirely independent from our actual as well as potential knowledge. 3 Infants may well start out with something like the “chaotic guesswork” I have gestured at in the main text; it is known as “exploratory activity”. Its function is precisely to acquire the kind of knowledge—of their own abilities and the affordances offered by their environment—that is needed for goal-directed agency. See Gibson and Pick (2000). 4 The reader may wonder at this stage how my appeal to phenomenology sits with the claim that the proposed modal epistemology is empiricist. I respond that empiricism has always included inner experiences—Locke and Hume include “reflexion” along with “sensation” as a source of our impressions and ideas. 5 I am here using the terms “manifestation” and “exercise” interchangeably. Since a disposition’s manifestation or exercise typically consists in an extended process, I am also assuming that we can often distinguish between the partial and full manifestation of a disposition; in the former case, the relevant process is started but not completed, while in the latter case it is completed. ­Alternative views might take the relevant dispositions to be multi-track and say, for instance, that loquaciousness already manifests in a person’s urge to talk, while the talking itself is just another manifestation of the same disposition. My point in the main text can be rephrased accordingly: the disposition manifests in one “track”, and the agent notices the pull for manifestation in another “track”, which she can then go along with or resist. 6 Such cases play a role, of course, in the philosophy of action at least since Anscombe (1957), and in more recent debates about the phenomenology of agency (see Bayne 2008; Roessler and Eilan 2003; Mylopoulos and Shepherd

66  Barbara Vetter forthcoming) or efficacy (Siegel 2005). But the emphasis here is typically on the actual agency, not on the ability; and the ramifications for modal epistemology have not, as far as I know, been noted. 7 I do not here assume any particular conception of imagination. What I say is so general that I believe it holds for any sense of “imagination” that modal epistemologists have appealed to. Note, however, that I am here still concerned with the role of imagination for ordinary modal knowledge, not for knowledge of remote metaphysical possibilities. Thus, my cases will be more relevant to the modal empiricist’s ideas of imagination, which tend to start from such ordinary cases, than to the modal rationalist’s, which tend to start with a priori conceivability. See Balcerak Jackson (2016) for a useful clarification of the notions involved. 8 Perhaps we need to make some adjustments to the proposed account since some of the guiding assumptions, such as folk physics, are false and therefore don’t qualify as knowledge at all (Williamson 2007, 145ff.). I will leave this complication aside here. 9 There is an important disanalogy: when using the imagination to affirm the truth of “possibly p”, the content of the relevant imaginings will typically be not the proposition whose truth is being probed (possibly p), but rather the embedded proposition, p. I suggest that this can be explained by a full consideration of what it is to imagine a modal property, and might even help extend the account toward de dicto modality. But I do not have the space to properly address this here. 10 Clarke-Doane is concerned not just to argue that we cannot simply think of metaphysical modality as absolute modality, but also that it is a mistake to attribute to metaphysical modality “unique metaphysical significance” (Clarke-Doane 2017, passim). As should become clear in “Metaphysical Cases and Moderate Modal Skepticism”, this is a point on which I may well agree with Clarke-Doane. 11 For knowledge of essence, see, for instance, Vaidya (2010) and Mallozzi (2021a). In Vetter 2021, I suggest that essence, understood in the Finean sense, is more closely related to grounding and dependence than it is to modality. Thus, I take the epistemology of essence not to be a special case of the epistemology of modality. That is, of course, a controversial claim, and those who disagree with it will justifiedly ask for my proposed modal epistemology to be supplemented with an epistemology of essence. 12 I would like to thank the editors of this volume for allowing me to draw this broad-brush picture here. For helpful feedback on the paper, I am grateful to Tom Schoonen, Anand Vaidya, Michi Wallner, Julia Zakkou, two reviewers for this volume, and patient audiences at Amsterdam, ­D ublin, and Berlin. I gratefully acknowledge funding from the Deutsche Forschungsgemeinschaft (German Research Foundation, DFG) within the Center for Advanced Studies in the Humanities “Human Abilities”, grant number 409272951.

References Alvarez, Maria. 2017. “Are Character Traits Dispositions?” Royal Institute of Philosophy Supplement 80:69–86. Anscombe, Gertrude Elizabeth Margaret. 1957. Intention. Cambridge, MA: Harvard University Press. Balcerak Jackson, Magdalena. 2016. “On the Epistemic Value of Imagining, Supposing, and Conceiving.” In A. Kind and P. Kung (eds.), Knowledge ­ Through Imagination, 41–60. Oxford: Oxford University Press.

An Agency-Based Epistemology of Modality  67 Bayne, Tim. 2008. “The Phenomenology of Agency.” Philosophy Compass 3:182–202. Chalmers, David J. 2010. “The Two-Dimensional Argument against Materialism.” In The Character of Consciousness. Oxford: Oxford University Press. Chalmers, David J. 2012. Constructing the World. Oxford: Oxford University Press. Church, Jennifer. 2016. “Perceiving People as People: An Overlooked Role for the Imagination.” In A. Kind and P. Kung (eds.), Knowing Through ­Imagination, 162–182. Oxford: Oxford University Press. Clarke-Doane, Justin. 2017. “Modal Objectivity.” Noûs 53:266–295. Clarke-Doane, Justin. 2019. “Metaphysical and Absolute Possibility.” Synthese 198 (Suppl 8):1861–1872. Edgington, Dorothy. 2004. “Two Kinds of Possibility.” Proceedings of the ­Aristotelian Society, Supplementary Volume 78:1–22. Fischer, Bob. 2016. Modal Justification via Theories. Cham: Springer. Gibson, Eleanor J. and Pick, Anne D. 2000. An Ecological Approach to Perceptual Learning and Development. Oxford/New York: Oxford University Press. Gibson, James J. 1986. The Ecological Approach to Visual Perception. New York: Psychology Press. Taylor and Francis Group. Hale, Bob. 2003. “Knowledge of Possibility and Knowledge of Necessity.” ­Proceedings of the Aristotelian Society 103:1–20. Hanrahan, Rebecca. 2017. “Imagination, Possibility, and Plovers.” In B. Fischer and F. Leon (eds.), Modal Epistemology after Rationalism, 197–219. Cham: Springer. Hawke, Peter. 2017. “Can Modal Empiricism Defeat Modal Skepticism?” In B. Fischer and F. Leon (eds.), Modal Epistemology after Rationalism, 281–308. Cham: Springer. Ichikawa, Jonathan Jenkins. 2016. “Modals and Modal Epistemology.” In A. Kind and P. Kung (eds.), Knowledge Through Imagination, 124–144. Oxford: Oxford University Press. Jaster, Romy and Vetter, Barbara. 2017. “Dispositional Accounts of Abilities.” Philosophy Compass 12:e12432. Kenny, Anthony. 1976. “Human Abilities and Dynamic Modalities.” In J. Manninen and R. Tuomela (eds.), Essays on Explanation and Understanding, ­209–232. Dordrecht: Reidel. Kind, Amy and Kung, Peter (eds.). 2016. Knowledge Through Imagination. Oxford University Press. Kratzer, Angelika. 1981. “The Notional Category of Modality.” In H. J. Eikmeyer and H. Rieser (eds.), Words, Worlds, and Contexts. New Approaches in Word Semantics. Berlin/New York: de Gruyter. Kung, Peter. 2017. “Personal Identity without Too Much Science Fiction.” In B. Fischer and F. Leon (eds.), Modal Epistemology after Rationalism, 133–154. Cham: Springer. Leon, Felipe. 2017. “From Modal Skepticism to Modal Empiricism.” In B. Fischer and F. Leon (eds.), Modal Epistemology after Rationalism, 247–261. Cham: Springer. Lewis, David. 1976. “The Paradoxes of Time Travel.” American Philosophical Quarterly 13:145–152. Maier, John. 2010. “Abilities.” In E. N. Zalta (ed.), The Stanford Encyclopedia of Philosophy. Spring 2010 edition. https://plato.stanford.edu/entries/abilities/

68  Barbara Vetter Mallozzi, Antonella. 2021a. “Putting Modal Metaphysics First.” Synthese 198 (Suppl 8):1937–1956. Mallozzi, Antonella. 2021b. “New Directions in the Epistemology of Modality: Introduction.” Synthese 198 (Suppl 8):1841–1859. Marcel, Anthony. 2002. “The Sense of Agency.” In J. Roessler and N. Eilan (eds.), Agency and Self-Awareness. Issues in Philosophy and Psychology, 48–93. Oxford: Oxford University Press. Mylopoulos, Myrto and Shepherd, Joshua.2020. “Agentive Phenomenology.” In U. Kriegel (ed.), Oxford Handbook of the Philosophy of Consciousness. Oxford: Oxford University Press. Nanay, Bence. 2011. “Do We See Apples as Edible?” Pacific Philosophical Quarterly 92:305–322. Nichols, Shaun. 2006. “Imaginative Blocks and Impossibility.” In S. Nichols (ed.), The Architecture of the Imagination, 205–236. Oxford: Oxford University Press. Nolan, Daniel. 2017. “Naturalised Modal Epistemology.” In F. Leon and B. Fischer (eds.), Modal Epistemology after Rationalism, 7–27. Cham: Springer. Roca-Royes, Sonia. 2017. “Similarity and Possibility: An Epistemology of de re Possibility for Concrete Entities.” In B. Fischer and F. Leon (eds.), Modal Epistemology after Rationalism, 221–245. Cham: Springer. Roessler, Johannes and Eilan, Naomi (eds.). 2003. Agency and Self- Awareness. Issues in Philosophy and Psychology. Oxford: Oxford University Press. Scarantino, Andrea. 2003. “Affordances Explained.” Philosophy of Science 70:949–961. Schrenk, Markus. 2014. “Die Erfahrung der Widerst¨andigkeit der Welt als Wahrnehmung kausaler Kraft.” In A.-S. Spann and D. Wehinger (eds.). Vermögen und Handlung. Der dispositionale Realismus und unser Selbstverst¨andnis als Handelnde, 23–62. Münster: Mentis. Siegel, Susanna. 2005. “The Phenomenology of Efficacy.” Philosophical Topics 33:265–284. Strohminger, Margot. 2015. “Perceptual Knowledge of Nonactual Possibilities.” Philosophical Perspectives 29:363–375. Thomasson, Amie. 2018. “How Can We Come to Know Metaphysical Modal Truths?” Synthese 198 (Suppl 8):2077–2106. Vaidya, Anand Jayprakash. 2010. “Understanding and Essence.” Philosophia 38:811–833. Vaidya, Anand Jayprakash. 2015. “The Epistemology of Modality.” In The Stanford Encyclopedia of Philosophy, Summer 2015 edition. http://plato.­stanford. edu/archives/sum2015/entries/modality-epistemology/. Vaidya, Anand Jayprakash and Wallner, Michael. 2021. “Modal Epistemology and the Problem of Modal Epistemic Friction.” Synthese 198 (Suppl 8), 1909–1935. van Inwagen, Peter. 1998. “Modal Epistemology.” Philosophical Studies 92:67–84. Vetter, Barbara. 2013. “Multi-Track Dispositions.” The Philosophical Quarterly 63:330–352. Vetter, Barbara. 2014. “Dispositions without Conditionals.” Mind 123:129–156. Vetter, Barbara. 2015. Potentiality: From Dispositions to Modality. Oxford: Oxford University Press.

An Agency-Based Epistemology of Modality  69 Vetter, Barbara. 2016. “Williamsonian Modal Epistemology, Possibility-Based.” Canadian Journal of Philosophy 46:766–795. Vetter, Barbara. 2019. “Are Abilities Dispositions?” Synthese 196:201–220. Vetter, Barbara. 2020. “Perceiving Potentiality: A Metaphysics for Affordances.” Topoi 39 (5):1177–1191. Vetter, Barbara. 2021. “Essence, Potentiality, and Modality.” Mind 130 (519):833–861. Viebahn, Emanuel and Vetter, Barbara. 2016. “How Many Meanings for ‘May’? The Case for Modal Polysemy.” Philosophers’ Imprint 16:1–26. Vihvelin, Kadri. 2013. Causes, Laws, and Free Will. Why Determinism Doesn’t Matter. Oxford/New York: Oxford University Press. Williamson, Timothy. 2007. The Philosophy of Philosophy. Oxford: Blackwell. Williamson, Timothy. 2016a. “Knowing by Imagining.” In A. Kind and P. Kung (eds.), Knowledge Through Imagination, 113–123. Oxford: Oxford University Press. Williamson, Timothy. 2016b. “Modal Science.” Canadian Journal of Philosophy 46:453–492. Wirling, Ylwa Sjölin. 2020. “Non-Uniformism about the Epistemology of Modality: Strong and Weak.” Analytic Philosophy 61 (2):152–173.

3 The Price of Sensitivity Rebecca Hanrahan

Most philosophers, especially modal realists, endorse the ­insensitivity counterfactual. They hold that the experiences we would have if a proposition were possibly true are no different from the experiences we would have if this same proposition were necessarily true, and vice versa. ­Consider proposition G. G: Rebecca is wearing a green shirt. G is possibly true, if it is true in some possible world. Since G is true in this world, we know that it is possibly true. G would be necessarily true if it were true in all possible worlds. So, what makes G necessarily true has to do with worlds other than this one. Yet, this world is the only world we experience. Thus, if G went from being possibly true to being necessarily true, we seemingly wouldn’t be any the wiser. For everything would stay the same in this world; the changes would occur in other worlds. But there are some philosophers who reject this bit of received wisdom. Carrie Jenkins (2010) in “Concepts, Experiences, and Modal K ­ nowledge,” and Nenad Miščević (2003) in “Explaining Modal I­ntuition,” argue instead for the sensitivity counterfactual. On this opposing counterfactual, the experiences we would have if a proposition were possibly true are different from the experiences we would have if this same proposition were necessarily true. My goal here is to define the costs of endorsing one of these counterfactuals over the other. On the one hand, I argue that accepting the sensitivity counterfactual results in the collapse of the modal. Most philosophers think that the scope of what is or is not possible1 changes depending on what notion of possibility is in play. Thus, something could be conceptually possible (like Philosophical Zombies), even though it is a physical or even a metaphysical impossibility. But if Jenkins is right, I argue that many of these differences in scope disappear. Specifically, I argue that, on her position:

DOI: 10.4324/9781003002192-5

The Price of Sensitivity  71 All conceptual possibilities are physical and hence metaphysical possibilities. All metaphysical and all physical possibilities are conceptual ­possibilities and hence in principle conceivable. This is what I refer to as the collapse of the modal. On the other hand, I argue that if we instead opt for the insensitivity counterfactual, we avoid this collapse, but we have to concede that what is conceptually possible is distinct from what is either metaphysically or physically possible. And either concession will come with a cost. We will either have to countenance categorical properties that lack causal powers or we will have to countenance a notion of strong metaphysical possibility that limits what can be instantiated in any possible world. Finally, I will argue that if a price has to be paid, it is best that we pay it by endorsing strong metaphysical possibilities, for doing so will preserve more of our modal intuitions. In what follows, I will be concentrating on Jenkins’s position. Jenkins and Miščević both argue that the actual world contains modal properties, and it is our sensitivity to these properties that informs us as to what is and is not possible. Hence, if the modal status of a proposition changed, we would be aware of the changes in these properties. Thus, their positions are quite similar. But Jenkins’s view is more developed than Miščević’s, and unlike his, is not limited to those worlds close to this world. Although I shall focus on Jenkins’s position, I take my arguments to apply more broadly to any view that endorses the sensitivity counterfactual. To understand her position, we must begin with her conception of concepts, so let’s begin.2, 3

Jenkins on Concepts According to Jenkins, our senses are sensitive to the individuals, properties, and relations that compose this world, and our concepts are sensitive to this sensory input. It is because of these two stages of sensitivity that Jenkins holds that our concepts are “epistemically grounded” (Jenkins 2010, 264). They are “relevantly accurate” and “non-accidental” pictures of the world (ibid.). They are “non-accidental” because these concepts are products of our reliably functioning sensory faculties. And they are “accurate” because they mirror facts of the world. Conceiving, for Jenkins, is limited by our concepts. I can conceive of a “blue cat,” because nothing about the concepts of blue or cat precludes this combination (Jenkins 2010, 263). But I can’t conceive of a cat that is a plant because “my concepts will not allow it” (ibid.). Because what I can conceive is limited by my concepts, via conceiving, I can discover the “structural relations” that hold between my concepts (Jenkins 2010, 268). Moreover, because these concepts are accurate and reliable pictures

72  Rebecca Hanrahan of the world, the relationships I discover will mirror the relationships found in the mind-independent world. So, by examining the relationships among our concepts, we can learn something about how the properties these concepts pick out are related to each other in the mind-independent world. And importantly, the information we recover will pertain not just to this world but to possible worlds as well. An example will help. My concepts of vixen and female fox mirror these properties as they are manifested in this world. Thus, I can grasp the relationship between these properties by grasping the relationship between these concepts. What I grasp is that the property of vixen “includes” the property of being a female fox (Jenkins 2010, 269). And understanding this enables me to conclude that there is no possible world in which there is a vixen that isn’t a female fox. So, by understanding how my concepts are related to each other, I can determine how the properties at issue are, could be, or even must be related to each other. In this way, we can see that for Jenkins conceivability is a guide to possibility. We can also see why Jenkins rejects the insensitivity counterfactual. For Jenkins, what justifies me in concluding that p is either necessarily or possibly true is my knowledge of the properties of this world. And this knowledge is gathered through an examination of my concepts, while these concepts are themselves the products of particular sensory experiences. Thus, for any proposition p, “if the modal status were different, then even if p’s truth value was the same, our sensory input would be different” (Jenkins 2010, 272). One question to raise at this point is how, according to Jenkins, modal facts are embedded in this world such that we can be sensitive to them. Jenkins wants to remain neutral on this question. But she sees at least two possibilities. It could be that modal facts depend on the non-modal properties of this world. For example, modal facts might depend on essences, and our sensitivity to a thing’s essence would then inform us of these modal facts. Alternatively, we might take a Lewisian approach and hold that while modal facts don’t depend on non-modal properties, the non-modal properties of this world are a guide to what is the case in other possible worlds. It might be, to give a simple example, that if the actual world is structured in such a way that the property of vixenhood includes that of females, then there is no possible world at which there is a non-female vixen, although this fact about worlds is not metaphysically reducible to, or otherwise dependent upon, the corresponding structural fact about our own world. (2010, 269–270) The problem with this second option is that it undermines one of ­Jenkins’s goals in this paper. She wants to resolve the integration challenge. That is,

The Price of Sensitivity  73 she wants to offer up a theory of modal epistemology that coheres with her modal ontology. And with the rejection of the insensitivity counterfactual, we can see how she intends to resolve this challenge. To know what is possible or necessary is to be sensitive to facts about the actual world, facts that are captured by way of our concepts. In so far as these facts are captured by way of our concepts and conceivability involves the examination of these concepts, our metaphysics is for Jenkins seemingly integrated with our epistemology. But this is a solution to the integration challenge only if we have reason to take these facts about the actual world to be revealing about possible worlds. But without there being a metaphysical connection of some kind between the structure of this world and the structure of possible worlds, the fact that what is is a guide to what could be becomes a mere brute fact. No explanation is available to us as to why our experiences of the actual world, and hence our concepts, are a guide to what is the case in possible worlds. Jenkins would object to this, for she does not want to rule out “a satisfying explanation of the correlation which does not mention metaphysical dependence at all” (Jenkins 2010, 27). But she declines to offer such an explanation. I have to say that I fail to see how such an explanation could be satisfying. Only an explanation that posited a metaphysical connection would meet the integration challenge. So that leaves us with the first option: modal facts are metaphysically dependent in some way on some non-modal properties to which we are sensitive. At this stage, it is right to wonder how and how much modal information can be embedded in our concepts by way of sensory experiences. And it is easy to trot out a series of objections inspired by Hume. But let us put aside these worries.4 Instead, let us assume that our perceptual faculties are sensitive to the facts of the actual world and that our concepts reflect this sensory input, so that we can consider our concepts as accurate pictures of this world. And let us also assume a metaphysical dependence between modal and non-modal facts. Specifically, let us also assume that somehow what these concepts encode are structural relations (like inclusion) that can support or justify our modal claims. Thus, our experiences would be different if the modal status of a proposition were different; hence, the sensitivity counterfactual is true. What are the costs of adopting this view? In what follows, I will argue that to reject the insensitivity counterfactual, you must accept that metaphysical, physical, and conceptual possibilities all share the same extension. But before arguing for this collapse of the modal, I want to discuss a different but related cost of endorsing this view.

Churches and Q-Churches To review, the sensitivity counterfactual holds that if the modal status of a proposition was different but yet its truth value stayed the same, our experiences would be different. But given the relationship Jenkins draws

74  Rebecca Hanrahan between concepts and experiences, a difference in experiences would also result in a difference in concepts. Consider proposition C. C: There is a church outside my window. If C had a different modal status, no longer would we be considering churches and windows but, as Jenkins acknowledges, Q-churches and Q-windows. And if this is the case, then the proposition at issue would be different as well. The proposition that is true if C is necessarily true isn’t “There is a church outside my window”; it is rather “There is a Q-church outside my Q-window.” But if the identity of a proposition differs depending on its modal status, then inquiring into a proposition’s modal status becomes impossible. Consider Z. Z: There are Philosophical Zombies. We don’t know the modal status of Z. But, on Jenkins’s account, there is a different proposition associated with Z if Z is necessarily true vs. possibly true. As with C, the conditions that make Z possibly true aren’t the same conditions that make Z necessarily true. But if this is the case, when seeking to determine Z’s modal status, it will be unclear which proposition is under investigation. But if we don’t know which proposition is under investigation, we then start having problems with the truth. The consensus at this time is that Z is false with respect to this world. But if we don’t know Z’s modal status, and hence we don’t know which proposition is at issue, then we seemingly can’t determine Z’s truth status. Our intuition is that we can judge a proposition’s truth status without knowing its modal status. But on ­Jenkins’s view, a proposition’s content is determined in part by its modal status. Thus, if we don’t know a proposition’s modal status, we can’t know its content and if we don’t know its content, we can’t adjudicate whether or not it is true. To this, someone might claim that while we don’t know which of the two propositions are at issue when it comes to establishing Z’s modal status, we could just consider both. But Jenkins claims that we have no words to articulate the difference between churches and Q-churches. She says: The differences in our sensory input would not be of an everyday kind, comparable to the difference between its looking like there’s a church outside the window and its not looking like there’s a church outside the window. Rather, the difference might manifest in the structure of all sorts of daily sensory interactions with the world. It is very hard to say in much detail what our sensory input would be like if it grounded different concepts. The kinds of concepts that would describe this scenario well, and would provide us with good

The Price of Sensitivity  75 ways of describing the sensory input we would get, are ones I do not have available to express here and ones which you probably wouldn’t be able to make sense of even if I did manage to express them. (Jenkins 2010, 273) But if we have no words to articulate the differences in this case, a proposition whose truth and modal status is known, aren’t we totally out to sea when it comes to Z, a proposition whose modal status is unknown? Jenkins does have a response to this set of objections. She argues that Q-churches and churches are counterparts of each other. She contends that the concept of a Q-church is “in significant ways similar to” our concept of a church in so far as both “play similar roles in our representation of our world” (Jenkins 2010, fn. 27). But I don’t see how they can be counterparts of each other on Jenkins’s theory. Namely, concepts are, according to Jenkins, “map-like” (Jenkins 2010, fn 18). That is, they convey meaning and information through the relationships they are in. But if this is the case, the concept of a Q-church would be in different relationships than the relationships our concept of church is in, so different that they would constrain us from being able to conceive of a world in which there isn’t a Q-church out my Q-window. So, the concept of church and Q-church must be “similar in significant ways” so that one can be a counterpart of the other, but yet these same concepts must be “different” so that one would ground a necessity claim, while the other wouldn’t (Jenkins 2010, 272, fn. 27). But how can there be the space to do both? Now, if things were individuated by their intrinsic properties, then churches and Q-churches could share intrinsic properties but differ with respect to their extrinsic properties. Hence, that they shared intrinsic properties would enable churches and Q-churches to be counterparts of each other, but yet their extrinsic properties could ground different modal relationships. But this response wouldn’t be an option for Jenkins, for if this division was allowed, the insensitivity counterfactual could still hold. I could be sensitive to the intrinsic properties that churches and Q-churches share but insensitive to those extrinsic relations by virtue of which Q-churches must necessarily be outside my Q-window. Moreover, if our concepts mirrored only intrinsic properties, we couldn’t read off from our concepts relations of necessity and possibility. Thus, conceivability couldn’t be a guide to possibility. So taking this tack would undermine Jenkins’s conclusion. Thus, as long as on Jenkins’s theory the identity of the proposition at issue is determined by the modal status of that proposition, epistemology is in a hopeless muddle. We seemingly can’t determine the modal status of that proposition or even which propositions are true. And the only way to avoid such consequences is via a solution that compels us to endorse the insensitivity counterfactual. While I want to largely put these objections aside in what follows, aspects of the above discussion will be relevant as we turn to the collapse of the modal.

76  Rebecca Hanrahan

Collapsing Possibilities As has already been noted, Jenkins contends that how our concepts are related to each other mirrors the “structural relations” between actual things (i.e., “individuals, properties, relations and whatever else there is in the world”; Jenkins 2010, 268, 267). Given this, if p is conceivable, then nothing about the structural relations associated with p preclude p from being true. But if this is the case, then if p is conceivable and hence a conceptually possibility, p must be possible. But what notion of possibility is at play here? Consider again that for Jenkins what concepts I have are a product of what sensory input I receive and what sensory input I receive is a product of the structure of this world. Thus, if a proposition is conceivable, it will reflect what is physically possible in this world. Now all physical possibilities are also metaphysical possibilities. Hence, if a proposition is conceptually possible, it will be a physical as well as metaphysical possibility. None of this should be particularly troubling to Jenkins or really anyone. But what about the reverse direction of this relationship? If p is a metaphysical or a physical possibility, must it also be a conceptual possibility? Must it also be (at least in principle) conceivable? Here, it is worth thinking through the implications of Jenkins’s endorsement of the sensitivity counterfactual. For Jenkins, what concepts we have are a product of the kinds of sensory input we receive. So, if our sensory experiences were different, our concepts would be different as well. Thus, if we are sensitive to differences in modal status, then it seems at least in principle that there are no physical or metaphysical possibilities that also aren’t conceptual possibilities. To see this, consider what would be the case if the opposite were true. Say there was either a physical or a metaphysical possibility that wasn’t as well a conceptual possibility. In such a situation, we couldn’t have the experiences and hence the concepts to describe this possibility. And none of the concepts we did or could have would preclude or allow for this possibility. (If either of these two options held, the possibility at issue would in fact be a conceptual possibility, contrary to the assumption just made above.) Thus, if the modal status of the proposition at issue were different, we couldn’t be sensitive to this difference. Which would mean, counter to Jenkins’s position, that the sensitivity counterfactual would be false. There is another way of getting at this same point. Here return to the difference between churches and Q-churches and since we are talking about churches, let’s also talk about God. While it is agreed that proposition C C: There is a church outside my window is merely possibly true, on Jenkins’s view this proposition couldn’t be necessarily true. A church can’t be a church and yet be outside my window

The Price of Sensitivity  77 in every possible world. So, God can’t make it so that there is a church outside my window in every world in which I have a window. If God is going to worlds installing churches outside my window, she has to stop when she comes to that last church so to make sure that the churches she has installed maintain their identities as churches. For if she puts that last church in, all the churches she has installed outside my window will stop being churches and start being Q-churches. Those who think that a church can be a church and yet be outside my window in every possible world in which I have a window must acknowledge that there could be some limit on modal space that works separately from what is allowed by our concepts. That’s what the talk of God is doing here. Those who think that God could put a church, and not a Q-church, outside my window in every world are conceding that the space of what is or is not possible and/or necessary could be defined by something other than what our concepts allow or preclude. They are endorsing what Chalmers and others have referred to as “strong metaphysical possibility” (Chalmers 1996, 137). Metaphysical possibility has often been defined as a conjunction of what is logically and conceptually possible. But some reject this notion and hold instead that “there are fewer metaphysically possible worlds than there are logically possible worlds, and the a posteriori necessity of certain statements can stem from factors quite independent of the semantics of the terms involved” (ibid). All such possibilities are both logically and conceptually possible but not all logical and conceptual possibilities are strong metaphysically possibilities. And importantly, the status of a proposition as strongly metaphysically necessary isn’t dependent on the terms involved rigidly designating. Thus, nothing in our concepts or logic requires that a church be outside my window, but might there be something else, for instance, God, that does or could?5 Jenkins has to answer this question in the negative. Jenkins can’t acknowledge such a limit. She can’t allow for strong metaphysical possibility. For if she did, the modal status of C could be different without our experiences or concepts being different. God could erect that last church in some possible world and the experiences I would have of the structure outside my window would stay the same. But if our experiences could stay the same while C’s modal status changed, the insensitivity rather than the sensitivity counterfactual would hold true. So, if there are propositions whose modal status was dependent on something other than what our concepts allow or preclude and hence on something other than our experiences, then the modal status of these propositions could be different and we wouldn’t be sensitive to these differences. Jenkins can’t allow for this. Thus, she has to reject strong metaphysical possibility and she has to hold not just that all conceptual possibilities are physical and hence metaphysical possibilities but that all physical as well as all metaphysical possibilities are conceptual

78  Rebecca Hanrahan possibilities. And she has to do so as a consequence of her endorsing the sensitivity counterfactual. Here then we have the collapse of the modal. Jenkins does seemingly have a way of avoiding this admission. Consider how she articulates the sensitivity counterfactual. “For propositions P whose modal status we know, if the modal status of P were different, then even if P’s truth value was the same, our sensory input would be different” (Jenkins 2010, 259). Notice that she limits this counterfactual to those propositions whose modal status is known. This allows for properties of the world that we aren’t sensitive to and hence certain propositions whose modal status is unknown to us. If there are, then there could be physical and hence metaphysical possibilities that are not within the scope of what is conceptually possible. And if that is so, then there would be no collapse of the modal. But this qualification is problematic. It prevents the collapse only if our inability to know the modal status of the proposition at issue is an in-principle barrier. If it is merely a barrier in practice, the collapse still occurs. Moreover, if it is a problem in principle, then though the collapse is avoided, troubling epistemic and metaphysical consequences follow. Let’s deal with each horn of this dilemma, in turn. Say our inability to know the modal status of a proposition is a problem in practice. While on Jenkins’s theory our concepts are for the most part “fitting,” the wheels can go off the track (Jenkins 2010, 263). Specifically, Jenkins concedes, our senses might fail to register correctly some fact of the world. Or while we might be sensitive to some fact of the world, our concepts might not correctly capture this input. Or finally, we might incorrectly relate certain concepts to each other. As a consequence, we will at times be mistaken in our modal beliefs. What is metaphysically possible will at times seem to us to be conceptually impossible. What we deem conceptually possible will at times be metaphysically impossible.6 But assume these mistakes are due to negligence on our part. Once this negligence is corrected for, there will be an alignment between what is conceptually possible and what is physically and hence metaphysically possible. There will be no physical and hence metaphysical possibility associated with this world that isn’t as well for us a conceptual possibility. But say our inability to know the modal status of a proposition is a problem in principle. Say there are properties associated with this world that are in principle unknowable to us. Then, there will be physical and hence metaphysical possibilities that are not also conceptual possibilities. But what is the relationship between these unknowable properties and the properties we are sensitive to? For Jenkins, again, our concepts are map-like. The meaning and information they convey are a consequence of the relationships they are in. But if our concepts are to mirror properties of the world, then the identity of a property must be defined in whole or in part by the relationship it is in with other properties. So, say these unknowable properties are in relationship with those properties we are sensitive to. Then, we have amplified our earlier problems

The Price of Sensitivity  79 about identifying the proposition under discussion. If the identity of the ­properties we are sensitive to is determined by the relationships they stand in, and yet we can’t know about all the relationships they stand in, even these properties become unknowable and their associated concepts equally unreliable. Suppose instead these unknowable properties are not in any relationship with the ones that we are sensitive to. On the bright side, the worry I just identified will not arise. For those properties we are sensitive to can in principle be known by us and the associated concepts will be accurate representations of those properties. But we are then countenancing a set of properties that are causally isolated from those properties we are sensitive to. This is clearly problematic and also largely vacuous. On this position, though there may be physical and hence metaphysical possibilities that aren’t conceptual possibilities, all those physical and hence metaphysical possibilities that are in any way relevant to us will be conceptual possibilities. In the knowable world, the collapse still occurs. And obviously, a whole set of causally isolated properties, whose presence or absence makes no difference at all to even one of the properties we can perceive, is a metaphysical extravagance.7

Accepting Insensitivity So, where do we stand? Jenkins argues that our concepts are grounded in sensory input that mirrors actual properties and their relations. As a consequence, we can justify our modal beliefs by considering what our concepts allow and preclude. If this is correct, then the sensitivity counterfactual must hold. If we can justify our modal beliefs by considering our epistemically grounded concepts, then if the modal status of a proposition was different, we would then have different experiences. I have argued that a consequence of this view is the collapse of the modal. Unless we want to countenance causally isolated metaphysically extravagant properties or the hopeless muddling of our modal epistemology, anything that is conceptually possible must also be physically and metaphysically possible, and anything that is metaphysically or physically possible must also be conceptually possible. For some (but not for all), this consequence is troubling. For if the modal collapses then this world defines the boundaries of what is and what is not possible. Here, it is worth considering how confining this boundary is on Jenkins’s theory. Consider cats. Our experiences of cats have been of earthen-toned creatures; brown, black, white, orange, and red; even Russian and Egyptian Blues are in fact gray. If our concepts are to mirror our experiences, then contrary to Jenkins’s claim it seems as if we can’t in fact conceive of a blue cat. We can picture to ourselves a blue feline-like beast, but our concept of a cat should preclude us from taking it to be a cat, for cats aren’t blue. Thus, there is a good deal that we have taken to be conceivable

80  Rebecca Hanrahan (blue cats, talking horses, dancing hippos) that is not conceivable, and if what is conceptually possible defines the limit of what is metaphysically possible, there is also a lot less that is possible on this theory.8 Those who do find this collapse troubling can of course backup and seek to preserve the insensitivity counterfactual. But just as there are costs to accepting the sensitivity counterfactual, there are costs to accepting the insensitivity counterfactual. So, what if, contrary to Jenkins’s view, we endorse the insensitivity counterfactual? What then follows about conceptual, metaphysical, and physical possibilities? To review, this counterfactual holds that if the modal status of p was different but p’s truth value stayed the same, our experiences of p would not be different. If our sensory experiences would not be different while the truth value of p holds steady, then there is no difference in the concepts at play between when p is possibly true vs. when p is necessarily true. And if the concepts aren’t different, then neither is the proposition. But if the concepts and hence the proposition hold steady across a difference in modal status, the source of the difference between when p is necessarily true vs. when p is possibly true has to be located in something other than our concepts. Thus, if the insensitivity counterfactual holds, what is conceptually possible must be distinct from what is either metaphysically or physically possible. And no matter whether you hold conceptual possibility to be distinct from physical possibility or instead distinct from metaphysical possibility, there will be a price to pay. To see this, let’s describe one way to distinguish conceptual possibility from physical possibility via a theory that countenances a specific notion of categorical properties. On this theory, categorical properties have their identity independently of their causal powers. The causal relations they figure in are fixed by the laws of nature that happen to obtain in the world(s) in which they are instantiated. So our experiences of these properties might be the same (after all, the properties are the same), even when the modal status of propositions referring to them differs. A simple example will help here. Salt and water each possess categorical properties that distinguish these substances as the kinds of substances they are. But on this theory, the categorical properties of a substance do not determine how that substance interacts with others; instead, it is the laws of nature instantiated in the world in question that determines these interactions. In this world, the laws at issue demand that salt dissolve in water; in other worlds, it does not. But what it is to be water and what it is to be salt are determined by their respective categorical properties and hence remain constant across worlds. Take proposition D. D: Salt dissolves in water. Whether or not this proposition is necessarily true vs. possibly true would depend on which laws held. Say there was a set of laws (Set-P) that

The Price of Sensitivity  81 made it such that D was merely possibly true, but as luck would have it consistently true in this world. Say there was another set of laws (Set-N) that made D necessarily true. From our perspective, the experiences associated with D’s being true when Set-P held would be exactly like the experiences we would have if Set-N held instead. Hence, if this theory of categorical properties is correct, then the insensitivity counterfactual would hold. But there are two problems with this theory. First, there is something troubling about what Helen Beebee has called the “governing” concept of laws (Beebee 2000, 573).9 It is difficult to imagine how two worlds could be indistinguishable with respect to their categorical properties but yet be governed by different laws. Even if we waive this difficulty, however, another remains. Sydney Shoemaker (1980) argues that properties are, like kinds, individuated by their causal powers. If so, then it is not the case that a property can retain its identity across worlds where different laws of nature obtain, at least not when the change in laws affects what the properties can do. So, we have a way of preserving the insensitivity counterfactual that, while not obviously false, is unappealing, for the identity of properties swings free from what those properties do. To avoid this problem, properties must be defined in terms of their causal powers. So, what makes a substance the kind of substance it is is its causal properties and these causal properties are described – rather than governed – by laws that detail how different substances are related to each other. On this view, the fact that salt dissolves in water is what makes salt the kind of thing it is, namely, salt. Thus, there is no world in which there is salt and yet it doesn’t dissolve in water. On this view, if concepts are to mirror properties and properties are differentiated by causal powers, there is no way to distinguish what is conceptually possible from what is physically possible. But what then happens to the insensitivity counterfactual on such a view? Do we have to conclude, with Jenkins, that it is false? We don’t, but only if we accede to a notion of strong metaphysical possibility. As I noted earlier, metaphysical possibility has often been defined as a conjunction of what is logically and conceptually possible. But again, some recognize strong metaphysical possibility. On this view, there is what our laws of logic allow and exclude, and this defines what is logically possible. And if concepts are still to mirror properties, then there will still be a collapse between conceptual possibility and what is physically possible. But what is metaphysically possible doesn’t reduce to a conjunction of the logical and the conceptual. There are some logical and conceptual possibilities that aren’t strong metaphysical possibilities. I think of strong metaphysical possibility in terms of instantiation. Consider tigers. To be a tiger, let’s say, an individual has to be striped. But nothing about our concepts or our logic mandate how many

82  Rebecca Hanrahan stripes an individual tiger will have. And yet there are no tigers with an ­indeterminate number of stripes. If this is so, we have a truth within the realm of what is or isn’t metaphysically possible that doesn’t reduce to a truth found within the realm of what is either logically or conceptually possible. Nor does it depend on rigid designation. More broadly, this truth is an instance of the principle that I: All determinable types have to be made determinate when a token of that type is instantiated. This principle narrows the space of what is possible; there can be no indeterminately striped tigers. If this is so, we have here a strong metaphysical necessity. Of course, objections can be raised to this notion of strong metaphysical possibility. But let’s put them aside and assume that there is such a thing as a strong metaphysical possibility that is distinct from conceptual and logical possibilities. If there is, then we can again make sense of the insensitivity counterfactual. Consider again proposition C. While nothing in our concepts of church and window require that there be a church outside my window, there could be some limits to how things get instantiated in possible worlds that result in there being a church outside my window in every possible world. (Enter again God. Maybe God makes this so.) Importantly, my experiences of the actual world would not change, if this limit were in place as compared to if it wasn’t. And if this is so, then the insensitivity counterfactual holds true. So, the price of buying into the sensitivity counterfactual is our collapsing conceptual, physical, and metaphysical possibilities into each other. But if this is correct, the price of the insensitivity counterfactual is that we must distinguish one of these possibilities from the other. What I have done above is identify two ways of making this distinction and the cost associated with each. The first involves our countenancing categorical properties. That is, it involves our believing that there are inert properties that nonetheless differentiate their possessors from each other. The second way involves our countenancing a notion of strong metaphysical possibility. That is, it involves our believing in some limit to what is possible that is not defined by our logic, concepts, or laws of nature and in fact can best be illustrated with a reference to God.

Conclusion Endorsing the sensitivity counterfactual runs counter to our intuitions about modality. What makes a proposition necessarily true are facts about possible worlds other than our own; thus of course, we would not be sensitive to any difference in a proposition’s modal status. So as ­Jenkins notes, we accept the insensitivity counterfactual without argument. But as counterintuitive as the sensitivity counterfactual is, the

The Price of Sensitivity  83 insensitivity counterfactual comes along with its own counterintuitive dimensions that we largely ignore because this counterfactual is so intuitively plausible. If the insensitivity counterfactual holds true, we must countenance either inert categorical properties or strong metaphysical possibility. Going down the route of inert categorical properties is problematic, as most philosophers would agree. So, that leaves us with only two options. Endorse the sensitivity counterfactual and collapse the modal or endorse the insensitivity counterfactual and endorse strong metaphysical possibility. That our senses are the ultimate source of knowledge about what is the case as well as what could be and what must be the case gives our senses too much credit. This option seems as unappealing as the route of inert causal properties. Thus, we seem forced to endorse the insensitivity counterfactual coupled with the recognition of strong metaphysical possibility. By going this route, we can preserve the intuition that there could be a church, and not a Q-church, outside my window in every possible world. And, relatedly, it allows for the identity of our propositions to be maintained under a change of modal status. But this view, like all the others, comes with cost. And one cost is epistemological in nature. We are trapped in this world, observing what is the case. And while what is the case might be a function of both what our concepts allow and how those various concepts get instantiated in this world, from our perspective, we seldom (if ever) are able to distinguish which of these two are doing the work when it comes to determining what is possible and what is necessary. Let’s say it is necessarily true that salt must dissolve in water. Is this because salt can’t be salt unless it dissolves in water? Or is it how salt gets instantiated that prevents there being a world in which salt doesn’t dissolve? There is seemingly no way we can tell between these two options. The other cost is metaphysical. Return to the example of God making it the case that there be a church outside of my window in every world that I have a window. Nothing in our logic or our concepts requires that there be a church outside my window. But couldn’t there be something else that shapes modal space? Couldn’t there be something that makes it necessary that there be a church outside my window? The problem is that in this example, that something is God. God requires that there be a church outside my window in every world. But God hides lots of the sins of this story.10 Why would God require that there be a church outside my window? If there is a reason and that reason is located in the nature of churches or windows, then we no longer have a strong metaphysical necessity here. If there isn’t a reason, then what is or is not a strong metaphysical necessity is a mere brute fact. Alternatively, there could be some requirements (by which God must abide) that define modal space, that aren’t brute facts and that aren’t captured by either our logic, laws, or concepts. But we have no idea what such requirements could be. Worse, they seem beyond our ken, hence the epistemological problem identified above.11

84  Rebecca Hanrahan Thus, accepting either counterfactual comes with serious costs. Given this, we could, of course, abandon both. After all, both counterfactuals are trivially true in so far as both are composed of an impossible antecedent. Jenkins, herself, suggests this option. Quoting Jenkins: Although I have been happy to consider the insensitivity counterfactual seriously as a threat in this paper, I would be equally happy if it were urged by a proponent of the current objection that the best thing to do is simply to say that talking about these kinds of counterfactuals just isn’t helpful for the purposes of understanding what sensitivity to experience amounts to, and get on with expounding modal empiricism in other ways. (2010, fn. 27) But while abandoning these counterfactuals is an option, the issues that arise as a consequence of our considering them don’t go away. Specifically, how should we consider the relations between our various different notions of possibility? Do all collapse into each other? Or can a notion of metaphysical possibility be separated from what our logic, concepts, and laws allow? But these questions are best left for another paper.12

Notes 1 Or necessary. 2 Some will argue that consideration of either counterfactual is a fool’s errand for they are both trivially true in so far as both are composed of an impossible antecedent. To this objection, I will follow Jenkins’s lead. She argues that “even a proponent of the traditional semantics can appreciate that some counterpossible conditionals are non-trivially interesting as well as trivially true” (Jenkins 2010, 258). For there might be reasons to accept such a conditional as true, despite the fact of its triviality. Such might be the case when it comes to one of these counterfactuals; hence, they are worthy of our consideration. 3 Jenkins articulates the sensitivity counterfactual in the following way. “For propositions P whose modal status we know, if the modal status of P were different, then even if P’s truth value was the same, our sensory input would be different” (Jenkins 2010, 259). I will discuss her version of this principle on pages 13–15 and there I will concentrate on the consequences of her limiting this counterfactual to those propositions whose modal status is known. 4 If you can’t set aside such objections, consider that for Jenkins it is the “inclusion-like relationship” that prevents us from conceiving of a situation in which vixens aren’t female, and hence it is our failure to conceive that grounds our belief that all vixens are necessarily female (Jenkins 2010, 268). But how do I learn that such a relationship holds? It can’t be via conceiving for it is my knowledge of this relationship that prevents me from conceiving the non-female vixen. So my knowledge of this relationship must precede my conceiving. Given Jenkins’s theory, she must argue that it is via the senses that I learn of this relationship, but my senses only provide me with experiences of the constant conjunction between femaleness and vixenhood. So how then do I learn of this inclusion-like relationship?

The Price of Sensitivity  85 5 In “Conceptual Truths, Strong Possibilities and Our Knowledge of Metaphysical Necessities,” Nimtz presents a conception of both strong metaphysical necessity (SMN) and strong metaphysical possibility (SMP). Nimtz’s position is that while there are no SMPs, he doesn’t dismiss the possibility of there being SMNs, though he has his doubts. Nimtz and I share the same conception of SMNs, where we differ is in our conception of SMPs. For Nimtz, S is a strong metaphysical necessity if and only if S is true in all counterfactual worlds but not true in all counteractual worlds. That is, S is an SMN iff “□CS + ~□AS” (Nimtz 2012, 50). Put in another way, S is an a posteriori necessity whose necessity doesn’t arise as a consequence of its terms rigidly designating. Hence, we will “judge” this proposition to be false in some possible worlds (ibid.). Consider the stripes on a tiger. Nothing in our concepts or logic demands that a tiger has a determinate number of stripes; thus, an indeterminately striped tiger isn’t conceptually incoherent. But no tiger can have an indeterminate number of stripes and yet this isn’t due to any of the associated terms rigidly designating. Nimtz holds that S is a strong metaphysical possibility if and only if S is true in all counteractual worlds but yet not true in all counterfactual worlds. That is, S is SMP iff “□AS + ~□CS” (Nimtz 2012, 51). Now, for Nimtz, if S is true in all counteractual worlds, then S is a conceptual necessity. But if this is the case, Nimtz argues, then to countenance a strong metaphysical possibility is to hold “that some conceptual impossibilities are metaphysically possible” (ibid.). To see this, assume that S is a strong metaphysical possibility. If this is so, then there is no counteractual world in which ~S is true, thus ~S is conceptual impossibility. But yet in some counterfactual world, ~S is the case, for S is not true in all counterfactual worlds. Thus, ~S is a conceptual impossibility that is nonetheless true in some counterfactual world. In so far as this is clearly abhorrent, Nimtz rejects SMPs. I do agree that as Nimtz defines, SMPs there can be no such SMPs, for the very reasons he gives. But I can’t at this point endorse Nimtz’s definition of SMPs. Again, for Nimtz if S is an SMP iff □AS + ~□CS. But… If S is an SMP as opposed to an SMN, then S is not an SMN. If S is not an SMN, then ~(□CS + ~□AS). ~(□CS + ~□AS) is logically equivalent to (~□CS v □AS). Thus, if S is an SMP as opposed to an SMN, then ~□CS v □AS. Or in other words, if S is an SMP as opposed to an SMN, then either S isn’t true in every counterfactual world or S is true in every counteractual world. Let’s assume this disjunctive condition is merely a necessary condition on being an SMP and not a sufficient condition. So, there is more to being an SMP than satisfying this disjunction. If this is correct, I don’t understand how Nimtz comes to define SMP via a conjunctive condition, that is as (□AS + ~□CS). Until Nimtz can answer this question, I will opt for a more deflationary account of SMP. As I said above, SMPs are both logically and conceptually possible, but not all logical and conceptual possibilities are SMPS. (See page 15.) If this is the case, then what follows? One thing that is clear is that at the very least all SMPs are conceptually possible. Thus, all SMPs are true in at least one counteractual world, so ◊AS We are dealing with metaphysical possibility here, albeit strong metaphysical possibility, and for Nimtz, counterfactual worlds are the vehicles by which we explicate the metaphysical. Thus, since we are dealing with mere possibility, S must not be true in all counterfactual worlds, so ~□CS. But since we are dealing with strong metaphysical possibility, S must be compatible with our SMNs. Thus, S is strongly metaphysically possible if ◊AS + ~□CS is compatible with our SMNs. Note that this definition is consistent with the disjunctive condition noted above.

86  Rebecca Hanrahan 6 It is for these reasons that Jenkins concludes that this process will only justify, but not prove, our modal beliefs. 7 Finally, there is something odd about limiting the sensitivity counterfactual to those propositions whose modal status is known. As I understand Jenkins’s argument, there is a tight connection between the sensitivity counterfactual and the conceivability principle. Conceivability is a guide to possibility because what we can conceive is limited by our concepts and our concepts are sensitive to modal and non-modal properties of this world. Moreover, since we are sensitive to these properties, if the modal status of a proposition was different, then our sensory experiences of this world would be different as well. Thus, for the very reason that conceivability is a guide to possibility, we should embrace the sensitivity counterfactual. But if the sensitivity counterfactual is limited to those propositions whose modal status is known, then the conceivability principle does no work. Nothing about experience tells us whether a proposition is merely possible or necessary; I must already know its modal status to begin with. 8 In response, Jenkins could argue that the relationship between our concepts of cat and color isn’t “structural” and thus, nothing precludes us from taking this beast to be a cat (Jenkins 2010, 268). In offering this reply, Jenkins is trying to loosen the connection between our concepts and our sensory experiences. But this reply invites the obvious question; what counts and what doesn’t count as a structural relation and how can we tell the difference (without begging the question)? More generally, how can we loosen the connection between our concepts and our experiences without already knowing what is possible? 9 For a defense of this strong concept of laws, see John Carroll (1990). 10 See Hanrahan (2009) for a consideration of our including God in our modal epistemology. 11 Of course, there is I, that is the requirement that all determinable types must be made determinate when a token of that type is instantiated. But there are questions to ask about this requirement. For example, might this requirement be a consequence of the requirement that possible worlds be complete and coherent? Might completeness coupled with coherence exclude indeterminately striped tigers? If they do, would this therefore mean that I isn’t a strong modal necessity but rather arises (somehow) from our logic, concepts, or laws? To this, some might argue that the order of argument is backward; it is because I holds true that possible worlds must be complete and coherent. If this is correct, then I is a strong metaphysical possibility, but it seems to start taking on the air of a mere brute fact. 12 I would like to thank the participants of the International Workshop: Directions in the Epistemology of Modality, held at the University of Stirling (October 2015) for their helpful feedback on an earlier draft of this paper. And I would like to thank Professor Walter Ott for his comments on all the drafts that came before and after that draft.

References Beebee, Hellen. 2000. “The Non-Governing Conception of Laws of Nature.” Philosophy and Phenomenological Research 62(3): 571–594. Carroll, John. 1990. “The Humean Tradition.” The Philosophical Review 99: 185–219.

The Price of Sensitivity  87 Chalmers, David. 1996. The Conscious Mind: In Search of a Fundamental T ­ heory, Oxford: Oxford University Press. Hanrahan, Rebecca. 2009. “Getting God Out of Our (Modal) Business.” Sophia 48: 379–391. Jenkins, Carrie. 2010. “Concepts, Experience and Modal Knowledge.” ­Philosophical Perspectives 24: 255–279. Miščević, Nenad. 2003. “Explaining Modal Intuition.” Acta Analytica 18(30/31): 5–41. Nimtz, Christian. 2012. “Conceptual Truths, Strong Possibilities, and Our Knowledge of Metaphysical Necessities.” Philosophia Scientiae 16(2): 39–58. Shoemaker, Sydney. 1980. “Causality and Properties.” In Time and Cause, edited by Peter van Inwagen, 109–135. Dordrecht: Reidel.

4 Modal Epistemology for Modalists Otávio Bueno and Scott A. Shalkowski

Introduction For any theory to be in good epistemic standing, its key claims must not only seem plausible to its adherents. The theory itself—or some wider theory within which it is embedded—must account for how its key claims are known. Failing that, there is no useful way of adjudicating philosophical disputes, leaving us with unsatisfying stand-offs or trading intuitions. Philosophers of modality have rightly embraced this constraint on respectable philosophical theorizing. There are two broad camps. Modal rationalists maintain that modal knowledge is either a matter of exploring suitable intuitions (Bealer 1996, 1999, 2002; Chudnoff 2013), of conceivability suitably idealized (Yablo 1993; Chalmers 2002), or of reasoning from proper principles (Peacocke 1999). Non-rationalists insist that modal knowledge results from theoretical inference based on explanatory considerations, such as inference to the best explanation (Fischer 2017) or similarity relations (Roca-Royes 2017). Both trends face serious difficulties. Rationalists following George Bealer’s claim that “modal reliabilism is the correct explanation of why intuitions are evidence” (1996, 121) requires that abductive inferences leading us to the “best” explanations are themselves reliable indicators of truth (contra Bueno and Shalkowski 2015 and 2020). Those following Chalmers require conceivability so idealized as to be inapplicable to the likes of us (Worley 2003, 19–23). Non-rationalists, such as Fischer or Roca-Royes, also require either inferential or similarity relations to be reliable indicators of truth, against which we argue below. Our critiques will apply to blends using rationalist methods for modal knowledge of abstracta and other methods for concreta. We provide a middle ground between rationalist and current forms of non-rationalist modal epistemologies. Rationalists are correct that modal knowledge requires reasoning from principles, but we maintain that these principles are local and specific to the relevant domains, not those arising from intuitions, conceivability, or meaning. ­Non-rationalists are correct that modal knowledge is local, domain-specific, and not idealized to a point beyond human ken, but inference to the best explanation and

DOI: 10.4324/9781003002192-6

Modal Epistemology for Modalists  89 similarity considerations are unnecessary. Better is a modal e­ pistemology emphasizing primitive modality within a metaphysically deflationary stance. A grasp of the implicit modal character of properties and empirical regularities provides a straightforward modal epistemology for modalists.

Modal Rationalism: Intuitions As part of his defense of modal reliabilism, George Bealer develops a two-pronged strategy. First, he argues that intuitions are evidence because they satisfy three significant epistemic conditions—consistency, corroboration, and confirmation (Bealer 1996, 124–125). He then defends the reliability of intuitions, despite their fallibility (Bealer 1996, 130). We argue that neither step succeeds, and that modal reliabilism is unable to provide a proper epistemology of modality. One’s intuitions are “largely consistent with one another”; “there is an impressive corroboration by others of one’s elementary logical, mathematical, conceptual, and modal intuitions”; and “intuition is seldom, if ever, disconfirmed by our experiences and observations” (1996, 125). Intuitions constitute evidence since they are mutually consistent, corroborated by others, and not disconfirmed empirically. By implication, intuitions about possibility are central to an epistemology of modality. Anand Vaidya (2021) uses Bealer’s corroboration condition to mitigate the internal indistinguishability of accurate and inaccurate (apparent) intuitions and the dependence of intuitive judgments on matters of presentation. By Bealer’s lights, intuitions are “seemings” (1996, 123–124). Someone has an intuition that something is so when it seems to that person that it is so, and that seeming satisfies the consistency, corroboration, and confirmation conditions. Rationalists are certainly well advised to respect the appearance/reality distinction enough to demand both the internal consistency condition and the external corroboration and confirmation conditions, on the modest assumptions that the relevant portions of reality are consistent and that each of us is epistemically fallible. Any intuition-based modal reliabilism requires, first, the identification of intuitions and, second, grounds for thinking them reliable about the modal. We determine the reliability of a theoretical model by identifying the consequences of a theory combined with plausible factual claims. With these consequences in hand—predictions, typically—we then compare with observations. Nuances aside, a theoretical model is more reliable than alternatives when its predictions are more often correct than are those of competing models. The entire practice of testing theories by observation assumes that the consequences of theoretical and factual assumptions can be discerned independently of the accuracy of those consequences. This central feature of determining reliability is absent

90  Otávio Bueno and Scott A. Shalkowski from Bealer’s account, given his characterization of intuitions. Each ­condition reveals that the stock of intuitions is unstable. Intuitions will appear to be consistent until such time as someone produces a proof that appears to show a contradiction among them. This results in competing (apparent) intuitions: the original “intuition” and the joint “intuition” that the proof’s premises are accurate and that the premises justify the conclusion. “Intuitions” are corroborated by others until such times as they are not, and they lack empirical disconfirmation until they are disconfirmed. Corroboration and fit with observation can wane as well as wax. Bealer’s criteria leave us unable to identify intuitions independently of their epistemic merits, thus rendering any claims about the reliability of intuitions trivial. Reliabilism’s rescue is both easy and philosophically unsatisfying. Simply declare offending “intuitions” to be inauthentic, merely apparent intuitions taken to be genuine by the misguided. Easily done, but unsatisfying in any philosophical dispute. This declaration is either unanimous but subject to later change in philosophical group-think, or else it is not unanimous and ill-suited to resolving any disputes of substance. Some illustration is in order. Consider first the consistency requirement and a variation based on a version of Frege’s Basic Law V. That every property corresponds to a set of objects having that property seems not only true, but analytically so. Despite that, Russell’s paradox exposed the principle as false because it is inconsistent and, hence, necessarily false. Consider now an “intuition” of Law V. It seems to be consistent with logical intuitions. Despite the appearances, the inconsistent intuition cannot be consistent with other intuitions. At any given stage, we have merely the appearance of consistency and, thus, the appearance of a genuine intuition. Similarly, with the Banach-Tarski theorem, given the axiom of choice. According to the theorem, a sphere may be decomposed into finitely many pieces and reassembled into two spheres of the same volume as the original. The process may seem to be consistent but proves to be impossible in any space containing ordinary objects such as oranges. Consider next corroboration. Until the beginning of the 20th century, the “intuition” that time is absolute was corroborated by many other logical, mathematical, conceptual, and modal “intuitions”. Time seems to have a structure and order that does not depend on the observer any more than the order of natural numbers does. It has a direction that is as inevitable as it is ubiquitous: the past is gone; the future has not yet emerged. One cannot, in fact, return to the past any more than one can jump to the future. Nonetheless, according to special relativity, the “intuition” of time’s absoluteness—the intellectual seeming that time is not relative to some frame of reference—and the network of corroboration surrounding it turned out to be false. Time is relative to a frame of reference. Furthermore, and finally, the empirical confirmation of special relativity provides

Modal Epistemology for Modalists  91 clear empirical disconfirmation of the “intuition” about time’s absolute character. All three modal reliabilist criteria face counterexamples. These criteria may fail all at once. For those in the 13th century, the Earth seemed not to move, as it still seems to us. They had a so-called intuition of the Earth’s immobility. This “intuition” was consistent with other “intuitions” about Earth, impressively corroborated by others, and not disconfirmed by 13th-century observation. Despite meeting all three criteria in the 13th century, the “intuition” is false, as established by 19th-century astronomy. It is no use for rationalists, with the benefit of hindsight, to maintain that there never was an intuition of a stationary Earth, thus saving modal rationalism from counterexample. Certainly, Earth’s movement was observable in some sense even in the 13th century, even if not observable to ­13th-century observers, limited as they were by 13th-century technology. Fit with observation, no less than consistency and corroboration, is unstable, so far as philosophical force is concerned, illustrated by the march of technology. Rationalists must reconcile the ever-present possibility of revising an apparent intuition with the alleged reliability that reasoning based on apparent intuition is supposed to display. Revisability regarding the status of any particular seeming as a genuine intuition undermines the presumed lack of experimental and observational disconfirmation of intuitions. Moreover, Bealer’s criteria do little, if anything, to establish that what satisfies the criteria are themselves reliable. It is crucial to Bealer’s argument (1996, 121) that intuitions are evidence. Nothing in the criteria of consistency, corroboration, and lack of empirical disconfirmation is evidence for truth. Appearing to satisfy all three is, at most, merely the lack of any grounds for doubting what satisfies them. The criteria must be supplemented to move properly those with apparent intuition from agnosticism to belief. Absent a well-supported story about how intuiting is reliable, rationalism remains magical or wishful thinking. Required is a principle linking the satisfaction of Bealer’s criteria with (likely) truth. Elijah Chudnoff’s (2013) perceptualist account of intuition requires that intuition experiences have veridical presentational phenomenology. If your intuition experience representing that p puts you in a position to know that p, then it does so because it has veridical presentational phenomenology with respect to p. Chudnoff (2013, 207) In turn, a perceptual experience has veridical presentational phenomenology regarding p provided that: it is true that p and your perceptual experience really makes you aware of the chunk of reality that makes it true that p. Chudnoff (2013, 21)

92  Otávio Bueno and Scott A. Shalkowski Presentational phenomenology is veridical as long as one’s perceptual experience makes one conscious of the portions of the world that makes the relevant belief true. Putting things together: an intuition of something provides knowledge when it has the right kind of presentation and the right kind of presentation is one that makes one aware of how things are. We register no objection. Intuitions will yield knowledge when their content is accurate and there is something about the accurate intuitions signalling to us that they are accurate. Call this something “verdical presentational phenomenology”, if you like. We nevertheless can be mistaken about whether any apparent intuition is genuine, since we may be mistaken about whether each criterion is satisfied. Chudnoff’s attempt to link intuition with perceptual experience is little help to the modal rationalist, since in any particular instance the question will be: is the presentational phenomenology veridical? If so, the intuition yields knowledge; otherwise, not. No experience, however, carries a s­elf-verifying ­“veridical” as a component. Anyone with an experience containing anything analogous to “veridical” cannot avoid the question of whether the sign itself is a truth teller. We might appeal to our memory about the character of presentations of other experiences that turned out to be correct or we might think that the experience contains some marker of its own veridicality. Knowing we are fallible, we know that appearances of consistency, corroboration, confirmation, and veridicality have failed us; we have made no progress in thinking that the appearances that we label “intuitions” are reliable. Establishing their reliability requires some means of checking, since no aspect of an apparent intuition’s phenomenology suffices. No trait of the experience itself helps, so long as rationalists intend to preserve the distinction between appearance and reality, testimony—even self-testimony of experiences—and accuracy. We conclude that a thermometer is accurate and that its reading is evidence about the temperature, once we have checked that it has been properly calibrated. The tools and procedures for this calibration are quite local, not easily transferrable even to other measuring devices. At the very least, calibration requires something for the sake of that procedure to be acknowledged by relevant parties to be sufficiently accurate. With that, we can compare what is to be calibrated. When trying to establish the evidential value of intuitions, rationalists are in the position of having no accepted standard at all, nothing granted by both rationalist and non-rationalist as (sufficiently) accurate, against which rationalists can at least compare the intuitions on which they wish to rely. Indeed, their circumstance is much worse than those with thermometers. We might find thermometers useful even without a standard against which calibrations can be performed. We can construct translations schemes between devices with very different metrics and we can use them to compare with the states of health and illness of patients. There

Modal Epistemology for Modalists  93 is nothing magical about 98.6°F or 37°C regarding accuracy. For some purposes, it suffices to pick one, stick with it, and compare with others when necessary. All we care about is establishing correlations between changes in the device with changes in human bodies or the environment. Rationalists are in an entirely different business. They require the accuracy of the contents of some intuitions, but without any prospect of peeking behind the intuition’s content to the reality represented. As such, there are no rationalist resources to do more than assert the accuracy of some intuitions, which may have some specifiable “look and feel”. Chudnoff’s linking intuition to perception is useful for the epistemology of modality only with suitable perceptual experiences of the chunks of reality that make modal claims true. The account fares better for possibility cases. How, though, does one perceive that water is necessarily H2O, if indeed it is? One may perceptually determine how it is, but not relevant modal qualifiers on how it is. Inferences are required to take us from water being H2O to its necessity, or even its possibility, of being H2O. Perception is a useful means of acquiring knowledge because we have learned to rely on it when its deliverances are robust, stable, and factive. Like Chudnoff, Bealer is not wholly unaware of the need to approximate factivity for intuitions. For concept possession, he maintains that: x determinately possesses the concept of being a multigon iff: x would have the intuition that it is possible for a triangle or a rectangle to be a multigon iff it is true that it is possible for a triangle or a rectangle to be a multigon. Bealer (1999, 41) We observed that Bealer’s criteria for intuitions are external to the intuitions, making their genuineness not perfectly transparent. The same predicament recurs here. Whether one determinately possesses a concept is a matter of making modal judgments about applicability if, and only if, those judgments are true. The simple example obscures the central difficulty. The general form is: one determinately possesses a concept if, and only if, one would make certain modal judgments if, and only if, those judgments are correct. So long as we focus on a few salient cases, it is hard to see how one could know what it is to be a multigon without judging that it is possible for triangles and rectangles to be multigons. Set aside the peculiarity of building into grasping this concept intuiting the possibility that triangles and rectangles be multigons, rather than the judgment that triangles and rectangles are multigons. The left-to-righthand direction of the inference shows that one determinately grasps what it is to be a multigon only if one has (only) correct modal intuitions about multigons. Possession depends on making modal judgments and making modal judgments depends on their accuracy.

94  Otávio Bueno and Scott A. Shalkowski This is no advance for rationalists. Philosophical disagreement about modal matters was the problem to which intuitions were to be the solution. When disagreements about modal matters are shifted to disagreements over the genuineness of intuitions or to whether one genuinely grasps a concept, we have no solution at all. This situation carries over to the more sophisticated social-disjunctive account of intuition-based modal epistemology recently advanced by Anand Vaidya (2021). Granting experimental results that threaten to undermine wholesale the epistemic value of intuitions, Vaidya adapts John McDowell’s disjunctive account of perceptual experience (McDowell 2008, 2009), according to which there is a metaphysical difference between veridical and non-veridical experiences. We pose no objection to the use of “metaphysical” in this context. Those minded to articulate this difference in terms of some relation—causal or otherwise—are free to do so. Vaidya correctly concedes that first-person assessments cannot distinguish between “a veridical intuitional state” and “a non-veridical appearance that is phenomenologically similar to an intuition, but because something has gone wrong it is not a genuine intuition” (Vaidya 2021, 215). We argued earlier that the proposed solution, i.e., incorporating a social dimension to modal epistemology by consulting others trained and competent in relevant matters, fails to solve the fundamental problem. Each party is, as it were, behind the same metaphysical veil. The social dimension simply adds more parties in the same plight: this is how things seem to me/us. Consulting others enhances our own epistemic state only when the others have been able to put themselves into positions to check the facts. When all, however, are limited to their own seemings, no one can help anyone else break out of the pool of seemings to verify whose seemings are veridical and whose are not. All of the talk of intuitions serving as evidence ignores a fundamental feature of evidence. Mostly typically, evidence for one state of affairs is another. Evidence of a mouse is scratching behind baseboards and missing cheese. Evidence of an approaching storm is a precipitous drop in barometric pressure, or the sudden cooling of the air. We may well observe various matters, but our evidence of the mouse is not that it seems to us that there is a mouse except insofar as our use of “seem” in that context is hardly experiential but inferential. The epistemic work is done by hearing the scratching and finding the cheese gone. We observe some worldly states statistically relevant to another. Knowing that statistical relevance is why we treat some states as evidence for another. Those wishing to treat intuitions as evidence provide nothing like this. Intuitions as seemings cannot serve as worldly state relevant to our belief in another unless those intuitions themselves are reasonably believed to be statistically relevant to their own contents, which is not established by observing that many (of the right sort, perhaps) are likeminded about matters. No one escapes Plato’s cave by consulting other

Modal Epistemology for Modalists  95 cave dwellers. Those wishing to treat intuitions as evidence must provide more. A design story may do. Theistic design of human cognitive machinery about modal matters would do. Naturalistic stories have the greater challenge of accounting for how reliability of intuitions about metaphysical matters would have developed. Observational reliability about cheese or air pressure is one thing; intuitive reliability about modal matters is another entirely. We do not claim that no such design stories can be developed, only that they must be, but have not yet been, developed. Bealer’s aim was to develop an account of the a priori. We articulated our best approximation of an epistemology of modality along his broader epistemological lines. The autonomy of the philosophical enterprise is crucial for the project. In fact: This kind of autonomy thesis is a modal claim; it posits only the metaphysical possibility of autonomous a priori knowledge, perhaps on the part of creatures in cognitive conditions superior to ours. But, if true, the thesis would nevertheless help to illuminate our own situation. For to the extent that we approximate the indicated cognitive conditions, we are able to approximate the sort of autonomous a priori knowledge contemplated in the thesis. Bealer (1999, 48) Acknowledged here is the highly idealized character of the overall account. Bealer helps himself to the claim that we approximate our cognitive superiors, but that must be established and not merely appropriated. In particular, it must be clearly specified and established for modal claims. Otherwise, the proposal has no force to accommodate how we might have such knowledge. If we know the truth, it is reasonably clear how to devise an account of the relevant approximation. The perfectly obvious difficulty, however, is that we are typically trying to determine the truth and cannot use knowledge of it as a guide to explain the relevant approximation. Bealer-style modal reliabilism is not equipped to provide a proper epistemology of modality. We conclude this section by considering a significant challenge to an influential form of modal rationalism, advanced, for instance, in Chalmers (2002), and that invokes conceivability. The challenge highlights a problematic consequence of the view (Vaidya 2008). According to this form of modal rationalism, an idealized kind of conceivability (the so-called ideal primary positive conceivability) entails a particular sort of possibility (primary possibility). As Vaidya argues, it follows from this view that “the space of logically possible worlds is coextensive with the space of metaphysically possible worlds” (Vaidya 2008, 191). The presumption of safety and plausibility of this consequence is unwarranted. One would expect that logical possibility and metaphysical possibility not to be coextensive. After all, nothing in logic precludes one of

96  Otávio Bueno and Scott A. Shalkowski the authors of this work being a fried egg, since the supposition is not c­ ontradictory. This does not, though, seem to be a metaphysical possibility. In order for that to be the case, two situations need to be possible: the author is generated inside an egg and the egg is fried. The latter situation is straightforward, but it is unclear that the former is metaphysically possible at all. Being human, a crucial feature of the author, at least given the DNA that humans have, does not seem to be compatible with coming from an egg. A very different creature from the author would end up in the frying pan. Vaidya’s argument challenges the adequacy of any modal rationalism conflating logical and metaphysical possibilities. Those conflating them owe us not just claims about the coextensiveness of “spaces” of worlds, but detailed accounts of how it is that a human being could be a fried egg, an oak tree, or a neutrino. What are the existence, identity, and persistence conditions for each? What are the parameters of variability? How could the tool (logic) for specifying one “space” be appropriate for all modalizing, since that tool is designed to wash out as much detail as possible when reasoning formally about things? Or, to put matters slightly differently, why think that the possibilities and the necessities for each thing are the same as they are for every other thing? Absent answers to those questions there is no way to assess that form of modal rationalism.

Modalism and Modal Epistemology Non-rationalism in the epistemology of modality. Important forms of rationalism, we argued, are inadequate for a proper epistemology of modality. Non-rationalists do better because they typically do not require idealization and they provide accounts more suitable to the epistemic circumstances faced by creatures like us. Two important developments are Bob Fischer’s (2017) account of justification of modal claims in which theories play a decisive role and Sònia Roca-Royes’s (2017) similarity-based account of knowledge of possibility for concrete objects. We consider each, in turn. Fischer (2017) rightly emphasizes the way in which justification of modal claims often relies on theories about the relevant domains. The theories encode significant information about the objects whose possibility or necessity is at issue. By relying on such theories one can be guided in the determination of what is possible for such objects. The difficulty is that inference to the best explanation is central to Fischer’s account. The notorious unreliability of this type of inference (see van Fraassen 1980, 1989; Bueno and Shalkowski 2015, 2020) makes it ill-suited to an epistemology of modality. Furthermore, there is a fundamental and unresolved ambiguity in the basic nature of IBE. When determining which theory is best, even if we set aside the prospect that the lot from which we select the best is a “bad” lot (van Fraassen

Modal Epistemology for Modalists  97 1989), IBE must select the best theory on the basis of either “internal” features of the theory itself or else features external to the theory. Treatments of IBE are not always explicit about which features are relevant and used. Consider them, in turn. The internal features are those of the theory, which is a representation of how things are. The theory may be brief, tome-like, simple, complex, stating that few or many (kinds of) things exist with few or many characteristics and relations, tied together in a unified explanatory mechanism or in a patchwork of independent explanations. If the project is to produce a reasonable approximation of a true theory, the representation itself is a distraction. The project is to determine whether there are few or many things, characteristics, relations, explanatory mechanisms, etc. To compare theories on the usual parameters of simplicity, coherence, unification, etc., is to preempt the task of determining how things are in favor of what the theories are like. Treating a simpler theory, however simplicity is understood, as thereby more plausible is to assume that reality is simple rather than to discover it. Otherwise, this is just a shift of attention from reality to representation. Avoiding this distraction, looking to external features is to remove anything distinctive about IBE. When answering “How well does each theory represent reality?”, we engage in the project of discovery. In doing so, we must then make the usual appeals to evidence and/or reasoning. We have no objection to either; we note only that when we cease assessing theories in light of assumptions about reality and look to reality itself to test those assumptions, we no longer engage in IBE. We examine candidates for data, theories, and permissible reasoning. We then assess what is true on the basis of old-fashioned evidence. Modal metaphysics and its epistemology are not well suited to justifications by assessing evidence in the usual ways, most obviously because we have no useful run of assessing hypotheses about it based on various justifying considerations and then in a theory-neutral manner, assessing track records, making probability assessments, and the like. We may use the language of evidence and probability, but there is nothing analogous to frequency distributions in light of which we can compare our stock of relevant evidence to discover how often evidence like this arose because reality was This Way instead of That Way. All we ever have is what we take to be evidence and the conclusions we draw from those takings, which are always disputed. For arguments about modal metaphysics and its epistemology ever to be persuasive, there must be some way of breaking out of partisan differences regarding what constitutes evidence and its force. There must be some agreement about relevant facts and success rates of inferential practices in light of those facts. None of this is available to modal metaphysicians and epistemologists. The moral for projects like Fischer’s is that either they rely on IBE and thereby focus on the wrong things or they focus on the right things, but with no prospect of weighing evidence that is analogous to its assessments in science.

98  Otávio Bueno and Scott A. Shalkowski Like Fischer, Roca-Royes (2017) correctly focuses on the relevant objects to determine what can be known about what is possible. Similarity relations are notoriously vague and hold indiscriminately; everything resembles everything else in some respect or other. Were merely some resemblance with something with characteristic F sufficient for knowledge that another could be G, then if anything is F, then everything could be G. Roca-Royes’s account correctly narrows the focus, selecting some resemblances to actual outcomes to justify claims about possible outcomes. That a person has a cardiovascular system and that some with a cardiovascular system have fatal coronary incidences provide knowledge that the person might have a similar incident. So far as that goes, it is not wrong. It does not, though, expose the central fact about modal knowledge. Central to her account, when dealing with concrete objects, is the reliance on induction on similarity relations to ground such knowledge claims. Inductive inferences justify claims only when some regularity underwrites the phenomena under study. There must be some known pattern in which being F is associated with being G for us to have our confidence justifiably raised that some particular is G on the grounds that it is F. If one has grounds for thinking that actual nature is uniform regarding being F and being G, inductions will be reliable as a matter of fact. Modal knowledge requires more, since the actual regularity may be a large-scale cosmic accident, which would have no implications whatever for how things could be. A non-rationalist alternative that relies neither on IBE nor on inductive inferences or on similarity judgments more generally is called for. We sketch such an approach, with some distinctive empiricist traits. Modal properties and modalism. Central to our modalist proposal is establishing access to the possible and the necessary by way of knowing the modal properties of the objects under consideration. These are properties that objects have given what they are. A wooden table is breakable, given the materials that constitute it. Breakability is, thus, one of its modal properties. By knowing a table’s composition and by knowing the behavior of its materials under various changing circumstances, we acquire knowledge of the table’s breakability. The table’s breakability is empirically accessible to anyone with knowledge of its composition. Engineers use sophisticated theories of materials when planning domestic, civil, and commercial projects. Those theories contain articulations of the results of stress tests, which exposed the limits in various conditions of materials and their combinations. The very point of attending to those theories when designing artifacts is to stay well away from the point at which the artifact would break, melt, or otherwise degrade. The same goes for abstract objects. Can every set be well-ordered? The answer depends on the relevant properties of sets. If the Axiom of Choice is in place, then the sets do have that property, since that axiom guarantees

Modal Epistemology for Modalists  99 the availability of a well-ordering even if none can be explicitly exhibited. Without Choice sets lack that property, since no such well-ordering is generally available. The modal content of this result is only superficially skewed by the use of purely extensional mathematical languages. Given Choice, it is possible to well-order any set; otherwise, it is not. As things stand, we can obtain only relative or conditional possibility. Only relative to Choice is well-ordering possible. Constructivists reject Choice and, so, universal well-ordering; classical mathematicians can have both. The point generalizes. The theories we have been considering, whether empirical or mathematical, typically involve no modal operators. ­Philosophers should find no deep lesson in this. Materials scientists and engineers are not modal metaphysicians. They ply their respective trades while being cavalier about categorical vs. modal matters. Nevertheless, there is no stretch to the imagination to consider the following: DESIGNER:  “Here’s my design for the table”. BOSS:  “This is a joke, right?” DESIGNER: “Why?” BOSS:  “Look. Right here at this stress point. Put your typical Thanksgiv-

ing turkey right there and the guests would have the entire dinner in their laps!” Any such conversation exposes that the apparently non-modal mathematics in the background is, by the lights of the practitioners, imbued with modal content. We decline any philosophically interesting account of these modal properties because we decline any philosophically interesting account of any properties. Paper tears, rubber bands stretch, and sugar cubes dissolve in water. We see these things. When convenient, we use more passive constructions to say how these things are; tear-able, elastic, soluble. When self-consciously distinguishing the item from its attribute, we use property constructions. The paper has the property of being tear-able, the band the property of being stretchable, the cube of being soluble. What each does has been observed. There is no need to invoke highly charged metaphysical descriptions of such properties. They, including modal properties, are nothing but features of objects; perfectly ordinary, observable features at that. Modal knowledge is an extension of knowledge of actual properties that objects have and how these properties change under changing circumstances. Knowledge of possibility arises from knowledge of actuality. Testing limits is testing the possible. Knowledge of the necessary results from what is precluded from such variations. Tables cannot fly on their own. Philosophical thought “experiments” purporting to show otherwise are almost always wholly inattentive to what makes things what they are. Wooden tables might be unbreakable were wood and steel nuts

100  Otávio Bueno and Scott A. Shalkowski and bolts to be very different from what they are. Metaphysical fantasies require the hard work of discerning other limits. How different from oak and maple can something be and still be wood? How different from iron’s standard chemical composition can something be and still be the iron that composes steel? These and related questions are wholly unanswered by those claiming the possibility of unbreakable tables or flying pigs. As with both rationalist seemings and the conflation of so-called logical and metaphysical possibilities, the hard work is left undone by thought experimenters. Induction plays no role in the distinctive move to warranted, even if limited, modal knowledge. In our examples, the properties of materials with their actual features suffice. Neither is access to modal properties the outcome of an inference to the best explanation. Most obviously in the mathematized portions of science, the mathematics permits the articulation of the scope of the possible. The breaking point of steel cables is the point beyond which one cannot exert force on an intact cable. Were the engineer’s formulae not encoding modal information, there would be no point in thinking counterfactually armed with engineering expertise. We modalists urge humility. Blue swans? Talking donkeys? Well, no. We have no doubt that there is consistency in things with the appearance of swans being blue. Indeed, that much has, no doubt, been an artificial reality. The envisioned natural reality? The case is never properly made. For convenience only, concede that surface appearance does not make a swan, but some “deep” structure. Tell us more about the tolerances of variation of that structure. Tell us exactly about the color-making features of that deep structure. Tell us when we have no longer a swan but a duck or something Dr. Seuss-like in kind. Similarly for talking donkeys. We should resist any pretense to knowledge in all such u ­ nder-specified cases. Whenever unestablished assumptions are needed to establish a given possibility or impossibility, some dose of modal skepticism is, therefore, expected (Bueno and Shalkowski 2015). Primitive modality. Also central to a modalist epistemology is the reliance on primitive modality. We noted that objects have properties and among those are modal properties, which need not, and should not, be characterized in more basic terms. It is an unnecessary philosophical addition to identify these modal properties with universals, dispositions, powers, propensities, laws, or related metaphysical posits. The tendency in metaphysical theorizing to reify gives the false impression of uncovering the underlying nature of the objects under consideration. As will become clear, reifying the relevant objects in terms of a preferred ontology offers only an illusion of understanding the relevant phenomena better. Modalists question the wisdom of ontologizing. The ontological inclination is driven by presumptions about the primacy of the categorical

Modal Epistemology for Modalists  101 or the drive for further explanation. What, though, are the grounds for thinking all reality to be categorical? It is difficult enough to banish the modal from the apparently categorical. Things are green not because of how they appear. It may be dark. Not because they are observed to be green. There may be no one about. Not because we think they are green. We may have confused green with blue. Not because our visual fields have a green feature. Some of us may be colorblind. Any useful account of frogs being green on a desolate night is shot through with modality. Were conditions to change this way or that, the frog’s skin, the light, a device, or eyes would do various things. Good luck with a non-modal account of an electron’s charge. Drives for deeper metaphysical explanation fare no better. Why does this do what it does? Because it has a disposition, a power, a propensity to do so. While that seems to take us one step further to a full explanation, we have simply traded one form of language for another. More carefully, such reifications fare no better than rationalist intuitions. There is no way to identify the reified portion of reality, save through the activity for which it is to be the explanation. Why does it do what it does? Because it is the kind of thing that does that, or it has a characteristic that makes it do what it does. Notice how the metaphysical explanation differs from a scientific one. Why did the container catch fire? Not because it is the kind of thing that catches fire or because it has the attribute of being fireable. No. It caught fire because it contained gasoline instead of water, the lid had been removed, and some fool thought it a good idea to toss a lit match into it. Here we have distinct facts explaining the one in question. Metaphysical reification manages only to use a different form of words to leave us where we were: none the wiser about why things are the way they are and how we might effect change in the future. A more promising approach is to understand modal properties directly as features of the objects. By taking the modal aspect of some properties as primitive, the modalist resists such reification and insists that some phenomena are better accommodated instead simply in terms of the modal. The problem of modal epistemic friction. In light of these considerations, a significant issue emerges. Given the variations involved in modal contexts, one needs to ensure that the possibilities (or necessities) entertained are genuine, and that they do not conflict with (the natures of) the objects in question. After all, if there are conflicts between the entertained possibilities (or necessities) and the relevant objects, the conjectured changes are not really possible. It would not be Socrates, under significant counterfactual variation, if the object under consideration is no longer human. Being essentially human seems to constrain the range of modal variations for Socrates. According to Anand Vaidya and Michael Wallner, the problem of identifying the constraints on modal variation and the problem of knowing

102  Otávio Bueno and Scott A. Shalkowski those constraints is the problem of modal epistemic friction (Vaidya and Wallner 2021). As they note: Suppose you know that Socrates is human. To get from there to the knowledge that it is metaphysically necessary that Socrates is human on the basis of essentialist deduction-theory, what you need to know is that being human is essential to Socrates and that what is essential (to some x) is metaphysically necessary. In other words, the epistemic friction creator […] in this case is the essentialist principle […] together with some essentialist proposition. Vaidya and Wallner (2021, S1919) Vaidya and Wallner thus generalize what we have urged throughout: there must be constraints on our modal theorizing and there must be points of epistemic contact with those constraints for warranted modal metaphysics. From a modalist perspective, the constraints involved in modal variation are given by the relevant modal properties. These properties are parameters that can be invoked to identify what remains constant across counterfactual scenarios. Clearly, something needs to be held fixed among the variation on pain of losing, among the changes, the very objects whose counterfactual shifts are at issue. A modalist who is also an essentialist would have no difficulty tackling the problem head-on, at least at an initial stage of analysis. No counterfactual variation violating the essential properties of an object and their implications is possible. The very object would be lost. Modalists, however, need not be essentialists. Our identification of both the object and its modal properties suffices. Non-essentialists typically complain that the essence of a given object is not obvious. We decline to litigate the matter here. There is no need. The modal epistemologist is not engaged in a process of inspecting alternative possibilities using a “Jules Vern-o-scope”, which would require knowledge of an object’s essence to identify it from among many options in an alternative possible world (Kaplan 1979). Rather, in any particular instance, we have identified the object and attributed to it a modal property. Rational and empirical grounds for both the identification and the attribution obviate the need for an essentialist theory about the object itself. We have picked it out. It is before us, as it were, even if only by way of reference rather than physical presence (Kripke 1980). Any reason to think that any change permitted by the modal property would obliterate the object is a reason not to attribute that modal property to that object. Any reason to doubt that we know that about which we speak is similarly grounds for not attributing any property to the object. We need not resolve questions about essences of artifacts to know that the table before us is breakable, nor do we need a philosophically robust theory of human nature to know

Modal Epistemology for Modalists  103 that Socrates could have forsaken the philosophical life. Anyone unable to make these and a multitude of similar assessments would have no business either ontologizing in terms of tables or Socrates or modalizing about those objects. A full modal metaphysics about tables or people may require more to decide edge cases. The devil will always be in the details. Any gains earned from quite general essentialist commitments are lost in particular judgments or disputes, if the essences of the relevant objects are insufficiently obvious. If the essence of Socrates is being human, unless it is known what it is to be human, one will be none the wiser. Identifications of the human essence with DNA, evolutionary genealogy, or social organizations each have their own challenges. Perhaps those challenges can be met. Here, we provide no grounds for essentialists to forsake their essentialist programs. We urge only that a great degree of knowledge about possibility for objects with which we are acquainted is achieved without these programs. Modalists armed with solidly empirically grounded theory can make progress in the epistemology of modality on bases much more modest than those typically wheeled out. Just as our inability to generate wellgrounded consensus about whether a particular shade is chartreuse save by stipulation is no barrier to warranted judgments that a typical rose is not chartreuse, so we may judge ranges of possibilities for the table in front of us and for the Greek philosopher famous for mentoring Plato. We may begin without E. J. Lowe’s desired knowledge of the appropriate ontological categorization of an object with philosophical sophistication (Lowe 2006), just as no one needs the sophistication of botanists to know that the wood of an oak may be used to make tables and may burn in a forest fire. As rough and ready as our modal judgments may be, we know enough to know what generates modal epistemic friction. Modal properties of the objects at hand. Skepticism and the scope of an epistemology of modality. Given these remarks, we close by considering the scope of any epistemology of the modal. Is an account of modal knowledge expected not to rely on any knowledge of the actual whatsoever? Or is it expected not to rely on any knowledge of the possible or the necessary? A positive answer to either of these questions—thus requiring an approach to modal knowledge that does not assume any knowledge of actuality or any knowledge of ­modality—seems to lead one directly into skepticism. Barry Stroud (2000) has examined the difficulties and inevitable dissatisfaction associated with the philosophical project understanding human knowledge in general. The aim is to understand human knowledge without first helping oneself to some knowledge. The question is: if one does not assume that one already has some knowledge and is using only knowledge-preserving inferences, how can one come to a completely general understanding of human knowledge? If such knowledge is

104  Otávio Bueno and Scott A. Shalkowski assumed, then skepticism is not defeated, since there is a remainder (that which is assumed) for which there is (yet) no account. If such knowledge is not assumed, skepticism is still not defeated, since there is no grist on which the epistemological mill may operate. To be left with either an unaccounted for remainder (either tout court or at any given stage in a hierarchy of accounts) or to be left with no data with which to work is to fail in the project of fully explaining knowledge. The result is an inevitable dissatisfaction familiar from philosophical skepticism (see also Stroud 1984, 2011). Earlier, in connection with modal rationalism, we rehearsed part of Stroud’s concern. Assuming arguendo that a range of our seemings track modal reality sufficiently, the modal rationalist’s project is not nearly complete. Advancing the rationalist thesis requires advancing reasons for thinking that seemings do track modal reality. Bealer rightly attempted to give conditions for these seemings that suffice for them amounting to evidence. His plight is that consistency is at any given moment indistinguishable from merely apparent consistency. Corroboration that one’s seemings might provide to another requires—if they are to provide any epistemically useful corroboration at all—that one’s seemings track modal reality. Without that assumption, Bealer-style rationalists cannot manage to set aside that all such “corroborations” are anything more than the blind leading the blind. They may avoid disaster. They may all end up happy. The stating of the three conditions on intuitions constituting evidence, however, merely assumes what should be placed in evidence, i.e., that any seemings at all ever track modal reality. The predicament to which Stroud points is generated by the project of a fully general account of knowledge, involving an explanation of how such knowledge is both possible and possessed. Externalist “if-then” strategies fail to tick all of the necessary boxes. Of course, if some externalist account is correct, be it Cartesian with a background story of divine design or secular with a different background, then those of us satisfying the relevant external conditions know some things. That is far too easy and satisfied by nearly any theory of knowledge. The sticking point is always to be in the position to assert the antecedent of that conditional, where externalists are always stuck, if trying to give completely general account of knowledge and not merely to maintain that skeptics have failed to demonstrate that we have none (Plantinga 2000). David Lewis’s appeal to a plurality of worlds falls afoul of similar difficulties (Lewis 1986). Of course, if (again) there is a plurality of such worlds with the character ascribed to it on the basis of principles that we use, such as recombination, then we have correct modal beliefs when judging and reasoning in ways validated by the plurality. Since such may also be said for any other modal metaphysics and epistemology, that is no advance. His use of inference to the best explanation to bolster his case does nothing, since IBE is just a variation on this conditional predicament

Modal Epistemology for Modalists  105 in another guise. A theory is said to be best when if true it would better explain things than would alternatives (Bueno and ­Shalkowski 2015). Without the IBE strategy, there is nothing but the assumption that the plurality has the requisite Goldilocks character of being just right (Shalkowski 1994). We modalists suffer no such plight. We have no overarching explanation of the modal and we have as yet attempted no general account of modal knowledge. We satisfy ourselves with acknowledging primitive modality and emphasize that the epistemology of the modal is concerned with modality in actuality. We find modal content in theories developed for other purposes, be they purely intellectual or practical engineering. Ignoring this modal content is to fail to grasp the use to which those theories are put. Were they to have no modal import we could not use them to think counterfactually nor to plan for the future. To take care in treating them wholly non-modally would be to treat their content as indistinguishable from theories about cosmic accidents, which would be no guides to either counterfactual reasoning about the past when thinking about, say, moral responsibility or for thinking about the future for which we may wish to take responsibility. Because the objects are ordinary and the grounds for the relevant theories that warrant the attributions of modal properties are not especially tendentious, we require no controversial substantive philosophical assumptions about what exists, how it exists, or how we come to know either. Such modesty is a virtue.

Conclusion We offered a critical assessment of central epistemologies of modality, especially of a rationalist sort, and sketched a modalist alternative. ­Central to the proposal we recommended is the identification of the relevant modal properties that ground the corresponding modal knowledge. There is no need to settle the metaphysical nature of such properties. We need only access to the relevant objects and grounds for attributing properties with modal content, even when that content is obscured by categorical representations. We are right to remain agnostic about edge cases and to be doubtful of claims based only on the thin reed of the failure of obvious contradiction. This is as it should be for those sympathetic to the fact that modal knowledge is knowledge of an important aspect of the actual world.

Acknowledgments Our thanks go to Anand Vaidya for extensive and extremely helpful comments on an earlier version of this work. They led to substantial improvements.

106  Otávio Bueno and Scott A. Shalkowski

References Bealer, G. (1996) “A Priori Knowledge and the Scope of Philosophy”, P ­ hilosophical Studies 81: 121–142. Bealer, G. (1999) “A Theory of the A Priori”, Philosophical Perspectives 13: 29–55. Bealer, G. (2002) “Modal Epistemology and the Rationalist Renaissance”, In T. Gendler and J. Hawthorne (eds.), Conceivability and Possibility, Oxford: Oxford University Press, 71–125. Bueno, O., and Shalkowski, S. (2015) “Modalism and Theoretical Virtues: Toward an Epistemology of Modality”, Philosophical Studies 172: 671–689. Bueno, O., and Shalkowski, S. (2020) “Troubles with Theoretical Virtues: ­Resisting Theoretical Utility Arguments in Metaphysics”, Philosophy and Phenomenological Research 101: 456–469. Chalmers, D. (2002) “Does Conceivability Entail Possibility?”, In T. Gendler and J. Hawthorne (eds.), Conceivability and Possibility, Oxford: Oxford U ­ niversity Press, 145–200. Chudnoff, E. (2013) Intuition, Oxford: Oxford University Press. Fischer, B. (2017) Justification via Theories, Dordrecht: Springer. Kaplan, D. (1979) “Transworld Heir Lines”, In M. J. Loux (ed.), The Possible and the Actual. Ithaca, NY: Cornell University Press, 88–109. Kripke, S. (1980) Naming and Necessity, Oxford: Oxford University Press. Lewis, D. (1986) On the Plurality of Worlds, Oxford: Blackwell. Lowe, J. (2006) The Four-Category Ontology: A Metaphysical Foundation for Natural Science, Oxford: Clarendon Press. McDowell, J. (2008) “The Disjunctive Conception of Experience as Material for a Transcendental Argument”, In A. Haddock and F. Macpherson (eds.), Disjunctivism: Perception, Action, Knowledge, Oxford: Oxford University Press, 376–389. McDowell, J. (2009) “Criteria, Defeasibility, and Knowledge” In A. Byrne and H. Logue (eds.), Disjunctivism: Contemporary Readings, Cambridge, MA: MIT Press, 75–81. Peacocke, C. (1999) Being Known, Oxford: Clarendon Press. Plantinga, A. (2000) Warranted Christian Belief, Oxford: Oxford University Press. Roca-Royes, S. (2017) “Similarity and Possibility: An Epistemology of de re ­Possibility for Concrete Entities”, In B. Fischer and F. Leon (eds.), Modal Epistemology after Rationalism, Dordrecht: Springer, 221–245. Shalkowski, S. (1994) “The Ontological Ground of the Alethic Modality”, ­Philosophical Review 103: 669–688. Stroud, B. (1984) The Significance of Philosophical Scepticism, Oxford: Clarendon Press. Stroud, B. (2000) Understanding Human Knowledge: Philosophical Essays, Oxford: Oxford University Press. Stroud, B. (2011) Engagement and Metaphysical Dissatisfaction: Modality and Value, New York: Oxford University Press. Vaidya, A. (2008) “Modal Rationalism and Modal Monism”, Erkenntnis 68: 191–212. Vaidya, A. (2021) “Intuition and Modality: A Disjunctive-social Account of Intuition-based Justification in the Epistemology of Modality”, In O. Bueno and S. A. Shalkowski (eds.), The Routledge Handbook of Modality, Abington: Routledge, 208–218.

