Kant on Freedom and Rational Agency 9780198873143, 019887314X

Citation preview

Kant on Freedom and Rational Agency

Kant on Freedom and Rational Agency M A R K U S KO H L

Great Clarendon Street, Oxford, OX2 6DP, United Kingdom Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide. Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries © Markus Kohl 2023 The moral rights of the author have been asserted All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by licence or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this work in any other form and you must impose this same condition on any acquirer Published in the United States of America by Oxford University Press 198 Madison Avenue, New York, NY 10016, United States of America British Library Cataloguing in Publication Data Data available Library of Congress Control Number: 2023935463 ISBN 978–0–19–887314–3 DOI: 10.1093/oso/9780198873143.001.0001 Printed and bound in the UK by TJ Books Limited Links to third party websites are provided by Oxford in good faith and for information only. Oxford disclaims any responsibility for the materials contained in any third party website referenced in this work.

Preface and Acknowledgments This book aims to develop a novel interpretation and rational reconstruction of Kant’s doctrine of freedom. Although later chapters build on previous ones, most of them can be read as self-­standing essays. In particular, Chapter X is an im­port­ ant addition to my overall argument which is of special interest to readers who wonder how Kant’s views on freedom of imagination in the third Critique relate to his more canonical ideas about freedom of will and freedom of thought. My interpretative methodology strongly prioritizes Kant’s mature published writings from the critical period of the 1780s and 1790s. I do not rely on pre-­ critical works, student lecture transcripts, or Reflexionen for my main claims or arguments. But I occasionally use such texts to further support my reading of passages from critical published works and to shed light on unclarities or seeming inconsistencies in these works. I have benefited from conversations with many philosophers who helped me gain a better understanding of Kant and the relevant philosophical issues. For feedback, support, and advice, I am especially indebted to Karl Ameriks, Richard Aquila, Anne Margaret Baxley, Luc Bovens, EJ Coffman, Alix Cohen, Adam Cureton, Bernd Dörflinger, Stephen Engstrom, Tom Hill, Dietmar Heidemann, Thomas Hofweber, Matt Kotzen, Marc Lange, Mariska Leunissen, Alan Nelson, Ram Neta, David Palmer, Ulrich Schlösser, Geoff Sayre-­McCord, Sarah Stroud, R.  Jay Wallace, Owen Ware, and Eric Watkins. I also thank OUP editor Peter Momtchiloff for his helpful, sympathetic guidance and two OUP referees for their illuminating, constructive comments on my manuscript. Above all, I am deeply grateful to Stephanie Basakis for her assistance with editing, but especially for her unwavering intellectual and emotional support. Some sections of Chapter I contain thoroughly revised and rewritten material from my 2016 article “Kant on Idealism, Freedom and Standpoints,” published in Archiv für Geschichte der Philosophie. Some sections of Chapter III contain thoroughly revised and rewritten material from my 2020 essay “Spontaneity and Contingency: Kant’s Two Models of Rational Self-­Determination,” published in The Concept of Will in Classical German Philosophy. I am grateful to Chicago University Press for granting me permission to re-­use some content from my 2015 article “Kant on Determinism and the Categorical Imperative,” published in Ethics, for Chapter  V.  I am grateful to Johns Hopkins University Press and Cambridge University Press to re-­use some content from my 2015 and 2020 art­icles “Kant on Freedom of Empirical Thought,” published in the Journal of the History of Philosophy, and “Kant on Cognizing Oneself as a Spontaneous Cognizer,” published in the Canadian Journal of Philosophy, for Chapters VI and VII.

Contents Notes on Sources and Key to Abbreviations and Translations

ix

Introduction 1 PA RT 1 .   T H E BA SIC F R A M EWO R K O F KA N T ’ S D O C T R I N E

I. Freedom, Idealism, and Standpoints

15

II. Human Action as the Effect of Two Causes

51

III. Freedom as Autonomous Self-­Determination

75

PA RT 2 .   T H E G R O U N D S O F KA N T ’ S I N C OM PAT I B I L I SM A B OU T F R E E W I L L IV. Legislative Freedom and Kant’s Genealogical Anxiety

103

V. Executive Freedom, Determinism, and the Categorical Imperative 131 PA RT 3 .   F R E E D OM O F T HOU G H T A S A SP E C I E S O F T R A N S C E N D E N TA L F R E E D OM Transition to Part Three

159

VI. Kant’s Free Thinker

161

VII. Freedom of Thought as a Condition of Theoretical Cognition

205

PA RT 4 .   KA N T ’ S J U S T I F IC AT IO N O F THE BELIEF IN FREE WILL VIII. Kant’s Moral Grounding of Free Will

253

IX. Kant’s Theoretical Defense of Moral Freedom

303

viii Contents

PA RT 5 .   F R E E D OM I N KA N T ’ S A E S T H E T IC S A N D  T H E U N I T Y O F KA N T ’ S D O C T R I N E Summary and Transition to Part Five

335

X. Freedom of Imagination and the “Autonomy of Taste”

339

Bibliography Index

375 391

Notes on Sources and Key to Abbreviations and Translations References to Kant’s Critique of Pure Reason (Critik der reinen Vernunft, Riga: Hartknoch, 1781 and 1787) are given in the standard way by citing pages of the first (“A”) and/or second (“B”) edition, and use the translation of Paul Guyer and Allen Wood, Cambridge (UK): Cambridge University Press (1998). Otherwise, references to Kant’s works use the abbreviations below and cite, in brackets, the volume and page of the Academy edition: Kant’s Gesammelte Schriften, Ausgabe der Preussischen Akademie der Wissenschaften (Berlin: De Guyter, 1900ff.). Details on translations are given in the list of references at the end of this volume. Anth Br EEKU FM GMS JL KpV KU MAN MS NG PND Prol Refl. RezS RGV TP VAMS V-­Anth/Mron V-­LO/Blom V-­LO/Phil V-­LO/Wiener V-­Met/Arn

Anthropologie in pragmatischer Hinsicht (7) Briefwechsel (10–12) “Erste Einleitung in die Kritik der Urteilskraft” (20) “Welches sind die wirklichen Fortschritte, die die Metaphysik seit Leibnitzens und Wolf ’s Zeiten in Deutschland gemacht hat?” (20) Grundlegung zur Metaphysik der Sitten (4) Immanuel Kants Logik, Ein Handbuch zu Vorlesungen (9) Kritik der praktischen Vernunft (5) Kritik der Urteilskraft (5) Metaphysische Anfangsgründe der Naturwissenschaften (4) Die Metaphysik der Sitten (6) Versuch, den Begriff der negativen Größen in die Weltweisheit einzuführen (2) Principiorum primorum cognitionis metaphysicae nova dilucidatio (1) Prolegomena zu einer jeden künftigen Metaphysik, die als Wissenschaft wird auftreten können (4) “Reflexion” (16–19) “Rezension von Schulz’s Versuch einer Anleitung zur Sittenlehre für alle Menschen” (8) Die Religion innerhalb der Grenzen der bloßen Vernunft (6) Über den Gemeinspruch: Das mag in der Theorie richtig sein, taugt aber nicht für die Praxis (8) “Vorarbeit zur Metaphysik der Sitten” (23) “Vorlesungen Wintersemester 1784/1785 Mrongovius (25) “Logik Blomberg” (24) “Logik Philippi” (24) “Wiener Logik” (24) “Metaphysik Arnoldt (K3)” (29)

x  Notes on Sources and Key to Abbreviations and Translations V-­Met/Mron V-­Met/Volck V-­Met-­K2 V-­Met-­K3E V-­Met-­L1 V-­Met-­L2 V-­Met/Mron V-­Met/Her V-­Met/Schön V-­MO/Mron 2 V-­MO/Col V-­MS/Vig V-­NR/Fey V-­Th/Baum V-­Th/Pöl V-­Th/Volck WA WDO

“Metaphysik Mrongovius” (29) “Metaphysik Volckmann” (28) “Metaphysik K2” (28) “Ergänzungen Metaphysik K3” (29) “Metaphysik L1” (28) “Metaphysik L2” (28) “Metaphysik Mrongovius” (29) “Metaphysik Herder” (28) “Metaphysik von Schön, Ontologie” (28) “Moral Mrongovius II” (29) “Moralphilosophie Collins” (27) “Die Metaphysik der Sitten Vigilantius” (27) “Naturrecht Feyerabend” (27) “Danziger Rationaltheologie nach Baumbach” (28) “Religionslehre Pölitz” (28) “Natürliche Theologie Volckmann nach Baumbach” (28) “Beantwortung der Frage: Was ist Aufklärung?” (8) “Was heißt sich im Denken orientieren?” (8)

Introduction 1.  Freedom in Kant and in Contemporary Philosophy The idea of freedom plays a central role in Kant’s philosophy. Kant explains that part of the “treasure” that his critical philosophy “bequeath[s] to posterity” is that it saves human freedom from the mechanism of nature (Bxxiv–­Bxxxi). The “system of the Critique of Pure Reason revolves around two cardinal points: as a system of nature and of freedom . . .” (Refl., 18:679); “the concept of freedom” even “constitutes the keystone of the entire edifice of a system of pure reason” (KpV, 5:3). When we compare Kant’s account of freedom with his discussion of God and immortality of the soul, we find that he assigns freedom a strongly privileged role: whereas God and immortality are matters of faith (mere credibile), the idea of freedom is the only metaphysical idea of reason that belongs to the “matters of fact” (scibile) (KU, 5:467–8). According to Manfred Kuehn’s seminal biography this difference was also reflected in Kant’s personal life: he had no personal faith in God and immortality,1 but the conviction that human reason must be free from fetters such as natural incentives, censorship, and prejudices was central to the way in which he lived and philosophized.2 This conviction highlights an important difference between the ways in which the concept of freedom figures in Kant and in contemporary philosophy. What is at stake in the question of whether we are free agents? In contemporary philosophy, this issue is typically consigned to specialized subfields such as the philosophy of action or cognate discussions of moral responsibility and punishment. While these topics are relevant to Kant as well, in his philosophy the notion of freedom has much larger ramifications. This is due to the inseparable connection that Kant forges between freedom and rationality. In Kant’s mature (mid-­1780s and later) view the question of whether we are free beings is tantamount to the question of whether we are autonomous subjects who can determine themselves by rendering and acting in accordance with objective rational judgments that yield genuine cognitive achievements. One central goal of this book is to explicate the Kantian link between freedom as autonomy and normative concepts like rationality, objectivity, cognition, or knowledge. 1  Kuehn 2001: 3.

2  Kuehn 2001: 146, 379–82.

Kant on Freedom and Rational Agency. Markus Kohl, Oxford University Press. © Markus Kohl 2023. DOI: 10.1093/oso/9780198873143.003.0001

2  Kant on Freedom and Rational Agency For readers who come to this book with the expectation that it focuses on how Kant’s view looks from the vantage point of the contemporary literature, it may come as a bit of a disappointment that moral responsibility does not figure as a major topic of its own right in my reconstruction of Kant’s view. This (relative) neglect is due to the Kantian link between freedom and rational agency. As I will show, for Kant free will is significant primarily because without free will we would not be bound by valid, rational moral norms (“oughts”). We are praise- or blameworthy only for complying or failing to comply with such norms. Thus, for Kant the topic of moral responsibility is only of secondary interest compared to the prior question of whether we have the freedom that qualifies us as proper addressees (and authors) of rational moral norms. There is a further reason why Kant’s conception of freedom is a bit hard to appreciate from a contemporary angle. Given the special systematic role that freedom plays in Kant’s philosophy, understanding Kant’s doctrine of freedom is central to understanding Kant’s philosophy on the whole. Conversely, an understanding of Kant’s doctrine of freedom also requires an understanding of other core elements in Kant’s philosophy—­specifically, his transcendental idealism with its (in)famous distinction between appearances and things in themselves. Kant states that if his idealism is false, i.e., “if appearances are things in themselves, freedom cannot be upheld” (A536/B564). While there have been attempts to consider Kant’s account of freedom apart from his idealism, I will argue that there is no self-­standing core of this account that can be profitably separated from his idealist claim that the sensible-­natural world depends (in some way) on the human mind. Despite these crucial differences, there are nevertheless important points of contact between Kant and contemporary philosophy. Kant’s rejection of compatibilism and naturalism; his attempt to defend a libertarian view of free agency; the link he forges between free will and moral agency; and his insistence that we cannot reach objective cognition (including, as I shall argue, theoretical cognition of nature) unless we are autonomous agents: these are topics of lasting importance which allow Kant to challenge and be challenged by contemporary views. These are also some of the central topics of this book. My overall aim is to develop a novel systematic interpretation and a qualified defense of Kant’s doctrine of freedom. My account centers on three major interpretive and philosophical issues. Sections 2–4 of this Introduction give a survey over these issues and suggest what is at stake in each case. This will both serve as a useful summary and highlight how my account differs from other interpretations.

2.  A Metaphysical Account without a Metaphysical Theory The first issue concerns the question: does Kant’s doctrine of freedom incorporate non-­trivial metaphysical commitments? In particular, does he posit the existence of an atemporal, “noumenal” form of causality that operates outside of nature and

Introduction  3 human experience? There is a sharp dividing line between interpretations which deny that Kant’s doctrine requires such commitments and interpretations for which these commitments are central to Kant’s view. This question is fundamental because how one answers it determines what interpretive and philosophical questions one subsequently faces. Those who side with the anti-­metaphysical reading have the challenging task of explaining away the passages where Kant talks about freedom in straightforwardly metaphysical terms. Proponents of a metaphysical approach must address the worry that the very notion of an atemporal form of agency (or causality) seems incomprehensible.3 Here one seems to face a choice between two programmatic responses: one can either take the unintelligibility of Kant’s metaphysical doctrine of freedom to show that this doctrine is untenable, or one can try to argue that Kant’s metaphysical doctrine of freedom is intelligible and thus philosophically respectable after all.4 The interpretation I develop falls in between these dividing lines. On the one hand, I side with metaphysical interpretations: I read Kant as holding that our belief in free will refers to an atemporal causal power. On the other hand, I use an interpretive tool that is standardly associated with anti-­metaphysical readings: I stress that the appeal to the causality of freedom is confined to a distinctive normative standpoint that is not concerned with explanation, theory-­formation, or prediction. On my reading the belief in freedom qualifies as “metaphysical” insofar as it posits the existence of a causally unconditioned causal capacity, but this belief must remain theoretically indeterminate: it does not lend itself to fruitful metaphysical theorizing about how the causality of freedom works. This means, among other things, that we cannot explain why a free cause produces or fails to produce particular effects such as observable human actions at any given time. The idea of freedom is, by Kant’s explicit admission, incomprehensible to us: because freedom lies beyond the bounds of our spatiotemporal human experience, and because the bounds of such experience are also the bounds of our theoretical understanding, we simply cannot theoretically comprehend the metaphysics of free will, no matter how hard we may try. In my view, the best route to take here for a sympathetic commentator is not to override Kant’s explicit disclaimers and to supply a comprehensible metaphysical theory of freedom on his behalf. Rather, we should question whether the theoretical incomprehensibility of Kant’s account of freedom strictly entails that this account is untenable. This issue bears on my goal to give a qualified defense of Kant’s account. My  defense has two cornerstones. First, Kant has intriguing arguments against compatibilist-­ naturalistic views that make free agency fully explicable and 3  This is perhaps the most common objection to Kant’s doctrine of freedom. In writing, it has been expressed (for instance) by Bennett 1974, 1984; Nagel 1989. I have also encountered it in many discussions. 4  For the first response, see again Bennett 1974, 1984; Nagel 1989. For the best attempt to make sense of Kant’s doctrine as a comprehensible metaphysical account (of free substance causation), see Watkins 2004.

4  Kant on Freedom and Rational Agency comprehensible (see Section 3). Given these arguments, the cost of theoretical inscrutability incurred by Kant’s model of freedom emerges as a price that might be worth paying. Second, neither Kant’s admission that freedom is theoretically inscrutable nor his claim that this raises no fatal problem are borne out of philosophical desperation. Rather, these points are grounded in a systematic conception of the cognitive resources, limitations, and needs of human reason. It is by demarcating the necessary limits of human reason that Kant reveals the necessary futility of our efforts to comprehend free agency. Thus, while Kant’s view of human reason does not allow us to comprehend freedom, it does allow us to “comprehend its incomprehensibility” (GMS, 4:463): it systematically explains why free causes are beyond the limits of our comprehension. Moreover, Kant’s systematic view of reason consigns the idea of freedom to a normative “standpoint” that considers only what ought to happen. The normative enquiry we conduct from this standpoint can “happily” (KpV, 5:49) set aside the purposes and constraints that apply (only) within the different context of observing, explaining, and predicting what does or will happen. Thereby Kant establishes that there is no legitimate rational need for us to pursue inscrutable theoretical questions about how free causes operate. There are several reasons why my book yields only a qualified defense of Kant’s doctrine. For one, I cannot show that Kant has conclusive arguments against every competing compatibilist or naturalistic account on which freedom is a theoretically respectable concept. Moreover, I acknowledge (more strongly than Kant does) that the theoretical inscrutability of freedom does impose a severe cost. Whether this cost is, in the end, too severe remains for the reader to judge.

3.  Kant’s Critique of Compatibilism and Naturalism The preceding remarks suggest that whether Kant’s account of freedom is defensible hinges crucially on whether he can motivate his rejection of accounts that (unlike his own) do not incur the cost of theoretical inscrutability. Both from a historical and a contemporary perspective, the main competitor for Kant’s model of freedom are compatibilist views on which human freedom can be integrated into our theoretical understanding of the natural world. Naturalistic compatibilism has been the dominant view in Anglo-­American philosophy with defenders such as Daniel Dennett, David Lewis, G. E. Moore, Peter Strawson, or R. Jay Wallace.5 Compatibilism was also the dominant view held by Kant’s most im­por­ tant historical predecessors including Hobbes, Hume, Wolff, and Leibniz.6

5  See Dennett 1984; Lewis 2003; Moore 1912; Strawson 2003; Wallace 1994. 6  Note, however, that these philosophers propose very different versions of compatibilism.

Introduction  5 One central worry about Kant’s view is that he fails to offer good reasons for rejecting compatibilism. Karl Ameriks has suggested that Kant’s incompatibilism is purely dogmatic, thus lacking rational support.7 If this assessment were correct, it would expose a damning deficiency in Kant’s view: Kant would be offering us an inscrutable model of free agency without explaining why we should reject alternative naturalistic models that allow us to understand and explain free agency. Ameriks correctly notes that Kant’s reasons for rejecting compatibilism are not readily perspicuous; what Kant explicitly says on the topic (e.g., at KpV, 5:96–7) can seem like an uncharitable dismissal revealing “a mere dogmatic attachment” to incompatibilism.8 However, I shall argue that one can reconstruct interesting, forceful, and original arguments for incompatibilism and anti-­ naturalism from Kant’s writings. Reassessing the grounds of Kant’s anti-­ naturalistic incompatibilism is one way in which this book contributes towards a more favorable understanding of Kant’s doctrine. For Kant, “if all causality . . . were mere nature” this would annihilate our freedom (A534/B562). Two central components of how Kant conceives “the causality of nature” are especially relevant here. First, natural causes determine their effect(s) with “hypothetical necessity” (A228/B280): if a natural cause is operative, then its characteristic effect(s) must occur. Second, since the operation of any natural cause is itself a natural event that is hypothetically necessitated by preceding causes, the exercise of every natural causal power is contingent upon (an indefinitely extended chain of) antecedent natural events and conditions. I argue that in Kant’s view these two features of natural causes threaten human freedom in two different ways. The first threat relates straightforwardly to traditional, ongoing debates between compatibilists and incompatibilists about questions such as the following: Does free agency require alternative possibilities? Does the fact that an agent is causally determined not to perform some action show that the agent cannot perform that action? I argue that Kant makes an important contribution to this debate at least as far as the first question is concerned because he provides an original motivation for the view that free human agency requires metaphysically real alternative possibilities. For Kant, this requirement follows from reflection on what it takes for imperfectly rational (e.g., human) agents to act under normative, specifically moral laws of reason: such agency involves both the option to act rightly and the option to inveigh against right reason. This presupposition of free human agency would be false if every human action was the result of natural necessitation because in that case human agents would be inevitably determined to perform one particular act at a particular time. This is one way in which Kant 7  Ameriks 2000a, 2013. Irwin 1984 and Wood 1984 express similar dissatisfaction with Kant on this issue. 8  Ameriks 2000a: 227.

6  Kant on Freedom and Rational Agency motivates the idea that natural determinism is incompatible with our human stance of free rational agency under normative laws. Kant’s second worry concerns the fact that every natural cause and its products are contingent upon preceding natural conditions; natural causes or their products never possess an absolute-­unconditional necessity. This link between natural causality and contingency becomes a problem if we consider human representations that purport to have absolute-­ unconditional necessity. For Kant, the alleged absolute-­unconditional necessity of a representation (an idea, a concept, or a judgment) would be illusory if this representation were the effect of natural causes whose efficacy is contingent upon a chain of preceding natural conditions. If a judgment has an empirical origin in natural causes, the judgment lacks necessary validity due to the contingency of its empirical origin. If we push the chain of natural conditions back far enough, we find that the judgment is ultimately contingent upon natural conditions that have no rational significance at all. This is an abstract point which requires (and will receive) much further clarification. But the intuitive pull of the second worry can perhaps best be illustrated by considering our acts of moral judgments. If those acts were just the causal upshot of our empirically given, contingent (psychological, biological) make-­up, then the resulting moral judgments could not prescribe with valid unconditional necessity how we (rationally speaking) must act whatever our empirical constitution (such as our contingent empirical desire-­base) happens to be. For Kant this concession would be detrimental to our moral self-­conception. Moreover, in his view our only basis for thinking that we have free will derives from our awareness that our will is governed by a priori necessary moral norms. If this moral self-­ awareness is illusory (because it is merely a product of contingent empirical sources), then our belief in free will is also mistaken. Since this worry concerns the contingency rather than the deterministic character of natural causes, it is not part of traditional debates among compatibilists and incompatibilists. But (as I will show) it does connect in interesting ways with contemporary debates about evolutionary debunking arguments. The two concerns I have sketched provide a respectable argumentative basis for Kant’s claim that our moral freedom of will cannot be understood in compatibilist-­naturalistic terms.

4.  Freedom of Will and Freedom of Thought Interpretations of Kant’s doctrine tend to focus exclusively on moral freedom of will. This is understandable because free will is clearly the focal point of Kant’s account. However, my interpretation denies that moral freedom of will is the only “real” kind of freedom that Kant acknowledges. It is a central component of Kant’s

Introduction  7 philosophy that we also exercise absolute freedom of thought qua rational autonomy in our theoretical reasoning and judgment. This point is philosophically significant (and controversial) in its own right. But on my reading, it is also relevant for appreciating how Kant seeks to justify our belief that we have moral freedom of will. A common worry about Kant’s view is that he ultimately abandons his previous attempts to provide a genuine justification for the belief in moral freedom of will. For some commentators Kant’s eventual (post-Groundwork) attempt to support our belief in noumenal freedom will through the moral “fact of reason” doctrine signals a lapse into dogmatic rationalism.9 By contrast, I argue that the “fact of reason” doctrine does not compromise Kant’s attempt to provide a rational justification for our belief in moral freedom of will. Kant’s account of the moral fact of reason as our epistemic basis for recognizing our noumenal free will does not simply ignore challenges from those who are skeptical about a priori moral laws or about supersensible free will. Rather, Kant defends his moral grounding of free will against those challenges by invoking our freedom of theoretical thought. The main point of this defense strategy is roughly as follows. Consider naturalists who doubt both the belief (1) that we are governed by a priori necessary moral norms and the (corresponding) belief (2) that we have absolute moral freedom of will. Kant argues that those who raise these naturalistic doubts are (however implicitly) committed to the belief (3) that they exhibit freedom of thought in their own naturalistically oriented theoretical reasoning. This is because their naturalistic outlook presupposes the validity of a priori necessary theoretical principles; this presupposition in turn requires the belief in the rational autonomy of our pure intellect as the source of these principles. Thus, the attempt to explain away morality and free will through naturalistic considerations is incoherent: cognizers who purport to give valid naturalistic explanations must rely on their free capacities for a priori theoretical cognition that (Kant argues) are analogous to our free capacities for a priori moral cognition and choice. Since theoretical cognizers must accept the validity of their consciousness as autonomous epistemic subjects, they cannot coherently debunk our analogous self-­consciousness as autonomous moral subjects. I will provide a rigorous reconstruction of this line of argument by showing how it connects with and contributes to Kant’s moral fact of reason doctrine. On this basis, I will also argue for the controversial view that in Kant’s considered view we possess objectively certain knowledge (not just subjectively rational faith) that we have free will.

9  See, e.g., Ameriks 2000a, 2003, 2011, 2013; Bittner 1983; Guyer 2007; Prauss 1983; Wood 2007.

8  Kant on Freedom and Rational Agency

5. Naturalism The above considerations indicate that the primary antagonist in Kant’s account is the “defender of an omnipotent nature” (A499/B477) who believes that “all causality is . . . nature.” This antagonist privileges the empirical standpoint of natural science and adopts a systematic naturalistic worldview that incorporates a package of related metaphysical and epistemological claims. These claims undermine corresponding anti-­naturalistic convictions that are central to our self-­conception as free persons.10 The dialectic between Kant and the naturalistic worldview is central for understanding and evaluating his doctrine of freedom. My account of this dialectic will frequently consider not just the historical (e.g., Humean) versions of naturalism that Kant was familiar with but also related contemporary views that combine a naturalistic ontology with an empiricist epistemology. This broader scope further contributes to showing that Kant’s defense of freedom does not lose its relevance in a contemporary setting. Since “naturalism” is a highly ambiguous term, it will be helpful to have a clear definition at hand. The kind of naturalism that Kant opposes affirms the following four theses:

(1) The only capacities that human beings possess are empirical powers whose existence we cognize on the basis of empirical evidence (perception, experiments, etc.). (2) Everything that happens, including every human action, is governed by laws of nature; there are no causal powers that do not operate in ac­cord­ ance with laws of nature. (3) The only actual (and really possible) efficient causes are those that figure in the explanations of the natural sciences (physics, chemistry, biology, psychology). (4) Causes inevitably necessitate their effects; causal laws are deterministic. Claims (1)–(4) yield the metaphysical position that all reality is exhausted by nature11 and the epistemological view that only empirical sources can legitimize beliefs about which things exist. Kant assumes that naturalists endorse (4) because of his own view that natural causes necessitate their effects. When Kant reflects on what would follow “if all causality . . . were mere nature” (A534/B562), he considers a systematic naturalistic worldview that comprises (1)–(4).12 10  I thus accept Ameriks’s important (2000b) point that Kant seeks to reconcile our scientific and our “manifest” image. But my understanding of how Kant conceives the manifest image (namely, as extending to our self-­image as theoretical cognizers) and how Kant seeks to achieve the reconciliation (namely, in part, by arguing that the scientific image itself presupposes the manifest image) differs from Ameriks’s view. 11 For the view that this point is central to naturalism in (Anglo-)American philosophy, see Kim 2003. 12  Kant’s naturalist-­empiricist opponent adopts the antitheses of the Antinomies of Pure Reason. The worldview that comprises (1)–(4) appears in Kant as one version of what he calls transcendental

Introduction  9 My definition of naturalism is stipulative: it tracks a systematic position that Kant was concerned to address and to refute. Here we must note two important qualifications. First, Kant does not commit his naturalistic opponent to materialism. His worries about the implications of naturalism do not include the concern that on a naturalistic view there are no mental phenomena or that our conscious deliberation is merely epiphenomenal. For Kant, the naturalist is fully entitled to the view that representations, volitions etc. can function as real natural causes (see e.g., KpV, 5:95–6). He accepts that empirical psychology is among the natural sciences mentioned in (3) (though he also places tight limits on what empirical psychology can accomplish). Second, some contemporary naturalists reject determinism (4), typically because of their views about the implications of quantum mechanics. It is controversial whether quantum mechanics really disproves determinism.13 Moreover, Kant’s account of natural causality may well allow for the possibility that (some) empirical laws have an inherently probabilistic or statistical character.14 In this book, I shall assume: first, that for Kant natural causes determine the occurrence of their effects in some sense that may allow for probabilistic interpretations; second, that the appeal to probabilistic natural causes does not mitigate (his) concerns about human freedom.15

6. Overview I have sketched in broad strokes certain key ideas that will receive a more detailed presentation, analysis, and defense over the course of this book. The book divides into four parts. Part 1, which comprises Chapters I–­III, presents the metaphysical, epistemological, and “semantic” framework of Kant’s doctrine. Here I show how Kant’s

realism. This version identifies the conditions of human experience with absolute ontological conditions of things in general. Kant sometimes uses the term “naturalism” in different senses, e.g., for a view on which common sense has greater authority than natural science in questions of empirical cognition (A855–6/B883–4). Bird (2006: 53) argues that Kant himself is a naturalist insofar as he accepts the authority of natural science regardless of whether philosophical reason can answer epistemological skepticism. Even if Kant were a naturalist in this specific sense (which I deny in Chapter IX), he would still be opposing naturalism in the different sense that I have defined. 13  For the view that determinism is still a live issue that cannot be refuted by quantum mechanics, see Earman 2004. 14  See Guyer 1987: 240–1. Some contemporary philosophers (such as Schaffer 2001) also hold that probabilistic causes and laws leave room for natural necessitation. I consider this issue in Chapter VII. 15  For a contemporary account of why an appeal to indeterminism cannot mitigate concerns about free will, see Loewer 1996. By contrast, Kane 1996 tries to develop an indeterministic-­naturalistic form of libertarianism. For Kant, if a causal relation or law is probabilistic, the probability of the effect following from the cause must be sufficiently high to allow for objective time-­determination and theoretical explanation. Thus, the empirical probability that a naturally caused immoral action might not have occurred must be correspondingly low; so low indeed that invoking this minimal chance fails to show that the agent was truly free not to perform this action.

10  Kant on Freedom and Rational Agency view is to be understood in the context of his idealism and what this implies for his answers to foundational questions such as these: How can we make room for an absolutely free causality alongside the deterministic causality of nature? How should we conceive the relation between the causality of freedom and the causality of nature? What does Kant’s abstract idealist defense of free will imply for our ordinary self-­conception as free agents? How can we adequately represent (i.e., think about and refer to) a supersensible type of causality via concepts that are essentially discursive functions for combining sensible data? What is distinctive about our human freedom in comparison with divine freedom? One main ambition of Part 1 is to develop and defend my aforementioned claim that Kant’s doctrine incorporates a metaphysically weighty belief in supersensible freedom that does not lend itself to metaphysical theorizing and that must, therefore, remain theoretically inscrutable. This invites a further central question: why should we prefer Kant’s account of noumenal freedom over metaphysically modest, theoretically fruitful compatibilist-­naturalistic models of freedom? Part 2, which comprises Chapters IV–­V, addresses this question by uncovering the sources of Kant’s incompatibilism about free will or practical reason. I reconstruct two separate arguments through which Kant seeks to show that his incompatibilist, anti-­ naturalistic view of free will is more adequate to our moral self-­conception than compatibilist-­naturalistic views. These are the two arguments whose main points I sketched earlier (in Section 3). They concern the absolute necessity of our moral judgments (Chapter IV) and the absolute contingency of the moral choices we make in light of these judgments (Chapter  V). Neither of these can be reconciled with the “causality of nature” in Kant’s view. Thus, compatibilist-­ naturalistic views cannot do justice to our moral self-­ conception as free agents who choose under necessary laws of practical reason. In Part 3, which comprises Chapters VI–­VII, I argue that Kant’s conception of transcendental freedom is not limited to our moral freedom of will but also extends to the freedom of thought that we exercise in theoretical cognition. I  show that Kant has (and never abandons) a viable, non-­mechanistic yet also non-­voluntaristic model of absolutely free epistemic agency. I also explain why Kant denies that can regard ourselves as thinking mechanisms whose theoretical judgments are determined by “foreign” causes and why he views freedom of thought as a sui generis type of transcendental freedom that differs from moral freedom of will. In Part 4, which consists of Chapters VIII–­IX, I reconstruct Kant’s two-­pronged strategy for justifying our belief in transcendental freedom of will as an objectively certain piece of knowledge (Wissen). Kant argues that our objectively ra­tional basis for adopting the belief in free will is our a priori knowledge of the moral law as the fundamental axiom of all practical justification. But Kant also defends this moral grounding of the belief in free will against the naturalistic

Introduction  11 attempt to debunk our a priori moral self-­consciousness as an invalid, empirically conditioned illusion. This defense rests on the abovementioned argument that naturalistic cognizers must presuppose the validity of their a priori theoretical self-­awareness as free thinkers. In Part 5 (Chapter  X), I consider Kant’s appeal to the aesthetic freedom of imagination that we show in our creation and reception of beauty. In contrast with moral freedom of will and theoretical freedom of thought, imaginative freedom is a freedom from a priori laws of reason. This makes it seem like a lawless, chaotic “liberty of indifference”—an implication that would compromise the systematic unity of Kant’s doctrine of freedom, because this doctrine standardly construes freedom as an essentially rule- or norm-­governed capacity. Against this, I argue that Kant’s appeal to aesthetic freedom can be taken to fruitfully expand and complete Kant’s doctrine: our imagination displays a further (third) species of transcendental freedom whose distinctive mark is that it brings the rational-­intellectual side of our nature in tune with the affective-­sensible aspect of who we are. Our free imagination enables a playful harmony between these two sides which grounds our aesthetically pleasant engagement with beauty in art and in nature. If this final component of my account succeeds, my interpretation establishes the idea of freedom as the proper anchor of all meaningful human activity (as Kant conceives it) in its practical, theoretical, and aesthetic dimensions. Transcendental freedom governs our moral striving to become more virtuous and to make the world a better place; it provides a rational basis for our theoretical efforts to understand, explain, and predict the world; and it enables our aesthetic engagement with the world of beauty, both artistic and natural.

PART 1

T HE BASIC F R A M EWOR K OF KA N T ’S D O C T R I NE

I

Freedom, Idealism, and Standpoints Kant’s position concerning free will has seemed elusive to many philosophers. Bernard Williams summarizes the “central mystery” as follows: [Kant] . . . was a determinist about all events; recognized that actions were events; believed in free will; and condemned the “wretched subterfuge,” as he called it, of making free will and determinism compatible with one another by ascribing free actions to a particular kind of cause.1

Williams correctly diagnoses that Kant’s view seems paradoxical in its attempt to combine a range of commitments that most philosophers deem mutually exclusive. In this chapter I develop a basic interpretive framework that explains how Kant can coherently accept these commitments.

I.1.  Six Interpretive Desiderata I begin by considering Kant’s commitment to natural or empirical determinism: “[T]hat all occurrences are empirically determined in a natural order . . . is a law of the understanding, from which under no pretext can any . . . appearance be exempted . . .” (A542/B570). This “law of the understanding” is the cognitive principle that Kant proves through the Second Analogy argument. According to this argument, it is a constitutive condition of any event or objective succession, where something changes from a temporally prior to a posterior state, that it be empirically determined (A200/B246). Events are ‘empirically determined’ when their occurrence is due to ‘the causality of nature’ (A532/B560). This mode of causality comprises three key features. First, natural causes render their effects necessary in a hypothetical sense: “Everything that happens is hypothetically necessary” (A228/B280), i.e., follows necessarily given some preceding causal condition. Second, natural causes are governed by invariable empirical laws (A536/B564): if cause C1 is operative and certain other natural conditions obtain, some characteristic effect E1 always follows. Third, the exercise of any natural causal power C1 that produces some natural event E1 is itself a natural event: this exercise (“the causality of the cause”) is the hypothetically necessary effect E2 of some further 1  Williams 1986: 211. In what follows, I will draw (but significantly expand on) Kohl 2016. Kant on Freedom and Rational Agency. Markus Kohl, Oxford University Press. © Markus Kohl 2023. DOI: 10.1093/oso/9780198873143.003.0002

16  Kant on Freedom and Rational Agency natural cause C2; the exercise of C2 that produces E2 is the necessary effect E3 of some further natural cause C3; and so on ad indefinitum. “Everything that happens must have a cause, and . . . the causality of the cause, as itself having happened or arisen, must in turn have a cause . . .” (A533/B561) so that all empirical events are connected in “the steady chain of nature” where “the law of natural necessity reigns” (KpV, 5:95). Since this law of natural necessity determines the course of events in a successive temporal order, where things have (and can be cognized as having) objective temporal positions, it applies to everything that we can experience in time: “. . . the principle of the causal relation in the sequence of appearances is valid for all objects of experience (under the conditions of succession) . . .” (A202/B248). Accordingly, the law of natural necessity is restricted to the specifically temporal mode of causality, the causality of nature, which governs all objects of our temporal experience (A181/B223).2 Hence, this law does not imply that every “causality of a cause” per se is the hypothetically necessary effect of some further cause. This leads to a first interpretive desideratum (D1): we must read Kant as endorsing determinism with respect to all natural-­temporal events including all exercises of natural causal powers, but also as allowing for the possibility (in a sense to be specified) of causally undetermined non-­temporal actions. Kant’s second commitment derives from his definition of transcendental freedom as “a spontaneity which can begin to act of itself, without requiring to be determined to action by an antecedent cause . . .” (A532–3/B560–1). Since compatibilist views allow that free actions may be determined by antecedent causes, we must read Kant as endorsing an incompatibilist conception of freedom (D2). When Kant considers the compatibilist idea that we act freely if our actions are causally determined by our (causally determined) inner representational states, he rejects it as a “wretched subterfuge,” a “petty word-­jugglery” (KpV, 5:96). But on the other hand Kant also says things that sound congenial to compatibilism: “. . . freedom and [the deterministic causality of] nature, in the full sense of these terms, can exist together, without any conflict, in the same actions” (A541/ B569); the deterministic causality of “nature does not contradict the causality of freedom” (A599/B587). Hence (D3), we must make room for the fact that in Kant’s view freedom is somehow compatible with natural determinism. The need to conjoin (D2) and (D3) as components of Kant’s view poses a serious problem because compatibilism and incompatibilism seem mutually exclusive. Some scholars argue that Kant’s view overcomes this opposition by combining

2  “The law of natural necessity” is the law of the understanding that Kant proves in the Second Analogy. I discuss some aspects of this proof and its implications in Chapters II (see especially n. 2) and VII (see especially n. 74).

Freedom, Idealism, and Standpoints  17 compatibilism and incompatibilism.3 But it is unclear how this combination of logical contradictories is supposed to work. The next two interpretive desiderata derive from the fact that Kant envisages two different philosophical opponents for his idealist doctrine. These opponents are both transcendental realists who accept that we know objects as they are in themselves, i.e., that our theoretical cognition has access to mind-­independent reality. Within the genus of transcendental realism Kant distinguishes between dogmatic empiricists who adopt a naturalistic worldview (see Section 5 in my Introduction) and dogmatic rationalists (A466/B494).4 He describes their characteristic errors as follows: “. . . to make principles of possible experience conditions of the possibility of things in general is just as transcendent a procedure as to assert the objective reality of concepts, the objects of which cannot be found anywhere save outside the limits of all possible experience” (A781/B809). The first point concerns the naturalistic view that whatever is not in accordance with conditions of experience or nature is absolutely impossible (A780/B808). One such condition is the abovementioned “law of natural necessity.” For naturalists, this law entails the impossibility of freedom (qua absolute spontaneity of action): “There is no freedom; everything in the world takes place solely in accordance with the laws of nature” (A445/B473). Kant’s second point concerns the rationalist view that freedom must exist because without free causes we cannot sufficiently explain what happens: since every natural causality is the effect of some further cause, a sufficient causal explanation requires a causally unconditioned causality as the unmoved mover of the chain of natural causes and events: “to explain . . . appearances it is necessary to assume that there is also another causality [apart from nature], that of freedom” (A445/B473). Here the rationalist assigns objective reality to the concept of a free cause although such a cause lies beyond the limits of our spatiotemporal experience. Kant agrees with naturalists that we must invoke the causality of nature to explain anything at all. This has a twofold implication. First, since “where determination according to natural laws cease, there all explanation also ceases” (GMS, 4:459), it follows that “from the concept of freedom nothing can be explained in the appearances” (KpV, 5:30); “the intelligible” (supersensible) “is . . . of no use to us for the explanation of appearances” (A562/B580). Second, the non-­natural causality of transcendental freedom not only fails to explain anything but is itself incomprehensible: “Freedom . . . is a mere idea, the objective reality of which can in no way be shown according to laws of nature, and consequently not in any possible experience, and which . . . can never be comprehended

3  See Allison 1990: 28 and 2011: 348–9; Proops 2021: 322; Watkins 2004: 335; Wood 1984: 74. 4  See Allison 2004: 28–35 for the point that transcendental realism is a general stance underlying competing views.

18  Kant on Freedom and Rational Agency or even understood” (GMS, 4:459).5 Thus, Kant accepts “the subjective impossibility to explain freedom of will” (GMS, 4:459). This leads to desideratum (D4): we must respect Kant’s insistence that freedom and free actions are theoretically inscrutable for beings endowed with our cognitive faculties. A corollary of this principle is that for Kant the actuality or real possibility of free will cannot be demonstrated (A557–8/B585–6)—at least not through a theoretical argument such as the abovementioned rationalistic causal sufficiency-­proof (the thesis argument of the Third Antinomy). However, Kant also opposes the naturalistic denial of freedom. We are licensed at least to “think of ” ourselves as free (GMS, 4:455) and to “presuppose” our free will (GMS, 4:447). Thus (this is (D5)), we must read Kant as trying to save our ordinary self-­image as free agents from “the mechanism of nature” (Bxxiv–­Bxxxi). As we shall see, one of the most central interpretive tasks here is to get clear about what kind of mental state or doxastic attitude Kant refers to when he says that can “think of ourselves as free.” Finally, Kant insists that a transcendental realist (of any variety) who identifies appearances and things in themselves is committed to denying that we are free: “if appearances are things in themselves, freedom cannot be upheld” (A536/B564; cf. Bxxxvi–­xxviii; GMS, 4:456–9; KpV, 5:95–9). Hence (D6), a viable interpretation of Kant’s doctrine must explain why saving our self-­image as free agents strictly requires Kant’s idealism. In the remainder of this chapter I employ these desiderata to settle some fundamental interpretive questions. I begin with the critique of an important and popular reading.

I.2.  Against Deflationary Interpretations Kant declares that we have “two standpoints” from which we can consider ourselves and our actions: we can either regard ourselves as subject to empirical laws of nature or as free from such laws (GMS, 4:452). According to an influential reading, Kant’s appeal to these two standpoints allows us to interpret his doctrine of freedom without controversial metaphysical commitments. For Thomas Hill, Kant’s view should be taken: . . . as only an attempt to distinguish two perspectives on human action, the theoretical/empirical perspective appropriate to natural science, and the practical/ evaluative perspective when we think about reasons for acting, obligation, and responsibility . . . the interpretative strategy is to admit that the practical 5  For the difference between mere einsehen (understanding) and begreifen (comprehending), see JL, 9:65.

Freedom, Idealism, and Standpoints  19 perspective is committed to irreducibly normative ideas, but deny that it is . . . committed to faith in mysterious entities outside of space and time.6

Likewise, for Christine Korsgaard Kant’s idea that we can regard ourselves as free from a non-­empirical standpoint “does not commit us to a belief in a mysterious form of supersensuous existence.”7 Although Kant’s distinction between a sensible and a noumenal world “has lent credence to the interpretation of the distinction as an ontological dualism,” this ontological interpretation is flawed: “Actually these two worlds are two standpoints, or ways we have of looking at things . . . they represent a practical and a theoretical standpoint.”8 In similar fashion Henry Allison argues that although Kant talks about free action “in dauntingly metaphysical terms,” he really requires only that we “view” or “regard” ourselves as absolutely free, atemporal noumenal beings.9 Properly understood, Kant’s doctrine of freedom requires (and licenses) “no ontological conclusions regarding the absolute spontaneity of the self.”10 Implicit in this approach is the conviction that if we can manage to understand Kant’s doctrine without ontological commitment to the (actual or possible) existence of supersensible causes, this is a decisive interpretive and philosophical benefit.11 This deflationary approach is often called “the two-­standpoint reading.”12 In this section I examine whether an interpretation that replaces a focus on ontology (existents) with a focus on standpoints can satisfy the interpretive desiderata I introduced in Section I.1. One central question here is whether the metaphorical expressions that we can “look at,” “view,” or “regard” ourselves as free entail a belief in freedom. Does “regarding myself as free” require that I believe that I am free, in one of the three modes of believing or holding-­to-­be-­true (Fürwahrhalten) that Kant distinguishes (knowledge, opinion, or faith; A820–32/B848–60)? The two-­standpoint reading can avoid “daunting” ontological commitments only if it can show that we need not believe in the (actual or possible) existence of a supersensible causality. Thus for Korsgaard, when Kant says that we must “act under the idea of freedom” (GMS, 4:448), his “point is not that you must believe that you are free, but that you must choose as if you were free.”13 For Allison, the suggestion “that one must act as if one were free, without necessarily having the belief ” should replace the

6  Hill 2002: 36. 7  Korsgaard 1996a: xi. 8  Korsgaard 1996a: 185. 9  Allison 2001: 608. Compare: Allison 1990: 46–52. 10  Allison 1996: 64. 11  This conviction rests (in part) on the identification of a metaphysically oriented interpretation with a “two-­world” reading which posits, as Korsgaard says, “an ontological dualism” between appearances and things in themselves as numerically different entities (1996a: 185). See also Allison 2011: 352–4; Beck 1960: 192. 12  Recent proponents of this approach besides Allison, Hill, and Korsgaard include Frierson 2011 and O’Neill 1989. For earlier versions, see Beck 1960; Matthews 1969; and Paton 1958: 266–7. 13  Korsgaard 1996a: 162.

20  Kant on Freedom and Rational Agency “implausible” idea that “one must believe . . . that one is free.”14 Does this proposal satisfy (D1)–(D6)? It neatly satisfies the compatibilist desideratum (D3). If Kant’s doctrine of freedom requires no belief that we are free in an incompatibilist sense, this clearly explains why he deems our self-­image as free agents compatible with natural determinism: choosing as if you were free “is quite consistent with believing yourself to be fully determined.”15 The two-­standpoint reading also gives an attractively simple, modest account of (D1) and (D5): when Kant appeals to the possibility of a non-­natural causality (D1) and says that we are entitled to “think of ” ourselves as free (D5), he only seeks to provide a model for conceiving ourselves as acting freely. He does not require any type of justified belief (Fürwahrhalten) that this conceptual model is or really (metaphysically speaking) could be realized. Finally, the suggestion that we conceive our freedom only from a practical-­deliberative standpoint which does not involve any theoretical beliefs nicely captures Kant’s emphasis that the idea of freedom has no legitimate theoretical use (D4). However, the deflationary reading falters with regards to the incompatibilist desideratum (D2).16 For incompatibilists, the assumption that our actions are causally determined raises a threat to our rational agency. If determinism does raise such a threat, then those who take up the standpoint of rational agency must believe that they have (or at least might have) causally undetermined capacities for deliberative choice. Conversely, if this belief is not required for our self-­ conception as rational agents, then the assumption that we are determined raises no threat to our self-­conception as rational agents: in which case incompatibilism is false. Suppose someone claims that my deliberative course is entirely fixed by the mechanism of nature; and suppose that I can rationally make my decision without any need to believe that this determinist claim is (or might be) incorrect. This shows that I am a compatibilist. In other words, compatibilism is a “wretched subterfuge” only if rational agents must adopt the belief that their will is (or might be) free from causal determination. Since the deflationary two-­standpoint reading dispenses with this belief, it cannot account for Kant’s incompatibilism. Allison correctly designates Kant’s claim that we incorporate desires into our maxims through an absolutely spontaneous act of choice as “a conceptual claim about the model . . . we are constrained to adopt insofar as we regard” ourselves as rational agents.17 But he goes on to suggest that this is a merely conceptual claim, “not a metaphysical thesis about the behind the scene activity of a noumenal agent”:18 “Kant is . . . claiming merely that it is necessary to

14  Allison 2011: 304. 15  Korsgaard 1996a: 162. 16  Here I am elaborating on a criticism that Ameriks (2003: 191) raises for Paton’s two-­standpoint reading. 17  Allison 2020: 278. 18  Allison 2020: 278.

Freedom, Idealism, and Standpoints  21 appeal to the transcendental idea of freedom in order to conceive of ourselves as rational . . . agents, not that we must actually be free in the transcendental sense.”19 However, if Kant does not claim that we must actually be transcendentally free in order to qualify as rational agents, then he has no legitimate basis for insisting that rational agents must nevertheless “appeal to” or care about their transcendental freedom. We are constrained to adopt a conceptual model that represents our choices as transcendentally free only if, and only because, the actual possession of transcendental freedom is required for being a rational agent. Conversely, if our status as rational agents does not hinge on whether we really are transcendentally free, then we have no need for Kant’s loaded conceptual model: we should instead conceptualize our free rational agency without appeal to supersensible causes, as being compatible with natural determinism. Proponents of the two-­standpoint view might respond here as follows: Kant’s incompatibilism requires not that we believe in undetermined causes but that (from the practical standpoint) we bracket or set aside the theoretical belief in determinism. Now, this cannot mean that as agents we need not consciously represent the absence of determining causes: for Kant this representation is in the foreground of our agential self-­consciousness, as an idea “under which” we act (GMS, 4:448). Korsgaard clarifies that when we bracket the belief in determinism we consciously set aside the truth of determinism as “irrelevant” for the practical purpose of agency.20 Even if you know that all your deliberative steps are determined by a device implanted in your brain, “in order to do anything, you must simply ignore the fact that you are programmed, and decide what to do—­just as if you were free.”21 Hence, our practical situation is exactly the same whether or not we are in fact free from causal determination: “Kant’s answer to the question whether it matters if we are in fact . . . free is that it does not matter.”22 But if it is does not matter to our agential self-­conception whether or not we are in fact free from causal determination, then there is also no need for us to bracket the belief in determinism or to act under an incompatibilist idea of freedom. The compatibilist view that Kant rejects as a “wretched subterfuge” (at KpV, 5:95–7) says: our actions are free if they result from our own representations and deliberative choices, regardless of whether the course of our deliberation is itself causally determined. If Kant held the conviction that it does not matter to our practical situation whether our deliberative processes are causally determined, his harsh dismissal of a compatibilist view which incorporates this very conviction would be plainly irrational.

19  Allison 1990: 57. Compare Allison 1996: 64; Allison 2011: 304; Allison 2020: 260. 20  Korsgaard 1996a: xi. 21  Korsgaard 1996a: 162–3. 22  Korsgaard 1996a: 176. Compare Hill 1992: 135–40; Korsgaard 1996b: 95.

22  Kant on Freedom and Rational Agency The deflationary two-­standpoint reading faces a further problem. As we saw, for Kant one cannot “uphold” freedom unless one rejects the realist identification of natural appearances and things in themselves that we encounter in (e.g.) naturalism (D6). For the “defender of an omnipotent nature” (A499/B477), the only actual and really-­metaphysically possible form of causality is the causality of nature (A445/B473). This naturalistic view surely permits us to “choose as if we were free” or to conceive of ourselves as acting independently of determination by natural forces. What the naturalist denies is that we can adopt (legitimate) beliefs in the actual or possible existence of non-­natural causal powers. But such “ontological” beliefs are not required by the deflationary two-­standpoint reading. Hence, this reading cannot account for Kant’s insistence that we need his idealism to save our freedom. Allison’s view suggests a response to this problem. On his view, what is “at issue” in Kant’s denial that all causality must be of the natural (causally determined) type “is merely whether it is . . . possible to find conceptual space for . . . a different conception of causality (transcendental freedom).” Kant’s idealism is  needed to “find conceptual space” for this “conception of ” transcendental freedom.23 This suggests that the naturalist (Antinomy-­antithesis) variety of transcendental realism renders the very concept of such freedom impossible; Kant’s idealism is thus required to save the logical possibility and conceivability of transcendental freedom.24 However, transcendental realists of a naturalist-­empiricist bent can surely grant the bare conceivability of transcendental freedom.25 For naturalists “transcendental freedom . . . is an empty thought-­entity” since “it is not to be met with in any experience” (A447/B475). An empty thought-­entity is a logically coherent representation that lacks objective reality, i.e., that does not refer to any actual or a really possible object. (Whereas, for instance, “non-­extended body” is not even a coherent thought-­entity.) Thus, naturalists should be happy to allow that we can represent spontaneous unconditioned causes, just as we can represent rapping unicorns. They can also permit the internal coherence of the “thought” that these causes operate outside of time and nature—­without accepting that this thought expresses a real possibility.26 If Kant “is . . . engaged in conceptual analysis rather

23  Allison 2004: 383. 24  Compare Allison 1990: 46: “transcendental idealism . . . is required to create the necessary conceptual space for [the idea of transcendental freedom]”; Allison 2020: 328: “establishing the logical possibility of freedom” is the main goal of Kant’s solution to the Third Antinomy. 25  One might argue that for an empiricist like Hume the very concept of transcendental freedom is impossible because it has no corresponding sense-­impression. But Kant’s Hume denies that the concept of causality, both in its natural and its non-­natural (free) sense, has objective validity (B19–20, A765–6/B793–4; KpV, 5:51). By contrast, the dogmatic empiricist-­ naturalist who defends the Antinomy-­antitheses believes in natural causal necessity. 26 Naturalists can even allow that we use the phenomena/noumena distinction to form this “thought”: they can conceive of the idea that a non-­temporal thing in itself underlies empirically

Freedom, Idealism, and Standpoints  23 than metaphysics” and is positing only “a transcendental concept of cause rather than a metaphysically distinct (noumenal) cause,”27 then his views on freedom are compatible with a starkly naturalistic ontology. Hence, a naturalistically inclined transcendental realist can accept the deflationary two-­standpoint doctrine of freedom while treating Kant’s idealism as no more than a curious but logically coherent fiction. One might suggest that Allison’s view should be more charitably understood as follows: in Kant’s idealism the spatiotemporal conditions that are essential to our understanding of nature are only epistemic conditions of knowledge rather than (as naturalists assume) absolute ontological conditions; thus, even though we cannot know that free (non-­spatiotemporal) causes exist idealism shows that such causes might exist. This creates “conceptual space” for a belief in causally undetermined powers of agency that comforts our incompatibilist worries, that is unavailable to naturalists, and that we can rationally adopt by taking up the practical standpoint from which our beliefs are not subject to the demanding evidential standards of knowledge.28 This gives us a legitimate practical faith (Glauben, as a mode of Fürwahrhalten) that we are free agents. This proposal may have its merits. (I discuss it in Chapter VIII.) But it does not yield a metaphysically deflationary interpretation. An attitude of faith towards the proposition that we have noumenal causal powers is a theoretical proposition that posits the existence of such powers (KpV, 5:133–5)—the practical sources of such faith notwithstanding. Indeed, the very idea that the existence of freedom is only a matter of faith attests to the metaphysical import of such faith: the reason why we are confined to practical faith here is that we cannot obtain certain theoretical knowledge of whether a noumenal causality of freedom exists in non-­sensible reality. There is yet another version of the two-­standpoint reading. On this version, we can believe that we are free because this belief, just like the belief in determinism, concerns only a standpoint-­relative fact. Whether we are free or determined depends upon our particular viewpoint or context of assertion: “Human beings are free insofar as they are thought of from a practical . . . standpoint, and determined

determined appearances, even if they believe that things in themselves collapse into empirically determined appearances. Bojanowski 2006: 136 also overlooks this point. 27  Allison 2020: 280. 28  Allison 1990: 44: “. . . by treating space, time, and the categories as epistemic rather than ontological conditions, transcendental idealism . . . opens up a ‘conceptual space’ for the nonempirical thought (although not knowledge) of objects . . . as they may be apart from these conditions.” The question is, however, what this “nonempirical thought” amounts to and whether it has any assertive force. Elsewhere (2011: 304), Allison explicitly denies that this “thought” has the status of belief or faith. When he says that Kant’s idealism “is required . . . to think without contradiction the applicability of both the causality of nature and of freedom” to human actions (2020: 263–4), he does not clarify whether this “thinking” involve any type of doxastic attitude (and, if so, which type).

24  Kant on Freedom and Rational Agency insofar as they are thought of from a scientific . . . standpoint.”29 “There is no [standpoint-­independent] fact of the matter regarding freedom.”30 There are two familiar difficulties with this proposal. First, the claim that we are free or determined simply insofar as we are thought from different standpoints is hard to understand. How could the mere fact that we consider ourselves from different perspectives make it the case that we are free or determined?31 Second, for Kant claims about our noumenal freedom are less subjective or relative than claims about our phenomenal character as causally determined beings: the latter claims describe “merely” an appearance and they depend upon our subjective, species-­ relative forms of sensibility whereas the former claims concern our “proper self ” and do not depend upon particular forms of sensibility (GMS, 4:451, 457, 461). Hence, Kant can hardly accept that both claims have only a relative truth-­value that depends on which subjective perspective we happen to adopt.32 The relativist approach faces a further difficulty. For Kant, practical reason supports judgments about supersensible beings that yield only subjectively rational faith rather than objectively certain theoretical knowledge (A829/B857; KpV, 5:122, 125–6). This demand of epistemic modesty for practical assertions about supersensibles would be unnecessary and unmotivated unless these assertions concerned the same (standpoint-­independent) proposition that theoretical reason also seeks but fails to prove as certain knowledge. Likewise, theoretical reason makes a “problematic” use of transcendental ideas (such as freedom), which benefits from (i.e., is rendered unproblematic by) the objective reality that practical reason confers on transcendental ideas (KpV, 5:49–50). Practical beliefs about supersensibles could not have this positive implication for theoretical claims about supersensibles if the truth (or objective reality) of such beliefs were strictly relative to the practical standpoint from which we adopt them. Even if we assume for the sake of argument that the truth of claims about freedom and determinism is purely relative to our perspectives or contexts of assertion, it is unclear why this should qualify as an anti-­metaphysical proposal. The proposal seems to suggest a highly controversial metaphysical orientation,

29  Frierson 2011: 83. 30  Allison 2013: 298. See also Allison 2004: 48–9 and Korsgaard 1996b: 124–5. 31  See Van Cleve 1999: 8. A journal referee once claimed that we can demystify the proposal by invoking familiar cases of relative truths: A can be called tall in the context of discussing C1 (the class of teachers) and not-­tall in the context of discussing C2 (the class of basketball players). But the assimilation of “free” and “determined” to vague, context-­sensitive predicates like “tall” is implausible. Moreover, even for such predicates the judgment that A is F is not simply made true by the fact that A is thought in relation to some comparison class C1. Rather, the truth-­conditions of such judgments partly involve a reference to properties that A has quite apart from being thought in relation to C1 or C2, e.g., (for “tall”) properties such as measurable height. 32 See Watkins 2004: 322–8. Frierson’s response to this worry (2011: 99–102) gives up on the standpoint-­relativity of the claims about freedom and determinism: if, as Frierson argues, the scientific standpoint itself must concede that we are free agents, then we are free not just “insofar as we are thought” from the practical standpoint.

Freedom, Idealism, and Standpoints  25 namely, an anti-­realist or relativist ontology.33 This is very different from the more common deflationary idea that Kantian idealism abjures ontological claims in favor of examining the epistemic conditions of human cognition: the claim that we are free or determined relative to whether or not we are considered in ac­cord­ ance with conditions of human cognition is much stronger than the modest claim that we cannot know anything about things apart from conditions of human cognition.34 One might reply that the appeal to standpoint-­relative facts is in the spirit of a metaphysically deflationary reading because it denies that there is an ontologically privileged perspective that concerns “really real,” absolute facts.35 But the claim, “The noumenal fact that we are free is ontologically on par with the empirical fact that we are not free” is no less metaphysically oriented and surely no less controversial than the opposite claim that one of these facts is ontologically privileged. The former claim might highlight an anti-­realist, relativist alternative to metaphysical realism, but it does not provide an alternative to metaphysics or ontology per se. Hence, it does not provide a metaphysically deflationary account. I conclude that deflationary two-­ standpoint readings face a dilemma. According to straightforwardly deflationary proposals, Kant’s account requires no ontological beliefs concerning the actual or really possible existence of supersensible freedom. But such proposals violate central interpretive desiderata ((D2), (D6)). If one concedes that Kant’s doctrine tries to legitimize beliefs in the actual or really possible existence of noumenal causal powers, then the deflationary veil evaporates and a metaphysically weighty interpretation of Kant’s doctrine of freedom becomes unavoidable.

I.3.  How Transcendental Idealism “Saves” Human Freedom As we saw, one crucial constraint on interpreting Kant’s doctrine stems from his claim that in order to “save” free will we need to accept his idealism (D6). In this section I consider how we should understand this claim. I have argued that a reading on which Kant’s idealism saves only the logical possibility or conceivability of freedom is too weak because a naturalist who rejects Kant’s idealism can gladly concede that “transcendental freedom” is, unlike

33  Allison at one point invokes Dummett’s anti-­realism as a source of inspiration (2004: 48). 34  The tension here is between (1) the modest idea that we have no warrant for holding that things are spatiotemporal and causally necessitated when we consider them independently of the conditions of our cognition (Allison 2004: 32) and (2) the ambitious idea that, relative to our practical point of view, we are warranted in affirming that we are not subject to conditions of spatiotemporality and natural causality. 35  See Allison 2004: 48–9, 120–1; 2013: 296–8.

26  Kant on Freedom and Rational Agency “non-­extended body,” a coherent (if empty) concept.36 Admittedly, some of Kant’s statements do suggest this reading: he says that his idealism ensures that “the representation of [freedom] is at least not self-­contradictory” (Bxxviii). However, this statement follows upon a more elaborate remark: if we could not make the idealist distinction “between things as objects of experience and those same things as things in themselves,” then “all things in general . . . would be determined by . . . the mechanism of nature” and in that case “I could not . . . without palpable contradiction, say of one and the same being, for instance the human soul, that its will is free and yet is subject to natural necessity” (Bxxvii). Hence, the “self-­ contradiction” that Kant seeks to avoid via his idealism does not concern a relation between the constituent marks of the concept “transcendental freedom.” Rather, it concerns a relation between the two assertoric propositions which predicate (on the one hand) “transcendentally free” and (on the other hand) “determined by the mechanism of nature” of the same subject.37 We can put this point by saying that Kant’s idealism creates logical space for freedom—­if by that we mean that idealism allows us to logically combine the two assertoric beliefs that we are free and that we are causally determined. Since each of these beliefs asserts the existence of a distinctive type of causality, we can also say that Kant’s idealism creates ontological space for the possible existence of the causality of freedom alongside the known (phenomenal) existence of the causality of nature. Idealism thereby enables an affirmative answer to the ontological question of “whether freedom . . . can exist along with the universality of the natural law of causality” (A536/B564). There are two different ways in which Kant’s idealism might save the belief in free will. Idealism might function as a necessary condition for affirming our freedom: it makes room for the belief in freedom by removing the threat of a contradiction between this belief and the belief in natural determinism, but it does not establish that freedom exists. On a more ambitious interpretation, Kant’s idealism

36  For Proops (2021: 279–80, 313), Kant’s resolution to the Third Antinomy shows not even the logical possibility of freedom because the concept of freedom might contain a “hidden contradiction.” However, such a theoretically possible contradiction would go beyond a mere logical conflict: it would involve a real repugnance among predicates that are not analytically contained in the idea of freedom (for the notion of real repugnance, see NG, 2:172; Chignell 2012: 643–7). The logical possibility of freedom requires only that the definition of freedom, as a spontaneous capacity to act on one’s own without being determined by a prior cause (A533/B561), be free of a logical contradiction. As I noted in Section I.2, this raises no problem: even the antithesis-­naturalist grants that “freedom” is a coherent, if empty, “thought-­entity.” In the second Critique, Kant says (looking back to the first Critique) that theoretical reason was able to establish the concept of freedom “problematically, as not impossible to think” (KpV, 5:3). This means, minimally, that theoretical reason showed the logical possibility of freedom. (As I show below, Kant intends a stronger claim here.) 37  Likewise, when Kant explains how his idealism solves the worry that “the thought of freedom contradicts . . . itself,” he clarifies that the threat of contradiction arises when a subject “calls itself free” and “assumes itself to be subject to the law of nature” (GMS, 4:498). Again, the worry here (and in similar passages, such as Prol, 4:344) concerns a clash between two judgments or assumptions that ascribe apparently contradictory properties to one subject.

Freedom, Idealism, and Standpoints  27 is sufficient for legitimizing the belief in freedom because according to transcendental idealism we exist as non-­spatiotemporal things in themselves which are not subject to “the mechanism of nature” and thereby qualify as free noumenal agents. The ambitious reading conflicts with how Kant presents his solution to the Third Antinomy (which yields his official account of how idealism contributes towards saving our freedom). As we saw, idealism allows us to assert that the causality of freedom “can take place” alongside the causality of nature. This is weaker than the assertion that the causality of freedom does take place. Accordingly, Kant concludes his discussion of the Third Antinomy with the reminder that he did not mean to show that freedom is actual. But even the weaker statement that freedom “can take place” may seem too strong in light of Kant’s disclaimer that “it has not even been our intention to prove the possibility of freedom” (A558/B586). The operative notion of possibility here is “real possibility,” which Kant equates with the “objective reality” or “objective validity” of a concept (Bxxvii; A223/ B270). The real possibility of a concept requires that there be either empirical or a priori grounds for supposing that the concept can have actual instances (Bxxvii; A221–3/B268–70). Kant’s idealism does not supply any positive reasons for supposing that the idea of freedom can have actual instances. There is nevertheless a sense in which his idealism allows us to assert that freedom “can” exist alongside the causality of nature: idealism shows that the possible existence of freedom is not precluded by the proven existence of natural causality. It thereby removes a metaphysical obstacle to the real possibility of freedom: this possibility would be conclusively negated if (as naturalists assume) the causality of nature placed an absolute constraint on which causal powers really can exist. But although transcendental idealism thus removes a metaphysical hindrance to the real possibility of freedom, it does not thereby entitle us to assume that freedom is really possible: “in order to assume something it is not enough that there is no positive hindrance to doing so” (A673/B701).38 Proponents of the ambitious reading might suggest that when Kant claims that he has not shown the actuality or real possibility of freedom, he is merely denying the theoretical explicability of freedom.39 However, this suggestion is textually implausible in light of Kant’s much stronger disclaimers. After stressing that “our intention has not been to establish the actuality of freedom,” he adds that “the only thing that” that his idealist solution to the Third Antinomy “was able to achieve” is the proof that the deterministic causality of nature “at least” does not rule out the (really possible) existence of freedom (A558/B586). These remarks 38  Allison argues (2020: 260) that a metaphysical reading would require Kant to solve the Third Antinomy by proving the real possibility of freedom. But one can hold, instead, that Kant’s solution carries metaphysical weight insofar as it removes a metaphysical obstacle to the real possibility of freedom, without thereby proving this possibility. 39  See Ameriks 2003: 166–7; Hogan 2009a: 379.

28  Kant on Freedom and Rational Agency would be disingenuous, indeed plainly false, if Kant took his idealism to yield a sufficient basis for establishing either the actual existence or the real possibility of freedom. Proponents of the ambitious reading might insist that in the course of proposing his idealist solution to the Third Antinomy, Kant commits himself to the actuality or (at least) the real possibility of freedom. They might invoke two passages here. First, Kant says that we can cognize that the intelligible cause of our observable actions “could be free” (“. . . daß sie frei . . . sein könne”; A557/B585). However, in the same paragraph he clarifies again that his only goal has been to show that freedom and natural necessity “could take place” alongside each other (“stattfinden können”). Hence, our theoretical cognition that the cause of our actions “could” be free reduces to the cognition that the real possibility of free causes is not precluded by the (known) existence of natural necessity. As we saw, this does not establish that free causes are indeed really possible. Second, at A546–7/B574–5 Kant says that we can cognize ourselves through pure apperception as intelligible objects whose spontaneous actions are empirically unconditioned. But here Kant refers to the spontaneity of thought that characterizes our theoretical faculties of “understanding and reason.” The fact that we are spontaneous thinkers does not entail that we also have the causality of a free will which we ascribe to ourselves only once we move beyond our purely theoretical self-­awareness and consider the practical imperatives that govern our productive efforts (A547/B575). (I give a detailed account of this distinction in Part 3 and 4.) The ambitious reading faces a more general problem. Transcendental idealism establishes only that non-­spatiotemporal things in themselves are not subject to the deterministic principle of natural causality. This is a purely negative statement which does not entail any positive claims about what powers things in themselves do (or really can) possess (B349, B307–10, A358–9). To see this gap, consider the following points. First, idealism does not prove that all things in themselves as such must possess causal powers. (Idealism might require that there be some things in themselves that ground appearances, but even if this grounding has a causal sense the set of grounding things might be very small or contain only a single member.) Second, assuming that our non-­sensible character does include causal powers, the fact that these powers are undetermined by natural causes does show that they are free powers: they might be determined by non-­natural causes such as a divine agency (KpV, 5:99).40 Third, in Kant’s idealism the

40  For Hogan (2009a: 370–2, 379–80), the idealist claim “Things in themselves lack a determining ground through which they could be discursively cognized” entails that things in themselves are free. But things in themselves might have a determining ground through which they could be non-­ discursively cognized by a divine intellect.

Freedom, Idealism, and Standpoints  29 appearances of stones and donkeys correspond to non-­natural things in themselves, but this does not entail that the things in themselves underlying stones and donkeys have any spontaneous capacities (A547/B575).41 I conclude that transcendental idealism is a necessary but insufficient condition for (justifiably) asserting that free will is actual or at least really possible. By restricting the deterministic principle of natural causality to the spatiotemporal world of sense, and by holding that there are things in themselves in addition to spatiotemporal appearances, Kant’s idealism makes ontological space for the real possibility of a free causality. But it provides no positive reasons for asserting that this space is filled, i.e., that we actually are or really can be free agents.

I.4.  A Framework for Interpreting Kant’s Idealist Doctrine of Freedom I shall now introduce a framework for interpreting Kant’s idealist doctrine of freedom. This framework satisfies (D1)–(D6) and thereby provides a general account (awaiting further clarification and specification) that unifies the six central commitments of Kant’s doctrine. Every event E1 that we can experience at some determinate time T1 is the effect of a natural cause C1 whose activity A1 makes the occurrence of E1 at T1 (hypothetically) necessary. A1 is itself a natural event E2 that begins to occur at some determinate time T2 and is hence the effect of a natural cause C2 whose activity A2 makes the occurrence of A1 at T2 necessary; and so on. In this “steady chain of nature,” no free causality exists because no cause can start to act spontaneously (“on its own”) without being determined to act by some prior causal condition. The natural-­hypothetical necessitation of events is inseparable from their temporal character because the principle that every event is causally necessitated is valid only as a general law of time-­determination (cf. Section  I.1). In Kant’s idealism time itself is (metaphysically speaking) nothing but a form of human sensibility (A35/B51), i.e., (roughly) a certain way in which the human mind orders empirical sensations. Hence, temporal determinations and the conditions for temporal determination depend on the cognitive faculties of the (generic) human subject. Since the “law of natural necessity” (KpV, 5:95) is a

41  Hence, Kant can escape Beck’s charge (1960: 190) of panlibertarianism (cf. Willaschek 1992: 293). By contrast, on a reading such as Hogan’s, which seems to equate the absence of determining spatiotemporal grounds with the presence of free will, it is difficult to see how Kant can avoid this charge. This problem is exacerbated by Hogan’s further suggestion (2009b) that noumenal affection is a relation between things in themselves that are free in a libertarian sense. For Kant noumenal affection is a completely general relation between finite minds and existent things in themselves (B72), including those that underlie the appearances of stones or donkeys.

30  Kant on Freedom and Rational Agency condition for temporal determination, it applies only to the spatiotemporal constitution that things have as objects of human cognition (appearances or phenomena), insofar as they “conform to” our cognitive faculties (Bxvi). The things that appear to us in the sensible-­spatiotemporal form imposed by our cognitive faculties must exist in­de­pend­ently of our mind, since our finite mind cannot create the things it cognizes (B72). Hence, these things must also have some character (some way of being) which differs from their mind-­dependent character as spatiotemporal appearances: namely, a non-­sensible or noumenal constitution that they possess in and by themselves, apart from their relation to our cognitive faculties. Among the characteristics included in this noumenal constitution there might be the spontaneous causality of freedom. At least this possibility is not ruled out by the phenomenal “law of natural necessity.” We can consistently attribute spatiotemporal actions to two different types of causes. We must regard such actions as effects of natural causes, namely, as products of the empirically conditioned causal capacities that acting subjects have as part of their natural-­spatiotemporal character. But we can also regard these actions as effects of non-­natural powers whose exercise is not determined by foreign causes—­these would be the free powers that some subjects might have as part of their non-­natural noumenal character. We can attribute spatiotemporal actions to two distinct types of causal powers because we can consider the acting subject in two ontologically distinct ways: first, as a spatiotemporal appearance whose actions result from the ontologically relative, mind-­dependent causality of nature; second, as a thing in itself whose actions might derive from the non-­ natural, noumenal causality of freedom. This account respects (D1): it acknowledges that observable actions have an objective temporal position and (therefore) are determined by natural causes whose exercise is itself causally determined, while also stressing that the deterministic character of natural causation does not have a metaphysically absolute, mind-­independent status. Because determinism (“the law of natural necessity”) is not an unrestricted metaphysical truth, there is ontological space for the ex­ist­ence of a non-­deterministic atemporal form of causality.42 For the same reason my account satisfies (D3): Kant’s claim that “freedom and nature, in the full sense of these terms, can exist together, without any conflict, in the same 42  Here my reading must be distinguished from interpretations that try to make room for freedom by supposing that Kant treats the deterministic law of natural necessity as a merely regulative principle. On such readings we are licensed only to represent events “as if ” they are necessitated by preceding causes (Allison 2013: 298; Beck 1960: 192–3; Bojanowski 2015: 98–100). However (as Beck admits), for Kant the law of necessity has a much stronger status than merely regulative ideas of reason: all natural phenomena as such must be causally determined and we have certain knowledge that they are so determined (A664/B692; cf. A542/B570). My interpretation respects this point: the property of being causally necessitated is constitutive of the character that sensible objects possess due to their conformity to our cognitive faculties. But objects also possess a different character that does not conform to our cognitive faculties and that might comprise a free causality.

Freedom, Idealism, and Standpoints  31 actions” (A541/B569) expresses the idea that the same action can be attributed both to a causally determined and to a free cause. It can be regarded as an effect of two distinct forms of causality which pertain to two distinct ontological characters of the acting subject. Since attributing observable actions to empirically unconditioned causes presupposes Kant’s idealist claim that subjects have a mind-­independent ontological constitution which differs from their mind-­ dependent constitution as empirically conditioned appearances, my reading also shows why adopting Kant’s idealism is necessary for adopting the belief that we are free (D6). Since on my reading Kant seeks to “uphold” the assertoric belief that our observable actions result from an unconditioned causality, my account requires no implausible deflation of Kant’s claim that we can regard ourselves as free (D5) in a squarely incompatibilist sense (D2). However, on my reading Kant’s idealism yields only a necessary, insufficient condition for adopting the assertoric belief that we are free in our noumenal character. Hence, my reading (thus far) leaves entirely open whether are rationally justified in adopting this belief and, if so, whether this justification grounds certain knowledge or merely faith (Glauben) that we are free. Finally, my reading also provides a systematic basis for (D4): since the capacity for free agency could only exist as part of a non-­sensible constitution that does not conform to the sensible conditions of human theoretical cognition, this capacity cannot be an object of human theoretical cognition. Our inability to comprehend or explain free agency is due to the (“subjective”) limitations of our finite cognitive faculties that delimit the boundaries of what we can theoretically comprehend. Anything that does not conform to the conditions imposed by these faculties—­anything that we cannot experience in space and time—­is unintelligible for our theoretical reason. But Kant’s epistemic claim that we cannot comprehend, understand, or explain the causality of freedom does not support the deflationary denial that he conceives freedom in straightforwardly ontological terms, as a causal power that exists (or might exist) outside of nature and time.43 My interpretive framework rests upon a metaphysical interpretation of Kant’s idealism. I read Kant as holding that features such as time and natural causality or necessity depend upon the (generic) human mind in an ontological sense: these features do not exist apart from the relation that things bear to human representational faculties. Moreover, I read Kant as holding that the beings which appear to us in spatiotemporal form also possess non-­sensible characteristics. Deflationary readings of Kant’s idealism deny these assumptions.44 There

43  Allison’s claim that because free acts are unexplainable Kant’s appeal to the independence of free acts from temporal conditions “is to be taken epistemically rather than ontologically” (2020: 480) is a non sequitur. 44  For paradigmatic instances of such readings, see Allison 1984, 2004; Bird 1962, 2006; Prauss 1974.

32  Kant on Freedom and Rational Agency is strong textual evidence that directly tells against such readings (see, e.g., A42/B59; B69; B308–9; A491–2/B520; B308–9; Prol, 4:315).45 But my argument in this chapter also entails a systematic argument against deflationary readings of Kant’s idealism, which goes as follows. (1) Since one of Kant’s key motives for proposing his idealism is his conviction that idealism is needed to save human freedom (D6),46 an interpretation of Kant’s idealism is adequate only if it can explain how Kant’s idealism contributes to saving human freedom. (2) As we saw in Sections I.2–3, Kant’s idealism saves human freedom by making ontological space for the existence of an unconditioned, atemporal form of causality. (3) Kant’s idealism makes ontological space for the existence of a free causality only if his idealism shows that the things which appear to us in spatiotemporal form have a constitution in and by themselves which differs from their mind-­dependent phenomenal features such as being temporally-­empirically conditioned. (4) Since deflationary readings deny that Kant’s idealism has this ontological implication, they cannot (via (2)+(3)) explain how his idealism saves human freedom (these readings violate (D6)). Thus (via (1)), these readings do not yield adequate interpretations of Kant’s idealism. Proponents of deflationary readings might respond that metaphysical readings are committed to an ontological dualism between appearances and things in themselves as numerically different objects which is a complete non-­starter for charitably interpreting Kant’s views on freedom. When deliberating “under the idea of freedom” I consider how I, not some other entity, should act in the spatiotemporal world of sense. Moreover, we hold agents morally accountable for spatiotemporal actions like thefts and lies: if the phenomenal subject that acts in space and time is unfree and numerically different from some free noumenal being, our moral imputations are misdirected and indefensible.47

45  In these passages, Kant explicitly denies that things in themselves are spatiotemporal. The modest idea that we are not epistemically warranted in attributing spatiotemporality to things in themselves (Allison 2004: 121) seems too weak for this denial. For criticisms of deflationary readings of Kant’s idealism along such lines, see Allais 2015: 77–97; Ameriks 2003: 73–9, 103–5; Guyer 1987; Langton 1998. 46  I do not see this as Kant’s only motive for idealism. Kant thinks that he can “directly” prove his idealism prior to considering the issue of freedom (in the Dialectic), namely, through the arguments he offers in the Aesthetic (cf. A506/B534). Hogan (2009a: 381–2; 2010: 36) holds that even Kant’s arguments for the non-­spatiotemporality of things in themselves in the Aesthetic are driven by his attempt to save freedom qua absolutely contingent agency. However, in the Aesthetic Kant argues for the non-­spatiotemporality of all things in themselves whereas (as we saw) he does not assert that all things in themselves as such exhibit absolute contingency. 47  See Beck 1987 and Bennett 1974, 1984. One might hold that we can blame the noumenon for phenomenal actions since it is the ultimate cause of those actions. But if the seat of the noumenal cause differs from the phenomenal being that performs these actions, then it is unclear why the noumenal cause is only responsible for these actions rather than for the spatiotemporal actions performed by a variety of different phenomenal beings. Moreover, our judgments of responsibility, reactive sentiments and physical acts of punishment are clearly directed towards phenomenal agents. This is incompatible with the idea that a phenomenal agent is caused to act by some numerically different (other-­worldly) entity. In our ordinary practice, the fact that A is caused to commit a crime by another being is a sufficient reason for excusing A.

Freedom, Idealism, and Standpoints  33 I agree that these are valid complaints against metaphysical two-­object readings. However, a metaphysical reading of Kant’s idealism is not committed to an ontological dualism between numerically different entities. The ontological contrast that follows from a metaphysical reading might instead concern a distinction between two types of properties (characters or constitutions; Kant frequently uses the term Beschaffenheiten) of one entity. This would yield a metaphysical dual-­aspect reading which allows that appearances and things in themselves can be numerically identical.48 Let me sketch why such a reading is a plausible interpretive option. One typical route towards a two-­object reading proceeds via the assumption that appearances are mere representations, i.e., mental states that must be numerically different from extra-­mental things in themselves.49 This assumption can be resisted as follows. Kant repeatedly stresses the following point: that which appears to us in spatiotemporal form has a further constitution that is in­de­pend­ ent of the character of our sensibility (A30/B45; Prol, 4:314–15; GMS, 4:451). When we represent objects something is “given to us,” i.e., affects our senses: this thing is “an unknown something.” We turn this unknown something into an object of our cognition by representing it in accordance with (thereby making it “conform to”) the character of our cognitive faculties such as our spatiotemporal forms of sensible intuition. In this manner we become acquainted with the given thing, though only as it appears, not as it is in itself (apart from its conformity to our cognitive faculties). Thus, in Kant’s view the very things that affect our senses acquire a spatiotemporal constitution only via their relation to our mode of sensibility; moreover, the same things that appear to us in spatiotemporal form exist and have some non-­spatiotemporal constitution independently of being related (“given”) to our mind (A28/B44; A35–6/B52). These claims are false with regards to our mental states: our representations do not affect our (outer) senses; they do not possess their temporal characteristics only in relation to our mind; and they neither exist apart from our mind nor do they possess a mind-­independent constitution. Hence, the things that appear to us as spatiotemporal phenomena cannot be identified with representations or mental states.50 Rather, our mental states 48  Allison concedes that “the notion of a single entity with two aspects can easily be considered as metaphysical” (2020: 343). Different metaphysical dual-­aspect readings have been proposed by Allais 2015; Langton 1998; Warren 2001; unpublished. 49  See Van Cleve 1999 and Guyer 1987: 281, 291, 334–5. For Guyer (1987: 335), Kant posits an “ontology” that “includes two classes of objects, namely things like tables and chairs and our representations of them.” 50  When Kant calls appearances mere representations, he uses the notion of representation in a transcendental rather than an empirical sense (see A372–3). Only the empirical sense of “representation” designates a mental state (see, e.g., A30/B45). Likewise, appearances are transcendentally speaking “in us” because objects (that exist and that have a mind-­independent character “outside us”) appear to us only by means of our inner states of consciousness and are thereby made to conform to the character of our representational faculties (see Longuenesse 1998: 20–1). Emphasizing the difference between the empirical and the transcendental meaning of terms like “representation” or “in us” is part of a strategy for resisting phenomenalist readings of Kant’s idealism (such as the one proposed by Van Cleve 1999). For powerful critiques of such readings, see Allais 2015; Allison 2004; Collins 1999.

34  Kant on Freedom and Rational Agency are representations of the spatiotemporal constitution of beings that also possess some further (unknown) constitution in and by themselves. Kant states his view as follows: “In order that something can appear outside us, there must be really something outside us, though not constituted in the way we have the representation of it, since other kinds of sense could afford other ways of representing the same thing” (emphasis mine).51 When something appears to us as external to our mental states, we represent this externality in a sensible mode that is peculiar to our cognitive constitution (and that might differ for other types of finite cognizers: cf. A26/B42; GMS, 4:451). The sensible mode in which we represent things as external to our representations is the form of space. But things can appear to us as spatially (empirically) external or “outside us” only if these same things are also “outside us” in another, non-­spatial (transcendental) sense: namely, if they exist and are constituted in a way that does not depend upon our spatiotemporal mode of representation. This yields a squarely metaphysical response to (one prominent version of) the two-­object reading. The distinction between appearances and things in themselves picks out two different sets of properties of one object: on the one hand, the properties that an object possesses independently of its relation to the (generic) cognitive human subject; on the other hand, the properties that this same object has (only) in relation to our cognitive faculties, insofar as it can be “given” to and represented by our mind: “in the relation of the given object to the subject, such properties [we ascribe to an appearance] depend upon the mode of intuition of the subject, [and hence] the object as appearance is to be distinguished from itself as object in itself ” (B69). Critics of the dual-­aspect (one-­object) reading might argue that it violates Kant’s commitment to noumenal ignorance: if we can know that every individual appearance is identical to an individual thing in itself, then we have the theoretical knowledge required for individuating and counting things in themselves.52 In response, one may distinguish between a weaker and a stronger version of the one-­object reading. On the weaker version, anything that we cognize as an individual appearance has a corresponding non-­sensible reality but this reality might (for all we can know) comprise a singular noumenon or a plurality of noumena.53 On the stronger version, individual spatiotemporal phenomena are numerically identical to corresponding individual noumena. While the worry concerning 51  This note (Refl., 18: 612–13) is viewed as central by Guyer (1987: 324–5). He claims that it expresses a post-­1787 turn to realism which is inconsistent with Kant earlier (1781) identification of appearances with representations. But I do not agree that Kant’s idealism ever equated spatial appearances with representations in an unqualified sense (see n. 50). 52  See Marshall 2013; Stang 2013; Walker 2010. 53  If I understand them correctly, Allais 2015 and Marshall 2013 endorse this reading.

Freedom, Idealism, and Standpoints  35 noumenal ignorance has no force against the weaker claim, it does affect the stronger claim which seems required for Kant’s account of free agency. The best interpretive option here is to adopt the stronger version of the one-­ object reading in a tightly restricted form. We have no general theoretical basis for claiming that every particular object that we individuate through spatiotemporal criteria corresponds to one particular thing in itself (A379–80; B427–8). Thus, we can neither rule out nor establish logically coherent but otherwise (cognitively speaking) empty possibilities (that we may raise in mere “plays” with our pure concepts; A239/B299) such as the following: one appearance (e.g., my office chair) has its basis in a plurality of things in themselves that are jointly given to our sensibility; or, many different appearances (all chairs on campus) are modes of one noumenon that affects our senses. It is only in the special case of rational agents that we can invoke objective criteria to identify things in themselves and appearances. We can use a valid principle of pure reason, the objective moral law, to classify spatiotemporal actions as permissible, required, or prohibited. On this basis we can regard the agent who performs or omits these acts as a proper subject of moral imputation. Since the true seat of moral imputation is the agent’s free noumenal character (their moral personality), the phenomenal beings that we distinguish from another via their morally assessable observable actions must (if these actions are fairly imputable to agents, as the objective moral law requires) also differ from one another qua free noumenal agents.54 The phenomenal being that performs a theft and the free noumenal being to whom this theft is imputed must be one and the same (KpV, 5:95–6; cf. GMS, 4:456–7). We lack this normative-­practical (i.e., moral) basis for identifying particular things in themselves and appearances whenever we cannot correlate the observable behavior of appearances with a noumenal seat of rationality and imputability, e.g., in the case of animals and inanimate matter (A546/B575). Hence, my interpretive framework respects Kant’s commitment to our theoretical ignorance of noumena: we have no general theoretical basis for identifying every particular appearance with a particular thing in itself; to make such an identification we must invoke normatively grounded criteria that apply only to a special class of beings.55 Accordingly, in the majority of passages where Kant clearly affirms the 54  For a similar approach to the issue of how appearances can be identified with things in themselves, see Schafer 2018. See B421–6 and A383–4 for the idea that theoretically uncertain metaphysical principles (such as the principle of the identity of a person with respect to different phenomenal states and actions) can be defended on practical grounds. The claim that “there is no viewpoint from which we can identify something as an empirical object and the same thing as a thing in itself ” (Allais 2010: 17) overlooks Kant’s appeal to the criteria of identity provided by the rules of assessment and imputation that we use from a normative viewpoint (see Section I.5). 55  Restricted one-­object readings have been suggested by Adams 1997; Ameriks 2003: 76–7, 83; Aquila 1979.

36  Kant on Freedom and Rational Agency stronger version of the one-­object view, he is concerned only with human beings as moral agents (Bxxvi; A537/B565; A538–9/B566–7; A540–1/B568–9; A 545/ B573; A546–7/B574–5; GMS, 4:456; KpV, 5:95–6).56 Another frequent worry about one-­object readings draws on the “Indiscernibility of Identicals” principle: if A and B are identical and if A has some property P, then B also has P. One-­object readings seem to violate this principle because they hold that an appearance can be identical to a thing in itself even though the appearance is spatiotemporal, empirically determined, etc. whereas the thing in itself is not.57 Some defenders of a one-­object view respond by retreating to the view that appearances and things in themselves stand in a relation of sameness that is weaker than strict numerical identity.58 But Kant’s doctrine of freedom requires and frequently asserts the strict numerical identity of phenomenal and noumenal human self. One-­object readings can satisfy the Indiscernibility Principle by stressing that spatiotemporal appearance-­properties have a relational character: they depend upon the relation between things and a certain (type of) cognizing subject. A being which has a strong smell in relation to creatures with discerning olfactory sense can be numerically identical to a being which is odorless in relation to creatures with impoverished olfactory sense. By (imperfect) analogy, a being which has the property of being causally determined/spatiotemporal in relation to the (generic) subject of human cognition can be numerically identical to a being which lacks this property apart from said relation.59 As Kant says, “in the relation of the given object to the subject, such [appearance] properties depend upon the mode of intuition of the subject,” and hence “the object as appearance is to be distinguished from itself as object in itself ” (B69). Spatial predicates like “extended” cannot be ascribed to things in themselves, i.e., to things insofar as they are considered “without relation to” (human) “outer senses” (A358): what things are in themselves concerns what they are “without all relation to senses” or sensible intuition (A380). The claim that a being (or action) A is P only in its relation to a specific type of finite cognition which involves a certain “mode of ” sensibility is consistent with the claim that A is not-­P in and by itself, apart from its relation to this (or any other) finite-­sensible mode of cognition.60 Notably, 56  Watkins 2004: 318 cites an unpublished note where Kant says that freedom and causal necessity relate to different objects. But surely unpublished Reflexionen do not carry the same interpretive weight as published writings, especially if these reflections conflict so sharply with so many central published texts. 57  See Bojanowski 2006: 141; Marshall 2013; Pogge 1991; Stang: 2013; Watkins 2004: 325. 58  This seems to be the strategy of Marshall (2013: 539–40). He construes appearances and things in themselves as “qua-­objects” “formed from the same original object” (540) which they “share.” 59  Kant suggests an analogy between secondary qualities like smell and appearance-­properties at Prol, 4: 288–94. It is unclear how exactly we should interpret this analogy and how we should construe its imperfections/limitations; I cannot (and need not) address this issue here. For helpful discussion, see Allais 2015, ch. 6. 60  See Allais 2015: 73 for what I take to be a similar point.

Freedom, Idealism, and Standpoints  37 Kant often invokes the relational status of appearance-­properties when he stresses that the same being can be causally determined and free: While “it would . . . be impossible to escape [the] contradiction [between freedom and determinism] if the subject . . . conceived itself in the same . . . relation when it calls itself free as when in respect of the same action it assumes itself to be subject to the law of nature” (GMS, 4:456), his idealism ensures that freedom and natural necessity might “both, each in a different relation” take place in the same occurrence (A536/B564). The claim that the relational character of sensible appearance-­properties is central to Kant’s idealism does not require a specific ontological view about what types of non-­sensible properties belong to things in themselves. Some commentators claim that things in themselves (1) must have non-­relational properties or even (2) can only have non-­relational properties.61 The interpretive framework that I have proposed is committed to neither (1) nor (2).62 Kant’s views on appearances and things in themselves raise many further complications, puzzles, and problems. I do not purport to have given a remotely comprehensive treatment of these issues here. My sole aim has been to show that a suitably qualified metaphysical one-­object reading has three important virtues: it is textually plausible, it yields a coherent doctrine, and it uniquely fits the bill presented by Kant’s doctrine of free agency.63

I.5.  The Empirical and the Normative Standpoint I now consider how the abstract philosophical reconciliation of freedom and natural necessity that I sketched in the preceding section bears upon the ordinary beliefs that human beings hold about themselves outside the context of subtle philosophical reasoning. One might suggest that Kant’s reconciliation project allows that the concepts of freedom and natural necessity are conjoined in our ordinary beliefs: we can consider ourselves as free and as causally determined at the same time, both when we decide how to act and when we try to explain or predict our actions. If Kant’s view had this upshot, it would make a significant nod towards compatibilism despite the fact that he uses an incompatibilist conception of freedom. Such a concession 61  (1) is affirmed by Langton 1998; Warren 2001 and unpublished. (2) is affirmed by Langton (1998: 50). 62  Thus, my interpretation is not subject to the worries that Allison (2004: 11, 52) and Ameriks (2003: 149) raise for Langton’s identification of things in themselves with intrinsic or non-­relational properties. 63 Note that some proponents of two-­object views make significant concessions to two-­aspect readings in order to accommodate Kant’s views on free human selves. For instance, while Jauernig argues extensively for a two-­object interpretation, she accepts that an “empirical self and the corresponding transcendental self are both aspects of the same entity, namely, a particular human being” (2021: 289).

38  Kant on Freedom and Rational Agency is suggested by commentators who think (cf. n. 3) that Kant’s view combines compatibilism and incompatibilism. Kant’s view that that an agent or action can be both causally determined and free seems to satisfy the lexical definition of the term “compatibilism.” However, this is merely a superficial agreement: for Kant freedom is compatible with causal determination only because freedom is compatible with the negation of freedom. For Kant “necessitated by natural causes” means (is synonymous with) “unfree”: the human will can be considered as both free and “as subject to natural necessity, that is, as not free” (Bxxvii); “transcendentally free” and “causally necessitated” are “contradictory concepts” (KpV, 5:95). These contradictory concepts can be applied to the same subject only because it has two different ontological constitutions (Bxxvii–­Bxxviii; KpV, 5:95–6; Prol, 4:343) and because causal necessitation belongs to that constitution of the subject which has an ontologically relative, mind-­dependent status: it is not an intrinsic property but rather depends on “the relation of the given object to” specific finite cognitive faculties (B69; cf. Section I.4). Thus, Kant’s view is irreconcilably opposed to the core compatibilist idea that freedom is compatible with determinism since “determined by natural causes” is not synonymous with “unfree.” This is not a merely semantic issue. If “free” and “causally determined” are contradictory predicates, it is hard to see how we can fruitfully use these predicates simultaneously in one and the same context of inquiry. Consequently, Kant holds that these concepts are consigned to essentially different fields of human reasoning. This is the crucial idea behind his distinction between two standpoints that I mentioned in Section II.2 and that I now re-­examine in more detail to clarify how the two contradictory predicates figure in common human reason. A rational being “has two standpoints from which it can regard itself and from which it can cognize laws for the use of its powers, consequently for all its actions.” On the one hand, a rational being can consider itself “so far as it belongs to the world of sense” where it “finds itself subject to laws of nature.” On the other hand, a rational being can consider itself “as belonging to the intelligible world, under laws which being independent of nature have their foundation not in experience but in reason alone” (GMS, 4:452). Kant uses two related criteria to define and distinguish these two standpoints.64 First, each standpoint pursues specific ra­ tional purposes, addresses distinctive kinds of questions, and is restricted by corresponding limitations: there are other legitimate purposes and questions that it cannot fruitfully pursue or address. “Each perspective has its necessary use, and limits.”65 Second, each standpoint employs a set of representations that it needs for answering its distinctive questions, but which are ill-­suited for addressing questions that can only be raised from the other standpoint. 64  In what follows, I give an abbreviated version of my account in Kohl 2018a. 65  Hill 2002: 20.

Freedom, Idealism, and Standpoints  39 One of these standpoints addresses observational, predictive, and explanatory questions about natural objects that stand under natural laws. From this perspective we have “only the course of nature in view” and our “understanding can . . . only cognize what there is, or was, or will be” and why things exist, did exist, or will exist (A547/B575). This is the only standpoint from which we can explain things, since “where determination according to natural laws ceases, there all explanation also ceases” (GMS, 4:459). We can also adopt this perspective towards ourselves when we consider our “empirical character” by virtue of which we “have to conform to all the laws of causal determination” and so are “nothing more than a part of the world of sense,” our actions being “the inevitable outcome of nature” (A540–1/B568–9). If we consider human beings as parts of the mechanistic order of nature, theoretical reason prohibits us from representing them as (truly) spontaneous (A542–4/B570–2) and denies that any actions other than those that actually occur at a certain time are really possible (KpV, 5:95). From this perspective judgments about what ought to happen are meaningless (A547/ B575) and agents are not morally responsible for their behavior (KpV, 5:95–7). But we can also adopt a different perspective towards ourselves from which we discard all explanatory, observational, or predictive purposes, when we represent human beings according to their “intelligible character . . . free from all influence of sensibility and from all [natural] necessity” (A540–1/B568–9). Since “ought has no meaning whatsoever if one has only the course of nature in view” (A547/ B575), we must adopt a standpoint which (at least partly) disregards the course of nature if we seek to cognize how we ought to act: from this standpoint we represent ourselves as subject to normative rather than natural laws.66 Here “reason does not follow the order of things as they present themselves in appearance, but in absolute spontaneity makes its own order according to ideas” by determining what actions ought to happen even though “they have not yet occurred and perhaps will not occur” (A548/B576). Thus, Kant distinguishes between an empirical or naturalistic and a normative standpoint. Neither of these perspectives affirms the compatibility of freedom

66  I identify the laws “which . . . have their foundation not in experience but in reason alone” (GMS, 4:452) with normative laws because Kant typically contrasts empirical “laws of nature . . . according to which everything happens” and normative “laws according to which everything ought to happen” (GMS, 4:388). Moreover, for Kant normative laws are, arguably, the only positive laws that we can cognize through pure reason. One might reply that the laws of general logic are non-­normative laws of pure reason because they are constitutive of any thinking activity (Tolley 2006). However, Kant explicitly assigns normative status to logical rules (JL, 9:13–14). When he says that pure logic contains rules without which no use of understanding takes place (A52/B76), he arguably means: “without which no right use of the understanding takes place.” His classification of the laws of pure logic alongside the normative laws of pure ethics (A54–5/B78–9) suggests that there can be illogical thoughts just as there can be immoral volitions. Due to my “lack of attention to the logical rules” (A296/B252) I might infer from “p implies q” and “not p” that “not q.” To criticize this inference as mistaken is to presuppose that I am engaged in an act of thinking that is liable to logical assessment via normative rules for right thinking.

40  Kant on Freedom and Rational Agency and determinism. Hence, the central aim of Kant’s reconciliation project is to show the compatibility of two incompatibilist stances: namely, the libertarian perspective from which we consider ourselves as norm-­governed agents who are free from natural necessitation, and the hard determinist perspective from which we regard ourselves as causally determined natural objects which lack freedom or spontaneity. The simultaneous conciliatory representation of freedom and determinism (i.e., the judgment that “freedom and nature . . . can exist together . . . in the same actions”) belongs only to speculative philosophy: it is external to the ordinary beliefs of both deliberating agents and empirical cognizers.67 My emphasis on the importance of Kant’s standpoint distinction may come as a surprise after my argument in Section I.2. It is thus important to be clear about the sense in which my interpretation counts as a two-­standpoint reading. I do not accept that appealing to standpoints can replace the ontological commitments of Kant’s doctrine.68 On my reading, each of the two standpoints involves an assertoric belief in the existence of a distinctive type of causality (nature and freedom). Kant’s standpoint distinction does not replace ontology but presupposes that “speculative philosophy” has done its “duty” to develop an ontological framework in which phenomenal natural causality and noumenal freedom can “well coexist” (GMS, 4:456–7) so that the existential commitments incurred by the two standpoints do not contradict one another. However, my reading does accept three crucial ideas that are usually associated with deflationary two-­standpoint readings. First, as we saw, I endorse a (qualified) one-­object account of the distinction between appearances and things in themselves: one and the same human being can be considered as causally determined or as free from such determination, depending on whether one represents the human being in relation to our cognitive faculties (according to its phenomenal constitution) or apart from this relation (according to its noumenal constitution). Likewise, I accept that the same human actions that are cognized as causally determined from the empirical standpoint can be considered as free from the 67  Proops sometimes presents Kant as rejecting an incompatibilist notion of freedom: he argues that Kant must be a compatibilist because he accepts event-­causal determinism and affirms the actuality of freedom (2021: 278, 282–8). Ultimately, though, Proops does not really deny the incompatibilist-­ libertarian orientation of Kantian freedom. He notes that although Kant is a determinist about all temporal events, the actuality of freedom requires that there be causally undetermined timeless actions, that free will must be “wholly spontaneous” (2021: 314). This is sufficient warrant for attributing to Kant an incompatibilist notion of freedom, which entails indeterminism at precisely the noumenal level where Kant locates the existence of the causality of freedom (as Proops also concedes; cf. 2021: 322). I am not sure why Proops insists that Kant’s view does not qualify as “a genuinely libertarian view” (2021: 323). In reading Kant as holding a libertarian conception of freedom, I follow Allais 2015; Ameriks 2000a, 2000b, 2003; Hogan 2009a, 2009b; Kosch 2006; van Inwagen 1993. 68  This deflationary claim is usually taken as definitive of two-­standpoint or two-­aspect readings, both by its proponents (Abela 2002: 38–9; Allison 2004: 37–8; Beck 1960: 192; Hill 2002: 36; Korsgaard 1996a: x) and by its opponents (Irwin 1984: 38; Proops 2021: 320–1; Van Cleve 1999: 147–50; Watkins 2004: 321–2).

Freedom, Idealism, and Standpoints  41 normative standpoint. On some readings Kant draws “a sharp metaphysical distinction” between causally determined spatiotemporal events and free actions,69 so that “that free actions cannot be events in the phenomenal world.”70 Such readings violate the “compatibilist” desideratum (D3) which arises from Kant’s claim that “freedom and [deterministic] nature . . . can exist together . . . in the same actions” (A451/B579). They sacrifice Kant’s ambition to save the commonsense idea that we can apply the concept of freedom to observable human behavior such as getting up from a chair (A557/B585), especially to spatiotemporal actions such as lies or thefts that are subject to legal and moral assessment. A human agent can “be called wholly free with regard to the same action” that is also a causally determined spatiotemporal event (KpV, 5:95; cf. GMS, 4:452). Depending on whether one attributes an action to natural causes from the empirical standpoint or to non-­natural rational causes from the normative standpoint, one considers different properties of that (same) action. The property of being causally determined belongs to the action as a mere appearance whereas the concept “right” “represents a . . . moral property of actions, which [can never be an appearance but] . . . belongs to them in themselves” (A44/B61). Third, in my view metaphysical speculation about how free causes operate or why free agents act as they do is pointless.71 Since the idea of freedom is consigned to a standpoint that systematically brackets theoretical questions, it cannot be invoked for the purposes of explanation or theory-­formation. This part of my interpretation goes back to the interpretive desideratum (D4): we must respect Kant’s insistence that the causality of freedom is incomprehensible, inexplicable, and (thus) cannot be used to explain anything at all.72 Kant’s standpoint distinction makes two important contributions to his doctrine of freedom. First, if the idea of freedom stakes its claim only from a standpoint that is solely focused on normative considerations, this yields a systematic reason for denying that we have any need to explain the causality of freedom. To insist on the need for an explanation of freedom or free actions is to misunderstand the only legitimate (namely, normative) role that the idea of freedom can play in our rational thinking. Second, even if Kant’s idealism provides an abstract insurance that there is no formal contradiction in believing ourselves to be both free and unfree, this may not suffice to dispel the worry that our self-­conception becomes deeply unstable when we ascribe to ourselves both of these “contradictory predicates.” But if this formal solution is combined with a standpoint 69  Hogan 2009a: 363. 70  Watkins 2004: 425. 71  See Korsgaard 1996a: 203. 72  By contrast, Watkins 2004 tries to reconstruct Kant’s conception of freedom as an explanatorily potent model of substance causation. He tries to respond on Kant’s behalf to the worry that “proponents of agent causation can give no explanation of why the effect that the agent brings about” occur at a given time (2004: 412). But since Kant so persistently concedes that his conception of free agency cannot explain why acts occur at the time they do or, indeed, why they occur at all (A550/B578; A556–7/B584–5; GMS, 4:459; KpV, 5:49–50, 54–5, 98–9), a sympathetic interpretation should rather question whether the inexplicability of free agency raises a fatal problem.