Modal Epistemology for Modalists  107 Vaidya, A. and Wallner, M. (2021) “The Epistemology of Modality and the ­Problem of Modal Epistemic Friction”, Synthese 198: S1909–S1935. van Fraassen, B. C. (1980) The Scientific Image, Oxford: Clarendon Press. van Fraassen, B. C. (1989) Laws and Symmetry, Oxford: Clarendon Press. Worley, S. (2003) “Conceivability, Possibility and Physicalism”, Analysis 63: 15–23. Yablo, S (1993) “Is Conceivability a Guide to Possibility?” Philosophy and ­Phenomenological Research 43: 1–42.

5 How (Meta-)Semantics Defuses Modal Pessimism Christian Nimtz

Introduction: Modal Pessimism and Why We Should Care True modal skepticism is out of fashion. Hardly anyone denies that we can attain metaphysical modal knowledge. (Unless indicated otherwise, please read my modal talk and my uses of ‘□’ and ‘◇’ as expressing metaphysical modality.) This does not hold for a weaker variety of modal skepticism I dub Modal Pessimism (van Inwagen 1998; Hawke 2010, 2017; Yli-Vakkuri 2013; Strohminger and Yli-Vakkuri 2018; see also Kung 2017). Modal Pessimism has it that our modal knowledge is characteristically limited: while we can reliably attain knowledge of some possibilities and necessities, other modal truths are beyond our epistemic ken.1 As Yli-Vakkuri points out, this position is “nowadays widely held, and even more widely treated as a serious threat to modal metaphysics” (2013: 607). The Modal Pessimist grants that we can reliably attain knowledge of possibilities such as ◇(I pick up that pencil) or ◇(There are craters on the Moon) that follow from empirically established truths (Hawke 2017: 296; Vetter this volume). The Modal Pessimist also grants that we can reliably attain knowledge of necessities such as □(p → p), □(5 > 2), and □(Grandmothers are female) since logical truths and provable mathematical truths hold with metaphysical necessity (Yli-Vakkuri 2013: 606–612). The same arguably holds true for conceptual truths, if you acknowledge them at all (Nimtz 2012). I call possibilities and necessities of this variety near-side modal truths. However, the Modal Pessimist denies that we can reliably learn about what Yli-Vakkuri (2013: 607) dubs ‘properly metaphysical’ modal truths and what I will call far-side modal truths. Following Yli-Vakkuri (2013: 607), I will assume that far-side modal truths comprise controversial metaphysical contentions such as □(Everyone has her actual parents) or ◇(There are philosophical zombies) alongside claims such as □(Cicero = Tully) or □(∀x: x is water ↔ x is H2O) that figure on just about any list of exemplary metaphysical necessities. What warrants the Pessimist to take a dim view of our ability to learn about such modal truths? Put simply, far-side modal truths appear to be modally autonomous in

DOI: 10.4324/9781003002192-7

How (Meta-)Semantics Defuses Modal Pessimism   109 that they apparently cannot be derived from factual, nomological, ­logico-mathematical, or conceptual truths even when combined. In contrast to ◇(There are craters on the Moon), a far-side possibility claim such as “◇(There are philosophical zombies)” embeds a statement p that is actually false and has no closely similar statement p* that happens to be true. We thus cannot infer ◇p from the actual truth of p or from the actual truth of p*, as Roca-Royes (2017) proposes. In contrast to the likes of □(p → p) or □(Grandmothers are female), a far-side metaphysical necessity claim such as “□(∀x: x is gold ↔ x is Au)” is neither a direct modal consequence from science (Williamson 2016), nor does it embed a statement p that is a logical, a mathematical, or a conceptual truth, nor follows from truths of these kinds. We thus cannot infer □p from the fact that p is a logical, mathematical, or conceptual truth, or follows from truths of these kinds, or so it appears. Why should we care about Modal Pessimism? Taking his cue from van Inwagen (1998), Hawke (2017) employs Modal Pessimism to overcome Humean skepticism about induction. My aims are less instrumentalist. I agree that defusing Modal Pessimism is of import for meta-philosophy. The Kripkean ‘remodalization’ (Hughes 2004: 84) of contemporary philosophy rests on the epistemic presupposition that we can reliably learn about those far-side modal truths our modal statements are about. Modal Pessimism entails that this presupposition systematically fails. It thereby contests the epistemic legitimacy of much of our contemporary modal dealings. Yet, I first and foremost think that assessing Modal Pessimism is of importance for modal epistemology itself. The last few years have seen a rich variety of proposals arguing that we can reliably attain knowledge of what amounts to near-side metaphysical possibilities and necessities (Yli-Vakkuri 2013; Fischer 2016; Roca-Royes 2017; Vetter 2016; Williamson 2016; Vetter this volume). The Modal Pessimist challenges us to substantiate our assumption that this knowledge projects to far-side modal truths. Thus broadening what Vaidya and Wallner (2021: 1909) dub the ‘navigation question’ of ‘[H]ow can we reason with justification from one kind of modality to another (…)?’, the Modal Pessimist insists on an answer fit to allay the – as I feel: legitimate – worry that our dealing in far-side modality may be tainted by an ‘institutionalised pretence to knowledge of remote matters’ (van Inwagen 1998: 69). This challenge is all the more serious for the fact that the Modal Pessimist is in a prima facie good position to defend her pessimistic contention, as I believe she is (see the next section). In this contribution, I develop a semantic answer to Modal Pessimism. I argue that science plus semantics – or rather: meta-semantics, but I won’t differentiate2 – allow us to reliably procure far-side modal knowledge beyond identities and rigidity. Kripke influentially argues that identity statements with rigid designators are what I call proto-necessities: these statements are necessary, if true. Drawing on the idea that predicates

110  Christian Nimtz such as ‘is water’ or ‘is an Allosaurus’ are what I call paradigm terms (Nimtz 2017, 2019, 2021) – that is, terms whose application is governed by a relation and anchored in specific actual paradigms – I show that select general statements beyond identities are proto-necessities, too. I argue that the Modal Pessimist may be able to accommodate modal knowledge arising from the original Kripkean result. But the expanded argument forces critical concessions that defuse her pessimism, or so I maintain. Here is what I will do. First, I argue that the Modal Pessimist can devise a prima facie forceful argument for her claim and is in a good position to defend its premises (Section “Modal Pessimism and the Argument from Autonomy”). Second, I rehearse the Kripkean argument and explain how it allows us to attain knowledge of truths such as □(Proxima Centauri = HIP 70890). (I call such truths necessary identities.) I add that if the Modal Pessimist were to accommodate such far-side knowledge, her weakened pessimism would still have bite (Section “Modal Knowledge from Science and Semantics I: Identities”). Third, I explain what paradigm terms are (Section “Paradigm Terms Semantics”), I show that select paradigm term statements such as “(∀x: x is an Antrodemus ↔ x is an Allosaurus)” or “(∀x: x is water ↔ x is H2O)” are proto-­necessities (Section “From Paradigm Terms to Proto-Necessities”) and I argue that semantics (by identifying paradigm terms) and science (by providing knowledge of empirical fact) allow us to reliably attain knowledge of farside truths such as □(∀x: x is an Antrodemus ↔ x is an Allosaurus) or □(∀x: x is water ↔ x is H2O) (Section “Modal Knowledge from Science and Semantics II: Beyond Identities”). Fourth, I conclude that our result defuses Modal Pessimism and I reply to four objections to my argument – that it commits to a controversial modal conventionalism, that it merely rehearses what Kripke and Putnam have taught us long ago, that it fatally relies on far-side modal assumptions and that it fails to translate into an effective epistemic procedure (Section “Conclusions and Objections”). My replies will bring the import of my argument from paradigm terms to modal epistemology at large into sharper focus, or so I hope.

Modal Pessimism and the Argument from Autonomy Far-side modal truths appear to be modally autonomous: they apparently cannot be derived from factual, nomological, logico-mathematical, or conceptual truths even when combined. This allows the Modal Pessimist to argue from the modal autonomy of far-side modal statements to their epistemic elusiveness. I believe that doing so generalizes the more specific pessimistic reasoning in Hawke (2017: 292–297), van Inwagen (1998), and Hawke (2010). The idea is this: sum up the epistemic means by which we can reliably attain knowledge of factual, ­ nomological, ­logico-mathematical, and conceptual truths (if you believe in those) as our ordinary epistemic means.3 These means encompass all epistemic

How (Meta-)Semantics Defuses Modal Pessimism   111 abilities, methods, mechanisms, etc., we draw on in empirical inquiry, logico-mathematical reasoning, and conceptual analysis (if you believe in it). Since autonomous modal truths cannot be derived from factual, nomological, logico-mathematical, and conceptual truths even when combined, it stands to reason that our ordinary epistemic means do not yield far-side modal knowledge. Such knowledge requires extraordinary epistemic means. Since we do not possess any of those, we do not have far-side modal knowledge. In bold generality, then, the Argument from Autonomy is this: 1. We can reliably attain knowledge of far-side modal truths only if we can reliably attain knowledge of autonomous modal truths. 2. We can reliably attain knowledge of autonomous modal truths only if we possess reliable extraordinary epistemic means. 3. We do not possess reliable extraordinary epistemic means. Hence: 4. We cannot reliably attain knowledge of far-side modal truths. I consider this a prima facie forceful argument – even though Kripkean ideas about identity statements will force the Pessimist to weaken her aspirations (see the next section). Its pessimistic conclusion covers all farside truths, including the likes of □(∀x: x is gold ↔ x is Au) rather than merely contentions claims such as ◇(There are philosophical zombies). In spite of this, the argument enlists premises that are not outright pessimistic. The Modal Pessimist is moreover in a good position to defend these, or so I believe. At least for standard challenges to her claims (see Vaidya and Wallner 2021: §1), the Pessimist can argue that these commit to epistemic principles, assumptions, or abilities that in all plausibility aren’t ordinary, or at least not clearly so. I will illustrate this claim presently, simplifying considerably in the process. This topical response raises the bar for standard responses to the Modal Pessimist, compelling the modal rationalist to argue that his principles are unmistakably ordinary and obliging the modal empiricists to explain how the ordinary principles or abilities she employs project to far-side modal truths without the assistance from non-ordinary assumptions. As for the first premise, it is widely held that necessities such as □(∀x: x is gold ↔ x is Au) can be derived from factual truths with the help of general principles (Kripke 1971: 153; Hale 2003; Elder 2004: 5ff; see McLeod 2005: 236 fn. 4 for further references). Bracketing identity statements until the next section, the Modal Pessimist can convincingly point out that it is none too clear that these general principles can themselves be derived from our factual, logico-mathematical, or conceptual knowledge (see also Strohminger and Yli-Vakkuri 2017: 827–830). But if it takes modally autonomous principles to derive the likes of □(Water has chemical composition H2O), far-side modal truths of this ilk are modally autonomous, too.

112  Christian Nimtz As for the second premise, modal empiricists have it that perfectly ordinary epistemic capacities such as counterfactual reasoning (Williamson 2004, 2007: chapter 5; Kment 2006; Kroedel 2017) or inference to the best explanation (Biggs 2011; Williamson 2013; Fischer 2016) allow us to reliably attain knowledge of far-side modal truths. Williamson (2007: chapter 5) rightly insists that once we see that the counterfactual “¬Gold has atomic number 79 □→ ⊥” is true, we can allow us to straightaway infer □(Gold has atomic number 79). However, counterfactually developing “¬Gold has atomic number 79” won’t let us arrive at a contradiction unless we take “Gold has atomic number 79” for a constitutive fact and hold it fixed in our counterfactual development. Taking a page from Roca-Royes (2010: 14–17), Yli-Vakkuri (2013: 616–620), and Vaidya and Wallner (2021), the Modal Pessimist can highlight that it is hard to see how ordinary means should bring about the requisite epistemic sensitivity to constitutive facts. Again deferring the discussion of identity statements to the next section, an abduction from exclusively near-side explananda to the likes of □(∀x: x is gold ↔ x is Au) needs to explain why the resulting modality should be read as metaphysical rather than as nomological. We could do so by assuming that the laws of nature themselves are metaphysically necessary (Fischer 2016: 238–239), or by assuming that nomological relations cannot just as well yield property constitution, as Biggs (2011) takes for granted. The Modal Pessimist can credibly question any such substantive assumption. However, deriving metaphysically modal truths without any calibration from far-side data points mainly on the strength of theoretic virtues or ‘abductive principles’ (Biggs and Wilson 2021: §2) is arguably problematic. The Modal Pessimist can quite rightly insist that when extrapolating from a specific region to the whole of modal space, we need to balance the methodological principle restrictions have to be justified against the principle ruling out variation needs to be justified. This is precisely why everyone will want to see an argument if you insist that the laws of nature hold with metaphysical, rather than with mere nomological necessity; allusions to simplicity won’t do. As for the third premise, Williamson influentially argues that “the postulation by philosophers of a special cognitive capacity exclusive to philosophical or quasi-philosophical thinking looks like a scam” (Williamson 2007: 136). Even modal rationalists who postulate apparently extraordinary capacities commonly explain why these are ordinary after all – say, because a rational intuition is anyway required to account for logico-mathematical knowledge (Ludwig 2010) or because we grasp essences by “understanding what things are, which we must in at least some cases be able to do, on pain of being incapable of thought altogether” (Lowe 2008: 46, 2011). The Modal Pessimist can convincingly reject such aspirations to the ordinary. Assuming that the intuition we supposedly need to account for logico-mathematical knowledge informs us about far-side modal truths is to claim that it

How (Meta-)Semantics Defuses Modal Pessimism   113 reaches beyond what ­follows from logic and mathematics, and thus arguably amounts to postulating an extraordinary capacity. Likewise, assuming that our capacity to understand what things are (which we supposedly need to think at all) to reach beyond how things do and would behave to uncover their essences amounts to assuming an extraordinary epistemic capacity. Although there is no guarantee that the Modal Pessimist’s topical response – charging commitment to de facto exceptionalist principles, assumptions, or abilities – will always work, it seems to work surprisingly well. I take this to warrant a dialectical moral: if we want to prove the Modal Pessimist wrong, we had to better avoid any epistemic principle, assumption, or ability that may be plausibly deemed to be non-ordinary. This is what I will do. I will argue that palpably ordinary means prove the Modal Pessimist’s first premise false even if it is weakened so as to accommodate knowledge of necessary identities. Before I turn to that, let me explain why the Modal Pessimist has to weaken her premise in the first place.

Modal Knowledge from Science and Semantics I: Identities The Argument from Autonomy overstates the Modal Pessimist’s case. Kripke has shown that we can reliably attain knowledge of truths such as □(Cicero = Tully) or □(Proxima Centauri = HIP 70890), or so everyone agrees. But necessary identities are far-side modal truths. So we can reliably attain knowledge of select far-side truths after all. In this section, I rehearse what I take to be the Kripkean argument. Taking my cue from Burgess (2014) and from Kripke (1980: 3, 143) himself, I understand Kripke to show that whenever we know that “a=b” is true and that ‘a’ and ‘b’ designate rigidly, we (are in position to) know that □(a=b). Since my own argument will take up this template, I will go over this well-trodden terrain in some detail. Introducing ‘designator’ as “a common term to cover names and descriptions” (Kripke 1980: 24), Kripke famously decrees: “Let’s call something a rigid designator if in every possible world it designates the same object, a nonrigid or accidental designator if that is not the case” (Kripke 1980: 48). This straightaway entails that “[i]f ‘a’ and ‘b’ are rigid designators, it follows that ‘a = b’, if true, is a necessary truth” (Kripke 1980: 3; see ibid.: 143). In other words: PN A statement “a = b” is a proto-necessity, given that ‘a’ and ‘b’ are rigid designators. Kripke’s PN is a principle about modal truth. But it arguably allows us to derive modal knowledge. Suppose we know these two non-modal truths: ( 1) Proxima Centauri = HIP 70890. (2) ‘Proxima Centauri’ and ‘HIP 70890’ are rigid designators.

114  Christian Nimtz PN allows us to infer from (2) that “Proxima Centauri = HIP 70890” is necessary, if true. Since (1) entails that this identity statement is true, it follows that its necessitation “□(Proxima Centauri = HIP 70890)” is true as well. However, disquoting a statement we understand from a language we speak is arguably knowledge-preserving: if ‘S’ is a statement of our language we understand and if we know that ‘S’ (as uttered by A in c) expresses p and if we know that ‘S’ (as uttered by A in c) is true, we (are in position to) know that p. By disquoting from “□(Proxima Centauri = HIP 70890)”, we arrive at (3) □(Proxima Centauri = HIP 70890). Knowing (1) and (2) thus puts us in a position to know the far-side necessity □(Proxima Centauri = HIP 70890). Generalizing, we find that once we know that “a = b” is true and that ‘a’ and ‘b’ designate rigidly, we (are in a position to) know that □(a = b). Let me dispel two worries. First, (2) requires us to identify rigid designators. This might seem problematic. Kripke’s well-known ‘intuitive test’ (Kripke 1980: 48; see Hughes 2004: 19) lets us identify ‘Nixon’ as a rigid designator by judging that no one other than Nixon could have been Nixon – which arguably is a far-side modal judgment. However, there is no need to rely on Kripke’s test. For many of our terms that designate rigidly, notably those that are de jure rigid (Kripke 1980: 21 fn. 21), semantics – by which I mean standard empirically based semantics – can uncover that they do. Semantics is arguably apt to reveal that we introduce and use ‘Proxima Centauri’ in a way that makes this term designate, for a specific actual object o, something x iff x = o. It may, for example, uncover that we introduce and use some term to designate something x iff x is nothing other than a specific object we picked out as that star. Attaining such a result may require assessing epistemic possibilities such as Kripke‘s Gödel case. But I see no reason to think that it requires far-side modal judgment. (Things are likely to be different for de facto rigid designators such as ‘3’, though.) However, any singular term whose designation is thus unvaryingly and unfailingly tied to a specific actual object designates rigidly (see Kripke 1980: 6, 57; Rami 2019; Nimtz 2019: 343–345). The same holds true for rigidified descriptions such as ‘the actual colour of the sky’ we know to be non-empty. Second, you might feel that there is no need to enlist semantics. Doesn’t “□(Hesperus = Phosphorus)” follow from “Hesperus = Phosphorus” by mere logic? Strohminger and Yli-Vakkuri appear to claim as much when they insist that “one can deduce □(H[esperus]=P[hosphorus]) directly from H[esperus] = P[hosphorus], without the aid of a conditional connecting the two, just as one can deduce p ∨ q directly from p without the aid of p ⊃ (p ∨ q)” (Strohminger and Yli-Vakkuri 2017: 829). Now we may accept “∀x∀y(x = y → □x = y)” as a theorem. But as Burgess stresses, this won’t yield “a = b

How (Meta-)Semantics Defuses Modal Pessimism   115 → □a = b” unless we make s­ ubstantial ­assumptions about the terms ‘a’ and ‘b’ (Burgess 2014: §3; Schwarz 2013: §3). These assumptions may take the form of a rule of instantiation ∀xΦ(x) → Φ(t) that excludes iota-terms ιyΘy (without specifically set scope markers) (Burgess 2014: 1576). Alternatively, they may take the form of an analogously restricted substitution principle a = b → (Φ → Φ[b/a]) (Schwarz 2013: 489). So there is no “short and easy route” (Burgess 2014: 1576) from the insight that “the relations of identity and (metaphysically) necessary identity are coextensive” (ibid.: 1578) to the claim that true identity statements hold with metaphysical necessity. We thus cannot drop the considerations about rigid designation from our case. Given its almost universal acceptance, our result may feel somewhat dull. But it underscores something that decidedly isn’t dull at all: science (by delivering true identity statements) and semantics (by identifying rigid designators) suffice to provide knowledge of far-side necessary identities; there is no need to invoke any additional epistemic principle, assumption, or ability. This neatly preempts the topical defense by the Modal Pessimist whose standard move, as we have seen, is to question any principle, assumption, or ability that fails to be demonstrably ordinary. Where does this leave the Modal Pessimist? I believe that she will have to grant that we can reliably attain knowledge of far-side modal truths. More precisely, the Modal Pessimist will have to grant that we can reliably attain knowledge of necessary identities to the extent that science yields knowledge of identities and semantics yields knowledge of rigid designators – provided, of course, that neither needs to rely on far-side modal judgment to do so. Necessary identities thus turn out to be non-autonomous after all. Accommodating this concession forces the Modal Pessimist to weaken the first premise of her Argument from Autonomy to the following claim: (1*) We can reliably attain knowledge of far-side modal truths beyond necessary identities only if we can reliably attain knowledge of autonomous modal truths. To do so is of course to weaken the conclusion of her argument to the claim that we cannot reliably attain knowledge of far-side modal truths beyond necessary identities. Does this weakened Modal Pessimism still have bite? I think it does. On the one hand, the Modal Pessimist may think that making allowances for necessary identities is not that much of a concession. That Hesperus = Phosphorus is arguably a substantial empirical insight. Yet, inferring □(Hesperus = Phosphorus) does not reveal anything further about Hesperus and very much feels like drawing a trifling logical consequence. On the other hand, the Modal Pessimist can quite rightly insist that just about any list of exemplary metaphysical necessities comprises the likes of □(∀x: x is water ↔ x is H2O). But given that we take statements such as

116  Christian Nimtz these at face value, rather than taking them to express identities between general entities (LaPorte 2013: 94; Martí and Martínez-Fernández 2011), these far-side truths are still covered by her weakened argument. To a good part, then, the challenge to far-side modal knowledge still stands.

Paradigm Terms Semantics The Kripkean argument establishes that science and semantics suffice to provide knowledge of necessary identities. Paradigm term semantics (see Nimtz 2017, 2019, 2021) allows us to expand on this theme. In this section, I explain how paradigm terms function. This will allow me to argue that there are proto-necessities beyond identity statements with rigid designators (Section “From Paradigm Terms to Proto-Necessities”) and that these yield far-side modal knowledge (Section “Modal Knowledge from Science and Semantics II: Beyond Identities”). The first Conférence Générale des Poids et Mesures in 1889 set the length of one meter as the length of a particular stick of platinum-iridium alloy procured for that very purpose and referred to as the prototype du mètre at the temperature of melting ice.4 We may capture the application conditions the conférence thereby fixed for the predicate ‘is one meter long’ thus: METER ‘is one meter long’ applies to something x iff x is of the same length as the actual prototype du mètre at 0°C. Going by METER, then, our predicate ‘is one meter long’ is peculiar in that it applies to something x just in case x stands in a specific relation R to a specific object o as it happens to be here in our actual world (under specific circumstances, or at a specific time). Note that the quantification is unbounded. When we consider an object x here in our actual world and want to get clear whether ‘is one meter long’ applies to it, we consider whether x bears R to o (as o is in our actual world). When we consider an object x in some counterfactual world, we do precisely the same: we also consider whether x bears R to o (as o is in our actual world). Let me express this by saying that ‘is one meter long’ applies to anything anywhere just in case it bears a specific relation R to a specific object as in our actual world (under specific circumstances, or at a specific time). Thus introduced, the predicate ‘is one meter long’ has application conditions that are relationally determined, object involving, and actuality dependent. I call any such expression a paradigm term. I employ the schema ‘F’ to indicate that ‘F’ is a paradigm predicate whose application is determined by the equivalence5 relation R and the actual object(s) collected in the set O @. I call the value structure of the predicate, and the object(s) in O @ the paradigm(s) of the predicate. I will say that the application of a paradigm predicate ‘F’ is governed by the relation R and anchored in the actual items serving as ‘F’-paradigms. On

How (Meta-)Semantics Defuses Modal Pessimism   117 this model, ‘is one meter long’ is a paradigm term with the value structure . Paradigm term semantics takes its inspiration from the Kripke-­Putnam tradition on natural kind terms. In fact, taking natural kind predicates to be paradigm terms neatly captures Putnam’s (1975) and Kripke’s (1980, 2013) own well-motivated commitments. It moreover accounts for semantic and epistemic peculiarities we find natural kind terms to have such as those highlighted in the Kripke-Putnam tradition (Nimtz 2017: §§3–4, 2019: §III, 2021: §6; see also Schwartz 2021). However, the paradigm term approach marks semantics in its own right. It allows us to construe the likes of ‘is water’ or ‘is gold’ as true predicates or ‘appliers’ (Devitt 2005: 147), just as Soames (2002: 246–249) urges us to, thereby avoiding their popular if problematic reduction to singular designators for general entities (see LaPorte 2013: chapter 13; Martí and Martínez-Fernandez 2011; Haraldsen 2018; Nimtz 2019: §I). The paradigm term approach also provides a powerful explanation as to how vernacular terms such as ‘is gold’ or ‘is a star’ fit in with actual research in chemistry or astronomy (Nimtz 2021). Finally, the paradigm term approach covers expressions well beyond natural kind terms, and it deepens our understanding of how externalist predicates such as ‘is gold’ or ‘is one meter long’ work. Let me focus on three aspects of paradigm terms to bring this out. First, paradigm terms share a (meta-)semantic structure. But their subject matter will vary wildly. There is no saying which objects we may use as paradigms and the relations employed could well be geometrical (is of the same length as or has the same shape as), functional (instantiates the same causal role as), natural (is of the same chemical substance as), or what have you. Amongst the paradigm terms, then, we are likely to find predicates such as ‘is one meter long’, ‘is Sydney Opera shaped’, or maybe ‘has been Tom Sawyered’ alongside the natural kind predicates such as ‘is gold’ the Kripke-Putnam tradition focuses on. Second, all we need to do in order to introduce a paradigm term ‘F’ is to pick some actual object(s) to serve as ‘F’-paradigm(s) and to indicate a relation R. For example, we can introduce a paradigm predicate ‘is Sydney Opera shaped’ by picking the relation is of the same shape as and choosing that building over there as the sole paradigm (which, let us assume, happens to be the Sydney Opera House). We thereby delegate the work of fixing what ‘is Sydney Opera shaped’ applies to non-linguistic reality, viz. to the actual shape of Sydney Opera House and to facts about sameness-of-shape. This externalist delegation has important consequences. For one thing, it renders paradigm terms epistemically indulgent. In introducing a paradigm term ‘F’, we can be profoundly ignorant about the relation R and the paradigms O@ beyond what we need to know to pick either out. This ceases to be a mere curiosity when we consider natural kind predicates such as ‘is water’ or ‘is gold’. As Putnam (1975: 232) emphasizes, it is a characteristic of natural kind terms

118  Christian Nimtz that their introduction, if successful, fixes determinate conditions of application together with an intension even though it takes subsequent empirical research to discover what these are. In fact, specifying what ‘is water’ applies to arguably requires two distinct empirical research projects – we need to ascertain the chemical structure of the ‘water’-paradigms and we need to determine what it takes for two samples to be of the same chemical substance (Nimtz 2021; see also Salmon 2005: 166). For another thing, it follows that the introduction of a paradigm term ‘F’ won’t be successful unless extra-linguistic reality is a certain way. Our introduction of ‘is Sydney Opera shaped’ will bestow n ­ on-defective conditions of application on the term only if the paradigm we employ does actually have a determinate shape. Suppose, rather, that Sydney Opera House slowly vacillates between two shapes. Then, no determinate conditions of application will have been fixed for the paradigm predicate. Third, a closer look at the relations R we find in value structures brings out precisely how extra-linguistic reality needs to be to render the introduction of a paradigm term successful. For just about any relation R we find in a value structure, there is a family of determinate properties Φ such that6: REL x bears R to y iff x and y instantiate the same unique determinate property from the family Φ. For example, has the same shape as holds between x and y iff x and y instantiate the same unique determinate property from the family of shape-properties. Let us say that a property φ realizes an equivalence relation R iff, given that x and y instantiate φ, x bears R to y. Let ‘Φ(R)’ stand for the family of R-realizing properties.7 This tells us precisely how non-linguistic reality needs to be for our introduction to be successful: we will have bestowed non-defective conditions of application on ‘F’ just in case the ‘F’-paradigms share a unique R-realizing property (Nimtz 2017: 135, 2019: 340–341). If the ‘F’-paradigms share no R ­ -realizing property, then ‘F’ has an empty intension. So the term will not even apply to its own paradigms, or it will lack paradigms altogether; either way, it proves defective. If the ‘F’-paradigms share more than one ­R-realizing property, then ‘F’ will have a systematically and thereby intolerably indeterminate intension, again proving that we have failed to successfully bestow non-defective conditions of application. If, however, the ‘F’-paradigms do share a unique R-realizing property, then ‘F’ will have non-defective conditions of application ensuring a non-defective intension. This, in turn, explains how the application conditions of a successfully introduced paradigm term depend on the properties of its paradigms. We have just seen that the paradigms of a non-defective paradigm term ‘F’ share a unique R-realizing property. Call it ‘φ’. Now ‘F’ applies to anything x anywhere iff x bears R to the ‘F’-paradigms. But REL assures us that something x bears R to the ‘F’-paradigms iff x instantiates the

How (Meta-)Semantics Defuses Modal Pessimism   119 unique determinate R-realizing property shared by the ‘F’-paradigms, which is none other than φ. Generalizing, we find that the application of any non-defective paradigm term ‘F’ is controlled, as I shall say, by the unique R-realizing property φ shared by the ‘F’-paradigms: ‘F’ will apply to anything x anywhere iff x is φ. By consequence, the intension of ‘F’ will yield for each world w the set of things that instantiate φ in w. In this vein, the determinate length property actually instantiated by the prototype du mètre controls the application of ‘is one meter long’, and the determinate chemical substance property actually shared by the ‘gold’-paradigms – which is arguably none other than the property of being Au (Nimtz 2021) – controls the application of ‘is gold’.

From Paradigm Terms to Proto-Necessities Kripke’s principle PN marks any identity statement that exclusively contains rigid designators as a proto-necessity. Paradigm term semantics reveals that proto-necessity reaches beyond identities and rigidity. As I will argue in this section, paradigm term semantics allows us to establish two principles marking select biconditionals that involve paradigm terms as proto-necessities.8 Here is the first proto-necessity principle we can derive from paradigm term semantics: PN* Any statement “∀x: Fx ↔ Gx” is a proto-necessity, given that ‘F’ and ‘G’ are two non-defective paradigm terms governed by the same relation R. In order to establish PN*, let us consider an example. Suppose that this holds true: (4) ‘is an Allosaurus’ and ‘is an Antrodemus’ are non-defective paradigm terms governed by the same relation R D. (For an intuitive illustration, assume that R D is is of the same kind of dinosaur as.) We want to show that “∀x: x is an Allosaurus ↔ x is an Antrodemus” is a proto-necessity. So we can assume that it is true. This ensures that ‘is an Allosaurus’ and ‘is an Antrodemus’ actually co-apply and hence have a shared extension in @. This shared actual extension cannot be empty, since a non-defective paradigm term needs to apply to its paradigms. Now ‘is an Allosaurus’ applies to x iff x bears R D to the ‘Allosaurus’-paradigms, and ‘is an Antrodemus’ applies to x iff x bears R D to the ‘Antrodemus’-paradigms. So any item(s) in the shared extension of these terms bear(s) R D both to the ‘Allosaurus’-paradigms and to the ‘Antrodemus’-paradigms. Since R D is an equivalence relation, it follows that the ‘Allosaurus’-paradigms bear the relation R D to the ‘Antrodemus’-paradigms.

120  Christian Nimtz This gives what we want. For the R D-relatedness of the ‘­Allosaurus’-paradigms and the ‘Antrodemus’-paradigms guarantee that our two terms co-apply everywhere. In order for ‘is an Allosaurus’ to apply to something x in world w, x in w has to bear R D to the ‘Allosaurus’-­paradigms in @. But if these paradigms bear R D to the ‘Antrodemus’-paradigms in @, the transitivity of R D ensures that x bears R D to the latter paradigms as well – and thereby satisfies ‘is an Antrodemus’. The same holds mutatis mutandis for ‘is an Antrodemus’. By consequence, ‘is an Allosaurus’ and ‘is an Antrodemus’ co-apply in all worlds, which entails that “∀x: x is an Allosaurus ↔ x is an Antrodemus” is necessary. So “Allosauruses are Antrodemuses” is indeed a proto-necessity. However, since all our assumptions are licensed by PN*, our result generalizes to prove this principle. Here is the second proto-necessity principle flowing from paradigm term semantics: PN** Any statement “∀x: Fx ↔ Gx” is a proto-necessity, given that the descriptive term ‘G’ attributes a property that realizes the relation governing the non-defective paradigm term ‘F’. Let us again consider an example. Suppose that the following holds true: (5) ‘is water‘ is a non-defective paradigm term governed by a relation R S and ‘is H2O’ is a descriptive predicate attributing a property φ# that realizes R S. (For an intuitive illustration, assume that R S is the relation is of the same chemical substance as and that φ# is being H2O.) This yields important consequences. As shown above, the non-defectiveness of the paradigm term ‘is water’ ensures that the ‘water’-paradigms share a unique ­R S-realizing property; call it ‘φ*’. This property controls the term’s application: ‘is water’ applies to anything x anywhere iff x is φ*. We thus find that the property φ* controlling the application of ‘is water’ and the property φ# attributed by ‘is H2O’ belong to the same family, viz. the family of properties realizing R S. We want to show that “∀x: x is water ↔ x is H2O” is a proto-necessity. So we may assume that this biconditional is true. Hence, ‘is water’ and ‘is H2O’ co-apply in @. But ‘is water’ applies to something x iff x instantiates φ*. And ‘is H2O’ applies to something x iff x instantiates φ#. It follows that any item in their shared actual extension instantiates both φ* and φ#. This holds true of the ‘water’-paradigms, which have to be in this shared extension. It follows that any of the ‘water’-paradigms instantiates both φ* and φ#. We thus can conclude that φ* = φ#. Since the ‘water‘-paradigms cannot instantiate more than one R s-realizing property without rendering the paradigm term ‘is water’ defective, φ* and φ# must be identical. But if the property φ* just is the property φ#,

How (Meta-)Semantics Defuses Modal Pessimism   121 ‘is water’ and ‘is H2O’ co-apply in all possible worlds. We thus find that “∀x: x is water ↔ x is H2O” is indeed necessary, if true. However, since all our assumptions are licensed by PN**, our result generalizes to prove this principle. Paradigm term biconditionals of this variety are indeed proto-necessities.

Modal Knowledge from Science and Semantics II: Beyond Identities PN* and PN** assure us that select biconditionals involving paradigm terms are proto-necessities. These results about modal truth yield results about modal knowledge. Assume that scientific investigation uncovers these non-modal truths: (6) ∀x: x is an Allosaurus ↔ x is an Antrodemus. (7) ∀x: x is water ↔ x is H2O. Assume further that semantic investigation allows us to see that (4) and (5) are in fact true. Then, PN* allows us to infer that the modal statement “□(∀x: x is an Allosaurus ↔ x is an Antrodemus)“ is true. And PN** allows us to infer that the modal statement “□(∀x: x is water ↔ x is H2O)” is true as well. Since disquoting a statement we understand from a language we speak is arguably knowledge-preserving, as stressed above, it follows that we know these two far-side modal facts: (8) □(∀x: x is an Allosaurus ↔ x is an Antrodemus). (9) □(∀x: x is water ↔ x is H2O). It follows that if semantics and science allow us to reliably attain knowledge of the sort given in (4)–(7), we can reliably attain far-side modal knowledge beyond identities and rigidity. We thus need to ask whether semantics and science can do so. As in the case of de jure rigid designators, I see semantics well-placed to discover that certain of our predicates are paradigm terms. Semantic inquiry is manifestly apt to reveal that we introduce and use, say, ‘is an Antrodemus’ in a way that makes this term apply to something x iff x bears a specific relation R to (a) specific actual object(s) – say, the relation is of the same kind of dinosaur as to the exemplar(s) (a) specific bone fragment(s) stem(s) from. But any term whose application conditions are thus governed by a relation and anchored in actual items is a paradigm term. As stressed above, semantic inquiry also shows that we can explain semantic and epistemic peculiarities we find natural kind predicates to have once we take these to be paradigm terms. This shows that natural kind predicates quite generally belong to the encompassing semantic category of paradigm terms, or so I contend.

122  Christian Nimtz Turning to science, I take it as given that science is well-placed to discover general truths such as (6) or (7). There is nothing trivial about finding out whether all Allosauruses are Antrodemuses (Madsen 1993), and vice versa, or whether any portion of water is a portion of H2O (Schulte 2020), and vice versa, or whether any sample of gold is a sample of Au (Nimtz 2021). Yet in each of these cases, science is arguably well up to the job. Note that my case does not rest on the specific examples. If you believe with Häggqvist and Wikforss (2018) that (9) is de facto false, just substitute a case of your choice (and see Martí and Hoefer 2019 for objections). Science is also arguably apt to procure the empirical insights requisite to showing that a paradigm term such as ‘is water’ is non-defective. As explained above, this will be so just in case the ‘water’-paradigms share a unique determinate property φ that realizes the relation is of the same chemical substance as. Chemistry is manifestly well apt to uncover the chemical characteristics of (paradigmatic) water-samples and to determine what it takes for portions to be of the same chemical substance, just as astronomy is arguably well apt to uncover the characteristics of (paradigmatic) stars and to determine what it takes for celestial things to be of the same variety of astrophysical objects (Nimtz 2021: §7). I conclude that we can reliably attain far-side modal knowledge beyond identities and rigid designation. These modal truths turn out to be non-autonomous after all. More precisely, I conclude that we can reliably attain far-side modal knowledge beyond identities at least to the extent that semantics and science reliably uncover truths such as (4)–(7) – again, provided that neither needs to rely on far-side modal judgment to do so.

Conclusions and Objections I have argued that by allowing us to procure the likes of □(∀x: x is an Allosaurus ↔ x is an Antrodemus) and □(∀x: x is water ↔ x is H2O), science and semantics allow us to reliably attain far-side modal knowledge beyond identities and rigidity. Given the Kripkean result about identity statements, it follows that those far-side modal truths that figure on just about any list of exemplary far-side modal truths are within our epistemic ken. I understand this to defuse Modal Pessimism. This upshot substantiates the assumption that our ability to attain modal knowledge projects to far-side modal truths, which is what the Pessimists challenged us to do. It moreover does so with demonstrably ordinary epistemic means, viz. science and semantics. To be sure, I have not argued that we can reliably pass judgment on more contentious far-side claims such as ◇(There are philosophical zombies). However, I find it hard to see how focusing on these could possibly allow the Pessimist to throw doubt on our general ability to reliably attain knowledge of farside modal truths.

How (Meta-)Semantics Defuses Modal Pessimism   123 Let me take up four potential misgivings about my argument. I hope that these bring the stance I take in modal epistemology into clearer focus. •

You argue from semantics to modal truth. Doesn’t that amount to a controversial modal conventionalism or modal normativism (Thomasson 2021)?

I don’t think so. Nobody believes that the Kripkean argument from rigid designation undercuts the realist credentials of □(Hesperus = ­Phosphorus). I plead the same for the far-side modal truths we can derive via paradigm term semantics. •

Kripke and Putnam have already convinced everyone that “Cicero = Tully” or “Water is H2O” hold with metaphysical necessity. So where’s the progress?

As for semantic externalism, I believe that paradigm terms semantics provides a fruitful general understanding of the Kripkean tradition (see Nimtz 2019), but that is controversial. As for modal epistemology, my argument is not meant to instill new modal beliefs. I rather aim to substantiate the contention that our ability to attain modal knowledge projects to truths such as □(∀x: x is gold ↔ x is Au). True, Kripke and Putnam have convinced many that we do know the likes of □(Cicero = Tully) or □(Water is H2O). But their reasoning is manifestly interwoven with far-side essentialist ideas (Hawke 2017: 297). In fact, this is true to an extent that their kind term semantics is often understood to rest on essentialist presuppositions, and thus to fail should these be proven false (Häggqvist and Wikforss 2018). Since my argument aims to defuse Modal Pessimism, it cannot be premised on any contentious far-side modal assumption. •

But doesn’t your own argument rely on far-side modal assumptions?

My argument enlists innocuous background assumptions about modal space. For instance, I assume that there are possible worlds, that individuals may occur in more than one world, and that items may stand in trans-world relations such that a heated metal bar in a nearby world will bear is longer than to the otherwise qualitatively indistinguishable prototype du mètre as we actually find it to be. My argument does not rely on specific far-side modal assumptions beyond these, or so I contend. This requires me to insist that in combining unbounded quantification with the idea that the application conditions of certain terms are fixed by whatever bears a certain relation to (a) specific actual object(s), I do not incur a far-side modal commitment. Suppose I pick an actual

124  Christian Nimtz individual o and decree that ‘Aristotle’ designates any x anywhere just in case x = this very individual o here. To do so is to build this very object as it is here and the relation is identical with into the designation conditions of ‘Aristotle’. Doing so yields modal consequences. But my stipulation is exclusively this-worldly, or so I hold. The analogous holds true for certain equivalence relations besides identity, or so I contend. Pace Steinberg (2020), given that we trust cladistics, we do learn from science rather than from metaphysics that sameness of clade (which of course includes sameness of an actual common ancestor) suffices for sameness of ­biological species. •

Deriving modal truths via paradigm term semantics requires an intricate argument from non-obvious semantic insights; doing so is also more complicated than deriving modal truths from identities. So how does the paradigm term result translate into an effective epistemic procedure?

This is important and requires a nuanced reply. First, the Kripkean argument provides a general assurance. There is no need to run through the argument whenever you acquire knowledge of specific instances such as, say, □(Cicero = Tully). Suppose you are disposed to believe the necessitation of any identity statement with typical proper names you know to be true. This should suffice for you to know that □(Cicero = Tully), provided that typical proper names de facto designate rigidly (compare Strohminger and Yli-Vakkuri 2017: 829). In much the same vein, the paradigm term argument provides a general assurance. There is no need to run through the argument to learn, say, that □(∀x: x is an Antrodemus ↔ x is an Allosaurus). Second, I agree that the paradigm term derivation is more complicated than any argument via identities. However, I contend that it is a much better fit with actual scientific practice than any hand-waving philosophical assurance that science procures property identities, for it allows us to see precisely how actual inquiry in the natural sciences bears on the truths of necessities concerning gold, water, or stars (Nimtz 2021). Third, and this is the key point, my aim was not to identify an epistemic procedure. My aim was to show that science and semantics provide all we need in order to reliably attain select far-side modal knowledge in a way that requires no additional epistemic procedure. That said, our result of course bears upon the procedures or heuristics we employ in modal epistemology. For our paradigm result provides precisely those ‘modally saturated data points’ that Clark-Doane (2019: 268) demands. Unlike the valuable modal data points we get from our empirical theories (see Fischer 2016), from scientific results (Williamson 2016) or from direct experience of modal properties (Vetter, this volume), the data points we get from our

How (Meta-)Semantics Defuses Modal Pessimism   125 paradigm term result are moreover far from home. So we can employ them to effectively calibrate whatever abductive procedures we use to navigate the depth of modal space.

Notes 1 I will talk of modal truths and our ability to attain modal knowledge rather than of modal propositions and our ability to pass judgment on these. I take it that my argument could be recast in the more cautious yet also more tortuous latter idiom. 2 Following Burgess and Sherman (2014: 1–8), I take it that semantics aims to uncover which intensions our terms have, whereas meta-semantics aims to explain why our terms have their respective intensions. Nothing in my argument hinges on this. 3 In my picture, the combination of factual, nomological, logico-mathematical, and conceptual truths plays the role of Hawke’s (2017: 294) ‘modally safe base’. 4 I ignore subsequent changes in the meter convention and pretend that the abrogation of the definition by the 11th Conférence Générale in 1960 never happened. 5 I assume that by default, R is an equivalence relation. This is a simplification. A paradigm term might well enlist a relation such as has double the length. 6 The relation is identical with is a plausible exception; see Nimtz (2019): §IV. Following Steinberg (2020: 405), REL is a logical truth. 7 Note that R-realizing properties may well be complex and/or multiply realized. 8 See Nimtz (2017) for a more comprehensive version of this argument; see also Nimtz (2019: 345–349). See also Steinberg (2020) for a condensed version.

References Biggs, S. (2011) “Abduction and Modality,” Philosophy and Phenomenological Research 83(2), 283–326. Biggs, S. and Wilson, J. (2021) “Abduction Versus Conceiving in Modal ­Epistemology,” Synthese 198, 2045–2076. DOI: 10.1007/s11229-019-02117-9. Burgess, A. and Sherman, B. (2014) “Introduction: A Plea for the Metaphysics of Meaning,” in A. Burgess and B. Sherman (eds.), Metasemantics: New Essays on the Foundations of Meaning, Oxford: Oxford University Press, 1–16. Burgess, J. P. (2014) “On a Derivation of the Necessity of Identity,” Synthese 191, 1567–1585. Clarke-Doane, J. (2019). “Modal Objectivity,” Noûs 53(2), 266–295. Devitt, M. (2005) “Rigid Application,” Philosophical Studies 125(2), 139–165. Elder, C. L. (2004) Real Natures and Familiar Objects, Cambridge, MA: MIT Press. Fischer, B. (2016) “A Theory-Based Epistemology of Modality,” Canadian Journal of Philosophy 46(2), 228–247. Häggqvist, S. and Wikforss, Å. (2018) “Natural Kinds and Natural Kind Terms: Myth and Reality,” The British Journal for the Philosophy of Science 69(4), 911–933. Hale, B. (2003) “Knowledge of Possibility and of Necessity,” Proceedings of the Aristotelian Society 103, 1–20.

126  Christian Nimtz Haraldsen, F. (2018) “Rigidity and Triviality,” Synthèse 195(5), 1993–1999. Hawke, P. (2017) “Can Modal Skepticism Defeat Humean Skepticism?” in Fischer, B. and Leon, F. (eds.), Modal Epistemology After Rationalism, ­Dordrecht: Synthese Library, 281–308. Hawke, P. (2010) “Van Inwagen’s Modal Skepticism,” Philosophical Studies 153(3), 351–364. Hughes, C. (2004) Kripke: Names, Necessity, and Identity, Oxford: Oxford ­University Press. Kment, B. (2006) “Counterfactuals and the Analysis of Necessity,” Philosophical Perspectives 20, 237–302. Kripke, S. (1971) “Identity and Necessity,” in Munitz, M. (ed.) Identity and ­Individuation, New York: New York University Press, 135–164. Kripke, S. (1980) Naming and Necessity, Oxford: Blackwell. Kripke, Saul (2013). Reference and Existence, Oxford: Oxford University Press. Kroedel, T. (2017). “Modal Knowledge, Evolution, and Counterfactuals,” in Fischer, B. and Leon, F. (eds.), Modal Epistemology After Rationalism, ­Dordrecht: Synthese Library, 179–195. Kung, P. (2017) “Personal Identity without Too Much Science Fiction,” in Fischer, B. and Leon, F. (eds.), Modal Epistemology After Rationalism, ­Dordrecht: ­Synthese Library, 133–154. LaPorte, J. (2013) Rigid Designation and Theoretical Identities, Oxford: Oxford University Press. Lowe, E. J. (2008) “Two Notions of Being: Entity and Essence,” Philosophy 83, 23–48. Lowe, E. J. (2011) “The Rationality of Metaphysics,” Synthese 178, 99–109. Ludwig, K. (2010) “Intuitions and Relativity,” Philosophical Psychology 23(4), 427–445. Madsen, J. H., Jr. (1993) [1976]. Allosaurus fragilis: A Revised Osteology. Utah Geological Survey Bulletin 109, 2nd edition, Salt Lake City: Utah Geological Survey. Martí, G. and Martínez-Fernández, J. (2011) “General Terms, Rigidity and the Trivialization Problem,” Synthese 181(2), 277–293. Martí, G. and Hoefer, C. (2019). “Water Has a Microstructural Essence after All,” European Journal for Philosophy of Science 9(1), 1–15. McLeod, S. (2005) “Recent Work on Modal Epistemology,” Philosophical Books 46(3), 235–245. Nimtz, C. (2012) “Conceptual Truths, Strong Possibilities and Our Knowledge of Metaphysical Necessities,” Philosophia Scientiae 16(2), 39–58. Nimtz, C. (2017) “Paradigm Terms: The Necessity of Kind Term Identifications Generalized,” Australasian Journal of Philosophy 95(1), 124–140. Nimtz, C. (2019) “Kripkean Meta-Semantics and Generalized Rigidity,” ­Philosophical Quarterly 69(275), 332–353. Nimtz, C. (2021). “How Science and Semantics Settle the Issue of Natural Kind Essentialism,” Erkenntnis 86, 149–170. Putnam, H. (1975). “The Meaning of ‘Meaning’,” in Putnam, H. (1975), Mind, Language, and Reality: Philosophical Papers Vol. 2, Cambridge: Cambridge University Press, 215–71. Rami, D. (2019). “Names and Their Kind of ­Rigidity,” Erkenntnis 84(2), 257–282.

How (Meta-)Semantics Defuses Modal Pessimism   127 Roca-Royes, S. (2010) “Conceivability and De Re Modal Knowledge,” Nous 45(1), 1–28. Roca-Royes, S. (2017) “Similarity and Possibility: An Epistemology of De Re ­Possibility for Concrete Entities,” in Fischer, B. and Leon, F. (eds.), Modal Epistemology After Rationalism, Dordrecht: Synthese Library, 221–246. Salmon, N. (2005) Reference and Essence, 2nd edition. Oxford: Blackwell. Schulte, P. (2020). “Why Mental Content Is Not Like Water: Reconsidering the Reductive Claims of Teleosemantics,” Synthese 197(5), 2271–2290. Schwartz, S. (2021) “Against Rigidity for Natural Kind Terms,” Synthese 198, 2957–2971. Schwarz, W. (2013) “Contingent Identity,” Philosophy Compass 8(5), 486–495. Soames, S. (2002) Beyond Rigidity. The Unfinished Semantic Agenda of Naming and Necessity, Oxford: Oxford University Press. Steinberg, A. (2020) “Paradigmatic Metaphysics,” Australasian Journal of ­Philosophy 98(2), 403–409. Strohminger, M. and Yli-Vakkuri, J. (2017). “The Epistemology of Modality,” Analysis 77(4), 825–838. Strohminger, M. and Yli-Vakkuri, J. (2018). “Moderate Modal Skepticism,” in M. A. Benton and J. Hawthorne, D. Rabinowitz (eds.), Knowledge, Belief, and God: New Insights in Religious Epistemology, Oxford: Oxford University Press, 302–321. Thomasson, A. (2021). “How Can We Come to Know Metaphysical Modal Truths?,” Synthese 198(Suppl 8), 2077-2106. Vaidya, A. J. and Wallner, M. (2021). “The Epistemology of Modality and the Problem of Modal Epistemic Friction,” Synthese 198(Suppl 8), 1909–1935. van Inwagen, P. (1998) “Modal Epistemology,” Philosophical Studies 92(1), 67–84. Vetter, B. (2016). “Williamsonian Modal Epistemology, Possibility-Based,” Canadian Journal of Philosophy 46(4–5), 766–795. Vetter, B. (this volume). “An Agency-Based Epistemology of Modality”, 44–69. Williamson, T. (2007) The Philosophy of Philosophy, Oxford: Blackwell. Williamson, T. (2013) “Methodological Afterword,” in T. Williamson (ed.), Modal Logic as Metaphysics, Oxford: Oxford University Press, 423–429. Williamson, T. (2016) “Modal Science,” Canadian Journal of Philosophy 46(4–5), 453–492. Yli-Vakkuri, J. (2013) “Modal Skepticism and Counterfactual Knowledge,” Philosophical Studies 162(3), 605–623.

6 How Things Have to Be1 Nathan Salmon

Putnam on Natural-Kind Terms According to Saul Kripke’s intended notion from his landmark Naming and Necessity (Kripke 1980), a term τ designates an object x rigidly (and is a rigid designator of x) if τ designates x with respect to every possible world in which x exists and does not (in the same use) designate anything else with respect to any possible world. In his classic and marvelously creative paper ‘Meaning and Reference’,2 Hilary Putnam proffers a theory according to which natural-kind terms, like ‘water’ and ‘tiger’, are rigid designators. He writes: When I say “this (liquid) is water,” … the force of my explanation [of the meaning of ‘water’] is that “water” is whatever bears a certain equivalence relation (the relation [x is the same liquid as y] we called “sameL” above) to the piece of liquid referred to as “this” in the actual world. We might symbolize [this] in the following way. …: [(OD)]  (For every world W) (For every x in W) (x is water ≡ x bears sameL to the entity referred to as “this” in the actual world W1) … If we extend [Kripke’s] notion of rigidity to substance names, then we may express Kripke’s theory and mine by saying that the term ‘water’ is rigid. The rigidity of the term ‘water’ follows from the fact that when I give the “ostensive definition”: “this is water,” I intend [(OD)] … (p. 707). Later in Putnam’s discussion, it emerges that (OD) is a misformulation, or at least highly misleading. Instead, his intent is that x is water in a possible world w if and only if x is the same liquid in w that the exemplar is in the actual world. Whether the exemplar is also water in w, as (OD) evidently requires, is altogether irrelevant. What matters is that x as it is in w be the same liquid as the exemplar is as it is in the actual world. A fourplace cross-world relationship is invoked: x is the same liquid in w1 that y is in w2. Putnam understands such a four-place cross-world relationship DOI: 10.4324/9781003002192-8

How Things Have to Be  129 as a binary relation between a pair of peculiar entities: x-in-w1 and y-in-w2. These are entities-as-they-are-in-particular-worlds. Although x itself exists in infinitely many possible worlds, such an entity as x-in-w exists in w and in no other world. In Reference and Essence (1981), I call such putative entities ‘possible-world slices’, on an analogy to time slices, or temporal stages (phases). Thus, (OD) should be reformulated in conformity with Putnam’s intentions as follows: (OD′) (w)(x)[ x existsw ⊃ (x is waterw ≡ x-in-w is the sameL as thisin-W@) ], where W@ is the actual world and where sameL is the equivalence same-liquid relation between possible-world slices of individuals. Putnam continues with characteristic vigor: What Kripke was the first to observe is that this theory of the meaning (or ‘use,’ or whatever) of the word ‘water’ (and other natural-kind terms as well) has startling consequences for the theory of necessary truth. … … we can understand the relation sameL (same liquid as) as a crossworld relation by understanding it so that a liquid in world W1 which has the same important physical properties (in W1) that a liquid in W2 possesses (in W2) bears sameL to the latter liquid. Then the theory we have been presenting may be summarized by saying that an entity x, in an arbitrary possible world, is water if and only if it bears the relation sameL (construed as a cross-world relation) to the stuff we call “water” in the actual world. … Once we have discovered that water (in the actual world) is H2O, nothing counts as a possible world in which water isn’t H2O. On the other hand, we can perfectly well imagine having experiences that would convince us (and that would make it rational to believe that) water isn’t H2O. In that sense, it is conceivable that water isn’t H2O. It is conceivable but it isn’t possible! Conceivability is no proof of possibility. … a statement can be (metaphysically) necessary and epistemically contingent. Human intuition has no privileged access to metaphysical necessity. (pp. 708–709) In the preface to the second edition of Salmon (1981) (the first edition was a revision of my doctoral dissertation), I said, ‘The central argument of Part II—that Kripke and Putnam made unsubstantiated claims, indeed false claims, to the effect that the theory of direct reference has nontrivial modal-essentialist import—initially met with controversy, but few writers today would dispute it’ (p. xvii).3 The year after these words were published saw (Mackie 2006), which does just that, or something rather

130  Nathan Salmon close to it. More accurately, Mackie argues that the central argument of part II of Salmon (1981) is correct in spirit but not in letter. The primary bone of contention between Mackie (2006) and Salmon (1981) is whether the case has been made that Putnam’s direct-reference theory of natural-kind terms, taken in conjunction with empirical or otherwise uncontroversial premises, yields conclusions that are both ­modal-essentialist and non-trivial, such as that the sentence ‘Water is H2O’ expresses a necessary truth about water. One of the central tenets of Putnam’s direct-reference theory is that the meaning of the substance term ‘water’ can be explained in an ostensive definition, ‘This liquid is water’, said with reference to a paradigmatic water sample. The ostensive definition itself, upon analysis, allegedly entails that ‘water’ rigidly designates a particular liquid substance. On my reconstruction of Putnam’s theory, the ostensive definition is deceptive in that its underlying logical form is modal, complex, and indeed modally complex. Given Putnam’s purposes, however, quantification over possible-world slices is neither desirable nor necessary. (Salmon (1981) includes an account of cross-world relations that eliminates all reference to world slices.) Salmon (1981) distinguished first of all between a general term, like ‘water’ or ‘tiger’, and the corresponding predicate formed from the term in combination with the ‘is’ of predication serving as copula: ‘is water’ or ‘is a tiger’. Drawing this distinction enables us to say that while ‘water’ designates a certain liquid substance, the extension of the corresponding predicate ‘is water’ is not this same substance but the set or class of water samples (or the set’s characteristic function, etc.). Here, I shall represent the copula by the partially artificial notation ‘is{___}’, where the space between the braces is to be filled with a general term. If ν is a general term, we thus represent the corresponding predicate as ┌is{ν}┐. If ν is a count noun, the predicate may be represented instead as ┌ is-a{ν}┐, though the supplementing indefinite article is regarded as nothing more than syntactic garnish (and not, for example, as indicating existential quantification). In either case, the predicate applies to all and only the instances of the kind (or other universal) designated by ν.4 The predicate ‘is water’, formalized as ‘is{water}’, might thus be read ‘is an instance of water’ or ‘is a water sample’. The intended underlying logical form of the ostensive definition ‘This liquid is water’ is now reformulated as follows, where the demonstrative ‘this’ again designates a paradigmatic water sample5: (POD water)  □(x)( x is{water} ≡ x is{dthat[the liquid substance ℓ: this is{ℓ}]} ). This Putnamian ostensive definition has the intended consequence that a substance sample is water in a possible world w if and only if that

How Things Have to Be  131 sample is an instance in w of the very same-liquid substance ℓ that the indicated exemplar is an instance of in the actual world. Thus, according to (­POD water), there is a particular liquid substance ℓ—the very liquid of which the exemplar is a sample—such that necessarily, any sample of ℓ is water and only samples of ℓ are water. By definition, water necessarily coincides with a liquid substance. That is not all. Putnam agrees with Kripke that ‘water’ rigidly designates the particular chemical compound H 2O. In Salmon (1981), I argued that there is no distinction among general terms, as there is for singular terms, between rigid and non-rigid designation, and that for general terms, designation and rigid designation are one and the same (pp. 44–54). I offered as a criterion for rigid designation of a general term ν that a sentence of the following form be true, where β is a term that designates a kind (sort, type, category) or other universal (pp. 73–75): (C)  □(x)( x is{ν} ≡ x is{dthat[β]}). This proposed criterion was encouraged by Putnam’s remark that the rigidity of ‘water’ follows from (POD water).6 Salmon (1981) defined notions of metaphysical extension and metaphysical intension for universals (pp. 46–47). The metaphysical extension of a universal U in a possible world w is simply the set or class of U’s instances in w. The metaphysical intension of U is then the function that assigns to each possible world w, U’s metaphysical extension in w. In these terms, the truth of (C) requires that the semantic intension of ┌is{ν}┐ and the metaphysical intension of the universal designated by β be the same. This is what it is, I argued, for ν to designate rigidly the same kind that β designates. The truth of (POD water) requires that the semantic intension of ‘is water’ be the same as the metaphysical intension of the liquid ℓ of which the exemplar is an instance. As Salmon (1981) notes (pp. 150–151), by the proposed criterion, (POD water) has the consequence that ‘water’ therewith rigidly designates a particular liquid substance—the very liquid ℓ in question.

Non-Rigid General Terms Putnam’s and my earlier claim that the rigidity of ‘water’ follows from (POD water) is, I have come to see, mistaken. Salmon (1981) overlooked a relevant logical possibility, brought to my attention by Linsky (1984). This is the possibility that the semantic content (or “meaning”) of ‘water’ is the same as that of a non-rigid general term. Can there be such a thing as a non-rigid general term? Even a descriptive common-noun phrase might be regarded as rigidly designating a

132  Nathan Salmon certain kind. For example, the phrase ‘unmarried man’ may be seen as designating a certain kind of man: the kind, bachelor. But if the phrase designates that kind, it does so rigidly. It is logically possible that ‘water’ is a non-rigid general term. To see why this is so, it must first be observed that even if definite descriptions are typically singular terms—as John Stuart Mill, Gottlob Frege, and many others have taken them to be, contrary to Russell’s theory—some definite descriptions are in fact general terms rather than first-order singular terms. (General terms might be regarded as second-order singular terms.) This is illustrated by the predicate-nominative use of ‘the color of the sky’ in. Henry’s favorite shirt is the color of the sky. The verb in this sentence is not the ‘is’ of identity. It is the familiar ‘is’ of predication, the same verb that occurs in ‘Henry’s favorite shirt is blue’. In fact, although it is a noun phrase the description plays the very same semantic role as the single word ‘blue’ in its adjectival use. Both are predicative, both designate a color (indeed the same color), etc. In fact, the definite description ‘the liquid substance ℓ: this is{ℓ}’ occurring in (POD water) is a general-term definite description rather than a singular-term definite description.7 One reason this is significant is that traditional theories have hypothesized that a term like ‘water’ abbreviates a non-rigid general description like ‘the colorless, odorless, thirst-quenching liquid in Earth’s rivers, lakes, streams, and oceans’. The theories of Frege and Russell were basically of this sort. Even Mill, whose theory of ordinary proper names has a remarkably contemporary flavor, held a theory of general terms very similar to Frege’s and Russell’s. Such descriptional theories of general terms are the foil to the direct-reference theory, which diametrically opposes them. Let us abbreviate the general (non-singular) definite description, ‘the liquid that covers most of the Earth’s surface’ as ‘earthliquid’.8 Then, the identity sentence ‘Water = earthliquid’ is true although not necessary. The new term ‘earthliquid’ is, like the general description it abbreviates, a descriptional general term that applies to something x with respect to a possible world w if and only if x is a sample in w of the liquid that covers most of the Earth’s surface in w. In the actual world, a liquid sample is earthliquid if and only if it is water. Let Wm be a possible world in which most of the Earth’s surface is covered by methane. In Wm , a bit of liquid is a sample of earthliquid if and only if it is methane. There is a special corresponding kind: Sample of the liquid that uniquely covers most of the Earth’s surface—or as it may now be abbreviated, Sample of earthliquid. This kind is to be sharply distinguished from water. Although the two kinds, water and Sample of earthliquid, have a common metaphysical extension, they differ in metaphysical intension. In Wm , a sample of water is an instance of the former kind and not the

How Things Have to Be  133 latter, whereas a sample of methane is an instance of the latter kind and not the former. Then, the following modal generalization may be asserted: □(x)( x is{earthliquid} ≡ x is-a{dthat[Sample of earthliquid]}). That is, in any possible world w, something is earthliquid in w iff it is an instance in w of the particular kind, Sample of earthliquid. By the criterion proposed in Salmon (1981), ‘earthliquid’ thus rigidly designates the kind, Sample of earthliquid. But this is mistaken. The term designates water (and nothing else). Water is a different kind from Sample of earthliquid. In the actual world, any liquid sample that is of one kind is also of the other, but there might have been liquid samples that were of one kind and not the other. That is to say, the two kinds share a common metaphysical extension but they differ in metaphysical intension. Furthermore, while ‘Sample of earthliquid’ designates rigidly, ‘earthliquid’, like the description it abbreviates, designates non-rigidly. There is thus a distinction even among general terms between rigid and non-rigid designation. Routinely, general definite descriptions are nonrigid designators. The truth of (C) cannot be taken as a criterion of (rigid) designation for general terms. Although the truth of (C) is indeed a necessary condition for a general term ν to designate rigidly the universal designated by β, it fails as a sufficient condition when ν is a non-rigid general definite description. However, where β is a non-rigid general description, the truth of (C), because of the presence of the rigidifying ‘dthat’ operator, precludes the possibility that ν abbreviates β. If (POD water) is true, then whatever ‘water’ designates, the semantic intension of ‘is water’ coincides with the metaphysical intension of a particular liquid substance, the substance of which the exemplar is an instance. The defect in Salmon (1981)’s criterion for general-term rigidity is reparable. Following the case of singular terms,9 we may say that a general term ν rigidly designates the universal designated by β if and only if the following is true: (C′)  □(dthat[β] = ν), The original (C) is a trivial logical consequence. But the two are not equivalent. The intended underlying logical form of the ostensive definition ‘This liquid is water’ is properly given not by (POD water) but by (OD water)  □(dthat[ the liquid substance ℓ: this is{ℓ}] = water). In English, ‘Necessarily, that [the liquid of which this is an instance] is water’.10

134  Nathan Salmon According to this new criterion, ‘water’ rigidly designates water, whereas the descriptive term ‘earthliquid’ (which also designates water) does not rigidly designate. The general term ‘sample of earthliquid’ also does not rigidly designate water, although it rigidly designates the kind Sample of earthliquid. This repair foregoes Putnam’s and my former claim that the rigidity of ‘water’ follows from (POD water) itself. It follows trivially from (OD water), which also trivially entails Putnam’s variant, (POD water). Perhaps the truth of (POD water) together with the observation that the liquid substance of which the exemplar is an instance is a natural kind precludes the possibility that ‘water’ designates only non-rigidly. For what can it non-rigidly designate if the semantic intension of ‘is water’ coincides with the metaphysical intension of a natural kind? (Notice that the semantic intension of the non-rigid general term ‘is earthliquid’ coincides with the metaphysical intension of a non-natural kind.) One way or another, (POD water) must be strengthened to yield the result that ‘water’ is rigid. However, (POD water) by itself is sufficient to yield the desired result that the semantic intension of the predicate ‘is water’ is exactly the metaphysical intension of a particular liquid substance, the very liquid of which the exemplar is an instance, so that if ‘water’ rigidly designates, it rigidly designates something whose metaphysical intension is exactly that of the liquid substance of which the exemplar is an instance.

Modal-Essentialism in the Putnam Theory Given that ‘water’ is rigid, (POD water) also yields the further result that one of water’s features is modal-essential, i.e., a feature it metaphysically could not fail to have: that x is a sample of it if and only if x is a sample of the very liquid that the actual exemplar is actually a sample of (whatever that liquid is discovered to be). This appears to be an example of trivial modal-essentialism.11 It is none too surprising that water has this feature by necessity, since ‘by definition’, i.e., by (OD water), water just is the very liquid of which the exemplar in question is a sample. But Putnam argued that his ­direct-reference theory of natural-kind terms has modal-essentialist consequences that are metaphysically substantive—in particular, that water could not fail to be the particular compound consisting of two parts hydrogen, one part oxygen. Suppose for the moment that ‘H2O’ is an abbreviation for a general definite description, e.g., ‘the chemical compound of two parts hydrogen, one part oxygen’ (or ‘the chemical compound molecules of which consist of two hydrogen atoms together with one oxygen atom’, or something similar). Then, Putnam’s focal sentence (1) Water is H2O.

How Things Have to Be  135 has the specific form of an identity sentence employing both a s­ imple term and a definite description, ┌ α = (ιβ)φβ ┐—except that both α and ┌ (ιβ)ϕβ ┐ are general rather than singular, so that the verb is an ‘is’ of identity to which general terms attach (hence, a kind of higher-­order dyadic predicate).12 Among the logical consequences of (1) is the following: (x)[x is{water} ⊃ x is{H2O}]. As Kripke and Putnam note, the truth expressed is evidently a posteriori, and its negation conceivable. If it is nevertheless a metaphysically necessary truth, and if ‘water’ rigidly designates water, then water has a further modal-essential feature: that all of its samples are also samples of H2O. This claim is evidently non-trivial modal-essentialism. It flies in the face of the metaphysical thesis that the chemical composition of water is a modal-accidental feature. How exactly does the necessity of (1) follow, as Putnam claims, from (POD water) taken together with (1) itself, or even from that together with the assertion that ‘water’ is rigid? The answer is that it does not. This is established by reinterpreting ‘H2O’ by means of a non-rigid general term, for example, ‘earthliquid’. It is a posteriori that water is earthliquid. Given that ‘water’ is rigid, the following analog of (POD water) is also true: □(x)( x is{water} ≡ x is{dthat[earthliquid]}). Yet if ‘water’ is indeed rigid, then it is not necessary that every sample of water is earthliquid. In some possible worlds, earthliquid is water; in others, it is methane. The reinterpretation of ‘H2O’ yields a counter-model to Putnam’s claim. Putnam provides in the passage quoted above a sketchy indication of how the necessity of (1) is supposed to follow from (PODwater). He points out, to begin with, that (PODwater) asserts a particular cross-world equivalence relation, sameL, between water samples in an arbitrary possible world w and the exemplar in the actual world. He then argues that we can understand the sameL relation as a cross-world relation by holding that a bit of liquid in a world w1 is the sameL as a bit of liquid in w2 if x in w1 has the same important physical properties in w1 that y has in w2. On this understanding, it follows from (PODwater) that a bit of liquid in an arbitrary world w is water in w if and only if it has the same important physical properties in w that the exemplar has in the actual world. ‘Once we have discovered the [microstructure] of water’, Putnam declares, ‘nothing counts as a possible world in which water doesn’t have that nature’ (p. 709). (See note 4.)

136  Nathan Salmon The sentence just quoted should arouse suspicion. It is fundamentally a statement of a special case of modal-essentialism: Water’s microstructure (whatever that is discovered to be) is a ­modal-essential feature of water. How did this bit of modal-essentialism find its way into our direct-­ reference theory of ‘water’? Is it really derived as a consequence of the theory? Or is it assumed as an independent and metaphysically laden supplement to the theory? According to an insightful analysis by Keith Donnellan, Putnam generates the necessity of water being H 2O by combining (POD water) with two other premises: the empirical assertion that the exemplar is composed of two parts hydrogen, one part oxygen; and the reductive assertion that the cross-world same L relation consists in having the same chemical analysis (the same ‘important physical property’) in the relata’s respective worlds. Let us construe the hyphenated term ‘H2O-matter’ as abbreviating the mass-noun phrase ‘matter composed basically of two parts hydrogen, one part oxygen’—where being ‘basically’ composed of such-andsuch is composition of such-and-such with relatively small amounts of impurities or other variation. Building upon Donnellan’s analysis, Salmon (1981) reconstructed Putnam’s largely implicit derivation as follows: (POD water)  □(x)( x is{water} ≡ x is{dthat[the liquid substance ℓ: this is{ℓ}]} ). (2) (∃!ℓ)[ℓ is-a{liquid substance} ∧ this is{ℓ}] ∧ this is{H2O-matter} ∧ is{H2O-matter} is-a{chemical analysis}. (3) □(F)(ℓ)( ℓ is-a{liquid substance} ∧ F is-a{chemical analysis} ∧ (∃x) [x is{ℓ} ∧ F(x)] ⊃ □(x)[x is{ℓ} ⊃ F(x)]). Therefore, (4) □(x)[ x is{water} ⊃ x is{H2O-matter}].13 The road from Putnam’s claims about startling consequences of his theory of natural-kind terms, and Donnellan’s analysis of the program, to the foregoing derivation is long and winding. The interested reader will find the details in Salmon (1981). The crucial point is this: Putnam’s proposal that the cross-world sameL relation ‘may be understood as’ the cross-world same-important-physical-property relation amounts, at a minimum, to (3).14

How Things Have to Be  137 With the derivation of ‘Water is necessarily H2O’ laid out in this way, its validity is obvious. Even in extremely weak modal logics (expanded to include ‘dthat’ and second-order validity), (POD water), (2), (3) ├  (4). But the principal weakness of the Putnam-Donnellan program is also thus laid bare: (3) is evidently a supplement to the theory of meaning. Even if (3) is true, and even if it is a priori, it is itself both ­modal-essentialist and evidently non-trivial. It asserts that, necessarily, if at least one sample of a given substance has a particular chemical analysis, then it is a necessary feature of that substance that all of its samples have that same basic chemical analysis. Small wonder it follows from this, taken together with (POD water) and the minor supporting premise (2), that it is a modal-essential feature of water, even though it is a posteriori, that its samples are basically two parts hydrogen, one part oxygen. We were investigating the consequences of the ostensive definition when the modal-essentialist premise (3) was surreptitiously brought into play. The case has not been made that (3) comes from direct-reference theory. Quite the contrary, it appears to be an independent, and metaphysically loaded, supplement to the theory of natural-kind terms—one that a committed anti-modal-essentialist will deny even if he/she buys Putnam’s direct-reference theory of ‘water’. In short, the case has not been made that (4) is a genuine consequence of the direct-reference theory of natural-kind terms.15

Reconfiguring the Putnam Program Not all binary relations can be construed as cross-world relations. ­Consider, for example, the relation, x kisses y. (Cf. Salmon 1981, pp. ­116–135.) Appreciating this, Putnam provides a purported reduction of the equivalence sameL relation: x’s being the sameL as y is reducible to x and y being liquid samples having the same important physical property, i.e., the same chemical analysis. This reduction allegedly enables us to understand sameL as a cross-world relation: liquid samples in different possible worlds are the sameL when they share the same chemical analysis across their respective worlds. The reconstructed derivation invokes this purported reduction, recasting it as premise (3). The very fact that Putnam’s proposed reduction is thus modal-essentialist, and yet evidently independent of the theory of reference, is precisely what casts serious doubt on his claim that his theory of reference has the startling consequences he claims it has.

138  Nathan Salmon But perhaps Putnam’s rationale might be better captured by avoiding talk of ‘sameL’ and ‘same liquid’ altogether and instead building the reduction directly into the ostensive definition of ‘water’, thus: (OD water′) □( dthat[the basic chemical analysis F: F(this)] = water ) (POD water′) □(x)( x is{water} ≡ dthat[the basic chemical analysis F: F(this)](x)). Mackie’s principal objection (pp. 185–186) to the letter of Salmon (1981) is that this alternative rendering of the ostensive definition has one evidently telling advantage over (OD water). We modify (2) as follows: (2′) (∃!F)[F is-a{chemical analysis} ∧ F(this)] ∧ this is{H2O-matter} ∧ is{H2O-matter} is-a{chemical analysis}. We now have the following: (POD water′), (2′) ├  (4). The former modal-essentialist premise (3) plays no role in the derivation of (4) from (POD water′) and (2′). Replacement of (POD water) and (2) by (POD water′) and (2′) renders (3) an idle wheel. To this extent, Mackie argues, Putnam’s claim is vindicated that the necessity of (1) is a consequence of his direct-reference theory taken together with (1) itself.16 In this, Mackie is evidently not alone. In an article obviously heavily influenced by Kripke’s and Putnam’s thought on the topic, as well as Donnellan’s analysis of Putnam (as presented in Salmon 1981), Scott Soames writes: Let us first examine the way in which the necessity of these sentences [like (1)] is related to their truth. As I see it, the crucial issue involves the nondescriptionality of simple manifest kind terms, and the way in which their reference is determined. … For example, we may imagine ‘water’ introduced by the following stipulation … The general term ‘water’ is to designate the kind instances of which share with all, or nearly all, members of the class of paradigmatic ‘water’-samples those properties that “make them what they are” (and that distinguish them from certain paradigmatic ‘non-water’ samples). These are properties that explain their most salient characteristics—e.g. the fact that they boil and freeze at certain temperatures, that they are clear, potable, necessary to life, and so on. Hence, the predicate ‘is water’ will apply (with respect to any world-state) to all and only those quantities of matter that have the properties that actually explain the salient features of all, or nearly all, of the paradigmatic ‘water’-samples (and that are lacking in all, or nearly all, the paradigmatic ‘non-water’- samples).

How Things Have to Be  139 … (i) We begin with the ostensive introduction of ‘is water’ by a stipulation that it is to apply (with respect to any possible state of the world) to all and only instances of the kind determined by certain properties—namely those possession of which by all, or nearly all, members of the ‘water’-sample in the actual state of the world distinguishes them from members of the ‘non-water’-sample, and (causally) explains the salient characteristics of the ‘water’-sample. (ii) It is then discovered scientifically that possession of the property expressed by ‘is H2O-matter’ distinguishes the members of the ‘water’-sample from the members of the ‘non-water’-sample, and (causally) explains the salient characteristics of the ‘water’-sample. (iii) From this it follows that the kind designated by the simple manifest kind predicate ‘is water’ is the kind determined by the property expressed by ‘is H2O-matter’. (iv) This is sufficient to establish the necessity of sentences like [(1)]. … As I see it, the necessity of … statements [like (1)] follows from their truth, plus the way in which the reference of the terms they contain is standardly fixed.17 The proposal appears to be that ‘water’ may be regarded as defined by, in effect, the following replacements for (OD water) and (POD water): (OD water″)  □( dthat[the explanatory profile F: F(this)] = water) (POD water″)  □(x)( x is{water} ≡ dthat[the explanatory profile F: F(this)](x)). Here, I use the phrase ‘explanatory profile’ as a term for the conjunction of those properties “possession of which by the exemplar in the actual world distinguishes it from the counter-exemplars, and (causally) explain the exemplar’s most salient characteristics” (wording extrapolated from Soames).18 We now modify (2′) as follows: (2″) (∃!F)[F is-a{explanatory profile } ∧ F(this)] ∧ this is{H2O-­ matter} ∧ is{H2O-matter} is-a{explanatory profile }. We thereby obtain a result exactly analogous to Mackie’s: (POD water″), (2″) ├  (4). Here again, the modal-essentialist principle (3) plays no role. These proposals raise the obvious question: Is the envisioned replacement of (OD water), whether by (POD water′) or by (POD water″), legitimate? This question must be answered in the negative. Given how the English word ‘water’ is actually understood, neither (POD water′) nor (POD water″) can serve as an appropriate ostensive definition, even if both fix the right semantic intension. Furthermore, if either (OD water′) or (OD water″) is

140  Nathan Salmon taken as a redefinition, providing the word ‘water’ with a new and different meaning specification, the modal-essentialism therewith contained in (4) then becomes trivial. We have already seen that (POD water) fails to deliver the consequence that ‘water’ is rigid. Both (POD water′) and (POD water″) fail in a further respect that is more fundamental to the theory. They fail to deliver a crucial consequence that (POD water) delivers: that if ‘water’ rigidly designates a kind, it rigidly designates a kind whose metaphysical intension is that of a particular liquid substance. Both (POD water′) and (POD water″) also fail to deliver the desired consequence that the semantic intension of the predicate ‘is water’ is the metaphysical intension of a liquid substance. The problem is not that (POD water′) and (POD water″) are false. Even if they are true, even if they are necessary, the fact that the semantic intension of ‘is water’ is the metaphysical intension of a substance does not follow from (POD water′) together with (2′), nor from (POD water″) together with (2″). Neither (ODwater′) nor (ODwater″) is a plausible rendering of the intended force of the ostensive definition, ‘This liquid is water’. Neither can be regarded as analytic. The question of whether (PODwater′) and (PODwater″) are even true is the very issue that separates the modal-essentialist about compounds from the anti-modal-essentialist. If a compound’s chemical analysis and its explanatory profile are modal-essential features, then (PODwater′) and (­PODwater″) get the right semantic intension for ‘is water’; otherwise, they do not. The Mackie-Soames proposal purports to settle a substantive metaphysical controversy while smuggling modal-essentialism into an alleged mere specification of the meaning of ‘water’. In a sense, (OD water′) and (OD water″) miss the very point of the ostensive definition. This is, in Putnam’s words, to explain the meaning of the English word ‘water’. A proper definition specifies meaning in a philosophically neutral manner, without begging any substantive questions. Insofar as this is its objective, (OD water′) and (OD water″) do not merely specify the English meaning of ‘water’ incorrectly. They evidently specify the wrong meaning. In effect, they miscast ‘water’ as a term for a particular basic chemical analysis or for a particular explanatory profile, or for an ‘important physical property’, rather than as a term for a particular liquid substance. Indeed, Soames (2004) characterizes the term ‘water’ defined in his proposed alternative to (OD water) as a term for a particular explanatory kind: As I have imagined their ostensive introduction, both [‘is water’ and ‘is green’] stand for what might be called explanatory kinds. In the case of ‘is water’, the kind is one that is determined by the properties possessed by paradigmatic ‘water’-samples that both distinguish them from paradigmatic ‘non-water’ samples and (causally) explain, in the actual state of the world, such salient characteristics of the

How Things Have to Be  141 members of the ‘water’-sample as their boiling point, freezing point, their properties as solvents, and so on. The intuitive, pre-theoretic concept of water is not that of an explanatory kind, nor is it that of a chemical-analysis kind. It is that of a liquid substance, that of a particular natural kind of bits of fluid matter. While (OD water′) cannot be plausibly regarded as a more explicit rendering of the ostensive definition ‘This liquid is water’, it might serve instead as a rendering of a distinctly different sort of sentence: ‘The chemical analysis of the liquid substance of which this is a sample is necessarily that of water’. The latter is not suited to define or introduce the word ordinary-language word ‘water’. It is better suited to define the particular phrase ‘the chemical analysis of water’. It is not especially well suited for that purpose either, but it might do. (The phrase ‘the chemical analysis of water’ presupposes an understanding of ‘water’. Its full meaning should then follow from a separate definition of ‘chemical analysis’. It is, rather, a formal consequence of (3) taken together with the authentic ostensive definition: ‘This liquid is water’. Analogous remarks apply in connection with ‘the explanatory profile of water’.) A more explicit variant of (OD water′) illustrates both the former’s inappropriateness as a definition of ordinary-language ‘water’ as well as the true nature of the derived modal-essentialism. If one were so inclined, one could introduce a name for the chemical analysis (whatever it might turn out to be) of water. One might coin the word ‘waterchemistry’ in a Kripkean reference-fixing stipulation on the model of ‘meter’, ‘Neptune’, ‘Newman 1’, or ‘π’: Let ‘waterchemistry’ be a mass-noun general name for (and thus a rigid designator of) the basic chemical analysis of water, whatever that analysis might be, if water has a unique basic chemistry; otherwise let it designate nothing. It should be emphasized that ‘waterchemistry’ is not defined to be a substance term as such; instead it is introduced as a general term for all those bits of matter that have a particular basic chemical analysis. The sentence ‘Waterchemistry is a substance’ is certainly not analytic, and similarly for ‘Every instance of waterchemistry is a sample of the same substance’. The reference-fixing introduction might be accomplished instead through something that resembles a more traditional ostensive definition: ‘The chemical analysis of this is waterchemistry’, said while pointing to a freshwater lake. The underlying modal force of the intended definition might be given by the following: (OD waterchemistry)  □( dthat[the basic chemical analysis F: F(this)] = waterchemistry).

142  Nathan Salmon This has exactly the form of (OD water′). Taking its Putnamian immediate consequence, (POD waterchemistry)  □(x)( x is{waterchemistry} ≡ dthat[the basic chemical analysis F: F(this)](x) ), together with (2) delivers the following result: (4′) □(x)[ x is{waterchemistry} ⊃ x is{H2O-matter}]. Given further that ‘waterchemistry’ is also rigid, (4′) yields the additional result that the exemplar’s basic chemical analysis has a particular modal-essential feature: something has that basic chemical analysis only if its chemical analysis is basically two parts hydrogen, one part oxygen. This is trivial modal-essentialism par excellence. It is analogous to, and every bit as trivial as, the consequence of (POD water) together with the rigidity of ‘water’, that the liquid of which the exemplar is a sample has as a modal-essential property that something is a sample of it if and only if it is a sample of the very liquid that the actual exemplar is actually a sample of (whatever that liquid is discovered to be). The philosopher who believes that the chemical analysis of a compound is a modal-accidental feature will insist that ‘Any sample of water is waterchemistry’ is true but contingent, that in some possible worlds there are samples of water that are not H 2O-matter, and hence not waterchemistry. One may decide to use the word ‘water’ henceforth not for water per se, but for the basic chemical analysis of its samples (assuming that those samples share a common basic chemistry). To do so would be misleading, but there is no law against it in free countries. Such re-labeling is precisely what (OD water′) and (OD water″) do. Let the buyer beware. In replacing the artificial term ‘waterchemistry’ with the pre-existing word ‘water’, (POD water′) conceals the trivial nature of (4′) while simultaneously misrepresenting the modal-essentiality of water’s chemical analysis as a trivial consequence of its English meaning. Explaining the meaning of (4) to be what (4′) expresses is not a way of making the modal-essentiality of water’s chemical analysis derivable from the meaning of ‘water’, any more than calling a medium-size pizza ‘extra large’ makes it so. The triviality of the derived modal-essentialism is perhaps better illustrated by a less arcane example. Suppose the word ‘X-color’ is introduced by means of the following modal ostensive definition and its Putnam-style immediate consequence: (OD X-color)  □( dthat[the color k: this is{k}]} = X-color ) (POD X-color)  □(x)( x is{X-color} ≡ x is{dthat[the color k: this is{k}]}).

How Things Have to Be  143 The word ‘X-color’ is a general term for a certain color. (OD X-color) has the same form as (OD water). It also has the same form as (OD water′), which, qua meaning-explanation, makes ‘water’ into a general term for a certain chemical analysis rather than a substance term. Given that ‘X-color’ is rigid, it follows from (POD X-color) that one particular feature is a modal-essential property of X-color: something has it if and only if it is the same color that the exemplar actually is. Suppose now it is discovered that the exemplar is blue. Assuming that ‘blue’ rigidly designates a color, it follows that necessarily, if a thing is X-color, it is blue. It also follows that X-color has a modal-essential property: anything that has it is blue. The derivation exactly parallels the derivation of (4) from (POD water′) and (2). The derived result is exactly analogous to the fact that necessarily, any instance of waterchemistry is H2O-matter. This kind of modal-essentialism is completely trivial. By logic alone, it is a modal-essential feature of any universal that any instance of it is an instance of it. It should not be startling that trivial results like this can be extracted from the direct-reference theory. They can also be extracted from a description of chicken soup, and in the same way. Whether one chooses to call waterchemistry ‘waterchemistry’, ‘water’, or something else, it is assuredly a modal-essential feature of it that all its samples are H2O-matter. The question remains whether (4) is true as ‘water’ functions in ordinary English. The anti-modal-essentialist answers in the negative. Direct-reference theory evidently has no horse in that race. That (OD water′), taken as an ostensive definition of ‘water’, renders the derived modal-essentialism trivial is perhaps most clearly illustrated through one final example. Suppose a new general name ‘2H-1O’ is introduced not ostensively but in a designatum-fixing stipulation: Let ‘2H-1O’ be a mass-noun general name for (and thus a rigid designator of) the particular kind, Matter composed basically of two parts hydrogen, one part oxygen.19 Suppose the intent is given by the following familiar (C′) and (C) sentences, respectively: (D 2H-1O)  □( dthat[H2O-matter] = 2H-1O ) (PD2H-1O)  □(x)(x is{2H-1O} ≡ x is-a{dthat[H2O-matter]}). Here, the ‘dthat’ operator is redundant. The phrase ‘matter composed basically of two parts hydrogen, one part oxygen’ (abbreviated as ‘H2O-matter’) may be regarded as a rigid designator of a particular kind. The semantic intension of the predicate ‘is 2H-1O’ is the metaphysical intension of this kind. More to the point, the term ‘2H-1O’ is a Millian general name; its semantic content is the very kind designated.20

144  Nathan Salmon 2H-1O is, by definition, a particular chemical-analysis kind. The ­ stensive definition (OD water′) fixes the semantic content of the familiar o term ‘water’ in such a way that it purportedly names the very same kind, H2O-matter, and is thereby a Millian synonym of ‘2H-1O’. But whereas the sentence ‘All 2H-1O is H2O-matter’ is derivable from the designatum-fixing definition of ‘2H-1O’, and is in that sense analytic, by contrast a mere ‘explanation’ of the English meaning of ‘water’, by itself or even together with the observation that a particular exemplar is ­H2O-matter, should not allow inference to the apparently synthetic sentence, ‘All water is H2O-matter’, let alone to (4). It is a small feat to derive from (PD 2H-1O) the consequence that necessarily, all 2H-1O is H2O-matter. Given that ‘2H-1O’ is rigid, it follows that it is modal-essential to 2H-1O—i.e., to the kind, H2O-matter—that all of its instances are H2O-matter. The triviality of this sort of modal-essentialism is as obvious as anything in this area can be.21 (POD waterchemistry) is simply the ostensive counterpart of the designatum-fixing (PD 2H-1O). The former requires one extra step in the derivation of the same necessary proposition, but insofar as Mackie transforms ‘water’ into ‘waterchemistry’—an ostensively defined Millian name synonymous with ‘2H-1O’—the modal-essentialism embodied in that proposition is exactly the same. Trivially, it is modal-essential to any substance that all of its samples are samples of it, and it is modal-essential to any chemical analysis that anything that has that analysis has it. It has not yet been established, as advertised, that as a consequence of the direct-reference theory of natural-kind terms, a substance’s chemical analysis is a modal-essential feature of that substance. Specifically, it has not been established that as a consequence of direct-reference theory, it is modal-essential to water that each of its samples are basically two parts hydrogen, one part oxygen.

The Failure of the Reconfigured Putnam Program The sameL relation is neither primitive nor monolithic. Like most equivalence relations that come readily to mind (e.g., x is wearing the same color shirt as y), there are intervening entities (colors) and an intervening relation R in virtue of which entities bear the equivalence relation to one another: x bears R to the same intervening entity as y. Equivalence relations of this sort are legitimately construed as cross-world relations if the intervening entities can be cross-world identified. Putnam’s reduction of the sameL relation is directly contrary to the spirit, if not also to the very letter, of Kripke’s haecceitism. The latter is Kripke’s doctrine that it is perfectly legitimate for one to stipulate, even by name if one chooses, which entities are present and have various properties in the class of possible worlds that one is singling out for consideration (e.g., the class of worlds in which Hubert Humphrey wins

How Things Have to Be  145 22

the 1968 U.S. presidential election). Among the things that one can legitimately stipulate are present in the possible worlds under consideration are material substances. According to haecceitism with respect to substances, one may legitimately fix which possible worlds are presently under consideration by stipulating that the very same-liquid substance is present in all of them. This by itself completely justifies our understanding of sameL as a cross-world relation. One might then assert that in a pair of worlds under consideration, w1 and w 2 , a particular bit of matter x is a sample of that substance in w1, while another bit of matter y is a sample of it in w 2. One is free to stipulate, in particular, that one is considering exactly those worlds in which some liquid samples are of the very same-liquid substance ℓ of which the exemplar is actually a sample. No further ‘reduction’ of the sameL relation is warranted by the direct-reference theory, and no further reduction is necessary to render the cross-world sameL relation legitimate. (For details, see Salmon 1981, section 13, pp. 116–135.) In short, according to Kripke’s haecceitism with regard to substances, the sameL relation is legitimately regarded as cross-world quite independently of any proposed reduction of that relation. Putnam’s proposed reduction of the cross-world sameL relation to the same-important-­ physical-property relation is non-trivially modal-essentialist, and it is no part of direct-reference theory proper. Without that modal-essentialist reduction, the necessity of water being H2O is not guaranteed by the modal ostensive definition of ‘water’ taken together with the rigidity of ‘water’ and the empirical fact that the exemplar is H2O. To repeat: the case has not been made that (3) is anything other than exactly what it appears to be: a non-trivial metaphysical sentence that is quite independent of direct-reference theory and that plays a pivotal role in the derivation of (4) from the modal ostensive definition of ‘water’ provided in Putnam’s theory. The case has not been made that direct-­ reference theory comes with substantive metaphysics as a free bonus.23

Notes 1 The present paper was written primarily in 2008. Portions were presented at Philosophy in an Age of Science: A Conference in Honor of Hilary Putnam’s 85th Birthday, Harvard and Brandeis Universities, May 31–June 3, 2011. I am especially grateful to several participants for discussion and comment. I am also grateful to Teresa Robertson Ishii for extremely fruitful discussion, for exposing significant errors in an earlier draft, and for much else. 2 Putnam (1973). 3 Modal-essentialism is the doctrine that some properties of some things are properties that those things metaphysically could not have lacked. This contrasts with quiddity-essentialism, the doctrine that each thing has a quiddity essence (a whatness or what-it-is). The unadorned term ‘essentialism’ has become ambiguous in contemporary philosophical English. Regrettably, this has led to a great deal of needless confusion. As with nearly all of the

146  Nathan Salmon modal-metaphysical literature in the Quinean and Kripkean eras (roughly the latter half of the 20th century), the present essay is concerned with modal-essentialism, and not at all with quiddity-essentialism. 4 Cf. Salmon (2003, 2005). 5 The present formulation involves an improvement over that provided in Salmon (1981), p. 145. See note 13. 6 Putnam: “The rigidity of the term ‘water’ follows from the fact that when I give the ‘ostensive definition’: ‘this (liquid) is water,’ I intend [(POD water)] and not [‘□(x)[x exists ⊃ (x is water ≡ x bears sameL to this)]’]” (Putnam 1973), p. 707. 7 Salmon (1981) mistakenly took it to be a singular term. This led to unnecessarily complicated formulations. 8 To distinguish the relevant description from a singular definite description, it might be formalized thus, where ‘k’ is a general-term variable, as opposed to a singular-term variable, and ranges over kinds (or other universals) whose instances are particular individuals: (ιk)[k is-a{liquid-substance} ˄ samples of k cover most of the Earth’s surface]. 9 Salmon (1981), pp. 40–41. 10 Salmon (1981) raised the question, but remained neutral, whether kinds having the same metaphysical intension are ipso facto the same kind (p. 53n9). Teresa Robertson Ishii has made observations that strongly incline me to believe that there are numerically distinct kinds that exactly coincide in metaphysical intension. See note 20. Salmon (1981) adopted an artificial use of ‘designate’, which I do not now favor, according to which a general term, η, designates each of the kinds, categories, etc., whose metaphysical intensions coincide with the semantic intension of the corresponding predicate ┌is ν ┐. The following formulation is stronger than (C) yet weaker than (C′), where ‘k’ is a general-term variable that ranges over universals whose instances are particular individuals: (C″) (∃k)□(ν = k) ˄ □(x)(x is{ν} ≡ x is{dthat[β]}). The left-hand conjunct requires that ν be rigid. The right-hand conjunct is just (C), which fixes the semantic intension of ┌is ν ┐. Unlike (C′), however, (C″) does not fix the designatum of ν. 11 Cf. Salmon (1981), pp. 82–87. 12 This deviates from the view of Scott Soames in one significant respect, and from my own view in another. Soames (2002), at p. 291, parses (1) as a universal-conditional, ‘(x)[ x is water ⊃ H2O(x) ]’, whereas I regard it instead as an identity employing general terms, ‘Water = H2O’. Soames regards ‘H2O’ as a predicate synonymous (at least roughly) with something like ‘is a liquid sample consisting of two parts hydrogen, one part oxygen’. I regard it ‘H2O’ (on an analogy with ‘Neptune’, ‘Newman 1’, and ‘π’) as a general name whose reference is fixed by a general description, perhaps ‘the chemical compound whose molecules are made up of two hydrogen atoms and one oxygen atom’. I note in this connection that it is at least distinctly odd, and very likely incorrect, to make substitutions for the term’s components—for example, the substitution of ‘the most plentiful element’ for the ‘H’ in ‘H2O’, or ‘3 – 1’ for the ‘2’. Cf. Salmon (1987). 13 See note 3. The present formulation likewise involves an improvement over that provided in Salmon (1981), pp. 178–184.

How Things Have to Be  147 14 Putnam’s claim that the cross-world sameL relation ‘may be understood as’ the cross-world same-important-physical-property relation must have at a minimum the force of (3′) (w1)(w2)(x)(y)( x existsw1 ˄ y existsw2 ⊃ [(x isw1{(℩ℓ)}( ℓ is-a{liquid substance}w2 ˄ y is{ℓ}w2)} ⊃ (F)( F is-a{chemical analysis} ⊃ [Fw1(x) ≡ Fw2(y)])]). Given trivial assumptions, (3) is equivalent. Compare this again to the last sentence of the passage quoted earlier. Cf. also Salmon (1981), at pp. 179–180, 183–189. 15 One of the criticisms in Mackie (2006) is that if (3) is a priori, then (4) is after all an ‘a priori consequence’ of (PODwater) and (2). This is mistaken. If a conditional ┌ ϕ ⊃ ψ ┐ expresses something that is knowable a priori, this in itself does not make ψ a consequence of ϕ. (The conditional can be a priori without being a logical truth.) Though the conditional is a priori, it might also be highly controversial, with some philosophers even insisting that its negation is a priori. If (3) is a priori, then the conditional ┌If (PODwater) and (2), then (4)┐ is also a priori. The relevant issue, however, is whether direct-reference theory can consistently accept this conditional’s antecedent while rejecting the consequent—even if the conditional itself is correct and even if it is a priori. 16 Mackie justifies the replacement of (POD water) by (POD water′) not on the ground that doing so better captures Putnam’s intentions and explanations, but on the ground that the replacement is merely consistent with Putnam’s theory of ‘water’. This is insufficient to vindicate Putnam’s claim. Otherwise, the replacement of (POD water) by the conjunction ┌ (POD water) ˄ (3) ┐ would count as vindication, so that the original reconstructed argument explicitly invoking (3) would already vindicate Putnam’s claim. It does not. 17 Soames (2004), at 160–165, 176. I have replaced Soames’s convention of using italics for quotation with the convention of single quotation marks. Soames (2004), pp. 165–166, offers two arguments for the principle that ‘manifest’ natural kinds that are metaphysically co-intensional are identical, but neither is conclusive. In any event, even if co-intensional manifest natural kinds are identical, it does not follow that, more generally, c­ o-intensional kinds (whether natural, gerrymandered, or otherwise) are the same. See notes 10 and 21. 18 I have taken considerable liberties in extrapolating these formulations from Soames’s actual wording. Nothing in my discussion depends on this. Soames (2004) asserts (pp. 174–175) that whenever a ‘dthat’-term, ┌ dthat[β]┐, is used in a designatum-fixing ostensive definition, where the semantic content of β is a descriptional concept the such-and-such, competence with the expression thereby introduced requires knowing that it designates the suchand-such (provided that there is exactly one such-and-such). For this reason, Soames might balk at my representing his proposal by means of (OD water″), depending on whether he disbelieves that semantic competence with ‘water’ requires knowing that it designates an explanatory profile. My criticism of Soames’s proposal does not turn on this issue. 19 My view of the chemical terms for water, ‘H2O’ and ‘dihydrogen monoxide’, is that they function not as abbreviations but in the manner of ‘2H-1O’— except perhaps that ‘2H-1O’ names the kind H2O-matter, whereas ‘dihydrogen monoxide’ names the substance that is the principal component of acid rain (viz., water). See note 20. 20 This same kind is arguably also the semantic content of the mass-noun phrase ‘matter composed basically of two parts hydrogen, one part oxygen’. It seems to me that in general, a gerrymandered general term of the form ┌ ν that is μ┐

148  Nathan Salmon (e.g., ‘girl who is brown-eyed’, or simply ‘brown-eyed girl’) may be regarded as a directly referential general name whose designatum is fixed by a general definite description, perhaps along the lines of ┌the kind corresponding to the propositional function x^ (x is a ν ˄ x is μ) ┐. (Caution is due in connection with words like ‘alleged’, ‘decoy’, ‘pretend’, ‘wannabe’, etc., as in ‘alleged embezzler’.) Unlike the typical cases of fixing-the-designatum-by-a-description (‘Newman-1’, ‘meter’, etc.), in the case of a typical compound general term, the property expressed in this designatum-fixing description is known a priori to hold of the designatum. A sharp distinction must be maintained between the general definite ­description ‘the liquid that covers most of the Earth’s surface’ (abbreviated as ‘earthliquid’) and the included common-noun phrase ‘liquid that covers most of the Earth’s surface’. Soames (2004), p. 167, takes it that the latter, like the former, non-rigidly designates water. Like any count-noun phrase, the latter rigidly designates a kind—in this case, a special kind of which water is the only instance in the actual world and methane is the only instance in Wm. I have introduced a variable-binding operator—the theta-abstraction operator—which works as follows: Where ϕα is an open formula (open sentence), both the designatum and the semantic content of the expression ┌(θα) [ϕα]┐ are the kind appropriately determined by the Russellian propositional function semantically associated with bound occurrences of ϕα (this propositional function being what I call the bondage content with respect to a of ϕα). Cf. (Salmon 2006). Then, a theta-abstract ┌(θα)[ϕα]┐ might be read: ┌thing α such that: ϕα┐. For example, the designatum/content of ‘thing x such that: x is brown-eyed and x is a girl’ is the kind, Brown-eyed girl, appropriately determined by the propositional function x^ (x is brown-eyed  ˄ x is a girl). Cf. (Salmon 1981), p. 51. For reasons related to this, Alonzo Church’s lambda-abstraction operator, as it occurs in ┌(λα)[ϕα]┐ where ϕα is an open formula, is susceptible of a contextual definition: ┌is-a{(θα)[ϕα]}┐. 21 The kind, H2O-matter, and the compound, water, coincide in metaphysical intension. Are they ipso facto the same kind? (See note 10.) Robertson Ishii observes that if so, it evidently follows that it is trivially modal-essential to water that all of its instances are H2O-matter—since this is trivially modal-essential to the kind, H2O-matter. While this may well be modal-essential to water, the prospect that this is trivial seems most implausible. Indeed, if water = H2O-matter, then, astonishingly, modal-essentialism concerning chemical compounds emerges as a truth of modal logic. For the proposition expressed by the sentence ‘Water = 2H-1O’ would then be nothing more than the classical-logical truth that water is water. In short, this formulation would be analytic, and (given that both terms are rigid) therefore necessary. So also would ‘Water = H2O-matter’. If it were correct that it is trivially modal-essential to water that all and only its samples are H2O-matter, would this vindicate the claim that ­direct-reference theory taken together with uncontroversial empirical facts yields modal-essentialism concerning chemical compounds? Even if so, the derived modal-essentialist sentence is not (4) but (PD2H-1O). One still needs ‘Water = 2H-1O’ to obtain (4) from (PD2H-1O). Furthermore, the obtained modal-essentialism would in that case be trivial. More plausibly, water and H2O-matter are distinct kinds—despite their coincidence in metaphysical ­intension, and despite the identity of water and H2O. (But compare the case of water and dihydrogen monoxide with the case of the property of being blue

How Things Have to Be  149 in color and that of reflecting light having a spectrum dominated by energy within a particular band of wavelengths (roughly 440–490 nm). Are these not the very same property?) 22 See Salmon (1996). 23 At the 2011 Conference in Honor of Hilary Putnam’s 85th Birthday, Putnam agreed that non-trivial modal-essentialism is not derivable from his theory of reference for natural-kind terms.

References Davidson, M. (ed) (2007) On Sense and Direct Reference. Boston: McGraw Hill. Kripke, S. (1980) Naming and Necessity. Oxford: Blackwell. Linsky, B. (1984) ‘General Terms as Designators’, Pacific Philosophical Q ­ uarterly, 65, pp. 259–276. Mackie, P. (2006) How Things Might Have Been: Individuals, Kinds, and Essential Properties. Oxford: Oxford University Press. Putnam, H. (1973) ‘Meaning and Reference’, The Journal of Philosophy, 70 (November 8), pp. 699–711. Salmon, N. (1981) Reference and Essence (Second Edition). Amherst, NY: Prometheus, (1981, 2005). Salmon, N. (1987) ‘How to Measure the Standard Metre’ Proceedings of the Aristotelian Society, 88, pp.  193–217; reprinted in Davidson (2007), pp. 962–980, and in N. Salmon, Content, Cognition, and Communication (Oxford: Oxford University Press, 2007), pp. 141–158, at p. 143n5. Salmon, N. (1996) ‘Trans-World Identification and Stipulation’ Philosophical Studies, 84, pp. 203–223; reprinted in Salmon (1981) (as “Cross-World Identification and Stipulation”), appendix V, pp. 345–369. Salmon, N. (2003) ‘Naming, Necessity, and Beyond’ Mind, 112 (447, July), pp. 475–492; reprinted in Salmon (1981), pp. 377–397. Salmon, N. (2005) ‘Are General Terms Rigid?’ Linguistics and Philosophy, 28 (1), pp. 117–134; reprinted in N. Salmon, Content, Cognition, and Communication (Oxford: Oxford University Press, 2007), pp. 100–112. Salmon, N. (2006) ‘A Theory of Bondage’ The Philosophical Review, 115 (4, ­October), pp. 415–448. Soames, S. (2002) Beyond Rigidity. Oxford: Oxford University Press. Soames, S. (2004) ‘Knowledge of Manifest Natural Kinds’ Facta Philosophica, 6, pp. 159–181.

7 In Search of a Structurally Complete Epistemology of Essence Michael Wallner

Introduction There are two different kinds of answers to the question of how we acquire knowledge of modality (i.e. knowledge of necessity and possibility): (MC) One can either appeal to specific mental capacities (like conceiving, imagining, intuiting, perceiving …) that provide us with warrant for beliefs about modality; or (SM) one can ground knowledge of modality in knowledge of some other subject matter, SM, arguing that it is in virtue of knowledge of SM that we come to have modal knowledge. In this paper, I investigate a particular, very influential (SM)-type answer to the central question in the epistemology of modality. According to essencebased accounts, we come to have modal knowledge in virtue of having knowledge of essences. Such an account has to answer two crucial questions: (i) Transition Question: How exactly does the transition from knowledge of essence to knowledge of modality work? (ii) Source Question: How do we acquire knowledge of essence in the first place? This paper focuses on (ii), the source question. With regard to (ii), one can, again, give a respective (MC)-type answer, appealing to some mental capacities, or a respective (SM)-type answer, grounding knowledge of essence in knowledge of something else. Looking at the two most influential essence-based accounts on the market, E. J. Lowe’s and Bob Hale’s, I argue that both deliver pictures of the epistemology of modality that are in a crucial, structural sense incomplete. In a section called The Road Map, I systematize the critical discussion of Lowe and Hale by breaking down general, structural desiderata any complete account of the epistemology of essence should fulfill. Concerning the epistemology of essence, we can, again, either take the (SM)-route, grounding knowledge of essence in knowledge of some subject matter, SM1, or the (MC)-route, identifying one or more mental capacities that give(s) us DOI: 10.4324/9781003002192-9

In Search of a Structurally Complete Epistemology of Essence  151 knowledge of essence. Since any (SM)-type answer in the epistemology of essence will raise a question about how we know the subject matter, SM1, in the knowledge of which we have grounded knowledge of essence, (SM)-type answers all the way down will be problematic on pain of a vicious infinite regress. That is to say that since (SM)-type answers prompt (ii)-type follow-up questions (source questions), we cannot always meet (ii)-type follow-up questions with (SM)-type answers. (SM)-type answers all the way down would just infinitely push the bump under the rug. Hence, at some point in the justificatory chain, or so I shall argue, there needs to be appeal to some (MC)-type answer to a (ii)-type follow-up question. What is more, I argue that not any (MC)-type answer will do the trick. The foundationalist’s (MC)-type answer will have to appeal to directly justifying mental capacities. From this, it should be clear that the notion of structural completeness, which is to be developed in detail in The Road Map, does not concern completeness in scope but rather completeness concerning the justificatory story an account tells. In other words, an account in the epistemology of essences is complete in the structural sense, not if it is able to account for the entire scope of essentialist knowledge but if it paints a map all the way down to the end of the justificatory or explanatory road for those pieces of essentialist knowledge it tackles. With these desiderata and the notion of structural completeness in hand, I discuss Edmund Husserl’s (1973) epistemology of essence, which gives an (MC)-type answer to the source question (ii) by appealing to imagination. Exploring a crucial problem this account faces, it becomes obvious that imagination (at least as it is applied here) is not a directly justifying mental capacity. Hence, the Husserlian appeal to imagination also falls short of providing a structurally complete account of the epistemology of essence. I argue, however, that supplying this imagination-based account with a broadly Husserlian epistemology of universals can yield a structurally complete picture of the epistemology of essence. This is due to the fact that Husserlian epistemology of universals assigns a central role to intuition—a mental capacity that fits the bill of being directly justifying. Finally, I will provide some concluding remarks.

Lowe’s Epistemology of Essence It seems that Lowe has a pretty straightforward answer to the question of how our epistemology of essence works. Following Fine (1994), he takes the essence of a thing x to simply be what it is for x to be. In this sense, the essence of x is given by the real definition of x. So, to Lowe, it is plausible, and sufficient, to say that we know the essence of x simply by understanding what x is. Lowe (2012: 944) gives a transcendental argument aiming to establish that we have essentialist knowledge: Knowing an entity’s essence is simply knowing what that entity is. And at least in the case of some entities, we must be able to know

152  Michael Wallner what they are, because otherwise it would be hard to see how we could know anything at all about them.1 Lowe (2012: 944) Note that, if successful, this transcendental argument establishes that we have knowledge of essence (in some cases). It does not, however, tell us how we come to know essentialist propositions. Yet, this was the question. Really, all Lowe can offer in answer to the latter question is that we know the essence of x by understanding what x is. This answer, however, immediately raises the question of how we understand what x is. It is not the case that Lowe’s epistemological story ends here. However, as Vaidya (2018: 232) correctly observes, Lowe’s theory of our knowledge of essence “for the most part is given by his theory of what essences are, what real definitions are, and what prospects there are for real definitions across different types of entities”. That is to say that Lowe offers some considerations as to how much of an entity’s essence one needs to know or understand for the resulting modal judgments about that entity to be reliable. So, Lowe offers some quantitative considerations about our understanding of what x is in relation to modal judgments about x. To illustrate, take the following of Lowe’s examples. Lowe (2012: 939– 940) argues that even if there is no complete verbal definition to be had of entities like a bronze statue and a lump of bronze, we have enough grasp on the essences or real definitions of those entities to conclude that they can coincide spatiotemporally and yet be numerically distinct.2 I take it, however, that the question of how we understand what x is remains largely unanswered without any qualitative considerations about how any quantity of understanding of what some x is could come about. Lowe’s account remains short on such qualitative considerations. With regard to the framework of possible kinds of answers to the source question, (ii), which I have offered above, Lowe can be taken to offer an (MC)-type answer to (ii). He appeals to the mental capacity of rationally understanding (what x is). Yet, Lowe does not satisfactorily answer the follow-up question of how understanding what some x is is supposed to work exactly.3 It is for that reason that Lowe’s answer to (ii) seems unsatisfactory and incomplete; it does not take us very far down the explanatory road. Let’s see if Hale’s account fares better.

Hale’s Epistemology of Essence A Priori Knowledge of Essence Hale considers both, a priori and a posteriori knowledge of essence.4 I will briefly sketch his views on both. Hale gives an (SM)-type answer with regard to a priori knowledge of essence. He thinks that a priori knowledge of essence is ultimately grounded in knowledge of meaning.

In Search of a Structurally Complete Epistemology of Essence  153 To connect essence to meaning, it is useful to look at the difference between real and nominal definition(s): While real definition is the definition of a thing, nominal definition is the definition of a word for a thing. Hale (2013: 254) emphasizes that in some cases the real and nominal definitions of some x—even though they are not identical— can be given by using the same words. For Hale, these are “[c]lear and straightforward cases in which knowledge of meaning suffices for knowledge of essence”. These clear cases are the ones where we are able to give an explicit definition of a word, i.e. where we are able to state analytically necessary and sufficient conditions for the application of the word. Here is Hale’s example for such a straightforward case where knowledge of meaning suffices for knowledge of essence: If we know that a plane figure is correctly described as “square” iff it is made up of four straight sides of equal length, meeting at right-angles, we know the essence of a square, i.e. the definition of the thing square, by knowing the (nominal) definition of the word “square”. In such cases, Hale takes the essence to be transparent. Note, however, that Hale does not build his entire case for grounding the knowledge of essence in the knowledge of meaning on these rare examples where essences are transparent. He (2013: 256) considers cases where the essence is less transparent. For some—arguably most—words, it is quite hard to give explicit definitions, i.e. to state analytically necessary and sufficient conditions for their application. So, for most words, we might only be able to give some form of implicit definition. Hale even considers basic or fundamental concepts, like basic logical words, that cannot be defined at all. However, he also holds that in these less transparent cases, knowledge of the meaning of the words grounds knowledge of the essence of the things they denote: Even though the basic logical word “and” cannot be defined at all “knowing what [the word] ‘and’ means is sufficient for knowledge of the essence of truth-functional conjunction” (Hale 2013: 258). It is plausible that an account in the epistemology of essence that grounds knowledge of essence in knowledge of meaning faces the same kind of follow-up questions as an account in the epistemology of modality that grounds knowledge of modality in knowledge of essence. For Hale’s picture of how we acquire a priori knowledge of essence to be complete, we need a story about how exactly the transition between knowledge of meaning and knowledge of essence works and a story of how exactly we acquire knowledge of meaning. Does Hale’s account provide those stories in a satisfactory way? Concerning the transition between knowledge of meaning and knowledge of essence, Hale merely remarks that in both cases, the transparent and the non-transparent ones, knowledge of meaning suffices for knowledge of essence. On the face of it, this should be especially obvious in the transparent cases. There is, however, a worry one might have about the transition between knowledge of meaning and knowledge of essence in

154  Michael Wallner these transparent cases: Suppose the essence of square is transparent in Hale’s sense, i.e. the nominal definition of “square” and the real definition of square can be given by using the same words. So, we are in a position to know the essence of square by knowing the meaning of “square”. However, one might think that in order to actually come to know the essence of square in that way, one needs to know (or at least have some justification for believing) that the essence of square is transparent in that sense. The point here is that knowing the real definition is one thing, knowing that it is the real definition is another.5 One might worry that in order for a subject S to realize that the essence of x is transparent in that way (i.e. the real definition of x can be given by using the same words as the nominal definition of “x”), S has to have some knowledge (or justified belief) about the essence of x already.6 If we take (ii) to be the question of how we come to know the essence of some x qua essence, i.e. how we come to know that some proposition (or property) is essential to x, then the mere fact that some essences are transparent will not suffice for an answer. We would also have to be given a way to tell the transparent ones from the non-transparent ones. That is to say that an answer to (ii) must not only contain a story of how we come to know the real definition but also a story of how we come to know that it is the real definition. The complaint then is that in transparent cases it does not suffice to know the meaning, i.e. the nominal definition, of the word to know the essence, i.e. the real definition, of the thing. We also have to know that we are dealing with a transparent case. One might think that Hale’s claim that knowledge of meaning suffices for knowledge of essence is less problematic in the non-transparent cases. Remember Hale’s example: Even though the basic logical word “and” cannot be defined at all “knowing what [the word] ‘and’ means is sufficient for knowledge of the essence of truth-functional conjunction” (Hale 2013: 258). I take it that this seems less problematic than the transparent cases for there is no explicit definition to be had such that we could neatly separate knowing the definition from knowing that it is the real definition. Both these points are, in a sense, lumped together in the implicit grasp of the meaning of the word “and”. The analogous worry, however, is that this implicit grasp of meaning might presuppose some essentialist knowledge. Hale seems to be aware of this point. If words are individuated semantically, rather than merely phonetically or typographically, facts about their meanings are essential to them, so that knowledge of word meanings is just a special case of knowledge of essence. If knowledge of meanings were a priori, there would have to be some a priori knowledge of essence that is not mediated by knowledge of meaning, on pain of apparently vicious circularity. But it is much more plausible that knowledge of meaning is a pre-condition for a priori knowledge in general, and is itself acquired

In Search of a Structurally Complete Epistemology of Essence  155 a posteriori. With much regret, I must defer further ­discussion of these questions to another occasion. Hale (2013: 258–259) To be perfectly clear, taking knowledge of meaning to be acquired a posteriori does not get Hale’s account out of the woods. Even if we grant that knowledge of meaning is a posteriori, the assumption that words are individuated semantically is still bad for Hale’s account. Semantic individuation of words would make knowledge of meaning itself knowledge of essence, as Hale concedes. On this picture, (a priori) knowledge of essence would be grounded in (a posteriori) knowledge of essence, which does not get us anywhere if we seek to answer (ii).7 So, the crucial question for Hale’s account is whether or not words are individuated semantically. Both the story of how knowledge of meaning works and the story of how exactly the transition between knowledge of meaning and knowledge of essence works crucially depend on these points, the discussion of which Hale defers to a different occasion. Hence, Hale’s overall account of a priori knowledge of essence remains crucially incomplete because it does not sufficiently answer or address the follow-up question about the knowledge of meaning and the transition between knowledge of meaning and knowledge of essence. A Posteriori Knowledge of Essence I will now turn to Hale’s account of a posteriori knowledge of essence. Here, Hale is guided by the Kripkean inference model in the epistemology of modality. According to Kripke, we can know some necessities a posteriori by the following inference model: (K1) S knows/has justified belief that p → □p (a priori) (K2) S knows/has justified belief that p (a posteriori) ∴ (K3) S knows/has justified belief that □p (a posteriori) Kripke’s idea is that since one of the premises is known a posteriori, the conclusion is known a posteriori as well. Hale’s strategy is to adopt this Kripkean deduction model for his epistemology of essence. In a nutshell, that strategy consists in formulating and arguing for general principles of essence—principles asserting, schematically, that such-and-such a property is essential to its instances—from which we may infer specific Kripke conditionals which, in their turn, may serve as the major premises for Kripke-style inferences to specific essentialist conclusions. Hale (2013: 269)

156  Michael Wallner This is what Hale’s deduction model in the epistemology of essence looks like: (a priori) (H1) S knows/has justified belief that Fx → □xFx (H2) S knows/has justified belief that Fx (a posteriori) ∴ (H3) S knows/has justified belief that □xFx (a posteriori)8 Hale is offering an (SM)-type account, grounding knowledge of essence in knowledge of (H1)-type principles of essences from which the former is to be deduced. In what follows, I will critically reflect on the viability of this approach, especially concerning the issue of structural completeness. Similar to Kripke, Hale takes knowledge of essence in (H3) to be a posteriori for it is derived from at least one a posteriori premise. It is plausible, however, that the way we come to know the major premise in Kripke’s deduction model, (K1), is also part of the overall story in the epistemology of modality.9 Analogously, the epistemic pathway to the major premise in Hale’s deduction model, (H1), is part of the overall story in the epistemology of essence. I take it that the story of how we come to know the major premise in these deduction models is in fact the most interesting part of the respective epistemology.10 On this picture, thus, the most interesting bit of the a posteriori part of Hale’s epistemology of essence are the a priori arguments for principles of the form of (H1). I do not intend to harp on about the a priori vs. a posteriori issue. Instead, the point I want to make is that Hale seems to tacitly presuppose some knowledge of essence in his case for (at least) one of the (H1)-type principles of essence he seeks to establish. If this is true, Hale’s deduction model of the epistemology of essence is crucially incomplete. The principle in question is called Kind membership (KM): (KM) Any object is essentially an object of a certain kind. Hale doubts that this principle admits of strict proof, but he still makes a case that there is sufficient reason to accept it. I cannot rehearse this case in full detail. To make my point, however, it suffices to discuss one crucial thread of Hale’s argument. For (KM) to not be obviously false, the principle cannot range over all kinds. Very roughly, for Hale (2013: 270–271) being of a certain kind is a sortal property. Sortal properties are determined by sortal concepts, which are expressed by sortal predicates.11 While the sortal concept horse determines a kind, membership of which is plausibly essential to its instances, the sortal concept horse owned by the Queen does not. It is plausible that for any horse owned by the Queen, belonging to the kind horse owned by the Queen is not essential to it (Hale 2013: 271). So, for the principle of (KM) to be plausible, the notion of kind in it has to be restricted to those that are determined by what Hale calls pure sortals.

In Search of a Structurally Complete Epistemology of Essence  157 Yet, how can we distinguish pure sortals from the impure ones? To s­imply answer that the former are those that are essential to the objects that instantiate them would amount to question-begging. Hale considers the possibility of characterizing pure sortals as those that are not themselves restrictions of others. This works for our example, since the plausibly impure sortal horse owned by the Queen can be understood as a restriction of the plausibly pure sortal horse. However, so the objection goes, the sortal horse can be understood as a restriction of the sortal animal, which, in turn, is a restriction of the sortal organism. Hence, characterizing pure sortals as those that are not restrictions of other sortals would crucially under-generate pure sortals, for it is plausible that horse and animal are pure sortals. Hale takes the point of this objection but believes that there is room for an appropriate refinement of his characterization of pure sortals. He takes note of a crucial distinction in the way in which horse owned by the Queen is a restriction of horse and the way in which horse is a restriction of animal. While horse owned by the Queen is a mere semantic restriction of horse, horse is a natural, and not merely semantic, restriction of animal (Hale 2013: 273). Accordingly, pure sortals are those that are not merely semantic restrictions of other sortals. So, in order to be able to tell the pure sortals from impure ones, we have to be able to distinguish merely semantic from natural restrictions. I take it, however, that natural restrictions are those that hold (not in virtue of the meaning of the terms involved but) in virtue of the nature (or essence) of the respective sortals or kinds involved. In this picture, distinguishing pure from impure sortals, which is crucial for Hale’s case for (KM), requires essentialist information and Hale’s account is guilty of begging the question. I think that Hale (2013: 270–271) is ultimately aware of this when he doubts that there is a non-question-begging proof for (KM). The moral to be drawn from this is not that Hale’s reasoning concerning (KM) does not illuminate the principle or give us reason to believe in it. I think that Hale is indeed successful in this regard. The moral here is rather that the epistemology of essences cannot take its start in inferring essentialist propositions from general principles about essence, for, at least in this case, the argument for the respective principle presupposes essentialist knowledge of some sort.12 A complete story of the epistemology of essence has to begin earlier, so to speak. In the worst case, Hale’s deduction model of the epistemology of essence is guilty of question-begging; in the best case, it can only offer us an incomplete account of how we come to know essences.

The Road Map I will now try to synthesize the critical points made so far, in order to get a more systematic picture of the possible routes in the epistemology of essence and the respective desiderata on each route. I have argued

158  Michael Wallner that both Lowe’s and Hale’s account of the epistemology of essence are in a crucial sense incomplete. Metaphorically speaking, none of these accounts takes us all the way down the explanatory (or justificatory) road. The purpose of this section, to stay in the metaphor, is to chart out the winding paths of the epistemology of essence and draw a map to see what could await us at the end of the explanatory road. Less metaphorically put, the aim is to determine, at least in principle and as generally as possible, what a structurally complete picture of the epistemology of essence would have to look like. Remember, I started out distinguishing two kinds of answers to the question of how we acquire modal knowledge? While (MC)-type answers appeal to mental capacities, the essentialist account in the epistemology of modality gives an (SM)-type answer, grounding the knowledge of modality in the knowledge of essence. As I said, such an account faces two follow-up questions: One about the transition of knowledge of essence and knowledge of modality (i), and one about how we come to know essences in the first place (ii). This picture is completely general. It applies equally to the epistemology of essence. So, again the question of how we come to know essentialist propositions can be given two different answers: (MC)  One can appeal to specific mental capacities (e.g. conceiving, imagining, perception, intuition, …) that provide us with warrant for beliefs about essence; or (SM) one can ground knowledge of essence in knowledge of some other subject matter SM, arguing that it is in virtue of knowledge of SM that we come to have essentialist knowledge. Both kinds of answers raise follow-up questions: Framing Question: How exactly does applying these mental (MC.i)  capacities work to yield modal or essentialist knowledge? Transition Question: How does the transition between knowledge (SM.i)  of SM and knowledge of essence work? Source Question: How do we acquire knowledge of SM in the (SM.ii)  first place? Lowe’s (MC)-type answer in the epistemology of essence is ultimately incomplete for it fails to answer or adequately address the follow-up question (MC.i). Hale’s (SM)-type account only offers a partial picture of the epistemology of essence because he falls short of fully addressing and answering the follow-up questions (SM.i) and (SM.ii). However, it is not my intention to point fingers here. Instead, I wish to make a positive and more general point. The critical discussion of Lowe and Hale can teach us something about what a structurally complete picture of the

In Search of a Structurally Complete Epistemology of Essence  159 epistemology of essence would have to look like. Supplying answers to the respective follow-up questions of (MC)- and (SM)-type approaches is a desideratum a complete account of the epistemology of essence should fulfill. This means that no (SM)-type story in the epistemology of essence is structurally complete without answering (SM.i) and (SM.ii), i.e. as we called them, the transition question and the source question. But, again, there are two different kinds of answers to (SM.ii): (MC)-type and (SM)-type answers. Suppose that we give an (SM)-type answer to (SM.ii), grounding knowledge of SM in knowledge of SM1. This would prompt another (SM.ii)-type follow-up question about the epistemology of SM1. No (SM)-type answer would ever end the chain of (SM.ii)-type follow-up questions. We cannot satisfactorily answer source questions by pointing at more sources indefinitely. So, in order to prevent a vicious infinite regress, eventually, at some point in this chain, we will have to give an (MC)-type answer, alluding to some specific mental operations or capacities that provide warrant for beliefs about the respective subject matter.13 Since (MC)-type accounts prompt (MC.i)-type follow-up questions, no story in the epistemology of essence is complete without an answer to some (MC.i)-type question. Somewhat more formally, the argument runs as follows. 1. We can either give an (MC)-type answer or an (SM)-type answer to the question of how we come to know some subject matter (like modality or essence). 2. Any (MC)-type answer will prompt an (MC.i)-type follow-up question. 3. Any (SM)-type answer will prompt (SM.i)- and (SM.ii)-type follow-up questions (i.e. transition questions and source questions). 4. For the respective answers to the question of how we come to know some subject matter (like modality or essence) to be complete, all follow-up questions need to be answered.14 5. Since (SM)-type answers prompt (SM.ii)-type follow-up questions (i.e. source questions), (SM)-type answers all the way down must be incomplete on pain of a vicious infinite regress. 6. Hence, for a structurally complete story, eventually (at some point in the chain) we must appeal to an (MC)-type answer. 7. Since (MC)-type answers prompt (MC.i)-type follow-up questions, no answer to the question of how we come to know some subject matter (like modality or essence) is complete without an answer to some (MC.i)-type question. So, these structural considerations about the two different ways to answer the question of how we come to know essences are able to put a general constraint on any structurally complete answer to this question. If we want to go all the way down the explanatory or justificatory road,

160  Michael Wallner we will, eventually, have to appeal to some mental capacities and explain how their application exactly works to yield knowledge of the respective subject matter SM. I think that similar considerations lead us to an additional constraint: Roughly put, not any mental capacity we can plug into an (MC)-type answer will do the trick. Ultimately, we will have to appeal to a directly justifying mental capacity or operation. But when exactly does a mental capacity qualify as directly justifying? (DJ) A mental capacity or operation, C, is directly (or immediately) justifying only if the respective belief yielded by applying C does not epistemically depend on some further (background) belief. Pryor (2005: 183) rightly emphasizes that “the fact that you have immediate justification to believe P does not entail that no other beliefs are required for you to be able to form or entertain the belief that P”. However, in order for you to have direct or immediate justification for P, P cannot depend on some further belief for its justification. That is to say, P cannot epistemically depend on some further belief. Correspondingly, a mental capacity, C, is directly justifying only if the belief yielded by applying C is not in need of epistemic support by anything other than the underlying experience of applying C.15 So, the kind of mental capacity appealed to in an (MC)-type answer matters for structural completeness. The respective belief yielded by the application of a non-directly justifying mental capacity would epistemically depend for its justification on some further (background) assumptions that are themselves in need of justification. Hence, appealing to a non-directly justifying mental capacity in an (MC)-type answer, we will not arrive at the end of the explanatory or, as it were, justificatory road. My discussion of Husserl’s account of the epistemology of essence will put some more flesh on these rather abstract issues. There, I will argue that imagination (at least as it is employed in Husserl) is not a directly justifying mental capacity, for the beliefs about essence we arrive at on the basis of imagination in Husserl epistemically depend for their justification on prior assumptions that are themselves in need of justification. Before I’ll get to that, however, let me mention two clarifications or disclaimers about this road map and the requirements for a structurally complete account in the epistemology of essence that come from it. First, I do not mean to suggest that no account in the epistemology of essence that falls short of being complete in this sense is useful. I have already said that I take Lowe’s and Hale’s accounts to be very illuminating about many issues. Of course, a partial answer is better than no answer. I take it, however, that breaking down the general desiderata of a structurally complete account of the epistemology of essence is a desideratum in itself. Second, the road map paints a foundationalist picture. This is most probably because the author of the map was tacitly presupposing a

In Search of a Structurally Complete Epistemology of Essence  161 broadly foundationalist picture of epistemology in general. It is a very interesting question of what the roadmap of the epistemology of essence would look like on a coherentist or infinitist picture. Unfortunately, I am unable to address this question in this paper. However, I wish to make the foundationalist assumption explicit by premising my argument here on (some suitable kind of) epistemic foundationalism.

Husserl’s Epistemology of Essence Eidetic Variation (EV) Husserl, like Lowe and Hale, defends an essentialist (SM)-type account in the epistemology of modality. He thinks that we have modal knowledge in virtue of having knowledge of essence. So, his account of how we know about modality is also subject to (SM)-type follow-up questions about the transition between essentialist and modal knowledge and about how we come to know essences in the first place. As stated in Section “Introduction”, in this paper, I am going to concentrate on the latter question.16 So, what is Husserl’s epistemology of essence? How is it, according to Husserl, that we acquire epistemic access or knowledge of essences? Husserl’s (MC)-type answer in the epistemology of essence has it that we are able to, as he sometimes puts it, “see” or intuit the essence of an object of a particular kind by varying the object in imagination. By imaginatively producing variants of the object under consideration, we will eventually be presented with an invariant structure that acts as a restraint upon the arbitrary variations (Mohanty 1991: 264). As soon as this invariant structure becomes evident, we have intuited the essence of the object qua member of a particular kind. This is roughly the gist of Husserl’s (in)famous method of EV and “eidetic seeing”. It helps to give a model of EV. I will discern four crucial steps of EV and then discuss the intricacies of each step, in turn: (Step 1) Start with an example of a particular object of a specific kind that you have experienced or imagined. (Step 2) Turn it into an arbitrary example of that kind (i.e. into a guiding model). (Step 3) Start voluntarily and arbitrarily varying the arbitrary example (i.e. the guiding model) in imagination by varying an arbitrary feature of the original, i.e. the guiding example. (Step 4) Eventually, a unity (an invariant structure) becomes evident in the multiplicity of variants, a general form without which an object of the kind in question cannot be intuitively imagined as such. Apprehending this unity, i.e. this invariant structure is intuiting (“seeing”) the essence of the object qua example of the kind (i.e. general essence of the kind).

162  Michael Wallner Step 1: Note that this method starts with a particular object of a ­specific kind. Correspondingly, the result of an EV is not the essence of the object simpliciter, but rather the essence of the object qua instance of a kind; that is to say the essence of the object-cumkind (Spinelli 2016: 3) or of the object-cum-­universal (Spinelli 2021: 149). An example might be helpful: When I start with my copy of Experience and Judgment as an instance of the kind “book”, I will end up with what essentially belongs to my copy of Experience and Judgment in as much as it is a book. However, if I start with my copy of Experience and Judgment as an instance of the kind “material object”, I will end up with what essentially belongs to my copy of Experience and Judgment in as much as it is a material object. In Husserl’s words, the first EV will present me with the eidos “book”, the second with the eidos “material object”. In a sense, it is really the essence of the kind that is under consideration here.17 Step 2: Turning the object we started with into an arbitrary example of that kind (i.e. into a guiding model) is not a modification of the object in question but a modification of the way this object is regarded (Kasmier 2010: 23). To stick with our example, the point is to regard my copy of Experience and Judgment as merely one possible instance of a kind (say, “material object”) among other instances of that kind. The purpose of Step 2 is to treat actualities as possibilities among other possibilities such that it becomes irrelevant that we began with an actual experience of a book. Step 3: After the example has been turned into a guiding model in Step 2, Step 3 consists in running through a multiplicity of variations of this guiding model in imagination. For Husserl, the notion of a guiding example has to be taken quite literally, since the variants all have to be “concretely similar” to the original example (Husserl 1973: 341). What does that mean? All of the variants have to be concretely similar to the guiding example with respect to the relevant properties pertaining to the type under considerations, though arbitrarily differing in all other respects (Kasmier 2010: 23).18 Two important aspects have to be mentioned here. It is important that the fabrication of the multiplicity of variants is conducted voluntarily and in a consciously arbitrary manner. This ensures that we do not need to actually produce all of the infinitely many variants to intuit the essence, i.e. to be evidently presented with the invariant structure. Due to the fact that we conduct the variation consciously in an arbitrary fashion we can come to know at a certain point in the variation that we could go on in this way, allowing us to actually stop the variation. If we know that the variants are produced arbitrarily, we can stop the variation once we have intuited the pattern that underlies the variation.

In Search of a Structurally Complete Epistemology of Essence  163 Step 4: Even though imagination plays a central role in EV, Step 4 makes it clear that much of the epistemic work is done by intuition. To put it in slogan form: Steps 1–3 lay out the groundwork on which we ultimately are able to “see” or intuit the essence. What is this essence that we intuit on the basis of EV? It is the eidos “k”, i.e. the essence of the object qua instance of kind k. This essence is presented to us as the invariant structure in the multiplicity of variants, or, as Husserl puts it, as “a necessary structure […] [or] necessary laws which determine what must necessarily belong to an object in order that it can be an object of this kind” (Husserl 1973: 352). That is to say that through EV, we apprehend essentialist laws about the kind under consideration. The invariant structure is given as applying to every instance of this kind. It is in that way that we come to know essentialist propositions about individual objects-cum-kinds. It is crucial to note that, according to Husserl, both the apprehension of essentialist laws and the knowledge of their application to particular instances are not acquired inferentially but intuitively.19 So, Husserl gives an (MC)-type answer to the question of how we come to know essentialist propositions. He does not ground essentialist knowledge in knowledge of some different subject matter SM. Instead, his answer (directly) appeals to a mental capacity—imagination. Husserl also answers the follow-up question (MC.i) by describing in detail how applying this capacity works as a means to get essentialist knowledge (EV). So, can Husserl’s EV be taken to offer a complete account of the epistemology of essence? Does Husserl bring us all the way down the explanatory road? Are we there yet? I will argue shortly that the answer is “No”, and that this is due to the fact that Husserl crucially appeals to a mental capacity that is not directly justifying—imagination. Before I can get to that, however, it seems that we have bigger fish to fry. One might think that the question about completeness is the least of our concerns, since EV seems to obviously contain a vicious circle that renders it useless as a method to acquire knowledge of essence. The Circularity Objection against EV The circularity worry I am talking about concerns Step 3. I spell this worry out in detail. (C) How are we to recognize whether or not each variant is an instance of the type under consideration, i.e. whether or not each variant is relevantly similar to the guiding example? In order to be able to restrict the arbitrary imaginative variation to those variants that are relevantly similar, I must be able to tell variants of the type from

164  Michael Wallner things that do not belong to the type. If we were not acquainted with the essential features in question, we obviously could not recognize which of the imaginatively produced variants belong and which do not belong to the essence or type in question. Thus, the method presupposes the kind of knowledge that it is purported to provide access to. This circularity problem, of course, has not gone unnoticed. In the literature on the issue, there are two different defense strategies available.20 Mohanty’s (1991) answer to the circularity charge ultimately bites the bullet. It admits the circular structure of EV. The point of this interpretation is that EV is not meant to discover new knowledge of essences, but rather to clarify our implicit essentialist knowledge by transforming it into explicit knowledge. If we consider EV in this sense as a method of clarification rather than discovery, the circle we are left with is rendered non-vicious. However, such an account is somewhat unsatisfactory, since it leaves open the question of how we gain implicit essentialist knowledge in the first place. Kasmier (2010) argues for a different defense strategy. His way of answering the circularity objection rejects that EV must presuppose knowledge of essences to be able to yield knowledge of essences. Crucially, this strategy trades on the Husserlian distinction between two kinds of universals: Empirical types and pure essences. It is argued that EV only requires acquaintance with the former universals (empirical types) to discover pure essences and yield modal knowledge. Kasmier claims that once we understand EV as a so-called “purification” of an antecedently apprehended empirical type, the circularity worry, (C), can be dispelled. In order to make sense of this strategy and of EV as a method of purifying empirical types to essences, we have to take a look at the difference between empirical types and essence and how knowledge about the former is acquired. Both pure essences and empirical types are universals. Pure essences are universals that exist necessarily. Empirical types are contingently existing universals. The idea of contingently existing universals is not very common in philosophy. Spinelli (2017) offers a great way to come to grips with this rather exotic Husserlian conception.21 I will be returning to the contingency of empirical types in a moment. It is best, however, to start with Husserl’s thoughts on the epistemology of empirical types. For now, just think of empirical types as universals. Husserl thinks that knowledge of empirical types is brought about by intuition.22 Here is a very rough sketch of how this intuitive awareness of universals works, according to Husserl: Suppose you are looking at two objects that are similar with regard to their color. Eventually, you realize that they are of the same color, say blue. So, the two objects no longer affect you merely for

In Search of a Structurally Complete Epistemology of Essence  165 themselves but as a unity. The focus of your attention is no longer directed toward the two particulars but rather toward that which makes them similar. Due to the similarity of the blue-trope in object 1 and the blue-trope in object 2, you become intuitively aware of a universal, a type, which can be predicated of both objects. As soon as we apprehend or intuit the universal “blue” (no longer taken as an individual trope of a particular object, but as the type), we realize that it is possible that more than just one thing instantiates this type, i.e. more than just one thing can be blue. If we intuit the universal in that way “the universal itself is given to us; we do not think of it merely in a significative fashion as when we merely understand general names, but we apprehend it, behold it” ­(Husserl 2001: 292). Intuiting the universal means intuiting the possible continuity of instantiations of the universal. This is not to say that intuiting the universal means intuiting all its possible instances. It means, however, that we intuit the possibility of an open infinity, i.e. an open continuity of instances of the universal. Apprehending a universal makes us realize that its instances outstrip what I have seen or imagined. Now, why are these universals called “empirical types”? This is, admittedly, a somewhat misleading term, for they are apprehended by a priori means, via rational intuition. Husserl’s reason to call them “empirical” is the following: Since the acquisition of the universal took its start in the experience of actual objects, the kind of universals we have acquired in this vein are empirical types or empirical generalities.23 And this is where the major difference lies to what Husserl calls pure essences or pure generalities (Husserl 1973: 330). Empirical types are still tied to the actual world. What does that mean exactly? In as much as empirical types are apprehended on the basis of contingently given particular objects in actual experience, these types are grounded in their actual instances. The unity of the empirically acquired type is, as Husserl (1973: 339) calls it, a “contingent” one.24 That is to say that the reason why Husserl takes those kinds of universals to exist contingently lies in the way they have been apprehended by a subject. This point is in need of explanation. Why should the epistemology of the universal affect its metaphysics? This has to do with Husserl’s views on metaphysics in general. Husserl’s phenomenology is characterized by a crucial assumption that has been called ideal verificationism. In Husserlian phenomenology, for an object o to exist just is for there to be a way for o to be possibly given to a (possible) subject. So, in Husserlian phenomenology, the way some object o is apprehended (or constituted) by a (possible) subject is intimately tied to the way o exists.25 This explains why the fact that the empirical type was apprehended on the basis of contingently given particular objects makes the empirical type “grounded in” (or, as H ­ usserl would say, “founded on”) its actual instances and, hence, contingent. Their being grounded in their instances, i.e. in empirical particulars,

166  Michael Wallner makes it the case that empirical types are not “capable of prescribing rules to all empirical particulars” (Husserl 1973: 340). This, however, is precisely the job-description of pure essences. Pure essences are supposed to have modal import or modal force. Their lawlike structure, their normativity is supposed to prescribe rules to all empirical particulars. Now, to apprehend such pure essences qua essentialist laws that are capable of doing exactly that, the tie to the actual experience and to the contingently given extension of particulars must be severed. For Husserl, this is done by free imagination in EV. Only by explicitly treating the object we start with (Step 1) as a mere arbitrary example of the kind under consideration (Step 2) and by voluntarily and arbitrarily varying the guiding example in imagination (Step 3), we can come to be intuitively aware of pure essences that determine the essential structure of all objects of the kind in question (Step 4).26 In this picture, EV is a method of “purification”, as Kasmier (2010) calls it. Its aim is to purify the antecedently apprehended empirical type and to free it from its essential attachment to the actual course of the experience of contingently given particulars. In other words, the purpose of EV is to turn an empirical type into a pure essence. But what does this mean, exactly? Ideal verificationism will, again, be illuminating. If the way some universal is apprehended is decisive for the way the universal exists, the distinction between empirical type and pure universal can be interpreted as about the way the same universal is epistemically given. Take some universal, U. If we apprehend U solely on the basis of its contingently given instances, we have apprehended an empirical type. If we further contemplate U (or some o-cum-U) in an EV via free imagination, we come to apprehend a pure essence. But, so the critic might ask, empirical types and pure essences differ metaphysically. While empirical types are grounded in their instances, pure universals can prescribe rules to their empirical particulars. How can the same universal U have both (conflicting) properties? Again, the answer lies in Husserl’s ideal verificationism. “Being grounded in p” here is not a perfectly objectivistic metaphysical property but a way in which a thing is (possibly) constituted by some subject(s). The difference with regard to the presumably conflicting properties becomes epistemic. Apprehending U qua empirical type on the sole basis of its instances, we are not yet aware of U’s law-like structure and its modal import. Only by severing the ties to actuality through free imagination in EV, we become aware of that feature of U qua pure essence. We are now ready to see how taking EV as a method of purification can provide an answer to the circularity charge (C). Answer to (C): Step 3 requires us to tell variants of the kind or type in question from things that do not belong to the type. Husserl’s point is that concrete similarity with the guiding example will do. However, the problem arises from the fact

In Search of a Structurally Complete Epistemology of Essence  167 that similarity in all respects would not give us variants of the guiding example but rather duplicates. So, we have to know which the relevant respects are, according to which concrete similarity is needed. This, however, is not circular, since it does not require us to know the essence that we seek to determine. All we have to be acquainted with is the empirical type that we seek to purify in a specific EV. Being acquainted with the type, say “material object”, already means knowing in which respect the variants have to be concretely similar, they all have to be material objects. In sum, Kasmier’s answer to (C) is that EV does not presuppose knowledge of (or acquaintance with) essence but rather knowledge of (or acquaintance with) empirical types. Thus, EV is not viciously circular. It is, however, in a crucial sense incomplete, as will become clear in a bit. In explaining the difference between empirical types and pure essences, I have often alluded to Husserl’s ideal verificationism. Does this mean that in order to accept Kasmier’s purification interpretation of EV, we have to buy into Husserl’s general picture of ideal verificationism? No. The important point that gets us out of the circularity charge is that there is a direct (non-circular) way or apprehending enough of a kind or universal U (through intuition of U qua empirical type) such that we can feed it into an EV, in order to get information about what the relevant respects of similarity are. This can be secured, even if we drop the talk of two different universals that makes sense only against the background of ideal verificationism.27 EV and Structural Completeness Having dealt with the circularity objection, we can finally come back to the question as to whether EV offers a structurally complete account of the epistemology of essence. The answer we gave to (C) entails that EV presupposes knowledge of empirical types. In this sense, EV cannot offer a structurally complete picture of the epistemology of essence. I argue that this is due to the fact that imagination, as it is used here, is not a directly justifying mental capacity. The circularity objection shows that the beliefs about essence we arrive at via EV epistemically depend on beliefs about empirical types. The problem which leads to the circularity charge is that we need to restrict our imagination in EV. Step 3 requires the imaginative production of only variants that are relevantly similar to the guiding example. Negatively put, Step 3 requires us to recognize when an imagined object is no longer a variant of the kind in question. It is for this reason that any belief about essence yielded by an EV epistemically depends for its justification on background assumptions that appropriately restrict the respective EV. These background assumptions concern, as we have seen,

168  Michael Wallner the empirical types that we seek to purify in an EV. Hence, essentialist beliefs yielded by EV epistemically depend on beliefs about the respective empirical types. EV and imagination as it is employed in EV are, therefore, not directly justifying mental capacities. Referring to the road map, i.e. the general and structural criteria of a potentially complete account in the epistemology of essence given in Section “The Road Map”, we can say the following: Husserl’s (MC)-type account succeeds in giving a quite detailed answer to (MC.i), i.e. his account of EV. However, since the latter crucially appeals to a mental capacity that is not directly justifying but is epistemically dependent on background assumptions concerning empirical types, the account is not complete without also accounting for these background assumptions. The point that imagination is a non-directly justifying mental capacity is not exactly new. In recent literature, several authors have argued for similar claims. Gregory (2010), Kung (2010), and Berto and Schoonen (2018) all argue, roughly, that in many cases of imagination (i.e. cases in which imagination does not work purely sensory), the justificatory power of imagination, say in the epistemology of modality, crucially depends on background assumptions about the objects or scenarios imagined, on the basis of which these imaginings operate.28 Vaidya and Wallner (2021) discuss this issue in terms of the input-question: To what degree do mental capacities like conceiving and imagination depend for their justificatory power on their inputs? They argue that these inputs, i.e. the background assumptions, on the basis of which conceiving and imagination operate, are that which creates epistemic friction for these mental capacities. In cases where conceivability or imagination are applied in the epistemology of modality, Vaidya and Wallner (2021) argue that these epistemic friction creators, i.e. the background assumptions, are essentialist propositions (or beliefs). In the Husserlian picture, where imagination is applied in the epistemology of essence, the epistemic friction creators are propositions (or beliefs) about the empirical type in question. The claim here is not that imagination is of no epistemic use. Also, I do not claim that absolutely no application of imagination is directly justifying. What is important, however, is that Husserl’s application of imagination depends for its justificatory power on background assumptions concerning empirical types. This makes Husserl’s EV structurally incomplete in the sense advanced in Section “The Road Map”, for it can be taken to ground knowledge of essence in knowledge of empirical types. Note, however, that there might be a way to complete the story with Husserlian means. Husserl indeed offers an account of how we come to know empirical types. In as much as Husserl appeals to our intuitive awareness of universals, he has an (MC)-type answer to that question. I already gave the rough outline of how intuition of universals works, according to Husserl.29 So, there is a Husserlian answer to the (MC.i)type follow-up question of how we know empirical types. What is more,

In Search of a Structurally Complete Epistemology of Essence  169 it appeals to intuition or intuitive awareness, a mental capacity that arguably fits the bill of being directly justifying. Unfortunately, due to space limitations, I cannot argue for the claim that intuition is directly justifying. I take it, however, that, at least among those who are sympathetic to the use of intuitions in philosophy, this is the received view.30 So, on the assumption that intuition is directly justifying, I conclude that Husserl’s EV plus his (or a suitably similar) account of intuitive awareness of universals like empirical types meet the general and structural criteria for a complete account in the epistemology of essence presented in Section “The Road Map”. If we take EV to be the whole story of the epistemology of essence, it is structurally incomplete. Yet, Husserl’s story might be completed, such that, at least structurally, it is capable of taking us all the way down the explanatory and justificatory road. The details of this completion, however, have to be left for a different occasion.31

Conclusion I have argued that Lowe’s and Hale’s account of the epistemology of essence, in one sense or the other, falls short of offering a complete picture of the epistemology of modality. After I have laid out what I take to be general conditions for a structurally complete account in the epistemology of essences, Husserl’s approach was presented. Using the work of Kasmier (2010), I have defended Husserl’s EV against a circularity objection. This defense showed that EV alone falls short of a complete picture of the epistemology of essence for it crucially appeals to a mental capacity that is not directly justifying. However, I have indicated that if we supplement EV with a broadly Husserlian epistemology of universals, appealing to intuition, the story of the epistemology of essence might be completed. In as much as this story finally appeals to intuition, a directly justifying capacity, it gets us as far down the explanatory (or justificatory road) as it can get: Immediate ‘seeing,’ not merely sensuous, experiential seeing, but seeing in the universal sense as an originally presentive consciousness of any kind whatever, is the ultimate legitimizing source of all rational assertions. This source has its legitimizing function only because, and to the extent that, it is an originally presentive source. If we see an object with full clarity, if we have effected an explication and a conceptual apprehension purely on the basis of the seeing and within the limits of what is actually seized upon in seeing, if we then see (this being a new mode of ‘seeing’) how the object is, the faithful expressive statement has, as a consequence, its legitimacy. Husserl (1983: 36–37) There are three caveats that need to be mentioned. First, it seems that, at least in the structural and general sense I appeal to in Section “The

170  Michael Wallner Road Map”, the Husserlian account presented here has the edge over the Lowe’s and Hale’s approach. Note, however, that I left it open whether Lowe’s or Hale’s account could be further developed so as to meet the criteria of a structurally complete epistemology of essence laid out in Section “The Road Map”. Even though I have raised some worries especially with regard to Hale’s account, I did not argue that either Lowe’s or Hale’s approach are in principle incapable of being adequately refined and expanded. Second, as already mentioned, the notion of completeness I had in mind here is justificatory (and foundationalist) in nature. This means that an account is complete in this sense, if it succeeds in tracing the justificatory chain of our essentialist beliefs back to the very end of this chain. An account in the epistemology of essence can be complete in this sense without, say, accounting for all the essentialist beliefs in all areas or domains. Completeness with regard to areas or domains is another, different sense of completeness that I have not discussed. That is to say that the Husserlian account might be incomplete with regard to the latter meaning of the term. Suppose that some essentialist knowledge is only to be achieved through natural sciences, not via some method involving intuition and imagination. (Think, e.g., of the essentialist proposition that it is essential to water to contain hydrogen.) It is compatible with my arguments that the Husserlian picture is incomplete in this sense. I did not mean to argue that the Husserlian approach can account for all the essentialist beliefs in all areas and domains but merely that it drives those it can account for all the way down the explanatory and justificatory road. Third, the notion of completeness I had in mind here is completely general and structural. Such considerations are largely independent from material considerations. So, one might agree that Husserl’s account is in the general and structural sense complete, while simultaneously disagreeing with one (or more) of the particular, material answers that the Husserlian picture provides. I am sure that there is quite some potential to disagree with Husserl’s material answers. Particularly, one might have qualms with the way ­Husserl tries to get rid of the ties to the actual world in EV and to purify empirical types to pure essences by using imagination. However, sorting out these material issues is a task for another paper. The conclusion of this paper is that if we accept EV as a method of purification, and ­Husserl’s universal intuition, we can escape EV’s circularity charge and the ensuing epistemology of essence is structurally complete.

Acknowledgments I am grateful to an audience at the Pacific APA in 2019, the Tübingen Metaphysics Reading Group, and the Phenomenology Forum Graz where prior drafts of this paper have been discussed. Special thanks

In Search of a Structurally Complete Epistemology of Essence  171 go to Philipp Berghofer, Gaétan Bovey, Eli Chudnoff, Martin Glazier, ­Antonella Mallozzi, Sonia Roca-Royes, Markus Seethaler, Nicola Spinelli, Anand Vaidya, Jonas Werner, and Tobias Wilsch for fruitful discussions on the topics of the paper. I am also indebted to an anonymous reviewer for productive suggestions.

Notes 1 A more elaborated version of this argument can be found in Lowe (2008: 35–36). See also Sgaravatti (2016) for criticism of this argument. 2 Note that Lowe (2012: 940) recognizes that this example is controversial. He concedes that there is substantive disagreement concerning the modal profile of entities like statues and lumps and the possibility of such numerically distinct entities to coincide. However, Lowe holds that this stems from the fact that there is substantive disagreement concerning the essence of such entities. 3 Maybe Lowe thinks that we have a primitive understanding of what x is. Note, however, that he does not provide an argument for this claim. 4 For a somewhat different discussion of this distinction, see e.g. Tahko (2017, 2018). For a view that considers knowledge of essence to be typically a posteriori, see Mallozzi (2021). 5 See Vaidya and Wallner (2021: §6) for further discussion of this distinction within the epistemology of modality. 6 I suspect that the epistemic externalist will be less moved by this worry than someone with more internalist leanings. 7 Moreover, this view would make it unclear as to whether we are really dealing with a priori knowledge of essence here. 8 From there, Hale goes on to argue for the following inference model in the epistemology of modality: (H3) S knows / has justified belief that □xFx (H4) S knows / has justified belief that □xFx → □Fx ∴ (H5) S knows / has justified belief that □Fx

(a posteriori) (a priori) (a posteriori)

See Vaidya (2018) for a discussion. 9 For Casullo (2012: 284), knowing (K1) amounts to knowing the “general modal status” of p. 10 This concurs with Hale’s (2013: 269) own verdict that “the most interesting question here concerns which such general principles [of essence] may be established”. 11 “A sortal concept is more usually characterized as one which differs from other, merely adjectival, concepts by its being associated with conditions or criteria of identity, as well as conditions or criteria of application” (Hale 2013: 271). 12 Note that this argument is specifically about (KM). Hence, it does not show that no principle of essence can be known without presupposing some essentialist knowledge. It does show, however, that Hale’s account of our knowledge of essences is incomplete. 13 This, of course, presupposes that (MC)- and (SM)-type answers exhaust the class of possible answers to questions like how we come to have knowledge of some subject matter SM. 14 Views according to which knowledge of some subject matter is primitive might be seen to disagree with this premise. Perhaps, according to the primitivist about some subject matter SM*, we cannot tell a story how we come to

172  Michael Wallner know SM*, so there is no answer to an (SM.ii)-type follow-up question and (4) needs to be rejected. I take it, however, that such a primitivist owes us a story about why knowledge of SM* is primitive, i.e. a story about why there is no answer to the respective (SM.ii)-type follow-up question. Nevertheless, this story might not appeal to a mental capacity but maybe only to the fact that knowledge of SM* is innate. So, adopting primitivism about some relevant subject matter here might be a way to block the conclusion in (7). In response, I am inclined to weaken my point to the claim that any non-primitivist answer to the question of how we come to know some subject matter (like modality or essence) is complete without an answer to some (MC.i)-type question. Thanks to Anand Vaidya for making me aware of this. 15 For further discussion, see, e.g., Berghofer (2018, 2020). Note that Berghofer speaks of experiences being immediately justifying. I expand his account in order to talk about mental capacities being directly (or immediately) justifying. 16 Let me just very briefly mention that while Lowe and Hale take the transition from essentialist to modal knowledge to function via deductive inference, there is reason to believe that a Husserlian approach regards this transition to be non-inferential. (See also fn. 19 and fn. 26.) 17 For a discussion of Husserl’s metaphysics of essences, see, e.g., Mulligan (2004), Spinelli (2016, 2021), Thomasson (2017), and Zhok (2011). 18 One can already see how this motivates a certain circularity objection. More on this soon. 19 This makes it plausible that the transition from essentialist to modal knowledge, on the Husserlian picture, is not inferential but intuitive. (See also fn. 16.) Essentialist and modal knowledge are thus in a sense entangled: It is by apprehending the essence of a kind as an essentialist law that applies to every instance of the kind that we apprehend what is necessary for every object that is an instance of that kind. (See also fn. 26.) 20 See Vaidya (2010) for a discussion of a somewhat different account that is very much inspired by Husserl and for a defense of this account against a similar charge. 21 See also Sowa (2010, 2011). 22 For a contemporary account of intuitive awareness of abstract objects, see Tieszen (2005) and Chudnoff (2012, 2013). 23 Note that Husserl also acknowledges the possibility of apprehending empirical types on the basis of the experience of imagined objects. However, such imagination is not yet the method of free phantasy since it is imagining actually possible instances of universals. That is to say that the important Step 2 of EV has not consciously been carried out. 24 It is important to see that the acquisition of empirical types that Husserl is elucidating has always already taken place in our everyday experience of the empirical world. That is to say that the factual world of experience is experienced as a typified world. Things are, as Husserl says, experienced as trees, bushes, animals, snakes, birds, etc. What is given in experience as a new individual calls to mind the similar and has a horizon of possible experience with corresponding prescriptions of familiarity and has, therefore, types of attributes not yet experienced but expected (Husserl 1973: 331). Since we already had previous experiences with dogs, when we see a new dog we immediately expect typical behavior and features. The empirical type, i.e. the empirical concept “dog”, however, is subject to continuous corrections and changes. “Empirical concepts are changed by the continual admission of new attributes but according to an empirical idea of an open, ever-to-be-corrected concept which, at the same time, contains in itself the rule of empirical belief and is founded on the progress of actual experience” (Husserl 1973: 333).

In Search of a Structurally Complete Epistemology of Essence  173 25 Whether ideal verificationism commits Husserl to metaphysical idealism is subject of considerable debate. Due to space considerations, I cannot go into this any further. See Zahavi (2010, 2017) for a good overview of the debate and Hopp (2020) for a recent plea for the compatibility of ideal verificationism and metaphysical realism. 26 A different way to put that point is the following: We can only intuit the lawlike structure, the normativity or the modal force of pure essences by severing the ties to the actual course of the experience of contingently given particulars through EV. On this Husserlian picture, the transition between knowledge of essence and knowledge of modality is not inferential but intuitive. We apprehend the modal force of essences by intuiting essences through EV. Contrary to what one might think, this does not make Husserl a modalist about essence. Spinelli (2016, 2021) argues that Husserl’s notion of essence is broadly Finean. Indeed, it is plausible that Husserl endorses something like Non-Reductive Finean Essentialism in Wallner and Vaidya’s (2020) sense. 27 See Wallner (2021) for a discussion of how Husserl’s ideal verificationism affects his thoughts on essence and modality. 28 See Lam (2018) for a defense of the view that non-sensory imaginings can (directly) yield prima facie justification. Williamson (2007) also holds that the imaginative evaluation of counterfactual suppositions has to crucially hold fixed some background assumptions. However, Williamson seems to disagree with a certain version of the claim that those background assumptions are themselves in need of justification. See Vaidya and Wallner (2021: §6) for a detailed discussion. 29 For a more detailed discussion, see Husserl (1973, 2001). For contemporary accounts of intuitive awareness of abstract objects, see Tieszen (2005) and Chudnoff (2012, 2013). 30 See, e.g., Bealer (2002), Berghofer (2018, 2020), BonJour (1998), Chudnoff (2012, 2013), and Huemer (2007). 31 If the end of the explanatory (or justificatory) road is marked by the appeal to a directly justifying mental capacity like intuition, one might wonder why EV is not the end of the explanatory road, for it crucially appeals to intuition in Step 4. Chudnoff’s (2012, 2013) notion of intuition experience might be useful here. Since Steps 1–3 are constitutively important for the intuition of essence in Step 4, the whole EV (Steps 1–4) forms an intuition experience in Chudnoff’s sense. Given that EV crucially appeals to imagination, a mental capacity that, at least in this application, needs to be restricted by background assumptions that are themselves in need of justification, the whole intuition experience that is EV cannot be the end of the explanatory (or justificatory) road. However, if we take on board Husserl’s (directly justifying) method of universal intuition, we can account for the justification of these background assumptions (about empirical types) in a way that makes the whole story (EV + universal intuition) a structurally complete picture of the epistemology of essence. However, one might wonder why the intuition of empirical types doesn’t suffice, especially since the distinction between empirical type and essence might just be about the way we epistemically regard the same universal. The answer is because intuition of the empirical type does not give us evidence or justification concerning the full modal import of the essence in question.

References Bealer, G. (2002): “Modal Epistemology and the Rationalist Renaissance”, in: Gendler, T. & Hawthorne, J. (eds.): Conceivability and Possibility. Oxford: Oxford University Press, 71–125.

174  Michael Wallner Berghofer, P. (2018): “Why Husserl is a Moderate Foundationalist”, Husserl Studies, 34, 1–23. Berghofer, P. (2020): “Intuitionism in the Philosophy of Mathematics: Introducing a Phenomenological Account”, Philosophia Mathematica, 28 (2), 2004–2235. BonJour, L. (1998): In Defense of Pure Reason. Cambridge: Cambridge University Press. Berto, F. & Schoonen, T. (2018): “Conceivability and Possibility: Some ­Dilemmas for Humeans”, Synthese, 195, 2697–2715. Casullo, A. (2012): Essays on A Priori Knowledge and Justification. Oxford: Oxford University Press. Chudnoff, E. (2012): “Awareness of Abstract Objects”, Noûs, 47 (4), 706–726. Chudnoff, E. (2013): Intuition. Oxford: Oxford University Press. Fine, K. (1994): “Essence and Modality”, Philosophical Perspectives, 8, 1–16. Gregory, D. (2010): “Conceivability and Apparent Possibility”, in: Hale, B. & Hoffman, A. (eds.): Modality. Metaphysics, Logic, and Epistemology. Oxford. Oxford University Press, 319–341. Hale, B. (2013): Necessary Beings. An Essay on Ontology, Modality, and the ­Relations between Them. Oxford: Oxford University Press. Hopp, W. (2020): Phenomenology. A Contemporary Introduction. New York & London: Routledge. Huemer, M. (2007): “Compassionate Phenomenological Conservativism”, in: Philosophy and Phenomenological Research, 74, 30–55. Husserl, E. (1973): Experience and Judgment. Investigations in a Genealogy of Logic. Revised and ed. by Ludwig Landgrebe. Transl. by James S. Churchill and Karl Ameriks. London: Routledge & Kegan Paul. Husserl, E. (1983): Ideas Pertaining to a Pure Phenomenology and to a ­Phenomenological Philosophy. First Book: General Introduction to a Pure Phenomenology. Transl. by F. Kersten. The Hague: Martinus Nijhoff Publishers. Husserl, E. (2001): Logical Investigations. Volume 1. Transl. by J. N. Findlay from the Second German edition of Logische Untersuchungen. Ed. with a new ­Introduction by D. Moran. London: Routledge. Kasmier, D. (2010): “A Defense of Husserl’s Method of Free Variation”, in: Luft, S. & Vandevelde, P. (eds.): Epistemology, Archaeology, Ethics: Current Investigations of Husserl’s Corpus. New York: Continuum, 21–40. Kung, P. (2010): “Imagination as a Guide to Possibility”, Philosophy and Phenomenological Research, 81 (3), 620–663. Lam, D. (2018): “Is Imagination Too Liberal for Modal Epistemology?”, Synthese, 195, 2155–2174. Lowe, E. J. (2008): “Two Notions of Being: Entity and Essence”, Royal Institute of Philosophy Supplement, 62, 23–48. Lowe, E. J. (2012): “What is the Source of Our Knowledge of Modal Truths?”, Mind, 121 (484), 919–950. Mallozzi, A. (2021): “Putting Modal Metaphysics First. From Knowledge of Essence to Knowledge of Metaphysical Necessity”, Synthese, 198 (Suppl 8), 1937–1956. Mohanty, J. N. (1991): “Imaginative Variation in Phenomenology”, in: Horowitz, T. & Massey, G. J. (eds.): Thought Experiments in Science and Philosophy. ­Lanham: Rowman & Littlefield, 261–272. Mulligan, K. (2004): “Essence and Modality: The Quintessence of Husserl’s Theory”, in: Siebel, M. & Textor, M. (eds.): Semantik und Ontologie. Beiträge zur philosophischen Forschung. Frankfurt & Lancaster: Ontos, 387–418.

In Search of a Structurally Complete Epistemology of Essence  175 Pryor, J. (2005): “Is There Immediate Justification?”, in: Steup, M.; Turri, J. & Sosa, E. (eds.): Contemporary Debates in Epistemology. 2nd edition. Malden, MA: Wiley Blackwell, 202–222. Sgaravatti, D. (2016): “Is Knowledge of Essence Required for Thinking about Something?”, Dialectica, 70 (2), 217–228. Sowa, R. (2010): “The Universal as ‘What is in Common’: Comments on the Proton-Pseudos in Husserl’s Doctrine of the Intuition of Essence”, in: Ierna, C., Jacobs, H. & Mattens, F. (eds.): Philosophy, Phenomenology, Sciences. Essays in Commemoration of Edmund Husserl. Dordrecht: Springer, 525–557. Sowa, R. (2011): “Deictic Abstractions: On the Occasional References to Ideal Objectivities Producible with the Words ‘This’ and ‘Thus’”, Journal of Phenomenological Psychology, 42 (1), 5–25. Spinelli, N. (2016): Husserlian Essentialism Revisited. A Study Essence, Necessity and Predication. PhD Thesis, University of Warwick. Spinelli, N. (2017): “Husserlian Intentionality and Contingent Universals”, Argumenta, 2 (2); 309–326. Spinelli, N. (2021): “Husserlian Essentialism”, Husserl Studies, 37, 147–168. Tahko, T. E. (2017): “Empirically-Informed Modal Rationalism”, in: Fischer, R. W. & Leon, F. (eds.): Modal Epistemology after Rationalism. Dordrecht: Springer, 29–45. Tahko, T. E. (2018): “The Epistemology of Essence”, in: Carruth, A., Gibb, S. C. & Heil, J. (eds.): Ontology, Modality, Mind: Themes from the Metaphysics of E.J. Lowe. Oxford: Oxford University Press, 93–110. Thomasson, A. (2017): “Husserl on Essence. A Reconstruction and Rehabilitation”, Grazer Philosophische Studien, 94, 436–459. Tieszen, R. (2005): “Consciousness of Abstract Objects”, in: Smith, D. W. & Thomasson A. L. (eds.): Phenomenology and Philosophy of Mind. Oxford: Oxford University Press, 183–200. Vaidya, A. J. (2010): “Understanding and Essence”, Philosophia, 38, 811–833. Vaidya, A. J. (2018): “Analytic Essentialist Approaches to the Epistemology of Modality”, in: Fred-Rivera, I. & Leech, J. (eds.): Being Necessary. Themes of Ontology and Modality from the Work of Bob Hale. Oxford: Oxford University Press, 224–244. Vaidya, A. J. & Wallner, M. (2021): “The Epistemology of Modality and the Problem of Modal Epistemic Friction”, Synthese, 198 (Suppl 8), 1909–1935. Wallner, M. & Vaidya, A. J. (2020): “Essence, Explanation, and Modality”, Philosophy, 95 (4), 419–445. Wallner, M. (2021): “Is Perception Essentially Perspectival? Modality in Husserlian Phenomenology”, History of Philosophy & Logical Analysis, 24 (2), 351–377. Williamson, T. (2007). The Philosophy of Philosophy. Oxford: Wiley-Blackwell. Zahavi, D. (2010): “Husserl and the ‘Absolute’”, in: Ierna, C., Jacobs, H. & ­Mattens, F. (eds.): Philosophy, Phenomenology, Sciences. Essays in Commemoration of Edmund Husserl. Dordrecht: Springer, 71–92. Zahavi, D. (2017): Husserl’s Legacy. Phenomenology, Metaphysics, and Transcendental Philosophy. Oxford: Oxford University Press. Zhok, A. (2011): “The Ontological Status of Essences in Husserl’s Thought”, New Yearbook of Phenomenology and Phenomenological Philosophy, 11, 96–127.

8 Morals and Modals Puzzling about the Dual Use of Modal Verbs Arindam Chakrabarti and Ian Nicolay

“Should the inside of this body be outside (imagine if you may), These people would have to take a stick to keep dogs and crows away.” —Abhinavagupta, to illustrate how the optative/injunctive (viddhi liṅ) “-et/-yāt” form of verb ending signifies fantastical counterfactual possibilities.1

“Should” vs. “Should”: Modality and the Dual-Use Subjunctive Verbs Modality, as we know it—the logic and metaphysics of possibility and necessity—was never directly theorized in classical Indian philosophy.2 Needless to say, where there is no theory of modality, there is no theory of modal knowledge either. An inquiry into classical Indian views on the epistemology of modality will therefore require more than a little creative reconstruction. In this paper, we try to glean from Indian epistemology, logic, grammar, and semantics materials for the construction of a “new” account of modal knowledge. This account prioritizes our use and understanding of deontic modals (e.g. “Imagine this”, “Assume that…”, “You should think …”) over our grasp of truth-claims about possibilities and necessities. It is not only meant as an historical-comparative exercise, but as an attempt at exploring deep links between two uses of “should”, and correspondingly between iffy thoughts and normative thoughts, between counterfactual conditionals and moral imperatives. Such a project runs the best chance of authenticity if it can find a strand of interconnected modal concerns to guide it. There is no guarantee that the picture of modality that emerges will correspond exactly with any of the familiar conceptions of modality available in contemporary philosophy—­ culturally and historically specific as those conceptions are. We must therefore be ready to think outside the box (□) of mainstream contemporary philosophy of modality. The puzzling dual use of modal verbs—illustrated succinctly by the different senses of “should” in “should this be the case…” and “should DOI: 10.4324/9781003002192-10

Morals and Modals  177 this be the case?”—is no mere accident of sloppy English usage. In ­Sanskrit, this duality is most clearly seen in the two distinct senses of the optative mood3 —called the liṅ by the ancient grammarian Pāṇini, who lists six basic uses: vidhi (injunction), nimantraṇa (invitation), āmantraṇa (address), adhīṣṭa (recommendation) saṃpraśna (query), and prārthana (request).4 As these uses show, the polite prescription is the basic sense of the optative; it is the voice of advice, entreaty, and instruction. But there is also an apparently non-prescriptive use of the optative which can be characterized as a kind of provisional or conditional assertion. This use is most often seen in the formulation of subjunctive conditionals, including counterfactuals. Wittgenstein started to puzzle over the link between “if” and “should”— conditionality and normativity, modality and morality—in Tractatus 6.422: “When an ethical law of the form, ‘Thou shalt …’ (du sollst…) is laid down, one’s first thought is, ‘And what if I do not do it?’ (Und was dann, wenn ich es nicht tue?)”. The cross-cultural recurrence of this duality transforms what would otherwise be a linguistic curiosity into a full-blown philosophical puzzle: what does prescription have to do with iffy positing as in the antecedent of a conditional? This paper follows the thread of this puzzling dual use of subjunctive verbs into a labyrinth of logico-linguistic issues about the meaning of the generic stative modal verb syāt. Syāt is the optative third-person singular conjugation of the root √as “to be/exist”. Depending on the context, it translates as “could be”, “would be”, or “should be”.5 The tightest web of authentically modal concerns across the Indian tradition as a whole is woven of this thread. Apart from pervasive non-technical use, there are three philosophical contexts where syāt is treated in a logically technical way, and therefore subjected to penetrating analytic scrutiny. The first is the logical core of the conciliatory6 Jaina philosophical methodology of intellectual non-­ violence (ahiṃsā): the canonical formulation of syādvāda in the so-called saptabhaṅgī or “seven-fold predication”. However, the word “syāt” used here is technically said to be an indeclinable adverbial particle (avyaya or nipāta), rather than the finite optative verb from which it is derived. Other avyaya-s include the connectives na (“not”), ca (“and”), vā (“or”), eva (“only”), iva (“as-if”), and so forth. As an adverb, it modulates the main verb of a sentence, and so also the meaning of the sentence as a whole. Existential quantification over intensional contexts—the trademark device of modern modal logic ever since Leibniz—is even used to explicate the logical meaning of syāt. It is standardly paraphrased by the adverb kathañcit “somehow”, which is analyzed by contrast with sarvathā “In every way”. On top of that, the distribution and scope of syāt in the conjunction syāt P & syāt ~P make it no more contradictory than ◇P & ◇~P. Thus, there are deep and genuine reasons to understand syāt as a monadic modal operator whose function falls in the range usually covered by ◇. Still, we should not let this strong similarity eclipse the undeniable

178  Arindam Chakrabarti and Ian Nicolay differences between the two. In particular, there is no non-psychologistic way to understand the intensional contexts involved here, and hence no temptation to provide an impersonal metaphysical account of them. Tarka—counterfactual reasoning—is the second context where syāt is used in a logically technical way. In this context, syāt occurring in the antecedent and consequent of a conditional changes it from an indicative conditional to a subjunctive one: “If …were, then …would”. Such a major premise is used to reason noncommittally about what one does not take to be the case. A famous example—the 29th verse of Nāgārjuna’s Vigrahavyāvartinī, where he responds to the claim that his view about the vacuity of all views must, in the spirit of consistency, be applied to itself—has become the slogan for the refutation-only philosophical tactic, known as viṭaṇḍa: If I had a thesis then this would be my problem But I have no thesis; so this is not my problem.7 In the Nyāya tradition, viṭaṇḍa is characterized by contrast with two positive styles of argumentation: vāda (arguing earnestly in support of a thesis which one honestly believes) and jalpa (arguing sophistically in support of a thesis which one does not believe, or in a manner which one knows to be unsound). The thesisless Vaitaṇḍika forswears positive arguments and independent premises, restricting herself to showing the untenability of the opponent’s thesis based solely on his own commitments (his premises and the principles of logic). In the opinion of later Prāsaṅgika Mādhyamika-s like Candrakīrti, this is the only legitimate method in philosophy: to demonstrate what the necessary but unwanted implications (aniṣṭa-prasaṅga-s) of the opponent’s thesis would be if it were true. In this technical sense, “prasaṅga” means one of several absurdities (contradiction, circularity, infinite regress, etc.) incurred by an argument.8 Hence, the inestimable methodological value of tarka for a universal skeptic like Candrakīrti: its noncommittal nature allows the skeptic to pursue the philosophical enterprise while renouncing reliance on positive theses and arguments. Without counterfactual reasoning, the only resort for such a thinker would be quietism. The meaning, truth, and our knowledge of counterfactual conditionals are crucial topics for contemporary philosophy of modality too. But while contemporary discussions foreground the issue of their meaning and truth—attempting, for instance, to supply a formal distinction between the indicative and the subjunctive conditional—the classical Indian discussion of counterfactuals focuses less on how they are true than on how they are used. It foregrounds conversational pragmatics and the ethics of belief over logic and formal semantics. The final context where optative verbs like “syāt, bhavet” are treated in a logically technical way is the Mīmāṃsā tradition's theory of the

Morals and Modals  179 conversational pragmatics of Vedic injunctions. The textbook example is svargakāmo yajeta “One who desires heaven should perform ritual sacrifice”. Mīmāṃsā thinkers like Kumārila, Prabhākara, and Maṇḍana Miśra pioneered a comprehensive theory of the psycho-semantics and pragmatics of the optative endings in ritual prescriptions like this. According to these thinkers, prescriptive grammatical forms like the optative conjugational endings convey a certain linguistic power or force: the power (when all auxiliary conditions are in place) to impel an agent to do an action. Prābhākara Mīmāṃsaka-s equated this power with kāryatā: “to-be-done-ness” or in a word, duty.9 As statements of duty, these injunctions may therefore be represented schematically as Osφ “It is obligatory for s to φ” or “s ought to/should φ”. The most extensive and systematic treatment of modal concepts—­ specifically, the deontic ones, obligatoriness, and permissibility—across the history of Indian ideas happened under the umbrella of this theory.10 The Mīmāṃsā theory is intended as a fully general theory for all Sanskrit injunctive verbs—including those which are not constituents of ritual prescriptions, and even, with a minor caveat, to stative optatives like syāt. This generates a puzzle which we can call the problem of circling the diamond-box. At the most basic level, “syāt” appears to approximate “could be” (◇) in the first context, “would be” (□) in the second context, and “should be” (O) in the third context. Is the word therefore modally equivocal? If not, can one of these be singled out as the single unequivocal meaning of “syāt” in each case? Applying the Mīmāṃsā view to the previous two uses of syāt would prompt an inversion of the usual hierarchical picture of the modal sphere. The primacy of alethic modality over deontic, epistemic, etc., is currently—with a few notable exceptions11—held as sacrosanct as the primacy of declarative speech over imperative, interrogative, etc. These are clearly two sides of the same coin. The prescriptive moral judgment that “s should φ” can be paraphrased by the non-prescriptive modal judgment “It is morally necessary for s to φ”, just as the command, “S, do φ!” can be paraphrased by the statement “the speaker wants s to φ”. It is hard to imagine how to formulate the reverse paraphrases. But imperative speech and deontic modality hold the same pride of place for staunch Mīmāṃsakas that declarative speech and alethic modality hold for mainstream contemporary philosophers of modality. So, what if deontic modality were treated as primary and alethic modality as derivative? How would our philosophy (including epistemology) of modality have to be revised to accommodate the theory of what is possible (◇) and necessary (□) within the theory of what is permissible (P) and obligatory (O)? And finally, it should already be clear that the issues surrounding these three uses of syāt are as closely connected to questions of philosophical methodology as are those surrounding the use of conceivability arguments and thought experiments. So, in light of this crucial

180  Arindam Chakrabarti and Ian Nicolay and universal importance of modal knowledge to the process of philosophy, what would the methodological consequences of this inversion be? This is the metaphilosophical thought experiment that this paper tries to motivate. The closest contemporary parallel for the classical-India-inspired view we develop here is the Modal Normativism of Amie Thomasson. In her words, “[The] diverse problems [with standard analyses of modality and modal knowledge] all arise from…treating modal claims descriptively, taking a modal claim to be an attempt to describe certain properties or features in this or another possible world…”.12 But, according to Thomasson, “modal terms do not function to describe or track special features of reality that we must discover to render our verdict. Instead, they serve as perspicuous ways of mandating or enforcing, reasoning with, and renegotiating rules”.13 Unfortunately, it falls beyond the scope of this paper to develop this intriguing comparison, which must therefore remain a direction for future research.

“May” and “Could”: Syāt vis-à-vis ◇ This section examines the use of syāt in the Jaina doctrine of syādvāda. We show that its use approximates that of an operator that quantifies existentially over intensional contexts; and further that it is used to parameterize apparent contradictions in order to show that they are merely apparent. Altogether, this supports understanding syāt as a possibility-operator. But the intensional contexts quantified-over are quite unlike any of those usually used to interpret ◇. The contexts are called “naya-s”—perspectives or ways of seeing—and are treated elsewhere in the Jaina philosophical system. But the difference between these “naya-s” and the intensional contexts standardly used to interpret ◇ is philosophically and methodologically significant. Pragati Jain has shown—through a close analysis of Malliṣeṇa’s Syādvādamañjarī on verse 23 of Hemacandra’s Anyayoga-vyavacchedatriṃśikā—that the understanding of “syāt” in the Jaina doctrine of syādvāda bears strong similarities with the contemporary analysis of “…is possible”.14 According to the Jaina metaphysical doctrine of ­anekāntavāda—illustrated by the famous parable of the seven blind men and the elephant—reality is multifaceted in its ultimate nature. It is quite possible for one and the same reality to satisfy two seemingly inconsistent descriptions15; for instance, “the mind is identical with the body” and “the mind is distinct from the body”. The logical device of syādvāda helps one corral this metaphysical plentitude. The so-called saptabhaṅgī (sometimes translated as “seven-fold predication”) is the canonical logical formulation of syādvāda. For a statement16 Ps (“Subject s has predicate P” or “s is P”), one should entertain and find a way to accept all of the following seven statements:

Morals and Modals  181 1. 2. 3. 4. 5. 6. 7.

Syāt Ps Syāt ~Ps Syāt Ps & Syāt ~Ps Syāt avaktavya-s Syāt Ps & Syāt avaktavya-s Syāt ~Ps & Syāt avaktavya-s Syāt Ps & Syāt ~Ps & Syāt avaktavya-s

Claim 3 represents the successive (kramataḥ) assertion of claims 1 and 2. The distribution and scope of syāt make 3 no more contradictory than ◇Ps & ◇~Ps. Claim 4 represents their simultaneous (yugapad, sahārpita) assertion.17 Since it is possible to say both that s is P and that s is not P, there is a genuine sense in which s is indescribable (avaktavya). Claims 5–7 are conjunctions of 1− 4 in each distinct combination. Claim 7 therefore represents the successive assertion of the successive and simultaneous assertions of 1 and 2—or equivalently, the successive assertion of 3 and 4. Although the conjunctions are no more informative than the unconjoined statements, their inclusion is considered nonredundant because it is essential to the instructive value and function of the saptabhaṅgī. Considering the same statements in these different configurations is thought to impart something, but not new information. As mentioned earlier, syāt in this context is held to be an indeclinable adverb (avyaya)—others of this category being the truth-functional connectives na (“not”), ca (“and”), vā (“or”), etc.—which is used to signify the multifacetedness of reality.18 Indeed, it must be taken as an adverb to avoid the ambiguity of a single atomic sentence (e.g. “syāt asti eva ghaṭaḥ”) having two finite verbs: asti “is” (present indicative third-person singular conjugation of the same root √as, “to be”, used existentially in this basic example but as a copula elsewhere) and the optative syāt “could be”. As an adverb, syāt modulates the main verb (typically a copula) in the sentence that follows it. Together, these points support the idea that syāt may fairly be understood as a modal operator in this context. But the strongest reason for interpreting syāt as a modal operator is that it is standardly glossed by the word kathañcit “somehow”. It amounts to saying that there is at least one way in which the modified statement is true—or assertible, at any rate. Here, we have unambiguous existential quantification over intensional contexts of some sort. But they are neither truth-value assignments, nor worlds (à la Leibniz, Lewis, and Lewis), nor times (à la Prior), nor subjects (à la Caspar Hare). They are epistemic stances or points of view (naya-s) which one and the same subject may take toward a proposition. Different types of stance are analyzed in the Jaina epistemological doctrine of nayavāda or perspectival pluralism. The exact number of distinguishable types (whether 7, 8, or 9) is a point of contention. The basic idea is this: from the practical, common-sense point of view, composite macroscopic objects possess color, but from the

182  Arindam Chakrabarti and Ian Nicolay equally legitimate point of view of their (colorless) atomic parts, they do not. There is no reason to reject one of these equally legitimate points of view in favor of the other. Thus, there is at least one way in which it is true that macroscopic objects possess color, and at least one way in which it is true that they do not. The availability of a point of view from which s is P is the assertibility condition for the formula syāt-Ps. Now, what about the philosophical interpretation of syādvāda? Naya-s are specific ways of standing in relation to a judgment. In this respect, they are like qualified doxastic attitudes (e.g. Believes from a common-sense point of view and believes from the perspective of atomic physics). At the same time, they are functions which take one from ­statement-stance pairs to assertibility. Ganeri characterizes naya-s as policies for the use of philosophical principles.19 As rule-governed procedures, they are comparable to Wittgenstein’s “games”. And they are tools for the systematic parameterization of apparent contradictions. Syād-Ps means that there is at least one rule-governed use of philosophical principles in which it is assertible that s is P—but implies that there is at least one from which its opposite is also assertible. The philosophical and imaginative exercise of identifying both points of view is left to the listener. This conception of modality offers a refreshing blend of logical, semantic, pragmatic, and metaphilosophical notions, but for that very reason, it is not possible to say that the type of possibility expressed by syāt is alethic and not (equally) epistemic or deontic. It should not be forgotten that syādvāda isn’t just a remarkable piece of formalism, but also a philosophical contemplative practice; and that this whole trio of doctrines—syādvāda, nayavāda, and anekantavāda—is often understood to be the intellectual counterpart of the cardinal Jaina moral virtue of ahiṃsā or non-harm. By offering a way to accept opposed views, it enables one to abstain from the intellectual violence of refuting another’s sincerely considered but necessarily limited opinion. The conciliatory philosophical methodology enshrined in this quartet of homologous doctrines is the distinctive contribution of Jaina thinkers to world philosophy.

“If” and “Would”: Syāt vis-à-vis □ This section examines the function of tarka (counterfactual reasoning) as it was understood and used across the several traditions of classical Indian philosophy. Counterfactual reasoning is reasoning which uses a counterfactual conditional as the major premise. In Sanskrit, one standard way of expressing a counterfactual conditional (as distinct from an indicative one) is to render the main verbs of the antecedent and consequent clauses in the optative mood. Frequently, the optative verb used is syāt. And, just like the formula □(P⊃Q), a Sanskrit counterfactual conditional thus formulated asserts that there is a necessary connection between P and Q, which obtains even when the antecedent and consequent are both false.