42  Kant on Freedom and Rational Agency distinction that systematically consigns these predicates to two different fields of enquiry, then the relevant beliefs that employ these predicates are not only con­ sist­ent in the (indispensable) formal sense that they do not contradict each other but also in the deeper sense that they do not interfere with one other. We are absolved from viewing ourselves at once as free from and as subject to natural necessity: in a context where the intentional focus of our rational thinking is on our freedom, our (phenomenal) lack of freedom is a non-­issue and vice versa. Critics of Kant’s standpoint distinction argue that the allegedly different standpoints and their respective beliefs cannot be neatly insulated from one another. This objection has a number of facets.73 In the present context, one central problem stands out: it seems that when we consider how we ought to act we must also, at the same time, regard ourselves as subject to causal laws of nature.74 For instance, in deciding when I should leave the house so that I can keep my promise to meet someone on campus, I must think of myself as an embodied agent who is subject to the laws of physics which prevent me from flying to campus in just a few seconds. One might respond that the normative-­noumenal standpoint includes no representations of natural laws or spatiotemporal determinations whatsoever. On this austere reading, Kant’s normative standpoint is not meant to be a perspective from which ordinary agents make concrete judgments about how they ought to act in the empirical world. This would render the normative standpoint an abstract philosophical stance that is far removed from our pre-­philosophical self-­ conception as free norm-­governed agents. But I see no evidence that Kant considers his standpoint distinction an artifice of abstract theorizing. Just like his conception of the empirical standpoint seeks to capture the perspective of ordinary spatiotemporal experience and natural science, so his conception of the normative standpoint seeks to capture how every ordinary human being (“even the commonest understanding”; GMS, 4:452) considers itself as the free cause of concrete spatiotemporal actions (KpV, 5:98–100). A more permissive, non-­ austere conception of the normative standpoint is suggested by Kant’s remark that “insofar as we consider ourselves obligated, we regard ourselves as belonging both to the sensible world and yet at the same time to the intelligible world” (GMS, 4:453; cf. KpV, 5:87). Arguably, Kant here articulates what adopting the normative standpoint really amounts to for beings like us, i.e., for sensible creatures endowed with spontaneous rational faculties. To consider oneself obligated is to regard oneself as subject to norms that one apprehends in the imperatival formula, ought, which applies only to finitely rational, that is: embodied and sensibly affected (though not determined) agency. (I discuss this point at length in Chapters III, V.) This suggests that the normative standpoint is a perspective from 73  See Kohl 2018a for detailed discussion. 74  This problem is raised by Nelkin 2000 and Watkins 2004.

Freedom, Idealism, and Standpoints  43 which we regard ourselves as members of both worlds, as free, norm-­governed beings that are also spatiotemporally situated and subject to natural laws. What, then, about the worry that the normative standpoint of freedom cannot be insulated from the belief that as members of the sensible world we are also subject to causal laws of nature and, as such, unfree? On the non-­austere reading, the normative standpoint need not insist on being completely insulated from the empirical standpoint: it can concede that our physical actions are constrained by natural conditions. This is because the normative standpoint does not regard our body as free from causal determination: it ascribes spontaneity only to the “higher faculties” that belong to us as “intelligences” (GMS, 4:451–2). The normative standpoint denies that we are determined by natural causes only with respect to rational faculties such as our will (considered as a noumenal power). We regard one aspect of ourselves, namely our higher mental capacities, as free and governed by rational norms (as sensibly affected but not determined), while we regard another aspect of ourselves (our body) as constrained by deterministic natural laws. Of course, we think that typically our physical actions depend upon our free will, i.e., on our freely chosen intentions. But from the normative perspective we must acknowledge that the free rational control that we exercise over our physical capacities is limited precisely because these capacities are also constrained by natural causes. This does not mean that our will always lacks causal efficacy with regard to what happens in nature, but it does mean that the extent to which our will has such efficacy is not fixed by our free volitional efforts. Our moral freedom of will is compatible with our inability to physically realize our freely chosen purposes and maxims (KpV, 5:36–7, 45); a morally good will can “sparkle like a jewel” even if it fails to realize its intentions in the physical world (GMS, 4:394). For example, from the normative standpoint I ascribe to myself the freedom to pursue the goal of being on campus at the promised time, regardless of how strongly I desire not to go because I feel very exhausted. But my self-­ awareness as a free finite agent also includes the recognition that my morally good choice might not, despite my best efforts, have the intended effects: when I seek to execute my intention to keep my promise and walk to the bus, my body might be empirically determined to suffer a stroke. More generally, which physical effects I can produce through my free agency is limited by empirical conditions that are outside of my free self-­control.75 Here one might ask: what accounts for the difference between cases in which our free choices do and cases where our free choices do not cause the intended bodily movements and effects? Kant regards this question as inscrutable. As we saw, for Kant all our explanations must defer to laws of nature (GMS, 4:459). Since the impact of our non-­natural free will on our bodily activity cannot be 75  For discussion of how this point affects Kant’s verdict that moral “ought” implies “can,” see Kohl 2015d.

44  Kant on Freedom and Rational Agency understood via natural laws, we cannot explain or comprehend why our free volition sometimes results and sometimes fails to result in intended bodily movement.76 But for Kant we should not be concerned about this incomprehensibility. Here we can return to the central idea behind his standpoint distinction: the two standpoints delineate distinctive fields of inquiry where only certain types of questions are legitimate, relevant, or worth pursuing. Questions about how, when, and why our free will causes intended physical movements are irrelevant when we adopt the normative perspective towards free rational beings: “Whether the causality of the will suffices for the reality of its objects or not is left for the judgments of . . . theoretical reason . . . in the practical task of reason . . . what matters is only the determination of the will and the determining ground of its maxim as a free will, not its success” (KpV, 5:45). When we take up the normative stance, we do not seek to understand or explain free agency but to cognize how we should act and to choose accordingly. Thus we can “happily admit being unable to comprehend how” the noumenal causality of reason influences the sensible world (KpV, 5:49). My interpretation suggests a middle path between readings that deflate the ontological implications of Kant’s doctrine and readings where Kant (not entirely unlike his rationalist predecessors) offers a metaphysical theory of how the causality of freedom works. Like other intended middle paths, this might seem to yield an unhappy compromise. If Kant denies that we can give a metaphysical theory which explains how the causality of freedom works, is it not, then, odd for him to nevertheless defend a metaphysical belief in free causal powers that have effects on the sensible world?77 Conversely, if Kant seeks to defend metaphysical beliefs in noumenal powers, what should bar him from also giving an account of how these powers operate? However, there clearly is an important difference between positing the ex­ist­ ence of something and giving a theory of how or why this thing exists and acts as it does. There is no prima facie tension between holding that we can do the f­ ormer and cannot do the latter. For instance, one might think that the best bet for traditional theism is to posit God’s existence and then declare that we cannot resolve the many mysteries surrounding God’s existence due to our cognitive limitations. Of course, one might then press theists on their reasons for believing in God, but this is a different challenge of justification; whether theists can justify their claim

76  Notably, this problem arises not just because of Kant’s appeal to the relation between noumenal will and empirical body: it remains when we switch to the relation between phenomenal mind and body because we cannot model an effect of phenomena of mere inner sense on phenomena of outer sense (A379–830; A683/B711). This is clearly not an idiosyncratic aspect of Kant’s view. For example, Hume also stresses that no “principle in nature is more mysterious than . . . the secret union of mind and body” (1993: Section VII, Part I). 77  See Allison 2020: 480.

Freedom, Idealism, and Standpoints  45 that God exists need not depend on their ability to comprehend how God exists or why He acts as He does. There is compelling textual evidence that Kant holds such a view with respect to freedom: “But how freedom is even possible and how we are to present this kind of causality theoretically and positively—­into this we do not . . . have insight; rather, only that there is such freedom is being postulated through the moral law and for its sake” (KpV, 5:133). When he discusses the role that metaphysics plays in his views on moral freedom, he admits that it may seem as if the very idea of a normative doctrine of virtue is inconsistent with the project of going back to the first metaphysical grounds of duty (MS, 6:375–6): the ordinary notion of virtue can hardly “borrow her arms from the armory of metaphysics” since metaphysics “is a matter of speculation that only few human beings can handle” (MS, 6:376). Nevertheless, he insists that a metaphysical consideration of the first principle of morality is necessary to prevent what he regards as a common, dangerous assumption among theorists: namely, the assumption that this principle is grounded upon empirical feelings. The moral law “is really nothing else than an obscurely thought metaphysics which inheres in every human being in its disposition to reason” (MS, 6:376). Common agents grasp (however unclearly) a connection between rational moral norms and supersensible freedom as a capacity to choose from purely rational motives, without motivational assistance from sensible causes such as empirical feelings. (I discuss this point in Chapters III, V, VIII.) The tendency to reject such ideas as illusory, to ground moral norms and motives instead upon empirical sources, is borne out of the frustration of those who believe in “the omnipotence of theoretical reason” and who balk at the fact that they are “unable [to] explain what lies wholly beyond th[e] sphere” of human understanding, namely our “freedom of choice” (MS, 6: 378). This shows Kant’s commitment to a metaphysical doctrine of moral freedom that abstains from offering a theory that explains free will. We cannot deflate the metaphysical content of the beliefs about moral freedom that are held, however obscurely, by ordinary agents. But neither can we subject this metaphysical content to the demands and standards of theoretical reason and its inherently “physiological” (MS, 6:378), i.e., naturalistic explanations. Philosophy can give a clear analysis of the metaphysical concept of moral freedom that is obscurely thought by every human agent, and it can (as idealist philosophy) defend the legitimacy of this concept against threats such as naturalism and fatalism. It can also expose the mistake of those who infer from the fact that freedom is subjectively (for human cognizers) incomprehensible that freedom is therefore objectively (“intrinsically”) impossible (GMS, 4:459; cf. Refl., 17:346–7). But human reason, including philosophical reason, cannot give an account that explains how the causality of freedom operates or why free agents do what they do at any particular time and place.

46  Kant on Freedom and Rational Agency

I.6.  Conclusion: Problems and Prospects for Kant’s Doctrine In Kant’s doctrine, freedom is a non-­natural, atemporal, hence (subjectively, for us) incomprehensible type of causality. For some, the fact that Kantian freedom is unintelligible yields a sufficient reason to reject Kant’s doctrine.78 However, Kant does not view the incomprehensibility of free noumenal causation as a fatal problem; rather, this incomprehensibility emerges as a systematic feature of his overall philosophy. Those who reject Kant’s doctrine because it self-­consciously posits an inscrutable form of causality owe us some further argument. It is not self-­evident that a view which renders freedom theoretically unintelligible is therefore unacceptable, especially if that view systematically shows why we cannot make theoretical sense of freedom and, moreover, defines a highly significant normative role for the idea of freedom outside the context of theoretical explanation. But it is of course true that the bars for accepting such an account are raised in comparison to views that do allow us to understand how the causality of freedom works. This may well be the crux of the issue: presumably, those who reject Kant’s doctrine of freedom do so not just because this doctrine renders freedom unintelligible but due to their additional conviction that there is no need for an unintelligible concept of freedom. That is, implicit in the rejection of Kant’s doctrine of non-­natural freedom as unintelligible is often, though not always,79 the assumption that there are more satisfying naturalistic alternatives which preserve the intelligibility of freedom. Thus, assessing the viability of Kant’s doctrine turns centrally on the question of whether there are satisfying alternatives to his view that do not impose the cost of theoretical incomprehensibility, or whether it might be worth paying this cost since these putative alternatives sacrifice central aspects of our self-­conception as free agents. Defenders of Kant’s view must address three separate issues. First, they must show that Kant has good reasons for dismissing compatibilist conceptions of freedom according to which the causality of freedom is identical with the deterministic causality of nature and is therefore fully comprehensible. Second, they must show that libertarian attempts to integrate an incompatibilist conception of freedom with our naturalistic understanding of the world qua indeterministic universe are unsatisfying. Third, they must show that we have good reasons to prefer the belief that we possess a theoretically inscrutable capacity for free agency over the hard determinist alternative that we lack this capacity: Kant owes

78  See Guyer 1987: 303–5; Nagel 1989: 119; Strawson 1966: 235–72. A more nuanced view is suggested by Van Cleve 1999: 138. He remarks: “I do not say that [atemporal causation] is unintelligible or incoherent, but I do find it difficult to comprehend.” For Kant, the atemporal causality of freedom is impossible to comprehend and thus unintelligible to us: but (as we saw) he denies that it is therefore incoherent to believe in freedom. 79  Nagel 1989 arguably does not envisage an alternative view that renders freedom comprehensible.

Freedom, Idealism, and Standpoints  47 us a compelling justification for the belief that we are free in his metaphysically demanding sense. If these three things could be shown, Kant’s doctrine might be attractive even though it comes at the price of theoretical unintelligibility. Kant’s view promises to synthesize: (1) the common view that libertarian accounts are inadequate for theoretical purposes; (2) the claim that naturalistic compatibilist models of freedom are inadequate for normative-­practical purposes; and (3) the conviction that we can legitimately regard ourselves as free, responsible agents. In the further course of this book I shall mostly bracket the question of whether a naturalistic libertarianism that is based on an indeterministic concept of natural causation yields a viable alternative to Kant’s anti-­naturalistic libertarianism. I do not pursue this topic for three main reasons. The first reason is pragmatic: considering this topic would add further intricate philosophical and scientific considerations to an already very complex (and voluminous) dialectic. The second reason is that naturalistic libertarianism may not be a genuine alternative to the Kantian model because naturalistic libertarianism likewise invites the charge of incomprehensibility. One long-­standing worry about libertarians of all colors is that libertarianism makes it inexplicable why an agent’s state at a given time leads her to perform one action rather than another. To be sure, there are central differences between the way in which this worry arises for Kant and for naturalistic libertarianism. But clearly, naturalistic libertarianism does face serious questions about how we could explain a free action if the processes that comprise the agent’s deliberation and choice involve chancy gaps (elements of sheer randomness). Third, the appeal to random distributions of probabilities in quantum mechanics, which may determine very low probabilities that agents make morally obligatory choices, arguably cannot speak to incompatibilist worries about whether natural causation leaves genuine room for free, responsible moral agency.80 My reconstructive interpretation will focus on two main issues: on Kant’s reasons for rejecting naturalistic compatibilism and on his attempt to legitimize the belief in supersensible freedom. If I can shed light on these issues, we can at least see why Kant felt so strongly committed to an (as he was well aware) incomprehensible idea of supersensible freedom.81 We may even be tempted to agree with him that this incomprehensibility does not pose a fatal problem.

80  See Loewer 1996 (cf. Introduction, n. 15). The issue here is not merely that for someone afflicted with very low probabilities it is very unlikely to act according to moral reasons. The more fundamental problem (which would persist even if the probability distributions turned out to be more favorable) is that the moral options for free agents wholly depend on how the contingent probability dices happen to roll. For Kant, this implication of naturalistic libertarianism puts it on par with naturalistic compatibilism. 81  There is a tendency even among broadly sympathetic commentators to leave Kant’s deep commitment to a noumenal, non-­natural free subject rather unmotivated. This tendency can be witnessed, for instance, in Longuenesse’s naturalistic Freudian take on the Kantian self (2017).

48  Kant on Freedom and Rational Agency Let me add a final clarification. On my reading Kant holds that the (real) possibility of freedom depends on a metaphysical form of idealism. One might wonder: even if we grant that the theoretical unintelligibility of Kantian freedom does not sufficiently undermine his doctrine of freedom, surely this doctrine is conclusively undermined by its reliance on an outdated form of idealism? Here I can only say that in my view transcendental idealism is not the obviously outdated non-­starter that many take it to be. As I argued in Section I.4, Kant’s idealism does not seem blatantly flawed or incoherent. I also do not see that recent scientistic philosophy has established that a naturalistic metaphysical realism, which affirms the absolute reality or mind-­independence of natural facts and causes, is our only live or our best option.82 Denying that transcendental idealism is an obvious non-­ starter does not require the stronger claim that it is ultimately correct. My project to provide a rational reconstruction of Kant’s doctrine of freedom need not endorse this ­stronger claim. If it could be shown that Kant’s account of freedom is worth taking seriously (because Kant’s transcendental conception of freedom is more adequate to our self-­image as free agents than naturalistic conceptions, and because Kant can offer good reasons for believing that we are free in the transcendental sense), this would also provide a reason to take seriously his idealism, since his account of freedom depends on his idealism. As we saw, Kant holds that the ability of his idealism to make room for the causality of freedom alongside the causality of nature provides one strong motive for endorsing his idealism. This does not seem implausible. If a philosophical idealism makes room for the existence of two seemingly incompatible types of causality, this is a good reason to accept such idealism: provided that it can be shown on independent grounds, without prior appeal to idealism, that both the causality of nature and the non-­natural causality of freedom play integral parts in our rational belief-­system.

I.7.  Appendix: Practical and Transcendental Freedom To avoid introducing further complications and technical vocabulary, I have focused on Kant’s conception of transcendental freedom. I have abstracted from the distinction between practical freedom (PF) and transcendental freedom (TF) which Kant draws in the first Critique’s Dialectic and then puts to some very controversial use in the Canon of Pure Reason. Kant’s discussion of PF in the Dialectic fits neatly into my interpretive framework. His PF-­TF distinction here is systematically important because Kant presents PF as a distinctive species of TF: namely, as the specific kind of TF that 82  For the entirely unargued but firmly held assumption that idealism is no longer a live option, see Rosen 1994.

Freedom, Idealism, and Standpoints  49 pertains to a finitely rational, sensibly affected yet undetermined faculty of choice.83 This suggests that there may be further species of TF such as divine freedom (cf. Chapter  III) or the epistemic freedom of our understanding (cf. Chapters VI–­VII). In any case, Kant asserts that PF existentially depends on TF: “the abolition of transcendental freedom would also simultaneously eliminate all practical freedom” (A534/B562). Therefore, my account of TF in this chapter applies equally to PF as it figures in the Dialectic (i.e., in Kant’s solution to the Third Antinomy).84 However, Kant’s PF-­TF distinction in the Canon of Pure Reason raises problems for my account because Kant here (at A801–4/B829–32) makes claims that seem at odds with some of my interpretive desiderata. He seems to suggest, in apparent conflict with (D4), that PF is an empirical concept which is given through experience and has explanatory power. He also seems to claim, in apparent conflict with (D2) and his view in the Dialectic (A534/B562), that we can ascribe PF to ourselves without considering whether we are transcendentally free from empirical determination. This seeming conflict between Canon and Dialectic has led some commentators to propose that in the Canon Kant is working with a pre-­critical, compatibilist-­empiricist conception of PF. This suggests a patchwork account of how Kant composed the first Critique.85 I do not accept this patchwork reading and its associated claim that Kant uses a compatibilist-­empiricist concept of PF in the Canon. When Kant says that PF can be proven through experience, we can arguably read him as employing a non-­ standard, practical conception of proof and experience (rather than an empirical notion of freedom).86 But fully supporting an anti-­patchwork, anti-­compatibilist reading of the Canon would require lengthy, subtle philological considerations. There are two reasons why it is unnecessary and unprofitable, in the context of my project, to devote much attention to the Canon-­distinction between PF and TF. First, Kant assigns no major weight to this distinction in other central texts: neither (as we saw) in his resolution of the Third Antinomy nor when he discusses freedom in central works such as the Groundwork, the second Critique or the Religion. In these later writings Kant does not even formulate any distinction between PF and TF: his focus is squarely on the incompatibilist notion of a transcendentally free will (or, as in the Dialectic, on PF qua existentially de­pend­ ent on TF).

83  I discuss this in Kohl 2014. 84  Allison (1990, 2020) argues that PF depends on TF only conceptually, not existentially or ontologically. This view does not (as Allison admits) accord with the squarely ontological language of A534/B562. It also invites the objections against deflationary readings that I raised in Section I.2. 85  See Carnois 1986: 29; Gunkel 1989: 94–105; Kvist 1978: 54; Schönecker 2005: 98–105; Steigleder 2002: 4–8. 86  See Kohl 2014. This point is indebted to how Allison 1990 and Ameriks 2000a solve the Canon problem.

50  Kant on Freedom and Rational Agency Second, quite independently of any patchwork or compatibilist suppositions, there is an important sense in which Kant’s Canon-­position fails to express his mature, critical views on freedom (which are the sole focus of my interpretation). The Canon relies on a heteronomous view according to which both human moral motivation and obligation depend upon our empirical wish for happiness in an afterlife (A811/B839; A813/B841; cf. A15; A589/B617; A633–4/B661–2; V-­Th/ Volck, 28: 1138, 1153).87 By contrast, it is a central feature of Kant’s mature doctrine that free will and moral autonomy are reciprocal concepts: as we shall see (in Chapter III), this doctrine centers on the idea that a free rational will is, essentially, an autonomous will that ought to and (therefore) can act on the basis of non-­empirical, purely rational or intellectual motives. Hence, we must separate the heteronomous Canon position (including its empiricist-­sounding views on PF) from Kant’s considered critical doctrine of freedom which is the topic of this book. 87  Here I agree with Allison 2011: 15–16, 55–6, but I disagree with Ameriks 2019: 43, 45 who claims that the Canon presents the standard Kantian doctrine that is also present in the later works. See my review of Ameriks 2019 in Kohl 2021b. I discuss this issue at greater length in Chapter VIII.

II Human Action as the Effect of Two Causes In chapter I, I proposed an interpretive framework for understanding Kant’s claim that his idealism makes room for the spontaneous causality of freedom alongside the deterministic causality of nature. In Kant’s view, our spatiotemporal actions can be regarded as both unfree and free because they can be ascribed to two different types of causes. Human actions are unfree as effects of causally determined capacities that belong to the empirical constitution which subjects have (only) in relation to finite cognitive faculties, as phenomenal objects of a spatiotemporal experience. Human actions would be free if they were also effects of causally undetermined capacities that might belong to the noumenal constitution which subjects have in and by themselves, apart from their relation to finite cognitive faculties. In this chapter I address various complications arising from this attempt to reconcile freedom and natural necessity.

II.1.  The Causal Dependency of Empirical Character on Noumenal Freedom One such complication arises if we consider how the same effect could result from two different causes. If each cause is individually sufficient for producing human actions, then it follows that such actions are causally overdetermined. The idea of causal overdetermination may not be inherently problematic. But there is something odd, to put it mildly, about a view that tries to save libertarian freedom by tracing our spatiotemporal actions to nondeterministic causes while conceding that deterministic natural causes are also sufficient for producing these actions. This view seems to entail that whatever we freely decide to do, our observable actions must happen anyway as effects of deterministic natural causes. How, then, does our freedom make any real difference to how we act in space and time?1 Kant’s response is to deny (or better, to qualify; see Section II.2) the assumption that natural causes are sufficient for producing our actions. As we saw (in Chapter I), in his idealist view the causality of nature has only a relative, mind-­ dependent ontological status. Since the causality of nature is not ontologically fundamental, it cannot be “the completely determining cause, sufficient in itself,

1  See Bennett 1974: 199–200; compare Beck 1960: 191–2 and Beck 1987: 42.

Kant on Freedom and Rational Agency. Markus Kohl, Oxford University Press. © Markus Kohl 2023. DOI: 10.1093/oso/9780198873143.003.0003

52  Kant on Freedom and Rational Agency of every occurrence” (A536/B564). Kant’s idealist account of natural causality makes room for the possibility that the natural causes which belong to our spatiotemporal constitution might causally depend on free causal powers which (potentially) belong to our noumenal constitution: “although for every effect in appearance there is required a connection with its cause in accordance with laws of empirical causality, this empirical causality itself . . . could . . . be an effect of a causality that is not empirical, but rather intelligible, i.e., an original action of a cause in regard to appearances” (A544/B572). If that were the case, then the empirically determining grounds of our observable actions which constitute our empirical character would be “determined in the intelligible character” (A551/ B579). If our free intelligible character were the “transcendental cause” of our “empirical character” (A546/B574), then our noumenal causality of freedom would make an indispensable contribution to how we act in space and time because it would control those natural causes that empirically determine our imputable phenomenal actions. This suggestion raises various questions which I address over the course of this chapter. I begin by clarifying how it relates to Kant’s general idealist commitments. In Kant’s idealism natural causality is a feature that pertains to subjects as part of their mind-­dependent, phenomenal constitution which they possess only in relation to our finite cognitive capacities. He assigns this ontologically relative status to natural causality primarily for epistemological reasons: if we thought of deterministic natural causality in realist terms, as part of the ultimate mind-­ independent fabric of the world, we could (being awoken from our dogmatic slumber by Humean doubts about necessary connections) never know that necessitating natural causes really exist (KpV, 5:53). This reveals a general motivation for Kant’s turn to idealism aside from his concerns about moral freedom: the epistemological aim to show that and how we can know non-­trivial (synthetic) necessity claims. (I discuss this issue at length in Chapter VII.) The principle that all natural events are causally determined is such a claim. Its truth does not follow analytically from the very meaning of the term “event,” and it cannot be justified through sense-­experience either. Kant justifies this principle roughly as follows: determining causes exist in the phenomenal world that is (partly) constituted by the conditions of temporally structured human cognition because our cognition of objective temporal succession among states (e.g., cold, warm) in an event or alteration (e.g., when a stove changes from cold to warm) requires that we represent the successive order among these states as fixed by some preceding cause.2 Give this cognitive-­temporal grounding of the causal 2  This is Kant’s Second Analogy proof of empirical determinism. Premises of this proof include the claim that temporal relations such as succession are not directly perceivable and the claim that the ever-­successive course among our given perceptions does not determine whether the states we perceive are objectively successive or simultaneous. For helpful discussion of Kant’s proof, see Allison 2004, ch. 9; Friedman 1992; Guyer 1987, ch. 10. I take this proof to show that every temporal

Human Action as the Effect of Two Causes  53 principle, its validity extends no further than to the temporal-­empirical character of acting subjects (cf. Chapter I). Since temporality is a feature that acting subjects have only in relation to human representational faculties (our “mode of intuition”), the same goes for the temporal-­causal determination of acting subjects. Thus, all causal powers of subjects which operate within the natural-­temporal order and whose exercise is empirically determined have a mind-­dependent ontological status. (If they had a mind-­independent status, they would be cognitively inaccessible and thus unknowable to us.) Hence, natural causal powers do not exist in and by themselves, apart from the manner in which the subjects that are the seats of these powers relate (“conform”) to our (generic) cognitive faculties. But this does not fully explain how Kant conceives the ontologically dependent status of natural causes. Our (generic) cognitive faculties impose only the general feature of being determined by some natural causes on the phenomenal objects of our spatiotemporal cognition. This does not determine the specific causal laws that govern particular spatiotemporal objects and that make up their specific empirical character (cf. A126–8; B163–5). The specific empirical-­causal character of spatiotemporal objects depends on some non-­empirical ground other than our generic cognitive faculties which impose the same general natural lawfulness on all appearances alike. (Similarly, the specific spatiotemporal properties—­e.g., the particular shape or duration—­of appearances depend on some non-­empirical ground other than our mode of intuition which imposes the same general spatiotemporal character on all appearances alike.) The specific empirical-­causal character of individual spatiotemporal subjects depends on some aspect of the non-­ sensible, noumenal reality that underlies spatiotemporal appearances. Hence, the specific empirical-­causal characters that individual human subjects exhibit as spatiotemporal phenomena might depend on the specific manner in

succession is determined by a cause that operates according to some universal causal law (for this reading see, e.g., Friedman 1992). Some (e.g., Allison 2004: 257–8) claim that the Second Analogy fails to establish that causes must have a law-­like character. However, for Kant there is an analytic link between “causality” and “strict lawfulness”: “The concept of causality itself contains . . . evidently the concept of . . . a strict universality of the rule” (B5); it “makes the strict demand that something, A, should be such that something else, B, follows from it . . . in accordance with an absolutely universal rule” (A91/B123–4; see also A549/B577). Thus, the (synthetic) proof that events have causes conceptually implies that events have causes that operate according to universal laws of nature. Allison at one point (1996: 86) concedes that the Second Analogy shows that there are rules governing events, but argues that these rules might be “laws with merely a single instance,” which would not be “genuine laws” that require “regularity and repeatability.” There are two different senses of “regularity” at issue here. I take Kant to hold that natural causes operate according to laws that are regular in the sense that the same causal conditions would always produce the same effect. Allison supposes that causes operate according to “genuine laws” only if the law is regular in the different sense that the same causal conditions are or can be operative repeatedly. But I see no evidence that Kant accepts this notion of regularity as a condition on empirical lawfulness. Some natural causes might produce their characteristic effects only under very specific conditions that obtain rarely or even just on a singular occasion; these causes are still governed by laws (e.g., the law governing the production of the Big Bang) because they would always produce the same effect whenever the relevant conditions obtained.

54  Kant on Freedom and Rational Agency which they (as non-­sensible things in themselves) exercise their noumenal causality of freedom. I say cautiously, “might depend,” because on my reading Kant’s idealist ­conception of natural causality (as lacking an absolute ontological status) does not prove that our specific empirical character in fact derives from the specific use that we make of free noumenal powers.3 Here we can consider a passage (A538–9/B566–7) where Kant makes two claims: (1) Because the phenomena that we cognize through our senses are not things in themselves, “a transcendental object must lie at their basis that determines them.” (2) Thus, “nothing prevents us” from ascribing to this transcendental object “a causality which is not appearance,” namely “an intelligible character” as the spontaneous cause of phenomenal actions. I read this passage as follows: (1) We know that some non-­ sensible reality which (as such) is not subject to natural causality (partly) determines the specific empirical character of sensible phenomena; hence, we can coherently suppose (2) that in the case of human agents the non-­sensible ground of their specific empirical character is their causality of freedom. Kant’s subsequent argument shows that (1) does not imply the truth of supposition (2). He stresses (at A546/B575) that with regards to animals or inanimate matter we have no reason to posit any unconditioned, spontaneous capacities even though “a transcendental object must lie at their basis” and determine their specific empirical character as well. Moreover, Kant emphasizes that his idealist account of natural causality does not prove the real possibility of a free causality (A557–8/B586–7; cf. Chapter I). Hence, when he says (in (2)) that “nothing prevents us” from assuming a free noumenal causality in human subjects as the non-­empirical ground which determines their specific empirical character, he means only: his idealist denial that unfree (causally determined) natural causes have a metaphysically absolute status makes ontological space for the existence of a metaphysically more fundamental free causality through which particular human subjects influence or control which specific natural causes comprise their empirical character. I do not see that Kant’s idealism entails that the noumenal features which influence the specific empirical character of natural phenomena must comprise any absolutely free causal powers. But even if Kant’s idealism had this implication, this would not show that our noumenal character involves the causality of freedom. For all that Kant’s idealism shows, the free causality that fixes our empirical character might belong to some non-­human noumenal being (like God). As I argued in Chapter  I, Kant’s idealism makes room for but does not prove the real possibility of our noumenal freedom. In the present context, this means: his idealism makes room for but does not prove the real possibility that

3  I discuss the contrary view (held, for instance, by Hogan 2009a) in Chapter I.

Human Action as the Effect of Two Causes  55 our non-­natural constitution involves a free causality that controls which specific natural causes make up our empirical character and empirically determine our actions. In this way, Kant’s idealist framework indicates how our observable actions might strictly depend on how we exercise our free will even though these actions, qua parts of the natural order, are also necessary effects of causally determined natural causes. However, natural determinism may still seem problematic for our self-­conception as free agents. Does it not entail that our observable actions are predictable, and would such predictability not undermine our freedom?4

II.2.  The Predictability Problem Twice in his mature works, Kant asserts that if we had sufficient insight into the empirical character of the human will we would be able to predict spatiotemporal human actions with certainty (A549–50/B577–8; KpV, 5:99). Kant claims that the empirical predictability of our actions poses no problem for our empirically undetermined noumenal freedom. But it is unclear how the appeal to our noumenal freedom helps to avoid the worries which arise from the admission that our spatiotemporal behavior is in principle predictable. First, it seems that a valid prediction that x will happen entails certain knowledge that x will happen. But if I can know with certainty how I will act on some future occasion, then my future is no longer open to me and I lack the alternative possibilities required for practical deliberation. Second, if it is predictable that I will tell a malicious lie at future time t1, I cannot comply with the duty not to lie. This threatens the Kantian principle that moral ought implies can. Finally, it is unclear how Kant can insist on freedom at the noumenal level while conceding the predictability of our observable actions. As we saw, for Kant those actions can be regarded as effects of noumenal freedom. Thus, if it is predictable that I will lie at time t1, this entails a corresponding prediction about how I exercise my noumenal causality. But then it seems that I am not free to exercise this noumenal causality differently. Can we sidestep these worries by stressing that we cannot actually predict human actions?5 In the abovementioned passages Kant expresses a counterfactual: if we had sufficient insight into our empirical character then we would be able to predict human actions with certainty. But we cannot reach this insight because empirical psychology cannot obtain a properly scientific status (MAN, 4:471; EEKU, 20:237–8). Kant deems empirical psychology methodologically impoverished in various ways. For instance, it lacks the a priori mathematical

4  See Beck 1960: 191–2.

5  This seems to be Wood’s response (1999: 181–2).

56  Kant on Freedom and Rational Agency foundations that ground the properly scientific status of physics. We need such foundations to cognize lawful regularities that go beyond the merely “comparative universality” of inductive generalizations which lack predictive certainty (B3–4). However, if our practical freedom hinges on the unpredictability of our actions, and if our actions are unpredictable just because we are ignorant of the relevant determining laws, then our freedom depends on our epistemic deficiencies: we can deliberate under the idea of freedom only because we fail to know the causes that fix our future behavior. This undermines the special dignity that Kant attaches to the idea of freedom as the “keystone” of his system of pure reason (KpV, 5:3–4), as part of rational science (KpV, 5:30). One might argue that our inability to empirically predict human actions rests not on mere ignorance but on the special character of mental, intentional concepts. There is a reading of Kant’s idealism according to which Kant anticipates Donald Davidson’s anomalous monism, the view that mental and physical states are token-­identical but differ in type. On this reading, we can describe physical causes in an atemporal “noumenal” fashion by using psychological predicates that do not support deterministic laws and predictive beliefs. Hence, we cannot make predictions about human actions when these actions are intentionally described in mental terms.6 However, for Kant mental or psychological concepts do not per se refer to atemporal or noumenal features. The temporal domain of nature includes both physical phenomena of outer sense and “thinking nature,” i.e., mental phenomena of inner sense (MAN, 4:467). The latter phenomena are temporally determined just like the former (“time determines the relation of representations in our inner state”; A33/B49–50). One might concede that our mental states belong to the temporal order of nature but deny that these states are causally determined.7 However, if mental states do objectively succeed one another in the order of time, then they must also be causally determined given Kant’s argument that all objective temporal succession conforms to the law of causal determination (B233–4).8 Proponents of the Davidsonian reading might respond that psychological states cannot be causally determined because empirical psychology cannot establish causal laws.9 However, we cannot infer from our inability to establish psychological laws that such laws do not exist. This inference requires a verificationism about the existence of specific empirical laws that Kant rejects (KU, 5:183; JL, 9:66–7; EEKU, 20:215).10 Moreover, there is incontrovertible textual evidence that 6  See Hudson 1994 and Meerbote 1984. 7  See Westphal 2005. 8  Westphal argues that this law applies exclusively to physical states and that we can discriminate objective from apparent changes only in the case of bodies (2005: 236). However, Kant’s view must allow for empirical self-­knowledge of objective order among our mental states; after all, this is the crucial premise in his attempt to refute Cartesian skepticism about bodies (see B155–9; A342/B400; B430). 9  Hudson 1994: 67–9; Westphal 2005: 235. 10  For criticism of this inference, see also: Hatfield 1990: 68–70, 1992: 217–23; McCarthy 2009: 61–7. In his recent work Allison seems to accept this inference (2020: 424), whereas in his earlier work he took Kant to be committed to psychological determinism at the phenomenal level (1990: 31).

Human Action as the Effect of Two Causes  57 Kant accepts psychological determinism: in his view, “the necessity of events in time, according to the law of causality, may be called the mechanism of nature” even when the mechanism operates through irreducibly psychological states (KpV, 5:97); our phenomenal actions “must be comprehended as determined through other appearances, namely desires and inclinations” (GMS, 4:453); and, his claim that “the human soul” is subject to natural necessity (Bxxvii) implies psychological necessitation since in his view the soul differs from the body in appearance, as a natural-­temporal phenomenon (A379). There is another reading where our inability to predict human actions does not rest on mere ignorance. According to this reading, the “exhaustive investigation” into the phenomenal will that would enable the predictability of human actions (A549–50/B577–8) is impossible because our actions are not sufficiently empirically determined. In Kant’s idealism the spatiotemporal order of nature does not exist in its entirety, as a complete whole, apart from our piecemeal efforts to synthesize appearances into one indefinitely comprehensive course of experience. Thus, the fact that a conditioned event is given to us in our experience does not entail that the entire series of its causal conditions exists (A495/B552). Now assume that in order to predict events we would need to cognize a complete(d) series of empirical causal conditions as the sufficiently determining causal basis of these events. If so, the ontological fact that no such basis is there to be discovered would entail that human actions are unpredictable: this would impose a “theoretical limit on causal explanation” and prediction which provides “room for practical deliberation.”11 However, this reading faces a decisive problem: in Kant’s view we can make certain empirical predictions, e.g., of lunar eclipses. Hence, our ability to predict natural events does not depend on our grasp of a non-­existent complete(d) causal series. The above reading passes over an ambiguity in the notion of a sufficient causal explanation. According to the rationalistic idea of sufficient explanation that is proposed by the Third Antinomy thesis, we cannot sufficiently explain or predict an event unless we know the entire closed chain of its causal conditions. If Kant did accept this rationalistic notion of sufficient explanation, then he would also accept that due to the incomplete character of the phenomenal world no spatiotemporal event (thus, no lunar eclipse) is ever fully explicable or predictable. However, in his own doctrine Kant employs a more moderate notion of causal-­ explanatory sufficiency: his critical version of the “principle of sufficient reason” (A200–1/B246) holds that “appearances must . . . be capable of complete causal explanation in terms of other appearances in accordance with natural laws” (A546/B574). To achieve this explanation we need not grasp an entire chain of causes: “Appearances require to be explained only insofar as their conditions of

11  Frierson 2011: 93–4. Compare Bojanowki 2015: 100; O’Neill 1989: 61–2, 68. Lucy Allais has also suggested a view along these lines in correspondence.

58  Kant on Freedom and Rational Agency explanation are given in perception” (A483/B511–12).12 To explain and predict observable events, it is enough that we cognize those causal conditions whose joint operation at some time hypothetically necessitates the relevant effects. Due to our ignorance of human psychology, we cannot cognize the empirical conditions that necessitate human actions; hence we cannot predict such actions. But the mere appeal to our (current) epistemic shortcomings cannot save a rational concept of free agency. So, the question remains how Kant can save this concept against the predictability problem(s).

II.3.  A Solution to the Predictability Problem Let me clarify what Kant needs to show here. The three abovementioned worries all center on the assumption that the (in principle) predictability of our spatiotemporal behavior would remove the epistemic options that we must presuppose as deliberating agents who seek to figure out what to do and who assume, among other things, the absence of any certainty that we will perform a future action which we morally ought not to perform. The predictability problem is thus primarily a threat to the epistemic openness of our future. To solve this problem, Kant must show that the predictability of our future behavior does not threaten the epistemic openness of our future; but his argument must not rely on the ep­i­ ste­mic shortcomings which (currently) prevent us from making accurate predictions. If Kant succeeds in this task, there is no further need for him to show that our future is also metaphysically open, i.e., that deliberating agents possess metaphysically real rather than just epistemic options. This is because our possession of metaphysically real alternative possibilities is already secured by his appeal to the causally undetermined powers we might possess as noumenal agents. Of course, I have not yet shown how Kant justifies the claim that we indeed do possess such powers, but this issue is beside the point in the current dialectical context. The current question is how Kant can (on the one hand) trace our freedom to (potential) causally undetermined powers that would provide us with metaphysically real options and yet (on the other hand) concede that our spatiotemporal behavior is in principle predictable so that we may lack the epistemic options which we need for practical deliberation. My account of how Kant might avoid the predictability problem draws on an argument that has been advanced by various recent philosophers, some of which 12  For a similar distinction between a priori and empirical sufficiency, see Pereboom 2005. This distinction allows Kant to claim both that particular natural causes are sufficient (in an empirical sense) to produce their effects and that natural causes are a priori insufficient to produce their effects since they further depend on non-­sensible grounds. Kant’s view that the notion of a sufficient ground is ambiguous since the notion of sufficiency is context-­dependent can be traced to his earliest writings (see, e.g., PND, 1:393).

Human Action as the Effect of Two Causes  59 take their inspiration from Kant.13 Suppose I consider whether to lie on some future occasion. Someone announces to me her prediction that I will lie. This prediction cannot settle or preempt my practical deliberation about what I should do. Either (scenario 1) my learning of the prediction has no impact on how I proceed to deliberate and choose, or (scenario 2) it does. If the prediction has no impact, this is because I choose to ignore and disbelieve it: I cannot continue to deliberate about whether to lie if I (fully) believe that I will lie.14 So, in scenario 1 the prediction of how I will act is irrelevant to my practical deliberation. In scenario 2, my awareness of a prediction of how I will act is causally relevant to my decision-­making because it alters my deliberative states. In this case the prediction is impossible, not because of mere ignorance but for logical or structural reasons: since the prediction figures among the causal data that the predictor must use for calculating the prediction, it would (per impossible) have to be known before it is calculated. If the prediction itself were a potential contributing cause of how I deliberate and choose, the predictor would have to “factor the result of [her] calculation into [her] calculations in order to arrive at a result, and [she] would have to know what that result was before arriving at it in order to do so.”15 So, if predictions of my actions have an effect on my deliberative course they become impossible; if they have no such effect, this is because I do not believe them and hence (since belief is a condition of knowledge) do not know their truth. Thus, predictions of my actions cannot threaten my freedom to deliberate about and choose among actions that remain epistemically open possibilities for me unless I decide not to perform them: I cannot know that I will (not) perform some action unless I deliberately choose (not) to perform it.16 It is not anachronistic to suppose that this line of argument illuminates Kant’s swift dismissal of the seeming predictability problem. In the first Critique, he denies that predictions can threaten our freedom because we can consider our 13  See Bennett 1974, 1984; Bok 1998; Korsgaard 1996a, 1996b; MacKay 1960; Ryle 1949; Scriven 1965. 14  See Wallace 1994: 2–4. 15  Bok 1998: 8 (cf. Korsgaard 1996b: 95–8; Ryle 1949: 175–8). One might respond that a well-­ informed prediction would be iterative, factoring in the laws about how people respond to predictions; hence, the alteration that the prediction causes in a predictee’s (P’s) deliberative states would not skew the prediction but instead conform to it. But this response does not get around the original difficulty: even if a prediction factors in how P responds to it, the fact that the prediction alters P’s deliberative state means that it is among the causally relevant predictive data, alongside other physical and psychological information. Clearly this other information must be cognized prior to the prediction, as it provides the input for the predictive calculation. The same point applies to the prediction (or its iterated version) as well: one would (per impossible) need to know it prior to calculating it. To illustrate this point, suppose P has counterpredictive motives. A maximally well-­informed bot-­device could compute that P will do something other than what the bot announces P will do, but this means that whatever the bot announces to P is not a valid prediction: any potentially successful prediction of how P will act must remain concealed from P. This shows that, due to the structure of practical deliberation, deliberators cannot gain predictive knowledge of how they will act prior to concluding their deliberation with a firm decision as to how they shall act. 16  See Bok 1998: 120; MacKay 1960: 31–2; Moore 1912: 94.

60  Kant on Freedom and Rational Agency future actions not just as empirical objects of theoretical reasoning but also in their relation to “reason in its practical bearing.” When we consider our actions in this practical vein “we find a rule and order altogether different from the order of nature.” This rule of practical reason concerns what ought to happen in contrast to what actually happens in the natural order of things (A550/B578). In the second Critique (KpV, 5:99), Kant says that predictability raises no threat to our freedom because the moral law assures us that our future spatiotemporal actions are under our spontaneous rational self-­control. Thus, in both passages Kant dismisses the predictability problem by emphasizing that we can shift our focus from theoretical judgments about what will happen to practical judgments about how we ought to act. The above argument provides a clear rationale for Kant’s point: it shows why theoretical predictions cannot take away or preempt the rational control that we exert over our future actions through our normative-­ practical deliberation about how we ought to act. Importantly, this argument is typically proposed by compatibilists who seek to show that an unrestricted causal determinism does not threaten human freedom.17 Hence, the argument does not capture Kant’s incompatibilist view of freedom. But here I am not yet concerned with Kant’s metaphysical reasons for rejecting compatibilism (this is the topic of Part 2). In the present context, my point is that the above argument can be integrated into Kant’s incompatibilist framework and is thus available to Kant as an argumentative resource for defusing the apparent ep­i­ ste­mic predictability problems. Let me flesh out how this integration works. The second lemma of the above argument is that an empirical prediction of a human action can succeed only if it can treat the agent’s state of mind as a static given. If the predictive effort has an impact on the agent’s deliberative self-­ consciousness, the predictive judgment itself becomes one of the empirically determining grounds that the predictive effort seeks to capture, which structurally undermines that effort. As Ryle stresses, a reflexive practical self-­consciousness is always one step ahead of theoretical reasoning.18 This argument does not rule out that a deliberator’s reaction to her awareness of a prediction is fully determined by empirical causes. Hence, it does not show that our practical reasoning involves an absolute spontaneity. However, the argument allows for the idea of absolute spontaneity. The fact that the practical deliberation of a subject who is aware of predictive efforts systematically eludes these efforts can be interpreted as an (to be sure, inconclusive) indicator (perhaps a “sensible sign”; A546/B574) of an empirically unconditioned spontaneity. The causal impact that my deliberative reaction to a prediction has on my empirical consciousness can (coherently) be attributed to the absolute spontaneity of my practical reason: by exercising this spontaneity 17  Bok, Moore, and Wallace openly endorse compatibilism; for Korsgaard’s somewhat veiled compatibilism, see Chapter I. 18  Ryle 1949: 175–8.

Human Action as the Effect of Two Causes  61 I attach a practical significance to the theoretical proposition that I will choose x, e.g., by endorsing a counterpredictive reason not to choose x. The first lemma of the above argument is that if am aware of a prediction concerning my future actions, the prediction can succeed only if I do not accept it as true (so that it has no impact on my deliberative course). Our choices are based on our judgments about how we ought to act, where this includes our assessment of whether we should attach any (if so, what) normative-­practical weight to all sorts of theoretical information.19 Thus, in our first-­person deliberative stance a practical assessment of our normative reasons for acting has priority over a theoretical prediction of how we will act: we can set aside, ignore, or disbelieve such a prediction through our assessment that there is no good practical reason why the theoretical prediction should have any relevance to our practical decision-­ making. Again, this argument allows (though of course does not show) that our agency is based upon an absolutely spontaneous causal power: since no theoretical cognition of any conditioned empirical fact about ourselves can preempt the spontaneity of our practical reason, it might (pending further arguments) be legitimate to regard this spontaneity as empirically unconditioned. I take myself to have shown two things. First, there is a textual basis for attributing to Kant something akin to the contemporary strategy for defusing the apparent epistemic threat that predictability poses to our free practical deliberation. Second, although this strategy is commonly used by compatibilists it does not entail a compatibilist view of freedom: it can be combined with (though it does not motivate or vindicate) Kant’s incompatibilist account of free will as an absolutely spontaneous causality of practical reason. I now show how this strategy defuses the three specific predictability problems that I mentioned in Section II.2. The first problem was: predictions of how I will act threaten the belief in deliberative options that is required for practical deliberation. The above argument shows that this threat is empty: theoretical predictions of how I will act cannot deprive me of my freedom to practically consider whether or not I should perform the predicted action and to choose accordingly. Even the prediction of a discursive cognizer C who had the deepest possible insight into my phenomenal will could not bypass my practical reasoning and the choices I make based on such reasoning: even C’s prediction of how I will act could succeed only if I was either unaware of C’s prediction or, once C’s prediction was brought to my attention, I did not accept it as true, i.e., as showing with certainty how I will act.20 If I 19  See Beck 1960: 31–2. 20  Note that I am conceiving C as a discursive cognizer who tries to predict behavior via piecemeal synthesis of in(de)finitely further given data (such as predictee’s responses to predictions). For Proops (2021: 300), Kant is rather alluding to non-­discursive divine prediction in the relevant passages. But since Kant here considers attempts to infer future behavior from (as Proops says) “the totality of laws and the state of the world at a given time,” he cannot be concerned with an essentially atemporal, non-­ inferential divine (immediately intuitive) mode of cognition. (I consider whether divine foreknowledge raises a problem for our freedom in Chapter IX.)

62  Kant on Freedom and Rational Agency did (fully) believe that the prediction shows how I will act, then I could not continue my deliberative efforts to figure out how to act and to choose accordingly. But my practical deliberation and choice are necessary causal conditions of what I end up doing. Thus, C’s prediction that I will perform action A1 cannot preempt my practical decision-­making; different actions A2 and A3 continue to be ep­i­ste­ mically live deliberative options for me unless I decide for practical reasons that they are not worth pursuing. The second worry was: predictability threatens the Kantian idea that we can do what morality demands from us. Suppose Paul considers defaming a colleague for his personal gain. Suppose, further, that a psychologist who has meticulously studied Paul’s past deliberative patterns, moods, decisions, etc. predicts that Paul will choose to defame his colleague. Even if we grant to the psychologist the deepest possible insight into Paul’s empirical character, the prediction cannot reveal to Paul that he, as a rational moral agent, lacks the power to decide against defaming his colleague. This is because Paul can only act if he reaches a decision for or against defaming his colleague; and he can only reach this decision through his practical deliberation about what he has reason to do. Paul’s awareness of the prediction that he will defame his colleague cannot settle this normative-­practical task for him because this awareness is just a further piece of theoretical input whose practical relevance Paul must determine through his practical deliberation. Most importantly, Paul’s awareness of the prediction that he will defame his colleague cannot inform him that he lacks the rational capacity to change, due to his awareness of moral reasons, the past causal dispositions embedded in his bad character traits that must be held as fixed in attempts to theoretically cognize his future behavior. The same diagnosis applies to cases where predictions of human actions take a probabilistic form. Suppose Paul is an alcoholic who learns of multiple studies which show that for people who share Paul’s sociological features and character traits, it is unlikely that they will succeed to quit drinking without relapse. The theoretical information Paul gains through these studies cannot disclose that it is unlikely for him as a rational agent to quit drinking without relapse. The predictions that come out of these studies take Paul’s empirical character, i.e., his past and present psychological dispositions, as a static given: they infer his future behavior from observational data about the behavioral patterns he has displayed thus far. But as a rational agent Paul has the capacity to (perhaps only gradually) change those patterns due to his recognition that he has good normative-­practical reasons for doing so. For instance, he might recognize a conclusive moral reason to stop drinking so that he can prevent the harms which he may cause (himself and others) under the influence of alcohol. Paul may of course decide that in light of these studies it is not worth trying to comply with the moral norm that he ought to quit drinking. But this decision rests upon Paul’s practical conception of what is worth doing. His verdict that there is no point in trying might betray a devious

Human Action as the Effect of Two Causes  63 commitment (maxim) to keep indulging his old bad habits. On the other hand, Paul’s awareness of the studies may also elicit a counterpredictive, defiant motive for redoubling his efforts to change his morally bad dispositions. In sum, when we take up the normative standpoint from which we consider human beings as free rational agents, we cannot in good moral faith assume that empirical data concerning their past behavior show that they are incapable of changing or are unlikely to change for the better. According to the third worry, an empirical prediction of how an agent will act in space and time entails certain knowledge of how that agent must be exercising her noumenal causality of reason (as the mediate cause of her spatiotemporal actions), which implies that this exercise cannot be absolutely free. But, as we saw, agents cannot know how they will act unless they firmly decide upon a course of action, a decision that derives from their normative-­practical reasoning about what is worth doing. Hence, theoretical predictions cannot give deliberating agents certain knowledge of how they exercise their noumenal causality of practical reason. One might object that my attempt to defuse the worrisome implications of Kant’s appeal to the predictability of human action is ultimately no different from the strategies that I criticized in Section II.2: like those strategies, my argument has the upshot that we cannot really accept Kant’s claim that our actions would be predictable with the same certainty as lunar eclipses if we had better insight into the phenomenal human will. However, there is a clear sense in which my argument does respect Kant’s claim. My argument just adds a further condition for the predictability of human action: predictive efforts can succeed only if they can relate to human predictees in the same way as they relate to (e.g.) planets, namely, without altering the predictee’s causally efficacious states. This further condition can actually be met if human predictees are unaware of or disbelieve (ignore) predictions of how they will act. But even in such cases, the added condition still defuses the epistemic threat that predictability seemingly poses to our deliberative freedom because it shows that deliberators can never know how they will act unless they choose an action based on their practical reasoning. Since different actions remain epistemic possibilities for a deliberator unless she decides against them for practical reasons, theoretical predictions cannot preempt the practical spontaneity of deliberative choice. Moreover, the added condition preserves the epistemic possibilities of deliberating agents without invoking the epistemic deficiencies of predictors: this condition concerns a structural feature of self-­reflexive agency that would persist even if predictors were not ignorant of psychological empirical causes.21 21  This structural feature might be partly responsible for the fact that empirical psychology cannot develop into a proper science: one methodological problem Kant sees for empirical human psychology is that those human beings who are the objects of psychological studies are typically aware of this fact; this awareness typically alters their mental state and thereby distorts the relevant psychological phenomena (MAN, 4:470–1).

64  Kant on Freedom and Rational Agency However, my argument in this section has relied on a controversial assumption. I have argued that theoretical knowledge of the causal dispositions which have thus far constituted our empirical character does not threaten our practical freedom because (so I assumed) we have the self-­reflective spontaneous power to change our empirical character or (at least) to choose a particular out of character-­ action. This assumption requires further examination and defense.

II.4.  The Mutability of Character Some commentators hold that our noumenal freedom determines a set of immutable natural laws that yield a permanently fixed empirical character: our freedom is exhausted by the singular noumenal choice of a permanent “nature” that henceforth inevitably necessitates our phenomenal activity.22 Once the empirical laws that constitute this “nature” are freely chosen, we are no longer capable of altering our empirical character. As Richard McCarthy puts it, “. . . we have each freely chosen the characters we now portray in this grand theater of the empirical world . . . . [O]n this cosmic stage we now always act on ‘character’, as dictated by some metaphysical script.”23 To assess this script-­reading we must recall a major point from Chapter I: Kant does not invite speculative philosophy to devise a model of freedom that is at odds with the self-­conception of common moral agents. Rather, for Kant philosophy has the “duty” to save the very idea of freedom that arises in “the practical use of common human reason” (GMS, 4:453: cf. Bxxviii). Common agents who “act under the idea of freedom” (GMS, 4:448) clearly do not take their actions to be “dictated” by a permanently fixed script. Rather, we assume that we are free to choose among different possible alternatives. Likewise, our deliberation often concerns “an immediately forthcoming” particular “free action” (RGV, 6:41) rather than the choice of a henceforth immutable general mindset. To confirm that the script-­reading flies in the face of Kant’s attempt to save our ordinary self-­conception as free agents, note that Kant’s doctrine seeks to defend our moral freedom against the threat of fatalism (GMS, 4:456). A reading on which our empirical character is unalterably fixed has striking fatalistic implications that clash with the presuppositions we adopt regarding moral agents. In holding others accountable for their past actions we deny that their actual empirical character inevitably constrained them to act as they did. Blaming someone for a bad action presupposes that their free causality of reason could and should have led them to behave differently “regardless of all . . . empirical conditions” that make up their empirical character (A555/B583). Likewise, when we deliberate

22  McCarthy 2009: 152–4; compare Watkins 2004: 426–7.

23  McCarthy 2009: xvii.

Human Action as the Effect of Two Causes  65 about what persons we morally ought to be we presuppose that we can step back from our engrained patterns of choice to scrutinize whether these patterns are appropriate; and we attribute to ourselves the rational power to change these patterns in accordance with our awareness of our moral duties. Rational agents cognize their freedom to comply with moral norms regardless of how strong their empirical desires may be and irrespective of how much their previous empirical behavior has flouted moral norms (KpV, 5:30). “It is always in everyone’s power to satisfy the categorical command of morality” (KpV, 5:36).24 Thus, we have the freedom to overcome (however gradually) even the most firmly engrained vices and passions that constitute our past and present empirical character (RGV, 6:41; MS, 6:408, 463). Another problem for the script-­reading arises from its claim that “we are . . . caused to act in character, through the psychological forces of our desires” that make up our fixed empirical character.25 Once we have freely set the script determining which desires are present in us with a specific degree of psychological force, we have no further deliberative control over them: “we cannot freely choose [which of our available incentives will move us to act] in deliberating between courses of action”; instead, “we are . . . free in a noumenal act, resulting in our having the empirical motives we have, and their relative strength.”26 This view conflicts sharply with how Kant characterizes our ordinary practical awareness of empirical desires or feelings (GMS, 4:457–8): insofar as these desiderative states result from our receptive sensible nature, their given presence and psychological force cannot be attributed to our free, active causality of reason at all—­unlike the normative weight we may grant to such states in our free choices, including choices about whether and how we intend to cultivate these states.27 As agents who deliberate under the idea of freedom, we view empirical desires or feelings as sensibly given states that affect and influence but do not causally determine our

24  For McCarthy, the “Ought Implies Can” principle can be falsified via empirical information about the actual character of agents which in his view inevitably determines what they can do (2009: 157–9). This illustrates how strongly his fatalistic account diverges from Kant’s view. 25  McCarthy 2009: xvii. 26  McCarthy 2009: 187. 27  The script-­view also gives an implausible account of the one feeling that is produced by the active causality of reason: the moral feeling of respect. According to McCarthy 2009 (cf. Frierson 2011; Scarano 2002) this feeling is part of the deterministic order of nature and causally explains the occurrence of morally good actions: such actions occur if the feeling of respect is strong enough to overpower conflicting feelings. But a view on which this feeling is a sensible, hence (in principle) discernible motive cause that explains our actions faces two problems. First, it implies that we can know (contrary to Kant’s view: see, e.g., GMS, 4:406, 419) when people act from respect for the law. Second, for Kant any causally efficient natural state must itself be the effect of some natural-­temporal cause (A532/ B560), whereas the feeling of respect is due to the atemporal causality of reason. The attempt to explain our actions via an alleged empirical cause C1 whose causally efficient activity cannot be connected to some temporally prior empirical cause C2 (that empirically determines when C1 begins to cause its effect) would destroy the unity of nature and experience (A542–3/B570–1). Hence, Kant’s account of moral feeling cannot inform an empirical-­causal theory of action. We should instead consider it as part of an a priori moral psychology (cf. Reath 2006). I consider this moral psychology in Chapter V.

66  Kant on Freedom and Rational Agency will when we freely choose our maxims or our physical actions (KpV, 5:117; A534/B562; MS, 6:213).28 McCarthy traces his idea that our freedom is exhausted by the singular choice of an unchangeable empirical character to Kant’s view that our particular choices are based upon maxims, general rules of choice that express a commitment to universal policies: to freely choose a maxim “is to determine oneself to follow such a rule in perpetuity.” The anti-­fatalist claim that we can change our maxims violates their universality: “how is something my policy if I can change it any time I choose, even every time I choose?”29 If we were free to change the maxims that make up our empirical character this would “refute the very idea of a character.”30 By sincerely adopting a maxim one indeed commits oneself to regular patterns of choice that signify universal policies: one shall choose certain kinds of actions under certain conditions for certain reasons. But the universality of maxims requires only that as long as one holds a maxim one is committed to acting in the general fashion that is articulated in that maxim; it does not require that one keeps holding that maxim “in perpetuity.”31 There is nothing incoherent or wanton about adopting a general rule of choice with the implicit understanding that one may give up the rule one day if one sees a good reason for doing so (e.g., when there is a change in the empirical circumstances that informed one’s original adoption of the rule). Likewise, the ideal of a stable character should not rule out a willingness to change one’s dispositions if one recognizes good reasons for this change. As finite moral agents we must acknowledge that our maxims might be based on honest and not so honest mistakes; we must be willing to correct our policies and (thus) to change our character if we discover such mistakes. The moral duties of self-­knowledge and self-­perfection under which we stand as free rational agents entail that we have the power to change our existing maxims and character traits. As Onora O’Neill puts it, “it is a corollary of taking the freedom of agents seriously that maxims are not unchangeable dispositions.”32 A further reason for denying that we can change our existing empirical character derives from the supposition that this would conflict with Kant’s view that empirical laws are permanent and invariable (see, e.g., B5, A536/B564; A766/B794; A798/ B826; Prol, 4:295, 343).33 As Eric Watkins puts it, “whatever grounds and causal laws have held in the past will not change in the future.”34 Since our empirical character consists of law-­governed causal grounds, the invariability of empirical laws seems to entail that a change in our empirical character is impossible. 28  For McCarthy, Kant’s notion of affection without causal determination is incoherent (2009: 69). But it seems perfectly coherent to suppose that a subject is influenced (e.g., tempted or engaged) by some given state without being determined to (re)act by that state in a specific way. See Kohl, forthcoming1 for an account of how this informs Kant’s views on moral education. 29  McCarthy 2009: 80. 30  McCarthy 2009: 164. 31  Here I rely on a rather common understanding of Kantian maxims. See, for instance, the illuminating discussion in Allison 2011: 95–8, 199–200 and in O’Neill 1989. 32  O’Neill 1989: 129–30. 33  See Beck 1960: 191–2. 34  Watkins 2004: 290.

Human Action as the Effect of Two Causes  67 However, we observe that human agents do change their empirical character: they lose or gain interests, alter their habits, fall into or get out of the grip of passions, discard and acquire preferences, etc. Hence, quite apart from issues concerning noumenal freedom, we have empirical evidence for the mutability of the causal dispositions that constitute our empirical character.35 Moreover, there is an ambiguity in the notion that empirical laws are necessary and invariable. This could mean (1) that the same causal powers and laws always remain in operation. But it could also mean (2) that lawful generalizations of the form “If A, then B” cannot be broken: this requires that the same effect B always follows from the same conditions A, but it does not require that these causal conditions permanently obtain. Kant’s remarks on causal laws do not support (1) for every type of causal power or law.36 In his empirical psychology, Kant holds that our desires, affects, and passions are in flux: he treats the psychological tendencies that constitute our empirical character as habitual dispositions that are gradually acquired and may change over the course of time (Anth, 7:286; RGV, 6:47). The moral presupposition that our noumenal freedom may induce changes in our empirical character does not conflict with Kant’s conception of empirical causality because this presupposition does not require a falsification of empirically lawful, invariable “If A, then B” generalizations. We can suppose, instead, that the free causality of practical reason influences our empirical character by enabling, disabling, strengthening, or weakening the causal conditions (e.g., the desiderative mental states) that figure in the antecedent of psychological laws.37 I will consider the cognitive status of this supposition in Section II.4. One might respond that my arguments for the mutability of our empirical character are beside the point. The real issue here, one might claim, is that our intelligible character involves an unalterably fixed bad moral disposition.38 This claim is likely based on Kant’s Religion, which may seem to support the following argument: (1) All human beings adopt a basic life-­governing evil maxim which constitutes our intelligible character, the moral disposition (Gesinnung) of our 35  Based on what Kant says at A549/B577, one might hold that these observable changes concern only the effects of empirical character and do not indicate a change in the underlying basic, “enduring” character itself. However, since we only know basic powers from their effects (KpV, 5:47), we cannot justifiably believe in an immutably uniform basic character if all we discern is variable, irregular external behavior. A weaker, more reasonable claim is that the empirical character of most human beings is relatively stable and enduring; accordingly, the behavior of most human beings shows a relative degree of uniformity. This weaker claim also allows us to accommodate Kant’s point that an absolutely stable character is only an exceptional “rarity” among human beings (Anth 7:291–2). 36  See n. 2. According to (2), there could be causes that produce an effect (e.g., the Big Bang) only on a singular occasion but which are still governed by laws which state general conditions under which these causes would always produce the same effect: it is just that these conditions do not repeatedly obtain. Other causes might be permanently operative because the general conditions under which they lawfully produce certain effects always obtain. Such causes might include the powers investigated by fundamental physics (cf. Watkins 2004: 270). 37  For a somewhat similar response to Watkins, see Indregard 2018. 38  For the worry that the atemporal, thus unchangeable nature of our noumenal character renders true moral progress impossible, see Pereboom 2005: 566; Wood 1984: 97.

68  Kant on Freedom and Rational Agency noumenal selves. (2) This evil intelligible character is immutable because there can be no change in our atemporal noumenal self. (3) Hence, we cannot freely adopt good moral principles. Conclusion (3) is so strikingly at odds with the anti-­fatalistic core of Kant’s ethics that we should be wary about an argument that leads to this conclusion. (3) cannot even be sustained in the context of the Religion where Kant also emphasizes that because we morally ought to overcome our evil character, we must be capable of doing so (RGV, 6:41, 45, 50). Kant here does not merely envisage a reform in our empirical character and behavior, which might occur despite our lasting moral corruption. He stresses that we are morally obligated to become, through our own free powers, “virtuous according to our intelligible character”; and moral duty commands nothing that is not possible (RGV, 6:47).39 Hence, there must be something wrong with the above fatalistic argument. I believe that premise (1) would need to be very strongly qualified before it could be deemed acceptable.40 But since this is a complex topic that cannot be addressed without detailed consideration of the difficult Religion text, I want to focus here on reasons for rejecting (2). We indeed cannot apply our empirical concept of change qua succession of states to atemporal noumenal beings, and we lack an alternative concept of change to represent the idea of a moral revolution in our supersensible character. But this just confirms how deeply our temporal form of sensibility constrains our thinking: when we try to transgress the representational bounds that are set for us by this form, we realize the poverty of our discursive conceptual resources. As Kant (reportedly) puts it, “we cannot take our concepts out of space and time” (V-­Met-­L1, 28:341). Thus, the idea of a moral progress in our intelligible character (or, likewise, the idea of a “relapse” from a good to an evil maxim; RGV, 6:93–4) is incomprehensible to us given the representational limitations that

39  This undermines McCarthy’s claim that while we can gradually reform our empirical character, we lack the power to abandon our intelligible evil maxim (2009: 204–18). Proops suggests that moral progress occurs when “later stages of ” the “empirical self are more steadily inclined than earlier ones to act in accordance with—­and for the sake of—­the moral law” (2021: 324). However, the capacity to act for the sake of the moral law cannot be ascribed to the empirical self: only the noumenal self has the power to determine its will for a priori moral reasons apart from empirical incentives. Hence, the proper subject of real moral virtue and vice, thus also of genuine moral progress or regress, can only be the noumenal self via its non-­empirical character and powers. 40  Kant’s doctrine of human evil raises severe problems which have led many readers to dismiss it as an uncritical attempt to appease Christian doctrine. The two main difficulties are: first, what ep­i­ste­ mic grounds could Kant have for predicating evil of the entire human species? Second, how is the claim that our membership in the general human species fixes our evil moral character compatible with Kant’s central idea (which he strongly affirms in the Religion) that a moral character is freely chosen by each individual finite agent? In my view, these problems can be resolved only if we interpret Kant’s appeal to universal human evil not as a factual claim but as a regulative idea to be employed in the service of our moral progress. I argue for this interpretation in Kohl 2017b. There I also discuss the different interpretations of Kant’s views on radical evil that have been offered by (e.g.) Allison 1990; Frierson 2003; Wood 1999.

Human Action as the Effect of Two Causes  69 afflict our finite intellect. But this subjective incomprehensibility does not entail an objective impossibility (RGV, 6:44–5, 50; cf. Chapter I). This is a general diagnosis, which is not limited to the idea of a moral “change of heart” in our intelligible character (RGV, 6:47–8). Let me mention two further cases where this diagnosis applies. First, when we try to conceptualize God’s eternity we must represent it as existence at all time because without the representation of time we cannot form any concept of the quantity of existence (as duration) (KU, 5:483–4). Although this representation cannot be literally correct (God is not in time), our subjective incapacity to properly represent God’s eternity does not tell against the objective possibility that God is eternal (or has some corresponding property).41 Second, when Kant contrasts noumenal agency with empirical alterations that are subject to “dynamical time determination” (A541/B569), he says that our noumenal faculty of “reason is the permanent condition of all voluntary actions under which the human beings appears” (A553/B581). This statement cannot be literally correct because permanence or endurance is a temporal schema for cognizing phenomenal substances (A143/B183; A186/B229). But it might serve as a legitimate metaphor for representing how a nontemporal being relates to the changing phenomenal-­temporal expressions of its noumenal moral character. As Karl Ameriks suggests, a human agent’s noumenal character might incorporate “a  non-­temporal sequence of qualitative gradations” that expresses itself in a temporal sequence of empirical acts. This qualitative sequence: . . . can express degrees of strength of a noumenal will that indicate various level of possible resistance, within one’s basic disposition, to various stages of temptation, that is, to lower degrees of respect for the moral law. Without being literally temporal, such a will can . . . correspond to, and even be the ground of, a series of temporal alterations in one’s phenomenal existence. One might then say, meta­ phorically, that one’s will or underlying character noumenally endures (that is, in a timeless way in itself) by having strength with respect to a certain range of temptations that life may present, while it is also true . . . that one’s phenomenal character endures in the literal sense of actually expressing this strength in reaction to a range of given temptations.42

If it is legitimate to make a metaphorical use of the concept of permanence or endurance for the purpose of representing our intelligible character, then it is also legitimate to make a metaphorical use of the concept of change for the purpose of representing a moral revolution in our intelligible character.43 Following 41  See Ameriks 2019: 81 and Kohl (forthcoming2). I return to this issue in Chapter III. 42  Ameriks 2019: 82–3 (my emphasis). 43  Moore 2013: 317 also invokes a legitimate metaphorical concept of supersensible change.

70  Kant on Freedom and Rational Agency Ameriks’s helpful lead, we might say that this revolution constitutes a qualitative reversal in one’s intelligible character (namely, an inversion of order among moral and non-­moral incentives in one’s highest maxim) which implies a corresponding difference in one’s empirical dispositions to respond to various phenomenal circumstances in the sequence of time. While our impoverished conceptual resources prevent us from literally representing or adequately comprehending the possibility of a moral revolution in an atemporal being, this subjective incomprehensibility cannot entail an objective impossibility because the real possibility of a morally good revolution is strictly implied by the objective moral command that evil agents ought to achieve it. (However, a quasi-­change in our intelligible character might also have a negative moral quality, as a “relapse” from virtue to vice; cf. RGV, 6:98.)

II.5.  Causes and Standpoints Revisited I have argued that Kant’s doctrine of freedom centrally involves the idea that the noumenal causality of practical reason controls our moral character and observable behavior. I now clarify the status of this idea by showing how it relates to the distinction between an empirical and a normative standpoint that I highlighted as a crucial aspect of Kant’s doctrine in Chapter I. This idea clearly does not belong to the empirical standpoint of theoretical cognition. From this standpoint we attribute observable human actions exclusively to determining natural causes (GMS, 4:453), since we focus solely on the empirical character that human subjects have as objects of sensible experience. Questions about the non-­sensible ground of this empirical character do not arise in this context: “This intelligible ground does not in the least concern the empirical questions” (A545/B573) that we raise, as theoretical cognizers of nature, about phenomenal human actions. Such actions can “be explained perfectly from their causes in appearance, in accord with natural laws” if we focus on “the merely empirical character as the supreme ground of explanation”; in this explanatory-­ theoretical context, “the intelligible character, which is the transcendental cause of the former, is passed over as entirely unknown” (A546/B574). More generally speaking, “in [our] experience no question is ever asked in regard to” non-­ sensible things that transcend nature and experience (A30/B45; cf. A393). The empirical standpoint that cognizes the phenomenal course of nature under natural laws provides the only context where we can legitimately raise explanatory questions and seek to obtain theoretical comprehension: “where determination according to natural laws ceases, there all explanation also ceases” (GMS, 4:459; cf. Chapter I). The idea that the natural laws which govern our empirical character depend on the non-­natural causality of freedom obviously does not lend itself

Human Action as the Effect of Two Causes  71 to explanation via natural laws. Hence, this idea has no explanatory power and must be deemed (subjectively) incomprehensible. The idea that our free intelligible character controls which natural causes empirically determine our phenomenal actions does not belong to the normative standpoint either. While the empirical standpoint ignores free intelligible causes, the normative standpoint focuses solely on our free intelligible character and ignores or brackets the fact that our (freely chosen) spatiotemporal actions are also determined by natural laws. This characterization of the normative standpoint does not signal a lapse into the deflationary view that I criticized in Chapter I, because on my account this standpoint requires a metaphysical belief in the existence (or at least real possibility) of an empirically undetermined causal power. When we adopt the normative standpoint we assume that this free power is the source of our observable actions. The abstract notion that our free causality produces our observable actions by influencing the relevant nexus of empirically determining causes does not and need not occur to us from the normative standpoint. The rational interest that we pursue when we adopt this standpoint does not concern intractable questions about the relation between noumenal and empirical causality. Rather, from this standpoint we seek to answer questions about how we should act and to ensure that our choices conform to our conception of the good. Practical reason “employs the notion of cause not in order to cognize objects . . . [but] only for a practical purpose”; hence, it can “happily admit being unable to comprehend how” the noumenal causality of reason influences the sensible world (KpV, 5:49; cf. A550/B578). Since the conviction that we freely cause our observable actions comes up solely in a context of rational enquiry that discards all purposes of theoretical understanding, we have no rational need to explain or comprehend how the causality of freedom affects the natural causes that empirically determine our observable actions. My claim that questions about the relation between the causality of freedom and the causality of nature are irrelevant for the two standpoints of common human reason is an application of a general principle that I proposed in Chapter I: the beliefs in freedom and natural necessity do not interfere with each other because they are consigned to essentially different fields of enquiry. But if the idea that the empirically determining natural causes of our actions depend upon our non-­empirical causality of freedom does not belong to the two perspectives of common human reason, where does this idea come up? The answer is that this idea is proposed by speculative philosophy when it pursues its “duty” to reconcile the two standpoints of ordinary human reason (GMS, 4:456–7). This reconciliation project requires not only the proof that the beliefs in freedom and natural necessity (which are constitutive of the normative and the empirical standpoint, respectively) are formally consistent. It also requires the defense of a presupposition that crucially informs our ordinary self-­conception as free norm-­governed

72  Kant on Freedom and Rational Agency agents: we assume that our free causality of reason makes a genuine difference to how we act in space and time. To show how this presupposition can be maintained despite the known fact that our spatiotemporal actions are the necessary effects of causally determined natural causes, speculative philosophy first draws on the idealist claim that all natural causes as such have only a relative, (transcendentally) subjective ontological status. Speculative philosophy then adds the further idea that (some of the) specific natural causes that determine our individual behavior might depend on our free intelligible character, although it cannot prove that we in fact possess such a character. If speculative philosophical reason is appropriately self-­conscious of its own cognitive resources and limits, then it cannot pretend that this idea has explanatory power or that we can truly comprehend it. Hence, philosophical reason must leave this idea maximally indeterminate; it must resist the urge to make this idea more concrete or determinate via further speculation about how exactly noumenal freedom influences empirical causes. Moreover, if speculative philosophical reason is appropriately self-­conscious of its legitimate purposes, it also cannot pretend to face or generate a legitimate need for such further determination. Speculative philosophical reason must acknowledge that its idealist doctrine of freedom together with all its corollaries (such as the idea that some specific natural causes causally depend on free powers) serves only one legitimate rational need: namely, the need to defend the integrity and coherence of the two standpoints of common human reason. Since the aim to achieve theoretical explanation can only be legitimately pursued from the empirical perspective but is not itself part of philosophy’s reconciliation project, this project cannot generate the aim to explain anything at all. A fortiori, philosophical reflection cannot reveal a rational need to explain or comprehend the metaphysics of freedom either. Philosophy can only allow us to comprehend the incomprehensibility of freedom (GMS, 4:463): it can provide a systematic investigation of the limits of human reason which demonstrates the necessary futility of all our efforts to understand or comprehend how the noumenal causality of freedom works. My account requires an important qualification: Kant allows that there are special contexts where we may indulge in speculation about how the intelligible causality of freedom works. These are contexts where speculative reason addresses ungrounded theoretical claims that threaten our practical self-­conception as free, morally responsible agents. To counter such claims, we can formulate what Kant calls “hypothes[e]s of pure reason” (A777–8/B805–6). These hypotheses are subject to three constraints: first, they must be logically coherent; second, they must not conflict with our theoretical knowledge; and finally, they must be tailored to our practical self-­conception as moral agents (cf. KpV, 5:133). This is part of what Kant means when he stresses that (pure) practical reason has primacy over theoretical reason (KpV, 5:119–22).

Human Action as the Effect of Two Causes  73 For instance, suppose someone challenges our stance of moral agency via the abovementioned claim that our character is unalterably fixed, or via the claim that through our timeless free agency we cause and are thus morally responsible for all past events (including atrocities like the Holocaust) that empirically precede our phenomenal actions.44 The initial Kantian reply to these challenges should be that they rely upon determinate theoretical assumptions about the causality of freedom that have no proper cognitive basis: since we cannot theoretically cognize supersensible causal powers, the theoretical claim that such powers produce an unalterable character or indefinitely extended chains of past events transcend the limits of human cognition. The challengers might respond that their assumptions yield the only way in which we can conceive how the causality of freedom operates. Even if that were true, it would not prove that the causality of freedom must conform to our thoughts: supersensible causal powers might operate in ways that we simply cannot capture through our finite conceptual resources. Moreover, the challengers’ assumptions do not yield the only coherent view of how free powers might operate. We can instead hypothesize that the exercise of a free power is a complex atemporal activity which produces effects like the gradual decrease or increase of those empirical states (e.g., passions and affects) that figure in the antecedents of psychological laws; this is how the free causality of reason might influence which specific laws govern our empirical character at a given present time.45 But we must be clear about the impoverished cognitive status that these and similar conceptual ruminations have in Kant’s system. The only proper use of such hypotheses lies in defending common moral reason against external (e.g., fatalistic) challenges. Apart from this strictly defensive purpose, the conceptual models we may devise when we “revel in the intelligible world” (GMS, 4:462) are pointless; in particular, they cannot inform a viable metaphysical theory of noumenal freedom. Speculative philosophy makes room for the belief that the empirical causality of nature depends upon a free noumenal causality, but it cannot thereby illuminate the ontological structure of that dependency relation: “how . . . one has to conceive [a free] form of causality theoretically and positively is not thereby comprehended” (KpV, 5:133).

44  See Walker 1978: 149 and Bennett 1974 for the claim that Kant’s appeal to a timeless cause of phenomenal actions has this implication. Some concede that Kant’s view has this implication but then try to show (unsuccessfully, in my view) that it does not threaten our ideas of moral responsibility (Proops 2021: 305–6; Vilhauer 2010). 45  One might raise the further challenge that if the laws governing my psychology are the same as the laws governing the psychology of others, then freely influencing these laws entails influencing the actions of others (an implication that would undermine our practices of moral imputation). However, here one can reply via the coherent (even empirically plausible) hypothesis that because there are innumerably many variations in the (genetic, biographical, etc.) circumstances of individual people, psychological laws differ for specific agents (at specific times of their development). Hence, by changing the laws of my individual psychology I do not necessarily cause others’ actions.

74  Kant on Freedom and Rational Agency

II.6.  Conclusion: Expanding the Interpretive Framework In Chapter  I, I proposed an interpretive framework for understanding Kant’s doctrine of freedom. According to that framework, the cornerstone of Kant’s doctrine is his claim that we can save the belief in human freedom by abandoning a realist conception of natural causality and by attributing our observable actions to two different types of causes which we represent from two different standpoints. In this chapter I have added two further components to this interpretive framework. First, I have argued that Kant can save the relevance of our noumenal freedom by supposing that the specific natural causes which empirically determine our observable actions depend on our free causality of practical reason. This supposition can be fruitfully combined with the argument that empirical predictions of how we will act cannot preempt the spontaneous self-­control that we exert over our agency and character traits through norm-­governed practical reasoning. Accordingly, Kant also rejects the fatalistic claim that our moral character is unalterably fixed by a singular noumenal choice: Kant’s view allows that both our empirical and our intelligible character may change (or “change”) as the result of how we exercise our noumenal freedom. Second, I have argued that we cannot and should not try to elaborate on the supposition that our empirical character depends on our noumenal freedom: this supposition must remain indeterminate because we cannot comprehend how the causality of freedom operates on the empirical world. This supposition also does not call for any further elaboration or determination (aside from the special contexts noted above) because the question of how the noumenal causality of freedom causes its empirical effects does not concern the two standpoints of common human reason. These two standpoints demarcate the only two contexts of enquiry in which human reason can legitimately use the concept of causality (either for empirical-­ explanatory or for normative-­ practical purposes). Hence, common human reason is absolved from the need to tell some incomprehensible story about how the causality of freedom affects the causality of nature and our empirical characters. Speculative philosophical reason cannot and need not come up with a theory of noumenal causation either: it is charged only with the “duty” to prevent a clash between the two standpoints of common human reason and their respective beliefs in the causalities of nature or freedom.

III Freedom as Autonomous Self-­Determination Thus far I have expounded the metaphysical core thesis in Kant’s doctrine of freedom: our observable actions can be regarded as free since they can be traced, from a normative standpoint, to a spontaneous type of causality. As we saw, Kant “happily” concedes that we can never comprehend or explain how the noumenal causality of freedom causes its effects. But surely Kant’s doctrine requires that we can meaningfully represent, i.e., think and judge about or refer to, our noumenal freedom as a unique type of causal power. In this chapter I examine how Kant seeks to meet this semantic requirement. As we will see, his intricate account of how we can adequately conceive our noumenal freedom has significant systematic implications. Among other things, it reveals a twofold motivation for his anti-­ naturalistic incompatibilism.

III.1.  Noumenal Freedom and the Limits of Discursive Representations On a common reading, Kant’s pure (unschematized) categories are ontological concepts whose intellectual content allows us to represent non-­sensible beings. Accordingly, we can make meaningful judgments about the atemporal causality of freedom simply by using the pure category of causality. We can only apply the categories to the sensible objects of our theoretical cognition by employing a temporal schema such as “succession of states according to a rule” for “causality.” Without these sensible schemata, the categories cannot yield theoretical cognition of objects. But the unschematized categories enable us to form meaningful thoughts about non-­sensible things we cannot cognize, such as noumenal freedom. This common interpretation has a strong appeal, not in the least because it can draw on textual support (see, e.g., B166).1 But its claim that the pure categories suffice for meaningful thought about noumena runs against Kant’s frequent in­sist­ence that apart from their relation to our sensibility the categories are 1 The core of this interpretation is accepted—­notwithstanding many important variations—­in: Abaci 2019; Adams 1997; Adickes 1924; Allais 2015; Ameriks 2000a, 2019; Baum 1986; Guyer 2010; Hogan 2009b; Jauernig 2008; Langton 1998; Stang 2016; Watkins 2002.

Kant on Freedom and Rational Agency. Markus Kohl, Oxford University Press. © Markus Kohl 2023. DOI: 10.1093/oso/9780198873143.003.0004

76  Kant on Freedom and Rational Agency “empty” concepts devoid of sense and meaning (see, e.g., A239–40/B299; B307–9). Proponents of the common interpretation respond that with such statements Kant only stresses, in hyperbolic fashion, that the unschematized categories have no meaning as theoretical cognitions:2 the pure category of causality lacks the determinate spatiotemporal content that we need for formulating causal laws, explanations, and predictions. Nonetheless, we can use this concept to form the abstract but perfectly meaningful thought of free noumenal causes as unconditioned grounds that produce certain consequences. In my view, this response does not accord with Kant’s critical doctrine of categories.3 Since the pure categories originate in a discursive intellect, their content essentially refers to a discursive manner of representation. More precisely, the categories derive from logical forms of judgment: for instance, “causality” arises from the hypothetical form of judgment which puts two propositions into a logical ground-­consequent relation. All logical forms are functions for combining different concepts or judgments into one complex, unified representational whole (A68–9/B92–4; Prol, 4:305). This essentially combinatory meaning of the logical judgment-­forms is inherited by the pure concepts derived from them: they are conceptual functions for combining given, uncombined manifolds of sensible intuition into one complex, unified representation of an object (A241–2; A349; B300–1; JL, 9:121–2). When we abstract from our human (spatiotemporal) mode of sensibility, we do not thereby free the categories from their discursive meaning as forms for combining and conceptualizing given data into cognitions of sensible appearances: a purely intellectual category represents “a logical function . . . to make a concept from . . . datis” (A239), where the specific character of these datis is left indeterminate. The unschematized categories are “nothing but forms of thought which contain merely the logical capacity for a priori uniting the manifold given in intuition” (B306); “they contain nothing other than the unity of reflection on appearances” in one empirical consciousness (A310/B367). They are empty forms, devoid of sense and objective meaning, because they are divorced from their complimentary partner: some specific, externally given, uncombined sensible manifold which provides the occasion as well as the material content for “the unity of reflection on appearances.”4 2  See Adams 1997: 806–9; Ameriks 2019: 146; Langton 1998: 18; Watkins 2002: 199. 3  My argument in what follows is an abbreviated version of the account I give in Kohl 2015a; Kohl forthcoming2. 4  Thoughts that use the categories without reference to a sensible manifold yield a mere “play” lacking in objective validity (A239/B299). Ameriks (2019: 146–7) argues that this play “is still within the context of a broadly objective framework.” However, Kant stresses that “without the data of sensibility” the categories are “merely subjective forms of the unity of understanding, but without object” (A287/B343). Ameriks suggests an analogy between the pure categories as forms of thought and the moral law, which does have objective validity despite its merely formal status. But the moral law has a formal normative content that qualifies as objective precisely because it abstracts from all sensible conditions: this is what ensures its applicability to all rational beings (see Section III.3 below). By contrast, since the pure theoretical categories are forms of unity that essentially refer to some sensible manifold

Freedom as Autonomous Self-Determination  77 Although the pure categories are independent of our human spatiotemporal sensibility, they refer indeterminately to (the combination of) some sensible manifold or other: they extend to objects of intuition in general, where this intuition “may be similar to our intuition or not, if only it be sensible and not intellectual” (B148; cf. B150). As concepts that represent “nothing but” formal rules for synthesizing uncombined sensible manifolds (A287/B343), the pure theoretical categories seem unsuitable for meaningful thoughts about non-­ sensible objects. Kant’s explicit denial that the categories extend to objects of an intellectual intuition is especially significant here. Intellectual intuition is the self-­active mode of cognition that would pertain to an infinite, omniscient divine intellect. This is the only mode of cognition that could cognize the constitution of noumenal objects in a positive, determinate fashion.5 The pure categories represent “nothing but” functions of synthesis: they “contain exclusively the synthetic unity of apperception” (B148; cf. A119; A138/B177). Since an intuitive intellect does not synthesize given intuitions, it is not subject to the principle of the synthetic unity of apperception (B137–8). The need to unify given data in one consciousness would not arise for “an understanding which should cognize its object, not discursively through categories, but intuitively in a non-­sensible intuition” (A256/B312). Hence, the representational content of the pure categories would not occur in an intellectual intuition that has an immediate, gapless, all-­encompassing cognition of noumenal objects.6

to-­be-­unified into one experience (GMS, 4:452), their objective reality is grounded “solely” on the fact that they “constitute the intellectual form of all experience”; as “functions of synthetic unity” they “have no objective meaning at all” unless they can be applied to some manifold of given intuition (A310/B367). 5  Because everything represented through sensible intuition is only an appearance, one could cognize even oneself as a subject in itself only if one possessed a non-­sensible intellectual intuition (B69; B153). Access to this mode of intuition would be required to form a positive conception of noumena (B306–7). Since a self-­active intuitive intellect is not limited by conditions of sensibility and does not need to synthesize given data, it “immediately intuits all objects as they are in themselves” (V-­Th/Pöl, 28:1052). 6  Marshall (2018: cf. Abaci 2019: 255–6) suggests that a divine mind would represent categorial features indirectly, by directly representing the non-­categorial features upon which categorial features “supervene.” But it is hard to get a grip on the supervenience-­relation envisaged here, especially since the fuzzy notion of supervenience itself seems to require the categorial representations (e.g., substance, quality) whose (noumenal) reality it is supposed to explain. The appeal to an indirect manner of divine representation is also problematic. It is obscure how concepts that are “nothing but” forms of synthesizing sensible data could adequately characterize, however indirectly, non-­sensible objects of a non-­synthesizing intellect. Moreover, a merely indirect mode of (re)presentation is peculiar to a finite, discursive intellect that lacks an immediate, complete grasp of reality and must therefore perform a special act of attention to get from the directly represented to the indirectly represented (e.g., from sensations to magnetic matter; A226/B273). In Kant and his tradition, indirect cognition is typically mediated by a logical act of reason, i.e., by an inference (A303–4/B359–60; Willaschek 2018: 29–36). Thus, the claim that a mind M represents some actual feature A only indirectly implies that M cannot immediately (non-­inferentially) grasp all that is actual. By contrast, an infinite divine mind that faces no representational limitations “must cognize all things . . . immediately at once” (V-­Th/Pöl, 28:1051). What it does not immediately-­directly cognize does not belong to noumenal reality.

78  Kant on Freedom and Rational Agency In particular, Kant denies that the pure modal categories “contingency” and “necessity” could occur in the representation of an intuitive intellect (KU, 5:402–3). If we regard our free will as a capacity for choosing between different options that are all genuinely open to us, we conceive our choices as contingent: we could choose (or have chosen) differently than we in fact do (did). Thus, the denial that contingency could pertain to noumena (as they would be positively cognized through an omniscient intuitive intellect) raises a deep problem for our self-­conception as free noumenal agents. So does the denial that necessity could pertain to noumena (as divinely intuited). The pure category of causality analytically involves the thought of necessity: our concept of a cause represents a ground from which a consequence must follow (B5). Thus, if our conceptual representation of necessity cannot pertain to noumena as cognized by an omniscient intuitive intellect, then our concept of causality cannot pertain to noumena either.7 A noumenal causality thus seems unthinkable for us. Hence, Kant’s view that the categories are essentially (“nothing but”) discursive forms for synthesizing sensible data threatens our ability to conceive our non-­ sensible noumenal freedom. To solve this problem, we must consider Kant’s conception of freedom as autonomy.

III.2.  Freedom as Autonomy: The Practical Concept of Noumenal Causation This conception derives from Kant’s view that freedom is essentially a law-­governed capacity. Every efficient cause, natural or free, “must . . . have a character, i.e., a law of its causality, without which it would not be a cause at all” (A539/B567). A lawless causality whose exercise is not governed by some universal rule is an absurdity, literally a “non-­thing” (Unding) (GMS, 4:446). Recall (from Chapter I) that we affirm our freedom when we adopt the normative “standpoint” from which “a rational being [considers itself] as belonging to the intelligible world,” as an intelligence.8 When a subject considers itself “as an intelligence,” it views itself “not from the side of its lower faculties” but from the side of its higher rational faculties (GMS, 4:452). Our lower faculties characterize 7  Abaci argues that only the modal categories are inapplicable to noumena (2019: 261–2). But since the other categories—­most obviously, the relational ones—­have irreducibly modal content, Abaci’s view that Kant is an amodalist concerning noumena affects the other categories as well. Abaci even claims that Kant defines transcendental freedom in non-­modal terms (2019: 268). However, since Kant defines transcendental freedom as an unconditioned cause that sufficiently determines a series of effects (A446/B474), this definition must involve the modal content that is analytically contained in the pure category of causality (B5)—if this category is indeed (as Abaci supposes) the same concept that Kant uses when he conceives free will as a causal power. 8  For the link between “intelligence” and “intelligible world/being” see, e.g., MS, 6:226; V-­Met-­L2, 28:583; cf. Puls 2015: 184–5. In what follows, I will draw (but significantly expand on) Kohl 2020b.

Freedom as Autonomous Self-Determination  79 us as passive beings that receive sense-­impressions from without. We have no active rational control over the exercise of these faculties: “The lower faculties cannot be instructed, because they are blind” (V-­MO/Col, 27:244). By contrast, the exercise of our higher faculties is a free rational activity that can be instructed since it is governed by laws which provide normative standards of correctness: an intelligence “cognize[s] laws for the employment of its powers and all of its actions” which “are independent of nature,” “founded in reason alone” (GMS, 4:452).9 Normative laws governing free powers cannot be externally imposed upon intelligences: a subject whose laws are imposed on it by foreign (e.g., natural) causes is not a truly spontaneous intelligence but an automaton which has only the “freedom of a turnspit” (KpV, 5:97). A normative law “founded in reason alone” is a law of autonomy that does not depend on conditions outside of the subject’s rational self-­control. It expresses an objectively-­universally valid normative standard that we give to ourselves for the proper exercise of our shared ra­ tional capacities.10 Our recognition of such a law is thus inseparable from our self-­conception as free rational beings: it “expresses a deep commitment from one’s ‘true’ nature as a rational and . . . free agent.”11 Hence, for Kant freedom as a law-­governed power is fundamentally a capacity for autonomous rational self-determination. As absolutely spontaneous intelligences “we are . . . determined by nobody, but determine ourselves” (V-­Met/Mron, 29:902–3) to act under self-­legislated laws that yield standards of correctness for our free rational agency. Although (as we shall see in Part 3) Kant’s notion of a spontaneous-­autonomous intelligence does not refer exclusively to the moral subject of free will, it has a predominantly moral connotation. A spontaneous intelligence is a practical intelligence when it possesses a higher faculty of will as the causality of a rational being. Given the link between causality and lawfulness, the causally efficient activity of a free practical intelligence must be governed by some law. This cannot be a law of nature, because for any power that is governed by natural laws “something else determines the efficient cause to [its] causality” (GMS, 4:446). Since the law governing a free rational will cannot be externally imposed on the will by any foreign causes, this law cannot be received by free agents from a super-­natural designer (“puppeteer”; KpV, 5:101) either. Instead, it must be a self-­given law of autonomy that derives “solely” (GMS, 4:452) from “universal practical reason” (GMS, 4:431): a faculty that is shared among all rational agents apart from the contingent external conditions that distinguish them from one another (e.g., apart from the natural conditions that determine which sensible desires different

9  On the identification of non-­empirical with normative laws, see Chapter I n. 66 and Kohl 2018a. 10  For helpful discussion of the link between autonomy and objective-­universal validity and of Kant’s point that we make autonomous laws for ourselves, see Ameriks 2019: 19–28, 58–9, 99–102, 113–17. 11  Hill 1992: 111.

80  Kant on Freedom and Rational Agency agents happen to have).12 The law of practical autonomy is the moral law. It yields a normative standard for what counts as a right, good, wrong, or evil volition or physical action. Thus, for Kant free will is a capacity for autonomous self-­determination under moral laws of practical reason. But how does this help with the problem that the causal power of free volition seems unthinkable for us since our pure theoretical concepts of contingency, necessity, and causality qua discursive forms of sensible cognition are inapplicable to our supersensible free will? The crucial point here is that the idea of free will as autonomous moral self-­ determination is not limited to beings with discursive intellectual capacities. The moral law that positively “expresses” our freedom as autonomy (KpV, 5:33) applies to all rational beings as practical intelligences, including a divine being as the “highest intelligence” (KpV, 5:32). Hence, the moral law would also be represented by a divine intellect which is not limited by discursive forms of cognition. Since the moral law governing a free will transcends all discursive forms of sensible cognition, our pure idea of the moral law enables us to adequately represent our supersensible free will. To explain how this works, I focus on Kant’s remarks in the second Critique about how pure practical reason can legitimately extend the categories beyond the bounds of sensibility that constrain the theoretical use of reason. The moral law is a rule of causality that posits the determining ground of the will beyond all conditions of the sensible world (KpV, 5:50). It is a rational norm which determines how we ought to and (thus) can choose regardless of what our sensible desires or feelings (neurological brain states, etc.) happen to be. Thereby, the moral law allows us to cognize the subject of a free will as a rational member of a non-­sensible, intelligible world (KpV, 5:50–1): it gives us cognition of our noumenal causal power to act from a purely intellectual motive. By considering ourselves subject to this law, we can thus get a positive grip on our ability to cause certain effects in a physically unconditioned manner, namely, via our purely rational appreciation of the normative standard governing the proper exercise of our free will. Kant’s account of how we can obtain this transcendent cognition of our free noumenal causality begins with a point whose importance is rightly stressed by the common interpretation (cf. Section III.1): since the categories originate in the pure understanding independently of all sensible intuition, they are not limited to phenomenal beings (see KpV, 5:55; cf. B166). The common interpretation takes Kant here to provide a sufficient condition for meaningful thought about non-­ sensible objects such as noumenal freedom. As I argued in Section III.1, this reading is hard to square with Kant’s repeated claim that as discursive forms of

12  As Reath puts it, the moral law “originates in the nature of rational volition per se” (2006: 104).

Freedom as Autonomous Self-Determination  81 thought the pure categories cannot extend beyond objects of sensible intuition.13 But Kant’s emphasis on the non-­sensible origin of the categories can also be taken in a weaker sense, as expressing a necessary condition for meaningful thought about noumena. Since our finite intellect needs the categories to form any thought whatsoever, we cannot think about our noumenal freedom without using the categories (better yet, anticipating my argument below, without employing some representation that is formed from a categorial basis). In the second Critique, Kant stresses—­as he does in the first Critique (A91/B123–4; A128) and the Prolegomena (Prol, 4:260, 312–13)—that unless the categories have an intellectual origin, they must be dismissed as invalid fictions: if the categories derived from empirical sources (as Hume claims), the concept of causality would be “false and deceptive” (KpV, 5:51) and could not yield valid representations of any, sensible or noumenal objects (KpV, 5:54; cf. KpV, 5:141).14 Since the categories have (as the first Critique shows) a purely intellectual origin, they are suitable for meaningful thoughts about all objects in general (KpV, 5:54). But this does not entail that the pure theoretical categories are on their own sufficient for the meaningful representation of noumenal objects. Considered in isolation, they are not meaningful representations of any objects at all. The categories can meaningfully represent objects only if they are supplemented with a further source of representational content. This can happen in two ways. First, the categories acquire a theoretical meaning that enables the representation of phenomenal objects if they are supplemented with forms of sensible intuition. Second, they acquire a practical meaning that enables the representation of noumenal objects if they are supplemented with the rational idea of the moral law.15 In his notes, Kant states that both our forms of sensibility and the moral law function as sources of objective content and meaning for the categories, adding: “There is only this difference here: in theoretical cognition the pure concepts have no meaning . . . except with respect to objects of experience, whereas in the practical they extend much further, namely to all rational beings in general” (Refl., 19:275). Likewise, in the second Critique he argues that in our representation of

13  Allison 2020: 391 also rejects the sufficiency-­reading. 14  See Kohl 2018c for discussion of Kant’s argument in relation to Hume’s empirical derivation of the categories. 15  Allison also stresses that a further source of representational content is needed for the objective meaning of the categories in their practical use, but in his view this content stems from the manifold of sensible desires which practical reason unifies via the form of the moral law. From this, Allison concludes that Kant’s account of positive judgments about noumena hinges “largely on an assumed parallelism between the claims of theoretical and practical reason,” which Allison deems “forced” (2020: 395). I agree that this parallelism is contrived, but I read Kant as holding that the moral law itself directly supplies the requisite objective-­practical content for positive categorial judgments about noumena without any reference to a synthesis of sensible manifolds. If the moral law owed its fundamental objective content to its role as a form for unifying sensible manifolds, then it would lose its advantage over the theoretical categories (cf. GMS, 4:452): it would be inapplicable to moral subjects (like God) who are not sensibly affected and do not synthesize any manifolds. See Bittner 1983: 151.

82  Kant on Freedom and Rational Agency noumenal freedom the moral “categories of freedom” take over the role of the “theoretical concepts as categories of nature.” The categories of freedom have an “evident advantage” over the theoretical categories: the latter, as “determinations of the synthetic unity of the manifold of given intuitions,” are “merely forms of thought which signify merely indeterminately objects in general for every intuition that is possible for us” (KpV, 5:65; cf. KpV, 5:103–4). Hence, the pure theoretical categories cannot signify noumenal objects of a non-­sensible intellectual intuition that is impossible for us (B148–50). The categories of freedom are exempt from this limitation since they do not depend for their objective content on “the form of intuition (space and time), which does not lie in reason itself but must be taken from elsewhere, namely from sensibility.” Instead, they draw their objective content from an a priori law of practical reason that determines the normative “form of a pure will” (KpV, 5:65–6). The moral law thereby gives a “positive determination” to what remains “a merely negatively thought causality” for theoretical reason (KpV, 5:48). Here Kant draws on his contrast between the negative and the positive concept of noumena. Through the pure theoretical categories we can think things in themselves as non-­sensible correlates of sensible appearances, but this representation is wholly negative and indeterminate (B306–7; A286/B343): the theoretical categories enable only “the thought of something in general outside our sensibility” (B307) which “is not indeed in any way positive” (A252). For instance, we can represent a physically unconditioned, non-­temporal causality as a cause which is not determined to its activity by some antecedent state or event (KpV, 5:48). But this thought has the grave limitation that “we cannot through any example make comprehensible to ourselves what sort of thing is to be meant by such a concept” (A241/B300); the pure theoretical category of causality does not even enable us to “distinguish cause and effect from each other” (A243/B301). Hence, the merely negative content that we can attach on a purely theoretical basis to the concept of a physically unconditioned cause is not sufficient for the meaningful representation of any real causal power or causal relation. A “merely negatively thought causality” does not suffice to represent a causal power because thoughts that contain only negations do not signify anything at all: “Negation signifies a mere want, and, so far as it alone is thought, represents the abrogation of all thinghood” (A575/B603). Accordingly, if we could not view our free will in positive terms as having a determinate law-­governed character that produces specific types of effects, we would have to regard free will as a non-­thing (GMS, 4:446). To grasp “what is meant” by the concept of an unconditioned cause, we must “attach meaning to this concept” through the idea of the moral law so that “the concept does achieve meaning . . . for a practical use.” The “meaning provided [for the pure concept of causality] by reason through the moral law is merely practical” (KpV, 5:49–50; cf. KpV, 5:56).