Morals and Modals  183 But the notion of necessity at play here is importantly ­different from mere logical necessity. An early example of tarka—formulated in the lṛṅ (conditional) rather than the liṅ (optative)—is found in the Chāndogya Upaniṣad (7.2.1): Had there been no Speech there would have been no knowing of duties and prohibited actions, no truth and no falsehood, no good, no bad, no one knowing-the-heart (what is enchanting) or ­not-knowing-the-heart (disenchanting). It is speech which makes all these known. Worship that Speech alone.20 Tarka is used here to discern the nature and value of Speech by pointing out how much we would stand to lose if we did not have it. Notice how pointing out unwanted consequences turns into contemplation, even veneration. Maitrāyanī Upaniṣad calls tarka the fifth limb of its six-limbed yoga. Tarka leads directly to Samādhi. We saw above that, according to Nāgārjuna, the methodological value of tarka stems from its ability to loosen the bonds of conditioned, conventional, and ultimately misleading ways of speaking and thinking. But the Nyāya tradition developed an elaborate method for the non-skeptical application of tarka. If a proposition implies any one of several absurdities, then it cannot be true—and its negation must be. This modal flavor of tarka is noticed by all interpreters, but, to ­foreshadow, this “must” is not the “must” of □, as it is usually interpreted. It became a key point of contention whether this kind of reasoning is a bonafide source of knowledge (pramāṇa) like perception and inference, or merely a helpful tool (sahakārīkāraṇa). The main reasons for thinking that it is not a pramāṇa are that it is parasitic on perception and inference, that its function is fictive not factive, and that its output is negative and presumptive, rather than positive and certain as knowledge is supposed to be. Naiyāyika-s like Udayana and Jayanta Bhaṭṭa held that tarka is not a pramāṇa. Dvaita Vedāntins like Vyāsatīrtha held that tarka is a sub-type of inference, and therefore a pramāṇa, but not a distinct and independent one. Jaina philosophers like Prabhācandra held that tarka is (and must be) a separate pramāṇa; specifically, it is the means of knowing the necessary connection between an inferential sign and what is logically signified by it. Tarka is enumerated as the eighth of 16 Nyāya padārtha-s (topics), knowledge of which leads to ultimate liberation from suffering: hence, very important but other than pramāṇa—the first of these 16 padārtha-s. The Nyāyasūtra defines tarka as follows: Tarka is a kind of ūha (=extrapolating conjecture necessitated by the context). When the real nature of an object is unknown, the conjecture that helps determine the truth of the matter by giving a causal justification for one alternative is called “Tarka”.21

184  Arindam Chakrabarti and Ian Nicolay Vātsyayana, author of the Nyāyasūtrabhāṣya, explains why tarka is not considered a pramāṇa in the Nyāya tradition: How is this [i.e. tarka] for the sake of knowledge of truth and not knowledge of truth itself? Because it is not ascertained with commitment. Between the two options, one property is favored, recommended (anujñā “this must be it”—notice the peculiarity of this “must”, which is like “should” or “has got to” be), but not firmly ascertained.22 In the 11th Āhnika of Nyāyamañjarī, Jayanta Bhaṭṭa gives the following example. The real nature of something standing erect at a distance is doubted: “Is it a man or a tree-trunk?” The context tells us that the area is a ground for horse-riding. Had there been a tree-trunk suddenly jutting out in the middle, it would not have been a horse-riding ground. Therefore, it must be a man. It is not as certain as seeing a man as a man, as Jayanta points out: 1. Doubt: Is it a man or a post? 2. Certainty: It is surely a man. 3. Necessitated Hypothesis: It must/should/ought to be a man. “The third—awareness of probability—is witnessed by our heart, surely”.23 But the doubt is removed by making the other alternative highly unlikely. This sounds like disjunctive syllogism: P or Q Not Q ∴P Later, Nyāya commentators see the process of eliminating Q as a reductio of that alternative: Let us assume that Q; let us suppose that is a tree-trunk over there. It follows that this place is unfit for practicing horse-riding smoothly (~R) But we already know that R. (It is a horse-riding ground.) To assume Q is to be forced to conclude R & ~R. This is intolerable. So the assumption (Q) also has to be dropped, on pain of contradiction. This is where prasaṅga (reasoning from threatening consequences) comes in. “Q” and “~R” are conceptually or causally related. If one (~R) is rejected, the other has got to be, must be rejected. Hence, Dharmakīrti: “Prasaṅga (=tarka) is the rejection of the other in absence of one, due to the relationship between the two”.24

Morals and Modals  185 Some features of tarka as understood by early Nyāya: It is a method of eliminating doubt. (tarkaḥ śaṅkāvadhiḥ “Counterfactual reasoning is the limit of doubt”. Udayana, Nyāyakusumāñjalī.) b Phenomenologically, it is between the oscillation of doubt and the full certitude of knowledge. It makes us feel strongly inclined to accept one option because the other option is threatened by an epistemic “punishment” of some kind, like incoherence. (“Must have fire, for firelessness would entail that the obviously smokey hill is smokeless!” But note that this “must” is doxastically weaker than the “is” of perception.) c It revolves around a counterfactual major premise connecting the possibility to-be-eliminated with an unacceptable consequence. This premise is typically formulated using optative forms like syāt: yadi vahniḥ na syāt tarhi dhumo ‘pi na syāt. If there were no fire, then there would be no smoke either. d It is not itself a pramāṇa. a

By the time of Udayanācārya (10th–11th century), a fully standardized account of tarka had emerged. I have to prove that a has G in the context of a debate. (I have or imagine an opponent who believes or argues that a does not have G.) I have given the basic positive argument: a has G (pratijñā) Because a has F (hetu) Whatever has F, has G, e.g. b (udāharaṇa) a has F which is pervaded by G (upanaya) Therefore, a has G (nigamana) But now the opponent raises a doubt about my claim of pervasion (∀x) (Fx ⊃ Gx). Maybe there are unobserved cases of F-possessing things which are not G-possessing! How do you know that a is not one of those unforeseen exceptions to the rule? Tarka (counterfactual reasoning) is deployed to eliminate this doubt: (T1) Had A been G-less, A would be F-less (T2) But A cannot be F-less, __________ (A’s being F-less would mean something undesirable even for my opponent.) (T3)Therefore A cannot be G-less T1, strictly speaking, is prasaṅga, a substitution-instance of the contrapositive of the original universal generalization or pervasion-rule. T2 is a demonstration of the undesirability of the consequence: ­aniṣṭatva-pradarśanaṃ. And T3 is “prasaṅga-viparyaya”: the false

186  Arindam Chakrabarti and Ian Nicolay assumption must be overturned on pain of some unwanted consequence (aniṣṭaprasaṅga). But again: this “must be” is of the nature of “ought to be admitted, otherwise we face an epistemic crisis of some sort”. It is significant that self-contradiction (logical impossibility) is not the only “undesirable consequence” brandished as a punishment in step T2. Aniṣṭaprasaṅga-s are of many kinds. So, to see the crucial counterfactual premise as being of the form □(φ⊃⊥) would be an oversimplification. In Nyāya and Vedānta, six (ṣaṭ tarkī) or more “absurd consequences” are listed.25 It could be: 1. 2. 3. 4. 5. 6.

Ātmāśraya (Self-Dependence)26 Anyonyāśraya (Mutual Dependence) Cakraka (Circularity) Anavasthā (Vicious Infinite Regress) Pratibandhi (tu quoque: you have the same fault!) Vinigamanāviraha (Absence of decisive proof—inability to tilt the dilemmatic doubt.)

Some other “unwanted consequences”, invoked in T2, include a “double noose” (ubhayataḥ pāśā) or catch-22 dilemma, utter senselessness/madness (vaiyātya), and pragmatic self-refutation (svakriyāvyāghata). This is not a failure to discriminate formal and informal considerations—a distinction to which classical Indian logicians were plenty sensitive. The form of reductio employed in tarka is not simply ad impossible, nor even ad absurdum; it works by pointing out an infraction of epistemic duty (e.g. coherence, but also explanitoriness, sensibility, and ingenuousness). This brings counterfactual reasoning under the purview of deontic logic and the ethics of belief. It would be strange to interpret this sort of “must be”— arising from the moral demands of the actual epistemic c­ ommunity—as expressing a metaphysical insight into all possible worlds. Two questions become urgent at this point. First, How much is included in Tarka, only T1 (The claim that the hypothetical assumption of the absence of the probandum (sādhya)/pervader (vyāpaka) would entail the absence of the prover (hetu)/pervaded (vyāpya)? or T1 + T2 + T3 —this entire package? Second, why is Tarka not itself a means of knowledge, given that it is such an essential remover of doubt and booster of the original inference, and a disputation trump-card? Of course, the concept of a “means of knowledge” cannot be mapped one to one on any Western epistemological concept, a reliabilist notion of justification coming somewhat close to it. Under the name of “ūha”, Jaina theorists of knowledge do give it the prestige of an independent method of knowing. Whether as a method of eliminating doubt concerning a possible negation (what if it is not the case that smoke must be absent

Morals and Modals  187 in a fireless place?) or as indirect inference to the contrary of a debating interlocutor’s hypothesis, Tarka deserves more attention in our theory of knowledge. Udayana does not give the traditional answer that Jayanta gave to the second question. He does not say that Tarka is not a pramāṇa because it falls doxastically short of certainty. He first makes it clear that only T1 is what is meant by tarka. But the “off-line” inference from the lack of f to lack of g is not seriously believed. It is a make-believe (āhārya) derivation of one (known to be false) conclusion from another (suspected to be false) premise. Hence, being wholly constituted by mock-beliefs (āhāryajñāna-s), it is not knowledge. In his Tarkatāṇḍava, the Mādhva philosopher Vyāsatīrtha devastates this argument and Udayana’s position. He establishes the Dualist Vedānta view that tarka is a variety of inference (anumāna)—specifically, a “refutational inference” (dūṣanānumāna)—and therefore a valid source of knowledge. He poses three responses to Udayana’s view that tarka is not a pramāṇa because it deals only in falsehoods or fictions, not truths: 1. Once we include T2 and T3 — closing the proof showing the absurdity of the consequence—the so-called “false superimposed hypothesis” (āhāryāropa), (~P) “has been eaten up by its own baby, like some crabs are supposed to be” (kulīrasya iva āpādakasya svaprasūtayuktyā pratihatatvāt).27 Since the false and un-believed assumption is overturned after all, there is no reason to withdraw the title of knowledge from the resulting belief that “Had ~P been the case, R & ~R would be the case, but that is absurd”. 2. Even if we take only T1, the conditional, the acknowledged falsity of both antecedent and consequent does not affect its truth-claim, because the whole thing is headed by an “If”. Its truth depends only upon the existence of a causal or conceptual relationship of inseparability between the two. 3. Even the antecedent “Should this hill be fireless” (yadi vahniḥ na syāt…) and the consequent “it would have to be admitted to be smokeless” (…tarhi dhumo ‘pi na syāt) are not individually false or erroneous. In the technical language of Navya-Nyāya, the relation that limits the hetu’s being in the pakṣa is “being assumed to be” (parābhyupagama) and the sādhya also only “has got to be admitted to be” (bhañjanīya=bhavitavya in Jayanta’s language) in the pakṣa. That is the point of using “syāt” instead of asti. “Syāt” does not carry existential or factive import. Thus, according to Vyāsatīrtha, tarka has all the credentials of a bonafide knowledge source, and may be classified as a sub-type of inference. In an iterated, reflexive, and meta-epistemological application of counterfactual reasoning, the Jaina philosopher Prabhācandra argues

188  Arindam Chakrabarti and Ian Nicolay that tarka must be acknowledged as a distinct and separate source of knowledge, on pain of universal skepticism. Prabhācandra poses his argument as a response to a Cārvāka (materialist/empiricist) pūrvapakṣin (imaginary opponent) who argues—roughly along the lines of Hume—that perception (pratyakṣa) is the only legitimate source of knowledge (pramāṇa), due to the irresolvable problems with the justification of induction.28 Prabhācandra agrees with the Cārvāka pūrvapakṣin that there are serious problems about the justification of induction, but shows that it is unacceptable to conclude that perception is the only source of knowledge. Apart from the problem of extrapolating from the observed to the unobserved, the Cārvāka opponent’s main reason for rejecting both inference and tarka is that they are parasitic or derivative (gauṇa) on perception. Tarka depends on the past observation of the smoke-fire connection; in addition to that, anumāna (inference) also depends on the present observation of smoke. Thus, due to the alleged independence or non-derivativeness of bonafide knowledge sources (agauṇatvāt pramāṇasya), two of the foremost candidates for pramāṇa status are not sources of knowledge at all; and by process of elimination, perception is the sole means of knowledge. This argument incurs at least the following six aniṣṭa-prasaṅga-s (not all of which are explicitly mentioned by Prabhācandra): 1. Perception is not non-derivative. We sometimes infer the existence of something and then become able to perceive it (pratibandhi). 2. Neither pramāṇa-hood nor (non-)derivativeness is perceptible. But if they are not perceptible and perception is the only means of knowledge, then we can’t know that perception is a means of knowledge in the first place, except by crass petitio (ātmāśraya). 3. Neither the anvaya (positive) nor vyatireka (negative) pervasion (vyāpti) required for this inference can be ascertained through perception alone. One would have to observationally check each member of the set, the extension of which is precisely the point in question (Vinigamanaviraha/vyāpyatva-asiddha). 4. Nor from inference, on pain of either circularity or infinite regress (cakraka, anavasthā). That would suppose another pervasion claim, which would need to be supported by another, in turn. 5. The act of inferring that inference is not reliable pragmatically contradicts itself (svakriyāvyāghata). 6. The inferred thesis that inference is not reliable contradicts its own content (svavacanavyāghata). Thus, according to Prabhācandra, if perception were the only source of knowledge then not even perception could be known to be a source of knowledge. Or in brief, inference, etc., must be accepted, or else we wind up with universal skepticism.

Morals and Modals  189 In the positive phase of the argument, Prabhācandra shows that just as inference is needed for the justification of perception, tarka (a.k.a. ūha) is needed for the justification of inference. He describes the function of tarka as upalabdhi-anupalabdhi29 “apprehension and non-apprehension”: not the mere observation of concomitance that motivated the original inductive generalization, but the understanding, based on careful consideration of the two properties serving as probans and probandum, that the latter can be present only if the former is, and must be absent if it is not30; or in other words, the probandum is a necessary condition of the probans, due to the inseparable causal or conceptual relationship between them (sādhya-sādhanayoḥ avīnabhāvaḥ). Provided that ◇ and □ are interpreted in a way that reflects the invocation of extra-logical prasaṅga-s, the outcome of tarka may be represented as ~◇(∃x)(~Fx & Gx), which is of course equivalent to the claim □(∀x)(~Gx⊃~Fx). The latter is the counterfactual contrapositive needed to safeguard the inductive generalization, (∀x)(Fx⊃Gx). Thus, upon Prabhācandra’s analysis, tarka is a distinct and independent source of modal knowledge. A puzzle is beginning to emerge—which, although translational and notational, is nonetheless interesting and important. The logical use of syāt in the saptabhaṅgī is best translated as “could be” and would best be symbolized as ◇. But the same word is best translated as “would be” in the context of tarka, whose counterfactual major premise (□(A⊃B)) is symbolized using □. Should we therefore understand syāt as a modally equivocal term, rather than one which is being used in a logically strict way across the two contexts? It would appear to only compound the problem to turn to the “official theory” of the optative, where syāt is best translated as “should be” and symbolized with the deontic modal operator O (“…is obligatory”). Nevertheless, that is the natural place to turn next.

“Ought” and “Should”: Syāt vis-à-vis O In the previous two sections, the sense of the optative verb syāt appeared to be the secondary, non-prescriptive one. Its assertoric force is, in both cases, is diminished (as compared with “is”), but neither context involves a strong sense that something ought to be done. (Though we have tried to indicate how, in fact, a tacit injunction is implied in each case.) However, prescription is taken to be the proper and primary sense of all liṅ forms by Mīmāṃsā thinkers, who developed a systematic account of it as a centerpiece of their philosophical system. This section surveys the Mīmāṃsā theory of prescriptive force, with an eye to applying it to the meaning of syāt. In the first aphorism of Jaimini’s Mīmāṃsāsūtra-s, the author defines mīmāṃsā as the inquiry ( jijñāsā) into dharma (duty or morality).31 For various reasons, it was held that this inquiry cannot proceed on the basis of perception alone, nor even perception aided by inference. Thus,

190  Arindam Chakrabarti and Ian Nicolay a separate means of knowledge must be called upon to account for the moral knowledge we humans have: namely śābdabodha, verbal or testimonial knowledge. Recall Chāndogya 7.2.1: Had there been no speech, there would be no knowledge of dharma or adharma. Mīmāṃsā thinkers consider the Vedic corpus to be the primary source of dharma: the foundational means of knowing what ought to be done. The core of the Veda, according to them, are the sentences enjoining ritual actions32 and giving instructions for their performance: who, when, why, how, etc. These ritual prescriptions are called “vidhi”; together with the religious prohibitions (niṣeda-s), they make up the distilled essence of the Veda. The goal of Mīmāṃsā is to refine a crystal-clear understanding of the Vedic teaching on what ought to be done, the practico-moral implications of those sentences for people’s lives. This requires, among other things, a comprehensive theory of language in general and of Vedic Sanskrit in particular. It is a common assumption of Mīmāṃsā and Vyākaraṇa (grammarian) philosophers that the verb is the heart of the sentence. The verbal form at the heart of a Vedic injunction is always a member of the class liṅādi (liṅ, etc., or in other words, prescriptive verbal forms). Hence the intense focus on the liṅ in Mīmāṃsā. The distinctive meaning of the optative is the center of the heart of the core of the Veda. Understanding the optative is the key to understanding what ought to be done. We can call the distinctive meaning of the optative “liṅ-ness” or liṅtva. Before discussing the Bhāṭṭa and Prābhākara theories of liṅtva, it is worth pausing to note that Mīṃāṃsā philosophers—especially the Prābhākaras—held a rather surprising conception of the Veda. According to them, the Veda is not a book like other bodies of sentences, including other (Hindu and non-Hindu) scriptures. The Veda is, according to them, authorless (apauruṣeya) and eternal (nitya). Not even God is credited with authorship of this scripture. In fact, some Mīmāṃsā thinkers denied the very existence of God in order to safeguard the infallible authority of the Veda. Couching the meaning of the Veda in the mind of an author would dilute this authority. For better or worse, this pious aversion to psychologism had a similar effect as Frege’s: the ontology of this system of moral-cum-modal relations approaches the Platonic vision of a pure, objective but extra-perceptual realm. If one is looking to do comparative metaphysics of modality, this is the place to go. The key contrast is that, in the Mīmāṃsā picture, modal reality consists not of what may or must be the case, but of what may or must be done. The initial discussion of the distinctive meaning of liṅādi forms occurs in the commentaries on Mīmāṃsāsūtra 2.1.1: “Words for actions (i.e. verbs) have occurrences/events as their meaning. From them actions should be understood. This indeed is the meaning enjoined”.33 Kumārila Bhaṭṭa in his Tantravārttika and Prabhākara Miśra in his Bṛhati on this verse proposed two opposing views, which were elaborated by later

Morals and Modals  191 philosophers of the Bhāṭṭa and Prābhākara schools of Mīmāṃsā. The stock example of a basic Vedic injunction is the sentence svargakāmo yajeta “he who desires heaven should sacrifice”. The project of identifying the distinctive meaning of the liṅ boils down to the task of saying, in full psycho-linguistic detail, how this sentence differs from the parallel one with the main verb in the indicative: svargakāmo yajati “he who desires heaven sacrifices”. According to both schools of Mīmāṃsā, verbal endings carry a ­psychosemantic force called bhāvanā (“realization”, literally “bringing into being”), a sort of causative import which brings about actions on the part of hearers. This action-propelling force is said to be of two types by later commentators. All verbal endings—indicative, optative, and the rest— convey what is called ārthībhāvanā “realization of meaning”. Hearing only the -ti in yajati informs one that someone does something; this naturally prompts one to wonder what action is done. Yāga “sacrificial ritual” is the lexical meaning of the root yaj-. Once one learns that someone sacrifices, one automatically wonders: Who? What? Why? How?, etc. This question-prompting power of verbal endings is the force of ārthībhāvanā . It operates until the full meaning of the sentence is comprehended, under two main conditions (one positive, one negative): 1. ākāṅkṣā: the semantic expectancy for specific complementary pieces of information, given a partial sentence. Its satisfaction is one condition for sentence comprehension.34 2. yogyatā, congruity in the emerging details.35 The absence or breach of yogyatā blocks sentence comprehension. For instance, one hears siñcati “he moistens”, and wonders, “with what?” One will not be able to understand the sentence if the only contextually available response is vahninā “with fire”. Similarly, if one asks “whose son?” about the sentence putraḥ siñcati and finds the response vandhyāyāḥ “a barren woman’s”. Like all verbal terminations, optative endings carry ārthībhāvanā. But they also have something that indicative forms do not: the power to prompt action. Unlike yajati, yajeta conveys prescriptive force and deontic modality. This obedience-commanding power is called śābdībhāvanā: it triggers a semanto-pragmatic process of determining what the prescriptive sentence requires. (Both the end and the means. Note: this word-powered process partly constitutes the possibility of the action enjoined.) The Mīmāṃsā spin of Wittgenstein’s remark is that a statement of the form “thou shalt…” also forces the question “And what if I were to?” This is the distinctive contribution of a liṅādi form to a sentence, the liṅtva (prescriptiveness) which is the essence of the Veda and its Dharma. It is in their accounts of śābdībhāvanā that the Bhāṭṭa and Prābhākara schools of Mīmāṃsā part ways. According to the more psychologistic

192  Arindam Chakrabarti and Ian Nicolay Bhāṭṭa theory, optative forms have the power to instigate action because they convey psychological motivation to the hearer. Specifically, an optative form like yajeta, under appropriate circumstances, coveys the following three pieces of information: 1. iṣṭasādhanatā: that accomplishing the enjoined deed is desirable for the hearer. 2. kṛtisādhyatā: that the deed can be accomplished through the will of the hearer.36 3. balavadaniṣṭānanubandhitva: that accomplishing the deed will not entail any overpoweringly unwanted consequences. According to Bhāṭṭa Mīmāṃsaka-s, learning these three facts suffices to bring about the action. For an appropriate hearer—a qualified person (adhikārin) with the requisite means and understanding of the injunction—there is no gap between “should”, “can”, and “does”. In the view of Maṇḍana Miśra, only the first element is really essential. To say that someone ought to do something (e.g. sacrifice) is already to say (implicitly) that it is both possible and desirable for them. Psychologically, this knowledge of ability and desirability suffices to remove one’s indifference and thus to motivate an action. So, liṅtva boils down to iṣṭasādhanatā. But Prābhākara-s found three problems with this understanding of liṅtva: (1) it doesn’t even capture the ordinary sense of liṅ; (2) it reduces the categorical imperatives of the Vedas to merely hypothetical ones (if you want heaven, then you should sacrifice); and (3) it psychologizes the realm of moral reality constituted by the Veda. They therefore proposed kāryatā (to-be-done-ness)—an irreducible deontic property— as the distinctive meaning of liṅtva.37 The concept of kāryatā respects the intuition that injunctive forms carry irrealis import and prescriptive force, in contrast to indicative ones which convey realis import and no prescriptive force.38 We are now in a position to apply the Mīmāṃsā theory of injunctions to the question of the dual meaning of syāt. In the midst of his long discussion of MS 2.1.1, Kumārila discusses the special case of generic stative verbs like asti, bhavati, etc.39 Hearing a verb-ending usually prompts the hearer to wonder what action is expressed; for this, she looks to the root of the conjugated verb. For instance, hearing the -ti in yajati (“sacrifices”) prompts the question kim karoti? (“Does what?”) to which the answer is yāga (“sacrifice”)—the lexical meaning of the root yaj-. But hearing the same endings attached to generic stative roots does not prompt the same kind of question. The sentence “ghaṭaḥ asti” (“There is a pot”) does not prompt one to ask what it is doing. Optative verbs are again a special case. In addition to ārthībhāvanā, they also convey śābdībhāvanā, which prompts the hearer to imagine how the sentence may be put into action. Thus, in the doubly special

Morals and Modals  193 case of stative optatives like syāt, assertoric force and existential/factive import are withheld (in accordance with liṅtva—ought does not imply is, but does imply can) and both śābdī- and ārthībhāvanā are conveyed. But as with all generic stative verbs, hearing “syāt” does not prompt one to wonder “kim kuryāt?” “what should be done?” Upon either the Bhāṭṭa or the Prābhākara theories, that piece of information is needed in order for śābdībhāvanā to impel physical action. Without it, the process stops short with ārthībhāvanā. The meaning of the sentence will be fully fleshed out and understood, but without the factive/existential import conveyed by indicative verbs. To put it another way, the practical activity brought about through syāt’s power of śābdībhāvanā (denoted by liṅtva) just is the full imaginative realization (but non-assertion) of the modified sentence’s meaning (ārthībhāvanā). We might say that the force of bhāvanā, in this exceptional case, turns in or loops back on itself.

Injunctive Modality and Our Knowledge of It We now have two questions on our hands: 1. Can the general Mīmaṃsā theory of optatives be applied to the— apparently non-prescriptive—uses of syāt in the contexts of syādvāda and tarka? Specifically, can it be applied in a way that attributes a more or less univocal meaning to “syāt” in all three contexts? 2. What would be the philosophical consequences of doing so? It appears to pose something of a doctrinal paradox that conciliatory Jaina syādvādin philosophers were also the only ones to accept tarka— which Vyāsatīrtha classified as dūṣaṇānumāna (refutational inference), and which Nāgārjuna used so effectively in the refutation of all views—as a distinct pramāṇa. But the compatibility and even complementarity of these doctrines are not too hard to see. In the saptabhaṅgī, syāt translates as “somehow, could be” and functions like a monadic modal operator. But it’s use (in service to syādvāda) is to impel the contemplator to identify a possible point of view from which the modified statement is assertible—and by doing so habitually, to inculcate a skill in the use of alternative philosophical principles, as well as a sense of the limits of one’s own perspective and the legitimacy of others’. Thus, it is easy enough to see syāt Ps as the consequent of a suppressed subjunctive conditional: if one were to look at it from point of view v, then it would be assertible that subject s has property P. This sets the intellectual task of imaginatively fleshing out the sufficient precondition—finding the value of the variable v—and thus “realizing” such an alternative point of view. But treating both Ps and ~Ps in this way has a tendency to “irrealize” both of them; thus, the irrealis moods of the optative in the canonical ­formulations of syādvāda and tarka are perfectly suited to reasoning

194  Arindam Chakrabarti and Ian Nicolay about each of the two. However, syāt Ps in the antecedent of a conditional sets the intellectual task of fleshing out the content Ps, systematically checking whether the implications of asserting it incur any infractions of epistemic duty—including, but not limited to, logical incoherence. Thus, the meaning of syāt in the first two contexts is not equivocal. And if a statement sets a task—mental or physical—for an agent, then it is prescriptive. As per the Mīmāṃsā theory, the generic stative optative syāt remains injunctive; specifically, it enjoins the hearer to do her intellectual duty in regard to the idea put forth. The sense of syāt thus approaches that of dhīmahi—“Let us reverentially consider, contemplate, meditate upon”)—and the function of bhāvanā comes closer to the Buddhist use of the term to denote a contemplative practice of imaginative realization. We can therefore settle on “should be” as the invariable, unequivocal meaning of syāt. In a conditional, the sense is “Should…be considered, … should be admitted”. Of course, the puzzling duality with which we began remains in these two instances of “should”. Now to the second question: what is the philosophical value of this interpretation? Apart from tracing a strand of connections between these three important classical Indian discussions of modal issues—in a way that partially address the problem of circling the diamond-box—we would just like to mention three key philosophical benefits: 1. It invites a broader understanding “modal epistemology”, in several ways: • It expands the sense of “epistemology” by drawing our attention to epistemic goods apart from acquiring JTB of modal facts/objects. The saptabhaṅgī and tarka are both, first and foremost, geared toward the elimination of error (e.g. about the limitations of one’s perspective and the tenability of one’s views), or at least doubt. • It expands the sense of “modal” by focusing on the question of our knowledge of deontic modality, i.e. morality. • It further expands the sense of “epistemology” by drawing our attention to less traditional epistemic goods like intellectual duty, responsibility, and virtue. Although flourishing themes in contemporary epistemology, these concepts have not yet been brought to bear on the theory of modal knowledge. 2. The Mīmāṃsā view on the primacy of prescriptive speech may help stir us from the dogmatic slumber of seeing language, logic, and modality primarily from the assertoric/alethic point of view. It also draws our attention from metaphysics and semantics, to pragmatics, theory of action, and ethics. 3. Together, 1 and 2 help to bring one of the most abstruse and ­ontologically prolific fields of philosophy back in touch with the practical, social, and interpersonal. Syādvāda—replete with a single-world

Morals and Modals  195 metaphysics of real alternative possibilities and modal semantics cashed out using existential quantification over intensional contexts— is, at the same time, a form of religious contemplative practice and of intellectual care and concern for the other (ahiṃsā). Finally, a word about the difficulties of cross-cultural philosophy of modality. The provocative claim with which we started—that there is no modality in classical Indian philosophy—is clearly an overstatement. There is no single Sanskrit word for modality (denoting the whole modal sphere and the current disciplinary angles on it)—just as there is no single word corresponding to “rationality”.40 And certainly nothing like possible worlds, metaphysically real but merely possible entities, and so forth. (It would be astounding if there were.) But there is no paucity of modal notions to be studied. The biggest challenge is that many of the most tempting leads are unconnected, and the important connections between others are, at times, hard to follow. This is the reason for the somewhat unusual approach we have taken in this paper—to propose a fresh angle on modality which is capacious enough to accommodate discussion of the innumerable special topics of classical Indian philosophy of modality. We have touched upon several of these special topics—e.g. invariable co-location (avīnābhāva), congruence (yogyatā), doability (kṛtisādhyatā), duty (kāryatā)—in order to indicate how they fit into the problem-space of classical Indian philosophy of modality. To conclude, let us suggest one provocative idea as a future research program. In exploring three different sub-traditions of Indian philosophy for the pragmatics and semantics of the optative/imperative/iffy uses of the phrase “syāt” in Sanskrit and in excavating the logical link between the two uses of the phrase “should” in English, we might have hit upon another radical theoretical possibility. Perhaps all our “if p…” or “Had p been the case…” kinds of unasserted antecedents of (subjunctive) conditionals are questions at heart: questions of the form “What if p?” The fact that a non-coercive imperative “Pass me the salt” is most felicitously expressed as “Will you please pass me the salt?” points toward this radical possibility: that the most fundamental unit of communicative action is the act of asking.

Notes 1 yadi nāmāsya kāyasya yad antas tad bahir bhavet | danḍam ādāya loko ‘yam śunaḥ kākāñśca vārayet || P. Shastri, ed. (1940). Dhvanyāloka of Ānandavardhana with Locana Commentary of Abhinavagupta. Varanasi: Kāśī Sanskrit Series Uddyota 1. 2 One section of the modal sphere has received heavy attention, for obvious reasons: logical necessity. Cf. Matilal, B.K. (1982). “Necessity and Indian Logic” in Logical and Ethical Issues of Religious Belief. University of Calcutta Press; Sarukkai, S. (2011). “Possible Ideas of Necessity in Indian Logic”

196  Arindam Chakrabarti and Ian Nicolay Journal of Philosophical Logic Vol. 40 No. 5; Gokhale, P. (2013). “The Concept of Necessity in Dharmakīrti’s Theory of Inference.” Presented as a guest lecture in the Buddhist Studies Center of Jadavpur University, Kolkata. But “modality” means the whole logically interconnected field, encompassing possibility, contingency, and necessity. 3 It is also present in prescriptive participles like kārya/kartavya, bhāvya/bhavitavya, sādhya, prameya, etc. The suffix -ya/-tavya has the sense of both -ible/ can and -worthy/should. 4 Vasu, S.C. trans. (1891) The Aṣṭādhyāyī of Pāṇini. Allahabad: Indian Press. 3.3.161. See Mahābhāṣya thereon for the distinction of each. 5 So Yablo’s contribution to Gendler and Hawthorne’s anthology Conceivability and Possibility could have been called “Syāt, Syāt, Syāt” in Sanskrit. 6 “Conciliatory” in the technical sense used in the epistemology of disagreement literature. The Jaina view is conciliatory because it strives to reconcile the two sides of the debate, without declaring a victor or a stalemate. 7 yadi kācana pratijnā syān me tata eṣa me bhaveddoṣaḥ/Nāstica mama pratijñā tasmānnaivāsti me doṣaḥ //. 8 The comparison with Sextus is obviously tempting. The basic ingredients (modes/prasaṅga-s) of ultimate goal (therapeutic suspension of all belief) are similar, but the method (antinomic proof vs. refutation of both contradictorily opposed hypotheses) is different. 9 But, as mentioned in a previous footnote, the ordinary use of the participle “kārya” ranges from doability to duty. 10 Elisa Freschi and associates have made substantial progress mapping the deontic-logical structure of this theory. Freschi, E. (2012). Duty Language and Exegesis in Prābhākara Mīmāṃsā: Including an Edition and Translation of Rāmānujācārya’s Tantrarahasya, Śāstraprameyapariccheda. Leiden: Brill; Freschi, E., Ollett, A., Pascucci, M. (2019). “Duty and Sacrifice: A Logical Analysis of the Mi ̄ mā m ̣sā Theory of Vedic Injunctions”. History and Philosophy of Logic; Freschi, E. Ciabattoni, A., Genco, F. A., Lellmann, B. (2017). “Understanding Prescriptive Texts: Rules and Logic as Elaborated by the Mīmāṃsā School”. Journal of World Philosophies. 11 The most prominent exception being G.H. von Wright. Thanks to Anand Vaidya for this reminder. 12 Thomasson, A. (2020). Norms and Necessity. Oxford: Oxford University Press, p. 14. 13 Ibid. p. 6. Thanks to Anand Vaidya for pointing out the strong affinity between our view and Thomasson’s. 14 Jain, P. (2000). “Saptabhaṅg i ̄ : The Jaina Theory of Sevenfold Predication: A Logical Analysis”. Philosophy East and West Vol. 50, No. 3, 390–393. 15 This may suggest that the saptabhaṅgī presupposes a non-classical (specifically, dialethic) logic. 16 A lot hangs upon whether this applies to all, most, many, or just a few statements. 17 The question whether Jaina logic is non-classical turns upon the question whether 4 is equivalent to “syāt Ps & ~Ps”. But 4, like 3, is said to be a conjunction of 1 and 2, where Ps and ~Ps are each in the scope of their own syād-operator. Perhaps the difference between 3 and 4 is one between two types of conjunction—successive (kramārpita) and simultaneous (sahārpita). 18 “Syād iti avyayam anekāntadyotakam…” Hemacandra, Syādvādamañjarī, under verse 25 of Malliṣeṇa’s Anyayogavyavacchedadvātriṃśikā. 19 Ganeri, J. (2019). “Epistemic Pluralism: From Systems to Stances”. Journal of the American Philosophical Association.

Morals and Modals  197 20 yad vai vāk na abhaviṣyan na dharmo na adharmo vyajñāpayiṣyan na satyam na anṛtam na sādhu na asādhu na hṛdayajño na ahṛdayajño | vāk eva etad sarvaṃ vijñāpayati vācam eva upāssva iti |. 21 avijñātatattve ‘rthe kāraṇopapattitaḥ tattvajñārtham ūhastarkaḥ (NS 1.1.40) Thakur, A, ed. (1997). Gautamīyanyāyadarśana with Bhāṣya of Vātsyāyana. New Delhi: Indian Council of Philosophical Research. pp. 36–37. 22 Katham punaḥ ayam tattvajñānārtho na tattvajñānam eva iti? anavadhāraṇāt anujānāti ekataraṃ dharmam, kāraṇopapattyā na tu avadhārayati | Ibid. 23 sthāṇur vā puruṣo vā iti pratītir ekā | puruṣa evāyam iti anyā | puruṣeṇa anena bhavitavyam (=“ought to be”) iti madhyavartinī | tritīyā sambhāvanāpratītiḥ svahṛdayasākṣikā eva | Śukla, S.N. ed. (2007 reprint) Nyāyamañjarī of Jayanta Bhaṭṭa, Vol. 2. Varanasi: Kāśī Sanskrit Series. 24 Prasaṅgo: dvayasambandhāt ekābhāve anya-hānaye. (Pramāṇavārttika, Parārthānumāna 12). 25 See S.S. Bagchi. (1953). Inductive Reasoning: A Study of Tarka and Its Role in Indian Logic. Calcutta: Calcutta Oriental Press. ch. 9. 26 As when the Sāmkhya position is refuted by the Tarka: “Had the effect been pre-existent in the cause, a thing would cause itself”. This is the example that is used in the Prāsaṅgika Candrakīrti’s discussion of prasaṅga under the MMK text: “na svataḥ”. 27 Madhvachar, V.V. ed. (1943). Tarkatāṇḍavam of Sri Vyāsatīrtha with the Nyāyadīpa of Ragavendratīrtha. Vol. IV. Mysore: Govt. Oriental Library. p. 144. 28 The relevant sections are translated and discussed in ch. 10 of K.K. Chakrabarti’s Classical Indian Philosophy of Induction: The Nyāya Viewpoint. Lanham: Lexington Books (2010). 29 Shastri, M.K. ed. (1990). Prameyakamala-mārtaṇḍa of Shri Prabhā Candra (A Commentary on Shri Manik Nandi’s Preeksha Mukh Sutra). Delhi: Sri Satguru Publications (3rd Ed.) p. 348. 30 “Idam asmin sati eva bhavati asati tu na bhavati eva iti” Ibid. p. 349. 31 Apte, V.G. ed. (1929). Mīmāṃsādarśana of Jaimini with the Bhāṣya of Śabara and Tantravārttika of Kumarila Bhaṭṭa. Pune: Ānandāśrama Sanskrit Series. p. 1. 32 karmacodanā; cf. MS/ŚBh 1.1.2. 33 bhāvārthāḥ karmaśabdās tebhyaḥ kriyā pratīyetaiṣa hi artho vidhīyate Apte, V.G. ed. Mīmāṃsādarśana. p. 370. 34 Cf. A. Chakrabarti, (1997). “Telling as Letting Know” in Matilal, B.K. and Chakrabarti, A. Knowing from Words: Western and Indian Philosophical Analysis of Understanding and Testimony. Dordrecht: Kluwer. 35 Cf. A. Chakrabarti, (1986). “On Understanding Falsehoods: A Note on the Nyāya Concept of Yogyatā.” Calcutta: Journal of the Asiatic Society. 36 A key topic for future studies in classical Indian philosophy of modality. 37 Freschi, E. (2012). Duty, Language, and Exegesis in Prābhākara Mīmāṃsā: Including an Edition and Translation of Rāmānujācārya’s Tantrarahasya, Śāstraprameyapariccheda. Leiden: Brill, pp. 22–46. 38 “Irrealis” is the linguist’s term for a grammatical mood (e.g. subjunctive) that lacks existential import or assertoric force. “Fictive” would be a close approximation, but not a perfect fit here. 39 Ollett, A. (2013). “What is Bhāvanā?”. Journal of Indian Philosophy Vol. 41, No. 3, pp. 221–262. 40 Chakrabarti, A. (1997). “Rationality in Indian Philosophy” in A Companion to World Philosophies. Deutsch, Eliot and Bontekoe, Ron, eds. Oxford: Blackwell. p. 259.

198  Arindam Chakrabarti and Ian Nicolay

References Apte, V.G. ed. (1929) Mīmāṃsādarśana of Jaimini with the Bhāṣya of Śabara and Tantravārttika of Kumarila Bhaṭṭa. Pune: Ānandāśrama Sanskrit Series. Bagchi, S.S. (1953) Inductive Reasoning: A Study of Tarka and Its Role in Indian Logic. Calcutta: Calcutta Oriental Press. Chakrabarti, A. (1986) “On Understanding Falsehoods: A Note on the Nyāya Concept of Yogyatā”, Journal of the Asiatic Society. Chakrabarti, A. (1997) “Rationality in Indian Philosophy” in The Blackwell Companion to World Philosophies. Deutsch, E. and Bontekoe, R., eds. Malden, MA: Blackwell. Chakrabarti, K.K. (2010) Classical Indian Philosophy of Induction: The Nyāya Viewpoint. Lanham: Lexington Books. Dhruva, A.B. ed. (1933) Malliṣeṇa’s Syādvādamañjarī with Anyayogavyavacchedatriṃśikā of Hemacandra. Mumbai: Bombay Sanskrit and Prakrit Series No. LXXXIII. Freschi, E. (2012) Duty Language and Exegesis in Prābh ākara Mīm āṃs ā: Including an Edition and Translation of R ām ānujācārya’s Tantrarahasya, Śāstraprameyapariccheda. Leiden: Brill. Freschi, E. Ciabattoni, A., Genco, F. A., Lellmann, B. (2017) “Understanding Prescriptive Texts: Rules and Logic as Elaborated by the Mīmāṃsā School”. Journal of World Philosophies, Vol. 2, No. 1, pp. 47–66. Freschi, E., Ollett, A., Pascucci, M. (2019) “Duty and Sacrifice: A Logical Analysis of the Mi ̄ mā m ̣sā Theory of Vedic Injunctions”. History and Philosophy of Logic, Vol. 40, No. 4, pp. 323–354. Ganeri, J. (2019) “Epistemic Pluralism: From Systems to Stances”. Journal of the American Philosophical Association, Vol. 5, No. 1, pp. 1–21. Gokhale, P. (2013) “The Concept of Necessity in Dharmakīrti’s Theory of Inference.” Presented as a Guest Lecture in the Buddhist Studies Center of Jadavpur University, Kolkata. Jain, P. (2000) “Saptabhaṅg i ̄ : The Jaina Theory of Sevenfold Predication: A Logical Analysis.” Philosophy East and West, Vol. 50, No. 3, pp. 385–399. Vasu, S.C. trans. (1891) The Aṣṭādhyāyī of Pāṇini. Allahabad: Indian Press. Madhvachar, V.V. ed. (1943) Tarkatāṇḍavam of Sri Vyāsatīrtha with the Nyāyadīpa of Ragavendratīrtha. Vol. IV. Mysore: Govt. Oriental Library. Matilal, B.K. (1982) “Necessity and Indian Logic” in Logical and Ethical Issues of Religious Belief. Kolkata: University of Calcutta Press. Ollett, A. (2013) “What Is Bhāvanā?” Journal of Indian Philosophy, Vol. 41, No. 3, pp. 221–262. Sarukkai, S. (2011) “Possible Ideas of Necessity in Indian Logic.” Journal of Philosophical Logic, Vol. 40, No. 5, pp. 563–582. Shastri, M.K. ed. (1990) Prameyakamala-mārtaṇḍa of Shri Prabhā candra (A Commentary on Shri Manik Nandi’s Preeksha Mukh Sutra), 3rd Ed. Delhi: Sri Satguru Publications. Shastri, P. ed. (1940) Dhvanyāloka of Ānandavardhana with Locana Commentary of Abhinavagupta. Varanasi: Kāśī Sanskrit Series. Śukla, S.N. ed. (2007 reprint) Nyāyamañjarī of Jayanta Bhaṭṭa, Vol. 2. Varanasi: Kāśī Sanskrit Series. Thakur, A, ed. (1997) Gautamīyanyāyadarśana with Bhāṣya of Vātsyāyana. New Delhi: Indian Council of Philosophical Research. Thomasson, A. (2020) Norms and Necessity. Oxford: Oxford University Press.

9 The Explanatory Power of Modal Rationalism Duško Prelević

Chalmers’s Modal Rationalism and Roca-Royes’s Essentialist Challenge Modal rationalism, broadly construed, is a view that rational reflection justifies our judgments about metaphysical modality. This reflection can be a priori or it can take corresponding empirical information into account. One popular way of spelling out this view comes from David Chalmers. In his account (see, for example, Chalmers 1996, 1999, 2002, 2010 for more details), ideal positive primary1 conceivability entails primary (or counteractual) possibility, which, together with established non-modal facts (and the semantics of concepts involved in propositions), enables us to gain corresponding knowledge about metaphysical (secondary or counterfactual) possibility. Here, ideal positive primary conceivability consists in conceiving of a counteractual scenario (or a counteractual situation) that verifies a proposition one is conceiving of, and which is undefeatable by better reasoning (Chalmers 2002: §§1‒3). Given that in this account conceivability is deployed in terms of conceptual coherence and of an ideal conceiver, it is not surprising that it is sometimes counted as “non-epistemic conceivability-based account” of our modal knowledge (Roca-Royes 2011a: §2). At first blush, Chalmers’s modal rationalism seems to be in accordance with a common practice in various fields of philosophy and sci­ onceivability methods are often used in metaphysics, conceptual ence. C analysis and mathematics, while predictions in science are often pursued by taking conceivable (and realistic) scenarios into account. Besides, Kripke’s necessary a posteriori statements (Kripke 1980), such as “Water is H2O”, are typically not counted as counterexamples to Chalmers’s account. Chalmers handles these cases by applying a two-dimensional semantic framework that enables him to claim that the statement “Water is H2O” is secondarily necessary and primarily contingent, which means that it is true in all counterfactual worlds and false in at least one counteractual world. That is because “water” and “H2O” have different intensions (functions from possibilities to extensions)2: the primary intension of “water” picks out (roughly) the colorless, drinkable liquid in all words

DOI: 10.4324/9781003002192-11

200  Duško Prelević considered as actual, whereas its secondary intension picks out the r­ ole-filler of such a liquid in the actual world (H2O) as well as in all counterfactual worlds (see Chalmers 1996: 59, 2006 for more details). However, Roca-Royes notices that Chalmers has overlooked that there are two sorts of Kripkean a posteriori necessities instead of one, since statements like “Water is H2O” can be read either in a de dicto manner (“Necessarily, water is H2O”) examined above or in a de re manner (“Water is necessarily H2O”). Nowadays, it is widely held that the explanatory power of modal rationalism is rather limited, for it cannot explain (non-trivial) de re cases, nor some important essentialist principles (see, for example, Roca-Royes 2011a: endnote 28, 2019; see also Vaidya 2008: 206; Hale 2013; Mallozzi 2018; Vaidya and Wallner 2021 for more details). For that reason, many philosophers in the field are more interested in pursuing alternative research programs (modal empiricism, in particular; see, for example, Fischer and Leon 2017). The essentialist criticism of Chalmers’s modal rationalism typically runs as follows. First, it is pointed out that we have good reasons to believe that some form of essentialism is true, as well as that Kit Fine’s well-known examples (Fine 1994), purported to show that essence is not reducible to necessity (although the former grounds the latter), are compelling. These examples, aimed to support a neo-Aristotelian modal metaphysics, have led many philosophers to believe that the epistemology of modality should be grounded on corresponding epistemology of essence (see, for example, Vaidya 2010; Roca-Royes 2011a, 2019; Lowe 2012; Hale 2013: §11; Tahko 2018). Relatedly, Roca-Royes (2011a: 39‒40) says that “if we have essentialist knowledge, conceivability methods do not account for it … as far as the epistemology of de re modality is concerned, conceivability cannot be the whole story” (cf. Roca-Royes 2011a: 38). She also says: “I am identifying the following, substantial explanatory deficit: on non-epistemic accounts, modal knowledge depends on a particular kind of knowledge, namely essentialist knowledge (whether modal or not)” (Roca-Royes 2011a: 42). In particular, Roca-Royes argues that modal rationalists are not capable of explaining our knowledge of interesting essentialist principles endorsed by many philosophers, such as the Essentiality of Origin, according to which individual’s origin is its essential (and therefore necessary) property, and the Essentiality of Kind, according to which individuals belong to corresponding kinds essentially. The Essentiality of Origin is Roca-Royes’s main illustration here. This principle states that origin is an essential property of an individual, which license a view that it is impossible for the individual to have a different origin from the one it actually has, given that essence grounds necessity. However, it seems perfectly conceivable that the very individual has different origins which, if true, corroborates the claim that conceivability does not entail metaphysical possibility, and that modal rationalism

The Explanatory Power of Modal Rationalism  201 is false thereof. Now, modal rationalists could try to show that it is ­inconceivable that an individual has a different origin. Yet, Roca-Royes thinks that this cannot be done without presupposing essentialism in advance. She (Roca-Royes 2011a: 38‒43) anticipates potential responses to her criticism arguing that they are not convincing. For example, if modal rationalists claim (by “crude stipulation”; see Roca-Royes 2011a: 41) that our conception of origin presupposes that individual’s origin is essential to it, this still, according to Roca-Royes, would not tell us anything about how we know that this is the case. Elsewhere, I have argued (see Prelević 2015a for more details) that Roca-Royes’s own account of our modal knowledge is not satisfactory. In this paper, I go a step further and argue that a modal rationalist of Chalmers’s kind can cope with the modal empiricist challenge that concerns the possibility of elucidating our knowledge of de re essentialist principles.3 It should be stressed from the very beginning that I won’t appeal in due course to conventionalism (see, for example, Carnap 1947; Sidelle 1989) or other anti-realist views about modality in order to handle potential counterexamples to modal rationalism. Were I to do so, the whole debate would be uninteresting, given that modal rationalism is compatible with anti-­realist views about modality. Had it turned out that modality is mind-­ dependent, that would pose a problem to modal empiricist accounts of our modal knowledge. Also, I will not try to diffuse Roca-Royes’s objection by claiming in advance that modality de re can be reduced to modality de dicto. However, I will assume in due course that a priori knowledge exists and that, for instance, logico-mathematical knowledge is of that sort (RocaRoyes does not dispute this). This assumption licenses a view that we know a priori that the Pythagorean theorem is true and that it tells us something about the essence of right-triangles (see Vaidya 2010: footnote 7). Although my aim is neither to challenge the validity of the Essentiality of Origin as such, nor the validity of the Essentiality of Kind, few words about the plausibility of those principles ought to be said at the outset. As for the Essentiality of Origin, Kripke himself famously offered a “proof” that is usually considered to be inconclusive (see, for example, Cameron 2005 for more details). Also, the very principle appears to be independent of Kripke’s influential views about rigid designators and corresponding identity statements, such as “Hesperus is Phosphorus” (see, for example, Bird 2004: §2; Mackie 2006: §10.6). Actually, it is the other way around: as Albert Casullo (2010: 356; cf. Kripke 1980: 110) reminds us, Kripke appealed to the Essentiality of Origin in order to illustrate how epistemic and metaphysical domains can be separated from each other, which, in turn, supports a view that the identity statements with rigid designators are necessary a posteriori. The Essentiality of Origin and the Essentiality of Kind are, as RocaRoyes indicates, accepted today by many philosophers who try to

202  Duško Prelević ground the epistemology of modality on a corresponding epistemology of essence; yet, these principles are far from being universally accepted and supported by our best scientific theories. When it comes to scientific justification of these principles, it should be stressed that many scientists and philosophers of science think that accepting them would actually depart from a common scientific practice. For example, cladism seems to be incompatible with essentialism, given that “a species becomes extinct whenever it sends forth a new side species” (LaPorte 2004: 54), which is something that occurs contingently. This amounts to the claim that both ancestry and belonging to a species are contingent. For that reason, Joseph LaPorte notices that “essentialists have simply been uninformed about systematics” (LaPorte 2004: 56; cf. Sober 1980; Dupré 1993; Christensen 2014: 83). He rather thinks that cladistic biologists merely stipulate that origin is an essential property (LaPorte 2004: 51‒52), which is by all means compatible with a modal rationalist approach. Critiques like the ones mentioned above seem to be forceful against those modal empiricists who adopt essentialism and believe that their view is supported by current scientific theories. It looks like they hold what Bas van Fraassen (2002) calls “the empirical stance”, that is, an empirically oriented set of commitments and beliefs as to the best way of approaching philosophical problems. However, modal empiricists might appeal to commonsensical examples instead. For instance, they might appeal to Aristotle’s distinction between substantial change and mere alteration (Aristotle 1984a: 319b 7‒24) based on our common belief that, for example, a tree would continue to exist, had it lost a branch, whereas it would cease to exist, had it been burnt down (Brody 1973: 353). The latter would be, in Aristotle’s terminology, a substantial change, while the former would be a mere alteration. Modal rationalists should take these examples seriously since, after all, they themselves typically think that many exotic scenarios,4 which are remote from everyday life and independent of currently accepted scientific theories, are both conceivable and metaphysically possible. Bearing this in mind, I will not appeal to the lack of scientific evidence in due course in order to diffuse modal empiricist epistemology of modality. Before I start elaborating my own criticism of Roca-Royes’s criticism of modal rationalism, let us recall that Chalmers himself (Chalmers 2010: footnote 3) provides a brief response to Roca-Royes’s objection trying to accommodate her purported counterexamples to modal rationalism. He thinks, contrary to Roca-Royes, that our knowledge of extensions of the terms like “origin” can be derived a priori from underlying non-modal truths about the world. Here, Chalmers appeals to “A Priori Scrutability” (defended extensively in Chalmers 2012), according to which all truths (including the truths about macro-physical objects, to which the Essentiality of Origin principle applies) can be inferred a priori from a compact class of truths (the scrutability base). Given that “A Priori Scrutability”

The Explanatory Power of Modal Rationalism  203 is aimed at amending Rudolf Carnap’s anti-realist stance, it is likely that the whole Chalmers’s strategy of dealing with essentialists’ worries about our knowledge of modality de re is conventionalist in spirit. Since my aim in this paper is to reconcile modal rationalism with a neo-Aristotelian account of modality (which relies upon accepting a realist view about modality and objects), I will not appeal to Chalmers’s response in due course either.5

Modal Rationalism and the Essentiality of Kind Given that Roca-Royes has argued that the explanatory power of modal rationalism is rather limited when it comes to our knowledge of widely accepted essentialist principles, it is a bit surprising that her paper is almost entirely devoted to the Essentiality of Origin, while the Essentiality of Kind has not been mentioned at all in the rest of it. However, the Essentiality of Kind is typically considered to be a hallmark of Aristotle’s essentialism. Here, it should be remembered that Bob Hale claimed that such a principle is “a central, and arguably fundamental case” in the epistemology of essence (Hale 2013: 269). Probably, Roca-Royes believes that the example of the Essentiality of Origin is representative enough, which means that if it turns out that modal rationalists cannot explain our knowledge of the instances of such a principle, it can easily be shown, mutatis mutandis, that this view cannot accommodate other interesting forms of essentialism either (see Roca-Royes 2011a: 30‒31). In order to show that this is not the case, let us start with our knowledge of the Essentiality of Kind (under assumption that it is true). In the previous section, we have seen that Roca-Royes challenges a view that conceivability alone can be our route to de re modal knowledge. However, it is not quite clear what “conceivability alone” means here. At one place, Roca-Royes (2011a: 31) mentions conceivability-under-the-scopeof-the-­pretense (presented by Janice Dowell in Dowell 2008) and argues that it cannot ensure that scenarios, in which essentialist principles are falsified, are inconceivable. Yet, there are good reasons to believe that such an assault on modal rationalism is making a straw man, since modal rationalists typically do not equate ideal conceivability with pretense understood (as in Dowell) as “trying to think ourselves into a state that mimics the belief that P” (Dowell 2008: 28). Nor should ideal conceivability be reduced to a mere describing (or supposing) a scenario (or a situation) in which certain proposition is true. Rather, ideal conceivability presupposes checking the coherency of the very scenario (or situation), since it is deployed in terms of undefeatability by a better reflection. For instance, the Naive set theory is not, according to Chalmers, ideally conceivable due to Russell’s paradox, and the like (Chalmers 2002: 155).

204  Duško Prelević Further, “better reflection” can be conducted in more ways than one, depending on which logical system one uses. In classical logic, reductio ad absurdum is a proof method commonly used. Here, it is worth recalling Gilbert Ryle’s well-known distinction between strong and weak reductio ad absurdum, of which the latter consists “in deducing from a proposition or a complex of propositions consequences which are inconsistent with each other or with the original proposition”, while the former presupposes a pertinent system of axioms that enable us to refute a statement that conflicts with them (or with their consequences; Ryle 2009: 206). This sort of reductio is typically used in logic and mathematics (Ryle appealed to some of Euclid’s demonstrations here), and it is also present in the case of ideal conceivability. For instance, “2+2=5” is not ideally conceivable since it contradicts to Peano’s axioms, the Naive set theory had been proclaimed as inconsistent due to Russell’s paradox, and the like. Both strong and weak reductio ad absurdum might be present when conceivability methods are applied. Equipped with Ryle’s notion of weak reductio ad absurdum, it should be remembered that there is more than one axiomatization in logic and mathematics. Pluralism in geometry is a good illustration here, since there is more than one consistent system of geometry to the effect that what is a theorem in one system need not be a theorem in another one. Some claims are, of course, true in all systems of geometry, such as the claim that a triangle is a figure with three edges and three vertices.6 It is in accordance with essentialism to say that these properties are essential to all triangles, as well as it is pretty much inconceivable that there is a triangle without three edges and three vertices. Moreover, in Euclidean geometry, the sum of angles of a triangle is equal to the straight angle, which is not the case in hyperbolic and spherical geometries (in the former, the sum of angles of a triangle is less than the straight angle, while in the latter, the sum of angles of a triangle is greater than the straight angle). Thus, within the Euclidean geometry, it is a priori true that the sum of angles of a triangle is equal to the straight angle, while in non-Euclidean geometries, the very proposition is inconceivable, which amounts to the claim that conceivability might depend on a previously accepted axiomatization.7 Elsewhere, I have pointed out that this can be represented in S5 system for modal logic, given that accessibility relation can be understood either as equivalence deployed in terms of reflexivity, symmetry and transitivity between possible worlds, or as universal accessibility, in which all worlds are accessible from one another (see Prelević 2020 for more details). Universal accessibility implies accessibility understood as equivalence, while vice versa is not the case. A model that represents accessibility relation understood as equivalence without universal accessibility is a case in which “multiple systems of worlds, such that within each system all worlds are accessible from one another, but across systems no worlds are accessible from one another” (Piccinini 2017: 85).

The Explanatory Power of Modal Rationalism  205 In view of the last fact, we can draw the distinction between absolute and relative ideal (positive/negative primary) conceivability (or inconceivability) in a sense in which the latter is, in contrast to the former, relative to a given system of axioms. For example, in the geometrical case above, it is absolutely inconceivable that a triangle does not have three angles, whereas it is relatively inconceivable that the sum of angles of a triangle is not equal to the straight angle. And so on. Likewise in geometry and logic, axiomatization in metaphysics is possible as well. Thus, it is reasonable to suppose that something might be conceivable with respect to one system of metaphysical axioms (or a set of beliefs) and inconceivable relating with some other system of metaphysical axioms (or a set of beliefs). In what follows, I argue that this is exactly the case with our knowledge of the Essentiality of Origin and the Essentiality of Kind. I will start with the latter principle, since I believe that this point can be put more clearly. First, lest us recall that there are three main views about the ontological structure of concrete particulars (under assumption that they exist)8: the substratum theory, according to which concrete particulars are bearers of properties (“bare particulars”) such that they themselves are not properties; the bundle theory, according to which concrete particulars are bundles of properties tied together by corresponding compresence relation; and the Aristotelian substance theory, according to which concrete particulars are irreducibly fundamental entities understood as instances of their proper kinds (see Loux 2006: chapter 3 for more details). All these theories had prominent adherents in the history of philosophy, and they can be elaborated in various ways (see Benovsky 2008 for more details). Now, let us sketch them briefly to see more clearly how our knowledge of the Essentiality of Kind and conceivability or inconceivability of the very principle depend on a previously accepted theory of concrete particulars.9 When it comes to the substratum theory, little reflection reveals that it can be modeled so that every property ascribed to a concrete particular could have been replaced by some other property. Thus, it is not strange at all that it is widely held that the substratum theory leads to anti-essentialism, as well as that from time to time philosophers who examine essentialist theories of modality explicitly refer to the substratum theory when addressing whether a situation in which the Essentiality of Kind does not hold is conceivable or not. For instance, Graeme Forbes says that when we believe that it is conceivable for a pot of marmalade to be a railway station, this means that “we would be conceiving of objects and their properties on the model of bare particular and inherence, according to which a thing is a propertyless substratum and can take on any nature you please via the inhering of appropriate properties” (Forbes 1997: 521).10 This view is quite compatible with modal rationalism, given that it is pretty much conceivable, under the assumption of the model of the

206  Duško Prelević substratum theory above, that the concrete particular above replaces any property that it actually has. Therefore, it is likely that the conceivability (and metaphysical possibility) of a scenario that falsifies the Essentiality of Kind depends on whether a corresponding model of the substratum theory has been already presupposed or not. Now, when it comes to the bundle theory, it is worth mentioning that this view can be modeled so that it leads to pan-essentialism, according to which all properties an object possesses are essential to it.11 But, like in the previous case, pan-essentialism is in accordance with modal rationalism under the assumption of a corresponding model of the bundle theory. Namely, if we define objects as bundles of properties, then it is inconceivable that the very objects could have possessed properties other than they actually possess and still be the same. Actually, Roca-Royes herself is well aware of the fact that pan-essentialists would render those scenarios impossible. She says (Roca-Royes 2011a: 39): It is probably safe to say that no one ‒ apart from pan-essentialists ‒ endorses the principle-like claim that if an arm is not broken, it is necessarily not broken. Nearly everyone, therefore will be happy with the conclusion that ◊a breaks her left arm. However, the relevant question is whether we can be said to know that conclusion via conceivability. Arguably, the answer is “no”. Be that as it may, it is likely that neo-Aristotelians in modal metaphysics (including Roca-Royes herself) prefer an Aristotelian view of concrete particulars over the bundle theory and the substratum theory. For example, Kripke himself simply takes such a theory for granted. He says: Philosophers have come to the opposite view through a false dilemma: they have asked, are these objects behind the bundle of qualities, or is the object nothing but the bundle? Neither is the case; this table is wooden, brown, in the room, etc. It has all these properties and is not a thing without properties, behind them; but it should not therefore be identified with the set, or “bundle”, of its properties. Don’t ask: how can I identify this table in another possible world except by its properties? I have the table in my hands, I can point to it, and when I ask whether it might have been in another room, I am talking by definition about it. Kripke (1980: 52‒53, cf. Witt 1989: 186‒188). However, when it comes to Aristotle’s own view of that matter, it is a commonplace that Aristotle’s hylomorphism, according to which concrete particulars are compounds of matter and form, corroborates the claim that some properties are essential to an object, while some other properties are accidental (see, for example, Brody 1973; Cohen 1978; Loux 2006: 107‒117; Witt 1989 for more details; cf. Aristotle 1984b).

The Explanatory Power of Modal Rationalism  207 Furthermore, many philosophers already noticed that the Aristotelian view of concrete particulars amounts to the claim that it is inconceivable that a concrete particular belongs to a kind different from the one to which it actually belongs. Here, it should be remembered that Marc Cohen has pointed out that “Aristotle’s theory neither provides us with nor allows us any way of conceiving of Socrates as being non-human” (Cohen 1972: 405). Cohen clearly emphasizes the distinction between Aristotle’s view of concrete particulars and that of the substratum theory: while the adherents of the substratum theory understand Socrates as that thing (which renders conceivable a situation in which, for instance, Socrates is an artichoke), Aristotle counted him as that man to the effect that, obviously, the sentence “That man is not a man” is inconceivable. Thus, it is in line with Aristotle’s account to hold inconceivable a scenario (or a situation) in which a concrete particular belongs to a kind different from the one to which it actually belongs, that is, a scenario (or a situation) in which the Essentiality of Kind is false. This is by all means in accordance with modal rationalism outlined in this chapter. Likewise, Timothy Sprigge, quoted by Kripke in a famous part of Naming and Necessity, in which arguments for the Essentiality of Origin are outlined, said that “it is logically inconceivable that the Queen should have had the property of, say, always being a swan” (Sprigge 1962: 203). Now, Kripke is inclined to think that there is no contradiction in a scenario in which we have been told by someone that “the Queen, this thing we thought to be a woman, was in fact an angel in a human form” (Kripke 1980: 111), which corroborates the claim that such a scenario is conceivable and metaphysically impossible. Yet, it is likely that this would be (in Chalmers’s account) a case of prima facie conceivability, rather than a case of ideal conceivability, because an ideal conceiver would by all means be in a position to check the trustworthiness of such a report. If these considerations are correct, then the whole case of the conceivability or inconceivability of a scenario that falsifies the Essentiality of Kind is parallel with the case of pluralism in geometry above: in the latter case, whether the sum of the angles of a triangle is equal to the straight angle is conceivable or not depends on whether one accepts Euclidean geometry or a non-Euclidean one, whereas in the former case, the conceivability of a situation in which Socrates is a non-human depends on which view of objects one presupposes. This corroborates the claim that Roca-Royes’s objection to the explanatory power of modal rationalism as to the Essentiality of Kind tacitly presupposes the Aristotelian view of objects, which, alas, implies that the negation of the very principle is inconceivable, making the whole enterprise self-defeating. Now, one might protest and say that our knowledge of the nature of concrete particulars (if we take it for granted) is already knowledge of their essences, which leads us back to Roca-Royes’s complaint outlined in the previous section.12 Indeed, Roca-Royes herself stresses that, in

208  Duško Prelević order to avoid trivializing inference from conceivability to m ­ etaphysical ­possibility, we should not understand concepts like that of origin as loaded with essentialist information (Roca-Royes 2011a: 41). However, modal rationalists need not accept such a constraint in its entirety. For example, had it turned out (counterfactually) that Pythagorean theorem for a triangle in Euclidean space (which, as we have seen previously, tells us something about the essence of right-angled triangle) is somehow relevant for elucidating our knowledge of the Essentiality of Kind, this would by no means be a question-begging explanation. So it is unlikely that modal rationalists cannot appeal to any kind of essentialist knowledge in order to explain how we could acquire knowledge of the Essentiality of Kind and the Essentiality of Origin. Besides, there are many other cases in which primary intensions do pick out essences of objects. For example, diamond and charcoal have the same chemical structure (which is, within the two-dimensional semantic framework, picked out by secondary intension), so it is likely that their primary intensions, which pick out their causal roles, make differentiating between them possible (see, for example, LaPorte 2004: 110). When it comes to water and its molecular structure, it is worth to mention Alfred Ayer’s remark that we would be surprised a lot had we been informed that water does not possess qualitative properties we typically ascribe to it. Ayer said: Suppose that in some part of this world we came upon stuff which had the chemical composition H2O but did not have the properties of falling as rain, allaying thirst, quenching fire and so forth, perhaps even failed to appear in liquid form. I certainly should not call it “water” and should be surprised if the majority of English speakers did so either. Ayer (1982: 270) Generally, determining the essences of things by referring to their causal roles (or functions) is not uncommon in science. For example, when we want to establish the essence of organisms, biologists typically appeal to their functions in terms of their causal interactions, ancestor/descendant relationship, growth, reproduction, metabolism, and the like (see Nichols 2010: section 2 for more details). Bearing this in mind, it is likely that primary intensions can encode information about essences of things, which is in accordance with a modal rationalist approach. Last but not least, the conception of essence that enables us to answer the question “What is it?”, and explain how constancy amid change is possible, might be regarded as non-modal (see, for example, Christensen 2014: 79). Roca-Royes (2011a: 42) is aware of that fact; yet, she claims that even then modal rationalists should explain how we acquire knowledge about non-modal essence and corresponding metaphysical

The Explanatory Power of Modal Rationalism  209 modality. However, we have seen at the very beginning that ­mathematical ­k nowledge, like the one that concerns the Pythagorean (essential) properties of right-angled triangles, is typically considered a priori and premised on corresponding axiomatization. If so, then it seems natural to suppose that our knowledge of the nature of objects might be understood along these lines too, as it has been presented above. At least the burden of proof is here on Roca-Royes’s side to show the opposite. Finally, if one wonders about whether modal rationalists can elucidate our knowledge of the nature of objects as such, that is, our knowledge of which theory of concrete particulars is true, modal rationalists might recall the mathematical case again. Namely, it is commonly held that our knowledge of mathematical truths is hypothetico-deductive since we know, for example, that in Euclidean geometry, the sum of angles of a triangle is equal to the straight angle in view of corresponding axiomatization and inference. Competent mathematician is the one who knows in which system of axioms one mathematical formula is a theorem, and in which this is not the case (or in which system of axioms the very formula is inconceivable). Adding further empirical information might enable them to prefer one system of geometry over another, like in a well-known episode in the history of science, in which the theory of relativity had been discussed (see, for example, Howard 2005: 38). Yet, when the nature of objects is at stake, it is likely that perception of ordinary objects is underdetermined by competing accounts, so we cannot know empirically which of them is true. Occasionally, the substratum theory is diffused as incoherent by some authors, given that, on the one hand, it is claimed that a bare particular does not have essential attributes, while, on the other hand, lacking essential attributes seems to be its essential attribute (see, for example, Loux 2006: 106). This, however, would be quite in accordance with modal rationalism, since in that case the whole theory would be just considered prima facie conceivable rather than ideally conceivable. However, the bundle theory is sometimes proclaimed as dependent on Leibnitz’s principle of the Identity of Indiscernibles13 which itself had been challenged many times. For example, if Max Black’s thought experiment (Black 1952) is correct, the bundle theory is not ideally conceivable, whereas if it turns out that the Identity of Indiscernibles is incompatible with quantum mechanics (see, for example, French and Redhead 1988), the bundle theory would not be counted as secondarily conceivable, which would be either way compatible with modal rationalist approach. Be that as it may, it is probably the case that, on the one hand, RocaRoyes herself simply assumes that the Essentiality of Kind is true and asks modal rationalists to explain such knowledge, while, on the other hand, neo-Aristotelians in modal metaphysics take the Aristotelian view of concrete particulars for granted.14 In her more recent paper, Roca-Royes (2019) has drawn the distinction between modally loaded and non-modally loaded concepts. She uses the

210  Duško Prelević concept human as illustration, telling that is a ­non-modally loaded concept defined as “a member of the genus Homo, especially Homo sapiens”, whereas and are modally loaded concepts deployed in terms of the Essentiality of Kind and the Essentiality of Origin, respectively (Roca-Royes 2019: section 2). According to Roca-Royes, the explanatory power of “concept-based epistemologies of modality”, to which Chalmers’s conceivability-based account belongs, is at best limited to modally loaded concepts. In order to show why this is the case, she refers to the following example: “Suppose there is an organism in front of me. Let’s call it ‘a’. Suppose also that I (reasonably) want to acquire (first-hand) knowledge as to whether a’s origin is essential to her”. Then, she proceeds by arguing that those accounts cannot elucidate such knowledge in a non-question-begging way. Setting aside that Chalmers’s account is more related to idealized knowledge, rather than first-hand knowledge, as well as that the essentialist knowledge is far from being an obvious datum in everyday life, it should be stressed that sentences like “There is an organism in front of me” or, more generally, “There is a physical object in front of me” are by no means metaphysically innocent, as it was suggested earlier in this section.15 So, it does not seem that appealing to the distinction between modally loaded and non-modally loaded concepts would render the conceivability-based account less convincing.

Modal Rationalism and the Essentiality of Origin Now, let us turn to the Essentiality of Origin case. At the very beginning, it should be remembered that Colin McGinn noticed a “haziness over what the origin of an individual is supposed to consist in” (McGinn 1976: 131). He himself agrees with Kripke by claiming that parents are essential to every individual, or, more precisely, that organisms are identical with their zygotes (McGinn 1976: 132; cf. Hale 2013: 278). Yet, the existence of identical twins is typically counted as a counterexample to the universal validity of the Essentiality of Origin (Witt 1989: footnote 18; Mackie 2006). Now, if one tries to distinguish between two identical twins by appealing to immunology (essentialists typically do not address this option) and argue that differences in their immune systems enable us to distinguish between them, it can easily be seen that the whole process of acquiring the “immune memory” is contingent (see, for example, Pradeu 2012: 7), and therefore by no means enables us to establish the essential difference above. Bearing this in mind, it seems more safe to say that they are identical with their embryos, since in that way the problem with monozygotic twins might be avoided, given that embryos, unlike zygotes, cannot be duplicated (see Alvargonzáles 2016 for more details). Be that as it may, it should be stressed that the Essentiality of Kind and the Essentiality of Origin are aimed at answering different questions

The Explanatory Power of Modal Rationalism  211 relating to the essences of things (see, for example, Christensen 2014: §2; Witt 1989: 191). Namely, Aristotle was mainly focused on addressing issues like how constancy amid change is possible, what determines the nature of things (such as concrete particulars), as well as whether appealing to essences provides good explanations of relevant phenomena. Given that one of Aristotle’s main aims was to answer the question “What is it?”, he was ready to endorse the Essentiality of Kind, while remaining silent on the Essentiality of Origin (see, for example, Witt 1989: 189). However, Kripke has been more interested in addressing the issues on identity and distinctness,16 so he was ready to endorse the Essentiality of Origin to make distinguishing an ordinary persisting individual from most individuals of the same kind possible (Mackie 2006: 94). However, although many philosophers think that the Essentiality of Origin and the Essentiality of Kind are independent of each other (see, for example, Roca-Royes 2011b: 70; Christensen 2014: 83), it is worth mentioning that not everyone agrees with it.17 For example, Bob Hale thought that “to be a member of a certain biological species is, in part, to have a certain sort of origin” (Hale 2013: 278). If one accepts this, then from the inconceivability of a scenario in which the Essentiality of Kind is falsified (once the Aristotelian view of concrete particulars is assumed), together with the Necessity of Identity (according to which the relation of identity holds necessarily, if it holds at all), which philosophers typically assume in the debate, it straightforwardly follows that the falsity of the Essentiality of Origin is inconceivable either. For if we assume that an adult person is identical with their embryo, and if the relation of identity holds necessarily, then a scenario in which the Essentiality of Origin is falsified turns out to be (secondarily) inconceivable either. In that case, modal rationalist justification of the Essentiality of Kind, presented in the previous section, would be (together with other background assumptions mentioned above) sufficient for justifying the Essentiality of Origin and diffusing Roca-Royes’s objection thereof. But let us suppose, for the sake of argument, that the Essentiality of Origin is independent of the Essentiality of Kind principle. In that case, the modal rationalist’s response above would not be available. Still, it might be the case that some other underlying assumptions relevant for elucidating our knowledge of the very principle are made. In order to see whether this is the case, let us recall Kripke’s famous passage in which the Essentiality of Origin is presented: How could a person originating from different parents, from a totally different sperm and egg, be this very woman? One can imagine, given the woman, that various things in her life could have changed … But what is harder to imagine is being born of different parents. It seems to me that anything coming from a different origin would not be this object. Kripke (1980: 113)

212  Duško Prelević Beside presupposing the validity of the Essentiality of Kind (“one can ­imagine, given the woman, that various things in her life could have changed”), it is noticeable in the passage above that the possibilities Kripke addresses are anchored in the actual world (cf. Mackie 2006: §6.2), which means that, within the two-dimensional semantic framework, the Essentiality of Origin is within the scope of the secondary intension. It is also likely that, as Penelope Mackie has noticed, the very principle is premised on some other theses as well, such as the branching model of de re possibilities together with “the overlap requirement”, according to which objects cannot have alternative histories completely different from the ones they actually have (see Mackie 2006: §§6.4‒6.5 for more details). But now, if we assume all these principles in Kripke’s example quoted above (together with the necessity of identity that Kripke independently endorses), then a scenario in which the Queen has a different origin from the one she actually has turns out to be not just (as Penelope Mackie notices) metaphysically impossible (Mackie 2006: §6.6),18 but secondarily inconceivable as well. If these considerations are correct, then modal rationalists can cope with Roca-Royes’s challenge. It is explicable, without begging the question, why scenarios (or situations) that falsify the Essentiality of Origin and the Essentiality of Kind are inconceivable. This makes the search for alternative accounts of our de re modal knowledge superfluous. Modal rationalism outlined in this chapter provides a viable solution once it is made explicit which theories of objects are tacitly assumed in using conceivability methods.19

Notes 1 This is what Chalmers (2002) calls “weak modal rationalism”. He also endorses “strong modal rationalism”, according to which ideal negative primary conceivability entails primary possibility (where a scenario is counted as negatively conceivable when it cannot be ruled out a priori), and “pure modal rationalism”, according to which ideal positive primary conceivability, ideal negative primary conceivability and primary possibility are coextensive. The existence of inscrutable truths, according to Chalmers (2002), would undermine the latter two versions of modal rationalism, leaving weak modal rationalism intact. Elsewhere (Prelević 2021), I have argued that Chalmers’s way of dealing with the problem of inscrutability in higher set theory (Chalmers 2002: 180, 2012: 261‒264) is not satisfactory. For that reason, I opt for weak modal rationalism. 2 According to Chalmers, primary intension is a function from counteractual worlds to extensions, while secondary intension is a function from counterfactual worlds to extensions. Chalmers endorses the epistemic version of two-dimensional semantics, according to which apriority and necessary primary intensions are coextensive (see Chalmers 2006 for more details). 3 In view of that, it should be stressed that the explanatory power of modal rationalism would not be decreased had it turned out that some modal truths can be known both a priori and a posteriori (this would be the case of what Albert Casullo calls “epistemic overdetermination”; Casullo 2010: 346). For this case would be parallel with a common situation in mathematics,

The Explanatory Power of Modal Rationalism  213 in which we consider mathematical theorems a priori because we can prove them ­i ndependently of whether they could be justified a posteriori too. 4 In philosophy of mind, in particular. Elsewhere, I have assessed in detail the interconnectedness between Chalmers’s modal epistemology and the zombie argument (see, for example, Prelević 2013, 2015b, 2017). Also, I have briefly assessed the problem of the Essentiality of Origin for artifacts in Prelević 2013: §4.3.2. In this paper, novel arguments for modal rationalism with respect to its explanatory power are presented. 5 Besides, inscrutable truths are potential counterexamples to A Priori Scrutability, which is another reason why I am not ready to accept Chalmers’s response above (see endnote 1). 6 I have used this common example in Prelević (2020: 349‒351) in order to show that modal rationalism is compatible with pluralism in geometry (and logical pluralism as well) independently of whether modal monism is true or not. This passage and next two passages are adapted from that paper. 7 It is likely that the same holds, mutatis mutandis, for analyticity and essentiality. As Christopher Kirwan (1971: 49) has pointed out, there is a vagueness at the boundary between the analytic and the synthetic (and the essential and the non-essential), although there are many clear cases as well (see Searle 1969: §1.2). I think it is more proper to say that ambiguity is present here rather than vagueness, for whether a statement is, say, analytically true or not depends on which axiomatization has previously been accepted. This, again, can be represented in S5 system for modal logic. As for essential properties, pluralism in geometry can be recalled again. For example, it is well known that equilateral right-triangles, which are impossible in Euclidean geometry, exist in Riemannian geometry; a Lambert quadrilateral (a figure with three right angles and one acute angle) exists in hyperbolic geometry, whereas it is impossible in Euclidean geometry; and so on.   Joseph Almog has made a similar point when addressing the debate between Descartes and Arnauld (Almog 2002: 27−8, 37; cf. Vaidya 2017, ­section 6.2.). I would like to thank Anand Vaidya for informing me about this. 8 Of course, some philosophers deny this (see, for example, Sidelle 2009) to the effect that the problem of essential properties of objects cannot be even posed. Given that such anti-realism about ordinary objects is in accordance with modal rationalism, it will not be addressed in due course. 9 Elsewhere, I have presented these views of material objects in more detail for the purposes of addressing issues on personal identity (Prelević 2022: 23‒27). 10 Here, it is safer to speak (in line with Forbes) about models of the substratum theory rather than the substratum theory as such, given that some philosophers, such as Jiri Benovsky (2008), claim that essentialist interpretations of the substratum theory are available as well. Benovsky’s arguments are criticized in Morganti (2009). Entering into details of this debate goes beyond the scope of this paper. 11 Again, Benovsky (2008: 182) argues that the bundle theory as such does not entail pan-essentialism since we can understand compresence relation so that it bundles different properties at different times. Nonetheless, it can be modeled (and philosophers typically understand it) in that way (see, for example, Loux 2006: 97). 12 Here, it is worth recalling that Chalmers (2011) counts the disputes over the question “What is it?” as merely verbal, while some other philosophers understand them as the cases of “metalinguistic negotiations” that concern pragmatic disagreements over appropriate uses of terms in dispute (see Plunkett and Sundell 2013; cf. Thomasson 2016 for more details). These two options are by all means in accordance with modal rationalism outlined above.

214  Duško Prelević 13 Although the bundle theory can be spelled out independently of Leibniz’s principle (see Barker and Jago 2018 for more details), the pan-essentialist version of this theory, scrutinized in this section, relies upon it. 14 Of note, if one argues that the problem of choosing adequate conception of concrete particulars is not substantial, this, even if true, would not be a way out for modal empiricist critique of modal rationalism, given that modal rationalism itself is compatible with anti-realist views about modality (see the first section). Relatedly, I can stay neutral here on the question of which theory of concrete particulars ought to be accepted. Also, it is worth mentioning that even if Roca-Royes’s criticism to conceivability accounts remains neutral on the very problem as well (as indicated in Roca-Royes 2011a: 30, this volume), the whole case would still be parallel with that of pluralism in geometry depicted above, which itself is in accordance with a modal rationalist approach (cf. Prelević 2020). 15 Eric Olson has recently stressed this point relating to organisms and biological individuality. Here is what he says: “Philosophers of biology have assumed a number of claims about the ontology of material things, often without realising it” (Olson 2021: 69). When it comes to non-modally loaded concepts, it is worth recalling that although, for example, mathematical formulas (such as 7+5=12) are, strictly speaking, non-modal, we typically hold them a priori and mathematically (and metaphysically) necessary (Williamson 2016: 464‒465). Likewise with the main accounts of physical objects sketched above, such as Aristotle’s hylomorphism: the fact that it is commonly expressed in non-modal terms (like in the sentence “Objects are compounds of matter and form”) does not prevent it, at the end of the day, from being modally loaded, given that, according to this view (cf. Cohen 1972), it is both inconceivable and metaphysically impossible for an actually existing concrete particular to persist amid substantial change. 16 For similarities and differences between Aristotle’s and Kripke’s views about essence, see Witt (1989: footnote 17, 189‒191). 17 Roca-Royes herself says that she thought earlier on that the Essentiality of Origin entails the Essentiality of Kind (Roca-Royes 2011b: endnote 10). 18 Namely, had the Queen had a different origin, she could have had a completely different history from the actual one, which would violate the overlap requirement (Mackie 2006: 111). Mackie assesses the plausibility of underlying assumptions mentioned above, and casts doubt on the overlap requirement. However, my aim in this paper is not to assess those assumptions as such, but rather to emphasize that the conceivability of a scenario in which the Essentiality of Origin is false depends on whether those assumptions have already been accepted or not. Mackie (2006: 110) also thinks that accepting the assumptions listed above would render the Essentiality of Origin metaphysically necessary; yet, it would still be epistemically possible that the very principle is false. This, however, does not apply to ideal (primary or secondary) conceivability since it can be understood, as we have seen in the first section, as a sort of objective modality rather than merely epistemic (subjective) modality. 19 An embryonic version of this paper was presented at the “Modal Epistemology and Metaphysics” conference in Belgrade (18‒20 September 2014). I would like to thank Bob Hale, Sonia Roca-Royes and Anand Vaidya for stimulating discussion of these matters. This research was supported by the Ministry of Education, Science and Technological Development of the Republic of Serbia (project: Logico-epistemological bases of science and metaphysics; No. 179067).

The Explanatory Power of Modal Rationalism  215

References Almog, Joseph. (2002). What Am I? Descartes and the Mind-Body Problem. Oxford: Oxford University Press. Alvargonzáles, David. (2016). The Constitution of the Human Embryo as ­Substantial Change. The Journal of Medicine and Philosophy 41: 172‒191. Aristotle. (1984a). On Generation and Corruption. In: J. Barnes (Ed.), The ­Complete Works of Aristotle (the Revised Oxford Translation), vol. 1. Princeton: Princeton University Press. Aristotle. (1984b). Metaphysics. In: J. Barnes (Ed.), The Complete Works of Aristotle (the Revised Oxford Translation), vol. 2. Princeton: Princeton ­ ­University Press. Ayer, Alfred. (1982). Philosophy in the Twentieth Century. New York: Vintage Books. Barker, Stephen and Jago, Mark. (2018). Material Objects and Essential Bundle Theory. Philosophical Studies 175: 2969‒2986. Benovsky, Jiri. (2008). The Bundle Theory and the Substratum Theory: Deadly Enemies or Twin Brothers? Philosophical Studies 141: 175‒190. Bird, Alexander. (2004). Kuhn on Reference and Essence. Philosophia Scientiae 8: 59‒92. Black, Max. (1952). The Identity of Indiscernibles. Mind 61: 153‒164. Brody, Baruch. (1973). Why Settle for Everything Less than Good ­Old-Fashioned Aristotelian Essentialism. Nous 7: 351‒365. Cameron, Ross. (2005). A Note on Kripke’s Footnote 56 Argument for the Essentiality of Origin. Ratio 18: 262–275. Carnap, Rudolf. (1947). Meaning and Necessity. Chicago: The University of ­Chicago Press. Casullo, Albert. (2010). Knowledge and Modality. Synthese 172: 341‒359. Chalmers, David. (1996). The Conscious Mind: In Search of a Fundamental ­T heory. Oxford: Oxford University Press. Chalmers, David. (1999). Materialism and the Metaphysics of Modality. ­Philosophy and Phenomenological Research 59: 473–496. Chalmers, David. (2002). Does Conceivability Entail Possibility? In: T. S. Gendler and J. Hawthorne (Eds.). Conceivability and Possibility. Oxford: ­ Oxford University Press. Chalmers, David. (2006). The Foundations of Two-Dimensional Semantics. In: M. García-Carpintero, and J. Macià, Josep (Eds.) Two-Dimensional Semantics: Foundations and Applications. Oxford: Clarendon Press. Chalmers, David. (2010). The Two-Dimensional Argument against Materialism. In: D. Chalmers (Ed.), The Character of Consciousness. Oxford: Oxford University Press. Chalmers, David. (2011). Verbal Disputes. Philosophical Review 120: 515–566. Chalmers, David. (2012). Constructing the World. Oxford: Oxford University Press. Christensen, Ryan. (2014). Essence, Essence and Essence. Studia Philosophica Estonica 72: 72‒84. Cohen, Marc. (1972). Essentialism in Aristotle. The Review of Metaphysics 31: 387‒405.

216  Duško Prelević Dowell, Janice. (2008). Empirical Metaphysics: The Role of Intuitions about Possible Cases in Philosophy. Philosophical Studies 140: 19‒46. Dupré, John. (1993). The Disorder of Things: Metaphysical Foundations of the Disunity of Science. Cambridge, MA and London: Harvard University Press. Fine, Kit. (1994). Essence and Modality. Philosophical Perspectives 8: 1–16. Fischer, Bob and Leon, Felipe (Eds.). (2017). Modal Epistemology after Rationalism. Cham: Springer. Forbes, Graeme. (1997). Essentialism. In: B. Hale and C. Wright (Eds.), A Companion to the Philosophy of Language. Oxford: Blackwell. French, Peter and Redhead, Michael. (1988). Quantum Physics and the Identity of Indiscernibles. British Journal for the Philosophy of Science 39: 233‒246. Hale, Bob. (2013). Necessary Beings: An Essay on Ontology, Modality & the ­Relation between Them. Oxford: Oxford University Press. Howard, Don. (2005). Albert Einstein as a Philosopher of Science. Physics Today 58: 34‒40. Kirwan, Christopher. (1971). How Strong Are the Objections to Essence? Philosophical Studies 71: 48‒59. Kripke, Saul. (1980). Naming and Necessity. Cambridge, MA: Harvard University Press. LaPorte, Joseph. (2004). Natural Kinds and Conceptual Change. Cambridge: Cambridge University Press. Loux, Michael. (2006). Metaphysics: A Contemporary Introduction (3rd edition). New York and London: Routledge. Lowe, Jonathan. (2012). What Is the Source of Our Knowledge of Modal Truths? Mind 121: 919‒950. Mackie, Penelope. (2006). How Things Might Have Been: Individuals, Kinds, and Essential Properties. Oxford: Clarendon Press. McGinn, Colin. (1976). On the Necessity of Origin. The Journal of Philosophy 73: 127‒135. Morganti, Matteo. (2009). Are the Bundle Theory and the Substratum Theory Really Twin Brothers? Axiomathes 19: 73‒85. Nichols, Peter. (2010). Substance Concepts and Personal Identity. Philosophical Studies 150: 255‒270. Olson, Eric. (2021). What Is the Problem of Biological Individuality? In: A. Sophie Meincke and J. Dupré (Eds.), Biological Identity: Perspectives from Metaphysics and the Philosophy of Biology, pp. 63‒85. London and New York: Routledge. Plunkett, David and Sundell, Tim. (2013). Disagreement and the Semantics of Normative and Evaluative Terms. Philosopher’s Imprint 13: 1‒37. Piccinini, Gualtiero. (2017). Access Denied to Zombies. Topoi 36: 81‒93. Pradeu, Thomas. (2012). The Limits of the Self: Immunology and Biological Identity. Oxford: Oxford University Press. Prelević, Duško. (2013). Modalna epistemologija i eksplanatorni jaz: Značaj argumenta na osnovu zamislivosti zombija (PhD Thesis). Belgrade: University of Belgrade. Prelević, Duško. (2015a). Modal Empiricism and Knowledge of De Re Possibilities: A Critique of Roca-Royes’ Account. Organon F 22: 488–498. Prelević, Duško. (2015b). Zombies Slap Back: Why the Anti-Zombie Parody Does Not Work. Disputatio 40: 25–43.

The Explanatory Power of Modal Rationalism  217 Prelević, Duško. (2017). Access Granted to Zombies. Theoria 60(1): 58‒68. Prelević, Duško. (2020). Modal Rationalism, Logical Pluralism, and the ­Metaphysical Foundation of Logic. In: A. Costa-Leite (Ed.), Abstract Consequence and Logics: Essays in Honor of Edélcio G. de Souza. London: College Publications (Tributes). Prelević, Duško. (2021). The Chalmers Trilemma Re-Examined. Journal of Philosophical Research 46: 345−361. Prelević, Duško. (2022). Lični identitet i problem svesti. Beograd: Srpsko filozofsko društvo. Roca-Royes, Sonia. (2011a). Conceivability and De Re Modal Knowledge. Nous 45: 22–49. Roca-Royes, Sonia. (2011b). Essential Properties and Individual Essences. Philosophy Compass 6(1): 65‒77. Roca-Royes, Sonia. (2019). Concepts and the Epistemology of Essence. Dialectica 73: 3−29. Roca-Royes, Sonia. (this volume). Conceivability: Still Not Enough: A Response to Prelević. Ryle, Gilbert. (2009). Philosophical Arguments. In: G. Ryle (Ed.), Collected Papers, vol. 2. London and New York: Routledge. Searle, John. (1969). Speech Acts: An Essay in the Philosophy of Language. ­Cambridge: Cambridge University Press. Sidelle, Alan. (1989). Necessity, Essence and Individuation: A Defense of Conventionalism. Ithaca, NY: Cornell University Press. Sidelle, Alan. (2009). Conventionalism and the Contingency of Conventions. Nous 43: 224‒241. Sober, Elliot. (1980). Evolution, Population Thinking, and Essentialism. Philosophy of Science 47: 350‒383. Sprigge, Timothy. (1962). Internal and External Properties. Mind 71: 197‒212. Thomasson, Amie. (2016). Metaphysical Disputes and Metalinguistic Negotiation. Analytic Philosophy 57: 1‒28. Tahko, Tuomas. (2018). The Epistemology of Essence. In: A. Carruth, A. Gibb and J. Heil (Eds.), Ontology, Modality, Mind: Themes from E. J. Lowe. Oxford: Oxford University Press. Vaidya, Anand. (2008). Modal Rationalism and Modal Monism. Erkenntnis 68: 191–212. Vaidya, Anand. (2010). Understanding and Essence. Philosophia 38: 811–833. Vaidya, Anand. (2017). Modal Knowledge: Beyond Rationalism and Empiricism. In: B. Fischer and F. Leon (Eds.), Modal Epistemology after Rationalism, pp. 85–114. Cham: Springer. Vaidya, Anand and Wallner, Michael. (2021). The Epistemology of Modality and the Problem of Modal Epistemic Friction. Synthese 198 (Suppl 8): 1909–1935. van Fraassen, Bas. (2002). The Empirical Stance. New Haven and London: Yale University Press. Williamson, Timothy. (2016). Modal Science. Canadian Journal of Philosophy 46: 453‒492. Witt, Charlotte. (1989). Substance and Essence in Aristotle: An Interpretation of Metaphysics VII-IX. Ithaca, NY: Cornell University Press.

10 Conceivability: Still Not Enough A Response to Prelević Sonia Roca-Royes

Introduction and Scope Prelević’s paper (this volume) is a thorough piece which presses me to clarify the argument in my 2011 paper ‘Conceivability and de re modal knowledge’. That paper deals with different types of conceivability-based epistemologies one finds in the literature, and constructs a dilemma for them (Roca-Royes 2011, §6), placing epistemic accounts on one horn and non-epistemic accounts on the other.1 The scope of this response however will be narrower, to match that of Prelević’s paper. Conceivability accounts can vary substantially from one another, but a common feature of them (as per Roca-Royes 2011, §5 and §7) is that, regardless of the specific conceivability notion at hand, the conceivability of p ultimately depends on the absence (or the unawareness) of a contradiction; and its inconceivability ultimately depends on the presence (or appearance as) of a contradiction. This is widely agreed upon, and it is something that Prelević captures too with remarks such as: ‘ideal conceivability presupposes checking the coherency of the very scenario (or situation) upon greater and greater degrees of reflection’ (this volume, 203). Beyond this common feature, the differences among the variety of accounts will manifest in (among other things) what they take this contradiction to be. For instance, in the case of Williamson (and arguably Yablo), the conceivability of p depends on counterfactual reasoning not yielding a contradiction when developing (counterfactually) the supposition that p. Something similar applies in the case of Chalmers’ ideal conceivability, whether primary or secondary. The disanalogy between them lies in that, for Chalmers (unlike for Williamson or Yablo), the nature of the concepts involved in p is sufficient for the existence (or non-existence) of a contradiction, be that at the level of primary or secondary intensions. This being so, for Chalmers (but not for Williamson or Yablo) the contradiction is of a conceptual sort (primary or secondary). This difference however, I argued, creates no difference with respect to their prospects qua conceivability accounts. The relevant difference is, instead, as suggested above, that between epistemic and non-epistemic accounts. And, given Prelević’s focus, I shall restrict myself in this response to the case of non-epistemic accounts. DOI: 10.4324/9781003002192-12

Conceivability: Still Not Enough: A Response to Prelević  219 The following explains my choice. Prelević’s ultimate goal in objecting to my argument is to mount a defence of the type of modal rationalism that Chalmers argues for. Because of this goal, Prelević engages with that part of my argument that applies to Chalmers’ conceivability account and, for brevity, I have restricted myself to that part too. But the epistemic ­status— as a rationalist or non-rationalist—of the accounts that are affected by that part of the argument is again neither here nor there. Instead, as suggested, it is the account’s non-epistemic nature which is decisive. It being non-epistemic means that the account delivers a notion of conceivability that is co-extensional with that of possibility (Roca-Royes 2011, §2). In particular, it differs from epistemic notions in that ­non-epistemic conceivability is absolute, rather than relative to subjects, as is the case with Yablo’s notions of conceivability. As such, of the several conceivability notions that Chalmers distinguishes, the ones under scrutiny here are the ideal ones, as these are the ones that guarantee the co-extensionality with possibility (primary or secondary). But, as anticipated, this co-extensionality feature is shared by accounts whose classification as rationalist is to be resisted. Salient among these is Williamson’s (2007). It is a well-known aspect of Williamson’s account that it endorses (and rests on) this logical equivalence: ◊p ≡ ¬(p □→ ⊥). For Williamson, in line with the rest of the literature, the right-hand side of this equivalence marks the (non-epistemic) conceivability of p (Williamson 2007, 163). On these bases, Williamson’s account is affected by the same horn of the dilemma, and same part of argument, as Chalmers’. And for the purposes of tailoring my response to Prelević’s pressing remarks, it will be helpful to focus on the class of non-epistemic conceivability accounts as a whole. I have structured the response as follows. I start (in the next section) by summarising my 2011-case for the claim that, when it comes to modality de re, non-epistemic conceivability is, per se, at most a trivial guide to possibility. Following this (in ‘Exegetical reconstruction of Prelević’s rescuing strategy’), I summarise Prelević’s criticism against my case, which contains a penetrating insight that I should take into account. And in the final section, ‘Still unexplained friction’, I show how Prelević’s insights provide nonetheless an opportunity for clarification more than they force a revision of my target claim in 2011.