Freedom as Autonomous Self-Determination  83 Kant’s claim that the moral law provides a distinctively practical meaning for the pure concept of causality implies that the concept lacks this meaning inde­ pend­ently of its relation to the moral law. Under the plausible assumption that a change in meaning involves a change of concept, it follows that the practical concept of an unconditioned cause signifies a new concept of causality. This new concept arises when the moral law is applied to the pure theoretical category of causality. This application enables an a priori-­practical transformation of the discursive-­logical content contained in the pure theoretical category. This discursive content signifies a ground-­consequent relation where the consequence follows necessarily from the ground (JL, 9:106): the ground is a sufficient reason for the consequent, so that given the ground the consequent cannot but follow. When the idea of the moral law is applied to this logical content (i.e., to the hypothetical form of judgment), the resulting conception of causality retains the idea of a sufficient rational connection between a ground and a consequence. The moral law supplies a sufficient normative-­justifying reason for performing a certain action which renders that action rationally speaking necessary. For a free rational will, its consciousness of this rational-­practical necessity can suffice as an intellectual motive ground for good choices (KpV, 5:25; MS, 6:393). But the practical transformation of the theoretical category transcends the strict logical-­inferential link between ground and consequence: it ejects the thought that given the ground the consequence cannot but follow. When our free rational will is conscious of how it morally ought to act, this consciousness does not inevitably lead to right action (GMS, 4:412–14; KU, 5:403–4; MS, 6:222). Thus, when we think of our free will as governed by the moral law, we represent it as a causal condition that can suffice for producing a certain effect, namely a good volition, without rendering that effect inevitably necessary. This designates the true constitution of our noumenal will that would also be represented by an omniscient intuitive intellect. A divine intellect is conceived as the just judge of our intelligible character—­as such, it must grasp (via its intellectual intuition) the complete intelligible moral disposition of finite agents (KpV, 5:123). The divine judge can justly distribute rewards and punishments according to our moral character only if it cognizes that we qua finite agents have imperfectly rational powers that are (on the one hand) sufficient for good agency but (on the other hand) do not inevitably determine us to good agency since they leave us with the option to make wrong choices. This account resolves the problem that an omniscient divine cognition of noumena does not involve modal content. Kant’s denial that divine cognition has modal content (at KU, 5:402–3) concerns only the theoretical modalities that we represent through our purely discursive modal forms of thought. These are second-­order concepts that designate the different ways in which our first-­order theoretical concepts of objects relate to the various cognitive faculties that we

84  Kant on Freedom and Rational Agency exercise when we synthesize given data into a complex whole of sensible experience (A234–5/B286–7).16 Theoretical modalities rest essentially on the split between concepts and given intuitions as two distinct stems of discursive cognition. This split forces upon our intellect the distinction between the possible (represented by the understanding through mere concepts), the actual (given through empirical intuition), and the necessary (thought by reason when it connects the actual with a determining sensible ground) (KU, 5:401–3). These theoretical modalities have no place in the cognition of a non-­discursive divine intellect whose spontaneous intuition does not face the need to integrate two difference sources of cognition. But such an intellect does cognize normative-­practical modalities: a holy or perfectly rational will represents the absolute normative-­ rational necessity or unconditional goodness that we also grasp (in imperatival form, as an ought) through our idea of the moral law (GMS, 4:412).17 One might object that although deontic noumenal modalities belong to divine intuition, this does not show how noumenal freedom could be divinely represented since such freedom requires alethic modalities—­e.g., the modality that we invoke when we say that a human agent could have done otherwise.18 However, in Kant’s view alethic noumenal modalities arise from deontic modalities. Since deontic modalities are based on practical laws of reason, they can be cast in terms of reasons for action: if an action-­type is rationally-­practically necessary there is a sufficient reason to choose relevant action-­tokens. It follows that all rational agents who act according to their representation of practical laws (cf. GMS, 4:412) have the power to perform such actions, the “power to satisfy the categorical command of morality” (KpV, 5:36–7). This grounds the alethic modality that it is possible for rational agents to act rightly due to their awareness of sufficient practical reasons for right action.19 16  For helpful discussion of Kantian theoretical modalities as second-­order concepts, see Abaci 2019, ch. 7. 17  One might claim that a divine intelligence does not represent deontic modalities since it does not distinguish between a practical ought and a theoretical is (KpV, 5:403–4). However, even though a divine intelligence does not represent the moral law governing its agency in imperatival ought-­form, the self-­representation of such an intelligence would involve non-­imperatival deontic terms such as absolute goodness or rational necessity (cf. GMS, 4:412; I discuss this issue in Chapter  VIII). Moreover, since an omniscient intelligence qua just judge of moral character would surely cognize the factual rational imperfection of other, e.g., human agents, it must also grasp that for those agents there is a real gap between what is rationally necessary or good and what is actual: it must grasp the actual versus the merely (but really) possible good and evil dispositions of finite, accountable noumenal wills. 18  See Stang 2016: 305–6; cf. Abaci 2019: 259–60. 19  Stang (2016: 316) seems to think that we can replace modal talk about possibilities through the idea of noumenal power. However, the category of causality and all its Prädikabilien, such as “power” (A82/B108), have irreducibly modal content. Kant defines power (Kraft) as something that contains the sufficient real ground for the actuality of some effect (V-­Met-­N/Her, 28:26–7). Thus, the belief that agents have a power to produce effect x just amounts to the belief that they can produce x all by themselves. Accordingly, Kant equates the idea that ought implies can (possibility) and the idea that moral actions are always within our power (KpV, 5:36–7; RGV, 6:49).

Freedom as Autonomous Self-Determination  85 Deontic modalities yield the (sole) basis for representing the non-­natural noumenal modalities of necessity and possibility required for free human agency. Practical imperatives articulate an “ought” which “expresses a species of necessity and a connection with reasons which does not occur anywhere else in the whole of nature.” The ought further “expresses a possible action, the ground of which is nothing other than a mere concept” (A547/B575). This “mere concept” is the intellectual representation of a normative-­practical necessity, i.e., of sufficient practical reasons for choosing an action just because it is good. For finitely ra­tional agents, their imperatival awareness that an action is practically necessary “expresses” only a possible action since they do not inevitably act in conformity with their awareness of conclusive practical reasons. Here we should also note that the various deontic “categories of freedom” are modes of just one transcendent category, namely of “causality” insofar as this concept signifies a supersensible free will governed by the moral law (KpV, 5:65). This means: the various deontic categories specify various action-­types, e.g., morally obligatory acts, as potential effects of a free rational causality. Thereby, they jointly determine what a rational causa noumenon (KpV, 5:55) is a capacity for. This enables us to conceive free agency in terms of determinate cause-­effect relations (whereas, as we saw, theoretical categories fail to signify such relations outside the sensible domain). By specifying which action-­types are (e.g.) rationally prohibited or required, the deontic categories signify corresponding alethic modalities, e.g., possible performances of obligations or omissions of prohibited actions as characteristic effects of free causation. My account here further confirms and develops the central point that our concept of noumenal causality is theoretically incomprehensible (cf. Chapters I–­II). We conceive our norm-­governed free will as a cause that is (when properly exercised) sufficient for a certain effect, namely a good volition, without making the occurrence of this effect inevitably necessary; the very same cause might also (when improperly exercised) lead to the contrary effect, namely a bad volition. This practical conception of free causation makes it impossible to explain why a particular effect, e.g., a morally bad rather than a good action, results from the exercise of our free will. Kant’s claim that the moral law provides a positive but only practical meaning for the concept of an unconditioned cause thus neatly corresponds to his view that the concept of free will can be used only from a normative standpoint that brackets theoretical purposes. Accordingly, when he sketches how the moral law gives the concept of freedom “objective . . . if only practical reality” he stresses that practical reason cannot represent in a theoretically intelligible manner how the ground-­consequence relation applies beyond sensibility (KpV, 5:49). As we saw (in Chapters I–­II), for Kant this poses no problem since practical reason “happily admits” being unable to comprehend what the practical notion of a causa noumenon contributes to

86  Kant on Freedom and Rational Agency theoretical cognition: practical reason does not and need not aspire to provide a theoretical meaning for that notion (KpV, 5:49–50). In sum, Kant’s conception of free agency as autonomous self-­determination under the moral law allows us to adequately conceive our free noumenal causality. Understood in a positive sense that designates not merely (negatively speaking) the absence of determination by foreign causes, freedom is the capacity of an intelligence to determine itself under self-­legislated laws of reason (V-­MS/Vig, 27:494). Since the practical idea of spontaneous self-­determination under the moral law is not limited to a discursive intellect, this practical idea provides an objective content or meaning for the concept of a free cause that extends beyond the bounds of discursive, sensibly restricted object-­representation. It thereby gives us a determinate, positive practical cognition of our supersensible freedom as a power of rational-­moral self-­determination.

III.3.  Divine versus Human Freedom My account of how we can adequately conceive our noumenal free will is based on a twofold claim. First, free rational agency is governed by the moral law. Second, this moral law applies to all free rational agents including a divine being whose representational powers are not confined to the phenomenal world of sense. This may seem to suggest that human freedom of will is essentially the same as divine freedom. But in this section I show that for Kant human and divine freedom yield two distinct species of free agency. If Kant did believe that human agents exhibit the same type of freedom as God, he would be following Leibniz. For Leibniz, three conditions are individually necessary and jointly sufficient for divine freedom: intelligence, spontaneity, and contingency. Since human minds are mirrors of God, human agents satisfy these three conditions, albeit to a lesser degree than God.20 These conditions play a central role in Kant’s account as well. For Kant the notion of an intelligence is coextensive with the notion of a free agent: only an intelligence and every intelligence is the subject of “higher” rational faculties whose exercise is governed by laws of freedom. Kant’s view that laws of freedom must be laws of autonomy implies a radicalized version of the Leibnizian view with regards to spontaneity. Kant distinguishes between a merely relative spontaneity (spontaneitas secundum quid) and an absolute spontaneity (spontaneitas simpliciter talis) and he associates the relative notion with the Leibnizian view (KpV, 5:96–101; V-­ MS/Vig, 27:503–5; V-­ Met-­ L1, 28:268). This relative notion applies when actions are based on an inner principle that is not within the agent’s control because it is “implanted” (Br., 10:131; B167–8) in the agent by a foreign cause. If 20  See Leibniz 1988: §288. Compare Adams 1994: 11; Jolley 2005: 5; Proops 2021: 314.

Freedom as Autonomous Self-Determination  87 the inner principle governing our volition were implanted in us by nature or God, our will would only have the “freedom of a turnspit.” Like a turnspit or clock, which moves on its own once it has been set up by its designer, our actions would be directed by externally given principles that are beyond our rational self-­control. This lack of genuine freedom would not be redeemed by the fact that the determining causes of our actions are representational rather than material states (KpV, 5:96–7). While this signals a central difference between Leibniz and Kant, it does not show that Kant also rejects the Leibnizian idea that human freedom is the same as divine freedom: Kant might accept a modified version of this idea by holding (as he indeed does) that both divine and human freedom involve an absolute, unconditioned spontaneity. Kant’s reason for denying the Leibnizian assimilation of human and divine freedom emerges only once we consider the third item in the Leibnizian trifecta, contingency. It is not easy to appreciate this point, because Kant’s view contains some tensions as far as the relation between freedom and contingency is concerned. Consider here the following two passages: The idea of freedom has its place solely in the relation of the intellectual, as cause, to the appearance, as effect. We [cannot] find any concept of freedom to fit a purely intelligible being, e.g., God, insofar as His action is immanent. For His action, although independent of causes determining it from outside, nevertheless is determined in His eternal reason, hence in the divine nature. Only if something should begin through an action, hence the effect be found in the time series, and so in the sensible world . . . does the question arise of whether . . . the concept of this causality is a concept of natural necessity . . . [or] of freedom. (Prol, 4:344) There is no difficulty at all in uniting the concept of freedom with the idea of God, as a necessary being: because freedom does not consist in the contingency of the action (that it is not determined through reasons), i.e., not in the indeterminism (that it be equally possible for God to do good or evil, if his action is to be named free), but in the absolute spontaneity, which faces a risk only with regard to predeterminism, where the determining ground of the action is in the preceding time, indeed in such a way that now the action is no longer within my power but in the hand of nature [which] irresistibly determines me; this difficulty now vanishes since in God no time series is to be thought.  (RGV, 6:50)

The first (Prolegomena) passage claims that we have no concept of freedom that fits a divine being because its actions would be determined “in His eternal reason” and would thus be absolutely necessary.21 The second (Religion) passage claims 21  In the Prolegomena, Kant says that we cannot find a concept of freedom for divine action “insofar as that action is immanent.” One might suggest that this allows us to regard as free at least the

88  Kant on Freedom and Rational Agency we can unproblematically think of God as free because we can conceive divine agency as rational self-­determination that is at once absolutely spontaneous and absolutely necessary. Thus, the first passage affirms while the second passage denies that contingency is essential to transcendental freedom. One might suggest that since the first passage is from 1783 and the second passage is from 1794, their conflict reflects a change in Kant’s view. But this is implausible. In Collins’s notes on Kant’s ethics lectures, which are usually dated to 1784–5, Kant already affirms the point of the Religion passage that divine actions are both absolutely necessary and free (V-­MO/Col, 27:267). In notes that belong to roughly the same period, Kant sometimes defines freedom (in the vein of the Religion passage) as the capacity to be only determined by reason (Refl., 18:181) but also explains (more in the spirit of the Prolegomena passage) transcendental freedom as the “complete contingency of the actions” (Refl., 18:183). Some commentators pass over this ambivalence and claim that the Religion passage commits Kant to the view that free will as such, not just divine freedom, is a form of necessary noumenal (rather than temporal pre-) determination by reason that does not involve contingency or the ability to do otherwise.22 There are two separate issues at stake here. First, proponents of this view must claim that in cases where human agents act as they are rationally required to, they (just like God) lack the option to act differently, i.e., wrongly. I argue against this claim below and (on systematic grounds) in Chapter V. Second, proponents of this view must claim that no contingency or alternative possibilities are present in cases where human agents act wrongly. But for Kant, phenomenal pre-­ determinism threatens our freedom (in part) because it implies that when I act immorally (e.g., tell a lie) the morally required action or omission “is no longer within my power” (RGV, 6:50; cf. KpV, 5:94–5). If my immoral actions were noumenally determined (e.g., by God), the result would be the same: it would not be within my free power to do as I ought. Thus, whether we are phenomenally pre-­determined or noumenally determined to wrong actions, our moral freedom and accountability are fatally impugned in either case. This also shows that the attempt to take divine freedom as a model for free agency per se fails (partly) because in the case of God as a perfectly rational being, the issue of imputing rationally deficient bad actions does not arise. By contrast, legitimately imputing bad actions to imperfectly ra­tional human agents presupposes a noumenal cause “which . . . could have determined the conduct of the human non-­immanent divine act of creating the world. Student notes on Kant’s lectures indeed contain the idea that God creates the world according to His freedom, in contrast to His inner actions that follow from the necessity of His inner nature (V-­Th/Pöl, 1092–7; V-­Th/Baum, 28: 1298–9). But even this externalist conception of divine agency fails the Prolegomena definition according to which a free action must have a sensible external effect that begins in the time series. We cannot regard the sensible world as coming into existence at a certain temporal moment, a first beginning (A427/B455ff.; see also KpV, 5:102 for the denial that God creates the sensible world). 22  See Abaci 2019: 207; McLear 2020; Pereboom 2005: 11.

Freedom as Autonomous Self-Determination  89 being . . . differently” (A555/B583) and whose operation must therefore involve noumenal contingency.23 In my view, the seeming conflict between the abovementioned passages derives from Kant’s attempt to accommodate two different models of free agency. As we saw, he defines free agency as the capacity for spontaneous rational self-­ determination. Given this definition, the absence of contingency cannot count against free agency per se. A divine being is inevitably determined by its own spontaneous reason to right or good actions; it thereby satisfies the conditions of autonomous rational self-­determination par excellence. Kant calls divine freedom the highest kind of freedom precisely because it involves an absolute moral necessity which is not in conflict with but rather arises from absolute rational spontaneity (V-­Met/K2, 28:806; V-­Th/Pöl, 28:1068; V-­Th/Baum, 28:1280–1). This is the point of the Religion passage. But now consider a subject that, like a divine being, has the capacity to act in accordance with objective self-­legislated norms of right reason but is also under the influence of non-­rational sensible conditions. For this subject “the actions which are recognized as objectively necessary are subjectively contingent” (GMS, 4:412–13).24 To be sure, the contingency of right action cannot enter into the definition of human freedom: a real definition must demonstrate the “objective reality” of a thing (A241–2), which requires that only positive determinations be included in the definition since negations abrogate realities (A575/B603). Freedom positively consists in the power to act in accordance with rational laws. Hence, the possibility to deviate from rational laws, i.e., our rational imperfection which renders our compliance with rational norms contingent, is not a positive capacity of its own but a mere want of a capacity, a “privation” that cannot figure in a definition of free agency (MS, 6:226–7). 23  Abaci (2019: 207) claims that the contingency affects only the phenomenal effects of our noumenal volitions, not these volitions themselves. However, the contingency of effects entails the contingency of their cause. If a rational agent A performed a morally bad phenomenal action, it was absolutely possible for A to omit this action by exercising their free noumenal will differently: the freely chosen evil character of a noumenal will is avoidable and therefore contingent (KpV, 5:100; RGV, 6:32). The phenomenal effects of an evil noumenal will (e.g., a lie and its consequences) are imputable only because they originate in a noumenal cause “which . . . could have determined the conduct of the human being . . . differently” (A555/B583); this requires that the noumenal cause itself is not inevitably determined to a particular mode of operation. McLear 2020 suggests that transcendentally free actions are causally determined by the agent’s own noumenal powers. But again, my moral freedom does not fare any better if I am causally determined by some aspect of my noumenal nature to violate moral norms than if I am causally pre-­determined to act wrongly by temporal states of nature. McLear does acknowledge a connection between transcendental freedom and the possibility to do otherwise: he concedes that the judgment that A ought not to have lied presupposes that A could have avoided their actual lie. But I do not see how this is consistent with the idea that A’s noumenal character causally determines A to lie—­unless the notion of free noumenal self-­determination is here understood in a way that allows for contingency and alternative possibilities, in sharp contrast to the inevitably necessary noumenal self-­determination of divine transcendental freedom. 24  Notice Kant’s language here: these actions really are subjectively contingent, namely, contingent for finitely rational subjects like us. He does not say that we merely subjectively represent them as contingent.

90  Kant on Freedom and Rational Agency Nonetheless, the possibility of deviation can be a necessary component of free agency (albeit one that is ontologically parasitic on the rational powers that are positively definitive of free agency) if the agent necessarily stands under the influence of conditions that entail this possibility. This is “really the case with human beings” who are necessarily affected by sensible conditions, by needs and inclinations that tempt us to violate moral norms so that our finite will “according to its nature” does not necessarily act from objectively necessary practical reasons (GMS, 4:412–13). The above Prolegomena passage does not explicitly invoke our rational imperfection as the reason why our concept of freedom cannot apply to divine agency. But it draws on a closely related point: it contrasts the “immanence” or self-­ sufficiency of divine agency qua “determined by His eternal reason” with the fact that our free agency aims towards external sensible effects. The free exercise of our will is always directed towards some external telos that lies outside the free exercise of our rational capacities properly speaking: even when we have done everything that is fully within our rational control, i.e., when we have chosen the right course of actions for the right reasons, we may still fail to perform the chosen action or to produce the intended effects (cf. GMS, 4:394). Whether we act externally in the way we internally intend and whether we produce all and only those physical changes we intend to produce depends upon vagaries of the empirical world such as our bodily state, which are always to some extent unforeseeable and beyond our rational control (cf. Chapter I). Our lack of self-­sufficiency affects our free agency with an additional layer of contingency: beyond the fact that it is contingent (because it is up to us) whether we choose for the right reasons. it is also contingent (because it is not wholly up to us) whether we succeed in realizing our freely chosen physical ends. Although our rational imperfection and our lack of self-­sufficiency generate two distinct layers of contingency, these layers are closely connected because they share a common root: they are both grounded in the fact that we exercise our free will under conditions of sensible affection. These conditions inform the content of the specific moral norms that govern our finitely rational agency. Our narrow, perfect duties prescribe the limits within which we can legitimately strive to realize the external-­empirical ends that appeal to us due to our sensible needs. In that way, narrow duties seek to curtail our tendency to transgress the bounds of practical reason when we pursue external things that we desire as rational creatures. Even when morality itself positively commands us to pursue certain ends, these ends depend on our constitution as imperfectly rational, sensibly affected agents lacking self-­sufficiency. The duty to promote the highest good is based upon our need to combine our rational-­moral vocation with our sensible desire for happiness. Our wide moral duties to pursue more concrete ends (such as promoting the happiness of others or cultivating our talents) also depend partly upon our sensible nature and rational finitude, albeit in a more indirect way: given our

Freedom as Autonomous Self-Determination  91 tendency to transgress moral demands when we strive to realize external ends that concern the satisfaction of our sensuous desires, our rational-­moral vocation requires the countervailing influence of external (“material”) moral ends the pursuit of which our practical reason legislates as wide moral duties. “Since the sensuous incentives incite towards ends (as the matter of the faculty of choice) which may be contrary to duty, the law-­giving reason can fend off their influence in no other way than, again, through an opposing moral end” (MS, 6:380–1). Hence, for Kant human and divine freedom yield two different species of free agency. While they both fall under the same generic notion of (transcendental) freedom as autonomous, absolutely spontaneous self-­determination, they instantiate this notion in different ways. Put in Leibnizian terms, divine and human freedom both involve intelligence and (absolute) spontaneity, but only human freedom involves two layers of contingency which both arise from the fact that we exercise our free will under conditions of sensible affection. By contrast, the free agency of a divine being is entirely unaffected by other things. It is therefore entirely unafflicted with any need or incentive to pursue external ends that lie outside the sphere of its immanent self-­governance qua absolutely necessary rational self-­determination.25 These differences between divine and finitely rational freedom have a significant upshot: we cannot apply our concept of a will and our practical concept of a noumenal cause to a divine being in any literal sense. Since our will is a faculty whose desiderative representations are intentionally directed towards and (at least potentially) productive of non-­actual objects or states of affairs (KpV, 5:15), it is essentially a capacity for striving to obtain things that we desire under the guise of some rational representation. We cannot at­trib­ute this faculty to a divine being because the concept of striving to obtain what is not yet actual cannot properly characterize a self-­sufficient, omnipotent divine agent—­ this concept signals a need for things which one still lacks, and only finite agents (rational creatures) are afflicted with needs and desires for things they lack. Our concept of will designates a practical capacity for producing the non-­actual in contrast with our theoretical capacity for merely cognizing (rather than producing) the actual. This contrast collapses for a divine being (KU, 5:403–4; V-­Met/K2, 25  Here we must again (cf. n. 21) note the complications arising from the non-­immanent divine act of creation which is directed towards an external end. Student lecture notes show Kant wrestling with this problem. Why does a self-­sufficient being create something outside of itself (V-­Th/Pöl, 28:1060)? It seems that to comprehend this we must attribute to God some kind of interest or incentive, but this is inconsistent with God’s absolute perfection (V-­Met/K2, 28:780–2; V-­Th/Pöl, 28:1060, 1065; V-­Th/ Volck, 28:1201–4; V-­Th/Baum, 28:1279). Kant sometimes suggests that we can attribute to God only something analogous to our concept of an interest or desire (V-­Th/Volck, 28:1279; A700/B728). He also entertains the different suggestion that God’s motive for creating things outside of Himself is His self-­sufficiency, namely, the non-­dependent pleasure deriving from His self-­consciousness as a self-­ sufficient ground of all possibility that possesses the power to produce the best of all possibility (V-­Th/ Pöl, 28:1061–2, 1100–2; V-­Th/Volck, 28:1204; V-­Th/Baum, 1257–77). I confess that I do not truly understand this suggestion.

92  Kant on Freedom and Rational Agency 28:803, V-­Th/Pöl, 1054–5; V-­Th/Baum, 1272). Accordingly, many concepts that are essential to our practical self-­consciousness as freely striving agents are inapplicable to a divine being, e.g., “imperative,” “incentive,” “interest,” or “maxim” (KpV, 5:79). Through these concepts we represent ourselves as de­pend­ent subjects who form their volitions in response to both given sensible needs and self-­imposed rational tasks. We empirically desire our happiness as a non-­actual state of our will which requires non-­existent external things. We rationally desire our virtuous moral character as a non-­actual state of our will which requires an infinite moral progress and a never-­ceasing struggle against corruption. If we abstract from these features and consider the immanent agency of a self-­sufficient, morally perfect being that has no needs and lacks in nothing, we lose any (literal) grip on our concept of a will qua capacity for rationally striving towards the non-­actual (V-­Met/K2, 28:780–2, 798; Refl., 18:444). Since our positive-­practical conception of noumenal causality is inherently tied to our idea of our free will as a potential cause of non-­actual effects, the fact that we cannot extend our idea of free will to God entails that we also cannot literally conceive divine agency in causal terms. Instead, we must resort here to an analogical, symbolical mode of representation. “I can think myself God as the cause of substances only per analogiam” (V-­Met/K2, 28:805); we can conceive a divine will, desire, or causality only through a “subtler” (i.e., legitimate) anthropomorphism (A700/B728; cf. KU, 5:464–5, 482–5).26 The practical concept through which we cognize (in a literal, non-­analogical manner) our noumenal causality signifies a cause-­effect relation that pertains only to finitely rational agency: while our free will is sufficient for producing morally good volitions, such volitions do not follow with inevitable necessity from our finitely rational choices. Moreover, a finite will does not necessarily produce the external (physical) effects it intends to produce since these effects are not fully under the control of the agent’s free rational striving. When we try to conceive a perfectly self-­sufficient being whose good actions flow with absolute inevitable necessity from its inner rational nature without being directed towards any external non-­actual effects to-­ be-­produced, we lose our grip on the distinction between cause and effect as two separate elements in a causal relation. I draw a twofold lesson from my discussion in this section which unifies the two passages cited at the beginning. First, we can positively conceive divine freedom (without resorting to merely analogical concepts) by considering God’s absolutely spontaneous, necessary conformity to self-­legislated rational standards. This is the point of the Religion passage. But (secondly) this yields only a 26  I discuss this issue at greater length in Kohl (forthcoming2). For Kant’s view that we can theoretically represent a highest being through the categories only “according to an analogy with the objects of experience,” see A696/B724. Furthermore, we can practically represent divine freedom as a cause only according to an analogy with the a priori practical (but non-­analogical) idea of our own finite will as a noumenal cause.

Freedom as Autonomous Self-Determination  93 highly generic conception which does not capture the kind of freedom that we ascribe to ourselves as finite agents. Our sense of freedom in inseparable from our awareness that we are members of both the intelligible and the sensible world (GMS, 4:453; cf. KU, 5:404): we are sensibly affected, dependent and striving rather than self-­sufficient agents; likewise, we are imperfectly rational, obligated creatures whose right actions are only ever contingent. Hence, the (practically) determinate idea of freedom under which we act designates a finite, non-­divine species of free agency. This is the point of the Prolegomena passage.

III.4.  Wille and Willkür Kant’s conception of free agency involves yet another tension that has important systematic upshots. In the first Critique Kant refers to free “actions through which [reason] prescribes laws” (A803/B831). This rational activity “yield[s] us laws” as “products of pure reason” which “are pure and determined completely a priori” because they “are prescribed to us not in an empirically conditioned but in an absolute manner” (A800/B828). In its prescriptive activity reason “frames for itself with absolute spontaneity an order of its own according to ideas . . . according to which it declares actions to be necessary” (A548/B576). But when Kant formulates his distinction between Wille and Willkür (“faculty of choice”) in the 1797 Metaphysics of Morals, he stresses that only the Willkür’s choice of maxims can be called free: the prescription of moral laws by the will qua practical reason “cannot be called either free or unfree, because it is concerned not with actions but immediately with the legislation for the maxims of actions” (MS, 6:226). It is doubtful whether Kant’s denial that the legislative-­prescriptive activity of practical reason qualifies as free agency expresses his considered view. In his Vorarbeiten for the Metaphysics of Morals we find him wavering, denying (VAMS, 23:248) and affirming (VAMS, 23:383) that the legislative will is free since it is undetermined by foreign causes.27 The unequivocal denial that the practical legislation of moral laws qualifies as a free action would have some dire consequences.28 One casualty would be the fact of reason-­doctrine according to which the moral law is self-­given by a legislative deed of reason, an autonomous Tathandlung that is not externally determined.29 The unconditioned, purely spontaneous formation of judgments expressing fundamental principles of moral autonomy must be a free action in some sense. In the Groundwork, Kant stresses that the freedom 27  See Allison 2020: 452–5 for helpful discussion of these and similar textual complications. 28  These consequences are overlooked by those (like Beck 1960: 172, 185; Bojanowski 2006: 242–4, 261) who deny that pure practical reason has any effects apart from its power to determine Willkür (in morally good choices). As we saw, Kant explicitly refers to moral laws as “products of pure reason.” 29  The systematic importance of the idea that the moral law is self-­given by a deed of reason is emphasized by Allison 2020; Ameriks 2019; Kleingeld 2010; Willaschek 1991, 1992.

94  Kant on Freedom and Rational Agency of practical reason consists in it not “receiving a bias from any other quarter with respect to its judgments” so that reason can “regard itself as the author of its principles independent of foreign influences” (GMS, 4:448). For these reasons, I do not believe that the Metaphysics of Morals passage unequivocally expresses Kant’s considered view. But we can make good sense of the spirit, if not the exact letter, of what Kant says here. I see an informative link between Kant’s ambivalence about whether our practical legislation is a free action and his ambivalence about whether a divine being is a free agent. In both cases, Kant recognizes that our ordinary self-­conception as free agents calls for a different notion of freedom: neither divine agency nor practical legislation exhibit the type of freedom that we represent when we exercise our Willkür in deliberative acts of choice which are directed towards external ends. Such deliberative activity yields the paradigm for our common practical experience of freedom. By contrast, both practical legislation and divine agency are entirely immanent actions: they have no external telos that lies outside the inner self-­determination of reason by reason. The practical legislation of reason “concerns nothing other than merely the law” (MS, 6:226): it is directed solely towards the internal self-­ representation of rational standards for good agency. Moreover, in its practical legislation the power of practical reason is as irresistible and inevitably determining for our acts of practical judgment as it is without exception for all divine agency. We cannot struggle against the rational self-­activity that generates our consciousness of the most fundamental moral standards: the moral law “forces itself upon us,” “whatever inclination may say to the contrary” (KpV, 5:32); the voice of reason here is “irrepressible” (KpV, 5:35). Thus, in the prescriptive act of legislating basic moral standards we are inevitably determined from within by our own spontaneous reason, just like a divine being inevitably determines itself through its own spontaneous reason in all its activity. The inevitable self-­determination by practical reason ends for finite agents like us with the mere consciousness of moral norms, which does not inevitably determine us to morally right choices. We need to be rationally necessitated to choose what is right and good. The concept of “rational necessitation” designates the force of reason in relation to sensible constraints or hindrances that inveigh against reason and thereby prevent the influence of rational considerations from being inevitably determining. But the presence of such obstacles to inevitable rational determination affects only our contingent acts of free choice: our inevitably necessary acts of moral legislation are not similarly afflicted. Consequently, the notion of rational necessitation does not apply to such acts: “the legislation for the maxims of actions . . . is . . . absolutely necessary and capable of no necessitation” (MS, 6:226). This further confirms that the immanent, absolutely necessary act of moral legislation does not exhibit the kind of freedom that we ascribe to ourselves when we deliberate and choose under the complimentary, contingency-­ inducing conditions of sensible affection and rational necessitation.

Freedom as Autonomous Self-Determination  95 Hence, we can interpret Kant’s claims in the Metaphysics as Morals as emphasizing that our ordinary self-­conception as free agents concerns acts of deliberative choice where we exercise our Willkür.30 Kant’s philosophical focus on this ordinary sense of freedom derives from his aim to define “preliminary concepts for the Metaphysics of Morals” and to find a concept of free “deed” that suits our ordinary conception of moral and legal accountability. The immanent act of legislative reason that provides us with a consciousness of the moral law is a necessary (perhaps also sufficient) precondition for moral and legal accountability, but we are not morally or legally accountable for this act because it is an absolutely necessary deed that “forces upon us” an “irrepressible” consciousness of how we are obligated to choose. Praise, blame, punishment, and other expressions of accountability concern our contingent free choices of morally good or evil maxims and the external behavior guided by these maxims (MS, 6:223). Kant’s conception of practical legislation entails that we partake in the divine model of free agency, albeit only to a strictly limited extent. A divine being has no separate faculty of Willkür for the choice of maxims. It is determined by reason with absolute necessity all the way through, apart from any need for rational necessitation. Likewise, our legislative will qua practical reason determines us, or better: we rationally determine ourselves to the absolutely necessary legislative deed that yields the moral law. But this is where the inevitably determining influence of reason ends and the need for a necessitating influence of reason begins for us. Whether or not we conform to the normative standards of our moral reason hinges on how we exercise a further volitional capacity, our Willkür, which is subject to conditions that inveigh against the moral law.31 Kant has another piece of terminology for capturing this difference. In its practical legislation our will acts as “pure reason,” as a “capacity of principles (and, here, of practical principles, indeed as a law-­giving capacity)” (MS, 6:214). By contrast, our Willkür is “affected, but not determined through [sensuous] incentives, and is thus in itself [. . .] not pure but can be determined through actions from a pure will” (MS, 6:213)—namely, when we choose in accordance with pure 30  Kant’s distinction between (1) the will’s legislation for maxims and (2) the Willkür’s choice of maxims raises a problem for intellectualist readings (Engstrom 2009; Reath 2006) which identify the legislative act of making a valid practical judgment with the choice of a maxim. (This textual issue aggravates philosophical problems that arise for such readings; for instance, it seems that we can legislate moral norms without intending to choose a maxim, e.g., when we make moral judgments about past actions or about the actions of other people.) Proponents of the intellectualist reading might respond that Kant’s distinction between (1) and (2) entails a gap between legislation and choice only in one special case, namely, with respect to the one fundamental legislative act that concerns the general moral law (which, due to its completely formal nature, cannot be adopted as a maxim all on its own). 31  I do not regard will and Willkür as ontologically separate faculties. Kant indicates that there is a conceptual and a functional difference between the two when he says: “Laws arise from the will, viewed generally as practical reason; maxims spring from Willkür” (MS, 6:226). This does not imply that we could have a will in isolation from a Willkür or vice versa. Here I agree with Allison 1990: 130–1 and Beck 1960: 190, who both emphasize that will (in the narrow, legislative sense) and Willkür are two interconnected aspects of one faculty (human will in a wide sense).

96  Kant on Freedom and Rational Agency self-­legislated moral laws. The legislative act of reason is “pure” because it is not even affected by non-­rational sensuous incentives that always influence our Willkür.32 Likewise, Kant also characterizes a divine will as entirely “pure” because it suffers no sensible affection. Accordingly, a divine being does not exhibit the kind of freedom that is characteristic of our Willkür: the freedom peculiar to a sensibly affected, imperfectly rational power of choice.

III.5.  Conclusion: Two Sources of Incompatibilism In this chapter I have examined how we can form a positive conception of our noumenal freedom of will. We can form this conception based on our a priori idea of the moral law which positively “expresses” our free will as the normative standard governing a supersensible type of causality. Since this moral law also governs divine agency, it may seem as if our freedom is the same as divine freedom. However, on closer inspection it turns out that Kant canvasses two distinct models of moral freedom qua spontaneous rational self-­determination. In each model, absolute spontaneity and intelligence are necessary conditions of free agency. In the first model, the absolutely spontaneous intelligence is inevitably determined by its own reason to certain acts. In the second model, the absolutely spontaneous intelligence is affected by sensible conditions whose influence inveighs against reason. This provides subjective hindrances which make it contingent whether or not the intelligence acts in accordance with right reason. A divine being would instantiate the first model in its entire agency. We partake in the first, divine model through our pure, sensibly unaffected acts of reason that legislate moral norms. We instantiate the second model in our sensibly affected acts of free choice that are always governed by but may either comply with or flout moral norms.33 I shall use the terms “legislative freedom” and “executive freedom” to designate the two kinds of free will.34 Since both terms designate absolutely spontaneous

32  See GMS, 4:457–8, where Kant also distinguishes pure (sensibly unaffected) reason that “gives” the moral law from the faculty that is responsible for choice, whose spontaneity is affected (not determined) by sensible inclination. 33  “Being governed by moral laws” does not mean “compliance with such laws.” An immoral choice is imputable because it violates its governing standard of correctness and the subject’s Willkür is negatively free (undetermined by foreign causes) in making this choice. Contrary to Reinhold or Sigdwick, Kant never claims that only morally good choices are free. In his phrase that “a free will and a will under moral laws are one and the same” (GMS, 4:447), the notion of standing under moral laws is broader than the notion of actually complying with such laws. For instance, Kant holds that in a civil state the citizens “stand under punitive laws” (TP, 8:290) even (indeed, especially!) when they violate these laws. Many commentators have argued that the Wille-­Willkür distinction (which Kant, arguably, implicitly uses long before the 1790s; cf. Willaschek 1992: 52) allows Kant to make room for free and imputable wrong choices: see Allison 2020: 456–65; Ameriks 2019: 26–34; Bojanowski 2006: 257; Wuerth 2014, ch. 7. 34  This choice of terminology is inspired by Allison 1990 and Beck 1960.

Freedom as Autonomous Self-Determination  97 capacities that are not determined by foreign causes, both executive and legislative freedom can be regarded as species of transcendental freedom in the negative sense of that term. Since both legislative and executive freedom are intimately connected to the positive law of moral autonomy (legislative freedom gives that law; executive freedom is governed by or “stands under” that law), the positive concept of transcendental freedom can also be applied to both kinds of free will.35 I conclude this final chapter of Part 1 with a suggestion that can serve as a bridge for Part 2. The distinction between executive and legislative freedom enables us to distinguish between two conceptually different threats that arise from the supposition that “all causality . . . [is] mere nature” (A534/B562). To adopt this supposition is “to make principles of possible experience conditions of the possibility of things in general” (A781/B809; cf. Chapter I). One such condition is the principle that everything is hypothetically necessary (A226–30/ B279–83): every event, state, or action is the necessary result of a preceding cause whose productive activity is in turn the necessary result of some further preceding cause, and so on in an indefinitely extended chain of natural-­hypothetical necessity. If all existing causality were of the natural type, this would have the following implications: hypothetical necessity is the only real (non-­logical) form of necessity; everything (every state and event) is the result of hypothetical necessity; and, all efficient causes operate according to the principle of hypothetical necessity so that all causal activity always depends on the causal activity of some further, foreign causal condition (A542/B570). Call these naturalistic suppositions (of “dogmatic empiricism”; A466/B494). Under these suppositions, we would lack both legislative and executive freedom of will. But naturalistic suppositions rule out these two kinds of free will for two different reasons. Naturalistic suppositions rule out the possibility of executive freedom because they remove the absolute contingency and thereby the metaphysically real alternative possibilities that are characteristic of such freedom. Under naturalistic presuppositions, our actions (as natural changes) are contingent only in a relative or empirical, not in an absolute sense. The difference between these two senses is as follows (cf. A460/B488; B290–1). If my action A1 at time t1 is absolutely contingent, then its contradictory opposite is possible, i.e., it is possible for me not to 35  There is some dispute over the extension of this concept. What is beyond dispute (see Allison 1990; Beck 1960; Bojanowski 2006; Willaschek 1992) is that a morally bad choice of Willkür does not display autonomy or positive freedom. The dispute arises over the question of whether the positive concept of freedom as autonomy applies to the will in the narrow sense qua legislator of laws (Beck 1960: 199–200) or merely to the determination of Willkür by the pure legislative will (Bojanowski 2006: 256–7; this is also the recent view of Allison 2020: 454–5, though he admits that this is only what “Kant should have said” rather than what he actually said). To avoid a reification of the Wille-­ Willkür distinction (i.e., to avoid the impression that these terms pick out two really distinct faculties), we should perhaps say that the one faculty of will as a whole (in the broad sense of “will”) is positively free or autonomous, by virtue of its interconnected capacities for pure rational self-­legislation and for rational choices in accordance with self-­legislated laws. (This is how I understand Allison’s earlier view: cf. 1990: 132.)

98  Kant on Freedom and Rational Agency perform A1 at t1. If A1 at t1 is only relatively contingent, this means: A1 could not have occurred at t1 without some preceding natural cause; but this does not show that the non-­occurrence of A1 at t1 was also (really, metaphysically) possible. Indeed, the fact that A1 at t1 was determined to happen by some preceding causal condition makes A1 at t1 not only empirically contingent but also empirically (hypothetically) necessary. Since the causal condition upon which the occurrence of A1 at t1 is contingent depends upon some further condition, etc. etc., it follows that our empirically contingent actions are necessarily determined by an indefinite chain of conditions whose temporally distant members pre-­date our own existence (V-­MS/Vigil, 27:505). By contrast, the absolute notion of contingency signifies that an action is not conditioned or determined to happen by any preceding states or events. Thus, if all our causal capacities stood under the principle of natural-­hypothetical necessity, the exercise of our Willkür would lack absolute contingency: we would always be constrained by an indefinitely extended chain of preceding natural conditions to make a particular choice at a particular temporal moment (A534/B562). Naturalistic suppositions rule out our legislative freedom because they rule out the special absolute necessity that is required for legislative acts of will. The legislative “actions through which [reason] prescribes laws” (A803/B831) are “absolutely necessary” (MS, 6:226): they occur “not in an empirically conditioned but in an absolute manner” (A800/B828). Such absolutely-­unconditionally necessary actions would be impossible if all our actions were exclusively subject to natural necessity. In the natural order of things, the only real necessity is the empirically conditioned (hypothetical) necessity of actions relative to some temporally preceding cause (A228/B280); everything in nature is empirically conditioned and therefore contingent upon further, foreign natural causes (A560/B588). Thus, in the natural order of things no action and no product of any action is absolutely-­ unconditionally necessary. If everything that exists or occurs were exclusively subject to the empirical conditions that govern the natural order of things, there would be no absolutely necessary actions. Consequently, there would be no absolutely necessary practical judgments either, because such judgments can only exist as “products of pure reason” (A800/B828), as resulting from empirically unconditioned acts of purely rational self-­legislation. Hence, for Kant the assumption that all our actions are exclusively the result of natural causality raises two conceptually different worries for our free will. The first worry concerns the naturalistic implication that we lack executive freedom of will because our acts of choice lack absolute contingency. The second worry concerns the naturalistic implication that we lack legislative freedom of will because our prescriptive acts of reason lack absolute necessity. This invites the question: why should we think that the ideas of absolute contingency and absolute necessity are integral to our self-­conception as free moral

Freedom as Autonomous Self-Determination  99 agents? Why could we not capture our legislative and executive sense of moral freedom through the merely relative, empirical notions of contingency and necessity that can be applied within a naturalistic framework? This is just another, but more informative and fruitful way of posing the central questions which arose at the end of Chapter  I: What motivates Kant’s metaphysically ambitious, anti-­ naturalistic conception of free will? What justifies his rejection of naturalistic compatibilism? My account in this chapter shows that Kant’s anti-­naturalistic conception of free will has a twofold motivation. In the next two chapters I reconstruct the two lines of reasoning that support this motivation. I examine why Kant holds that our self-­conception as free moral agents requires both the idea of absolute necessity (with respect to our legislative freedom of practical reason; Chapter IV) and the idea of absolute contingency (with respect to our executive freedom of Willkür; Chapter V).

PART 2

T HE GROU N D S OF KA NT ’S INC OMPAT IBI LI SM A B OU T F R EE W I LL

IV Legislative Freedom and Kant’s Genealogical Anxiety At the end of Part 1, I suggested that Kant’s twofold model of free agency implies a twofold motivation for his view that there could be no human freedom if our actions were (exclusively) the effect of natural causes. In this chapter I examine the first part of this motivation: I consider Kant’s view that on a naturalistic world­view we would lack the legislative freedom that secures the unconditional necessity of our practical-­moral judgments since naturalism implies that such judgments are empirically conditioned and (therefore) contingent. This is one way in which naturalism threatens an integral component of our moral self-­conception. Kant’s view here relates in interesting ways to contemporary genealogical debunking arguments. Amia Srinivasan suggests that such arguments are driven by genealogical anxiety: “We suffer from genealogical anxiety when we worry that the contingent origins of our representations, once revealed, will somehow undermine or cast doubt on those representations.”1 In the course of this chapter I will compare Kant’s genealogical anxiety with contemporary views about how naturalistic evolutionary theory invalidates our moral beliefs.

IV.1.  Origin and Validity I am assuming that for Kant, the origin of our representations can bear upon the validity of those representations. This assumption is controversial.2 On a common reading, Kant’s frequent claim that moral principles derive from our pure autonomous reason does not concern their genesis; rather, Kant seeks to stress only that our moral judgments cannot be justified by appeal to external sources such as nature, religion, or tradition. Christine Korsgaard argues that there is “nothing” in Kant’s view that “requires any ontological claims” about the origins of our normative judgments.3 One can endorse these judgments as valid even while conceding that one’s reasoning is controlled by an electronic device implanted in one’s brain.4 Daniel Warren points out that “many commentators” like Graham Bird or 1  Srinivasan 2019: 127. 3  Korsgaard 1996a: 183.

2  But it is accepted, for instance, by Henrich 1989: 35–6. 4  Korsgaard 1996a: 162–3.

Kant on Freedom and Rational Agency. Markus Kohl, Oxford University Press. © Markus Kohl 2023. DOI: 10.1093/oso/9780198873143.003.0005

104  Kant on Freedom and Rational Agency Henry Allison “stress the irrelevance of investigations of origin to questions of justification”; Warren agrees that as far as a priori concepts are concerned, their origin is indeed irrelevant to their justification.5 This approach follows a venerable tradition: prominent nineteenth-­century Neo-­Kantians saw it as a hallmark of Kant’s philosophy that he distinguished sharply between the two issues. According to Alois Riehl, “the question as to the origin and development” of rational principles “has not” much if any “importance for the critique of knowledge.”6 For Wilhelm Windelband, Kant declared “the fundamental difference between ‘origin’ and ‘justification.’ ”7 Windelband argues that those who overlook this difference succumb to a faulty psychologism.8 This has inspired the now common view that those who use a judgment’s origin to assess its validity commit the so-­called genetic fallacy. I consider the genetic fallacy-­objection in Section IV.4. For now, I shall ­confirm that in Kant’s view genealogical considerations are highly relevant for questions of validity. Here we can consider how he explains the need for a transcendental deduction of the categories. According to the abovementioned commentators, Kant merely denies that we can justify the categories by investigating, in a Lockean vein, their empirical origin; such investigations cannot settle the issue of validity (the quid iuris question) for a priori concepts. This reading supports the claim that Kant considers genealogical questions simply irrelevant to questions of (a priori) justification. However, Kant does not merely argue that the appeal to empirical origins cannot validate the categories. He makes the stronger claim that an empirical origin would invalidate the categories. As objectively valid concepts, the categories require a non-­empirical pure origin: they “must show a certificate of birth other than descent from experience” (A86–7/B119). The notion of a “descent from experience” covers more than just the Lockean attempt to abstract concepts from sense-­perceptions: the categories would also descend from experience if they arose (as Hume claims) from the mental processes that empirical psychology studies (B127–8).9 Such an empirical-­psychological origin would undermine the objective validity of the categories (cf. B5; B127–8). Kant clarifies his general view on this issue in the following passage: (P1) If . . . a judgment . . . asserts a claim to necessity, then . . . it would be absurd to justify it by explaining the origin of the judgment psychologically. For . . . if the attempted explanation were completely successful it would prove that the judgment could make absolutely no claim to necessity, precisely because its empirical origin can be demonstrated.  (EEKU, 20:238)

P1 confirms that for Kant, if one could demonstrate that an allegedly a priori judgment derives from empirical origins this would show that the judgment is 5  Warren 1998: 218. He refers specifically to Bird 1962: 10–11 and Allison 1984. 6  Riehl 1894: 78–9. 7  Windelband 2015: 273. 8  See Windelband 2015: 284. 9  Hence, Kant holds that Hume tried to derive the categories from experience (A94/B127; Prol, 4:258).

Legislative Freedom and Kant ’ s Genealogical Anxiety  105 not a priori and that its claim to necessity is unjustified. He also expresses this view in the following passages: (P2) If the determining grounds were empirical and were given in an a posteriori subjective fashion, then the judgment of reason could not be regarded as a priori and thus could not be regarded as absolutely necessary. In order to judge in objectively universal, and indeed in an apodictic fashion, reason must be free from subjectively determining grounds; for if those grounds did determine [reason], then the judgment would be merely as it is contingent, i.e., according to its subjective causes. Hence, reason is conscious of its freedom in objectively necessary judgments a priori . . . (Refl., 18:176) (P3) . . . the understanding alone (and the will, insofar as it can be determined through the understanding) is free and pure self-­activity which is determined through nothing other than itself. Without this original and immutable spontaneity, we would cognize nothing a priori; because we would be determined to everything, and our thoughts themselves would stand under empirical laws . . . (Refl., 18:182–3) (P4) The title to freedom of the will claimed by ordinary reason is based on the consciousness and the conceded presupposition of the independence of reason from merely subjectively determining causes which together constitute what belongs only to sensation and is included under the general name of sensibility. (GMS, 4:457) (P5) [Even the fatalist always presupposes that] the understanding has the capacity to determine its judgment according to objective reasons that are valid at any time, and does not stand under the mechanism of merely subjectively determining causes that may change subsequently. (RezS, 8:14)

I reconstruct the argument that Kant sketches in these passages in Section IV.3. Here I just want to note two preliminary clarifications. First, Kant’s denial that valid a priori concepts or judgments can be traced to empirical origins concerns a scenario where natural causes determine the representational content of concepts or judgments. By contrast, he sees no problem with the different scenario where natural causes merely occasion our power of cognition to bring forth a priori representational content through its own spontaneous self-­activity. He even encourages a Lockean psychological investigation of a priori concepts that searches “not for the principle of their possibility” but rather “for the occasional causes of their generation, where the impressions of the senses provide the first occasion for opening the entire power of cognition” (A86/B118).10 10  See A454/B482 for the general point that all representations, including those that have objective (epistemic) priority over empirical sense-­impressions, depend on such impressions as their subjective occasioning conditions.

106  Kant on Freedom and Rational Agency Second, the above passages jointly entail that Kant’s genealogical anxiety concerns reason in general, both our theoretical and our practical faculty of reason. In some passages Kant explicitly includes both faculties (cf. (P2)) or talks indiscriminately about a priori judgment as such (cf. (P1), (P2); in (P1) Kant also includes a priori aesthetic judgments based on the reflective power of judgment). Depending on the context, his emphasis is sometimes on theoretical reason in a broad sense as including the understanding (cf. (P3)) or on practical reason, the rational will (cf. (P4)). I will consider his genealogical concerns regarding theoretical reason in Chapters VII and IX. In this chapter I focus on practical reason or the rational will as the source of moral judgments.11

IV.2.  Moral Judgment, Objectivity, and Unconditional Necessity In this section I consider which features of moral judgments are threatened by genealogical considerations. In contemporary discussions, the target of genealogical anxiety is (typically) a view on which our moral attitudes have the following features: (1) they are cognitive, truth-­apt beliefs (as opposed to mere expressions of feelings or emotions); (2) they can be true, when they correctly apprehend the relevant moral facts; and (3) the truths or facts they represent are objective in the sense of mind-­independent.12 Contemporary proponents of genealogical anxiety argue that genealogical considerations falsify the idea that our moral judgments have properties (1)–(3), although they draw different conclusions from this. Some (like Sharon Street) suggest that we should retain (1) and (2) but reject (3): we can accept that our truth-­apt moral attitudes are often true because the moral facts they apprehend are (pace (3)) dependent on or constituted by our subjective empirical attitudes. Others (like Allan Gibbard) argue that we should reject (1) and instead accept some form of non-­cognitivism that surrenders the truth-­ aptness of our moral attitudes (though recent non-­cognitivism seems to have gravitated towards the idea that our moral attitudes are truth-­apt in some deflationary sense). Error theorists like James Joyce reject (2): they concede that our ordinary moral practice is firmly committed to (1)–(3) and that our moral beliefs purport to capture mind-­independent moral facts, but in their view genealogical considerations show that there are no such facts so that our moral judgments are false. 11  According to some readings, Kant’s “metaphysics of morals” concerns merely the a priori content of moral principles rather than “a metaphysical conception of the moral agent” (Guyer 1995: 358; cf. Beck 1960: 172, 185; Bojanowski 2006: 242–4, 261). But as the passages cited above clearly show, Kant holds that one cannot endorse the (synthetic) a priori content of a representation without incurring a metaphysical commitment regarding the pure, non-­empirical origin of that representation. This commitment is inseparable from a metaphysical conception of the moral subject as a noumenal intelligence, a self-­active, empirically unconditioned legislator of a priori principles (cf. Chapter III). 12  Here I am drawing on Gibbard 1990; Joyce 2001, 2006; Kitcher 2006; Street 2006.

Legislative Freedom and Kant ’ s Genealogical Anxiety  107 Kant accepts (1) and (2): in his view, moral judgments are cognitive states that represent which types of actions are good or what we ought to do. While he does not deny that certain feelings (respect for the moral law) are essential to our moral agency, in his view these feelings are dependent on our prior cognitive representations of moral good—­or rightness (GMS, 4:401; KpV, 5:75). These representations may be true or false (GMS, 4:410, 445; MS, 6:401), and they often are true in his considered view. However, Kant denies (3) that moral judgments refer to objective mind-­independent facts. He accepts that moral judgments purport to cognize non-­empirical evaluative or normative facts concerning what is good or evil (right or wrong), what ought and ought not to be done. He also stresses that moral judgments purport to be objective (KpV, 5:19–21). But he denies that objective moral judgments refer to mind-­independent facts. Contemporary philosophers tend to assume that the only kinds of truths or facts that merit the term “objective” are those which are entirely independent of the mental attitudes (such as beliefs) that can be held by a subject of representations.13 What is objective in this sense is the mind-­independent natural world or (for some normative realists) a realm of non-­natural mind-­independent normative facts.14 But from a Kantian perspective, the idea that whatever is fully objective must not essentially relate to the attitudes of rational subjects is puzzling. The standards for determining what counts as objective are not simply intuited by us as part of the mind-­independent fabric of the world. Instead, these standards (including a standard claiming that only what is entirely independent of our mental faculties counts as objective) are themselves dependent on our mental faculties: namely, on our rational assessment of what criteria we should accept for classifying various beliefs or facts as objective or subjective. For Kant, the very idea of objectivity is an inherently evaluative or normative notion that stems from our own legislative reason. Hence, one cannot coherently deny that normative moral judgments and facts which are grounded “in reason alone” might also have a fully objective status. Kant employs different conceptions of objectivity/subjectivity.15 The purest, most uncompromising kind of objectivity concerns what is objective in a transcendental sense (A32/B49; A36/B52). This notion applies to principles or facts whose truth, validity, or reality does not depend on the cognitive faculties and representations of finite (e.g., human) beings, such as: sensible faculties that impose a formal structure on given sense-­impressions; a discursive understanding

13  See, e.g., Kahane 2011: 116. He takes Street and Joyce to target an “objectivist” moral discourse that consists of claims about “attitude-­independent norms.” Likewise, Enoch presupposes that only “response-­independent truths” can be “perfectly objective” (2010: 414–15). Schafer (2010: 473) seeks to defend “a naïve sort of metaphysical realism about normativity that views normative facts as mind-­independent.” 14  See, e.g., Enoch 2010: 414–15; Schafer 2010: 473. 15  See Kohl 2021a for detailed discussion of this point.

108  Kant on Freedom and Rational Agency that combines disparate sensible intuitions via general concepts; or, a faculty of reason whose theoretical representations of the unconditioned are valid only as rules guiding the progress of our finite cognitive efforts. The most uncompromising, cognitively worthless form of subjectivity concerns what is subjective in an empirical sense. This notion applies to representations which depend on contingent circumstances that vary from one subject to another, such as: optical illusions; psychological associations (like connecting black cats with malfortune); or mere feelings like the pleasant taste of wine (A28–9/B44; B139–40; B142). For Kant our fundamental moral principles are objective in the strongest transcendental sense because they originate in pure practical reason or the pure will “regardless of all subjective differences” that may obtain between different agents (KpV, 5:32). In the domain of practical normativity, the “moral law . . . alone is truly (namely, in every respect) objective” (KpV, 5:74). Importantly, moral principles do not depend on the needs or limitations of our finite reason. If they did, they would be subjective in one though not the most disparaging (empirical) sense of the term: they would be analogous to the subjectively rational convictions that God exists or that nature is a unified system of laws.16 As we saw in Chapter  III), valid moral principles originate in “universal practical reason” (GMS, 4:431), a rational faculty that is shared among all rational subjects including an infinite mind, a “highest intelligence” (KpV, 5:32). Since the most basic moral principles are not restricted by any finite cognitive or conative limitations, they are objective in the strongest sense (“in every respect”). To confirm this, consider the tight connection Kant sees between objectivity and intersubjectivity or universal validity. Generally speaking, the type of objectivity that can be attributed to a representation of mine depends on the extent to which the representation is accessible to, and can be affirmed as valid by, other rational subjects. Since the representation of the moral law is accessible to, and can be affirmed as valid by, all rational subjects including one whose intellectual faculties are unhampered by any needs or limitations, this representation exhibits the strongest kind of objectivity qua universal validity (GMS, 4:417; KpV, 5:19–21, 26, 32). While Kant does not posit a constitutive link between moral truth and finite human reason, the objective truth and validity of moral principles cannot be understood apart from the moral judgments that spring from the exercise of autonomous subjects’ “universal practical reason.” Hence, there is a non-­trivial sense in which moral truths are mind-­dependent: if there were no autonomous subjects of universal practical reason, there would be no moral truths either. The constitutive link between objective moral truth and universal practical reason crucially informs Kant’s genealogical concerns about moral judgments. Due to their origin in pure, universal practical reason, moral judgments represent 16  Kant seems to be flirting with such a subjectivist view of morality in the Naturrecht Feyerabend (see Kohl 2021a).

Legislative Freedom and Kant ’ s Genealogical Anxiety  109 a distinctive type of necessity “that is not to be found in the whole of nature” (A547/B575). The only necessity to be found in nature is the hypothetical, conditional necessity that is expressed (e.g.) by the claim, “when a stone is dropped, it must fall to the ground” (cf. Chapter III). By contrast, the necessity represented in moral judgments is an unconditional rational-­ normative necessity that (for finitely rational agents) concerns what ought to happen rather than what inevitably does or will happen if certain empirical conditions obtain.17 My judgment that I must keep a promise represents a normative reason that sufficiently determines what I rationally speaking must do even if I actually fail to do it. In its practical-­normative judgments, reason “does not follow . . . the order of things as they present themselves in appearance, but frames for itself with perfect spontaneity an order of its own according to ideas” (A548/B576). The “order” framed by moral ideas is a standard of universal practical reason that classifies actions as rationally required, prohibited, or permissible. Implicit in these concepts is a claim to unconditional objective necessity: e.g., the judgment that I am required to keep my promise entails that I have an objectively necessary, universally valid reason to do so whose normative weight does not depend on (is not conditioned by) contingent empirical facts about myself, e.g., facts about what my subjective desires or feelings happen to be. Hence, for Kant the most significant property of moral judgments is that they represent universally valid, objective practical laws that prescribe with unconditional rational necessity how every rational agent must (or has sufficient reason to) act.18 It is this property that Kant sees threatened by the claim that our moral judgments result from the causality of nature.

IV.3.  Kant’s Genealogical Anxiety Argument In this section I consider how Kant supports a stance of genealogical anxiety concerning moral judgments which purport to be (1) truth-­apt, (2) true, and (3) objective. But let me first briefly sketch the genealogical argument proposed by contemporary philosophers (see n. 12). The main premise of this argument is that our moral beliefs are the effect of causal mechanisms that are epistemically unreliable. This means, roughly, that these causal mechanisms produce moral beliefs completely irrespective of whether these beliefs are objectively true. This premise is typically based on evolutionary theory: the evolutionary processes that cause our moral attitudes do not select these attitudes for their truth but rather for their 17  As Ameriks stresses (2019: 20–5), for Kant pure universal reason is the only source of unconditional necessities. 18  There are some complications here: for instance, imperfect or wide moral duties leave room for leeway or discretion in favor of subjective inclinations. But it is rationally necessary in an unconditional sense to adopt the general ends that are prescribed by wide duties.

110  Kant on Freedom and Rational Agency usefulness, e.g., their conduciveness to survival and reproduction. Or, these attitudes might be non-­adaptive by-­products of genetic mutations. In either case, the genealogical skeptic concludes that moral judgments which are caused by such processes are unjustified if they refer to objective moral facts. Some genealogical skeptics (such as Joyce or Street) proceed to argue that moral attitudes understood in terms of (1), (2), (3) are not only unjustified but false. This further argument may rest on the claim that it would be an unintelligible mystery, a wildly improbable coincidence, if processes that are totally insensitive to mind-­ independent moral truths nevertheless led us to acquire true beliefs concerning mind-­independent moral facts.19 Or, the further argument may go as follows: since we can explain everything (including our moral attitudes) without appeal to mind-­independent moral facts, and since such facts are deeply mysterious from a naturalistic standpoint, we can conclude that there are no such facts so that our moral beliefs which purport to apprehend such facts are false.20 The appeal to evolutionary theory is somewhat incidental to these arguments. The main thrust of contemporary genealogical anxiety is simply that our moral attitudes are invalid (unjustified or false) because they result from processes which are insensitive to objective moral truths. Hence, it is not anachronistic to view the above arguments in relation to Kant’s worries concerning the naturalistic origins of moral thought.21 As we saw, in Kant’s view moral judgments purport to represent (mind-­ dependent yet) objective, a priori moral truths that are absolutely necessary. In the argument that he sketches in passages (P1)–(P5), Kant asks us to assume a naturalistic worldview where everything that exists and happens is solely the product of natural laws and causes. He argues that on this view our moral judgments would result from the “mechanism of merely subjectively determining causes” (RezS, 8:14). He concludes that if our moral judgments had this empirical-­ causal origin, they would lack an a priori character and would (therefore) fail to express objectively rational, unconditional necessities: “the judgment of reason could not be regarded as a priori and thus could not be regarded as absolutely necessary” (Refl., 18:176). To understand Kant’s argument, we must first note his denial that objectively necessary practical truths might be conveyed to us by natural causes via perception. For Kant, objective necessities in general, including descriptive necessities, cannot be sensibly intuited (A137–8/B 176–7; A 732–3/B 760–1). The reason why perception cannot generate moral representations which present 19  See Street 2006. 20  See Joyce 2001 for this naturalistic route into error theory, which is modeled on Mackie 1991. 21  Arguably, genealogical anxiety arguments that draw on evolutionary theory revive an old philosophical theme in a new scientific framework: as Srinivasan point out (2019: 131), there is little in contemporary “evolutionary debunking arguments that is not already present in Xenophanes’ genealogy of Greek theology.”

Legislative Freedom and Kant ’ s Genealogical Anxiety  111 us with normative-­practical necessities is that we cannot apprehend a norm which unconditionally obligates the will, a categorical imperative, through the senses or through analogous passive-­receptive faculties. The passive apprehension of external facts, including commands such as state laws or strictures viewed as a divine commands, can only support a conditional normative judgment: if one has an antecedent desire or aversion related to these facts (e.g., an aversion against suffering state or divine punishment), then and only then can the passive apprehension of external facts make one recognize a normative reason for acting in a certain way. A practical norm whose validity is conditional on whether agents have an antecedent desire for or an aversion against external states of affairs is a merely hypothetical imperative. Thus, our perception (or intuition) of things or facts that are external to our rational will (qua “universal practical reason”) cannot ground unconditionally necessary, categorically binding moral norms that obligate our will regardless of what our antecedent desires happen to be (GMS, 4:443–4).22 Since perceptions of external reality are not sufficient to provide moral representations, such representations must be generated from within, by faculties that determine whether and how we respond to perceptual input. On a naturalistic worldview these would be psychological, neurological, or biological mechanisms that process the intake of sensations. As Kant argues in the passages cited earlier ((P1)–(P5)), such mechanisms have three interrelated features that invalidate our moral judgments: they are contingent, mutable, and subjective in the pejorative (empirical) sense that implies a lack of cognitive validity. The naturalistic mechanisms that process the intake of sensations into moral judgments are contingent in the empirical-­relative sense of contingency that I explained in Chapter III: the occurrence, character, and representational output of these processes depends on whether certain empirical (environmental, genetic, neurological, psychological) conditions obtain; whether these conditions obtain in turn depends on further empirical conditions; etc. Naturalistic mechanisms are mutable as a result of being contingent: their character and representational output “may change subsequently” (RezS, 8:14) due to changes in their natural conditions, e.g., in our gene pool or physical environment. These mechanisms generate merely subjective, cognitively invalid practical thoughts 22  For helpful exposition of this point, see Korsgaard 1996b (especially Lecture 1). Mackie 1991 also accepts that categorical oughtness cannot be perceived through the senses. On some contemporary views, we can perceive moral truths or facts through the senses because moral properties are reducible to natural-­empirical properties (Brink 1989; Railton 1986). Proponents of such reductive-­ naturalistic moral realism concede that sense-­perception of natural properties cannot generate normative reasons for acting because they hold externalist views on which moral judgments state purely descriptive truths and thus do not engage the rational will at all—­one can sincerely make such judgments without seeing any reason to perform or omit the relevant action. This view sacrifices what Kant regards as the most integral feature of moral judgments: their capacity to bind the rational will via the representation of objectively necessary practical reasons.

112  Kant on Freedom and Rational Agency because they are wholly insensitive to the objective moral properties that we ascribe to actions in our practical reasoning when we deem them required or prohibited. What does it mean to say that naturalistic mechanisms are insensitive to ra­ tional moral properties? For contemporary philosophers it means that such mechanisms do not track moral properties or facts.23 But the notion of tracking is unsuitable with regard to Kant’s view because it suggests a model of practical reason that (as we saw) supports only hypothetical imperatives: namely, a model on which our practical judgments are passive responses to facts that are external to our rational will and therefore cannot categorically obligate us.24 I now suggest three different (though related) ways in which Kant might explicate, without recourse to the heteronomous notion of tracking, the idea that naturalistic mech­ an­isms are insensitive to the rational properties (such as being required) that our moral judgments purport to determine. First, naturalistic mechanisms are insensitive to such rational properties insofar as these mechanisms operate in a non-­intentional, sub-­personal, subconscious manner. Examples of such mechanisms include bio-­chemical processes that wire our brain, deep-­seated psychological instincts, and associative habits. On the naturalistic view, empirical mechanisms of this sort generate the basic moral categories (such as the idea of inviolable dignity, or the idea that certain actions are strictly prohibited) that eventually structure our conscious practical reasoning. But sub-­personal processes cannot generate objectively binding moral ideas that may ground a rational normative discourse. More generally, an internal mech­an­ ism over which we have no rational control and whose workings are entirely removed from our conscious intellectual awareness cannot produce representations that have any legitimate rational claim on us, i.e., that provide us with objectively valid reasons for thinking or acting in a certain way. This way of unpacking the notion of insensitivity is suggested by Kant’s conception of how Hume construes the genesis of our allegedly a priori concepts, namely, as deriving from psychological processes that occur below the threshold of consciousness. For Kant, such non-­conscious mechanisms are the cause of empirical associations that lack objective validity (B139–40). His examples often concern the subjective association of words (e.g., “lemming”) with certain properties (e.g., suicidal tendencies), but the general naturalistic model here is broader and can also be applied to our moral thinking: sub-­personal mechanisms whose operations bypass our rational self-­consciousness generate conscious ideas such as “unconditional wrongness” and make us strongly associate these ideas with certain action-­types 23  The tracking-­metaphor is standardly used in contemporary discussions; see, e.g., Kahane 2011: 118; Schafer 2010: 471; Srinivasan 2019: 132; Street 2006: 125. 24 Accordingly, Kant would also reject reliabilist accounts of moral knowledge or justification according to which our moral judgments are non-­arbitrarily caused by facts that are independent of and external to our rational will.

Legislative Freedom and Kant ’ s Genealogical Anxiety  113 and (e.g., unpleasant) sentiments. After some time-­period that involves frequent associative repetition (plus other, e.g., social mechanisms that further enforce the associative connections), this conditioning process eventually grounds habitual dispositions to morally judge, feel, and act in certain ways which are triggered by certain external stimuli. Second, the naturalistic processes that generate our moral thoughts are insensitive to objective rational necessities insofar as the character of these processes (i.e., how they operate and what output they produce) is determined by causes that are devoid of rational significance. From a contemporary naturalistic perspective, our current moral thinking is the product of internal processes whose character is partly determined by past genetic mutations that, over time, generated dominant (biological, psychological) traits. For early modern thinkers who consider a proto-­naturalistic worldview, the (as Descartes puts it in the First Meditation) purposeless “continual succession of events” randomly happened to generate a type of consciousness with certain innate psychological dispositions. In both scenarios, the processes that produce moral representations result from natural developments which are, rationally speaking, entirely arbitrary: it makes no sense to say that a genetic mutation rationally speaking ought (not) to have occurred. In Wilfried Sellars’s famous terms, such events belong to the space of causes rather than to the space of reasons. Causal conditions that are rationally speaking random or arbitrary cannot imbue our mental faculties with the power to generate universally valid, objectively necessary practical judgments. To illustrate: if we must admit that our disposition to regard lying for personal gain as strictly impermissible is a mere product of a-­rational factors such as biochemical syntheses in the brain, then we cannot ascribe to this practical representation any objective rational force or validity. We must then also admit that some random change in the natural causes which determine our practical reasoning in the actual world, e.g., the abruption of some brain mutation process due to a sudden climate drop, might have disposed us to endorse lying for personal gain as permissible, even commendable. This illustrates the fortuitous, (rationally speaking) arbitrary character of the actual mechanisms that control our practical reasoning in the naturalistic scenario. There is a further sense in which naturalistic causal processes are insensitive to the normative properties that our moral judgments seek to determine: these processes involve causal factors that are directly opposed to the rational authority these judgments purport to have. Here the point is not merely that the naturalistic explanation of our moral beliefs invokes (rationally speaking) arbitrary conditions or that this explanation makes no reference to corresponding moral facts or truths.25 Rather, the naturalistic explanation of our moral judgments refers to 25  This is how White 2010 reconstructs the main worry underlying contemporary genealogical skepticism.

114  Kant on Freedom and Rational Agency precisely those causal factors from which moral judgments purport to be free and independent. This strong type of insensitivity would obtain if the causal processes that generate our moral representations essentially involved our empirical desires or feelings. This is how Kant conceives of attempts to naturalize our moral representations (MS, 6:378). It is plausible for naturalists to invoke such desiderative states as playing a central role in the genesis of our moral judgments if they accept that moral judgments as practical judgments are inherently connected to motivation and action. If one sincerely makes a judgment about what one has reason to do, then one has at least some motivation to act in accordance with this practical representation. In Kant’s own view, the motivational-­emotive disposition to act in accordance with our moral judgments depends on our prior rational cognition of how we ought to act. This relation is reversed on the naturalistic view where our moral judgments are conditioned by (and owe their motivating force to) antecedent, empirically given desires and feelings.26 For Kant this view implies that our moral judgments fail to express objective rational necessities. To see why, let us return to Kant’s aforementioned claim that moral ought judgments express “a species of necessity and a connection with reasons which does not occur anywhere else in the whole of nature” (A546/B575). He explains this claim as follows: However many natural grounds or sensible stimuli may be that impel me to will, they cannot produce the ought but only a willing that is yet far from necessary but rather always conditioned, over against which the ought that reason pronounces sets a measure and goal, indeed, a prohibition and authorization.  (A548/576)

As we saw (in Section IV.2), for Kant moral judgments purport to represent an objective normative standard which classifies actions as obligatory, prohibited, and permissible. Each of these concepts involves a representation of practical necessity: an action or end deemed obligatory is such that every rational agent must perform or adopt it; an action or end deemed prohibited is such that every rational agent must not perform or adopt it; when an action or end is deemed permissible, this verdict relies on the idea that there are rational criteria which an action or end must satisfy so that agents rationally speaking may perform the act or adopt the end at their discretion. When we apply these normative concepts, our moral judgments purport to have rational authority over empirically given desires or feelings: judgments about which actions and ends are obligatory,

26  Influential contemporary proponents of this view include Williams 1997 and Schroeder 2007. The major historical inspiration here is Hume’s claim that reason is motivationally inert: judgments of “practical” reason depend for their normative content and motivational force on empirically given passions.

Legislative Freedom and Kant ’ s Genealogical Anxiety  115 prohibited, or permissible “confront our desires with a limit and an end,” they “forbid or authorize” acting on those desires. The idea that persons are morally obligated to perform some action entails that they rationally speaking must perform the action regardless of what their desires and feelings may be. Our practical reason could not legislate claims that have a legitimate absolute authority over our empirical desires if our practical reasoning were conditioned by those very desires. Hence, “natural grounds” such as our empirical desires or “sensible stimuli” cannot imbue practical judgments with the kind of necessary, desire-­independent normative force that we represent through moral concepts such as “obligatory,” “prohibited,” or “permissible.”27 Empirically given desiderative states paradigmatically exemplify the three features that comprise Kant’s pejorative concept of “merely subjectively determining causes.” First, empirical desires are inherently contingent (in the relative-­empirical sense): on the naturalistic view, the desires that prompt our moral judgments depend on natural conditions which in turn depend on prior empirical developments, etc. etc. Second (and consequently), these desires are mutable: they might change at any time along with the natural conditions on which they depend. Third, these desires are subjective in a pejorative sense because they are insensitive to the normative properties of actions that our moral judgments purport to determine. The term “insensitivity” here designates a direct opposition or incompatibility between contingent empirical desires and moral claims that purport to have a desire-­independent, necessary rational authority and normative force. This third notion of insensitivity fruitfully connects with the two previous notions I have suggested, which stress (1) the dependency of our moral thought on rationally arbitrary causes, and (2) the sub-­conscious character of the mech­an­ isms that produce our moral representations. Regarding (1), the presence and strength of the empirical desires that (on the naturalistic view) produce our moral representations depends upon causal processes that are rationally speaking random or arbitrary, such as our genetic development. Desires with such a causal ancestry “cannot produce the ought”: they cannot yield a judgment that prescribes with objective rational authority and necessity. I could not view myself as having an objectively necessary reason to act in a certain way if I had to concede that my conviction that I have such a reason is conditioned by desires whose presence and strength is due to processes such as 27 Guyer argues that the real possibility of the categorical imperative can be shown prior to addressing “metaphysical” questions about the freedom of moral agents (1995: 384; see n. 11 above): it requires only that one can establish the “independence of the principle of morality from any contingent and empirically given end” (1995: 369). However, one cannot proclaim such independence while conceding that our representation and acceptance of moral principles is due to the causal influence of contingent, empirically given desires. When Kant says (at A548/B576) that such desires “cannot produce the ought,” he immediately invokes a squarely metaphysical property of our will, namely the absolute spontaneity of practical reason, as the only legitimate source of a priori moral oughts (“reason frames for itself with absolute spontaneity an order of its own . . .”).

116  Kant on Freedom and Rational Agency genetic mutations which lack any rational significance. Here I would also have to concede that if (say) the genetic mutation that is a partial cause of my current desires had been prevented by (say) a sudden drop of temperature, I might have ended up with different desires that might have led me to adopt a (perhaps very) different moral outlook with just the same degree of subjective conviction. This illustrates how our desire-­based moral judgments on the naturalistic view depend on arbitrary coincidences which “cannot produce the ought” of objective rational necessity. Regarding (2), the relevant point is that on the naturalistic picture the sense of necessity that accompanies our moral representations is illusory. It is imparted onto our conscious moral representations surreptitiously, through a causal influence of desires, instincts, or feelings that is not transparent to our conscious reasoning. These non-­conscious causal factors give rise to a sentiment of psychological compulsion, a strong impression that one must (not) do certain things which has no objective normative-­rational force but which masks itself as having such force and which we erroneously take to have such force. In Kant’s words, sub-­personal naturalistic processes would generate “a subjective necessity, arbitrarily implanted in us” (B168), “which is subsequently falsely held to be objective” (B127). With regards to our allegedly a priori practical representations, this means: due to the “covert” way in which empirical incentives “have an influence on the will,” “the so-­called moral imperative, which as such appears to be categorical and unconditional, would in fact only be a pragmatic precept” lacking in objective rational necessity (GMS, 4:419). This picture is quite congenial to the way in which contemporary naturalists like Joyce account for the false appearance of categorical oughtness in our moral thinking: the sense of urgency and categoricity that we tend to associate with the term “moral ought” derives from sub-­personal processes of natural selection which imbue our practical representations with a desiderative motive force that enhances our reproductive fitness since it maximizes the chances that we are socially cooperative, help family members, etc.28 In sum, Kant sees a complete misfit between the causes, such as most prominently empirical desires or feelings, that produce our moral representations on a naturalistic worldview and the features that our moral judgments purport to have. Contingent, mutable, subjective states like empirical desires cannot generate moral ideas that correctly represent objectively necessary, desire-­independent, timelessly valid normative prescriptions. I will now suggest a further (related) take on Kant’s genealogical anxiety argument, which focuses on the link between objectivity and intersubjectivity that I mentioned in Section IV.2.

28  See Joyce 2001: 135–40.

Legislative Freedom and Kant ’ s Genealogical Anxiety  117 As we saw, Kant’s strongest notion of objectivity pertains to judgments that apply to all rational beings. This requires that such judgments arise from universally shared rational capacities. Accordingly, the cognitively speaking most worthless form of subjectivity pertains to judgments that agents make in the grip of private empirical conditions such as their given desires or feelings which in turn result from individual empirical circumstances such as genetics, psychological development, and upbringing. Practical judgments that are conditioned in this manner cannot lay claim to objective validity because they entirely lack intersubjective validity: they apply to agents only if these agents happen to have a particular idiosyncratic desire-­base. The issue here is most vivid in cases where people make conflicting practical judgments. Suppose B judges that stealing from one’s sick father is prohibited whereas A judges that stealing from one’s sick father is permissible. If both judgments derive from evaluative outlooks that are empirically conditioned by A’s and B’s respective desires, which in turn result from A’s and B’s varying empirical circumstances (e.g., their genes, upbringing, development), then neither judgment can lay claim to universal-­intersubjective validity. Accordingly, neither judgment can purport to have a more rational or objective status than the other. Both judgments have only what Kant calls a subjective, private validity (KpV, 5:19–21). One might respond that the naturalistic picture is not as such incompatible with the notion of intersubjective validity because different agents may and sometimes do share roughly the same empirical psychology so that roughly the same practical judgments apply to them on the naturalistic view. Kant anticipates this response in the second Critique: But suppose that finite rational beings were thoroughly agreed with respect to what they had to take as objects of their feelings of pleasure and pain . . . even then they could by no means pass off the principle of self-­love as a practical law; for, this unanimity itself would still be only contingent. The determining ground would still be only subjectively valid and merely empirical and would not have that necessity which is thought in every law, namely objective from a priori grounds, unless one had to say that this necessity is not practical at all but only physical, namely that the action is as unavoidably forced from us as is yawning when we see others yawn.  (KpV, 5:26)

Part of Kant’s stance here appears to be (surely unwittingly) echoed by David Lewis: What comfort would it be if all mankind just happened to be disposed alike? Say, because some strange course of cultural evolution happened to be cut short by famine, or because some mutation of the brain never took place? Since our dispositions to value are contingent, they certainly vary when we take all of

118  Kant on Freedom and Rational Agency mankind into account, all the inhabitants of all possible worlds . . . . The spectre of relativity within our own world is just a vivid reminder of the contingency of value.29

Kant’s and (perhaps) Lewis’s point can be put as follows: contingent sameness of subjectivity does not equal intersubjective validity. Even if we happened to have the same empirical psychology and desire-­base, this would not make it the case that we are governed by truly shared normative standards. The idea that we are governed by shared normative standards which have a universal, objective (rather than merely private, subjective) validity implies that we can hold each other to these standards: implicit in my moral judgments about what is required, prohibited, or permissible is the demand that others ought to agree with me. But on the naturalistic picture I can only predict that others will morally judge as I do since, and for as long as, we happen to desire the same things. The central difference between the normative demand for agreement and the prediction of happenstance agreement is that the normative demand involves a refusal to let others off the hook if they disagree. If we are governed by truly shared, intersubjectively valid moral norms, the mere fact that others might begin to disagree with my moral verdicts cannot imply that the moral norms I accept no longer apply to them. But this implication does hold on the naturalistic view. Unless practical disagreement is based on one party’s theoretical ignorance about the causal means to a desired end, such disagreement indicates that our empirical desires are no longer the same (or no longer have the same strength), and contingent sameness of desires is the sole basis for agreement about moral matters on the naturalistic picture. These desires, as part of our empirical-­temporal constitution, are per se mutable: for instance, my basic desires and therefore my moral outlook might “change subsequently” due to (say) my prolonged use of some medication. The implication that if this happens others must let me off the hook and concede that their moral judgments are no longer binding for me shows that we are not governed by truly shared normative standards and that our individual moral judgments lack intersubjective validity. Suppose A and B agree that stealing from one’s sick father is strictly prohibited. On the naturalistic account of why both A and B accept this moral representation, their agreement is based on nothing but the contingent sameness of their respective empirical desires. Thus, if B were to incur a change in their contingent desire-­base the above moral judgment would no longer apply to them. Suppose B encounters a financial emergency which prompts a strong desire to look for new sources of income; moreover, this also causes B to suffer from depression and the anti-­depressant medication unsettles B’s previous sympathetic feelings for their

29  Lewis 2000: 89 (his emphasis).

Legislative Freedom and Kant ’ s Genealogical Anxiety  119 family. Under the influence of these new empirical conditions, B’s “subjective motivational set” changes and B adopts the new conviction that stealing from their sick father is permissible.30 On the naturalistic view this judgment expresses B’s new desire-­based practical reasons. Here, the empirically conditioned moral judgment that both A and B accepted prior to B’s psychological changes cannot lay claim to the intersubjective, necessary, immutable (empirically-­temporally unconditioned) normative validity that we express when we say that one must (ought) not steal from one’s sick father. Indeed, on the naturalistic picture the very idea of an immutable or unconditional practical necessity is itself a mere product of non-­rational, contingent empirical mechanisms such as social conditioning procedures, evolutionary processes that hard-­ wired the brain by grounding adapt­ive dispositions to support relatives, or non-­adaptive brain mutations. Thus, our firm conviction that we are rationally required not to perform certain actions arises from cognitively blind habits, from a “subjective necessity which is subsequently falsely held to be objective” or universally binding.

IV.4.  Responses to Kant’s Argument In this section I consider a range of responses to the argument that underlies Kant’s genealogical anxiety. I begin by considering a remark from Harry Frankfurt: The judgments a person makes concerning rationality are manifestly no less dependent than are any other occurrences in his life upon contingent features of his nature and his circumstances. What renders these judgments impersonal is that the claims they make are not limited to the person who makes them; rather, it is implicit that anyone who disagrees with the claims must be mistaken.31

Frankfurt holds that a judgment about rationality or reasons has an impersonal normative scope because it is not limited to the judging subject. This is in line with Kant’s point that our judgments about (moral) reasons purport to have a more than private, namely a universal validity. But Frankfurt’s appeal to the impersonal scope of our normative judgments is incompatible with his subsequent concession that such judgments “manifestly” depend on contingent features of our varying natural circumstances. If my moral judgment strictly depends on my private, contingent empirical conditions, then it cannot lay claim to a validity that extends beyond myself to others including those whose diverging judgments depend on their private, contingent empirical conditions. Likewise, it is unclear how I can claim that someone who disagrees with me about what 30  The notion of a subjective motivational set is taken from Williams 1997. 31  Frankfurt 1988: 90.

120  Kant on Freedom and Rational Agency ra­tion­al­ity requires is mistaken if I must admit that my conception of rational requirements is a mere product of empirical circumstances that are beyond my rational control, while others’ diverging conceptions of such requirements result from diverging empirical circumstances that are beyond their rational control. We might all have the same view of rational requirements because our thinking depends on roughly the same contingent natural conditions (e.g., on a similar empirical psychology). But, as I argued above, a happenstance coincidence of subjective-­private conditions cannot secure a normative claim to objective, intersubjective validity. Kant’s view is also rejected in Thomas Scanlon’s conception of “our process of critical reflection.” For Scanlon such reflection requires only “that that process itself be sensitive to reasons, and that later stages of the process be importantly dependent on conclusions reached at earlier stages. But there is no reason . . . that this process itself not be a causal product of antecedent events and conditions.”32 Scanlon rightly stresses that the later stages of a process of (valid) practical reasoning must be considered rationally dependent on earlier stages of this process: B’s judgment that it is wrong to steal from one’s sick father logically depends on, and derives its normative force from, B’s prior recognition that (say) persons have a supreme worth. But if our reasoning and judging is always “the causal product of antecedent events and conditions” construed in naturalistic terms, then our acceptance of ideas that are fundamental to our practical reasoning (such as the idea that persons have a supreme worth) ultimately depends on our contingent natural constitution and, thereby, on whatever non-­rational processes caused us to have this constitution. This dependence cannot be construed as a rational dependence that obtains between the contents of more general and more specific normative propositions. Rather, it is a merely causal dependence of our practical concepts and judgments on contingent factors that are, rationally speaking, wholly arbitrary, contingent, thus subjective in Kant’s pejorative-­empirical sense. If our basic practical concepts are the products of such non-­rational factors, then the specific judgments which apply these concepts also lack objective rational force and normative authority. Scanlon’s metaphor that our critical reflection is “sensitive to reasons” is unhelpful here since it is unclear how we could be sensitive to valid reasons if our mental faculties and their representational outputs arose from causal processes that are wholly insensitive to rational considerations. Another strategy for dismissing genealogical anxiety is to deny that our normative judgments require any external-­ genealogical validation—­ instead, one might insist, the authority of these judgments derives entirely from our internal reflective perspective. Korsgaard argues that “Reflection gives us a kind of distance from our impulses which both forces us, and enables us, to make laws for 32  Scanlon 2003: 370.

Legislative Freedom and Kant ’ s Genealogical Anxiety  121 ourselves, and it makes those laws normative.”33 As we saw, for Korsgaard this reflective normative stance is immune to considerations concerning the origin of our normative judgments. However, we cannot claim to have reflective distance from our impulses if we must concede that our reasoning and judgments are (however surreptitiously) controlled by such impulses: the notion of reflective distance is hollow unless the reflecting subject is free from the determining influence of those factors that the subject seeks distance from. Thus, we cannot obtain genuine reflective distance on a naturalistic view where our representations of putatively necessary normative laws are determined by contingent desires or similar (psychological, biological) causal factors, even if these factors do not operate in the foreground of our conscious reflection and (thus) remain unnoticed by our naïve reflective self-­awareness. Indeed, as we saw, on the naturalistic view the illusory sense of necessity that informs our representation of alleged normative laws is itself the upshot of sub-­personal empirical mechanisms. David Owens defends Korsgaard’s claim that our normative judgments can be regarded as valid even “if our judgments as to what we should . . . do are themselves determined by factors beyond our control.” For Owens the “assumption . . .  that true control must be total and unconditional . . . is unwarranted: control can emerge from a psychological background which we do not control . . .  human beings have reflective control over their actions, precisely because we can form a view of what we ought to do and implement our practical judgment in action.”34 Owens is correct to stress that reflective control per se does not require independence from a psychological background that we do not ­control. If the term “reflective control” designates no more than a capacity to guide our actions through a correct assessment of how we can satisfy our given desires or maximize our overall desire-­satisfaction, then we can concede that our practical reflection is conditioned by an empirically given psychological desire-­based (a subjective motivational set) that lies beyond our rational control. But we cannot make this concession if we think that our moral judgments have universal validity and rational authority over our empirically given desiderative states. Hence, Owen’s claim that we possess reflective control “precisely because we can form a view of what we ought to do,” even though this normative view itself results from contingent conditions beyond our rational control, is acceptable only when it is coupled with a subjectivist, desire-­based model of normative-­practical judgment. If our “view of what we ought to do” aspires to objective-­universal validity and to rational necessity, we must have normative powers of practical reflection whose character and output is not determined by private, contingent, and arbitrary conditions.

33  Korsgaard 1996b: 129.

34  Owens 2000: 11.

122  Kant on Freedom and Rational Agency The responses considered thus far share a common core: they suggest that the causal origin of our practical judgments is simply irrelevant to their justificatory force or rational content. Hence, I suspect that these responses ultimately assume that those who infer the invalidity of a representation (concept, judgment) from its causal origin commit the so-­called genetic fallacy. The appeal to the genetic fallacy suggests a pattern of inference that is commonly recognized as mistaken, just like (say) the statistical fallacy. However, we routinely deem inferences from origin to invalidity quite acceptable. A few examples: X expresses the belief that professor Q has an affair with a student; when asked where they got this idea from, X refers to a gossipy internet blog and their belief is duly dismissed. Or, Y expresses the belief that chicken which has been fried at 100 degrees for thirty minutes protects against Alzheimer; once we learn that this belief originates in the home-­schooling doctrines imposed upon Y by their unhinged father we no longer take it seriously. Or, Z stops trusting their belief that their parents always treated them with love and compassion once Z finds out in psychotherapy that this belief results from subconscious coping mechanisms. Finally: a biologist abandons a long-­held theory once they recognize that their interpretation of relevant data has been strongly influenced by a surreptitious type of confirmation bias. In all these cases, we infer from the (unreliable, irrational, etc.) origin of a belief that the belief is invalid. In each case this inference seems entirely warranted. The term “invalidity” here is ambiguous: it could refer either to a lack of ra­ tional justification or, more strongly, to a lack of truth. A genealogical examination which uncovers the cognitively insensitive or non-­rational sources of a belief shows that the belief is unjustified. This does not refute the belief in the stronger sense of proving it to be false. But this raises no major issue for proponents of genealogical critique: for them, the conclusion that a belief is rationally unjustified since it derives from cognitively blind mechanisms is strong enough for all intents and purposes. Moreover, the claim that the relevant beliefs might yet be true is idle unless it can be backed up by further considerations that themselves rely, however implicitly, on certain genealogical suppositions. For instance, the possibility that the abovementioned biological theory might yet be true is worth taking seriously only if there is a process of interpreting the relevant data as supporting that theory which is free from confirmation bias; or, the belief that professor Q has an affair with a student is worth taking seriously only if this belief originates in a veridical perceptual process (which can in turn underwrite credible testimony). These remarks do not exhaust the complex issues surrounding the topic of genealogical critique, but they suffice at least for shifting the burden of proof to those who declare that genealogical considerations cannot, as a matter of principle, be relevant to questions of validity or justification. I return to this topic in Chapter VII.

Legislative Freedom and Kant ’ s Genealogical Anxiety  123 It is instructive to briefly consider how the genetic fallacy issue arises for Nietzsche, a famous proponent of genealogical critique. In an attempt to deflate the critical importance that genealogy plays in Nietzsche’s philosophy, Bernard Reginster argues that “a moral truth would be no less a truth . . . for having been discovered by immoral means,” e.g., by a mind filled with ressentiment.35 However, Nietzsche (arguably) denies that moral truths are simply there to be “discovered” or that moral facts exist by themselves apart from any relation to human attitudes and (thereby) to the causal processes that give rise to such attitudes. This is why he takes his genealogy of moral beliefs to undermine the validity of common morality. If the common belief that positive, peaceful attitudes toward others (such as respect, benevolence, compassion) are good arose and gained acceptance not because of rational, truth-­directed cognitive enquiries but due to the influence of non-­rational, sub-­conscious psychological mechanisms that centrally involve hostile, negative attitudes (like ressentiment), then we have no reason to take that belief seriously as a justified claim to moral truth. Could the belief still be true, its genealogy notwithstanding? Perhaps Nietzsche would hold that if a belief can be naturalistically explained without reference to the alleged moral facts which this belief purports to apprehend, and if these alleged moral facts play no explanatory role in other contexts either, then considerations such as parsimony imply that there are no such facts so that the belief is false.36 Or, perhaps Nietzsche would say that the question is misplaced because as far as beliefs in moral values are concerned we have no realistic handle on their truth or significance apart from considering the psychological needs and drives of those who accept these values as true and binding. There is a further (though related) critical dimension to Nietzsche’s genealogical critique. For Nietzsche, common morality essentially involves a claim to necessary, universal validity and bindingness on everyone: it represents itself as the only possible and legitimate form of morality. This claim to absolute validity rests on the conviction that the content and acceptance of common morality does not depend on contingent historical or psychological conditions and (hence) does not arise from the partial needs, desires, and interests of specific individuals and peoples. As Alexander Nehamas puts it, for Nietzsche every morality is a historically contingent, subjective interpretation of human life and meaning, but “an interpretation can appear to be binding on everyone only if the fact that it is an interpretation remains hidden” since it “is presented as a view that is objectively true of the world and is addressed to all human beings simply as human beings, as

35  Reginster 2009: 198 (cf. 292); cf. Leiter 2015: 139, 249. 36  For Reginster and Leiter, Nietzsche employs the principle of parsimony in this vein. But the appeal to parsimony cannot work without attributing a critical role to the genealogical method as a tool for showing that alleged facts of a certain kind play no role in explaining the genesis of our beliefs in these facts.

124  Kant on Freedom and Rational Agency rational agents . . . or . . . as children of God.”37 The ascetic ideal which Nietzsche takes to be the common core of Christian, Kantian, Utilitarian (etc.) moralities owes its common acceptance partly to its success in masking its contingent, partial origins: “In order to make a claim to unconditional acceptance, the ascetic ideal conceals its will to power and its partial and specific origins and goals”; it “denies the radical contingency of history.”38 As Nietzsche puts it, “all these values are empirical and merely conditioned. But he who believes in them, reveres them, does not want to acknowledge precisely this character . . .” (NF 1888: Gruppe 14, 109). Nietzsche’s genealogical method has strong critical potential partly because it threatens to undermine common morality’s claim to universal, necessary validity or bindingness by exposing how this morality manifests and derives from empirically conditioned, contingent, partial needs that characterize only specific modes of human life and culture rather than humanity per se. I have engaged in this brief excursus on how the genetic fallacy-­charge falls flat with respect to Nietzsche because I take Nietzsche and Kant to be in agreement about two crucial points. First, both deny that moral beliefs refer to facts or values that exist out there to be discovered or “tracked” by mental processes whose internal character stands in no essential relation to the objects of belief. Second, both recognize that our common morality is closely aligned with a claim to universal, necessary, objective validity which cannot survive the naturalistic exposure that common morality originates in entirely partial, contingent, subjective psychological needs and drives that operate below the threshold of rational consciousness. This common ground explains why for both Nietzsche and Kant, the method of assessing the validity of moral beliefs in terms of their causal origin is not inherently fallacious.

IV.5. A Rational Causality of Nature? I now consider a different response, call it R, to Kant’s genealogic argument.39 According to R, we can cure Kant’s genealogical anxiety by assuming that the natural causes which are causally responsible for our representational faculties and (thus) for our moral judgments differ from Kant’s starkly mechanistic, non- or a-­rational conception of natural causality: nature may include forms of causation which operate in an intentional, rational, cognitively sensitive fashion and which therefore can provide us with the mental capacities that we need for representing objective rational necessities. In my view, R fails for three reasons. First, R seems dialectically inappropriate since it does not capture the naturalistic worldview of “dogmatic empiricism” that Kant assumes (for the sake of 37  Nehamas 1987: 126. 38  Nehamas 1987: 128. 39  I am grateful to Karl Ameriks for suggesting this response to me.

Legislative Freedom and Kant ’ s Genealogical Anxiety  125 argument) when he discusses what would follow if all causality were “mere nature.” According to this naturalistic worldview, a purely mechanistic, non-­ intentional, or non-­teleological character is constitutive of natural causality per se. In Kant’s own view, a mechanistic form of causality is the only natural causality that can figure in objective theoretical cognition and explanation (see, e.g., KU, 5:397, 410). Likewise, contemporary forms of naturalism that rely on evolutionary theory cannot allow for the existence of natural causes, other than human (perhaps also animal) agency, that operate in an intentional or teleological fashion. Second, even if we assume (with R) that some form of rational causality is immanent to the natural order of things, we must still regard its existence and internal character as irreducibly contingent. In the chain of natural causality nothing has absolute necessity: every cause depends for its characteristic activity and output on preceding conditions which are in turn contingent on further conditions, and so on ad indefinitum. It seems implausible to suppose that every preceding part of this indefinitely extending natural chain is endowed with a rational character that it can impart onto its effect. Hence, it seems that whatever rational character we can attribute to some later member of the chain is ultimately the upshot of (however remotely) preceding cognitively insensitive, rationally speaking arbitrary empirical conditions—­which is precisely what R qua response to Kant’s genealogical anxiety-­argument sought to avoid. In any case, since the supposedly rational character of natural causes cannot prevent their contingency and mutability, it also cannot prevent the contingency and mutability of the relevant causal output such as our mental faculties and moral representations. The natural causal processes that are responsible for our practical reasoning “may change subsequently” whenever there is a change in the natural circumstances upon which the presence, character, and output of these processes depend. Thus, if our practical reason were a temporally conditioned member of the chain of natural causes, our acceptance of specific moral judgments and values would only be a fickle moment in the causal passage of time. Insofar as a moral judgment results from contingent natural processes, it cannot (validly) represent a timeless (absolute) necessity; instead, this “judgment would be merely . . . contingent . . . according to its . . . causes” (Refl., 18:176). A third Kantian reason for rejecting R can be gleaned from B167–8. Kant’s argument here implies that the appeal to an intelligent natural cause which conditions us to make certain judgments invites the same twofold problem as the appeal to an intelligent super-­natural (i.e., divine) cause that disposes us to make certain judgments (via some kind of “preformation system” or pre-­established harmony). The first problem is epistemic: the existence or character of any ra­ tional but non-­human form of causality lies beyond our ken. Hence, invoking a foreign rational causality as the ultimate source of our moral judgments cannot secure our epistemic entitlement to the idea that our moral judgments express an objective rational necessity. For Kant, the appeal to an unknowable external

126  Kant on Freedom and Rational Agency causality that sets up our mind in a benign fashion cannot provide us with ­ra­tional justification for believing in non-­trivially necessary (synthetic a priori) knowledge claims (such as the claims of morality) because this appeal amounts to nothing more than a hopeful guess or leap of faith, which is “precisely what the skeptic wishes most” (B168). (I elaborate on this argument in Chapter VII.) The second problem concerns the metaphysics of moral normativity. If our minds were set up by some intelligent (in R, natural) design to represent and endorse moral ideas, these ideas would rest “only on a subjective necessity . . .  implanted in us” (B168) by a foreign cause. Hence, these ideas would not stand under our autonomous, absolutely spontaneous rational self-­control; in accepting these ideas we would exhibit the merely relative spontaneity of an automaton spirituale whose goals and representational activity are determined from without. In this case, we would have a merely heteronomous will that could only be governed by hypothetical imperatives. In response to this second problem, a proponent of R might argue that the foreign intelligent design need not be regarded as causing our moral ideas or judgments: rather, such design only provides us with an innate rational disposition that we exercise spontaneously and autonomously when we bring forth our moral representations. This might be a suitable way of conceiving how our noumenal constitution, including its absolutely spontaneous capacity for rational self-­ determination, is created by some noumenal cause such as God. But within the temporal order of natural causality (which tolerates no absolute spontaneity: A451/B479), each and every exercise of some externally implanted innate disposition is conditioned or triggered by antecedent foreign causes (cf. GMS, 4:446). This leaves no room for autonomous self-­determination or (more than relative) spontaneity in our acts of reasoning and judgment: as mere parts of the natural order of things we are always externally conditioned, both in receiving and in exercising our capacities for practical reasoning and judgment.

IV.6.  The (Potential) Instability of Evolutionary Moral Skepticism In this section I consider a final strategy for dismissing genealogic anxiety that some contemporary philosophers have used against evolutionary moral skepticism. This objection can take several forms, but its general core is roughly as follows. The objector typically concedes the naturalistic premise (1) that the causal (e.g., evolutionary) mechanisms which generate our moral judgments are insensitive to the truth of these judgments.40 But the objector denies that the genealogical 40  This is not always the case: Schafer 2010 and (in a more roundabout way) Enoch 2010 argue that the mechanisms selected by evolutionary processes do track properties that have normative significance, such as pain or survival. The appeal to features like pain and survival may (perhaps) show how

Legislative Freedom and Kant ’ s Genealogical Anxiety  127 skeptic can infer from (1) the conclusion (2) that our moral judgments are unjustified or false. The rejection of this inference does not rest on the vague appeal to some alleged genetic fallacy. Rather, the objector’s point is that the inference from (1) to (2) is self-­undermining for naturalists: naturalistic-­genealogical skeptics must accept certain scientific beliefs (such as the beliefs that constitute evolutionary theory) or epistemological beliefs (such as the beliefs that yield a reliabilist view of knowledge or justification) which are or (at least) might also be generated by causal mechanisms which are insensitive to the truth of those beliefs.41 Thus, if genealogical skeptics insist that generally speaking, all beliefs that (might) have a truth-­insensitive causal ancestry are unjustified or false, they undercut the validity of the scientific or epistemological beliefs that ground their genealogical anxiety concerning moral beliefs. This argument hinges on the claim that our theoretical beliefs, just like our moral beliefs, are (or might be) generated by causal mechanisms that are insensitive to the truth of these beliefs. It is not obvious why naturalistically minded genealogical skeptics concerning moral judgments should concede this.42 (I consider this issue in Chapter VII.) But let us grant that genealogical skeptics must make this concession. Even so, the above case against genealogical skepticism about morality boils down to a mere companions in guilt type of argument. Arguments of this type have the strong limitation that they do not engage with the negative, undermining force of the skeptical considerations that they seek to defuse. Hence, such arguments cannot secure our positive entitlement to the beliefs targeted by the skeptic. In the present context, this means: the claim that the generalized inference from (1) to (2) would threaten both our theoretical and our moral beliefs cannot show that either type of belief is immune against the substantive challenge that is raised by the skeptical argument. This claim rather leaves us with the live possibility that both our moral and our theoretical beliefs lack objective rational credentials. Perhaps this shows (as I shall confirm in Chapter VII) that a genealogical skepticism concerning the reliability of our cognitive faculties which relies on evolutionary considerations is incoherent since the

the validity of desire-­based normative judgments can be reconciled with evolutionary theory, for such judgments focus on factors like pain that have a strong impact on our welfare or desire-­satisfaction. But the appeal to processes that track natural features like pain cannot account for the validity of essentially non-­hedonic moral judgments that prescribe how one must act regardless of one’s future pleasure and pain or even (for Kant; cf. KpV, 5:30) regardless of one’s survival. 41  Srinivasan 2019: 134 focuses on epistemological beliefs. Schafer 2010: 475–6 focuses on scientific beliefs. 42  For Schafer 2010, the naturalist needs an a priori assumption that our perceptual faculties are reliable. Presumably, his point is that naturalists cannot verify their theories without assuming that their perceptual faculties generate reliable results, where this assumption cannot (on pain of circularity) be perceptually justified. However, naturalists may not actually need to endorse this assumption; for instance, they might think that our perceptual faculties generate only misleading, inaccurate data (e.g., regarding secondary qualities) that reveal the relevant causal factors only indirectly via abstract theorizing (inferences to the best explanations, etc.).

128  Kant on Freedom and Rational Agency acceptance of evolutionary theory itself presupposes the reliability of our cognitive faculties. But the diagnosis that a genealogical skepticism which relies on a specific (e.g., evolutionary) naturalistic theory is unstable does not remove the threat that our judgments do have some naturalistic origin. Even if genealogical skeptics cannot consistently give a theoretical account of the specific naturalistic (e.g., evolutionary) mechanisms that generate our judgments, they can raise the specter that our (theoretical-­scientific and/or moral) judgments are the mere upshot of some (unspecified) naturalistic processes which, simply as such (namely, qua contingent, mutable, etc.), are insensitive to the rational features (e.g., objective necessities) that these judgments purport to represent. This is sufficient to cast a serious doubt on the rational validity of these judgments. Here is another way of articulating the limitations of the companions in guilt-­argument. One need not accept the naturalistic premise (1) as true in order to accept that its truth would establish (2). Kant himself rejects (1), but he invites us to consider the dire consequences that would obtain if (1) were true. One might accept (as Kant indeed does; cf. Chapter VII) that a generalized naturalistic genealogy for all our mental activity would undermine our theoretical judgments in addition to our moral judgments. But even so, one can consistently affirm the significant counterfactual: if an exclusively naturalistic worldview were correct, then we would indeed lack any objective, both practical and theoretical cognition. Some critics of genealogical anxiety-­arguments hold that if such arguments imply that we have no objective cognition at all, this reduces genealogical arguments to absurdity.43 I will argue in Chapter VII that Kant accepts a similar point. But for Kant we cannot avoid the “absurd” conclusion that all our representations lack objective rational credentials without rejecting the naturalistic premises which ground this conclusion. To deem this conclusion absurd is to presuppose a confident stance regarding our cognitive or rational powers, and it is precisely this stance which genealogical anxiety arguments challenge. To regain or substantiate a confidence in our cognitive powers one must show that genealogical anxiety is unwarranted because we can reject the naturalistic premises which produce such anxiety. In his version of genealogical anxiety, Kant accepts the inference from (1), the naturalistic claim that all our judgments result from the “mech­an­ ism of merely subjectively determining causes,” to (2), the conclusion that all our judgments lack cognitive or rational validity. For Kant, we cannot avoid this drastic conclusion without rejecting (1) and the associated naturalistic worldview where “all causality is mere nature.” As I will argue in Chapters VII and IX, for Kant the patent absurdity of (2) is a helpful tool for illustrating why a naturalistic worldview is unacceptable.

43  See White 2010.

Legislative Freedom and Kant ’ s Genealogical Anxiety  129

IV.7.  Conclusion: Kant’s Incompatibilism about Legislative Freedom of Will In this chapter I have reconstructed the first rationale for Kant’s claim that a naturalistic view of the human will is inadequate to our self-­conception as free moral agents. I have focused on the incompatibility Kant sees between naturalism and our legislative freedom of will qua practical reason. This is the kind of freedom we ascribe to ourselves when we engage in “actions through which [reason] prescribes laws” (A803/B831) that we deem “absolutely necessary” (MS, 6:226). Our acts of practical judgment can “yield us laws” as “products of pure reason” only if these prescriptive acts occur “not in an empirically conditioned but in an absolute manner” (A800/B828). A view on which “all causality” is “mere nature” entails that our prescriptive acts of reason and the resulting moral judgments are empirically conditioned by and contingent upon the “mechanism of subjectively determining” natural causes. Hence (according to the argument I reconstructed in Section IV.3), on a naturalistic view our moral judgments have no objective ra­tional force or absolute (unconditional) necessity. This result would be fatal to our ordinary moral self-­conception: as moral agents we presuppose that we are governed by objectively necessary norms which articulate intersubjectively valid practical reasons and which exert legitimate rational authority over our contingent, subjective-­private empirical character (including our empirically given desires). Thus, “if all causality were mere nature” our moral self-­conception would be illusory, a mere “figment of the imagination,” a “chimerical idea without truth” (GMS, 4:445). These considerations do not, of course, show that the naturalistic view of the human will is false or that we indeed have the requisite legislative freedom. (This is the topic of Part 4.) But they support Kant’s view that our moral self-­ conception is incompatible with a naturalistic view of the human subject. This, in turn, explains how Kant can motivate his metaphysically ambitious, anti-­ compatibilist and anti-­naturalistic conception of the human will qua norm-­ giving practical reason. Hence, my account in this chapter expounds one key source of Kant’s anti-­naturalistic incompatibilism about free will. In the following chapter I consider a further source by considering Kant’s views on our executive moral freedom of choice: the kind of freedom we ascribe to ourselves when we exercise our sensibly affected faculty of Willkür.

V Executive Freedom, Determinism, and the Categorical Imperative In this chapter I reconstruct the second argument grounding Kant’s rejection of a naturalistic- compatibilist conception of free will. According to the first argument, an exclusively naturalistic worldview implies that we lack the legislative freedom to prescribe absolutely necessary laws of practical reason. Now the question is: how must we conceive the executive freedom that we exercise in our acts of choice that stand under practical laws? I argue that for Kant the executive freedom that belongs to our law-­governed faculty of choice (Willkür) requires an absolute contingency that is incompatible with the deterministic causality of nature.

V.1.  Recent Views on Kant’s Incompatibilism about Freedom of Choice I begin by considering how Kant’s incompatibilism about freedom of choice (or action)1 has been received by recent commentators. On some readings Kant fails to offer good reasons for holding that a choice cannot be free insofar as it is determined by natural causes. According to one influential interpretation, Kant’s incompatibilism is based on a dubious hedonistic conception of our natural motive causes: Kant worries that if our choices were determined by natural desires we would always act for the sake of pleasure and a non-­hedonistic (“pure”) moral motivation would be impossible.2 Karl Ameriks makes an even dimmer assessment: Kant’s rejection of compatibilism reflects a mere, unmotivated “dogmatic attachment” to incompatibilism.3 Henry Allison traces Kant’s incompatibilism to the idea that free agents exercise “a certain control over” their desires by incorporating them into their maxims, i.e., by forming a rule determining which desires are worth satisfying. It is only once a given desire has been reflectively assessed and endorsed that the 1  For Kant, exercising the will in choices of general maxims, ends, or particular outer (physical) actions is itself an “inner” action (MS, 6:218). See Korsgaard (1996a: 179): “Making something your end is a kind of internal action.” In what follows, I will draw (but significantly expand on) Kohl 2015b. 2  Irwin 1984: 39–40; Wood 1984: 82–3. I consider this reading in Section V.4. 3  Ameriks 2000a: 227.

Kant on Freedom and Rational Agency. Markus Kohl, Oxford University Press. © Markus Kohl 2023. DOI: 10.1093/oso/9780198873143.003.0006

132  Kant on Freedom and Rational Agency desire (qua part of a maxim) can move us to act.4 This “Incorporation Thesis” is clearly an important part of Kant’s doctrine. However, it cannot by itself justify Kant’s incompatibilist commitments. Compatibilists might well accept this thesis and hold that the acts of choice through which we incorporate desires into our maxims are free insofar as they are causally determined by the agent’s reflection on reasons. As Allison notes, Kant’s conception of natural causality allows “for a rich and potentially attractive form of compatibilism” that includes among the determining causes not just material (e.g., brain) states or non-­ intentional impulses but also representations of reasons (cf. KpV, 5:96), including maxims qua subjective principles of choice (cf. KpV, 5:20; A546/B574–A550/B578).5 But this undermines the motivation that Allison supplies for Kant’s incompatibilism, because it gives compatibilists the resources they need to endorse the claim that we exercise reflective control over our desires when we incorporate these desires into our maxims. A “rich” kind of compatibilism seems especially compelling in cases where our reflection on reasons causally determines us to make the right choice (e.g., not to act upon a desire when doing so would be impermissible). Why would an incompatibilist insist that a free, morally appropriate choice must not be causally determined by our practical reflection and the resulting desiderative states?6 The difficulties in seeing the point of Kant’s incompatibilist pronouncements might be one source of inspiration for a compatibilist reading where Kant holds that determinism merely appears to threaten our freedom of choice via the predictability of our actions: we can only adopt the practical standpoint of agency if we assume that we have deliberative options, and predictive knowledge of how we will act would undermine this assumption. As we saw in Chapter II, this threat can be defused through the compatibilist argument that even in a deterministic world, predictions of our actions are either impossible (when the purported prediction has a causal impact on our decision-­making) or must not interfere with our deliberative choices, which requires that we set them aside and do not accept them as true. In a deterministic world, my practical deliberation and choice remain indispensable conditions of my agency that theoretical predictions cannot  bypass or preempt: I cannot know that I will (not) perform some action unless I  choose (not) to perform that action based on my representation of 4  Allison 1990: 39. 5  Allison 1990: 34; cf. Chapter II. Pace Hatfield (1990: 16–17), for Kant a compatibilist qua metaphysical naturalist is not committed to materialism (or to the merely epiphenomenal character of the mental). 6  In his recent work, Allison says that for Kant our choices must be transcendentally free since the unconditional spontaneity of reason “is the essential presupposition in Kant’s account of the free power of choice” (2020: 261–2). This is clearly true, but what justifies this presupposition? Allison again defers to the Incorporation Thesis but again does not explain why the incorporating act of choice “through which a reason to act is made into a sufficient reason” (2020: 277) must be regarded as free from determination by natural processes of reflection.

Determinism, and the Categorical Imperative  133 normative-­practical reasons. Hence, determinism does not deprive us of alternative epistemic possibilities: for a deliberating agent different actions are “possible for all [she] could possibly know.”7 In Chapter II, I argued that Kant can avail himself of this argument to defuse the epistemic threat that determinism, via predictability, seemingly raises for free will. But this does not show that Kant deems our freedom of choice compatible with determinism or that “Kant’s answer to the question whether it matters if we are in fact . . . free [from natural necessitation] is that it does not matter.”8 For Kant free will is an absolutely spontaneous capacity that, as such, cannot be determined by natural causes. This is clear from his insistence that our freedom of choice requires “independence from everything empirical and hence from nature in general,” including representational or deliberative states conceived as parts of the natural order. If we did not consider our freedom as “transcendental, i.e., absolute” then our so-­called freedom would be no better than the pseudo-­freedom of a turnspit (KpV, 5:97). Since the epistemic threat that determinism raises to our freedom of choice via predictability can be defused in compatibilist terms and since Kant rejects compatibilism, it follows that Kant does not regard the predictability of our actions as the most significant threat that determinism poses to our freedom of choice: Kant’s incompatibilism about freedom of choice does not rest on the epistemic worry that “[predictive] knowledge could somehow take away our freedom.”9 But why, then, does Kant think that determinism is incompatible with our freedom of choice? What are his grounds for insisting that such freedom requires metaphysically real (rather than merely epistemic) alternative possibilities?

V.2.  The Conditions of Law-­Governed Finite Agency In my view, Kant indicates the grounds of his incompatibilism when he declares that “ought has no meaning whatsoever when one has only the course of nature in  view” (A547/B575). For Kant “the course of nature” as such is deterministic (A533/B561). So, when he assumes that one has “only the course of nature in view” he envisages a world where all activity results from deterministic causes since the exercise of every capacity “itself stand[s] under another cause determining it in time” (A533/B561). He claims that practical imperatives do not apply in such a world: “it is impossible that in [nature] anything ought to be different from the way it actually is in those [deterministic] time relations.” Thus “we cannot ask: 7  Bok 1998: 120 (cf. Moore 1912: 94; Wallace 1994: 2–4). 8  Korsgaard 1996a: 162; 176 (cf. Hill 1992: 135). For early versions of this compatibilist reading, see Matthews 1969 and Bennett 1974, 1984 (who, like Ameriks, denies that Kant provides good incompatibilist arguments). 9  Korsgaard 1996b: 95.

134  Kant on Freedom and Rational Agency what ought to happen in nature” (A547/B575). In a deterministic world, oughts would be meaningless and so our behavior would lack the normative dimension provided by meaningful oughts.10 We would then also fail to be accountable agents: we would not be proper subjects for moral praise, blame, and punishment.11 To see why Kant holds this view, notice first that deterministic natural causes render their effects nomologically (hypothetically, empirically) necessary (A201–2/B246–7): in an exclusively deterministic world all causes necessitate their effect(s) in accordance with natural laws (A534/B562). This includes processes of deliberative reflection and choice as the typical natural causes of our actions (KpV, 5:96–7): our empirical-­temporal mental states causally necessitate our acts of compliance or non-­compliance with practical norms. Now assume that it is impossible to do something other than what one is causally determined to do. (I discuss this assumption below.) If so, our practical deliberation never leaves us the double option to choose in conformity with or in violation of normative laws. Normative laws are principles that provide necessary reasons for acting which apply to every rational agent. The only normative practical laws are moral norms,12 which we apprehend as categorical imperatives. On my reading, Kant holds that such categorical imperatives (representing practical laws) govern our actions only if we have the metaphysically real options to choose in conformity with or in violation of these laws. Kant is led here by two commitments. First, Kant holds that normative (as opposed to natural) laws apply only to agents who can act according to their representation of these laws, i.e., who can correctly respond to the reasons such laws provide (GMS, 4:412). If A is truly incapable of correctly responding to normative reasons, i.e., if A is causally necessitated not to comply with a valid norm (because A’s reflection on norms is completely inefficacious or because it causally determines A to go against reason) then the norm fails to govern or direct A’s behavior.

10 Pereboom 2005 reads A547/B575 as asserting that oughts would be false in a deterministic world. But for Kant causally determined agents are not the types of beings about which one can make true or false ought claims. For instance, the judgment that we ought to lie when this benefits us is a false, yet meaningful claim about free agents. 11  As I stressed in the Introduction, although Kant’s concerns about (moral) responsibility play an important role in his discussion of free will they do not yield his primary motive for incompatibilism. Rather, these concerns are based upon his views about the possibility of ought-­governed agency. For Kant, legitimate judgments of praise or blame (and related expressions of accountability, such as reactive sentiments) are grounded in the prior recognition that agents did or did not act as they ought to have acted. In other words, we praise, blame, or punish agents for complying or failing to comply with moral oughts. Here we presuppose that agents are governed by meaningful oughts. When Kant says that without free will “no moral law, no imputation according to it is possible” (KpV, 5:97), he indicates that unfree subjects lack imputability because they are not bound by a normative law that could yield the rational basis for deeming people praise—­or blameworthy. 12  This claim is controversial. I defend it in Section V.6 (and in Kohl 2018b).

Determinism, and the Categorical Imperative  135 Second, Kant sees an irreducible difference between the way in which ra­tionally perfect (holy) and imperfect wills are influenced by their recognition of normative laws (cf. Chapter III). A perfectly rational will’s recognition of a normative law leaves that will with no option but to do what the law says must be done. By contrast, an imperfect will’s ability to conform to the reasons provided by normative laws has an ontological shadow or privation (MS, 6:227), namely, the propensity to violate these laws. Due to this propensity an imperfect “will is by its nature not necessarily obedient” to laws of reason (GMS, 4:413). An imperfect will’s ­recognition of the right reasons does not necessarily lead to right action because such a will faces alternatives or obstacles to right action (GMS, 4:394). Given the presence of these alternatives, the rational influence of normative laws cannot make it impossible for finite agents to violate such laws: the idea that imperfectly rational agents stand under normative laws presupposes the option of non-­ compliance. When imperfectly rational deliberators recognize the force of normative reasons, any rational influence would leave them with the option to go against those reasons: “inevitable” determination by reason is the privilege of a perfectly rational being (GMS, 4:412–13). If a finite deliberator’s attendance to normative laws does inevitably determine them to act and hence deprives them of alternative options, it cannot be reason that determines them: instead, they are necessitated by non-­rational factors that belong merely to the space of ­natural causes. To accommodate the fact that imperfect wills cannot be inevitably determined by reason, determinists may suggest that an imperfect will’s appreciation of the right reasons can only bring that will close(r) to right action: compliance with norms occurs only when further causal factors other than the agent’s rational acceptance of norms contribute to the deterministic nexus. But this suggestion renders our rational powers impotent rather than imperfect: since our appreciation of the right reasons cannot by itself lead us to act, our actions fail to fully accord with normative laws. Here we lack the ability to act from duty, which requires that “the thought of duty be of itself a sufficient incentive” (MS, 6:393) for making the right, rationally required choice.13 For Kant, the presumption that we possess the sufficient rational powers required for truly lawful agency is central to our deliberative self-­awareness.14 13  This problem persists if we conceive natural causes in an indeterministic fashion. If natural causes are probabilistic, they make it (perhaps highly) improbable for immoral agents to act from duty. Moreover, when agents act rightly, their rational recognition of moral norms qua merely probabilistic cause is never quite sufficient for compliance—­it merely renders compliance more likely. These implications of naturalistic indeterminism also turn our rational imperfection into rational (quasi-)impotence. 14  Hence his frequent appeal to the idea that pure practical reason can of itself determine our will (KpV, 5:15, 19, 28, 42). This is consistent with saying that determination by reason is the privilege of the Holy Will because these are two different notions of determination. Determination of the Holy Will by reason is “inevitable” whereas for imperfect wills “the determination of [our] will according to objective laws is a constraint” (GMS, 4:412–13). Determination of the will via practical constraint by

136  Kant on Freedom and Rational Agency Thus, for Kant the deliberative plight we face as finitely rational agents is that our recognition of normative laws of reason can by itself suffice for our lawful agency but also leaves us the option to defy such laws (cf. Chapter III). The difference between the way in which perfect and finite wills are influenced by their recognition of normative laws comes to light in the fact that only finite agents apprehend such laws as oughts. For a divine being whose “volition is already by itself necessarily in unison with the law . . . the ought is . . . out of place”; ought “represents the practical rule in relation to a will which does not immediately perform an action simply because it is good” (GMS, 4:414). Thus, the function of a categorical ought is to represent objectively right reasons to agents who have the freedom to choose rightly or wrongly. This freedom of Willkür is “the ground of the possibility of categorical imperatives” (MS, 6:222).15 Kant holds that in a deterministic world, agents are always deprived either of the option to act rightly or of the option to act wrongly. Hence, he denies that categorical imperatives exist in such a world. Kant’s incompatibilism thus rests on his acceptance of two claims: (1) If one is under a normative obligation to act in a certain way, then one can act otherwise than one in fact does. (2) If one is causally determined to act in a certain way, one cannot do otherwise. For Kant, (1) and (2) show that causal determinism is incompatible with the kind of freedom of choice that allows imperfectly rational agents to be governed by normative-­practical laws. Kant accepts (2) as a consequence of the realistic-­naturalistic supposition that all efficient causes are of the natural type and thus empirically determine their effects. Under this supposition, we cannot act differently than we in fact do. Kant here assumes that natural causes and their governing laws possess genuine modal, nomic force. As phenomenal parts of the natural order we cannot act contrary to the influence of natural causes that render their effects nomologically necessary: we cannot defy causal conditions whose presence indicates that some effect must occur as a matter of necessary natural law, where these causal conditions are themselves the nomologically necessary effect of preceding causal conditions, and so on in an indefinitely extending chain of causes whose temporally distant members pre-­date our existence. reason is rational necessitation, which (unlike inevitable divine determination) results in a merely contingent conformity to normative laws: “actions which are recognized as objectively necessary are subjectively contingent” (see Chapter III). 15  One might object (cf. Lavin 2004: 448) that at MS, 6:226–7 Kant rejects the idea that we can define free will “through the capacity to choose to act for or against the law.” But here Kant wants to deny the proposal that we can explain free will through the experience of wrong actions. My analysis respects his denial: my appeal to a non-­deterministic influence of reason on the will seeks to capture our practical self-­conception but not to provide a theoretical understanding of free will. Moreover, my account is not based on the experience (in Kant’s technical sense) of wrong action but on an a priori consideration of the different ways in which laws of reason influence perfect and imperfect wills. Finally, my account does not purport to define the positive power of free rational choice through the negative-­privative option to go against reason (cf. Chapter III).

Determinism, and the Categorical Imperative  137 Compatibilists might object that taking nomological modalities to fix the real options of empirical agents begs the question against compatibilism. This objection places the burden on compatibilists to explain in what sense agents who are nomologically necessitated not to do x nevertheless can do x. Kant’s critique of the comparative concept of freedom (at KpV, 5:96–7) implies how he would respond to one prominent compatibilist proposal. According to the comparative concept, freedom is the ability to do what one wants in the absence of external impediments that would otherwise stand in one’s way. This implies that agents are free to (can do) x just in case they would do x if they wanted or chose to do x. This the core idea behind the compatibilist conditional analysis of “can.” Kant rejects this idea because the internal empirical mental processes, wants, and choices which determine our physical acts belong to the deterministic chain of nature and are therefore themselves determined by preceding states. Suppose R chooses not to do x. Insofar as R’s choice not do x is nomologically necessitated by internal states such as R’s representations and desires which are themselves nomologically necessitated by antecedent natural states (etc. etc.), R cannot choose to do x.16 But if R cannot choose to do x, then the counterfactual that R would do x if they chose to do x fails to show that they really can do x.17 Recent compatibilists have tried to reject (2) via more refined, increasingly technical analyses of “can.”18 But considering how Kant might respond to such maneuvers would take me too far afield. The more pertinent question is how Kant defends the claim (1) that our agency under normative laws requires that we have metaphysically real alternative possibilities. Arguably, the best strategy for compatibilists is to resist (1): as Jay Wallace puts it, compatibilists give away the game when they accept the significance of alternative possibilities because they “are then compelled to rest the case for compatibilism on distinctions between

16  As Kant puts it, “internal” and “psychological” causes, conceived as parts of nature, stand “under the necessitation of conditions of past time, which therefore, when the subject has to act, are no longer in his power” (KpV, 5:96). 17  To illustrate: suppose that a phobia determines R not to choose x (cf. Lehrer 1968). In response, compatibilists might appeal to a further conditional, as G. E. Moore does (1912: 94) when he holds that R can choose to do x if: R would choose to do x if they chose to choose to do x. But this invites an infinite regress, “a proliferation of conditionals and a proliferation of objections” (Wolf 1980: 154). See also Chisholm 2003: 28–30. 18  An exception to this trend is the strategy (employed by Reath 2006: 255–66; Scanlon 2003; Wallace 1994; Wolf 1990, 2003) to focus on a general unanalyzed notion of a capacity: we have a general capacity to respond to objective norms (and a corresponding “counter-­normative capacity”: cf. Wallace 2006) whose existence is not threatened by determinism. However, general capacities have no real significance for agents apart from their ability to exercise them in concrete choices under specific conditions. As Wolf admits (2003: 387), “according to our commonsense understandings, having [an] ability is one thing and exercising it is another.” Thus, the appeal to general abilities is insufficient to account for an agent’s real options: for instance, my general capacity to run does not provide me with a real option to run on a specific occasion where the reigning natural conditions make it such that I have a muscle spasm which prevents me from moving my legs. For critiques of general capacity compatibilism, see Locke 1973–4: 179; van Inwagen 1983: 13.

138  Kant on Freedom and Rational Agency different senses of opportunity . . . that seem too fine-­grained and technical to do the normative work required of them.”19 (1) has two sub-­components. First, ought-­governed agency requires the ability to act for the right reasons; second, such agency requires the option to go against reason. The first point is perhaps more intuitive. Kant here assumes that if an agent is truly incapable of responding to the reasons provided by a normative law there is no relevant sense in which the law governs her action: for an action to be governed by a normative law of reason the agent must be capable of understanding and correctly responding to the right reasons. Many philosophers, including compatibilists, accept this claim or the cognate principle Ought Implies Can.20 But Kant has a unique way of justifying this principle, which connects the sources of moral normativity with the sources of rational motivation.21 As we saw (in Chapters  III–­IV), for Kant the origin of moral principles lies in the faculty of pure, “universal practical reason” that brings forth the representation of moral norms through a self-­legislative deed. If we lacked the rational power to choose in accordance with moral norms, then we could not legislate these norms as valid principles for our rational choices: our rational will would undermine its legislative authority if it tried to impose on itself normative laws that it cannot execute or conform to. In commanding itself to do things that it cannot do, the subject would misapprehend its own rational powers and would thus be the victim of a deep illusion or self-­deception. The pseudo-­norms arising from such an inherently confused attempt at self-­legislation would lack rational validity, rendering our deliberative awareness of self-­imposed normative laws a mere figment of our imagination (GMS, 4:446). Thus, the idea that rational agents have the power to choose in accordance with normative principles derives from the idea that these

19  Wallace 1994: 223. Dennett 1984 also holds that the best compatibilist strategy is to deny the significance of alternative possibilities. As Leiter puts it, compatibilist analyses of “can do otherwise” appeal only to “certain academic philosophers who think the need to be a self-­caused agent is superfluous, something that can be finessed via some adroit dialectic moves” (2015: 72) which Kant would dismiss as “petty word-­juggleries” (KpV, 5:96). Consider, for instance, Lewis’s local miracle compatibilism (Lewis 2003), which rests on the distinction between causing and (merely) entailing a violation of the laws of nature. Beebee 2003 argues that this fails to track any relevant difference. For Lewis, a counterfactual choice was possible for a causally determined agent if the possible world where this choice occurs is sufficiently close to (nearby) the actual world. But it is hard to know what to make of these technical modal ideas or how they could save our ordinary normative self-­conception. 20  Compatibilists who accept the “Ought Implies Can” principle include Nelkin 2011 and Wallace 1994. This principle is not uncontroversial: for recent criticisms, see Graham 2011. However, these criticisms do not pertain to Kant’s version of the principle. Contemporary debates focus on the idea that moral ought implies the ability and opportunity to perform specific physical acts (Graham 2011: 341–2; Haji 2002: 21–2), whereas for Kant moral ought implies the volitional power to choose the right motive for the correct reasons (GMS, 4:394; KpV, 5:36–7). This is due to Kant’s conception of morally right action, which requires that one earnestly strives to produce certain physical effects but not that one succeeds in doing so (see Kohl 2015d). 21  In what follows I give an abbreviated version of my argument in Kohl 2015d. I return to this point in Chapter VIII.

Determinism, and the Categorical Imperative  139 principles are valid laws of autonomous reason that rational agents legislate for themselves, rather than heteronomous rules which are externally imposed upon agents by foreign causes (and which are thus not inherently attuned to the ­ra­tional powers of norm-­addressees). “For this moral law is founded on the autonomy of  the will of [the subject of the moral law], as a free will, which according to its own general laws must, at the same time, necessarily be able to conform to that to which it ought to submit” (KpV, 5:132). But this does not explain why Kant endorses the second sub-­component of (1): the view that ought-­governed agency also requires that ought-­addressees can go against reason.

V.3.  Against Anodyne Analyses of Rational Imperfection As Douglas Lavin has shown, the notion that governance by norms requires the possibility of violation is accepted by many contemporary philosophers.22 However, the vague idea that finite agents can go against reason may be interpreted in different ways.23 For Kant, our rational imperfection implies the real option to go against reason whenever we act under practical laws: we can never be inevitably determined by reason to act rightly. Call this Strong Imperfection.24 By contrast, according to Weak Imperfection we can understand the possibility to go against reason in a weaker sense that undercuts the incompatibilist implications of (1): we can honor the fact that meaningful oughts apply only to imperfectly rational agents without treating the possibility of violation as a condition on all individual ought-­governed action. On this view finite agents can be inevitably determined by reason to act rightly, but this leaves two senses in which it is still possible for them to go against reason. First, the notion of possibility may be interpreted diachronically: even if finite agents are on particular occasions causally determined to comply with normative laws, one may account for their imperfection by considering their agency across time, i.e., by picking out instances of non-­compliance which attest to the fact that they sometimes have the option to act wrongly. Second, “possibility” may be interpreted as mere conceivability. If our recognition of a normative law inevitably determines us to compliance, this is 22  See Lavin 2004: 424–5. 23  See Lavin 2004: 426–7. 24  A journal referee claimed that Kant’s view here is similar to that of Haji (2002, 2012), who also argues that if one morally ought to do x one can refrain from doing x (2002: 34–5). However, Haji arrives at this view through a semantic analysis of deontic concepts: the idea of rational imperfection plays no role in his view. His argument relies on the claim that “ought”/“ought not” are conceptually equivalent to “right”/“wrong,” as well as on the following principle: if concepts (“ought,” “right,” “wrong”) are in the same deontic family, then, barring a good reason to think otherwise, they have the same freedom condition (2012: 25, 50). But for Kant, “right” and “wrong” are not equivalent to, and do not involve the same freedom condition, as imperatival representations of right—­or wrongness because “right” (or “good”) applies to a Holy Will whereas “ought” applies only to finite agents.

140  Kant on Freedom and Rational Agency a hypothetical necessity: our compliance depends on determining conditions that are empirically contingent and that might, conceivably, have been otherwise so that we might have acted wrongly. By contrast, a perfectly rational being’s compliance with practical laws is unconditionally necessary and its non-­compliance is inconceivable. This, one might suggest, suffices to distinguish imperfectly rational agency from divine agency under normative laws.25 In this section I raise two worries about Weak Imperfection that draw on Kant’s conception of our human condition. The first worry is that Weak Imperfection downplays the impact that our ra­ tional imperfection has in cases where our appreciation of norms leads us to choose correctly. Weak Imperfection allows that if an imperfectly rational deliberator acts rightly due to her recognition of the right reasons, such recognition removes her option to choose wrongly. This conflicts with the idea that even in those cases where our deliberation leads us to choose correctly, the conditions that render our rationality imperfect (such as proneness to temptation) inveigh against the rational influence exerted by our consciousness of normative laws. Weak Imperfection allows that in such cases the conditions that seek to divert us from choosing correctly are rendered powerless by our recognition of normative laws. One might respond that the influence of these conditions is not negated if the right choice occurs as the result of a process of deliberative competition between various motives. But the notion that the factors which account for our rational imperfection truly compete against our power to choose for the right reasons requires that these factors have a chance of success, i.e., that they may sway us to choose the wrong thing. If it is impossible for us to choose the wrong thing, there is no such chance of success. The concession that our rational imperfection cannot successfully interfere with our consciousness of rational norms conflicts with the idea that for finitely rational agents like us, responding correctly to ra­ tional norms involves an open-­ended struggle against the temptation to deviate from laws of reason (KpV, 5:32, 74–5). One might wonder in what sense a competing desire has more influence in a case where one chooses freely not to act upon the desire than in a case where one is causally determined not to act upon the desire. But a desire whose presence yields a genuine option to deviate from reason clearly affects (and afflicts) 25  While Weak Imperfection is not explicitly advocated in the literature, it is implied by a popular view. That view grants the “trivial” (Lavin 2004: 441) conceptual fact that ought entails the possibility of violation but denies that agents governed by oughts have the liberty to accept or reject the right reasons. Thus, on this view the notion of a possibility to act wrongly contained in the concept “ought” must be weaker than the notion that agents have the liberty to act wrongly in all their ought-­governed actions. My account of Weak Imperfection seeks to articulate what this weaker notion amounts to. The view I have in mind is well-­put by Lavin 2004: 441–9. Lavin follows Wolf ’s “Reason View,” according to which the ability to act in accordance with reason is sufficient for freedom (Wolf 1990: 61–2). A weak conceptual sense of the possibility to go against reason is also suggested by McDowell’s appeal to a merely “potential gap” between ideal and actual motivation (1998b: 105).

Determinism, and the Categorical Imperative  141 deliberative choice in a much deeper way than a desire that one cannot act on. To see this, consider how the fact that one is causally determined not to act on a tempting desire affects our deliberative self-­awareness. A deterministic causal process may involve, as parts of the causal chain, events that cause the psychological experience of temptation or struggle. But there is something illusory about this experience in cases where one is causally necessitated not to act upon the allegedly tempting consideration: it seems essential to genuine temptation and struggle that what one is tempted by, or what one struggles against, can sway one to act. Similarly, it seems constitutive of the (practical) experience of genuine temptation that one takes oneself to be aware of a real (not merely epistemic) possibility that one may give in to the tempting motive. Consider a paradigm case of temptation: Tom feels drawn to cheat on his trusting wife despite recognizing that doing so would be wrong. Here Tom’s experience of temptation revolves around a lively representation that he may yet choose to favor his carnal desire, which presupposes his belief that this option is truly open to him. If Tom lacked that representation, he would not regard himself as being engaged in a struggle against a genuine threat of deviating from reason. Moreover, if Tom manages to resist the desire to cheat due to his rational awareness of the practical norm that declares cheating to be wrong, his choice to prioritize right reason over his desire is accompanied by a sense of constraint (KpV, 5:32; 80). Acting under the awareness of constraint implies that one represents oneself as renouncing an attractive option that one in fact possesses. Hence, if it is not really possible for Tom to act on his desire to cheat, there is something deeply misleading about the awareness of constraint that accompanies his choice to prioritize normative laws of reason over the bidding of this desire.26 Could defenders of Weak Imperfection not be happy to account simply for the illusory experience of struggle, temptation, or constraint (in cases where one is causally determined to choose correctly)? Weak Imperfection seeks to show that doing justice to the idea of rational imperfection does not require Kant’s stronger interpretation. If Weak Imperfection entails (via its allowance that imperfect wills can be inevitably determined by reason to act rightly) that characteristic aspects of our practical self-­awareness are illusory, this shows that Kant’s stronger account of rational imperfection is more adequate to our deliberative self-­conception. There is a further Kantian worry about Weak Imperfection. As we saw, this view locates our imperfect rationality in two features: in acts of non-­compliance across time and in the fact that our compliance is only ever hypothetically necessary so that non-­compliance is always conceivable. The first of these features cannot be definitive of our imperfection (it can only illustrate it): the extent to which we are non-­ compliant across time is itself contingent whereas our rational 26  I take the “bidding” metaphor from Schapiro’s helpful 2009 account of how inclinations impinge upon deliberation.

142  Kant on Freedom and Rational Agency imperfection is a necessary aspect of the human condition. Hence, only the second feature allows Weak Imperfection to specify an idea of imperfection that applies to all finite agents as such. Now imagine Steph: due to her place in the causal nexus, she is always causally determined (via her deliberative reflection) to comply with normative laws. Steph’s will is distinguishable from a Holy Will since we can imagine that she would be susceptible to real temptations if the causal (e.g., psychological) conditions that in the actual world necessitate her compliance were different. But Steph’s case makes vivid that according to Weak Imperfection, depending on how the natural world happens to go the presence of real options to deviate from reason may completely vanish for human agents. This conflicts with the conviction that our rational imperfection, including our susceptibility to genuine temptations, is a deep structural feature of the human will whose grip on a free human agent cannot be reduced to merely counterfactual relevance by the contingent course of events. Weak Imperfection’s sense in which Steph can still be considered imperfectly rational seems too anodyne precisely because her imperfection has been turned into an object of idle counterfactual speculation that does not genuinely afflict her actual powers of choice or her (veridical) practical self-­ awareness.27 Weak Imperfection cannot account for the conviction that there is something missing here, something which deliberators who are aware of real options to deviate from normative laws deem integral to their self-­conception as imperfectly rational agents. According to Weak Imperfection Steph counts as imperfectly rational in the one sense that defines what it means to be imperfectly rational, even though none of her choices involve any real option to go against laws of reason.

V.4.  The Grounds of Strong Imperfection Even if the preceding considerations suggest that Weak Imperfection is too weak, they do not positively identify those structural features of finite agency that motivate Kant’s stronger interpretation of our rational imperfection: the features which account for our pervasive real option to deviate from laws of reason. How does Kant conceive of these features? Consider a mundane example of law-­governed action: A finds a lost wallet with five thousand dollars and returns it to its owner. If Strong Imperfection implied that every finite agent must overcome a temptation to keep the money, it could easily be dismissed as resting on a false psychological generalization.

27  One cannot respond here (in the vein of Lewis 2003) that invoking nearby possible worlds where Steph acts wrongly shows that she has the real option to act wrongly: the whole point of Weak Imperfection is that we can use the counterfactual possibility of wrong action (as designating mere conceivability) to replace the idea that human agents always possess real options to act wrongly.

Determinism, and the Categorical Imperative  143 However, for Kant Strong Imperfection is true because we may always choose to act upon some one of the empirically given desires that affect us as finite agents: this may be (i) a greedy desire to keep the money, or (ii) a desire to return the money that is fueled by a fear of being reprimanded, or (iii) a desire to return the money that is based on the want that others be happy, etc. If A acts on the basis of (ii) or (iii), A’s act of returning the money contains the letter but not the spirit of the moral law (KpV, 5:71–2; RGV, 6:30). But why does Kant posit an opposition between acting on the basis of benevolent empirical desires such as (iii) and acting on the basis of moral-­practical laws? For some, Kant’s claim that acting on benevolent desires displays a rationally deficient motivation derives from his view that benevolent empirical desires aim at the at agent’s own pleasure so that acting on the basis of moral reasons is the only way to counteract our self-­centered hedonistic motives.28 If this were true, Kant’s endorsement of Strong Imperfection and thereby his incompatibilism would rest on the controversial hedonistic view that all natural-­empirical desires aim at personal pleasure whereas only pure moral motivation is independent of the egocentric desire for pleasant sensations. Whether Kant is committed to such hedonism is a matter of debate.29 But we can sidestep this debate here since Kant’s endorsement of Strong Imperfection is not driven by concerns about hedonistic motivation. Suppose that my benevolent desire (iii) aims not at the pleasure I anticipate getting from making others happy, but (solely) at their happiness. This desire, qua empirical state, depends on my  contingent psychological make-­up, on my “love for people and affectionate benevolence” (KpV, 5:82). These emotional states lead me to identify empathetically with the feelings of others, which makes it hard for me to bear their ­(anticipated) distress. Now, normative laws have universal objective validity: they provide reasons for acting that apply to every rational agent (cf. Chapter IV). The validity and authority of the norm that prescribes returning lost money to its owner is independent of whether the addressee happens to have benevolent sentiments. Thus, if my choice to return the money is based on such sentiments, it is not responsive to the objective reasons for returning the money that are represented by practical laws: rather, my motive (iii) is “only subjectively valid and

28  See Foot 1972; Irwin 1984; Wood 1984. 29  See Allison 1990: 102–3. Following Reath 1989, Allison denies that Kant is committed to he­don­ ism about non-­moral motivation. (But in a more recent appendix to his earlier essay, Reath clarifies that he wants to attribute to Kant a “sophisticated hedonism” after all; 2006: 56.) For the view that Kant accepts a hedonistic view of non-­moral motives, see Beck 1960: 92–102; Engstrom 2009: 70f.; Johnson 2005; Wood 1984. I think this view is correct. There is strong textual evidence that for Kant our general end of happiness, which comprises all our more specific empirical desires, aims at a maximum of pleasure or contentment (see GMS, 4:418; KpV, 5:22; KU, 5:208; MS, 6:480). This need not imply that we always aim at pleasure in all our intentional (non-­moral) agency: an overall policy of maximizing pleasure might require bracketing thoughts about pleasure in one’s regular everyday activity (see Railton 1984). I discuss Kant’s views on non-­moral motives in Kohl 2017b.

144  Kant on Freedom and Rational Agency merely empirical, and . . . lack[s] the necessity which is represented in every law, namely, an objective necessity arising from a priori reasons” (KpV, 5:26). A motive that aims at satisfying a contingent empirical desire cannot coincide with (i.e., cannot capture the spirit of) a motive that is based on the recognition of necessary reasons, because those reasons apply to every rational agent and so the motives they supply must differ from the motives supplied by contingent natural (including benevolent) desires that merely happen to affect only some agents depending on their private conditions. Kant’s view can be confirmed in the following twofold manner. First, if the motivating ground of my benevolent acts is my affectionate want that others be happy, then all it would take for me not to act benevolently is a lapse of affection, due to (say) a sudden depression induced by a drop of my hormone level which dampens my desire to make others happy and thus interferes with the motivational grounds of my benevolent acts. So, my willingness to make others happy here strictly rests upon egocentric (subjective), non-­rational states and circumstances (such as the development of my hormone level). Moreover, assuming my benevolent psychology remains as it is, my willingness to choose from my affectionate sympathy with others may well prompt morally wrong physical acts (that do not even contain the letter of the moral law; cf. RGV, 6:30). Suppose an irresponsible student explains the terrible consequences that failing my class would have for them and asks, with tears in their eyes, for the chance to re-­take the final. If my affectionate desire that others do not suffer constitutes the motivational ground of my choices, I shall (impermissibly, unfairly) grant the re-­write. Cases like these illustrate Kant’s insistence (e.g., at KpV, 5:19) that true responsiveness to objectively necessary moral reasons cannot be simulated by motives which depend on subjective psychological conditions such as contingent desires, even if those desire happen to be benevolent ones.30 The other side of this coin is that a free agent whose empirical psychology happens to be such that the suffering of others leaves her emotionally cold has the rational power to fully comply with rational demands of beneficence (GMS, 4:398). Kant’s basic point here can be summarized through the following argument. (1) Finite agents are always (via their sensible nature) affected by empirical desires that impel them choose on their behalf, i.e., to make their satisfaction a condition of choice.31 (2) Practical laws give necessary reasons for acting that are not based 30  However, Kant denies that when I (e.g.) return a lost wallet to its owner from a motive that accords with the normative spirit of the moral law, the moral worth of my action is decreased if I also have a benevolent desire to see the owner happy, as long as my act is not based on this desire. Likewise, the moral worth of my good action here is not increased if it involves a struggle against my greedy want to keep the money. See Herman 1996: 12, 21. 31  “. . . we find our nature as sensible beings constituted in such a way that the matter of the faculty of desire . . . first presents itself to us; and our pathologically affected self, although it is in its maxims quite unfit for universal legislation, yet, just as if it constituted our entire self, strives to put its pretensions forward first, and to have them acknowledged as the first and original” (KpV, 5:74).

Determinism, and the Categorical Imperative  145 on empirical desires. (3) Thus, finite agents are always affected by desires that impel them to choose contrary to the spirit of practical laws.32 According to Strong Imperfection, we always have the real option to go against reason. The above argument traces this omnipresent tendency to the fact that we are persistently affected by non-­rational, empirically given desires. This argument identifies the sources of Strong Imperfection without invoking a hedonistic conception of non-­moral, natural motives: Kant’s point is that all our empirically given desires, regardless of their content or their object, incline us to choose on their behalf and hence impel us not to choose according to necessary reasons whose validity and normative force is independent of those very desires. The supposition that empirical desires impel us towards their satisfaction and thereby provide us with potential motives for choice seems plausible: to have a sensible need or desire just means that one feels attracted or drawn towards realizing its object, towards obtaining something that one wants but lacks. I now clarify the implications of Strong Imperfection by discussing some objections that one might raise against it. A common complaint is that Strong Imperfection implies that we can always “make a choice for any principle” and that a human agent “has the ability to act against everything he believes in and everything he cares about.”33 This objection rests on a misunderstanding. While it is true that on the account I am ascribing to Kant human agents always have the option to deviate from laws of reason, our alternatives to acting for the right ­reasons are delimited tightly by the bounds of our empirical psychology, i.e., by our contingent desire-­base which (partly, not wholly) determines what we care about.34 Thus, Kant can concede that for someone with a strong natural affection for children, it is very hard (perhaps impossible) to ignore the plea of an injured child. But for Kant there is still the question of whether their choice is based on contingent benevolent feelings towards children or rather on the rational recognition of the value that the child has as a person, i.e., of the norm that prescribes that one must help the child whether or not one happens to have affectionate 32  “If a rational creature could ever reach this point, that he thoroughly likes to do all moral laws, this would mean that there does not exist in him even the possibility of a desire that would tempt him to deviate from them; for to overcome such a desire always costs the subject some sacrifice and therefore requires self-­compulsion, i.e., inward constraint to something that one does not wholly like to do. But no creature can ever reach this stage of moral disposition. For since he is a creature and therefore always dependent with respect to what he demands for complete satisfaction, he can never be quite free from desires and inclinations: and as these rest on physical causes, they can never of themselves coincide with the moral law, the sources of which are quite different . . .” (KpV, 5:83–4). 33  See Lavin 2004: 447 for the first and Wolf 1980: 153 for the second quote. 34  On Kant’s view rational agents also care (deeply, indeed) about the necessary ends and objectively right reasons supplied by morality (GMS, 4:457–8). Wolf (1980: 152–4) claims that it is incoherent to say that agents who act rightly are not psychologically determined and that they could do otherwise. She assumes that an agent can act on the basis of the right reasons only if he is determined by “his interests” which “are determined by his heredity or environment.” But this assumption seems based on the subjectivist idea that what counts as the right reason for R depends on R’s empirically given interests and circumstances.

146  Kant on Freedom and Rational Agency feelings toward children. For Kant, free rational agents do have the freedom to choose between these different motives. Resistance towards Strong Imperfection also comes from the Aristotelian tradition where ethical virtue enables a perfect harmony of reason and desire. The conflict between Kantian and Aristotelian moral psychology raises complex issues that I cannot discuss here in depth, but I want to suggest one fundamental reason why Kant rejects the Aristotelian view. A perfect harmony of ethical reason and natural desire requires that the objects of desire may completely coincide with the objects of ethical norms. For Kant this is impossible because contingent desires whose presence and strength rest on subjective conditions cannot impel us to act for objectively necessary reasons whose normative force and validity is independent of whether we possess those very desires (KpV, 5:34).35 Aristotelians might respond that there are objective, non-­contingent facts about what our natural desires truly aim at:36 these desires find their proper satisfaction and we achieve the universal human goal of happiness (eudaimonia) only when we follow the voice of practical reason and exercise the virtues (in a complete life; EN 1098a15–20). But this view rests on the idea that human beings have a natural telos that determines what the human good and the satisfaction of our natural desires truly, objectively consists in (EN 1097b22–3). Kant rejects this idea: for Kant, what makes individual people happy depends entirely upon their subjective, contingent, empirically given preferences (KpV, 5:20–1).37 Thus, Aristotle’s controversial, objectivist-­ teleological conception of true desire-­satisfaction and happiness grounds his view that there can be (for virtuous agents) a perfect harmony between objective rational norms and the intentional structure of our non-­rational desires. Accordingly, some Neo-­Aristotelians who leave behind Aristotle’s teleology ensure this harmony by assuming that the

35  Pace Wood (2013: 157), this means that Kant cannot allow for “cases where action for [necessary] reason[s] is easy and natural, harmonizing with empirical desires.” See KpV, 5:83–4, as cited in n. 32. 36 We can ascribe to Aristotle something akin to Kant’s concept of “non-­rational, contingent empirical desire.” Aristotle posits desiderative states that belong to the “irrational element in the soul,” to the “appetitive and . . . the desiring element” (EN, 1102a32–1103a3). He further accepts that these desires are contingent upon (what we would call) empirical factors such as upbringing (EN, 1103b6–25; 1114a4–21) which determine how one subjectively (perhaps mistakenly) conceives of happiness or living well (EN, 1095a12–25; 1097b1–23). 37  The disagreement here is not merely verbal. For Aristotle as for Kant, happiness is the ideal of a life that is pleasant on the whole (EE, 1249a18; EN, 1153b14–5; KpV, 5:22); and both hold that our conception of happiness is based upon what we find pleasant (EN, 1104b4–17; EE, 1227a31–1227b12; KpV, 5:23). For Aristotle, vicious agents are confused both about what gives them true pleasure and about what happiness consists in; consequently, their life is bound to be wrecked by pain and grief (EN, 1166b13–29) whereas agents who (continuously) exercise the virtues are immune against misery (EN, 1100b33–4) and derive true pleasure from their activities. By contrast, for Kant the claim that virtue yields the pleasure required for happiness is false as a judgment of experience (KpV, 5:114); both the (relative) happiness of (manifestly) bad people and the misery of (seemingly) good people are data of experience.

Determinism, and the Categorical Imperative  147 norms of practical reason depend upon the agent’s subjective desire-­ base.38 Hence, Kant would challenge Aristotelians to specify how the ends of our empirically given desires can completely coincide with the ends of morality without (on the one hand) sacrificing the objective necessity of moral norms and without (on the other hand) relying on questionable (pre-­modern) teleological assumptions about true happiness or desire-­satisfaction. Kant departs from Aristotle by endorsing a conception of virtue as continence, i.e., as a “moral disposition in struggle” (KpV, 5:84) against our contingent sensible desires.39 This invites a further worry: does Kant’s view imply, rather implausibly, that our deliberative activity is typically fraught with the experience of struggle against temptation? Struggle against temptation presupposes a reflective awareness of the threat that our grounds of choice might be usurped by unlawful motives. Kant denies that such an awareness accompanies our everyday activity: common human agents do not typically reflect on their reasons for acting because they typically act on the basis of maxims (KpV, 5:74), habitual volitional patterns that incorporate general policies to choose in certain ways. Such policies tend to foreclose reflection on many specific practical questions. Moreover, the extent to which agents become aware of (i) a temptation to choose from motivational grounds that are contrary to the spirit of the moral law is oftentimes contingent upon the (typically) rare occasions where they become aware of (ii) a temptation to act contrary even to the letter of the moral law. To illustrate: my maxim to tell the truth may have been based, all along, on my fear of bad consequences (in which case I have been violating the spirit of the law). But my reflective awareness of this fact may be sparked only when I find myself in circumstances where no such consequences loom and where I feel a conscious temptation to lie (thus to violate the letter of the law). Many agents are, because of luck and prudent risk-­adverseness which they mistake for real virtue, rarely confronted with temptations of the latter sort (MS, 6:392–3). Hence, the idea that it is always possible for finite agents to choose contrary to the spirit of practical laws does not entail that this possibility

38  I have in mind here chiefly Foot 1972 and Anscombe 1958, who reject the idea that there can be necessary laws of practical reason (categorical imperatives). By contrast, McDowell 1978 treats moral requirements as categorical imperatives that, in the case of virtuous agents, “silence” competing motives. But this is puzzling: the Aristotelian virtuous agent altogether lacks competing motives, so there is no need for any silencing here. Hence, as Anscombe notes, such agents do not grasp ethical norms as imperatives that constrain their activity. 39  Allison (1990: 114), Baron (1999: 199), and Herman (1996: 32) all deny that for Kant the struggle against inclination is a mark of virtue or moral worth; here they react against Henson’s claim (1979: 48) claim that Kant adopts a “battle citation” model of moral worth. Kant indeed denies that agents typically need to struggle against inclinations to perform external acts that conflict with the letter of the moral law. But, as we saw, for Kant we do have a pervasive tendency to adopt grounds of choice that violate the spirit of the moral law. And, for Kant it is the defining mark of human virtue that we incessantly engage our sensible inclinations in battle to overcome the “obstacles” confronting our attempts to respect (in our motives) the spirit of practical laws (MS, 6:394, 409).

148  Kant on Freedom and Rational Agency is always an object of conscious awareness in the form of temptation or struggle. Rather, the most typical manifestation of our rational imperfection is smug self-­ deception about our motives that derives from a want of reflection on the grounds of our choices and that makes us confuse prudent yet morally unlawful patterns of choice with genuine moral virtue (RGV, 6:38). One might propose that the appeal to self-­deception suggests a moderation of Strong Imperfection: instead of positing the metaphysical possibility that we may always choose contrary to the spirit of practical laws, why not say that it is always (merely) epistemically possible for us to choose incorrectly since we can never be certain that our particular acts or general maxims are truly lawful? Kant clearly accepts this epistemic point (GMS, 4:419; MS, 6:392), but only as the epistemic upshot of the metaphysics of imperfectly rational agency: for Kant the ubiquitous doubt about whether we respect the spirit of practical laws derives from the ubiquitous metaphysical possibility that we may fail to respect this spirit. One might argue that even if it is not always metaphysically possible for us to act on the basis of inadequate motives, we can always doubt the adequacy of our motives because we do not know whether or not this metaphysical possibility obtains in a given case. However, doubts about whether our grounds of choice are pure or corrupt are idle unless these doubts have a real basis. For Kant, the metaphysically real basis for such doubts is the possible corrupting influence of empirical desires on our grounds of choice (GMS, 4:419). Since we are always (to some degree) drawn to satisfy some such desires, the metaphysical possibility of choosing incorrectly generalizes to all cases of law-­governed finite agency. For Kant, the omnipresence of empirical desires that impel us to choose on their behalf is guaranteed by the fact that we are dependent beings or rational creatures. As such, we are always affected by some contingent, subjective want for something we lack that inclines us towards its satisfaction and whose motivational influence cannot coincide with the spirit of necessary, objective laws of reason (KpV, 5:83–4).

V.5.  Rational Imperfection and Akratic Action In explicating the grounds of Strong Imperfection, I have suggested that our pervasive tendency to violate (the spirit of) normative laws typically manifests itself in our unreflective, self-­deceptive prioritization of contingent empirical desires over necessary a priori moral reasons. Kant acknowledges related though subtly different types of practical failure: for instance, he discusses reflective processes of rationalizing, as permissible exceptions, choices whose moral badness we notice with a sense of discomfort.40 But does the rational imperfection in our moral 40  See GMS, 4:424, where he characterizes an agent who suppresses via rationalization their clear recognition that their choice is unjustified in the eyes of impartial moral reason. They persuade

Determinism, and the Categorical Imperative  149 agency reduce to the threat of self-­ deception or rationalization? Pre-­ philosophically, it seems that our rational imperfection may also, if not typically, manifest itself when we reflectively confront temptations to violate good principles of choice that we sincerely hold and continue to hold even while knowingly violating them: namely, in cases of akratic choice. Some commentators deny that Kant’s view allows for truly akratic choice. This denial arises from the conjunction of two claims. First (1), every free choice is: either, a choice of a general principle (maxim) that incorporates some desire, i.e., that authorizes acting on that desire; or, a choice of a specific intention or physical action that falls under a previously chosen general principle. This implies that we cannot freely choose any particular action apart from our principled, maxim-­ based endorsement of that action. Second (2), morally akratic choices occur when agents choose a particular action without such principled endorsement, by freely and intentionally violating their sincerely held good maxims.41 Claim (1) is essentially what Allison calls the Incorporation Thesis, which many commentators (following Allison) deem absolutely central to Kant’s views on free will. Hence it is not surprising that such commentators tend to deny that akrasia as defined by (2) is really possible on Kant’s view.42 By contrast, I think that Kant’s view does leave room for weak-­willed choices because he does not hold that each and every free choice involves either the adoption of a general maxim or the choice of some specific action that falls under (is approved by) that maxim. There is just one passage where Kant clearly espouses (1) (RGV, 6:23–4). It occurs within a special context: Kant here argues against the idea that there is a middle ground between acting from good or evil moral principles. According to Kant’s rigorism, human agents have either a good or an evil character (Gesinnung) which consists in a highest life-­governing maxim that either subordinates the pursuit of individual happiness to the moral law or vice versa. The adoption of this highest maxim must consist in a free choice, for otherwise it would lack a rational-­moral significance and thus could not ground an imputable moral character (RGV, 6:21–2). Agents freely incorporate both the moral law and the incentive to pursue their own happiness into their highest maxim by subordinating the themselves that they can allow themselves an exception to general moral rules because this exception is so very minor (unerheblich) and their individual circumstances and plights are so very special that others must see their legitimizing force. This differs from the kind of self-­deception described in the preceding section because rationalization may be a singular slip that is later exposed and condemned as such by the agent. But it may also lead to an engrained pattern of rationalization which gradually induces agents to adopt thoroughly corrupt general principles of choice that they (via habitual self-­ deception) deem morally good and praiseworthy. 41  Against (2), McCarthy (2009: 38, 215) holds that morally akratic agents act from evil principles. But if agents adopt evil principles, their morally bad choices accord with their general (misguided) conception of the good. There is nothing akratic about acting in accordance with one’s mistaken practical convictions. 42  See Allison 1990: 159.

150  Kant on Freedom and Rational Agency one under the other (RGV, 6:24), which makes for a good or an evil Gesinnung. Admittedly, what Kant says here suggests that he generalizes from this special case (the choice of a live-­governing moral character that involves ordering our two most general motives into a highest maxim) to every free choice. This would commit him to the view (1) that agents cannot freely, knowingly choose to act against the general principles that express their considered evaluative outlook unless they replace these principles with other maxims that express their newly adopted values. This would render akratic choices (as defined by (2)) impossible.43 However, Kant often implies that wrong actions are not always based on bad maxims that people mistakenly (via self-­deception) conceive as good. People may also knowingly act against truly good principles which they sincerely accept. He refers to the “deliberate transgression” of moral principles (Anth, 7:163); he discusses the self-­awareness of people who violate what they acknowledge as their duties (GMS, 4:424); and he considers cases where one recognizes the authority of the moral law in the very act of transgressing it (GMS, 4:455). In the Religion (the primary source text for (1)), Kant says that via introspection we can note actions that “are consciously contrary to the law” (RGV, 6:20) and he acknowledges our weakness or “frailty” as one distinctive type of moral imperfection that afflicts the human condition: namely, our failure to muster the self-­control that we need for complying with our sincerely held good maxims in hard cases (RGV, 6:29).44 The appeal to frailty also partly grounds his claim that a perfectly stable character is a rarity among human beings (Anth, 7:291–2). We act unstably insofar as we fail to exercise sufficient inner strength or self-­control to abide by the maxims that express our general moral character. 43  Some suggest that we can reconcile (1) and (2) by assuming that agents who act against their general good maxim choose to adopt an additional bad maxim as a principle for just this particular bad action (Hill 2011; Morrisson 2005; Timmermann 2022; Willaschek 1992: 242–3). But the idea of a singular particularist maxim violates the generality that attaches to every maxim-­content. It also violates Kant’s rigorism (since it allows that the same agent may simultaneously hold both good and evil principles). Moreover, if an agent performs a bad action on the basis of a newly formed maxim, then the agent is endorsing this action as justifiable and is therefore not (pace (2)) acting akratically against their better judgment. Timmermann (2022) tries to block this implication by arguing that the incorporation of a desire into a maxim is an elective act that need not be accompanied by a sense that one’s choice is justified. But for Kant, maxims are principles of volition that at least subjectively (from the agent’s point of view) justify the acts that fall under them (GMS, 4:400; KpV, 5:19–20). 44  Those who propose (1) at the expense of (2) suggest that when agents bemoan their frailty this betrays their adoption of evil principles and their attempt to feign (or deceive themselves into accepting) that morality is too hard for us (Allison 1990: 159). But nothing in RGV, 6:29–30 suggests that frailty is a merely feigned or imagined state of will. What Kant says here (and elsewhere: cf. KU, 5:264) implies that he views frailty as a genuine feature of the human condition, just like the other types of moral failures that he mentions (e.g., impurity of motivation). He defines frailty as “the weakness of the human heart in the following of adopted maxims in general” (RGV, 6:29). His further characterization of frailty draws on the standard contrast he also invokes elsewhere (e.g., at GMS, 4:412–13) between the objectively, rationally sufficient incentive to follow the moral law and our subjective tendency not to follow this incentive. This suggests that agents become truly evil—­namely, adopt evil maxims—­only once they go beyond frailty, i.e., by entering the second and third stage of evil (see Carnois 1986: 105). Of course, an initial state of frailty may well lead to further moral regress that culminates in adopting evil maxims.

Determinism, and the Categorical Imperative  151 If we take into account these textual data and acknowledge the special context in which Kant proposes (1), we can assume (exercising both interpretive and philosophical charity) that for Kant the Incorporation Thesis describes the typical but not the only possible way in which we exercise our capacity for free choice. Under this assumption, Kant’s doctrine of free choice can accept the pre-­ philosophical idea that finite agents occasionally act out of character by freely opting to act against their sincerely held good maxims.45 This is fully compatible with Kant’s rigorist idea that agents are either good or evil: in cases of akratic choice, agents who sincerely hold a morally appropriate highest maxim but choose to act against their better moral judgment do not thereby adopt any further, evil maxims that would then (according to Kant, per impossible) have to co-­exist with their good principles. There are two further worries, apart from those that derive directly from (1), that one may raise against the possibility of akratic choice in Kant’s framework. First, one may insist that Kant’s account of free choice cannot allow that a free will is ever too weak to follow acknowledged moral principles or that a free will may be overpowered by the strength of sensible inclinations. This is correct. But Kant’s appeal to the frailty of the human will does not imply that our faculty of choice is ever truly incapable of complying with moral principles because it is overpowered by sensible impulses. Rather, akratic agents are not trying hard enough to summon the moral strength of rational self-­control that is fully available to them qua free agents and that would allow them to abide by their good principles, to resist the temptation posed by a particular impermissible act in favor of their sensible inclinations. Hence, a free akratic choice is fully imputable to agents as their moral failing. One might ask why free agents would fail to summon their available strength of will, or what explains the difference between free agents who do and do not give in to temptations. But those are futile questions that cannot be answered within Kant’s framework (cf. Part 1). Given the incomprehensibility of transcendental freedom, a free conscious violation of sincerely held good maxims is as inexplicable (for us) as other types of moral failings (such as the free adoption of an evil maxim where one deceives oneself into thinking that the maxim is morally justifiable). A second worry goes as follows: acting against a maxim which expresses one’s convictions about what makes actions good cannot be a case of a free (even deficiently) rational choice because it is unintelligible from the agent’s deliberative perspective why an action that they do not regard as good is at all choiceworthy. This worry might derive from the common idea that in Kant’s view all free

45  My conception of moral frailty denies the claims that agents can never act out of character and that human agents are always, inevitably, in the grip of evil principles. Both claims are espoused by McCarthy 2009. As I argued earlier (Chapter II), attributing these claims to Kant raises insuperable problems (see also Kohl 2017a).

152  Kant on Freedom and Rational Agency intentional agency occurs under the guise of the good.46 (This idea may just be another expression of the Incorporation Thesis or (1), but I am not sure.) Proponents of this view might attribute to Kant a strict dichotomy between (on the one hand) freely choosing an action that one represents as good and (on the other hand) being pulled and pushed by wholly non-­ intentional or non-­ representational impulses, like Warren Quinn’s Radio-­Man.47 I believe that this is a false dichotomy and that Kant does not accept the guise of the good-­view of intentional agency.48 When Kant discusses this view (at KpV, 5:57–9), he remarks that the Latin word “bonum” conceals an ambiguity between two different evaluative notions that the German language differentiates: namely, “good” as a concept of reason and “weal”/“well-­being” (Wohl) as an empirical value-­concept derived from sensations (KpV, 5:62). When we act for the sake of non-­moral aims, the value we strive to attain “is not a good, but a weal” (KpV, 5:62).49 This is a distinction between the value of the empirical state of a person (whether the person feels or anticipates feeling weal or woe) and the moral value or integrity of the person (KpV, 5:60). As Kant remarks when he considers the Stoic plagued by gout, we can evaluate people (and their actions) along these two irreducibly different evaluative dimensions. Only our a priori moral evaluation involves the concepts of good and evil ends. Hence, not all intentional human action occurs under the guise of the good: we also act under the guise of our weal and woe, when we seek to secure for ourselves a state of contentment or at least one devoid of displeasure or frustration. Morally akratic agents choose an action that they conceive as producing a weal (or preventing a woe) against their better rational judgment that the action is not supported by valid reasons since it is morally impermissible and since moral values are absolutely superior to, i.e., cannot be rationally weighed against, non-­moral values (GMS, 4:434–5). Although the desire they akratically choose to act upon fails to ground a good purpose or a valid reason for acting, it is not therefore an unintelligible urge devoid of intentional evaluative content. Its motivationally intelligible grip on free deliberation and choice is due to the rational creature’s representation of the sensible, desire-­based value/disvalue of weal/woe.50

46  See Engstrom 2009; Korsgaard 1996a; McCarthy 2009; Reath 2006; Wood 1999. 47  See Quinn 1994. 48  In what follows I offer an abbreviated version of my argument in Kohl 2017b. 49  Wood (1999: 128–9) argues that the representation of weal/woe is prior to rational deliberation. However, for Kant the good/weal (evil/woe) distinction arises within the sphere of rational deliberation: it is a distinction between two ways of judging the value of an action (KpV, 5:59–60). 50  Kant’s notion of a categorical gap between reason- and desire-­based motivation arguably yields a more compelling account of desiring and acting against reason than anodyne intellectualist views of desire. According to such views, the conflict between desire and reason can only be one of degree because desiring as such involves a rational judgment representing reasons for acting (see Scanlon 2000: 20–40; see Schapiro 2009 for a forceful defense of Kantian anti-­intellectualism about desiring).

Determinism, and the Categorical Imperative  153

V.6.  The Scope of Kant’s Incompatibilism about Freedom of Choice In this section I further clarify the basis and scope of Kant’s incompatibilism about free choice. Contemporary incompatibilists often hold that deliberation per se commits us to the falsity of determinism.51 By contrast, Kant’s motivation for incompatibilism concerns only deliberative choices under practical laws. Why does Kant not appeal to the more general intuition that practical deliberation as such presupposes that determinism is false? As we saw, Kant accepts that determinism threatens neither the causal efficacy nor the epistemic openness of practical deliberation. He thus accepts a point stressed by some compatibilists: different options are always open to us in a ­subjective epistemic sense because we cannot know how we will act before we actually make our deliberative choices.52 For Kant, the stronger presupposition that different courses of action are open to us in an objective metaphysical sense is based on the idea that two irreducibly different poles of motivation jointly ­constitute our nature as rational creatures: this accounts for the contrast between reason-­based and desire-­based motives that pervades our reflective ­self-­awareness. The (pervasive) existence of this contrast is guaranteed solely by the antagonism between the objective moral reasons that are provided by practical laws and the subjective desires that spring from our sensible nature. The non-­ moral reasons of finite agents and the prudential norms that represent those ­reasons themselves depend on the contingent natural desires whose satisfaction brings weal and happiness (KpV, 5:25–7). Hence, the reasons provided by subjective non-­moral norms of desire-­satisfaction are (unlike moral reasons) not inherently opposed to the contingent natural desires that impel us to choose on their behalf. Since our choices under subjective (desire-­based) norms do not necessarily confront an opposed source of motivation, these choices do not inherently involve a real possibility of violating the norms that govern them.53 Accordingly, deliberation under subjective non-­moral practical rules does not presuppose the existence of metaphysically real alternative possibilities. For Kant, a world in which no categorical moral imperatives exist (so that our moral self-­consciousness is an illusion) does not call for an incompatibilist concept of free choice. If all our reasons for acting did depend on our empirical desires, then a volition that is determined by such desires and by correct 51  For a seminal defense of this view, see van Inwagen 1983. 52 This is Bok’s (1998: 96–114) compatibilist response to van Inwagen . See Section  V.1 and Chapter II.3. 53  On a subjectivist view of reasons such as the one proposed by Williams 1997, the occurrence of temptation or weakness of will is entirely contingent: the “subjective motivational set” of agents may be such that they have just a few dominant desires which never conflict and which agents are always moved to satisfy.

154  Kant on Freedom and Rational Agency deliberation about the means for satisfying them could be considered free in a compatibilist sense because it would conform to our only practical purposes and norms.54 This is why Kant denies that all practical deliberation as such, including prudential reasoning, requires transcendental freedom of will and is thereby committed to incompatibilism.55 Against this, some commentators argue that in Kant’s view our pursuit of non-­ moral ends which is governed by hypothetical (as opposed to categorical, moral) imperatives also implies our transcendental freedom of will.56 However, this view faces numerous problems.57 It is unclear (given the preceding arguments) why Kant would hold that deliberative agency under subjective, desire-­based norms requires anything but a compatibilist (naturalistically acceptable) notion of freedom as a capacity to act in accordance with our representations of the means for satisfying our empirically given desires. Accordingly, Kant repeatedly stresses that our self-­conception as having transcendental freedom of will strictly depends upon our consciousness that we are governed by the moral law whose validity is independent of our contingent sensible inclinations (KpV, 5:29–30; RGV, 6:49). For Kant, absolute freedom of will and morality are reciprocal concepts that mutually entail one another (GMS, 4:450). He discusses transcendental freedom of will exclusively in the context of considering the presuppositions of moral as opposed to prudential agency. When he imagines a being endowed with a capacity for impeccable prudential calculation of how to satisfy “the greatest sum of incentives” but without the capacity to choose from the purely rational incentive provided by the moral law, he stresses that the will of such a being would be wholly determined by sensible incentives that derive from external objects; hence, the moral “law is the only thing that makes us conscious of the independence of our faculty of choice from the determination of all other incentives (our freedom) . . .” (RGV, 6:26).58 In defense of his view that agency under non-­moral practical rules implies transcendental freedom of will, Allison argues that the essential link between freedom and moral agency concerns only the positive conception of freedom (qua

54  Kant imagines a similar scenario at GMS, 4:395. He grants that beings whose deliberation determines them to act in ways that satisfy their desires have the “psychological freedom” (KpV, 5:96) that is central to the empiricist tradition (Hobbes, Hume) but that he also associates with Leibniz (the “freedom of a turnspit”). 55  Here I agree with Frierson 2003: 51. 56  See Allison 1990 and 2013: 293–5; Wood 1999: 119, 365. 57  In what follows I give an abbreviated version of my argument in Kohl 2017b, 2018b. 58  Allison (2020: 371) reads this passage very differently: in his view, Kant is not concerned here with the gap between prudential rationality and awareness of moral freedom but with the gap between representing what a perfectly rational being would do and our finite sense of moral obligation. However, Kant’s appeal to “the greatest sum of incentives” is a clear reference to prudential reasoning. Moreover, Kant refers to “the most rational being of this world” (Weltwesen). This clearly suggests a prudentially rational agent rather than a perfectly rational being which is not a Weltwesen afflicted with incentives and a need to calculate how to maximize their satisfaction.

Determinism, and the Categorical Imperative  155 autonomous self-­governance under moral laws). This, Allison claims, leaves room for the idea that non-­moral agency is transcendentally free in the negative sense that denotes independence from determination by foreign causes.59 This idea is correct: agents can only be held responsible for morally impermissible choices aimed at satisfying their desires if these choices are not determined by their empirical desires or other foreign causes (KpV, 5:95–9; RGV, 6:20–2, 24–5, 31, 34–5, 228). But this argument for holding that prudential agency under hypothetical imperatives must be regarded as transcendentally free in the negative sense is parasitic on our moral vocation as agents who act under categorical imperatives that yield conditions of moral permissibility for prudential agency. Hence, this argument fails to support the claim that the ability to be guided by prudential rules on its own, apart from the appeal to moral laws, implies transcendental freedom of will. To support this claim, one would have to show that negative freedom of will is metaphysically and conceptually independent from positive freedom of will as a capacity to choose under self-­legislated moral laws.60 But this strategy flies in the face of Kant’s view that a purely negative freedom of will qua lawless causality is an absurdity (cf. Chapter III). The idea of negative freedom necessarily leads to the idea of positive, law-­governed freedom as moral autonomy (GMS, 4:46; KpV, 5:29). Since Kant so clearly ties the awareness of free will to the awareness of practical laws, proponents of the view that non-­moral agency by itself implies transcendental freedom of will must deny that moral laws are the only practical laws: law-­ governed free agency does not coincide with agency under moral laws. To defend the claim that our agency is also governed by non-­moral practical laws, one might appeal to what some commentators call the Hypothetical Imperative: an a priori, objective, but non-­moral norm of instrumental reason.61 However, I do not believe that Kant posits an objective a priori law of instrumental reason. While

59  See Allison 2013: 293–5. Compare Allison 1990: 88–9. 60  Allison (2020: 261–2; 274–5) tries to show this by appeal to A548/B576 where Kant seems to say that all practical oughts presuppose transcendental freedom, even when the will is focused on objects of sensible inclination (“the agreeable”) rather than on the morally good. But there are two reasons why this passage does not conclusively support Allison’s reading. First, it is from the 1781 A-­edition of the first Critique where Kant has not yet developed his critical account of pure moral motivation and positive freedom as autonomy (he still holds a heteronomous view on which moral motivation ultimately depends on sensible desires for happiness in an afterlife: cf. the appendix to Chapter I). Second, at A548/B576 Kant might be stressing that the validity of non-­moral oughts ultimately hinges on their authorization by moral oughts within the overall structure of practical justification: when the will is focused on objects of sensible inclination, the presupposition of transcendental freedom is due (solely) to the moral ought which determines whether pursuing this sensible object is morally permissible or prohibited (and thereby does or does not authorize pursuing “the agreeable”). 61  A number of commentators (e.g., Hill 1973, 1992; Korsgaard 1997; Willaschek 1992) claim that Kant posits such an a priori practical law of instrumental rationality. But only Wood (1999: 65, 119, 365) draws the (valid) inference that if our agency is governed by such a non-­moral law, this suffices to make us transcendentally free.

156  Kant on Freedom and Rational Agency Kant acknowledges various types of empirically grounded hypothetical imperatives that guide our instrumental reasoning and our non-­moral agency, he argues emphatically that these rules lack objective rational necessity or universal validity and hence do not qualify as practical laws.62 Moreover, the idea that Kant invokes a non-­moral practical law of free agency does not accord with his repeated in­sist­ ence that our awareness of the moral law is our only basis for ascribing to ourselves a law-­governed transcendentally free will (KpV, 5:4, 5:29–30; RGV, 6:269). If Kant did posit a further non-­moral practical law, his frequent inference from the idea that free agency must be governed by some law to the idea that free agency stands under the moral law of autonomy would be fallacious. In sum, Kant’s conviction that metaphysically real alternative possibilities are essential to free will applies only to beings that satisfy two separate conditions: first, they are inclined to choose on behalf of their contingent, subjective desires; second, they are governed by moral laws that provide objectively necessary, desire-­independent reasons for acting. In a scenario where there are no objective moral norms, the second condition is violated: here agents would lack a faculty of pure practical reason qua autonomous source of necessary ends, in relation to which their contingent sensible desires yield a strictly opposed source of (potential) motivation so that a choice on behalf of such desires counts as rationally deficient. A perfectly rational agent would violate the first condition: God’s guidance by reason is immune to the possibility of resisting reason because there is no metaphysical basis for such resistance. God has no sensible nature that afflicts Him with options to go against reason: His Holy Will does not involve a faculty of Willkür that is constrained to choose between different (real) options. My interpretation suggests how Kant’s view can be situated in contemporary debates about the connection between norm-­governed action and the possibility of norm-­violation. Kant, like Lavin, rejects Korsgaard’s view that free norm-­ governed agency as such requires the possibility of violation.63 For Kant divine

62  See Kohl 2018b and Timmermann 2022. Kant seems to suggest a law of instrumental reason just once, in Groundwork II (GMS, 4:417) but never refers to this law again (in published or unpublished texts). Even in Groundwork II he does not treat it as an actual practical law that really applies to our will. The instrumental proposition he considers here concerns a connection between a given end and the means that are necessary and sufficient for securing this end. But the norms that actually govern our non-­moral agency cannot yield this connection. All hypothetical imperatives that prescribe actions as means to a given end ultimately refer to the end of happiness (V-­NR/Fey, 27:1324); but we cannot cognize necessary or sufficient means towards happiness that would support objectively necessary practical laws (GMS, 4:419). All actual hypothetical imperatives are “material” principles, empirically grounded “counsels” that have only inductive (“comparative”) universality and thus lack a priori necessary practical lawfulness (KpV, 5:19–27). Accordingly (pace Pollok 2007, 2017), moral and hypothetical imperatives have no common “root” in pure practical reason. I return to this issue in Chapter VIII. 63  See Korsgaard 1997: 240; 1996b: 161. She suggests that we can only describe the acts of God. Her point seems to be that when we try to conceive a being that cannot fail to comply with normative rules, we lose our grip on the idea of normative guidance and must represent an automaton. But it is not clear why that should be (see Lavin 2004: 443–6). For Kant, an automaton is programmed to

Determinism, and the Categorical Imperative  157 freedom consists in the absolutely spontaneous capacity to act in accordance with self-­legislated norms of right reason (RGV, 6:50; cf. Chapter III). But Kant also rejects Lavin’s or Wolf ’s view that our freedom as finite agents coincides with the ability to act for the right reasons.64 For Kant, the idea that divine agency is completely exempt from the possibility of error and the idea that free human action is afflicted with a pervasive possibility of error are two sides of the same coin: finite agents always choose under conditions of sensible affection by empirical desires that direct them away from the right reasons, whereas God must be conceived as entirely unaffected by sensible conditions. Kant here denies a presupposition shared by Lavin and Korsgaard: namely, the idea that we can give a unified account of what it takes to act freely under normative laws which applies indiscriminately to finite and perfect agents.65 For Kant, finite and perfect agents exhibit different forms of free agency because their respective rational capacities are constituted differently: a Holy Willy is identical to pure practical reason whereas the human will comprises not just pure practical reason but also a sensibly affected, hence impure faculty of choice (MS, 6:213, 222; cf. Chapter III). Kant’s incompatibilist account of human freedom of choice does not suggest that having the option to go against reason is a good thing. There is nothing desirable about falling short of practical perfection. The concept of a perfectly rational Holy Will represents a moral ideal which can inspire us towards moral self-­ improvement (RGV, 6:61), but it cannot be anything but an ideal for those who find themselves in the human condition of finitely-­imperfectly rational agency. For Kant, an adequate account of the free human will must do justice to this condition.

V.7.  Conclusion: The Second Source of Kant’s Incompatibilism about Free Will In this chapter I have reconstructed Kant’s argument for claiming that the executive freedom of choice we ascribe to ourselves as moral agents requires a causally unconditioned, absolutely contingent spontaneity. The key point in Kant’s argument is this: finite agency under normative laws presupposes that the agent qua rational creature is in a position to favor either their sensible nature or their higher rational vocation. Our ability to act on the basis of objective moral reasons has a shadow, namely our propensity to choose against such reasons for the sake of sensible motives, that perpetually hovers over our acts of practical function as it does by an external designer (KpV, 5:99), whereas God is entirely self-determined (absolutely spontaneous) by His own rational nature. 64  See n. 25. 65  See Lavin 2004: 453; he and Korsgaard seek to answer the question, “what is agency?” in the most general sense.

158  Kant on Freedom and Rational Agency self-­governance. This conception of the human condition grounds the significance of metaphysically real alternative possibilities and absolute contingency in human agency.66 For Kant, such alternatives would never exist if (unrestricted) determinism were true, if “all causality were mere nature.” I have explicated why Kant is an incompatibilist with regards to our freedom of will qua Willkür: without the spontaneity to make absolutely contingent choices for or against moral laws of reason, finite agents would fail to be governed by such laws (qua categorical imperatives). This complements my account in Chapter IV of why Kant is an incompatibilist with regards to our freedom of will qua pure practical reason: without unconditioned spontaneity rational subjects could not render absolutely necessary practical judgments in acts of autonomous legislation, and so there would be no unconditionally necessary moral laws. These two arguments yield two reasons why our self-­conception as moral agents cannot be captured by a compatibilist-­naturalistic conception of free will. Thereby, they jointly ground Kant incompatibilism about free will.

66  One might wonder how Kant would respond to Frankfurt’s influential (2003) argument against the relevance of alternative possibilities for moral responsibility. Suppose A impermissibly kills B out of her own volition. Clearly, A is responsible for doing so. Suppose, further, that if C had foreseen that A will not voluntarily decide to kill B, C would have (e.g., by means of mind control) caused A to choose to kill B. Frankfurt argues: since A’s responsibility for killing B is not undermined by the fact that A lacked the option not to kill B, it follows that moral responsibility does not require alternative possibilities. To see how Kant can respond here, recall that for Kant choices are “inner actions.” Moreover, A’s free choice to kill B is a different type of inner action than A’s choice to kill B that would result from C’s determination of A’s volition. This is because only the former choice is a law-­governed act of free self-­determination that involves A’s autonomous rational capacities; accordingly, only the former choice reflects A’s moral character and A’s failure to acknowledge the force of moral reasons. A is morally blameworthy for the outer act of killing B only if this outer act is based upon an inner act of free moral choice that A was in a position to omit because A could also have freely chosen not to kill B (due to her recognition of sufficient moral reasons against killing B). For related criticisms of Frankfurt, see Widerker 2003.

PART 3

FREED OM OF T HOU G H T AS A SPECIE S OF T R A N SCE NDE NTA L F R E E D OM Transition to Part Three Thus far I have focused on Kant’s account of free will. In Chapters I–­III, I have explored fundamental metaphysical, epistemic, and semantic aspects of Kant’s account. I have shown, among other things: how Kant can coherently combine the belief in phenomenal determinism with the belief in the unconditioned noumenal causality of free will; how we can view this noumenal causality as making a difference to what happens in the sensible-­phenomenal world; how we can adequately conceive the theoretically incomprehensible notion of free will; and, how our freedom of will as spontaneous-­autonomous self-­determination relates to divine freedom. My examination of the latter topic led to a distinction between two kinds of free will: the legislative freedom that our will (qua practical reason) exhibits in its prescriptive law-­giving acts of practical judgment and the executive freedom that our will (qua Willkür) exhibits in law-­governed choices. Via this distinction, it turned out that Kant’s rejection of naturalistic-­compatibilist conceptions of free will rests on two worries about the implications of the naturalistic assumption that “all causality” is “mere nature.” I have explicated these worries in Chapters IV and V. Having reconstructed the basic contours of Kant’s doctrine and the grounds of his incompatibilism about free will, I still need to address the crucial epistemic questions that I flagged in Chapter I. How does Kant try to justify the belief that we have the legislative and executive freedom of will that we presuppose in our practical reasoning and moral agency? And, what type of doxastic strength does Kant envisage for this justification? Do our reasons (whatever they may be) for believing that we have transcendental freedom of will entitle us to objectively certain knowledge (Wissen) or merely to a subjectively rational faith (Glauben)? But before I address these issues in Part 4, I consider a further topic in Part 3: I examine the freedom or spontaneity of thought that we exhibit in our theoretical

160  Kant on Freedom and Rational Agency reasoning and cognition. This is a philosophically important topic in its own right that must figure in a systematic-­comprehensive account of Kant’s doctrine. But it will turn out that this topic also bears crucially on Kant’s epistemology of free will: in Chapter IX, I show that Kant’s appeal to freedom of thought plays a central role in his strategy for justifying the belief in free will. My analysis of Kant’s views on freedom of thought proceeds as follows. In Chapter VI, I argue that in Kant’s considered view such freedom is a species of transcendental freedom that involves an absolute epistemic spontaneity. In Chapter  VII, I examine the grounds of Kant’s incompatibilism about epistemic spontaneity by reconstructing his argument for the claim that we must presuppose our transcendental freedom of thought in our theoretical efforts to understand and explain the natural world. This presupposition qualifies as certain knowledge that our theoretical intellect is indeed transcendentally free.

VI Kant’s Free Thinker VI.1.  Free Thinkers versus Thinking Mechanisms Many commentators believe that for Kant free agency requires the practical faculty of volition. On this view the domain of free agency does not include the mental acts of making theoretical judgments and holding propositions to be true (Fürwahrhalten, which I translate as “belief ” or “assent” and whose species are opinion, knowledge, faith; cf. A820/B832). While it is uncontroversial that these acts result from the spontaneity of thought, for many commentators such ep­i­ste­mic spontaneity is much thinner than the moral spontaneity of will.1 Henry Allison argues (commenting on A546–7/B574/B575) that in Kant’s view “we might be rational beings, fully capable of conceptual thought and yet lacking in genuine agency or will (practical reason) . . . . We would, if this possibility were realized, be rational beings but not rational agents.”2 For Allison, “[t]he necessity of acting under the idea of freedom is affirmed [by Kant] not of rational beings überhaupt but only of beings who possess both reason and will,” since will “amounts to the same thing [as] rational agency.”3 According to Andrew Chignell, “sufficient objective grounds typically not only license but also necessitate firm assent . . . once we acquire sufficient objective grounds for p, we typically just find our assent to p following along” without any agency on our part.4 If sufficient objective grounds are perceptual states, this echoes Robert Hanna’s claim that perceptions “necessitate . . . the perceiver’s assertoric belief in a corresponding propositional content.”5 The idea that our beliefs are causally necessitated by empirical states is also in line with Patricia Kitcher’s claim that the “I that thinks” is “phenomenal and causally determined.”6 For Wilfried Sellars, “Kant is leaving open the possibility that the being which thinks might be something ‘which is not capable of imputation.’ It might, in other words, be an automaton spirituale or cogitans, a thinking mechanism.”7 1  See Friedman 1996: 438; Kitcher 1984: 122. In sections VI.1–VI.2, I will draw (but significantly expand on) Kohl 2015c. 2  Allison 1990: 63. 3  Allison 1990: 218. See, however, n. 7. 4  Chignell 2007a: 327. 5  Hanna 2005: 264. Unlike Chignell, Hanna claims that false judgments are perceptually necessitated as well. 6  Kitcher 1990: 140. This encourages the idea that Kant’s views on the thinking subject are intended as a kind of functionalist proto-­neuroscience (cf. Brook 1994: 5; Hoppe 1983: 19–20; Kitcher 1991; Meerbote 1991). 7  Sellars 1970: 25. I should note that it is unclear to me how Allison’s overall view relates to this mechanistic picture. In some of his works (1996, 2013), Allison criticizes Kitcher’s naturalistic reading

Kant on Freedom and Rational Agency. Markus Kohl, Oxford University Press. © Markus Kohl 2023. DOI: 10.1093/oso/9780198873143.003.0007

162  Kant on Freedom and Rational Agency But this view is hard to square with important textual data. In the passage Allison cites as evidence for his reading (A546/B574), Kant says: In lifeless or merely animal nature, we find no reason for thinking that any faculty is conditioned otherwise than in a merely sensible manner. Only the human being . . . cognizes itself also through pure apperception: and this, indeed, in actions . . . which it cannot ascribe to impressions of the senses. It is thus to itself . . . in respect of certain faculties the action of which cannot be ascribed to the receptivity of sensibility . . . a purely intelligible object. We entitle these faculties understanding and reason . . . .

This suggests that we have an apperceptive awareness of ourselves as noumena (“a  purely intelligible object”) that possess sensibly-­empirically unconditioned theoretical faculties such as the understanding. Our awareness of these faculties consists in our consciousness of actions that we perform when we engage in theoretical thought. Kant here does not invoke a thin, deflationary notion of ep­i­ste­mic spontaneity that excludes genuine agency: as we saw, the idea that we can act in a sensibly unconditioned manner is also central to Kant’s conception of a transcendentally free will. This passage raises important complications that I will consider in Sections VI.2–3. For now, I just note that it is not an outlier: in other parts of the Critique Kant also stresses that pure apperception involves a consciousness of an unconditioned spontaneity, an “intellectual representation of the self-­activity of a thinking subject” (B278; cf. B157–9). Kant does not use the concept of freedom in these passages. But elsewhere he does ascribe “real” freedom and agency to our theoretical intellect: “[t]o abstract from a representation of which I am conscious . . . is a real act of the faculty of cognition” that “proves a freedom of the faculty of thought and the power of the mind to control the state of its representations” (Anth, 7:131). If “an appearance is given to us, we are still completely free as to how we want to judge things from it” since the “appearance was based on the senses, but the judgment on the understanding” (Prol, 4:290). Kant characterizes this “complete” freedom of theoretical judgment in the same metaphysical terms as freedom of will: our cognitive spontaneity enables the “self-­activity” of a noumenal “intelligence” (B157–9; cf. B278); “the understanding . . . is free and pure self-­activity which is determined through nothing other than itself ” (Refl., 18:182–3). Hence, theoretical cognizers must

and suggests that for Kant the spontaneity of thought is absolute or unconditioned. This seems to contradict his 1990 view, cited above, that without practical spontaneity we would fail to be free agents. In his recent work, Allison appears to combine these two strands, insisting that epistemic spontaneity is unconditioned (2020: 223, 272, 330) and signaling his agreement with Sellars’s thinking mechanism-­interpretation (2020: 321–3). This strikes me as an inconsistent view. If rational beings possess an absolutely spontaneous, unconditioned capacity for thought, they cannot be mere mech­an­ isms but must qualify as free cognitive agents (see my review of Allison 2020 in Kohl 2020c).

Kant ’ s Free Thinker  163 assume their “freedom of thought . . . [i]n the same way” as moral agents must presuppose their freedom of will (RezS, 8:14). Accordingly, Kant ascribes autonomy not only to practical reason but also to the theoretical understanding (EEKU, 20:225, 241). In this chapter I argue that we should take these claims at face value: for Kant we are transcendentally free and exhibit absolute epistemic spontaneity (“self-­ activity”) in our acts of theoretical judgment. This reading encounters a variety of problems. Some commentators would concede that it captures a view Kant held prior to the late 1780s but not his fully considered view at it emerges by 1787. I will address this challenge in Sections VI.2 and VI.3. But the view I am ascribing to Kant also invites an important philosophical worry: the claim that we are “completely free” in our empirical judgments seems to grant us an implausibly high amount of control over what we can believe about the natural world. If we can assent to empirical propositions without being determined by our perceptual awareness, does this not imply the untenable voluntaristic idea that we can simply choose what we want to believe, regardless of or even contrary to our evidence? My first step in this chapter is to show that Kant offers us a plausible, interesting model of free epistemic agency that steers clear of both doxastic voluntarism and determinism.

VI.2.  Freedom of Thought as Reflective Self-­Control I begin by considering the negative (empirically undetermined) epistemic freedom that we exhibit in all imputable acts of thought including our mistaken theoretical assents. Kant designates the understanding as the capacity to judge (A69/B94). He repeatedly says that the exercise of this capacity consists in the subject’s combination of the objective and subjective grounds of judgment. This idea is crucial to Kant’s conception of empirical error. Consider the following central passage: For truth or illusion is not in the object, in so far as it is intuited, but in the judgment about it, in so far as it is thought. It is therefore correct to say that the senses do not err—­not because they always judge rightly but because they do not judge at all. Truth and error, therefore . . . are only to be found in the judgment . . . in a representation of the senses—­ ­ as containing no judgment whatsoever—­there is . . . no error . . . the [understanding] would not [of itself fall into error], since, if it acts only according to its own laws, the effect (the judgment) must necessarily be in conformity with those laws . . . . Now since we have no sources of cognition besides [understanding and sensibility], it follows that error is brought about solely by the unobserved influence of sensibility on the understanding, through which it happens that the subjective grounds of the

164  Kant on Freedom and Rational Agency judgment enter into union with the objective grounds and make these latter deviate from their true function . . . .  (A294–5/B350–1)8

Kant here equates an empirical judgment’s “objective grounds” with the laws and concepts of the understanding and its “subjective grounds” with impressions of the senses. To understand this equation, we need to consider some core tenets of Kant’s epistemology. The impressions we receive through the senses do not by themselves represent objects (B130–1, 137–8, 219, 233–4; Prol, 4:290). A sense-­ impression is “subjective” because it is just a private episode in a person’s mind that cannot, as such, be shared with other perceivers (B139–40, 142).9 In order for given sense-­impressions to yield objective empirical cognition, they must be brought under concepts of two types: first, under pure concepts (categories) that enable the representation of something distinct from private mental states, a public empirical object in general (A105; A111–12; B165; A93/B126); second, under empirical concepts that cognitively determine the specific character of this object. So, the contrast here is between subjective sensible data and the conceptual rules for combining such data in a manner that enables objective empirical thought.10 Kant says that in cases of error the subjective grounds of an empirical judgment “enter into union” with its objective grounds. This is puzzling: sensibility and concepts must “enter into union” for there to be empirical cognition because sensible intuitions and concepts are individually necessary and (only) jointly sufficient for cognition of empirical objects (A51–2/B75–6; A92–3/B125). Kant explains (A295/B351) that when sensibility is “subordinated” to (the rules of) the understanding, “it is the source of real modes of cognition.” It becomes a source of error when the proper sensibility-­understanding union is reversed: namely, when there is no adequate subordination of sensible representations under correct conceptual rules but sensibility instead “influences the operation of the understanding, and determines it to make judgments.” This phrase is ambiguous: does Kant mean only that the senses “influence” an erroneous judgment by providing the understanding with an occasion for error, i.e., with merely subjective 8  Kant also discusses objective and subjective grounds at A820–31/B848–58. But there he applies these notions indiscriminately to all kinds of judgments, including practical judgments about supersensible objects. 9  For Kant, perceptions are not fully conceptualized because they do not fall under the dynamical categories (such as causality) and hence do not represent an intersubjective object of experience (B219; Prol, 4:297–8). I am here only concerned with the spontaneity we exercise in conceptually determinate thought when we form judgments of experience. I cannot discuss Kant’s difficult views on “judgments of perception” or the kinds of syntheses that epistemically precede judgments of experience (see n. 17). For a seminal discussion of those syntheses and the teleological idea that they are geared towards valid judgments of experience (as the fundamental goal of the understanding), see Longuenesse 1998. 10  For some (e.g., Abela 2002; Collins 1999), Kant systematically rejects the notion of given subjective sense-­data. However, such readings owe us an account of what Kant means when he invokes merely subjective grounds of judgment and when he appeals to a subjective form of awareness that cannot, due to insufficient conceptualization, be shared with others (Prol, 4:298–9; B139–40).

Kant ’ s Free Thinker  165 data that the understanding wrongly takes to be objective (see V-­LO/Wiener, 24:825)? Or does he intend the stronger point that the senses causally determine the understanding to err? The stronger interpretation is ruled out by Kant’s further claim in above passage (see also JL, 9:53) that “the senses do not err,” since this claim would be false if the senses were causally sufficient for the occurrence of an erroneous judgment. The weaker interpretation is confirmed by Kant’s remark that error “is not ascribed to the senses, but to the understanding, whose lot alone it is to render an objective judgment from the appearance” (Prol, 4:291). This implies that empirical errors are imputable to the understanding: “is not ascribed to the senses, but to the understanding” reads “kommt nicht auf Rechnung der Sinne, sondern [auf Rechnung] des Verstandes” and the German word for imputability is “Zurechnung.” Since the understanding is not a self-­ subsisting faculty but the theoretical intellect of a thinking subject, the idea that a false judgment is imputable to the understanding means that it is imputable to us as thinking subjects who freely exercise their capacity to judge: “we are . . . completely free as to how we want to judge things” (Prol, 4:290). This puts significant pressure on Sellars’s claim that the understanding qua thinking mechanism is incapable of imputation. Sellars’s claim derives from his synthesis of two passages: in the Dialectic Kant refers to the “I or he or it (the thing) which thinks” (A346/B404), and in the Metaphysics of Morals he defines “Sache” as a “thing that is not capable of imputation” (MS, 6:223). But it is far from clear whether Kant’s (polemical) use of the term “Ding” in the Dialectic corresponds to the narrow definition of “Sache” that he gives in the Metaphysics of Morals. Sellars’s reading also rests on concerns about the non-­critical rationalistic implications that he suspects Kant’s view would have if Kant did regard the thinking self as a free, accountable epistemic agent. I will consider these worries in Section VI.2 (and in Chapter VII). For now I just note that Kant does not seem to share these concerns in passages such as Prol, 4:290. Let us now consider how Kant conceives of cases where the influence of perception leads us to form erroneous judgments that are imputable to us as free thinkers. Empirical error arises when a judgment has its ground only “in the special character of the subject” rather than “the character of the object” (A820/ B848). This terminology refers us to Kant’s discussion of merely subjective sensations such as taste: “The taste of a wine does not belong to the objective determinations of the wine . . . but to the special character of sense in the subject that tastes it” (A28). I suggest that Kant’s picture here is roughly as follows. Suppose I recognize an object as wine: I conceptualize it in terms of the valid empirical concept of an alcoholic liquid that is made from grapes. Drinking the wine, I have an impression of sourness. So far there is no error because I have not yet assented to this impression. The error occurs when I judge that sourness is a quality that inheres in the wine, i.e., when I carelessly mingle a private sensation of taste with a valid, public empirical concept. Kant has another example illustrating this point:

166  Kant on Freedom and Rational Agency [I]t is not the fault of the appearances at all, if our cognition takes illusion for truth, that is, if intuition, through which an object is given to us, is taken for the concept of the object . . . which only the understanding can think. The course of the planets is represented to us by the senses as now progressive, now retrogressive, and herein is neither falsehood nor truth, because as long as one grants that this is as yet only appearance, one still does not judge at all the objective quality of their motion . . . however, if the understanding has not taken good care to prevent this subjective mode of representation from being taken for objective, a false judgment can easily arise . . . .  (Prol, 4:290–1)

The perceptual appearance of backward or forward motion depends on contingent facts about us (say, on our position in the solar system); it does not reflect objective qualities of empirical objects. But, again, perceptions do not involve judgments about the way things are. They lead to error only insofar as a thinking subject carelessly takes the content of these perceptual data to present an objective quality that she ascribes to public empirical objects (e.g., the planets).11 In that way, the false judgment is influenced (affected) but not determined by the perceptual illusion. Empirical error may also involve the faculty of imagination, when “the faculty of judgment is misled by the influence of imagination” (A295/B352). The empirical, reproductive imagination generates contingent habitual patterns of associating representations (A100; B140). The fact that we may be misled by such associations does not imply that the imagination is responsible for our cognitive mistakes. For example, as a result of socio-­cultural conditioning I might find myself connecting the thought of lemmings with the thought of suicidal tendencies (since I played the computer game Lemmings as a kid, etc.). I have no control over this associative pattern. But an error occurs only once I take this pattern to indicate an objective quality and judge that lemmings as a matter of fact exhibit suicidal tendencies: this judgment is not a mere causal upshot of psychological customs beyond my control. While Kant agrees with Hume that our minds are led to associate representations by subjective empirical habits (B139–40), he does not accept that those habits determine us (considered as spontaneous thinkers) to render judgments which ascribe properties to empirical objects (B142). Hence, for Kant empirical judgments and assents are “effects” (A294–5/ B350–1) that result from imputable actions of thinking subjects. These actions and their effects are imputable to us as thinking subjects because we have a 11  What Kant says at Prol, 4:290–1 commits him only to the view that empirical error typically results from carelessness. This leaves open the possibility that there may be cases where thinkers make an erroneous judgment despite their careful examination of the case; here the judgment would still be imputable to them as free thinkers, but they would not be epistemically blameworthy. However, Kant often suggests that all error can be avoided if we exercise sufficient care and attend to the limits of our knowledge (JL, 9:54; WDO, 8:136).

Kant ’ s Free Thinker  167 freedom of thought that consists, negatively speaking, in the fact that our cognitive activity is not determined by sensible-­empirical causes such as given perceptions or psychological habits of association. One might object that the formation of judgments and corresponding assents involves alterations of a thinker’s mental states which, like all temporal changes, fall under the principle of the Second Analogy and hence must be attributed to a determining empirical cause.12 But this does not raise any problems other than those which also arise for Kant’s doctrine of free will and which I addressed in Part 1. There, I explained how Kant’s idealism makes room for the idea that our actions can be attributed both to a determining natural cause and to a free cause that stands outside the deterministic order. This account applies not only to acts of will but also to acts of thought including the formation of empirical judgments or assents. We attribute acts of thought to a determining natural cause when we adopt an observational-­explanatory-­predictive standpoint towards human beings considered as empirical phenomena; we ascribe acts of thought to a free intellectual capacity to judge when we adopt a normative standpoint towards human beings considered as spontaneous noumenal intelligences. There is clear textual evidence that Kant’s crucial distinction between an empirical and a normative standpoint also applies to acts of thought and judgment that do not involve the will. Aesthetic judgments of reflection “do not state that everyone judges in this manner—­if they did, they would be a task for the explanations of empirical psychology—­but rather that one ought to judge in this manner” (EEKU, 20:239; cf. KU, 5:182). In pure “logic, we do not want to know: how the understanding is and thinks and how it has thus far proceeded in thinking, but how it ought to proceed in thinking” (JL, 9:14). There is thus a close analogy between the normative standpoints of pure logic and pure ethics (cf. A53–5/B77–9).13 If we adopt this normative perspective towards our acts of thought, we view these acts (and their effects, judgments and assents) as free from determination by foreign empirical causes. Having shown that in Kant’s view our cognitive activity proceeds under the negative idea of freedom from sensible-­empirical determination, I now consider whether freedom of thought can also be understood in an informative positive fashion. Does Kant have a plausible model of how we exercise our intellectual freedom in forming beliefs about the world? Here Kant faces a challenge arising from the view that belief unlike choice is an inherently passive state whose

12  The claim that our thoughts (as inner phenomena) are empirically determined is controversial. Some commentators argue that since our mental activity is the source of the temporal-­deterministic order of nature, such activity cannot (on pain of circularity) be understood in temporal-­deterministic terms; see Wolff 1963: 115–16. For an attempt to resist this argument, see Kitcher, 1999. I discuss this issue in Chapter VII. For others, the Second Analogy Principle only applies to physical alterations (see Melnick 1973; Westphal 2005; see Chapter II for my critique of this view). 13  For a detailed account of Kant’s extended standpoint distinction, see Kohl 2018a.

168  Kant on Freedom and Rational Agency formation cannot be genuinely free. David Wiggins eloquently states this view as follows: It is up to me to make up my mind. But how far is this from the platitude that if it is my mind then only I can make it up, or the truistic interpretation of “you must believe what you think true”? It does not seem to follow . . . that it is up to me what to think true. And it cannot be up to me if belief is to retain its connection with the world. And this connection must, from the nature of belief and its onus to match the world, be subjection. If belief does not retain this connection and subjection it could not be belief. For . . . reality is most of it precisely what is in­de­pend­ ent of me and my will . . . . From which it would seem to follow that if my beliefs are to relate to the world at all, I simply have to lay myself open to the world in order to let the phenomena put their print upon me . . . . If my state is one which seals itself off from the outside, this surely enfeebles its claim to be a state of belief. For there is subtracted everything which distinguishes belief from fantasy . . . .14

This suggests the following argument: (1) Belief is a state that aims at objective truth, at matching a reality that is independent of my will. (2) Thus, insofar as my representations of the world are under the immediate control of my will that expresses how I want reality to be, they are “fantasies” rather than beliefs. (3) So, my beliefs must be independent of my will. (4) Hence, my beliefs cannot be genuinely up to me: they are passive states that occur when I “let the phenomena put their print on me.” The inference from (3) to (4) rests on the idea that any (genuinely) free agency as such must directly involve the will as a faculty of choice. As Williams puts it, since “it is not the case that belief is connected with any decision to believe” we must endorse “the picture offered by Hume of belief as a passive phenomenon, something that happens to us.”15 Likewise, Chignell argues that since Kant rejects doxastic voluntarism he must “side with Hume” and hold that our assents are passive states that we “find following along” the acquisition of evidence.16 Kant indeed rejects doxastic voluntarism. His argument here seems remarkably similar to Wiggins’s line of thought: The will does not have any influence immediately on assent; this would be quite absurd . . . . If the will had an immediate influence on our conviction concerning what we wish, we would constantly form for ourselves chimeras of a happy condition, and always hold them to be true, too. But the will cannot

14  Wiggins 1969: 143. 15  Williams 1973: 147–8. 16  Chignell 2007a: 327; cf. Chignell 2007b: 36; Kitcher 2011: 169.

Kant ’ s Free Thinker  169 struggle against convincing proofs of truths that are contrary to its wishes and inclinations.  (JL, 9:73–4)

This argument can be reconstructed in terms of (1)–(3) just like Wiggins’s argument, “wish” corresponding to “fantasy” and “assent” to belief. But Kant does not move from (3) to (4): from the fact that beliefs are not under the immediate control of the will he does not infer that beliefs are passive states. As we saw, he accepts that our empirical judgments and assents result from our (“completely”) free, imputable cognitive activity. Hence, he must accept a non-­volitional sense in which we are free and active when we assent to empirical propositions. I suggest that (part of) Kant’s point here is this. How we judge the natural world depends upon our conception of how we ought to judge: our beliefs about the world are governed by, or subjected to, our intellectual representation of what our evidence is. It is through my endorsement of some perceptual datum as a good reason for believing that p that I form a belief that p; my doubt about whether a perceptual datum supports the belief that p effects my suspension of judgment; and my verdict that a given perception which represents p as being the case is illusory leads me to form the belief that not-­p. Thinkers are active in forming assents insofar as they govern their doxastic states through reflection on their epistemic reasons, thereby exercising “the power of the mind to control the state of its representations” (Anth, 7:131). This form of agency may be called reflective control.17 The question is whether we really have such control over our theoretical assents and whether exercising such control can be regarded as a genuine kind of free agency. David Owens rejects the idea of reflective control because he thinks that there is simply no need for subjects to exercise such control over their empirical beliefs: “[I]f anything is in control of the belief-­ forming process . . . it looks to be the . . . experiences that motivate belief rather than the subject who enjoys these . . . experiences”; thus, reflective judgments about what we ought to believe seem like “an idle wheel in our motivational economy.”18 To see how Kant can respond here, consider first cases where we have non-­ veridical perceptions of p which are subjectively indistinguishable from veridical perceptions, without being compelled to believe that p. In such cases, one can acknowledge that the senses present p as being the case and yet withhold assent to 17 I borrow the term “reflective control” from Hieronymi 2009, though the account of cognitive agency I am ascribing to Kant corresponds to what Boyle 2009 calls intrinsic control, rather than to the extrinsic control model Boyle finds in Hieronymi. Recall (see n. 9) that I am concerned with determinate judgments of experience. For Kant, the reflective nature of these judgments goes beyond the original synthesis of intuitions (B133), which prepares intuitions for full conceptual determination via empirical concepts (B143): it “precedes a priori all my determinate thought” (B134). I am sympathetic to the idea that (pace Kitcher 1990) this synthesis can also be regarded as an (implicitly) conscious process which involves reflective control (see McLear 2020); but this raises further (e.g., textual) complications. 18  Owens 2000: 3, 18.