Non-Epistemic Conceivability and De Re Modal Knowledge The Problem, in a Nutshell Regardless of their status as a rationalist or not, what creates the problem for non-epistemic conceivability accounts is that the cognitive

220  Sonia Roca-Royes capacities to assess the conceivability (or inconceivability) of p involve a capacity for essentialist knowledge; whether this capacity is exercised at a personal level, a sub-personal one, or else simply assumed to have been exercised somehow. The problem boils down to the fact that this sensitivity to (the essentiality of) essentialist facts cannot be, in turn, explained in conceivability terms and, because of this, such accounts offer at best an incomplete account of our modal knowledge. More specifically, if such sensitivity to essentialist facts is not explained, the conceivability account risks trivialisation. For in that case, what we are being offered is an explanation of how the reasoner’s sensitivity to essentialist facts can reliably guide their possibility-judgements. But this falls critically short of the sought-after explanation: the sought-­ after explanation takes our sensitivity to essentialist facts as part of the explananda, not the explanans. Vaidya and Wallner (2021) have identified this problem for ­non-epistemic conceivability accounts as one instance of a more general problem that they label ‘The problem of modal epistemic friction’. In their own terms, ‘conceivability cannot be a reliable guide to what is really possible, if it is based on ignorance of what things are’ (S1915, their emphasis). In what follows, I shall unfold exactly how this reliability is indeed achieved, in non-epistemic accounts, at precisely the price of essentialist cognisance. The Problem, Unfolded: Secondary Intensions Are to Be Held Fixed I shall unfold the case by means of examples involving de re modality, as this is the modality that lies at the centre of the current exchange with Prelević. (Because of this focus on de re modality, bear in mind that whenever thinking of the problem as it manifests in Chalmers’ case, it will be secondary (ideal) conceivability, rather than primary, that will be the salient notion. I will not always stress this in what follows.) Water is H2O and, let us grant, it cannot be H2O2. On non-epistemic accounts, as per the equivalence above, this impossibility means that ‘water is H2O2’ is inconceivable: the counterfactual supposition that water is H2O2 (plus counterfactual reasoning developing the supposition) would deliver a contradiction: (1) water is H2O2 □→ ⊥ For Williamson as well as for Chalmers, the truth of (1) marks the inconceivability of ‘water is H2O2’; secondary (ideal) inconceivability in the case of Chalmers. Similarly, Socrates is a human being and, assuming for the sake of argument the essentiality of kind, he cannot be a dog. Thus, developing the counterfactual supposition that Socrates is a dog will yield a

Conceivability: Still Not Enough: A Response to Prelević  221 contradiction, and (2) marks the (secondary) inconceivability of ‘Socrates is a dog’: (2) Socrates is a dog □→ ⊥ Chalmers and Williamson would provide different answers to the questions ‘what grounds the truths of (1) and (2)?’ and ‘how do we know them?’. In fact, we know more about Chalmers’ answer to the first question (than to the second) and we know more about Williamson’s answer to the second question (than to the first). For Chalmers, (1) is true because the secondary intension of ‘water’ is H2O and, as such, the secondary intension of the antecedent of (1) is H2O is H2O2, from where the contradiction that H2O is not H2O follows. Similarly for the case of (2): for the property of being human would be involved in the secondary intension of ‘Socrates’. For Williamson, we know about the truths of (1) and (2) because (regardless of any considerations about the nature of concepts), ‘part of the general way we develop counterfactual suppositions is to hold such constitutive facts fixed’ (Williamson 2007, 164). As such, when developing the counterfactual supposition that water is hydrogen peroxide, we will not drop (we will not imagine away) the fact that water is H2O, thus arriving at the contradiction that makes (1) true; namely, that H2O is not H2O. And similarly for the case of (2): in assessing it, we would hold fixed that Socrates is human, thus arriving at the contradiction that Socrates is and is not human. It is not difficult to envision from this how Chalmers can answer the second question too: we know the truths of (1) and (2) by knowing the secondary intensions of the concepts involved in their antecedents. To put it schematically: Chalmers (1) water is H2O2 □→ ⊥ In virtue of H2O being the secondary intension of ‘water’, we get: H2O is not H 2O (2) Socrates is a dog □→ ⊥ In virtue of humanity being part of the secondary intension of ‘Socrates’, we get: Socrates is and is not human

Williamson (1) water is H2O2 □→ ⊥ by means of holding fixed the fact that water is H2O, we arrive at: H2O is not H 2O (2) Socrates is a dog □→ ⊥ by means of holding fixed the fact that Socrates is human, we arrive at: Socrates is and is not human

With this, we’ve got the seeds we need to see how the problem of modal epistemic friction manifests in their cases. To see it, the contrast with conceivable contents is instructive (as developed in Roca-Royes (2011, §5)). Indeed, a whole battery of other contents, equally involving de re modality, are nonetheless conceivable, even by the lights of Williamson and Chalmers. For instance, the water in my bottle is cold, and Socrates was a philosopher. But that this water is hot and that Socrates is a

222  Sonia Roca-Royes shoemaker are nonetheless conceivable, and the truths of (3) and (4) mark such conceivability: (3) ¬ (that water is hot □→ ⊥) (4) ¬ (Socrates is a shoemaker □→ ⊥) We can figure out what answers would be given to the questions: what grounds the truths of (3) and (4)? And: how do we know them? In Chalmers’ terms, it’s not part of the secondary intension of ‘that water’ that it is cold, and it’s not part of the secondary intention of ‘Socrates’ that he’s a philosopher. Similarly, in Williamson’s terms, it would be part of the way we assess counterfactuals that we do not hold fixed non-constitute facts when they contradict the antecedent of the counterfactual at hand. Because of this, neither we get, nor do we arrive at, a contradiction. Now, on the basis of the co-extensionality between conceivability and possibility which is inherent, as we have seen, to non-epistemic accounts, one can declare conceivability to be a guide to possibility. This is how Chalmers does this: ‘[o]ne argues that some state of affairs is conceivable, and from there one concludes that this state of affairs is possible’ (2002, 146). And these are Williamson’s terms: ‘we assert ◊A when our counterfactual development of the supposition A does not robustly yield a contradiction’ (Williamson 2006, 165). How good a guide conceivability is to possibility (or inconceivability to impossibility), therefore, depends on the modal reasoner’s level of sensitivity to conceivability facts (or else to inconceivability facts). This being so, the thesis that conceivability is a guide to possibility (or inconceivability to impossibility) thus brings with it a commitment that us, modal reasoners, are at least reliably good at tracking conceivability and inconceivability facts. At this point, the contrast between (1) and (2), on the one hand, and (3) and (4), on the other, serves us to illustrate that the cognitive capacities responsible for such reliability-success include a capacity to tell apart the essential facts from the non-essential ones. For the facts that water is H2O, that Socrates is human, that water is cold, and that Socrates is a philosopher, all appear innocently non-modal enough. What is not so modally innocent is a theory’s commitment either to the claim that, of them four, only the first two have (somehow!) made it into secondary intensions of concepts, or else to the claim that only the first two are (somehow!) being held fixed by counterfactual reasoners, even when those facts are contradicted by the antecedent of the counterfactual at hand. In other words, according to these theories, in being good at telling apart conceivability from inconceivability facts, we are somehow sorting out the essential from the non-essential, as argued in Roca-Royes (2011).

Conceivability: Still Not Enough: A Response to Prelević  223 This is not modally innocent because our successes as modal reasoners are critically dependent, in these non-epistemic conceivability accounts, on us having this capacity for essentialist knowledge. As the above s­ uggests, different conceivability theories will vary in respect to at which point (and in what form) they incur a commitment to such capacity. And, although I can’t unfold it further here, the above suffices to see that, in the case of Chalmers, it is not essentially neutral to consider x to be a given concept C’s secondary intension; and, in the case of Williamson, it is not essentially neutral to hold x fixed (even) when it contradicts the antecedent of the counterfactual at hand. These are the places where their theories build in a non-explained capacity for essentialist knowledge as that which constrains our conceivability judgements. And two things are to be noted about this capacity: first, by the nature of the case, it cannot be explained by conceivability; second, it is part of our explananda. Consequently, the target claim of my 2011 paper is achieved: ‘conceivability cannot be the whole story in de re modal epistemology’ (2011, 38). The complaint is thus one about incompleteness. This summarises my case against non-epistemic accounts developed in Roca-Royes (2011, §6.2 and §7).

Exegetical Reconstruction of Prelević’s Rescuing Strategy Prelević rightly notes that Chalmers’ own response to my line of criticism (in Chalmers 2012) is tinted with (Carnapian-inspired) modal anti-realism. This is an aspect of the view that I was overlooking in my paper (2011). In effect, the question I assessed there, and that is still the one I’m most interested in, is the following: on the assumption of (mind-independent) realism, can Chalmers’ (or anyone else’s) non-epistemic conceivability be our route to modal knowledge? And my answer, as per the above, is: only to the extent that essentialist knowledge (or a capacity for it) is somewhere operative in the conceiver, as that which makes them reliable with their conceivability exercises. This being so, the recommendation to the realist about modality is not to endorse a conceivability-based epistemology of modality, unless they are prepared to provide us with what they would owe us as a result: an explanation of our capacity to discriminate the essential facts from the non-essential. Modal anti-realism is an aspect of Chalmers’ view that Prelević is ready not to appeal to when tailoring his response to my criticism. Therefore, he has kindly put us both on the same page, so that we can focus on our disagreements past the realism/anti-realism debate. Against this dialectic, he has taken on the task of showing that my recommendation to the modal realist is premature: non-epistemic conceivability notions can still support an attractive, workable epistemology, even for the realist. I shall here reconstruct what I take to be the core of Prelević’s argument to this effect.

224  Sonia Roca-Royes The Running Example and Its Impact: Relative vs. Absolute Conceivability Prelević bases his rescuing strategy on the following case, which he uses as his running example: there is more than one consistent system of geometry to the effect that what is a theorem in one system need not be a theorem in another one. Some claims are, of course, true in all systems of geometry, such as the claim that a triangle is a figure with three edges and three vertices. Prelević (this volume, 204) But with Euclidean geometry as our default operative conception, we would be induced to deem it inconceivable (thus impossible) that the sum of the angles of a triangle be other than 180°. And yet, we know that that sum is less in Hyperbolic geometries and more in Spherical ones. On these bases, Prelević recommends the distinction between absolute and relative conceivabilities. Contents such as that triangles have four edges would thus be absolutely inconceivable (and ­impossible), whereas contents such as that the sum of the angles of triangles are less than 180° are not absolutely so. Instead, this content is only inconceivable (i.e., contradictory) relative to Euclidean and Spherical geometries, and it’s conceivable (coherent) relative to Hyperbolic geometry. How best to spell out this distinction between absolute and relative conceivabilities is something I’d like to see more about. In particular, it is important that we’re offered clarification as to how this distinction differs from the primary vs. secondary one.2 I will return to this question in my response (in the section ‘Still Unexplained Friction’). For now, to sum up at surface level the running example, we would have three subsets of worlds, each corresponding to the class of worlds that satisfy one of the three different geometries: Euclidean, Hyperbolic, and Spherical; and, Prelević suggests, our world belongs to only one of them, depending on how our space turns out to be. With this example in place, Prelević envisions a strategy to rescue conceivability accounts that I reconstruct as a three-step argument. I state the three main claims from the outset, and I unfold them one by one afterwards. First, on the basis of his running example, he reasons by analogy to reach the conclusion that (1) what is and what is not conceivable—in a sense of conceivability relevant when wondering about de re modal facts—depends on what is the underlying conception of object that is operative when carrying out the conceivability exercise; and he illustrates this by distinguishing three rival conceptions: the substratum, the bundle, and the Aristotelian ones. To this conclusion, he adds the claim that, (2) in the philosophical context to which the current exchange belongs, it is the Aristotelian conception that is operative. Finally, (3) with that conception of object, conceivability delivers the right modal results. Let me now take these claims, in turn.

Conceivability: Still Not Enough: A Response to Prelević  225 Analogy: Conceptions of Objecthood Likewise in geometry and logic, axiomatization in metaphysics is possible as well. Thus, it is reasonable to suppose that something might be conceivable with respect to one system of metaphysical axioms (or a set of beliefs) and inconceivable relative to some other system of metaphysical axioms (or a set of beliefs). Prelević (this volume, 205) Prelević fleshes out this suggestion with three competing conceptions of objecthood. First, the substratum theory, in considering objects to be bare particulars that have properties but are not themselves properties, is taken by Prelević to correlate with anti-essentialism: any loss or replacement of properties would be compatible with the persistence of the object. Second, the bundle theory, in identifying objects with a mere bundle of properties (and no underlying bare particular), would correspond to pan-essentialism: the loss of even one property would result in a different bundle and thus no object would survive losing or gaining any property. And finally, Aristotelian hylomorphism, in ­c onsidering objects to be composites of matter and form, would correlate with a middle-point position (which we can call ‘non-trivial essentialism’) according to which objects survive some changes but not all. Developing the analogy to the geometric example, we would have the total space of worlds, TSoW, partitioned into different equivalence classes, with worlds falling (uniquely) into one or another depending on whether their objects satisfy the substratum, the bundle, or the Aristotelian metaphysical axioms; and which class our world belongs to depends on what objects really are (in the actual world). Again, it is an important question here too how exactly to understand the distinction between absolute and relative conceivabilities, which I postpone until the last section (the one titled ‘Still Unexplained Friction’). But to wrap up the analogy, at least at surface level, we might assume that it would be absolutely inconceivable that Socrates lacks any spatial extension, but it would be only relatively inconceivable that he is a dog, as this content would be conceivable relative to the substratum theory (axiomatisation) of objecthood. Living and Thinking in an Aristotelian World Now, in an attempt to situate the context of the current exchange, Prelević notes that It is likely that neo-Aristotelians in modal metaphysics (including Roca-Royes herself) prefer an Aristotelian view of concrete particulars over the bundle theory and the substratum theory. Prelević (this volume, 206)

226  Sonia Roca-Royes I think that my criticism of conceivability accounts is neutral on this matter. Indeed, the conditional nature of the main claim defended in 2011 must not be overlooked: The conclusion I am aiming at is that the suggested conceivability method cannot establish essentialist truths and, generalizing from (EO), I will argue for this by showing that, if there are any such truths, we could establish them neither with epistemic nor with non-epistemic notions of conceivability. Roca-Royes (2011, 30) I grant however that neo-Aristotelianism—if by this we mean the n ­ on-trivial essentialism it can be taken to correlate with—is more often than not a background assumption in debates on the epistemology of modality. Now, I’m not clear whether Prelević’s focus on the Aristotelian conception amounts to the endorsement of the view, whether it is (less committingly) the assumption of its truth for the sake of argument or whether it is the assumption that we think with such a conception. As we shall see next, however, Prelević’s case (for resisting my incompleteness complaint against conceivability-based epistemologies) is the strongest when at least the last two assumptions can be exploited. Conceivability Delivers the Right Modal Results Prelević reminds us that: Many philosophers already noticed that the Aristotelian view of concrete particulars amounts to the claim that it is inconceivable that a concrete particular belongs to a kind different from the one to which it actually belongs. Prelević (this volume, 207) Since not only Essentiality of Kind but also Essentiality of Origin have historically been at the centre of metaphysical and epistemological discussions, Prelević is interested in motivating how the Aristotelian conception would render it inconceivable too that an entity has origins different from the actual ones. And yet, he’s also ready to grant that, to achieve such inconceivability, further supplementary principles might be needed too. But now, if we assume all these principles […] then a scenario in which the Queen has a different origin from the one she actually has turns out to be not just […] metaphysically impossible […] but secondarily inconceivable as well. Prelević (this volume, 212; his emphasis)

Conceivability: Still Not Enough: A Response to Prelević  227 So far, though, as far as I can see, things are harmless to my case. In fact, this is how things should be with non-epistemic conceivability accounts, which, to recall, are the focus of the current exchange. On such accounts, what is conceivable is co-extensional with what is possible. Thus, if, in line with the Aristotelian account, it is impossible for Socrates to be other than a human being, such thing is inconceivable (in the non-epistemic sense of the notion). Similarly, if, in line with the (maybe-supplemented) Aristotelian account, origins are essential, then any content contradicting this is inconceivable (in the same non-epistemic sense). Despite this harmlessness, Prelević seems to conclude his case here: If these considerations are correct, then modal rationalists can cope with Roca-Royes’s challenge. It is explicable, without begging the question, why scenarios (or situations) that falsify the Essentiality of Origin and the Essentiality of Kind are inconceivable. Prelević (this volume, 212) Now, if, in addition to this noted co-extensionality, it also happens that we think within the Aristotelian conception as well (supplemented or not), then our conceivability exercises (our assessment of counterfactuals) are aligned with the conceivability facts. I shall assume that Prelević takes it to be part of the case that this is how we think; although I don’t think he emphasises this enough. This being so, the reasoning would continue, we are getting modal facts right. And I shall assume too that, by his lights, the fact that this conception (whether with supplementary principles or not) is the operative one constraining our conceivability exercises does not trivialise the epistemic route from conceivability to possibility. And yet, this is what I’m not seeing.

Still Unexplained Friction There are two types of responses that I want to make in response to Prelević’s challenge. A bold one and a lengthier one that elaborates on his insightful analogy with the case of geometry pluralism. The bold response is that if, as presumed in Prelević’s scenario, we live in an Aristotelian world, and if we think with an Aristotelian conception, then, the Aristotelian principles are what creates the required modal epistemic friction, making us thereby reliable with our conceivability exercises. These principles play two roles: they define the extension of the correct (relative) conceivability and they also govern our conceivability exercises. Regardless of whether these principles are entirely encoded in our conceptual network, or else are the pieces of information that we hold fixed no matter what, in assessing counterfactuals, the fact that they play these two roles indicate a capacity for essentialist knowledge: our capacity to track essentialist truth. To re-state the complaint from above

228  Sonia Roca-Royes (at the end of the sub-section ‘The Problem Unfolded’): this capacity cannot be explained in terms of conceivability, so we haven’t finished explaining the explananda. The lengthier response requires me to return to an issue that I postponed above (in the ‘Exegetical Reconstruction’ section) about how best to understand Prelević’s distinction between absolute and relative conceivabilities. From the outset: I receive it as providing us with a useful way of generalising Chalmers’ distinction between primary and secondary conceivabilities, such that those who don’t endorse the ­two-dimensionalist framework (e.g., Williamson or Yablo) can still avail themselves of an analogous distinction. Let me unfold. The geometry model suggests that (in the neatest case) when we speak of competing relative conceivability notions, there’s a weaker theory, and a number of stronger theories, such that each of the stronger theories includes the weaker one, and any two of the stronger theories are incompatible. Here’s a toy example: the weaker theory would be ‘water is watery stuff’, and two stronger theories would result by adding either ‘water is H2O’ or ‘water is XYZ’. Another example: the weak theory ‘triangles have three angles’ can be turned into stronger theories by adding in either the axioms of Euclidean geometry, or those of Hyperbolic geometry, or those of Spherical geometry. This much is suggested by Prelević’s brief remarks about his distinction. If this is so, someone like Chalmers, who already has available the distinction between primary and secondary conceivabilities, might not gain much by adding Prelević’s distinction into his theoretical repertoire. For instance, it would be absolutely, but also primarily, inconceivable that a triangle has other than three angles, or that a material object lacks extension. Similarly, absolute conceivability (i.e., conceivability relative to any coherent stronger axiomatisation) seems to imply primary conceivability. Still, two notions call for special attention, and these are that of ­relative-but-not-absolute conceivability and that of relative-but-not-­ absolute inconceivability. If something, p, is relatively-but-not-absolutely conceivable (relative to, say, axioms AS1), then, had the world conformed to AS1, p would be secondarily conceivable, and there’s a competing set of axioms, AS2, such that p is inconceivable relative to them, and such that, had the world conformed to AS2, p would have been secondarily conceivable. To unfold an example: if the world is Aristotelian, then that Socrates is a dog is indeed (as Prelević suggests in his piece) secondarily inconceivable. But had the world turned out to be as bundle theory tells us, it would have been secondarily conceivable that Socrates is a dog. (Similarly, with that water is XYW, had the world turned out to be twin-Earthy.) This distinction deserves a lot more scrutiny than I can afford here. But if the above is on the right track, for the two-dimensionalist, Prelević’s distinction easily amounts (simply?) to a shortcut to think counterfactually

Conceivability: Still Not Enough: A Response to Prelević  229 about secondary intensions. For someone like Williamson or Yablo, who participate of conceivabilism without endorsing two-dimensionalism, the theoretical gain might be higher, allowing them to reason counterfactually about conceivability facts. But this shall be material for another occasion.3 For current purposes, more urgent is to note that the notions of relative-but-not-absolute conceivability/inconceivability show neatly where the epistemic risk lies for the modal reasoner as envisioned by a non-epistemic conceivability account, Williamson or Chalmers alike. If the principles that generate their modal epistemic friction (or, more aptly in this case: doxastic friction) are not those of the correct relative conceivability (i.e., if they are not the ones to which our world conforms), such modal reasoner will be led astray with her conceivability exercises/ judgements. This will manifest in the reasoner wrongly holding fixed, when reasoning counterfactually about a given object, o, truths that are not constitutive of o (or, in Chalmers’ terms: truths that are not secondary intensions); or, alternatively, it will manifest in the reasoner imagining away things that are constitutive. To be at least reliable, as stressed in Section ‘The Problem, Unfolded: Secondary Intensions Are to Be Held Fixed’, we ought to hold fixed secondary intensions (and only them). Admittedly, the core claim of a non-epistemic-conceivability epistemology of modality is that we are not (massively) led astray: ‘Gaps between secunda facie positive conceivability and ideal positive conceivability seem to be very rare’ (Chalmers 2002, 160). Their claim is thus that we do largely hold fixed secondary intensions (and only them). My complaint, given the epistemic risk, is (as was already in 2011) that an explanation is owed (but not forthcoming in conceivability terms) for that cognitive success.

Notes 1 This terminology is borrowed from Worley (2003). 2 Prelević does provide some exciting remarks to work on. If we think about the total space of worlds, TSoW, he considers S5 as satisfied, not in virtue of the accessibility relation within TSoW being the total one (each world accessible from any) but rather in virtue of that relation constituting an equivalence relation on TSoW: in particular, providing isolated subsets of worlds such that the relation is total within each subset, but no two worlds in different subsets are related by it. But are we now (contrary to Chalmers) endorsing modal dualism, given that the modal space is only one of these subsets (the one that contains our world), and it’s outrun by the (epistemic?) TSoW? More clarification is needed. 3 It should not be overlooked, in reasoning counterfactually about conceivability facts, that Williamson is committed to the vacuous truth of all counterpossibles, and this might undermine, by his lights, the interestingness of such reasoning.

230  Sonia Roca-Royes

References Chalmers, David. (2002). Does Conceivability Entail Possibility? In T. S. G ­ endler and J. Hawthorne (eds.), Conceivability and Possibility (pp. 145–200). Oxford: Oxford University Press. Prelević, Duško. (this volume). The Explanatory Power of Modal Rationalism. Roca-Royes, Sonia. (2011). Conceivability and De Re Modal Knowledge. Nous 45: 22–49. Vaidya, Anand, and Wallner, Michael. (2021). The Epistemology of Modality and the Problem of Modal Epistemic Friction. Synthese 198(S8): S1909–S1935. Williamson, Timothy. (2007). The Philosophy of Philosophy, Oxford: Blackwell Publishing. Worley, Sara. (2003). Conceivability, Possibility and Physicalism. Analysis 63: 15–23. Yablo, Stephen. (1993). Is Conceivability a Guide to Possibility. Philosophy and Phenomenological Research 53: 1–42.

11 Reviving the Modal Account of Essence1 Rebecca Chan

Introduction Socrates has many properties. Some of these are necessary, or such that he cannot lack them in any possible scenario, while others are contingent, or such that he can lack them. For example, Socrates has the properties of being short, snub-nosed, and human. While being human is a necessary property for Socrates because Socrates must be human in all possible scenarios, being short and snub-nosed are contingent properties.2 Had Socrates had the technology of the 21st century available to him, he could have undergone hormone treatment as a child or had plastic surgery, and thus had a different stature and nose. In addition to necessary and contingent properties, Socrates also has essential and inessential properties. For preliminary purposes, let’s very roughly conceive of a thing’s essence as “what it is” to be that thing. So, for instance, the essence of Socrates provides an answer to the question “what is Socrates?” or “what is it to be Socrates?”3 (Since the purpose of this paper is to arrive at a precise, correct account of essence, we’ll begin with the intentionally rough conception to start the discussion without begging the question in favor of a particular conception.) Essential properties are those that make up Socrates’s essence, while inessential, or non-essential, properties are ones that do not make up Socrates’s essence. The properties of being short and of being snub-nosed are undoubtedly inessential properties as well as contingent ones. They are properties Socrates could lack, and thus do not get at what it is to be Socrates. The same is true for all other contingent properties. In light of the fact that all contingent properties are inessential, are all inessential properties contingent? Or to put it another way, are all necessary properties essential properties? According to philosophers who favor a modal conception of essence, the answer to this question is “yes.” However, philosophers who favor a definitional conception of essence answer the question negatively. The modal account of essence rose to prominence in the decades following Quine’s anti-essentialism because it offered a simple, fully reductive story about what a thing is. But since Fine’s “Essence and Modality,” the modal

DOI: 10.4324/9781003002192-13

232  Rebecca Chan account has fallen into disrepute and, generally, is no longer regarded as a viable account of essence. On the current mainstream view, the definitional conception is superior because (i) there are counterexamples to the modal account that definitional ones avoid, and (ii) there is important metaphysical work that only definitional accounts can do. Though Fine and the defenders of the definitional accounts are right to point out flaws in the traditional modal accounts of the 1960s and 1970s, it would be hasty to completely abandon the modal conception. My goal in this paper is to revive the modal account of essence and restore it to its former glory by modifying it in a way that neutralizes the force of both (i) and (ii). First, I’ll offer two modifications that yield a revised modal account that is extensionally similar to a definitional account; this revised account thus avoids the Finean counterexamples and dispenses of (i). More significantly, if these modifications can be given in purely modal terms— which is what I aim to do—then the modal account can push back against (ii) as well. While dispensing with (ii) isn’t as clear-cut since it will still be the case that the definitional account offers theoretical benefits the modal one can’t, it will also be the case that the modal account has its own theoretical benefits to offer. I’ll highlight the theoretical tradeoffs involved, and leave it to others to settle which account comes out on top. Here’s the plan for the paper. I start by getting clear on what the competing conceptions of essence are and why the definitional conception is thought to be superior. I then offer my positive proposal on which all of a thing’s necessary properties still count as essential ones. The revision to the standard modal account will come in two stages: determinateness and sparseness. In my analysis of determinateness, a property is more (or less) determinate in proportion to how narrow (or wide) the scope of what a thing that has that property could be. A thing’s essential properties still are all and only its necessary properties, but these essential properties come in degrees corresponding to determinateness. In my analysis of sparseness, necessarily coextensive properties are the same property. This analysis of sparseness is more serviceable than other competing analyses of sparseness in the literature on essence because it is given in purely modal terms. Given these two modifications, the extension of a thing’s most essential properties by the lights of the revised account will be the extension of a thing’s properties given as essential by the definitional account. Finally, since my revised modal account is given in purely modal terms, it is fully reductive and offers two theoretical upshots, parsimony and non-arbitrariness. Before diving into things, I want to address a couple of methodological preliminaries. First, I’m going to speak as if there are individual essences. Some philosophers object to talking about, e.g., Socrates’s essential properties on the grounds that there are no individual essences, only kind essences. Nothing in this paper rests on this issue since everything said applies to kinds and individuals (assuming there are individual essences). Thus, I’m going to talk about Socrates’s essential properties since that’s an example that runs throughout the literature.

Reviving the Modal Account of Essence  233 Second, one might worry that “essence” is a technical term introduced by philosophers for the purposes of the projects with which they are associated. If this is right, then what essence ends up being is a matter of stipulation, and the “dispute” between the two conceptions is merely verbal. There are at least two reasons to reject this diagnosis of the dispute. First, the parties to the dispute do not take themselves to be engaged in a merely verbal dispute. For instance, both Plantinga and Fine are clearly responding to Quinean anti-essentialism, and Fine is clearly responding to the modal conception of essence.4 Newcomers to the dispute who propose revisions to the modal account in response to Fine also clearly take themselves to be engaged in meaningful dispute.5 Second, all parties agree that what is at issue is whether essences have a special role—namely, the answering of “what” questions—to play in metaphysics. The anti-essentialist, like Quine, will deny this, mostly because he wants to reject de re modality and holds that something else, like nature, can answer these questions.6 Those on the other side who hold that there are essences will say that essence does provide an answer to metaphysical “what” questions, and then fight amongst themselves about whether the modal or definitional conception provides the better answer. With these preliminaries out of the way, let’s turn to our two conceptions of essence.

Two Conceptions of Essence On the simplest modal conception of essence, essentiality is analyzed in terms of modality. For example, Plantinga (1974) offers this sort of account when he says that “an object x has property F essentially if and only if there is no world in which x has not-F” (60). This conception of essence is appealing because necessary properties are ones that Socrates must have, and this necessity appears particularly salient. Furthermore, many necessary properties, such as the property of being human, get at what Socrates is. For ease of reference, let us formulate the simple modal conception of essence (“SMC”) as follows: (SMC) Where F is a property, F is essential to a iff necessarily, Fa if a exists.7 Socrates’s essential properties are his necessary properties, and vice versa, since his necessary properties are the ones that he will have so long as he exists. The most significant strength of SMC is that it offers a fully reductive account in virtue of analyzing essence in terms of modality. Reduction is desirable for at least two reasons. First, as Cowling (2013) notes, ­reduction is more ideologically parsimonious than adopting new primitives. Second, as Denby (2014) argues, essence is exactly the sort of thing we would expect to be reducible. Essence doesn’t seem to be primitive,

234  Rebecca Chan and if it were, it would not be clear enough to provide a solid grasp of related notions. But despite its elegant, reductive analysis of essence, many have argued that SMC is inadequate. It is now widely believed that a conception of essence modeled on definition is superior to the modal conception. On the competing definitional conception, though all essential properties are necessary, not all necessary properties are essential. As its name suggests, the definitional conception of essence likens the essence of a thing to its definition; just as a definition concisely and informatively explains what a word means, a thing’s essence explains what that thing is in a concise, informative way. As Fine (1994) puts it, “just as we may define a word, or say what it means, so we may define an object or say what it is” (2). Crucially, defining and explaining what a thing is do not involve listing all of that thing’s necessary properties. Let us formulate the definitional conception as follows: (DC) F is essential to a if and only if F is part of the “real definition” of a. Unlike SMC, DC does not count all of a thing’s necessary properties as essential since not all of them are part of the thing’s definition. Also unlike SMC, DC does not have a fully reductive analysis since “is part of the real definition” is taken as primitive. Even if one attempts to analyze “is part of the real definition” in terms of fundamentality or metaphysical priority, that analysis will require taking fundamentality as a primitive. However, definition is more fine-grained than modality. This fine-grainedness allows for a narrower conception of essence that is thought to provide a better account of what a thing is. Returning to the case of Socrates, many of his necessary properties do not quite get at what he is. For instance, Socrates is not identical to his student Plato, and never could be. Thus, he has the necessary property of being distinct from Plato. But this property is not particularly informative with respect to what Socrates is; many other diverse things such as Aristotle and Mt. Olympus have this property as well. There are even more necessary properties that seemingly have nothing to do with what he is such as the property of being the sole member of Socrates’s singleton set and the property of being a man or a mountain. This problem is exacerbated by properties relating to necessary truths. Socrates also has the property of being such that 2 = 2. Since 2 = 2 in all possible worlds, and thus in all worlds in which Socrates exists, Socrates is always such that 2 = 2. In fact, any other necessary truth can be substituted in for “2 = 2” to yield yet another property he has necessarily. In light of cases like these, there seem to be necessary but inessential properties. Consequently, there is a case for moving toward a definitional account of essence.

Reviving the Modal Account of Essence  235

The (Alleged) Virtues of Definition The model of definition is more narrow than that of necessity. While SMC counts all of a thing’s necessary properties as essential, DC excludes some necessary properties from a thing’s essence. This exclusivity purportedly makes DC superior. This section highlights the alleged virtues of the definitional model. Seemingly Inessential but Necessary Properties Fine (1994) offers a wide range of cases in which a thing’s necessary properties do not appear to be essential. I’ll break these cases up into two broad groups. I employ these groups out of convenience—as we’ll see, they line up with the stages of my positive proposal—and not because I take the distinction to be metaphysically significant. The first group of necessary but seemingly inessential properties is that of non-discriminatory properties. Recall that Socrates necessarily has the property of being such that 2 = 2. He also necessarily has the properties of being red or not red, being self-identical, and being such that 1 + 1 = 2. All of these properties are ones that all things have necessarily. They don’t get at what Socrates is because they don’t distinguish him from any other objects. Disjunctive and negative properties can also be non-discriminatory. A paradigm case of a non-discriminatory disjunctive property is the necessary property of being P or ~P. In the case of disjunctive and negative properties, non-discrimination can come in degrees. There are marginally more discriminating properties such as being a man or a mountain, being non-mountainous, or being not-Plato. All of these are still too broad to give a good sense of what Socrates is. More generally, non-discriminatory properties are problematic in this context because they are not specific enough to get at what a thing is. A special case of non-discriminatory necessary properties, the property of existence, is worth mentioning. I keep existence separate from the other cases because it is peculiar and potentially controversial, given one’s prior commitments regarding existence. For instance, a Meinongian might hold that there are some things that do not exist, and Kantian might hold that existence is not a property. But if we set aside those sorts of views, then according to Fine (1994), counting existence as an essential property is problematic because trivially, Socrates exists in every world in which Socrates exists. He necessarily has the property of existing. There is something peculiar about the property of existing being essential to Socrates, but modal accounts of essence are committed to existence as an essential property if all necessary properties are to be essential ones. Similarly, it turns out that all things have the property of existence essentially since they all have that property in all scenarios in which they exist. Thus, existence is also non-discriminating and faces the worries that the other non-discriminatory properties do.

236  Rebecca Chan The second group of seemingly inessential necessary properties is that of necessarily coextensive and derivative properties. Necessarily coextensive properties are properties that are necessarily coinstantiated: properties P and Q are necessarily coextensive if and only if necessarily, P is instantiated when and only when Q is. For example, trilaterality and triangularity are coextensive since they are both instantiated in all and only triangles. Derivative properties are coextensive properties that exhibit a dependence relation where one property is more fundamental and the other is derived from it. (All derivative properties, as I define them, are coextensive, but I leave open the possibility that the converse does not hold.) Unlike the properties in the first group, coextensive and derivative properties can be perfectly discriminating. Socrates, for instance, has many necessary properties that are unique to him, such as the property of being Socrates and being the sole member of Socrates’s singleton set. But all of these properties are coextensive, and some of them, purportedly, are derivative. Take being Socrates and being the sole member of the singleton set containing Socrates. Necessarily, the singleton set containing Socrates exists if and only if Socrates exists. Thus, an essential property of the singleton is the property of having Socrates as its sole member. This result is intuitive. However, because Socrates only exists when the singleton exists, the modal account also yields the result that the property of being the sole member of Socrates’s singleton set is essential to Socrates. And according to champions of definitional accounts, this result is peculiar because Socrates is metaphysically prior to his singleton. As Fine (1994) puts it, “it is no part of the essence of Socrates to belong to the singleton” (6). The problem can be iterated with the singleton containing Socrates’s singleton, the singleton containing the singleton containing Socrates’s singleton, and so forth. On the definitional conception of essence, all of these derivative necessary properties, and many of the coextensive non-derivative ones are not counted as essential. Virtues of Exclusivity What are the virtues of exclusivity when it comes to qualifying as an essential property? First, there is something worrying about counting the non-discriminating and coextensive properties as essential. Imagine Pat who is taking a philosophy class for the first time and who asks the professor what it is to be Socrates. Now suppose that Pat’s professor answers by saying that it is to be red or not red, be self-identical, and be such that 1 + 1 = 2. Though Pat’s professor has said some true things about what Socrates must be, none of these properties really get at what Socrates is because they don’t distinguish Socrates from other objects. Pat would be justified in feeling as if his question has not been answered. The same rationale holds for the slightly more discriminating properties such as being distinct from Plato or being a man or a mountain. Though these

Reviving the Modal Account of Essence  237 properties are slightly more discriminating than the necessary ones, they still are not discriminating enough to be sufficiently informative. Now imagine that Pat’s professor gives a more discriminating answer and says that to be Socrates is to be the sole member of his singleton set. Again, if Pat were to feel that his question has not been answered, his feelings would be apt. Though that property is necessary, it is a round-about way of getting at what it is to be Socrates. The more narrow definitional conception of essence fits our intuitions about which properties answer the question of what it is to be Socrates. Second, essence is a useful tool for individuating amongst kinds and, potentially, individuals. In the case of kinds, the essence of a kind explains what it is for a thing to be of that kind. For instance, in Aristotle’s definition, human beings are rational animals. The combination of rationality and animality explain what it is to be human. Furthermore, the differentia of rationality distinguishes human animals from other animals. If we suppose that the properties of being rational and being an animal are the essential properties of what it is to be human, we can see that the essential properties not only tell us what humans are but also how they are different from other animals and other kinds of things. The case of individuals is more controversial, but if there are individual essences, then they would provide an answer for what it is to be a particular individual.8 Just as essence could specify what type of thing something is, it could also specify what particular it is. Third, because it is more narrow, the definitional conception of essence is a conception on which essence does important metaphysical work. For Fine, one of the upshots of the definitional conception is that essence is able to give an account of ontological dependence. For instance, with the definitional conception of essence, we can have an account where “one object depends upon another…if its essence prevents it from existing without the other object” (Fine 1994, 2). Recall the case of Socrates and the sole member of his singleton set. Plausibly, Socrates is more fundamental than this singleton. On modal conceptions, being the sole member of his singleton set is essential to Socrates and being Socrates is essential to the sole member of Socrates’s singleton set; there won’t be an asymmetry between Socrates and his singleton that captures the difference in fundamentality. On the definitional conception, being the sole member of his singleton is not essential to Socrates, so there is an asymmetry in essences that yields a difference in metaphysical dependence. Other sorts of metaphysical work attributed to the definitional conception abound. For instance, definition might be part of a story upon which essential truths explain modal ones.9 Since this paper focuses primarily on Finean counterexamples, I won’t go into detail about these issues here.10 It is, however, worth noting that this third virtue regarding metaphysical work goes beyond showing that the extension of modal essence is overly inclusive in counting certain properties as essential. Thus, the

238  Rebecca Chan attack on the modal account is two-pronged. As the first two virtues suggest, the Finean counterexamples show that there are serious problems that arise from counting all necessary properties as essential. And as the third virtue alleges, there are theoretical advantages to be gained if essence is not given in terms of modality.

The Revised Modal Account I now turn to my proposal for how the modal account ought to be revised. In a nutshell, essential properties can be sorted according to determinateness and sparseness. This proposal departs from competing revisions to the modal accounts in two significant ways. First, via determinateness, essentiality turns out to be scalar rather than binary. All of a thing’s necessary properties are its essential properties, but some are more essential than others. Second, I offer an account of sparse properties. But unlike other sparse accounts that attempt to cut down on the candidate essential qualities by disqualifying certain types of properties from being essential ones, my account cuts down on the candidate essential properties by holding that there are far fewer properties than initially thought. My account adopts a view of sparseness on which all necessarily coextensive properties are the same property. Since the analyses of determinateness and sparseness are purely modal, my revised account is also purely modal and retains the virtue of reduction that the simple modal account has. In what follows in this section, I elaborate on these two parts of my proposal and explain why the way in which they depart from other revisions to the modal accounts are advantageous. Stage 1: Determinateness In the first stage of my proposal, necessary properties are ordered according to determinacy. A property is more (or less) determined in proportion to how narrow (or wide) the scope of what a thing that has that property could be. At the most narrow extreme, a property is maximally determinate if there is only one kind (or individual) that a thing could be given that it has the property in question since the scope of what that thing could be includes just one kind (or individual). For instance, the property of being Socrates is maximally determinate since given that a thing has that property, the scope of what that thing could be—namely, Socrates—is extremely narrow. The property of being Socrates or Plato is slightly less determinate since given that a thing has that property, it the scope of what it could be—Socrates or Plato—is slightly wider. The property of being human or mountain is even wider, and thus even less determinate. At the other extreme, properties that all things have, like being such that 2 = 2, are the least determinate properties since given that a thing has that property, the scope of what that thing could be covers all things. Like the traditional modal account, this revised account

Reviving the Modal Account of Essence  239 counts all necessary properties as essential ones, and vice versa. However, unlike the traditional account, which only lists the essential properties, the revised account orders the essential properties in degrees in accordance with determinacy. The more determinate a property is, the more essential it is. Significantly, determinacy is not merely an accidental feature of a world. Imagine a world where the only things that exist are humans. In this world, all things have the property of being human as well as the property of being such that 2 = 2. Because the properties are instantiated in the same number of things, one might be tempted to think that they are equally determinate in this world. However, the property of being human is still more unique because a particular thing’s having that property limits the scope of what that thing could be to just humans. The property of being such that 2 = 2 does not limit the scope of what that thing could be whatsoever. Thinking about determinacy in terms of possibility leads to a potential worry for its analysis. To explain this worry, I’ll use possible worlds heuristics, though I’m not wedded to any particular conception of possible worlds and the force of the worry and my response do not turn on it. Given sufficiently many worlds, there may be an infinity of humans and an infinity of animals. The scope of being human and the scope of being an animal thus turn out to be rather wide—they cover an infinity of things. Furthermore, the infinities appear to be of the same cardinality. If this is right, then it appears that the two properties have the same degree of determinacy and essentiality, which is precisely what the revised account wants to avoid. Fortunately, this worry is not devastating to the account. Some philosophers have argued that infinities of the same cardinality can be compared. For example, Vallentyne and Kagan (1997) present some compelling examples that involve comparing infinite values. We might imagine two worlds that have a first time slice but no last time slice (i.e., the worlds extend infinitely into the future). Now imagine that at all times in the first world, the world has one unit of value. At all times in the second world, the world has two units of value. Adding the values at all the times for a world results in the overall value of the world. For both of these worlds, that value is infinite. Furthermore, those infinite values have the same cardinality. Nevertheless, it seems that the second world contains more value than the first. Kagan and Vallentyne propose establishing an isomorphic mapping between the times of the two worlds. If one world is such that each time slice has more value than its corresponding time slice in the other world, the world with the time slices that have more value than their corresponding slices in the other world also has more value than the other world. The general strategy implemented by Kagan and Vallentyne is applicable to degrees of determinacy. Suppose that the two properties in question are being human and being an animal. Instead of counting up all the humans and animals in all possible worlds, look instead at the number

240  Rebecca Chan in each world. For every world, the number of things that are humans is equal or less than the number of things that are animals. Since humans are a subcategory of animals, it is impossible for the number of humans to exceed the number of animals. Even if there are infinite numbers of both in all possible worlds, we can use the numbers at each world to create an isomorphic pairing between the numbers of humans and numbers of animals at each world. Thus, even if there are infinite animals and infinite humans, it can still be shown that there are more things that are animals than things that are human. Consequently, the property of being an animal is less determinate than the property of being human. It is also worth noting that even if this solution isn’t applicable to all cases—perhaps it won’t work in the case of being a moral patient and being chordate—determinacy can still give us a thing’s most determinate and least determinate properties. An individual’s most determinate properties are the ones that only it has. In the case of kinds, the most determinate properties are the ones that only that kind has. The least determinate properties of an individual or kind will be those that all things have. The cases in which it may be difficult to assess which of two properties is more determinate (either because we’re epistemically impoverished or because there’s no fact of the matter) are cases in between these two extremes. And since what matters for extensional similarity with the definitional conception is the thing’s most essential properties, as long as we can figure out what a thing’s most determinate properties are, the modification will have done the crucial work with which it was tasked. With just this preliminary sketch, we can see how the revised modal account avoids the problems associated with non-discriminating necessary properties. It also avoids the problem of essential existence, assuming that existence turns out to be a trivially necessary property. Equipped with determinateness, the revised modal account begins to resemble the definitional account insofar as it can distinguish properties that seem to get at what a thing is from properties from properties that all objects have. The most essential properties of a thing are the properties that definitional accounts would tend to say are essential. It can also deal with the disjunctive and negative properties that are necessary but seemingly inessential. For example, the property of not being Plato is shared by everything but Plato. It is thus not very determinate, and, in turn, not very essential to Socrates. The property of being a human or a mountain is slightly more essential than not being Plato, but not as essential as being human since the scope of what a thing with that property could be is still somewhat broad. By appealing to determinacy, the revised account gives the intuitively correct result that disjunctive properties are going to be less essential than the property involving just one of the disjuncts. In addition to the completely non-discriminatory necessary properties, the slightly more discriminatory properties that still

Reviving the Modal Account of Essence  241 seem inessential are also accounted for by determinacy. However, the challenge posed by coextensive and derivative properties still remains. Stage 2: Sparseness The second stage of revision, sparseness, addresses the challenge posed by coextensive and derivative properties. Several proponents of the modal conception have proposed revising it by restricting the necessary properties that are candidate essential properties. This revision based on sparseness has the following schema: (Schema) a is essentially F iff (i) necessarily, Fa if a exists, and (ii) the property of being F is a sparse property. What counts as a sparse property differs wildly. Schaffer (2004) offers two conceptions of sparse properties—one that includes metaphysically fundamental properties and one that includes scientifically fundamental properties. Wildman (2013) picks up on the metaphysically fundamental conception and gives a sparse modal account of essence on which sparse properties are the ones with metaphysical priority. Denby (2014) offers a different account of sparseness on which intrinsic properties are the sparse ones. There are other variations as well that analyze sparseness in terms of non-triviality (Della Rocca 1996) or non-accidentalness (Gorman 2005). Though these analyses of sparseness differ, they function similarly once sparseness is incorporated into a modal account of essence. There are fewer essential properties than necessary ones. Thus, sparse accounts avoid the Finean counterexamples based on necessary coextension and determinateness, such as being the sole member of Socrates’s singleton set. My analysis of sparseness departs from all of these. Like the other sparse theories, I offer an account on which there are fewer candidate essential properties than initially assumed. For instance, it also turns out that on my account, there is only one essential property amongst being Socrates and being the sole member of Socrates’s singleton set. However, unlike the other sparse accounts, I don’t draw a distinction between sparse and non-sparse (or abundant) properties. Instead, there is only one property under consideration in the first place. In my analysis, necessarily coextensive properties are the same property. For convenience, let’s call this claim the “necessary coextension thesis.” According to the thesis, the property of being Socrates and the property of being the sole member of Socrates’s singleton set are necessarily coextensive, and thus turn out to be the same property. I admit that this view of coextensive properties is somewhat counterintuitive. For instance, triangularity and trilaterality are necessarily coextensive, and the former appears to be a property about angles, while the

242  Rebecca Chan latter appears to be a property about sides. But that they are in fact the same property has been defended by several different philosophers. For instance, on the Lewisean account of properties, properties are classes. To have a property is simply to be a member of a class.11 Where two properties are coextensive, the corresponding classes contain the same members. Since members of classes wholly determine the identity conditions of their classes, coextensive classes (and thus properties) turn out to be identical to each other. Oddie (2005) defends the necessary coextension thesis in a different way. In his view, there are distinct procedures or specifications of properties, but only one property where there is necessary coextension. For instance, just as subtracting two from five and taking the square root of nine are two different procedures for arriving at the number three, being the difference between two and five and being the square root of nine are two ways specifications of the property of being three. Similarly, trilaterality and triangularity are two specifications of the same condition—the condition had by all triangles. A procedure or specification might turn out to be prior to some other procedure or specification, but there is only one property in question. My goal is not to mediate between different views on why necessarily coextensive properties are the same property, but to show that if the thesis is true, then there is nothing objectionable about saying that the property is essential. If being the sole member of Socrates’s singleton set is the same property as being Socrates, and the latter is essential, then so is the former—after all, they are the same property! The upshot of this analysis of sparseness is that it avoids several problems that its competitors encounter. First, as Skiles (2015) points out, on accounts where sparseness is analyzed in terms of metaphysical or scientific fundamentality, many seemingly essential properties end up being inessential. For instance, being human is usually taken to be an essential property of Socrates. But the property of being human is likely not metaphysically or scientifically fundamental, which thus rules it out as being essential. Problems also arise for accounts like Denby’s that are based on some other feature like intrinsicality. Some objects have essential relational properties; for instance, the Eiffel tower is essentially man-made or designed by so-and-so. Since my account does not rely on any sort of fundamentality, intrinsicality, or the like, it avoids these problems. Second, giving an analysis of sparseness that is not purely modal runs dangerously close to removes the main advantage of having a modal account in the first place. The main advantage of the modal conception is that it is reductive. But if one gives an account of sparseness that introduces new primitives—as, for instance, Wildman and Gorman do—then the account of essence is no longer reductive. My analysis of sparseness is purely modal. It is thus able to enjoy the advantage of being purely modal and does not introduce unnecessary primitives.

Reviving the Modal Account of Essence  243 Summing Up Now that the two stages of my revised account are in place, we can see that my revised account is a purely modal one that is extensionally similar to the definitional account. By using determinateness to order necessary properties in degrees of essentialness, the account can distinguish between more or less discriminatory necessary properties just as the definitional account does. Furthermore, via sparseness, it avoids the problem of necessarily coextensive and derivative properties altogether. My revised modal account is not perfectly extensionally similar to definitional accounts. As I’ll explain in the next section, this leads to two theoretical advantages. However, my account is still similar enough to do all of the same work that definitional accounts do. First, it yields similar results with respect to properties that are necessary but seemingly inessential. On my account, these seemingly inessential properties turn out to be the least essential ones. Second, because we can talk about the most essential properties of a thing, we can enjoy the metaphysical explanation that definitional conceptions have to offer. For instance, the most essential properties are the ones that tell us about a thing’s nature. Third, my account individuates kinds and individuals just as well as the definitional account does. And finally, to the extent that the definitional conception is joint-carving, the revised modal conception is as well since it can approximate those joints.

Two Theoretical Advantages Before concluding, I would like to point out two theoretical advantages that my revised account has and that the definitional account lacks. First, my account is reductive, and thus does not introduce any new primitives. Definitional accounts, however, either take essence as primitive or must use some other primitive such as fundamentality to distinguish between essential and inessential necessary properties. Thus, my account is more parsimonious. Insofar as parsimony is a virtue, my account has a virtue that definitional accounts do not. Second, my account avoids arbitrariness when it comes to borderline cases. Plausibly, there are cases where it might be vague whether a certain property gets at what a thing is. For instance, being capable of moral reasoning seems to be one of the most significant properties that humans have. However, it is possible that this property does not rise to the level of essentiality for humans under the definitional account since it might follow from the property of being rational. On the definitional account, borderline definitional properties have to be either essential or ­non-essential. Classifying them as one rather than the other appears arbitrary. A similar problem arises in cases of necessarily coextensive properties. Without adopting the necessary coextension thesis, there can

244  Rebecca Chan be two properties that seem like equally good candidates for essentiality. For instance, both triangularity and trilaterality get at what a triangle is, and neither is obviously more fundamental than the other. But definitional accounts are not inclined to say that both triangularity and trilaterality are essential to triangles. Yet designating one rather than the other seems arbitrary in light of the fact that there aren’t good grounds for holding that one is more essential than the other. On the revised modal account, we don’t need to make judgments in any of these cases. All necessary properties are essential, and then the property of being capable of moral reasoning is simply more essential than properties like being an animal and roughly as essential as being rational. Admitting degrees of essentialness is more advantageous than attempting to draw a sharp cut-off between necessary essential properties and necessary non-essential properties. It avoids the implausible result that there is a sharp cut-off between properties that get at what a thing is and properties that do not. While the definitional account is committed to this implausible sharp cut-off thesis, the revised modal account is not. The revised modal account also is not committed to distinguishing between necessarily coextensive properties, since, via sparseness, they are the same property. My account thus handles close cases better than the definitional account does.

Concluding Thoughts These last two points show that my revised modal account goes beyond other attempts to repair the simple modal account. My revised account is extensionally similar to the definitional account, and the most essential properties of a thing can be called upon to do similar work to that of the definitional account’s essential properties. Furthermore, the revisions I have proposed are better at approximating the extension of the definitional account than those proposed by competing revised modal accounts. Thus, my account handles the Finean counterexamples, and does so in a principled way that avoids the arbitrariness to which definitional accounts are vulnerable. Finally, because my account is still purely modal, it preserves the main advantage of the traditional modal account—a fully reductive analysis of essence. This reductive analysis confers theoretical advantages (such as parsimony) to the modal account over the definitional one. Whether the theoretical advantages of reductiveness are greater than the purported theoretical advantages of the definitional account (such as those discussed in “Virtues of Exclusivity”) is an open question—at least as far as this paper is concerned. Obviously, there are tradeoffs that need to be carefully weighed. However, showing that the question is in fact open is significant, for it shows that the modal conception is, contrary to popular opinion, a legitimate contender to the definitional one and deserves to be revived.

Reviving the Modal Account of Essence  245

Acknowledgments This project is based on a paper I wrote for my very first graduate seminar and is a decade in the making—so I have a decade’s worth of kind and generous people to thank. First and foremost, I want to thank Kathrin Koslicki. Her seminar on ontological dependence birthed the core ideas in this paper, and her encouragement and feedback over the years kept me from abandoning it. I am also deeply indebted to Michael Rea for reading far too many versions of this paper and providing careful comments on each one. My dissertation committee, especially Christopher Shields and Peter van Inwagen, gave me invaluable feedback, especially toward the later stages of the project. And, along the way, Michaela McSweeney, David Barnett, Bradley Monton, Graham Oddie, and Jason Hanschmann helped me sharpen the ideas contained here. I am also grateful to Anand Vaidya and Sanna Mattila for the opportunity to think through these ideas one last time this volume. Finally, I both appreciate and apologize to anyone else I’ve forgotten to mention who deserves credit for any virtues this paper may contain.

Notes 1 For the purposes of this paper, I take for granted that the property of being human being is a necessary and essential property of Socrates. If, like Plantinga, the reader is inclined to think that Socrates could have been an alligator (or, more accurately, could have had an alligator body) and that the property of being human is not essential or necessary, the reader can substitute his or her favored prototypical essential property for that one. Also for the purposes of this paper, I set aside anti-essentialism and assume that things, such as Socrates, have essential properties. 2 For Correia (2006), the two questions are different. The former gets at an “objectual essence,” or what that particular object is, while the latter is a “generic essence” and gets at what it is to be a certain sort of thing. For our purposes, we need not make the distinction. The answers to “what is Socrates?” and “what is it to be Socrates?” will both be given by referencing his essence. The answer to the former will be “a thing that has such-and-such essence” and the answer to the latter will be “to have such-and-such essence.” 3 Quine (1960), who was skeptical of de re modality, rejected essentialism altogether. Part of Plantinga’s project was to respond on the behalf of de re modality and thus modal essence. Fine gives a different response on which he wants to preserve essence because of their special role in metaphysics. 4 For example, Gorman (2005) and Wildman (2013). 5 See Cowling (2013), especially section 5, for an illuminating diagnosis of the Quinean, Finean, and modal essentialist positions. 6 I’ve given the existential formulation. There are also other ways of formulating the modal conception (cf. Chisholm 1967), but since all are vulnerable in the same way to the definitional challenge, having just the existential formulation as our exemplar of the simple modal conception serves our purposes. 7 Whether essences can individuate individuals is controversial. Koslicki (2020), for example, argues that unless essences include haecceities or forms, they will not be able to individuate. I won’t take a stance on this issue in this paper since if there are individual essences, they’ll function similarly to kind

246  Rebecca Chan essences in a way that does not change the character of the dispute between the modal and definitional conceptions. 8 For example, Kment (2014) has an interesting account on which modal truths depend on essential ones. 9 For a detailed discussion of Fine’s view and objections to it, see Wallner and Vaidya (2020). 10 To be clear, in “New Work for a Theory of Universals,” Lewis (1983) distinguishes between properties and universals. The former are what I describe above. The latter are sparse and comprise the minimal basis for characterizing the world completely. Some people refer to universals as “necessary properties.”

References Chisholm, Roderick (1967), “Identity through Possible Worlds: Some ­Questions”, Nous, 1:1–8. Correia, Fabrice (2006), “Generic Essence, Objectual Essence, and Modality”, Nous, 40:753–767. Cowling, Sam (2013), “The Modal View of Essence”, Canadian Journal of ­Philosophy, 43(2):248–256. Della Rocca, Michael (1996), “Essentialism”, Philosophical Books, 37:1–13 and 81–89. Denby, David (2014), “Essence and Intrinsicality”, in R. Francescotti, ed., ­Companion to Intrinsic Properties, Berlin: De Gruyter, pp. 87–109. Fine, Kit (1994), “Essence and Modality”, Philosophical Perspectives, 8:1–16. Gorman, Michael (2005), “The Essential and the Accidental”, Ratio, 18:276–289. Kment, Boris (2014), Modality and Explanatory Reasoning, Oxford: Oxford University Press. Koslicki, Kathrin (2020), “Essence and Identity”, in Mircea Dumitru, ed., ­Metaphysics, Meaning and Modality: Themes from Kit Fine, Oxford: Oxford University Press. Lewis, David (1983), “New Work for a Theory of Universals”, Australasian ­Journal of Philosophy, 61:343–377. Oddie, Graham (2005), Value, Reality, and Desire, Oxford: Oxford University Press. Plantinga, Alvin (1974), The Nature of Necessity, Oxford: Oxford University Press. Quine, W.V.O. (1960), Word and Object, Cambridge: MIT Press. Schaffer, Jonathan (2004), “Two Conceptions of Sparse Properties”, Pacific Philosophical Quarterly, 85(1):92–102. Skiles, Alexander (2015), “Essence in Abundance”, Canadian Journal of Philosophy, 45(1):100–112. Vallentyne, Peter and Shelly Kagan (1997), “Infinite Value and Finitely Additive Value Theory”, Journal of Philosophy, 94(1):5–26. Wallner, Michael and Anand Vaidya (2020), “Essence, Explanation, and Modality”, Philosophy, 95:419–445. Wildman, Nathan (2013), “Modality, Sparsity, and Essence”, The Philosophical Quarterly, 63(253): 760–782.

12 A Neo-Aristotelian Reply to A Modalist Sanna Mattila

Introduction There are two main rival camps in contemporary essentialist discussions: on the one hand, modalists and, on the other hand, non-modalists – or as I call then, neo-Aristotelians. Both agree on the central essentialist thesis: that things have essences or essential properties independently of the ways we think about or refer to them. Yet, they disagree on how this thesis should be articulated: modalists aim to account for essence in terms of necessity, and often the modal notion of necessity is defined in terms of possible worlds. Neo-Aristotelians deny the reduction of essence to necessity and maintain that necessity should be explained in terms of essence. Thus, the central disagreement between the two camps concerns the relationship between the two notions, essence and necessity. The discussion has been very much shaped by Kit Fine’s seminal paper “Essence and Modality” (1994) which argues that necessary properties such as being distinct from the Eiffel tower and being the sole member of Socrates’s singleton set should not be counted as essential properties of Socrates. According to Fine, this shows that we need a notion of essence that is not given in terms of necessity. Modalists, subsequently, have aimed to answer Fine’s criticism by modifying their accounts so that they would render the same properties essential as Fine’s theory does, i.e., so that their theories would be extensionally similar to Fine’s.1 In this chapter, I discuss a novel modalist account, put forward by Rebecca Chan, and I defend neo-Aristotelian essentialism against it. My aim is to extend the discussion between modalists and neo-Aristotelians beyond Fine’s counterexamples, and to show that assessing their respective strengths requires taking into account other significant differences between them. This is also why I find the more general label – “neo-­Aristotelian account” – more apt than labels like “definitional account” or “non-modalist account”: in order to appreciate the power of neo-­Aristotelian essentialism, we should see it as a broader metaphysical theory. Like other recent modalist accounts, Chan’s account is shaped by Kit Fine’s counterexamples. Although offering a rival account to Fine’s essentialism in the sense that she aims to account for essences purely in

DOI: 10.4324/9781003002192-14

248  Sanna Mattila modal terms, Chan agrees with Fine’s strategy. Fine starts out with a pre-philosophical conception of essence as what a thing is and uses that conception in order to exclude cases that are not part of the essence of a thing. In this way, he claims, we find out that certain necessary properties should not be counted as the essential properties of Socrates, because these properties are inconsistent with the pre-theoretical conception of essence. Given that Chan aims to accommodate most of Fine’s intuitions about non-essential properties, she is committed to Fine’s strategy and to most of the outcomes Fine’s intuitions reveal. What they disagree about is how to account for the data we have gained this way: in contrast to Fine, Chan shows that a modalist framework can accommodate these intuitions too. In Chan’s proposal, the extensional similarity with Fine’s view is accomplished by adding two modifications to the simple modalist view which claims that all necessary properties are essential to a thing. First, she introduces the notion of determinacy of a property as a tool to make essentiality a matter of degree. Second, she adopts a special minimalist or sparse view of properties, where necessarily coextensive properties are taken to be identical. With these modifications, she claims, a modalist will get to have it all: necessary properties are essential to a thing but also some of them are more essential than other necessary properties. Moreover, it seems that she can handle most of Fine’s counterexamples and explain why it is good to depart from Fine’s intuitions in the cases where the views are at odds. While Chan’s elegant and important proposal should be taken seriously by other modalists and merits further discussions, I will argue that Chan’s account – despite its virtues – is not able to do the same work as neo-Aristotelian essentialism. Moreover, Chan’s own account suffers from problems neo-Aristotelians can avoid. I start by introducing the neo-Aristotelian account of essentialism in “The Neo-Aristotelian Account”. In the following sections, I focus on Chan’s two modifications and evaluate how her account fares with respect to the neo-Aristotelian account.

The Neo-Aristotelian Account In this section, I will characterize three features that I take to be central to neo-Aristotelian accounts of essentialism in addition to Fine’s distinction between essential and necessary properties. While my characterization is inspired by E.J. Lowe’s views, I take these central features to be such that all neo-Aristotelians, including Kit Fine, endorse. First, I maintain that a neo-Aristotelian understands the notion of essence in terms of the identity of a thing. Here, the notion of identity is a metaphysically substantial notion of identity, which is not the mere logical relation everything bears to itself, i.e., it is not a matter of

A Neo-Aristotelian Reply to A Modalist  249 numerical identity. This is a notion of identity which comes with specific ­metaphysical commitments about the nature of things. E. J. Lowe captures the central idea by quoting Locke: the identity of a thing is “the very being of anything, whereby it is, what it is” (Lowe 2008, 23; Locke ECHU III, III, 15). Things are not bare particulars escaping any qualitative characterization, but they are taken to be something in themselves, i.e., we can describe what they are and not just how they are. The reason it is important to understand essence in terms of the identity of a thing in this way is that it enables us to distinguish between representational entities, such as propositions and definitions, and worldly entities, such as Socrates or electrons or other non-representational entities: the essence or the identity of a thing is expressed by a real definition and represented by essential propositions. It is in virtue of the essence or identity of a thing that the essential propositions or real definitions are true. It is true that neo-Aristotelians sometimes run together the notion of real definition and essence. Fine (1995, 275), for example, suggests that “we may identify the being or essence of x with the collection of propositions that are true in virtue of its identity (or with the corresponding collection of essential properties)”. Still, Fine’s formulation implies that we need the distinction between a collection of propositions and the source of the truths of those propositions. Thus, the identity of a thing turns out to be ontologically prior to its real definition or essential truths also in Fine’s theory.2 Second, the neo-Aristotelian conception of essence allows for a distinction between individual essences and general essences. The individual essence of an entity X “is what it is to be the individual of kind K that X is”, whereas a general essence is “what it is to be a K” (Lowe 2008, 35). These both can be seen as the essence of an individual, although the individual essence is what it is to be that individual and the general essence is what it is to be an individual instance of a kind. It is also important to distinguish the essence of a kind from the essence of an individual: a kind essence is what it is to be a kind K. While the general essence of an individual and the essence of a kind are often considered to be closely connected, it is still a different thing to ask what the kind tiger is than to ask what a tiger is. Neo-Aristotelian essentialism need not be committed to individual essences, i.e., essences of the sort that explain, for example, the uniqueness of Socrates but one can allow there to be only general essences of individuals and essences of kinds. For example, a real definition in Aristotle’s own account expresses the essence of a thing but (at least according to the traditional interpretation) there are no individual essences for Aristotle. The essence of a thing is general, determined by the species-form the thing instantiates. It is the essence of the kind that the real definition captures, and thus there is strictly speaking no real definition of Socrates as an individual.3 For Aristotle, the individuation of a

250  Sanna Mattila concrete particular object is explained by other means, namely by the matter of the ­hylomorphic compound. In contemporary discussions, some are skeptical about individual essences (e.g. Tahko 2022), whereas some others explicitly endorse them (e.g. Lowe 2008). Third, I take it that Aristotelian essentialism, in its historical as well as more contemporary form, is interested in explaining reality as a whole, and not just in answering a specific metaphysical question in isolation of other metaphysical questions. In Aristotle’s own account, essences play multiple roles in explaining reality: they help to account for the structure of reality, for our knowledge of it, and for the dynamic nature of different things behaving in different ways. On the one hand, our own essences as rational beings explain our desire and ability to understand reality, and, on the other hand, it is because things have essences that reality itself is understandable.4 This ambition is alive within neo-Aristotelian essentialism. For example, questions about the structure of reality – fundamentality, grounding and ontological dependence – are still explored and answered in terms of essence-based metaphysics.5 Likewise, questions about the metaphysical nature of natural kinds and laws of nature are answered in terms of essences, and there is a growing interest in essence-based epistemology of modality as well.6 While a neo-Aristotelian holds that modality is grounded in essences of things, she uses the notion of essence in many other metaphysical roles too. This is what gives neo-Aristotelian essentialism its explanatory power: it is not used only to account for Fine’s intuitions about essential truths or “carve nature at its joints” but it is also used to account for the different levels of reality and for the multiple dynamic features of it.7 I take it that these three very broad characteristics of neo-Aristotelian essentialism will allow for a lot of variation among the specific theories. These characteristics do not imply any commitment to a certain ontology, although it is rather natural, and also preferable, that essentialist theories would not keep ontology at arm’s length. If one is to account for different features of reality, one should also say whether there are concrete, macrolevel objects or kinds or properties, and if so, what one takes them to be. Moreover, ontological issues have an important impact on what one takes essences themselves to be.8 Thus, the ontology matters for neo-Aristotelian essentialists, but they do not need to agree on ontology. Given the ambitions of neo-Aristotelian essentialism, it is worth asking to what extent a modalist account of essentialism is a genuine alternative to it. Essences have explanatory roles in the neo-Aristotelian framework that a purely modal conception of essence lacks. In what follows, I will argue that this general mismatch between the two essentialist camps also bears on the specific question of what essential properties are supposed to be and of whether Chan’s version of modalism is really extensionally similar to neo-Aristotelianism.

A Neo-Aristotelian Reply to A Modalist  251

The Determinacy Modification and the Need to Individuate As said, Kit Fine (1994) argues against modalist essentialism by showing that not all necessary properties of a thing are part of its essence. For example, Fine claims that intuitively it is not essential to Socrates to be the sole member of singleton Socrates nor is it essential to him to be distinct from the Eiffel Tower. The modalists, who claim that all necessary properties are essential to a thing, cannot capture this intuitive difference, and this makes their essentialist account implausible. Chan proposes to answer Fine’s counterexamples to modalism while still accounting for essences purely in modal terms. Instead of just claiming that all necessary properties of a thing are essential to it, Chan introduces two modifications which make her account more sophisticated than a simple modal conception of essence. The first modification is an appeal to determinacy, which allows Chan to claim that essentiality comes in degrees, giving her room to say that some necessary properties are among the most essential properties of a thing. This enables Chan to explain why all necessary properties are not on a par when we are considering their essentiality to a thing: it is because not all necessary properties are determinate enough. She defines the notion of determinacy as follows: A property is more (or less) determined in proportion to how narrow (or wide) the scope of what a thing that has that property could be. Chan (this volume, 238) As I understand Chan’s proposal, we can think of the scope of a property as its extension conceived as a set of actual and possible things, and we can order different properties in accordance to how many things occupy their respective extensions. I will refer to this as the transworld extension, and I am using the set theoretic account of properties as a useful tool, not implying a commitment to realism about possible worlds and possible individuals. Those properties that have the smallest number of things in their extensions are the most determinate, whereas properties whose extensions include all things whatsoever are the least determinate. The next step for Chan is to build the notion of determinacy into the notion of essentiality: the more determinate a necessary property is, the more essential it is. Since the property being such that 2 = 2 has any thing that does or could exist in its extension, such a property is not determinate at all and thus not relevant to what Socrates is, for example.9 However, the property being Socrates has only Socrates in its extension, and thus it seems to be a very relevant property to Socrates: it distinguishes him from other things.10 Following this intuition, Chan suggests that a necessary property with a smaller transworld extension, i.e., a property with a smaller number of actual and possible things in its extension, is more

252  Sanna Mattila essential to a thing than a property with a bigger extension. The aim of the modification is to make the most essential properties of a thing in Chan’s view the same set of properties as what the neo-Aristotelian account would hold to be the essential properties of a thing (Chan this volume, 232). This elite set of properties define what the thing is. In what follows, I will call the most essential properties in Chan’s account: the core essential properties.11 Understanding essentiality in terms of determinacy raises two important questions. First, how well does the determinacy-driven notion of essentiality fit with our pre-philosophical intuitions about essences? Second, how does it compare with the neo-Aristotelian version of essentialism? Given that Chan seems to be sharing the Finean methodology of starting out with our everyday intuitions about essences, she needs to take these two questions seriously. Chan’s determinacy modification is, for example, vulnerable to Fine style counterexamples the same way the traditional modalist account is. For example, electrons, as fundamental particles, have three fundamental properties, which, in conjunction, will be the most determinate property of the kind: to be an electron is to have a -e charge, ½ spin and mass of 91,093,837,015 × 10−31 kg. Thus, the property which is the conjunction of the specific charge, spin, and mass will have only the kind electron in its transworld extension. This is a viable candidate, then, to be the core essential property of the kind electron, given that it fulfills the demand to be the most determinate necessary property of the kind. But this leads to a very unintuitive result. Given that the property having a -e charge is not an exclusive property of electrons – also muons and tauons have the same charge – the property of having a -e charge will turn out to be less essential to an electron than having a -e charge and ½ spin and mass of 91,093,837,015 × 10 −31 kg. In other words, a property that enters into the core essence of a kind loses its status as a core essential property when taken apart from the conjunction. Chan’s determinacy modification functions in the same way with respect to individuals and their essential properties. Socrates’s property, for example, of being a human being is essential, given that it is a necessary property of him, but it is not as essential as a property whose extension contains only Socrates in all possible scenarios. If we accept that there is such a property as being Socrates, such a property would be the most essential property of Socrates, while being a human being, being an organism, being a thinking being are less essential to him, and thus do not enter into his core essence. Given that the core essence of an individual, in Chan’s account, excludes kinds altogether, one might wonder to what extent her view is extensionally similar to the neo-Aristotelian view. In fact, Fine (1994, 8) seems to hold, along with many other neo-Aristotelians, that Socrates is essentially a human being, a member of a kind. When taken as the most

A Neo-Aristotelian Reply to A Modalist  253 determinate necessary properties, the core essential properties are more limited than what a neo-Aristotelian would consider to be the essential properties of an individual. Transworld Individuation Focusing on the most determinate property of an individual is understandable, however, if one thinks of essentialism as an answer to the so-called problem of transworld identity, and Chan’s view is well-suited to answer the problem. I will argue, however, that the significance of the problem itself can be questioned, and more importantly, that formulating essentialism around the problem of transworld identity will make the view significantly different from the neo-Aristotelian view. The problem of transworld identity is a recent problem that arises within possible world semantics, where the framework of possible worlds is a tool to interpret our de re modal expressions. Thus, “Socrates could have survived his trial and lived 90 years” translates into “there is a possible world where Socrates survives his trial and lives 90 years”. But if actually existing individuals can and do exist in other possible worlds, how are we to account for their sameness? If Socrates is here in the actual world and there in some other possible world, what are the principles according to which his sameness is accounted for? After all, we want to investigate the possibilities of Socrates, and not of some other entity, so we better have a story to tell about what makes it the case that the individual in question is Socrates. It is important to pay attention to what notion of identity is at play here. As noted earlier, the notion of identity is an important factor in neo-Aristotelian essentialism to the extent that the notion of essence is understood in terms of the identity of a thing. This, however, is a metaphysically substantial notion of identity, and hence a very different notion than the one relevant to the problem of transworld identity which concerns numerical identity of a thing, in logic expressed by the identity or equality sign. While everybody, commonly, agrees that all things are self-identical – for example, that Socrates=Socrates – this does not entail any specific view about how identity should be accounted for. One could hold, for example, that things have non-qualitative thisness properties which explain their numerical identity, or that their identity-conditions are primitive and thus not be accountable in terms of anything else. Or one can appeal to qualitative, essential properties of individuals, which provide necessary and sufficient conditions for being identical with, for example, Socrates. To preserve clarity in distinguishing between the two different notions of identity, I will reserve the term “identity” only for the metaphysically substantial notion of identity. When talking about the problem of numerical identity, I will talk about individuation.

254  Sanna Mattila The problem of transworld individuation is a request for the necessary and sufficient conditions for individuation of particular objects, such as Socrates, and Chan’s modalist proposal does very well in answering that demand. The core essential properties of individuals in her account are such that they have only one individual in their transworld extensions12 (Chan this volume, 240). For example, the property of being Socrates has only Socrates in its extension. So, not only is it the case that Socrates necessarily has that property, it is also the case that whatever has that property is Socrates. Chan’s account is fully individuative, making it a strong rival to other essentialist accounts in this respect. Although historically it has not been the central aim of essentialism to answer the problem of individuation (as we saw, for Aristotle it is not essence but matter that explains individuation), one can maintain that our philosophical thinking has advanced in this respect and that the problem has a legitimate place in shaping essentialist theories: the possible worlds framework has just made the significance of the problem visible. Yet, this is controversial. For example, Saul Kripke, who is often placed in the modalist camp, takes the problem of transworld identity to be a pseudo-problem created by the apparatus of possible worlds. He argues that rather than losing an individual object into the plurality of possible worlds and needing some qualitative clues to recognize the same object again, we can stipulate from the start that we are considering this very object (Kripke 1980, 47–53). At least cognitively and epistemically, the object is prior to possible world construction: we are directly thinking about Socrates and considering what could have happened to him if things had been different. While Kripke’s point is not metaphysical, his insight fits with Aristotelian-minded essentialism too. If things have identities in a metaphysically substantial sense, as neo-Aristotelians hold, and these things or their identities ground the essential and thus the necessary truths, then one could think that the problem is not even metaphysically a genuine one. A neo-Aristotelian could stress that the problem of individuation should not be put in terms of possible worlds in the first place, because possible worlds are representational entities, which are dependent on how actual things are or at least on how the essences of things are.13 Asking questions about the metaphysical nature of these things by starting out with possible worlds is, once again, putting the cart before the horse. One does not need, of course, to agree with this assessment of possible worlds: one can also hold that possible worlds have priority over actuality or over the essences of things.14 But the nature of possible worlds is a contentious issue, and it is problematic if the central motivation for contemporary essentialism depends on adopting certain specific views about possible worlds.15 A modalist could, however, claim that the problem of individuation doesn’t need to rely on the commitment to possible worlds, but that it has

A Neo-Aristotelian Reply to A Modalist  255 16

a more intuitive ring to it. What we are after is an informative answer to the question: how could things have been different for a certain individual? For example, could Socrates have been a farmer or could he have had ten children? Could he have lived in the 21st century or could he have been a poached egg? These questions seem not only sensible enough, but also very central to any account of essentialism, and furthermore, these questions do not involve reference to possible worlds. But it could still be urged that these questions can be answered by appealing to the general essence of an individual: one could say, for example, that Socrates could have not been a poached egg, given that he is essentially a human being and no human being could have been a poached egg. Eggs, whether they are poached or not, come from chickens, and no human being can come from a chicken. Or one could say that Socrates could not have lived in the 21st century, given that for any entity whatsoever its actual position in the world’s causal history is essential to it (as Kripke 1980, 114–115 seems to assume). Even questions that come closest to transworld identity formulations, such as could Socrates have been Descartes, can be answered in this way: no, because Socrates and Descartes are distinct individuals and it is due to the general essence of things that if the two are distinct, they are necessarily distinct. This is where the explanatory ambitions of neo-Aristotelians are relevant to modal metaphysics: a neo-Aristotelian can appeal to general essences when explaining what is possible and what is not for individuals. In contrast, a modal account offers different modal metaphysics because it lacks the explanatory roles the neo-Aristotelian essences have. What is possible and what is necessary is primitive. Given the dual nature of the notions – either one can be defined in terms of the other –necessity, strictly speaking, cannot explain possibility. Moreover, it is not so clear whether Chan can make a distinction between, on the one hand, a general essence of a thing and, on the other, not-so-essential necessary property of a thing: assuming that being a thing is part of the general essence of Socrates, it is still in Chan’s account as determinate, i.e., as essential, as the property being such that 2 = 2. It does not discriminate at all and thus, it is not essential in any interesting respect in Chan’s account, but a neo-Aristotelian can take it to be a very significant general essence.17 This analysis reveals an important difference between Chan’s modalist account and the neo-Aristotelian account: Chan’s modalist essentialism seems to be shaped by the problem of transworld individuation, and it is well-suited for answering the problem. Neo-Aristotelians, however, can argue that they do not need to take the technical problem seriously, and yet, answer the non-technical questions about the possible vicissitudes of individuals in terms of general essences. This is also why the determinacy modification makes Chan’s view extensionally different from the neo-­Aristotelian view. With the modification, the core essential properties turn out to be more exclusive than what neo-Aristotelians would take to be essential to a thing.

256  Sanna Mattila

The Sparseness Modification and Borderline Cases While the determinacy modification is introduced in order to make the modalist account more exclusive and more in line with Fine’s counterexamples, another motivation for Chan’s modifications comes from the so-called borderline cases. Chan points out (this volume, 244), for example, that neither of the properties being triangular and being trilateral seem more fundamental than the other. Yet, the neo-Aristotelians will need to choose one over the other – and the choice seems to be arbitrary. Chan’s solution to overcome this kind of cases is to say that one is not more fundamental than the other: they are the same property and thus on equal footing. Chan’s second modification to the simple modalist view is, then, a notion of sparseness where necessarily coextensive properties are taken to be identical. This makes her account of sparseness different from the accounts that follow Lewis’s notion of sparseness, where naturalness or fundamentality explains the elite nature of sparse properties.18 Such an account allows the existence of abundant properties in addition to sparse properties, whereas for a proponent of the so-called necessary coextension thesis there are no abundant properties at all, but just a smaller number of properties altogether. Yet, while taking the notion of sparseness to be something different from the one Lewis introduces, the necessary coextension thesis is naturally supported by the Lewisian account of properties, where properties are taken to be sets of things. As Chan (this volume, 242) points out, in such a view, necessarily coextensive properties are identical, given that they have the same members across possible worlds. Chan also considers another type of borderline cases. According to Chan (this volume, 243–244), there are some necessary properties of individuals and kinds that the neo-Aristotelian account would not include in the essence of a thing, but which nevertheless are closely connected to the essence and clearly not just trivial necessary properties of a thing. She mentions the property of being capable of moral reasoning, which she (this volume, 243) thinks is “one of the most significant properties that humans have”, and she formulates the problem as follows 19: [I]t is possible that this property does not rise to the level of essentiality for humans under the definitional account since it might follow from the property of being rational. On the definitional account, borderline definitional properties have to be either essential or non-essential. Classifying them as one rather than the other appears arbitrary. Chan (this volume, 243–244) Thus, the problem is that the neo-Aristotelian account makes too sharp a distinction between essential and non-essential necessary properties

A Neo-Aristotelian Reply to A Modalist  257 without any good grounds for doing so. Chan’s own account with the determinacy modification in place will do better, given that it involves the idea of degrees of essentiality. It is not, however, so clear whether neo-Aristotelians are in such a bad position with respect to either of the borderline cases Chan introduces. It seems plausible, for example, that being trilateral is more fundamental than being triangular, because angles are by their nature dependent on lines. This would make the lines that are used to construct a triangle more basic than the angles. Likewise, one could claim that being capable of moral reasoning is dependent on rationality, thus making the capability of moral reasoning closely tied to the essence of the kind human. In general, if one is from the start happy, as neo-Aristotelians are, to account for the structure of reality by appealing to ontological dependencies or to grounding, then one has a repertoire of tools to explain connections between the properties that are part of the essence of a thing and those that follow or flow from the essential properties. This is not to say that it is an easy task to do but that the neo-Aristotelians are in a very good place to account for such a distinction. Moreover, Aristotle explicitly endorses such a distinction. Aristotle acknowledges that there are properties which are closely connected to the essences of a thing, and he introduces the distinction between the essence of a thing, which is to be expressed in a real definition, and the propria or the necessary accidents, which are necessary for a certain kind of thing, but which are not part of the definition (Topics 101a17–21). For example, being capable of moral reasoning or being capable of grammar are propria of human beings: they are closely connected to the essence of human beings but they are not part of the definition of the kind human. So, there is a clear answer to Chan’s example of a borderline case: no, they are not part of the essence of a thing but this does not mean that they are on a par with any trivial necessary property of a thing. It is due to what the thing is that it has the propria it has, but these propria still do not define the thing. It is because human beings are rational animals that they have the proprium being capable of moral reasoning. Thus, not only is Aristotle able to make the distinction that Chan is after but he is also suggesting that there is an explanatory relationship between essence and propria. Kit Fine (1995, 276) defends another distinction by distinguishing the constitutive essence from the consequential essence of a thing. The definition of constitutive essence is given negatively, defining it as a property “not had in virtue of being a logical consequence of some more basic essential properties”. The consequential essence of a thing, however, is exactly what is denied from the constitutive essence: properties belonging to the consequential essence of a thing are those that are the logical consequences of the constitutive essential properties. Although Fine’s distinction is problematic because it allows any logical consequence to be

258  Sanna Mattila relevantly connected to a thing’s essence, and thus fails to be fine-grained enough (Koslicki 2012, 189–195), one could still make use of the overall idea when accounting for Chan’s borderline cases.20 Moreover, it is in fact unclear whether Chan’s own account can deal with this kind of borderline cases. Consider the property of being capable of moral reasoning. Let’s assume that this property is connected to the property of being rational so that anything that is capable of moral reasoning is also rational. This makes rationality a necessary condition for being capable of moral reasoning, and thus the extension of the latter property cannot be bigger than the extension of the former. Hence, the property of being capable of moral reasoning is either as determinate or more determinate than the property of being rational. Chan needs to say either that the properties are necessarily coextensive and thus identical, or that the property of being capable of moral reasoning is more essential to an individual human being than the property of being rational. Both conclusions are problematic. On the one hand, if she claims that the properties are identical, she cannot appeal to the graded nature of essential properties when explaining borderline cases. The work would be done solely by the necessary coextension thesis. Note that neo-­Aristotelians could allow these properties to have the same extension, and still explain why one follows or flows from the other. Due to her necessary coextension commitment, Chan cannot do this. On the other hand, if she claims that they are distinct, then she has the unintuitive result that being capable of moral reasoning is more essential to some human being than being rational and not at all essential to some other human beings. There is also a third option available to Chan. She could claim that the extension of the property being rational is smaller than the extension of the property being capable of moral reasoning but this suggests that the capacity of moral reasoning is not dependent on rationality. Then, she would need to claim that there are things that are not rational but that are, nevertheless, capable of moral reasoning. In either case, a claim that the extensions of the two properties are not the same needs further support: Chan needs to tell a story about what these properties are as well as a story about what possible worlds are and what are the things inhabiting them if she takes properties solely to be sets of actual and possible things. These kinds of metaphysical questions about the nature of things, as pointed out in “The neo-Aristotelian Account”, are the central motivations for neo-Aristotelian essentialism in the first place, and now it seems that Chan cannot avoid them either. Just like the determinacy modification, the sparseness modification is also problematic in itself. First, as already mentioned, one can argue that being trilateral is more fundamental than being triangular: it is because angles are ontologically dependent on lines. Thus, these are two distinct properties, unlike the necessary coextension thesis assumes, because one property is more fundamental than the other. This line of argument

A Neo-Aristotelian Reply to A Modalist  259 can be put in a more general form: the assumed examples of identical ­properties violate Leibniz’s law, and thus, they cannot be identical. The law states that if x and y are identical then they have the same properties. But if we can show that necessarily coextensive properties, such as being triangular and being trilateral, have different properties, then the necessary coextension thesis collapses. It has been argued, for example, that the properties being triangular and being trilateral have different causal roles. We can think of a device that is designed to detect how many pointy corners – i.e., interior angles that stick out – a figure has, and which then goes on to show the number of the corners on a screen. When the device operates on a triangle, the result will be three. When it operates on a pentagon or on a five-pointed star, the result will be five, although these figures differ in the number of sides. This suggests that what causes the result on the screen is the number of angles or corners, not the number of sides. Therefore, when the device is operating on a triangle, it is the property being triangular and not the property being trilateral that causes the outcome three, making the two properties causally different and thus distinct (Audi 2016, 831; Sober 1982, 185). Moreover, the property of being Socrates seems to be an intrinsic property of Socrates, whereas being the sole member of Socrates’s singleton set is an extrinsic property, involving a relation to a thing that is distinct from Socrates. This again suggests that the properties are distinct, even though they are necessarily coextensive. The second problem for sparseness is that adopting Lewisian account of properties will create a problem for essentialists, and because the necessary coextension thesis most naturally fits the Lewisian account of properties, it is not so clear whether the coextension thesis itself suits the essentialist agenda. Consider Chan’s explanation of Lewisian properties21: For instance, on the Lewisian account of properties, properties are classes. To have a property is simply to be a member of a class. Where two properties are coextensive, then the corresponding classes contain the same members. Since members of classes wholly determine the identity conditions of their classes, coextensive classes (and thus properties) turn out to be identical to each other. Chan (this volume, 242) The problem I pose, then, arises from the following commitments: 1 Properties are sets. 2 Sets are individuated by their members. 3 Objects are individuated by their most determinate necessary properties.

260  Sanna Mattila The first commitment is the Lewisian account of properties and the ­second is a matter of mathematical definition: sets A and B are numerically identical if and only if they have the same members. For example, Socrates’s singleton set is individuated by its member, i.e., Socrates. Thus, the singleton set is dependent on Socrates for its numerical identity. The third commitment is the central thesis of Chan’s essentialism. As we saw, the most determinate necessary properties – the core essential properties, as I called them – individuate objects. The outcome is that Chan’s account becomes circular if she adopts the Lewisian account of properties: on the one hand, properties are dependent for their numerical identity on objects, and objects are dependent for their numerical identity on properties. For example, Socrates is the individual he is because he has the property being Socrates, and the property being Socrates is the property it is because it is a set that has Socrates, the man, as its member. The circularity problem is peculiar to essentialists who are willing to adopt the Lewisian account of properties. The neo-Aristotelians explain the nature of properties by other means and thus, they can avoid the problem. To sum up, the sparseness modification relies on a rather controversial thesis about necessarily coextensive properties. Moreover, the borderline cases Chan raises against the neo-Aristotelians can be answered within the neo-Aristotelian framework too, while it is not so clear how exactly the determinacy modification is supposed to help a modalist in answering borderline cases.

Conclusion I have argued that appreciating the metaphysical ambitions of ­neo-Aristotelian essentialism reveals important differences between the modalist account Chan is proposing and the neo-Aristotelian view. Chan’s overall strategy is to retain some neo-Aristotelian insights about essences without taking onboard the whole metaphysical package of neo-Aristotelian essentialism. This makes her view more parsimonious, but, as I have argued, Chan’s modalist account fails to be extensionally similar to the neo-Aristotelian view and a purely modal account of essences lacks the explanatory power the neo-Aristotelian view has. Moreover, the borderline cases Chan uses against neo-Aristotelians are, in fact, explainable within the neo-Aristotelian framework, whereas it is not so clear whether Chan’s gradual account of essentiality helps answer these cases. I emphasized that a general feature of the neo-Aristotelian view is that the notion of essence can be used for many different purposes. This is especially relevant to modal metaphysics because recognizing the different roles essences play in neo-Aristotelian essentialism helps us to see that a purely modal account cannot do the same work: neo-Aristotelians can make use of the general essence of an individual in explaining the possible outcomes for an individual. The general essence can also explain

A Neo-Aristotelian Reply to A Modalist  261 why certain necessary properties are closely tied to the essence of a thing, although they are not part of the definition of the thing. Moreover, the neo-Aristotelian account offers an interpretation of possible worlds where things or essences enjoy ontological priority, which puts the problem of transworld individuation in a new light. It is quite common to think of essentialism as an answer to the problem, but, as I argued, it is not so clear from the neo-Aristotelian perspective that the problem is a genuine one. Given these differences, it is hard to see how a purely modal account could be an alternative to a neo-Aristotelian account: the two different camps of essentialism are driven by different motivations and background assumptions, even if they agree on the central thesis that there are essences or essential properties.

Acknowledgments I am grateful to Anand Vaidya for helpful feedback on my essay. ­Likewise, I like to thank Tuomas Tahko, Donnchadh O’Conaill, Markku Keinänen, Peter Myrdal, Arto Repo, Veli Virmajoki, and Rebecca Chan for discussions and comments. My research is funded by the Finnish Cultural Foundation (projects 00180709 and 00200718).

Notes 1 For example, Cowling (2013), Wildman (2013), Gorman (2005), and Della Rocca (1996). See Livingstone-Banks (2017) who rejects the need for modalism to be extensionally similar to Fine’s essentialism. 2 Likewise, when drawing an analogy between analyticity and necessity, and between essence and meaning, Fine (1994, 9) claims that “[j]ust as a necessary truth may be true in virtue of the identity of certain objects as opposed to others, so an analytic truth may be true in virtue of the meanings of certain terms as opposed to others”. The analogy is between the identity of an object (or of certain objects) and meaning, not between real definition and meaning. 3 See, for example, Metaphysics VII.4; Lear (1988, 268–269); Witt (1989, 175–176). 4 See, for example, Metaphysics I, VII, Analytic Posterior, Physics II. An excellent entry point to the discussion is Lear (1988). 5 See, for example, Tahko and Lowe (2020), Correia and Skiles (2019), O’Conaill (2018), Wallner (2021), and Fine (2015). 6 On laws of natures and natural kinds: see, for example, Lowe (2006), Ellis (2001). On epistemology of modality, see, for example, Lowe (2012), Mallozzi (2021), Jago (2021), Vaidya and Wallner (2021). 7 E.J. Lowe’s metaphysics is an excellent example of a theory where this approach is taken seriously. 8 Although the discussion is often had in terms of essential properties, it is not unproblematic to assume that essences are properties or sets of them. I will, however, leave this issue aside here. 9 Note, however, that the property being such that 2 = 2 might be intuitively counted as very relevant and interesting to number 2 (if such an object exists) despite the fact that the property is not determinate and thus not among the most essential properties of anything at all in Chan’s account.