170  Kant on Freedom and Rational Agency the proposition that p. But then it must be the subject rather than her “experiences” who is in control of the belief-­forming process. Moreover, we can (and sometimes should) withhold assent also in cases of veridical perception, when we doubt whether a sensory appearance is indeed veridical.19 Suppose, for instance, I hear a  sound that I immediately associate with the thought of warning cries uttered by  jaybirds. The sound in fact stems from a jaybird. Still, the conjunction of (i)  my  perception of the sound and (ii) my association of this sound with the concept of jaybirds does not automatically induce or warrant the belief that there is a jaybird nearby. Suppose I have epistemic grounds for mistrusting my ability to interpret bird sounds; I rightfully do not take my association of a sound with jaybirds to provide me with a good reason to believe that the sound really comes from a jaybird. On this basis, I suspend judgment about what caused my perception. This suggests that if I adopt the belief that there is a jaybird nearby, this depends on the fact that I, however implicitly, endorse my perceptual states as evidentially viable: I take myself to have evidential grounds that entitle me to judging that there is a jaybird nearby. Or, consider beliefs based on testimony: here, too, our belief-­formation presupposes that we actively endorse the verdicts of others as trustworthy, as providing us with sufficiently strong objective grounds for taking some proposition to be true. What distinguishes a judgment that lays claim to objective, intersubjective validity (and thus represents what is the case in a public world shared with other thinkers; B141–2) from a merely subjective, private association of representations is precisely that thinkers who put forward a judgment of experience (and thereby hold an empirical proposition to be true) takes themselves to possess valid reasons that entitle them to demand assent from other thinkers: If I want it to be called a judgment of experience, I then require that this connection [of perceptions] be subject to a condition that makes it universally valid . . . [and] that I, at every time, and also everyone else, would necessarily have to conjoin the same perceptions under the same circumstances. (Prol, 4:299)

For Kant, our “freedom to think” requires that we judge, if not in relation to actual other thinkers, at least “as it were in community with others to who we communicate our thoughts, and who communicate theirs with us” (WDO, 8:144). Thus, we exercise freedom of thought qua reflective control over our doxastic states only if we, however implicitly, regard ourselves as members of a community of rational thinking subjects. This in turn requires that we, however implicitly, form our 19  See McDowell (1994: 26): a perceptual datum “becomes the content of a judgment” only “if the subject decides to take” the perception “at face value.” Note, however, that his notion of “deciding” is problematic since it may be taken to have voluntaristic implications.

Kant ’ s Free Thinker  171 judgments on the basis of what we (perhaps to some extent mistakenly) regard as universally valid epistemic reasons which can be communicated to, assessed and criticized by others. Note that one can legitimately form a judgment of experience without deeming one’s epistemic grounds wholly conclusive or objectively sufficient, as long as one rightly takes oneself to possess some degree of universally valid evidence. In such cases one’s assent should take the form of mere “opinion” rather than “conviction” (A822–3/B850–1; JL, 9:66–7). The degree of the subjective strength of one’s opinion should be proportionate to the degree of objective justification one takes oneself to possess for assenting to the relevant empirical proposition. These considerations support the idea that as rational thinkers we exercise a free reflective over our theoretical assents. But this idea invites a further worry: it might seem to imply that we cannot form a belief without going through actual episodes of reflection through which we form a conception of what beliefs are required or prohibited by our evidence; and this clearly misdescribes our everyday patterns of belief-­formation. Fortunately, Kant is quite explicit about the relation between reflection and judgment-­formation: Every prejudice is to be considered as a principle of erroneous judgment, and . . . erroneous judgments arise from prejudices . . . . The reason of this illusion is to be looked for in subjective grounds’ being falsely taken for objective ones, from a want of reflection that must precede all judging . . . we cannot and must not judge anything without reflecting . . . . If we assume judgments without this reflection . . . prejudices, or principles for judging from subjective reasons that are falsely taken for objective ones, arise.  (JL, 9:76)

This shows that for Kant it is a norm of theoretical cognition that we ought not (“must not,” rationally speaking “cannot”) judge without reflecting on whether our perceptions provide us with objective reasons. He does not claim that our judgments always are preceded by such reflection, for he is well aware that we do not typically comply with his normative demand and thus oftentimes form our judgments on the basis of cognitive prejudices. Here we must note that for Kant our empirical judgments are under our reflective control even when they are not preceded by actual reflection. Consider his analogous view in the practical domain: our specific choices typically depend on our ideas about we have reason to do even when they are made unreflectively. This is because we typically choose on the basis of general principles of ­volition  (maxims) which embody standing commitments to choose in certain ways and thereby foreclose our answers to many specific practical questions (cf. Chapters II, V). For instance, my choice to put on shoes before I leave the house is surely under my reflective control but is rarely preceded by active reflection. Kant also invokes this notion of maxims in the theoretical domain: we assent on the

172  Kant on Freedom and Rational Agency basis of freely adopted general “principles for judging” that attribute evidential significance to recurring perceptual patterns and thereby control the unreflective formation of specific beliefs (see e.g., JL, 9:74–6).20 Hence, for Kant our particular cognitive responses to perceptual (or testimonial, etc.) input are typically based on general commitments to treat certain data as grounds for making certain judgments; these explicitly reflective commitments foreclose our (only implicitly reflective) answers to a variety of theoretical questions.21 The idea that a person’s practical choices and theoretical assents are typically under her reflective control can be supported by the fact that typically a person is at no loss (when pressed by others or when entering into a reflective mood) to make explicit the reasons that implicitly guide her choices or theoretical assents. Accordingly, when people find that they hold beliefs which they cannot support with reasons, they are alienated from (and not free with regard to) their doxastic states.22 General epistemic maxims that control the formation of particular doxastic states are not per se problematic. For Kant, the adoption of maxims that lead to habitual suspension of judgment is an epistemic virtue: such maxims are informed by a norm of epistemic caution which leads to a sustained, reflective (explicitly conscious) appreciation that we lack adequate evidential grounds (JL, 9:74–5). There might also be more concrete policies of judgment that are perfectly valid, such as the maxim to form an opinion that there is a dog nearby whenever one hears a barking noise. However, it seems that for Kant our subjective general principles of theoretical judgment are typically mistaken because they perpetuate careless, epistemically irresponsible habits of thought. The general principles for judging unreflectively that Kant tends to focus on (e.g., at JL, 9:76) are prejudices which rely on inadequate (perceptual, testimonial, etc.) grounds of assent; if we adopt such principles we incur epistemic vices. This raises the question of why Kant deems it an epistemic virtue to judge on the basis of explicit reflection or at least on the basis of general maxims of caution that prescribe policies of suspending judgment or weak opining until further reflection.23 I address this question later in this section.

20  Note, however, that there are crucial disanalogies here as well: for instance, unlike practical maxims of choice epistemic maxims of judging need not be traced to a highest life-­governing maxim. 21 Thus, for Kant reflective control over belief-­formation is exercised typically but often only implicitly (in the background of thought). This is different from the weaker idea (cf. McDowell 1998a: 434) that we always can exercise reflective control over our beliefs even though we typically do not exercise such control at all. Boyle 2011 develops an intriguing view of doxastic self-­determination as energeia which, I believe, captures the general spirit of Kant’s account in some crucial respects. 22  For this notion of alienation, see Moran 2001. 23  To clarify: the notion of theoretical maxims at issue here must be distinguished from the notion of a regulative maxim of theoretical reason that Kant defines in terms of reason’s speculative interests (A666/B694). It is always rational (an epistemic virtue) to adopt the latter maxims as norms for ep­i­ste­ mic reflection. By contrast, theoretical maxims of the former type may be mistaken or even exhibit epistemic vices.

Kant ’ s Free Thinker  173 A further worry about the view that I am ascribing to Kant is that it implies a vicious regress. On Kant’s view empirical judgments are free insofar as they are controlled by thinkers’ reflective conception of what their theoretical reasons are. One might argue that this reflective conception itself requires reflective judgments (concerning what one’s reasons are) which in turn require further reflective judgments (providing reasons for one’s judgments about what one’s reasons are) and so on ad infinitum. To see how Kant can respond here, recall how he motivates the need for reflective control in the first place. For thinkers to judge what is the case in the public empirical world, they must take their judgment to be based on objectively valid, communicable reasons that entitle them to demand assent from other thinkers; this sets a mental state of assent (Fürwahrhalten) apart from a merely private association of representations. Since perceptual states cannot interpret themselves as constituting objective reasons (“the senses do not judge”), thinkers must perform an act of thought through which they interpret their perceptions as providing valid reasons for stating that something objectively is the case. Since the judgment that one’s perceptions provide sound epistemic reasons for assent already constitutes a representation of universal-­objective validity and thereby grounds the thinker’s demand on others’ assent that Kant deems integral to public judgments of experience, there is no need for further judgments here: so the alleged regress does not get started. Moreover, the act of taking (e.g.) one’s perception of a sound pattern to provide one with objective reasons for judging that there is a jaybird nearby need not be construed as a separate act of judgment over and beyond judging that there is a jaybird nearby. To judge that one has valid reasons to believe that there is a jaybird nearby just is to judge that there is a jaybird nearby; more generally, the belief that I ought to believe that p is simply “transparent to” the belief that p.24 This is not to deny that on Kant’s view our empirical judgments are governed by further judgments: namely, by a priori judgments that express general principles of empirical thought. These principles are epistemically fundamental: all our specific verdicts about what our perceptual evidence entitles us to say about the public natural world must rely on the general principles of the understanding25 (and on other principles as well).26 But since the judgments articulating these 24  For related notions of transparency, see Moore 1993 and Moran 2001. 25  These principles include the “analogies of experience,” which are required for interpreting our subjective sensations in such a way that we can represent an objective temporal order. This is a basic condition for making specific empirical judgments. The analogies of experience are thus “rules of universal time-­determination” (A178/B220; A180/B222). For instance, all our empirical thought about events must presuppose that events have determining natural causes because positing such causes is required for interpreting successive sensations (e.g., of cold and warmth) as perceptions of two objectively successive states that follow upon each other in the alteration of a public empirical object (B233–4; A94–6/B239–41). Our empirical thought thus proceeds from the general, fundamental presupposition that events have determining causes to specific causal judgments: “For instance, a room is warm while the outer air is cool. I look around for the cause, and find a heated stove” (A202/B247–8). 26  Namely, the a priori regulative principles of theoretical reason and reflective judgment.

174  Kant on Freedom and Rational Agency principles are a priori, they do not presuppose any further reflective interpretation of what our sense-­perceptions entitle us to judge (cf. A158–60/B197–9). Hence, our epistemic need to rely on these principles does not generate a reflective regress either.27 A final objection to the Kantian conception of epistemic freedom as reflective control runs as follows. Recall the original worry that it cannot be “up to us” what to believe since beliefs aim at the truth. One might argue that since our reflective conception of how we ought to judge is also aimed at the truth, it is not up to us in any deeper sense either; thus, the idea of reflective control cannot ground a non-­trivial notion of free belief-­formation. This objection assumes that we exercise free control over an action only if we can alter (at will) our verdict about whether this action is justified or not. But for Kant free rational agency is compatible with the inescapability of normative verdicts about how one must act. Our freedom of will is not in the least impugned by the fact that we cannot alter at will our self-­legislated conception of what we morally ought to do. As we saw (Chapter III), Kant holds that our legislative freedom of will (qua practical reason) does not involve alternative possibilities; it “irrepressibly forces” upon ourselves (“whatever inclination may say to the contrary”) basic moral laws prescribing how we ought to choose. Moreover, a choice that conforms to these laws is free in the positive sense of autonomous rational self-­ determination. Hence, the fact that it is not up to us to reject valid a priori theoretical principles that prescribe how we ought to judge cannot tell against our epistemic freedom either. For instance, I cannot give up the principle prescribing that one must suspend a judgment that p if one suspects that one’s perceptual awareness of p might be misleading. If my suspension of belief is governed by this epistemic norm, it is a free act of rational self-­control. Likewise, when I seek empirical cognition of nature I am “irrepressibly” bound by the a priori rule that events must be attributed to some determining cause. If my belief that a pen is determined to fall once dropped in mid-­air is based on my “recognition of [this] rule” (A196/B241),28 it stands under my rational control and is thus free. I would lack such freedom if, as Hume claims, this belief is a passive mental event that

27  These “principles” of the understanding are objective laws of cognition, whereas the “principles for judging” discussed earlier are maxims that express our subjective conception of our epistemic reasons for judging. For an analogous ambiguity of “principles” in the practical domain see GMS, 4:420. 28  When Kant refers to the “recognition of the rule” that is presupposed by specific causal judgments, he does not mean any particular causal rule but the general rule “supplied” by the category of causality: namely, “the rule . . . that everything which happens has a cause” (A196/B241). In our experience of an event (and therefore also in our specific empirical judgments about the causes of events) we must “always presuppose that something precedes it, on which it follows according to a rule”: i.e., we presuppose the general rule that events are determined by some causal condition which operates according to some specific causal rule. See Longuenesse 1998: 368–71.

Kant ’ s Free Thinker  175 occurs apart from my recognition of intellectual rules, thus regardless of whether I consider causal beliefs rationally justified.29 Let me summarize my argument thus far. Kant agrees with Wiggins and others that forming an empirical belief requires that we “lay ourselves open to the world to let objects put their print upon us.” But Kant denies that the passive reception of sensory data is sufficient for theoretical assent. On Kant’s view, there is a reflec­ tive gap between taking in perceptual data and judging what these data reveal about the world. Kant further accepts that our cognitive responses to perceptual input are not under the immediate control of the will (and so the resulting states of assent do not collapse into mere “fantasies” or “wishes”). But for Kant this does not imply that we are passive in our theoretical assents: assenting is an act of free control since it depends on our conception of what our theoretical reasons are (rather than, as in Hume, on sub-­personal psychological processes that are completely divorced from our intellectual self-­consciousness). Thus, we can give positive content to Kant’s notion of freedom of thought: such freedom not only consists (negatively speaking) in the absence of determination by natural causes but also involves our “capacity to determine [our] judgments according to objective reasons” (RezS, 8:14), through our reflective appreciation of epistemic reasons and norms. Thanks to this capacity, we qualify as genuine epistemic agents.30 Does our freedom of thought solely consist in our capacity to think in ac­cord­ance with the right reasons?31 Or does it also involve a propensity to think incorrectly? In Chapter  V, I argued that our executive freedom of will essentially involves a propensity to violate moral laws. In the practical domain, there is a gap between our recognition that we (e.g.) morally ought to tell the truth and actually telling the truth, a gap that exists (in part) because of the weakness (“frailty”) inherent in the human Willkür. But it seems that there can be no analogous gap in the theoretical domain because Kant denies that the faculty of choice may directly interfere with a thinker’s representation of 29  For Hume’s view that causal beliefs are passive states that do not involve our intellectual faculties, see An Enquiry Concerning Human Understanding, Section V. On some readings, Hume in the Treatise of Human Nature holds that causal beliefs are based on reason (see Loeb 2002: 53). But even on this reading, Hume reconfigures the notion of reason or understanding as a purely associative faculty that operates apart from thinkers’ active reflective endorsement; moreover, the representation of causal necessity remains excluded from reasonable belief. I discuss the dialectic between Hume and Kant regarding these issues in Chapter VII. 30  As Hampshire says (1971: 91), “it would be a crude metaphysics that implied that an action is necessarily a physical movement.” Kant generally denies that “real” agency requires the use of our bodily capacities: he distinguishes between “inner” and “outer” actions (MS, 6:214, 218; cf. Chapter V). In the practical domain he holds that only inner acts of choosing maxims are always fully within our control (KpV, 5:36–7; cf. Chapters I, III). Hence, we should also take his reference to inner “actions of the understanding” (A330/B387) or “real act[s] of the faculty of cognition” (Anth, 7:131) as designating a type of genuine thinking agency, rather than holding (with Allison 1990: 63, 218; Kitcher 2011: 178) that “action” contrasts with “thought.” For unapologetically agential conceptions of rational thinking inspired by Kant, see Boyle 2009, 2011; Burge 1998: 251; McDowell 1998a: 434. 31  See Pettit and Smith 2003 for this notion of freedom of belief as “orthonomy.” Remember also Wolf ’s 1990 “Reason View” that I considered in Chapter V with regards to free will.

176  Kant on Freedom and Rational Agency theoretical reasons to believe that p. Such a representation (typically) compels a belief that p. This rules out any form of epistemic akrasia that is structurally analogous to volitional akrasia. Nonetheless, for Kant our epistemic power to get things right is shadowed by a tendency to adopt rationally illegitimate doxastic states. Theoretical belief as such aims at the truth: this provides a constraint on what may count as an act of conscious belief-­formation.32 But concepts such as “truth” or “evidence” are empty in the sense that they do not, on their own, provide concrete guidance as to what one should believe. To adopt a belief, thinkers must form for themselves a representation of their evidence, of what considerations bear on the truth of some proposition. Hence, the notions of truth or evidence call for interpretation. If it is always possible for us to (carelessly and imputably, though not knowingly) misinterpret our evidential situation, then it always possible for us to judge improperly. There are three aspects of Kant’s epistemology which jointly explain why this possibility characteristically arises in our free doxastic agency. First, Kant’s epistemology is holistic: individual perceptions depend for their evidential significance on how and where they fit into a unified framework of further perceptions, concepts, and laws. This assumption is based on the (regulative) idea that the disparate parts of nature which our perceptions disclose to us are systematically connected: hence “we must endeavor, wherever possible, to bring systematic unity into our cognition” (A650/B678).33 The holistic character of Kant’s epistemology illuminates why he posits a reflective gap between receiving perceptual input and judging. It also illuminates why he endorses a norm demanding that our judgments be based on either active reflection or maxims of caution that prescribe policies of suspending judgment (or weak opining) until further reflection. If empirical propositions had their confirmation conditions in isolation from one another, the need for reflection and caution would be greatly diminished: having sufficient evidence for accepting a certain proposition would amount to receiving perceptions that individually confirm the proposition.34 By contrast, holism requires that thinkers actively work out what, if any, implications a given perception has in light of their overall system of cognition. This holistic pursuit of empirical knowledge is a complex task that requires an informed grasp of nature as a whole. Accordingly, the holistic structure of empirical justification also increases the potential for error or careless misjudging. Thus, a principled 32  See Shah 2003. 33  Without the search for systematic unity, we would lack “a sufficient criterion of empirical truth” (A651/B679). What exactly this means is a matter of debate. At least one thing Kant likely has in mind here is that although we know that there is a determining cause for every event, we cannot cognize whether an observed regularity has the status of a necessary causal law of nature without invoking considerations of systematic unity. See KU, 5:179–80; Guyer 2003a: 285–6. For a helpful summary of Kantian holism, see Friedman 1999: 99. 34  For an example of how an atomistic conception of evidence encourages a deterministic picture of belief-­formation, see Brewer 1995. I return to this issue in Chapter VII.

Kant ’ s Free Thinker  177 stance of taking active, cautious reflective care not to misconstrue one’s evidence is an epistemic virtue. Here one might object that whatever the merits of such holism may be in the domain of systematic scientific reflection, surely our ordinary everyday beliefs concerning particular empirical states of affairs are not subject to holistic evidential constraints: there is no need here for active reflective control that counteracts a pervasive tendency to err. This brings me to my second point: the pervasiveness of empirical illusion. For Kant, empirical illusion is not just a feature of certain special cases, as when we see a straight stick as bent in the water. Rather, our perceptions always present empirical objects in ways that depend on the contingent subjective constitution of the perceiver (“the special character of the subject”). Whenever we are affected by outer objects, we cannot help having “[S]ensations of colors, sounds, and heat [that] do not of themselves yield cognition of any object” and thus “cannot rightly be regarded as properties of things, but only as changes in the subject . . . which may . . . be different for different human beings” (A29/B45). Even an objective property such as motion is displayed by our senses in a distorted fashion: “The course of the planets is represented to us by the senses as now progressive, now retrogressive”; this is a merely “subjective mode of representation” (Prol, 4:290–1). Since our senses constantly give us subjective impressions of this sort, we are globally susceptible to empirical illusion whenever we represent empirical objects. Moreover, this susceptibility continues to affect us even once we see through sensory illusions. If we recognize in hindsight what led us to make some empirical judgment, “. . . without having to take account of the character of the object, we expose the illusion and are no longer deceived by it, although we are always still in some degree liable to come under its influence, in so far as the subjective cause of the illusion is inherent in our nature” (A821/ B850). To prevent our assents to merely subjective appearances of color, taste, etc., we must exercise constant reflective care: even those who have recognized the deceptive nature of these appearances must keep bringing this realization to the foreground of their reflective consciousness in order to resist the influence of empirical illusion upon their judgments. Thus, Kant’s rejection of a naïve realism according to which our perceptions accurately represent the real empirical character of objects further explains why he posits a reflective gap between the reception of perceptual input and the formation of empirical judgments, and why in his view our empirical assents ought to be regulated by reflection about what perceptual states truly reveal about nature.35 35 Chignell (2007b: 46) denies that for Kant epistemic justification requires reflection: “. . . Joe’s assent that the sky is blue can be justified simply by way of his having the experience of the sky as blue; he doesn’t also need actively to introspect and determine that his perceptual experience is a sufficient objective ground for his assent. Kant would presumably agree with this.” But for Kant, Joe’s assent that the sky is blue is not only unjustified but even wrong, since “colors are not properties of . . . bodies [but] . . . are . . . only . . . effects accidentally added by the particular constitution of the sense organs” (A29).

178  Kant on Freedom and Rational Agency My third and most complicated point in explicating why our free doxastic agency involves a susceptibility to theoretical error concerns the influence of what Kant calls transcendental illusion. Unlike empirical illusion, which may depend on the contingent constitution of our sensibility, this is a form of illusion which finite “reason must necessarily encounter in its progress” (A422/B450).36 I cannot examine the difficult notion of transcendental illusion in any greater detail here; I will merely sketch what “harmful influence” (A642/B670) such illusion has on our doxastic agency.37 This influence has several different facets. First, we have the tendency to misuse the idea that represents a complete totality of explanatory conditions for a given explanandum (A308/B365; A462/ B490), namely, when we claim to know how this totality is constituted. This tendency may lead either to the empiricist notion that the totality consists in an infinite, never-­ceasing series of empirical conditions that cannot have a first beginning or cause (A470/B498), or to the rationalist notion that this totality must consist in a finite series terminating in a privileged member (a first beginning or cause). Neither of these notions is supported by our objective evidence because a totality of empirical conditions is not itself an object of our experience (A482–4/B510–12). Moreover, both notions have highly problematic upshots (cf. Chapter I): the rationalist picture posits free causes as explanatory conditions and thereby destroys our understanding of nature (given the incomprehensibility of freedom); on the empiricist view all there is must be part of the empirically conditioned order of nature, which means that free causes are absolutely impossible (A472/B500). Second, we have a tendency to misuse the idea that represents the purposiveness of organisms, namely, when we take this idea to designate an objective (“constitutive”) feature of nature. This may lead to the claim that purposiveness is strictly impossible because it conflicts with the objective mechanistic structure of nature (KU, 5:187–8). This claim hinders our quest for achieving systematic unity in our empirical cognition of nature because the (regulative) idea that nature is purposively arranged helps us to find systematic connections among disparate Chignell might respond that generally speaking (leaving aside the special case of color), perceptions can justify our assents independently of reflection. But for Kant reflection on perceptions (at least implicit reflection that is regulated by adequate, explicitly reflective epistemic maxims) is generally required for empirical justification because whenever we make judgments about physical objects, we face the need to determine whether the content of our perceptions indicates an objective quality that falls under public concepts or rather derives from the private character of sensibility. Owens (2000: 1, 5, 9) notes that Kant, Locke, and Descartes all demand that we should determine our empirical judgments through reflection. In my view, this agreement derives (in part) from the fact that all these modern philosophers deny that our unreflective perceptual awareness adequately represents the true, empirically real constitution of physical reality. 36 For the distinction between empirical and transcendental illusion, see A296/B352. For the necessity and the indestructibility of transcendental illusion, see A421–2/B449–50. 37 Like perceptual illusion, transcendental illusion does not necessitate erroneous judgments. Unlike perceptual illusion, transcendental illusion can also have a positive epistemic function (see Grier 2001 for helpful discussion).

Kant ’ s Free Thinker  179 (both organic and mechanistic) phenomena (A651/B679; KU, 5:183). If, on the other hand, the constitutive misinterpretation of natural purposiveness leads to the conviction that nature is objectively purposive, this may encourage epistemic vices such as the “laziness” of reason which cuts off the arduous search for mechanistic causes by appealing directly to the alleged purposes ordained by an all-­ wise creator (A689–95/B717–23). Such epistemic vices may have profoundly negative effects on our thinking, both within and outside the context of scientific reflection (consider, e.g., the impact of lazy reason upon members of fundamentalist religious communities). If we consider specifically the context of scientific reflection, we must note the further problematic tendency to objectify the ideas that (1) various different natural objects can be systematically understood by reference to common properties (A652/B680) or that (2) there is sheer endless “manifoldness and diversity” in nature (A654/B682). If these ideas are misconstrued as describing the objective constitution of nature, they come into conflict: the notion that nature is intrinsically heterogeneous is hard to square with the notion that nature is intrinsically unified.38 Thus, under the influence of this objectifying illusion researchers neglect the search for either homogeneity or heterogeneity. This hinders their pursuit of empirical truth (see A667–8/B696–7) because we need to look for both unifying and diversifying properties to achieve the widest possible expansion of our unified web of empirical cognition: i.e., we must comply with the norm that “prescribes that we ought to study nature as if systematic and purposive unity, combined with the greatest possible manifoldness, were everywhere to be met with” (A701/B729). The point of my brief discussion of transcendental illusion was only to show that for Kant finitely rational thinkers are prone to think about reality in a subtly but gravely distorted fashion. This tendency is rooted in the very nature of finite theoretical reason, which provides a transcendental (i.e., a necessary, a priori) ground for the possibility of theoretical error. The ineradicable force of transcendental illusion, the holistic character of empirical knowledge, the global presence of empirical illusion, plus other factors such as concealed wishful thinking or misremembering (see A53/B77) jointly afflict us with a propensity to misinterpret our theoretical reasons for belief. These factors influence (but do not determine) us to adopt inadequate, e.g., prejudicial maxims of judging which in turn lead us to make particular judgments on the basis of merely subjective epistemic grounds that we mistakenly regard as 38  See A666–7/B694–5. Guyer (1990: 32) argues that the conflict between (1) and (2) could only be practical, relating to limits on time and resources. But Kant’s point pertains rather to the general view of nature that, however implicitly, guides thinking subjects in their theoretical inquiries: if (1) and (2) are each taken as objective insights into the constitution of nature they cannot be combined into a coherent reflective stance because the claim that nature is objectively completely homogenous clashes with the claim that nature is objectively completely heterogeneous.

180  Kant on Freedom and Rational Agency objective. For Kant, systematic reflection or adopting maxims of epistemic caution (that block firm assent until further thorough reflection) are our only safeguards against our propensity for theoretical error. I do not take myself to have conclusively shown already that for Kant epistemic agency is transcendentally free or involves an absolutely unconditioned cognitive spontaneity. But my argument in this section shows that there is nothing absurd or inherently implausible about the idea that we are free in our theoretical judgments and beliefs (assents): such freedom of thought can be understood negatively as the absence of sensible-­empirical determination and positively as a reflective, norm-­governed control that we exercise over our cognitive states. Moreover, if we indeed have such free control over our beliefs, this puts at least significant pressure on the idea that a thinking subject is a mere thinking mech­ an­ism. For Kant, our free rational powers always enable us to correctly interpret our evidential situation: cognitive errors are never forced upon us.39 But the imperfections inherent in our finite rational powers also frequently afflict us with the privative option to misapprehend our theoretical reasons: in many cases of theoretical, especially empirical judgment the truth is not forced upon us either. Thus, we are (typically) not necessitated to judge either correctly or incorrectly. By contrast, a thinking mechanism would always be necessitated to respond to external stimuli in one particular manner: it would thus always be determined either to function or to malfunction.

VI.3.  Cognizing Ourselves as Spontaneous Noumenal Cognizers According to an influential interpretation, Kant ascribes absolute freedom of thought to human cognizers only in his (partly pre-­critical) views during the 1770s and the early 1780s but then abandons this position and moves towards a thinking mechanism-­conception of theoretical cognition during the mid- to late 1780s. On this interpretation, some of the passages that I cited at the beginning of this chapter cannot be invoked as conclusive evidence against the thinking mechanism-­reading since they do not express Kant’s considered critical view. Here we must consider first and foremost an apparent tension between the 1781 A-­edition and the 1787 B-­edition of the Critique of Pure Reason. One of the passages that I cited earlier stems from the A-­edition discussion of the Third Antinomy (A546–7). There, Kant says that a human being theoretically “cognizes itself through pure apperception,” the “I think,” “as a merely intelligible” or noumenal “object” 39  This does not entail, absurdly, that we can always know the truth about any given empirical subject matter: “correctly interpreting our evidential situation” often means suspending judgment due to the recognition that we lack sufficient evidence, or forming an opinion with an appropriately weak degree of conviction, etc.

Kant ’ s Free Thinker  181 whose understanding and reason must be “distinguished . . . from all empirically conditioned powers.” By exercising these empirically unconditioned epistemic powers, we determine ourselves through spontaneous “actions and inner determinations.” By contrast, in the 1787 B-­Paralogisms Kant argues that it is “impossible” for the thinking self “to cognize it[s] [existence] as noumenon” (B430): my apperceptive self-­awareness of my “sheer spontaneity” gives me no noumenal self-­ cognition because the application of this spontaneity “to sensible intuition . . . would . . . be demanded if I wanted to cognize myself,” and sensible “intuition always makes available the object . . . merely as appearance” (B429). Kant now stresses that we require our practical self-­awareness as an “occasion for presupposing ourselves to be . . . self-­determining in [our] existence”: only the moral law reveals “a spontaneity through which our actuality is determinable without the need of conditions of empirical intuition” (B431–2). These passages raise two separate issues. First, can we cognize our noumenal (“intelligible”) selves as subjects of absolutely spontaneous thinking activity? Second, can we ascribe to ourselves the capacity for spontaneous self-­determination only on a priori practical-­moral grounds or does our a priori theoretical self-­ consciousness also reveal that we possess this capacity? I address the first issue in this section and the second one in Section VΙ.3.40 The first issue concerns the seeming conflict between Kant’s statements that we can (1781 A-­edition) and cannot (1787 B-­edition) cognize our noumenal thinking self. One common response to this tension is that by 1787 Kant has abandoned his former view that we can cognize our noumenal capacity for absolutely spontaneous self-­determination on a solely theoretical basis (via pure apperception).41 For others, Kant’s appeal to apperceptive cognition of our noumenal thinking self (at A546–7) is just a careless slip: Kant “here misstated his position.”42 I believe that Kant’s 1781 and 1787 claims about the (im)possibility of apperceptive noumenal self-­cognition are consistent: they express two central components of his critical philosophy. My guiding idea is that Kant has two different notions of theoretical cognition.43 Cognitions in a narrow sense (N) are representational states that combine concepts with sensible intuitions and thereby, unlike “empty” concepts or “blind” intuitions, inform us about the properties of actual (“given”) objects (A50–2/B74–6; B146–7). Kant’s wider notion of cognition(W) contrasts cognitions with sensations. Empty concepts (and blind intuitions) qualify as cognitions(W) because they are conscious representations that relate to

40  Here I will draw (but significantly expand) on Kohl 2020a. 41  See Ameriks 2000a: 217; 2003: 166–7; Hogan 2009a: 379; Keller 1998: 159. 42  Allison 2020: 270. Similarly, for Proops (2021: 84) Kant here invokes (when he makes similar-­ sounding claims in the B-­deduction, which I consider below) “an epistemic state that has no official place in his epistemology.” 43 For this point I am indebted to Grüne 2007: 29–30 and to the discussion in Watkins and Willaschek 2017a. I discuss this distinction at greater length in Kohl (forthcoming3).

182  Kant on Freedom and Rational Agency objects. Empty concepts relate to objects in the sense that they have intentional aboutness: they represent something distinct from the occurrent mental state of conceiving. This intentional relation to an object need not involve any existential commitment, but via cognition(W) we can at least represent something as existing apart from (external to) our mental states. Thus, all purely intellectual representations, including even the transcendental “general cognitions of reason” (A421/B448), are cognitions(W). By contrast, sensations like feelings of pleasure and pain have no cognitive content: as mere occurrent modifications of some particular mind they are not intentionally related to (are thus not representations of) any object or property (A320/B376; JL, 9:91; V-­Met/Schön, 28:471). In the B-­Paralogisms Kant uses the concept “cognition(N)” when he argues that since pure apperception lacks intuitive content it cannot yield cognition of our noumenal self. This does not contradict his claim (at A546/B574) that through pure apperception a human being can cognize itself as an intelligible (noumenal) object if we assume that here Kant uses the different concept “cognition(W).” This assumption does not seem excessively charitable: since Kant repeatedly stresses, in both editions, that pure apperception lacks the intuitive content required for self-­ cognition(N) (A382; A400; B159), we can conjecture that his appeal to self-­ cognition through purely conceptual apperception involves the broader notion of cognition(W) that does not require intuition. In designating pure apperception as a kind of cognition despite its lack of intuitive content, Kant points out that pure apperception is not a private, non-­intentional mental state but a universally shared intellectual representation through which every thinking being conceives itself as the transcendental subject of thoughts (A346/B404; B423). Since cognition(W) is defined as a conscious representation that relates to an object, pure apperception can yield cognition(W) of our noumenal self only if it represents our thinking self “as a merely intelligible object” (A546). Kant’s conception of object-­representation allows for this possibility. To be sure, he has a narrow conception of object-­representation that applies only to mental states which combine intuitive and conceptual components into determinate, informative cognition(N) of phenomena and their spatiotemporal properties. When Kant denies that pure self-­consciousness represents the thinking self as an object (e.g., at B407), he uses this narrow notion of an object that designates a determinate object of cognition(N). But if one abstracts from all sensible data for cognition(N), one can still represent an indeterminate something via the pure concept of “an object in general” that is the highest, i.e., most generic concept in all cognition (A290/B346; V-­Met/L2, 28:544). Accordingly, pure apperception indeterminately represents the thinking self as “something in general” (A400; cf. A345–6/ B403–4; A383; B421). Because my pure apperception intentionally relates to a non-­sensible (“merely intelligible”) something that is conceived as distinct from my occurrent mental states, pure apperception yields cognition(W) of myself as a noumenal object.

Kant ’ s Free Thinker  183 A problem for my reading arises from Kant’s B-­edition assertion that through the representation “I think,” “I am conscious of myself not as I appear to myself, nor as I am in myself, but only that I am” (B157; cf. B429). Here, one might argue, Kant explicitly rules out that pure apperception yields cognition of an intelligible object (of myself “as I am in myself ”).44 But we must consider what Kant means by his claim that in pure apperception I am not conscious of myself “as” I am in myself. He argues for this claim by first stressing that I require sensible intuition to cognize(N) how I exist as a thinking being; from this, he infers that non-­sensible apperception does not yield “cognition of oneself.” “I therefore have no cognition of myself as I am, but only as I appear to myself ” (B158). Thus, when Kant denies that pure apperception gives me a consciousness of myself as I am in myself, he means that pure apperception fails to yield the specific kind of self-­cognition(N) which requires sensible intuition. In pure apperception I am not conscious of myself “as” I am in myself because my purely intellectual self-­cognition(W) provides no determinate information about the manner in which I exist as a thing in itself (B420). In the B-­edition Kant still accepts that pure apperception represents the thinking self as an “object in general” (B158; B429) and that it provides an indeterminate ontological self-­characterization: “I exist as an intelligence that is merely conscious of its faculty for combination.” Although I cannot cognize(N) my spontaneous intellect, “yet this spontaneity is the reason I call myself an intelligence” (B158); “the I that I think . . . differ[s] from the I that intuits itself ” because the I that I think is the “I as intelligence and thinking subject” (B155). Our self-­awareness as an intel­ ligence concerns our noumenal self: to consider human beings as noumena is to consider them according to the capacities they have as non-­sensible intelligences (MS, 6:226; cf. Chapter III).45 Thus, Kant’s remarks about pure apperception in the B-­edition are compatible with his claims in the A-­edition Dialectic: my pure self-­consciousness gives me indeterminate cognition(W) that I exist as a spontaneous noumenal intelligence. Nevertheless, some commentators insist that in Kant’s critical doctrine pure apperception cannot represent a noumenal object. Here I focus on Michelle Grier’s influential reading. She makes two claims that are especially relevant for our topic. First, the concept “I” that one employs in pure apperception relates merely to the formal or logical subject of thought, which “is not an object of any sort.”46 Second, “the conflation of the ‘I’ of apperception with a fictitious object”47

44  See Kitcher 1984: 121. 45 For the analytic link between “intelligence” and “noumenal being,” see GMS, 4:452, 457–9; V-­Met-­L2, 28:583; V-­Met-­K2:773; V-­Met-­K3E, 29:1020. See also Puls 2015: 184–5. Against Schönecker (1999: 287–9), Puls argues that “intelligence” designates the capacity for moral agency rather than the capacity for theoretical thought. However, B158, MS, 6:226 and GMS, 4:452 jointly suggest that the term designates both capacities. 46  Grier 2001: 160. This is also a major theme in Rosefeldt 2000. 47  Grier 2001: 170.

184  Kant on Freedom and Rational Agency occurs under the influence of a transcendental illusion, when pure reason hypostatizes the logical subject of thought into an “illusory idea of the soul” as a thing in itself or noumenon.48 When Grier claims that the pure concept “I” represents no object, she means that it does not represent a determinate object of intuitive cognition(N) because “it is unacceptable to think of this ‘I’ as an object given in . . . intuition.”49 She accepts that the transcendental subject of thought can be conceived as an “object in general.”50 Accordingly, in the passages she cites (B407, B409–10) Kant makes the same point as in the B-­deduction (B157–8): without self-­intuition I cannot cognize(N) myself as a determinate object (that has specific properties or characteristics). But for Grier the purely conceptual representation of a non-­sensible object in general is an illusory thought of a mere pseudo-­object that lacks real ontological import: an illusion of reason is needed to create the fictitious idea that pure apperception represents the thinking self as an existent non-­sensible being. These claims rest on Grier’s general conviction that Kant’s appeal to things in themselves involves no commitment to the existence of non-­sensible objects: “. . . critically understood, ‘things in themselves’ are not objects at all but certain ways of considering objects” in abstraction from the sensible conditions of human cognition(N).51 For Grier, “the consideration of things as they are in themselves is basically an illusory way of representing appearances.”52 My account of Kant’s views on self-­cognition presupposes that this deflationary reading is mistaken: Kant’s idealism does posit the existence of non-­sensible things in themselves in addition to sensible appearances.53 I have argued for a metaphysically weighty interpretation of Kant’s idealism in Chapter I. For Kant, “beings of the understanding certainly correspond to sensible beings” (B308): if one accepts the existence of sensible appearances, one must accept not merely (as proponents of deflationary readings suggest) the concept of non-­sensible beings but “the ex­ist­ ence of things in themselves” that “underlie the appearances” (Prol, 4:315) as their “true correlate” (A30/B45) (cf. Bxx; A538/B566; A496/B524, Prol, 4:355, 361). The link between a metaphysical reading of Kant’s idealism and my reading of his views on self-­cognition is as follows. On a metaphysical reading, Kant’s idealism holds: for everything that we determinately cognize(N) as a sensible appearance there exists some non-­sensible being that we can represent only 48  Grier 2001: 168–9. 49  Grier 2001: 161. 50  Grier 2001: 160, 170. 51  Grier 2001: 90. Here she follows Allison’s take on Kant’s idealism. Allison also reads Kant as warning against conflating “the indeterminate concept of a thinking being” with a “transcendent entity” (2004: 350). Similarly, for Rosefeldt (2000: 73–5) the pure “I” refers to a non-­sensible being only qua logical pseudo-­object which he understands as an ens rationis; this point seems based on his general view that all things considered apart from our sensibility (noumena in the negative sense) are mere entia rationis since we cannot cognize whether they exist. 52  Grier 2001: 278. 53  Here my reading is similar to that of Martin (1969: 207–8), who argues (against Cohen and Natorp) that a deflationary interpretation of the transcendental subject of thought is based on an implausibly deflationary understanding of Kant’s appeal to things in themselves.

Kant ’ s Free Thinker  185 indeterminately, via purely conceptual cognition(W), “as a something in general outside of our sensibility” (B307).54 Here Kant combines his wide or indeterminate object-­ concept with his negative concept of noumena; via this combination, we can conceive some existent object in general that does not conform to our forms of sensibility and is thus not a sensible appearance. This yields a negative-­indeterminate conception of an intelligible being: the “intelligible [is] that in an object of sense which is not itself appearance” (A538/ B566). When we negatively represent intelligible beings, we refer indeterminately to those (unknown) features of things which differ from the sensible constitution that they have as appearances. Accordingly, we can determinately cognize(N) ourselves as sensible appearances and we also have purely conceptual cognition(W) of our intelligible (noumenal) thinking self as an object in general whose constitution does not conform to our sensibility.55 Kant thus distinguishes between the sensible cognition(N) that a human being has of itself as an empirical phenomenon and the purely intellectual-­apperceptive cognition(W) that a human being has of itself as a “purely intelligible object” (A546–7/B574–5), “the I as intelligence and thinking subject” (B155). Grier is right to stress that for Kant the “I” of pure apperception designates only a logical subject. But in Kant logical characterizations may well refer to real actions or beings.56 Pure apperception (unlike inner sense) involves an action awareness, “a consciousness of what the human being does” (Anth, 7:161), an “intellectual representation of the self-­activity of a thinking subject” (B278). Moreover, Kant specifies “this logical meaning of the I” through the ontological notion of inherence: the pure “I” refers to myself as the transcendental bearer of all my thoughts (A349–50). The pure “I” is a merely logical notion because it supplies “no determinate predicate” for conceiving this bearer (A479/B507), i.e., no predicate like permanence, persistence through change, absolute subjecthood (where what is conceived as a subject of inherence cannot, in turn, inhere in some more fundamental subject; cf. Prol, 4:333–4), or absolute separability from

54  Recall from Chapter I that a metaphysical reading is not committed to the idea that noumena and phenomena are separate entities. See Allais 2015: 3–36; Ameriks 2003: 35–8; Marshall 2010, 2013. 55  One might argue that pure apperception cannot represent the self as a non-­sensible object because Kant characterizes the “I think” as an empirical proposition (B423). However, it is unclear what he means by this (and how seriously he means it: at A346/B404 he refers to an “empirical-­ seeming proposition”). He might seek to stress the contingency of the existential assertion that is implied by the “I think,” or (perhaps: and) that apperceptive self-­consciousness requires empirical intuitions as data for spontaneous thinking activity (Allison 2004: 354; Proops 2021: 82–4; see A343/B401 where Kant clarifies that the “I think” is the transcendental condition of all inner empirical cognition(N)). All this is compatible with the idea that the “I think” indeterminately represents my thinking self as an intelligible being. Longuenesse (2017: 87–8) distinguishes between (1) the “I think” qua “pure action-­awareness” that (indeterminately) represents the thinking being in itself and (2) the indeterminate empirical intuition that I think. I am concerned only with (1) here. 56  Here I concur with Kitcher’s critique of “logical” readings of apperception that seek to remove “real” implications concerning our mental activities (1990: 94–5). See also Longuenesse 1998: 74–8.

186  Kant on Freedom and Rational Agency material beings.57 Since the logical meaning of “I” abstracts from all these predicates, Kant concludes that “apart from this logical meaning of I, we have no ac­quaint­ance with the subject in itself that grounds this I as a substratum, just as it grounds all thoughts” (A351). This statement implies that the logical meaning of “I” does relate to the thinking subject as a thing in itself that grounds all (its) thoughts, though without yielding any determinate information about how this subject is constituted in itself: “the I, as the general correlate of apperception . . . designates . . . a thing of undetermined meaning” (MAN, 4:542). Through the “I think” the thinking self is “only transcendentally signified” as a “something in general.” While this transcendental signification does not reveal “the least property” (A355), it does refer to a noumenon (cf. A358): to “the real self, as it exists in itself, or the transcendental subject” (A492/B520). In the B-­edition Paralogisms, Kant likewise says that “in the consciousness of myself in mere thinking I am the being itself, about which, however, nothing yet is thereby given to me for thinking” (B429). Through the concept “I,” I can “signify” myself as an object in itself but I cannot cognize the manner in which I exist as a thing in itself (B430). The indeterminate apperceptive representation of myself as a non-­sensible something does not entail the problematic characterizations of the thinking self that rational psychology aspires to. This is because the concept of a “thing,” as the highest concept in all cognition, is less determinate than even the concept of substance (B6; JL, 9:97) which plays a crucial role in rational psychology.58 While the notion of a spontaneous intelligence goes beyond the mere thought of something in general, Kant denies that my pure consciousness of myself as a spontaneous intelligence gives me a determinate concept of how I exist as an intelligible being (B158). Since the notion of a spontaneous intelligence is not a target of Kant’s critique in the Paralogisms,59 the self-­conception of a spontaneous noumenal intelligence does not entail the informative metaphysical characterizations of the thinking self that are the targets of Kant’s critique.60 These points should go some 57  Kant’s denial that we have cognitive grounds to settle the separability issue rests on his view (which looms large in the Paralogisms) that we cannot know whether the transcendental subject that grounds bodily appearances is identical to the transcendental subject that grounds mental appearances (A379–89; Ameriks 2000a: 34). 58  It is controversial whether Kant can avoid the idea that the noumenal self is an (ultimate) substance. Proops argues that Kant avoids this “dogmatic” idea (2021, chs. 4, 5). For Ameriks, Kant must concede the substantiality of the noumenal soul (2000a: 72–3; cf. Wuerth 2014: 167–8), but this forces him to dismiss Kant’s A-­edition Paralogism argument as “spurious” (2000: 66–7). For Longuenesse (2017), we are subjectively constrained to regard ourselves as ultimate substances but we cannot know whether we really are. 59  See Ameriks 2000a: 190; Rosefeldt 2000: 150–7. 60  Against this, Grier and Allison might invoke B426–7 where Kant says that the conflation of the indeterminate concept of a thinking being with the idea of a pure intelligence betrays a transcendental illusion. However, if Kant meant that our self-­awareness as a spontaneous intelligence is illusory, he would have fallen prey to that illusion at B155–8. The illusory idea Kant targets at B426–7 is not merely that of a spontaneous noumenal intelligence but the more determinate (thus controversial)

Kant ’ s Free Thinker  187 way towards answering Sellars’s worry (cf. Section VI.1) that if Kant allowed for cognition of ourselves as spontaneous noumenal cognizers who are capable of imputation, he would be committed to a dogmatic rational psychology. I return to this issue in Chapter VII.

VI.4.  Two Senses of “Self-­Determination”: Cognitive versus Causal Spontaneity Kant’s distinction between two senses of “object-­cognition” allows us to see how his A-­edition appeal to our apperceptive self-­cognition as spontaneous noumena is compatible with his more skeptical remarks in the B-­edition Paralogisms. However, there is a further seeming tension between the two editions which concerns Kant’s notion of spontaneous self-­determination. At A546 Kant asserts that pure apperception reveals our capacity for spontaneous “actions” and “inner determinations” that derive from our empirically unconditioned intellectual capacities. Student notes on Kant’s lectures from the 1770s and early 1780s indicate similar claims: “Now the ‘I’ proves that I myself act . . . I am conscious in myself of determinations and actions, and a subject who is aware of such actions and determinations has libertatem absolutam” (V-­Met-­L1, 28:266–9); “The intellectual capacity of cognition rests upon spontaneity or the capacity to determine oneself ” (V-­Met/Mron, 29:881). As we saw, in the 1783 Prolegomena Kant stresses that “we are . . . completely free as to how we want to judge things” through “the understanding” (Prol, 4:290). However, by 1787 Kant deems it “impossible” for the thinking self “to determine its kind of existence” via the “sheer spontaneity” of understanding: we can only regard ourselves as “self-­ determining in [our] existence” if we presuppose laws of practical reason (B430). Here Kant seems to claim that our pure theoretical self-­awareness as thinking beings cannot reveal our capacity for free, absolutely spontaneous self-­ determination.61 For some commentators this is also his considered view: according to Allison, Kant came to recognize that the “sheer spontaneity” of thought is insufficient for genuinely free agency or self-­determination.62 Kant’s various remarks about spontaneous self-­determination are certainly confusing and seem to contradict each other.63 Consider, for instance, the following crucial but very difficult passage from the B-­deduction (B157–8):

notion of a separable “pure intelligence” that could exist apart from my empirical self and thus from all material, corruptible beings (see B409; B420; A741–2/B769–70). 61  This is how Ameriks reads the B-­Paralogisms (2000a: 217; 2003: 166–7). 62  Allison 1990: 63 (for complications in his view, see n. 7). 63  For Kitcher (1984: 121), these remarks are “individually perplexing and mutually inconsistent.” For helpful discussion of various meanings of “self-­determination,” see Ameriks 2019: 14–18.

188  Kant on Freedom and Rational Agency The I think expresses the act of determining my existence. . . . Now I do not have yet another self-­intuition, which would give the determining in me . . . even before the act of determination, in the same way as time gives that which is to be determined, thus I cannot determine my existence as that of a self-­active being, rather I merely represent the spontaneity of my thought, i.e., of the determining, and my existence always remains only sensibly determinable, i.e., determinable as the existence of an appearance.

Here Kant makes three assertions:

(1) Through pure apperception, I represent myself as a self-­active thinker whose spontaneity is “the determining in me.” (2) Through pure apperception, I cannot determine my existence as a “determining” thinker. (3) If I combine pure apperception with empirical intuition, I can determine my existence (actuality) as a temporal appearance. Furthermore, in the B-­Paralogisms (at B430) Kant claims: (4) Only our awareness of practical laws “discloses a spontaneity through which our actuality is determinable without the need of conditions of empirical intuitions”; in this awareness “something is contained a priori that can serve to determine our existence.” Finally, we have the abovementioned remarks from lecture notes, A546 and the Prolegomena: (5) The capacity for theoretical cognition is the spontaneous “capacity to determine oneself ” via “actions and inner determinations” that result from the free exercise of our empirically unconditioned intellectual powers (theoretical reason and understanding). I propose that a first step towards an interpretation where (1)–(5) yield a coherent view is the disambiguation between two different senses of “spontaneous determination.” At A92–3/B124–6 Kant notes that there are two different senses “in which the [a priori] representation alone makes the object possible.” In the first case, “the representation by itself ” through “its causality by means of the will . . . produce[s] its object as far as its existence is concerned.” I call this practical-­causal determination (PCD). In the second case a representation “does not produce its object as far as its existence is concerned” but “is still determinant of the object a priori if it is possible through it alone to cognize something as an object.” By

Kant ’ s Free Thinker  189 “cognition” Kant here means object-­cognition(N) through the conceptual synthesis of sensible intuition: “With us understanding and sensibility can determine an object only in combination” (A258/B314). I call this theoretical-­cognitive determi­ nation (TCD).64 The distinction between PCD and TCD is illuminating with regards to (1)–(5). In (1) and (5), Kant stresses that through pure apperception I represent myself as a thinking being which has the capacity for spontaneous TCD of myself. In (2), he denies that pure apperception enables the TCD of myself as a spontaneous thinker: since TCD of any object requires sensible intuition, the purely conceptual representation of my spontaneous thinking self is insufficient “to determine its kind of existence, i.e., to cognize it as a noumenon” (B430). In (3), Kant says that if I combine pure apperception with empirical intuition I can “determine my existence” in the sense of TCD: I can cognitively determine myself as a non-­ spontaneous, empirically conditioned temporal phenomenon via empirical self-­ cognition(N). In (4), Kant stresses that moral agents may ascribe to themselves a capacity for purely a priori self-­determination in the sense of PCD. I can causally determine myself when I exercise the capacity of pure reason to be “practical,” to “determine the will by itself, independently of everything empirical” (KpV, 5:42), e.g., of all contingent empirical desires. This is a causal self-­determination because the pure exercise of my will effects my spatiotemporal actions and thereby causally influences the course of my phenomenal existence (B430). This analysis removes some of the tensions between (1)–(5). The claim (in (4)) that only our practical self-­awareness under the moral law reveals our capacity to spontaneously determine our existence does not contradict the claim (in (1), (3), and (5)) that pure apperception also reveals a capacity to spontaneously determine our existence, because these two claims concern two different types of spontaneous self-­determination (PCD vs. TCD). Moreover, since our capacity for spontaneous self-­determination in the theoretical-­cognitive sense (for obtaining self-­cognition(N)) requires empirical self-­intuition, Kant’s claim (in (5)) that pure apperception reveals our epistemic capacity for TCD is perfectly consistent with his claim (in (4)) that only the moral law reveals our capacity for spontaneous self-­determination in a different, practical-­causal sense that does not depend on “conditions of empirical intuition.” One might wonder: if there is no genuine conflict between the two editions, then why does Kant so strongly emphasize in the re-­written B-­Paralogisms but not in the earlier A-­edition that we have a solely practical capacity for spontaneous self-­determination that is independent of all empirical conditions? I concede that there is a significant difference between Kant’s 1781 and 1787 doctrine. But this difference concerns Kant’s conception of the spontaneity of will that is 64  See Watkins and Willaschek 2017a for helpful discussion of cognition(N) as conceptual “determination” of intuition.

190  Kant on Freedom and Rational Agency involved in PCD, not his conception of the spontaneity of thought that is involved in TCD. As we saw (in the appendix to Chapter  I), in the A-­edition Kant still holds a heteronomous view of moral motivation that he came to reject only by 1785. According to that view, we are not capable of PCD independently of empirical conditions because our incentive to follow the moral law rests on our expectation of divine rewards in the afterlife, a motive which depends on our empirical desire for happiness. Hence, Kant’s 1787 conception of PCD was not yet available to him in 1781. Once he has developed his doctrine of moral autonomy, he is in a position to stress that our capacity for PCD, unlike our capacity for TCD, does not depend on the limiting conditions of empirical intuition or sensibility, at least when we exercise our capacity for PCD in morally good choices: “The essential point in every determination of the will by the moral law is that being a free will it is determined simply by the moral law, not only without the co-­operation of sensible impulses, but even to the rejection of all such” (KpV, 5:75).65 There is a further potential conflict between (4) and (5): according to (4), the spontaneity of understanding depends on conditions of empirical intuition whereas in (5) Kant refers to our empirically unconditioned theoretical faculties. I suggest that in (5) he means to stress the pure origin of intellectual representations that govern acts of TCD. Our theoretical categories and ideas are not abstracted from passively received empirical data; rather, they arise spontaneously from our purely intellectual self-­activity. But the determinate use or application of pure concepts in acts of TCD (including acts of self-­cognition(N)) always requires empirical data: “the empirical is only the condition of the application or the use of the purely intellectual capacity” (B423). The cognitive determination of objects that results from the application of purely intellectual representations to empirical data differs depending on whether these representations originate in pure reason or the pure understanding. Ideas of reason cannot cognitively determine objects or yield cognitions(N) directly. But they contribute to acts of TCD by unifying the various cognitions(N) of the understanding into cognitive determinations of more complex objects or relations among objects. Thereby theoretical reason regulates and directs the empirical use of the understanding. This is what Kant means when he says that reason “determines the understanding” (A547/B575) through its ideas. “For even if no object can be determined through [these ideas], they can still . . . serve the understanding as a canon for its extended and self-­consistent use” (A329/B387); these ideas “determine the use of the understanding according to principles in the whole of an entire experience” (A321/B378).

65  My reading can accept that when Kant developed his view of moral autonomy he changed his strategy for proving that we have absolute moral spontaneity of will. Thus, my reading can agree with Ameriks 2000a, 2003 that there is a reversal in Kant’s argumentative strategies during the 1780s. I discuss this issue in Chapter VIII.

Kant ’ s Free Thinker  191 I have argued that the distinction between TCD and PCD allows us to make coherent sense of Kant’s confusing talk about spontaneous self-­determination in (1)–(5). This distinction promises a further interpretive benefit. If the “determining in me” that I represent via pure apperception is a capacity for cognitive rather than causal self-­determination, then ascribing the capacity for TCD does not have the controversial metaphysical implications that arise for PCD. With PCD one ascribes a productive causal power that operates independently of all sensible conditions; this raises difficult questions about how we can extend the categories beyond the limits of sensibility and how (or whether) we can demonstrate the (real) possibility of such a noumenal cause. The assumption that positing the capacity for spontaneous TCD does not invite these questions because this capacity lacks the controversial metaphysical features of PCD might explain why Kant in (both editions of the) first Critique constantly invokes our non-­empirical capacity for spontaneous TCD while insisting that we cannot prove the (real) possibility of the causal spontaneity that is involved in PCD (A557–8/B585–6). Furthermore, this assumption might also explain why Kant does not treat the “sheer” non-­causal spontaneity of mind in the Paralogisms alongside metaphysically weighty features of the thinking self. However, here my reading runs into a significant complication. I have argued: when Kant says that the spontaneity of thought in conjunction with empirical intuition enables me to “determine” my actual existence, he means that I can exercise my capacity for TCD to obtain self-­cognition(N). This is the correct interpretation for central passages such as B157–8. But when Kant designates the understanding as a spontaneous “capacity to determine oneself ” (V-­Met/ Mron, 29:881), “spontaneous self-­determination” cannot mean just “self-cognition”: a cognizing being spontaneously determines itself in all acts of TCD, whether it cognizes(N) itself or other phenomenal objects such as tables or stones. In any act of TCD, one exercises reflective control over one’s mental states by forming certain representations, combining them in a judgment, and adopting some doxastic state of assent (e.g., an opinion that a table is round). Following Kant’s language in (5), we might call this inner (self-)determination. The idea that our thinking activity in TCD spontaneously determines our inner representational states seems like a causal notion: judgment is the “effect” of “actions” of the understanding (A295/B350), experience is the “product” of the understanding (A1); the synthetic unity of intuition which yields object-­ cognition(N) is “effected” by the understanding (A105). But if we must regard TCD as a kind of causal self-­determination after all, this threatens the interpretive benefits that I tried to reap from the disambiguation between the two types of spontaneous self-­determination. A causal conception of TCD not only seems to undermine my suggestion that ascribing a capacity for spontaneous TCD involves less of a metaphysical commitment than ascribing a capacity for spontaneous PCD. It also threatens the idea that there is any decisive

192  Kant on Freedom and Rational Agency difference at all between TCD and PCD. To be sure, we can still say that the causality of the cognizing intellect, unlike the causality of the will, depends on empirical conditions. But the significance of this contrast is unclear if the contrast concerns a difference between two non-­sensible (hence non-­natural) forms of  causal self-­determination rather than a difference between a causal and a merely cognitive sense of spontaneous self-­determination. Finally, if Kant ascribes a causal status to TCD in (5) when he designates our understanding as a non-­ empirical capacity for producing actual “inner determinations,” this refuels the tension between the early 1780s view expressed in (5) and the 1787 view expressed in (4) (the B-­Paralogisms): Kant’s remarks in (4) suggest that only our moral will exhibits a non-­empirical form of causal spontaneity that produces effects in the actual world (through which “our actuality is determinable”). For some commentators who defend the thinking mechanism-­reading, we can simply dismiss the claims that Kant makes in (5): in their view TCD is indeed a causal capacity for determining our actual mental states, but this requires no metaphysically loaded claims about noumenal causality because the capacity for TCD belongs to the empirical causality of nature. For Kitcher, “the I that thinks is the phenomenal self ” whose synthesizing cognitive activity is determined by natural laws.66 These laws “govern synthesis only as the law of gravity governs the movements of the planets.”67 On this view, the thinking activity which causes cognition(N) belongs to the empirical order of nature and is therefore itself an object of cognition(N). But Kant denies these claims, as we can see in (1)–(3) (B157–8). In (1) Kant says that through pure apperception I represent my spontaneity as the agent of TCD, “the determining in me”: “the I think expresses the Actus of [cognitively] determining my existence” when I obtain self-­cognition(N) through my spontaneous conceptual synthesis of empirical self-­intuitions. Kitcher construes this synthesizing activity as a phenomenal process that can itself become an object of cognition(N). But in (2), Kant denies that the active, cognitively determining aspect of myself can become an object of cognition(N): I am “merely conscious” of the spontaneity that is the “determining in me” and this purely conceptual “consciousness of myself is by far not a cognition of myself.” Since I cannot intuit my spontaneous cognizing activity, “I cannot [cognitively] determine my ex­ist­ ence as that of a self-­active being.” According to (3), I can cognitively determine my existence as a subject of mental states, but this concerns only my existence as a passive temporal phenomenon: TCD of myself yields “not the consciousness of the determining self, but only that of the determinable self ” (B407; cf. A402). Thus, for Kant the self that can become an object of empirical cognition(N) is not 66  Kitcher 1990: 140. Accordingly, she argues (1984: 122; 2011: 270) that our faculty of cognition lacks genuine spontaneity or freedom. 67  Kitcher 1990: 83. Kitcher’s view seems to be endorsed by Grüne 2007: 246–50.

Kant ’ s Free Thinker  193 the self qua spontaneous (determining) cognitive agent.68 He sees a vicious circularity in the attempt to cognize(N) as phenomenal states or processes the necessary a priori conditions for cognizing(N) any phenomenal states or processes: conditions such as the unity of apperception and the spontaneous acts of thought that unite given intuitions in one consciousness (A402; B422).69 In light of these complications, it would be desirable to have a reading that satisfies two desiderata which seem to pull in different directions. First, we should respect Kant’s idea that in acts of TCD we exercise a non-­empirical capacity for spontaneously effecting our actual cognitive states. Secondly, we should also respect Kant’s point that our only non-­empirical capacity for causally determining our actuality is the spontaneity of will that we exercise in acts of PCD. I suggest that we can satisfy both desiderata if we ascribe to Kant the following view: while the non-­empirical spontaneity involved in acts of TCD produces cognitive mental states as “inner determinations,” it lacks a genuine or “true” causality that we exhibit only in acts of PCD. Kant expresses this view repeatedly, especially in the Dialectic: “human reason shows true causality [wahrhafte Kausalität] . . . where ideas become efficient causes (of actions and their objects), namely in morality” (A317/B374). Through its practical ideas, “pure reason even has causality, to actually produce what its concept contains” (A328/B386). Immediately after claiming that in pure apperception we cognize(W) our capacity for spontaneous “actions and inner determinations,” Kant adds that we ascribe to ourselves a rational “causality” only once we consider “the imperatives that we propose as rules to our powers of execution in everything practical” (A547/B575). In its practical imperatives, reason presupposes that it “could have causality in regard to appearances” and thus can “expect its ideas to have effects in experience” (A548/B576). The notion that practical “reason has causality in regard to appearances” implies a “causality of reason in the intelligible character” (A551/B579). By “appearances,” Kant here means outer, spatial appearances: namely, the external effects of morally good volitions. When I choose to act on the basis of moral imperatives, I expect that I can cause (and witness in experience) physical effects such as repaying my loans, helping others in need, or arriving on campus at the time I promised. Since Kant designates only 68  Ameriks 2000a: 289 also stresses this point. 69  Kitcher’s response to these passages is that Kant is confused here; in his considered view the subject of cognitive synthesis must be the empirical self that cognizes(N) its own phenomenal cognitive activity (1984: 122–6). Her argument seems to be as follows: it is a known fact that all representations must be united in one I that thinks; since this is a known fact, “the doctrine of apperception must present a phenomenal aspect of the self ” (1984: 123). This argument might work if Kant were claiming that all representations must, as a matter of known empirical necessity, be united in one consciousness. But for Kant the (objective) unity of apperception signifies an a priori epistemic condition, thus a normative rather than an empirical-­causal sense of “must.” It is a condition that all my representations must meet in order to qualify as empirical knowledge, but not a condition that all my representations automatically meet as a matter of empirical law (hence Kant’s appeal to a merely empirical-­subjective unity of consciousness that lacks objective cognitive validity, e.g., at B139–40; cf. Wuerth 2014: 27).

194  Kant on Freedom and Rational Agency this capacity for PCD as a “true causality,” he must hold that the capacity for TCD lacks true causality because it is not an efficient cause of what happens in the (empirically) real outside world apart from our mere inner representations of that world.70 Unlike practical reason, our spontaneous theoretical intellect does “not actually produce what its concept contains.” This distinction between the spontaneity of thought in TCD and the spontaneity of will in PCD is also reflected in Kant’s definitions of the relevant faculties. He defines the understanding or faculty of cognition as “the capacity to produce representations” (A51/B75) whereas he defines the will or faculty of desire as the “capacity to . . . produce objects corresponding to representations” (KpV, 5:15), as “the causality of the power of representation in respect of the actuality of objects” (V-­Th/Baum, 28:1275; cf. Bix–­x), or as the power “to actually produce what its concept contains” (A328/B386). For example, TCD merely gives me cognition(N) of an existent table whereas via PCD I can bring a table into existence. This contrast plays a central in Kant’s system. He relies on it, for instance, when he argues that a deduction of the moral law is impossible because the moral law is conceived as an a priori causal ground of objects whereas a deduction of a priori theoretical principles is possible because these are a priori principles for cognizing objects that already exist and that are given to reason from elsewhere (KpV, 5:46–7; cf. Br., 10:131). The claim that only PCD qualifies as a “true causality” is congenial to our common conviction that agents show causal efficacy only if they can realize their purposes by producing changes in the outside world (GMS, 4:394). But Kant’s view rests on two further considerations. First, it is significant that Kant defines TCD as a capacity to produce mere representations as opposed to “objects corresponding to representations.” Although every representation can be called an object insofar as our consciousness is directed towards it (A189/B234), Kant typically reserves the notion of an object (when he is concerned with sensible phenomena) for the public items of our shared cognition(N) or experience, as opposed to mere representations that are private to the mind of individual subjects (A104–5; A191/B236). As Peter Strawson argues, “objects, in the weighty sense” are “essentially spatial.”71 This is because only spatial phenomena are truly public objects of intersubjective cognition(N) and have a lasting existence that our ever-­fleeting representations lack 70  One might object that outer objects, as spatial appearances, are themselves mere inner representations. This objection rests on a literal reading of Kant’s claim that outer appearances are representations, which leads to a phenomenalist reading of Kant’s idealism. But as I argued earlier (see Chapter I, n.50), Kant distinguishes between a transcendental and an empirical sense of expressions like “representation,” “appearance,” “in us,” or “outside us” (see, e.g., A30/B45; A372–3). Only the empirical sense of “representation” or “in us” entails that something exists literally as a non-­permanent, fleeting mental state of some particular thinking subject. Permanent physical things are “representations” or “in us” in a different transcendental sense. 71  Strawson 1966: 88 (cf. 24).

Kant ’ s Free Thinker  195 (see, e.g., Bxli): they persist through changes including the continuous changes in our ever-­successive representational states. Our capacity for TCD can produce cognitive representations (cognitions(N)) of weighty spatial objects, but it cannot produce the weighty spatial objects of these representations.72 We can also conceive (though not cognize(N)) non-­phenomenal, hence non-­ spatial things as objects in a weighty sense. Significantly, Kant relies on the contrast between mere representation and genuine objecthood here as well: when he considers the idea of God that purports to represent a divine being which is “outside” us in a non-­spatial (transcendental) sense, he stresses that this idea is a “mere representation” as opposed to a corresponding object (A583/B611), “a  mere self-­creation of . . . thinking” as opposed to an “actual entity” (A584/ B612). The theological idea of God does “not signify the objective relation of an actual object to other things, but only that of an idea to concepts” (A579/B607). Since “mere representations” (including ideas and concepts) lack the status of metaphysically “weighty” (phenomenal or intelligible) objecthood, it follows accordingly that our epistemic capacity to produce mere representations lacks “true” causality in a metaphysically weighty sense.73 Second, for Kant the fact that our theoretical intellect is merely a capacity to produce representations carries additional metaphysical weight because it marks a decisive contrast between our mind and an infinite, divine understanding. Our theoretical intellect can produce only cognitive representations that depend on the prior existence of their objects. By contrast, an infinite intellect produces the objects of its representations just by representing these objects (B72; B138–9). Consequently, for a divine intellect there is no distinction between its capacities for theoretical cognition and practical agency (KU, 5:403–4; V-­Met/K2, 28:803; V-­Th/Pöl, 28:1054–5; V-­Th/Baum, 28:1272). But for a finite mind like ours, a

72  One might object that due to the interaction between mind and brain, the fact that our power for TCD effects representations entails that it also causes corresponding spatial brain states. However, this requires that there be laws governing efficient mind-­body causation. Kant denies this: he accepts empirical mind-­body dualism (A379). He further endorses a regulative principle of cognizing mental phenomena of mere inner sense (namely, the rule to view the empirical thinking self as if it were a mental substance) to ensure that “empirical laws of corporeal appearances, which are of an entirely different species, will not be mixed up in the explanation of what belongs merely to inner sense” (A683/B711). To be sure, for Kant soul and body do interact and “the soul affects the brain . . . by thinking” (V-­Met-­L1, 28:259). But interaction via mere affection is not the same as law-­governed causal determination or production. For instance, Kant reportedly says that “nerves will be affected . . . by thinking; for all of our thoughts are accompanied by bodily motions” (V-­Met-­K2, 28:757). This suggests that the reason why bodily nerves are affected by thoughts is that these thoughts are accompanied by bodily motions—­thus, what causes changes in the nerves are not the thoughts themselves but their accompanying bodily motions. 73  Kant ascribes even “subjective causality” (causality with respect to mere representations) not to our faculty of cognition but, instead, to our faculty of feeling which strives to produce and preserve the mental state of pleasure (V-­Met-­K2, 28:741, 815; EEKU, 20:206; KU, 5:220) to give it a more than fleeting existence. By contrast, the capacity for TCD does not aim to produce or preserve mental states: our cognitive activity is intentionally directed towards the objects of mental states, but as objects to-­be-­understood rather than as objects to-­be-­produced.

196  Kant on Freedom and Rational Agency practical faculty of desire that is irreducibly distinct from its theoretical faculty of cognition is the only power to produce objects in a weighty sense as opposed to mere representations of existent objects (Br., 10:130).74 These considerations should go some way towards validating my proposal that the distinction between TCD and PCD concerns a difference between a merely cognitive and a causal capacity for self-­determination. Our capacity for theoretical cognition is not a “true” causal power because subjects have genuine causality only if they possess the ability to produce genuine objects: namely, the external objects of their representations. This also yields a final confirmation that there is no conflict between (4) and (5). When Kant claims (in (4)) that “our actuality is determinable” only through the capacity for PCD, he seeks to emphasize that only our faculty of will allows us to produce actual (weighty) objects of our (normative-­ practical) representations, whereas the “sheer spontaneity” of thought in TCD gives us merely (descriptive-­ theoretical) representations of actual (weighty) objects. Moreover, whereas the spontaneity of will in PCD can legislate and execute its purposes independently of empirical desires that result from our affection by external objects, the spontaneity of thought in TCD can achieve its cognitive telos only by conceptualizing empirical intuitions that we receive from already existent (actual) objects.

VI.5.  Kant’s Commitment to Transcendental Freedom of Thought In this section I confirm that for Kant our theoretical intellect is an absolutely spontaneous faculty which exhibits transcendental freedom of thought. I have argued that the spontaneity of thought that we exercise in acts of theoretical-­cognitive self-­determination (TCD) is not a “true” causal power. For some, the fact that our faculty of theoretical cognition lacks “true causality” entails that this faculty has only a “relative” spontaneity (spontaneitas secundum quid): according to Allison and Kitcher, Kant’s denial that the epistemic spontaneity of thought involves absolute spontaneity simply follows from his claim (at A547/B575) that only practical imperatives disclose a free causality of human reason.75 On their view, the causal capacity to produce the external objects of

74  The fact that PCD and TCD involve essentially different types of spontaneous determination undermines interpretations that model the spontaneity of understanding on the spontaneity of will (see Heimsoeth 1956; Martin 1969: 207, 212). Likewise, readings where our theoretical categories “in a sense cause their objects” so that our theoretical intellect is “a securalized version” of the divine intellect (Forster 2010: 42) cannot be squared with Kant’s view that our theoretical intellect differs from a divine mind precisely because it lacks (true) causality with respect to the things it cognizes. 75  Allison 1990: 63; Kitcher 2011: 246.

Kant ’ s Free Thinker  197 one’s representations is a necessary condition for absolute spontaneity or transcendental freedom. However, Kant defines transcendental freedom in more general terms as a spontaneity to initiate new states independently of foreign causes (A533/B561). Our theoretical intellect satisfies this definition if it can perform causally unconditioned cognitive actions that effect new representational states, even if these actions lack the external effects that Kant deems necessary for ascribing true causality to a spontaneous faculty.76 Thus, Kant’s claim at A547/B575 that only our awareness of practical imperatives reveals a noumenal causality of reason need not entail that our theoretical intellect lacks transcendental freedom. Instead, Kant might seek to stress here that our absolutely spontaneous faculty of theoretical cognition differs from our absolutely spontaneous will in that it lacks true causality. When Kant considers freedom as an ability to initiate a series of outer (physical) events, he is already concerned with a specifically practical sense of transcendental freedom that involves the will (A534/B562). The passages I cited at the beginning of this chapter strongly support a reading on which Kant accepts that our theoretical intellect satisfies the general definition of transcendental freedom as a spontaneous capacity to act independently of foreign causes: we are “completely free” in our cognitive acts of judgment (Prol, 4:290); “the understanding . . . is free and pure self-­activity which is determined through nothing other than itself ” (Refl., 18:182–3). These quotes date back to the early 1780s, but in the 1787 B-­edition Critique Kant still designates our cognitive spontaneity as the “self-­activity” of a noumenal “intelligence” (B157–9; B278). Moreover, in the 1790 Critique of Judgment Kant ascribes “autonomy” not just to practical reason but also to the theoretical understanding (EEKU, 20:225, 241). Kant’s appeal to the autonomous self-­activity of the understanding decisively undermines the thinking mechanism-­reading. To see why, consider how Sellars tries to support this reading: “[F]or practical reason to be autonomous, there must be a practical premise which is . . . intrinsic to reason . . . . For, surely, if all its premises come from without, then it is indeed ‘set in motion’ from without—­its

76  Kitcher (2011: 169–70) argues that a transcendentally free cognizer would have to see “his judgment . . . as a new beginning” and thus “would not see it as dependent on his other representations . . . would not through the exercise of his cognitive faculties come to understand his states as belonging to the unity of apperception.” But a judgment that p (as a “new” representational state) can cognitively depend upon a unity of other representations without being causally determined by these representations. A free judgment is based upon the subject’s spontaneous act of taking other representations to support this judgment. Free acts of judging in light of other representations typically involve some leeway (because reflection on empirical representations typically does not rationally necessitate a particular cognitive response, or because rationally imperfect thinkers may misapprehend how certain representations support a certain proposition). But we can also sometimes be rationally determined by our own intellectual faculties to freely accept a judgment as conclusively warranted by other representations: as we saw (in Chapter III), for Kant freedom of (even practical) judgment allows for inevitable rational determination.

198  Kant on Freedom and Rational Agency ‘causality’ is caused; its ‘spontaneity’ relative.”77 On Sellars’s view, the “premises” of the understanding may come “from without” and therefore the understanding may lack absolute spontaneity and autonomy: “[W]e can conceive Kant to argue that although we are conscious of ourselves as spontaneous in the synthesizing of empirical objects, this spontaneity is still only a relative spontaneity, a spontaneity ‘set in motion’ by ‘foreign causes.’ ”78 Sellars correctly stresses that the absolute spontaneity of practical reason requires that practical reasoning be guided by an autonomous, self-­given “premise” concerning what we ought to do: namely, by the moral law which has an a priori source in the will qua pure, universal practical reason rather than (say) in our contingent empirical nature or in God (cf. Chapters  III–­IV). But Sellars goes wrong in denying that Kant holds a similar view about the epistemic spontaneity of the understanding. Earlier (in Section VI.1) I mentioned that in Kant’s view our specific empirical judgments are governed by fundamental a priori “premises” that concern the general constitution of nature and that (accordingly) require us to think about nature in certain basic terms. These “premises” include the law that for any event there must be a determining cause: this law is presupposed by all empirical reflection aimed at specific causal judgments. When Kant invokes “[the autonomy] of the understanding, in view of theoretical laws” (EEKU, 20:225) he refers to the spontaneous legislation of understanding as the original source of cognitive laws such as the general causal principle (KU, 5:186; A127). By “spontaneous legislation” he means self-­legislation that is unconditioned by external foreign causes: as we saw, he stresses that the understanding here engages in intellectual “self-­activity.” Moreover, he explicitly denies that theoretical categories like “causality” and the corresponding cognitive principles of the understanding (like the general law of causality) could be regarded as objectively valid if the categories were merely “subjective dispositions of thought, implanted in us” by external causes (B167). (I consider why he denies this in Chapter VII.) Hence, the distinction between a spontaneity that has true causality and a spontaneity that lacks true causality is orthogonal to the distinction between absolute and relative spontaneity. The faculty of desire that belongs to a squirrel has “true causality” because it produces external effects (such as digging up nuts), but this causality involves a merely relative spontaneity because it lacks autonomy. The inner states (e.g., instincts) that proximately determine the squirrel’s behavior are outside the squirrel’s control; they are determined by foreign causes such as the evolutionary processes which fix the squirrel’s genetic make-­up. Conversely, while our faculty of cognition lacks true causality it does have 77  Sellars 1970: 26. 78  Sellars 1970: 23. Compare Allison 1990: 63, 218, 228; Allison 2020: 321–3; Chignell 2007b: 43; Friedman 1996: 438; Henrich 1975: 65–7; Kitcher 1990: 140.

Kant ’ s Free Thinker  199 autonomy and (thus) absolute freedom. Our theoretical intellect can determine its cognitive activity in accordance with cognitive laws that arise from its own spontaneous self-­activity rather than from external foreign sources. These autonomous laws include not only the constitutive laws of the understanding (such as the general causal principle) but also the regulative laws of theoretical reason that prescribe how “we ought to study nature” and that arise from the “legislation” (A700/B728) of our empirically unconditioned theoretical reason. These regulative laws give a general orientation to our cognitive activity by providing us with fundamental cognitive aims or “interests” (KpV, 5:120): for instance, they set the systematic unity of our disparate cognitions(N) as “the goal of the actions of the understanding” (A644/B672) and thereby direct us to pursue “the highest purpose of the speculative use of reason” (Prol, 4:350). I now consider some further reasons for denying that in Kant’s considered view our faculty of theoretical cognition exhibits transcendental freedom of thought. Allison and Dieter Henrich reject this claim based on a note (Reflexion 5442; Refl., 18:183) where Kant seemingly distinguishes between “logical freedom in actions of reason” and transcendental freedom. For Allison, this note even indicates “the key to [Kant’s] change of mind” about whether our theoretical intellect is transcendentally free.79 However, an alternative reading of this note is that Kant here seeks to deny only that absolute epistemic or “logical” freedom involves the true causality that pertains to a transcendentally free will. Moreover, it seems questionable to use a sketchy private note to support a reading that conflicts with the abovementioned passages from Kant’s published writings.80 This particular note is especially untrustworthy as evidence for Kant’s considered view because Kant here defines transcendental freedom as the “complete contingency of actions.” This definition contradicts his considered view that there is no strict link between transcendental freedom per se and contingency (cf. Chapters III–­IV). Another common motive for viewing the spontaneity of cognition as merely relative or conditioned derives from the fact that our capacity for TCD depends on empirical data that we receive from without.81 This may also be a further reason for Sellars’s claim that the spontaneity we show “in the synthesizing of empirical

79  Allison 1990: 62; cf. Henrich 1973, 1975. In his recent work Allison initially shows greater reluctance about using this private note as a basis for controversial interpretive claims (2020: 209). But he still claims that Kant eventually came to reject his earlier (alleged) conflation of “a merely logical with a transcendental freedom” (2020: 232; cf. 303, 320)—without offering any support for this claim apart from Refl. 5442. 80  Allison himself urges caution about drawing controversial interpretive conclusions from mere Reflexionen (2020: 233). 81  See Hanna 2009: 109; Keller 1998: 158–9. For Keller (and Sgarbi 2012: 48–9), the understanding has absolute spontaneity in a priori logical thinking but then shows only relative spontaneity in empirical thinking, when it operates under conditions of sensible affection. This has the odd implication that the understanding shows a more elevated form of spontaneity when it fails to realize its cognitive telos (to obtain determinate cognition(N)).

200  Kant on Freedom and Rational Agency objects . . . is . . . only a relative spontaneity, a spontaneity ‘set in motion’ by ‘foreign causes.’ ”82 Earlier (in Section VI.3) I conceded that we cannot exercise our capacity for TCD in a purely a priori-­intellectual fashion apart from the (pure or empirical) conditions of sensibility. However, our noumenal capacity for thought is merely affected by these conditions (e.g., by empirical intuitions). As we saw, Kant places a lot of weight on the difference between mere sensible-­empirical affection and sensible-­empirical necessitation or causal determination (cf. Chapters  III–­IV; A534/B562; MS, 6:213). The mere fact that a capacity is affected by sensible conditions cannot entail that this capacity lacks absolute spontaneity, since our absolutely free faculty of choice is essentially affected by sensible inclinations that yield the “matter” for our free choices. Both our higher practical and epistemic powers are “set in motion” by foreign non-­ determining, merely “occasional causes” when we are affected from without with “impressions of the senses” that “provide the . . . occasion” (A86/B118) for the free exercise of these powers. Even when we exercise our will to causally determine our external behavior apart from sensible motives, on the basis of a priori moral reasons, our practical self-­ awareness is affected both by empirical thoughts (e.g., thoughts about how to use our physical capacities) and by the pull of empirical desires that impose a sense of constraint on our morally pure choices (cf. Chapter V). Moreover, some (perhaps most or all) of our absolutely free choices are not based solely, or at all, upon pure moral reasons: we make many morally permissible and impermissible choices for the sake of satisfying the sensible desires that affect our free will. Even morally wrong choices that entirely prioritize such desires over a priori moral reasons are still imputable and thus transcendentally free in the negative sense that rules out causal determination by empirical desires (see Chapter III, n. 33). Likewise, as I argued in Section VI.1, our misguided (e.g., careless, prejudiced) cognitive responses to given empirical (perceptual, testimonial, etc.) data are transcendentally free in the negative sense that rules out determination by perceptual states or psychological habits. If our capacity to judge were not merely affected but strictly determined by the sensible data we receive from foreign causes, then it would indeed exhibit a merely relative form of spontaneity. But Kant explicitly denies that acts of TCD are so determined when he stresses that the input of sensibility leaves us “completely free” to consider what our cognitive responses to given perceptions ought to be (Prol, 4:290). For instance, we must freely make up our mind as to whether a given empirical intuition (that sensibly “affects” our mind) warrants a judgment that p or rather calls for suspension of judgment. This mental act is free not only in the negative sense that rules out necessitation by sensible conditions such as

82  Sellars 1970: 23.

Kant ’ s Free Thinker  201 perceptions, psychological habits, etc., but also in the striking positive sense that designates our reflective self-­control over our judgments and assents (cf. Section VI.1). Most significantly, our reflective epistemic self-­control is governed by our a priori representations of cognitive laws (such as the law of causality) and purposes (such as the purpose to bring systematic unity to our disparate empirical cognitions) that are spontaneously legislated by our autonomous theoretical intellect (as comprising the higher faculties of understanding, theoretical reason, and reflective judgment). Hence, in our acts of theoretical (suspension of) judgment and assent we exercise our capacity for autonomous, spontaneous cognitive self-­determination. I concede to Sellars that the legislative powers of the understanding are inferior to, or less dignified than, the legislative powers of the will qua pure practical reason. As we saw (Chapter  III), the moral self-­legislation of practical reason is entirely pure and (not just undetermined but even) unaffected by sensibility. Accordingly, the prescriptive content of the moral law is entirely independent of sensibility; this ensures that the moral law applies to all rational beings including the highest divine intelligence. By contrast, the contents of the cognitive laws of our understanding refer to the specific forms of our human sensibility. Therefore, these schematized laws do not apply to all cognizing beings. Even in their unschematized version, the pure categories or forms of synthesis that (partly) ground these laws do not apply to a divine (non-­synthesizing) intelligence (cf. Chapter III). There is a corresponding functional difference between the pure ideas of reason and the pure concepts of the understanding: the understanding “cannot produce by its activity any other concepts than those which serve to bring the sensible representations under rules,” whereas “reason shows such a pure spontaneity in the case of ideas that it far transcends anything that sensibility can give to consciousness” (GMS, 4:452). Hence, the legislative spontaneity of universal practical reason has a higher degree of purity (and a correspondingly greater scope of validity) than the legislative spontaneity of our understanding. But this does not entail that the theoretical legislation of the understanding lacks absolute spontaneity. Even though the theoretical application of concepts like “causality” requires empirical intuitions “that supply the material” (B423) for object-­directed thoughts (for cognitions(N)), the categories and their corresponding principles arise from our spontaneous intellectual self-­activity that is undetermined by empirical laws or other foreign causes. These concepts have a non-­empirical, pure origin (which Kant views as a condition for their objective validity; cf. B166–7): “pure synthesis . . . gives us the pure concept of the understanding” (A78/B104). This synthesis, “inasmuch as . . . [it] is an expression of spontaneity,” does not stand under empirical laws (e.g., of imagination) (B151–2). Thus, “the empirical is only the condition of the application, or of the employment, of the pure intellectual faculty” (B423). Likewise, even though the (schematized) content of the understanding’s

202  Kant on Freedom and Rational Agency theoretical laws refers to sensible conditions, this does not entail that the understanding is determined by these conditions (or by other conditions outside its own intellectual self-­control) to legislate its own laws. Here we must distinguish two different concepts of “dependency” or “condition.” First, there is a non-­causal notion according to which the validity and application of some principle epistemically depends on sensible conditions. Dependency upon sensibility in this epistemic sense is fully compatible with absolute spontaneity because epistemic dependency does not imply determination by foreign causes. To illustrate: the specific moral laws of freedom that Kant considers in the Metaphysics of Morals arise from the absolutely spontaneous self-­legislation of practical reason even though their content depends upon empirical-­anthropological conditions which are far more contingent than the purely formal conditions of sensibility that epistemically restrict the laws of understanding. Second, there is the causal sense of being conditioned by sensible-­empirical objects that undermines any kind of transcendental freedom or absolute spontaneity. Our theoretical intellect is conditioned by or dependent upon sensibility in the first, epistemic but not in the second, causal sense. Hence, it possesses the absolute, causally unconditioned spontaneity that is required for transcendental freedom. I have sometimes encountered (in discussion) the claim that our understanding cannot be transcendentally free since it cannot pick any different laws other than its specific constitutive principles. But this claim overlooks the fact that Kantian freedom does not essentially involve the liberty of indifference: our transcendentally free pure practical reason is not at liberty to legislate any highest principle other than its constitutive moral law either (cf. Chapter III).

VI.6.  Conclusion: Two Species of Transcendental Freedom My argument in this chapter shows that there are some deep parallels between freedom of thought and freedom of will. First, both kinds of freedom involve, negatively speaking, the absence of determination by sensible-­empirical causes. Consequently, sensibility is not a source of either practical or theoretical failure: the senses are responsible neither for theoretical error nor for evil choices.83 83  Kant’s point that “it is not the fault of the [sensible] appearances . . . if our cognition takes illusion for truth” (Prol, 4:290) has a clear practical analogue: “[T]he source of . . . badness cannot . . . be placed in the sensibility of man and the natural inclinations springing therefrom” (RGV, 6:38; cf. RGV, 6:21). However, this is only an analogy. Pace Henrich (1954–5: 29–34), the claim that both practical and theoretical errors are imputable because they result from the free exercise of our higher faculties does not imply that both types of errors call for the same reactions. If theoretical error is typically a result of mere epistemic carelessness, it need not involve any disrespect for the dignity of persons; thus the rich vocabulary of moral blame which tracks moral character flaws does not apply here. Kitcher (2011: 247–8) suggests that all epistemic accountability ultimately reduces to moral responsibility. I agree that we can be morally accountable for epistemic errors when such errors impair our moral agency;

Kant ’ s Free Thinker  203 Second, both freedom of thought and freedom of will involve the positive capacity to act in accordance with our awareness of objective reasons that derive (partly but essentially) from foundational a priori, self-­legislated, hence autonomous laws: “freedom in thinking,” like freedom in willing, “signifies the subjection of reason to no laws except those which it gives itself ” (WDO, 8:145). Finally, in both cases our positive ability to act on the basis of objectively right reasons is accompanied by a privative propensity to choose or think incorrectly. These parallels ground Kant’s claim that cognizers of nature must presuppose their freedom of thought “in the same way” as moral agents must assume their free will (RezS, 8:14). Despite acknowledging these parallels, my reading also shows that freedom of will and freedom of thought are two specifically different kinds of transcendental freedom. As we saw over the course of this chapter there are some crucial differences between free volition and free thinking. Most significantly, only our will involves a (“true”) causality of freedom. Moreover, our free will or practical reason is less dependent on sensibility than our free theoretical intellect (though, as we saw, this point requires important qualifications). Finally, our free thinking typically directly produces just those judgments (and assents) that are licensed by our (perhaps mistaken) consciousness of what our theoretical reasons are, of how we ought to judge. By contrast, (partly) due to the “frailty” inherent in our free Willkür, our consciousness of how we morally ought to choose does not directly lead to good choices but leaves us the option to choose wrongly. I have shown that in Kant’s (considered) view we possess transcendental freedom of thought. But my account leaves us with two important questions. First, why does Kant insist that we possess a freedom of thought that is incompatible with determination by foreign causes—­what grounds his incompatibilism about cognitive spontaneity? Second, can we justifiably believe (in the mode of either opinion, faith, or even knowledge) that we indeed have such freedom—­and if so, what entitles us to this belief? I will take up these questions in Chapter VII.

this is due to the indirect sense in which theoretical judgments stand under the control of the will (JL, 9:74). (To illustrate: I may negligently choose not to perform physical acts that would give me morally relevant theoretical information, e.g., when I fail to check the bank schedule to see when the bank opens so that I can pay my loan on time.) But Kant also leaves plenty of room for imputable epistemic errors, e.g., beliefs based on perceptual illusions, that reflect sui generis doxastic vices (e.g., lack of epistemic caution) rather than underlying moral flaws. Accordingly, for Kant the kind of personality required for imputability is not simply identical to moral personality (this is the joint implication of MS, 6:223 and Anth, 7:127).

VII Freedom of Thought as a Condition of Theoretical Cognition In Chapter VI I argued that in Kant’s view our theoretical intellect has transcendental freedom of thought. In this chapter I address two central questions about this view. First, why does Kant insist that our epistemic spontaneity must be absolute, i.e., that our intellectual freedom is incompatible with determination by foreign causes? Second, how does Kant seek to justify the belief in absolute freedom of thought, and what kind of strength or certainty does this justification achieve?

VII.1.  Kant’s Epistemic Incompatibilism: Preliminary Considerations and Complications In considering the grounds of Kant’s epistemic incompatibilism, the first question is whether these grounds are similar to those supporting his practical incompatibilism. As we saw, one reason why Kant rejects a compatibilist-­ naturalistic account of free will concerns our executive freedom of choice: our choices are governed by categorical moral imperatives only if we are free to choose between the metaphysically real options to comply with or to violate moral norms, options that we would lack if all causality were “mere nature” (cf. Chapter V). Might there be a parallel incompatibilist argument with regards to our executive freedom of thought? By “executive freedom of thought,” I mean our capacity to freely make particular theoretical judgments and assents or to freely adopt general epistemic maxims that govern particular assents (cf. Chapter VI). An incompatibilist argument regarding executive freedom of thought that runs parallel to Kant’s incompatibilist argument regarding executive freedom of choice would have to draw on two considerations. First, it would have to draw on the idea (1) that finitely rational cognizers are always in a position to judge correctly and to form rationally appropriate doxastic states (where this includes suspending judgment, forming opinions whose subjective strength is appropriately weak, etc.). Second, this argument would have to draw on the further idea (2) that finitely rational cognizers always have the (privative) option to judge incorrectly because their cognitive capacities are inherently susceptible to the influence of factors such as perceptual and

Kant on Freedom and Rational Agency. Markus Kohl, Oxford University Press. © Markus Kohl 2023. DOI: 10.1093/oso/9780198873143.003.0008

206  Kant on Freedom and Rational Agency transcendental illusion, wishful thinking, biases and prejudices, lack of epistemic care and caution, etc. In Chapter VI, I suggested that the conjunction of (1) and (2) shows that the Sellarsian thinking mechanism-­view is implausible because a mechanism would always be determined either to function or to malfunction so that either (1) or (2) would always be false. However, there are two reasons why I shall not utilize this line of argument to expound the grounds of Kant’s epistemic incompatibilism. First, there is no clear textual basis for assuming that this argument underlies Kant’s incompatibilism about epistemic freedom. There are many textual clues suggesting that Kant relies on the analogous argument for grounding his incompatibilism about moral freedom of choice (cf. Chapter V): his repeated appeal to the “Moral Ought Implies Can” principle, his frequent emphasis that for finite agents the compliance with moral norms is only ever contingent, and his claim that the imperfectly rational freedom of Willkür is “the ground of the possibility of categorical imperatives.” There are no passages indicating the analogous view that we can always go either way with regards to objective epistemic oughts. Moreover, Kant has good philosophical reasons for not relying on this line of argument when he tries to justify the significance of absolute epistemic spontaneity. Let us consider here the two aspects that jointly constitute this argument, (1) and (2). Consider first (1), Ought Implies Can. Kant does accept this principle with regard to human beings qua theoretical cognizers who are governed by epistemic oughts (RezS, 8:14). However, one main reason why he stresses the importance of this principle in the practical domain does not apply with the same force in the theoretical domain. The assumption that every agent has the ability to respond correctly to moral norms underlies our practice of moral imputation, blame, and punishment. In Chapter VI I argued that for Kant the subject of theoretical cognition is capable of imputation as well; we must regard ourselves as accountable to other rational thinkers in our theoretical judgments since these judgments lay claim to intersubjective, universal validity. But ideas such as imputation, responsibility, and personality seem to be of major interest to Kant only in the context of his moral philosophy (cf. A365–6). The epistemic significance of these ideas is limited because in many standard cases of mistaken theoretical judgments it does not seem very pertinent to hold subjects accountable for their errors; e.g., it would be rather strange for someone to blame me for my mistaken belief that my cat has the orange quality I see displayed in my visual field. In Kant’s view many theoretical errors are the result of epistemic carelessness and a corresponding lack of reflection; although this carelessness is avoidable and imputable, such imputations do not play the same central role in our lives as moral imputations. There are complications here. For one, we have a moral duty to adopt informed, carefully reflected theoretical opinions about morally relevant empirical facts: e.g., I am morally on the hook for knowing when the bank opens so that I can

Freedom of Thought as a Condition of Theoretical Cognition  207 repay my loan on time. Moreover, some epistemic vices are ultimately based on moral vices of characters: for instance, the arrogant refusal to consider other thinkers’ reasons for their beliefs, racist or misogynist biases, perhaps also the lazy complacency that Kant laments in his Enlightenment essay. In such cases, the will’s immoral qualities indirectly (“secretly”) influence the understanding to adopt mistaken theoretical stances. This shows that there are important connections between moral character and epistemic maxims, but it does not show that epistemic accountability considered in its own right, apart from its connection to moral accountability, is of major concern to Kant. Regarding (2), I have argued (in Chapter VI) that human cognizers are afflicted with a propensity to violate epistemic oughts. But I do not see that Kant invokes this propensity to argue that in our epistemic agency we can never be inevitably determined by reason since we must confront a real possibility of error in all our theoretical judgments. The argument (in Chapter V) that our finitely rational will always, “by its nature” (GMS, 4:413) leave us the real option to choose incorrectly does not extend to our theoretical intellect. This argument relied on an irreducible contrast between desire-­based and reason-­based motivation and on the additional premise that we are always subject to the motivational influences of desires that impel us to choose on their behalf. There is no analogous contrast in the epistemic case. One can make a good case for the idea that we are typically subject to factors (perceptual illusions, etc.) that account for a real possibility of epistemic error. But I see no basis for the stronger idea that we are necessarily, in every norm-­governed theoretical judgment, afflicted with real obstacles to right judgment that preclude our inevitable determination by reason. For example: if I know that skunks are mammals and I come across a particular skunk, I may well be rationally compelled to judge that this animal is a mammal. One reason why Kant puts so much emphasis on the pervasive real-­practical possibility of deviating from right moral reason in favor of our sensible nature is that this possibility looms so large in his a priori moral psychology. For Kant both our awareness of the right moral reasons and our awareness of options to choose against these reasons are central to our reflective moral self-­consciousness: the temptations posed by sensible desires have a deep impact on our reflective moral choices, and the sense of constraint that accompanies even morally good choices is a consciously felt constraint. These factors are absent from our epistemic self-­ consciousness. This is in part because we cannot judge akratically, contrary to our (perhaps mistaken) conception of how we ought to judge (cf. Chapter VI). Moreover, in moral deliberation our reflective focus is on ourselves: our values, goals, and imperfections. By contrast, in theoretical reflection our focus is oftentimes on objects other than the self (skunks, cats, numbers, protons, . . .). Hence, an awareness and certainly any felt awareness of our subjective rational imperfection is typically overshadowed by our reflective attention to the external objects of theoretical judgment.

208  Kant on Freedom and Rational Agency For these reasons, I do not want to assimilate Kant’s case for incompatibilism about freedom of thought to his grounds for incompatibilism about executive freedom of choice. However, I see a strict parallel between Kant’s main argument for epistemic incompatibilism and his argument for incompatibilism about legislative freedom of will (qua practical reason). I already intimated this parallelism in Chapter IV: when considering the passages which highlight Kant’s concerns about the naturalistic idea that our law-­giving practical reason is empirically determined, I stressed that in many of these passages Kant is focused not just on practical reason but on reason as such, in a broad sense that encompasses all our “higher” intellectual faculties including the understanding. Admittedly, Kant’s main argumentative point regarding the absolute freedom of understanding is far from obvious. This point is missed, for instance, by a common interpretation of his denial in the Schulz review (RezS, 8:14) that we can take our theoretical judgments to be causally determined. This interpretation was first proposed by the Neo-­ Kantian Heinrich Rickert and has recently been endorsed by various commentators.1 According to this reading, Kant claims that a determinist or fatalist view of human judgment is self-­undermining because (as Wood puts it) the determinist “would be unable to represent herself or those to whom she offers the arguments [for fatalism] as holding fatalism rationally on the basis of those arguments.”2 In Rickert’s words, “the determinist is determined in his decision [for determinism] not through the value of truth, but solely through the force of causality.”3 Rosefeldt reads Kant as holding that “in judgments where one says why one has a certain conviction, one must not give a causal explanation for this conviction but must always give a justification for it, namely, reasons that speak in favor of it being correct.”4 On this reading, Kant argues: the claim that our theoretical beliefs are causally determined implies that the belief in determinism is itself causally determined, can thus be explained “through the force of causality” and therefore cannot be based on truth-­oriented reasons. This leaves Kant with a weak argument for his epistemic incompatibilism. It is unclear why being determined by the force of causality should conflict with being rationally determined. Likewise, giving a causal explanation for one’s theoretical convictions seems perfectly compatible with holding that these convictions are based on truth-­oriented justifying reasons.5 It seems that one can rationally 1  See Rickert 1921: 302. Horn (2015: 143–51) endorses Rickert’s view as the proper way of understanding Kant’s argument in the Schulz review and in Groundwork III § 2. Wood (1999: 177–80) does not mention Rickert but his understanding of Kant’s argument is very similar to Rickert’s and Horn’s; see also Collins (1999: 199). 2  Wood 1999: 177. 3  Rickert 1921: 302, cited after Pothast 1980: 253. The translation is mine. 4  Rosefeldt 2000: 178. The translation is mine. 5 Wood immediately notes these problems (1999: 179). Likewise, when Rosefeldt considers whether Kant’s argument is defensible (2000: 182–4), he ends up weakening Kant’s position so that it

Freedom of Thought as a Condition of Theoretical Cognition  209 endorse determinism by viewing one’s belief that determinism is true as determined by the rational considerations that speak in favor of determinism. Naturalists might hold that our theoretical beliefs are (often) based on truth-­ oriented reasons precisely because they are (often) causally determined by the natural states that these beliefs refer to: paradigmatically, by the perceptual states that we receive in our causal interaction with nature and that evidentially determine our beliefs about nature. A view along these lines has been endorsed by some prominent recent philosophers. They do (or may) accept Kant’s concern that determinism threatens our practical self-­image as deliberating agents. But they deny that determinism poses a similar threat to our self-­image as theoretical cognizers whose beliefs must conform to the natural world. This conformity, they suggest, is plausibly understood through the idea that the natural world determines our beliefs by conveying evidence to us. For Thomas Nagel, “in forming beliefs we may hope for nothing more than to be determined by the truth.”6 For David Wiggins “the libertarian ought to be content to allow the world . . . to convey to the free man who has questions . . . and is suitably placed to answer them, how the world is”;7 “The world must not dictate how a man should deliberate to change the world, but . . . the world must dictate how the world is.”8 As Roger White puts it, since “action and belief are importantly different,” “causal determination of belief is not worrisome as such”; indeed, the determination that occurs when I look at a computer and end up believing that there is a computer in front of me is “just what we want for our beliefs.”9 Kant accepts that the strict determination of theoretical cognition “is not worrisome as such.” For Kant divine freedom consists in the fact that divine cognition (and action) is inevitably determined by reason (cf. Chapter III). Hence, he is not in principle opposed to the idea that causal determination can be the same as free rational determination “by the truth” or by objectively valid reasons. Moreover, as I suggested earlier, Kant does not deny the possibility of inevitable causal-­rational determination for finite cognizers either: he does not rule that we may

becomes fully compatible with a naturalistic-­deterministic worldview. By contrast, Horn objects to Kant that even though epistemic subjects must take their judgments to be based on objective reasons, this might be an illusion that can be explained in terms of naturalistic processes (2015: 150–1). But this alleged objection concedes the very point of Kant’s argument. If the assumption that our theoretical judgments are determined by naturalistic processes did entail that our awareness of objective ep­i­ ste­mic reasons is illusory, then naturalists could not consider themselves rationally entitled to theoretical beliefs that are determined by naturalistic processes. Allison comments that in the Schulz review Kant brings up “the familiar trope that the skeptic refutes himself when, in arguing for his position, he necessarily assumes the objective validity of his own argument” (2020: 302). But he does not explain why this “trope” applies to Schulz or why Schulz should be considered a “skeptic” about theoretical reason. The crucial issue is precisely: why should someone (like Schulz) who adopts a theoretical belief in absolute determinism be committed to skepticism about the objective validity of theoretical cognition? 6  Nagel 1989: 116. 7  Wiggins 2003: 96. 8  Wiggins 1969: 143. 9  White 2010.

210  Kant on Freedom and Rational Agency occasionally, albeit not typically, be rationally compelled by our appreciation of conclusive evidence to make correct theoretical judgments. So the question remains: why does Kant deny that we can rationally endorse the claim that our theoretical beliefs about the natural world, including the belief in natural determinism, are causally determined by the natural world (“determined by the truth”)? In addressing this question, I shall argue first (in Section VII.2) that the appeal to determination by the truth is problematic even for beliefs involving empirical concepts and perceptual evidence. This discussion will lead (in Section VII.3) to the more central case of a priori judgments involving pure concepts—­judgments that express necessary cognitive laws and thereby exhibit our legislative freedom of thought.

VII.2.  Empirical Belief and “Determination by the Truth” In Chapter VI we already encountered the idea that empirical belief-­formation is a passive matter since it is responsive to the flow of information that causally runs from states of the world to the subject’s mind: as a believer “I simply have to lay myself open to the world in order to let the phenomena put their print upon me.”10 White illustrates his claim that evidential-­causal determination “is just what we want for our beliefs” with a case where a computer in front of me causes me to have perceptions which determine my belief that there is a computer in front of me. Here it seems plausible to say that I am caused to have a belief with a certain content (that p) by the external state of affairs that this belief represents as obtaining (p). One might suggest that belief-­acquisition on the basis of (veridical) perception is the paradigmatic case of acquiring true, justified beliefs about the world. If in this paradigmatic case we have no trouble understanding how being naturally determined to believe that p equals being afforded with good reason to believe that p, we might then extend the notion of external, rational determination by the truth from perceptual belief to other kinds of cognition. As Bill Brewer puts it: A person’s direct perception of the way things are in the world around him provides a paradigm case of the integration of causation and rationalization constitutive of compulsion in thought by reason, which is essential, quite generally, if the capacities . . . for language use, deductive reasoning, and now perception, are to make a genuinely cognitive impact on their subject, in the growth of knowledge and understanding.11

10  Wiggins 1969: 145. Similarly, Galen Strawson claims that “there is no action at all in reasoning and judging . . . action, in thinking . . . is waiting, seeing if anything happens, waiting for content to come to mind, for the ‘natural causality of reason’ to operate in one” (2003: 232). 11  Brewer 1995: 17.

Freedom of Thought as a Condition of Theoretical Cognition  211 As we saw in Chapter VI, Kant denies that this yields an adequate model for conceiving rational belief-­formation. He sees a reflective gap between the passive intake of perceptual states and the active formation of belief. This gap exists because subjects must take themselves to possess sufficient objective grounds for believing that p and must thus endorse their perceptual states as providing such grounds. This renders their assents public, communicable states that place them in (potential) dialogue with other thinkers from whom they can demand agreement and to whom they are, in turn, accountable for providing epistemic reasons and for considering rational criticisms. Kant’s view that our empirical beliefs are not causally-­rationally determined by passively received perceptual states also rests on his denial (discussed in Chapter VI) that perceptions adequately represent the physical world. When I believe that there is a computer in front of me, I also represent the computer as being a certain way: as square, silver, vibrating, etc. For Kant, we need to be epistemically careful here not to immediately assent to such perceptual appearances because they do not accurately portray physical things as they really (empirically speaking) are in all respects. For instance, my computer does not have the very quality of silver that I see displayed in my visual field. The fact that our naïve awareness of sensory states does not disclose how things really are in the external world points to a more fundamental problem with the idea that our empirical beliefs are perceptually and rationally determined by their intentional external objects. Such beliefs involve empirical concepts that allow us to represent things as being a certain way. The notion of determination by the truth may suggest that the external world determines us, by causing us to have certain sensory experiences, to form and use the correct set of empirical concepts. But this is an instance of what Sellars calls the myth of the given. Being causally affected with sensory data does not yield conceptual classifications: such classifications aim to tell us what different particulars have in common, which requires that we reflectively compare and combine various sensory experiences. This reflective activity must be carried out by the cognizing subject: “combination does not lie in the objects, and cannot be borrowed from them” (B134). Moreover, the sensory data that we receive from without underdetermine our cognitive commitment to one among many possible conceptual schemes.12 As Ralph Walker puts it, with respect to any conceptual classification “there is nothing in the perception that compels us to make this classification rather than some other.”13 If the causal impact that external objects have on our sense organs underdetermines which conceptual schemes we employ in our beliefs about 12  To clarify: I do not mean to suggest here that the data we receive from without can inform our conceptual classification while lacking in any conceptual content. For Kant, given data cannot ra­tion­ ally inform empirical concept-­formation unless they are conceptually determined via certain a priori concepts. 13  Walker 1978: 128–9.

212  Kant on Freedom and Rational Agency nature, then we are not determined to form empirical beliefs by the very objects that our beliefs are about. Instead, it must be something about us that leads us to classify objects in one way rather than another. But then we have already left behind the idea that the natural conditions which cause us to conceptualize the empirical world are (solely) external objects and states of affairs. It rather seems that the natural determination of our conceptualizations and (thus) our empirical beliefs occurs (partly, but significantly) from within. On a naturalistic view of the human mind, this may invite the Humean idea that we are led to our empirical conceptualizations by the sheer non-­rational force of psychological “habit.” Here it is instructive to consider Nelson Goodman’s influential neo-­Humean discussion of empirical concepts. Goodman points out that the bare content of our sensory experiences supports both the judgment that emeralds are green and the judgment that emerubies are grue.14 (“Grue” applies to all things examined before time t just in case they are green but to other things just in case they are blue; “emeruby” applies to emeralds examined for color before t and to rubies not examined before t.) His Humean sympathies come out in his claim that the only relevant difference between a classificatory scheme that employs “grue”/“emeruby” and our classificatory scheme is that our concepts are actually “entrenched” as a  matter of habit. This implies that our commitment to our actual conceptual scheme is not based on our intellectual “compulsion” by objectively valid reasons.15 The same implication follows from the idea that our adoption of particular classificatory schemes depends on our given pragmatic interests.16 On these views, our use of specific sets of empirical concepts depends on our contingent natural dispositions. If the laws determining our neurological endowment or our pragmatic interests had turned out differently, we might—­in response to the same sensory data that inform our actual “green/emerald” scheme—­have judged that emerubies are grue. Thus, how we conceptually carve up the world depends on (i.e., is explained by) subjective facts about us rather than on objective facts about the external things that our concepts refer to.17 14  Goodman 1955: 74–5, 103. 15 Goodman might deny this implication. He claims that the appeal to (degrees of) habitual entrenchment can legitimize the conceptual schemes that we in fact use: “The reason why only the right predicates happen so luckily to have become well entrenched is just that the well entrenched predicates have thereby become the right ones” (1955: 98). This attempt to derive legitimacy from actual entrenchment (“ought” from “is”) seems highly puzzling, especially since Goodman concedes that the initial entrenchment and “projections” of predicates are not by themselves reasonable but become so only through increasing degrees of actual projection and habitual entrenchment (1955: 90). To see why this view is problematic, consider what it implies for a community where predicates such as “witch,” “flat earth,” or “phlogiston” are and continue to be very well entrenched. 16  Both de Sousa 1984 and Hacking 1991 emphasize that conceptual classification is relative to pragmatic interests. Hacking, in particular, stresses the practical dimension of carving up the world (1991: 113). However, he seems to think that this point applies only to natural kind concepts rather than to the (in his view) cognitively more robust concepts of advanced, e.g., physical science (1991: 115). See n. 17 below. 17  The challenge deriving from Goodman’s Humeanism cannot be answered simply by insisting (with Lewis 1983) that certain ways of conceptually carving up the world are intrinsically more

Freedom of Thought as a Condition of Theoretical Cognition  213 This problem afflicting the naturalistic determination by the truth picture is exacerbated by a further complication regarding our rational belief-­formation. In Chapter VI, I noted that the idea that our beliefs are “determined by the truth” suggests an atomistic construal of the evidential relation between a particular sensory experience and a particular empirical belief. This relates to what Quine calls the (second) dogma of empiricism: “to each . . . statement, there is associated a unique range of possible sensory events such that the occurrence of any of them would add to the likelihood of the truth of the statement.”18 Such atomistic dogmas are implicit in the claim that there is a tight causal-­evidential connection between (i) a particular worldly fact; (ii) a particular sensory experience; and (iii) a particular belief that represents this fact as obtaining, so that the belief is both causally and rationally (evidentially) necessitated by the specific sensory impact that the external world makes upon the subject. This conception clashes with the view that empirical cognition has a holistic structure. Jane Heals puts the basic intuition that is common to different forms of holism (including Kant’s) as follows: The central factor underlying it [holism] is the potential complex interconnectedness of things, both causally and evidentially. The world may present itself to us much of the time as more or less isolated subsystems, further features of which can be inferred on the basis of information about only the current or preceding states of that subsystem. But each such subsystem is, we believe, embedded in a wider spatio-­temporal framework which may impinge on it or provide clues about it.19

According to holism, the evidential significance of a particular sensory experience depends on its place within an extensive framework of further experiences, concepts, and laws. Thus, the passive element in perceptual awareness—­namely, the reception of sensory input—­cannot rationally determine us to make empirical assents: rational belief formation requires that we actively reflect on how given sensory data relate to our comprehensive body of empirical knowledge, e.g., whether some unusual perceptual episode should be taken to disconfirm a generalization that we take (in the assent-­mode of opining or hypothesizing) to have a law-­like status, or whether we should rather dismiss this perception since the relevant generalization plays a central (e.g., explanatory or unifying) role in our cognitive framework of interconnected empirical laws and concepts. objective or “natural” than others, i.e., that they enable us to “carve nature at its joints” and thus show up in the best (“final”) physical theory. The basic issue here remains: which properties we designate as the “perfectly natural” ones, or what we regard as the “best” (“final”) physical theory, depends not on the objective physical world but on our subjective dispositions. 18  Quine 1951: 38. 19  Heal 1996: 80; compare the helpful summary of Kant’s holism in Friedman 1999: 99.

214  Kant on Freedom and Rational Agency This holistic picture further supports Kant’s claim that there is a reflective gap between our reception of sensory data from external objects and our theoretical belief-­formation. Accordingly, it further diminishes the plausibility of the naturalistic claim that our beliefs are determined by just those external natural objects that our beliefs seek to cognize: the assumption that all our mental processes are naturally determined implies that our patterns of holistic reflection and the beliefs we form on this basis are (partly) determined by natural causes that operate within us, namely, by our contingent neuropsychological constitution. The considerations that I have adduced in this section show that we cannot easily defuse worries about the notion that our empirical beliefs are determined by natural causes via the idea that these beliefs are determined by the perceptual impact that external nature makes upon our mind. This impact does not obviate the need for active contributions of the epistemic subject to belief formation. If this active contribution is the upshot of natural causes that operate internally, we face serious questions about the objective rationality of our empirical beliefs. In what way can our actual cognitive responses to perceptual data be deemed objectively, rationally superior to our (perhaps vastly) divergent cognitive responses to the same data in alternative scenarios where our brains and minds have been wired differently by the chain of natural causality? Perhaps our rational entitlement to use specific empirical concepts and to endorse specific patterns of holistic empirical reflection is secured by certain a priori concepts that have a privileged epistemic standing in our cognitive efforts because they are indispensable to any rational process of organizing sensory data into an objectively valid belief-­system. The privileged epistemic role of a priori concepts includes their capacity to render sensory intake conceptually articulate so that it can rationally justify our empirical concept-­formation.20 Such priori concepts include chiefly the representations of self (“I think”), substance, causality, and the systematic unity or simplicity of nature (alternatively, the unity and simplicity of theories about nature). But this suggestion leads us right into the heart of the question of whether naturalists can take their theoretical beliefs to be based upon objective reasons: on a naturalistic worldview, are we rationally entitled to use concepts like “causality,” “systematic unity,” or “simplicity”?21 In Section VII.3, I argue that this central question of seventeenth-­century philosophy retains its relevance in a contemporary setting. If we understand the 20  This suggestion raises thorny issues about Kant’s views on empirical concept-­formation (for helpful discussion, see Grüne 2007), which I bracket for the purposes of my discussion here. 21  Loewer (2004: 186) argues that our rational entitlement to prefer our conceptual scheme over Goodman’s “gruesome” scheme rests on the appeal to simplicity: we can rationally adopt our system of beliefs over a “gruesome” system because “it seems likely that the gruesome system will have to be a bit more complicated to equal an ungruesome system in informativeness.” The question is, however: can we regard (our acceptance of) criteria such as simplicity as rationally grounded? Or must we rather concede that our commitment to such criteria is just another subjective, contingent, rationally speaking arbitrary preference or projective habit?

Freedom of Thought as a Condition of Theoretical Cognition  215 ep­i­ste­mic concerns that prompt this question, we can charitably understand Kant’s incompatibilism about freedom of thought.

VII.3.  Natural Determination and Causal Belief For Kant the rational construction of a system of empirical cognitions from given sensory data requires that there be objective a priori epistemic concepts and principles that govern our mental activity. The a priori concept of “causality” is especially relevant here because it plays a fundamental role both in our cognitive efforts to understand the natural world and in Kant’s denial that we can regard our theoretical intellect as determined by foreign causes. Kant’s discussion of causality is crucially influenced by Hume, who famously argues that our beliefs in causal relations are not based on reason because the representation of causal necessity cannot be derived from the content of sense-­ perceptions. Kant accepts the latter point and generalizes it to all the other a priori categories: “. . . no one will say that a category, such as that of causality, can be intuited through sense” (A137–8/B 176–7); no empirical (or pure: cf. A 732–3/B 760–1) intuition can exhibit a necessary connection among two different events or states of affairs.22 I will now suggest that the resulting problem for our causal thinking, which leads Hume to conclude that our causal thoughts have no ra­ tional basis, is still a live issue in contemporary philosophy. Few philosophers have challenged Hume’s and Kant’s basic premise that the content of our perceptions underdetermines the modally thick content of causal judgments. Those who purport to disagree with Hume about the unperceivability of causality typically give a deflationary account of causation that dispenses with the idea of necessity.23 A notable exception is Evan Fales. He argues that production, involving necessity, is an object of direct perception. Fales focuses on the experiences we have when “our bodies are pushing and pulling, or being pushed or pulled by, other bodies,” which gives us a “feeling of pressure” and a “sensation of force.”24 But the fact that we have sensations with a certain phenomenal content or “feel” is, by itself, a merely psychological occurrence which does not indicate the existence of natural powers, forces, or necessary causal connections. My 22  See Guyer 1987: 164–6. 23  This tendency is exemplified by Anscombe 1993 and Ducasse 1993. Anscombe considers causal efficacy a primitive feature that we directly observe in in instances of “cutting . . . drinking . . . purring” (1993: 93). At the same time, she repudiates “the equation of causality with necessitation” (1993: 90). But if we leave out the idea of causal necessitation, all that remains is the fact that (e.g.) the butter parts after (or when) the knife touches it. If Anscombe insists that we see the causal efficacy of the knife’s activity, the burden is on her to explain what causal efficacy minus some thick modal connection between cause and effect amounts to. Similar problems can be raised for Ducasse (see especially 1993: 127, 131). 24  Fales 1990: 12–16.

216  Kant on Freedom and Rational Agency subjective feelings of pushing and being pushed cannot justify (e.g.) my belief that if a strong person pushes a plastic chair then (other things equal) the chair must move as a matter of objective causal necessity. Fales responds that if Humeans are unwilling to accept necessary cause-­effect relations, they are not entitled to believe in other kinds of non-­logical necessities either.25 But committed Humeans are prepared to deny that we can justifiably believe in non-­logical necessary connections. In any case, this partners in crime move fails to show that there really are any necessary causal connections in nature that we could directly perceive via subjective feelings of pressure. A more popular way to defend the claim that our causal beliefs can be justified on the basis of sense-­perception is the regularity theory, which holds that a causal connection obtains when (as Hume says) “an object [is] followed by another, and where all the objects similar to the first, are followed by objects similar to the second.”26 This account—­ or its contemporary variations27—makes causation unmysterious because the frequent, regular occurrence of one event after another seems to be directly observable (though Kant denies this; cf. Section VII.5). However, the conception of causation as mere regularity faces several problems. It arguably cannot distinguish causal from coincidental regularities or genuine causes from inefficacious epiphenomenal by-­products of a causal process.28 It is unclear how it can deal with cases where cause and effect occur simultaneously, or with cases where a cause produces an effect only on a singular occasion. Moreover, the mere fact that A-­type events are regularly followed by B-­type events fails to explain and thus does not allow us to comprehend the occurrence of B-­events.29 Assuming a tight link between causation and (causal) explanation, this means that the regularity account cannot yield a satisfying concept of causation.30 These problems for the modally deflationary regularity view suggest that our causal thought is committed to modal properties such as necessitation. One might respond that an interest in this property wanes if we focus on ­indeterministic, probabilistic rather than deterministic causation.31 However, accepting the possibility of indeterministic causation hardly shows that the idea 25  Fales 1990: 25. 26 Hume, An Enquiry Concerning Human Understanding, Par. 7. 27  See, e.g., Ayer 1963 and Baumgartner 2008. 28  See Lewis 1973a for a good account of these and similar problems. Baumgartner tries to defend the regularity analysis against these worries. He appeals (2008: 5–7) to the idea that (e.g.) singing a song can be eliminated from the conjunction of factors that regularly precede the lighting of a match because it is not part of a “minimally sufficient” conjunction of factors. But the notion of minimally sufficient factors involves modal content and thus undermines the reduction of causal efficacy to mere regularity. 29  See Dretske 1977 and Stroud 2011: 26–9. 30  Similar issues arise for the neo-­Humean view that causal laws merely describe observable matters of fact, as axiomatizations of the “list” of facts that have happened and do happen in the world (see Beebee 2000). It is hard to see how descriptions of facts can serve to explain those very facts (see Lange 2013). Humeans have responses to these worries, such as the appeal to counterfactual analyses of causation; but see n. 34. 31  See Salmon 1984.

Freedom of Thought as a Condition of Theoretical Cognition  217 of physical necessity is irrelevant to our ordinary or scientific thought. For one, causal laws themselves retain their necessity even if these laws are statistical or probabilistic. Moreover, regarding the relation between a cause and its effect, the indeterministic implications of quantum physics do not seem to tell against the notion that if someone chops off my head I must die, or the notion that procreation among two mice cannot produce a dog.32 Finally, an appeal to probabilistic causal relations faces an analogue of the problem Hume raises about the perceivability of necessary connections, because the concept of an objective probabilistic relation between cause and effect implies a real connection that cannot simply be read off from given perceptual data; in particular, it cannot be analyzed in terms of observed regularities.33 The contemporary literature on causation is so vast and fragmented that I cannot profitably discuss all views that try to reduce causality to more basic features. However, generally speaking such attempts fail to preempt Hume’s problem because they all accept that our causal thinking involves the representation of some objective yet unobservable modally thick connection between cause and effect. This diagnosis applies to counterfactual theories;34 to views of causation as “process-­linkage”;35 and to views which purport to agree with Humean skepticism about causal necessity but then fall back on modally thick representations after all.36 32 For the view that deterministic relations are prevalent at the macroscopic level, see Baumgartner 2008. 33  This is because indeterministic causes are not regularly followed by their effects. Note also that an indeterministic view leaves room for necessary causal relations: causes are sufficient conditions for a higher probability that certain effects occur (Schaffer 2001: 76). A cause (absent other factors) raises or at least changes the objective chance of the effect occurring and this changing/raising is naturally necessary. One might try to reduce these notions to talk about mere frequencies, but this would raise problems analogous to those which afflict the regularity view. Some might seek to analyze causal probabilities subjectively as degrees of belief, but this would undermine the idea that our causal thoughts refer to objective connections. 34  Counterfactual theories in the vein of Lewis (1973b) analyze causal claims in terms of subjunctive conditionals. These are modal claims whose truth-­conditions cannot be drawn from sense experience. For instance, the Lewisian assessment of counterfactuals requires the modal notion of “possible worlds,” where some of these worlds are “closer” to the actual world than others. The epistemic credentials of such talk are surely no less mysterious than the credentials of notions like causal necessity or power that worried Hume. Fales (1990: 3) aptly diagnoses that “talk of physically possible worlds, and the like, only defers the day of reckoning.” 35  On this view, causation is rooted in the existence of mark-­transmitting processes from cause to effect. It is controversial whether such views are committed to de re modality. Salmon rejects what he calls the “modal conception” of causal explanation and suggests (1984: 202–3) that causal processes “carry” probability distributions for different interactions. But once again, the epistemic burden for the idea that objective probabilities are “carried” (or that “marks” are “transmitted”) equals the burden afflicting beliefs in de re modal connections. Moreover, Salmon’s attempt to distinguish causal from pseudo-­causal processes invokes (1984: 147–8) the de re modal notion of what “characteristics” a proc­ess would manifest over time in the absence of certain conditions. 36  Woodward (1990: 558–9) applauds Humean skepticism and claims that “scientific discussion typically says nothing about metaphysical relations of necessitation.” But his account of causation (i) centers on counterfactual claims about what would happen in case of counterfactual interventions and (ii) appeals to what is physically impossible (2003: 71). If Humean epistemology puts a ban on causal

218  Kant on Freedom and Rational Agency There is an influential anti-­Humean view that espouses a non-­reductive realism about causal relations involving nomic necessity.37 Its core claim is the idea that a causal law is a relation between universals: the law that all Fs are G is grounded in the fact that the universal F stands in a relation of nomic necessitation to the universal G.  However, this account answers Humean doubts about causal necessity merely by fiat.38 Epistemic worries about how we can know that nomic necessitation relations exist in nature are only aggravated by shifting the issue to relations among abstract entities (universals). Moreover, this shift invites a new challenge: “why should some state of affairs in the lofty realm of universals . . .  translate into hard facts below, facts about earthly particulars that fall under those universals?”39 My discussion suggests that Hume’s problem about causality does not simply go away in a contemporary setting. When we assume a causal relation among natural objects, the modal content of our causal belief is strongly underdetermined by the perceptual data that we receive from the physical world: hence, we are not perceptually “determined by the truth” here. Moreover, the concept of causality is so fundamental to our cognition that an eliminativist stance in the vein of Russell is not a real option either.40 Barry Loewer puts the issue as follows: On the one hand, causality seems to be so intertwined with so many of our concepts (indeed, with the concept concept) that if it fails to refer then most of our thoughts would also fail to refer. On the other hand, look as hard as one might, we just don’t find causal relations among the fundamental properties of physics or in the dynamical laws of physics. So we would be in a dilemma of either rejecting our conceptual scheme or rejecting physicalism, at least in its austere formulation.41

necessity, it is hard to see how the modal concepts employed in (i) and (ii) survive this ban. Moreover, as Woodward admits, the concept of “intervention” is itself a causal notion; as such it cannot provide a complete reductive analysis of causal relations. 37  For variants of this view see Armstrong 1983; Dretske 1977; Tooley 1987. 38  See Sider (1992: 262) for the point that Tooley’s method for identifying the necessitation relation is a “solution by stipulation”: Tooley postulates some properties that the necessitation relation has by definition and then invokes the Ramsey-­Lewis account of defining theoretical terms (Tooley 1987: 78–80). Tooley and Sosa claim (1993: 4–5) that Hume’s “crucial handicap” was that he lacked a proper method for analyzing theoretical terms that refer to unobservable entities. But the problem Hume raises is an epistemic rather than a semantic issue. The Ramsey-­Lewis method for defining theoretical terms cannot show that we have any legitimate epistemic grounds for believing in unobservable nomic necessitation relations. For Armstrong (1983: 55), the rational basis for the belief that it is a law that all Fs are G is that all observed Fs are G plus “further observational circumstances.” He does not indicate what those circumstances are, and it is obscure how mere observation could support the ex­ist­ence of genuine law-­likeness over and beyond mere regularities. Armstrong might respond that we know that necessary laws exist via an inference to the best explanation. See n. 52 for discussion. 39  Sider 1992: 261. For an early version of this challenge, see van Fraasen 1989. 40  See Russell 1992. 41  Loewer 2004: 199–200.

Freedom of Thought as a Condition of Theoretical Cognition  219 Loewer may yet be underappreciating the centrality of causality, for the appeal to fundamental properties or dynamical laws of physics itself seems to presuppose the legitimacy of modal and causal notions (unless one accepts, with Russell, that there can be laws without causes).42 If Hume’s problem persists, so does the central question that he raises after posing this problem: since our causal beliefs are evidentially underdetermined, what determines us to believe in causal relations? Hume’s answer is this: since causation is not a property that we can detect through any sensory information we receive from without, our concept of causality is forced upon us from the inside by psychological mechanisms (the principles of custom or habit). We cannot help but believing that events are causally connected because it is in our given psychological nature to believe this. Since the belief in causal necessity (or its probabilistic counterpart) is a result of sub-­personal mental processes over which we have no rational control, this belief is itself non-­ rational: “when the mind . . . passes from the idea or impression of one object to the idea or impression of another, it is not determined by reason, but by certain principles, which associate together the idea of these objects, and unite them in the imagination”;43 our causal beliefs are due to natural processes that “are a species of natural instinct, which no reasoning or process of . . . the understanding is able either to produce or prevent.”44 One can devise similar Humean accounts for other concepts that seem central to our cognitive orientation and whose content cannot be derived from sense-­ perceptions, such as the ideas of systematic unity or simplicity.45 These considerations provide the background for charitably understanding Kant’s epistemic incompatibilism. As we saw, commentators tend to read Kant as making the sweeping claim that causal determinism as such rules out the possibility of truth-­oriented or reason-­based judgment per se. Now, however, we can see that Kant is primarily concerned with certain a priori components of theoretical judgment that he deems indispensable to our entire theoretical cognition: namely, the concept of causal necessity, other categorial features such as 42  It seems plausible that causal relations are among the fundamental properties of physics. For the view that the dynamical laws of classical (or even quantum) physics refer to causal relations, see Frisch 2005. 43  Hume 2009: Book 1, Part 3, Section 6. 44  Hume 1993: Section V, Part 1. 45  Simplicity and unity are of special relevance here because these ideas are invoked by some who try to distinguish between laws and accidental regularities without appeal to nomic necessitation. They hold that “a regularity is a law iff it is a theorem of the best system” where the best system is “one that strikes as good a balance as truth will allow between simplicity and strength” (Lewis 1994: 478; cf. Loewer 2004: 189). Since the assessment of what “strikes a good balance” between competing criteria is subject-­dependent, this account suggests that what we regard as a law ultimately depends in part on our subjective, contingent natural psychology (for Loewer, the very property of being a law is “partly constituted by psychological factors”; Loewer 2004: 191; cf. Lewis 1994: 479). But then concepts such as “simplicity” or systematic unity’ call for the same diagnosis as “causality”: we must view them (and our tendency to employ them in our empirical inquiry) as mere upshots of contingent, non-­rational psychological dispositions (“customs”).

220  Kant on Freedom and Rational Agency substantiality or community, or regulative ideas such as systematic unity, homogeneity, or simplicity. For Kant, these a priori concepts are fundamental to all our theoretical representations including our empirical concepts and judgments because they ensure that sensory data can be conceptualized and combined in an objectively rational, universally valid fashion. In the Schulz review, Kant makes essentially the same point about theoretical reason that also informs his anti-­naturalism about practical reason (cf. Chapter IV): an a priori judgment that purports to express an objective, timeless rational necessity cannot result from empirical causes that are contingent, mutable, sub-­ personal, and thus altogether insensitive to rational considerations. Conversely, if a judgment is determined by such causes, then it cannot make a valid claim to an a priori, rational-­objective necessity (cf. the passages (P1–P5) cited in Chapter IV). This point affects the epistemic self-­awareness of theoretical cognizers because everyone who seeks to reach objectively valid theoretical judgments must presuppose that “the understanding has the capacity to determine its judgment according to objective reasons that are valid at any time, and does not stand under the mechanism of merely subjectively determining causes that may change subsequently” (RezS, 8:14). Again, Kant’s concern here lies primarily with a priori cognition (which must be “valid at any time”). It extends to empirical cognition because all valid empirical judgments must be formed “according to objective reasons” that are partly but essentially based upon a priori cognitive sources such as categorial principles (e.g., “every event follows from some cause”) or regulative rules (e.g., “study nature as a unified system of properties and laws”). Thus, Kant’s argument in the Schulz review is crucially informed by Hume’s attempt to “derive” putatively a priori concepts from empirical conditioning proc­ esses, “from a subjective necessity.” Such an “empirical derivation . . . cannot be reconciled with . . . a priori cognition” (B127–8). For instance, the a priori concept of causality . . . . . . . so manifestly contains the concept a necessary connection with an effect and of the strict universality of the rule, that the concept would be altogether lost if we attempted to derive it, as Hume has done, from a repeated association of that which happens with that which precedes, and from a custom of connecting representations, a custom originating in this repeated association, and constituting therefore a merely subjective necessity.  (B5)

The term “subjective necessity” refers to a habitual-­instinctual compulsion of our mind due to some underlying sub-­personal, intellectually blind empirical mech­ an­ism. If our causal judgments were determined by such mechanisms, they would depend on the way in which our mind happens to be wired by the course of nature: we would represent causal necessities just because of the contingent way in which we are naturally disposed to react to given perceptual input. If our

Freedom of Thought as a Condition of Theoretical Cognition  221 empirically given psychology or gene pool (etc.) changed “subsequently” we would no longer believe in causal necessities. Moreover, in this scenario our firm conviction that necessary causal connections exist in the natural world is likewise a product of cognitively blind mechanisms that operate below the threshold of consciousness and whose non-­rational influence on our beliefs we fail to notice: here the merely subjective psychological necessity which induces in us the firm belief in causal connections “comes mistakenly to be regarded” by us “as objective” (B127). In sum, I read Kant as endorsing the following argument: (1) As thinking subjects who aim at objectively valid judgments about the natural world, we must take our judgments to be based on objective (theoretical) reasons. (2) These objective reasons stem partly but essentially from our consciousness of a priori laws, such as the general law of causality, which yield an indispensable cognitive framework for objectively valid theoretical judgments. (3) A priori cognitive laws that express an objective rational necessity cannot result from contingent, mutable natural mechanisms which are “merely subjective” since they completely bypass our conscious rational awareness and our (“higher”) intellectual faculties. We would not be rationally entitled to accept these cognitive laws as true or objectively valid if all our mental activity stood under “the mechanism of merely subjectively determining causes that may change subsequently.” In order for such laws to have objective a priori validity, they must be grounded in our empirically unconditioned spontaneity.46 “Without this original and immutable spontaneity, we would cognize nothing a priori; because we would be determined to everything, and our thoughts themselves would stand under empirical laws” (Refl., 18:182–3). To save the objective validity of our causal thought we must regard the principle that natural events have determining causes as deriving from our intellectual self-­activity, as a “self-­thought first a priori principle of cognition” (B167). This allows us to deny that we are non-­rationally compelled to believe in causal necessity by contingent, sub-­personal dispositions (“habits,” “customs”). Thus, the concept of causality (and the corresponding causal principle) “must either be grounded completely a priori in the understanding, or must be entirely given up as mere figment of the brain” (A91/B123). Earlier in this section I expressed doubts about the Neo-­Humean proposal that the representation of causal necessitation is irrelevant to an enlightened empiricist model of causal reasoning. But even if this proposal were defensible, it would not be available to the hard determinist view that Kant critiques in the Schulz review. For this view rests on the conviction “that . . . everything stands under the strict law of necessity” (RezS, 8:11). If determinists like Schulz concede that their 46  Why in our spontaneity rather than in some other spontaneous source such as God’s intellect? I consider this question in Section VII.6. For present purposes, remember that Kant’s main dialectical opponent here is a dogmatic naturalist who would deny the existence of God.

222  Kant on Freedom and Rational Agency firm endorsement of the “strict law of necessity” is a mere product of contingent mental habits which bypass their rational capacities, they compromise their ra­tional grounds for believing in natural necessity. Hence, one cannot rationally endorse an unrestricted (as opposed to merely phenomenal) natural determinism about all human actions including those cognitive actions through which we form, accept, and apply the concept of causal necessity.

VII.4.  Etiology, Rationality, and Evolution On my interpretation, Kant argues that the non-­empirical, purely spontaneous intellectual origin of our epistemically basic a priori concepts is required for the objective validity of these concepts. Hence, Kant sees a tight link between the origin and the validity or justification of certain concepts or beliefs. This invites the objection that etiological considerations regarding the origin of our concepts or beliefs are irrelevant to their epistemological properties. In this section I examine some facets of this objection. (This supplements my earlier discussion of the supposed genetic fallacy in Sections IV.1 and IV.5.) Here I focus on a recent discussion by Roger White who considers and rejects various principles that might be endorsed by those who think that etiology has important epistemological implications.47 One of these principles (call it P) seems reminiscent of Kant’s (and also Hume’s) view: “if one is justified in believing that the ultimate explanation of my belief makes no reference to whether p, then my belief that p is not justified.” White rejects P since in his view we are justified in beliefs about future facts and necessary truths even though the relevant future or necessary facts play no role in explaining our present contingent beliefs. Surely proponents of P should not require, absurdly, that our beliefs in future truths must be explained by reference to not yet actualized future facts. But they can plausibly demand that our beliefs about the future derive from our awareness of past or present facts that rationalize such beliefs, e.g., from our perception of causal powers that determine what happens in the future: or else we lose our rational entitlement to beliefs about the future. After all, Hume argues that our beliefs about the future are “not based on reason” precisely because in forming these beliefs we are not determined by our awareness of considerations that ra­tion­ally support them, such as a perceptual awareness of causal powers or a non-­circular inferential process. Rather, we are determined to adopt these beliefs by sub-­ personal associative customs that bypass our intellectual powers.48 47  White 2010. 48  One might suggest, in the vein of Goldman’s causal theory of knowledge (1967: 364), that we know future facts if our belief and the fact believed have a common cause, e.g., volcanic activity as the cause of a belief in, and the fact of, future eruption. However, citing volcanic activity as the cause of this belief is insufficient to justify it: a perception of such activity may set in motion a rationally flawed

Freedom of Thought as a Condition of Theoretical Cognition  223 Likewise, White’s presupposition that we have rational beliefs in necessary truths, whatever their origin may be, begs the very question that animates Hume and Kant. They regard the issue of whether we possess epistemic justification for (non-­trivial) modal beliefs as one of the most pressing philosophical problems. Consequently, they take etiological skepticism about such beliefs very seriously. It is dialectically inappropriate to dismiss P by taking for granted our epistemic justification for the very types of beliefs that proponents of P consider most problematic. Conversely, Kant’s (and Hume’s) etiological worries may seem irrelevant to philosophers who feel that they can simply take the objectively rational character of their modal beliefs and so-­called intuitions for granted—­e.g., philosophers who pursue the modal metaphysics inspired by Lewis and (remotely) by Leibniz. As Robert Adams puts it: It was characteristic of much early modern philosophy, and especially the critical philosophy of Kant, to assign a dominant role to epistemology. Of the great early modern philosophers . . . Leibniz was probably the least preoccupied with epistemology. He was typically willing to begin an argument with whatever seemed true to him and might seem true to his audience, without worrying too much about whether epistemology would present it as something we can really know. Much of the development of modal metaphysics in the 1970s and since has proceeded in this Leibnizian way . . . .49

Adams’s remark raises intricate methodological issues. But Kant would (indeed, did) consider the invocation of brute modal intuitions describing “whatever seems true” to rationalists and their favored audience unacceptably dogmatic: if one has been woken from one’s dogmatic slumber by Hume, the philosophical task of considering how non-­trivial necessity claims might be justified cannot be avoided. Hence, the etiological worry that such modal claims might be a mere upshot of contingent, non-­rational psychological habits cannot be dismissed as philosophically irrelevant. White imposes a further challenge on arguments that seek to draw skeptical conclusions from etiological considerations: such arguments can be dismissed if they have the radical, general implication that we do not know anything at all. I consider the merits of this challenge in relation to Kant’s view in the concluding section of this chapter. For now, I note that those who accept P can deny that P by itself yields the general skeptical implication that White deems absurd. Consider mental process that entirely fails to appreciate why the activity licenses the predictive belief. But the more fundamental problem here is that the attempt to defend the rational credentials of future beliefs by appeal to a common cause is viciously circular when the rational credentials of causal representations themselves are under dispute (as they are for Hume and Kant). 49  Adams 1994: 3.

224  Kant on Freedom and Rational Agency here White’s point that the causal determination of belief is not per se ep­i­ste­ mically worrisome because the right kind of causal determination “is just what we want for our beliefs.” (In his example, looking at a computer determines me to believe that there is a computer in front of me.) Proponents of P can accept this point: causal determination of belief per se raises no epistemic problems because it is (at least) conceivable that in some cases the causal explanation of a belief rests on those very facts that the belief seeks to apprehend. Since P acknowledges ep­i­ ste­mically harmless or even beneficial causal etiologies for belief, P does not by itself entail a fully general skepticism. The epistemic worries raised by P concern, in the first place, epistemic domains that have long seemed problematic anyway: beliefs about unobservables, about the future, or about (synthetic) necessities.50 In these domains our beliefs lack the “right” kind of causal determination that would give us just what we epistemically “want for our beliefs,” namely, an evidential determination by the appropriate external facts or objects (“by the truth)”. The more global skeptical implications that White deems absurd do not arise, in Kant, from some brute intuition that natural determination of belief is always epistemically bad. Rather, they arise from the further argument that the objective rationality our empirical belief system is epistemically grounded in those (allegedly) a priori representations (like “causality”) which are vulnerable to etiological skepticism. Note, moreover, that the reason White gives for denying that the natural determination of belief is per se problematic presupposes the link between etiology and justification that he professes to deny. White (just like Nagel, Wiggins, and others) accepts that there are certain causal influences on belief that are ep­i­ste­ mically beneficial, which implies that etiological considerations do have epistemological implications. The idea that there is a kind of causal influence that we, epistemically speaking, want for our beliefs because it can be regarded as direct evidential necessitation “by the truth” makes sense only against the backdrop of the complimentary view that there are other kinds of belief-­determination which we do not want because we consider them epistemically undermining (cf. Section IV.5). For these reasons, Kant’s argument that a naturalistic picture of the human mind is incompatible with the objective validity of our beliefs in causality (substantiality, systematic unity, simplicity . . .) cannot be dismissed as resting on a confusion of etiology and epistemology.

50  These worries also affect mathematical beliefs. I am bracketing this issue here because it raises additional complications, especially in connection with contemporary views. For example, the famous Benacerraf-­Field challenge rests on a Platonist view of mathematical facts as causally inert with regards to our mathematical beliefs. Kant is not a Platonist since for Kant mathematical facts or properties depend upon our a priori forms of sensible intuition. In his view, worries about the objective necessity of mathematical beliefs arise from the assumption that these forms are empirical rather than a priori, i.e., are part of the contingent natural order.

Freedom of Thought as a Condition of Theoretical Cognition  225 The passages cited earlier suggest that instead of P Kant endorses a more complex principle (P*): if the explanation of our belief that p makes crucial reference to factors that are incompatible with the content of that belief, then we cannot consider our belief that p rationally valid or justified.51 If all our beliefs could be explained as the upshot exclusively of contingent (“a posteriori”) processes, then we would not be rationally entitled to beliefs that purport to represent strict (a priori) necessities. Of course it is a contingent fact that we believe anything at all or, for that matter, that we exist. But Kant is concerned with the factors that lead us to adopt beliefs given that we exist as cognizing subjects. If these factors are wholly contingent, mutable, and subjective (i.e., non-­rational), then their products, namely our judgments, cannot lay claim to an objective a priori necessity. The intuitive force of P* is that a valid judgment representing a strict necessity cannot result from precisely those contingent conditions that the judgment purports to be independent from. In Chapter IV, I argued: if we had to concede that we adopt our basic moral convictions just because our practical reasoning stands under the (concealed, surreptitious) influence of contingent empirical desires, this would undermine the rational credentials of our belief that we must (ought to) act in certain ways regardless of what we happen to desire as a contingent matter of fact. This is why Kant ties the validity of our moral judgments to the empirically unconditioned legislative freedom of our practical reason. His appeal to the empirically unconditioned freedom of our theoretical intellect rests on an analogous point. Our causal beliefs represent necessary connections among natural objects that obtain regardless of how our empirical psychology happens to operate. That is, our causal beliefs about what must happen in nature as a matter of objective nomic necessity purport to be independent from subjective contingent habits of thought over which we have no rational control. Thus, if we had to concede that we adopt causal beliefs just because our theoretical reasoning stands under the (concealed, surreptitious) influence of such habits, this concession would undermine the rational credentials of our causal beliefs: it would debunk our causal representations as a “mere figment of the brain” (A91/B123). Some might deny Kant’s premise that if our causal thought were empirically determined these determining processes would be merely “subjective” or non-­ rational. One might argue that the neurological-­psychological mechanisms which make us believe in causal necessities are sensitive to objective evidence because they evolved from natural processes that do have (contrary to Kant’s premise) a cognitive-­ rational significance: evolution selected for a human cognitive psychology which

51  In reading Kant as endorsing a factual version of this principle (which concerns the factors that actually explain our beliefs rather than what we justifiably believe about those factors), I construe Kant’s view as invoking what White calls “blocking” rather than “defeating” debunkers. (But remember that Kant is not a debunker.)

226  Kant on Freedom and Rational Agency tracks objective features of our environment such as causal relations, since our ability to track these features enhances our reproductive fitness. This proposal invites the question: which mental processes enable us to track causal relations? This is just Hume’s question in new disguise. If we assume (first) that perception is the only device we have for tracking real features of external nature and (second) that the content of our perceptions underdetermines the modal content of our causal beliefs, it follows that we cannot view the natural evolution of our mind (or brain) as selecting for a cognitive mechanism which tracks real causal relations or powers. In response, one might appeal to processes of inference rather than perception: we track causal connections by first using our past experience of regularities as a basis for inferring future events and then, if these inductive expectations are continuously confirmed, for inferring the ex­ist­ ence of causal powers. This proposal faces two related worries. First, the fact that some of our inductive expectations about the future continue to be confirmed does not show that we are tracking modally thick causal connections or laws: what we are tracking (via direct observation) is only the frequent correlation of certain event-­types. Second, even if one might plausibly argue that evolution selected for cognitive mechanisms that infer future events from observed regularities, it is hard to see why evolution should also have selected for cognitive mechanisms that further (correctly) infer the existence of law-­governed causal powers. In response, one might argue as follows: (1) The capacity to deliberate successfully to achieve various ends is selected for. (2) Successful deliberation requires knowledge of various subjunctive conditionals’ truth. (3) Subjunctive conditionals are intimately connected to law-­governed causal powers. However, step (2) in this argument is dubious: the capacity for successful deliberation does not seem to require more than the tendency to infer future events from observed regularities. Animals process repeated conjunctions well enough to achieve their ends without representing subjective conditionals or law-­ governed powers. Such representations are abstract and complex, far removed from our immediate perceptual awareness and not directly tied to how our nerv­ ous system regulates its behavior. Thus, it is unclear why or how correct causal representations would have enhanced the deliberative skills or reproductive fitness of our ancestors (whose evolutionary development is, on a naturalistic view, largely responsible for our mental dispositions and belief-­forming mech­an­isms). But then we again face the worry that our disposition to infer the ex­ist­ence of law-­governed causes is rationally speaking arbitrary since it arises from natural processes that do not select for cognitive aptitude and which are therefore “blind” to objective epistemic reasons. To illustrate: if some mutation of the human brain had not occurred, we might not have become disposed to believe in lawful causal

Freedom of Thought as a Condition of Theoretical Cognition  227 necessities even though our disposition to infer future A-­B patterns from frequently observed A-­B correlations might have been left intact.52 This worry highlights a key problem for attempts to defend the rationality of our causal beliefs by appeal to evolutionary processes: if our cognitive faculties developed via natural selection, genetic drift, or random mutations, it is unclear why these faculties should have turned out to be rational or cognitively reliable. The problem here is twofold. First, even if we assume that our adaptive mental faculties or beliefs (i.e., those faculties/beliefs which do confer an advantage with regard to survival) are cognitively reliable or aim at the truth, it is unclear to what extent our mental faculties/beliefs are mostly adaptive (as opposed to, mere byproducts of processes that select for adaptive faculties/beliefs). Second, it is unclear why our adaptive mental faculties/beliefs should have turned out to be rational or cognitively reliable: it seems quite possible, perhaps even likely, that false or unreasonable beliefs confer crucial advantages for survival. This is because there are factors other than cognitive reliability (such as efficiency) which ­constrain the adaptive development of our neurological make-­up. As Patricia Churchland puts it: Boiled down to essentials, a nervous system enables the organism to succeed in the four F’s: feeding, fleeing, fighting and reproducing. The principle chore of nervous systems is to get the body parts where they should be in order that the organism may survive. . . . Truth, whatever that is, definitely takes the hindmost.53

52  Armstrong (1983: 73) claims that “the postulation of laws is a case of inference to the best explanation. The thing to be explained is the observed regularity of the world.” However, a naturalistic view of why we make certain inferences suggests that if our brain arose from a different natural history we might not be prompted to respond to the observation of regularities by inferring causal laws, and we might not endorse this as an inference to the best explanation. If our actual tendency to infer causal laws from observed regularities is the result of processes like brain mutations that are insensitive to objective epistemic reasons, these inferences cannot be regarded as rational or as yielding the objectively “best” explanation. A similar problem arises for the Putnam-­style argument that unless our causal beliefs were true the predictive success of our scientific theories would be an inexplicable miracle (see Lyons 2003 for critical discussion). Whatever may or may not strike us as “inexplicable” or “miraculous” itself depends (given a naturalistic view of our mind) upon contingent, rationally arbitrary conditions. Moreover, the claim that positing causal laws of nature yields a valid “non-­ miraculous” or “best” empirical explanation of natural facts (like the observed regularity of the world) already assumes that we can rationally believe in natural causes and their governing laws as factors which (unlike supernatural entities) do genuine explanatory work in a viable empirical account of why things happen. But it is precisely this assumption that is under dispute given a naturalistic view of how we come to believe in natural causes. To illustrate why this is a serious problem, consider how it afflicts Hume’s critique of beliefs in miracles. This critique rests on the presupposition that our beliefs in natural causes and laws are rationally-­evidentially superior to beliefs in supernatural forces. But Hume’s naturalistic picture of the non-­rational processes which lead us to believe in natural causes undermines this presupposition. 53  Churchland 1987: 548.

228  Kant on Freedom and Rational Agency Even if this “definite” diagnosis seems too strong, there are more cautious ideas in the vicinity: e.g., Darwin’s “horrid doubt” about “whether the convictions of man’s mind, which has been developed from the mind of the lower animals, are of any value or at all trustworthy”54 or Stephen Stich’s point that “there are major problems to be overcome by those who think that evolutionary considerations impose interesting limits on irrationality.”55 At the very least, we cannot place much confidence in or assign a high probability to the hypothesis that our disposition to represent causal necessities was naturally selected for its rationality or cognitive value. One might suggest that we can defend our causal representations via their adaptive benefits by invoking their practical or pragmatic value. Suppose that our chances for survival or well-­being are enhanced by our tendency to represent modally thick causal connections even if these representations do not reliably track such connections: “a fancier style of representing is advantageous so long as it is geared to the organism’s way of life and enhances the organism’s chances of survival.”56 However, an appeal to pragmatic benefits does not confer the right kind of justification upon our causal beliefs (it invokes the wrong kind of reason). The Humean-­Kantian question is whether we are epistemically justified, by the standards of theoretical reason, in our beliefs that thick causal connections exist in the natural world. The preceding discussion of the evolutionary tracking hypothesis raises a complication: are our beliefs about how human cognitive capacities evolved not themselves based upon empirical evidence for the existence of certain biological causes and laws? One might argue that our actual rational knowledge of evolutionary causes proves that our mental faculties and causal representations are adaptive and have been selected for their cognitive or rational properties. However, we cannot simply take for granted that we do have rational knowledge of evolutionary causes. If evolutionary theory suggests that our mental faculties and causal representations in general might well be cognitively unreliable, this suggestion also undermines the rational credibility of those specific causal representations that figure in our evolutionary theories. It thereby undermines the rational credibility of evolutionary theory itself (qua product of our mental faculties, whose cognitive aptitude is under dispute).57 This would mean that evolutionary theory is rationally unstable. I return to this issue in Sections VII.6–7. 54  This quote is from an 1881 letter, contained in Francis Darwin’s 1887 book (1:315–16). 55  Stich 1990: 56. 56  Churchland 1987: 548 (her emphasis). As Nietzsche puts it: “the time has finally come to replace the Kantian question ‘How are synthetic judgments a priori possible?’ with another question, ‘Why is the belief in such judgments necessary?,’ that is, to understand that for the purposes of preserving beings of our type we must believe that such judgments are true, although, of course, they could still be false judgments!” (1966: Par. 11). 57  This concern is especially salient since the cognitive value of complex, abstract scientific beliefs about unobservables (such as evolutionary causes) is much harder to connect with adaptive advantages than the cognitive value of ordinary, concrete perceptual beliefs (cf. Plantinga 1993: 232–3).

Freedom of Thought as a Condition of Theoretical Cognition  229

VII.5.  Causality, Idealism, and Naturalism In this section I consider how Kant’s epistemic incompatibilism relates to his idealism. From a Kantian perspective, the epistemic problem exposed by Hume stems from the ambition to legitimize necessary causal judgments about a mind-­ independent natural reality. For Kant this is a hopeless endeavor since we cannot know necessary (synthetic) truths about things that are entirely independent of our cognitive faculties. The only reality that we can cognize on a determinate theoretical basis depends on (“conforms to”) our representational capacities (cf. Chapters I–­II). A bit more precisely: the (sensible, phenomenal) objects of our cognition depend for their general character on the necessary general conditions under which a being endowed with our representational capacities can obtain objective theoretical cognition (experience) of objects. Among these conditions is the representation of things as causally connected in a determinate spatiotemporal order. Hence, for Kant we can defend a (restricted) version of the principle of sufficient reason by assuming that this principle governs a natural world that is structured by our cognitive capacities. On the Kantian approach, concepts such as “causality” or “substance” that enable us to experience the order and regularity of nature are not taken from nature but imposed on nature by us (A114; A125), namely, by our spontaneous understanding which is “itself the source of the laws of nature and hence of the formal unity of nature” (A128). The categories which arise from the intellectual self-­activity of our understanding “are concepts which prescribe . . . to nature laws a priori” so that “nature must conform to them” (B163).58 Even if one disagrees with the details of Kant’s approach, its general contours might seem to provide an attractive strategy for defending the legitimacy of a concept like “causality” which is both epistemically problematic and cognitively indispensable.59 Patricia Kitcher argues that if one takes seriously Hume’s epistemological problem one may be led to a Kantian conception of causality as 58  Note here two important qualifications. First, our understanding is only the source of fundamental laws of nature such as the general a priori causal principle; it does not determine the specific nomic regularities and causal powers that exist in the natural order of things (cf. Chapter II). Second, we must distinguish between a priori concepts such as the categories which are constitutive of nature (in its general-­formal unity) and a priori ideas of theoretical reason, such as the idea designating the systematic unity of different empirical laws, which only regulate the way in which we must rationally study nature without thereby determining the objective constitution of nature. 59  The resulting mind-­dependence of causality cannot per se be considered a reason for rejecting the Kantian program: contemporary naturalists such as Loewer endorse such mind-­dependence as well (see n. 45). Loewer, following (his reading of) Lewis, stresses (2004: 191) that in his framework the mind-­dependence of causal laws (which results from the subject-­dependence of the criteria for lawfulness, such as simplicity) is compatible with a robust realism: “which generalizations are the laws is mind-­independent.” As we just saw (n. 58), Kant can make a very similar point: in his view the spontaneous mind imposes only the general feature of causal lawfulness upon nature without thereby determining what the specific laws of nature are.

230  Kant on Freedom and Rational Agency mind-­dependent, especially since the concept of causality is fundamental to externalist epistemologies where our knowledge is based upon reliable causal connections: How can epistemologists appeal to causal relations to explain knowledge if they have no theory of our epistemological access to causal relations? . . . dodging Hume’s criticisms about mysterious or unanalyzed notions of causation may lead to an analysis where “causation” is (partially) “mind-­dependent.” Hence, Hume’s celebrated critique seems to confront causal epistemologists with an awkward choice: carry on with an unanalyzed (and so suspect) notion of cause, or acknowledge that causation may turn out to be (partially) “mind-­dependent . . . .”60

I have argued that the spontaneous cognitive faculty which is the source of natural causality cannot itself be a part of the empirically conditioned order of nature. However, in Kitcher’s view the idealist claim that causality is mind-­dependent can justify our causal beliefs even if our intellect is conceived in a wholly naturalistic fashion. On her view, the central aspect of Kant’s justification strategy is his (Second Analogy) argument that the concept of causality is a necessary condition of objective temporal experience. If that argument is correct, it is rational and legitimate for us to make causal judgments; we can then happily concede that the mental processes which generate causal representations are themselves determined by natural causes. I will raise two separate problems for this attempt to combine an idealist view of natural causality with a wholly naturalistic view of the human mind. My account of the first problem draws upon but amplifies some of the considerations that emerged in Sections VII.2–4. The first thing to note here is that our valid epistemic reasons for thinking that events are causally connected must be independent of how we actually think as a matter of fact. That is: even if we failed to represent events as causally connected, there would still be valid epistemic reasons for representing events as causally connected, assuming (with Kant and Kitcher) that representing things in this way is indeed a necessary epistemic condition of objective temporal experience. Now, if our actual tendency to think of events as causally connected is the result of natural causes, then our causal thinking merely happens to coincide, as a matter of contingent empirical circumstance, with our epistemic reasons: that we happen to think in causal terms is entirely unrelated to the fact that we have good ep­i­ste­ mic reasons for thinking in causal terms. This is because the natural evolution of our mind (or brain) which is responsible for the fact that we happen to think in

60  Kitcher 1999: 418.

Freedom of Thought as a Condition of Theoretical Cognition  231 causal terms occurred entirely regardless of the fact that there are good reasons for thinking in causal terms: clearly, the contingent empirical processes (e.g.,  genetic mutations) that gave rise to our mental dispositions were not attuned to a priori necessary epistemic conditions for achieving objective temporal experience. Thus, if our causal judgments are determined by mechanisms whose character depends on the natural evolution of our mind, it follows that  we would also make such judgments if there were no valid epistemic ­reasons supporting these judgments. And, if the neurological hardware that now determines us to make causal judgments changed (say, because of a global change of atmosphere altering our gene pool), then we would no longer be disposed to make causal judgments even though there would still be the same epistemic reasons supporting those judgments. This illustrates that our contingent natural disposition to habitually make causal judgments in response to external stimuli bypasses our awareness or appreciation of non-­contingent, timelessly valid epistemic reasons. Hence, on Kitcher’s naturalistic view our intellect lacks “the capacity to determine its judgment according to objective reasons that are valid at any time”; instead, it stands “under the mechanism of merely subjectively determining causes, which may change subsequently” (RezS, 8:14). We are thus thrown back to Hume’s debunking diagnosis. A mental process that is altogether insensitive (intellectually blind) to considerations which might rationally support the use of some concept cannot be regarded as a rational proc­ ess that confers objective epistemic justification on judgments which apply that concept. One might insist that all that matters here is whether the concept of causality can be objectively justified from an external theoretical perspective that evaluates the internal mental processes of those who apply this concept.61 But the question is: why does this matter, and to whom is it supposed to matter? If our internal use of the concept of causality is entirely insensitive to considerations that validate this concept, then it is hard to see what a putative external justification is supposed to be doing for us. If a verdict delivered from a perspective which is wholly external to our cognitive practice confers legitimacy on this practice, and if it must be conceded that the practice would continue in exactly the same way even if the verdict turned out to be less favorable, the external verdict seems irrelevant as far as the rational or non-­rational character of our cognitive practice is concerned. The resulting sense in which we are supposedly justified in applying the 61  This response is suggested not only by Kitcher but also by Longuenesse (1998: 34, 42–2, 64, 245, 366) and Anderson (2001: 292). They see Kant as adopting an external perspective towards mental processes that are not within the thinker’s rational control but that can be normatively assessed from the outside. One can note, in such approaches, an affinity with externalist-­reliabilist models of knowledge or justification (see, e.g., Goldman 1967; Papineau 1992) that I view as antithetical to Kant’s epistemology (see n. 69).

232  Kant on Freedom and Rational Agency concept of causality is similar to the sense in which a thermometer which has been programmed to track the temperature works appropriately when it successfully tracks the temperature. In both cases, we are dealing with a mere mechanism that operates according to some internal structure which has been imposed on the mechanism from without and over which it has no intellectual control. Its intellectually blind internal processes occur in response to external stimuli completely regardless of whether or not there are good reasons for responding in this manner. Thus, the external assessment that these internal processes are appropriate or justified has no implications for the rational character of the process or the processing subject. Moreover, there is an important disanalogy between mechanical devices like thermometers and our human mind qua naturally determined mechanism. The verdict that confers appropriateness on a thermometer’s internal processes relies on two facts: (1) the device has been intentionally set up by an external agent to serve a certain purpose (calculating the temperature); (2) the device operates as designed and reliably serves this purpose. Neither (1) nor (therefore) (2) apply to our causal thought processes on a naturalistic view: on this view the contingent constitution of our brain that prompts our causal thinking derives from non-­ purposive empirical (e.g., evolutionary) processes that were not guided by any rational intentions. So, we cannot deem our causal thought processes appropriate or justified by proxy, namely, by deferring to the rational intentions of an external agent that has purposively designed our inner hardware. One might still insist that none of this matters as long as one can certify the validity of our causal thought-­patterns from an external theoretical perspective which establishes that these thought-­pattens are epistemically required for objective temporal experience. However, those who purport to establish this are in the same boat as those whose causal thought-­patters they seek to evaluate, because their own mental capacities are determined by the same natural mechanisms. When they devise concepts like “objective cognition,” “experience,” or “necessary epistemic conditions,” their epistemological and modal representations are not based on any perceptual data; they can only come from within, namely, from the internal set-­up of the human mind that is (on the naturalistic view) imposed on us by natural causes like our genetic development, environmental circumstances, etc. Thus, if our human intellect belongs to the natural order, then our conceptions of “necessary epistemic conditions” for “objective cognition” are also a mere upshot of contingent, subjective habits emanating from non-­rational causes that have wired the human mind during the antecedent course of nature. There is no way to break out of the naturalistic box here, i.e., no truly external perspective (like that of the engineer towards the thermometer) that we can take towards our causal thinking in order to assess its rationality. The claims that we happen to come up with from an allegedly external epistemic perspective are also (like our internal causal thinking itself) arbitrary upshots of our contingent sub-­personal hardware.

Freedom of Thought as a Condition of Theoretical Cognition  233 I now expound a second problem for the naturalistic-­idealist view. A naturalistic view of our causal thinking suggests the following claim N: “Our causal thought is determined by natural laws that might (at least conceivably) have been different and thus might have caused us not to think in causal terms.” There would be no problem with N if we could confidently represent causality as a feature of mind-­independent reality: in that case we could say that if the causal laws of nature had turned out differently, the human mind—­qua part of the natural order—­would have failed to apprehend the inherent causal structure of nature. But if we envisage causality in an idealist vein as a mind-­dependent feature (so that we can solve Hume’s problem and secure our cognitive access to causal relations), then N looks gravely confused. According to N, if the laws of nature had been relevantly different the human mind would have been endowed with a causal structure that prevents it from forming causal representations. Now combine N with the idealist view of natural causality as dependent upon our representational faculties. This implies: the natural world might have been such that our mind stands under causal laws which prevent it from contributing the property of lawful causality to nature. But if our mental activity is the very source of lawful natural causality, then our mental activity cannot stand under causal laws of nature that might determine us not to contribute this property to the ­natural world. A similar point can be made regarding concepts such as simplicity or unity. Some contemporary philosophers accept that the property of natural law-­likeness depends on our representations of nature’s simplicity and unity: the property of being a law is “partly constituted by psychological factors.”62 Here mind-­ dependence is equated with dependence on our contingent natural psychology. But if the property of law-­likeness depends in part on our commitment to simplicity and unity, then the notion that this commitment is itself a mere matter of contingent natural (psychological) law is incoherent. If the very feature of causal lawfulness depends upon our psychological disposition to think in terms of unity or simplicity, then we cannot conceive our psychology as being governed by causal laws that might dispose us not to think in terms of unity or simplicity. But we must be able to conceive our mental activity in this way if (as naturalists insist) our mind is a part and parcel of the contingent natural order. Naturalist-­idealist views face a related problem. Any valid causal law must explain the occurrence of the relevant effects by reference to conditions that are causally sufficient for these effects. Thus, those who combine the mind-­ dependence of lawful causality with the naturalistic idea that our causal thinking itself stands under natural causal laws must accept that there is a natural causal law which explains why and how the human mind generates the property of 62  Loewer 2004: 191. He traces this point to Lewis: “So it is . . . a consequence of Lewis’s account that which propositions are laws depends on mental facts about us.” See nn. 45, 59 above.

234  Kant on Freedom and Rational Agency lawful causality. But if the human mind is supposed to generate the property of lawful causality via its cognitive representations, then one cannot invoke this property to explain (by appeal to specific causal mechanisms) how it arises from (or gets contributed to reality by) those cognitive representations. This is a vicious circle where the alleged explanans presupposes the explanandum. Put in metaphysical terms: the naturalist-­ idealist approach implies both a metaphysical dependence of our causal thinking on specific natural causes and a metaphysical dependence of natural causality in general on our causal thinking. This is incoherent because an activity cannot be the source of x while already instantiating and (causally) requiring the existence and operation of x. This problem undermines Kitcher’s claim that even though “the causal relation is itself contributed by the actions of our faculties,” these actions can be explained by appeal to natural laws in a naturalistic cognitive psychology. In response to the worry that this explanation is viciously circular since it presupposes the explanandum, Kitcher claims that we can regard this as a coherent “reapplication of the model to itself,” which shows that “Kant’s epistemological theory has . . . the same status as a scientific theory.”63 To be sure, any ordinary scientific theory that attempts to discover empirical causal laws will (quite rightly, in Kant’s view) take it for granted that lawful causal mechanisms exist. But the peculiar scientific theory that Kitcher envisages cannot take this for granted: in trying to explain how our mental actions contribute lawful causality to nature in the first place, the theory cannot invoke (in the content of its explanation) lawful causal mechanisms which already govern the causality-­contributing mental processes. Accordingly, for Kant his philosophical account of natural causality as originating in our noumenal spontaneity has metaphysical and epistemological priority over scientific claims that invoke natural causes in order to explain phenomenal processes. To further illustrate this problem, consider how Kant argues for the cognitive indispensability of causality: in his view causality is a condition for objective time-­determination. The mental act of contributing the causal relation to an empirical reality which (only) thereby acquires objective temporal order cannot itself causally depend upon temporally ordered events.64 One cannot treat the mental activity that makes temporal order among events possible in the first place as a given datum for a naturalistic cognitive science, because any such datum must already exhibit the determinate temporal structure that scientific investigations (rightly) take for granted. But Kant’s view here suggests a more general point which does not require his specific claim that causal thinking is necessary for objective time-­determination. The concept of causality is so fundamental to our cognitive efforts that no serious attempt to understand the natural world can dispense with this concept (cf. Section VII.3): any attempt to explain any natural 63  Kitcher 1999: 434–5. 64  This is a major theme in Wolff ’s classic (1964) treatment of Kant’s views on mental activity.

Freedom of Thought as a Condition of Theoretical Cognition  235 phenomena must already presuppose the existence of lawful causal relations. This means that one cannot give a causal explanation of how causality or causal lawfulness as such comes to exist in nature. A fortiori, one cannot give a causal account of how the human mind contributes a (general) causal structure to the natural world either. The common source of these problems is the attempt to reap the benefits of the Kantian approach to a priori concepts like “causality” while refusing to pay the price that this approach demands. The Kantian view traces properties such as causal lawfulness to our mental capacities that (as it were) shape the natural world in their own cognitive image by making nature conform to the general conditions under which we can have objective cognition of nature. This approach can avoid the Humean worry that we have no cognitive access to necessary causal relations and thus no rational basis for believing that such relations exist in nature. But this approach also comes at a price: it requires us to forsake the ambition to cognize an absolute mind-­independent reality and to abandon the concomitant idea that all reality, including the human mind, is determined by natural causes. A view on which the mental actions that contribute causality to nature are themselves governed by natural causes attempts to have it both ways. It purports, in an idealist fashion, to trace the very property of natural causality to the human mind (thereby making natural causality cognitively accessible to us). But it also claims, in a naturalist-­realist fashion, that everything there is stands under and can be explained via natural causes. I have argued that this attempt to marry Kantian idealism with anti-­Kantian naturalism is incoherent. With my arguments in Sections VII.2–5, I have reconstructed the motivation for Kant’s view that our epistemic rationality requires an empirically unconditioned spontaneity of thought. This reconstruction addresses Karl Ameriks’s important compatibilist challenge “that although what we intend when we use our understanding and reason may appear transcendently intended, our intending it none the less might be sufficiently explained by a natural process that involves nothing transcendent in us.”65 The attempt to sufficiently explain our use of understanding and reason in terms of natural processes presupposes that we are rationally entitled to the belief that nature indeed contains the necessitating causal conditions which we cite in our explanations. Now assume that our mental acts of causal-­explanatory judgment “involve nothing transcendent in us”—they are mere products of nature. These mental acts cannot, in light of Hume’s problem, be regarded as products of rational, evidential-­perceptual determination by external natural causes. Instead, these mental acts are determined by internal natural processes, namely, by psychological or neurological mechanisms that operate in response to external stimuli. But in that case our causal judgments,

65  Ameriks 2000a: 207 (his emphasis).

236  Kant on Freedom and Rational Agency including our general conviction that nature contains necessitating causes, result from contingent, sub-­personal, cognitively blind mechanisms beyond our ra­ tional control. This undermines our rational entitlement to causal judgments and to sufficient naturalistic explanations. Kant’s idealist conception of natural causality suggests a solution to Hume’s problem: it treats causality as a mind-­dependent feature whose existence we can know by examining the structure of our own cognitive faculties (namely, by investigating the conditions under which these faculties can achieve objective experience of nature qua structured by our cognitive faculties). But in treating natural causality as a mind-­dependent, thus (in principle) knowable feature, Kant has a conclusive reason for denying that all our cognitive activity can be sufficiently explained via natural causes: it is incoherent to suppose that the cognitive spontaneity which is the very source of natural causality is itself sufficiently determined and explained by natural causal processes.

VII.6.  Causal Belief and Divine Determination I have explained why Kant insists that our theoretical intellect cannot be empirically conditioned by natural causes. But, especially in the context of early modern Rationalism, an empirical-­natural conditioning relation is not the only relevant threat to the absolute spontaneity of the human mind. What about the possibility that our theoretical intellect is determined by a super-natural divine being? Here, one might hold, Kant’s reasons for denying the natural determination of our intellect no longer apply: a perfectly rational divine cause would not set up our intellect in a contingent, mutable, and merely subjective (i.e., rationally arbitrary) fashion. The idea that our intellect is set up to judge in certain ways by a divine cause might even re-­invite the notion of determination by the truth, albeit in a new transcendent guise: while this notion seems (for the reasons given in Sections VII.2–5) indefensible when we consider the relation between our judgments and natural causes, a supremely perfect (rational, benevolent) non-­natural being would determine our mind to judge in ways that are reliably correlated with the way things really are. Thus, our theoretical beliefs would be determined by the truth, which is “just what we want” for such beliefs. However, Kant deems our theoretical rationality incompatible with divine determination as well. The key passage here is the conclusion of the B-­deduction, B167–8.66 The divine determination view says that categories like “causality” and their corresponding principles (e.g., “every event has a cause”) are “predispositions for thinking, implanted in us along with our existence by our author in such a way that their use would agree exactly with the laws of nature along which

66  My argument in what follows picks up and expands on my discussion of B167–8 in section IV.5.

Freedom of Thought as a Condition of Theoretical Cognition  237 experience runs (a kind of preformation-­ system of pure reason)” (B167). Descartes’s clear and distinct ideas come to mind, but Kant probably refers chiefly to Crusius’s suggestion that we can account for the objective validity of our basic theoretical concepts by assuming a kind of pre-­established harmony between our mind and the world: a divine creator makes the basic structure of our thought agree with the structure of reality (cf. Br., 10:131; Prol, 4:319). Kant rejects this proposal for the same reason, even in the same terms that also inform his critique of the naturalistic Humean account. The divine preformation hypothesis implies: [That] the categories would lack the necessity that is essential to their concept. For, e.g., the concept of cause, which asserts the necessity of a consequent under a presupposed condition, would be false if it rested only on a subjective necessity, arbitrarily implanted in us . . . . I would not be able to say that the effect is combined with the cause in the object (i.e., necessarily), but only that I am so constituted that I cannot think of this representation otherwise than as so connected; which is precisely what the skeptic wishes most, for then all of our insight through the supposed objective validity of our judgments is nothing but sheer illusion.  (B167–8)

For those who are sympathetic to theistic Rationalism in the vein of Descartes or Leibniz, this argument must seem puzzling for two main reasons. First, they would argue that there is nothing “arbitrary” about God implanting true cognitive principles in our minds—­rather, this is something God must (cannot but) do given His perfectly rational and benevolent character. Second, one might wonder what benefit there is to grounding the relevant principles in our transcendental freedom. Our freedom may provide us with alternatives in selecting our basic cognitive principles, but it is unclear why we would want such alternatives: all we could want here is that such principles are necessarily true, a result that would surely obtain if they derived from divine preformation. One might even argue that Kant’s claim that these principles derive from our freedom incurs the main problem that he raises for the naturalistic view: transcendental freedom also involves a kind of contingency which would (like the contingency of evolutionary causal chains) undermine the objective justification of our beliefs. If it is entirely up to our freedom how we judge and assent, then we might always be wrong. The second worry passes over Kant’s distinction between executive and legislative freedom (cf. Chapter III). Only executive freedom involves contingency and alternative possibilities (and, as we saw in Section VII.1, alternative possibilities are metaphysically essential only to our executive freedom of will; they accompany our executive freedom of thought typically rather than essentially). The a priori categorial principles of our understanding (and the regulative principles of our theoretical reason) spring necessarily from our legislative cognitive freedom: we are compelled to form these principles by our empirically unconditioned

238  Kant on Freedom and Rational Agency autonomous intellect (just like we are necessarily compelled to form the moral law by our pure, autonomous practical reason). Thus, Kant denies that we either want or possess alternatives when we use our legislative freedom to prescribe correct a priori principles. Likewise, he denies that freedom qua autonomous self-­ legislation implies the possibility of error: free self-­legislation rather involves a necessary rational self-­compulsion to get things right. But this does not answer the first worry: why is grounding our basic cognitive principles in our autonomous intellect preferable over grounding them in divine preformation? Why does Kant claim that the divine preformation hypothesis is on par with the natural determination hypothesis insofar as both imply that our (e.g.) causal judgments lack objective validity and rest on a merely arbitrary subjective necessity? One salient commonality is that both hypotheses trace the categories to “foreign” causes that operate outside of our rational self-­control. In the divine preformation case, categories like “causality” “rest only on a subjective necessity, arbitrarily implanted in us” in the sense that our causal thoughts are mental events that happen to (or in) us due to the way in which a divine “puppeteer” (KpV, 5:101) programs our mind like an engineer constructs a thermometer. We judge in causal terms just because we cannot help but doing so. Exactly the same is true in the naturalistic scenario. The only difference between the two scenarios lies in how one conceives the external genesis of our thoughts. Accordingly, the only rational character that our causal thoughts supposedly have in the super-­ natural scenario is due to an alleged external rational influence on our thought processes, deriving from the alleged rational insights and benevolent intentions of the external cause that sets up our mental dispositions. This external influence has no bearing on our internal perspective, our epistemic self-­awareness as cognizing subjects: since our causal thought does not derive from our own autonomous rational insight, our causal judgments are not governed by our awareness of objective reasons for making these judgments. Rationalists might respond that God wires our mind in such a way that our judgments are based on our reflective recognition of the objective validity of the principles that God instills in us. However, if our appreciation of objective reasons for reaching causal judgments depends on divine preformation, then we can have such an awareness only if we have objective reasons for believing in divine preformation. For Kant, claims about the existence, rational character, and benevolent intentions of God are mere conjectures that can never be justified (not even as a matter of legitimate faith) on a purely theoretical basis. Accordingly, his critique of Crusius’s divine preformation hypothesis is that since we lack certain criteria for distinguishing a “genuine” from a deceptive external origin, we can never know whether we are set up to think by the “spirit of truth” or the “father of lies” (Prol, 4:319). If we wanted to ground our entire cognitive (categorial) framework on an unverifiable hypothesis about the character of the external origin that is

Freedom of Thought as a Condition of Theoretical Cognition  239 responsible for this framework, then our theoretical cognition would lack objectively certain, rational foundations. Thus, from our internal epistemic perspective things look exactly the same in the divine preformation and the naturalistic scenario: our causal thoughts arise just because they are implanted in us by foreign causes via a given necessity that strikes us as an arbitrary non-­rational compulsion. Since we have no active ­ra­tional control over our causal thoughts, we cannot help but thinking in causal terms wholly regardless of whether we deem such thoughts justified or not. The only putative difference between the two scenarios is one that we cannot ra­tion­ ally ascertain. This explains why Kant suggests that the divine preformation hypothesis is “precisely what the skeptic wishes most.” “The skeptic” here refers to an empiricist Humean skeptic who doubts the objective validity and truthful origin of a priori concepts such as “causality” and who denies that we possess the non-­trivial (synthetic) a priori cognition that objectively valid a priori concepts would provide.67 We can imagine the following dialogue between such a skeptic and divine preformation theorists. Divine preformation theorists concede that on their view the only reason why we causally connect an A-­type representation with a regularly succeeding B-­type representation is “that I am so constituted that I cannot think of this representation otherwise than as so connected.” The Humean skeptic gleefully observes, “Right on! We hold exactly the same view about our causal thought processes.” The preformation theorist responds, “But on my view, I am constituted to causally connect A with B because my mind is set up by a perfectly ra­ tional, benevolent divine cause!” We can imagine the Humean shrugging his shoulders, retorting: “Maybe, or maybe not, we can never know, and so it does not matter. The unknowable external origin of our mental constitution makes no difference to the rationality of our thinking: we will judge in causal terms regardless of whether or not we deem judging in this manner rationally justified, regardless of what we may or may not believe about some alleged divine puppeteer. Thus our causal judgments are, in any case, a function of mental habits which no ra­ tional thought process on our part is able either to produce or prevent.” This is how we should understand Kant’s concluding statement at B168: I “would not be able to quarrel with anyone about that which merely depends on the way in which his subject is organized.” One might insist that the truth of the preformation theory would make a genuine difference to our theoretical knowledge: if a benevolent God is indeed the common cause of our concepts and their objects, then these concepts do have

67  For confirmation that Kant’s transcendental deduction engages Humean skepticism, see A128; B5; B127–8; Prol, 4:260, 312–13. See Kohl 2018c for detailed discussion. See also Allais 2015: 162; Ameriks 2003: 60–1; Longuenesse 2017: 181.

240  Kant on Freedom and Rational Agency objective validity and our (e.g.) causal judgments do provide us with knowledge.68 But this claim rests on an externalist, e.g., reliabilist conception of knowledge that Kant rejects. For Kant, the very idea of knowledge and especially a priori knowledge requires that our knowledge be grounded in our own reason, namely, in our consciousness of objective universal validity, necessity and certainty (JL, 9:70–1).69 Thus, our categories cannot obtain their objective validity, and our categorial judgments cannot acquire the sufficient objective grounds required for knowledge, from some external cause to which we have no cognitive access from within our epistemic self-­awareness. If an externally sustained reliable connection between representations and objects were sufficient for knowledge, then higher animals should also be granted knowledge of causal relations just in case their associations are reliably correlated (e.g., via genetic wiring) with causal relations. There is a further problem with the preformation hypothesis and with the reliabilist conception of knowledge that one might adduce on its behalf. Since “the skeptic” Kant addresses doubts the objective validity of our causal thought, the attempt to defeat such skepticism with a causal hypothesis about the origin of our causal thought is viciously circular.70 That is: one cannot counter the Humean worry that the concept of causality is inherently “false and deceptive” because it is not rightfully acquired (KpV 5:51) by claiming that there is some external cause that truthfully sets up our causal thoughts so that they are reliably connected with causal facts. It does not matter here whether this claim appeals to a benevolent super-­natural cause or to a reliable naturalistic (e.g., evolutionary) causal mech­ an­ism (that allegedly tracks causal relations): Hume’s skepticism targets both our empirical and our noumenal causal judgments (KpV, 5:56). However, Humean naturalists cannot gleefully endorse this dialectic situation. Their appeal to the non-­rational origin of our causal thinking is itself a causal hypothesis. Causal hypotheses are integral to any naturalistic science including Hume’s own natural science of the human mind. According to the Humean

68  See Stang 2016: 192–4. 69  Chignell reads Kant as an externalist (who only gives an “emphatic nod” to internalism) because he thinks that some of the facts that provide sufficient objective ground for knowing assents “will typically be inaccessible to a normal subject” (2007b: 49). However, if the objective grounds for knowing assents were inaccessible to subjects, then they would also be incommunicable. For Kant, the universal communicability of one’s reasons for assent is a strictly necessary condition of knowledge qua “certainty for everyone” (A821/B849; A829/B857; JL, 9:70; see Kohl (forthcoming3). For Chignell, it is specifically facts about the objective probability of our grounds for assent which are often inaccessible to cognizing subjects. I do not agree that Kant holds this view. But in any case, this point is moot when it comes to the issue of knowledge, because probabilistic assents yield only (comparatively) uncertain opinions (A775/B803; JL, 9:66–7, 81–5; V-­LO/Phil, 24:433). 70  My point here differs from Stang’s suggestion (2016: 192–4) that Kant’s reason for setting aside the divine imposition theory at B167–8 is that it would be circular to rely on claims about God in the context of examining whether our categories yield a priori cognition of noumena. This is implausible because Kant has already denied categorial cognition of noumena at B147–9. At B167–8, he argues that the divine imposition hypothesis fails because it implies that the categories rest on a non-­rational subjective necessity that undermines our objective categorical cognition of empirical phenomena.

Freedom of Thought as a Condition of Theoretical Cognition  241 hypothesis, our mind is determined by instinct or habit to judge in causal terms; one can predict that if a human being has seen 100 A-­B correlations it will, upon perceiving an actual A, judge (due to the sheer force of habit) that B will and must happen. This naturalistic theory puts forward inductive and causal claims about how the human mind works and why we think as we do. But according to this theory, such claims have no rational basis since they are the upshot of sub-­ personal mechanisms that do not involve our higher intellectual faculties. Hence, this naturalistic theory implies its own lack of rational grounds. Barry Stroud aptly diagnoses the key issue here: “what makes it hard to accept Hume’s results” is the fact that “when human beings are both the agents and the objects of naturalistic study, what is found to be true of the objects studied must somehow also be understood to be true of those of us who conduct the study.” Humeans cannot coherently regard themselves as trying to construct a rational natural science of the human mind while admitting that in forming their naturalistic hypotheses they are nothing “other than a credulous victim of the blind workings of certain ‘principles of the imagination.’ ”71 As we saw (in Section VII.4), similar problems arise for those who use evolutionary theory as a basis for doubting the rationality or cognitive reliability of our mental faculties which are the very source of evolutionary theory itself. We arrive here again at a Kantian (but hardly only Kantian) insight that has repeatedly emerged over the course of this chapter: since the concept of causality is fundamental to all our explanatory efforts, we cannot pretend to give a causal explanation of why our causal judgments have or lack a rational justification. One cannot adjudicate over the rational credentials of our causal thought by helping oneself to a causal explanation of why we think in causal terms. This is so regardless of whether the explanation is framed as a conciliatory cognitive psychology (as in Kitcher), as a debunking cognitive psychology (as in Hume), or as a transcendent appeal to divine causation (as in Crusius, Descartes, or most recently in Plantinga).72 71  Stroud 2006: 348–9. 72 Plantinga makes a transcendent causal claim to explain the objective validity of our causal thought: our cognitive faculties function properly (by virtue of aiming at and reliably grasping the truth) because they are designed by a divine being which intends us to obtain knowledge (1993: 197, 237). Plantinga would presumably say that Kant’s argument at B167–8 does not apply to his account since he does not seek to address Humean skepticism: he takes the reliability of our cognitive faculties for granted (1993: 235–7). However, Kant can give the same reply here as against White (cf. Section VII.4): we cannot simply take the validity of our causal thought for granted once we have been awoken from our dogmatic slumber by Humean doubts about non-­ trivial necessity claims. Moreover, Plantinga’s approach also ignores Kant’s point that if the alleged rational credentials of a judgment depend upon an external origin whose existence and character lies beyond our rational insight, then we are not rationally entitled to this judgment. Plantinga might respond that from his initial premise that we can trust the reliability of our cognitive faculties, we can rationally prove that these faculties must have been designed by a perfect being. However, it seems that Plantinga’s argument rather yields a reductio of the initial premise: if the trust we are initially inclined to place in our cognitive faculties turned out to require metaphysical claim about a perfect God, this initial trust would (or should) be

242  Kant on Freedom and Rational Agency

VII.7.  How We Can Know Our Transcendental Epistemic Freedom Having reconstructed the grounds of Kant’s epistemic incompatibilism, I now examine the doxastic status of the judgment that our theoretical intellect has the absolute spontaneity of thought that is (as I argued in Chapter VI) necessary and sufficient for transcendental freedom. For Kant, every instance of holding some proposition to be true can be classified as opinion, faith, or knowledge.73 Opinions are impermissible regarding intelligible, non-­sensible properties or beings because opinions are inherently probabilistic and empirical (A772/B800; KU, 5:465). Hence, our doxastic attitude towards the proposition that our noumenal thinking self is an absolutely spontaneous intelligence must be either faith or knowledge. In my view, Kant holds that we can know this proposition to be true. My argument here is as follows. Assume that I have cognitive achievements that amount to knowledge. Assume, further, that I can know that these achievements presuppose the spontaneous thinking activity of myself qua non-­sensible intelligence. If so, I can infer that “I exist as an intelligence that is . . . conscious of its” spontaneous “faculty for combination” (B158). We have actual empirical cognition (cognition(N)) and knowledge of nature. We can know that this requires our spontaneous intellectual capacity for combining given data, which cannot be ascribed to our natural-­phenomenal (“determinable” rather than “determining”) self. Hence, reflection on the conditions of our actual empirical cognition and knowledge of nature gives us knowledge that our noumenal thinking self is a spontaneous intelligence. The kind of cognition and knowledge that this argument presupposes as actual does not include controversial items such as synthetic a priori knowledge of necessary causal connections. Kant needs to presuppose only that we know the objective temporal order among some natural (e.g., our phenomenal mental) states: an assumption that is also accepted by empiricist skeptics about a priori cognition. On this basis, Kant can prove the a priori causal principle (that every natural event has a cause) as a necessary condition for our actual knowledge of objective temporal order (via the Second Analogy argument). He can then infer the absolute spontaneity of our theoretical intellect from the insight (cf. Sections

shattered. We cannot reasonably take for granted any claims to knowledge that depend upon theoretically inscrutable assertions concerning a perfect, benevolent external world-­designer. Such assertions raise grave challenges for our theoretical understanding even apart from their unverifiability; consider, for instance, the serious puzzles arising for theists from the observed imperfections of the world, including the striking imperfections of our cognitive faculties that are allegedly caused by a perfect design-­plan. 73 I examine this classification in Chapter VIII. See also Kohl (forthcoming3) and Chignell 2007a, 2007b.

Freedom of Thought as a Condition of Theoretical Cognition  243 VII.2–6) that such spontaneity is the only viable cognitive source of the a priori necessary causal principle. To confirm that this is indeed Kant’s view, consider once more his conclusion to the B-­deduction where he designates the categories and their corresponding laws as “self-­thought a priori first principles” (B167) that originate in the pure (non-­empirical), spontaneous syntheses of our understanding (A78/B104). The categories are “self-­thought” only insofar as they arise from an intellectual “self-­ activity” that is not determined by foreign causes. Accordingly, at B167 Kant rules out two scenarios where the categories are the product of foreign causes. These are the two scenarios that I have considered in this chapter. First, as a priori necessary conditions of experience the categories cannot have “an empirical origin” in contingent natural processes that may be given in experience and studied in empirical psychology (B152). Second, the categories cannot be “implanted in us” by a super-­natural designer. If one concedes that foreign causes determine us to make categorial judgments one cannot avoid Humean skepticism about the objective validity of these judgments, because due to this concession one lacks “certain criteria” for establishing and thus one “cannot know with certainty” whether the external origin of the categories is truthful or rather unreliable, non-­ rational, and deceptive (Prol, 4:319). Conversely, if we have certain knowledge that our causal judgments are objectively valid, we must also have certain knowledge that these judgments do not depend on the way in which our mind is set up by some foreign causality whose rational or non-­rational, deceptive character lies beyond our ken. Now we do have certain knowledge that our causal thought has objective validity. A priori categorial principles such as “every event has a cause” are apodictically certain (A160–1/B199–200), and apodictic certainty is a mark of knowledge (JL, 9:66). Since we cannot know these principles unless they are a priori “self-­ thought,” the claim that they are a priori self-­thought by our spontaneous understanding must also qualify as certain knowledge. Thus, our intellectual self-­activity reveals itself to us as the source of known a priori cognitive laws such as the general causal principle, which in turn provides objectively valid grounds for our empirical causal judgments.74 74  This does not mean that the known a priori causal principle grounds knowledge of, or provides sufficient reasons for, empirical causal judgments. Whether and how we can know specific natural causes is an intricate question. But the crucial point here is that for Kant, the known a priori truth that every event has a cause provides our causal hypotheses with a rational basis and an objective validity. Kant’s concept of “objective validity” is not the same as “knowledge” or even “truth” (cf. Allison 2004: 88). A causal hypothesis can be false and yet have objective validity insofar as it is a viable candidate for empirical truth, since it is based on public, intersubjectively accessible reasons for assent (cf. Chapter VI.2). Our general knowledge that there are causes in nature which make events happen provides the intersubjectively certain basis for legitimate, true or false, hypotheses about what these causes are. Such hypotheses may qualify as highly probable opinions that approximate the status of certain knowledge claims (JL, 9:82–5). But no opinion that A is the specific cause of B-­events could have this dignified epistemic status unless it was based on the certain knowledge that all events in

244  Kant on Freedom and Rational Agency Kant’s commitment to this line of reasoning can also be witnessed in the second Critique. There (at KpV, 5:30) he argues that if we attend to the a priori necessity of first theoretical principles, we can become conscious of our pure, empirically unconditioned understanding as the source of these principles. Our pure, non-­empirical self-­consciousness pertains to ourselves as things in themselves (KpV, 5:6). Hence, the known a priori necessity of theoretical principles allows us to grasp that these principles originate in our spontaneous noumenal intellect. In these two passages (B167–8; KpV, 5:30), Kant argues that we have an inferential basis for knowing that our theoretical intellect is absolutely spontaneous or self-­active: this knowledge is derived from our cognitive awareness of a priori representations whose content (an objective necessity) cannot originate in an empirically (or otherwise) conditioned intellectual capacity. Thus, our consciousness of the a priori causal principle is the ratio cognoscendi of our noumenal cognitive spontaneity. The inference from our objective consciousness of a priori necessary principles to the existence of our spontaneous intellect as the source of this consciousness might exemplify the special mode of “transcendental reflection” that considers which of our different sources of cognition is the origin of some (perhaps self-)given representation (A260/B316). How is the knowledge that our noumenal self possesses an absolutely spontaneous intellect compatible with Kant’s critical epistemology, in particular with his commitment to noumenal ignorance? This question is often combined with or driven by the complaint that an appeal to noumenal self-­knowledge would commit Kant to the dogmatic rational psychology that he so vigorously attacks in the Paralogisms. We have already seen how this complaint leads to the thinking mechanism interpretation in (e.g.) Sellars. Likewise, Walker claims that the appeal to our noumenal spontaneity would open the floodgates for metaphysical

nature must result from some causal condition or other. This general knowledge is prior to all our empirical knowledge, including our knowledge of events or changes: the Second Analogy proof shows that we can have empirical knowledge that a change of states C-­D occurred only based on our a priori knowledge that some cause or other made the C-­D succession objectively necessary (see Chapter II n. 2). On Guyer’s reading (1987: 252–9), we need to know specific empirical causes and laws to confirm that we are perceiving events. This reading has the implausible implication that subjects who do not know specific causal laws therefore also lack knowledge of whether events happen in nature. It further invites a strong concern about circularity: arguably, we can establish causal laws only if we already know that certain types of events regularly succeed each other (see Allison 2004: 257–8). Kant frequently expresses the conclusion of his proof as the claim that our objective experience of events requires only the known a priori presupposition that events follow from some cause or other. “If, then, we experience that something happens, we in doing so always presuppose that something [irgend etwas] precedes it, on which it follows according to a rule.” (A198–9/B244) “There is an order in our representations in which the present, so far as it has come to be, refers us to some preceding state . . . and though this correlate is, indeed, indeterminate, it none the less stands in a determining relation to the event as its consequence . . .” (A195/B240). For instance, if “wax, which was formerly hard, melts, I can know a priori that something must have preceded . . .” (A766/B794) (emphases are mine).

Freedom of Thought as a Condition of Theoretical Cognition  245 speculation in the vein of rational psychology.75 I believe that these worries can be significantly defused by emphasizing three central points. First, there is an important methodological difference between the way in which rational psychology tries to establish its claims and the way in which Kant establishes the absolute spontaneity of the noumenal thinking self.76 The rational psychologists (as Kant conceives them) try to reach non-­trivial claims about the self either via some kind of immediate self-­intuition or via analytic judgments that examine concepts like “self ” or “thinking.” By contrast, Kant derives his knowledge claim that our noumenal thinking self has absolute spontaneity through an inference from a synthetic basis: namely, via transcendental reflection on the origin that our representations must have if they can supply objectively valid synthetic (a priori and empirical) cognition. This methodology respects Kant’s critical denial that we can know ourselves as we are in ourselves via intuition or analytic judgments.77 One might object that if our noumenal self is not given to us in any intuition then it cannot be an object of cognition and knowledge.78 However (this is my second point), Kant sometimes allows for rational knowledge of x that is not grounded in intuitive cognition(N) of x.79 Consider the belief that there are things in themselves. Arguably, this belief amounts to knowledge: we know (through sense-­perception) that certain objects exist and we know (through philosophical reflection) that we experience only the mind-­dependent sensible constitution of these objects. Hence, we can knowingly infer that these objects also have some mind-­independent non-­sensible constitution even though this constitution is not given to us in any kind of intuition.80 Our knowledge that there are non-­sensible aspects of reality is justified as a certain, necessary entailment of our sensible cognition(N): “. . . beings of the understanding certainly correspond to sensible

75  Walker 1978: 131–5. 76  For a systematic reading of Kant as rejecting primarily the methods of rational psychology, see Ameriks 2000a. 77  Wuerth argues that Kant allows for an immediate, non-­inferential awareness of ourselves as spontaneous noumena. When he discusses Kant pre-­critical early 1770s views, he attributes to Kant the claim that we possess “an intuition of our self ” which is “the only intuition of a thing in the noumenal sense” (2014: 77). But an intuition of our noumenal self would have to be a non-­sensible, intellectual self-­intuition. In Kant’s critical view, our finite mind cannot possess this divine type of intuition (see, e.g., B159). Thus, when discussing the critical view Wuerth says more cautiously that we stand in a “nonsensible manner of contact, in pure apperception, with ourselves” as noumena (2014: 122). However, if this contact is to be immediate rather than (as on my reading) inferential, then it is hard to see how it could be anything but (per impossible) intuitive. For Wuerth, pure apperception yields an indeterminate consciousness (though not knowledge or cognition, which for Wuerth require intuition) not just of our spontaneity but also of our noumenal substantiality, simplicity, etc. (2014: 165–73). This is hard to square with Kant’s critique of rational psychology (see Proops 2021: 134–6). 78  See Proops 2021: 84–5; McLear 2020; Wuerth 2014 (cf. n. 77). 79  See Chignell 2014; Kohl (forthcoming3); Schafer 2017; Watkins and Willaschek 2017a, 2017b. 80  See Allais 2010: 15 for a similar point. This reasoning relies on further premises, such as the claim that our finite mind does not create the objects that it sensibly cognizes (cf. B72).

246  Kant on Freedom and Rational Agency beings . . .” (B308); “. . . beyond the sensible world there must necessarily be found something that is thought only by the pure understanding . . .” (Prol, 4:361). Yet (this is my third point), in Kant’s critical doctrine these knowledge claims about non-­sensible reality face severe epistemic limitations. Since we lack cognition(N) of things as they are in themselves, we cannot turn the abstract knowledge that there are non-­sensible aspects of reality into positive, determinate knowledge of what these non-­sensible aspects are like or how things are constituted in themselves, i.e., in their intelligible character (B149): “. . . we . . . cannot know anything determinate about . . . intelligible beings” (Prol, 4:315). Similarly (cf. Chapter VI), even though we have abstract knowledge that our spontaneous noumenal self is responsible for our cognition(N) and for our determinate knowledge of nature, our inability to intuitively cognize(N) our spontaneous thinking self prevents us from acquiring determinate knowledge about this self. This restriction has two dimensions. First, we cannot know anything about our noumenal thinking self apart from what we can infer about it from its representational output. Our determinate, objective cognition(N) and knowledge allows us to infer that our noumenal self has the capacity for absolutely spontaneous, autonomous cognitive self-­determination; but we lack determinate theoretical insight into how this capacity works. Our knowledge that we have the capacity for spontaneous cognitive self-­activity is theoretically fruitless in the sense that we cannot explain, understand, or predict how this capacity operates, whereas we can theoretically explain the empirical self ’s combination of representations according to the law of reproductive association or the movement of bodies according to the law of gravity. Second, we cannot cognize(N) or know the ontological constitution of our noumenal thinking self. For instance, we cannot know whether the intelligible subject of “sheer spontaneity” is a simple (thus incorruptible and immortal) or a complex being (A351–61). Hence, for Kant our apperceptive self-­cognition(W) that our thinking self has absolute spontaneity does not suffice to (cognitively) “determine its existence” as a noumenon (B158). The spontaneous “subject of the categories cannot by thinking the categories acquire a concept of itself as an object of the categories” (B422). Specifically, our knowledge that we possess a noumenal cognitive spontaneity does not entail “what one really wants to know” and what rational psychologists really want to establish: namely, that the noumenal self is a self-­subsistent, simple, and incorruptible (hence immortal) being (A400). Thus, pace Walker, Kant can permit knowledge that our cognizing self has absolute noumenal spontaneity without thereby granting the determinate metaphysical self-­knowledge to which rational psychology aspires.81 Here one might wonder: does the (on my reading, known) fact that the noumenal thinking self is absolutely spontaneous not entail that it must also be an

81  See Ameriks 2000a: 287–90 for a similar point.

Freedom of Thought as a Condition of Theoretical Cognition  247 ultimate subject of inherence? I deny that we can know whether this is a valid entailment. A defense of this entailment would have to rely on the claim (C) that an absolutely spontaneous agent cannot inhere in another, more ultimate subject. But there is no conclusive argument for (C). Kant distinguishes (1) the relation between subject and accidents from (2) the relation between a determining cause and its effects: only (2) is to be thought as involving a regressive series from condition to the unconditioned whereas in (1) accidents are coordinated with one another and hence do not constitute a series (A414/B441). Given this distinction, there is no obvious reason to hold that if X inheres in another thing A, X’s actions must be part of a causal series which terminates in A as the ultimate cause of that series (so that X’s actions cannot be absolutely spontaneous). When Kant (seemingly) equates (1) and (2) and thus affirms (C) (e.g., at KpV, 5:101–2), he is specifically concerned with things that are both accidents and effects of another being within a temporal framework. This does not establish (C) with regards to atemporal noumena.82 One might insist that Kant is committed at least to some of the determinate metaphysical claims about our thinking self that are proposed by rational psychologists. While he officially denies that we can know the personal identity of our thinking self, he sometimes claims that the understanding is a capacity for self-­conscious thought which allows a thinker to become aware of different thoughts as hers and which thus grounds our status as persons (Anth, 7:127). Perhaps the belief in our personal identity as spontaneous thinking subjects is yet another case where Kant disagrees chiefly with the methods through which ra­tional psychology seeks to defend this belief (namely, by invoking inner intuition or through analysis of the concept “thinking subject”). As we saw (in Chapter I), we are entitled to a notion of moral personhood that grounds ascriptions of moral responsibility; this entitlement derives from our awareness of our capacity to comply with the moral law (KpV, 5:37, 98–9). By analogy (this would be the very

82  According to student notes, in some lectures Kant infers from the fact that thinking results from my spontaneous self-­activity and thus inheres “in me” that I am an ultimate noumenal substance (V-­Th/Pöl, 28:1042). But it is far from clear whether Kant here intends to make a claim to certain knowledge. Moreover, we should be cautious when considering student notes as evidence for Kant’s considered view: such notes are even more problematic than Reflexionen, since they are third-­party reports about what Kant says in contexts where he may well be relaxing (e.g., for pedagogical purposes) his strict critical-­epistemic standards. As we saw (in Chapter III), Kant’s considered critical view denies that we can use the cognitively empty, indeterminate pure categories to establish determinate metaphysical claims about noumena or relations among noumena. A fortiori, he must also deny that purely conceptual rumination involving the bare categories of substance-­accident and cause-­ effect can conclusively establish the non-­trivial claim (C) that if a noumenon X inheres in another noumenon A, then X’s actions must be the effect of A’s causal determination. One might respond that this is a purely conceptual, analytic truth. But I do not see how (C) could be established via conceptual analysis of the unschematized categories, since they lack the determinate meaning that one would need to analyze if one sought to establish how the relations of inherence and causal determination line up among noumenal objects. In Kant’s view we lack any clear (theoretical) handle on what it means to say that non-­sensible things exist in such relations (B288–9; A241/B300).

248  Kant on Freedom and Rational Agency analogy Kant suggests at RezS, 8:14), Kant might say: our awareness of our cognitive capacity to comply with objective laws of thought entitles us to the belief that we have a cognitive personality which allows us to own our assents in a way that implies epistemic accountability, e.g., when we face the criticism that we have been epistemically careless, insufficiently reflective, guilty of “lazy reason” (A689/ B717) or of “perversa ratio” (A692/B720). I am unsure about whether the idea that we qua subjects of theoretical cognition have a personality that makes us epistemically accountable should be ­classified (like the less determinate-­informative claim that we are absolutely spontaneous cognizers) as objective knowledge or, instead, as a subjectively ra­tional belief (i.e., faith) that lacks objective certainty.83 However we may classify this belief, it calls for the same important qualification that also applies to the belief in our moral personality: since the seat of an accountable cognitive or moral personality is the atemporal noumenal self, the idea that we have a persisting, semi-­permanent personality must be understood in a metaphorical, non-­ literal sense (cf. Chapter II). One might raise another related problem for my reading: if we had theoretical knowledge that our theoretical intellect is absolutely spontaneous, would this not imply our theoretical knowledge that our moral will is absolutely spontaneous as well? This would be a fatal implication given Kant’s insistence that we cannot theoretically demonstrate the (real) possibility of free will (cf. Chapter I). Robert Pippin responds to this worry by suggesting that the ascription of absolute spontaneity in one (epistemic) context does not entail that one must also ascribe such spontaneity in other (moral) contexts.84 For Kitcher, this response is too weak since it provides no reason that would block the generalization from one context to the other.85 My interpretation strengthens Pippin’s point because it supplies a twofold principled reason why knowledge of absolute epistemic freedom does not entail knowledge of absolute moral freedom.86 First, for Kant the spontaneity of thought that is involved in theoretical-­cognitive determination (TCD) does not include the most controversial, metaphysically weighty feature that we attribute to a spontaneous moral will: namely, a non-­sensible form of efficient (“true”) causality that allows us to produce actual external objects as opposed to mere representations of such objects (cf. Chapter VI). Second, for Kant we can gain indeterminate cognition(W) and knowledge of our absolute epistemic spontaneity only because 83  This is how Longuenesse classifies the belief that the thinking self is an ultimate substance (see Chapter VI, n. 58). 84  Pippin 1987: 473. 85  Kitcher 1990: 253. 86  Rosefeldt 2000: 167–8 suggests that we can theoretically cognize our freedom of thought but not our freedom of will because we cannot cognize whether our volition-­based actions are done from duty. However, this suggestion threatens to imply the Sidgwickian view that our actions are free only if they are done from duty—­an implication that Rosefeldt indeed seems to endorse (2000: 159, 165). This view is both philosophically implausible and textually indefensible (see Chapter III, n. 33).

Freedom of Thought as a Condition of Theoretical Cognition  249 we can recognize that this spontaneity is—­unlike the moral spontaneity of our will—­required for our actual determinate theoretical cognition(N) and knowledge of nature. For these reasons, we can know that we possess absolute epistemic spontaneity, but we cannot know (at least not without further, practical-­moral arguments) that we also possess the truly causal absolute spontaneity of a noumenal moral will.

VII.8.  Conclusion: What Does It Take to Reject Naturalistic Etiological Skepticism? In this chapter I have shown why Kant holds that we must regard ourselves as absolutely spontaneous subjects of theoretical cognition. I have argued further that this theoretical self-­conception qualifies as a priori self-­knowledge. My overall interpretation in Part 3 allows me to clarify how, precisely, Kant’s view relates to a contemporary line of reasoning that has been suggested (in different forms) by Schafer, Srinivasan, or White and that we first encountered in Section IV.6. This line of reasoning tries to disarm etiological debunking arguments regarding moral cognition through the claim that such arguments undermine, incoherently or absurdly, the possibility of theoretical cognition (e.g., of evolutionary causal processes) as well. I can now diagnose to what extent Kant agrees and disagrees with this line of reasoning. He agrees that a skeptical view which implies our lack of any cognition including theoretical cognition of nature is implausible. The argument for our knowledge of our cognitive spontaneity that I have reconstructed in this chapter begins from the unargued premise that we have some objective empirical cognition(N) of nature (cf. A93–4/B126–7)—a premise that is also accepted by Kant’s empiricist or naturalist opponents. Thus, however ambitious one may find Kant’s conclusions regarding our cognitive spontaneity, Kant’s view is not voraussetzungslos or intended as a response against a most radical skeptic who is prepared to accept that we cannot cognize(N) or know anything at all.87 For Kant, as for the abovementioned contemporary philosophers, such a radical skepticism is not worth engaging. (I return to this important point in Chapter VIII.) But there is also a significant point of disagreement between Kant and the contemporary critics of etiological skepticism. As I noted in Section IV.6, these critics claim that they can disarm etiological skepticism simply by exposing the implausibly radical implications that such skepticism has once it is fully generalized: because etiological skepticism about morality leads to the “absurd” view that we have no objective cognition at all, we can fully grant the naturalistic premises on 87  This is a major recurring theme in Ameriks’s work, to which I am indebted here. See Ameriks 2003, 2019.

250  Kant on Freedom and Rational Agency which etiological skepticism relies and yet block the debunking inferences that etiological skeptics seek to draw from these premises. As I also noted in Section IV.6, Kant would reject this view as unsatisfying because it does not engage with the reasons that support the inference from a naturalistic conception of the human mind to our complete lack of cognitive achievements. In Kant’s view, we cannot blithely concede that all our judgments are the causal upshot of contingent, non-­rational natural processes and yet insist that these judgments provide us with objective cognition. The conclusion that we lack any cognition, however absurd, can be established through a valid argument whose central premise is the naturalistic claim that all our mental activity is empirically conditioned.88 To be sure, this argument also undermines the specific causal (e.g., evolutionary) claims whose rational credentials naturalistic-­ etiological skeptics take for granted in the specific etiological premises of their skeptical reasoning. But this added complication only shows that naturalistic theorizers are pulling the rug from under their own feet. Reaching this diagnosis is fully compatible with conceding the validity of a genealogical skeptical argument that implies our lack of any including naturalistic cognition. One can endorse this argument in counterfactual terms, as showing only but significantly what would follow for our cognition if our mental faculties were constituted solely in some unspecified naturalistic fashion. For Kant, we can reject this argument as unsound only if we deny, first, its general etiological premise asserting the empirical origin of all our thinking and, second, the corresponding exclusively naturalistic picture of the human mind. That is: we can claim to possess objective theoretical cognition (including naturalistic cognition of how the phenomenal human mind evolved) only if we trace our a priori theoretical representations (which epistemically ground all our empirical cognition(N) and knowledge of natural phenomena) to our spontaneous noumenal self-­activity. But what about our belief that we have objective practical cognition and knowledge? And, what about the belief that we have the freedom of will or practical reason that would be required for objective practical cognition and knowledge? How does Kant seek to justify these beliefs, and how strong or certain is this justification? What, if anything, does Kant’s justification of the belief in moral freedom of will have to do with his justification of the belief in epistemic freedom of thought? I will address these and related questions in Part 4.