262  Sanna Mattila 10 I will assume such a property to be unproblematic in order to preserve the example Chan is using. However, one might wonder whether being Socrates turns out to be haecceitist property, expressing a non-qualitative thisness of an object. 11 This comes close to Fine’s (1995) distinction between the constitutive essence and the consequential essence of an object. I will get back to Fine’s distinction toward the end of the paper. 12 The same applies to kinds: the core essential property of a kind individuates a kind (Chan this volume, 240). For the sake of simplicity, I will focus here on the individuals only. 13 Lowe, for example, wants to maintain that modalities are grounded on essences, but he denies that only actually existing things have essences. 14 See, for example, Lewis (1973) and Plantinga (1974). Although Lewis’s metaphysics does not allow for transworld identity, he nevertheless takes the problem itself seriously and answers it in terms of counterpart relations (Lewis 1973, 39–41). 15 See also Oderberg (2007) on possible worlds and the problems they pose to modalists. 16 See also Koslicki (2018) for an elaborate defense of the genuinity of the problem. She is a neo-Aristotelian, who is committed to non-modal essences, but she answers the transworld individuation problem in terms of individual forms. 17 See Fiocco (2019) for the metaphysical significance of the question: what is a thing? 18 See Cowling (2013) and Wildman (2013). Lewis’ notion of sparseness is given in terms of naturalness: he thinks that there are properties that are more natural than others, and some properties are perfectly natural (Lewis 1983). Given that Chan’s modifications do not adopt this notion of sparseness, her account is more appealing to those who find Lewis’ formulation of sparse properties unsatisfactory. 19 I read this as the claim that all human beings are capable of moral reasoning. I think this is a legitimate assumption, although it needs elaboration. One might question, for example, whether psychopaths are in fact capable of moral reasoning. This depends on the notion of capacity and the notion of moral reasoning: a child, for example, cannot often tell difference between right and wrong, but we might still say that she is capable of learning moral reasoning. Likewise, one could hold that a psychopath has the capacity for moral reasoning, but it is not exercised for one reason or another. Or one could think that psychopaths are exercising moral reasoning just like other people are, but they just do not care for the outcome – thus, they are behaving as if they lacked moral reasoning altogether. 20 See Koslicki (2012) for further discussion of Fine’s and Aristotle’s accounts. 21 It is true that Chan is not committing herself to Lewisian properties but merely pointing to different ways of supporting the necessary coextension thesis. I still take the Lewisian account of properties to offer the strongest support for the thesis and thus it is worth evaluating the consequences of adopting such a view.

References Aristotle (1995) The Complete Works of Aristotle. The Revised Oxford Translation I-II. Edited by Jonathan Barnes, Princeton, NJ: Princeton University Press. Audi, Paul (2016) “Property Identity”, Philosophy Compass 11: 829–840. Chan, Rebecca (this volume) “Reviving the Modal Account of Essence”.

A Neo-Aristotelian Reply to A Modalist  263 Correia, Fabrice and Skiles, Alexander (2019) “Grounding, Essence and ­Identity”, Philosophy and Phenomenological Research 98 (3): 642–670. Cowling, Sam (2013) “The Modal View of Essence”, Canadian Journal of ­Philosophy 43 (2): 248–256. Della Rocca, Michael (1996) “Essentialism”, Philosophical Books 37: 1–13 and 81–89. Ellis, Brian (2001) Scientific Essentialism, Cambridge: Cambridge University Press. Fine, Kit (1995) “Ontological Dependence”, Proceedings of the Aristotelian ­Society 95: 269–290. Fine, Kit (2015) “Unified Foundations for Essence and Ground”, Journal of the American Philosophical Association 1 (2): 296–311. Fiocco, M. Oreste (2019) “What Is a Thing?” Metaphilosophy 50 (5):649–669. Gorman, Michael (2005) “The Essential and the Accidental”, Ratio 18: 276–289. Jago, Mark (2021) “Knowing How Things Might Have Been”, Synthese 198: 1981–1999. Koslicki, Kathrin (2012) “Essence, Necessity, and Explanation” in T. Tahko (ed.), Contemporary Aristotelian Metaphysics (pp. 187‒206). Cambridge: C ­ ambridge University Press. Koslicki, Kathrin (2018) Form, Matter, Substance, Oxford: Oxford University Press. Kripke, Saul (2007) Naming and Necessity, Oxford: Blackwell. Lear, Jonathan (1988) Aristotle: The Desire to Understand. Cambridge: ­Cambridge University Press. Lewis, David (1973) Counterfactuals. Oxford: Blackwell. Lewis, David (1983) “New Work for a Theory of Universals”, Australasian ­Journal of Philosophy 61 (4): 343–377. Lewis, David (1986) On the Plurality of Worlds, Oxford: Blackwell. Livingstone-Banks, Jonathan (2017) “In Defense of Modal Essentialism”, Inquiry 60 (8): 816–838. Locke, John (1979) An Essay Concerning Human Understanding, ed. P. H. ­Nidditch, Oxford: Clarendon Press. Lowe, E. Jonathan (2006) The Four-Category Ontology, Oxford: Oxford University Press. Lowe, E. Jonathan (2008) “Two Notions of Being: Entity and Essence”, Royal Institute of Philosophy Supplement 83: 23–48. Lowe, E. Jonathan (2012) “What Is the Source of Our Knowledge of Modal Truths?”, Mind 121: 919–950. Mallozzi, Antonella (2021) “Putting Modal Metaphysics First”, Synthese 198: 1937–1956. O’Conaill, Donnchadh (2018) “Grounding, Physicalism and Necessity”, Inquiry 61 (7): 713–730. Oderberg, David S. (2007) Real Essentialism, London and New York: Routledge. Plantinga, Alvin (1974) The Nature of Necessity. Oxford: Clarendon Press. Sober, Elliot (1982) “Why Logically Equivalent Predicates May Pick Out Different Properties”, American Philosophical Quarterly 19 (2): 183–189. Tahko, Tuomas E. (2022) “Possibility Precedes Actuality”, Erkenntnis. https:// doi.org/10.1007/s10670-022-00518-w. Tahko, Tuomas E. and Lowe, E. Jonathan (2020) “Ontological Dependence”, in E. N. Zalta (ed.), The Stanford Encyclopedia of Philosophy, https://plato.stanford.edu/archives/fall2020/entries/dependence-ontological/.

264  Sanna Mattila Vaidya, Anand and Wallner, Michael (2021) “The Epistemology of Modality and the Problem of Modal Epistemic Friction”, Synthese 198: 1909–1935. Wallner, Michael (2021) “The Ground of Ground, Essence, and Explanation”, Synthese 198: 1257–1277. Wildman, Nathan (2013) “Modality, Sparsity, and Essence”, The Philosophical Quarterly 63 (253): 760–782. Witt, Charlotte (1989) Substance and Essence in Aristotle: An Interpretation of Metaphysics VII-IX. Ithaca, NY: Cornell University Press.

13 Semantic Rules, Modal Knowledge, and Analyticity Antonella Mallozzi

Introduction Amie Thomasson (2007, 2018, 2020) has put forward an ambitious account of metaphysical modality, which she calls Modal Normativism— henceforth (MN). According to (MN), the function of metaphysical modal discourse isn’t to describe modal facts or properties in the world (or in other possible worlds). Instead, in thinking and talking about what is metaphysically possible and necessary, what we are doing is expressing, applying, and renegotiating semantic rules. As a consequence, knowledge of metaphysical modality is to be explained in terms of our understanding and use of semantic rules—i.e., semantic mastery—rather than our epistemic grasp of some independent modal reality. Here is an outline of the chapter. In Section ‘Introduction’, I present Thomasson’s (MN). In Section ‘What Is Modal Normativism?’, I aim to show, against (MN), that semantic mastery is not sufficient for gaining knowledge of metaphysical modality. For a competent subject could always wonder whether something is metaphysically possible (or necessary) only based on her mastery of the semantic rules (and possibly empirical information). In reasoning aimed at establishing what is metaphysically possible and necessary, a competent subject further needs to rely on essentialist principles and information. In response (Section ‘Semantic Mastery vs. Metaphysical Modal Knowledge’), modal normativists might deny that any such ‘extra’ principle and information is needed for knowledge of metaphysical modality. Instead, on analogy with quasi-realism about morality, they might argue that a competent speaker will only need to appeal to specific independence counterfactuals. Those conditionals fix the meaning of our terms at the actual world, independently of the particular context in which a statement is evaluated. If such conditionals hold, a subject who masters the rules couldn’t rationally wonder whether something is metaphysically possible (or necessary) while still counting as competent. However, in Section ‘Semantic Mastery vs. Metaphysical Modal Knowledge’, I show that this strategy is problematic for (MN) in that it appeals to metasemantic principles that go beyond one’s semantic competence. Additionally, the strategy raises

DOI: 10.4324/9781003002192-15

266  Antonella Mallozzi a worry of vicious circularity and seems potentially question-begging. In Section ‘Independence Conditionals for Knowledge of Metaphysical Modality’, I explore a possible reply by modal normativists, which rests on identifying logical-conceptual modality with metaphysical modality at the level of worlds—a thesis known as ‘Modal Monism’. But Modal Monism involves notorious problems, which the normativist will have to address. Lacking convincing reasons to think that Modal Monism is true, it is doubtful that semantic mastery alone (or aided by empirical information) can yield knowledge of metaphysical modality. Still, (MN) can account for some modal knowledge without committing to Modal Monism. As I show in Section ‘A Possible Way-Out: Modal Monism’, semantic mastery may suffice for gaining knowledge of logical-­conceptual necessity or analyticity. I introduce Timothy Williamson’s challenge to knowledge of analyticity (2007) and argue that normativists can successfully address it by adopting Paul Boghossian’s recent reply (2020).

What Is Modal Normativism? Thomasson aims to explain knowledge of metaphysical modality, since this is the modality that is at stake within central debates in philosophy. While she doesn’t explicitly characterize metaphysical modality or what counts as a metaphysical necessity or possibility, she offers a number of examples of metaphysical necessities (and derivatively, possibilities). Those include not just a priori analytic truths such as ‘Necessarily, all bachelors are unmarried’, but also traditional Kripkean a posteriori necessities, e.g., ‘Necessarily seals are mammals’, including de re ones, e.g., ‘Water is necessarily H2O’ (see esp. 2020: Ch. 4). Thomasson frames modal normativism as a main alternative to traditional descriptivism, according to which modal discourse is aimed at tracking or describing certain features of our world (or of other possible worlds), namely modal facts and properties that exist independently of us. For Thomasson, descriptivism is false. She argues that metaphysical modal discourse is distinctively normative, in that it ‘serves the function of expressing, teaching, conveying, or (re-)negotiating semantic rules (or their consequences) in particularly advantageous ways’ (2018: S2087, Also 2020: 64). Semantic rules include ‘application conditions’, which concern the conditions under which a term is to be applied or refused, as well as ‘co-application conditions’ namely rules dictating when a name or sortal term may be applied again to one and the same entity (2007: 140).1 On (MN), claims of metaphysical possibility and necessity should thus be understood as simply expressing our semantic rules, as opposed to tracking independent modal truth-makers. Take, for example, (B) Necessarily, all bachelors are unmarried males

Semantic Rules, Modal Knowledge, and Analyticity  267 The descriptivist says that (B) describes the necessary fact about ­bachelors that they are unmarried males, or that bachelors (qua the kind, not the individuals independently of the kind) necessarily possess the properties of being males and unmarried. Such properties or facts make (B) true. By  contrast, according to Thomasson, (B) is to be explained in terms of the rules for correctly applying the term ‘bachelor’, particularly, the (metalinguistic) rule: ‘Apply “bachelor” only where “unmarried male” applies’. By contrast with the metalinguistic rule, (B) explicitly states a necessity. Still, giving an account of metaphysical modality in terms of semantic rules apparently doesn’t prevent the normativist from talking of modal truths, facts, and properties. One will only need a few more steps to get to those from the relevant linguistic expressions. From (B), one can infer ‘ is true’ by applying the equivalence schema ‘

is true IFF p’. Furthermore, Thomasson holds that from true modal claims we can ‘trivially infer’ that modal facts and properties exist. For example, (B) implies ‘It is a fact that it is necessary that all bachelors are unmarried males’. And from a de re modal claim such as ‘Water is necessarily H2O’, one may trivially infer ‘Water has the modal property of being necessarily H2O’. Thomasson’s thesis is that once (MN) is in place, we can explain knowledge of metaphysical modality in terms of semantic mastery.2 More precisely, on (MN), the ability to use our terms appropriately is tantamount to tacit modal knowledge. For being able to correctly apply the semantic rules governing our terms shows that one at least implicitly knows what the rules require and allow, namely, as we might put it, what must be (necessity) and might be (possibility) in cases that fall under the scope of the relevant rules. However, speakers will have explicit modal knowledge when they ‘gain an explicit understanding of the rules’ (2018: S2091), which enables them, in turn, to articulate and communicate the rules. This goes beyond one’s ability to correctly apply or refuse to apply our terms in relevant contexts. What is required for explicit modal knowledge is that a speaker can explicitly express the semantic rules themselves in the object language. In Thomasson’s words, explicit modal knowledge is the ability to mov[e] from mastering the rules for properly applying and refusing expressions (as a competent speaker), to being able to explicitly convey these rules (and what follows from them) in the object language and indicative mood. (2018: S2091) Importantly, empirical information will sometimes contribute to the acquisition of modal knowledge, in addition to one’s conceptual mastery.

268  Antonella Mallozzi On (MN), knowledge of empirical facts and empirical discoveries often contributes to knowing derivative modal facts. Consider, for example, (W) Whatever microstructure the baptized sample has, water necessarily has that microstructure For Thomasson, (W) is a ‘conceptual truth that we can know via c­ onceptual analysis’. However, the fact that water necessarily has microstructure H2O is a derivative modal fact that one may come to know via empirical investigation (2018: S2092. Also 2020: 163–164).

Semantic Mastery vs. Metaphysical Modal Knowledge Here is my main worry. Does (MN) succeed in explaining knowledge of metaphysical modality? Can one’s mastery of semantic rules (possibly together with empirical information) yield knowledge of metaphysical possibility and necessity?3 A natural concern that one might have is that coming to know metaphysical modal truths requires investigating philosophical issues that seem to go beyond matters of semantic competence and linguistic practice. For instance, issues involving the nature of things or their essence, what sorts of grounding relationships things are involved in, the modal status of the laws of nature, and so on. How can knowledge of such issues solely derive from semantic mastery (and possibly empirical information)? Semantic mastery allows speakers to use words correctly. It may well enable them to formulate complex modal questions with great precision but arguably it doesn’t per se put one in a position to answer such questions. Consider an example. Patty is a chemist who works on water quality around the world. She has mastered the application and co-application conditions for the term ‘water’ as well as the relevant empirical information about actual water, while also being a rigorous reasoner. Could Patty rationally wonder whether water is necessarily H2O, or whether, say, it could have contained carbon instead? Modal normativists would say no. Remember principle (W): (W) Whatever microstructure the baptized sample has, water necessarily has that microstructure According to (MN), Patty should know (W) solely from mastering the rules for the ingredient terms (that is, a priori). Then, by combining principle (W) with the other semantic rules as well as with what she empirically knows about water, she should be able to conclude that water couldn’t have contained carbon. Water is necessarily H2O. One main difficulty with this approach is that it isn’t clear what exactly in the semantic rules for ‘microstructure’, ‘baptized sample’, and ‘water’

Semantic Rules, Modal Knowledge, and Analyticity  269 would justify conclusions concerning the metaphysical necessity of a ­substance’s microstructure. It isn’t clear, in other words, what would compel a thinker to connect the rules for using those terms in such a way that she would infer principle (W). Which rule or combination of rules would Patty have to master in order to come to draw such a metaphysical conclusion? That’s hard to say. But, without principle (W) and by only appealing to the rules for using ‘water’ and ‘H2O’, Patty couldn’t rule out the possibility that water might have contained carbon if the actual world had been different. She could wonder whether water is necessarily H2O without counting as irrational, incompetent, or poorly informed for doing so. For such a scenario seems consistent with what the rules for using those terms strictly dictate. I think that in order for Patty to come to know (W), she would need to rely on certain further assumptions. Specifically, she needs to deploy a general essentialist bridge-principle: (E) If it is essential to x being F that it is G, then necessarily anything that is F is G together with the information that (C) Having chemical composition C is essential to being a certain kind of substance s (E) and (C) are the additional premises that together would allow a competent speaker like Patty to rationally derive (W) from her semantic base. But these sorts of metaphysical principles also seem hardly reducible to the semantic rules for the correct application of our terms. How would Patty retrieve (E) and (C) from her semantic repertoire? In this case, too, it isn’t clear how the rules for using the ingredient terms would dictate (given adequate translations) principles (E) and (C). It seems indeed consistent with those rules that the principles are false. However, without the contribution of (E) and (C), it would seem that (W) was simply introduced in the language ‘by hand’ or arbitrarily: for nothing in the rules for correctly using ‘microstructure’, ‘baptized sample’, and ‘water’ appears to indicate that a substance necessarily possesses its actual microstructure. Note further that empirical information won’t help fill the gap here between semantic knowledge and metaphysical modal knowledge, no matter how accurate that information is. Empirical information concerns how things actually are on Earth but says nothing about other possible worlds—especially worlds where the laws of nature are quite different from our world’s. But if Patty can’t infer (E), (C), or (W) solely from her mastery of the ingredient terms, that means that solely based on her semantic

270  Antonella Mallozzi competence she could legitimately wonder whether water is necessarily H2O, or whether it could have contained carbon instead. Indeed, if she could only rely on her mastery of semantic rules, she would likely have many other queries concerning metaphysical modality. She might wonder whether it is in virtue of the nature or essence of water that it cannot contain carbon. She might wonder whether the fact that something is a sample of water is grounded in its being H2O. She might consider some distant possible world where the laws of nature are very different from our own: could there have been water in such a world, or perhaps something in its core properties is tied to the actual chemical and physical laws in such an intimate way that makes that scenario metaphysically impossible? Which semantic rules in Patty’s repertoire could answer such questions? Again, it’s hard to say. The rules seem simply silent about these issues. Crucially, as mentioned, we wouldn’t take Patty’s queries as an indication of some flaw in her semantic competence, relevant empirical knowledge, or reasoning. We’d still regard her as a perfectly competent speaker of English, who’s also very knowledgeable about chemistry, and who reasonably wonders about such difficult issues. However, on (MN), we would be forced to conclude that Patty isn’t a competent speaker of English or a good chemist after all. Or even that she’s irrational.4 Perhaps the modal normativist might cook up some principles analogous to (W) containing the terms ‘essence’, ‘grounding’, ‘natural laws’, etc., which are devised to address each of those issues. But note that won’t be much progress. For an analogous challenge to the one I raised in the case of (W) arises, namely to show how the purported principles could themselves be derived solely from our semantic rules.

Independence Conditionals for Knowledge of Metaphysical Modality I suggested that (MN) needs to integrate knowledge of essentialist or other metaphysical principles and information besides one’s conceptual mastery (and possibly empirical information) in order to explain knowledge of metaphysical modality. The modal normativist might reply that the objection begs the question in favor of certain ‘heavy-weight’ descriptivist metaphysics and against modal normativism. For the objection claims that notions such as essence or the nature of things, grounding, etc., escape the kind of semantic reduction (MN) proposed. But those notions might themselves have a plausible treatment on the normativist view. However, modal normativists might also reply that they don’t need to resort to any such notions at all. There is a different strategy to ensure that a rational, competent, and well-informed speaker won’t have doubts concerning metaphysical modal matters such as whether water is necessarily

Semantic Rules, Modal Knowledge, and Analyticity  271 H2O. The strategy consists in addressing objections like the one involving chemist Patty in the same way as normativists address the traditional objection against conventionalism. The objection against conventionalism denies that we can know metaphysical modal truths by extrapolating them from our semantic competence, since our linguistic conventions might have been different.5 But normativists stress that in their view modality isn’t contingent on the particular linguistic conventions we happened to adopt. On the contrary, they claim that they are entitled to accept specific independence counterfactuals, on analogy with quasi-realism about morality. Like ­quasi-realists, modal normativists also […] accept that these moral/modal facts are—in a relevant and important sense—mind-independent. For both accept certain independence conditionals. The moral quasi-realist accepts, for example, that it would still be wrong to kick dogs for fun, even if it were the case that I (and others) approved of it. (2018: S2096) Analogously in the case of (MN), The modal normativist can tell a parallel story, entitling her to accept independence conditionals—accepting, for example, that it is necessary that seals are mammals, and that this would still be the case even in worlds in which there were no speakers or thinkers (and so in which we don’t use the relevant terms) at all. (2018: S2096) The independence conditionals specify that certain truths hold independently of contingencies about how we use language and the semantic rules we have adopted. They fix the meaning of our terms at the actual world, independently of the particular context in which a statement is evaluated. In this way, they secure that the modal statements we derive from our semantic rules effectively range over all metaphysically possible worlds. The general form of an independence counterfactual is the following: given that it is necessary that xs are F, then even if we didn’t use terms A and B to refer to xs and F, respectively, it would still be necessary that xs are F. Accordingly, it would still be the case that ‘seals are mammals’ is true in (i) worlds where there are no speakers or thinkers, as well as (ii) worlds where people use the terms ‘seals’ and ‘mammals’ differently—worlds, let’s say, where those words pick out lizards and birds, respectively. This strategy should help the modal normativist explain how one could gain knowledge of metaphysical modality solely based on her semantic

272  Antonella Mallozzi competence, since the conditionals entail that we should take our (actual) semantic rules to hold at all possible worlds. As Thomasson explains: When we evaluate a counterfactual conditional, we must evaluate its truth at another world, leaving its meaning fixed as the actual meaning at our world. For we want to know whether this same claim (“Necessarily, all seals are mammals”), with the same meaning, would be true at another world, in other circumstances or given other suppositions. (2020: 89) Going back to chemist Patty, based on the independence counterfactuals she should conclude without hesitation that water is necessarily H2O, likewise that it is necessary that seals are mammals (assuming she is also minimally knowledgeable in biology). Patty would know that (actual) semantic rules such as ‘Apply “water” only where “H2O” applies’, or ‘Apply “seal” only where “mammal” applies’ are meant to hold at all possible worlds. Thus, she couldn’t rationally wonder whether water is necessarily H2O or necessarily seals are mammals, given her semantic competence and empirical background—which blocks our original objection. However, this reply raises several problems. Let us leave aside the issue of whether morality and modality are in fact relevantly similar, such that metaethical arguments can be successfully recast for the case of modality by simple analogy and go through as well. There are three additional main problems. First, endorsing the independence counterfactuals requires accepting a meta-rule concerning how our semantic rules work at all possible worlds. This indicates that the conditionals carry a primitive or unexplained modal element, which raises a worry of vicious circularity. Second, the need to integrate the independence conditionals shows, against (MN)’s main tenet, that conceptual competence alone isn’t strictly sufficient for knowing metaphysical modality. Third, it is not clear on what grounds one should accept such conditionals. On the contrary, the modal normativist doesn’t seem to have good reasons to endorse them. That such conditionals hold is something normativists assume, rather than defend, which raises the worry that the view begs the question in their favor. Let us discuss these issues, in turn. The Threat of Circularity As we saw, modal normativists may invoke suitable independence conditionals to ensure that a competent subject will reach the correct judgments concerning metaphysical necessity and possibility in all sorts of counterfactual scenarios. These conditionals secure that the meaning of

Semantic Rules, Modal Knowledge, and Analyticity  273 our words won’t change in different contexts but will stay stable across all possible worlds. But then it appears that endorsing such conditionals entails in effect accepting a second-order rule concerning the scope of our own semantic rules. As I shall formulate it, the meta-rule establishes that (N) Actual semantic rules should be held fixed at all possible worlds, i.e. they should be taken to hold necessarily If this is correct, (MN)’s account of knowledge of metaphysical modality turns out to involve an implicit modal element. A competent speaker would have to have some prior grasp on what ‘necessarily’ means in order to be able to apply the independence conditionals and reach the correct modal judgments. To further elaborate, the problem is that in order for a subject to accept the independence counterfactuals, she would need to know not just what our semantic rules are and how to express them explicitly. But also that the rules should be kept fixed at all possible worlds. Otherwise, how could she conclude that rules such as ‘Apply “water” only where “H2O” applies’, or ‘Apply “seal” only where “mammal” applies’ are meant to regulate uses of ‘water’, and ‘seal’ at all possible worlds, not just in actual world contexts? Being able to formulate the independence counterfactuals thus requires applying meta-rule (N). But that means that some modal knowledge is presupposed by one’s correct application of the semantic rules, rather than being derived from the rules like (MN) holds. Thus, my first worry is that (MN)’s account of modal knowledge is threatened by vicious circularity. That a competent subject must have some prior knowledge of the concept of metaphysical necessity seems problematic, since on (MN) knowledge of metaphysical modality should be derived from the semantic rules, not presupposed by them. Perhaps modal normativists can find some way to accommodate this issue and avoid circularity. But note that explaining how a subject could have such a concept of metaphysical necessity is no trivial task. Surely, she will translate the semantic rules into the corresponding explicit modal statements: ‘Water is necessarily H2O’, and ‘Necessarily, seals are mammals’. But, as mentioned, she might naturally take ‘necessarily’ to simply regulate correct linguistic usage, and so to express what the semantic rules dictate for all contexts of use that might actually occur, i.e., in our world. After all, we don’t typically teach and learn a language by assessing whether our utterances would be true or false on TwinEarth, or other sorts of possible worlds. In order for a subject to gain metaphysical modal knowledge from semantic mastery, she’ll further need to understand ‘necessarily’ not simply as a ‘must’ regulating correct linguistic usage, but rather as the ‘must’ of metaphysical necessity, capturing what it is for something to be true ‘absolutely’ or ‘no matter what’ or, in the usual terminology, ‘at all possible worlds’. Understanding

274  Antonella Mallozzi correctly the modal force of the statements translating the semantic rules is essential for gaining knowledge of metaphysical modality like (MN) holds. But that is substantive modal knowledge (knowledge regarding what metaphysical necessity captures) that’s presupposed by the account not explained by it. Lacking some grasp of the second-order rule (N), it would be rational for Patty to wonder whether water is necessarily H2O, or could have contained carbon; or whether seals are necessarily mammals, or could have been birds. And so on. Thus, modal normativists should address the following problem. Knowing the meta-rule that governs the modal behavior of semantic rules is required for knowing about metaphysical necessity and possibility but knowing the rule itself involves prior modal knowledge. Meta-Rules Governing Semantic Rules Let us assume that modal normativists can manage the circularity problem. Still, appealing to one’s second-order knowledge of the independence counterfactuals shows, against (MN)’s main thesis, that semantic competence (and possibly empirical information) isn’t strictly sufficient for knowledge of metaphysical modality. In addition, a subject would need to know meta-rule (N) governing our semantic rules, or otherwise grasp it, in order to formulate suitable independence counterfactuals and thereby reach the correct metaphysical modal judgments. Importantly, rule (N) isn’t itself something one can derive from one’s concepts, since it is a rule concerning how to properly apply our semantic rules. Furthermore, it doesn’t seem that (N) could be discovered empirically, either. What kind of empirical evidence could one possibly gather in support of a rule that establishes the validity of our semantic rules at all possible worlds? So, the modal normativist should explain how to accommodate the need for such metasemantic knowledge within an account that purports to explain knowledge of metaphysical modality solely in terms of mastery of our semantic rules (and possibly empirical information). The Justification of the Independence Conditionals A final problem for (MN) that derives from introducing the independence counterfactuals is that it is not clear on what grounds one should accept such conditionals. While Thomasson stresses that ‘indeed it is crucial that the modal normativist be able to accept this kind of independence counterfactuals’ as well as that she ‘can justifiedly accept independence conditionals to the effect that metaphysical modal truths aren’t contingent on our adoption of certain linguistic rules’ (2018: S2099), we are not told

Semantic Rules, Modal Knowledge, and Analyticity  275 on what grounds modal normativists have such an ability and what their justification for accepting such conditionals is. We saw that the conditionals themselves rest on a meta-rule, (N), which states that our semantic rules should be taken to hold at all possible worlds or necessarily. But that only pushes the question back to what, in turn, justifies (N). Why are we entitled to reason on the assumption that our rules won’t change across possible worlds? Indeed, within the normativist framework, it might seem false that the independence conditionals hold. Suppose that the actual world never contained any speakers or thinkers. Contrary to what (MN) claims, it would follow that, say, ‘seals’ and ‘mammals’ were never terms that came about to refer to seals and mammals, since there wouldn’t have been any language or thoughts to start with. How could it still be true on (MN)’s account that ‘Necessarily, seals are mammals’? Note that the descriptivist has a significant advantage here. While she is happy to endorse the independence counterfactuals, she doesn’t have a problem acknowledging the existence of modal facts and properties that are independent of any conceptualization and linguistic expression, and that support such conditionals. She can indeed justify endorsing the conditionals in a straightforward way, by simply appealing to what being a certain (kind of) thing is: it is just part of the nature or essence of seals that they are mammals, likewise it’s part of the nature or essence of water that it is H2O. That guarantees that ‘Necessarily, seals are mammals’ is true not just at worlds where individuals use those terms differently but even assuming there never had been any individuals to start with. On this view, modal facts and properties have nothing to do with the rules for using our terms, beyond the simple fact that our terms are inter alia aimed to correctly track and express such facts and properties. As we know, modal normativists explicitly reject descriptivism. So they owe us an explanation as to what justifies endorsing the independence conditionals. If, however, those conditionals are simply assumed to hold, that seems just arbitrary. The worry is that (MN) would be begging the question in its favor. Although the theory should demonstrate that the modal normativist can legitimately integrate such conditionals for reaching correct judgments about metaphysical modality, it rather presupposes that this is the case.

A Possible Way-Out: Modal Monism I have argued that semantic competence is not sufficient for knowledge of metaphysical modality. In particular, essentialist principles and information should be integrated in a successful account of how we know about metaphysical necessity and possibility. However, attempting to solve the problem by appealing to independence conditionals that fix the meaning of our terms at the actual world generates a host of other problems. It

276  Antonella Mallozzi raises a worry of vicious circularity; it shows that semantic competence isn’t strictly sufficient for knowledge of metaphysical modality; and, finally, it requires further justification, on pain of being question-begging. However, the modal normativist might contend that those criticisms miss the point of the normativist account. She might insist that one doesn’t need any ‘extra’ essentialist or other metaphysical knowledge on top of one’s conceptual competence in order to reach correct judgments of metaphysical modality. Likewise, one doesn’t need to integrate any second-order knowledge or meta-rules. Instead, one could formulate the relevant independence conditionals by simply mastering the rules themselves. It should be part of one’s ability to master the rules that ‘Necessarily, all seals are mammals’ derives from rules that should be taken to hold at all possible worlds, i.e., necessarily. Perhaps all there is to metaphysical necessity is fully captured by the explicit modal translation of the semantic rules. Put otherwise, the modal normativist might deny that there is a distinctive concept of metaphysical necessity, which is different in particular from logical-conceptual necessity. This is a familiar distinction in the modal metaphysics literature. Metaphysical modality concerns possibilities and necessities that roughly depend on the nature or identity of things and the laws of metaphysics (i.e., essentialist principles, relations of grounding, ontological dependence and laws of mereology, and so on). Logical-conceptual modality, however, concerns the possibilities and necessities that depend on the meaning of our terms or the rules for applying our concepts, while also respecting the laws of logic and the truth-preserving patterns of inference. Based on this distinction, we could recast our interpretation of (MN) in the following way. The object-language translations of the semantic rules strictly express matters of logical-conceptual necessity, whereas the meta-rule underlying the independence conditionals ensures that those statements are also metaphysically necessary. However, modal normativists might deny that there is a genuine distinction between logical-conceptual vs. metaphysical modality. They might instead join those philosophers who believe that at the level of worlds or propositions, the two modalities coincide—a thesis that’s called ‘Modal Monism’ (Chalmers 2010: ch. 6; Kment 2017). By identifying logical-conceptual modality and metaphysical modalities in this way, monists deny that there are two different and irreducible kinds of necessity. But how do they deal with the usual modal distinctions? On the one hand, it is widely acknowledged that propositions may be logically conceptually possible though not metaphysically possible. For example, it is logically conceptually possible that water isn’t H2O (that doesn’t imply a contradiction), although it is metaphysically impossible. On the other hand, logical-conceptual necessities might not be metaphysically necessary. We could slightly adapt Gareth Evans’ ‘Julius’ case (1979). If

Semantic Rules, Modal Knowledge, and Analyticity  277 the name ‘Julius’ refers rigidly to the person who is in fact the inventor of the zip, then ‘Julius (if he exists) invented the zip’ is logically conceptually necessary. But it’s also metaphysically contingent, since Julius could have become a car dealer, say, rather than an inventor. To account for those data, Monists will have to introduce a distinction at the level of meaning or content: sentences may come with different associated descriptions or express two different propositions having different modal status, necessary vs. contingent. That’s the core thesis of two-dimensional semantics. For example, in David Chalmers’ ­two-dimensional framework (2010), ‘Water is H2O’ is secondarily necessary but primarily contingent, whereas ‘Julius (if he exists) invented the zip’ is primarily necessary but secondarily contingent. Thus, Monists hold that all the data can be explained by a single notion of modality coupled with semantic distinctions. At the level of modal metaphysics, they maintain that logical-conceptual possibilities (and necessities) are also metaphysical possibilities (and necessities) (see, e.g., Chalmers 2010: ch. 6). It follows that if Modal Monism is true, no extra component besides conceptual mastery and rigorous reasoning (and sometimes empirical information) is needed for gaining knowledge of metaphysical modality. If we can gain knowledge of logical-conceptual necessity from opportunely translating the rules, we eo ipso gain knowledge of metaphysical necessity—which is (MN)’s desired outcome. But accepting Modal Monism comes with its own costs. I shall just mention two main issues Monists must face. First, while Monism simplifies the modal metaphysics, it requires that we endorse a complex semantic theory, i.e., two-dimensionalism, which many find highly controversial. (For example, it is doubtful that two-dimensionalism can provide a plausible analysis of names and natural kind terms. For a discussion, see Schroeter 2021.) Additionally, accepting Monism typically involves endorsing the view that what a possible world could look like is a matter of the descriptive content of our expressions. But language can hardly be the general source of the modal status of propositions, as Kripke’s cases of the necessary a posteriori have long shown. (For a discussion, see Chalmers 2010: ch. 6; Soames 2002; Vaidya 2008; Mallozzi 2021a,b.) Modal Monism might offer some insight into the modal metaphysics underlying (MN)’s account of knowledge of metaphysical modality. So, the modal normativist might want to clarify if she is in fact committed to such a view. If that’s the case, however, she would have to further explain how the account addresses Modal Monism’s notorious problems.

Modal Normativism and Knowledge of Analyticity In this final section, I aim to show how (MN) might still explain some modal knowledge, without committing to Modal Monism. Suppose Modal Dualism is true (i.e., roughly the thesis that there is a genuine

278  Antonella Mallozzi distinction between logical-conceptual vs. metaphysical m ­ odality). Modal normativists can at least explain how semantic mastery alone may suffice to gain knowledge of logical-conceptual modality. For on this view translating the semantic rules into the object language will strictly express matters of logical-conceptual necessity, and so a competent, rational thinker will be able to gain at least some modal knowledge by simply mastering the rules. Let us define the analytic truths as those sentences that may be known to be true solely through grasp of their meaning (this is what Boghossian 1996 calls ‘epistemic analyticity’).6 Knowledge of logical-conceptual necessity might then be taken to coincide with knowledge of analyticity thus defined. The thesis that semantic competence is sufficient for knowledge of analyticity might perhaps strike one as trivial. It’s not. Indeed, knowledge of analyticity has been extensively criticized by Williamson (2007: ch. 4). In what follows, I outline Williamson’s challenge against knowledge of analyticity and how it might affect (MN). While Thomasson’s own reply to Williamson (2015: 7.3) stresses the differences between their respective views and appeals to normative notions that the opponent might find unsatisfying, I show that (MN) can still successfully answer Williamson’s challenge on his own (descriptive) terms, as originally formulated. I do so by adapting Boghossian’s recent reply to Williamson (2020). I conclude that (MN) can still explain knowledge of analyticity or logical-conceptual modality. As Williamson characterizes it, epistemological analyticity is such that one’s understanding of a sentence guarantees that one assents to the sentence (2007: 73). For example, ‘Every vixen is a female fox’ is an analytic sentence in this sense since necessarily, whoever understands it assents to it. However, failure to assent to sentence s shows that one doesn’t understand s. Against epistemological analyticity, Williamson has raised the problem of ‘Competent Dissent’ (as Boghossian 2020 labeled it). Williamson argues that a fully competent native speaker of English might not assent to an elementary logical truth such as ‘Every vixen is a vixen’, while still understanding it. There might be various reasons for this. One might have somehow developed the conviction that vixens do not exist, while also holding that universal quantification is existentially committing. Or one might believe that there are borderline cases, such that it is neither true nor false that ‘Every vixen is a vixen’ (2007: 86–88). Indeed, the subjects in Williamson’s Competent Dissent scenarios are often experts in relevant fields—such as in the case of the distinguished logician Vann McGee, who has famously argued against modus ponens, thus refusing to assent to modus ponens while certainly understanding it (92). A further example involves philosophical expertise. Take, for example, sentence (KB): ‘It is necessary that whoever knows p believes p’. Williamson

Semantic Rules, Modal Knowledge, and Analyticity  279 remarks that ‘many philosophers, [and] native speakers of English, have denied [KB]. They are not usually or plausibly accused of failing to understand the words “know” or “believe”’ (168). Additionally—we will see the significance of this point shortly—according to Williamson, the subjects in question might not just refuse to assent to a certain sentence s, but even lose the disposition to assent to s while allegedly retaining full understanding of s. Williamson draws a provoking general conclusion to the question of ‘what is epistemically available simply on the basis of linguistic and conceptual competence. To a first approximation, the answer is: nothing’ (77). According to (MN), a subject may come to know modal truths solely based on her ability to correctly use and convey the semantic rules for our terms and what follows from them. Assuming, as it is plausible, that being able to use and express the semantic rules entails grasping their meaning, then the modal truths we may come to know on (MN) are (epistemic) analytic truths. But then (MN)’s account of modal knowledge (qua restricted to logical-conceptual necessity or analyticity) is also open to Williamson’s Competent Dissent problem. Assuming that understanding a sentence is a matter of normal semantic competence, possessing the relevant competence is all that is required for assenting to an analytic sentence. (MN) is concerned with knowledge not assent, as we have seen. But since knowledge is a stronger notion than assent, (MN)’s account should entail that mastering our semantic rules suffices for assenting to analytic truths. So, for example, on (MN) someone who masters the semantic rules for ‘vixen’ (and the other ingredient terms) should assent to analytic truth (or logical-conceptual necessity): (V) Necessarily, all vixens are female foxes In a Williamsonian spirit, one could argue that a competent speaker might not assent to (V), while mastering the rules for the ingredient terms and thus perfectly understanding (V). Thomasson is aware of the challenge and maintains that modal normativists can easily avoid it. What they need isn’t the (descriptive) claim that semantic competence entails a disposition to assent but rather the (normative) claim that semantic competence entails that one ought to assent to the relevant sentences and can be rebuked if one refuses. Semantic competence entitles one to the relevant conclusions (2015: 238–239). Additionally, a competent subject who deviates from the normal practice might have revisionary goals: she might be aiming to renegotiate the relevant semantic rules (243). But Williamson will likely find this normativist strategy unsatisfying. As he put it elsewhere, ‘if one ought to reason in some way, should not something deeper explain why one ought to reason in that way?’ (2003:

280  Antonella Mallozzi 291). What’s missing from the normativist story is an explanation of the entitlements in question and of our obligation to reason according to the relevant rules. Unless the normativist can integrate her account with that kind of justification, Williamson’s original challenge is still open. Fortunately, there is a response to the Competent Dissent problem that preserves Williamson’s non-normative setup in terms of one’s (disposition to) assent, which can be fruitfully adopted by (MN). Boghossian (2020) points out that in certain cases assenting to a sentence s is in fact constitutive of one’s understanding s—for example, in the case of logical constants, or of traditional analytic truths. In such cases, one is justified in assenting to s based on understanding, according to Boghossian, because assenting plays such a constitutive role for understanding. Thereby, in cases where a competent subject refuses to assent to s, she will at least retain the disposition to assent to s if she does still understand s. Boghossian takes up the example of conjunction: A natural description of Williamson’s expert, who develops theoretical misgivings about CE, is that she retains the disposition to assent to CE but refuses to act on that disposition as a result of the theoretical misgivings. As we may put it, she may continue to find CE primitively compelling, even as she now finds it derivatively uncompelling. (Boghossian 2020a p. 190) We saw that Williamson denies that competent speakers, indeed experts, need to retain even any disposition to assent. For him, they might well lose it altogether. However, Boghossian distinguishes two ways in which one might lose such a disposition and end up thinking that s is no longer plausible: (a) all things considered vs. (b) independently of any consideration. Williamson’s Competent Dissent scenarios are not really counterexamples to epistemological analyticity because they are instances of (a) not (b). So they strictly miss the target. To elaborate, in Williamson’s scenarios, experts reach their dissident conclusions based on lots of theoretical considerations, namely all things considered (a). That means that they lose their disposition to assent to s not solely based on their linguistic competence (or ‘primitively’), but in virtue of further considerations stemming from their expertise (or ‘derivatively’). In order to have a genuine counterexample to epistemological analyticity, the speaker should lose her disposition independently of any contribution from further considerations (b), while somehow still retaining full understanding of s. But cases of the latter sort are quite implausible. As Boghossian points out, we might legitimately doubt that a subject who loses her disposition to assent to, say, conjunction-­elimination for no particular reason would still count as understanding ‘and’ (Boghossian 2020b). Boghossian’s reply can in effect be cast as a dilemma for Williamson. Once it is clear that a counterexample to epistemological

Semantic Rules, Modal Knowledge, and Analyticity  281 analyticity should involve one’s losing her disposition to assent to s primitively or independently of any considerations, while also retaining her understanding of s, Williamson’s candidate cases either (i) don’t work (because they fail to satisfy the independence requirement) or (ii) are utterly implausible (because it is not clear that they satisfy the latter requirement that the subject still understands s). For similar reasons, I think that (MN)’s account of modal knowledge (again, as restricted to knowledge of analyticity or logical-conceptual necessity) is also safe from Williamson’s criticism. For speakers might well gain knowledge of analytic truths by rigorous reasoning solely based on their semantic competence. A competent speaker who loses her disposition to assent to (B) or (V) solely on that basis (or primitively), while still retaining normal reasoning capacities, no longer appears to understand those truths—somehow, she would have lost her mastery of the basic semantic rules for the ingredient terms.7 In conclusion, although (MN)’s account of modal knowledge faces several problems as an account of knowledge of metaphysical modality, it may still successfully explain knowledge of logical-conceptual modality or analyticity, while being safe from the sort of criticisms raised by Williamson’s Competent Dissent problem.8

Notes 1 Note that (MN)’s vocabulary seems neutral between the level of language and the level of thought. Its main tenets are cashed out in terms of semantic rules, concepts, and mastery, as well as linguistic rules and linguistic competence. Accordingly, my terminology is also meant to be neutral between the two levels. 2 Like Thomasson, I use ‘conceptual mastery’ and ‘conceptual competence’ interchangeably. 3 A further problem that I won’t pursue here is that requiring that one explicitly expresses the rules in order to gain modal knowledge seems to imply that there must be fixed rules for all cases. However, Wittgenstein’s commentary on rule-following gives reason to think that this is wrong. Some concepts may not have fixed rules all the way out; yet, one can have explicit modal knowledge (thanks to Anand Vaidya for raising this issue). 4 My argument here is reminiscent of Moore’s Open Question argument. Patty’s rational wondering shows that more than conceptual mastery is needed for metaphysical modal knowledge. Analogously, Boghossian has recently argued against the thesis that knowledge of normative truths can be explained by understanding alone, by pointing out that a competent thinker can always doubt whether, e.g., some candidate substantive characterization of a bad act-type is what actually plays the role indicated by an obvious definition of ‘wrong’. That shows that more than conceptual mastery is needed for substantive moral knowledge. Notably, Boghossian’s argument reaches the same conclusion as Moore’s Open Question argument, but through a different route (see Boghossian and Williamson 2020: Ch. 7; Boghossian 2021).

282  Antonella Mallozzi 5 The objection against conventionalism assumes that appealing to something contingent (i.e., a linguistic convention) to explain a metaphysical necessity fails because necessary truths are necessarily necessary, as per the S4 axiom of modal logic. Although most philosophers hold that S4 is correct for metaphysical modality, some have denied that (Salmon 1989; Vaidya 2008). 6 For Boghossian, an analytic sentence may be known to be true solely through grasp of its meaning, ‘provided that grasp of its meaning alone suffices for justified belief in its truth’ (1996: 363). As he explains, epistemic analyticity should be sharply distinguished from traditional metaphysical analyticity, according to which an analytic sentence is true purely by virtue of its meaning. Quine’s criticism of analyticity should be understood as undermining the stronger, metaphysical notion, whereas the epistemological notion can be preserved. (Plausibly, if the stronger notion is true, the weaker one is true a fortiori.) Note also that there are additional conceptions of analyticity. In Kant’s original definition, an analytic truth is one in which the predicate is already (covertly) contained in the subject (e.g., in the case of ‘All bachelors are unmarried’, the predicate ‘unmarried’ makes explicit what was already contained in the subject ‘bachelor’). Yet, another notion—I shall call it ‘essentialist analyticity’—takes analytic truths to hold in virtue of the identity or essence of the concepts (Fine 2005). Whether there might be additional notions, and whether they may all be interconnected, are interesting issues that I set aside here. 7 The strategy should be welcomed by normativists, as Thomasson seems open to the possibility of treating basic norms regarding acceptance (and rejection) as ‘constitutive norms for thought’ (2015: 240, fn. 8). 8 Many thanks to Paul Boghossian, Theodore Locke, Amie Thomasson, and Anand Vaidya for helpful comments on an earlier draft.

References Boghossian, Paul. 1996. Analyticity Reconsidered. Noûs 30, 360–391. Boghossian, Paul. 2020a. Intuition, Understanding, and the A Priori. In P. Boghossian and T. Williamson (Eds.), Debating the A Priori (pp. 186‒207). Oxford: Oxford University Press. Boghossian, Paul. 2020b. Reply to Williamson on Intuition, Understanding, and the A Priori. In P. Boghossian and T. Williamson (Eds.), Debating the A Priori (pp. 215‒226). Oxford: Oxford University Press. Boghossian, Paul. 2021. Normative Principles are Synthetic A Priori. Episteme 18(3), 367‒383. Boghossian, Paul and Williamson, Timothy. 2020. Debating the A Priori. Oxford: Oxford University Press. Chalmers, David. 2010. The Character of Consciousness. Oxford: Oxford ­University Press. Evans, Gareth. 1979. Reference and Contingency. The Monist 62: 161–189. Fine, Kit. 2005. The Varieties of Necessity. In K. Fine (Ed.), Modality and Tense: Philosophical Papers (pp. 235–260). Oxford: Oxford University Press. Kment, Boris. 2017. Varieties of Modality. Stanford Encyclopedia of Philosophy. https://plato.stanford.edu/archives/spr2017/entries/modality-varieties/. Mallozzi, Antonella. 2021a. Two Notions of Metaphysical Modality. Synthese 198, 1387‒1408.

Semantic Rules, Modal Knowledge, and Analyticity  283 Mallozzi, Antonella. 2021b. Putting Modal Metaphysics First. Synthese 198, 1937‒1956. Salmon, Nathan. 1989. The Logic of How Things Might Have Been. ­Philosophical Review 98(1), 3–34. Schroeter, Laura. 2021. Two-Dimensional Semantics, The Stanford E ­ ncyclopedia ofPhilosophy.https://plato.stanford.edu/archives/fall2021/entries/two-dimensionalsemantics/. Thomasson, Amie. 2007. Modal Normativism and the Methods of Metaphysics. Philosophical Topics, 35(1/2), 135–160. Thomasson, Amie. 2015. Ontology Made Easy. Oxford: Oxford University Press. Thomasson, Amie. 2018. How Can We Come to Know Metaphysical Modal Truths? Synthese 198, 2077‒2106. Thomasson, Amie. 2020. Norms and Necessity. Oxford: Oxford University Press. Vaidya, Anand J. 2008. Modal Rationalism and Modal Monism. Erkenntnis 68, 191–212. Williamson, Timothy. 2003. Blind Reasoning. Aristotelian Society ­S upplementary 77, 249–293. Williamson, Timothy. 2007. The Philosophy of Philosophy. Oxford: Blackwell.

14 Modal Knowledge and Modal Methodology Theodore Locke and Amie Thomasson

Introduction Many of the challenges and puzzles surrounding the ontology and ­epistemology of metaphysical modality stem from the assumption that modal claims—e.g., that it is metaphysically impossible for a statue to survive a drastic change in shape—are quasi-scientific claims attempting to describe or track modal features or possible worlds. Modal normativism argues against this descriptivist assumption and for the view that the function of modal claims is primarily normative. More specifically, normativism argues that the function of metaphysical modal claims is to mandate, convey, or renegotiate the semantic rules governing the terms involved in those claims, while staying in the object language, using rather than mentioning the relevant terms. By rejecting descriptivist assumptions, modal normativism demystifies the ontology of modality and offers a unified account of deontic and alethic modal claims, while also clarifying the methodology of modal metaphysics. Normativism shows how descriptivist assumptions make so many modal disputes feel endless and offers strategies for resolving those disputes in a way that is intellectually honest—for conceptual proposals are not passed off as objective metaphysical discoveries. But perhaps the most important advantage of modal normativism is the epistemological advantage of demystifying our knowledge of metaphysical necessities and possibilities. For on this view, we need nothing more than mastery of the relevant terms and concepts, ability to reason well, and ordinary empirical knowledge in order to know those metaphysical modal truths that there are. Antonella Mallozzi (this volume) casts doubt on this central advantage of modal normativism, arguing that the normativist instead must presuppose further metaphysical knowledge of “essentialist principles and information” to account for knowledge of metaphysical necessities. She also considers a couple of potential replies on behalf of the n ­ ormativist— one that appeals to our knowledge of independence conditionals, and another that embraces the “modal monist” view that denies a difference between conceptual and metaphysical modalities. She argues that both

DOI: 10.4324/9781003002192-16

Modal Knowledge and Modal Methodology  285 responses are problematic, though she ultimately allows that normativists might at least be able to account for knowledge of merely logical/­ conceptual modalities. In this paper, we aim to respond to these objections and to clarify the modal normativist’s treatment of how we can acquire modal knowledge. We begin by addressing the worry that being semantically competent and being empirically informed is not enough to explain knowledge of metaphysical necessities. Specifically, we will discuss how normativism can explain our knowledge of metaphysical necessities without appealing to talk of essences, but will also discuss how normativism can accommodate talk of essences and metaphysical laws, for those who would like to make use of these “post-modal” tools. Given the response here, there is no need for modal normativists to reach for either of the possible replies Mallozzi offers them. Nonetheless, “independence counterfactuals” do play a central role for modal normativists in showing why their view is not subject to the classic problems of conventionalism. So we will go on to clarify what justifies the modal normativist in accepting these independence counterfactuals. Similarly, it may be useful to clarify where the modal normativist stands on the issue of “modal monism”, so, we show how the normativist preserves a distinction between what are thought of as “conceptual” vs. “metaphysical” necessities, and yet can do so without abandoning their demystifying approach to modal knowledge. Next, we address concerns about conceptual analysis, and argue that modal normativism provides important insights into the methodology of modal metaphysics. Finally, we turn to discuss how our account of modality meets important constraints on modal epistemology.

Worries about Modal Knowledge One of the central advantages of modal normativism is its ability to give a clear view of how we can come to know metaphysical modal truths. For on this view, it takes nothing more than conceptual mastery, reasoning abilities, and (for some cases) empirical knowledge to come to know such metaphysical modal truths as there are.1 Mallozzi (this volume), however, denies this, arguing that this is not enough for metaphysical modal knowledge. In defense of this claim, she has us consider Patty the chemist, who is an expert on the use of chemical terms such as “water” and is informed about all of the relevant empirical information concerning water. Nonetheless, Mallozzi argues that this is not enough for Patty to know that: (W) Whatever microstructure the baptized sample has, water necessarily has that microstructure.

286  Theodore Locke and Amie Thomasson Nor is it enough for her to know the more specific modal truth that: ­ ecessarily, water is H2O. For, she suggests, even having mastered the n rules of use for the terms “water”, “microstructure”, “baptized sample”, etc., and even having all of the relevant empirical information, Patty might still reasonably wonder whether it is metaphysically possible that water contains carbon. Mallozzi suggests, instead, that Patty could rule this out and come to know that it is impossible for water to contain carbon only by using the essentialist bridge principle: (E) If it is essential to x being F that it is G, then necessarily anything that is F is G and the essentialist fact: (C) Having microstructure C is essential to being a certain kind of substance.2 But if we need knowledge of essentialist principles and facts, she argues, then we cannot maintain the normativist’s demystifying account of modal knowledge—instead, we need some distinctively metaphysical knowledge as well. Our response to this worry is twofold. First, we will try to show clearly how the normativist can account for the ability to acquire the relevant modal knowledge, and will also suggest how we should respond to the claims about the chemist Patty’s wonderings. Second, we will show that, even where we might need to know essentialist bridge principles and facts, these, too, can be given a normativist explanation, so that we might also explain Patty’s knowledge of (E) and (C) in terms of semantic competence, reasoning abilities, and empirical knowledge.

The Normativist Account of Modal Knowledge The trouble, as Mallozzi puts it, is supposed to be that “it isn’t clear what exactly in the semantic rules for “microstructure”, “baptized sample”, and “water” would justify conclusions concerning the metaphysical necessity of a substance’s microstructure” (this volume, 268–269)— enabling a competent speaker to know: (W) Whatever microstructure the baptized sample has, water necessarily has that microstructure. So what is it in the rules that would justify these conclusions? As discussed in Chapter 4 of Norms and Necessity (Thomasson 2020b), on the normativist view, what it is for a term to be a chemical kind term (that is: a term that purports to or is supposed to name a chemical kind, rather than to name a superficial observable kind, functional kind, artifactual

Modal Knowledge and Modal Methodology  287 kind…) is for it to be governed by a rule which, in the metalanguage, we could state roughly as follows: the relevant chemical kind term is to be applied only to whatever has the microstructure the relevant sample has, whatever that microstructure might be (Thomasson 2020b, 110). There are good reasons for having terms that follow these rules, since microstructures play a central unifying role in chemical theories and permit strong inductive inferences.3 Since Patty is a chemist, we can assume she has mastery of how chemical kind terms work. It is this mastery that puts her in a position to know (should she come to reflect on matters in these terms) that a chemical kind necessarily has whatever microstructure it actually has, and more specifically (given also her empirical knowledge that water actually has microstructure H2O) to know that water is necessarily H2O. Pace Mallozzi, Patty is entitled to infer both the general claim and her specific claim solely from her “mastery of the ingredient terms” (Mallozzi, 269), combined with her empirical knowledge. What then can we make of the claim that Patty might nonetheless wonder whether water could have contained carbon? First, it is worth noting that someone’s seriously wondering whether not-P does not show that they do not know that P, unless we also assume the controversial view that knowing something entails a kind of subjective feeling of certainty that would rule out further wondering. Still less does wondering whether not-P show that someone is not in a position to know that P, entitled to infer that P, etc. But can we respect Mallozzi’s intuition that Patty could nevertheless “legitimately wonder whether water is necessarily H2O or whether it could have contained carbon instead” (270)? It is hard to really imagine a chemist with mastery of chemical kind terms wondering this—other chemists might simply fail to understand what she is asking if (with presumed chemical competence) she asks whether water could contain carbon, or whether this (sample known to contain carbon) could be water. If a real chemist did raise these questions, we might think that she is implicitly shifting to thinking of “water” as a superficial kind term (“watery stuff”) or a functional kind term (“stuff we can drink, swim in, etc.”), or some combination of those. Or we might assume she has an epistemic understanding of modality in mind—that, for all we knew at one time, water could have turned out to contain carbon. Or she might be engaged in a metalinguistic bit of wondering—wondering how we ought to use the term “water”; if it ought to be treated as a chemical kind term, or (given its central role in our lives) as a superficial and/or functional kind term.4 Or finally, though she is entitled to the relevant modal inference, she might fail to make that inference (and so be left wondering whether water could have contained carbon) if she fails to master use of the relevant modal terminology (the introduction rules for “necessity”) or if she holds deviant philosophical beliefs about modality (not deviant chemical beliefs) that motivate her to resist the inference.

288  Theodore Locke and Amie Thomasson But none of these would give us reason to doubt the normativist’s claim that all one needs to acquire metaphysical modal knowledge is semantic mastery, reasoning capacities, and (often) empirical knowledge—and that given those, she is entitled to make explicit the rules of use for the chemical terms she has (ex hypothesi) mastered and come to know (W). No extra peering into deep, non-empirical, metaphysical facts of the world is required for this modal knowledge.

A Normativist Account of Essences But while normativists need not rely on some separate knowledge of essentialist facts and principles to account for our knowledge of metaphysical modal truths, it is nonetheless useful to be able to acknowledge that we may (for various purposes) want to make essentialist claims, and that we can acquire knowledge of them. For there are well-known reasons to think that speaking just of metaphysical necessity may be insufficient to say all we want to say in metaphysics.5 The normativist can agree that individuals and kinds have essences that metaphysically constitute or ground what those things are, how they might be, and how they must be.6 But the normativist will insist that we need to be careful with our talk of essences, constitution, and grounding, lest we reintroduce the same metaphysical and epistemological puzzles we started with.7 So, while the normativist can agree that justified beliefs about essences, constitution, metaphysical laws, or grounding are important aspects of our modal knowledge, we should not be led astray again by descriptivist assumptions into treating claims about essences and the like as if they were quasi-scientific claims tracking substantive features of the world.8 So, again, the question to begin with is: what are the functions of discourse about essences and metaphysical laws? What does that language do for us that we cannot do (so well) without it? Providing a full normativist account of essences and the like is beyond the scope of this paper, but there is a developed normativist approach to metaphysical explanations that gives an account of metaphysical laws, real definitions, and grounding.9 First, one normativist option is to say that metaphysical laws illustrate very general linguistic rules governing, say, a domain of proper names or terms that are important for some area of inquiry. For example, object language generalizations about natural kinds—e.g., generalizations provided in terms of inductively strong properties—function to illustrate rules governing all natural kind terms.10 In addition, on the view put forward in Locke (2020), knowing how to introduce the term, say “essential”, amounts to being able to make object language distinctions between metaphysical necessities (i.e., make hyperintensional distinctions), which amounts to being able to make important distinctions between semantic rules of use, such as which rules effectively govern a term and which do not. On the view

Modal Knowledge and Modal Methodology  289 put forward, a rule effectively governs the use of a term ­whenever it actually moves a person or community to conform their use of that term to the rule regardless of other possible collateral beliefs or linguistic plans (or actually moves a person or community to correct violations of the rule by others). For example, Smith, might be an eliminativist about numbers and so reject the sentence “Necessarily, if water is H 2O, then there are infinitely many primes”, but is still subject to semantic correction on independent grounds (i.e., regardless of their philosophical views about numbers) if Smith tries to apply “water” to an imagined sample that contains carbon. This is because Smith is still subject to certain effective rules governing the use of “water”, which are expressed in the object language using essentialist terms, i.e., water has its microstructure essentially. Overall, the normativist can agree that the essence of a chemical kind just is its microstructure, and so can agree that a chemical kind’s essence is independent of how we choose to describe or conceptualize it—­m icrostructures simply do not metaphysically depend on how we conceptualize the world. But the important question for the normativist is: why do we say that the microstructure is the kind’s essence? Why introduce talk of essences at all? Here, the normativist can agree with Mallozzi (2021b): we single out the microstructure and call it “essential” because the microstructure is shared by the many samples we have observed using straightforward empirical methods, and enables us to give a unified causal explanation of certain properties and behaviors of observed members of the kind, which means that we can then predict the properties and behaviors of members of the kind not yet observed. However, for the normativist, we introduce talk of essences here because it is useful to our scientific practices.11 It is certainly true that the correlations between microstructure and other properties and behaviors hold (to varying degrees) regardless of how we choose to conceptualize or talk about members of a given kind. But it is a mistake to think that there is anything metaphysically significant about that beyond our scientific practices, i.e., it is a mistake to think that we must appeal to “deep” facts about metaphysical fundamentality or naturalness to ground or further explain why the microstructure is important to scientific practice. Knowledge of an individual’s or kind’s essence does not require any special metaphysical insight. Like modal knowledge, it only requires conceptual mastery, reasoning abilities, and empirical knowledge. Given this brief sketch of a normativist account of essence and metaphysical laws, we can see that the normativist is free to accept bridge principles connecting those to metaphysical necessities. Specifically, the normativist can accept: (E) If it is essential to x being F that it is G, then necessarily anything that is F is G.

290  Theodore Locke and Amie Thomasson However, it is important to stress that, as for the essences of ­individuals and kinds, we need not discover any “deep” truths about the nature of metaphysical necessity in order to come to know these bridge principles, and we need not suggest that essences provide any substantive metaphysical ground for metaphysical necessities. For the normativist, these bridge principles reflect rules for how to appropriately apply modal terms given the successful use of essentialist terms—and knowing such general bridge principles requires nothing more than semantic mastery and reasoning. All in all, essentialist claims, for the normativist, are just ways of expressing finer-grained nuances on the appropriate use of non-modal terms. So, while it is an option for the normativist to argue that Patty can know (W) along the lines of the explanation provided earlier, which did not invoke essences or metaphysical laws, we have just shown how Patty can also come to know essentialist bridge principles and facts. She knows these essentialist claims in a manner similar to how she knows (W): by having the relevant empirical information, semantic mastery of non-modal terms, and semantic mastery of modal, essentialist, and grounding terms.12 But just as Patty’s considered beliefs about metaphysical necessity (rather than her considered beliefs about chemistry) might undermine her justified modal beliefs, Patty’s considered beliefs about essences and grounding might undermine her beliefs in claims like (C) and (E). Again, she might hold philosophical beliefs about essences and grounding (not chemical beliefs) that motivate her to resist certain essentialist inferences. If Patty has philosophical views whereby questions about grounding and essences cannot be answered by appeal to conceptual analysis and empirical inquiry alone, then, while Patty might in fact have sufficient justification for her beliefs about what is metaphysically possible, what is essential, what grounds what, etc., she has collateral beliefs that undermine that justification. So, the normativist can agree that Patty can be fully semantically competent and have the right empirical information, but still reasonably wonder about necessities, essences, grounding, and the like, given her other philosophical beliefs. Yet, none of this interferes with the modal normativist’s demystifying account of modal knowledge—on the contrary, as we have shown, this account can be extended to include a demystifying account of our knowledge of essentialist facts and principles.

Worries Surrounding Independence Conditionals So far, we have given reason to resist Mallozzi’s claim that modal normativists must appeal to a separate knowledge of essentialist bridge principles or facts in order to account for our knowledge of metaphysical modality. We have aimed to show not only how one can acquire knowledge of

Modal Knowledge and Modal Methodology  291 metaphysical necessities based on nothing more mysterious than semantic mastery, reasoning ability, and empirical knowledge, but also how a deflationary normativist account can be applied to knowledge of essences as well. The remainder of Mallozzi’s paper consists of suggesting and problematizing different potential lines of reply to the problem raised in the first section—by appealing to knowledge of independence counterfactuals or appealing to a form of “modal monism”. Since we have given a response (based on the original account in Thomasson 2020b) that appeals to neither, to that extent, we need not worry about whatever problems these appeals might bring. Nonetheless, it is worth discussing the points raised. The first, because modal normativists do need to appeal to independence counterfactuals not to buttress their account of modal knowledge, but rather to avoid the classic problems of conventionalism. The second, because it is worth clarifying the relation between modal normativism and the “modal monist” view Mallozzi discusses. The standard criticism of a “conventionalist” approach to modality (often paraphrased as saying “modal truths depend on our linguistic ­conventions”) is that it makes necessary truths hold merely contingently on our accepting linguistic conventions. So (the objection goes) we have to say that it is merely contingent that “Seals are necessarily mammals”, since it’s contingent that we adopt the relevant rules of use for “seal”. And so, the objection goes, the conventionalist has to accept that, if our linguistic conventions had been different (or if there had been no language), then it wouldn’t have been necessary that seals are mammals. However, Thomasson has argued that the modal normativist avoids this problem—by not taking modal claims to be attempted descriptions of semantic rules (which require the adoption of those rules as a truthmaker), but rather as object language formulations or mandates of such rules. And the rules, say, for “seal”, do not require for their application conditions that there be any language or speakers whatsoever in order for “seal” to apply.13 But they do require that there be something of the same biological kind as was ostended in the relevant sample. If that sample was a group of aquatic mammals, the rules require that the term only be reapplied to mammals (and not to fat tarpons on rocks). Another way to put this is to say that the normativist can accept certain independence counterfactuals, such as “seals are necessarily mammals, and this would be true even in situations in which speakers adopted different linguistic rules (or in which there were no speakers)” (for the full reply and discussion, see Thomasson 2020b, 88–90 and 170–172). Mallozzi’s worry is that in order to evaluate any metaphysical modal claims, speakers would have to know these independence counterfactuals, too—so that they know to keep our actual semantic rules fixed in

292  Theodore Locke and Amie Thomasson evaluating claims like “necessarily, seals are mammals”. But Mallozzi argues that this requires presupposing that speakers already have some modal knowledge—knowledge of what she labels as a meta-rule: (N) Actual semantic rules should be held fixed at all possible worlds; i.e. they should be taken to hold necessarily (p. 273, italics ours) But let’s look at this more closely. What Mallozzi calls “meta-rule N” and thinks speakers must have knowledge of is (on the normativist view) itself a metaphysicalized way of stating a norm: that (unless explicitly stated otherwise) speakers are to hold the language fixed in assessing counterfactuals. But this is not a requirement that speakers have knowledge of some metaphysical modal facts (say, that semantic rules hold in all possible worlds). The basic norm at issue is one we must master to be competent speakers in a public language: in engaging in conversation, or question and response of any kind, in order to preserve communication, keep the meanings of the terms fixed as far as possible (unless you explicitly signal otherwise, or unless the meaning of terms is what’s at issue, or unless you have other purposes that override the communicative one—e.g., humor or social change).14 This norm governs our engagement in normal question and answer situations, and also governs our evaluation of ordinary counterfactuals— we keep the language we are using in evaluating alternative scenarios fixed, unless explicitly stated otherwise. If we don’t, we get into all sorts of muddles. We assess counterfactuals in order to work out, and communicate with others about, what would be the case in alternative scenarios. To do this, we have to be able to use our extant language in stating, evaluating, and communicating about these scenarios. And that requires holding the meaning of a claim intact, as we evaluate (and communicate about) whether it would be true in another world. “If seals evolved gills, would seals still be mammals?”—interesting question. To meaningfully discuss and evaluate it, we better be sure that in the relevant world, the meaning of “seal” hasn’t changed to “brick” and “gills” to “potato”. To answer our original query and communicate with others about it, we have to be sure we do not shift meanings (rules of use) when we evaluate counterfactuals. The same applies when we ask, “If there were no languages and no speakers, would seals still be mammals?” This is an easier question (the answer is: yes!), but one again that we must answer using our extant language (even though we are considering a situation in which there is no language) and holding its meanings intact, asking (using that language) how to properly describe the counterfactual scenario.15 Another way to put matters is this: when we make basic modal claims, e.g., “seals are necessarily mammals”, we generally use the relevant terms rather than mentioning them. Similarly, when we characterize possible

Modal Knowledge and Modal Methodology  293 worlds—i.e., when we talk about and describe non-actual, hypothetical scenarios—we do so using the terms, not mentioning them, e.g., “in all possible worlds where seals exist, seals are mammals”. To talk about what norms govern the term “seal” at a non-actual possible world is to mention the term. Importantly, this is not to ask about seals at that world but to ask about the term “seal” and what linguistic practices (if any) are in force at that world rather than the actual world. But this is to hypothetically consider linguistic possibilities and not biological possibilities. So, when we express actual semantic rules in terms of possible worlds, we characteristically do so using the relevant terms, e.g., there is no possible world where seals are not mammals (which logically follows from: seals are necessarily mammals).16 Furthermore, it is worth stressing that the idea that we tend to hold meanings fixed, given a context of utterance, is a standard assumption in semantics. For example, suppose that it’s sunny in Montreal and raining in Seattle and that we are typing in Montreal. It is arguably false that “If we were in Seattle, then it would be raining here” when uttered in Montreal, because “here” uttered in Montreal picks out Montreal and not Seattle. Otherwise, we would need to shift the context of utterance, which fixes the meaning of the terms in the conditional, so that “here” would pick out Seattle instead. Kaplan (1977) calls such expressions monsters and argues that there are no monsters. In other words, in standard communication, we tend to evaluate truth at alternative circumstances using actual linguistic conventions.17 This is a standard (and useful) semantic assumption (and practice) rather than a claim about the meaning of specific terms such as “water” or “seal”. In short, when we evaluate independence conditionals, we hold the actual meaning of our terms fixed (perhaps along with certain empirical assumptions) no matter what actual or imagined cases we are considering. And the normative practice of holding such meanings fixed is justified by the purposes of communication generally, and more specifically by the purposes of uttering and discussing counterfactuals, which would be undermined if we simply let the meanings of our terms slip around as we describe the counterfactual scenarios. So, the modal normativist does have good reasons to endorse independence conditionals. Those reasons appeal to general communicative norms mastered by competent speakers. Once again, the basis for our knowledge of modalities traces to our mastery of normative practices—those involved in being competent speakers and reasoners—not to insight into some deep metaphysical facts. This is a matter of thoroughgoing consistency, not circularity. The normativist also can endorse the claim that it is necessary that necessarily seals are mammals—so that our necessary truths don’t hold merely contingently. The normativist’s reasons for accepting that what is necessary is necessarily necessary come from the rules of use for ­“necessary” (not by a claim or assumption that all semantic rules hold necessarily). For

294  Theodore Locke and Amie Thomasson given the introduction rule, (I), for “necessarily”, making one’s mastery of such rules of use explicit in the object language seems only to require that speakers understand that: if one has an expression of a semantic rule, p, then one has “Necessarily p” by (I), and so again one has an object language expression of an actual semantic rule, and so again, by (I), one can introduce “Necessarily, Necessarily p” (see Thomasson 2020b, 170–172).18 More generally, the normativist can argue that our knowledge of: (S4) If necessarily p, then necessarily, necessarily p is based on our mastery of the use of “necessarily, p”, i.e., our mastery of the introduction and elimination rules for modal terms. But, importantly, for the normativist, (S4) is not a description of the status of the rules that govern p, i.e., it should not be translated as “The semantic rules that govern the use of p are metaphysically necessary, i.e., hold across all possible worlds”. In short, the normativist does have good reason to endorse independence conditionals, and to endorse the corresponding claim of S4. And to do so does not require holding that speakers must know or assume that our semantic rules hold necessarily. It requires only mastering basic norms governing our communicative practices, along with mastering the rules of use for “necessarily”.

Modal Monism vs. Modal Dualism Mallozzi finally suggests that, in order to account for knowledge of metaphysical necessities, a modal normativist might resort to “modal monism” (though, she argues, this brings its own problems and limitations). Again, the normativist’s deflationary modal epistemology has no need to appeal to “modal monism” as Mallozzi understands it. Moreover, Thomasson (2020b, Chapter 4) provides an extensive discussion of how normativism can handle Kripkean necessary a posteriori propositions (Kripke 1980). So the normativist view cannot be undermined by suggesting that it relies on modal monism, and then arguing that modal monism fails to account for Kripkean necessary a posteriori truths. Nonetheless, it may be worth clarifying the relation between normativism and modal monism. Mallozzi characterizes modal monism as the view that denies that there is a distinction between logical/conceptual modality (understood as concerning “possibilities and necessities that depend on the meaning of our terms or the rules for applying our concepts,19 while also respecting the laws of logic and the truth-­preserving patterns of inference” (this volume, 276)) and metaphysical modality (understood as concerning “possibilities and necessities that roughly depend on the nature or identity of things and the laws of metaphysics (i.e., essentialist principles, relations of grounding, ontological dependence and laws of mereology, and so on)” (this volume, 276)).

Modal Knowledge and Modal Methodology  295 If this is how modal monism is understood, then the modal ­normativist will insist that it presents a false dichotomy. For the normativist resists thinking that metaphysical modality concerns possibilities or necessities that metaphysically depend on the “natures” of things or on essentialist or metaphysical principles.20 Nonetheless, the normativist also does not think that all metaphysical possibilities and necessities can be known simply through “conceptual mastery and rigorous reasoning” (Mallozzi, this volume, 277).21 Importantly, Thomasson (2020b) stresses the crucial role empirical knowledge plays in acquiring some kinds of modal knowledge. Once we acknowledge the central role empirical information plays in acquiring metaphysical modal knowledge, it is easy to see how the normativist can and does recognize a distinction between those sorts of modal claims that can be known merely through conceptual competence (made explicit) plus reasoning capacities—say, knowledge that it is necessary that squares have four sides or that bachelors are ­u nmarried—and modal claims that can only be known a posteriori. These latter Kripkean modal claims—say, that water is necessarily H2O or that Kamala Harris is necessarily human—are the central cases classified as “metaphysical necessities” as opposed to merely logical-conceptual necessities.22 Thus, modal normativism does acknowledge a distinction between what are typically thought of as (merely) conceptual modal truths and what are typically thought of as metaphysical modal truths.23 But that is not because the latter require some form of metaphysical discovery. Instead, the difference lies in whether one needs only conceptual and reasoning competence (including competence with rules that are conditionalized on states of the world or deferential to worldly facts to fill in schematic rules), or whether one also needs empirical knowledge, to know the relevant modal truths. Pace Mallozzi, the “desired outcome” for modal normativists isn’t to maintain that we can “gain knowledge of metaphysical necessity” from merely “opportunely translating the rules” (this volume, 277). Instead, the goal is to demystify our knowledge of modality by showing how we can come to acquire modal knowledge by nothing more mysterious than a combination of conceptual mastery, reasoning abilities, and ordinary empirical knowledge.

Worries about Knowledge of Conceptual Truths Mallozzi closes by offering a consolation to modal normativists—­ allowing that they can at least provide a path for knowledge of logical/ conceptual modal truths (though not, she alleges, of deeper metaphysical modal truths). As Mallozzi points out, some have been skeptical of the usefulness of conceptual analysis as a way of coming to know conceptual truths. For example, Timothy Williamson (2007) argues that someone

296  Theodore Locke and Amie Thomasson might understand the semantic rules for “vixen”, “all”, “if-then”, etc., and yet refuse to introduce or endorse the following: (V) Necessarily, all vixens are female foxes. For example, someone might refuse if they take all universal affirmatives to be existentially committing but also has the empirical belief that there are no vixens. Or perhaps someone takes the term “vixen” to be vague and the quantifier “all” and the implicit material conditional to be governed by a three-valued logic and so take the truth value of (V) to be indeterminate. Against this, and in defense of the limited account of modal knowledge she credits normativism with, Mallozzi argues that the normativist may make use of Paul Boghossian’s (2020) responses to Williamson. For the most part, the normativist can agree with the spirit of Boghossian’s response to Williamson’s proposed counterexamples: in cases where an expert dissenter counts as understanding (V) but nonetheless refuses to introduce or endorse (V), we need not say that the dissenter has lost all disposition to assent to (V), but only that their disposition is undermined by what the dissenter takes to be sophisticated reasons for rejecting (V).24 But there is an earlier and more distinctively normativist response to the Williamsonian objections to knowledge of analytic or conceptual truths, which is developed at length in Thomasson (2015). As Thomasson points out there, what we need is not the claim that semantic competence entails a disposition to assent, but that semantic competence entails that they ought to assent to the relevant claim and can be rebuked if they refuse—so that semantic competence (plus reasoning and empirical knowledge) gives us all we need to be entitled to the conclusions and to come to have modal knowledge.25 So, when someone refuses to assent to (V), they are subject to correction by other competent language users. This is especially true in cases where the refusal to assent to (V) is based on faulty empirical information, e.g., the person accepts a conspiracy theory that all vixens are secretly robots. Mallozzi says that the normativist response is “missing” an “explanation” of why we ought to reason according to the relevant rules—and she seems to be looking for a metaphysical “explanation”. But the normativist of course won’t reach for an epistemologically suspect “metaphysical explanation”, but rather for a pragmatic explanation. As Thomasson discusses (2020b, 169), it is easy to see pragmatic reasons speakers generally26 ought to follow the rules governing the natural language they present themselves as speaking. For only by doing so can they gain the benefits of being part of a community of users of a common language—enabling them to share information, engage in joint planning, and so on.27 In any case, the normativist also wants to acknowledge that someone can refuse to assent to (V) for another reason: because they are implicitly aiming to renegotiate accepted semantic rules, e.g., the semantic rules

Modal Knowledge and Modal Methodology  297 for universal quantifiers or conditionals. In this second case, the person is offering a revisionary semantic theory because they think that there are important considerations that mean we should change the standard semantic rules for logical terms.28 This part of the normativist response to Williamson’s cases highlights an important methodological insight of modal normativism. The semantic rules that govern the use of our terms are often open-ended or indeterminate. This means that we will often encounter cases that do not have a settled answer, given the current rules of use. For example, do I apply the term “Kamala” in an imagined case where Kamala Harris has used Parfit’s teletransportation machine? Maybe the actual semantic rules for “Kamala” are not of any help. Perhaps there is significant disagreement about whether we should focus on bodily or psychological continuity when answering important questions about legal and social responsibilities surrounding personhood. In these cases, it is up to speakers to negotiate or re-engineer the concepts to fit our needs. The difficult issues surrounding such a renegotiation might suggest that we should suspend belief about “Kamala could possibly survive the use of a teletransporter” until the semantic rules have been re-engineered or precisified. So, rather than thinking that there are deep, unresolved metaphysical mysteries lurking, we can approach questions of indeterminacy: in a more clear-headed fashion by recognizing that the nature of the problem is how to re-engineer and precisify our concept to serve a new purpose, or to serve an old purpose in a new situation. This kind of approach allows us to lay our cards on the table, and make public our reasons for choice in a way that disguising answers as “metaphysical discoveries” could not do.29 Acknowledging that meaning is often underdetermined and that semantic rules are open for renegotiation clarifies the methodology of modal metaphysics. Many modal disputes, e.g., about the nature of persons or the nature of death, have often felt endless because conceptual proposals are passed off as “metaphysical discoveries” that are subject to “counterexamples”, rather than noting that these are competing proposals based not on alleged metaphysical “discoveries” but rather reflecting competing goals and values. Normativism offers strategies for resolving those disputes in a way that is intellectually honest, for conceptual proposals are not passed off as objective metaphysical discoveries, but rather can be evaluated as the proposals they are.

Conclusion The problem of how we could come to know modal facts has been notorious for centuries. A chief advantage of the modal normativist approach is its

298  Theodore Locke and Amie Thomasson ability to demystify that knowledge—by showing how we can come to know those metaphysical modal facts that there are, via nothing more mysterious than conceptual mastery, reasoning abilities, and empirical knowledge. Like many serious metaphysicians, Mallozzi suspects that we cannot deflate modal knowledge in that way—and so thinks that the modal normativist must somewhere implicitly presuppose knowledge of deep metaphysical facts to justify our claim to know metaphysical modal facts. We have aimed to show that that is not so, by clarifying the ways in which modal normativists think we can come to acquire modal knowledge—and showing how the account can be extended to demystify our knowledge of essentialist principles and facts as well. We have also aimed to make it clear how the normativist is entitled to the independence counterfactuals needed to avoid the problems of conventionalism, and have aimed to show how normativists can respect the distinction between cases we think of as merely “conceptual” vs. “metaphysical” necessities—all without presupposing (an ability to know about) some more “metaphysically serious” form of modality. We hope this makes it clear that, however much serious metaphysicians may be accustomed to thinking of these things in metaphysicalized terms (and so tend to suspect a circle), there is no need for the modal normativist to think in these terms, or to appeal to some deeper sense of “metaphysical modality” (or knowledge of it) in order to capture crucial distinctions in varieties of modal claim, and in order to account for our knowledge of those metaphysical modal truths there are. Seen from a neutral perspective, this is good news—for it means that a demystifying account of modal knowledge and of the methodology of metaphysics remains available. And that was the goal of modal normativism. Those who, for some reason, still prefer to think of things in metaphysicalized terms are welcome to do so. But they owe us an account of why we should or must think of things in that way, and (most crucially) of how we can come to know such allegedly deep metaphysical facts, and how we can hope to resolve such modal debates in metaphysics.

Notes 1 For a full account of how the normativist handles modal epistemology, see Thomasson (2020b), chapter 7. 2 See also Mallozzi (2021b). 3 Locke (2020, 40). We will return to this idea below. 4 See Thomasson (2020b), Chapter 8 for a discussion of how certain modal questions and debates might be better understood metalinguistically. 5 See Fine (1994), Rosen (2015), Kment (2014), Nolan (2014) among others. 6 See Mallozzi (2021a, 1394). 7 Locke (2020). 8 Modal normativism does not require taking talk of essence, grounding, etc., seriously. The point being made here is that it is an available option, provided an account of that talk can be given that meets the ontological and epistemological constraints of modal normativism. This work is done in Locke (2018, 2019, 2020).

Modal Knowledge and Modal Methodology  299 9 Ibid. 10 These general semantic rules would be stated in a metalanguage, if stated at all. Importantly, these general rules are not linguistic or conceptual hypotheses but reflect broader norms often determined by the functions the relevant family of concepts or terms serve, e.g., permitting strong inductive inferences. 11 See Locke (2020) and Thomasson (2020a). 12 For example, following Fine (1994), this mastery would be reflected in accepting that “Necessarily, Socrates exists just in case the singleton {Socrates} exists” while rejecting “The existence of the singleton {Socrates} is essential to the existence of Socrates”. See Locke (2020) for a more detailed normativist explanation. 13 This puts them in contrast to the application conditions for “novel”, “money”, “government”, and many other social terms, the application conditions for which plausibly do require that there be language. 14 Of course in metalinguistic negotiations, this norm may be flouted in order to achieve other goals. (In some cases, it will involve not so much of a total shift of meaning as a renegotiation of how to understand or fix a vague or openended meaning.) 15 Of course, we can also ask questions about what would happen if we changed language in various ways (in fact, we might want to do this to evaluate different proposals in conceptual engineering: what if we were to change to using “woman” with Haslanger’s proposed meaning, for example? Would that change society in desirable ways?) But then to avoid confusion we had better be very careful in putting the relevant language in quotes to show that it’s being mentioned, and where we simply use the term we had better annotate to make it clear which use we are employing. 16 Of course, this doesn’t interfere with us, in special circumstances or given special needs, asking instead, supposing our language were different in the following ways, what could be truly said using the new language? 17 Kaplan (1977). See Einheuser (2006), Locke (2019), and Kocurek et al. (2020) for more recent engagement with this idea. This isn’t to say that the normativist is committed to the view that we always hold actual meanings fixed. Locke (2019) argues that semantic mastery of actual rules might require some tacit knowledge of the semantic rules for alternative languages. Such tacit knowledge of alternative languages is needed to evaluate metaphysical counterpossibles and does in fact reflect competence with terms in our actual language, e.g., knowing that the impossibility where Socrates has a different biological origin is more relevantly similar to the actual world than the impossibility where Socrates is the number 5 does in fact reflect an important kind of competence with our actual use of “Socrates”. 18 The rule (I) is: If p is an object language expression of an actual semantic rule (or a logical consequence of actual semantic rules), then you are entitled to introduce Necessarily p, regardless of any subjunctive supposition. 19 The normativist rejects this way of putting things—indeed, this is the standard uncharitable way of expressing classical conventionalism, which Thomasson (2020b, 27–31) argues against. For this talk of dependence is precisely what makes it sound as if we must reject the independence conditionals—and hold, wrongly, that if our linguistic or conceptual rules were different, then the modal facts would be, too. 20 Unless or except to the extent that such talk of natures, essences, and the like can itself be reinterpreted in deflationary normativist terms—see Locke (2020). 21 The normativist also would not describe logical or conceptual modality as “possibilities and necessities that depend on the meaning of our terms or the rules for applying our concepts while also respecting the laws of logic and

300  Theodore Locke and Amie Thomasson the truth-preserving patterns of inference” (Mallozzi this volume, 276, italics ours). For that would leave the view subject to the old problems of conventionalism that the normativist view is specifically designed to avoid. For discussion, please see Thomasson (2020b, 24–31 and 88–91). 22 See Thomasson (2020b), chapter 4. 23 Furthermore, the normativist is not necessarily committed to the view that there is a single space of worlds. Specifically, the normativist can retain a distinction between the space of epistemically, or logically conceptually, possible worlds and the space of metaphysically possible worlds. For example, it is open to the normativist to argue that while scenarios in which water is not H2O are in some sense epistemically possible, there is no sense in which they are metaphysically possible, because the appropriate use of the term “water” is always constrained by the microstructure of whatever sample was used in the actual baptism grounding use of the term “water”, whether we know what that composition is or not. Therefore, there are epistemically possible worlds that are metaphysically impossible and so not members of the space of metaphysically possible worlds. For more on normativist approaches to impossible worlds, see Locke (2019), which assumes modal normativism and argues that a scenario counts as metaphysically impossible when we must “misuse” our terms to describe it and goes on to show how the normativist can provide an account of non-trivial counterpossibles. 24 Boghossian and Williamson (2020, 191). 25 Thomasson (2015, 238–239). 26 Again, this is not to say that this general norm cannot be overridden in special cases—where a speaker aims for humorous, aesthetic, or social effects by flouting the rules. But for this to be successful, the interlocutors must see the speaker as generally mastering the rules but flouting them for effect on this occasion (otherwise, the speaker may just be ignored or dismissed as mistaken or incompetent). 27 If the question isn’t why speakers should (in general) follow the rules of the language they present themselves as speaking, but rather why (or whether) those particular rules should be as they are, then that, too is to be assessed pragmatically. See Thomasson (2020a). 28 Ibid: 243. 29 Thomasson (2020b, 199). See also Thomasson (2020a).

References Boghossian, P. and Williamson, T. (2020). Debating the A Priori. Oxford: Oxford University Press. Nolan, D. (2014). “Hyperintensional Metaphysics.” Philosophical Studies, 171(1), 149–160. Einheuser, I. (2006). “Counterconventional Conditionals.” Philosophical Studies, 127(3), 459–482. Fine, K. (1994). “Essence and Modality.” Philosophical Perspectives, 8, 1–16. Kocurek, A., Jerzak, E. and Rudolph, R. (2020). “Against Conventional Wisdom.” Philosophers’ Imprint 20(22), 1–27. Kaplan, D. (1977). “Demonstratives.” In J. Almog, J. Perry and H. Wettstein (eds.) Themes from Kaplan (pp. 481–563). New York: Oxford University Press. Kment, B. (2014). Modality and Explanatory Reasoning. Oxford: Oxford University Press.

Modal Knowledge and Modal Methodology  301 Kripke, S. (1980). Naming and Necessity. Oxford: Blackwell. Locke, T.D. (2018). Counterpossibles for Modal Normativists. Ph.D. dissertation, University of Miami, 2018. Locke, T.D. (2019). “Counterpossibles for Modal Normativists.” Synthese. 198(2), 1235–1257. https://doi.org/10.1007/s11229-019-02103-1. Locke, T. (2020). “Metaphysical Explanations for Modal Normativists.” Metaphysics, 3(1), 33–54. https://doi.org/10.5334/met.35. Mallozzi, A. (2021a). “Two Notions of Metaphysical Modality.” Synthese 198(S6), 1387–1408 https://doi.org/10.1007/s11229-018-1702-2. Mallozzi, A. (2021b). “Putting Modal Metaphysics First.” Synthese, 198(S8), 1937–1956. https://doi.org/10.1007/s11229-018-1828-2. Rosen, G. (2015). “Real Definition.” Analytic Philosophy, 56(3), 189–209. Thomasson, A. (2015). Ontology Made Easy. New York: Oxford University Press. Thomasson, A. (2020a). “A Pragmatic Method for Conceptual Ethics.” In H. Cappelen, D. Plunkett and A. Burgess (Eds.), Conceptual Engineering and Conceptual Ethics. Oxford: Oxford University Press. Thomasson, A. (2020b). Norms and Necessity. New York: Oxford University Press. Williamson, Timothy (2007). The Philosophy of Philosophy. Oxford: Blackwell.

15 Gettier’s Thought Experiments Joachim Horvath

Introduction Edmund L. Gettier III sadly passed away on March 23, 2021 at the age of 93.1 Gettier was most famous for his two-and-a-half-page classic ‘Is Justified True Belief Knowledge?’ (Gettier, 1963)—and for the fact that he barely published anything apart from that.2 Almost every philosopher knows that Gettier, in his much-cited gem, presented two ingenious counterexamples to the sufficiency of the justified-true-belief (JTB) analysis of knowledge that sparked an intense debate about the correct analysis of knowledge, which still continues today (see, e.g., Borges et al., 2017). It is somewhat less well-known that Gettier’s miniature also prompted another intense debate in recent metaphilosophy that concerns the method of hypothetical cases, which Gettier used to such great effect. Trying to reconstruct the reasoning behind Gettier’s method has proven to be almost as intractable as finding a better analysis of knowledge in light of his famous counterexamples. And while hypothetical cases with a similar structure can already be found in medieval (see, e.g., Hilpinen, 2017) and Indian philosophy (see, e.g., Matilal, 1986), no one had ever developed them with as much care and philosophical acumen before Gettier—and very few have matched him in this respect since then. In contrast, the ‘Gettier cases’ that are used in recent epistemology and metaphilosophy tend to be overly simplified versions of Gettier’s original cases that often leave out a lot of philosophically relevant details (see, e.g., Malmgren, 2011; Weinberg et al., 2001; Williamson, 2007). Against this trend, I will argue that a careful analysis of Gettier’s actual method can be a source of insight and inspiration even for contemporary accounts of Gettier’s thought experiments—although some of its aspects are also ‘merely historical’ or in need of improvement.3 More specifically, I will offer a suppositional reconstruction of Gettier’s thought experiments on the basis of Gettier’s (1963) original presentation. In doing so, I will take those aspects of Gettier’s method at face value that still seem defensible, but I will also fill a number of substantial gaps in Gettier’s argument on the basis of interpretative charity and the particular dialectical context of Gettier’s thought experiments. Moreover,

DOI: 10.4324/9781003002192-17

Gettier’s Thought Experiments  303 I will aim to avoid unnecessary technicalities in order to preserve the felt ­simplicity and fluency of Gettier’s method at the metalevel as well— which, however, requires some revisions of Gettier’s actual procedure (but so do other, more technical reconstructions). The basic contours of the proposed suppositional account are as follows: we first have to explicitly assume that Gettier’s cases are (metaphysically) possible, which is at best only implicit in Gettier’s own presentation; following Gettier’s (explicit) invitation, we then suppose his cases in hypothetical mode (in contrast to supposing them as actual or counterfactual); Gettier now argues, under the supposition, that the protagonist in each of his two cases does not know a certain proposition P, despite having a JTB that P; in order to draw Gettier’s envisaged (implicit) conclusion that JTB without knowledge (K) is possible, which is required to refute the sufficiency of JTB for K, we need to make the further assumption that Gettier’s suppositional case judgments are true because they are entailed by the supposition in question; in combination with the (metaphysical) possibility of his cases, this allows us to finally ‘export’ the (metaphysical) possibility of JTB without K from our supposition. The last step in particular has puzzled metaphilosophers enormously. For how exactly does one infer the possibility of JTB without K from judgments about merely hypothetical cases (see, e.g., Ichikawa and Jarvis, 2009; Malmgren, 2011; Williamson, 2007)? The now standard metaphilosophical answer is to offer a modal reconstruction of Gettier’s case judgments that facilitates a deductive inference to the conclusion that JTB without K is possible. For example, according to Williamson’s (2007, Chapter 6) highly influential proposal, the relevant sense of possibility is metaphysical possibility, and Gettier’s case judgments should be understood as counterfactual conditionals that—in combination with the metaphysical possibility of his cases—allow for a logically valid inference to the conclusion that JTB without K is metaphysically possible. The suppositional reconstruction of Gettier’s thought experiments to be developed in this chapter aims for a simpler and more straightforward solution which does not require any intermediate conditional step, and therefore comes much closer to how Gettier actually presents his thought experiments. In the final section, I will discuss potential advantages of my suppositional reconstruction over Williamson’s counterfactual approach in particular, but I will also note a number of commonalities.