88  One might hold that this view is inconsistent because it presumes that we have valid cognition of our lack of cognition. However, this response is unsatisfying for various reasons (cf. Chapter IV). First, it cannot (by itself) restore confidence that we do have valid cognition, since it does not engage with the argumentative force of the skeptical reasoning. Second, one can—­like Kant—­endorse the skeptical argument from a non-­skeptical perspective (which does not deny that we have valid cognition) as valid but unsound (see the next paragraph).

PART 4

KAN T ’S J U ST IF IC AT ION OF T H E BE L IE F IN F RE E W I LL

VIII Kant’s Moral Grounding of Free Will In Part 3, I argued that our theoretical intellect has transcendental freedom of thought and that we can know this to be the case. This does not entail that we also have or that we can know to have transcendental freedom of will. The will, unlike the capacity for thought, is a power to produce actual external objects (as opposed to mere representations of actual objects). Thus, the idea that we have absolute freedom of will implies that we have a causal power to change the external world and to direct the course of our lives as we see fit, without being determined to do so by foreign causes such as external objects, empirical desires, or brain states. This is clearly a metaphysically loaded, controversial assumption. In this chapter I examine the doxastic status this assumption has in Kant’s post-­1785 view as it emerges (especially) in the Critique of Practical Reason.

VIII.1.  The Parity Thesis and Kant’s Moral Proof of Free Will Free will, God, and immortality of the soul are the three objects of traditional special metaphysics. On a common reading, a central legacy of Kant’s philosophy is that we cannot know whether these objects exist: “traditional (transcendent) metaphysics, which [Kant] sees concerned with God, freedom, and immortality . . . deals with topics with respect to which knowledge is not possible for us.”1 On this view, Kant allows only for a legitimate Glauben (typically translated as “faith” or “belief ”; I shall always use “faith”) that these objects exist. With regards to assertions that postulate the existence of these objects, “Kant holds that it is rational for us to believe in their truth even though they lie completely outside the reach of human knowledge.”2 To assess this view, we first need to consider Kant’s notion of faith. Kantian faith is not only (like Kierkegaardian faith) a firm, unwavering (“­subjectively sufficient”) state of holding some proposition to be true. It is also (unlike Kierkegaardian faith) a rational conviction: finite agents have a rational basis for believing (e.g.) that God exists. Rational faith falls short of knowledge because it is based on subjective reasons that do not guarantee the truth of what one believes. Thus, beliefs in the mode of faith lack the “objective sufficiency” that is required 1  Allais 2015: 5. Compare Bird 2006: 11; Wood 1978: 16. 2  Willaschek 2010: 169 (his emphasis).

Kant on Freedom and Rational Agency. Markus Kohl, Oxford University Press. © Markus Kohl 2023. DOI: 10.1093/oso/9780198873143.003.0009

254  Kant on Freedom and Rational Agency for knowledge (A822/B850). Faith in the truth of an assertion is not supported by any considerations (such as perceptual evidence, testimony, or a rational proof) which certify that the assertion is or even must be true as a matter of objective fact. Hence, legitimate faith always involves a “consciousness of . . . the possibility of the opposite” (JL, 9:66). The reasons supporting faith are subjective in a rather special sense. The relevant notion of subjectivity here is not the kind of empirical, psychological subjectivity that entails arbitrariness, privacy, and contingency (cf. Chapter IV). Suppose my psychology happens to be such that I can only satisfy my desire to win some competition if I firmly believe that I will win. This is not a sufficient basis for adopting rational Kantian faith that I shall win.3 Kantian faith is based on two different kinds of non-­arbitrary, intersubjectively valid reasons. First, such faith can be adopted only if it is rationally connected to morally required ends; second, it rests on shared subjective needs that reflect our common rational finitude, namely, the limits of human reason as such. It is a matter of debate how, precisely, Kant conceives the rational structure of moral faith. Generally speaking, we must adopt rational faith that p if we must believe that p in order to act under objectively valid practical norms. The necessity to believe that p here does not derive from the intentional object of belief (from p itself; cf. KpV, 5:11) but from the rational constitution of the believer, namely, from a need imposed by the believer’s finite hence “subjective” rationality (“Bedürfniß der reinen Vernunft”; KpV, 5:4; cf. KpV, 5:125): I cannot regard myself as acting rationally under a valid moral norm unless I hold the descriptive proposition that p to be true. In the case of God and immortality, the story goes roughly as follows. We have a moral duty to promote the highest good. This is a state of affairs where there is a maximum degree of moral virtue that corresponds to a maximum amount of happiness and where the correspondence between virtue and happiness is necessary (non-­accidental): for every agent, their amount of happiness is conditioned by their degree of virtue. Now, we cannot rationally promote the highest good unless we believe that this state of affairs is really possible.4 And, we cannot recognize the highest good as a real possibility unless we assume (first) that our souls are immortal so that we can reach the highest degree of virtue and (secondly) that an omnipotent, omniscient God exists who ensures that every finite agent has precisely the amount of happiness that they morally deserve according to their 3  Willaschek (2010: 193–5) suggests that this is an instance of the rationally permitted pragmatic faith that Kant considers in the Canon, which does include “arbitrary and contingent ends” (A823/ B851). But this notion of faith seems too arbitrary to fit naturally into Kant’s systematic conception of faith. Accordingly, in Kant’s later works the notion of practical faith becomes identical with moral faith; his systematic accounts of faith in the second and third Critique (cf. KU, 5:469) do not mention arbitrary pragmatic faith anymore. Kant now stresses that faith as such must be connected to moral duty (KU, 5:472). 4  For helpful discussion of some complications regarding this point, see Willaschek 2010: 180–3.

Kant ’ s Moral Grounding of Free Will  255 degree of virtue. Hence, in order to act rationally under the moral duty to promote the highest good we must assume that God exists and that we are immortal. But we have no objective certainty and hence cannot know that these assumptions are in fact correct. Kant does not directly connect the belief in free will to the highest good. Those who hold that Kant licenses only rational faith in free will might take him to argue roughly as follows. We cannot regard ourselves as acting rationally under valid moral principles unless we believe that we have a capacity for absolutely spontaneous choice that elevates us above mere automata (“turnspits”). But for all we can know, our subjectively rational self-­image as free moral agents might be ­illusory—­we might be mere automata as a matter of objective fact. As Allison puts it, for Kant “the idea of spontaneity is inseparable from . . . our conception of ourselves as . . . agents; but this does not license any ontological conclusions, since it leaves in place the epistemic possibility that we are nothing more than automata.”5 On the common interpretation, we are entitled only to subjectively rational faith with respect to free will, God, and immortality. This implies that our belief that we are free moral agents is epistemically on par with our beliefs that we have immortal souls and that God exists. I call this the “Parity Thesis.”6 One major goal of this chapter is to show that the Parity Thesis is false. The Parity Thesis is compatible with Kant’s frequent appeal to our practical cognition of freedom (e.g., at KpV, 5:30). This is because “cognition” [Erkenntnis] and “knowledge” [Wissen] are different concepts.7 Cognitions need not have the propositional structure that is required for knowledge qua certain assent to a proposition. Even those cognitions that do have a propositional content may not yield knowledge because propositional cognitions can also have the status of opinion or faith (JL, 9:66–7). For instance, our “practical cognition” of God grounds only “faith” that God exists (KU, 5:467). One might suggest that the function of our practical cognition of freedom is not to justify (as knowledge) the belief that freedom exists but, rather, to provide a determinate representational content for our idea of freedom as a supersensible causality.8 In Chapter III, I gave a detailed account along those lines: I showed how we can form a determinate-­ positive conception of our free noumenal causality via our practical cognition of the moral law. But this account has no obvious implications for the epistemic

5  Allison 1996: xiv. See also Wood 1974: 36–7, whose reading I discuss in Section VIII.3. 6  The Parity Thesis is endorsed by Allais 2015: 5; Chignell 2007b: 354–7; Della Rocca 2008: 287; Frierson 2003: 100; Kuehn 2001: 312–13; McLear 2020; Pasternack 2011: 309; Pereboom 2005; Rauscher 2002: 493; Walsh 1975: 4–5, 197; Willaschek 2010: 169, 193 and 2018: 271–2; Wood 1974 (see Section VIII.3 below). Hogan (2010: 29) seems to accept a different Parity Thesis: we have practically grounded knowledge not only of freedom but of God and immortality as well. Dissenters from the Parity Thesis include Carnois 1986; Kain 2010; and Schafer 2018. 7  See Chignell 2014; George 1982; Kohl (forthcoming3); Watkins and Willaschek 2017a, 2017b. 8  See Allais 2015: 304.

256  Kant on Freedom and Rational Agency credentials of the belief in free will; it does not show that our moral cognition of free will yields knowledge rather than faith. My evidence for denying the Parity Thesis derives from various key texts. Of particular importance here is the Critique of Practical Reason. In its Preface (at KpV, 5:3–4) Kant distinguishes between the assent (A1) we may give to the proposition that we have free will and the assent (A2) we may give to the propositions that God exists/that we are immortal. The “reality [Realität]” of freedom “is proven through an apodictic law of practical reason”; “freedom is actual [wirklich], for this idea reveals itself through the moral law.” But “freedom is also the only among all ideas of speculative reason whose possibility we know [wissen] a priori . . . because it is the condition of the moral law, which we know. However, the ideas of God and immortality are not conditions of the moral law” (cf. KU, 5:474). They are only conditions for complying with the specific duty to pursue the highest good. Hence, (A2) is based on “a merely subjective ground of assent, but one which is objectively valid for . . . practical reason.” This means, roughly: since we cannot regard ourselves as rationally pursuing the highest good unless we believe in God and immortality, we have a subjective practical need to adopt those beliefs. These beliefs are “objectively valid” for practical reason because they are legitimized (in part) by objective practical laws: we can have faith (or “trust”; KU, 5:472) in the truth of assumptions which we (subjectively) must adopt so that we can comply with objective moral commands (KpV, 5:145; cf. A828–30/B856–8; KpV, 5:122–32; KU, 5:450–1).9 But for all we can know, the beliefs in God and immortality might yet be mistaken. By contrast, since we can know the possibility and prove the actuality of free will, we need not confine ourselves to a merely subjective ground of assent with regards to (A1) that could yield only faith. The ensuing argument of the second Critique confirms these prefatory remarks: Kant provides a moral deduction of freedom but not of God or immortality (KpV, 5:47–8). Later on, he recapitulates that thanks to this deduction freedom could be “assertorically cognized” so that the “actuality of the intelligible world could be given to us” whereas “we were unable to take such a step with regard to the idea of a necessary being” (KpV, 5:105), i.e., with regard to God.10 9  Using the term “objective validity” for assumptions that are based on a “subjective . . . rational necessity” (KpV, 5:11) seems odd, but the fact that these assumptions are also grounded in objective moral commands might explain Kant’s usage. His terminology here is admittedly a bit messy, partly because of the general ambiguities in his notions of objectivity/subjectivity (cf. Chapter IV) and also because of his tendency to (sometimes merely implicitly) index terms like “objective validity/reality” to theoretical or practical contexts of use. 10  One might suggest that the comparatively modest status Kant assigns to the idea of God here pertains only to the purely theoretical, deistic conception of God as a necessary being, i.e., to the dynamical idea of God that figures in the first Critique’s Dialectic. However, when Kant says that we “were unable to take” the step of assertoric cognition for God he clearly refers back to his earlier argument in the second Critique’s Analytic; thus, he stresses a difference between the ideas of free will versus God in our practical thinking. Accordingly, he adds (at KpV, 5:105; cf. KpV, 5:48) that the idea of freedom, in contrast with the theistic idea of God, “is immanent in a practical respect.”

Kant ’ s Moral Grounding of Free Will  257 What does Kant mean when he says that we know the “possibility” of freedom? Kant generally distinguishes between the “logical” and the “real” possibility of a concept (Bxxvii; A244/B302–3). The logical possibility of a concept requires only that its various predicative marks are logically consistent; the concept “non-­ extended body” fails whereas the concept “rapping unicorn” meets this condition. But the mere logical possibility of the idea of freedom cannot be at issue here (at KpV, 5:3–4): Kant stresses that freedom is the “only among all ideas of speculative reason whose possibility we know a priori,” and we surely know that the ideas of God and immortality are logically possible as well (see, e.g., A596–7/B624–5). Thus, Kant’s point must be that through the moral law we know the real possibility of free will. This shows that the second Critique goes beyond the idealist defense of free will that Kant provided in the first Critique. In Chapter  I, I argued that Kant’s idealism removes a metaphysical obstacle that would decisively rule out the real possibility of free will. Transcendental idealism restricts the scope of the deterministic causality of nature to sensible appearances and posits a supersensible reality that is not subject to the causality of nature. But this yields only an indeterminate ontological framework which does not positively establish the real possibility of determinate objects such as noumenal free will (A558/B586). In the second Critique Kant recapitulates that theoretical reason could establish the concept of freedom merely “problematically, as not impossible to think” (KpV, 5:3). This is imprecise (cf. Chapter I): the possibility to think the concept of freedom poses no problem at all. Kant means: theoretical reason, via its appeal to transcendental idealism, shows that the belief in our noumenal free will does not contradict our knowledge that we (as phenomena) are causally determined (A538–40/B566–8; GMS, 4:455). Theoretical reason shows this only “problematically” because it provides no positive reason for believing in the real possibility of free will. But it clears the path for practical reason, which supports the assertion that free noumenal causes are really possible: “the possibility [of freedom], which previously was only a problem, now becomes assertion” (KpV, 5:5). Kant anticipates this development in the first Critique’s B-­edition when he notes that our basis for ascribing real possibility to a concept “can . . . lie in the practical sources of cognition” (Bxxvii).11 A further complication here is that the second Critique seems to collapse the difference between the real possibility and the actuality of freedom: Kant equates the claim that through the moral law we know the real possibility of freedom with the claim “that freedom is actual” (KpV, 5:4) since the moral law “proves not only the possibility but the actuality” of freedom (KpV, 5:47; cf. KpV, 5:133). This equation of real possibility and actuality occurs in other contexts as well (cf. Bxvii; 11 Kain 2010 also stresses the importance of Bxxvii for the argument structure of the second Critique.

258  Kant on Freedom and Rational Agency A244/B302–3). In the case of free will, it might be due to the status of free will as a fundamental power (Grundkraft). Generally speaking, the real possibility of x can only be ascertained from something actually given. Since we cannot cognize the real possibility of fundamental (as opposed to derivative) powers from our knowledge of other, more basic actual powers (RGV, 6:31; A449/B577), we must derive their real possibility from considerations that attest to their own actual existence (KpV, 5:47–8). In the case of empirical fundamental powers, knowledge of their real possibility requires sensations of effects through which such powers are “given” to us as actual.12 But sensible experience cannot attest to (“give” us) the actual efficacy of a supersensible power such as free will (KpV, 5:47; GMS, 4:459). Instead, our a priori consciousness of the moral law proves the actuality (and objective reality) of free will (KpV, 5:3–4, 47): through the moral law, free will “is given as actual in the subject” (FM, 20:292). Thereby, the “actuality of the intelligible world” is also “given to us” (KpV, 5:105). Since what is actual is (a fortiori) also really possible (cf. A460/B488), the proof that free will is actual entails that free will is really possible. Kant makes another strong claim about freedom in the second Critique: “. . . one would never have committed the daring deed of introducing freedom into science had not the moral law, and with it practical reason, come in and thrust this concept upon us” (KpV, 5:30). If freedom belongs to science [Wissenschaft], then it must be an object of knowledge [Wissen]. Of course freedom cannot belong to theoretical science. But in Kant’s view “the properly and objectively so-­called sciences” include “the practical cognition of reason” (Bix–­x).13 These strong claims about freedom are not limited to the second Critique. The Jäsche Logic echoes the second Critique argument for the special status of the idea of freedom: “One cannot provide for the objective reality of any theoretical idea, or prove that idea, except for the idea of freedom, and this is because this idea is the condition for the moral law, whose reality is an axiom,” whereas we can only “act as if there was a God” (JL, 9:93).14 Likewise, in his (1793–5) Preisschrift concerning the progress of metaphysics Kant lists three articles of moral-­practical faith: the belief in God, the belief in the highest good, and the belief in a future eternal life. These beliefs rest on a subjective-­practical ground that provides no

12  Here I am indebted to the insightful discussion in Warren 2001: 85–7. 13  Willaschek (2018: 44) concedes that ethics for Kant has the status of a science but denies that the belief in freedom can be part of that science (or can be a matter of knowledge). This is in direct conflict with KpV, 5:30. 14  This passage is further evidence (adding to my argument in Chapter I) against the “as if ” reading regarding freedom that has been proposed by Korsgaard (1996a: 162, 176). Her reading is chiefly based on GMS, 4:448, where Kant does seem hesitant about asserting the actuality of freedom. But such passages can be read as stressing only the theoretical inexplicability of freedom (see Ameriks 2000a: 194; 2003: 169–71). Moreover, the Groundwork might not incorporate Kant’s considered views on freedom, if (following Ameriks 2000a: 226) there is “a great reversal” in Kant’s practical writings after 1785. See n. 48 and Section VIII.4 for further discussion.

Kant ’ s Moral Grounding of Free Will  259 “objective instruction concerning the actuality of the objects” of those beliefs; this subjective ground entitles us only “to act as if we knew that these objects were actual” (FM, 20:298). Freedom of will is not included among these three “as if ” articles of moral faith.15 A further highly relevant text is §91 of the third Critique where Kant lists three kinds of “cognizable things”: matters of opinion, matters of fact [scibile], and matters of faith (KU, 5:467). He stresses that there is “even one idea of reason” that belongs to the matters of fact, “namely, the idea of freedom, whose reality, as a special form of causality . . . can be proven through practical laws of pure reason” (KU, 5:468; cf. KU, 5:474). By contrast, the highest good, God, and immortality are “a mere matter of faith for pure reason.” The cognition of God and immortality as conditions of the highest good yields “neither knowledge nor opinion” but “faith” (KU, 5:470). The triad of objects of doxastic attitudes that Kant lists here, namely (1) matters of opinion, (2) matters of fact, and (3) matters of faith, corresponds to his canonical triad of doxastic attitudes: opinion, knowledge, and faith. Since freedom is classified as (2) a matter of fact, it follows that our assent to the proposition that we are free has the status of knowledge. Proponents of the Parity Thesis might respond that freedom, God, and immortality are postulates of practical reason. They might argue that the very notion of a practical postulate entails an attitude of mere faith that rests on subjective needs of finite reason.16 However, in his account of the postulates Kant does not trace the assumption of free will to subjective needs of reason. Instead, he repeats his earlier claim that “the reality of freedom is demonstrated through the moral law” (KpV, 5:133). While he dedicates special sections to explaining why we have a subjective need to assume the existence of God and immortality in relation to the highest good (KpV, 5:122–32), he offers no corresponding “subjective need” explanation for the belief in free will. Still, this leaves the question of why Kant classifies freedom as a postulate alongside God and immortality. I suggest that in Kant’s view there are two different kinds of practical postulates.17 In a generic sense, something qualifies as a 15  Kant clearly views free will as a matter of faith in texts before the mid- to late 1780s, i.e., before he develops his considered view on morality and freedom. These texts include the first Critique’s Canon of Pure Reason and some lecture transcripts (e.g., V-­Th/Volck, 28:1192). I discuss how this bears on my reading in Section VIII.4. 16  See Willaschek 2018: 271–2; Wood 1974: 138, 146. 17  Another common suggestion (cf. Allison 2020; Carnois 1986; Kain 2010) is that Kant operates with two different concepts of freedom in (1) the second Critique’s Analytic and (2) the Dialectic’s Postulates section. On this view, the kind of freedom (1) whose reality is proven by the moral law is freedom as autonomy whereas the kind of freedom (2) that remains a mere postulate is freedom qua “autocracy” of reason, as a capacity for obeying the moral law (Kain 2010: 225) or as a necessary condition for the attainment of virtue (Allison 2020: 408–9). This suggestion is implausible for numerous reasons. First, Kant never mentions freedom as autocracy in the relevant Dialectic section—­instead, he repeats his earlier definition of free will as autonomous, causally unconditioned self-­determination according to moral laws (KpV, 5:132–3; cf. KpV, 5:3–4). Second, he stresses that the idea of freedom which is at issue in the Dialectic is precisely the one that has already been proven via the moral law in

260  Kant on Freedom and Rational Agency practical postulate when our belief in its possibility is based on pure practical reason (without evidential support from theoretical reason or experience). Some practical postulates have a quasi-­mathematical status because they are apodictically certain. Moral laws themselves are postulates in this sense: they are “to be found improvable and yet apodictically certain, akin to mathematical postulates,” and this epistemic property allows them to “irrevocably display” (MS, 6:225) or “reveal” (KpV, 5:3) our free will. Kant does not mention God and immortality in this context. He defines another species of practical postulates in sharp contrast to mathematical postulates: something S is postulated when the belief in S lacks apodictic certainty but is subjectively necessary for finite moral agents who seek to follow moral laws. What is postulated in this subjective sense is the possibility of “God and the immortality of the soul”; Kant does not mention freedom in this context (KpV, 5:11). Since moral laws themselves are postulates in the first non-­ subjective sense that involves apodictic certainty, and since freedom is indubitably “proven through practical laws of reason” (KU, 5:468), the belief in free will belongs to the non-­subjective species of practical postulates as well.18 I have provided some strong textual evidence that in Kant’s post-­1785 view we can know that we have noumenal free will. The remainder of this chapter is devoted to a reconstruction of how Kant argues for this view. In this reconstruction, I will consider various objections to the claim that Kant truly does, or can, allow for human knowledge of noumenal free will. As we saw, Kant claims that we know our free will because free will (unlike God and immortality) “is a condition of the moral law . . . which we know” (KpV, 5:4; cf. JL, 9:93). More precisely: free will is a ratio essendi, a necessary real ground or condition of the existence of the moral law. The moral law is the ratio cognoscendi, the epistemic ground of our cognition of free will. I propose that Kant’s view can be understood in terms of the following argument:

(I) If we know that the existence of x is a necessary real condition for y, and if we know y, then we also know (inferentially) the existence of x. (II) We know the moral law. (III) We know that our noumenal freedom of will is a necessary real condition for the moral law. (IV) Hence, we know the existence of our noumenal freedom of will.

the Analytic. Third, if the actuality of freedom as autonomy has already been proven (in the Analytic), there is no need for a further step (in the Dialectic) where one postulates the possibility of freedom as autocracy or the possibility of virtue: the capacity to autonomously determine one’s will under moral duties essentially includes the capacity to acquire a volitional state of autocracy and virtue (cf. MS, 6:383). 18  For a related distinction between two senses of “practical postulates,” see Wolff 2009: 522–4.

Kant ’ s Moral Grounding of Free Will  261 Since (I) is both clear and rather uncontroversial, I shall not discuss it any f­ urther.19 Instead, I will focus on (II), (III), and (IV). Each of these claims invites a different type of objection. Some deny (II) that we know the moral law or that we know the moral law in the normatively thick sense that would entail knowledge of free will. I discuss this issue in Section VIII.2. Others might be sympathetic to (II) but reject (III): even if we know the moral law in the fullest sense (as normatively binding us to certain duties), this does not entail knowledge of free will because there is no reason to think that our knowledge of moral duties presupposes that we actually have free will. I discuss this view (which rests on the Parity Thesis) in Section VIII.3. Finally, for some (IV) may seem incompatible with Kant’s critical epistemology. I address this worry in Section VIII.4.

VIII.2.  Kant’s Conception of Moral Knowledge VIII.2.1.  Moral Assent: Neither Opinion nor Faith Since knowledge entails truth, (II) requires first of all that moral judgments are truth-­apt, i.e., that our moral discourse does not reduce to emotive expressions of feelings or pro-­attitudes that lack genuine cognitive content. Kant clearly accepts this. He distinguishes between objective judgments of cognition that are based on reason and subjective judgments that are based on mere feeling; moral judgments paradigmatically belong to the former class (KU, 5:214–17; cf. Chapter  IV). Moreover, Kant refers to the “truth” of the categorical imperative (GMS, 4:444–5) and holds that moral judgments can be false (MS, 6:401). But can we know the truth of moral judgments in Kant’s view? The idea that Kant allows for genuine knowledge of normative-­practical propositions may seem odd if one assumes that knowledge is solely a matter of theorizing, when our descriptive beliefs seek to match the world, but has no place in our practical efforts to make the world correspond to our will. But this assumption is not forced upon us since we can regard practical reason as a cognitive power which guides our practical efforts to change the world by providing us with knowledge of normative truths about how we ought (not) to act and live. Still, the assumption that Kantian knowledge is confined to theoretical propositions is widely accepted. It is well captured by W. H. Walsh’s statement that “the cognitive powers of reason” just are “the powers of theoretical knowledge of the world of sense.”20 19  Some (e.g., Nozick 1983) might doubt (I) because of concerns about the principle of epistemic closure, to which Kant is here implicitly committed (as Proops notes: 2021: 312). But this principle is rather widely accepted by early modern philosophers (e.g., Descartes) and contemporary epistemologists (see e.g., Hawthorne 2003; Neta 2003). 20  Walsh 1975: 1. Smith (2003) also excludes morality from the domain of possible knowledge. Ameriks (2013: 155) claims that for Kant all knowledge “in a strict sense . . . requires proof that is

262  Kant on Freedom and Rational Agency This assumption is especially prevalent among commentators who discuss Kant’s account of doxastic attitudes.21 I shall therefore examine this account in some detail. Kant provides an exhaustive threefold classification of doxastic attitudes: whenever we hold a judgment to be true, we do so either in the mode of opinion, faith, or knowledge (A822/B850; JL, 9:66). If a person has an opinion that p, they are not fully convinced that p is the case (their assent lacks “subjective sufficiency”) and their judgment is not conclusively supported by objective reasons either (opinions also lack “objective sufficiency”). Since opinions are assents to probabilistic judgments that rely on varying degrees of empirical evidence, a priori judgments of pure reason, including moral judgments, can never be opinions: opining “occurs in no sciences that contain a priori cognitions, thus neither in mathematics, nor in metaphysics, nor in morality” (JL, 9:67; cf. A823/B851). Hence, those who deny that Kant’s concept of knowledge applies to moral judgments must claim that such judgments are a matter of faith.22 However, the different species of faith that Kant considers do not include moral judgments. This is clear in the case of what he calls doctrinal and pragmatic faith; more importantly, what he calls “moral faith” does not comprise prescriptive moral judgments either. Rather, moral faith pertains to descriptive metaphysical claims made in a moral context, such as the claim “that there is a God and a future world” (A828/ B856). Moral faith concerning what is “is grounded upon a knowledge with regard to laws” (Refl., 16:392) concerning what ought to be. Moral ought-­ judgments are not mere matters of faith: “Matters of faith are . . . not objects of the cognition of reason . . . neither of the theoretical cognition of reason (for instance in mathematics . . .), nor of the practical cognition of reason in morality . . . regarding rights and duties, there can be . . . no mere faith” (JL, 9:69–70). Since all our assents to propositions are either knowledge, faith, or opinion, and since our assents to moral judgments are neither faith nor opinion, it follows that such assents (if true and justified) yield knowledge. This includes our assent to the highest principle of morality, “the moral law, which we know” (KpV, 5:4). One might respond that concepts such as “knowledge” take on a special non-­ epistemic, pragmatic meaning in the context of Kant’s practical philosophy.23 However, it is unclear what that special meaning would be or how a non-­epistemic pragmatic knowledge state is supposed to differ from practical faith. Moreover, when Kant says that “in mathematics . . . as in metaphysics and morality, it is essential either to know or not to know” (JL, 9:67), he clearly uses the term “to grounded only on a priori theoretical considerations.” But see Engstrom 2009 for a detailed account of “strict” practical knowledge. 21  See Chignell 2007a: 327; 2007b: 39–44, 47; Pasternack 2011: 307–14; Stevenson 2003: 88. 22 See Bird 2006: 237–9; Chignell 2007a: 356–7; Pasternack 2011: 311; Rauscher 2002: 486–92, 498–9. 23  Baiasu 2013a attributes this view to Chignell 2007a, 2007b.

Kant ’ s Moral Grounding of Free Will  263 know” univocally. Here, as elsewhere (cf. A823/B851; Bix–­x), Kant refers to morality as one domain of a priori knowledge (Wissen) and science (Wissenschaft), namely, as one of the “pure rational sciences” alongside mathematics whose a priori rational principles cannot allow for any degree of uncertainty (A480/B508). The fact that Kant’s generic concept of knowledge applies univocally across theoretical and practical domains is compatible with the fact that practical knowledge has distinctive features. As we saw, practical unlike theoretical knowledge can be a cause of objects (A92–3/B124–6; KpV, 5:42–6).24 Likewise, we do not seek moral knowledge for its own sake but for the purposes of acting well: “it does not only matter to know what it is one’s duty to do” (MS, 6:375)—one’s moral knowledge must also lead to dutiful action. What do assents to moral propositions have in common with assents to theoretical propositions so that both (may) qualify as species of knowledge? In Kant’s view, knowledge has two features that jointly distinguish it from faith and opinion (A822/B850). First, knowledge (unlike opinion) involves “subjective sufficiency,” which means that one holds a proposition to be true with the fullest degree of conviction that leaves no room for sincere doubt. Our practical assents to moral propositions carry an “irrepressible” (KpV, 5:35) degree of conviction that is fully on par with our conviction that every phenomenal event has a cause or that two plus two equals four. Second, knowledge (unlike both opinion and faith) involves “objective sufficiency.” This requires that one’s assent be based upon conclusive, objectively-­universally valid reasons that are fully communicable to other thinkers “regardless of the difference among the subjects” (A821/B849; cf. A829/B857; JL, 9:70). Our assents to moral rules satisfy this condition: a correct moral rule is “objectively and universally valid,” “valid without . . . subjective conditions that distinguish one rational being from the other” (KpV, 5:21) since the moral law applies to “every one . . . who has reason and will” (KpV, 5:36). The objective sufficiency of a knowledge-­state is also called “certainty (for everyone)” (“die objektive [Zulänglichkeit] [heißt] Gewißheit für jedermann”; A822/B850). The emphasis on certainty is central to Kant’s conception of knowledge. Although (as we saw) cognition as such does not entail knowledge, the entailment holds for cognitions that involve an awareness of certainty: “to know something means nothing other than to cognize it with certainty” (V-­LO/Phil, 24:242), since “the word knowledge [Wissen] is the . . . stem of certainty [Gewissheit]” (V-­LO/Phil, 24:227–8). For a priori knowledge, the certainty of a knowledge-­state involves an awareness of necessity: a “certain holding to be true” is connected with “the consciousness of necessity” (JL, 9:66). In our moral judgments about how agents rationally speaking must act we have “precisely such certain cognitions, indeed wholly a priori, in practical laws” as in well-­grounded 24 For detailed discussion of what is distinctive about practical knowledge, see Engstrom 2009: 118–28.

264  Kant on Freedom and Rational Agency theoretical propositions (JL, 9:67). Since the moral law is “apodictically certain” (KpV, 5:47) and since “knowledge is an apodictic judging” (JL, 9:66; cf. Refl., 16:386–7), we know the moral law. What I have argued thus far yields a strong textual case for Kant’s commitment to (II). But the fact that so many commentators are opposed to acknowledging (II) as Kant’s considered view imposes the need for further consideration. In the remainder of Section VIII.2, I will consider three worries that one might raise against the possibility of objectively certain moral knowledge.

VIII.2.2.  Moral Knowledge and Kant’s Epistemology According to the first worry, there are central tenets of Kant’s epistemology which rule out the possibility of moral knowledge. This worry seems to be based on something like the following argument. All our assents to moral judgments ultimately rest on the moral law. But, despite of what Kant says (at KpV, 5:4), the moral law itself cannot be known since the moral law cannot (according to the second Critique) be derived from any further ground, i.e., cannot be supported or justified through a proof or deduction. If Kant’s account of knowledge imposes the general requirement that every knowledge claim must be grounded in some further epistemic basis that is prior to the knowledge claim itself, then there can be no moral knowledge.25 But Kant’s account of knowledge does not incorporate this requirement. The “certainty for everyone” that is required for knowledge need not be based upon further cognitive grounds (on pain of a potentially vicious regress or circle).26 In addition to the “mediate” certainty that is based on some further ground or proof, “there must also be something indemonstrable or immediately certain, and our entire cognition must proceed from immediately certain principles” (JL, 9:71; cf. V-­LO/Phil, 24:226). Synthetic a priori principles that can be cognized immediately without proof play a crucial role in Kant’s philosophy of mathematics where they are designated as axioms (A732/B760). As we saw (cf. JL, 9:93), for Kant the reality of the moral law is an axiom as well. Moral principles are “akin to mathematical postulates” since they are “improvable and yet apodictically certain” (MS, 6:225).27 These characterizations relate to Kant’s account of the moral law as a “fact of pure reason” (KpV, 5:47) “of which we become immediately conscious” 25 See Baiasu 2013b: 40; Chignell 2007a: 356–7; Grenberg 2014: 42, 56, 272–3; Pasternack 2011: 311. 26  While Kant defines knowledge as assent “from a ground of cognition” that has objective sufficiency (JL, 9:70), he does not demand that the ground of a knowledge claim must be prior to that claim itself. The objective ground of cognition for assenting to p might just be a rational awareness of the content of p itself. 27  “Axiom” and “Postulate” are tightly connected terms (cf. Wolff 2009: 514). See (e.g.) V-­LO/Phil, 24:231–2: for any improvable yet absolutely certain cognition, one “postulates” agreement; such

Kant ’ s Moral Grounding of Free Will  265 (KpV, 5:29), which cannot be proven but which also “needs no justifying reasons” (KpV, 5:47). Hence, the moral law exemplifies a type of principle that is firmly engrained in Kant’s general epistemology: an axiomatic principle that can be cognized with “immediate certainty” since it “is neither capable of nor in need of a proof ” (JL, 9:71). As Marcus Willaschek explains, principles of this type are such that “their truth becomes clear from themselves, as soon as we have cognized them distinctly” where such distinct cognition immediately commands assent.28 The ascription of axiomatic status to the moral law does not conflict with Kant’s point in the Doctrine of Method (A732–3/B760–1) that only mathematics contains axioms: here he seeks to contrast mathematics with pure reason “in its merely speculative employment” (A736/B764). This leaves open the possibility that in its practical employment pure reason contains first axiomatic principles. The epistemic significance of axiomatic principles in mathematics is that these axioms make mathematics possible as a science because they are general principles from which all more specific mathematical cognitions can be systematically proven. A science, for Kant, requires the systematic unity of various specific cognitions under common general principles (see, e.g., Prol, 4:306). Such systematic-­ scientific unity is possible in the domain of practical reason (only) because “all imperatives of duty can be derived from this one [categorical] imperative as a principle” (GMS, 4:421) so that we can “measure out the entire scope of rational cognition of this kind” (GMS, 4:412) in a metaphysics of morals which is “an indispensable substrate of all theoretically sound and definite cognition of duties” (GMS, 4:410). This cognition yields a system of moral duties (MS, 6:205). Our assent to a principle which grounds a rational science must itself qualify as knowledge: “from knowledge [Wissen] comes science [Wissenschaft], which is the epitome of cognition as a system” (JL, 9:72). Thus, our assent to the moral law yields knowledge because it uniquely enables “the practical cognition of reason” as “one of the properly and objectively so-­called science” (Bix–­x): the moral law is known as the rational ground of moral science qua unified system of a priori practical cognitions (JL, 9:67; A480/B508).29 The morality-­mathematics analogy has its limits. Mathematical axioms are intuitively certain because they can be constructed in pure intuition (A733–4/ B761–2). By contrast, the practical cognition of the unconditionally commanding

cognitions are “Axiomata.” Here Kant invokes his mathematical notion of postulate that involves apodictic certainty (cf. Section VIII.1). 28  Willaschek 2018: 30–1. He shows that this is a common, uncontroversial view in seventeenth-­ century German philosophy. 29  To confirm that Kant’s oft-­repeated classification of morality as a science is not a mere facon de parler, note that he also characterizes the doctrine of duties as more than a mere “aggregate” (MS, 6:375) since it is systematically organized according to rational principles. This is his standard definition of science (JL, 9:72; A64–5/B89–90).

266  Kant on Freedom and Rational Agency moral law is “grounded on no, neither pure nor empirical intuition” (KpV, 5:31).30 Mathematics describes what is, namely the formal intuitive structure of sensible space, whereas morality cognizes what ought to be (cf. A547/B575). One might claim that for Kant all objective cognition as such requires some sensible-­intuitive content; on this basis, one might suggest that the moral law owes its objective cognitive content to moral feeling.31 However, this suggestion does not accord with the fact that moral feeling (humiliation by and respect for the law) is both causally and epistemically posterior to the certain recognition of the moral law as the sole fact of pure reason (KpV, 5:75–6). This feeling is produced through “an intellectual ground” (KpV, 5:73), namely, through our purely intellectual awareness of the moral law. We feel humiliated by and respectful towards the moral law due to our prior recognition that this law has unconditional normative authority over all the sensible incentives that make up our “self-­ love.” It is by casting down our empirical feelings as normatively inferior that the moral law has an a priori influence on our faculty of feeling (KpV, 5:74). Since the resulting moral feeling is only the “effect of the consciousness of the moral law” (KpV 5:75), i.e., of our rational awareness that the moral law has objective normative authority over our self-­love, the moral law cannot owe its objective normative content to moral feeling. Kant restricts the condition that objective cognition requires sensible content to theoretical cognition (FM, 20:273). Practical cognition of the moral law must be exempt from this condition because otherwise this law would lose its unrestricted normative scope, i.e., its claim to a truly unconditional validity. The fact that the moral law has a purely intellectual, entirely non-­sensible cognitive-­ normative content ensures that it applies to all rational beings including a divine intelligence whose cognition is unrestricted by any subjectivizing cognitive constraints that affect our finite nature, such as sensible intuition or moral feeling (cf. KpV, 5:79; see Chapters III–­IV).32 This concludes my response to the worry that Kant’s appeal to moral knowledge violates core tenets of his epistemology. I now consider a second worry against (II).

30  I thus cannot follow Wolff ’s claim that there is an “exact analogy” between mathematical construction and deed of pure reason that effects our consciousness of the moral law (2009: 532–3). The notion of “practical construction” (also employed by Allison 2020: 373) is, in my view, a very misleading (thus a bad) metaphor. 31  See Schönecker 2013: 2. Eric Watkins has also suggested this view in correspondence. 32  This does not sever all connection between (our finite) objective moral cognition and sensible intuition. While the moral law itself as “the sole fact of pure reason” is cognized apart from all sensible conditions, the application of this formal law in specific moral judgments requires the “material” input of sensibility, e.g., when we cognize whether or not it is morally permissible to satisfy some sensible inclination or to perform some empirical action.

Kant ’ s Moral Grounding of Free Will  267

VIII.2.3.  Against a Deflationary-­Theoretical Interpretation of “Moral Knowledge” According to this second worry, my discussion thus far has passed over a crucial ambiguity in the claim (II) that we know the moral law. Some commentators distinguish between a theoretical representation of moral law’s content and the practical representation of the moral law as a normatively binding categorical imperative.33 On this view, the claim (IIa), “We have theoretical knowledge that fully rational agents would comply with the moral law” must be distinguished from (IIb), “We have normative-­practical knowledge that we are bound by the moral law and (hence) qualify as rational agents.” Based on this view, one might argue that Kant is entitled to and endorses only (IIa). Without further proof he cannot establish (IIb), the (knowledge) claim that the moral law binds our will so that we ought to comply with it. This undermines the argument (presented in Section VIII.1) that we know our noumenal freedom will, because (IIb) rather than (IIa) is needed to establish the (knowledge) claim that we have free will as a supersensible power that enables us to be (self-)bound by and comply with a priori moral oughts. Our purely theoretical knowledge of (IIa) does not imply or require that we really are transcendentally free moral agents. This deflationary account of Kant’s view faces a striking problem. Since one can know the moral law as a purely theoretical representation even if free will does not exist, it follows that free will is not a real condition of the moral law construed as a known theoretical representation. But Kant stresses that free will is a real “condition of the moral law, which we know” (KpV, 5:4). Hence, his claim here is correct only if we know the moral law as a normative-­practical proposition. Those who reject (IIb) in favor of (IIa) must hold either that Kant is confused (in key texts) about his own views on what type of moral knowledge we can legitimately claim to have or that he makes illegitimate claims to normative-­practical moral knowledge. The latter dismissive take on Kant’s express view suggests a separate (third) skeptical objection which I consider in Section VIII.2.4. By contrast, the second objection that I am concerned with in this sub-­section is an attempt to charitably interpret Kant’s appeal to moral knowledge as an uncontroversial theoretical knowledge claim to which he is fully entitled. But since this reading must take Kant to be confused in key texts, its claim to interpretive charity cannot be sustained. There are further reasons to resist the deflationary reading of (II) as (IIa). This reading relies on a distinction between our (theoretical) representation of the moral law per se and our (practical) representation of the moral law as a normatively binding categorical imperative. Kant indeed distinguishes between the

33  See especially Allison 2011, 2020; Schönecker 1999, 2013; Schönecker and Wood 2002.

268  Kant on Freedom and Rational Agency moral law as a principle that applies to all rational beings including a perfectly rational divine intelligence and the imperatival ought-­form in which this law is apprehended by finitely rational, sensibly affected agents (cf. Chapters III, V). But the deflationary reading requires the further claim that we, qua finitely rational agents, have a separate purely theoretical-­descriptive consciousness of the moral law over and beyond our apprehension of this law in imperatival form.34 I see no textual basis for this further claim. Kant holds that for beings like us the representation of the moral law just is the practical idea of a categorically binding norm. After all, “the moral law . . . which we know” is essentially a law for the practical use of reason: its very possibility as a law depends on whether pure reason has the practical power to determine our will via moral ideas of how we ought to act: “only [the fact] that reason can be practical as pure reason makes it possible for it to be legislative,” law-­giving (KpV, 5:25). Thus, the only representation that we have of the moral law derives from our “deed” of practical self-­legislation through which our pure will binds itself to follow its own rational standards. This self-­ legislative act gives us the apodictically certain fact of reason (KpV, 5:47), the practical cognition of a moral principle that is “as law unconditionally commanded” (KpV, 5:31). Hence, there can be no gap between our legislative knowledge of the moral law and our normative-­practical cognition of a moral imperative that unconditionally commands or binds our will.35 Proponents of the deflationary reading must hold that we can reduce the normative-­practical content of the moral law to a theoretical proposition that does not presuppose our capacity for practical self-­legislation. They might say that we can know the moral law in its non-­imperatival form as a causal law that merely describes how a perfectly rational being would act.36 However, the idea of a perfectly rational agent who necessarily conforms to the moral law is not a merely descriptive proposition devoid of normative-­practical content. To see this, notice first that the moral law retains an irreducibly evaluative content in its application to a perfectly rational will: such a will would “stand under objective laws (of the good)” and could “be determined solely through the representation

34  See Allison 2020: 370–2; Willaschek 1992: 176–7. 35  Schönecker (2013: 32) concedes that “reason alone . . . formulates the content of the moral law and thereby determines which actions are morally required” but insists that we need moral feeling to recognize that the moral law yields a valid command for our will (2013: 2–4). This is puzzling: a ra­ tional determination of what we are required (hence, ought) to do must entail valid commands for us. Terminological issues aside, a central problem with Schönecker’s view is that in its practical self-­ legislation reason cannot simply “formulate” a law without also giving it the normative-­practical force of a command for our finite will (hence Kant’s point that simply “as” practical “law,” the moral law is “unconditionally commanded” for us; KpV, 5:31). In addition, his claim that we need moral feeling to grasp the commanding authority of morality runs into the problems I raised in Section VIII.2.2. 36  Thus Allison asks, “why should the laws describing the behavior of a perfectly rational agent be the grounds of obligation for us?” (2011: 360). Compare Schönecker and Wood 2002: 106, 181–3, 208.

Kant ’ s Moral Grounding of Free Will  269 of the good” (GMS, 4:414).37 Furthermore, “practically good” is that which “determines the will objectively, i.e., according to reasons that are valid for every rational being as such” (GMS, 4:413). Hence, the very idea of the moral law, quite apart from its imperatival form, has an evaluative content that further entails a normative representation of objectively-­universally valid reasons for acting or (alternatively) of an objective practical necessity (since “practically necessary” is equivalent to “unconditionally good”; cf. GMS, 4:412). Practical normativity per se is thus not the same as imperatival normativity: a perfectly rational being is not an automaton but a self-­governing agent who inevitably acts from a spontaneous representation of the absolutely good or of the right (objectively-­universally valid) reasons (RGV, 6:50), though without apprehending these reasons as oughts.38 The representation of these normative reasons turns into an imperatival representation just in case the addressees of moral norms are finite beings who are subject to an inner constraint that makes their conformity to moral norms contingent (a matter of intellectual necessitation; cf. Chapters  III,  V). In our knowledge of the moral law as a normative-­evaluative representation (of what is practically necessary or unconditionally good), our finitely rational mode of apprehension transforms this representation into knowledge of a categorical ought, obligation, or duty (MS, 6:214). Accordingly, for Kant our moral knowledge is essentially knowledge of what our obligations or duties are, of what we are normatively bound (not) to do. “One must be able to know what is right or not right in all possible cases in accordance with a rule, because it concerns our obligations, and we cannot have any obligation to do what we cannot know” (A476/B504). The doctrine of duties is a doctrine of knowledge [“Pflichtenlehre ist . . . Wissenslehre”] which allows us “to know what it is one’s duty to do” (MS, 6:375); the moral knowledge contained in the “properly and objectively so-­called” science of practical reason (Bix–­x) pertains to our “rights and duties” (JL, 9:69).39 The concept “moral duty” analytically implies being bound by the moral law (GMS, 4:439; MS, 6:222). Likewise, we know for certain that some types of actions are morally permitted (A823/B851),

37  This tells against Willaschek’s claim (1992: 328) that the expression “good” does not occur in the pure Sittengesetz. Allison claims that the following proposition yields a purely descriptive representation of the moral law: “an absolutely good will is one whose maxim can always contain itself regarded as a universal law” (2020: 334). But a purely descriptive account cannot use deontic concepts of practical reason. Moreover, the concept “maxim” does not apply to a perfectly rational will (KpV, 5:79). 38  Pace Korsgaard (1997: 240; 1996b: 161) who argues that a perfectly rational being is a mere automaton. For criticisms of her view that align with my account here, see Lavin 2004: 443–6. See also Chapter V, n. 63. 39  One might respond that Kant means only: we can know what our duties would be if we were really bound by the moral law. But this is a complete misrepresentation of the text where Kant makes assertoric rather than hypothetical or counterfactual claims. Besides, Kant calls the doctrine of duties a “true science” (MS, 6:375), and a true science cannot be a system of merely problematic-­hypothetical statements that lack any assertoric force.

270  Kant on Freedom and Rational Agency and “moral permissibility” is also defined in terms of “bindingness” (“that action is permitted, which is not contrary to bindingness”; MS, 6:222). The idea that we can know the moral law as a purely theoretical proposition runs into a further problem. If our moral knowledge did leave open whether or not the moral law applies as a valid practical norm to us as rational agents, it would follow that we fail to know whether or not the moral law applies to any actual being. What, then, could we claim to know here? At best, it would be something like this: “Our concept of a perfectly rational agent includes the proposition that such an agent would act in conformity with the moral law.”40 I say “at best” because it is doubtful whether Kant extends his dignified notion of knowledge to the analysis of potentially empty concepts that present mere logical possibilities, such as (according to the deflationary reading of (II)) the concept “rational agent.” If such feats of analyzing merely logical possibilities do count as knowledge, the resulting knowledge is solely knowledge about our concepts rather than knowledge of any actual or really possible facts. Therefore, such conceptual analyses cannot provide us with knowledge of the normative fact that the moral law has real (as opposed to merely imagined) rational validity. Likewise, such conceptual analysis cannot afford us any true insight into what rational agency is or requires. It can only tell us what we have built into the logically possible construct (the “arbitrary combination of thoughts,” A223/B270) that we call “rational agency.” Let me illustrate why this raises a major problem for the deflationary reading of (II) as (IIa). Suppose there is a popular science fiction/fantasy series X that has a huge impact on our shared conceptual repertoire—­think of Star Wars or Harry Potter but with larger cultural impact across generations. After a decade or so, there emerges a shared conceptual understanding that the X-­concept “Martian rational agent” analytically includes the proposition that “A fully rational Martian rational agent would always act in conformity with Martian Moral Law and perform three cartwheels every ten minutes.” Surely, in this scenario we do not know Martian Moral Law as a real normative fact or truth. Likewise, we do not know that our concept of Martian rationality gives us true insight into the objective structure of (a real species of) rational agency. Our concepts of Martian Moral Law or Martian rational agency are mere subjective figments of our imagination. Now, our concept “the moral law” as a law of what we term “rational agency” is—­ for all we can tell—­precisely in the same boat if (per the deflationary reading) we do not know whether our conceptual construct “the moral law” applies, as a normative-­practical truth, to us or to any actual being. One might respond that our concepts of “moral law” or “rational agency” have, unlike their fictitious counterparts, a source in our universally shared pure practical reason. This 40  See Allison (2011: 168): “it is an analytic truth that a perfect will would do whatever the moral law requires.”

Kant ’ s Moral Grounding of Free Will  271 response is correct according to the thick reading of (II) as (IIb) which I endorse. But it is unavailable to those who propose a thin deflationary account of (II) as (IIa) since they are agnostic about whether a faculty of pure legislative practical reason really exists—­hence their treatment of (II) as the merely theoretical proposition (IIa) that could be established without any normative-­practical legislation, simply via theoretical reason’s feats of conceptual analysis. To confirm that in Kant’s view our putative knowledge of the moral law lacks real validity unless we conceive it as an imperative that actually binds our will, consider what he says at the end of Groundwork II (GMS, 4:445): unless the moral law binds our autonomous will as a categorical imperative, the idea of morality is a mere “figment of the imagination,” a “chimerical idea without truth” which designates nothing at all rather than a real “something” (“etwas”). When Allison defends the deflationary view that we know the moral law only as a non-­binding, purely theoretical-­descriptive truth, he holds that “the bindingness of the” (categorical) “imperative presupposes the validity of the moral law.”41 But there is no way for us to certify “the validity of the moral law” unless we grasp the moral law as a normative truth that really applies to some actual (rather than a merely hypothetical, logically possible, thus potentially fictitious) will. Since our human will is the only known actual will to which the moral law could apply, and since our will is finite, we cannot know the validity of the moral law without grasping its validity as an imperative that really binds our will. Conversely, if we fail to grasp the moral law as a true or valid practical ought that binds our will, we cannot know whether our representation of this principle has any rationally valid content. Instead, we must concede that our idea of the moral law might, for all we can tell, be a mere figment of our imagination, just like “Martian Moral Maw.” There is a further strategy for denying that we can know the moral law in the thick sense of (IIb). For some commentators, the conviction (IIb) that the moral law obligates our will is an irreducibly first-­personal subjective matter that, as such, cannot yield objective-­ intersubjective knowledge. Lawrence Pasternack defends this view by appealing to Kant’s notion of “moral certainty.”42 For Kant, such certainty “rests upon subjective reasons” which constrain us to express our moral convictions in the first person (“I am morally certain . . .”). Since subjective moral certainty is not a shared, communicable “certainty for everyone” and since “all knowing . . . can be communicated,” moral certainty falls short of knowledge (A828–9/B856–7). But the strategy of invoking Kant’s notion of moral certainty to show that in his view we cannot have normative-­practical (as opposed to purely theoretical) moral

41  Allison 2020: 332. 42  Pasternack 2011: 311–15. The alleged first-­personal, subjective character of Kant’s moral theory is also a major theme in Grenberg 2014. My contrasting anti-­solipsistic approach is inspired by Ameriks 2019 and O’Neill 1989.

272  Kant on Freedom and Rational Agency knowledge fails for the same reason as the attempt (considered in Section VIII.2.1) to convert moral knowledge into moral faith: it conflates two different kinds of “moral” propositions. Earlier, I pointed out that what Kant calls “moral faith” does not comprise prescriptive moral judgments but rather pertains to descriptive metaphysical claims made in a moral context, such as the claim “that there is a God and a future world” (A828/B856). Likewise, Kant ascribes a subjective moral certainty to the belief in God but not to beliefs in moral obligations. For Kant moral judgments are cognized as objectively certain prescriptions of reason, in contrast with our subjective attitudes and incentives (GMS, 4:412). The first-­ personal, private or incommunicable element that (according to the Canon)43 attaches to the moral certainty that God exists and that prevents this certainty from becoming objective knowledge is due to the fact that this certainty “rests on subjective reasons” that concern one’s “moral disposition [Gesinnung])” (A829/B857), especially the strength or magnitude of that disposition (cf. JL 9:70; Refl., 16:392–3). The notion of a moral disposition is distinct from concepts such as “bindingness” or “duty” that express the objective normative-­practical force of morality. My awareness of my moral disposition and its degree of strength concerns my personal maxim as a “subjective principle of volition that guides my deliberation and choice,” whereas “the objective principle . . . is the practical law” (GMS, 4:400; cf. KpV, 5:26). This subjective-­objective contrast persists when “practical law” is replaced with the imperatival term “principle of duty”: “Maxim is the subjective principle of acting . . . by contrast, the principle of duty is that which reason commands . . . objectively (how [the subject] ought to act)” (MS, 6:225). Since practical laws, qua laws, are independent of subjective, contingent conditions that distinguish one agent from another (KpV, 5:21–2), they fully satisfy the “universal communicability”/“certainty for everyone” condition for knowledge. Hence, we can know them as objectively-­universally valid categorical imperatives that apply to every finite agent. By contrast, hypothetical imperatives are “subjectively conditioned” and thus apply only (if at all) to some agents depending on their special private, contingent empirical circumstances (KpV, 5:19–21; I discuss this contrast in Chapter VIII).44 I conclude that the second objection that one might raise against (II) fails since it relies on a deflationary conception of (II) as (IIa) which cannot be sustained. For Kant our knowledge of the moral law is not a knowledge of some purely 43  I do not see that Kant’s later (second/third Critique) account of rational moral faith in God retains the Canon-­idea that such faith involves an element of incommunicability. But this is irrelevant to my argument here. 44  Pasternack’s claim that “there is something subjective about our belief in the Fact of Reason” (2011: 315) does not accord with Kant’s aforementioned claim that we are a priori conscious of the moral law as a fact of reason with apodictic certainty (KpV, 5:47). Apodictic certainty is precisely what is lacking in faith or in the subjective “moral certainty” that there is a God (A830/B858). Apodictic certainty is an objective “rational certainty,” the mark of genuine knowledge (JL, 9:71), also characteristic of mathematical cognition (B41; A46–7/B64; A713/B741).

Kant ’ s Moral Grounding of Free Will  273 theoretical proposition that can be established through the analysis of potentially empty concepts raising mere logical possibilities. Rational knowledge of the moral law cannot be anything but the irreducibly normative-­practical knowledge (IIb) that we are categorically bound to certain types of actions (or omission) as our duties.

VIII.2.4.  Moral Knowledge and Moral Skepticism I now turn to a third objection against (II). Those who raise this objection may concede that Kant seems committed to the idea that we have moral knowledge in the thick sense of (IIb). But in their view, Kant is unacceptably dogmatic when he insists that we can know without any proof or deduction that our will is categorically bound by moral principles. The term “dogma” has a technical meaning in Kant that need not ha76ve non-­ critical implications.45 In what follows I use “dogmatic” in its ordinary sense as a pejorative term indicating a non-­argumentative stance of merely pounding on the table. We must then distinguish between two different versions of the worry that Kant’s practical philosophy is unacceptably dogmatic. This worry could relate to Kant’s putative normative dogmatism, his claim that we know without proof the absolute bindingness of moral prescriptions. Or, it could relate to Kant’s putative metaphysical dogmatism, his attempt to infer knowledge of noumenal freedom from normative moral knowledge.46 I will consider the latter worry in Section VIII.4 and (especially) in Chapter  IX. I address the former, normative issue in this sub-­section. The criticism that Kant’s appeal to our knowledge of the moral law as a binding “fact of reason” rests on mere dogmatism has recently been expressed as the claim that Kant is here engaged in mere “foot stomping” or “moralistic bluster.”47 These phrases suggest that the charge of dogmatism is fueled by moral skepticism about whether we really are subject to absolutely binding moral obligations that may

45  See Ameriks’s discussion of the “dogmatic” character of Kant’s moral theory (2000a: 218; 2000b: 75; 2013). Kant does argue that a “dogma” qua direct (immediately known) synthetic proposition from concepts fails to accord with his critical strictures (A736–7/B764–5), but here he seeks to deny only that speculative reason contains principles that are “dogmatic” in this sense. Hence, it is not clear that he would also take the term “dogma” as indicating a non-­critical stance in his practical fact of reason doctrine. Ameriks (2000a: 218, 231) cites Kant’s remarks in the late (1793–5) Preisschrift as evidence that Kant eventually deemed his own practical philosophy dogmatic. Kant here indeed talks a good deal about “practical-­dogmatic” principles, but he also calls his critical account of a priori (transcendental) cognition of sensible objects “theoretical-­dogmatic” (FM, 20:311; cf. FM, 20:274). Thus, here the term “dogmatic” does not indicate a non-­critical stance. 46  Ameriks’s worry about dogmatism is often focused on the metaphysical implications of Kant’s moral views regarding free will; this is especially clear in Ameriks 2000a: 226, but somewhat less clear in Ameriks 2013. 47  Guyer 2007: 462; Wood 2008: 135.

274  Kant on Freedom and Rational Agency require us to frustrate our prudential self-­interests (our wish for happiness): unless Kant can offer us a valid proof that (and why) every rational agent has a conclusive reason to follow the moral law, his appeal to certain knowledge that we stand under binding moral duties is mere foot stomping, moralistic bluster. Along with most commentators, I believe that Kant has no (successful) proof of this sort.48 But does this conclusively show that his appeal to our normatively binding moral knowledge is clearly unacceptable? One might propose a rather modest defense of Kant’s view: his account of moral knowledge pertains only to agents who partake in and are committed to our moral practices—­it is a moral epistemology for moral agents which has no implications for those agents, if they exist, that are sincerely indifferent to moral considerations.49 By rough analogy, it seems possible to account for the logical knowledge possessed by those who reason in accordance with standard rules of inference without trying to prove that such rules are also binding for those thinkers, if they exist, who opt out of the kind of reasoning that is governed by these rules or who even profess not to engage in any kind of inferential reasoning. The bindingness of principles for agents who are firmly committed to a certain practice is not impugned by the fact that there are or might be different kinds of agents who stand outside that practice. One might add that a more ambitious deduction project aimed at refuting or converting radical moral skeptics is unreasonable since there is simply not much that abstract philosophical reasoning can say to defend the authority of morality against those who (purport to) stand outside our moral practices.50 I believe that a somewhat more ambitious strategy is available to Kant as a way of mitigating the charge of dogmatism. This strategy is best understood an attempt to raise the stakes for moral skepticism and thereby to defend the moral knowledge of non-­skeptical moral agents. The strategy I propose goes back to my earlier suggestion that our assent to the moral law has the status of knowledge because it grounds a systematic body of objectively certain practical prescriptions. I now consider what a practical orientation would look like that did not accept the moral law and that, consequently, 48  Some argue that in Groundwork III Kant tries to give a “deduction” of morality on the basis of non-­moral premises. This is a controversial reading. Others regard the Groundwork III position as akin to the second Critique’s fact of reason doctrine (see, e.g., Brandt 1988; Freudiger 1993; Henrich 1975; Ware 2017; Wolff 2009). Kleingeld (2010: 55) aptly diagnoses the “wildly divergent” views that commentators have about the relation between Groundwork III and second Critique. Even those who maximize the contrast between the two works (Allison 1990, 2011, 2020; Ameriks 2000a, 2000b, 2003, 2019) concede that the proof strategy of Groundwork III fails and that Kant gave up the attempt to provide a deduction of morality in his post-­1785 views. My account here focuses on the post-­1785 views and (mostly) brackets Groundwork III because in my view Groundwork III is too unclear and ambiguous to warrant confident interpretive claims about Kant’s argument structure, about what (moral or non-­moral) premises he argues from, or about the intended strength of his “deduction” of morality. 49  See Willaschek 1992: 191–2. 50  See Ameriks 2000a: 225–6; 2013: 155.

Kant ’ s Moral Grounding of Free Will  275 had no access to the action-­guiding practical cognitions that can be derived from this axiomatic principle. Kant argues that this would be a cognitively impoverished practical orientation devoid of rational foundations.51 A morally indifferent or skeptical practical outlook would need to rely, in figuring out what to do and how to live, on counsels of prudence that represent means towards happiness. These counsels constitute a privileged class of technical-­hypothetical imperatives because all other kinds of hypothetical imperatives presuppose the prudential determination of what our happiness consists of (EEKU, 20:201; V-­NR/Fey, 27:1324): the causal rule that I must do x in order to realize some end y has normative-­practical relevance for me only if prudence determines that I should realize end y because this would contribute to my happiness. However, the prudential determination of what would make us happy lacks all the essential components of rational practical knowledge: prudential rules have neither certainty, necessity, nor objective universal validity.52 Rules of prudence lack certainty and necessity because they are based on the inductive observation of empirical circumstances under which human agents tend to be more or less happy (GMS, 4:418; MS, 6:215–16). These circumstances include both our external material conditions and our internal psychology (our desires and preferences). The lack of certainty and necessity here reflects not just the usual Kantian caveat that inductive evidence is merely probabilistic and (thus) cannot show that a rule must apply without exception (cf. B3–4; JL, 9:66–7, 82–5). There are three reasons why the limits of inductive reasoning have far more serious implications for the rationality of prudential judgments than for theoretical judgments that are based upon objective empirical evidence (e.g., “All ravens are black”). First, because we have only a very poor theoretical understanding of human empirical psychology in general (MAN, 4:471–2; EEKU, 20:237–9), we do not properly understand the psychological conditions of human happiness either. Second, both the internal-­psychological and the external-­material conditions of happiness are highly individualistic, i.e., they vary strongly across different agents. Hence, rules of prudence that are focused on giving general advice to human beings must abstract from precisely those specific circumstances that may turn out to be crucial for the happiness of individual agents. The lack of a common basis for prudential reflection, i.e., the fact that prudential counsels depend on subjective circumstances (such as egocentric empirical preferences) that vary from agent to agent, is also the reason why such counsels lack objective-­universal validity (GMS, 4:416; KpV, 5:20–5; MS, 6:215–16): what might make me happy might well make someone else miserable and vice versa. This severely limits the 51 In what follows I draw on my discussion of instrumental (pseudo-)rationality in Kohl 2017b, 2018b. 52  Kant does not strictly identify universal (intersubjective) with objective validity (his aesthetics appeals to a form of universal subjective validity; cf. Chapter X), but he does equate these notions with regards to practical cognition.

276  Kant on Freedom and Rational Agency possibility of an intersubjective discourse about how we should act in the pursuit of our happiness: there is no shared normative framework for rationally discussing and settling such questions in a purely prudential, non-­moral fashion. Third, even if prudential rules focus just on the specific circumstances of a single agent, the problem remains that both the external and the internal conditions of individual happiness are highly fickle: they are subject to many changes over the course of time that we cannot anticipate with any degree of certainty (KpV, 5:25). What seems prudent for me to do in relation to my present material conditions and my present preferences may turn out to be very bad advice when measured against the unforeseeable changes that affect my material realities and my empirical psychology over the course of time. This is a significant problem because our wish to be content and happy concerns our life as a whole, as a temporally extended process (KpV, 5:22). For Kant the lack of certainty, necessity and objective-­universal validity is essential to non-­moral practical rules.53 This is the upshot of an important passage from the second Critique (KpV, 5:26) that I already considered in Chapter IV. Kant asks us to imagine that all finitely rational beings have the same conception of happiness and of the means through which they can attain their happiness. That is, he asks us to imagine a wildly implausible scenario (a mere logical possibility) that seems most congenial to the certainty, necessity, and universal validity of prudential rules. But even in this scenario, prudential rules cannot provide rational practical knowledge. First, there is no way around the fact that the conditions which (in this scenario) make it the case that everyone right now has the same conception of what their happiness involves and how to obtain it are fickle conditions which are subject to numerous changes over time, both inter- and intrapersonally. This  inherent susceptibility to change follows inevitably from the empirical, contingent-­temporal character of the relevant material or psychological circumstances that determine any finite conception of what would make people happy. Accordingly, the rules of prudence which are based upon these circumstances inherently lack certainty and necessity. Kant’s second point is a bit harder to appreciate but equally important. Even if the same prudential rules did apply to all of us this agreement would be merely accidental, i.e., it would be due to the empirical coincidence that we all happen to have the same psychological proclivities. As I argued in Chapter IV, contingent sameness of subjectivity cannot ground genuine intersubjectivity and objectivity. In the fictional scenario, the same prudential rules apply to all of us only because each of us happens to have the same subjective desires (plus, the same physical capacities, material means etc.). This cannot ensure that we act for truly shared

53  Pace Wood 1999: 69–70.

Kant ’ s Moral Grounding of Free Will  277 reasons. Psychological states like empirical desires or feelings are non-­rational in two related senses. For one, we have no rational control over their (continuing) presence and strength; hence there can be no rational discussion, debate, or argument about whether one should (continue to) follow the relevant prudential rules—­either one has the relevant desire(s) with a sufficient degree of strength, or one does not, and in either case there is nothing further to be said apart from asking, “well, do you strongly want this, or not?” Moreover, the presence, content, and strength of empirically given desires and feelings (like pleasure and pain) is not fully communicable: we may sometimes feel like we want to obtain or avoid something, but it is never fully clear to us what exactly attracts or repels us and how strong the degree of attraction or repulsion is (especially when viewed over the course of time). As we saw (in Section VIII.2.3), such a lack of communicability blocks the possibility of knowledge since it entails a lack of “certainty for everyone.” Based on these considerations, Kant holds that non-­moral practical norms are weak, fickle, subjective opinions devoid of objective necessity and certainty (KpV, 5:19–21, 35). Hence, our agency would be cognitively and rationally impoverished if it were guided solely by precepts of happiness: if our actions were based only on prudential opinions, we would never know what we ought to do; our norm-­governed agency would be like a continuous shooting in the dark. We might even develop a “misology, hatred of reason” (GMS, 4:395) when we find that the more we seek to rationally plan out our happiness the farther we fall short of true contentment (GMS, 4:418–19). Let me illustrate with two examples that this is not merely an academic issue. First, consider the case of Paul who seeks to invest money with the aim to generate a good twenty-­year investment return. The empirical data Paul consults to make a rational choice include facts about past fund performances, economic studies and trends, and so on. But these all depend upon empirical conditions that may well “change subsequently” (RezS, 8:14), e.g., if there is a stock market crash as in 2008 or a pandemic as in 2020. Bracketing those sorts of issues, the real question Paul seeks to answer is not what he should do to generate a good twenty-­year investment profit but what he should do to live a content and happy life. Even if we suppose for the sake of argument that Paul has a rational basis for expecting that choosing investment plan X increases the objective probability for a good twenty-­year investment return, the question remains: what is Paul’s basis for expecting that choosing plan X will be conducive to his overall life-­ satisfaction? In forming this expectation Paul must rely on many fickle, uncertain assumptions. He must hope that his material circumstances, his job situation and relationship, etc. remain constant over the next twenty years so that he does not want to pull out the money before the investment pays off. He must take it for granted that he does not fall ill or die prematurely, for if he did it would be better to spend the money right away while he still can enjoy things. He must count on

278  Kant on Freedom and Rational Agency the assumption that he adequately discerns his real current preferences (he must discount, e.g., the possibility that donating the money might make him happier because he has dormant benevolent tendencies). He must, furthermore, presuppose that those current preferences which he can discern do not change subsequently: e.g., that he does not develop a new passion whose pursuit would require accessible funds, or that in twenty years he no longer has a strong interest in the amenities that the investment is supposed to secure for him, etc. etc. And now suppose, finally, that there are no rational criteria of moral permissibility that constrain Paul’s decision making. Why wait for twenty long years before obtaining more money? Paul may instead contemplate getting a larger amount immediately by killing his overbearing wife who has a huge life insurance policy. That of course raises the risk of being caught. But how high is that risk really? Paul considers himself clever. Is he clever enough to get away with it? Who knows. Paul is currently intrigued by the thrill of risk: it adds to the spice of his life. How long is that thrill going to last? Maybe Paul should go read some psychology studies on that question. But do those studies really have predictive value that applies to his personal psychology? Who knows. Maybe Paul should consider, instead, embezzling money from his company? That might be less risky and the consequences if he was caught would be less severe. And so the questions continue indefinitely. Imagine now a young talented undergraduate student considering whether they should pursue a humanities PhD degree. Do they have a rational basis for judging whether doing so will be conducive or detrimental to their overall life-­ satisfaction (viewed as a temporal continuum)? There is a vast variety of relevant but wholly unpredictable variables here, pertaining both to the future state of the external world (economic up- or downturns; developments in the landscape of higher education; changes in one’s relationships with family, friends, and partners; etc. etc.) and to the student’s psychology (whether they will enjoy writing about their now-­favorite topics when they must devote sustained scholarly attention to them in professional settings, under peer pressure and “Publish or Perish!” norms; how the realities of graduate school and post-­grad life will affect their emotions and stress-­level; what other interests they might pick up in the meantime; etc. etc.). Perhaps one can make some educated guesses regarding these variables. But they won’t be more than guesses, and it is questionable just how educated or well-­informed they could possibly be. Moreover, the general content that such guesses will take qua putatively educated, inductive counsels (that are based on tests, studies, statistical averaging, etc.) may need to abstract from precisely those highly specific (external and internal) conditions that account for the unique situation of individual agents. There is popular generic prudential advice where it may seem that its abstractness allows it to cover all sorts of individual cases, such as “follow your passion.” But here the generic nature of the putative advice prevents it from being informative. What exactly does it take to “follow” a particular passion? If one tries to turn one’s passion (e.g., for philosophical

Kant ’ s Moral Grounding of Free Will  279 thinking) into a source of regular income and employment, that might be the best way to kill it off. If one tries to pursue it as an aside to regular patterns of work, that might also make it more of a burden or increase dissatisfaction with one’s work routines. Moreover, passions (desires, inclinations, interests) are highly opaque states: it is not clear that we know what our strongest passions truly are (there is much room for self-­deception and bias here) and we cannot know how the passions we ascribe to ourselves now will develop over the course of times, or what other interests will crop up that may weaken or undermine our present proclivities. Thus, Kant surely has a point when he argues that prudential norms fail to guide our agency in a truly rational manner that would involve certainty, necessity, and objective-­universal validity. Our practical agency would be cognitively impoverished if it were not governed by our certain knowledge of objective moral norms. Now, radical skeptics about rational agency and normative-­practical knowledge may dig in their heels and respond that our practical orientation is indeed cognitively impoverished, that we are shooting in the dark when we make our practical decisions and that we lack a shared normative framework for intersubjectively valid practical discourse. I believe that for Kant this view does not merit serious consideration: the fact that moral skepticism leads into general skepticism about rational agency and normative-­practical knowledge serves as a reductio of moral skepticism. I understand his argument in the second Critique’s Analytic as follows: since empirically based rules for obtaining happiness fail to give our practical agency a secure rational foundation, there must be non-­empirical practical rules that do provide us with objectively certain rational guidance. These rules are moral norms that derive their a priori rational normative force from the moral law as the first axiom of all objective practical cognition. The moral law can fulfill this role because its prescriptive content abstracts from all those conditions that prevent certain rational guidance: the demand that our maxims be suitable for universal legislation does not hinge on our fickle empirical desires or material circumstances (KpV, 5:27). In the domain of practical reason the moral “law alone implies the concept of an unconditional and objective and hence universally valid necessity” (GMS, 4:417). It has “that necessity, which is thought in each and every law, namely, an objective necessity from a priori reasons” (KpV, 5:26) that it imparts onto all concrete moral duties that can be derived from it. On my reading Kant’s argumentative strategy presupposes the existence of objectively necessary practical laws. This runs against the assumption that Kant in the second Critique seeks to prove that such laws exist.54 This assumption 54  For Allison, the second Critique’s Analytic aims to answer the question of “whether there is in fact such a” (formal, objective) “law” of practical reason (2020: 356). See also Kleingeld 2010: 57; Willaschek 1992: 170–2, 195–202.

280  Kant on Freedom and Rational Agency sometimes leads commentators to finesse some argument that is absent from the text so that the fact of reason doctrine seems less dogmatic.55 The expectation that the second Critique seeks to disprove empiricist (say, Humean)56 skepticism about the existence of practical laws might be fueled by Kant’s proclamation that the critique of practical reason must “keep the empirically conditioned reason from presuming to seek to provide, alone and exclusively, the determining basis of the will” (KpV, 5:16). But this proclamation is best understood as anticipating the argument of the Analytic which aims to show that empirically conditioned reason is incapable of yielding valid practical laws. The Analytic begins with “Theorem 1” (KpV, 5:21) where Kant argues that practical principles whose motivating force depends on a given object of desire are “one and all” empirical. After establishing that all empirical-­practical principles are ultimately concerned with happiness, Kant infers the “Corollary” that the empiricist view precludes the ex­ist­ence of necessary practical laws (KpV, 5:21–2). In the following “Comment I” Kant criticizes empiricist accounts of practical reason for being confused “in their own explanations” (KpV, 5:24). These explanations try to secure the lawfulness of practical reason as a higher faculty of desire through the misguided claim that within the class of practical representations that presuppose subjective feelings of attraction or repulsion towards given objects, there is a qualitative difference depending on whether these representations originate in the intellect (KpV, 5:23). Kant rejects this claim because in his view all object-­based, receptive desiderative states or feelings belong to the same empirical (“pathological”) kind. His criticism of the empiricist view culminates in his claim that even the least dependency of practical rules on such empirical states would diminish “the strength and merit (Vorzuge)” of practical reason, just like the dependency of a mathematical demonstration on empirical conditions would destroy its “force and dignity” (KpV, 5:25). This appeal to a parallel between practical and mathematical reasoning would be totally pointless if Kant were trying to prove against empiricists that reason has the same lawful force or dignity in practical matters that it has (even according to empiricists; cf. KpV, 5:13) in mathematics. But his critical observation that an empiricist view of practical reason undermines the lawfulness and necessity of practical rules makes good argumentative sense if his target here are “otherwise acute men” (KpV, 5:22), such as Epicurus (KpV, 5:24), who accept that there are necessary practical laws but who also subscribe to empiricist presuppositions that vitiate this belief. Against such views, Kant claims that empirical practical principles “cannot possibly be regarded as a law” (KpV, 5:25). He supports this claim

55  See, e.g., Kleingeld 2010 and Sussman 2008. Ameriks (2013: 161–73) provides a strong critique of these readings. 56  Note, however, that Kant may well have been unfamiliar with Hume’s radical skepticism about practical reason due to his lack of familiarity with the Treatise (cf. Guyer 2003b: 22).

Kant ’ s Moral Grounding of Free Will  281 with the abovementioned argument that even in a fictional scenario where everyone has the same desires, empirical practical principles “would still be only subjectively valid” (KpV, 5:26). Since all empirical-­material practical rules are only ever subjective and contingent, there must be an a priori-­formal practical principle that can rationally guide our agency as an objectively necessary, certain law (KpV, 5:27). Thus, the second Critique argues from, not to, the claim that objective practical laws exist (it “begin[s] from the possibility of a priori practical principles”; KpV, 5:89–90). After presenting his main argument in the Analytic, Kant repeats that empiricists are themselves committed to this claim but cannot sustain this claim due to philosophical obfuscations: “it is odd how it could have occurred to intelligent men” that maxims based on empirical conditions would “pass . . . off as universal practical laws” (KpV, 5:28). Fortunately these philosophical obfuscations do not affect “the commonest understanding” (KpV, 5:27) which has immediate cognitive access to the moral law. Through their assent to the axiomatic principle which grounds all their practical knowledge, common agents immediately grasp how they (generally speaking) ought (not) to act. Thus, in my view Kant’s fact of reason doctrine is best understood as an appeal to the common (often obscure, indistinct) practical awareness of the fundamental principle that universal practical reason gives to itself so that it can govern our productive efforts in an objectively rational fashion. Viewed in this way, the fact of reason is simply the pre-­philosophical (self-)consciousness that we have practical reason: that we are capable of objective normative-­practical knowledge and (consequently) of genuinely rational agency. An appeal to the ordinary moral self-­consciousness of common agents is not a new strand in Kant’s practical philosophy. Already in the (A-­edition of) the first Critique, he declares: I assume that there really are pure moral laws which determine completely a priori (without regard to empirical motives) . . . what is and is not to be done . . . and that these laws command in an absolute manner . . . and are therefore in every respect necessary. I am justified in making this assumption, in that I can appeal . . . to the moral judgment of every human being . . .  (A806–7/B834–5)

In the second Critique Kant says (when recapitulating his argument in the Analytic) that the “justification of . . . moral principles . . . could . . . with sufficient certainty be provided through mere reliance on common human sense” (KpV, 5:91–2), because in morality “the voice of reason [is] so clear, so irrepressible . . . even to the commonest person” (KpV, 5:35).57 One might object that Kant’s 57  For some, Kant argues that the common acceptance of the moral law by different people yields evidence that it is truly a priori and does not depend on empirical conditions (Proops 2003: 224–9;

282  Kant on Freedom and Rational Agency appeal to common sense as providing a justification of moral principles conflicts with his claim that the moral law “itself needs no justifying reasons” (KpV, 5:47). But these points are compatible if we take him to mean: the moral law needs no further philosophical justification (specifically, no proof or deduction: cf. KpV, 5:46–7) apart from the common pre-­philosophical awareness that this law is the basic standard underlying all and any objective, universally valid practical justification or rational decision-­making.58 With all that said, where do we stand on the issue of whether Kant’s appeal to our knowledge of the moral law (in the thick sense of (IIb)) is dogmatic? If one thinks that Kant owes us a proof that we are truly rational agents who act on the basis of objectively certain normative-­practical knowledge, then my account admittedly does nothing to mitigate the charge of dogmatism. If, on the other hand, one thinks that Kant needs an argument to show why we qua rational agents cannot do without the moral law, or why the moral law alone has the special epistemic properties that make for certain practical knowledge, then my account shows that Kant’s fact of reason doctrine is not dogmatic: while he presupposes that there are objectively-­universally valid laws of practical reason, he does not simply presuppose that such laws are to be identified with moral laws—­ rather, he argues for this identification (in the ways I have sketched). He thereby supports, without begging the question, his claim that rational agents who make their decisions based upon objectively certain normative standards must accept the moral law as an immediately certain, axiomatic fact of reason which grounds all mediate, non-­axiomatic practical knowledge.59 Wolff 2009: 539–40). But an empirical survey that many different agents accept the moral law yields only an empirical, inductive epistemic ground which could establish no more than a limited comparative universality rather than the strict universality/necessity that is required for an a priori principle (see Grenberg 2014: 150–5). 58  For a similar point, see Timmermann 2010: 82. The Analytic of the second Critique gives a philosophical “exposition” rather than a proof of the moral law (KpV, 5:46). An exposition provides a “true and useful” analysis which clearly exhibits certain characteristics of a given concept (A730/B758; JL; 9:43). So, when Kant states that the exposition of the moral law shows “that it exists by itself, wholly a priori and independently of empirical principles” (KpV, 5:46), he means: an analysis of the concept “practical law” clarifies that this concept contains representations, e.g., strict necessity or the absence of material conditions (KpV, 5:34), which entail an a priori origin in pure reason. An exposition that clarifies what agents obscurely think through the idea of a practical law is not a philosophical proof that argues from neutral premises to the conclusion that there is a valid practical law. 59  Ware also defends the fact of reason doctrine against the charge of dogmatism. While there is much I agree with in his thoughtful argument, I cannot follow his claim that there is “a continuity in the argument-­structure” of the first two Critiques (2014: 16). Kant repeatedly points out a discontinuity, a “peculiar contrast” between the two argument-­structures (KpV, 5:16, 42, 44–5, 89–90). Ware argues that the abstraction procedure Kant invokes in the first Critique at B5–6 is also at work in the second Critique: in each case, Kant uses thought experiments to show that representations such as space, substance, and the moral law are necessary and a priori. However, Kant’s thought experiments in the B-­introduction rely on the as of yet unargued claim that we “cannot” remove space and substance from our experience of bodies. This is only a preliminary anticipation of Kant’s subsequent attempts to prove that space, substance etc. are necessary conditions of experience. By contrast, the second Critique denies that the moral law can or needs to be proven. Ware argues, further, that “[J]ust as examples from mathematics and science confirm the reality of synthetic a priori cognition,

Kant ’ s Moral Grounding of Free Will  283 I conclude my discussion of this issue by mentioning three further considerations which add to my view that Kant’s post-­1785 position on moral knowledge is non-­dogmatic in the qualified sense that I just explained. First, the view I am ascribing to Kant does not imply an inability or unwillingness to engage with traditional expressions of philosophical moral skepticism or with common moral vices. Skeptical views often share Kant’s presupposition that we have some objectively certain, universally valid practical knowledge. For instance, Thrasymachus in the Republic arguably proposes a substantive conception of the good, of what is really the best kind of life, which he thinks can be defended through valid practical reasons in a rational debate. Similar points apply to Hobbes’s Fool or to Hume’s Sensible Knave. Leaving aside such philosophical characters, Kant’s view that all common human agents accept the moral law as binding acknowledges the fact of widespread vice: even those who strike us as blatantly immoral typically believe (due to self-­deception) that their maxims are universalizable and rationally defensible, or that everyone should agree that their special circumstances morally warrant an exception from standard moral rules (cf. Chapter V). Second, one might argue that even radical skeptics about a priori practical knowledge or objective practical cognition must implicitly rely on the authority of pure reason that they purport to undermine.60 The skeptical view clearly goes beyond experience or empirical evidence: it involves a claim to necessity in proclaiming that we cannot have certain a priori cognition of how we ought to act, or that it is impossible for practical reason to determine how everyone ought to act. Moreover, the skeptical denial of practical knowledge also lays claim to universal validity: the skeptics argue that this denial ought to be shared by everyone who clearly reflects on the epistemic credentials of their practical judgments. Hence, they seem committed to the idea that we can obtain an objective rational insight regarding the limits of practical rationality. But this attempt to undermine the authority of reason through rationally conclusive, universally valid arguments may be unstable. For Kant, it is “tantamount to someone’s wishing to prove by reason that there is no reason” (KpV, 5:13). Third and finally, those who insist that Kant must argue for the claim that we possess objective practical cognition and knowledge should be careful not to impose upon Kant’s practical philosophy a burden that his theoretical philosophy judgments from ordinary people confirm the reality of moral consciousness” (2014: 16). However, examples from mathematics confirm the reality of synthetic a priori mathematical cognition not because mathematical cognition “neither admit[s] nor require[s] further proof ” (2014: 16) but because these examples illustrate an established method of proof (construction in pure intuition) that is inapplicable to discursive principles of pure reason (A734–5/B762–3), including practical ones. Transcendental philosophy must explain how actually given theoretical sciences such as mathematics are possible (B20), whereas an explanation of how the moral law is really possible lies beyond the reach of pure reason. 60  Here I am picking up on a suggestion that Engstrom 1994 makes in a different context.

284  Kant on Freedom and Rational Agency is unable and unwilling to bear. Kant complains that it would be a “scandal of philosophy” to accept the existence of physical things outside our mind as a mere matter of faith without addressing relevant skeptical doubts through a proof (Bxl). The proof Kant offers (in the Refutation of Idealism) begins from the premise that we possess experiential cognition(N) and knowledge of our empirical self, specifically of the temporal relations among its fleeting mental states (B275–8). He argues that such empirical self-­cognition and self-­knowledge presupposes our cognition and knowledge of permanent material substances. A radical skeptic could reject this proof simply by denying the premise that we can know objective temporal relations among our mental states.61 Since Kant simply ignores this response in the Refutation, he must think that such radical skepticism does not merit serious philosophical discussion. More generally, Kant’s theoretical philosophy presupposes that we do have some objective empirical cognition(N) (cf. Chapter VII).62 Kant even presupposes, as an actual “fact,” that we have synthetic a priori mathematical cognition (A95/B128). For these reasons, we should be wary of the expectation that Kant’s practical philosophy must confront the claim that we lack any objective normative-­ practical cognition or knowledge. Likewise, those who insist that there is something incommunicably subjective or “first-­personal” about the fact of reason face a burden to explain why the same issue does not arise for theoretical cognition. A suitably radical skeptic can always insist that our immediate awareness involves only private, subjective representations that cannot rise to objective cognition. In this regard, our moral self-­awareness is on par with our inner empirical self-­ awareness, which Kant assumes to yield objective cognition of temporal order (this is his premise in the Refutation and in the Second Analogy proof that causality is a condition of our actual objective temporal cognition). This brings out an important feature of Kant’s approach to philosophical questions: neither his theoretical nor his practical philosophy begins from some essentially private, egocentric starting point. Rather, Kant’s critical methodology accepts the authority of “universal reason” as the object and subject of critique: the court of reason that adjudicates questions about what we can know and how we should act is a public court inviting intersubjective debate.63 This explains why Kant in his practical philosophy feels entitled to presuppose that we have shared access to the moral law as an axiomatic fact or formal standard of reason that enables different persons to engage in fruitful normative-­practical discourse about how we ought (not) to act.

61  See Brueckner 1984: 218. 62  This supports the reading of Kant’s transcendental deduction as a regressive argument from the fact of (some) empirical cognition to its a priori conditions; see Ameriks 2003 for a classic defense of that view. 63  O’Neill 1989 presents and develops this anti-­solipsistic theme in illuminating fashion.

Kant ’ s Moral Grounding of Free Will  285

VIII.3.  From Moral Knowledge to Knowledge of Free Will VIII.3.1.  The Inferential Structure of Our Knowledge of Free Will In this section I consider a view that accepts (II) but denies (III). On this view, our knowledge that we are bound by real moral duties does not presuppose that we have free will: it tolerates the epistemic possibility that we lack free will just as it tolerates the epistemic possibility that there is no God. Our justification for believing that these supersensible objects exist strictly depends upon our subjective need to view ourselves as moral agents. (This is, in essence, the Parity Thesis.) Allen Wood provides the classic articulation of this view. He reconstructs Kant’s argument concerning God and immortality as a reductio ad absurdum practicum: (1) Unless I accept that God exists and that I am immortal, I cannot conceive the highest good as possible. (2) If I cannot conceive the highest good as possible, I cannot (rationally) pursue the highest good. (3) If I cannot pursue the highest good, I cannot regard myself as obeying the moral law. (4) If I cannot regard myself as obeying the moral law, I must regard myself as a moral scoundrel, a conclusion that I cannot live with as a rational moral agent.64 On Wood’s view, “Kant’s argument for freedom can . . . be stated analogously to the moral arguments for God and immortality,”65 albeit without appeal to the highest good: (1’) Unless I accept that my will is free, I cannot conceive of myself as willing autonomously. (2’) If I cannot conceive of myself as willing autonomously, I must deny that I can obey the moral law. (3’) If I deny that “I can do what I am unconditionally obligated to do,” I must accept “a conclusion about myself as a moral agent which I cannot tolerate.”66 Wood’s analogy, specifically (3’), presupposes that moral agents can simultaneously deny their free will and yet (still) consider themselves subject to moral obligations. But for Kant, “if there were no freedom, the moral law would not be encountered in us at all” (KpV, 5:4): hence, a denial of freedom would also entail a rejection of the moral law. Those who denied their free will could not (pace (3’)) reach an intolerable conclusion about themselves as moral agents who stand under what they (still) recognize as valid, unconditionally binding moral norms. This response to Wood presupposes that moral agents recognize their free will as a real condition of moral normativity. One might object that this implies an over-­intellectualization of ordinary moral consciousness. If common agents typically do not recognize the conditional, “If the moral law is valid, then I must be

64  Wood 1974: 29–30 (cf. V-­Th/Pöl, 28:1083–4; A828/B856). 66  Wood 1974: 36–7.

65  Wood 1974: 36.

286  Kant on Freedom and Rational Agency free,” then someone might doubt or even deny their freedom while still accepting the normative validity of the moral law, as Wood’s argument suggests.67 Kant does not believe that common agents who lack philosophical sophistication clearly formulate abstract conditionals regarding the different (epistemic, ontological) relations between freedom and morality. But in his view common agents do recognize (however implicitly) that free will is required for the objective normative validity of the moral law. The protagonist in Kant’s case of someone who is threatened with execution unless he bears false witness stands proxy for ordinary moral agents; this person “judges that he can do something” (namely, refuse to bear false witness) “because he is conscious that he ought to do it” and thus “cognizes in himself the freedom that would have remained unknown to him otherwise without the moral law” (KpV, 5:30). Hence, the ordinary moral consciousness that one morally ought to do x entails that one has the freedom to do x. Since common moral agents are disposed to reason via modus tollens (an elementary rational capacity), they must also be disposed to recognize that if they lacked freedom to do x their consciousness that they morally ought to do x would be illusory. Kant espouses an even stronger view: ordinary moral agents are (rationally) compelled to recognize both their subjection to moral demands and, on this basis, their freedom of will. We cannot willfully choose to withhold our assent to judgments when we are conscious of sufficient proofs for these judgments, even if they conflict with our strong desires or wishes (JL, 9:74).68 While our consciousness of the moral law is (as we saw) not based on any proof, we recognize the moral law as an apodictically certain, indubitable fact of reason (KpV, 5:29–31, 47; MS, 6:225). Hence, we cannot refuse to recognize the supreme normative authority that this law has over us when we are inclined to transgress it: the moral law “thrusts itself upon us,” “whatever inclination may say to the contrary” (KpV, 5:31–2); the voice of reason here is “irrepressible” (5:35).69 Since rational agents recognize that being morally obligated implies having free will, the fact that they are rationally compelled to recognize that they are morally obligated entails that they are also rationally compelled to recognize that they have free will. Those who are conscious of moral norms “must concede without hesitation” that they have the freedom to act as morality demands, even if they would like to pretend that

67  This is implied by Pereboom 2005: 563–4; I discuss his view in the concluding section. Likewise, for Frierson (2011: 89–90) one can act according to moral judgments without believing that one has free will. In support of this claim Frierson cites KU, 5:451–2, but here Kant says only that one can be a righteous moral agent without believing in God. Presumably, Frierson here implicitly relies on the Parity Thesis. But Kant stresses that agents can only act from the moral law if they are conscious of their free will (FM, 20:306). 68  For Kant’s anti-­doxastic voluntarism, see Chapter VI; Buroker 2017; Chignell 2007a. 69  This is of course compatible with making immoral choices, through akrasia, self-­deception, or rationalization (RGV, 6:32, 35–7; MS, 6:379–80). See Chapter V and Hill 2011: 7–59.

Kant ’ s Moral Grounding of Free Will  287 their empirical desires are irresistible so that they can find an excuse for perpetuating a morally corrupt (e.g., hedonistic) lifestyle (KpV, 5:29–30). Thus, rational agents who sincerely denied their freedom of will would lose not only their knowledge that they are free (since knowledge of p requires assent to p) but would also, thereby, lose their knowledge that they are subject to valid moral demands. One might worry that this reverses the epistemic dependency relation Kant sees between morality and freedom: if we require knowledge of freedom to sustain our knowledge of moral duties, does this not undermine Kant’s point that knowledge of the moral law has epistemic priority over our knowledge of freedom? To see why this worry is misplaced, consider an (imperfect) analogy. I can know that (A) there is water on planet Omega without yet knowing that (B) the atmosphere of Omega has the chemical composition XYZ. Assume that (B) is a necessary real (e.g., causal) condition of (A). If so, my denial of (B) would undermine my knowledge of (A).70 Yet my knowledge of (A) might well be part of the inferential basis that is the epistemic ground for my knowledge of (B). This illustrates that the dependency of (B) on (A) in the order of (finite) knowledge is compatible with the dependency of (A) on (B) in the order of being and also with the fact that a denial of (B) would undermine one’s knowledge of (A). Likewise, for Kant the fact that knowledge of the moral law provides the (only) epistemic ground for our knowledge of free will is compatible with the fact that a (sincere) denial of free will (qua necessary real condition for the existence of the moral law) would undermine our moral knowledge. I have argued that Kant’s moral epistemology of freedom has the following structure: since we are indubitably certain that we are bound by the moral law (II) and since we can recognize that we would not be bound by the moral law if we lacked free will (III), we can infer with indubitably certainty that we have free will. Before examining (in VIII.3.2) how Kant defends the mediating premise (III) that is needed for that inference, I consider an alternative reading on which Kant allows us non-­inferential knowledge of freedom. On that reading it is a mistake to drive a wedge between knowledge of the moral law and knowledge of free will, i.e., to posit a gap between two separate items of knowledge that needs to be bridged via an inference. On this view, consciousness of the moral law is itself constitutive of the existence of free will since the causality of freedom is essentially a form of self-­conscious activity governed by the moral law. Since my consciousness of acting under the moral law is just what my having free will amounts

70  This might not follow according to externalist views of knowledge. But in Kant’s internalism (see Chapter VII, n. 69), S knows some proposition (A) only if S’s assent to (A) is based on S’s grasp of considerations that objectively support the truth of (A): this is the objective sufficiency plus communicability condition of knowledge. If S denies some fact that is a necessary real condition of the truth of (A), S’s assent lacks objective sufficiency.

288  Kant on Freedom and Rational Agency to, my moral self-­awareness is at the same time the awareness that I have free will.71 This reading correctly stresses that moral law and free will are objectively inseparable. However, in the subjective order of finite human cognition only the moral law is directly (self-) “given” to our consciousness: morality has epistemic priority over freedom because we have an immediately certain cognition of moral norms that we lack with regards to supersensible properties like freedom (KpV, 5:29).72 We can only grasp our noumenal free will via a cognitive route from something more familiar, namely from our normative-­practical knowledge. This route involves an inference: “. . . Duty commands unconditionally: one ought to stick to [one’s firm moral principles]; and from this one infers rightly: one must be capable of doing so, and that one’s faculty of choice is therefore free” (RGV, 6:49). The non-­inferential reading also has a problematic upshot. On this reading, our moral self-­consciousness yields immediate cognition of free will because this consciousness is itself constitutive of freedom and thus makes us free. This looks suspiciously like the model of cognition that Kant reserves for God, whose creative intellectual intuition is both an epistemic ground of cognition and an ontological ground of what it cognizes.73 By contrast, in our finite cognition of free will our consciousness of the moral law allows us to recognize a property that we do not create. Our (objective, non-­illusory) moral self-­awareness already presupposes, as its real ground (ratio essendi), that we possess a free autonomous will.

VIII.3.2.  Why Free Will Is Not on Par with God and Immortality Proponents of the Parity Thesis deny that Kant justifies the mediating premise which is needed for the inference from moral knowledge to knowledge of free will: namely, the claim (III) that we can recognize our freedom as an objectively necessary real condition for our certain moral duties. They ask: why should our known moral duties objectively depend on the actual existence of free will? Why would Kant not hold, instead, that our known moral duties depend merely on our

71  For this reading, see Schafer 2018. 72  The moral law is “inseparably connected, even identical to consciousness of freedom of will” (KpV, 5:42). But this means only that the concept of the moral law is in fact [“in der Tat”], objectively, the same as the (positive) concept of freedom (KpV, 5:28). In the subjective order of our cognition the moral law is nevertheless prior to the concept of freedom: “. . . it is the moral law, of which we become immediately conscious, which presents itself to us first,” and which (only) then “leads us straight to the [positive] concept of freedom” (KpV, 5:28–9). 73  Schafer 2018 accepts that on his account, our cognition of our freedom is “akin to the maker’s knowledge,” giving us something “very like a faculty for intellectual intuition.” But Kant stresses that we “must not at all” assume anything like an intellectual intuition for our cognition of freedom (KpV, 5:31).

Kant ’ s Moral Grounding of Free Will  289 subjective belief that we are free, just like they depend on our subjective (objectively uncertain) beliefs that we are immortal and that God exists?74 Some answer this challenge by focusing on the kind of spontaneity that is involved in the self-­legislation of pure practical reason: our legislative freedom of will (cf. Chapters III–­IV) is the “ground of being” of the moral law since the moral law is based on the spontaneous capacity of our rational will (qua “universal practical reason”; GMS, 4:431) to give a law to itself.75 The moral law existentially depends on the noumenal capacity for spontaneous self-­legislation that rational beings possess as supersensible “intelligences” (GMS, 4:452–4). By contrast, the moral law does not depend on our immortality or on specifically divine legislation (cf. GMS, 4:442). I agree that (III) rests in part on Kant’s appeal to the capacity for spontaneous moral self-­legislation. But in my view Kant’s claim that free will is a necessary real condition for the moral law also refers to our capacity for spontaneous choice, i.e., to our executive freedom of Willkür. Here I am not just relying on the passage where Kant claims that the legislative will (pure practical reason as a law-­giving faculty) cannot be called “free” since the concept of freedom applies only to the faculty of Willkür (MS, 6:226; see my discussion in Chapter  III). More importantly, our immediate consciousness of the moral law discloses what kind of freedom is the ratio essendi of the moral law (KpV, 5:4). As Kant’s explication of the fact of reason makes clear (cf. KpV, 5:30), the ratio essendi of the moral law includes our freedom of choice. Our consciousness of the moral law reveals to us (via the modus tollens inference explicated in Section VIII.3.1) that the moral law would not exist as a valid norm that governs our will if we lacked the capacity to obey the law, i.e., if we lacked the freedom to choose in accordance with self-­ legislated moral commands.76 The objective reality of the moral law depends both on our legislative spontaneity and on our executive freedom of choice. This is because our capacity to legislate objectively valid laws for our choices and our capacity to choose in conformity with these laws cannot come apart (cf. Chapter V). If the source of moral normativity were independent of our rational will (say, a divine authority or a Platonic realm of intrinsic values), there might well be a gap between how we should choose and what our will can accomplish.77 But the rational will cannot normatively bind itself to actions that it cannot perform, for this would undermine its legislative authority: it would render its putative normative law a vain “figment of the imagination” (GMS, 4:445). If the very source of our moral 74  I take Wood (1974: 35–6) to raise this crucial question. 75  This approach is suggested by Carnois 1986 and Kain 2010; see also Rawls 2003: 325. 76  Likewise (pace Kain 2010: 225), Kant stresses that from the validity of the categorical imperative we can infer our executive volitional capacity for “autocracy,” our ability to master the sensible inclinations that conflict with morality (MS, 6:383). See n.17. 77  See Darwall 2006: 139–40.

290  Kant on Freedom and Rational Agency obligations is our own free rational will, then these obligations must be tailored to our powers of free rational choice. Thus, the ratio essendi of the moral law is the autonomy of our will in a wide sense that comprises both our power to legislate this law and our power to choose in agreement with or “from” this law (as a sufficient a priori rational motive or “incentive” for our choices): For, this moral law is grounded in the autonomy of [the human being’s] will, as a free will which, according to its universal laws, must necessarily be able at the same time to agree with that to which it ought to subject itself.  (KpV, 5:132) So far as morality is based on the conception of the human being who is free but who also, just because of that, binds himself through his reason to unconditional laws, it is in need . . . of [no] incentive other than the law itself.  (RGV, 6:3)

If the normative standard that our will imposes on itself commands that we ought to do something, it follows (due to the inescapable normative logic of rational self-­legislation) “inevitably” that we can do it (RGV, 6:50). The fact that moral norms derive from our rational self-­legislation is the (for Kant, only) reason why moral ought necessarily implies can. This suggests an argument for the claim (III) that we can recognize our free will as an objectively necessary real condition for our actual, known moral duties. Objectively valid moral norms that truly bind or govern rational agents can only arise from autonomous self-­legislation. A rational being cannot legislate valid rational norms for itself unless it can also conform to these norms. We cannot conform to rational moral norms unless we possess a free executive power of rational choice. Hence, we cannot be truly governed by valid moral norms unless we have free will as a rational power for moral legislation and execution. However, proponents of the Parity Thesis might respond here as follows. In Kant’s view, we can conform to valid moral norms only if we believe that God and immortality exist. But (as Kant stresses) this does not entail that God and immortality actually do exist. Hence (so the response goes), our ability to comply with valid moral norms requires just our belief in the existence of supersensible entities such as God or also free will—­it does not require or entail that these entities actually exist. So, our knowledge that we are governed by and (thus) can conform to valid moral norms does not entail or give us knowledge that we actually have a free noumenal will. To answer this objection, I must show: our knowledge that we are governed by and (thus) can conform to valid moral norms objectively presupposes that we really do possess a noumenal free will but does not objectively presuppose that we really are immortal or that God really exists. The first thing to note here is that Kant posits two levels of conformity to the moral law, a “material” and a normatively prior, fundamental “formal” level: “The moral law is first of all a formal condition that reason imposes for the use of our

Kant ’ s Moral Grounding of Free Will  291 freedom of will; as such, the moral law obligates our will by itself, without depending on any purpose as material condition. But the moral law also determines for us . . . a final purpose . . .”—namely, the purpose to pursue the highest good (KU, 5:450). Those who did not believe in God could not comply with the moral demand to pursue this final purpose, but this would not absolve them from being formally obligated: No! Only the intending of the . . . final purpose . . . would in that case have to be abandoned. Every rational being would still have to cognize itself as rigorously bound by the prescriptions of morals; for the laws of the latter are formal and command unconditionally, without regard to purposes (as the matter of volition).  (KU, 5:451)

One might wonder what morality could prescribe while abstracting from determinate purposes. Kant explains that there are basic duties which do not “­concern . . . a determinate purpose (matter, object of the faculty of the choice) but merely the formal in the determination of the will (for instance, that the action that accords with duty must also happen from duty)” (MS, 6:383; cf. MS, 6:395, 410). The highest formal command of morality is that our choices must accord with the spirit rather than merely the letter of the moral law (KpV, 5:71–2; RGV, 6:30): we must choose on the basis of pure practical reason, due to our respect for the objective a priori moral law, thus independently of our subjective, empirically given sensible inclinations (cf. Chapter V). The recognition of this formal requirement is central to the moral self-­consciousness that constitutes the fact of reason: “the moral law . . . presents itself as a principle not to be outweighed by any sensible conditions, nay, wholly independent of them” (KpV, 5:29–30). Accordingly, someone who is aware of being formally obligated can infer that he possesses the power of choice “to overcome his love of life, however great it may be” (KpV, 5:30). The term “love of life” here designates both our generic sensible wish for happiness and the various specific empirical desires that are subsumed under this wish and provide it with concrete desiderative content.78 The command that we subordinate our empirical desires to the normative weight of a priori moral reasons is “formal” because it can be understood without reference to any particular effects or states of affairs that we might intentionally strive to produce as our concrete “material” purposes or ends. Accordingly, successful conformity to this formal demand does not depend on whether any material purpose is in fact realized (GMS, 4:394; KpV, 5:36–7). Formal conformity to the moral law depends exclusively on the internal character of our will rather

78  We seek fulfillment of these desires in this sensible life—­hence our love of life qua wish for maximal sensible gratification is jeopardized by the threat of being executed if one refuses to defame an innocent person (KpV, 5:30).