A Suppositional Reconstruction of Gettier’s Thought Experiments Gettier (1963, p. 121) presents the JTB-analysis of knowledge in the standard lingo of ‘necessary and sufficient conditions’, and given that he introduces his two cases not as actual counterexamples, but as merely

304  Joachim Horvath supposed hypothetical cases, he must assume that his cases are possible in order to refute the sufficiency of the JTB-analysis. Likewise, Gettier must take the JTB-analysis to have a modal strength that goes beyond a mere generalization over actual cases of knowledge. For otherwise, it would be mysterious why merely possible counterexamples should even be relevant for his argument (‘All humans are mortal’, for example, cannot be refuted by the mere possibility of immortal humans). Most of this, however, is highly implicit in Gettier’s paper, and he gives us no real hint which sense of possibility and which precise modal strength he has in mind. Therefore, this is a clear case of a gap in Gettier’s argument that needs to be filled somehow. When Gettier wrote his paper in the early 1960s, the natural choice would have been to understand his argument in terms of conceptual or logical possibility (broadly understood). However, since Kripke (1980) explicitly distinguished metaphysical possibility from conceptual possibility and other modalities in his famous 1970 lectures Naming and Necessity, we cannot simply interpret Gettier according to the modal mindset of the 1960s, on pain of being philosophically naïve. For this reason, I will follow the recent metaphilosophical debate and assume, somewhat anachronistically, that metaphysical possibility is the relevant sense of possibility here, and that metaphysical necessity is the intended modal strength of the JTB-analysis (see, e.g., Ichikawa and Jarvis, 2009; Malmgren, 2011; Williamson, 2007).4 This interpretative choice is also justified by the fact that Gettier nowhere indicates that he wants to establish his conclusion only in the formal mode, for example, as a claim only about our concept of knowledge and related conceptual possibilities (see also Horvath, 2018). Moreover, even contemporary philosophers who prefer an understanding of thought experiments in terms of conceptual modality will typically argue that conceptual possibility entails metaphysical possibility in all or most relevant cases, including Gettier’s thought experiments (see, e.g., Chalmers, 2002). In any case, Gettier’s method is flexible enough to be used for exploring conceptual possibilities and—maybe—other modalities as well. However, the assumption that Gettier’s two cases are metaphysically possible requires that the way in which they are characterized expresses a truth-evaluable content, because metaphysical possibility is standardly understood in terms of truth in some metaphysically possible world. In this respect, Gettier’s use of arbitrary proper names, like ‘Smith’, ‘Jones’, and ‘Brown’, raises some tricky philosophical issues. It seems clear that Gettier, in using these proper names, has no actual Smith, Jones, or Brown in mind. For it is widely known that these names are among the most common English last names, and so if Gettier really had some actual people in mind here, he should have given us a lot more information to single them out as specific individuals. Instead, Gettier’s highly generic use of ‘Smith’, etc., suggests that these names are rather intended as arbitrary names of some sort.

Gettier’s Thought Experiments  305 As it turns out, prominent accounts of the semantics of arbitrary names are all extremely contentious (see, e.g., Cappelen and Dever, 2018, Chapter 6). For example, Fine (1983) postulates a kind of objects in their own right, arbitrary objects, as the referents of arbitrary names, while Breckenridge and Magidor (2012) claim that an arbitrary name refers to a specific ordinary object, but we do not and cannot know which object it is. So, if we hesitate to rely on such contentious views about arbitrary names for the primarily methodological purpose of doing and reconstructing thought experiments, then all the remaining alternatives for obtaining truth-evaluable versions of Gettier’s cases are to some extent revisionary. As metaphilosophers, it therefore seems that we cannot just take Gettier’s and other’s use of arbitrary names in thought experiments at face value.5 For example, Williamson’s (2007, p. 184) revisionary interpretation of Gettier’s ‘Smith’, ‘Jones’, and ‘Brown’ is to understand them as ‘picturesque substitutes for variables’. To actually replace these names with variables and quantifiers would, however, make the thought experiment a lot more cumbersome and difficult to follow—in striking contrast to Gettier’s natural and easy-to-follow rendering with arbitrary names. Moreover, representing the content of Gettier’s cases and our related reasoning in terms of quantifiers and variables gives rise to tricky technical issues of its own, such as the problem of donkey anaphora (as in, e.g., ‘If a farmer has a donkey, then he beats it’). This problem arises because we want to ‘keep track’ of one and the same individual, Smith, when we reason about Gettier’s cases, which is not easy to capture in terms of quantifiers and variables (cf. Williamson, 2007, pp. 195–199, Appendix 2). For these reasons, I propose another revisionary treatment of this issue, the substitution gambit, which both avoids distracting technicalities and preserves the ‘cognitive fluency’ of Gettier’s use of arbitrary names. The basic idea is as follows: instead of mulling over the semantics of arbitrary names, we can simply resolve the issue by replacing the likes of ‘Smith’ and ‘Jones’ with ordinary proper names of arbitrary people. For example, we could replace ‘Smith’ with ‘Edmund Gettier’, ‘Jones’ with ‘Keith Lehrer’, and ‘Brown’ with ‘Alvin Plantinga’—and Gettier’s thought experiments would work just as well, and would be just as easy to process as in the original version with arbitrary names.6 In addition, we now have ordinary propositional contents for Gettier’s cases that are nicely truth-evaluable.7 What potential problems could there be with this solution, that is, how risky is the substitution gambit after all? Now, we cannot rule out for sure that we might run into trouble or counterintuitive consequences for some choice of ordinary proper names, especially if we ‘import’ a lot of facts about the relevant people into our supposition. However, the substitution gambit gives us all the flexibility in the world to select the relevant ordinary names in such a way that they best serve the philosophical purpose of the thought experiment in question. Moreover, such worries are less pertinent for the mode of supposing—supposing-as-hypothetical—that

306  Joachim Horvath is the most natural mode of supposing for Gettier’s thought experiments, as I will argue below. For the remainder of this chapter, I will therefore assume that the arbitrary names in Gettier’s cases are, or can be, (tacitly) replaced with ordinary proper names in the way suggested above. However, for ease of presentation, I will mostly continue with Gettier’s arbitrary names (unless the difference really matters). Let us now consider Gettier’s crucial step of introducing two hypothetical cases ‘in which the conditions stated in [the JTB-analysis] are true for some proposition, though it is at the same time false that the person in question knows that proposition’ (p. 122).8 Gettier clearly intends these hypothetical cases to be considered in suppositional mode, for he repeatedly uses formulations like ‘suppose that’ or ‘let us suppose that’ throughout his presentation (sometimes interchangeably with ‘imagine that’). What is not so clear, however, is how Gettier wants his cases to be supposed. To shed more light on this issue, let us first consider two wellknown modes of supposition, namely, supposing as actual and supposing as counterfactual (compare Adams, 1970). First, when we suppose something as actual, we add our supposition as a further piece of hypothetical knowledge to our existing beliefs and knowledge about the world, and we hypothetically suspend or adjust those beliefs that are in conflict with our hypothetical new piece of knowledge—roughly corresponding to the way in which we evaluate indicative conditionals (cf. Ramsey, 1931). For example, if we suppose-as-actual that Oswald did not kill Kennedy, we hypothetically adjust our Oswald-­ related beliefs, yet still hold fixed our knowledge that Kennedy was killed by someone, and therefore judge, under the supposition, that someone else must have killed him. There is no indication that Gettier wants us to suppose his cases in this primarily epistemic sense of supposing something as actual. Also, this mode of supposing would at best help to establish the epistemic possibility of JTB without K, which is too weak if the modal strength of the JTB-analysis is taken to be metaphysical necessity. Second, when we suppose something as counterfactual, we suppose that things basically are as in the actual world, except for those facts that need to be changed or adjusted in order to make room for the hypothetical truth of our supposition—roughly corresponding to the way in which we evaluate counterfactual conditionals (cf. Lewis, 1973; Stalnaker, 1968). For example, if we suppose-as-counterfactual that Oswald did not kill Kennedy, we make room for the hypothetical fact that Oswald did not kill Kennedy and its downstream consequences, while holding fixed other, unrelated facts about the actual world, and so we judge, under the supposition, that Kennedy was not killed at all (unless we assume, for example, the existence of a ‘backup-killer’).9 Should we suppose Gettier’s cases as counterfactual? Let us again consider the instructive example of Oswald and Kennedy. In a counterfactual

Gettier’s Thought Experiments  307 supposition of ‘Oswald did not kill Kennedy’, we judge that Kennedy was not killed at all on that tragic day in Dallas, because Oswald was a lone perpetrator. However, if we should happen to learn from historians that there were in fact several backup-killers who were ready to step in for Oswald in case of his failure to kill Kennedy, then we would rather judge that Kennedy was killed on that day even under the counterfactual supposition that Oswald did not do it. Thus, it is a hallmark of supposing-as-counterfactual that our relevant suppositional judgments depend on various facts about the actual world, often of a highly particular kind (such as the existence of backup-killers), that are not explicitly mentioned in the counterfactual supposition at issue, but that we nevertheless implicitly assume or ‘import’ when we make the relevant suppositional judgments. However, we do not find this kind of dependence on actual facts in Gettier’s thought experiments, and so supposing-as-counterfactual is not the most plausible mode of supposing for these thought experiments. For example, assume that we learn that there is an actual person, called ‘Smith’, who perfectly satisfies Gettier’s first case description, and that we also learn that the president of the company who ‘assured [Smith] that Jones would in the end be selected [for the job]’ (Gettier, 1963, p. 122) is widely known, including by the actual Smith, to be a pathological liar. In a counterfactual supposition, we should now be ready to judge that Smith’s belief that he will get the job is unjustified. However, this would disqualify the case for Gettier’s purpose of refuting the JTB-analysis, which requires that Smith has a JTB.10 But in the context of a Gettier thought experiment, we are not ready to change our judgments about Gettier’s Smith in light of what is going on with some actual Smith who happens to satisfy Gettier’s case description (cf. Ichikawa, 2009; Ichikawa and Jarvis, 2009; Malmgren, 2011). Rather, we would dismiss such (empirical) facts about the actual world as irrelevant to Gettier’s hypothetical case, and instead focus on the case as he actually describes it, which does not include, for example, that the president of the company is a pathological liar.11 Fortunately, there is a more fitting mode of supposition that I will call supposing-as-hypothetical, which also captures the common characterization of thought experiment cases as hypothetical cases.12 In contrast to the previous two modes of supposing, when we suppose something as hypothetical, we only hold fixed the explicit propositional content of what we suppose (plus things that are clearly semantically implicated or presupposed by that content). For example, if we suppose-as-hypothetical that Oswald did not kill Kennedy, we merely hold fixed this particular content—and nothing else beyond that (except semantic implicatures and presuppositions). It immediately follows that there is no good answer to the question ‘Did someone else kill Kennedy?’, because this issue is underdetermined by the mere fact that Oswald did not kill Kennedy (see Table 15.1 for a comparison of the three modes of supposition in

308  Joachim Horvath light of this example). The only judgments we can reasonably make if we ­suppose-as-hypothetical that Oswald did not kill Kennedy are relatively trivial things like: there is someone who did not kill Kennedy, there is someone who was not killed by Oswald, and so on. The reason why we cannot go beyond near-trivial judgments in this toy example is that the content of our supposition is relatively simple—unlike in philosophical thought experiments, which usually characterize a particular scenario in some detail. However, when it comes to philosophical thought experiments, the ‘sparseness’ of supposing-as-hypothetical is also a strength because this mode of supposing avoids the above-mentioned problem with supposing-as-counterfactual, but it can still help us to establish conclusions about metaphysical possibilities (in a way to be explained below). In contrast to counterfactual supposition, facts about actual people are now irrelevant to the assessment of Gettier’s protagonists’ beliefs in terms of knowledge because only the explicit content of our supposition matters. This nicely explains why facts about actual people who happen to satisfy Gettier’s case descriptions strike us as irrelevant to his thought experiments, and why we are therefore not prepared to revise our suppositional case judgments in light of such (empirical) facts about the actual world (unlike in the case of Oswald and Kennedy). My discussion still raises a pressing question about the truth-­conditions of suppositional judgments in hypothetical mode. So far, I have simply made the assumption that suppositional judgments about Gettier’s cases can be true or false (or indeterminate), which is clearly shared by Gettier himself (1963, p. 122; see quote above) and by most other philosophers dealing with his cases. For example, most philosophers would say that someone who judges that the Smith in Gettier’s cases knows the relevant proposition is somehow mistaken or misjudges these cases (see, e.g., Williamson, 2020, p. 236). Genuine disagreement of this kind, however, requires objective truth-conditions, or at least objective correctness-­ conditions of some sort, for the suppositional judgments in question. Moreover, philosophers typically argue for their suppositional judgments about thought experiment cases (cf. Cappelen, 2012; Deutsch, 2015; ­Horvath, 2022), and Gettier does so as well, even if only quite briefly (in Table 15.1  T  ruth-value differences for the suppositional judgment ‘Someone else killed Kennedy’ in the three modes of supposing as actual, counterfactual, and hypothetical Supposing as …

Supposed ‘case’

‘Someone else killed Kennedy’

… actual … counterfactual … hypothetical

Oswald did not kill Kennedy Oswald did not kill Kennedy Oswald did not kill Kennedy

true false indeterminate

Gettier’s Thought Experiments  309 his already very short paper). For example, concerning his second case, Gettier (1963, p. 123) highlights ‘the sheerest coincidence’ of the truth of Smith’s belief as the crucial reason why Smith’s belief falls short of knowledge (i.e., Gettier appeals to a form of epistemic luck here). It is a platitude, however, that arguing for a judgment J typically just is arguing for the truth of J, and it would be surprising if arguing had a very different point or function in the case of suppositional judgments. So, let us then ask: when Gettier judges (about his second case) that ‘Smith does not know that (h) [= “Either Jones owns a Ford, or Brown is in Barcelona”]’ (Gettier, 1963, p. 123), what actually fixes the truth of this judgment? Given that we are considering the case in hypothetical mode, the truth of Gettier’s suppositional judgment can only be fixed by the explicit content of our supposition of the case (and its semantic implications or presuppositions). The natural suggestion therefore is that Gettier’s suppositional judgment is true iff it is entailed by the relevant hypothetical supposition of his second case (see also Barnett, 2010, pp. 277–279).13 Given the initial assumption that Gettier’s thought experiments aim at metaphysical possibility (see above), we may further assume that the relevant kind of entailment is metaphysical entailment.14 However, it should be emphasized that the proposed view about the truth-conditions of suppositional judgments does not require that the metaphysical entailments in question are known or even explicitly endorsed by philosophers (or other people) who justifiedly make these judgments. From an epistemological point of view, it is only required that one’s suppositional judgments are somehow sensitive to the relevant entailments, if only by means of fallible heuristics that track these entailments reasonably well. For example, the fallible but often reliable connection between conceptual entailment and metaphysical entailment seems to work just fine for Gettier’s cases, even if it may break down elsewhere, for example, in the case of suppositional judgments about natural kinds, like water or gold (see Horvath, 2015; 2016, for discussion). The fact that no explicit appreciation of the incompatibility of epistemic luck with knowledge is epistemically required for successfully performing Gettier’s thought experiments is an important aspect of Gettier’s method, which makes it epistemically less demanding than, for example, extant reconstructions of Gettier’s case judgments in terms of strict conditionals (see below for further discussion). Let us now ask which suppositional judgment(s) must be true in a hypothetical supposition of Gettier’s cases for his overall argument to succeed? What Gettier ultimately needs is the existentially generalized judgment ‘there is JTB without K’. For, in order to be genuine counterexamples to the sufficiency of JTB for K, it must be true in Gettier’s cases, supposed as hypothetical, that there is JTB without (propositional) knowledge. Under a supposition of Gettier’s second case, for example, the intermediate steps that lead to this conclusion are ‘Smith has JTB

310  Joachim Horvath that (h)’ and ‘Smith does not know that (h)’, which entail that ‘Smith has JTB that (h) without knowledge that (h)’ and, in a final step—by existential generalization—that ‘there is someone who has a justified true belief without (propositional) knowledge’ (which we may conveniently abbreviate as ‘there is JTB without K’). So, we need to ask whether it is true in a hypothetical supposition of Gettier’s second case both that (1) ‘Smith has JTB that (h)’, and that (2) ‘Smith lacks K that (h)’—which translates into the question whether (1) and (2) are metaphysically entailed by a hypothetical supposition of Gettier’s second case (mutatis mutandis for Gettier’s first case). So, we need to ask whether (1) and (2) are true in all metaphysically possible instantiations of Gettier’s second case, that is, whether there is no such instantiation on which (1) or (2) come out false. In other words, we want to know whether there are deviant realizations of Gettier’s second case, such that the metaphysical entailment of (1) or (2) is falsified by them. This is nothing but the problem of deviant realizations, which is one of the driving forces behind the recent metaphilosophical debate about thought experiments, and which originates from Williamson’s (2007, Chapter 6) The Philosophy of Philosophy (see below). Thus, we need to consider the following question: what basic strategies are there to ensure that a supposition of a thought experiment case, such as Gettier’s second case, entails the intended suppositional judgments and thereby avoids the problem of deviant realizations? We can distinguish top-down and bottom-up strategies here. The top-down strategy, suggested by Grundmann and Horvath (2014a,b), is to explicitly set up the case in such a way that ‘first, […] Smith really has the relevant justified true belief, and second, […] that Smith does not know the relevant proposition in some other way’ (Grundmann and Horvath, 2014b, p. 528). This requires us to (A) stipulate that Smith has JTB that (h), to (B) include some feature that prevents Smith from knowing that (h) in the described way, and to (C) stipulate that Smith does not know that (h) in some alternative way. Below, I will argue that Gettier applies all three measures (A), (B), and (C) of the top-down strategy to some extent, but that he also does not go far enough in this respect, and therefore fails to exclude some very far-fetched deviant realizations. For example, what if Smith’s future self has traveled back in time and forms the belief that (h) at the very same moment when Smith comes to believe that (h), but in a way that actually constitutes knowledge? It would then be true in the supposed situation that Smith knows that (h), simply because his future, time-traveling self knows that (h). This is a bizarre development of Gettier’s second case, for sure, but it seems hard to deny that it is at least metaphysically possible. How fatal this is depends on the availability of a bottom-up strategy in order to supplement Gettier’s incomplete use of the top-down strategy. The bottom-up strategy is to marshal broadly inductive reasons why the remaining deviant realizations do not threaten the philosophical

Gettier’s Thought Experiments  311 purpose of the thought experiment at issue. More specifically, the idea is that, given that all deviant realizations considered so far are dialectically non-threatening to Gettier’s overall argument, we have good reason to believe that there is a properly amended version of, for example, Gettier’s second case such that a hypothetical supposition of this amended version actually entails (1), (2), and the crucial claim that ‘there is JTB without K’. Now, what does it mean for a deviant realization to be dialectically non-threatening? Roughly, it means that once a deviant realization has been made dialectically salient, it can be ‘stipulated away’ by adding further clauses to one’s supposition that do not compromise its metaphysical possibility or the intended case judgment. As far as I can see, all deviant realizations of Gettier’s cases that have been proposed so far are of this dialectically non-threatening kind. For example, consider the deviant realization of a time-traveling Smith suggested above: this can be easily stipulated away by explicitly adding it to our supposition that Smith does not travel in time—without casting any doubt on the metaphysical possibility of the case, or on Smith’s having JTB without K. In fact, this is exactly how most philosophers would react if someone brought up the possibility of a time-traveling Smith as a problem for Gettier’s case judgment (and it is also how we respond to our undergraduate students when they bring up dialectically irrelevant possibilities on their first encounters with Gettier’s cases). Of course, due to the inductive character of the bottom-up strategy, a certain epistemic risk remains that there are still dialectically threatening deviant realizations ‘out there’ that cannot be stipulated away. However, this is the same epistemic risk that basically affects inductive judgments in general—from ‘all swans are white’ to ‘all electrons are negatively charged’. For sure, little of this can be found in Gettier’s short paper, even on a very charitable interpretation. The present proposal to use the top-down and/or bottom-up strategy for dealing with the problem of deviant realizations is thus clearly a revisionary amendment to Gettier’s actual method. But it is hard to see how one could successfully deal with the problem of deviant realizations in a way that does not go substantially beyond Gettier’s own presentation. However, what matters from a metaphilosophical perspective is that the bottom-up strategy, as well as a full-fledged version of the top-down strategy, is available to us now, as present-day performers of Gettier’s thought experiments. To that extent, they can also vindicate Gettier’s actual method in retrospect—as a method that is basically sound, although incomplete and in need of further refinement. To complete my discussion of the problem of deviant realizations, I will now discuss to what extent Gettier can be seen as applying the top-down strategy for avoiding deviant realizations, with its three measures (A), (B), and (C) sketched above. As will become clear, Gettier’s own development of his cases—in particular of his second case—displays a much higher sensitivity to this problem than many subsequent presentations of

312  Joachim Horvath ‘Gettier cases’ in epistemology and metaphilosophy. In this respect, Gettier’s short paper can still be seen as an exemplary—albeit i­mperfect— development of the thought experiments in question. Let us begin with measure (A) of the top-down strategy, which is to stipulate that Smith has JTB that (h). Gettier arguably stipulates or near-stipulates all three of the conditions J(ustification), T(ruth), and B(elief), and thus comes close to a full implementation of this (A)-part of the top-down approach to deviant realizations. First, what about the T(ruth) of the target proposition (h) [= ‘Either Jones owns a Ford, or Brown is in Barcelona’] in the supposed case? Gettier (1963, p. 123) effectively stipulates that this proposition is true, by saying that ‘the place mentioned in proposition (h) happens really to be the place where Brown is’, which entails that Brown is in Barcelona and, as a consequence, that the proposition (h) is true (with the help of the truth-conditions for disjunction and the T-schema: ‘p’ is true iff p). Second, for an analysis of Gettier’s top-down strategy for B(elief), let us first consider the following key passage: Imagine that Smith realizes the entailment of each of [the] propositions he has constructed by (f) [= ‘Jones owns a Ford’], and proceeds to accept (g) [= ‘Either Jones owns a Ford, or Brown is in Boston’], (h) [= ‘Either Jones owns a Ford, or Brown is in Barcelona’], and (i) [= ‘Either Jones owns a Ford, or Brown is in Brest-Litovsk’] on the basis of (f). Smith has correctly inferred (g), (h), and (i) from a proposition for which he has strong evidence. Smith is therefore completely justified in believing each of these three propositions. Gettier (1963, p. 123) Here, Gettier stipulates that ‘Smith […] proceeds to accept (g), (h), and (i) on the basis of (f)’, and the surrounding context makes it quite clear that Gettier treats ‘proceed to accept’ as synonymous with—or at least as entailing—‘come to believe’, because after the following sentence, he simply continues with ‘Smith is therefore completely justified in believing each of these three propositions’.15 Third, Gettier’s treatment of J(ustification) amounts to the ­near-stipulation that Smith’s true belief that (h) is also justified. How so? Gettier (1963, pp. 122–123) first stipulates that Smith has ‘strong evidence’ for ‘Jones owns a Ford’ at the very beginning of the case, and later continues with ‘Smith has correctly inferred (g), (h), and (i) from a proposition [= (f) = “Jones owns a Ford”] for which he has strong evidence. Smith is therefore completely justified in believing each of these three propositions’. With respect to Smith’s critical inference from (f) to (h), Gettier (1963, p. 121) explicitly assumes that deductive inferences of this kind are justification-preserving, from which it follows that Smith’s belief that (h) is justified if his belief that (f) is justified. This only leaves

Gettier’s Thought Experiments  313 the small step from ‘Smith has strong evidence for his belief that (f)’ to ‘Smith’s belief that (f) is justified’. In the specific dialectical context of Gettier’s thought experiment, the most charitable reading of ‘strong evidence’ seems to be in terms of Smith’s total evidence concerning ‘Jones owns a Ford’, which plausibly entails that Smith’s belief that ‘Jones owns a Ford’ is also justified. On an alternative reading of ‘strong evidence’, for example, as ‘some evidence that is strong but possibly other evidence as well’, Gettier would actually leave it open whether Smith’s belief that (f) is really justified. But given that Gettier has the explicit dialectical aim of refuting the JTB-analysis of knowledge, an alternative reading of this kind would be quite uncharitable, especially because a natural reading of ‘strong evidence’ that does not run counter to Gettier’s dialectical purpose is easily available here. Moreover, Gettier seems to presuppose that Smith’s belief that (f) is justified when he applies his explicit principle that justified belief is closed under deduction (Gettier, 1963, p. 121) to Smith’s inference from (f) to (h), and this presupposition only makes full sense if the ‘strong evidence’ that Smith has for (f) is understood as entailing that Smith’s belief that (f) is also justified.16 But even if there should be no plausible interpretation of Gettier’s presentation on which ‘having strong evidence’ metaphysically entails ‘being justified’ in a fallibilist sense (as explicitly assumed by Gettier), we could simply enforce the top-down strategy more rigorously by stipulating outright that Smith’s belief that (f) is justified.17 Arguably, such an outright stipulation would be a legitimate move in Gettier’s thought experiment, given that the critical ‘test question’ is about knowledge under the circumstances of the case—and not about justification (cf. Grundmann & Horvath, 2014b). The comparison to scientific experiments is instructive here, because it is a standard practice in empirical experiments to control for potentially confounding factors as much as possible, such that—­ ideally—only the relevant test question remains genuinely open. Let us now turn to measure (B) of the top-down strategy, which is to include some feature in the case that prevents Smith from knowing that (h) in the way described. More specifically, what we need is a feature that metaphysically entails that Smith does not know that (h) in the way described, because only then would it be true in a hypothetical supposition that Smith’s belief that (h) falls short of knowledge. However, is there such a feature in Gettier’s second case? In fact, Gettier explicitly tells us what this feature is: But imagine now that two further conditions hold. First, Jones does not own a Ford, but is at present driving a rented car. And secondly, by the sheerest coincidence, and entirely unknown to Smith, the place mentioned in proposition (h) [= ‘Either Jones owns a Ford, or Brown is in Barcelona’] happens really to be the place where Brown is. If these two conditions hold then Smith does not know that (h)

314  Joachim Horvath is true, even though (i) (h) is true, (ii) Smith does believe that (h) is true, and (iii) Smith is justified in believing that (h) is true. Gettier, (1963, p. 123) Gettier argues here that it is the epistemic accidentality or luck—‘the sheerest coincidence’—of the truth of Smith’s belief that (h) which makes it true that Smith does not know that (h) in the way described (i.e., on the basis of a deductive inference from Smith’s belief that (f)—see the earlier passage quoted above). Gettier’s argument presupposes, of course, that the relevant kind of epistemic luck is indeed metaphysically incompatible with knowing that (h) in the way described. Otherwise, the key suppositional judgment that Smith does not know that (h) would not be true, because it would not be metaphysically entailed by a hypothetical supposition of the case. However, the metaphysical incompatibility of epistemic luck with knowledge is an independently plausible claim that is still widely shared in contemporary epistemology (see, e.g., Deutsch, 2016; Engel Jr., 2015; Pritchard, 2005). This brings us to measure (C) of the top-down strategy, which is to stipulate that Smith does not know that (h) in some alternative way. Gettier clearly displays some sensitivity to this issue, because he deliberately forecloses a deviant, knowledge-involving interpretation of the case when he stipulates that Brown’s actual location—Barcelona—is ‘entirely unknown to Smith’ (see quote above). So, Gettier explicitly rules out that Smith knows the disjunction ‘Either Jones owns a Ford, or Brown is in Barcelona’ by knowing its (true) second disjunct in some independent way. However, could Smith also know the whole disjunction in some alternative way that is not ruled out by the explicit content of the case? A reliable and trustworthy informant might have told Smith, for example, that it is true that ‘Either Jones owns a Ford, or Brown is in Barcelona’ (e.g., the informant might know where Brown is, but—for some strange reason— only communicate the weaker disjunction to Smith).18 However, Gettier (1963, p. 123) also forecloses a more far-fetched deviant realization of this kind, because he explicitly stipulates that Smith ‘proceeds to accept’ the disjunction (h) [= ‘Either Jones owns a Ford, or Brown is in Barcelona’] ‘on the basis of’ his justified belief that (f) [= ‘Jones owns a Ford’] (my emphasis). As argued above, Gettier uses the phrase ‘proceeds to accept that p’ more or less synonymously with ‘comes to believe that p’. It follows that Smith didn’t believe (h) before he accepted it on the basis of (f), and so it seems that, at the moment when he proceeds to accept (h), Smith’s belief that (h) is only based on his belief that (f). Thus, the deviant realization sketched above seems to be ruled out by how Gettier sets up the case. But could Smith have come to believe that (h) on the basis of (f) and also on the basis of testimony? No, because the most charitable interpretation of Gettier’s use of the adverbial phrase ‘on the basis of’ is in terms of Smith’s sole basis for coming to believe that (h), for it would run

Gettier’s Thought Experiments  315 counter to the dialectical purpose of Gettier—which is to argue that JTB without K is possible—to leave it open that Smith may have a larger basis for believing—and thus potentially knowing—that (h). So, there is only room for a deviant realization here if Smith could have two distinct beliefs with the same content that (h) at the same time (assuming that knowledge-ascriptions are implicitly time-indexed), one of which amounts to knowledge that (h). One way for this to happen is if Smith had a split-­ personality or a radically compartmentalized mind that genuinely allows for the ascription of two tokens of the same belief-type. Another way is my scenario of a time-traveling Smith sketched above. Concerning such very far-fetched deviant realizations, I think it would be disingenuous to claim that they are also ruled out by Gettier’s case description (given that they are clearly not dialectically salient), or even to deny that they are metaphysically possible in the first place. So, arguably, there are indeed deviant realizations that are not even ruled out by Gettier’s careful presentation of his second case. To sum up, Gettier clearly intends to rule out some of the most salient or nearby deviant realizations with respect to Smith’s lack of knowledge that (h), but he still falls short of excluding very far-fetched deviant realizations of the kind just discussed. As present-day thought experimenters, however, we can easily amend—and improve on—Gettier’s method with the two strategies outlined above. For example, we can use the top-down strategy and simply stipulate that Smith has only one token-belief that (h), and that this one belief-token is solely based on his other belief that (f) by means of a deductive inference. Independently of that (or in addition to it), we can also pursue the bottom-up strategy and ‘stipulate away’ all salient deviant realizations, for example, by explicitly adding it to our supposition that Smith does not have a split-personality, or a radically compartmentalized mind, or a time-traveling future or past self. Finally, a mutually reinforcing combination of these two strategies would give us a strong reason to believe that there is indeed a properly amended version of Gettier’s second case which metaphysically entails that there is JTB without K. With the proposed understanding of supposition and suppositional judgments in Gettier’s thought experiments in hand, we are now in a position to consider how the overall argument for the envisaged conclusion should be understood. First, the goal of the thought experiments is to establish the metaphysical insufficiency of JTB for knowledge. To that end, Gettier invites us to suppose two cases in hypothetical mode that are assumed to be metaphysically possible, which they uncontroversially are. Gettier then argues, under the supposition, that it is true in both cases that their main protagonist, Smith, has a JTB that falls short of knowledge. To side-step problems with arbitrary names, like ‘Smith’, we can also apply the substitution gambit, for example, by replacing ‘Smith’ with ‘Edmund Gettier’ or ‘Timothy Williamson’. Moreover, to assure ourselves that Gettier’s suppositional conclusion ‘there is JTB without K’

316  Joachim Horvath is entailed by a hypothetical supposition of his cases—and thus true or correct under the supposition—we also need to exclude far-fetched deviant realizations of his cases that would falsify the relevant entailments. To this end, we can use the top-down strategy and explicitly set up the cases in a deviance-proof way, or the bottom-up strategy of stipulating all salient deviant realizations away—or some mutually reinforcing combination of both strategies. If successful, we can conclude that, given that it is true in these metaphysically possible cases, supposed as hypothetical, that there is JTB without K, it is also metaphysically possible simpliciter that there is JTB without K, which contradicts the metaphysical sufficiency of JTB for knowledge.19 The final step in this suppositional reconstruction of Gettier’s argument might still raise some eyebrows: the key modal inference from ‘it is true in the supposed-as-hypothetical cases that P’ to ‘it is (metaphysically) possible that P’. So, why is it plausible that one can ‘export’ claims about metaphysical possibility from suppositional judgments that are only made relative to cases which are supposed as merely hypothetical? This question has a simple answer that, I think, is nevertheless correct. If we assume—as we do with Gettier’s cases—that they are metaphysically possible, then a judgment that is true in these cases, when supposed as hypothetical, is a judgment that is metaphysically entailed by those cases. However, if P is metaphysically entailed by something that is itself metaphysically possible, then P is metaphysically possible as well (cf. Williamson, 2007, p. 184)—and so the critical modal inference is warranted.20

Williamson’s Counterfactual Reconstruction In this section, I will contrast and compare the suppositional reconstruction of Gettier’s thought experiments developed above with Timothy ­Williamson’s (2007, Chapter 6) celebrated counterfactual reconstruction. In doing so, I will highlight both points of contention and commonality. Let me first consider how Williamson relates his own approach to Gettier’s actual method. After briefly sketching Gettier’s actual method in the following passage, Williamson (2007, pp. 183–184; my emphasis) quickly opts for a more schematic approach: Gettier presents his specific counterexamples to the target analysis through short fictional narratives, in the present tense indicative, with fictional uses of proper names (‘Smith’ and ‘Jones’), all introduced by ‘suppose that.’ Beyond their conformity to the abstract pattern just explained, their details do not concern us. Let us construct another example to the same pattern. […] For present purposes, in formalizing Gettier’s argument […], we can ignore most of the structure specific to his cases, and concentrate on the logical structure they share with most other imaginary counterexamples to philosophical analyses.

Gettier’s Thought Experiments  317 What Williamson assumes here is that the details of Gettier’s p ­ resentation are largely irrelevant to an adequate understanding of his thought experiments. Thus, Williamson continues with suggesting a general logical structure that he takes to be shared by most other philosophical thought experiments of the counterexample-type. Against this, I have argued that some of the details of Gettier’s presentation are in fact crucial to Gettier’s method, for example, the explicit inclusion of the key feature of epistemic luck, or the way in which Gettier sets up his cases vis-à-vis the JTB-condition. Therefore, a more cautious approach seems advisable here, which is to begin with detailed analyses of seminal philosophical thought experiments, and only then look into more general patterns. Setting the issue of generality to the side, what is Williamson’s main motivation for his own counterfactual account? Williamson (2007, p. 184) starts from the plausible assumption, which I share, that Gettier’s cases must be regarded as metaphysically possible in order to raise a challenge for the metaphysical sufficiency of JTB for K. What Gettier really needs, however, is the metaphysical possibility of JTB without K, because only this refutes the metaphysical sufficiency of JTB for K. The key metaphilosophical question for Williamson therefore is which additional premise we need in order to deductively infer the metaphysical possibility of JTB without K from the metaphysical possibility of Gettier’s cases. As a natural starting point, Williamson suggests that, ‘[t]o a first approximation, we can formalize that as the claim that, necessarily, anyone who stands in the Gettier relation to a proposition has justified true belief in that proposition without knowledge’ (Williamson, 2007, p. 184). This firststab reconstruction amounts to the strict conditional that, by metaphysical necessity, if things are as in Gettier’s cases, then there is JTB without K—from which it follows, together with the metaphysical possibility of Gettier’s cases and ‘[b]y elementary modal reasoning’ (ibid.), that it is metaphysically possible that there is JTB without K. Williamson rejects this first-stab reconstruction, however, because it runs into a serious obstacle that I have already discussed: the problem of deviant realizations (cf. Williamson, 2007, p. 185, 2020, pp. 230–231; Grundmann and Horvath, 2014a,b; Malmgren, 2011). The reason why this problem arises here is because the strict conditional is falsified by various unintended or deviant metaphysically possible realizations of Gettier’s cases, of the kind considered above, that render its antecedent true but its consequent false (cf. Malmgren, 2011, p. 275). This problem is Williamson’s main motivation for weakening the missing premise from a strict conditional to a counterfactual conditional, which equally suffices for a deductively valid modal argument from the possibility of Gettier’s cases to the possibility of JTB without K.21 According to Williamson (2007, p. 186), the relevant counterfactual allows us to ‘leave the world to fill in the details of the story, rather than trying to do it all ourselves’, which would be an elegant way to avoid deviant realizations.

318  Joachim Horvath The main extant criticism of Williamson’s counterfactual reconstruction is that it runs into a ‘revenge problem’ of deviant realizations. For what if the closest possible world where the antecedent of Williamson’s counterfactual is true just is one of the problematic deviant worlds (see, e.g., Malmgren, 2011, section 1.5)? Even worse, some critics argue, this holds Williamson’s counterfactual hostage to opaque going-ons in the actual world that can only be known empirically—which is at odds with the common metaphilosophical view that philosophical thought experiments can be evaluated from one’s armchair alone (cf. Ichikawa, 2009; Malmgren, 2011). Irrespective of whether one thinks that this ‘empirical version’ of the problem of deviant realizations is worse than the original, more aprioristic version, the fact remains that the counterfactual reconstruction does not fully resolve the problem—even though it seems to be Williamson’s main motivation for ‘going counterfactual’ in the first place.22 This raises the question whether he has really achieved so much in comparison to the problem as it arises for Gettier himself. For Gettier, too, manages to exclude a number of deviant realizations—those that are especially salient from an epistemological point of view—by the way in which he constructs his cases. And, just like Williamson, Gettier does not manage to exclude all deviant realizations in a principled way. Above, I have argued that we can improve on Gettier in this respect by using a top-down and/or bottom-up approach to the problem of deviant realizations—and especially the top-down approach promises to deliver a systematic and complete exclusion of all deviant realizations (cf. Grundmann and Horvath, 2014a,b). Williamson could use these or similar strategies as well, of course, but this would also undermine his main motivation for rejecting the strict-conditional account and other entailment-based accounts, such as my suppositional reconstruction. Williamson (2007, pp. 200–201) actually comes close to endorsing a bottom-up approach in his ‘official’ response to the revenge problem of deviant realizations, where he suggests that one could ‘repair’ the relevant case description or make further stipulations about the case when confronted with a deviant realization. The top-down proposal of Grundmann and Horvath (2014a,b) would seem to work for Williamson’s counterfactual account as well because one could simply ‘plug’ a deviance-proof version of Gettier’s cases into the antecedent of his counterfactuals. ­However, using any of these approaches would, to some degree, undermine ­Williamson’s main motivation for preferring the counterfactual account over other approaches. Therefore, it seems that the problem of deviant realizations, being a problem for almost everyone, does not help the counterfactual account to gain a decisive advantage over other accounts. Turning to a different issue, Williamson seems to accept that both a strict conditional and his counterfactual conditional are at least prima facie plausible reconstructions of Gettier’s case judgment, that is, of ‘the

Gettier’s Thought Experiments  319 verdict that the subject in the Gettier case has justified true belief without knowledge’ (Williamson, 2007, p. 184). This assumption is also shared by most other reconstructions of Gettier’s thought experiments, for example, by Malmgren (2011, section 1.3), who regards all of these reconstructions as potential answers to the ‘content problem’, which is to provide an adequate reconstruction of the content of Gettier’s case judgments. Both Williamson and Malmgren explicitly note, however, that the face-value content of Gettier’s case judgment—something like ‘Smith has a justified true belief without knowledge’—is very different from either a strict conditional or a counterfactual judgment (and also from Malmgren’s (2011, p. 281) alternative possibility judgment). But despite the fact that suppositional reasoning can in principle be used to justify both a strict conditional and Williamson’s counterfactual (cf. Balcerak Jackson and Balcerak Jackson, 2013; Williamson, 2007, pp. 152–153), it is quite implausible to identify either the strict conditional or the counterfactual with only one part of the suppositional reasoning in question, for example, with a particular suppositional judgment like ‘Smith has a justified true belief without knowledge’. If we consider a conditional proof as analogy, in which we first assume A, then show that C follows from A, and finally conclude that A → C, this would amount to a plain confusion of what follows from the conditional proof as a whole, namely, A → C, with what only follows from the assumption A, namely, C (see also Balcerak Jackson, 2018). However, this is exactly what both the strict conditional and the counterfactual reconstruction effectively do here: they wrongly identify the concluding part of the relevant suppositional reasoning with the conditional as a whole. Another reason why it seems odd to reconstruct the content of Gettier’s case judgments in terms of either a strict or a counterfactual conditional is that Gettier manifestly argues for his case judgments, but not for the relevant conditionals (cf. Deutsch, 2010, 2015, 2016; Horvath, 2022). In his second case, for example, Gettier explicitly cites the accidental truth of Smith’s belief as the key reason why Smith does not know that ‘Either Jones owns a Ford, or Brown is in Barcelona’ (see above). From the perspective of my suppositional reconstruction, the point of Gettier’s argument can simply be interpreted as a (successful) attempt to identify the (most salient) feature of his case that makes it true that Smith lacks the relevant knowledge (see also Horvath, 2022)—which I have spelt out in terms of being entailed by a hypothetical supposition of the case. From Williamson’s perspective, however, the only way of accommodating ­Gettier’s argument would be as part of developing a counterfactual supposition of the case, in which ‘one supposes the antecedent and develops the supposition, adding further judgments within the supposition by reasoning, offline predictive mechanisms, and other offline judgments’ and ‘one asserts the counterfactual conditional if and only if the development eventually leads one to add the consequent’ (Williamson, 2007,

320  Joachim Horvath pp. 152–153). But because Gettier’s case judgment clearly belongs to the part of ‘adding the consequent’, this just reinforces my point that the content of Gettier’s case judgment is not happily captured by the counterfactual conditional, which would seem to confuse the content of the counterfactual as a whole with the content of its consequent. I therefore conclude that neither a strict conditional nor a counterfactual conditional qualifies as acceptable solution to the content problem. However, I want to close this section on a more positive note, namely, by exploring the common ground between my suppositional reconstruction and Williamson’s counterfactual reconstruction, which is a lot larger than one might think at this point. To begin with, it is common ground that a reconstruction of Gettier’s thought experiments requires an (implicit) assumption or premise that affirms the (metaphysical) possibility of his cases (see also Williamson, 2020, p. 234), and it is common ground as well that Gettier (implicitly) concludes that it is (metaphysically) possible that there is JTB without knowledge. The remaining controversy thus only revolves around the key metaphilosophical question how Gettier can plausibly establish his conclusion on the basis of his case judgments and the (metaphysical) possibility of his cases. ­Furthermore, it is also common ground with Williamson’s counterfactual reconstruction—in contrast to various other approaches—that supposition and suppositional reasoning play an important role in answering the key metaphilosophical question. The ‘only’ substantial difference is that Williamson understands the relevant suppositional reasoning as counterfactual, and thus takes it to support a counterfactual conditional, while I understand it as hypothetical reasoning that—together with the metaphysical possibility of Gettier’s cases—allows one to directly infer the metaphysical possibility of JTB without K. Moreover, if one were in a position to explicitly acknowledge that Gettier’s case judgments are metaphysically entailed by a (suitably amended) hypothetical supposition of his cases, then one would also be in a position to endorse the corresponding counterfactuals, which simply follow from these entailments (cf. Williamson, 2007, pp. 185–187). Williamson’s counterfactual reconstruction is thus neither incompatible with my suppositional reconstruction, nor with a reconstruction in terms of a strict conditional. Thus, one could still resort to Williamson’s counterfactual approach as a nearby fallback option if one should have doubts about the suppositional reconstruction or other entailment-based accounts.

Conclusion In this chapter, I have developed a suppositional reconstruction of Gettier’s famous thought experiments that is inspired by a close reading of Gettier’s (1963) seminal presentation. This suppositional reconstruction comes closer to Gettier’s actual method than other extant reconstructions,

Gettier’s Thought Experiments  321 and thus provides a descriptively more adequate reconstruction of ­Gettier’s thought experiments. For example, Gettier arguably invites us to suppose his cases in hypothetical mode, while Williamson’s influential counterfactual account requires that we suppose them in counterfactual mode. Moreover, Gettier’s case judgments are best understood as non-conditional suppositional judgments, and not as conditional categorical judgments of some sort—as Williamson’s and other extant reconstructions would have it. Despite the merits of Gettier’s seminal presentation, it also turns out that Gettier’s method, as it stands, is not fully defensible or at least incomplete in several respects. For example, Gettier does not make the kind of modality and the modal reasoning that underlies his argument sufficiently transparent. Thus, a somewhat anachronistic reconstruction in terms of metaphysical modality seems to be the most plausible option from a contemporary point of view. More specifically, I have proposed that the overall argument is best understood in terms of supposing Gettier’s cases as hypothetical, which allows one to ‘export’ certain claims about metaphysical possibility under suitable conditions. Second, Gettier’s use of arbitrary names, like ‘Smith’ or ‘Jones’, while seemingly innocuous and easy to process, raises thorny philosophical issues for a rational reconstruction of his thought experiments. Here, I have made the pragmatic methodological suggestion that this issue can be side-stepped with the substitution gambit, that is, by replacing ­Gettier’s arbitrary names with ordinary names of arbitrary people. Third, while Gettier shows some sensitivity to the problem of deviant realizations and manages to exclude some epistemologically salient deviant realizations, he nevertheless fails to exclude various more farfetched deviant realizations of his cases. For this reason, I have argued that Gettier’s method needs to be supplemented with a top-down and/ or bottom-up approach to the problem of deviant realizations. Given that most other reconstructions—including Williamson’s counterfactual account—also do not manage to exclude all deviant realizations in a principled way, they would have to rely on similar strategies for providing a satisfactory solution to the problem. Finally, I contrasted and compared my suppositional reconstruction with Williamson’s influential counterfactual account. Apart from being more faithful to Gettier’s actual way of doing thought experiments, the suppositional reconstruction can be supplemented with various tools, e.g., for dealing with arbitrary names and the problem of deviant realizations, that make it no less defensible than other reconstructions, including Williamson’s account. From a more ecumenical perspective, the suppositional reconstruction and Williamson’s counterfactual account, as well as various other reconstructions, may be much closer to each other anyway than the controversial debate about them often suggests. In light of these commonalities, the descriptive adequacy to the

322  Joachim Horvath method of thought experimentation as it is actually practiced by Gettier and many other philosophers, often with great historical success, suggests itself as an important tiebreaker in the metaphilosophical debate. This alone seems to be a good reason to carefully reconsider Edmund L. Gettier III’s exemplary and still instructive way of doing Gettier thought experiments.23

Notes 1 For an obituary, see https://www.umass.edu/philosophy/memoriam-edmundl-gettier-iii-1927-2021. 2 Gettier published only two papers and one book review in the 1960s—and maybe one hard-to-track paper in a Spanish translation (see https://www. umass.edu/philosophy/memoriam-edmund-l-gettier-iii-1927–2021). 3 Methodologically, this approach is somewhat inspired by recent work of Max Deutsch (2015) and Herman Cappelen (2012), who argue that appeals to intuition play no crucial role in seminal presentations of philosophical thought experiments, including Gettier’s short paper (see, e.g., Deutsch, 2016). According to them, Gettier and other philosophers instead argue for their judgments about hypothetical cases (see also Horvath, 2022). 4 Thanks to Tim Williamson for discussion here. 5 Thanks to Tim Williamson for very helpful and illuminating discussions on this issue. 6 Historical fun fact: Gettier, Lehrer, and Plantinga were all colleagues at Wayne State University during the ‘golden age’ of its philosophy department in the late 1950s and 1960s (cf. Sleigh, 1988). 7 I have to leave it open here to what extent the substitution gambit is also applicable to other thought experiments beyond Gettier’s. My conjecture would be that it works well in epistemology and moral philosophy but might have certain limitations, e.g., in metaphysics (thanks to Tim Williamson for discussion). 8 Gettier (1963, p. 121) also makes two general assumptions: first, that justification is fallible, which is fairly uncontroversial, and second, that justified belief is closed under deduction, which I will come back to below. 9 In order to work as planned, this example presupposes that Oswald did in fact kill Kennedy and that Oswald was the only person aiming to kill ­Kennedy at that time. If you happen to disagree with these presuppositions, you can still accept them for the sake of the example here (thanks to Alex Wiegmann for discussion). 10 This is basically a version of the problem of deviant realizations (see below). 11 What happens if we apply the substitution gambit to Gettier’s cases (see above), for example, by replacing ‘Smith’ with the ordinary proper name ‘Edmund Gettier’? It would basically make no difference here because we would still ignore corresponding (empirical) facts about the actual Edmund Gettier as irrelevant to the thought experiment, even if the actual Gettier should happen to satisfy the case description in question. 12 This is not to suggest that the three modes of supposition discussed in this chapter are already exhaustive. There might be other distinctive modes of supposing apart from them, such as supposing-as-probable, supposing-as-­ possible, or supposing-as-fictional. 13 It should be emphasized that this makes suppositional truth a species of relative truth: suppositional judgments are true only relative to a particular

Gettier’s Thought Experiments  323 supposition. Or, as Barnett (2010, p. 274) puts it in slightly more general form: ‘[…] a suppositional statement itself has a given value just in case an assignment of that value to what is stated by the statement, relative to what is supposed by the statement, is objectively correct’. It follows that suppositional truth differs from plain truth in some important respects. For example, if we suppose a contradiction in hypothetical mode, then this contradiction is true under the supposition in question because it is (trivially) entailed by it. However, as long as we stay in suppositional mode in such cases, no harm is done concerning what we accept as true simpliciter. The different behavior of suppositional truth only requires that we are extra-careful when we ‘export’ something from our supposition in order to assert it categorically (thanks to Tim Williamson for discussion). Finally, what really matters in the present context is that there are objective correctness-conditions for suppositional judgments that bear some important similarities to truth simpliciter. How happy one is to characterize these correctness-conditions in terms of truth might depend on one’s more general views about truth. For example, while this might be highly congenial to pluralists about truth (see, e.g., Lynch, 2009), it might be more of a stretch for correspondence theorists. 14 Mutatis mutandis for thought experiments that aim at conceptual possibility: a suppositional judgment would then be true iff it is conceptually entailed by the explicit content of the supposition. 15 One might object that even if Gettier treats ‘accept’ as entailing ‘believe’ in his paper, this does not establish that the entailment actually holds (thanks to Tim Williamson for pressing this point). However, if we interpret Gettier charitably, we should at least assume that he speaker-means ‘accept’ in a way that entails ‘believe’, for his immediate transition from ‘proceed to accept’ to ‘believing’ the relevant propositions would otherwise be left quite mysterious. 16 In an earlier version of this paper, I was under the illusion that one could interpret Gettier as directly stipulating that Smith’s relevant true belief is justified. I am very grateful to Tim Williamson for alerting me to this mistake. Thus, a more nuanced interpretation is required here that takes the dialectical context of Gettier’s thought experiments into account as well. A more ambitious option would be Saint-Germier’s (2021) pragmatic account on the basis of general neo-Gricean principles. In a nutshell, Saint-Germier argues that the default interpretation of Gettier cases—even outside their proper dialectical context—is resistant to deviant realizations in just the way proposed by Grundmann and Horvath (2014a, 2014b). Saint-Germier thus boldly defends the claim that the top-down approach to deviant realizations is already fully realized by the default interpretation of Gettier’s case descriptions. In this paper, I only make the more modest assumption that the top-down strategy is at least available to us now, as present-day thought experimenters. 17 One might worry that this stipulation makes the metaphysical possibility of the case more dubitable (thanks to Tim Williamson for this point). While this is a legitimate worry in principle, it nevertheless does not apply here because the metaphysical possibility of having a justified belief in a mundane proposition like ‘Jones owns a Ford’ seems beyond reasonable doubt (at least from the fallibilist perspective of Gettier’s paper). 18 Thanks to Tim Williamson for pointing out a deviant realization of this kind. 19 This rational reconstruction of Gettier’s argument is in considerable agreement with a brief sketch that Max Deutsch (2016, p. 93) provides in footnote 6 of his paper ‘Gettier’s Method’. Another suppositional reconstruction of Gettier’s argument is Balcerak Jackson (2018). Unlike Balcerak Jackson, however, I think that Gettier’s

324  Joachim Horvath case judgments do have truth- or correctness-conditions of some sort, that these case judgments are not inferential judgments, and that they are not best ­understood in analogy to the conclusions of a conditional sub-proof—which is Balcerak Jackson’s preferred model for reasoning in a thought experiment. What is more, Balcerak Jackson lacks a good explanation of how Gettier’s case judgments ‘follow’ from his cases, because she does not offer a solution to the problem of deviant realizations. Finally, Balcerak Jackson also fails to explain how the (metaphysical) possibility of JTB without K is actually established in her conditional sub-proof model, because a conditional sub-proof typically only supports a conditional judgment of some sort (cf. Horvath, 2015, pp. 402–403). Other suppositional reconstructions of thought experiments that I lack the space to discuss here can be found in Rescher (2005), Balcerak Jackson and Balcerak Jackson (2013), and Horvath (2011, 2015). 20 This can be seen as a (hopefully improved) successor of my earlier proposals in Horvath (2011, 2015). 21 Williamson’s counterfactual approach also has predecessors in Sorensen (1992) and Häggqvist (1996). 22 In Suppose and Tell, Williamson (2020, section 14.3) argues that this ‘revenge problem’ of deviant realizations is substantially mitigated by his novel account of counterfactuals developed in this book. 23 First and foremost, I would like to thank the editors of this volume, Anand Vaidya and Duško Prelević, for inviting me to contribute this chapter, and for their incredible patience with my ever-growing pandemic-related delays in delivering the manuscript. Second, I am very grateful to Tim Williamson for the opportunity to develop my ideas about Gettier’s thought experiments in relation to his counterfactual account, for pointing out various shortcomings and mistakes in my earlier drafts, and for his willingness to respond to my chapter in this volume. I would also like to thank Max Deutsch, Steffen Koch, and Alex Wiegmann for discussion and for very helpful and encouraging comments on a first draft of this chapter, as well as the following people for helpful discussions of my work on thought experiments over the years: Magdalena Balcerak Jackson, Al Casullo, Earl Conee, Richard Feldman, Harmen Ghijsen, Thomas Grundmann, Sören Häggqvist, Peter Hawke, Jens Kipper, Steffen Koch, David Mark Kovacs, Errol Lord, Daniel Nolan, Christopher Ranalli, Luis Rosa, Gerhard Schurz, Ori Simchen, Lukas Skiba, Margot Strohminger, Tim Williamson, Richard Woodward, and Juhani Yli-Vakuri. Finally, I would like to thank several audiences of talks that can be seen as predecessors or variants of the present chapter: at the international conference Modal Epistemology and Metaphysics, University of ­Belgrade, September 2014, at the Epistemology of Metaphysics ­Workshop (EMW 1), University of Helsinki, May 2015, at the 1st Cologne/ Leuven Epistemology Meeting, University of Cologne, October 2015, at the Research Seminar – Chair of T ­ heoretical Philosophy, Heinrich Heine University Düsseldorf, April 2016, at the Bonn-Cologne Research Colloquium in Epistemology, University of Cologne, January 2017, at ECAP9, LudwigMaximilians-­University Munich, August 2017, at the 12th Cologne Summer School in Philosophy – Advantages of Evidentialism with Earl Conee and Richard Feldman, University of Cologne, October 2017, at the Department Colloquium of the Department of Philosophy, University of Hamburg, November 2017, at the workshop Imagination as a Source of Knowledge, CONCEPT – Cologne Center for Contemporary Epistemology, University of Cologne, February 2018, and at the online Rhine-Ruhr Epistemology Meeting in July

Gettier’s Thought Experiments  325 2021. Last but not least, my work on this paper was generously funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – project number 391304769.

References Adams, E. W. (1970). Subjunctive and Indicative Conditionals. Foundations of Language, 6(1), 89–94. Balcerak Jackson, M. (2018). Intuitions as Inferential Judgments. Philosophical Issues, 28(1), 7–29. Balcerak Jackson, M., & Balcerak Jackson, B. (2013). Reasoning as a Source of Justification. Philosophical Studies, 164(1), 113–126. Barnett, D. (2010). Zif Would Have Been If: A Suppositional View of Counterfactuals. Noûs, 44(2), 269–304. Borges, R., de Almeida, C., & Klein, P. D. (Eds.) (2017). Explaining Knowledge: New Essays on the Gettier Problem. Oxford: Oxford University Press. Breckenridge, W., & Magidor, O. (2012). Arbitrary Reference. Philosophical Studies: An International Journal for Philosophy in the Analytic Tradition, 158(3), 377–400. Cappelen, H. (2012). Philosophy without Intuitions. Oxford: Oxford University Press. Cappelen, H., & Dever, J. (2018). Puzzles of Reference. Oxford: Oxford University Press. Chalmers, D. (2002). Does Conceivability Entail Possibility? In T. S. Gendler & J. Hawthorne (Eds.), Conceivability and Possibility (pp. 145–200). Oxford: Oxford University Press. Deutsch, M. (2010). Intuitions, Counter-Examples, and Experimental Philosophy. Review of Philosophy and Psychology, 1(3), 447–460. Deutsch, M. (2015). The Myth of the Intuitive: Experimental Philosophy and ­Philosophical Method. Cambridge, MA: MIT Press. Deutsch, M. (2016). Gettier’s Method. In J. Nado (Ed.), Advances in Experimental Philosophy and Philosophical Methodology (pp. 69–97). London: Bloomsbury. Engel Jr., M. (2015). Epistemic Luck. In J. Fieser & B. Dowden (Eds.), Internet Encyclopedia of Philosophy. http://www.iep.utm.edu/epi-luck/. Fine, K. (1983). A Defence of Arbitrary Objects. Proceedings of the Aristotelian Society, Supplementary Volumes, 57, 55–77. Gettier, E. L. (1963). Is Justified True Belief Knowledge? Analysis, 23(6), 121–123. Grundmann, T., & Horvath, J. (2014a). Erratum to: Thought Experiments and the Problem of Deviant Realizations. Philosophical Studies, 170(3), 535–536. Grundmann, T., & Horvath, J. (2014b). Thought Experiments and the Problem of Deviant Realizations. Philosophical Studies, 170(3), 525–533. Häggqvist, S. (1996). Thought Experiments in Philosophy. Stockholm: Almqvist & Wiksell International. Hilpinen, R. (2017). Sed ubi socrates currit? On the Gettier Problem before ­Gettier. In R. Borges, C. de Almeida, & P. D. Klein (Eds.), Explaining ­K nowledge: New Essays on the Gettier Problem (pp. 135–151). Oxford: Oxford University Press. Horvath, J. (2011). In Defense of Conceptual Analysis [Dissertation]. Universität zu Köln.

326  Joachim Horvath Horvath, J. (2015). Thought Experiments and Experimental Philosophy. In C. Daly (Ed.), The Palgrave Handbook of Philosophical Methods (pp. 386–418). London: Palgrave Macmillan. Horvath, J. (2016). Conceptual Analysis and Natural Kinds: The Case of ­K nowledge. Synthese, 193(1), 167‒184. Horvath, J. (2018). Philosophical Analysis: The Concept Grounding View. ­Philosophy and Phenomenological Research, 97(3), 724–750. Horvath, J. (2022). Mischaracterization Reconsidered. Inquiry. DOI: 10.1080/0020174X.2021.2019894 Ichikawa, J. J. (2009). Knowing the Intuition and Knowing the Counterfactual. Philosophical Studies, 145(3), 435–443. Ichikawa, J. J., & Jarvis, B. (2009). Thought-Experiment Intuitions and Truth in Fiction. Philosophical Studies, 142(2), 221–246. Kripke, S. (1980). Naming and Necessity. Cambridge, MA: Harvard University Press. Lewis, D. (1973). Counterfactuals. Cambridge, MA: Harvard University Press. Lynch, M. (2009). Truth as One and Many. Oxford: Clarendon Press. Malmgren, A.-S. (2011). Rationalism and the Content of Intuitive Judgements. Mind, 120(478), 263–327. Matilal, B. K. (1986). Perception: An Essay on Classical Indian Theories of ­K nowledge. Oxford: Clarendon Press. Pritchard, D. (2005). Epistemic Luck. Oxford: Oxford University Press. Ramsey, F. P. (1931). General Propositions and Causality. In R. B. Braithwaite (Ed.), The Foundations of Mathematics and Other Logical Essays (pp. 237–255). London: Kegan Paul, Trench & Trubner. Rescher, N. (2005). What If?: Thought Experimentation in Philosophy. New Brunswick, NJ and London: Transaction Publishers. Saint-Germier, P. (2021). Getting Gettier Straight: Thought Experiments, ­Deviant Realizations and Default Interpretations. Synthese, 198(2), 1783–1806. Sleigh, R. C. (1988). Knowing Edmund Gettier. In D. Austin (Ed.), Philosophical Analysis: A Defense by Example (pp. xiii–xv). Dordrecht: Kluwer Academic Publishers. Sorensen, R. (1992). Thought Experiments. Oxford: Oxford University Press. Stalnaker, R. C. (1968). A Theory of Conditionals. In N. Rescher (Ed.), ­Studies in Logical Theory (American Philosophical Quarterly Monographs 2) (pp. 98–112). Oxford: Blackwell. Weinberg, J. M., Nichols, S., & Stich, S. (2001). Normativity and Epistemic Intuitions. Philosophical Topics, 29(1/2), 429–460. Williamson, T. (2007). The Philosophy of Philosophy. Oxford: Blackwell. Williamson, T. (2020). Suppose and Tell: The Semantics and Heuristics of ­Conditionals. Oxford: Oxford University Press.

16 Horvath on Gettier’s Thought Experiments Timothy Williamson

Introduction Edmund Gettier gave two hypothetical examples to turn the tide of epistemology against the analysis of knowledge as justified true belief (Gettier 1963). That episode has long been a canonical test case for the philosophical role of thought experiments. In The Philosophy of Philosophy, I developed a rational reconstruction of the argumentative structure underlying such philosophical uses of thought experiments, in the framework of quantified modal logic with counterfactual conditionals (Williamson 2007: 179–207). Now Joachim Horvath in his characteristically thoughtful chapter ‘Gettier’s Thought Experiments’ proposes a new reconstruction of what Gettier was doing. Horvath claims that his account is more faithful to the text of the 1963 paper than the available alternatives, and that it is ‘simpler and more straightforward’ than my account. He also acknowledges some features common to his reconstruction and mine. In this response, I will discuss various problems for his account, and explain how my alternative avoids them. In particular, I will note ways in which the apparent simplicity of his reconstruction is achieved at the cost of introducing elements alien to Gettier’s paper and neglecting to develop a metalogical framework to make sense of his reconstructed argument.

Hermeneutic Difficulties A complicating factor is that although Gettier’s brief paper is mostly clear, it is severely elliptical, and provides little explicit evidence for deciding some key points of interpretation. Horvath is well aware of this difficulty. He writes: ‘Trying to reconstruct the reasoning behind Gettier’s method has proven to be almost as intractable as finding a better analysis of knowledge in light of his famous counterexamples’. For instance, as Horvath and I agree, Gettier’s arguments tacitly presuppose that his hypothetical scenarios are possible in some sense, but Gettier’s paper does not indicate in what sense. Horvath follows my 2007

DOI: 10.4324/9781003002192-18

328  Timothy Williamson account in treating the modality as metaphysical, although we also agree on historical grounds that if Gettier had been asked in 1963 what kind of possibility he intended, he would probably have said that it was ‘logical’ or ‘conceptual’. Furthermore, Horvath identifies various subtle loopholes in Gettier’s arguments, which Gettier seems to have missed; to fill them, he adds corresponding constraints to the description of the hypothetical scenario. Presumably, Gettier did not positively intend those additional constraints; otherwise, he would have made them explicit, given the care with which his paper is written, though he might well have welcomed them if the need for them had been pointed out to him. Horvath concludes that his reconstructive strategies ‘give us strong reason to believe that there is indeed a properly amended version of Gettier’s second case which metaphysically entails that there is JTB without K’; on his interpretation, the soundness of Gettier’s arguments requires such a metaphysical entailment. Thus, Horvath offers at best a seriously qualified vindication of Gettier’s original arguments. On his interpretation, they are unsound when taken at face value. Naturally, Gettier himself does not present his arguments as merely first approximations, but as the finished article. Of course, even if an argument is sound—valid and the premises true—it does not follow that we are in a position to know that it is sound. If the explicit ‘properly amended version’ of the hypothetical case must metaphysically entail the case verdict (in this instance, that the subject of the thought experiment does not know), it may be very hard for us to know that no subtle metaphysical possibility has been overlooked, just as Gettier overlooked several—unless we put so much into the description of the case that its possibility trivially amounts to the intended conclusion. One fall-back for Horvath is that although we know that there is an adequate explicit case description, there is no particular explicit case description which we know to be adequate. The unspecific knowledge would be based on our track record of finding appropriate plugs whenever we find holes. Indeed, Horvath repeatedly describes the ‘bottom-up’ part of his strategy as ‘inductive’. Still, that would be a rather disappointing outcome: we know that there are proper Gettier cases out there somewhere, even if we will never knowingly find one. By contrast, the counterfactual strategy does not require metaphysical entailment from explicit case description to case verdict. For Horvath, these disadvantages of his account are outweighed by what he takes to be its advantages over its rivals. In particular, he takes it to do more justice to the role of supposition in Gettier’s arguments. The section ‘Horvath on suppositional truth’ below will analyze that issue in detail. Before that, a difference between two ways of implementing the counterfactual strategy must be explained.

Horvath on Gettier’s Thought Experiments  329

Theories of Counterfactuals As Horvath notes, I have revised my view of counterfactuals, in my more recent work on conditionals. In 2007, I followed the general approach of David Lewis’s classic work (1973), parsing the counterfactual conditional as a two-place modal operator, taking the antecedent and the consequent as separate inputs, in a framework of possible worlds semantics with a suitable similarity relation: very roughly, A □→ C (‘if it were that A, it would be that C’) is true at a world w if and only if the most similar A-worlds to w (if any) are C-worlds (an A-world is a world at which A is true). For a thought experiment, A is the stipulated scenario (the case description) and C is the verdict (the case judgment). That approach worked, but it posed some tricky problems for formalization, with clunky solutions. The scenario for a thought experiment is typically that there are some entities with specified properties and relations, so A is existentially quantified. The verdict on the scenario is typically meant to be about the very entities introduced in the scenario, so one wants singular terms in C to be anaphoric on singular terms in A. But one cannot achieve that with bound variables, because the scope of the variable-binding existential quantifiers in A does not extend to C; one gets something like (∃x Fx) □→ Gx, where ‘x’ in the consequent is left unbound. One can try a universal quantifier outside the scope of the counterfactual conditional, as in ∀x (Fx □→ Gx), but that does not have the intended effect, since some values of ‘x’ may satisfy F only at very far-out worlds, making the truth of the universal generalization vulnerable to what happens in bizarre possibilities, even when the original scenario, ∃x Fx, is true in some quite ordinary possibilities (2007: 195– 99). The compromise I settled on was to use the complex counterfactual (∃x Fx) □→ ∀x (Fx ⊃ Gx) (‘If something were F, everything F would be G’). That more or less does the job, but it feels much clumsier than the English ‘If something were F, it would be G’. The formula cumbersomely copies all the material in Fx from the antecedent into the consequent, and has no anaphora in the consequent corresponding to the economical English ‘it’. In an appendix, I experimented with a very non-standard variable-­binding counterfactual conditional, and obtained a more elegant treatment, but the framework seemed rather ad hoc (2007: 305–8). The key to my new treatment of counterfactuals is that ‘would’ has an independent life as a modal operator, outside conditionals. It is a necessity operator, restricted in contextually determined ways, as in ‘She wouldn’t betray a friend’ and ‘He would, wouldn’t he?’ I will symbolize ‘would’ as □R, where ‘R’ represents the contextual restriction, which we may treat as a proposition, true in all and only the contextually relevant possible worlds. We can analyze the English ‘If A held, C would hold’ by treating the bare ‘if’ (controversially, but defensibly) as the material conditional, and ‘would’ as scoping over ‘if’. The result is simply □R (A ⊃ C), a contextually restricted

330  Timothy Williamson strict conditional (Williamson 2020: 167–213). Arguably, the phenomena Lewis explained semantically in terms of his complex account of similarity relations are better understood pragmatically in terms of the restriction R to worlds relevant in the context (op. cit.: 222–8). The restriction to contextually relevant worlds enables the truth of the counterfactual conditional to be insulated against what happens in unwanted far-out worlds; there can also be restrictions on the quantifiers, if desired. This new parsing of counterfactual conditionals gives more flexibility than Lewis’s account in avoiding irrelevant realizations of the scenarios of thought experiments, because the restrictions are imposed pragmatically rather than semantically. This flexibility gives the (2020) account a general defense against the ‘revenge’ problem of deviant realizations of a hypothetical scenario, a problem which Horvath and others use against the (2007) account. A deviant realization of a thought experimental scenario A is an A-world w which is ‘irrelevant’ in the context of the thought experiment. Pre-theoretically, the falsity of C at a deviant realization does not impugn the correctness of C as a verdict on the thought experiment. By using the Lewis counterfactual A □→ C rather than the (unrestricted) strict conditional □(A ⊃ C) in formalizing the arguments underlying thought experiments, the 2007 account prevented many deviant realizations from undermining the arguments, because those deviant realizations were not amongst the most similar A-worlds to the actual worlds. However, since the Lewis semantics imposes comparatively inflexible constraints on the similarity relation between worlds, that point did not completely solve the problem. For a deviant realization w of A may nevertheless be one of the most similar A-worlds to the actual world; if so, and C is false at w, then the Lewis counterfactual A □→ C is false at the actual world, which on the 2007 account undermines the thought experiment: an unwanted result. In particular, if the actual world happens to be a deviant realization, then it is automatically one of the most similar A-worlds to the actual world, for any world is at least as similar to itself as any world is to it (by Lewis’s Weak Centering constraint). Thus if A is true and C false at the actual world, then the Lewis counterfactual A □→ C is actually false. Horvath emphasizes the ‘revenge’ problem for the 2007 account. Although my 2007 discussion anticipated the revenge problem and gave a defensible response, the latter was somewhat inelegant. By contrast, my 2020 account has a much more straightforward and general response to the revenge problem: if w is a deviant realization in the context of a thought experiment, that is because w violates some contextually relevant restriction, which the contextual restriction R thereby encodes, so R is false at w, so the truth of A and falsity of C at w does not falsify the contextually restricted strict conditional □R (A ⊃ C), which is the relevant counterfactual conditional. Even if w is the actual world, the contextual restriction R can still be false at w, because the actual world

Horvath on Gettier’s Thought Experiments  331 may be deviant in ways which make it contextually irrelevant (op. cit.: 185–6). For some restrictions R, □R (A ⊃ C) does not entail A ⊃ C. To put the general point the other way round, if A is true and C false at a world w which satisfies the contextual restriction R, then w is not a deviant realization; it is simply a counterexample to the counterfactual □R (A ⊃ C). After all, false conclusions are sometimes drawn from thought experiments. Thus, the 2020 account is untroubled by the revenge problem of deviant realizations. A special case of the contextually restricted strict conditional □R (A ⊃ C) is worth noting: when R is true at every possible world, □R (A ⊃ C) is equivalent to the unrestricted strict conditional □ (A ⊃ C). For example, in a metaphysics seminar, counterfactuals can be used to generalize over all possible worlds, with similarity to the actual world irrelevant for those purposes. In principle, the availability of this reading of ‘would’ undermines Horvath’s contrast between supposing-as-counterfactual (corresponding to counterfactual conditionals) and supposing-as-hypothetical (corresponding to unrestricted strict conditionals), which he treats as mutually exclusive. But in a metaphysics seminar, supposing-as-counterfactual may constitute supposing-as-hypothetical. For present purposes, however, that point makes little difference. In any context with deviant realizations, R rules them out and so is false at some worlds, so the counterfactual conditional is not equivalent to the unrestricted strict conditional. Since Gettier’s scenarios have deviant realizations, the relevant R is false at some worlds, so Gettier’s attitude cannot be both supposing-­ascounterfactual (as on my reconstruction) and supposing-as-hypothetical (as on Horvath’s reconstruction). Toward the opposite extreme to a universally satisfied restriction, R may be satisfied only within a very narrow range of worlds. For example, imagine an actual conversation about the conspiracy theory that Oswald had an invincible back-up, ready to kill Kennedy if Oswald failed. The interlocutors need not believe the theory; they may know it to be false. They may simply be interested in exploring its consequences, to the point where any world in which it is false (including the actual world) is irrelevant for conversational purposes. Thus, only worlds in which the theory is true satisfy the contextual restriction R. In such a context, the counterfactual ‘If Oswald had not killed Kennedy, someone else would have’ is actually true (not just believed true). Horvath follows orthodoxy in simply classifying that counterfactual as false, as it is in normal contexts in the actual world, but in principle ‘would’ is more flexible than that, and my 2020 version of the counterfactual approach to thought experiments exploits that flexibility. For applications to thought experiments, we need quantified counterfactuals. The new approach handles them much more easily than its 2007 predecessor. For example, the natural formalization of the English ‘If something were F, it would be G’ is simply □R ∀x (Fx ⊃ Gx); likewise for

332  Timothy Williamson multiple quantifiers. That yields the desired anaphora with the bound variable ‘x’, without awkward repetition of ‘F’. Unlike Lewis’s binary modal operator □→, □R does not separate two inputs to the modal operator and so force a quantifier binding variables in both inputs outside the modal operator’s scope. This streamlining eliminates much of the inelegant complexity of the 2007 account. The arguments presented in The Philosophy of Philosophy as underlying thought experiments can all be adapted without difficulty to the new framework (op. cit.: 229–241).

Horvath on Suppositional Truth Horvath describes his reconstruction of Gettier’s argument as ‘suppositional’. Where I offered a counterfactual conditional as the key premise established by the thought experiment for the underlying argument against the sufficiency of the analysis, Horvath insists that Gettier’s argument involves only the ‘case judgment’, which is just the consequent rather than the whole conditional, but made on the supposition of the antecedent. This matches the style in which Gettier writes of his thought experiments, non-modally and unconditionally, but under the supposition of his case description. On Horvath’s account, no conditional statement linking the case description to the case verdict figures in the argument. Horvath treats this as a central advantage of his account over mine, in both simplicity and faithfulness to Gettier’s text. A first observation is that by keeping the case judgment separate from the case description, and treating each as expressing a self-standing content, Horvath reintroduces the problem of anaphora between description and judgment, solved by my 2007 account at the cost of some complexity and by my 2020 account quite simply. However, the simplicity of the latter account depends on treating the verdict as a necessitated and universally quantified conditional, which Horvath’s suppositional approach rules out. Gettier himself handled the problem by using names such as ‘Smith’ and ‘Jones’ in a lightly fictional way. In order to avoid controversial issues in the semantics of fictional and arbitrary names, Horvath resorts to what he calls the ‘substitution gambit’, which involves notionally substituting actual names of people for Gettier’s fictional names, for instance, ‘Donald Trump’ for ‘Smith’ and ‘Joe Biden’ for ‘Jones’. Since Gettier’s examples are presumably metaphysically possible for any human, this substitution should not affect the possibility of the scenario at issue. Horvath does not pretend that the substitution gambit works for all philosophical thought experiments. Consider this scenario: Lump is a piece of gunk (every part of Lump has a proper part). Given that gunk is metaphysically possible but not actual, it is by no means obvious that there are any actual names of possible pieces of gunk. Thus, substituting an actual name for ‘Lump’ may yield a metaphysically impossible scenario. One may suspect that an adequate semantic treatment of the

Horvath on Gettier’s Thought Experiments  333 fictional name ‘Lump’ above would generalize to an adequate s­ emantic treatment of Gettier’s fictional names, thereby obviating any need for Horvath’s artificial and ad hoc substitution gambit. Another concern is that the substitution gambit has no basis whatsoever in Gettier’s text. It hardly supports Horvath’s claim of faithfulness to Gettier’s text. Indeed, on the descriptive approach to the semantics of proper names dominant at the time, the consistency of the case descriptions would be a tricky issue once actual proper names were substituted in, making it even less likely that Gettier would have been willing to adopt the substitution gambit in 1963. The gambit is no mere optional extra to Horvath’s account. Without it, his suppositional approach lacks a solution to the problem of anaphora. If he replaced the substitution gambit with a less artificial, less unfaithful treatment of Gettier’s fictional names, he might easily find himself back with complexities like those of rival accounts, such as mine of 2007. Thus, his claims for the simplicity and straightforwardness of his account partly rest on the substitution gambit. However, in what follows, I will mainly bracket the gambit and turn to the overall structure of Horvath’s suppositional account. Horvath wants to describe Gettier’s case judgment C as objectively ‘true’ in the context of (a properly amended version of) Gettier’s case description A. Of course, A and C may both be false in the actual world— in which Gettier made his judgment—so Horvath needs some other sort of truth. Following David Barnett (2010), he opts for something called ‘suppositional truth’. As already noted, he takes the type of supposition relevant to Gettier’s arguments as hypothetical (in his sense), and explains that C is true under the hypothetical supposition A if and only if A metaphysically entails C. Thus, truth under a hypothetical supposition does not require ordinary truth. Horvath adds: ‘It should be emphasized that this makes suppositional truth a species of relative truth: suppositional judgments are only true relative to a particular supposition’ (his italics). A few elementary structural features of Horvath’s suppositional truth are worth noting. First, a contradiction is sometimes true under a supposition. For example, since B  ∧  ¬B metaphysically entails itself, it is true under the hypothetical supposition of itself. That sounds like a dose of dialetheism. Second, a disjunction is sometimes true under a supposition even though neither disjunct is true under that supposition. For example, if B is contingent, B ∨ ¬B metaphysically entails neither B nor ¬B, although it metaphysically entails itself; thus although B ∨ ¬B is true under the hypothetical supposition of B ∨ ¬B, neither B nor ¬B is true under that supposition. For similar reasons, if ¬B is true under a supposition, it does not follow that B is not true under that supposition, and if B is not true under a supposition, it does not follow that ¬B is true under that supposition. Horvath writes in a way which implicitly identifies the falsity of B with the truth of ¬B. Thus, he is committed to both truth-value gaps

334  Timothy Williamson and truth-value gluts for truth under a supposition. For gluts: since a contradiction metaphysically entails everything, every sentence is both true and false under a contradictory hypothetical supposition. For gaps: whenever A is metaphysically compossible both with B and with ¬B, B is neither true nor false under the hypothetical supposition A. In the latter case, Horvath describes B as ‘indeterminate’, slightly misleadingly, since B is determinately neither true nor false under the supposition. There are many different ways of adapting logic to the presence of truth-value gaps and gluts. For example, we might define a valid argument as one which preserves truth from premises to conclusion under all assignments, or as one which preserves non-falsity. The argument from two atomic sentences p and q to their conjunction p ∧ q is valid in the former sense but not in the latter. Horvath does not specify what approach he would adopt. Of course, he might say that truth under a supposition is irrelevant to validity, but that would involve not taking truth under a supposition seriously as a form of truth, which is inconsistent with his overall account. To support the ascription of truth to case judgments in successful thought experiments, Horvath comments, ‘It is a platitude […] that arguing for a judgment J typically just is arguing for the truth of J, and it would be surprising if arguing had a very different point or function in the case of suppositional judgments’. By apparent analogy, it is also a platitude that arguing against a judgment J typically just is arguing against the truth of J, and again it would be surprising if arguing had a very different point or function in the case of suppositional judgments. But ­Horvath’s approach violates the latter platitude, for one can argue decisively against the truth under a hypothetical supposition A of J by showing that A does not metaphysically entail J, but that does not constitute a decisive argument against J itself, for that would require a decisive argument for ¬J, but the fact that A does not metaphysically entail J establishes neither ¬J nor that A metaphysically entails ¬J. Thus, Horvath’s platitude is in danger of indirectly undermining his case for suppositional truth. Since truth under a supposition behaves so radically differently from ordinary truth, we will need to be exceptionally careful in evaluating arguments whose component sentences are to be assessed in terms of truth under a supposition rather than ordinary truth. To complicate matters further, Horvath’s account of Gettier’s argument involves both kinds of assessment. The case judgment C is to be assessed for truth under supposition A, but the final conclusion about the possibility of justified true belief without knowledge is to be assessed for ordinary truth, since it is drawn under no supposition. Thus, we need a metalogical framework that explains how the two kinds of assessment interact. Horvath offers one or two isolated remarks but no framework. The simplest solution would be to reduce the truth of C under the hypothetical supposition A to the ordinary truth of the strict conditional

Horvath on Gettier’s Thought Experiments  335 □(A ⊃ C), and then do the underlying metalogic in terms of ordinary truth alone. But that is too close to sneaking in the very conditionals that Horvath prides his suppositional approach on avoiding. To vindicate the spirit of his approach, he needs to do his metalogic in terms of the combination of truth under a supposition and ordinary truth. Presumably, that could be done, although Horvath has not shown how, and it would surely involve considerable complications in the metalogic. Two comments are in order. First, we can now see the apparent simplicity and straightforwardness of Horvath’s account to be illusory (as the discussion of the substitution gambit already hinted). He has not eliminated complexity; he has merely shifted it into the metatheory, which has to deal with all the complications of truth-value gaps and gluts. Second, nothing in Gettier’s article suggests that his arguments require any sort of non-standard metalogical framework, let alone one with truthvalue gaps and gluts. If Gettier had thought that way, he would have been grossly irresponsible in not warning readers of such controversial commitments. The introduction of a weird sort of relative truth leads away from what Gettier was doing, not toward it. Together with the substitution gambit, it undermines Horvath’s claim ‘This suppositional reconstruction comes closer to Gettier’s actual method than other extant reconstructions’.

Supposition and Conditionals Nothing I have said means that I reject Horvath’s emphasis on the role of supposition in Gettier’s thought experiments. Supposition has long been central to my account of how we assess counterfactuals and other modal claims (2007: 147–65, 185–9; 2020: 189–213). Our primary way of assessing a counterfactual is by counterfactually supposing its antecedent, and then assessing its consequent on that supposition, typically by an exercise of imagination; we then make our conditional assessment of the consequent on the antecedent into a corresponding unconditional assessment of the whole conditional. Gettier’s style of presentation is silent over that last stage of assessment. But it is also silent over the possibility of the antecedent, which is needed to derive the possibility of the consequent, as Horvath and I agree. Since the function of the counterfactual conditional in the overall argument is to get us from the possibility of its antecedent to the possibility of its consequent, the absence of an explicit conditional in Gettier’s account is no surprise. He leaves the final stage implicit, rightly confident that readers will take it for granted. The absence of an explicit counterfactual conditional no more shows that it is not implicit than the absence of an explicit possibility ascription to the antecedent shows that it is not implicit. Schematically, in my 2020 framework (with a single variable for simplicity), we derive Gx under the counterfactual supposition Fx with background restriction R. That enables us to discharge the suppositions and

336  Timothy Williamson assert the counterfactual conditional □R ∀x (Fx ⊃ Gx) unconditionally. Given the possibility premise ◊R ∃x Fx, we can then derive the required conclusion ◊R ∃x Gx by standard quantified modal logic. The possibility at issue has the same contextually supplied restriction as the counterfactual conditional itself, for if the case description were metaphysically possible but not compossible with the restriction, we could not use the conditional to derive even the metaphysical possibility of the case judgment. Once we have the restricted possibility ◊R ∃x Gx, a fortiori we have the unrestricted possibility ◊ ∃x Gx too. In principle, the conditional is not indispensable in that reasoning. Having derived Gx under the counterfactual supposition Fx with background restriction R, one could invoke a special rule which allowed one to move directly from the premise ◊R ∃x Fx to the conclusion ◊R ∃x Gx, with no mediating conditional. But that rule is rather ad hoc. It is more perspicuous to obtain the counterfactual conditional in the standard way, and then apply standard quantified modal logic to reach the same final conclusion. The restriction R corresponds to nothing in Horvath’s account, since he claims that the suppositions in Gettier’s paper are hypothetical (in his sense), so no background assumptions are allowed. Only what a hypothetical supposition strictly implies can be accepted on that supposition (he makes an exception for ‘semantic implicatures and presuppositions’, which we can ignore for present purposes). In his view, Gettier’s argument is sound only if his case description metaphysically entails his case judgment. We have already seen how difficult that makes the problem of deviant realizations for Horvath. By contrast, if a realization is genuinely deviant, it is ipso facto ruled out by the contextually supplied restriction R, as explained in the section ‘Theories of counterfactuals’ above. Since Horvath presents his reconstruction as more or less vindicating Gettier’s arguments in their original form, it is worth emphasizing that (as he acknowledges) his reconstruction makes the deviant realizations guarantee those arguments’ unsoundness, while my 2007 reconstruction does not (since the deviant realizations may well be confined to distant worlds), and the 2020 version avoids the problem altogether. In that way, Horvath’s suppositional approach creates a far more hostile environment for Gettier’s arguments than does the counterfactual alternative. In short: Horvath, despite himself, vividly illustrates the dangers that motivate understanding thought experiments in terms of counterfactual conditionals.

Acknowledgment I am grateful to Joachim Horvath for stimulating discussion and correspondence on these topics, and for giving me the opportunity to write this piece.

Horvath on Gettier’s Thought Experiments  337

References Barnett, David. 2010: ‘Zif Would Have Been If: A Suppositional View of ­Counterfactuals’, Noûs, 44: 269–304. Gettier, Edmund. 1963: ‘Is Justified True Belief Knowledge?’, Analysis, 23: 121–123. Horvath, Joachim. this volume. ‘Gettier’s Thought Experiments’. Lewis, David. 1973. Counterfactuals. Oxford: Blackwell. Williamson, Timothy. 2007: The Philosophy of Philosophy. Oxford: Blackwell. Enlarged edition, 2021. Williamson, Timothy. 2020: Suppose and Tell: The Semantics and Heuristics of Conditionals. Oxford: Oxford University Press.

17 Challenges for an Experimentalist’s Skepticism about Cases Margot Strohminger and Juhani Yli-Vakkuri

Introduction In Philosophy Within Its Proper Bounds (PPB), Edouard Machery (2017) argues that philosophical practice is badly in need of change. Machery uses data surveying folk responses about “cases” or examples used in philosophical arguments to argue that philosophers should—at least for the most part—avoid what he and some other metaphilosophers call the “method of cases”. As a result, he argues, philosophers will have to abandon the quest for knowledge of non-trivial truths about what is and is not metaphysically necessary. Instead, we will have to adopt new aims, chief among which is the study and revision of concepts. One of the greatest virtues and vices of PPB is its vagueness. The vagueness is a virtue from our point of view because it makes it possible for us to construe a great deal of what Machery says in such a way that we agree with it. In some areas of philosophy, in some literatures, there is something resembling Machery’s description of the method of cases at work: a theory—say, the justified true belief (JTB) theory of knowledge—is articulated and then modified to accommodate a litany of apparent counterexamples, leading to increasingly inelegant versions of the original theory, and sometimes sprawling subliteratures about the merits of particular proposed counterexamples and variations on them. Machery is not a fan of this process and neither are we. (One of us has had critical things to say about it in relation to certain debates in the philosophy of mind1 and in fundamental metaphysics.2) Both Machery and we think that philosophy needs to be more scientific. We even agree with Machery that debates about what is and isn’t metaphysically necessary is sometimes pointless and should be abandoned, and we have said so in print.3 We certainly don’t want to express any disagreement with the conclusions of two of Machery’s main arguments, which are as follows: There are many philosophical issues that we cannot resolve (PPB, 187). We lack knowledge of many metaphysical necessities that are of philosophical interest (ibid.). DOI: 10.4324/9781003002192-19

Challenges for an Experimentalist’s Skepticism about Cases  339 And while we don’t think that the revision of concepts should be the aim of philosophy, we think that good philosophy almost inevitably involves the revision of concepts in some sense—as when one replaces a vague concept with a less vague one that is better capable of doing the theoretical work for which the original concept was adopted. We think philosophers should be doing more of just that kind of thing: identifying dysfunctional concepts and replacing them with better ones. But we think philosophers are already doing it and have been since the dawn of analytic philosophy (Frege, Russell, and Church on numbers and other mathematical notions, Russell on definite descriptions, Church on computability, Tarski on truth, definability, and logical consequence), and on the whole we don’t see much to complain about. PPB’s vagueness is a vice insofar as it makes it difficult for us to say just where we disagree with its characterization of contemporary philosophical methodology and its calls for reform. (It is determinate that we disagree about a lot but there is very little such that it is determinate that we disagree about it.) Part of the reason, no doubt, is the subject matter itself, which makes it very difficult to be precise enough to have a productive exchange. Often, in surveying the philosophical landscape, the best one can do is use language that is so vague that it’s hard to know how to argue for or against claims couched in it: for example, vague generics about what different kinds of “philosophers tend to” do (8, 90, 113, 153, 193) and about what their “cases tend to” do (91, 111, 115, 119, 122). We don’t fault Machery for using them. We have some vague generics of our own in what follows. Some of our disagreement with Machery likely boils down to whether we see the glass as half-empty or half-full, but we may also be looking at different glasses or different parts of the same glass. From where we stand, it looks like the kind of methodological reform we would like to see is already under way, and the areas of philosophy we follow actively are making rapid progress. For example, five years ago, philosophers who specialize in the topic knew significantly more about metaphysical necessity than they did in the 1980s, and now they know significantly more about it than they did five years ago. (One change that made a significant difference is the shift of focus from objective necessities to necessities simpliciter—a conceptual reform. As a result, we are now in a position to say that our best theories tell us that there is a unique strongest necessity and that it is at least an S4-necessity, and its being an S5-necessity is about as certain as the axiom of choice.4) But one might well begin to feel discouraged if one focused one’s energies on areas of philosophy where progress is less rapid. But let’s accentuate the negative, as is customary in exchanges like this. In what follows, we will pose some challenges for Machery. In the first section, we try to understand what is meant by “the method of cases”. In the second section, we ask why Machery’s attack on the method of

340  Margot Strohminger and Juhani Yli-Vakkuri cases doesn’t defeat itself if it defeats anything. In the third section, we question some assumptions Machery makes about the epistemology of disagreement and the import he takes them to have. In the fourth section, we ask whether the method of cases is needed to evaluate “modally modest” theories by Machery’s own lights. In the fifth section, we ask why Machery’s critique of “modally immodest” philosophy isn’t itself modally immodest. In the sixth and final section, we consider Machery’s prescriptions for reform and ask some further questions.

What Is the Method of Cases? According to Machery, the method of cases is seldom capable of producing knowledge (PPB, 7–8). The rough idea is that experimental philosophy (x-phi) studies about some of the most well-known cases in contemporary philosophy show that the method fails to produce knowledge of what is true in those cases, and we should expect the same to go for other cases not yet examined by x-phi. Machery’s skepticism is not total: in his view, the method of cases can still produce knowledge albeit in very limited circumstances. To begin to evaluate these claims, we need to get clear on what the method of cases is. Here is how Machery describes the method: Philosophers rely on the method of cases when they consider actual or hypothetical situations (described by cases) and determine what facts hold in these situations. These facts then bear, more or less directly, on competing philosophical views (PPB, 3). Machery then characterizes cases as follows: “Cases are descriptions of actual or hypothetical situations” (PPB, 11). If we rely on this characterization alone, any judgment we make will count as the product of the method of cases: when we judge that p, we consider the actual world (the actual maximal situation) and we judge p holds in it. We suggest that a better characterization of the method construes cases as examples (especially but not only attempted counterexamples). On this construal, we are relying on the method of cases when you tell us about someone’s epistemic situation and we judge that they have a JTB that isn’t knowledge—that is, we take your example to be a counterexample to the JTB theory of knowledge. This “someone” might be a fictional character but it might also be an actual person. By contrast, we are not relying on the method of cases when we judge that objects that are identical are necessarily identical, and we do so on the basis of a proof from logical axioms, such as the one made famous by Kripke in “Identity and Necessity” (1971). This characterization seems to better fit philosophers’ usage of the term “method of cases”. It is also more charitable to

Challenges for an Experimentalist’s Skepticism about Cases  341 Machery, who illustrates the method exclusively using judgments about alleged counterexamples to philosophical theories (e.g. PPB, 3–4). Being an example is not a property of states of affairs but a relation in which a state of affairs stands to a context (such as a position in a philosophical text) in which it is discussed. For example, the necessity of identity (NI) is not an example in “Identity and Necessity”. But NI is an example in the text you are reading now: in the previous paragraph, we used it as an example of a state of affairs that is not a case in one text, and in this paragraph, we just used it as an example of a state of affairs that is an example in another text. We agree with Machery that overindulgence in examples can be a problem in philosophy. On some topics in philosophy, one cannot expect to find exceptionless generalizations, and a model-building approach similar to what we find in economics is appropriate (attempted counterexamples to models simply miss the point of models). Some parts of epistemology are like this, especially and unsurprisingly the ones with significant overlap with economics. As Williamson’s (2013, 2015) recent work on Gettier cases in epistemic logic shows, the model-building approach can lead to surprising insights also on questions that have traditionally been approached in an example-driven way. In the philosophy of mind, the vast literature on semantic internalism vs. externalism presents itself as another case study in how a topic can “go stale” (to use David Chalmers’ [2018] term) when philosophers focus their energies on the details of apparent counterexamples to proposed theories of subject matter for which the model-building methodology is more appropriate.5 In contrast, in fundamental metaphysics, we can sometimes expect to find exceptionless generalizations: modal metaphysics is a good example. But in that area, it seems to us, philosophers don’t spend much time pondering the details of alleged counterexamples to the proposed generalizations, and the methodology is often self-consciously and explicitly abductive, as we think it should be. (We’ll come back to this.) If we use “the method of cases” as a label for the activity of studying the details of particular examples of the phenomena of interest instead of building models or theories of the phenomena, then we are on board with the idea that the method of cases is widely applied in some corners of philosophy that could make more progress by abandoning it and focusing on either model-building or theory construction, depending on which is more appropriate. This strikes us as a reasonable criticism of some literatures in analytic philosophy, and it isn’t a mile away from what we see as the most charitable reconstruction of Machery’s attack on the method of cases, but the two are different in two important ways. First, Machery is not advising us to carry on with our model-building and theory construction in an abductive spirit and to reject the philosophical analogue of the kind

342  Margot Strohminger and Juhani Yli-Vakkuri of “naïve falsificationist methodology in natural science, which treats a theory as refutable by a single observation” (Williamson 2016, 279–280), whereas that would be our advice. Second, Machery tells us that the use of the method of cases is permissible in some situations, while the methodological stance in the vicinity that we find plausible does not allow that there are special circumstances in which the philosophical analogue of naïve falsificationist methodology is appropriate. We don’t pretend to know whether Machery uses “the method of cases” for the kind of activity just gestured at. Perhaps he does, and we only disagree on what should replace it. But we want to try to be clearer about what we think is problematic about example-driven philosophizing. It’s not that we think that a single fairly specific counterexample can never decisively refute a theory. We share what seems to be the majority view on Gettier’s cases and standard Twin Earth cases.6 It’s rather that we think that time spent debating whether a particular proposed counterexample really is one and trying to come up with better counterexamples is usually better spent by retreating to a greater level of generality and constructing models or theories and evaluating them by the usual standards of model or theory choice. The necessitism/contingentism debate, to be referenced later, is a good example of such a successful retreat.

Does PPB Apply the Method of Cases Appropriately by Its Own Criteria? Let’s now try to understand what the situations in which Machery thinks reliance on the method of cases is permissible have in common. Machery develops two arguments for the position that we should seldom rely on the method of cases. The first argument focuses on the extent to which untrained judgments about cases are influenced by demographic variables and the way in which the cases are presented (this is the argument labeled “Unreliability” in Chapter 3). The second argument is a dilemma which has us consider whether experimental data reveals genuine disagreement between philosophers and ­non-philosophers about cases (this is Chapter 4’s dilemma between “Dogmatism” and “Parochialism”). Both the Unreliability and Dogmatism arguments conclude that philosophers ought to suspend judgment in all assumptions about cases with the exception of those known to have certain features. According to the conclusion of the Unreliability argument, “except when a philosophical case is known to elicit a reliable judgment, philosophers ought to suspend judgment about the situations described by philosophical cases” (PPB, 103). And according to the conclusion of the Dogmatism argument, “except for those philosophical cases known not to elicit disagreement among peers,

Challenges for an Experimentalist’s Skepticism about Cases  343 ­ hilosophers ought to suspend judgment about the situations described p by philosophical cases” (PPB, 127). These conclusions suggest that Machery thinks that one may only rely on the method when one knows both that the judgment it elicits is reliable and that one’s epistemic peers share the resulting judgment (for the moment we are not going to worry too much about what this means exactly7). We can identify three kinds of judgments in PPB that appear to be products of the method of cases. First, Machery makes various assumptions about how to interpret x-phi survey data. For example, Machery thinks that the judgments reported on by certain x-phi studies are influenced by order and framing effects to degrees large enough to threaten their reliability (PPB, 88, 93), and that philosophers and the folk are “epistemic peers” when it comes to determining what is true in various cases (PPB, 130–131). In making these estimations, Machery seems to be employing the method of cases. Machery usefully tells us that “epistemic peerness is always relative to some topic—x and y can be epistemic peers relative to one topic, but not to other topics”, and that x and y are not epistemic peers with respect to a topic if x is “influenced by reasoning biases that do not influence y” or “has less evidence than she seems to have” or has “not considered [the issue] as carefully as y has” (PPB, 130). The complex judgment calls one has to make in determining whether these conditions hold between the philosopher and lay respondents with respect to each issue on which x-phi surveys query them look very similar to paradigm applications of the method of cases. Would an epistemology paper that described in some detail a series of hypothetical x-phi studies and asked each of them whether the philosopher and lay respondents in the studies were epistemic peers count as thereby applying the method of cases? It seems to us that the answer would have to be “Yes”. If it is, it’s hard to see why Machery’s judgments about epistemic peerhood in actual x-phi studies should not count as applications of that method. Second, Machery sometimes describes hypothetical scenarios and uses them as counterexamples to philosophical views. Consider, for example, the following passage: From the fact that V influences J, it does not follow that J is unreliable. Suppose that the color of the room in which one is asked a question about global warming influences one’s opinion: People in a bright room are more likely to assert that global warming is real than people in a dark room. Suppose, however, that the difference is small: 80 percent of people in a bright room assert that global warming is real, for 75 percent of people in a dark room. And suppose that, as is the case, global warming is real. Then, if people are equally likely to be in a bright and dark room, 77.5 percent of people make a

344  Margot Strohminger and Juhani Yli-Vakkuri true judgment despite the normatively inappropriate influence of the color of the room. People in this fictional example are reliable. PPB (104) Here, Machery is using a hypothetical example of a judgment in defense of the view that a judgment can be influenced by a factor of a certain type without being unreliable. He is not, it seems, using the case simply to illustrate a conclusion he has reached by some other means. He argues in a similar vein against the principle that “You ought not to decide to φ in order to ψ if you believe you are not justified to believe that φ-ing is likely to bring about ψ”, which he labels “Premise 4” (PPB, 140): There are some counterexamples to Premise 4. If you are in a maze and must choose between turning right and left (or must prove a theorem by embracing one of two different strategies), it seems rational to choose either turn despite the fact that you believe not to be justified to believe that either will allow you to exit the maze. PPB (147) Machery is arguing against the principle in question by appealing to an assumption about what it is rational for one to do if faced with a choice between two paths in a maze. Machery also argues against the view that “judgments about hypothetical situations are unreliable” by using a case: “Simply consider the judgment that this book would fall instead of going up if you released your grip on it” (PPB, 113). Third, Machery apparently employs the method of cases indirectly: by relying on assumptions made by other philosophers, which they seem to support in part by the method of cases. At various points, Machery relies on views defended in other philosophical work, especially in epistemology. The Dogmatism argument assumes that one or another of the “leading views” in the epistemology of disagreement is true. Those views were developed in large part by thinking about various hypothetical cases involving disagreement.8 Christensen’s (2007, 193) “Restaurant Check Case” is an especially prominent example but the literature is full of hypothetical cases.9 Machery offers us no assurance that he can defend his stance on the epistemology of disagreement without relying on this prior work in epistemology. Machery rightly points out that it is easy to overlook how one’s arguments rely on the method of cases: chains of reasoning can be long and stretch across several papers and books and use as assumptions claims (or disjunctions of claims, as above) argued for by other authors who use the method.10 Thus, by his own lights, Machery’s reliance on the method of cases would seem to be much more widespread than a simple search for cases in his text reveals.

Challenges for an Experimentalist’s Skepticism about Cases  345 We seem to have identified three kinds of applications of the method of cases in PPB. If that is what they are, then to be justified in making these applications of the method of cases, Machery, by his own lights, must know that the judgments produced by them are reliable and shared with his epistemic peers. Does he? Let’s see what Machery himself has to say about the matter. First, Machery discusses work that rejects his interpretations of x-phi surveys, and this suggests that his interpretations are not shared by his epistemic peers. Consider, for example, Machery’s judgment that judgments reported on by certain x-phi studies are influenced by order and framing effects to degrees large enough to threaten the reliability of the method used. According to Machery (PPB, 88, 93), the literature has already questioned whether this is so (see e.g. Demaree-Cotton 2016; Horne and Livengood 2017). Second, it’s safe to assume that Machery’s hypothetical counterexamples to other philosophers’ views will not end the disagreement about the correctness of those views. Third, Machery supports his key claim about the epistemology of disagreement by telling the reader that “it follows from the leading views in the epistemology of disagreement” while acknowledging that “some other views about disagreement may lead to another conclusion” (PPB, 131). Are these in fact applications of the method of cases, and if so do his epistemic peers share his judgments about them? We’ll let Machery respond.

Epistemic Peerhood, Disagreement, and X-Phi In the previous section, for the sake of argument, we went along with Machery’s ideas about the conditions under which reliance on the method of cases is permissible, but in fact those ideas seem pretty wrong-headed to us. First, there are the “leading views” in the “epistemology of disagreement” that Machery assumes are true (more precisely, that the disjunction of them is true). We don’t think they are true, but we don’t want to start arguing against them here. Machery doesn’t want to argue for them in his book either—he simply assumes them (or their disjunction)—so we assume it’s dialectically appropriate for us just to register our disagreement and leave the matter at that. Second, there is the idea that we should have such high confidence in what we take to be the views of the experts in some very recent, niche area of epistemology (the “epistemology of disagreement”) that we should suspend judgment in virtually all of the rest of philosophy because of what Machery thinks (but the relevant experts likely don’t think) follows from those views. This idea we find on the face of it bizarre and desperately in need of defense. We already noted that the niche area in question does not seem to be methodologically sound by Machery’s lights. And Machery apparently rejects the view that philosophers who do that kind of

346  Margot Strohminger and Juhani Yli-Vakkuri example-based epistemology have any expertise to defer to (PPB, chapter 5, section 3). But these ad hominem points are not the ones we want to highlight. What we want to highlight is simply the prima facie bizarreness of selecting a recent niche sub-area, deciding to defer to the presumed expertise of the philosophers working in that niche, and concluding on that basis that we should suspend judgment in everything we take ourselves to know about standard cases that are widely accepted as having refuted important philosophical theories (Gettier, Gödel-Schmidt, Twin Earth…). Of course, deferring to the leading views in every recent niche literature would lead to intellectual paralysis (because some combinations of such views are inconsistent). So there must be something very special about the “epistemology of disagreement” that justifies treating it with this kind of deference. What does Machery think is the special feature of the “epistemology of disagreement” that confers on its practitioners such awesome expertise? Third, we find approximately equally prima facie bizarre the idea that philosophers and non-philosophers are “epistemic peers” when it comes to any questions that philosophers specialize in. Since others have already argued for our view on this matter capably,11 we won’t add anything to what they have said. But we do want to emphasize that learning to figure out what’s true in counterfactual scenarios relevant to competing theories is part of a good philosophical education. It’s to be expected that people with no training in it won’t be as good at it as people who have a lot of training in it. Most importantly, philosophers have a degree of expertise about the subject matter of the cases—knowledge and reference, for example—that non-philosophers lack. We might put this in Machery’s language as follows: non-philosophers’ concepts of (e.g.) knowledge and reference are very different from, and worse than, philosophers’ concepts of knowledge and reference because philosophers have been trained to think systematically about knowledge and reference, to defend and to criticize competing theories of knowledge and reference, to formalize arguments about knowledge and reference in particular ways (for example, with “S knows that” treated as a sentential operator), and so on. (See the final section of this chapter for discussion of what Machery means by “concept”.) But it does not follow that non-philosophers mean different things by the relevant words. While philosophers often use “reference” and “refer” as terms of art, they use “know” to mean exactly what all normal English speakers mean by it.

What Does the Method of Cases Have to Do with Modal Immodesty? And How Much Modally Immodest Philosophy Is There? Machery uses his skepticism about cases to defend a form of modal skepticism, which he, in turn, uses to argue that philosophers ought to

Challenges for an Experimentalist’s Skepticism about Cases  347 “[set] aside modally immodest philosophical issues” (PPB, 186). Modally immodest issues are those whose resolution requires “epistemic access to metaphysical necessities and possibilities”—apparently, claims of the forms “□ϕ” and “◇ϕ” that are not theorems of logic and are not logical consequences of known non-modal facts (or do not “hold in virtue of logic”, as Machery puts it).12 The idea is that in such cases determining whether “□ϕ” is true would require the use of the method of cases, and in particular its application to “unusual and atypical” cases which might witness the truth of “◇¬ϕ”, and we cannot reliably judge what is true in such cases (PPB, 193–194). There is a problem here: if the method of cases, and especially its problematic applications to unusual and atypical cases, is required for evaluating modally immodest theories, as Machery claims, then they are also required for evaluating modally modest theories. Let’s consider an example: the modally modest version of the JTB theory of knowledge, which only says that something is an instance of knowledge iff it is an instance of JTB (a universally quantified material biconditional—no modal operators). This is a theory that only purports to tell us about “what knowledge is like in the actual world”—it says nothing about what knowledge would have been like if things had been otherwise. But, as Machery recognizes, there are actual Gettier cases, which by his lights are unusual and atypical (PPB, 113, 183). And in any case there are cases of knowledge that are unusual and atypical by anyone’s lights, and the modally modest JTB theory of knowledge is false if even one of those unusual and atypical cases is a counterexample to it. So how could we ever know whether the theory is true? Apparently, by Machery’s reasoning, we couldn’t. What Machery ought to be objecting to, if he wants theories to which unusual or atypical cases are irrelevant, is not the necessity operator but the universal quantifier. But he doesn’t object to it. Why? There is another reason to doubt that Machery would be satisfied with modally modest theories, at least if we take his characterization of them as telling us only what their subject matter is like “in the actual world” at face value. No other science only cares about whether its generalizations hold in the actual world—worthwhile scientific generalizations generally support counterfactuals.13 Philosophy shouldn’t either, and doesn’t, perhaps with some marginal exceptions, such as the law ∀p(p ↔ @p) (“Everything is as it actually is”), which is quite literally true only in the actual world: if anything had been different from the way it actually is, it would have been false. (However, the sentence that expresses that law would have expressed a truth in a wide variety of counterfactual circumstances, and that is what makes it interesting.) Whether modal modesty is what Machery ought to be calling for or not, it still bears asking to what extent work in contemporary philosophy is modally immodest. There is a great deal of work even in modal metaphysics that is extremely modally modest. Williamson’s Modal Logic as

348  Margot Strohminger and Juhani Yli-Vakkuri Metaphysics (2013) is perfectly modally modest, even though it argues for a very bold and counterintuitive thesis about what is metaphysically necessary: necessitism, the thesis that it is metaphysically necessary that everything metaphysically necessarily exists (in the sense of □∀x.□∃y. (x = y)). It follows that you, the reader, necessarily exist. But these are consequences of the logic Williamson defends, so they are not “metaphysical necessities” in Machery’s sense, as we read him. Work that defends contingentism (the negation of necessitism)14 is not perfectly modally modest, since it argues that some things do not necessarily exist, which is not a theorem of any contingentist logic, but it does not use any “unusual or atypical” cases to motivate this view but only cases like: ( 1) You could have failed to exist (if you had never been born). (2) The Eiffel Tower could have failed to exist (if it had never been built). And so on. And you will find nothing like the paradigms of the method of cases in this literature: in it, examples like (1) and (2) are not described in any detail, and most importantly no one in the literature (that we have read) takes them to directly refute necessitism. Rather, the contributors to the literature develop necessitist and contigentist theories, and it is common ground that the theories are to be judged by their theoretical virtues—strength and simplicity—not by how well they fit with our pre-theoretic intuitions about cases like (1) and (2). To take a more recent example, Dorr, Hawthorne, and Yli-Vakkuri’s The Bounds of Possibility (2021) not only assumes nothing beyond principles of basic modal logic (what they call “our logic” in Chapter 1) but it is not even primarily concerned with metaphysical modality: it is about all modalities, and the arguments it finds to be the most compelling are couched in terms of the most scientifically respectable of all modalities (by the standards of natural science): objective chance (Chapter 8).15 It takes seriously the “hypertolerantist” idea that just about anything not ruled out by basic modal logic is metaphysically possible: perhaps, for example, it is metaphysically possible for all of us to be poached eggs. And the book is not an outlier in this respect: starting with Armstrong (1986, 1989) and Lewis (1986), much of the rest of the literature seriously considers or even endorses some variant of that view. The idea that any pattern of instantiation of fundamental (physical) properties and relations that is not ruled out by logic is metaphysically possible has been precisified in a number of defensible ways, 16 and, if it is correct, then it would seem that there are no metaphysical necessities or possibilities at all in Machery’s sense. Naming and Necessity isn’t the last but the first word—or one of several first words—on these matters. Kripke’s core ideas about the structure of metaphysical necessity are upheld by current work at the cutting edge, but his examples of particular metaphysically necessary truths remain highly controversial.

Challenges for an Experimentalist’s Skepticism about Cases  349

Is PPB Modally Immodest? Another question we have for Machery is this: How are his own modal claims not modally immodest? Among the main conclusions of the book are two forms of skepticism (one about the method of cases and the other about metaphysical modality), and it is hard to see how to avoid using modals like “can” in stating them. Machery himself uses them: for example, in writing that “modally immodest issues cannot be resolved” (PPB, 2, emphases added). Machery’s view is not merely that we do not know various claims but that we cannot know them. Moreover, he invokes modal notions such as reliability and non-accidentality in defending assumptions on which these conclusions rest. One of his two arguments against the method of cases is stated in terms of “reliability”, the key claim being that certain applications of the method of cases are “unreliable” (e.g. PPB, 10). Or consider another central assumption: “it is not an accident that philosophical cases tend to have these characteristics, since to bear on some central material-mode issues … philosophers must appeal to cases displaying them” (PPB, 111, emphases added). All of the “can” claims entail metaphysical-possibility claims: since metaphysical necessity is the strongest necessity, metaphysical possibility is the weakest possibility. Less obviously, all of the “must” claims entail metaphysical-necessity claims, by the logical equivalence of restricted necessity claims with corresponding unrestricted (metaphysical) necessity claims. The relationship between metaphysical and restricted necessity is structurally analogous to that between restricted and unrestricted quantification. One cannot escape commitment to claims of absolute generality by restricting one’s quantifiers, say, to things on Earth: in saying that there are (Earth-­restrictedly) no talking donkeys, one is saying that everything (unrestrictedly) is either not a talking donkey or is not on Earth. Similarly, in saying that ­(X-restrictedly) one cannot know such and such one is saying that it is metaphysically necessary that, if the X conditions hold, then one does not know such and such.17 This is a serious problem for Machery: his “must” and “cannot” claims are not about “what knowledge is like in the actual world” (PPB, 2) but about what knowledge is like in all metaphysically possible worlds in which the X conditions hold—where X is whatever restriction he has in mind. How, if not by examining “unusual and atypical cases” in which the X conditions hold, can we determine what knowledge is like in all metaphysically possible worlds in which the X conditions hold?18 For a concrete example, consider the claim that we cannot know whether zombies are metaphysically possible. And suppose the relevant conditions are the actual current cognitive capacities of the human species. The claim then amounts to this: no metaphysically possible world in which humans have their actual current cognitive capacities contains any case—no

350  Margot Strohminger and Juhani Yli-Vakkuri matter how atypical or unusual—in which we know whether zombies are metaphysically possible.

Prescriptions for Reform How should philosophers be going about their work, if not by applying the method of cases, however exactly that is construed? Machery promotes two kinds of projects in particular, which are supposed to be possible without any troubling reliance on the method of cases. One project is the construction of what he calls “modally modest” theories. A second is the investigation and reform of concepts. We already made our concerns about the first project known: unusual and atypical cases are still relevant to modally modest theories. As far as we can tell, the only way to ensure that one’s generalizations are not vulnerable to unusual and atypical counterexamples is to not make any generalizations at all. (One way to do so would be to engage in model-building. Models aren’t true or false, so there are no counterexamples to them.) While Machery cites different examples than we do, it’s unclear to us whether there is any disagreement between us about matters of principle. For example, Machery contrasts Woodward’s work on causation (methodologically sound, by his lights) with work that proposes ever more complex counterfactual analyses of causation in response to apparent particular counterexamples (not methodologically sound, by his lights). We are sympathetic with these judgments. However, pace Machery, what’s distinctive about Woodward’s work is not that it isn’t in the business of looking for generalizations about causation that are metaphysically necessary (PPB, 191–192), but rather that it isn’t in the business of looking for generalizations that are true. It’s in the business of studying causation via simplified models of causation—an approach that is often appropriate in non-fundamental metaphysics.19 Similarly, work in epistemology that makes use of epistemic logic typically “assumes” that the knowing subject is logically omniscient, but it would be wrong to characterize it as attempting to tell us “what knowledge is like in the actual world” as opposed to in all metaphysically possible worlds. Idealizing assumptions are made without regard for their truth in the actual world. As for the investigation and reform of concepts, the recent work on metaphysical necessity we mentioned in the introduction to this chapter contains some good examples. If Kripke meant anything definite by “necessity” and “in the highest degree” when he characterized metaphysical necessity as “necessity in the highest degree”, he didn’t tell the reader what it was, and much of the subsequent literature followed Kripke in using an intuitive, informal notion of metaphysical necessity. The recent work enacts major conceptual reforms by replacing the intuitive, informal notion with a variety of candidate formal notions that all agree on the key results. It also replaces, as we noted, the notion of an “objective

Challenges for an Experimentalist’s Skepticism about Cases  351 necessity”—a holdover from Naming and Necessity—with a better notion: “necessity”.20 While “objectivity” could only be characterized in terms of other intuitive notions like “guise-sensitivity” and “non-epistemicity”, “necessity” can be given an explicit definition in higher-order logic.21 The results have been encouraging. This new work shows that (i) the existence of a unique strongest necessity, □, is provable in a natural strengthening of Church’s (1940) higher-order logic (a strengthening that is still weaker than what is nowadays standardly called “classical higher-order logic” and applied routinely in work on automated theorem proving), and other things provable in that strengthening include (ii) NI, (iii) □ obeys a normal modal logic at least as strong as S4, and (iv) □ that can be given a reductive analysis—a case of a modal notion being definable in terms of purely quantificational and truth-functional logical notions (something previously widely believed to be impossible).22 These are very recent examples—easy to get excited about, but too early to claim lasting success for. For examples of lasting success, consider the list of examples of conceptual reforms from the introduction to this chapter: • • • •

Frege, Russell, and Church on numbers, Church on computability, Russell on definite descriptions, and Tarski on (sentential) truth, definability, and logical consequence.

The list could be extended in many ways. For example, we could add: •

Higher-order logic as a way of regimenting our talk of propositions, properties, classes, relations, functions, and (propositional) truth.

These are conceptual reforms that have been influential not only within philosophy but in other fields as well, including mathematics, linguistics, automated theorem proving, and computer science. For example, natural-language semantics and model theory (a major area of mathematics) are both offshoots of Tarski’s work on truth, definability, and logical consequence. However, when Machery speaks of the investigation and reform of concepts, he seems to have in mind something very different than we do. He advocates “naturalized conceptual analysis” (PPB, 209), which investigates concepts, how they vary between people and how they should be modified. Machery’s “concepts” are to be understood as subsets of individuals’ beliefs and other belief-like states (which he calls “bliefs”) that are retrieved by default from long-term memory (PPB, 210−211). The kind of surveys carried out by experimental philosophers would seem to be tailor-made to the project of discovering what subjects’ concepts are. The thought is that the survey responses tell us about the respondents’ concepts (PPB, 239).

352  Margot Strohminger and Juhani Yli-Vakkuri According to Machery, “one of the points of naturalized conceptual analysis is to assess the validity of concepts” (PPB, 223). Here is how Machery defines invalidity for concepts: A concept is invalid if and if only if the inferences it disposes us to draw are deficient in some way or other. Concepts can be invalid for many different reasons. When the validity of concepts depends on how the world is (e.g., when we are concerned with the reliability of the inferences underwritten by concepts), I’ll say that the point of naturalized conceptual analysis is to assess the empirical validity of concepts. Determining whether the inferences underwritten by a concept are empirically valid, for instance, whether they are reliable, requires two things: understanding the content of a concept (i.e., what bliefs constitute it) and relevant empirical knowledge about the world. PPB (223) Machery illustrates how to assess the validity of concepts with the concept of innateness. But consider how we might assess the validity of a more central philosophical concept like that of knowledge or of necessity. To do this, we would need to have relevant knowledge about knowledge or necessity to serve as a benchmark against which to compare our concepts. But where does that knowledge come from? In the case of necessities, the best approach would seem to be the one taken in the recent literature we cited. In the case of knowledge, the model-­building approach taken in the literature we cited seems appropriate. Would Machery recommend an alternative approach, and what would it be like? We’ll let him speak for himself on this question. But suppose that we have been able to acquire some knowledge about knowledge or necessity, and we discover that some group of people—the non-philosophers, for example—possess invalid concepts of knowledge or necessity. What’s the next step? To change (or “modify”,23 “reform”24) those concepts—in Machery’s sense of “concept”—we need to change the relevant subjects’ belief-like states about the topic that they retrieve by default from long-term memory. How? Machery cites no research on the effectiveness of any interventions that might have this effect. In fact, he gives no evidence that there are any such possible interventions, in any sense of “possible” restricted enough to be relevant to decisions about how to allocate scarce research dollars and euros. Instead of speculating about what Machery thinks about these matters, we’ll end by asking him some questions: First, by what kind of interventions can we change someone’s concept of, say, knowledge? Second, what is the evidence that those interventions would have the desired effect? Third—never mind whether it can be done or how—why should the revision of concepts in Machery’s sense be an aim of philosophy as opposed to

Challenges for an Experimentalist’s Skepticism about Cases  353 a welcome byproduct? No doubt, a good philosophical e­ ducation results in some conceptual revision in Machery’s sense. For example, we conjecture that philosophers are far more likely, as a result of their education, to unthinkingly make inferences like “false; therefore not known” and “there are no Fs; therefore all Fs are G” than non-­philosophers. Knowing more about X can improve the default inferences one makes about X, and we welcome any good byproducts of knowledge, but why should delivering the byproducts become part of the main business? Fourth, in light of the fact that academics in any field are at least as well-placed, as far as we can tell, to evaluate the validity of concepts of whatever they do research on, why is this a task for philosophers specifically? Why shouldn’t mathematicians and theoretical physicists, for example, also dedicate themselves to the study and revision of mathematical and physical concepts, in Machery’s sense of “concept”? To the fourth question, we would like to add a few words of motivation, concluding with two more questions: The kind of conceptual revision exemplified by Frege’s, Russell’s, Church’s, and Tarski’s work isn’t an activity only philosophers engage in. Every science introduces new concepts (by which we mean pairings of words with meanings) and revises them with some regularity. We take it we all agree that scientific progress wouldn’t happen if scientists were only allowed to use words from ordinary language with their ordinary meanings. However, there is no precedent in any science, as far as we know, for the practice Machery recommends of doing surveys in which non-specialists—or specialists for that matter—are asked to weigh in on questions of the relevant science and then basing recommendations for revising concepts—in either sense of “concept”, ours or Machery’s—in part on the data produced by those surveys. None of the major successful conceptual reforms on which the progress of other sciences has been built, and none of the successful conceptual reforms in philosophy cited above, were informed by such studies. We wonder why Machery recommends this radical break from established scientific practice. Does he have evidence to support the view that it would help philosophy but not other fields make progress? Or does he think that it would help other fields make progress as well? (If so, what is the evidence of that?)25

Notes 1 Yli-Vakkuri 2018. 2 Dorr, Yli-Vakkuri, and Hawthorne 2021. 3 Strohminger and Yli-Vakkuri 2018. 4 The axiom of choice, which says that there is a function that maps each   property that has an instance to one of its instances (formally: ∃f ( → t) → .  ∀X ( → t).(∃X → X( fX))), is not at all certain, but there is a powerful abductive argument for it that cites its usefulness in a variety of mathematical applications. Its applications in metaphysics and in the foundations of semantics have only recently begun to be explored: See Goodsell and Yli-Vakkuri in

354  Margot Strohminger and Juhani Yli-Vakkuri preparation a and b for a discussion. When the axiom of choice is added to any of the logics in which Bacon (2018) shows the derivability of S4 for metaphysical (or what he calls “broadest”) necessity, the result is a logic in which S5 is also derivable for metaphysical necessity, as Goodsell and Yli-Vakkuri (in preparation a, Ch. 3) show. 5 For this reason, in the work Chalmers is reviewing, Yli-Vakkuri and Hawthorne (2018) dedicated three of its six chapters (Chs. 1, 5, and 6) to developing models of mental content and narrowness. 6 To clarify, mainstream opinion in the philosophy of mind is that standard Twin Earth cases do show that the kind of belief content we report with sentences of the form “x believes that p” is not narrow. Yli-Vakkuri (2018) is engaging with the minority view that says they don’t. In contrast, Yli-Vakkuri and Hawthorne (2018) are engaging with the more mainstream view (held by Chalmers, among others) that accepts the standard judgment about Twin Earth cases but claims that there is additionally some other kind of mental content that is narrow. 7 Machery idiosyncratically talks about judgments themselves—in addition to methods or processes that yield judgments and other beliefs—as bearers of reliability (see e.g. PPB, 96–97). The assumption seems to be that a judgment-token is reliable just in case it is produced by a reliable method. 8 Hughes (2019) makes this observation in order to develop a similar worry about Machery’s strategy. 9 See e.g. Lackey 2010, 283 and Kelly 2010, 125–126. 10 Against the view that “if cases play a role in philosophy, it is a small one”, he writes: “In a sense, their role is indeed limited. The facts assumed to hold in the situations described by cases must be brought to bear on philosophical conclusions (about free will, meaning, moral permissibility, or causation), and doing this often requires long chains of reasoning. Much of philosophers’ ingenuity goes into developing these long chains of reasoning. Acknowledging this point is, however, consistent with the claim that, if we did not know which facts hold in the situations described by philosophical cases, we would not be able to draw the philosophical conclusions we hope to draw” (PPB, 6). 11 Williamson (2011, 2016). 12 This is our interpretation, which may be incorrect. One thing that made interpreting the text difficult is that Machery speaks (like many other philosophers) as if “logical necessity” was a necessity stronger than metaphysical necessity, presumably because it is (for example) metaphysically but not logically necessary that Hesperus = Phosphorus. But this kind of “necessity” is best thought of as a predicate of sentences; it certainly isn’t a property of propositions since, by Leibniz’s law, since H = P, (H = P) = (H = H). In any case, we are assuming that the metaphysical necessities are not just the relevant (sentences) that are theorems of logic, such as NI, but anything that logically follows from them together with humdrum truths such as “H = P”. 13 Supporting counterfactuals is a necessity: see Williamson (2016). 14 Yli-Vakkuri and McCullagh (2017) contains a representative sample: see especially the critical contributions by Kit Fine and Robert Stalnaker. See also Fritz and Goodman (2016, 2017a, 2017b) and Fritz (2017a, 2017b, 2017c). 15 It also contains some polemics against practices in the vicinity of Machery’s description of the method of cases (see especially the discussion of “ontological particularism” at pp. 277–78). 16 See Salmon (1989), Dorr (2004), Russell and Hawthorne (2018), and Bacon (forthcoming). Salmon and Bacon don’t use the term “metaphysical possibility”, but this does not indicate any disagreement of substance: each is

Challenges for an Experimentalist’s Skepticism about Cases  355 concerned with the strongest necessity (which Salmon calls “ancestral” and Bacon “broadest”). 17 See Strohminger and Yli-Vakkuri 2019 and Roberts 2020. 18 Rejecting the equivalence of restricted necessity claims with metaphysical necessity claims in favor of a syntactic conception of restricted necessity (as consistency with some set of sentences) is not a serious option unless one is willing to give up the principle that necessity implies truth: see Montague (1962). 19 Woodward (2017) doesn’t see himself as doing “metaphysics”, but this is a terminological matter. His imaginary opponent, who rejects model-building and insists on being told about “truthmakers” and “grounds” has one approach to what we call the “metaphysics” of causation, while Woodward has another approach. We can resolve to use the term “metaphysics” in such a way that it applies to both approaches or to only one. No issue of substance turns on this decision. But for what it’s worth, one of us (Yli-Vakkuri) is on record maintaining that “truthmaking” and hyperintensional metaphysical notions in general (“essence”, “grounding”, etc.) are not in good logical standing. The philosophers who hold that reality is hyperintensional (i.e., it is not the case that necessarily coextensional relations are identical) have never told us which system of higher-order logic they accept but they cannot maintain neutrality on systems of higher-order logic: the claim that there are hyperintensional distinctions is inconsistent in all of the known systems of higher-order logic that are strong enough for the formalization of basic mathematics, metaphysics, and semantics. See Goodsell and Yli-Vakkuri (in preparation a and b), and Yli-Vakkuri and Hawthorne (2022b, Sec. 5) for a discussion, and see also Fritz 2021 for a discussion of grounding with which Yli-Vakkuri is sympathetic. For hyperintensional metaphysics to be a serious option, it must be developed within a higher-order logic that is strong enough for the formalization of the more elementary parts of mathematics, metaphysics, and semantics on which its advocates tacitly rely, and so far there has been no attempt to do this. 20 Beginning with Bacon (2018). Williamson (2017) and Strohminger and Yli-Vakkuri (2017) are still talking about objective modalities. 21 For example, as in Bacon (2018). Dorr, Hawthorne, and Yli-Vakkuri (2021, 212–13) consider six different definitions (including Bacon’s) and show that their key results can be obtained using any of them. 22 All of these results are from Bacon (2018). For further results, see Dorr, Hawthorne, and Yli-Vakkuri (2021: Appendices A–D), Dorr and Goodman (2021), Goodsell and Yli-Vakkuri (2020), and the chapter on necessity in Bacon (forthcoming). 23 PPB, 217, 223, 224. 24 PPB, 217, 245. 25 This discussion occasionally draws on and expands upon points made in Strohminger 2018. Margot Strohminger’s work on this essay was funded by the European Union’s Horizon 2020 research and innovation program under the Marie Sklodowska-Curie grant agreement no. 747658.

References Armstrong, David. 1986. “The Nature of Possibility.” Canadian Journal of ­Philosophy 16: 575–594. Armstrong, David. 1989. A Combinatorial Theory of Possibility. Cambridge: Cambridge University Press.

356  Margot Strohminger and Juhani Yli-Vakkuri Armstrong, David. 1997. A World of States of Affairs. Cambridge: Cambridge University Press. Bacon, Andrew and Dorr, Cian. Forthcoming. “Classicism.” In P. Fritz and N. K. Jones (eds.), Higher-Order Metaphysics. Oxford: Oxford University Press. Bacon, Andrew. 2018. “The Broadest Necessity.” Journal of Philosophical Logic 47: 733–783. Bacon, Andrew. 2020. “Logical Combinatorialism.” Philosophical Review 129: 537–589. Bacon, Andrew. Forthcoming a. A Philosophical Introduction to Higher-Order Logics. Oxford: Oxford University Press. Chalmers, David J. 2018. “Review of Yli-Vakkuri and Hawthorne, Narrow Content.” Notre Dame Philosophical Reviews [online]. Accessed December 3, 2020. Available from: https://ndpr.nd.edu/news/narrow-content/. Christensen, David. 2007. “Epistemology of Disagreement: The Good News.” Philosophical Review 116: 187–217. Church, Alonzo. 1940. “A Formulation of the Simple Theory of Types.” Journal of Symbolic Logic 5: 56–68. Demaree-Cotton, Joanna. 2016. “Do Framing Effects Make Moral Intuitions Unreliable?” Philosophical Psychology 29: 1–22. Dorr, Cian, Hawthorne, John, and Yli-Vakkuri, Juhani. 2021. The Bounds of Possibility: Puzzles of Modal Variation. Oxford: Oxford University Press. Dorr, Cian. 2014. “Non-Symmetric Relations.” In Oxford Studies in Metaphysics, vol. 1, edited by Dean W. Zimmerman, 155–192. Oxford: Oxford University Press. Fritz, Peter, and Goodman, Jeremy. 2016. “Higher-Order Contingentism, Part 1: Closure and Generation.” Journal of Philosophical Logic 45: 645–695. Fritz, Peter, and Goodman, Jeremy. 2017a. “Counterfactuals and Propositional Contingentism.” The Review of Symbolic Logic 10: 509–529. Fritz, Peter, and Goodman, Jeremy. 2017b. “Counting Incompossibles.” Mind 126: 1063–1108. Fritz, Peter, and Jones, Nicholas K., eds. Forthcoming. Higher-Order Metaphysics. Oxford: Oxford University Press. Fritz, Peter. 2017a. “Higher-Order Contingentism, Part 2: Patterns of Indistinguishability.” Journal of Philosophical Logic 47: 407–418. Fritz, Peter. 2017b. “Higher-Order Contingentism, Part 3: Expressive Limitations.” Journal of Philosophical Logic 47: 649–671. Fritz, Peter. 2017c. “Logics for Propositional Contingentism.” Review of Symbolic Logic 10: 203–236. Fritz, Peter. 2020. “Metaphysical Necessity.” Unpublished Manuscript. Fritz, Peter. 2021. “Ground and Grain.” Philosophy and Phenomenological Research, published online-first, 11 August 2021: https://doi.org/10.1111/ phpr.12822. Goodsell, Zachary, and Yli-Vakkuri, Juhani. In preparation a. Logical Foundations of Philosophy. Goodsell, Zachary, and Yli-Vakkuri, Juhani. In preparation b. “Defining Meaning”. Hall, Ned. 2004. “Two Concepts of Causation.” In Causation and Counterfactuals, edited by John Collins, Ned Hall, and L. A. Paul. Cambridge, MA: MIT Press.

Challenges for an Experimentalist’s Skepticism about Cases  357 Horne, Zachary, and Livengood, Jonathan. 2017. “Ordering Effects, Updating Effects, and the Specter of Global Skepticism.” Synthese 194: 1189–1218. Hughes, Nick. 2019. “Disagreement, Dogmatism, and the Bounds of Philosophy,” International Journal of Philosophical Studies 27: 591–596. Kelly, Thomas. 2010. “Peer Disagreement and Higher-Order Evidence.” In ­ Disagreement, edited by Richard Feldman and Ted A. Warfield. Oxford: Oxford University Press. Kripke, Saul A. 1971. “Identity and Necessity.” In Identity and Individuation, edited by Milton K. Munitz. New York: New York University Press. Kripke, Saul A.1980. Naming and Necessity. Oxford: Blackwell. Lackey, Jennifer. 2010. “What Should We Do When We Disagree?” In Oxford Studies in Epistemology, vol. 3, edited by Tamar S. Gendler and John ­Hawthorne. Oxford: Oxford University Press. Lewis, David K. 1973. Counterfactuals. Oxford: Blackwell. Lewis, David K.1983. “Mad Pain and Martian Pain.” In Philosophical Papers, vol. 1. Oxford: Oxford University Press. Lewis, David K.1986. On the Plurality of Worlds. Oxford: Blackwell. Machery, Edouard. 2017. Philosophy within Its Proper Bounds. Oxford: Oxford University Press. McPherson, Tristram. 2019. “Supervenience in Ethics.” In The Stanford Encyclopedia of Philosophy (Winter 2019 Edition), edited by Edward N. Zalta. https:// plato.stanford.edu/archives/win2019/entries/supervenience-ethics/. Putnam, Hilary. 1975. “The Meaning of ‘Meaning’.” Minnesota Studies in the Philosophy of Science 7: 131–193. Roberts, Alexander. 2020. “Relative Necessity and Propositional Quantification.” Journal of Philosophical Logic 49: 703–726. Russell, Jeffrey Sanford, and Hawthorne, John. 2018. “Possible Patterns.” In Oxford Studies in Metaphysics, vol. 11, edited by Karen Bennett and Dean W. Zimmerman, 149–93. Salmon, Nathan. 1989. “The Logic of What Might Have Been.” Philosophical Review 98: 3–34. Smith, Michael. 2004. “Does the Evaluative Supervene on the Natural?” In Ethics and the A Priori: Selected Essays on Moral Psychology and Meta-Ethics. Cambridge: Cambridge University Press. Strohminger, Margot, and Yli-Vakkuri, Juhani. 2017. “The Epistemology of Modality.” Analysis 77: 825–838. Strohminger, Margot, and Yli-Vakkuri, Juhani. 2018. “Moderate Modal Skepticism.” In Knowledge, Belief, and God: New Insights in Religious Epistemology, edited by Matthew A. Benton, John Hawthorne, and Dani Rabinowitz, 302–321. Oxford: Oxford University Press. Strohminger, Margot, and Yli-Vakkuri, Juhani. 2019. “Knowledge of Objective Modality.” Philosophical Studies 176: 1155–1175. Strohminger, Margot. 2018. “Review of Philosophy within Its Proper Bounds.” Notre Dame Philosophical Reviews [online]. Viewed 20 November 2019. Available from: https://ndpr.nd.edu/news/philosophy-within-its-proper-bounds/. Williamson, Timothy. 2007. The Philosophy of Philosophy. Oxford: Blackwell. Williamson, Timothy. 2011. “Philosophical Expertise and the Burden of Proof.” Metaphilosophy 42: 215–229. Williamson, Timothy. 2013. “Gettier Cases in Epistemic Logic.” Inquiry 56: 1–14.

358  Margot Strohminger and Juhani Yli-Vakkuri Williamson, Timothy. 2015. “A Note on Gettier Cases in Epistemic Logic.” ­Philosophical Studies 172: 129–140. Williamson, Timothy. 2016a. “Modal Science,” Canadian Journal of Philosophy 46: 453–492. Reprinted in Yli-Vakkuri and McCullagh 2017. Williamson, Timothy. 2016b. “Philosophical Criticisms of Experimental Philosophy.” In A Companion to Experimental Philosophy, edited by Justin Sytsma and Wesley Buckwalter, 22–36. New York: John Wiley & Sons. Woodward, James. 2017. “Interventionism and the Missing Metaphysics: A Dialogue”. In Metaphysics and the Philosophy of Science, edited by Matthew Slater and Zanya Yudell. Oxford: Oxford University Press. Yli-Vakkuri, Juhani, and Hawthorne, John. 2018. Narrow Content. Oxford: Oxford University Press. Yli-Vakkuri, Juhani, and Hawthorne, John. 2022a. “Counterpart Theory and Counterfactuals.” Forthcoming in Oxford Studies in Metaphysics. Oxford: Oxford University Press. Yli-Vakkuri, Juhani, and Hawthorne, John. 2022b. “Intensionalism and Propositional Attitudes.” Forthcoming in Oxford Studies in the Philosophy of Mind. Oxford: Oxford University Press. Yli-Vakkuri, Juhani, and McCullagh, Mark, eds. 2017. Williamson on Modality. London: Routledge.

18 In Defense of Modest Modal Skepticism Edouard Machery

Recent work in metaphilosophy has regularly touched upon modal ­epistemology. Among other points of contact, it is not clear about how we can come to know what holds in the counterfactual situations described by thought experiments (Williamson, 2007, chapter 5): For instance, how do we know whether Twater on Twin-Earth is water or whether Swampman has any belief? More recently, metaphilosophical debates have focused on the role of thought experiments in coming to know metaphysical necessities and strict metaphysical possibilities, i.e., possibilities that are not nomologically possible (Machery, 2017; Hawke and Schoonen, 2021; Nado, 2022; Symons, 2022). Thought experiments are arguably the most commonly used tool for learning about the modal facts that play a central role in century-old philosophical controversies, but many have concluded that this tool is severely deficient (Weinberg, 2007; Machery, 2011, 2017; Stich, 2018). In particular, I have argued that philosophers should suspend judgment when they discover the flaws of the judgments elicited by thought experiments, exactly as scientists set aside their measurements when they find out that these are artifacts of their instruments (Franklin, 2013). If there is no other way of learning about the modal facts of philosophical interest, then we should set aside the philosophical controversies that depend on them and reorient our efforts toward more productive philosophical projects. Unsurprisingly, this line of argument has been met with resistance (Levin, 2019; Strevens, 2019; Akagi, 2019; Hugues, 2019; Springle, 2019; Alexander and Weinberg, 2020; Baz, 2020; Cappelen, 2022; Deutsch, 2020; Nado, 2022; Symons, 2022). In their thought-provoking chapter, Strohminger and Yli-Vakkuri (this volume) add their insightful voices to this distinguished list of critics, and engage with important issues at the intersection of metaphilosophy and modal epistemology, including how to characterize the use of thought experiments in philosophy (what philosophers sometimes call “the method of cases”), the self-defeating nature of arguments against the method of cases, and the prospects of a naturalized conceptual analysis as well as conceptual engineering (what I have called “prescriptive conceptual analysis” in Machery, 2017). As they note in the Introduction, we agree about important metaphilosophical questions despite DOI: 10.4324/9781003002192-20

360  Edouard Machery coming from very different starting points: the abuse of counterexamples in philosophy, the exclusive concern with metaphysical necessities, and the importance of conceptual reform. For all that, they envisage a future for philosophy, for its methods and research questions, that is very different from what I am myself envisaging and hoping for. In the first section, I will discuss the method of cases. In the next section, I will examine whether the arguments provided in Philosophy Within Its Proper Bounds are self-defeating. In the third section, I will examine whether I am making claims that by my lights I could not know. Finally, in the last section, I will assess the prospects of a renewed, naturalized conceptual analysis, and assess the feasibility of conceptual engineering.

The Method of Cases and Modal Epistemology Since the very early days of experimental philosophy (roughly, the use of experimental and other empirical methods to collect data taken to be relevant to philosophical issues), experimental philosophers have claimed that their results challenge some common ways of doing philosophy. The expression “the method of cases” is now commonly used to refer to what experimental philosophers have had in mind. Characterizing the method of cases turns out to be difficult. At the beginning of Chapter 1 (“The Method of Cases”) of Philosophy Within Its Proper Bounds, I first specify what I take cases to be—“descriptions of actual or hypothetical situations”—and I then state: “philosophical cases are cases put forward by philosophers. Philosophical cases are almost always meant to elicit a judgment or some other mental state about the situations they describe” (Machery, 2017, 11). Strohminger and ­Yli-Vakkuri pick a bone with this characterization (this volume, 340): “If we rely on this characterization alone, any judgment we make will count as the product of the method of cases: when we judge that p, we consider an actual situation (state of affairs) and we judge p holds in it.” They then go on proposing their own characterization, appealing to the notion of example. It would be a mistake to “rely on this characterization alone” to see what I mean by “the method of cases.” In Philosophy Within Its Proper Bounds, I illustrated this characterization with the example of Russell’s clock case, quoted verbally and described in some detail, one of the many examples given in the Introduction and Chapter 1. I then enriched the characterization of the method of cases with three important distinctions: the target vs. superficial content of cases, the implicit vs. explicit content of cases, and the material vs. formal use of cases. The superficial content of a case is its narrative setting, which does not matter philosophically; the target content consists in the facts described by the case that matter philosophically. Cases are used to answer formal questions when they bear on the meaning of words or

In Defense of Modest Modal Skepticism  361 the content of concepts; cases are used to answer material questions when they bear on what these words or concepts are about. Most importantly, I distinguish different uses of cases in philosophy, some of which are entirely innocuous, such as illustrating a definition or a theory and producing philosophical puzzlement. I contrast these uses with what I call “the method of cases”: “the formal and material use of cases in philosophical discourse” (16). So, in my mind, the method of cases is characterized by the use of stories (“cases”), by what the stories are meant to do—elicit the application of a concept in order to determine whether a particular fact holds in this situation—and by the way the cases are used—to learn about meaning or about some topic of philosophical interest. What’s more, much of Philosophy Within Its Proper Bounds is concerned with the material use of cases, and their formal use is left to the last chapter of the book. Incidentally, I introduced the method of cases this way to highlight the similarity between what philosophers do in response to a case—judge about an actual or fictional situation—and mundane activities such as judging that a character lies in a novel or that Trump is a bad person based on a newspaper article—in line with the anti-exceptionalism defended in Chapter 1 of Philosophy Within Its Proper Bounds (i.e., cases simply elicit everyday judgments that are justified the way other everyday judgments are). In any case, the important lesson here is that the method of cases isn’t the kind of things that can be defined: The method of cases is a social practice that philosophers learn by reading and imitating other philosophers or by philosophizing in graduate seminars. Like most social practices (e.g., games, democracies, and even philosophy or science), instances of the method of cases share a family resemblance, and there is thus no point engaging in the (so often misguided) philosophical game of definitions and counterexamples. Rather, the point of the characterization briefly summarized above and of the many examples given at the beginning of Philosophy Within Its Proper Bounds is to help the reader, presumably a philosopher, to identify what the book is about—presumably most philosophers are acquainted with the use of cases in philosophy—and to highlight what I take to be important features of this practice. Furthermore, my description of the practice is anchored by prototypical examples of cases and their use in philosophy (see the various examples given in the Introduction and Chapter 1). Strohminger and Yli-Vakkuri’s disagreement with my characterization of the method of cases is not substantial, and it appears ultimately half-hearted: Characterizing the method of cases by means of an example and characterizing it as the description of a situation in order to elicit the application of a concept seem a distinction without a difference. And we seem to agree that some areas of philosophy such as the post-Gettier discussion of knowledge rely on this method.

362  Edouard Machery

Are Arguments against the Method of Cases Self-Defeating? Metaphilosophers have sometimes asserted that arguments against some particular philosophical methods were self-defeating (Bealer, 1992; Hughes, 2019). Strohminger and Yli-Vakkuri (this volume, 343) retool this concern by claiming to “identify three kinds of judgments in PPB that appear to be products of the method of cases.” In their view, I rely on the method of case to conclude that judgments elicited by philosophical cases (“case judgments” for short) are “unreliable” and that lay people and philosophers are epistemic peers with respect to case judgments. In particular, they note that I must decide whether lay people and philosophers are “epistemic peers,” the application of a philosophical concept— viz. the concept of epistemic peer—that involves “complex judgment calls’” that “look very similar to paradigm applications of the method of cases (this volume, 343).” (Similarly, I must apply the concept of unreliability to the judgments elicited by philosophical cases, which is very similar, one may think, to what epistemologists do when they assess thought experiments.) The application of the concept of epistemic peer to lay people and philosophers is in some way similar to the application of the concept of knowledge to Smith in Gettier’s well-known case: After all, both involve the application of a concept, indeed of a philosophical concept, and the relevant concepts are applied to make a philosophical point (about knowledge or about the methods of philosophy). What’s more, in my anti-exceptionalist approach to the method of cases, case judgments are just everyday judgments, and it is thus not surprising that the judgments singled out by Strohminger and Yli-Vakkuri look like the judgments made in response to cases. But they also look like other kinds of judgment, for instance, the judgment that the assessment of the quality of wines or peer reviewers’ judgments about the quality of submitted articles are not reliable, or that the physicists disagreeing about the reality of dark matter are epistemic peers. So, are these judgments instances of the method of cases? As noted earlier, the method of cases is a social practice that is unified by at best a family resemblance, and if they are instances of the method of cases, they are at best marginal cases. What they definitely are not is prototypical cases. The reliability judgment results from a generalization on the basis of the unreliable judgments elicited by those cases that experimental philosophers have experimentally examined, the epistemic peerness judgment from various lines of arguments, including a generalization from empirical studies showing that philosophers and lay people suffer from the same epistemic vices when confronted with some cases. These two forms of arguments have little to do with the prototypical instances of the method of cases in philosophy. I do not describe a particular

In Defense of Modest Modal Skepticism  363 situation and prompt people to apply a concept of philosophical interest (e.g., the concepts of permissibility, causation, or knowledge) in order to learn about the referent of this concept (e.g., permissibility, causation, or knowledge), as is often the case in the method of cases. In any case, Strohminger and Yli-Vakkuri mischaracterize my ­position. As explained in Chapter 3 (“Fooled by Cognitive Artifacts”), my proposal is that we should suspend judgment when cases possess “disturbing characteristics” (which are surprisingly not explicitly mentioned by Strohminger and Yli-Vakkuri): Characteristics that make the judgments they elicit unreliable. I identify three disturbing characteristics: Unusualness: Philosophical cases describe unusual situations. Separating what goes together in everyday life: The properties that are possessed by instances of a concept in everyday concept applications are separated in philosophical cases. The entanglement of target and superficial content: The features of cases that are philosophically relevant and the narrative settings both influence judgment. The basis for the two generalizations singled out by Strohminger and Yli-Vakkuri does not possess any of these characteristics: They are not even the right kind of things to have these characteristics! Strohminger and Yli-Vakkuri also notice that I occasionally “describe hypothetical scenarios and use them as counterexamples to philosophical views.” No doubt I may have slipped here or there—the method of cases has a powerful hold on the way we do philosophy—but even if this happened a few times, none of the crucial steps in the book rely on it. It is also important to keep in mind the different uses of cases in philosophy, which are discussed at the beginning of Philosophy Within Its Proper Bounds: Strohminger and Yli-Vakkuri’s first example of a “hypothetical scenario” merely applies the concept of reliability, as defined earlier in the book (2017, 96), to a fictional situation, a use of cases that I explicitly characterize as innocuous in Chapter 1. That said, Strohminger and Yli-Vakkuri’s reaction to this example shows that it is often not easy to determine what function—e.g., illustrative or argumentative—a case serves. Their second example is a better instance of the method of cases, and perhaps I should not have presented it. However, it does not play any important dialectical role in the book. In particular, it is presented as a familiar objection to my view, an objection someone might raise, not as a tool I use to support my view. Finally, Strohminger and Yli-Vakkuri correctly note that some of the philosophical views I appeal to (e.g., dogmatism in the epistemology of disagreement among epistemic peers) have been “supported” or, alternatively, “developed” by means of the method of cases, but they do not explain why this is an issue, and it isn’t, although the reason varies from case to case. Let’s focus on dogmatism. Cases played a role in

364  Edouard Machery “developing” the view, but surely one could have formulated this simple idea without considering them. Epistemologists also appeal to cases to “support” dogmatism (as they are generally prone to do), but I do not endorse dogmatism in Chapter 4 (“Enshrining Our Prejudices”); rather, I argue that if any of the two dominant approaches in the epistemology of peer disagreement is right, then we should suspend judgment in response to cases. Because I do not endorse dogmatism, I do not need to accept that it is supported by the cases epistemologists consider.

Is Modal Skepticism Coherent? It is an open philosophical question how we come to learn modal facts, assuming that modal language is best understood as genuinely assertive, robustly truth-evaluable (in contrast to truth-evaluable according to a minimalist understanding of truth predication), and not fictional: For instance, how do we know that this pen would fall if I had released it? How can we know the apparent modal implications of physical theories? How can we know whether one is necessarily responsible for an action only if one could have done otherwise? Philosophy Within Its Proper Bounds isn’t concerned with all types of modal knowledge, and I am happy to concede that we know many possibilities and necessities. My interest is in metaphysical necessities and strict metaphysical possibilities. I define metaphysical necessities as unrestricted necessities: While other necessities (e.g., nomological necessities) are relative to some assumptions (e.g., the assumption that the laws of nature hold), something is metaphysically necessary if and only if it is necessary absolutely. A strict metaphysical possibility is a possibility that is not nomologically possible. I interpret many philosophical debates as being concerned with metaphysical necessities and strict metaphysical possibilities (2017, 184–185). For instance, physicalists assert, and non-physicalists deny, that every event is a physical event; action theorists debate whether necessarily one is responsible for an action only if one could have done otherwise. As I put it, these philosophical debates are “modally immodest.” Chapter 6 of Philosophy Within Its Proper Bounds (“Modal Ignorance and the Limits of Philosophy”) defends a moderate form of modal skepticism: I grant that we can learn many modal facts, particularly of the restricted kind; I even concede that some metaphysical necessities may be knowable, but I argue that many metaphysical necessities and strict metaphysical possibilities that are of philosophical interest are beyond our epistemic reach: We cannot know whether they hold. Strohminger and Yli-Vakkuri charge Philosophy Within Its Proper Bounds of inconsistency: Noting, correctly, that many of the claims made in the book have a modal content, they ask (this volume, 349), “How are his own modal claims not modally immodest?” They aren’t, however, for a straightforward reason: None of the modal claims they single

In Defense of Modest Modal Skepticism  365 out are about metaphysical necessities. Consider, for instance, the claim that “modally immodest issues cannot be resolved” (Strohminger and Yli-Vakkuri’s emphasis, this volume, 349). It is curious to interpret this claim as making an unrestricted modal claim. Rather, it holds only for creatures like us, given the type of psychology and limitations we have. In response, Strohminger and Yli-Vakkuri insist that restricted necessities are equivalent to some metaphysical necessities: “in saying that (X-restrictedly) one cannot know such and such one is saying that it is metaphysically necessary that, of the X conditions hold, then one does not know such and such.” Strohminger and Yli-Vakkuri think that the equivalence between claims about restricted necessities and some metaphysical or unrestricted necessities renders the appeal to disturbing cases unavoidable: They take these cases to be needed to come to know restricted necessities themselves. As they ask (this volume, 349), “How, if not by examining ‘unusual and atypical cases’ in which the X conditions hold, can we determine what knowledge is like in all metaphysically possible worlds in which the X conditions hold?” I confess not understanding their argument. The first argument against the method of cases, Unreliability (Chapter 3 of Philosophy Within Its Proper Bounds), goes roughly as follows: When people, including philosophers, are presented with disturbing cases, their judgments are unreliable, and one should suspend judgment when one learns about this unreliability. If there are no other ways than the method of cases to come to know them, many metaphysical necessities of interest to philosophers (what knowledge essentially is or what responsibility entails) cannot be known. Of course, there is a sense in which disturbing cases are “used” to reach this conclusion about the limits of our modal knowledge: Disturbing cases are used to learn about the unreliability of the judgments they elicit. But this is not the sense of “using cases” that is at stake (let’s call this use of cases “the experimental use of cases”): The relevant sense concerns considering cases to learn things about knowledge, responsibility, or causation. I fail to see why the equivalence of the modally restricted proposition that we cannot know many metaphysical necessities of philosophical interest and of the modally unrestricted proposition that in all possible worlds where we have our actual cognitive limitations we cannot know many metaphysical necessities of philosophical interest changes the epistemic situation: The experimental evidence bearing on the unreliability of judgments—established by using cases experimentally—is what allows us to know it. After all, every nomological generalization, including scientific generalizations, is equivalent to some unrestricted necessities, and this mere fact does not change what is needed to know these nomological generalizations. Strohminger and Yli-Vakkuri would have had a better case if they had focused on some of the principles that play a role in my arguments against the method of cases such as Premise 1 of Unreliability:

366  Edouard Machery “Unreliable judgments are severely deficient from an epistemic point of view” ­(Machery, 2017, 102). One might wonder whether this and other principles are metaphysical necessities, and if they are how I can come to know them (Hugues, 2019; Nado, 2022). As I have explained elsewhere, however, this concern can be alleviated in various ways (Machery, 2019, 2022). My preferred answer goes as follows: This principle is a piece of non-ideal theorizing. So understood, I am asserting that for us, unreliability is sufficient for severe epistemic deficiency, while suspending judgment about whether it is for creatures different from us and even about whether this question is meaningful. It might or might not be true in all possible situations that unreliability results in severe epistemic insufficiency, but that is totally irrelevant for the claim on which Unreliability rests on. In effect, I follow those who refuse to look for modally immodest normative principles (Pogge, 2008, 468): Those (…) are not typically lazily neglecting to unify their moral views about the various factual contexts by tracing them back to ultimate principles that cover all possible worlds. Rather, most of them seem to be (…) reluctant to try to cover, with their moral principles, factual contexts substantially different from our own, like the described outlandish context of regressing fetuses.

Conceptual Analysis and Conceptual Engineering The resistance to the message of Philosophy Within Its Proper Bounds comes partly, I believe, from the threat it poses to what many philosophers do for a living: I argue that if my moderate modal skepticism is right, then the philosophical issues that depend on knowing the relevant modal facts should be set aside and philosophers would have to reorient their philosophical activity, a radical conclusion, no doubt. But it isn’t a threat to philosophy at all: As I explain in Chapter 6 and in Chapter 7 (“Conceptual Analysis Rebooted”), there is much left to do for philosophers even if many modal facts of philosophical interest are beyond our epistemic reach. First, many modally immodest philosophical issues have modally modest counterparts and I give two examples in Chapter 6: Polger and Shapiro’s (2016) work on psychological and neuroscientific categories as a modally modest way to examine multiple realizability and Woodward’s (2005) work on causation as an alternative to the attempt to reduce causation to some form of counterfactual dependence. Second, I propose to breathe a new life into an activity that was once central to philosophy: conceptual analysis and its normative cousin, conceptual engineering. Perhaps surprisingly, Strohminger and Yli-Vakkuri are not hostile to this message, although they disagree with some aspects of my views. They understand Woodward’s work on causation differently than I do

In Defense of Modest Modal Skepticism  367 (this volume, 350): “it isn’t in the business of looking for generalizations that are true. It’s in the business of studying causation via simplified models of causation.” Not much hangs on the proper exegesis of Woodward’s large and diverse work, and it is not clear what part of Woodward’s work Strohminger and Yli-Vakkuri have in mind (I suspect Hausman and Woodward, 1999). More interesting is Strohminger and Yli-Vakkuri’s opposition between searching for true generalizations about the properties of causal systems and “studying causation via simplified models of causation.” Much has been written about scientific modeling in the philosophy of science, and there is no room here to engage with this subtle and controversial literature, but on most views Strohminger and Yli-Vakkuri’s opposition is hard to sustain. And for good reasons: For instance, models in mechanics (e.g., the linear oscillator) study the consequences of various forces applied to idealized systems, and aim at representing actual physical systems, at least to some degree of approximation. They also embrace conceptual engineering, quoting their favorite achievements: “The recent work [on necessity] enacts major conceptual reforms by replacing the intuitive notion with a variety of candidate formal notions that all agree on the key results.” However, as Strohminger and Yli-Vakkuri correctly note, there are substantial differences between the kind of conceptual reform I advocated in Philosophy Within Its Proper Bounds (what I called “prescriptive conceptual analysis”) and the paradigms of conceptual reform they embrace: “When Machery speaks of the investigation and reform of concepts, he seems to have in mind something different than we do.” I take concepts to be psychological entities, constituted by those inferences that are context-insensitive, or, more precisely, “by default” (Machery, 2009, 2015). The goal of naturalized conceptual analysis is to describe, in part by means of empirical methods, the inferences that are constitutive of concepts in order, among other things, to assess the validity of these concepts. I use the concept of innateness to illustrate a naturalized conceptual analysis in Philosophy Within Its Proper Bounds. Prescriptive conceptual analysis reforms concepts, so understood. Strohminger and Yli-Vakkuri assert that for “more central philosophical” concepts1 such as the concept of knowledge, assessing the inferences that are constitutive of the concept of knowledge would require knowing what knowledge is, a project best fulfilled by a “model-building approach.” (I will leave aside the fact that it is not clear how their defense of the use of disturbing cases in philosophy fits with their embrace of the model-building approach.) I am not entirely sure how serious the disagreement is between, on the one hand, myself and, on the other, Strohminger and Yli-Vakkuri, in part because they say very little about how they conceive of the m ­ odel-building approach. In particular, they do not distinguish two different issues: first, the clarification of notions of philosophical interest such as necessity by

368  Edouard Machery developing (typically formal) abstract and idealized models (they refer to Yli-Vakkuri’s work on the matter) and, second, the goal of discovering what the referents of these notions essentially are (what causation is, what necessity is, etc.). The latter issue concerns the goals of philosophical theorizing, the former its means. It would seem that Strohminger and Yli-Vakkuri take the development of abstract and idealized models to be a way of answering the very issues that have motivated more traditional approaches in metaphysics: Revealing what causation is, etc. If that is the correct reading (but I am far from sure it is), their approach is similar to the methodological views expressed by some metaphysicians, which are perhaps best captured by Paul (2012): Metaphysicians develop idealized and abstract models “of features of the world” of philosophical interest (causation, necessity, ontological relations such as supervenience or grounding, etc.) and choose among them on the basis of theoretical virtues (simplicity, elegance, etc.). As I read Paul, the goal is to meet the traditional goals of metaphysical theories. This way of developing the model-building approach was criticized in Chapter 6 of Philosophy Within Its Proper Bounds. I have rejected Paul’s and others’ appeal to theoretical virtues to choose the best models on the grounds that there is no context-independent justification for using simplicity and other theoretical virtues in theory choice, and that there is little justification for using them in the context of philosophy. Paul’s is not, however, the only way of developing the model-building approach. Glymour has usefully contrasted the Socratic and the ­Euclidean approach to philosophy, and the latter can be viewed as one development of the model-building approach. In an article that is a wonderful companion to my own attack on the method of cases, Glymour, Danks and colleagues characterize the Euclidean approach as follows (2009, 20): Rather than trying to find necessary and sufficient conditions for actual causation, a “Euclidean” strategy aims to provide reliable indicators for discovering actual causal relations. Those indicators might provide a definition of actual causation, but they need not. The justification of a Euclidean account of actual causation is provided by its fruitfulness in generalization, inference, control, and so forth. Perfect fit with intuition, or applicability in all possible situations, are not desiderata for a Euclidean account, precisely because it does not try to provide necessary and sufficient conditions. There are of course more and less sophisticated versions of the Euclidean strategy, depending on both the scope of application, and the criteria for assessing fruitfulness. It is clear that philosophers who cling to the goal of learning what causation and necessity are will not be sympathetic to the Euclidean approach. (Again, I am not sure where Strohminger and Yli-Vakkuri stand on the matter.)

In Defense of Modest Modal Skepticism  369 In any case, Strohminger and Yli-Vakkuri’s conception of what t­heorizing about knowledge, causation, or whatever happens to appear currently “central” to some philosophers seems too narrow. One does not need a theory of what causation is, based on the traditional method of cases or based on a model-building approach, to distinguish useful from useless inferences with the concept of causation, as is made clear by the quotation from Danks and colleagues. Indeed, much of the recent work on causal inference in philosophy and statistics proceeds without any metaphysical theory about causation at all. Philosophers and statisticians identify properties such as the Causal Markov Condition and Faithfulness and examine what follows for causal inference when causal systems have these properties without taking on any commitment about whether these properties are essential, necessary, or possessed by all actual instances of causation (in fact, they typically deny they are any of those). I now turn to my defense of prescriptive conceptual analysis.2 First, I am a pluralist about conceptual reform, and I am open to the idea that it can take different forms depending on the nature of the explicandum and the goals of explication (to use Carnap’s terminology). Conceptually reforming a mathematical concept such as the concept of limit (to give an important historical example) is plausibly different from reforming the concept of gene in biology, and both plausibly differ from reforming a lay concept like that of marriage or for reforming a concept like that of ideology for political purposes (Machery, 2022). Strohminger and Yli-Vakkuri question the practical possibility of conceptual engineering. They ask (this volume, 352), “by what kind of interventions can we change someone’s concept of, say, knowledge? (…) What is the evidence that those interventions would have the desired effect?” These are excellent and fair questions. The simple answer is that we do not know since the idea that concepts, as I understand them, can be intentionally changed, has never been the focus of a systematic research program in the social sciences. Concerns about the feasibility of conceptual engineering are thus well-taken (e.g., Machery, forthcoming), and too little is known about the issue. Ignorance however is no ground for skepticism: Actual examples of conceptual engineering in science as well as in the social realm show that a broad skepticism is out of place. The concept of bias, at the intersection of science (measuring bias) and policy (addressing bias), illustrates the possibility of conceptual engineering: With the emergence of the literature on implicit bias in social psychology in the 1980s and 1990s and its influence on social policy in the 2000s, what it means to be biased has been thoroughly redefined. Other social examples include marriage, husband and wife, sex and gender, as well as man and woman (but see Byrne, 2020; Dembroff, 2021); scientific examples include limit, information, computing, gene, and many psychological concepts, now applied to subpersonal systems and even plants (for a discussion, see, e.g., Figdor, 2018; Machery, 2020). More generally, there

370  Edouard Machery is a large literature about scientific concepts as tools in the history of science (Feest, 2010; Rheinberger, 2010), and feasibility concerns remain empty until confronted with this and other relevant literatures. Strohminger and Yli-Vakkuri also wonder whether reforming concepts should be the task of philosophers instead of other academics: Aren’t biologists, for instance, the best situated to reform biological concepts such as the concept of gene? This too is a fair question. Whether one answers it positively or negatively depends, at least in part, on one’s broader attitude toward the objects of philosophy. Some philosophers—I suspect Strohminger and Yli-Vakkuri are among those—hold that philosophy is most successful when it focuses on proprietary topics; ­others— including me—believe that philosophy is most successful when it looks outward, and finds its food for thought in the sciences, the social issues of the time, or the arts. Those who share my leanings are willing to engage with, e.g., the relevant sciences to a degree such that they are as well situated as any biologist to put some order in biologists’ conceptual scheme. Now, I have no a priori argument that the latter approach to the task of the philosopher is superior to the former, but I believe that the history of philosophy shows that fruitful and exciting philosophy tends to emerge when philosophers look outward rather than inward. Finally, Strohminger and Yli-Vakkuri note that in my view prescriptive conceptual analysis involves describing the concepts to be reformed (“the explicanda”) and that I take this task to be empirical, and to often require experimental tools. They take this to be a “radical break from established practice (this volume, 353).” Importantly, I do not restrict the kind of empirical investigation to experiments (2017, 240): “Historical, anthropological, or sociological linguistics can provide important evidence relevant to describing concepts of philosophical interest. Historians of science have often relied on this type of evidence to study past scientific concepts.” And there is no doubt that scientists take into account the history of their concepts and possible variation in the relevant scientific communities when they propose conceptual reform. Consensus conferences aim, among other things, at assessing such variation. And surveys themselves have actually been used by scientists attempting to reform scientific concepts: For instance, the reform of the concept of impairment in clinical neuropsychology was informed by a survey of neuropsychologists (Guilmette et al., 2020).

Conclusion Like most social objects, a philosophical practice such as the method of cases is multifarious, and can take various, more or less prototypical forms across contexts, defying any attempt to define it. For all that, cases have been deployed throughout philosophy to argue for and against modally immodest philosophical conclusions. In fact, they play an important,

In Defense of Modest Modal Skepticism  371 perhaps irreplaceable role to establish or undermine many metaphysical necessities and strict metaphysical possibilities of philosophical interest. How else could we know whether there could be a psychological event that is not a physical event? How else could we know whether one is necessarily responsible for one’s action only if one could have done otherwise? Doubts about the method of cases, particularly doubts about the reliability of the judgments they elicit, justify a modal skepticism, which, while restricted, should lead philosophers to redirect many of their intellectual efforts.

Notes 1 Strohminger and Yli-Vakkuri’s position is surprisingly conservative: Why are these concepts “more central”? 2 Here, I gloss over the differences between conceptual engineering, explication, and prescriptive conceptual analysis.

References Akagi, M. (2019). Finding the bounds of Machery’s critique. International ­Journal of Philosophical Studies, 26, 584–591. Alexander, J., and Weinberg, J. (2020). You just can’t count on (un)reliability Analysis, 80, 737–751. Baz, A. (2020). On discontinuity and its discontents: Comments on Edouard Machery’s Philosophy within Its Proper Bounds. Analysis, 80, 751–758. Bealer, G. (1992). The incoherence of empiricism. Proceedings of the Aristotelian Society Supplemental, 66, 99–138. Byrne, A. (2020). Are women adult human females? Philosophical Studies, 177, 3783–3803. Cappelen, H. (2022). Experimental philosophy without intuitions: An ­i llustration of why it fails. Philosophical Studies, 179, 309‒317. Dembroff, R. (2021). Escaping the natural attitude about gender. Philosophical Studies, 178(3), 983–1003. Deutsch, M. (2020). The method of cases unbounded. Analysis, 80, 758–771. Feest, U. (2010). Concepts as tools in the experimental generation of knowledge in cognitive neuropsychology. Spontaneous Generations: A Journal for the ­History and Philosophy of Science, 4, 173–190. ­ redicates. Figdor, C. (2018). Pieces of Mind: The Proper Domain of Psychological P New York, NY: Oxford University Press. Franklin, A. (2013). Shifting Standards: Experiments in Particle Physics in the Twentieth Century. Pittsburgh: University of Pittsburgh Press. Glymour, C., Danks, D., Glymour, B., Eberhardt, F., Ramsey, J., Scheines, R., … and Zhang, J. (2010). Actual causation: A stone soup essay. Synthese, 175(2), 169–192. Guilmette, T. J., et al. (2020). American Academy of Clinical Neuropsychology consensus conference statement on uniform labeling of performance test scores. The Clinical Neuropsychologist, 34(3), 437–453. Hausman, D. M., and Woodward, J. (1999). Independence, invariance and the causal Markov condition. British Journal for the Philosophy of Science, 50(4), 521–583.

372  Edouard Machery Hawke, P., and Schoonen, T. (2021). Are Gettier cases disturbing? Philosophical Studies, 178, 1503–1527. Hugues, N. (2019). Disagreement, dogmatism, and the bounds of philosophy. International Journal of Philosophical Studies, 26, 591–596. Levin, J. (2019). A case for the method of cases. Philosophy and Phenomenological Research, 98, 230–238. Machery, E. (2009). Doing Without Concepts. New York: Oxford University Press. Machery, E. (2011). Thought experiments and philosophical knowledge. Metaphilosophy, 42(3), 191–214. Machery, E. (2015). By default. In E. Margolis and S. Laurence (Eds.), The ­Conceptual Mind: New Directions in the Study of Concepts (pp. 567–588). Cambridge, MA: MIT Press. Machery, E. (2017). Philosophy Within Its Proper Bounds. Oxford: Oxford University Press. Machery, E. (2019). Response to Akagi, Hughes, and Springle. International Journal of Philosophical Studies, 26, 608–623. Machery, E. (2020). What do plants and bacteria want? Commentary on Carrie Figdor’s Pieces of mind. Mind & Language, 35(5), 677–686. Machery, E. (2022). Responses to Herman Cappelen and Jeniffer Nado. Philosophical Studies, 179, 329‒342. Machery, E. (forthcoming). A new challenge to explication. Inquiry. Nado, J. (2022). Philosophizing out of bounds. Philosophical Studies 179, 319‒327. Pogge, T. (2008). Cohen to the rescue! Ratio, 21, 454–475. Paul, L. A. (2012). Metaphysics as modeling: The handmaiden’s tale. Philosophical Studies, 160(1), 1–29. Polger, T. W., and Shapiro, L. A. (2016). The Multiple Realization Book. New York: Oxford University Press. Rheinberger, H. J. (2010). An Epistemology of the Concrete: Twentieth-Century Histories of Life. Duke University Press. Springle, A. (2019). The method of cases in context. International Journal of ­Philosophical Studies, 26, 597–608. Stich, S. P. (2018). Knowledge, intuition and culture. In J. Proust and M. Fortier (Eds.), Metacognitive Diversity: An Interdisciplinary Approach (pp. 381–394). Oxford: Oxford University Press. Strevens, M. (2019). Philosophy unbound. Philosophy and Phenomenological Research, 98, 239–245. Strohminger, M., and Yli-Vakkuri, J. (2019). Knowledge of objective modality. Philosophical Studies, 176, 1155–1175. Symons, J. (2022). Experimental philosophy within its proper bounds. Inquiry, 65(5), 586‒606. Weinberg, J. M. (2007). How to challenge intuitions empirically without risking skepticism. Midwest Studies in Philosophy, 31, 318–343. Williamson, T. (2007). The Philosophy of Philosophy. Oxford: Blackwell. Woodward, J. (2005). Making Things Happen: A Theory of Causal Explanation. New York: Oxford University Press.

Notes on Contributors

Otávio Bueno Professor of Philosophy, University of Miami Arindam Chakrabarti Professor of Philosophy, University of Hawaii, Manoa Rebecca Chan Associate Professor of Philosophy, San Jose State University Rebecca Hanrahan Associate Professor of Philosophy, Whitman College Joachim Horvath Assistant Professor of Philosophy, Ruhr University Bochum Boris Kment Professor of Philosophy, Princeton University Theodore Locke Visiting Assistant Professor Trinity University, San Antonio Edouard Machery Distinguished Professor of History and Philosophy of Science, University of Pittsburgh and Senior Research Fellow, University of Johannesburg Antonella Mallozzi Assistant Professor of Philosophy, Providence College Sanna Mattila PhD Candidate, University of Helsinki Ian Nicolay Graduate Student, University of Hawaii, Manoa Christian Nimtz Professor of Theoretical Philosophy, Bielefeld University Duško Prelević Professor of Philosophy, University of Belgrade

374  Notes on Contributors

Sonia Roca-Royes Senior Lecturer of Philosophy, University of Stirling Nathan Salmon Distinguished Professor of Philosophy, University of California, Santa Barbara Scott Shalkowski Professor of Philosophy, University of Leeds Margot Strohminger Lecturer of Philosophy, Australian Catholic University Amie Thomasson Professor of Philosophy, Dartmouth University Anand Vaidya Professor of Philosophy, San Jose State University Barbara Vetter Professor of Theoretical Philosophy, Freie Universität Berlin Michael Wallner Lecture of Philosophy, University of Graz Timothy Williamson Wykeham Professor of Logic, New College, Oxford University Juhani Yli-Vakkuri Senior Research Fellow of Philosophy, Australian Catholic University

Index

abduction/abductive accounts of modal knowledge 9, 35–36, 45, 58, 88, 112, 124–125, 341 ability/abilities 44, 47–49, 51–55, 63, 65–66 accessibility relation (equivalence/ universal accessibility) 204 affordance 44, 47–49, 52–54, 63, 65 Almog, Joseph 213 analyticity/analytic truths 1–3, 17, 61–62, 90, 140–141, 144, 148, 153, 261, 266, 280, 282; absolute/ relative analyticity 213; epistemic analyticity/the problem of “Competent Dissent” 278–281, 296; epistemic/metaphysical/essentialist analyticity 282 Anscombe, Gertrude Elizabeth Margaret 65 a priori/a posteriori 1–4, 6–8, 36, 46, 66, 96, 137, 147–148, 152–157, 165, 171, 199–204, 209, 212–214, 266, 268; identity statements 2, 15, 60–61, 93, 108–116, 119–123, 124, 128–143, 145–148, 199–201, 208, 220–222, 228, 266–269, 359; contingent a priori 3, 116–117, 276–277; necessary a posteriori 3–4, 77, 85, 124, 135–137, 155–156, 199–201, 266, 269–277, 285–287, 289–290, 294–295, 300; synthetic a priori 2 Aristotle 7, 202–203, 206–207, 211, 214, 237, 249–250, 254, 257, 262 armchair methodology 1, 5–6, 49, 63, 318; exceptionalism/antiexceptionalism 45–46, 59, 112–113, 361–362 Armstrong, David 348 Arnauld, Antoine 213

Ayer, Alfred 208 Bacon, Andrew 354–355 Balcerak Jackson, Magdalena 323–324 Barnett, David 323, 333 Bealer, George 5–6, 8, 88–91, 93, 95, 104 Beebee, Helen 81 Benovsky, Jiri 213 Berghofer, Phillip 172 Berto, Francesco 168 Bhaţţa, Jayanta 183–184, 187 Biggs, Stephen 112 Black, Max 209 Boghossian, Paul 266, 278, 280–282, 296 Breckenridge, Wylie 305 Bueno, Otávio 10 Burgess, Alexis 113–114, 125 Cārvāka 188 Casullo, Albert 171, 201, 212 causal explanation 30, 33, 35, 289 Candrakīrti 178, 197 Cappelen, Herman 322 Carnap, Rudolf 203, 369 Chalmers, David 4, 6, 16, 77, 88, 199–203, 207, 210, 212–213, 218–223, 228–229, 277, 341, 354 Chan, Rebecca 17, 247–248, 250–262 Christensen, David 344 Chudnoff, Elijah 91–93, 173 Church, Alonzo 148, 339, 351, 353 Clark-Doane, Justin 66, 124–125 Cohen, Marc 207 conceivability 4, 6, 10, 16, 45, 65–66, 71–73, 75–76, 79–80, 86, 88, 95, 129, 135, 168, 202, 204, 209, 219–223,

376 Index 227–229; absolute/relative ideal positive primary conceivability 205–208, 224–226, 228–229; ideal positive primary conceivability 4, 199–200, 204, 209–212, 228 conceptual engineering 105, 297, 299, 359–360, 366–371 conceptual/semantic mastery 17, 37–38, 265–270, 273–274, 276–279, 281, 284–295, 298–300 Correia, Fabrice 245 counterfactual conditional/ counterfactual reasoning 5–6, 11, 14, 16, 19–22, 25, 29–30, 38, 46, 53, 100– 101, 105, 112, 176–178, 208, 218, 220–223, 227–229, 272, 292–293, 303, 306–308, 317–320, 324, 327, 329–332, 335–336, 346–347, 350, 354, 359; counterfactual robustness 23–24; independence counterfactuals 265–266, 271–275, 285, 291–293, 298; sensitivity/insensitivity counterfactuals 14, 70–73, 75–84, 86; tarka 178, 182–189, 193–194, 197 Cowling, Sam 233 Danks, David 368–369 Denby, David 233, 241–242 Descartes, René 213 Deutsch, Max 323 Dharmakīrti 184, 196 disposition 44, 49–55, 63, 65, 100–101 Donnellan, Keith 136–137 Dorr, Cian 348 Dowell, Janice 203 Edgington, Dorothy 59 the epistemology of disagreement 196, 340, 345–346, 363 essence 7–8, 11, 14, 16–17, 19, 24. 30, 36–37, 61–62, 66, 72, 102–103, 145, 151–153, 155, 157, 162, 165–166, 171–173, 200–201, 208, 210–211, 214, 231–261, 268–270, 275–276, 282, 288–290, 294–295, 299, 355, 369; the essentiality of kind 7–8, 16, 32, 80–81, 100–101, 119, 121, 123, 129–130, 132–134, 137–140, 143– 144, 148, 156–157, 161–164, 166, 171–172, 200–212, 214, 220–221, 226–227, 232, 237, 243, 245–246, 249–250, 252–253, 255–257, 262,

269–270, 276, 286–291, 299; the essentiality of origin 3, 16, 200–203, 205, 207–214, 226–227, 299; individual essence 232, 237, 245–246, 249–250, 253–254 Evans, Gareth 276 experimental philosophy 1, 8, 94, 338–339, 342, 351, 360–363, 365, 370 Fine, Kit 7, 17, 24, 66, 151, 173, 200, 231–238, 241, 244–252, 256–257, 261–262, 299, 305 Fischer, Robert William 9, 64, 89, 96–97 Forbes, Graeme 205, 213 Franklin, James 9 Frege, Gottlob Friedrich Ludwig 132, 190, 339, 351, 353 Frege’s Basic Law V 90 Freschi, Elisa 196 Ganeri, Jonardon 182 Gettier Edmund 17–18, 64, 302–324, 327–328, 331–336, 341–342, 346–347, 361–362 Gettier’s thought experiments 17–18, 64, 302–304, 307, 312–314, 327–328, 335–336, 341–342, 347, 362; deviant realizations 17–18, 310–312, 314– 318, 321–324, 330–331, 336; Russell’s clock case 360; the counterfactual reconstruction 17–18, 316–321, 324, 331, 335–336; the substitution gambit 18, 305, 315, 321–322, 332–333, 335; the suppositional reconstruction 17–18, 303–316, 320–321, 323–324, 332–335 Glymour, Clark 368 Gorman, Michael 242 Gregory, Dominic 168 Grundmann, Thomas 310, 318, 323 grounding 7, 30–33, 35–36, 39, 66, 71, 75, 79, 102, 105, 150, 153, 155–156, 158–159, 163, 165–166, 184, 200, 202, 221–222, 250, 254, 257, 262, 268, 270, 276, 288–290, 294, 298, 300, 355, 368 haecceitism 144–145, 245, 253, 262 Häggqvist, Sören 122 Hale, Bob 6–8, 15, 150, 152–158, 160–161, 169–172, 203, 211

Index  377 Hare, Caspar 181 Haslanger, Sally 299 Hausman, Daniel 367 Hawke, Peter 109–110, 125 Hawthorne, John 348, 354 Hemacandra 180 Horvath, Joachim 18, 310, 318, 323, 327–336 Hughes, Nick 354 Hume, David 65, 73, 109, 188 Husserl, Edmund 15, 151, 160–170, 172–173 Ichikawa, Jonathan Jenkins 46 imagination 6, 11, 22–23, 52, 56–58, 66, 81, 99, 129, 138, 140, 150–151, 158, 173, 176, 182, 193–194, 211–212, 221, 229, 236–237, 239, 287, 289, 293, 306, 312–313, 316, 331, 335; imagination and eidetic variation 160–170, 172 induction 9, 44, 46, 55–56, 58, 63, 98, 100, 109, 188–189, 287–288, 299, 310–311, 328 intuition 6–10, 63, 71, 74, 82–83, 88–95, 101, 104, 112, 114, 119–120, 129, 141 Jaimini 189 Jainism 177, 180–183, 186–187, 193, 196 Jain, Pragati 180 Jenkins, Carrie 70–82, 84–86 Kagan, Shelly 239 Kant, Immanuel 1–3, 9, 282 Kaplan, David 293 Kasmier, David 164, 166–167, 169 Kenny, Anthony 53–54 Kind, Amy 57 Kirwan, Christopher 213 Kment, Boris 10, 246 Knobe, Joshua 8 Koslicki, Kathrin 245 Kratzer, Angelika 39 Kripke Saul 2–4, 24, 99, 109–111, 113–114, 116–117, 119, 122–124, 128–129, 131, 135, 138, 141, 144– 146, 155–156, 199–201, 206–207, 212, 214, 254, 266, 277, 294–295, 304, 340, 348, 350 Kumārila, Bhaţţa 179, 190–192 Kung, Peter 57, 168

Lange, Marc 23 LaPorte, Joseph 202 Legg, Catherine 9 Leibniz, Gottfried Wilhelm 177, 181, 209, 214, 259, 354 Lewis, Clarence Irving 180 Lewis, David 10, 18, 20, 38, 72, 104, 181, 242, 246, 256, 259–260, 262, 329–330, 332, 348 liṅ 177, 190, 192 Linsky, Bernard 132 Livingstone-Banks, Jonathan 261 Locke, John 65, 249 Locke, Theodore 299 Lowe, Jonathan 6–7, 15, 103, 150–152, 158, 160–161, 169–172 Machery, Edouard 8, 18, 338–354, 367 Mackie, Penelope 15, 130, 138–140, 144, 147, 212, 214 Magidor, Ofra 305 Mallozzi, Antonella 9–11, 17, 284–287, 289–292, 294–296, 298 Mallișeṇa 180, 196 Malmgren, Anna-Sara 319 Maṇḍana, Miśra 179, 192 McCullagh, Mark 354 McDowell, John 94 McGee, Vann 278 McGinn, Colin 210 metaphysical intension/metaphysical extension 131–134, 140, 143, 146, 148 the method of cases 18, 338–350, 354, 359–363, 365, 368–371 Mill, John Stuart 132–133, 143–144 Mīmāmsā 178–179, 189–191, 193–194 Mīmāmsāka, Bhaţţa 192+ Miščević, Nenad 70–71 modal conventionalism 110, 123, 201, 203, 271, 282, 285, 291, 298–300 modal empiricism 8–10, 45, 47–48, 56, 63–66, 84, 88–89, 108, 111–112, 188, 200–202, 214 modal epistemology/the epistemology of modality 1–3, 6–11, 14–16, 44–47, 52–53, 58–59, 62, 65–66, 72–73, 75, 79, 83, 86, 88–89, 93–105, 109–110, 123–124, 150, 155–156, 158, 161, 168–169, 171, 176, 179, 194, 200, 202, 213, 223, 226, 229, 250, 261, 275–276, 284–286, 294, 299, 359–361;

378 Index the epistemology of essence 15, 17, 66, 112–113, 150–173, 201–203, 208, 250, 276, 284, 288–291, 295, 299; the problem of modal epistemic friction 11, 101–103, 168, 220–221, 227–229 modal essentialism 6–9, 15, 17, 102– 103, 129–130, 134–146, 148–149, 151–155, 157–158, 161–168, 170, 172, 199–214, 220, 223, 225–228, 231–238, 245, 247–261, 265 modalism 10, 16–17, 89, 96, 98, 100– 105, 173–174, 247–248, 250–252, 254–256, 260–262 modality de dicto/modality de re 44, 46, 66, 93, 200–201, 203–204, 212–213, 219–220, 223–224, 233, 245, 253, 266–267; absolute/relative (restricted) modality 10, 21, 28, 46, 52–54, 58–62, 66, 99, 112, 156–157, 273, 316, 329–330, 336, 349, 352, 355, 364–365, 371 modal monism/modal dualism 45–46, 213, 229, 266, 275–278, 285, 291, 294–295; the collapse of the modal 14, 70–71, 73, 75–84 modal normativism 17, 123, 180, 265–268, 270–280, 282, 284–286, 288–300 modal rationalism 4, 8, 15–16, 46, 56, 66, 88–96, 104–105, 111–112, 199–203, 205–214, 219, 227 modal reliabilism 5, 88–95, 109–113, 121–125, 186, 220, 222– 223, 227, 229, 299, 344, 349–350, 362–363 modal skepticism/modal pessimism 10, 63, 100, 103–104, 108–110, 115, 122–123, 178, 188, 245, 250, 338, 340, 346–347, 349, 359–360, 364, 366, 369, 371 Mohanty, Jitendra Nath 164 Moore, George Edward 281 Morganti, Matteo 213

205–206, 213–214, 224–225, 228; hylomorphism/the Aristotelian view of objects 16, 206–207, 209, 211, 214, 224–228, 248–250; the substratum theory 205–207, 209, 213, 224–225, 249 Oddie, Graham 242 Olson, Eric 214

Navya-Nyāya 187 Nāgārjuna 178, 183, 193 Nichols, Shaun 8, 10 Nimtz, Christian 84–85 Nolan, Daniel 25, 49 Nyāya 178, 183–186

Pānini 177 paradigm terms 15, 110–111, 116–125 Paul, Laurie Ann 368 Peacocke, Christopher 4, 8 perception 9–10, 46–48, 55, 57, 73, 91–94, 158, 183, 185, 188–190, 209 perspectival pluralism 181 philosophical zombies 63, 65, 70, 74, 108–109, 111, 122, 213, 349–350; conceivability arguments/ conceivability methods 179, 199–200, 204, 212, 226 Plantinga, Alvin 233, 245 pluralism in geometry/pluralism in logic 16, 204, 207, 213–214, 227 Polger, Thomas 366 possible worlds 1, 7, 10, 14, 19–20, 25–26, 29, 39, 45–46, 59, 70–73, 77, 82–83, 86, 95, 102, 113, 116, 121, 123, 128–133, 135, 137, 142, 144–145, 176, 180, 186, 195, 199–200, 204, 206, 234, 239–240, 247, 251, 253–256, 258, 261–262, 265–266, 269–277, 284, 292–294, 304, 318, 329–331, 349–350, 365– 366; impossible worlds 25–27, 38, 300; modal space/space of possible worlds 45–46, 55, 77, 82–83, 95–96, 112, 123, 125, 204, 225, 229, 300 potentiality 44, 52–55, 59, 62 Prabhācandra 183, 187–189 Prabhākara 179, 190–193 Prabhākara, Miśra 190 Prāsaṅgika Mādhyamika 178 Prelević, Duško 16, 218–220, 223–229 Prior, Arthur 181 Pryor, James 160 Putnam, Hilary 15, 110, 117, 123, 128–131, 134–139, 142, 144–147, 149

objecthood 16, 36, 202, 213–214, 225–226, 250, 255; the bundle theory

Quine, Willard van Orman 146, 231, 233, 245, 282

Index  379 rigid designators/rigid designation 2, 15, 77, 82, 85, 109, 113–116, 119, 121–124, 128, 130–135, 140–146, 148, 201, 266–267; rigid/non-rigid general terms 15, 130–135 Robertson Ishii, Teresa 148 Roca-Royes, Sonia 8–9, 16, 44, 46, 88, 96, 98, 109, 112, 199–203, 206–212, 214, 225, 227 Russell, Bertrand 132, 339, 351, 353; Russell’s paradox 90, 203–204; Russellian propositional function 148; Russellian propositions 26 Ryle, Gilbert 204 Saint-Germier, Pierre 323 Salmon, Nathan 130–131, 133, 136, 138, 146, 354–355 Sāmkhya 197 Schaffer, Jonathan 241 Schoonen, Tom 168 Schrenk, Markus 51–52 Sextus, Empiricus 196 Shalkowski, Scott 10 Shapiro, Lawrence 366 Sherman, Brett 125 Shoemaker, Sydney 81 Siegel, Susanna 51 Skiles, Alexander 242 Soames, Scott 15, 117, 138–140, 146–148 sparse/abundant properties 16–17, 233, 238, 241–244, 246, 248, 256–260, 262 Spinelli, Nicola 164, 173 Sprigge, Timothy 207 Steinberg, Alex 124–125 Stich, Stephen 8, 10 Strohminger, Margot 9, 18, 114, 355, 359–371 strong metaphysical possibility/strong metaphysical necessity 15, 71, 77, 81–86 Stroud, Barry 103–104 syāt 16, 177–182, 193–196 syādvadā 180, 182, 193–194 Tarski, Alfred 339, 351, 353 the Banach-Tarski theorem 90

Thomasson, Amie 10, 16, 63, 180, 196, 265–268, 272, 274, 278–279, 281–282, 291, 294, 296, 299 thought experiments 17–18, 38, 99–100, 179–180, 209, 304–305, 307–308, 310–312, 315, 317–318, 322–324, 327–334, 336, 359, 362 two-dimensional semantics 16, 45, 199–200, 208, 212, 228–229, 277 Udayana/Udayanācārya 183, 185, 187 Vaidya, Anand 11, 55, 59, 89, 94, 96, 101–102, 109, 112, 152, 168, 172–173, 196, 213, 220, 281 Vallentyne, Peter 239 van Fraassen, Bas 202 van Inwagen, Peter 10, 63, 108–109 Vātsyayana 184 Veda 179, 190–192 Vedānta 186–187 Vetter, Barbara 10 Viebahn, Emanuel 46 von Wright, Georg Henrik 196 Vyākaraṇa 190 Vyāsatīrtha 183, 187, 193 Wallner, Michael 11, 55, 59, 101–102, 109, 168, 173, 220 Weinberg, Jonathan 8, 10 Wiegmann, Alex 322 Wikforss, Åsa 122 Wildman, Nathan 241–242 Williamson, Timothy 5–6, 10–11, 16–18, 23, 45–46, 56, 112, 173, 218–223, 228–229, 266, 278–281, 295–297, 303, 305, 310, 316–324, 341, 347–348, 355 Wirling, Ylwa Sjölin 9 Wittgenstein, Ludwig 177, 182, 191, 281 Woodward, James 350, 355, 366–367 Worley, Sara 229 Yablo, Stephen 4, 10, 16, 196, 218–219, 228–229 Yli-Vakkuri, Juhani 18, 108, 112, 114, 348, 354–355, 359–371