292  Kant on Freedom and Rational Agency than on its external manifestations.79 To illustrate: suppose Tom has promised to help his sister move but on the moving day he feels exhausted and would much prefer to relax at home. Tom fully complies with the formal moral demand to choose independently of sensible conditions if he makes a choice (and a most sincere effort) to keep his promise that is based on the a priori motive of duty, i.e., that is motivated simply by his awareness that there are conclusive, objective practical reasons which normatively bind him to keeping his promise. The formal conformity of his will to the moral law is not impugned if he then has a car accident that prevents him from helping his sister move. Now, we can comply with the formal command to act purely from a priori moral reasons only if we really possess the rational freedom to choose independently of all sensible-­empirical conditions. If we lacked this rational power of choice, i.e., if our choices were determined by our empirically given incentives, then we could not meet the most fundamental normative demand that the moral law makes upon our will. By contrast, our beliefs in God and immortality are irrelevant for obeying the moral law at this basic formal level. Moreover, the material command to pursue the highest good presupposes not only the belief in God/immortality but also our actual freedom to choose on the basis of necessary moral reasons. This is because we can be subject to a strict rational command or duty to pursue the highest good only if we possess the capacity to pursue the highest good from a purely rational motive: namely, from respect for the moral law rather than from our sensible desire for happiness in an afterlife. If we could pursue the highest good only due to the motivating force of this desire, we would not be acting under a strict duty but under a comparatively uncertain, non-­ binding prudential counsel recommending (in the vein of Pascal’s Wager) a strategy to increase our prospects for obtaining happiness in the afterlife. Thus, our conformity to all (formal and material) duties presupposes that we actually possess the power of free rational choice, whereas we need the belief in God/immortality only for complying with a specific material duty. This is one important asymmetry between the actual possession of free will versus faith in God and immortality qua conditions for complying with valid (self-­legislated) moral norms. But there is a further crucial asymmetry here: our need to believe in God and immortality hinges on two components of our subjective finitude that are extraneous to the objective moral law and that do not affect the relation between morality and free will. First, the very purpose of pursuing the highest good depends upon the conative limitations of finitely rational agency. Our need to act under the idea of a state of affairs in which virtue and happiness are appropriately combined rests, 79  Likewise, the dignity of rational persons (who are capable of autonomous legislation and choice) is a “formal motive”: this dignity cannot be physically effected but is rather to be “esteemed, preserved and furthered” (Wood 1999: 115) through an inner attitude of caring respect (GMS, 4:437).

Kant ’ s Moral Grounding of Free Will  293 partly but essentially, on our sensible wish for happiness.80 The original formal content of the moral law abstracts entirely from the concept of happiness and consequently from the concept of the highest good because the moral law applies to all rational beings including a self-­sufficient divine will that does not strive for happiness (KpV, 5:32). By contrast, the original formal content of the moral law does not abstract from the power of free rational volition. In its normative essence, this law is a formal prescription to act on the basis of objective reasons that are valid for every being which has (universal) practical reason. As a self-­ legislated norm of reason, the normative content of this law must be tailored to the rational powers of its (self-)addressees. Hence, all addressees of the moral law (including God: RGV, 6:50) must have the free rational power to act on the basis of objectively-­universally valid reasons. Second, the belief that divine assistance and immortality are required for the real possibility of the highest good is not based on an objective insight into the necessary conditions of the highest good. Our finite reason can only conceive the highest good as a really possible outcome if it connects this outcome to our immortality and divine assistance, but for all we can objectively know this might be a merely subjective conceptual constraint which need not coincide with the objectively necessary conditions for the real possibility of the highest good (KU, 5:455). Since our finite “reason cannot objectively decide” (KpV, 5:145) whether God and immortality are real conditions of the highest good, it is epistemically possible that the highest good might come to be without God and immortality. Hence, it is also epistemically possible that the moral law could validly command us to pursue the highest good if God and immortality do not exist.81 By contrast, it is not epistemically possible that the moral law could issue a valid formal command that we act from purely rational motives, or that we pursue (from such motives) any material moral purposes (such as the highest good), if we lacked free will. We know for certain that we could not legislate a valid normative law for our choices if we lacked the freedom to comply with the formal normative essence (“spirit”) contained in this law: namely, with the directive to

80  More precisely: happiness is a partial component (in addition to virtue) of the highest good qua necessary object of finitely rational wills; it is not part of the (purely rational) motive that virtuous agents adopt when pursuing the highest good (KpV, 5:109–10). A divine being wills that those who need and are worthy of happiness obtain such happiness (KpV, 5:110). But since a self-­sufficient divine being does not need or desire happiness for itself, the purpose of the highest good depends, essentially, on the conative neediness of finitely rational wills. 81  This point is important for interpreting a controversial passage (KpV, 5:114) where Kant might be taken to assert that the truth of the moral law depends on whether the highest good can really (come to) exist. I do not agree with this strong reading since it conflicts both with the axiomatic status of the moral law (KpV, 5:29–32, 46–7) and with Kant’s abovementioned point that even if one could not pursue the highest good one would still be bound by the true moral law (KU, 5:451). But even if the strong reading (of KpV, 5:114) were correct it would not follow that the truth of the moral law depends on whether God and immortality exist, because (for all we can know) the highest good which the moral law obliges us to pursue might (come to) exist without God and immortality.

294  Kant on Freedom and Rational Agency choose on the basis of objective, universal, a priori necessary practical reasons that are independent of contingent, egocentric empirical motives. Since the only possible source of a valid practical law is the self-­legislation of pure practical ­reason and since legislative and executive freedom are inseparably intertwined, a valid practical law could not apply to our will if we lacked the executive power of purely rational choice, namely, the (“positive”) ability to adopt an objectively necessary, desire-­independent normative standard as a sufficient rational “incentive” (RGV, 6:3) for our choices. This requires our (“negative”) freedom to choose without being determined by contingent, subjective empirical desires. In other words: if practical reason could not by itself determine our will as “pure” (KpV, 5:3), “empirically unconditioned” (KpV, 5:15) reason, then we would also lack a “higher faculty of desire” as the legislative source of priori necessary practical norms (KpV, 5:25). Conversely: since we have certain knowledge that we are bound by self-­legislated, a priori necessary practical norms, we can knowingly infer that we also possess the power to satisfy these norms: namely, the power to choose on the basis of a priori necessary practical reasons and thus independently of determination by empirical desires (RGV, 6:49). This power is our transcendental freedom of will (GMS, 4:461; KpV, 5:29; RGV, 6:29).82 To summarize: we can recognize the existence of our free will as an objective real condition for the existence of all valid (basic formal and derived material) moral duties. By contrast, our need to believe in God and immortality only attaches to a specific material duty and further depends on our subjective (conative and cognitive) limitations. Therefore, our certain knowledge that we are governed by valid moral norms removes the epistemic possibility that we might lack free will but leaves us with the epistemic possibility that God and immortality might not exist.

VIII.4.  Knowledge of Noumenal Free Will and Kant’s Critical Epistemology I have argued that in Kant’s considered (post-­1785) view we can know that our noumenal selves have the capacity for transcendental (legislative and executive) 82  My account of Kant’s moral proof of free will differs from the following reconstruction suggested by Proops (2021: 312): (1) We know our moral duties; (2) We know that ought implies can; (3) We know that we sometimes fail to act as we ought to; (4) Hence, we can know that we could have acted differently when we performed a bad action. There are two problems with this argument. First, it introduces (2) as a mere premise without explaining on what grounds we can know that moral duty and the power of compliance are inseparably intertwined. Second, with (3) this argument makes our knowledge of free will dependent on the empirical cognition of our bad actions. But Kant’s a priori proof that we possess free will as the ratio essendi of the moral law cannot depend on the empirical appeal to actions that have a merely negative (privative) moral quality: this cannot reveal the existence of our supersensible positive power to determine our will from pure grounds of choice (cf. MS, 6:226–7 and Chapters III, V).

Kant ’ s Moral Grounding of Free Will  295 freedom of will. In this section I consider the worry that the claim to know the existence of this metaphysical, supersensible property is inconsistent with Kant’s critical epistemology. This worry can take several forms. Here is one straightforward way of framing the worry. In order to know our noumenal free will, we would need objective cognition of noumenal free will. But we cannot have objective cognition of noumena because all objective cognition must have some sensible content. In Chapter  VII, I argued that Kant’s concept of “knowledge” incorporates indeterminate judgments and is thus broader than his concept of determinate, sensibly informed “cognition(N).” I further argued that via this distinction Kant’s epistemology can grant us indeterminate knowledge that we have an absolutely spontaneous theoretical intellect even though we have no determinate cognition(N) regarding our freedom of thought. The abstract idea of a free theoretical intellect does not entail the determinate metaphysical characteristics which figure in Kant’s critique of rational psychology. However, this argument cannot secure our entitlement to knowledge of free will because the idea of a free moral will does have a determinate metaphysical content: it designates a temporally unconditioned causal power (a “true causality”) and it strictly implies the existence of persons who are loci of moral accountability, whose identity persists in some (analogical) sense, and who are thus substantial unities of some sort. Because our belief in free will has this determinate metaphysical content, it requires a determinate and (qua knowing assent) objective cognitive basis. But not all objective-­determinate cognition must have intuitive-­sensible content. As I mentioned earlier (in Section VIII.2.2), Kant restricts the condition that our objective-­determinate cognition requires an intuitive component to theoretical cognition(N) (FM, 20:273). Our objective-­determinate cognition of our noumenal free will is a practical cognition that derives from our purely intellectual idea of the moral law. Although this idea is independent of all sensible intuition (GMS, 4:452; KpV, 5:31), its degree of objective-­universal validity equals (or even surpasses) that of our objective theoretical cognitions (JL, 9:66–7). This invites the question of how the idea of the moral law can supply a purely intellectual, non-­ sensible, yet objective cognitive content for the idea of an unconditioned causality. I have already addressed this question in Chapter III where I gave a detailed account of how the moral law provides practical meaning and objective reality for the idea of a free causality. I believe that similar accounts can be given to show how the moral law provides objective practical content for cognate ideas such as those of non-­temporal substantiality and personal identity. But this requires further complex considerations which would take me too far afield from the main topic. One might insist that my reading conflicts with Kant’s denial that we can cognize the real possibility of objects on a purely conceptual basis apart from sensible

296  Kant on Freedom and Rational Agency intuition.83 However, in the B-­edition of the first Critique Kant explicitly allows that pure reason can be our basis for cognizing that a concept has “real possibility” or “an object corresponding to it”: “To cognize an object I must be able to prove its [real] possibility, either from its actuality as attested by experience, or a priori by means of reason.” He adds that the grounds for attributing real possibility to a concept “need not . . . be sought in the theoretical sources of cognition; it may lie in those that are practical” (Bxxvii). Our a priori rational knowledge that we are governed by moral norms gives us practical cognition of our free will. It thereby takes over the epistemic role that sensible intuition plays in our theoretical cognition of objects (KpV, 5:56; cf. Chapter III).84 Since the moral law is a source of objective cognition that provides for “real” (normative) rather than merely “logical” representational content, Kant’s appeal to our practical cognition of freedom does not violate his admonition “not to conclude immediately from the possibility of concepts (logical [possibility]) to the possibility of things (real [possibility])” (A597/ B625). Likewise, since the judgment that we are morally obligated is synthetic a priori (GMS, 4:445–7; KpV, 5:31), the inference that noumenal free will exists (in us) derives from a synthetic ground of moral cognition. Thus, this inference does not violate Kant’s principle that existence claims cannot be analytic. The worry about a conflict between my reading and Kant’s critical epistemology might take another form. On my interpretation, Kant argues that our knowledge of the descriptive metaphysical claim that we have supersensible free will can be inferred from normative judgments concerning our moral duties. One might argue that there is something illegitimate about inferring a metaphysical “is” from an “ought.”85 Now, Kant clearly does accept this general form of (ought-­is) inference: in his view our hope for the highest good and its noumenal conditions rests on “the inference that something is (regarding the final possible purpose) because something ought to happen” (A806–7/B833–4). So, the worry must be that a practical ought cannot ground knowledge of a non-­normative noumenal fact. I accept that our normative moral knowledge cannot establish knowledge of our noumenal free will all by itself. On my reading, Kant’s inference from morality to noumenal free will depends on a background view that already incorporates metaphysical assumptions: namely, the idealist framework whose relevant aspects I examined in Part 1. I argued that this idealist framework is necessary but not yet sufficient to ground the legitimate assertion that we possess noumenal freedom of will. When Kant in the second Critique invokes the moral law as the basis for inferring (knowledge of) our noumenal free will, he thinks he has already shown in the first Critique that there is ontological space for the existence of a sensibly

83  See Chignell 2014: 595–7. 84  On this point, see also the helpful discussion in Kain 2010: 219–20. 85  This the inverse version of the worry that in Groundwork III, Kant illegitimately seeks to derive an “ought” from an “is” (for that worry, see Ameriks 2000a, 2003; Guyer 2005).

Kant ’ s Moral Grounding of Free Will  297 unconditioned form of causality. There he established that the spatiotemporal, causally determined things we know through experience are only sensible phenomena which must have some non-­sensible constitution in and by themselves. Our knowledge of the moral law further reveals to us that the non-­sensible constitution of rational beings like us whose will is governed by this law includes the causality of freedom. Thus, the metaphysical conclusions regarding free will that Kant seeks to derive from our normative-­practical self-­awareness do not come out of nowhere: our consciousness of moral duty determines an indeterminate metaphysical framework that is already in place.86 These considerations may still be deemed insufficient to (objectively) justify the strong metaphysical-­theoretical commitments that one incurs when one infers the belief in an atemporal, inscrutable type of causality from a normative-­ practical basis. I will return to this important lingering worry about Kant’s moral proof in the conclusion of this chapter and in Chapter IX. For now, I want to consider yet another version of the objection that Kant’s critical system does not permit knowledge of noumenal free will. My argument in Section VIII.2 for the claim (II) that we know the moral law relied partly on Kant’s appeal to our moral knowledge in the first Critique’s Canon of Pure Reason. However, in the Canon Kant does not grant us knowledge that we have transcendental freedom of will (cf. A802/B831). This may seem to provide a textual case for skepticism about whether the appeal to moral knowledge (II) can support (III) a knowledge claim about supersensible freedom in Kant’s critical epistemology. I concede that the overall view Kant articulates in the Canon is incompatible with the position that I have reconstructed in this chapter. But this incompatibility does not arise from Kant’s critical epistemology. Rather, the Canon view rests on a claim about morality that Kant rejects in his later works: namely, on the ­heteronomous claim that the moral law can obligate us only via “promises and  threats” concerning our divinely regulated future happiness (A811/B839) (cf. A589/B617; see the appendix to Chapter I). This non-­critical view of morality vitiates the argument for (III) that I have reconstructed in Section VIII.3 since it has two (related) implications which block the inference from moral knowledge to knowledge of noumenal free will. First, the heteronomous view that the moral law can obligate us only via our sensible desire for our future, divinely ordained happiness entails that moral obligations do not presuppose a capacity to choose from pure practical reason alone, independently of any empirical incentives that affect us as members of the

86  This point is particularly relevant regarding Prauss’s criticism that Kant’s moral proof of freedom is a lapse into dogmatic rationalism (1983: 62–70). This criticism is based in part on Prauss’s deflationary understanding of Kant’s idealism (cf. Prauss 1974): for Prauss, Kant’s appeal to things in themselves in the first Critique does not already entail that sensible phenomena have supersensible correlates.

298  Kant on Freedom and Rational Agency sensible world. Thus, according to the Canon view there can be no normatively fundamental formal duty to choose solely from a purely rational-­moral incentive (the a priori desire-­independent “spirit” of the moral law). As I argued in Section VIII.3, our knowledge of this fundamental formal duty is required for the inference that we have a corresponding noumenal power for purely rational, desire-­ independent, thus empirically unconditioned choice. Second, if the moral law can obligate us only via threats and promises concerning our divinely ordained future suffering or happiness, then “God and a future life . . . are inseparable from . . . obligation” as such (A811/B839; cf. A633–4/ B661–2). Since we cannot know whether God and immortality really exist, the Canon view entails that our knowledge of our moral obligations cannot ground knowledge claims about the existence of supersensible beings which are inseparable from these obligations. This undermines the strategy to derive knowledge that free will exists from the claim that free will is inseparably tied to our known moral obligations. What is missing from the Canon view is the doctrine of autonomy that is a cornerstone of the two later Critiques where Kant claims that we know our free will on a moral basis. In those works, Kant argues that the moral law itself “is grounded in” (KpV, 5:132) an act of autonomous self-­legislation. Through this act, the will prescribes to itself how it rationally must choose in accordance with how it rationally can choose without needing heteronomous motivational assistance from non-­ rational foreign causes such as the goading “promises and threats” concerning our happiness or misery in the afterlife. Since our moral obligations originate in the self-­legislation of our autonomous rational will, our formal moral knowledge that we ought to choose on the basis of purely rational motives (“from duty”) allows us to infer that we have the corresponding rational power of free choice. Qua autonomous law that our rational will legislates for itself in accordance with its own executive capacities, the a priori moral law itself is essentially connected with our executive power to choose independently of sensible-­empirical conditions. This renders free will “inseparable” from moral obligation in a way that God and immortality are not. Based on this autonomous conception of morality, Kant can now say (e.g., at KpV, 5:3–4) what he could not yet say in the Canon: free will, unlike God and immortality, is a real condition of all moral normativity. My interpretation makes only a selective use of the first Critique’s Canon. It relies (first) on the Canon’s tripartite classification of doxastic attitudes including the basic definition of “knowledge” and (second) on the Canon’s claim that our assent to moral judgments is a species of knowledge. These are both mainstays of Kant’s thought from (at least) the 1770s to the 1790s.87 The Canon’s refusal to 87  For instance, the Blomberg logic lectures which are usually traced to the early 1770s give an account of doxastic attitudes that closely resembles the Canon account (V-­LO/Blom, 24:142–50).

Kant ’ s Moral Grounding of Free Will  299 allow for knowledge of transcendental freedom is not dictated by Kant’s critical epistemology but by his (as of yet) uncritical, heteronomous moral philosophy. Against my claim that Kant changed his view on whether we can know our transcendental freedom of will in between the 1781 Canon and the 1788 second Critique, one might object that Kant did not revise the Canon in the 1787 B-­edition of the first Critique. Here my response is twofold. First, the issue of why Kant did not revise the Canon in the B-­edition arises quite independently of my interpretation. Apart from any issues concerning free will, there is a more basic question of why Kant did not revoke the Canon’s heteronomous account of moral motivation in 1787 even though he had already abandoned that account in the 1785 Groundwork. Second, I suspect that the reason why Kant did not revise the Canon is perhaps rather trivial or pragmatic. Kant indicates (in the Preface to the B-­edition: Bxl–­xli) that his revisions do not extend beyond the first chapter of the Dialectic because “time was too short to allow for further changes” and because “competent critics” did not object to later parts of the work. There are other intricate questions one can raise about the development of Kant’s thoughts on freedom and morality. For instance, there is much debate about whether Kant’s 1788 view that the moral law itself needs no deduction but can provide a deduction of free will is contradicted or anticipated by the 1785 Groundwork III. I cannot enter this debate here. I am inclined to view Groundwork III as a transitional writing.88 If it is indeed incompatible with the post-­1785 view that I have reconstructed in this chapter, this shows that in Groundwork III Kant had not yet arrived at his considered view concerning the relation(s) between morality and freedom. What about Kant’s famous B-­edition claim that he had to “cancel knowledge in order to have room for faith” (Bxxx), which he applies indiscriminately to God, immortality, and free will? The fact that Kant here does not discriminate between these objects is understandable because the first Critique, even in the 1787 B-­edition, provides no basis for such discrimination. Making that discrimination would have forced Kant to make the “further changes” (e.g., to the Canon) for which he found no time. The first Critique only shows that our knowledge that we are causally determined as sensible phenomena does not preclude our legitimate belief in free will (Bxxix; cf. Chapter I). This is the point of the departure for the more ambitious argument in the second Critique (KpV, 5:6). It is only by drawing Morality is here already viewed as a source of immediate apodictic certainty which is the mark of knowledge (V-­LO/Blom 24:225–8). 88  See n. 48. For the view that the Groundwork is a work in transition that falls in between the first Critique’s critical theoretical philosophy and Kant’s struggle to devise a correspondingly critical practical philosophy, see Ameriks’s pioneering work (2000a, 2000b, 2003) and more recently Baum 2014. Whereas Ameriks views the Groundwork strategy more favorably than the “dogmatism” he finds in the later fact of reason doctrine, Baum harshly dismisses Groundwork III as an immature “Kinderversion” (2014: 217) of the properly critical practical philosophy he finds in the second Critique.

300  Kant on Freedom and Rational Agency on considerations concerning the relation between freedom and morality which lie outside the scope of the first Critique that Kant can show that we know our free will whereas we are confined to mere faith in God and in our immortal souls.

VIII.5.  Conclusion: The Lingering Worry about Kant’s Moral Proof My argument in this chapter denies the common view that Kant’s critical philosophy leaves no room for determinate knowledge of specific transcendent objects. But something in the ballpark of this view remains true on my reading. Most obviously, this view remains untouched with regards to God and immortality, which on my reading are only matters of faith. Moreover, my claim that Kant derives knowledge of noumenal free will from moral considerations is congenial to the common view that Kant’s critical philosophy rules out (determinate) theoretical knowledge of noumenal objects. Accordingly, Kant denies that we can put our practically grounded knowledge of our noumenal free will to any theoretical use: through the moral law we can recognize that there is noumenal freedom of will, but “how freedom is possible, and how this kind of causality is to be conceived theoretically . . . is not thereby comprehended” (KpV, 5:133; cf. KpV, 5:3–4). Since our knowledge of free will does not stem from the theoretical sources of cognition, we can use this knowledge only from a perspective where we seek to determine how we ought to choose. In Kant’s critical epistemology we cannot theorize about the metaphysics of free agency: we cannot comprehend how the causality of freedom operates on the sensible world or explain why free agents perform certain acts at certain times (cf. Part 1). This is a severe epistemic limitation. Because he accepts it, Kant’s moral epistemology of free will differs decidedly from (how he conceived) the rationalistic attempt to demonstrate that an unconditioned causality of freedom is required for successful theoretical-­causal explanations (A445/B473). My reconstruction of Kant’s moral proof shows that Kant can preempt an important moral worry that some might raise about our belief in free will. Consider the following recent hard incompatibilist argument. Since our practices of holding people accountable typically involve sanctions and punishment, they are harmful to the people we hold accountable. Since these practices presuppose the belief that people were free to do otherwise, this belief is a source of harm and is therefore morally problematic. According to hard incompatibilists this significantly increases the justificatory burden for the belief in free will; a burden that (they argue) is already rather stiff since the belief in transcendental freedom is not supported by scientific evidence.89 89  For this worry, see Pereboom 2006 and (specifically in relation to Kant) Pereboom 2005: 263–4.

Kant ’ s Moral Grounding of Free Will  301 My account shows that Kant can respond to this objection in a twofold manner. First, he can deny that our belief that we are free, responsible agents is a matter of  responding to scientific theoretical evidence: the justificatory basis for this belief is our a priori practical self-­awareness (buttressed by the metaphysical-­ idealist assurance that the deterministic world examined by natural science does not exhaust all there is). Secondly, Kant can expose the hard determinist critics’ position as incoherent. The critics purport to raise a moral objection against our belief that people are free, responsible agents: they argue that this belief and the practices informed by it are problematic because they yield morally bad consequences such as causing people harm. But, as we saw, for Kant the belief that we are free and responsible (for complying with self-­legislated moral norms) is inseparably intertwined with the rational belief in morality itself, since free will is a ratio essendi of moral normativity. Thus, our rational acceptance of the normative standards that we presuppose in any moral criticism inescapably commits us to the belief in free will and responsibility. Hence, we cannot invoke these normative standards to raise doubts about our self-­conception as free and responsible agents.90 But although Kant’s moral proof of free will shields his view against such normative-­practical challenges, it still faces a lingering theoretical worry. Even if one accepts my argument (in Section VIII.2) that Kant’s appeal to a moral fact of reason is not guilty of a problematic normative dogmatism, one might insist that my reconstruction does not get around the worry about metaphysical dogmatism that afflicts the inference from morality to noumenal free will. This worry may persist among those who concede that we have highly generic, indeterminate knowledge (secured by Kant’s idealism) that there are non-­sensible aspects of reality. Even so, one might argue, purely practical-­moral premises cannot enable us to know the highly controversial determinate, theoretical-­ metaphysical claim that we possess an atemporal, unconditioned form of noumenal causality which supposedly produces specific effects in the sensible world. By Kant’s own admission, the causal relation envisaged here is completely inscrutable for us: the causality of freedom cannot be explained and cannot explain anything at all. This may be seen as an unwitting admission of absurdity, a vivid illustration of why only theoretical reason’s methods and principles can yield a legitimate epistemic basis for adopting beliefs about the causes of sensible events. According to this objection, we must not posit the existence of any causal 90  This is not to deny that Kantians should be sensitive to the worry that holding people accountable can have harmful consequences (especially given Kant’s view that outer behavior is not a reliable indicator of moral character); they should be open to (re)considering the extent to which retributive practices are morally defensible. Kant himself defends such practices mainly in the context of his political philosophy. I cannot consider this defense here. But even if it turns out to be flawed, this does not impugn Kant’s account of moral responsibility because blaming people for their bad behavior can take many different guises, some of which (perhaps the most basic ones) do not involve retributive actions or practices such as punishment.

302  Kant on Freedom and Rational Agency powers unless we can theoretically understand these powers and cite them in genuine explanations that allow us to make sense of why things happen as and when they do. It is not the business of practical reason to render existential, (thus) theoretical assertions about what exists or to introduce specific causal powers into our ontology, especially since these alleged noumenal powers cannot be integrated with our most fruitful theoretical reasoning: the causal belief regarding supersensible freedom that practical reason pretends to certify cannot play any part in and is even contrary to our ever more successful, expanding scientific understanding of why things in nature happen as and when they do. In Chapter IX, I argue that Kant acknowledges this worry and offers an in­tri­ guing response.

IX Kant’s Theoretical Defense of Moral Freedom Apart from the controversial claim that Kant grants us (not just faith but) knowledge that we have free will, my interpretation in the preceding chapter is fairly orthodox insofar as it takes Kant’s post-­1785 works to legitimize the belief in noumenal free will via practical-­moral premises. This seems congenial to the common idea that Kant’s argumentative strategy regarding free will undergoes a crucial shift in the 1780s: Kant initially thinks that he can prove our free will on theoretical grounds but then abandons this approach sometime during the mid-­1780s.1 As Desmond Hogan puts it, “Kant eventual[ly] opt[s] for a purely moral grounding of freedom.”2 If this were Kant’s final word, it would leave him open to the charge that the attempt to derive controversial metaphysical knowledge (a determinate noumenal is) from a normative-­practical basis is highly problematic, perhaps unacceptably dogmatic. In this chapter I argue that Kant is aware of this worry and addresses it through an intriguing theoretical defense of his practical-­moral proof that we have noumenal free will. This defense is based on considerations that are often seen as yielding Kant’s (alleged) pre-­critical theoretical proof of free will.

IX.1.  A Theoretical Defense versus a Theoretical Proof of Free Will We first need to disambiguate between two different ways in which one might understand the idea that Kant gives a theoretical proof of free will. First, this might refer to the role that Kant’s theoretical doctrine of transcendental idealism plays in legitimizing the belief in free will. Rather few commentators believe that Kant takes his idealism to provide a sufficient basis for adopting the belief in free will.3 As I argued earlier (Chapters I–­II), Kant’s idealism on its own yields no positive grounds for thinking that our noumenal selves have free will. Rather, Kant

1  Ameriks 2003: 176. See also Allison 1990 (chs. 10, 11); Henrich 1975. 2  Hogan 2009a: 380. 3  But see Ameriks 2003: 166–7 and especially Hogan 2009a, 2010. I discuss their views in Chapter I.

Kant on Freedom and Rational Agency. Markus Kohl, Oxford University Press. © Markus Kohl 2023. DOI: 10.1093/oso/9780198873143.003.0010

304  Kant on Freedom and Rational Agency views his idealism only (but significantly) as an indispensably necessary condition for adopting a legitimate belief in our supersensible free will. But the idea that Kant tries to defend the belief in free will on theoretical grounds can be taken in another way: various commentators argue that up until the mid-­1780s Kant seeks to establish free will by invoking our self-­consciousness of the spontaneity that we exercise in theoretical cognition and judgment.4 This strategy is commonly rejected via the claim that only our will or practical reason has the absolute spontaneity required for transcendental freedom whereas our theoretical intellect exhibits a merely relative kind of spontaneity that is compatible with determination by foreign causes.5 In Part 3, I have argued that our theoretical intellect shows absolute spontaneity and autonomy in its epistemic activity. Nevertheless, I agree with the majority of commentators that Kant cannot prove our free will by appealing to our a priori theoretical self-­consciousness (pure apperception). I am not convinced that Kant ever intended such a proof. The texts that commentators adduce to show that Kant pursued this proof-­strategy prior to 1787/8 (chiefly, Groundwork III, the Schulz review, and student lecture notes) do not, to my mind, conclusively establish their case. But my argument in what follows does not depend on this point (though I shall confirm it with regard to the Schulz review). I will argue for two claims. First (here I agree with common readings), Kant does not try to justify the belief in free will by invoking the epistemic spontaneity of theoretical apperception in his post-­1785 works. Second (here I disagree with common readings), the appeal to our epistemic spontaneity (freedom of thought) nevertheless serves an important argumentative function in Kant’s post-­1785 defense of moral freedom. A proof that infers free will from theoretical apperception might go as follows. (1) Our pure theoretical self-­consciousness justifies our belief that we have absolute spontaneity. (2) Absolute spontaneity is necessary and sufficient for free will. (3) Hence, we are justified in believing that we have moral freedom of will.6 Commentators tend to reject this argument as unsound by denying (1). In my view, (1) and (2) are true but the argument is invalid since it equivocates on the notion of absolute spontaneity: the absolute spontaneity of our understanding differs from and hence does not entail the absolutely spontaneity of our will. The main consideration blocking this entailment has come up many times in Chapters VI–­VIII: the absolute spontaneity we exercise in our theoretical cognition lacks the “true causality” that would pertain to an absolutely spontaneous will. A free noumenal will is the unconditioned cause of objects (events, states of 4  See especially Allison 1990 (chs. 3, 11); Ameriks 2003; Henrich 1975. 5  See Allison 1990: 69, 227–8; Henrich 1975: 72; Kitcher 1990, 2011; Schönecker 1999: 209, 299; Sellars 1970. 6 This inference is similar to what Horn (2015: 148, following Quarfoot 2006: 285) calls “the straightforward argument” for free will based on apperceptive-­theoretical self-­consciousness.

Kant ’ s Theoretical Defense of Moral Freedom  305 affairs) whose very existence (or occurrence) is produced by the will’s purposive representations. By contrast, a free finite noumenal understanding is merely a capacity to effect (without being determined from without) representations of objects, where these objects exist prior to the understanding’s representational activity. Hence, the fact that we have a capacity for absolutely spontaneous theoretical cognition does not show that we also have the “true causality” involved in free volition. Thus, there is no contradiction in imagining a being that has freedom of thought but lacks freedom of will. Such a being would be an autonomous source of a priori theoretical laws that govern spontaneous acts of judgment, but its practical judgments and choices would be heteronomous, i.e., would be determined by sensuous incentives. Kant himself seems to raise something akin to this logical possibility in the Religion (RGV, 6:26). But the failure of the above proof should not lead us to dismiss freedom of thought as irrelevant to Kant’s project of defending the belief in free will. Not every philosophical defense of a belief must provide a justificatory basis, such as a major premise in a logical inference, for that belief. A philosophical argument may also defend a belief by warding off an important challenge against it, without thereby supplying any positive reasons for adopting this belief. For instance, consider a popular defense of metaphysics against the positivist challenge that a statement is meaningless unless it is either a tautology or empirically verifiable. The defense consists in the argument that the verificationist principle itself satisfies neither of those conditions, so that the positivists have no coherent basis for their rejection of metaphysics. But this defense provides no reason for thinking that there are any meaningful or justified metaphysical statements. The positive reasons for asserting metaphysical statements must come from some other source. I shall argue that Kant’s theoretical defense of moral freedom fits roughly this pattern. My reconstruction has three component claims. First, Kant justifies our belief in free will without invoking our freedom of thought. Second, Kant acknowledges that this justification faces an important challenge. Third, Kant invokes our freedom of thought to ward off this challenge.

IX.2.  Kant’s Moral Proof and “the Omnipotence of Theoretical Reason” Since I considered the first component at length in Chapter VIII, I shall here only briefly recapitulate how Kant explicates the justificatory basis for the belief in free will (in the second Critique). When we adopt the normative-­practical standpoint we have an “irrepressible” (KpV, 5:35) consciousness of the moral law as the fundamental normative standard for all our specific rational choices. This consciousness is “pure” (a priori) because the moral law presents itself as independent of

306  Kant on Freedom and Rational Agency sensible conditions such as our contingent, empirically given desires or feelings. This independence enables the moral law to prescribe with strict necessity how we rationally speaking must, i.e., ought to choose regardless of what those empirical mental states happen to be. The moral law need not (and cannot) be derived from any further premises. It has the status of an immediately certain practical axiom because it provides the objectively-­universally valid basis for any mediate rational-­practical justification. This axiom gives us apodictically certain practical cognition that we ought to determine our will independently of empirically given incentives. We can further recognize that the moral law, qua autonomous principle that our rational will legislates to itself, cannot demand anything from us that does not lie within our rational power. Hence, we can infer that it must lie within our rational power to determine our will through our a priori conception of moral duty, regardless of and even contrary to our empirically given incentives or feelings. Thus, we can legitimately believe (indeed, know) that we have transcendental freedom of will, the power of rational choice which is free from empirical determination by sensible causes. Philosophical reason articulates this moral justification of the belief in free will in a clear, explicit form that is typically not transparent to common practical cognition. Moreover, common practical cognition is typically unaware of the fact that the inference from moral ought to freedom of will is defensible only within the metaphysical framework provided by Kant’s idealism (KpV, 5:95–6). Nonetheless (cf. Chapter VIII), common moral agents draw the inference from morality to free will and thereby grasp the objectively sufficient moral justification for believing in free will, however implicitly or obscurely. Thus, our positive reasons for believing that we have noumenal free will stem exclusively from our ordinary, pre-­philosophical moral self-­consciousness (KpV, 5:30–7). The second component in my reconstruction is the concession that this moral grounding of freedom invites a serious challenge. One must be careful, though, in how one frames this challenge. Some object to Kant’s moral justification of free will via the circle worry that Kant (arguably) raises in Groundwork III, i.e. by arguing that the objective validity of the moral law already presupposes and hence cannot serve to justify the belief in free will.7 Here Kant can reply that (whatever view he may have held in Groundwork III) the circle is broken through the main innovation of the second Critique: namely, the insight that while moral law and free will indeed condition each other they do so in different ways (KpV, 5:4; cf. Chapter VIII).8 Since we can recognize the moral law as a certain fact of reason that grounds all objectively-­universally valid practical justification, the moral law itself “needs no justifying reasons” (KpV, 5:47) apart from our common practical knowledge that we ought (not) to act in certain ways. Our objectively certain 7  Wood (2002: 33–4) raises this objection for Kant’s argument in the second Critique’s Preface. 8  See Bojanowski 2006: 191.

Kant ’ s Theoretical Defense of Moral Freedom  307 awareness of the moral law as a foundational (axiomatic) normative-­practical truth can thus serve as a sufficient epistemic basis for inferring the existence of whatever is a necessary real (as opposed to epistemic) condition for this truth. Since we know that our noumenal free will is such a condition, we can infer that we have free will from our epistemically prior moral knowledge. There is no circle here. As I see it, a potent challenge to Kant’s moral proof of free will is best framed as a doubt about whether normative-­practical judgments can serve to justify highly controversial theoretical assertions that are inscrutable to theoretical reason.9 This is not simply a generic concern about moving from an ought to an is. Rather, it is a more specific worry about invoking normative-­practical considerations as a basis for certifying a determinate theoretical, namely existential-­causal assertion that is at odds with our spatiotemporal experience and our theoretical understanding. Kant tries to justify, on a solely practical basis, the claim that we possess a supersensible, atemporal power of unconditioned choice that produces temporal effects in the sensible-­natural world. By Kant’s own admission, we cannot comprehend how this power could exist and how it is supposed to operate, i.e., how and why it produces or fails to produce specific sensible effects at specific times. This invites the protest that practical reason here oversteps its bounds by meddling in questions that only theoretical reason can legitimately and fruitfully address: claims about the causes of natural effects must be grounded in theoretical reason’s methods (e.g., testing hypotheses via experiments) and evidential standards (which ultimately rely on sensible experience). Similarly, one might argue that if one cites a cause for what happens in the natural world there must be a corresponding explanation that allows us to comprehend why things in nature happen as and when they do as the effect of said cause. Since the belief in transcendental freedom of will as an atemporal cause of observable human behavior fails these conditions, one cannot legitimize it (let alone ground it as knowledge) via practical reason. Leave it to theoretical reason to establish the causes of sensible phenomena! Only theoretically grounded beliefs can serve the integral function of causal assertions: namely, to explain and enable us to comprehend what happens in the natural order of things by citing causes whose temporal operations can be certified through the testimony of experience. This is, in any case, how Kant conceives the theoretical challenge confronting his moral justification of free will. To confirm this, consider his discussion of a naturalistic view that seeks to derive the moral law from empirical feelings like pleasure. This derivation entails that all our choices according to the moral law are also based on empirical mental states so that all our “conduct follows the

9 My framing of the challenge indebted to Ameriks’s critique of Kant’s account in the second Critique. See Ameriks 2000a, 2000b, 2002, 2003, 2013.

308  Kant on Freedom and Rational Agency natural order” and we lack free will. Kant traces this naturalistic view to more general commitments: Those who are accustomed merely to physiological explanation will not get into their heads the categorical imperative, from which these [moral] laws arise dictatorially . . . . But being unable explain what lies wholly beyond this sphere [of physiological explanation] (the freedom of choice) . . . they are stirred by the proud claims of speculative reason, which feels its power so strongly in other fields, to band together to a general call to arms (as it were) of those who are allies to the omnipotence of theoretical reason, in order to revolt against this idea and thereby to assail . . . and, wherever possible, make suspect the moral concept of freedom now and perhaps for a long time.  (MS, 6:378)

Naturalists believe in the “omnipotence of theoretical reason” and accept only “physiological explanations.” These are the explanations of natural science in general which include those of empirical psychology (KpV, 5:96–9). Such explanations view every state or event as the necessary effect of an empirical cause that operates according to some law of nature, where the operation (“causality”) of the cause is itself the effect of some further empirical cause, etc. (A543–4/B571–2; GMS, 4:446–7; cf. Chapter I). By contrast, our ordinary moral self-­consciousness entails that we can spontaneously determine ourselves to choose on the basis of a priori moral reasons regardless of what our preceding mental states, empirical desires, or occurrent feelings may be. That is, our ordinary moral self-­ consciousness supports the belief that we have a non-­empirical power of free choice that is inscrutable to the physiological explanations of theoretical reason. Those who are engaged in the naturalistic project of making the entire world theoretically comprehensible via physiological explanations are prone to see their efforts compromised by this belief. They insist that whatever existential or causal beliefs we can legitimately adopt from a normative-­practical perspective must meet the evidential, explanatory standards of the empirical standpoint. In imposing this demand naturalists feel empowered (“are stirred”) by the ever-­growing success that theoretical reason enjoys in explaining the natural world. Thus, naturalists are prone to challenge the moral basis for the belief in free will by trying to debunk the allegedly “pure” consciousness of an a priori moral law, i.e., by arguing that this is really an empirical form of consciousness that results from natural conditions such as given desires or feelings. This would mean that our moral consciousness is illusory: due to the “covert” influence that empirical desires have on our will or practical reason “the so-­called moral imperative, which as such appears to be categorical and unconditional, would in fact only be a pragmatic precept” that lacks unconditional, objective-­rational necessity (GMS, 4:419; cf. Chapter  IV). If this “naturalization” of our moral self-­consciousness were to succeed, it would conclusively undermine the inference Kant envisages in

Kant ’ s Theoretical Defense of Moral Freedom  309 his justification of free will, namely, the inference from our pure moral self-­ awareness to our supersensible freedom of will. Via this naturalization, both our moral judgments and our choices could be sufficiently explained as the upshot of empirical causes: we would lack both legislative and executive freedom of will. Kant’s recognition that the belief in moral freedom faces this theoretical-­ naturalistic challenge was likely sparked by philosophers like Schulz who raised something akin to this challenge.10 One might suggest that the naturalistic position is already refuted by Kant’s transcendental idealism because this doctrine entails that there are non-­sensible grounds of appearances that cannot be explained via natural causes. But this suggestion faces two (related) problems. I have repeatedly stressed the first problem: while transcendental idealism does entail that there are non-­natural grounds of sensible appearances, it does not show that these grounds involve an unconditioned causality of freedom. The appeal to transcendental idealism can mitigate the metaphysical burden incurred by the normative-­practical justification of free will (cf. Chapter VIII): given transcendental idealism, this justification can be seen as further determining an indeterminate non-­ natural (noumenal) reality whose existence has already been established on independent (non-­moral) grounds. But the challenge arises precisely for the idea that we can determine an indeterminate supersensible reality on practical grounds, where this determination is supposed to have specific implications for the ultimate causes of sensible phenomena. Naturalists insist that practical reason lacks the authority to posit theoretically inscrutable supersensible causes of sensible effects. On their view, the immense success theoretical reason has had in explaining what exists and happens in nature dictates that we must subject our normative-­practical self-­awareness to the standards of naturalistic theoretical reason and its physiological explanations; we cannot permit practical reason to legitimize beliefs in noumenal capacities that are inexplicable and that cannot explain their alleged natural effects. Second, while transcendental idealism indeed rules out a metaphysical naturalism which dogmatically denies the possible existence of non-­natural, empirically unconditioned causal powers (A445/B473), it is compatible with and indeed supports a methodological naturalism: a view on which all our explanations must refer to natural laws and causes.11 For Kant, the indeterminate non-­sensible reality that his idealism posits cannot serve our theoretical understanding because it lies beyond the bounds of our sensible experience. In his critical epistemology, all our explanatory efforts must invoke spatiotemporal causes governed by natural laws: “we can explain nothing but what we can reduce to [natural] laws whose object can be given in a possible experience” (GMS, 4:459). Therefore, 10  Another likely inspiration is Garve (cf. TP, 8:285). 11  For this distinction, see Leiter 2015: 2–5.

310  Kant on Freedom and Rational Agency transcendental idealism does not interfere with a naturalistic commitment to the exclusiveness of “physiological explanations.” To be sure, Kant sees in human reason a tendency to slip from a merely methodological naturalism to naturalism as a metaphysical worldview: those who are engaged in the project of explaining the sensible world via natural laws are prone to accept (under the influence of “transcendental illusion”) that the sensible world and the natural order are all there is (A468–72/B496–500). But since methodological naturalism does not strictly entail metaphysical naturalism, it can be combined with Kant’s idealist notion that there is some supersensible reality that we cannot cognize. Indeed, one might argue that a stance which combines a methodological naturalism concerning the determinate character of sensible reality with strictly agnostic epistemic humility concerning the determinate character of supersensible reality is most in line with Kant’s critical epistemology, whereas his moral proof of supersensible freedom as a cause of sensible phenomena signals a lapse into dogmatic rationalism.12 This seems sufficient to motivate a theoretical-­naturalistic challenge for Kant’s moral justification of freedom.

IX.3.  Kant’s Theoretical Defense Strategy: 1783 and 1788 I believe that Kant seeks to defuse this challenge by appealing to our freedom of thought. This is the third component in my reconstruction of how he legitimizes the belief in free will. One of the most significant passages here stems from Kant’s 1783 Schulz review. I already discussed this passage in Chapter  VII when I reconstructed Kant’s reasons for claiming that we must presuppose our freedom of thought. In the present context I consider this passage once more to explicate how the appeal to freedom of thought figures in Kant’s defense of free will against the theoretical-­ naturalistic challenge. Schulz argues, in Kant’s words, “that there is . . . no free will; rather, everything stands under the strict law of necessity” (RezS, 8:11). He adopts a “general fatalism” (RezS, 8:13) about all human thought and behavior. Kant traces Schulz’s denial of free will to his “speculative principles” (RezS, 8:14). Schulz’s position thus embodies the commitment to the “omnipotence of theoretical reason” that Kant regards as a threat for the belief in free will.13 Kant responds as follows: He [Schulz] has presupposed . . . that the understanding has the capacity to determine its judgment according to objective reasons that are valid at any time, 12  See Ameriks 2003; Prauss 1983: 62–70. 13  Schulz is not a full-­blown naturalist since he accepts a theistic grounding of natural laws. This is compatible with accepting “the omnipotence of theoretical reason” if one believes that God’s existence can be theoretically proven.

Kant ’ s Theoretical Defense of Moral Freedom  311 and does not stand under the mechanism of merely subjectively determining causes that may change subsequently; thus he always assumed freedom of thought, without which there would be no reason. In the same way, he must presuppose freedom of will in acting, without which there would be no morals, if he seeks to . . . proceed in accordance with the eternal laws of duty rather than being a play of his instincts and inclinations.  (RezS, 8:14)

Kant here does not infer our freedom of will from our freedom of thought.14 Rather, he regards our practical awareness that we are subject to “eternal laws of duty” as the epistemic ground of our conviction that we possess free will. Kant invokes freedom of thought only in response to Schulz’s speculatively grounded denial of this conviction. This response goes as follows. The “speculative principles” that are supposed to undermine the belief in free will, such as (chiefly) the general causal principle that every event is causally determined, are principles of theoretical reason (in the broad sense of “reason” which encompasses the faculty of understanding). One can only take these principles to be based upon objectively valid reasons if one assumes one’s “freedom of thought, without which there would be no reason.” Moreover, the belief in rational freedom of thought that one (however implicitly) adopts when exercising one’s capacity for theoretical judgment is analogous to the belief in rational freedom of will that one adopts when exercising one’s capacity for moral judgment and choice. Since the speculative principles of theoretical reason that one must presuppose to undermine the belief in free will depend on our freedom of thought, and since the capacity for free thought is analogous to the capacity for free volition, one cannot coherently seek to undermine the belief in free will. The fact that Kant employs this strategy in 1783 does not show that he still relies on it when he has fully developed his moral grounding of free will by 1788. But Kant supplements this moral grounding with the theoretical defense strategy from the Schulz review. To see this, consider a key passage where Kant explicates the moral fact of reason (KpV, 5:29–30). The passage begins with the reciprocity thesis:15 “Freedom [of will] and the unconditional practical law . . . reciprocally refer to each other.” Kant then raises the question of whether “our cognition of the unconditionally practical starts . . . from freedom or from the practical law.” He rejects the former alternative: we have no immediate grasp of the unconditioned causality of freedom. But we do have an immediate consciousness of the moral law as a normative axiom that has unconditioned practical necessity. “Therefore it

14  Pace (among others) Horn 2015 and Quarfoot 2006 (see n. 6) who read the Schulz review as presenting the “straightforward argument” from theoretical to practical freedom that I criticized in Section IX.1. 15  I take the term “reciprocity thesis” from Allison 1990: 7.

312  Kant on Freedom and Rational Agency is the moral law of which we become conscious directly . . . and which . . . leads straight to the concept of freedom.” Kant continues as follows: But how is even the consciousness of that moral law possible? We can become conscious of pure practical laws just as we are conscious of pure theoretical principles, by attending to the necessity with which reason prescribes them to us, and to the separating [from them] of all empirical conditions, to which that necessity points us. The concept of a pure will arises from [the consciousness of] pure practical laws, as the consciousness of a pure understanding arises from [the consciousness of] pure theoretical principles.

To understand this intricate passage we must keep in mind the overall dialectical situation. Kant argues: from our valid consciousness that we are bound by an a priori moral law which prescribes with strict practical necessity how we must choose independently of all sensible conditions, we can infer that we have a noumenal capacity to choose independently of all sensible conditions. A significant challenge to this argument arises from the denial that we can take our consciousness of being bound by an a priori moral law at face value. As we saw, Kant is well aware that those who accept the “omnipotence of theoretical reason” try to reduce everything to “physiological explanations” and thus seek to naturalize our moral self-­awareness. They claim that our allegedly a priori representation of a putatively unconditional practical necessity is really (secretly) conditioned by contingent empirical states (MS, 6:378; GMS, 4:419). If this transformation of our allegedly “pure” moral self-­consciousness into a form of empirical consciousness were to succeed, it would also undermine the belief in free will: if our beliefs in moral duties are themselves conditioned by sensible conditions such our contingent empirically given desires, then these moral beliefs cannot entail that we have the power to choose independently of empirical conditions. The question that Kant poses at the beginning of the above passage, “But how is even the consciousness of that moral law possible?” refers to such theoretical doubts regarding our moral self-­awareness. What is being questioned here is the possibility of a pure, non-­empirical self-­consciousness that (as such) cannot be theoretically explained as the upshot of sensible phenomena and thus relates to our noumenal selves. Those who are grappling with this “how possible?” question may well be led to the answer that a pure moral self-­awareness is really impossible if they also have (perhaps because they are awed by the immense success of natural science) naturalistic sympathies. According to naturalists, we should not countenance the possibility of a pure moral self-­consciousness because such a consciousness and the capacity for free choice that it entails are inscrutable to theoretical reason. We should rather treat our moral representations as belonging to our empirical consciousness of our phenomenal self, which is determined by

Kant ’ s Theoretical Defense of Moral Freedom  313 laws of nature and is thus susceptible to the “physiological explanations” of theoretical reason.16 In the remainder of the above passage Kant seeks to defuse this challenge in a way that is reminiscent of the Schulz review. It is noteworthy that Kant does not directly answer the “how possible?” question that he has posed: he does not explain how a pure moral self-­consciousness comes about. Such an explanation is impossible from Kant’s critical perspective because Kant is a methodological naturalist who accepts that our capacity for explaining things depends on our ability to trace the explanandum to empirical causes. This cannot be done in the present case without exposing our allegedly pure moral self-­awareness as an illusion. Rather than answering the “how possible” question directly, Kant points out that our pure moral self-­consciousness is structurally analogous to our pure theoretical self-­consciousness and that we can recognize our pure (empirically unconditioned) will in the same way that we can recognize our pure (empirically unconditioned) theoretical intellect: namely, by attending to the strict a priori necessity that is contained both in our pure practical and in our pure theoretical principles. For Kant, this analogy preempts the need for a direct answer to the “how possible?” question and defuses the theoretical-­naturalistic challenge to the possibility of a pure moral self-­consciousness. The theoretical cognizers who raise this challenge must accept that we do possess a pure, objectively valid self-­ consciousness of our empirically unconditioned rational understanding which defies all naturalizing efforts. Thus, they cannot coherently deny that we might also possess a pure, objectively valid self-­consciousness of our empirically unconditioned rational will. This of course invites the question of why theoretical cognizers of nature must accept that we have a valid pure self-­consciousness of our empirically unconditioned understanding. Since I have addressed this question at length in Chapter VII, I shall here only briefly recapitulate Kant’s main point. Unlike other commentators,17 I do not read Kant as claiming that if our theoretical judgments are causally determined they cannot be based on truth-­oriented reasons. If that were Kant’s point, naturalists could simply respond that our judgments about the natural world are based on truth-­oriented reasons precisely because they are causally-­evidentially determined, via perceptual data, by the natural objects that these judgments refer to. This response is encapsulated in Nagel’s claim that “in forming beliefs we may hope for nothing more than to be determined by the truth”18 or in White’s suggestion that “causal determination of belief is not

16  See the footnote at KpV, 5:6 for confirmation that the distinction between our pure and empirical self-­consciousness corresponds to the difference between our noumenal and our phenomenal self. 17  Such as Rickert, Horn, and Wood; see Section VII.1. 18  Nagel,1989: 116. Compare Wiggins 2003: 96, 1969: 143.

314  Kant on Freedom and Rational Agency worrisome as such” since determination by evidential states is “just what we want for our beliefs.”19 For Kant causal determinism as such does not rule out the possibility of truth-­oriented or reason-­based judgments. Rather, for Kant the idea that theoretical judgments are determined by natural, empirical conditions conflicts with a specific component of theoretical judgment that Kant deems essential to our entire theoretical cognition: “If the determining grounds were empirical and were given in an a posteriori subjective fashion, then the judgment of reason could not be regarded as a priori and thus could not be regarded as absolutely necessary” (Refl., 18:176). This Reflexion accords with the published texts cited above. In the second Critique Kant says that we derive the concept of our pure theoretical understanding from our consciousness of “the necessity with which reason prescribes” to us “pure theoretical principles,” a necessity which signals the independence of theoretical reason from contingent, mutable empirical conditions. In the Schulz review Kant argues that theoretical cognizers must assume their freedom of thought because only such intellectual freedom ensures that their “understanding has the capacity to determine its judgment according to objective reasons that are valid at any time, and does not stand under the mech­ an­ism of merely subjectively determining causes that may change subsequently.” Kant’s basic point here is that as cognizers who aim at objectively valid judgments about nature, we must take our judgments to be based on objective reasons which stem partly but essentially from our consciousness of a priori necessary cognitive laws that originate in our theoretical intellect. Such laws would be impossible if our intellect were determined by contingent empirical conditions that belong to “the mechanism of merely subjectively determining causes.” Naturalists might reply that we can cognize nature without a priori principles since all our objective theoretical reasons for judgment stem from the empirical evidence that we receive when natural objects causally affect our senses. But (cf. Chapter VII) this reply does not work for epistemically fundamental representation such as the concept of causality and the corresponding principle that every event is determined by some cause. Our causal representations are, due to their thick modal content, underdetermined by the perceptual data that we receive from the physical world. If naturalists revert to the idea that our representation of causal necessitation derives from the empirically given internal constitution that has been naturally “implanted in us” (e.g., when our brain got hard-­wired during the course of evolution), Kant’s response is that in this scenario our causal judgments have no objective rational basis: The concept of cause . . . which expresses the necessity of an event under a presupposed condition, would be false if it rested only on an arbitrary subjective

19  White 2010.

Kant ’ s Theoretical Defense of Moral Freedom  315 necessity implanted in us . . . . I would not then be able to say that the effect is connected with the cause in the object, that is to say, necessarily, but only that I am so constituted that I cannot think this representation otherwise than as thus connected. This is exactly what the skeptic most desires. For if this be the situation, all our insight, resting on the supposed objective validity of our judgments, is nothing but sheer illusion.  (B168)20

To save the objective validity of our causal judgments we must treat the principle that natural events are causally determined as a “self-­thought first a priori principle of cognition” (B167) that derives from the autonomy of our free intellect. This allows us to deny that our representation of causal necessity is the upshot of “arbitrary” psychological “customs” or other empirical (e.g., neurological) conditions beyond our rational control.21 Some naturalists might double down here and claim, in a (Neo-)Humean fashion, that the representation of causal necessitation is irrelevant to an enlightened empiricist model of causal reasoning.22 But this response is unavailable to theorists like Schulz who seek to challenge the belief in moral freedom of will via their conviction “that . . . everything stands under the strict law of necessity” (RezS, 8:11). If naturalists concede that their theoretical commitment to the “strict law of ” natural “necessity” is a mere upshot of contingent non-­rational mental habits, they compromise their rational grounds for denying that our practical judgments and volitions are free from natural necessity. Let me summarize the defense strategy I have ascribed to Kant. Freedom of thought is the capacity (negatively speaking) to judge without being determined by contingent, non-­rational empirical conditions or (positively speaking) to judge from one’s awareness of rationally necessary theoretical laws. Likewise, freedom of will is the capacity to choose without being determined by contingent, non-­ rational empirical conditions, from one’s awareness of rationally necessary moral laws. Kant utilizes this analogy when he argues (at KpV, 5:30) that we can recognize our pure, empirically unconditioned will in the same way that we can recognize our pure, empirically unconditioned understanding: namely, by attending to the a priori necessity that is expressed both in our fundamental theoretical and in our fundamental practical principles. The ratio essendi of this strict rational necessity is the unconditioned rational spontaneity or autonomy of our higher mental faculties. Hence, our objectively valid consciousness of a priori necessities is the ratio cognoscendi of both our epistemic and our practical spontaneity. Unless one is prepared to doubt the validity of necessary principles in general, 20  See Section VII.6 for my account of why this argument applies to both natural and divine imposition views. 21  The fact that this point is more pronounced in the B-­edition of the first Critique than in the earlier (A) edition (cf. B5; B127; B167–8) further confirms that Kant was keenly focused on these considerations in the late 1780s. 22  As in Woodward 1990: 558–9. For doubts about the viability of this approach, see Section VII.3.

316  Kant on Freedom and Rational Agency one must accept that our higher mental faculties in general are not determined by contingent empirical conditions. Those who accept that causal necessity reigns in nature cannot doubt the validity of necessary principles. Hence, they must accept the very idea they seek to debunk: they must concede that we possess a pure form of consciousness which entails our noumenal capacity for free rational self-­ determination under a priori necessary intellectual laws. If they stubbornly refuse this concession and insist that all our mental states and actions are the upshot of contingent empirical conditions beyond our rational control, they must also accept that all our representations of non-­trivial necessities, whether theoretical (e.g., causal) or practical, are mere “figments of the brain” (A91/B123–4; GMS, 4:445) to which we have no rational entitlement. I should emphasize once more that Kant’s appeal to freedom of thought here is not meant to prove our moral freedom of will. This is because (for the reasons given earlier) the appeal to freedom of thought supplies no positive reasons for adopting the belief in free will. Those reasons stem exclusively from the premises (examined in Chapter  VIII) that constitute Kant’s moral proof of free will. The appeal to freedom of thought protects this proof against the theoretical-­ naturalistic challenge considered above. The argumentative impact of this defense can be considered in rough analogy to the philosophical defense of metaphysical statements against the challenge raised by logical positivism. Positivists claim that metaphysical statements are meaningless since they are neither tautologies nor empirically verifiable. The philosophical defense strategy argues that the positivists’ own verificationist principle of meaning is neither tautological nor empirically verifiable either, from which it follows that the positivists have no coherent basis for attacking metaphysical beliefs. But this does not positively establish that there are any meaningful, justified metaphysical beliefs. Likewise, in his philosophical defense of the belief in free will Kant argues that naturalists who deny this belief must presuppose the rational validity of fundamental a priori theoretical claims (e.g., about the existence of necessary causal relations), a presupposition which commits them to believe in freedom of thought. Hence, they have no coherent basis for attacking the analogous belief in free will. But this does not positively establish that the belief in free will is true and justified. The justification for believing in free will rests on considerations that lie implicit in the ordinary a priori practical self-­consciousness of moral agents, apart from abstract philosophical arguments.

IX.4.  In Defense of Kant’s Defense Strategy In this section I consider two objections that one may raise against my reconstruction of Kant’s overall justification of the belief in free will.

Kant ’ s Theoretical Defense of Moral Freedom  317 According to the first objection, the attempt to invoke freedom of thought in a theoretical defense of the practical proof of free will fails for the same reason as the attempt to invoke freedom of thought in a direct theoretical proof of free will (cf. Section IX.1): the fact that these are two different species of freedom undermines the analogy on which Kant’s defense strategy relies. As I conceded earlier, one cannot directly prove freedom of will from freedom of thought because the idea of a free noumenal will is metaphysically richer and thus more controversial than the idea of a free noumenal understanding: a free noumenal will, if it did exist, would exhibit a “true” unconditioned causality that a free noumenal understanding lacks. But, so the objection goes, this disanalogy shows that naturalists can coherently deny that we have free will while accepting their freedom of thought. By accepting their non-­causal freedom of thought, they salvage their rational entitlement to the a priori theoretical principles which they need for undermining the belief in free will as an unconditioned cause of natural events. This objection fails for the following reason. Naturalists cannot (without further argument) deny the validity of Kant’s moral proof of free will, which infers that we have free will from our pure consciousness of normative-­practical necessity (cf. Section IX.1 and Chapter  VIII). So, they must show that this proof is unsound because its premises are mistaken. Without further argument, they cannot reject Kant’s premise that our empirically unconditioned will is a real condition for our pure moral self-­consciousness of normative-­practical necessity (see Chapter VIII, Section IX.3). Hence, to undermine Kant’s moral proof naturalists must argue that we lack a pure, objectively valid consciousness of normative-­ practical necessity. Their argument is based on their belief in the “omnipotence of theoretical reason” which leads them to impose the demand that everything, or at least (to prevent the slip from methodological into metaphysical naturalism) everything that we are consciously aware of, must be susceptible to “physiological explanations.” This demand is the guiding principle behind their attempt to naturalize and thereby debunk our moral representations by tracing them to contingent empirical conditions such as sensible desires or feelings (MS, 6:378). But they cannot reconcile this guiding principle with their self-­conception as theoretical cognizers: their naturalistic enquiries are governed by a priori necessary theoretical (e.g., causal) representations whose content and origin they cannot naturalize without debunking these representations as rationally groundless figments of the brain. Thus, the theoretical self-­ awareness of naturalistic cognizers eludes the “omnipotence of theoretical reason” because it involves epistemically fundamental, a priori component representations that one cannot subject to “physiological explanations.” This is sufficient to show that the attempt to naturalize and debunk our pure moral self-­awareness as an empirically conditioned phenomenon is incoherent: since their own theoretical self-­awareness forces naturalists to admit

318  Kant on Freedom and Rational Agency the actuality of a pure form of consciousness and thereby forces them to deny that everything we are consciously aware of can be naturalistically explained, the naturalistic attack on the possibility of a pure moral self-­awareness loses its sting. Once a naturalistically oriented theoretical reason is compelled by itself to abandon its claim to omnipotence, it can no longer rely on this claim in its attempt to compromise the integrity of pure practical reason. The argumentative force of this point is not diminished by the fact that our pure moral self-­consciousness has metaphysical-­causal implications that our pure theoretical self-­consciousness lacks. Since any pure form of consciousness relates to our noumenal rather than our empirical-­phenomenal self (cf. KpV, 5:6), our pure theoretical self-­awareness has some (indeterminate) metaphysical implications as well. While the unconditioned spontaneity of thought that lies at the basis of our pure theoretical consciousness is not a “true causality,” it is a noumenal, thus naturalistically inexplicable and theoretically incomprehensible source of representations and mental syntheses. Naturalists cannot cherry-­pick and endorse a non-­naturalistic metaphysical component in our pure theoretical self-­awareness while insisting that when it comes to our moral self-­awareness, naturalistic constraints compel us to explain away (i.e., debunk) any seemingly non-­sensible components as empirically conditioned illusions. Once the basis for the naturalistic attack on the validity of our pure moral self-­consciousness has been compromised, i.e., without the guiding principle that every aspect of our self-­consciousness must yield to naturalizing physiological explanations, naturalists can no longer doubt our pure moral self-­consciousness just because it cannot be naturalistically explained or because it entails, via a valid inference, the ex­ist­ence of a naturalistically inexplicable causality. They can, of course, try to show that this entailment is false because the pure contents of our moral self-­awareness do not require (as their ratio essendi) the existence of a supersensible causality. Showing this would require an argument against premise (III) of Kant’s moral proof which asserts this requirement (cf. Section VIII.3). Or, naturalists can try to argue that our moral self-­consciousness is flawed in some other way that goes beyond its purity and non-­susceptibility to naturalistic explanations. But these objections differ from the challenge that Kant’s inference from moral ought to noumenal freedom fails simply because of its inscrutable metaphysical-­noumenal implications. A second objection to my reconstruction of Kant’s view goes as follows. I have argued that we must distinguish between two different components: first, Kant’s moral proof of free will; second, his theoretical defense of this proof via the appeal to freedom of thought. One might object that these two components cannot be separated: if Kant’s moral proof of free will requires a theoretical defense against a relevant naturalistic challenge, then this defense must be incorporated into the proof strategy as an additional premise—­in which case we are no longer dealing with a purely moral proof. Thus, according to this objection, the argumentative schema I supplied in Section VIII.1 must be amended by adding a further premise, III*:

Kant ’ s Theoretical Defense of Moral Freedom  319





(I) If we know that the existence of x is a necessary real condition for y, and if we know y, then we also know the existence of x. (II) We know the moral law, i.e., we have an objectively valid pure moral self-­consciousness. (III*) Our knowledge of the moral law (the objective validity of our moral self-­consciousness) is, owing to the analogy between our pure moral and our pure theoretical self-­consciousness, immune to the naturalistic challenge that our seemingly pure moral (self-)representations are subject to physiological explanations and are therefore empirically conditioned illusions. (IV*) We know that our noumenal freedom of will is a necessary real condition for the moral law, i.e., for the objective validity of our pure moral self-­consciousness. (V*) Hence, we know the existence of our noumenal freedom of will.

The demand that Kant’s theoretical defense of the belief in free will must (as III*) be part of the justificatory basis for the knowledge claim that we have free will implies a more general epistemological constraint: one cannot be ­sufficiently justified in adopting a belief unless one’s justificatory basis includes an awareness of responses to relevant skeptical challenges. But this constraint seems implausibly demanding. To see why, consider our ordinary belief in bodies. Steph believes that she is sitting on a chair. Her justificatory basis for this belief derives from her (objectively conceptualized) sensory states of sight, touch, etc. According to the epistemic constraint under consideration, Steph cannot claim to know that she is sitting on a chair since she is not aware of and has no adequate response to external world skepticism; she would have to grasp such a response in order to have objectively sufficient grounds for her belief that she is sitting on an actual chair. So, those who have never engaged in abstract philosophical reflection lack knowledge that they sit on chairs, have hands, and eat bread! This implausibly demanding view can be contrasted with two alternatives, a very permissive account and a position that occupies a middle ground between these two extremes. On the permissive view one can know that p on the basis of justifying reasons that typically derive from our ordinary (e.g., perceptual or practical) consciousness. One’s justification for believing that p does not require an awareness of skeptical challenges or philosophical responses to such challenges. Moreover, one’s knowledge that p is left intact even if there are skeptical challenges that receive no satisfying answer from philosophical reflection. Of course, the permissive view accepts that one’s justification for believing that p can be undermined by specific epistemic defeaters that arise in specific cases. But it denies that a completely general, a priori skeptical challenge can serve as an ep­i­ ste­mic defeater. To illustrate: my sensory representation of a chair loses its

320  Kant on Freedom and Rational Agency justificatory force if my tea has been spiked with hallucinogenic drugs, but the justificatory force of this representation is left untouched by a general skeptical (e.g., evil genius-­style) argument contending that all our sensory representations are always doubtful, even if this argument receives no satisfying philosophical response.23 The middle ground view agrees with the permissive view that ordinary knowledge claims are typically justified sufficiently by the subject’s awareness of non-­philosophical considerations that objectively support those claims. This justificatory basis can sufficiently support ordinary beliefs even when the subjects who grasp it are oblivious to relevant skeptical doubts or of adequate philosophical responses to such doubts. But on the middle ground view our justificatory basis supporting our ordinary knowledge claims would be undermined if there was no adequate response to relevant skeptical challenges. One is only justified in believing that p if there is, in fact, something wrong with skeptical philosophical arguments that call one’s justifying reasons into doubt. Moreover, on this view the error that afflicts these skeptical challenges must be in principle accessible to our rational insight, i.e., must lie within the purview of critical human reason: it must be such that one would discover it if one were to engage in the right kind of reflection. This is compatible with conceding that most ordinary thinking subjects cannot, due to contingent factors, fully grasp the considerations that defuse relevant skeptical challenges. Such contingent factors might include a lack of patience or time, a lack of conceptual sophistication, a lack of talent for abstract reasoning, or some combination of these and further factors that render many thinkers ill-­disposed for philosophical reflection. On the middle ground view there is something like an epistemic division of labor between common and philosophical cognition. The positive, objectively sufficient justificatory basis for our ordinary knowledge claims does not derive from philosophical reason; it stems from cognitive sources other than abstract philosophical reflection. But such reflection (e.g., the mode of reasoning Kant calls critique) is required to defend ordinary knowledge claims against key philosophical challenges or problems. In my view, Kant adopts this model of epistemic justification for both our theoretical and our practical knowledge claims. To confirm this, consider two central cases from the theoretical domain. First, Kant takes it as a given fact that we possess synthetic a priori mathematical ­cognition and knowledge (B128). Explaining how such knowledge is possible—­ ­in particular, solving the difficult puzzle of how we could know non-­trivial necessity claims about the spatial properties of objects—­requires an appeal to

23  This is, roughly, the view of Pryor 2000. There are more permissive views in the current epistemological landscape, namely those of certain externalists such as Williamson 2000 who seem to allow that one can be justified in believing that p even without any awareness of the justificatory basis itself.

Kant ’ s Theoretical Defense of Moral Freedom  321 transcendental idealism.24 But mathematicians can obtain mathematical cognition and knowledge without grasping this rather involved answer to the “how possible?” question. What constitutes the (objectively sufficient) justificatory basis for their claims to mathematical knowledge is their intuitive awareness of mathematical axioms and their ability to understand the constructive mathematical proofs which are governed by these axioms. Second, in Kant’s view my assent to the proposition that I have two hands qualifies as knowledge since it is based on my immediate outer experience of my body which provides an objectively sufficient, certain epistemic ground for assent.25 This common empirical knowledge would be undermined, namely transformed into mere faith, if there was no adequate response to Cartesian external world skepticism: “it . . . remains a scandal of philosophy and universal human reason that the existence of things outside us . . . should have to be assumed merely on faith, and that if it occurs to anyone to doubt it, we should be unable to answer him with a satisfactory proof ” (Bxxxix). But for Kant, common-­universal human reason is typically unaware of these doubts and is also unable to quell them through what he regards as a satisfactory proof, namely, the abstract argument that he offers in the Refutation of Idealism. The account I have reconstructed in Sections IX.2–3 can also be seen as an instance of the middle ground view. Common agents have a pre-­philosophical knowledge that they possess noumenal free will which is based on their immediate knowledge of the moral law and their (at least implicit) recognition that free will is a real condition of morality. Since they typically do not accept the omnipotence of theoretical reason and its physiological explanations, they are typically oblivious to the challenge that our common moral self-­awareness must be susceptible to the physiological explanations of theoretical reason and must therefore be viewed as an empirically conditioned illusion (A473–4/B501–2). This naturalistic challenge would undermine common knowledge of moral duties and (thereby) of free will if it was left unaddressed. Philosophical reflection defuses this challenge by arguing that since naturalistic cognizers must presuppose the actuality of their empirically unconditioned, hence naturalistically inexplicable theoretical self-­ awareness, they cannot coherently insist that an empirically unconditioned moral self-­awareness is impossible. But common human reason is typically unable to appreciate the subtle reasoning that underlies this philosophical defense of the ordinary belief in free will.

24  For my account of why Kant holds this view, see Kohl 2021c. 25  Kant does not explain the epistemic standings of such assents in any detail, but he does appeal to various kinds of “empirical certainty” (e.g., at JL, 9:70–7). One species of empirically certain cognition concerns what I can immediately ascertain through my own experience, e.g., my experience of my physical self and (some of) its states or properties. Empirically certain knowledge differs from a priori (apodictically) certain knowledge in that it is not accompanied by a consciousness of necessity. I discuss this distinction in Kohl (forthcoming3).

322  Kant on Freedom and Rational Agency We can confirm that Kant adopts something like this view by considering a passage from the first Critique’s B-­edition Preface, where he contrasts two ways of recognizing freedom of will: first, via the philosophical-­speculative “proof of freedom of the will against universal mechanism drawn from the subtle though powerless distinctions between subjective and objective practical necessity” (Bxxxii); second, via “the mere clear exposition of our duties in opposition to all claims of the inclinations leading to the consciousness of freedom” (Bxxxii–­Bxxxiii). The second route leads to the conviction in the “public” whereas the first route is too “subtle,” hence “powerless” to influence the (practically efficacious) convictions of public, common reason. The second route is precisely the non-­speculative, moral proof of free will that infers our freedom to choose independently of sensible-­natural determination from our formal moral duty to choose independently of sensible-­natural inclinations. This is the proof that Kant expounds in the second Critique and that I have reconstructed in Chapter VIII. It provides sufficient “rational grounds” (Bxxxiii) for adopting the belief in free will which are readily accessible to common human reason; on my reading, these practical grounds yield certain knowledge of free will. Kant’s characterization of the first, speculative route invokes the kind of language that is central to his theoretical defense of free will. His distinction between “subjective and objective practical necessity” refers to the philosophical argument that if our practical faculties were exclusively subject to the mechanism of nature we would be unable to cognize and choose on the basis of objective practical reasons that yield rational necessities. Instead, our practical judgments and choices would be determined by contingent, egocentric states such as our naturally given desires, whose presence and strength is beyond our intellectual self-­control and whose empirically conditioned desiderative force masks itself under the false guise of a categorical-­objective normative necessity (cf. Chapter IV). This argument carries no weight against speculative philosophers who are prepared to deny (qua spectators) that we truly can cognize and choose on the basis of objectively necessary practical norms. Such philosophers are concerned to uphold the omnipotence of theoretical reason’s physiological explanations. This concern drives their claim that all our practical judgments and choices are empirical effects of the “mechanism of subjectively determining causes” (RezS, 8:14). Against this, philosophical reason (purified by Kant’s critique) runs a parallel argument which employs the “subtle” distinction between “subjective and objective necessity” once more but now in relation to theoretical reason. This argument yields the theoretical defense of free will I have reconstructed in this chapter. Naturalistic cognizers are (implicitly) committed to a non-­naturalistic conception of their higher mental capacities because they must presuppose the rational validity of their causal, explanatory judgments. If these causal judgments were empirical effects of the “mechanism of subjectively determining causes,” they would not be grounded in our objective intellectual determination by reason: they would

Kant ’ s Theoretical Defense of Moral Freedom  323 express merely “a subjective necessity arisen from frequent association in experience, which is subsequently falsely held to be objective” (B127). This problem undermines the naturalistic principle that everything we are consciously aware of must be susceptible to naturalizing causal-­physiological explanations. By losing their entitlement to this principle, naturalists also lose their basis for insisting that our ordinary, seemingly a priori awareness of objective practical necessities (which, if valid, entails the existence of free will) can be debunked as an empirically conditioned illusion. The common moral self-­consciousness which provides the objective rational grounds for the belief in free will does not include an awareness of the naturalistic challenge. Nor does it grasp the “subtle” philosophical argument from freedom of thought that defends the integrity of common moral self-­awareness against this challenge: “for just as little as the people want to fill their heads with fine-­spun arguments for useful truths, so just as little do the equally subtle objections against these truths ever enter their minds” (Bxxxiv). Thus, naturalistic skepticism about morality has little bearing on common people outside “the school” (A474/B502). But this does not render the theoretical defense of free will superfluous: through this defense philosophical reason defuses a potentially unsettling skeptical argument. If this argument were left unaddressed, it would undermine the justificatory force of common moral cognition, since it does contain many “disadvantages against supreme practical principles” (A474/B502). The naturalistic view might even “become generally injurious” (Bxxxiv) if the idolatry of theoretical reason’s physiological explanations seeped from “the school” into public consciousness (or at least into its educated segments, as in our time).

IX.5.  Does Naturalistic Cognition Presuppose Morality and Free Will? In this section I show how my reconstruction of Kant’s defense strategy differs from a related approach. For Onora O’Neill, the moral law is the highest principle not just (as on my reading) of practical reason but also of theoretical reason. If that were the case, then one could reject the naturalistic attack on free will in a very straightforward manner: no complicated theoretical defense of free will would be needed. The shorter defense would run simply as follows: naturalistic cognizers are committed to accepting the moral law as the highest principle of all including their own naturalistically oriented theoretical reason; moral law and free will reciprocally imply one another; hence, naturalistic cognizers must endorse their freedom of will in their cognitive efforts to understand and explain the world. Call the claim that the moral law is the highest principle of all including theoretical reason the “HP-­thesis.” One reason to doubt this thesis is that Kant never

324  Kant on Freedom and Rational Agency clearly states or commits to it. One would expect him to call explicit attention to the HP-­thesis if he did endorse it, since it would be of utmost importance for his philosophical system. For instance, the HP-­thesis would afford him a straightforward solution to the problem he designates as the unity of reason. Yet, he does not appeal to the HP-­thesis in crucial passages that raise the unity issue (cf. GMS, 4:397; EEKU, 20:201–5). Moreover, whatever solution to the unity problem Kant eventually lays out in the third Critique, he does not invoke the HP-­thesis as part of that solution either.26 O’Neill suggests three arguments for the HP-­thesis. One argument concerns the unity issue just mentioned: since the practical use of reason has primacy over the theoretical use, and since the categorical imperative (CI) is the supreme principle of practical reason, the CI must be the supreme principle of all reason as such.27 But this is a non-­sequitur. The primacy of practical reason might rest on the claim that practical reason sets us goals that override the aims of theoretical reason. Or (perhaps, and), it might rest on the claim that in those rare, special cases where theoretical reason must remain agnostic about whether some theoretical proposition is true and where pure practical reason demands assent, theoretical reason cannot refuse assent. Neither of these claims hinges on or implies the further contention that the highest principle of practical reason also constrains the theoretical use of reason in all other, standard cases where theoretical reason on its own supplies grounds for affirming or denying some theoretical proposition. O’Neill’s second argument for the HP-­thesis invokes the agential components of theoretical cognition. Commenting on Kant’s claim that the authority of reason in adjudicating disputes yields a “tribunal,” she stresses that this tribunal is a collective, ongoing task in a community of rational agents. Moreover, it must be a “practical task” since it involves “the genuinely practical task of judging.”28 Rational judging as an essentially public act of shared communication among people with different interests poses a “practical problem”29 whose solution requires the CI as “a fundamental strategy . . . not just of morality but of all activity that counts as reasoned.”30 This practical conception of judgment per se grounds her straightforward rebuttal of the naturalistic challenge to free will: she argues that in Kant’s Second Analogy “agency is . . . taken to be the presupposition of causal judgment” and that Kant here “invokes practical freedom—­not the mere spontaneity of understanding, which is revealed within theoretical knowledge — to account for the possibility of causal explanation” so that practical “freedom is presupposed by theoretical understanding.”31

26  Klemme 2014 also raises this problem for O’Neill’s view. 27  O’Neill 1989: 3, 29. 28  O’Neill 1989: 18. 29  O’Neill 1989: 43. 30  O’Neill 1989: 59. 31  O’Neill 1989: 62–3.

Kant ’ s Theoretical Defense of Moral Freedom  325 I accept that our causal judgments and explanations are due to our free, spontaneous cognitive agency. But I deny O’Neill’s claim that “the mere spontaneity of understanding” is insufficient for such agency. I have argued (in Part 3) that for Kant our theoretical intellect exhibits a sui generis form of intellectual self-­activity in its theoretical syntheses and judgments. If this view of epistemic spontaneity is correct, one need not invoke practical freedom (and thereby the principle governing such freedom, the moral law or CI) to account for the free cognitive agency that yields theoretical judgments.32 Likewise, O’Neill is right to stress that theoretical enquiry is an ongoing “task” that cognizers collectively pursue under the guidance of rational norms and goals. But this does not require practical reason and its moral norms, since our theoretical intellect (in a broad sense, as comprising the understanding, theoretical reason, and the power of judgment) is itself a source of rational norms and goals (cf. Chapter VI).33 O’Neill’s claim that our acts of theoretical judgment rest on practical freedom of will governed by moral norms suggests that these judgments derive from our moral or immoral choices: “It is in choosing how to act, including how to think, to understand and to interpret, that we embody or flout the only principles that we could have reason to think of as principles of reason.”34 This raises two problems. First, recall that the moral law is essentially a law governing the “true” causality of the will as a power to produce external objects according to its purposive representations. Given this characterization of what the moral law is and what specific type of activity it governs, the idea that it also governs how we form mere theoretical representations of already existent objects seems a bit like a category mistake.

32  O’Neill claims that (1) we can deploy categories like causality “in complete acts of judgment” only if we follow regulative ideas and maxims of reason; from this she infers (2) that “a practical principle must guide all complete acts of judgment” so that “practical reason is fundamental to all reasoning” (1989: 19). Claim (1) is problematic because it is unclear what O’Neill means by “complete acts of judgment.” She might mean the use of judgment in an interconnected, rationally unified system of judgments. But the application of categories in singular acts of (e.g.) causal judgment does not (obviously) depend upon systematically oriented reasoning. Even if it did, this would not establish (2) since the regulative norms governing such reasoning can be drawn from theoretical reason or the reflective power of judgment. This is precisely what Kant suggests in those sections of the Dialectic (e.g., the Appendix) where he clarifies the positive function of theoretical ideas and also in the introductions to the third Critique—­texts which O’Neill surprisingly tends to ignore. 33  One should not equate (as O’Neill sometimes seems to; see, e.g., 1989: 44–5) moral requirements on communication, such as prohibitions against polemic or lying, with the appeal to publicity and communicability as a formal normative constraint on theoretical judgment. It is one thing to impose the epistemic norm that judging subjects must be disposed to demand assent from others and to reconsider their own assent in light of public evidence and criticism (cf. Chapter  VI). It is quite another thing to hold that such norms of theoretical judgment involve a moral prohibition against lying. A liar’s belief that their lie allows them to obtain some benefit relies on objective causal judgments that do satisfy relevant epistemic norms of communicability. (Liars are disposed to revoke these judgments in light of public evidence suggesting a mistake in their causal reasoning.) The lying claim itself does not satisfy these norms, but this yields no basis for sui generis epistemic (as opposed to moral) criticism since liars do not even intend this claim as an objective theoretical assent that makes a valid claim on the agreement of others. 34  O’Neill 1989: 27.

326  Kant on Freedom and Rational Agency Second, the idea that “in choosing how to think” we “embody or flout” the moral law implies that our theoretical assents, like our practical maxims, are based on our choices for or against the moral law. As such, they would be under our direct voluntary control and would involve our struggle against the corrupting influence of sensible inclinations. But, as we saw (in Chapter VI), Kant rejects a voluntaristic conception of theoretical assent: the “will does not have any influence immediately on assent” since it “cannot struggle against convincing proofs of truths that are contrary to its wishes and inclinations” (JL, 9:73–4). One might respond that the formation of theoretical assent is a special case where our volitional execution of moral principles encounters no struggle against sensible inclinations. But this is merely ad hoc. As we saw (in Chapter  V), for Kant such struggle is inherent to the act of finitely rational choice under moral oughts; this entails (among other things) the possibility of akrasia. Since the sensibly affected will cannot struggle against its awareness of valid reasons for theoretical assent and since theoretical assent is (thus) not afflicted with the threat of volitional akrasia, acts of theoretical assent cannot be construed as acts of choice under moral oughts.35 O’Neill’s third route towards the HP-­thesis proceeds as follows. She stresses that for Kant all our rational activity must be governed by standards deriving from the self-­reflexive critique of reason: namely, by norms through which reason determines the limits for its proper, permissible exercise and thereby disciplines itself against overstepping its boundaries. O’Neill further emphasizes that reason can exercise the legitimate authority that it needs for critical self-­discipline only insofar it has autonomy. This is because reason can determine, in a non-­arbitrary or universally valid fashion, the limits of its proper use only by using principles that are not imposed upon it by some foreign source. Only autonomous principles that derive from our shared “universal reason” are such that all rational thinkers can, indeed must accept them regardless of their contingent, varying circumstances. Conversely, rules that are externally imposed upon rational thinkers depend for their (limited) validity on contingent, subjective conditions which vary across different subjects. Hence, such rules cannot govern the universally valid self-­discipline of reason. “Reason’s discipline cannot be alien; it must be autonomous.”36 O’Neill here makes a deep, important point in showing the connections between the collective project of rational self-­critique or discipline, the claim to universal validity, and the autonomy of reason. However, this does not establish 35  Kitcher 2011 tries to have it both ways. She follows O’Neill in endorsing the HP-­thesis (2011: 248), but she also denies that theoretical cognizers are morally responsible for their acts of theoretical judgment. The latter point seems correct (cf. Chapter VI), but I do not see how thinking subjects can fail to be morally responsible for their cognitive acts if (as the HP-­thesis asserts) these acts are governed by the moral law. 36  O’Neill 1989: 57.

Kant ’ s Theoretical Defense of Moral Freedom  327 her further claim that the autonomous discipline of our rational judging activity must be governed by the moral law of practical reason (“for us practical reasoning is fundamental”).37 This does not follow because the normative authority and autonomy of reason is not simply identical to the normative authority and autonomy of practical reason. Kant attributes autonomy to the faculty of understanding which prescribes general laws to nature, and he also attributes (he)autonomy to the faculty of reflective judgment (cf. Chapter VI). Moreover, if we focus on the self-­discipline through which reason imposes limits upon its metaphysical speculation, we must chiefly consider the autonomy that theoretical reason (in the narrow sense) exhibits by legislating its own sui generis regulative-­ normative standards (cf. A699/B697).38 For these reasons, the HP-­thesis cannot be sustained. This undermines the short argument against naturalistic attempts to debunk the belief in free will which claims that naturalistic reasoning presupposes the moral law and thereby practical freedom of will. The more complicated argument I have suggested (which draws on crucial analogies between moral freedom of will and epistemic freedom of thought) seems like a better contribution to Kant’s defense strategy.

IX.6.  Kant’s Defense Strategy in Action In this section I illustrate the force of Kant’s defense strategy by showing how it works in an actual case of a naturalistic, Nietzschean attack on the moral belief in transcendental freedom of will. I say “Nietzschean” because I am not convinced that Nietzsche endorses all the commitments which yield the naturalistic view I am about to describe. But we can set aside this complication here. The Nietzschean view has been stated by Brian Leiter. He reads Nietzsche as a naturalist in the vein of Hume and Freud who seeks to make his views continuous with the methods and results of the natural sciences. Nietzschean naturalism aims at (what Kant calls) “physiological” causal explanations which show that human beings and their faculties are exclusively products of nature.39 On this view, the human personality including all its judgments and choices is causally determined by an “interplay of certain drives . . . over which the conscious self exercises no control.”40 Since our human personality, conscious evaluative 37  O’Neill 1989: 29. 38  As I pointed out in n. 32, O’Neill tends to ignore texts such as the Transcendental Dialectic or the introductions to the third Critique. Her arguments focus strongly on the Discipline of Pure Reason. But as Klemme (who also criticizes her move from autonomy to moral autonomy) points out, the concepts of moral duty, moral law, or categorical imperative do not come up in that section (2014: 122). Moreover, in the A-­edition Discipline (which remains unchanged in the B-­edition) Kant cannot be invoking the concept of moral autonomy because he still espouses a heteronomous conception of moral agency (cf. appendix to Chapter I, Section VIII.4). 39  Leiter 2015: 2–6. 40  Leiter 2015: 80.

328  Kant on Freedom and Rational Agency outlook, and behavior derives from the workings of “certain definite physiological and psychic traits” beyond our conscious rational self-­control, we have no capacity for autonomous agency or transcendental freedom in Kant’s sense.41 The allegedly pure moral self-­consciousness that is supposed to yield the rational basis for the belief in a non-­natural type of free will is merely another natural phenomenon, the empirically conditioned upshot of sub-­personal, non-­rational psychic forces. This naturalistic position relies on a variety of causal-­explanatory claims. But what justifies these claims? That is: what grounds our rational entitlement to the belief that there are “certain definite” natural states operating below the threshold of consciousness which causally suffice to determine our conscious states and our external behavior? On what basis can we rationally endorse the universal belief that “everything about our will” and our actions “is causally determined,” that one always “becomes what one is” through the force of causal necessity?42 It would seem that these beliefs, qua conscious mental states, are also the upshot of unconscious psycho-­physical forces which are insensitive to objective epistemic reasons: “agents hold any belief . . . because it favors or supports the type of life those agents are capable of living”;43 how our elementary non-­rational drives “play out determines what” one “believes.”44 But in this case, naturalists must concede that they have no good theoretical (non-­pragmatic) reasons for holding their causal and explanatory beliefs to be true. Nietzsche has often been read as a skeptic about truth and objectivity; there are many passages where he disparages, specifically, the objectivity and explanatory potency of our causal judgments: “one should use ‘cause’ and ‘effect’ only as . . . conventional fictions for the purpose of designation and communication—­ not for explanation” (Beyond Good and Evil, Par. 21; compare The Gay Science, Par. 112). Leiter tries to accommodate these claims by arguing that they undermine only a “pure” concept of causality rather than the empiricist notion of causality and causal explanation that Nietzsche uses in his naturalistic attack on morality and free will.45 However, Kant’s pure concept of causality contains precisely those features that Leiter’s Nietzsche also presupposes in his naturalistic account: namely, the representation that a cause sufficiently determines its effect through natural necessity, with the implication that since our conscious states are determined by unconscious natural drives we cannot avoid those conscious states. If Nietzscheans hold that beliefs in sufficient causal determination or necessity and corresponding explanatory judgments are mere upshots of non-­ rational desiderative “drives,” then they end up in the same untenable position as Humeans (cf. Chapter VII): they propose a natural science of the mind according 41  Leiter 2015: 69–72. 42  Leiter 2015: 68–72. 43  Leiter 2015: 62 44  Leiter 2015: 80. 45  Leiter 2015: 18; for his reading of Nietzsche as an empiricist, see 2015: 11–12.

Kant ’ s Theoretical Defense of Moral Freedom  329 to which our causal beliefs have no (epistemically) rational basis, but this picture of the human (thus Humean, Nietzschean) mind undermines the rational credentials of the causal explanations that are essential to any natural science which is established through our cognitive agency. In response, one can try to play up the alleged empiricist commitments of the Nietzschean view and hold that our causal beliefs are not, after all, the upshot of non-­rational drives but are rather directly responsive to the empirical information that we receive through the senses. However, since the senses do not convey to us features such as sufficient causal determination or necessity, the remaining conception of causality would be too thin to underwrite the Nietzschean attack on free volition. An austerely empiricist notion of causality suggests a mere regularity view, which Leiter indeed seems to be flirting with.46 But such a view leaves no room for genuine causal necessitation or determination. Moreover (cf. Section VII.3), a mere regularity account is fraught with so many problems that it cannot inform a viable program of naturalistic causal explanation. In roughly this manner, Kant’s theoretical defense strategy shields our moral self-­awareness and the corresponding belief in free will against Nietzschean naturalism. Once again, for Kant it is solely our moral self-­awareness which positively justifies the belief in free will. So there is certainly room for debate between Kantians and Nietzscheans about whether the moral grounds for the belief in free will are compelling. But this debate must be conducted in terms of the normative-­practical considerations that were the focus of Chapters  IV,  V, and VIII. Nietzscheans cannot (and, I would argue, Nietzsche does not)47 rely on the blanket claim that our moral self-­consciousness (alongside its implication regarding free will) is illusory simply because it does not accord with a naturalistic causal framework or an empiricist epistemology.

IX.7.  The Problem of Divine Foreknowledge In this section I consider how Kant might defend his moral grounding of free will against a further theoretical challenge. I have discussed a challenge arising from the view that we must trace everything (that we are consciously aware of) to natural causes and conditions. But one might also challenge Kant’s view from a super-­natural theistic perspective. Transcendental freedom of will is our capacity to choose independently from determination by all foreign causes, including the super-­natural agency of a divine being. As we saw (Chapter VIII), Kant tries to  legitimize a rational faith in the existence of a divine being endowed with traditional theistic attributes such as omniscience, which implies divine ­

46  Leiter 2015: 8–9, 257.

47  I further discuss this issue in Kohl (fortcoming4).

330  Kant on Freedom and Rational Agency foreknowledge of human actions. Hence, one might argue, Kant must confront the traditional problem of how divine omniscience could allow for absolute freedom of human choice. If God’s creative act supplies Him with certain cognition that I shall choose to be evil, how can I nevertheless be free to choose between good and evil? In suggesting how Kant can address this challenge, I will diverge a bit from the methodology I have adopted throughout this book. I have typically supported my reconstruction of Kant’s arguments with fairly direct textual evidence. By contrast, in presenting what I take to be his best defense against the theistic challenge I will largely bracket his most explicit attempt to defuse this challenge. I will suggest a line of reasoning that is not fully explicit in his writings, though it does utilize central aspects of his explicitly stated philosophical theology. Kant’s most explicit discussion of the theistic challenge in his published writings occurs in the second Critique (KpV, 5:100–3). It is not altogether clear whether he is concerned here with the problem that divine foreknowledge raises for human free will since he frames the issue in metaphysical-­causal terms: if God is the self-­sufficient creator of everything and is thus also the cause of my ex­ist­ence as a substance, then all my actions must have their determining ground in God. This does not strike me as a cogent challenge: one can respond that God might create me as a substance endowed with a power of choice which counts as free precisely because God abstains from determining how I shall use this power. But once we invoke the property of divine om­nis­ cience, it seems that God’s creation of me entails His complete cognition of how I shall use the powers that He endows me with. From there it is only a short step to the further claim that God cognizes how I shall act by determining my actions, so that God causes both my existence and my agency. Student lecture transcripts suggest that Kant may want to block this step by holding that God’s foresight of our actions does not determine that these actions must happen (­V-­Th/Pöl, 28:1054–5). It is unclear, though, how there could be divine cognition of x without divine determination of x: in a divine, non-­receptive mode of cognition, the theoretical cognition of what is fully coincides with the practical creation of what shall be (cf. Chapter III). One might (with Crusius) insist that divine foreknowledge does not imply that our actions are determined because God cognizes our actions immediately rather than mediately via an inference from a determining ground to its consequence. For Kant divine cognition is indeed immediate (cf. Chapter III). But he cannot use this as a reason for denying that our divinely foreseeable actions are predetermined. God can immediately cognize our actions as immediately determined in a manner that is inscrutable to our intellect precisely because we cannot capture it through our discursive, inferential-­mediate ground-­consequent logic. One might add that God could not have foreknowledge of our actions unless these actions were determined in some way: if they were entirely undetermined, divine

Kant ’ s Theoretical Defense of Moral Freedom  331 cognition of our actions would be indeterminate—­but this is absurd because cognitive indeterminacy implies cognitive finitude.48 In the second Critique Kant seems to be arguing that his transcendental idealism via its distinction between our temporal existence as sensible phenomena and our atemporal existence as noumena is sufficient to resolve the problem, because the act of divine creation concerns only our atemporal noumenal existence (KpV, 5:102). I must confess that despite many efforts, I fail to understand Kant’s argument here.49 To be sure, idealism complicates the issue: God’s atemporal cognition of the noumenal beings He creates cannot literally involve a divine foreknowledge of their future actions. But when we deliberate about how to exercise our freedom of choice, we are focused on our future phenomenal actions that we regard as (absolutely) contingent. If those temporal actions are grounded in our noumenal freedom, and if God cognizes how we exercise our noumenal freedom, then the temporal framing of the problem creeps back into the picture and thus cannot be explained away in an idealist fashion. Moreover, the problem remains in a purely non-­temporal framework: if God creates me as a noumenal being and in doing so immediately cognizes with absolute certainty my atemporal choice of an evil maxim, then how am I free to make an absolutely contingent, metaphysically open noumenal choice between good and evil?50 Although I do not quite understand Kant’s official solution, I shall propose a way of solving or better dissolving the problem of divine foreknowledge which Kant may be hinting at after having presented his official solution (namely, at KpV, 5:104–6). Let us first compare the naturalistic assumption (A) that every object of our conscious awareness is empirically conditioned with the theistic assumption (B) that everything (in the noumenal world) is the product of divine creation (and thus of God’s omniscient, creative-­intuitive cognition). (B) is at a striking disadvantage here which limits its potential to raise, like (A), a serious challenge to Kant’s moral proof of free will: unlike (A), (B) gains no support from theoretical reason. To be sure, (A) enjoys no conclusive support from theoretical reason either: (A)’s claim that every object of our conscious awareness is empirically conditioned goes beyond the weaker, more restrictive Second Analogy proof that everything we can theoretically comprehend as part of our sensible experience (i.e., as a determinate spatiotemporal phenomenon) is empirically conditioned. But the Second Analogy principle can nevertheless seem to yield support for (A) if one accepts the “omnipotence of theoretical reason.” Defenders of such omnipotence can rightly 48  See Allison 2020: 32–3. 49  Allison regards Kant’s solution as “curious” (2020: 404). 50  One might reply that freedom is compatible with divine foreknowledge since it does not involve alternative possibilities or contingency (Abaci 2019: 2070). But (cf. Section III.3) this reply conflates divine freedom (which indeed involves no contingency) with our imperfectly rational executive freedom of choice (most dramatically, our absolutely contingent choice between a life-­governing good or evil maxim).

332  Kant on Freedom and Rational Agency insist, at least, that taking our pure moral self-­consciousness at face value comes at a severe theoretical cost: our pure moral self-­consciousness implies the ex­ist­ ence of a causal power which is inscrutable to our essentially “physiological” explanations since it does not conform to the empirically comprehensible model of causation licensed by the Second Analogy. Hence, the re­sist­ance towards Kant’s moral proof of free will that arises from (A) has some (though, in the end, inconclusive) basis in theoretical reason. By contrast, (B) has no basis whatsoever in theoretical reason. This point has a twofold dimension. First, theoretical reason does not support a belief in a divine creator since all possible theoretical proofs of God fail (in Kant’s view). One might argue that in the Canon of Pure Reason Kant invokes a “doctrinal faith” in God that is based upon the needs of speculative reason. But Kant strongly qualifies the status of such speculative faith when he says that it is merely a “wavering” attitude (A825–9/B853–6). Moreover (secondly), Kant holds that on a purely speculative basis we cannot form a determinate conception of God that includes the predicates which have traditionally been ascribed to a divine creator: speculative reason cannot ground a rational theology (A636/B664) as opposed to a merely deistic view which posits a highest being without implying the notion of a personal God endowed with intelligence, reason, and will (A632–3/B660–1; cf. V-­Th/Pöl, 28:1020; Prol, 4:355). This deistic conception is sufficient to satisfy our speculative need for the regulative notion of a God that arises in our theoretical efforts to achieve a systematic unity of empirical cognitions (A674–5/B702–3). In the later Critiques, Kant repeats that a purely speculative-­transcendental metaphysics cannot yield a determinate, theistic, personal conception of God. An adequate idea of a personal God endowed with traditional theistic properties (like omniscience) can only arise from pure practical reason; likewise, only pure practical reason may ground a firm, unwavering faith in the existence of God (KpV, 5:131–3, 138–40; KU, 5:470, 473). Pure practical reason is moral reason: both the theistic conception of a personal God and the firm faith that such a God exists derive from our moral duty to pursue the highest good, through which the moral law “gives meaning . . . to the theological concept of the highest being” (KpV, 5:133). In particular, the firm belief in an omniscient and omnipotent God is justified only because of our ra­tional need to conceive God as an infallible judge of our complete moral character who guarantees the real possibility of the highest good through His infallibly just distribution of happiness according to moral virtue (A814–15/B842–3; KpV, 5:140; KU, 5:444). This view affords Kant the following simple but effective strategy for defusing the theistic challenge to our belief in free will. Our only legitimate basis for the firm rational belief in an omniscient God is our knowledge of the formal moral law and of the material duties, such as the duty to pursue the highest good, that can be derived from this law. Thus, unless the moral law has rational validity we

Kant ’ s Theoretical Defense of Moral Freedom  333 have no good reason to accept the idea of a personal God who has divine foreknowledge. But we cannot regard the moral law as rationally valid unless we regard ourselves as having transcendental freedom of will, since such “freedom . . .  is the condition” (ratio essendi) “of the moral law” (KpV, 5:4). Since the moral law  yields the only rational basis for legitimate faith in an omniscient God, and since our free will is a (real, necessary) condition of the moral law’s rational validity, it follows that one cannot legitimately believe in an omniscient God without accepting that one has free will. Therefore, the appeal to an omniscient God cannot yield a coherent challenge to the epistemically prior belief in our freedom of will. This strategy has no ambition to solve the puzzle of how human freedom is compatible with divine foreknowledge. Kant should be happy to concede that the co-­existence of human freedom and divine foreknowledge, just like human freedom itself, is subjectively incomprehensible to finite human reason.51 The strategy I have suggested does not explain how human freedom and divine omniscience go together; it does not even purport to show that there cannot be a real problem or conflict here. But it does show that we cannot coherently raise such a problem or entertain the possibility of a real conflict. That, to my mind, is quite sufficient to do the job—­at least the kind of job that Kant needs done. I mentioned earlier that Kant may be hinting at this strategy for defending the belief in free will against the theistic challenge. Let me suggest what I have in mind here. Kant explicitly posits free will as a condition for our rational belief in God: the theistic belief in God depends on the actuality of our free will because without the morally proven belief that we have free will the representation of God would remain a “mere” idea without firm rational grounding (KpV, 5:3–4). The argumentative strategy I have suggested yields a plausible account of why Kant makes this (prima facie surprising) claim: the idea of God must be grounded in morality, and free will is a ratio essendi of morality. Kant repeats a version of his claim that the belief in God depends on the appeal to our free will immediately after he has presented his official idealist solution to the problem of how divine creation may allow for human freedom (at KpV, 5:104–6). He says that the moral law grounds the assertoric cognition that we have free will. Although we cannot reach such an assertoric cognition with regards to God, the assertorically cognized notion of free will allows for a “mediation” (Vermittelung) between the sensible world and God as an intelligible being. Kant’s point here seems to be (roughly) as follows. Our rational cognition that we possess a free supersensible power to cause morally desirable effects in the sensible world enables us to take a securely (rationally) grounded first “step” into the field of the supersensible. This 51  Here I agree with Allison: the issue of how divine creation (and, I would add, omniscience) is compatible with human freedom is insoluble for theoretical reason, but this raises no fatal problem—­ rather, it “clear[s] the decks for pure practical reason” to which Kant “gives primacy” (2020: 406).

334  Kant on Freedom and Rational Agency is a safe step insofar as it does not require that we posit something external to our own rational-­moral selves (“außer uns”). Once we have made this step and thus have gained a secure foothold in the intelligible world due to our self-­conception as free moral agents, we can go a step further, still guided by the same moral law that enabled us to take the initial step: by reflecting on the implications of the moral norm to pursue the highest good that governs us qua free finite agents, we can legitimately posit a noumenal intelligence that is external to our free selves and whose theistic predicates (such as omniscience) secure the real possibility of the highest good.

IX.8.  Conclusion: Kant’s Theoretical Defense and the “Great Reversal” Reading I have argued that the ambition to defend our belief in free will against key theoretical challenges remains a central component of Kant’s view even once he develops his practical-­moral proof of free will. This interpretation is not incompatible with a great reversal reading of how Kant’s moral philosophy develops during the 1780s. For all I have said, there may well have been a serious shift in Kant’s views on how to justify the belief in moral freedom of will during the 1780s, though on my reading the shift would not be quite as dramatic as it is on readings where Kant completely abandons the theoretical defense strategy that I have suggested. My reading does not rule out the possibility of a serious shift or reversal because it leaves open the possibility that in Groundwork III Kant tries to prove that we have free will and/or that we are subject to moral laws on a purely theoretical basis (perhaps by using an argument akin to the direct theoretical proof that I considered in Section IX.1). If that were his approach in Groundwork III, then the second Critique would yield a considerable reversal because Kant here argues first that our moral self-­awareness sufficiently-­objectively justifies our belief in free will and (only) then uses theoretical arguments to shield this moral justification against key theoretical challenges. My overall interpretation allows for the possibility of such a reversal because in my reconstruction of Kant’s view I have refrained from taking a stand on the (in my view, intractably difficult) argument structure of Groundwork III.

PART 5

F R E E D OM IN KA NT ’S AEST HET IC S A N D T H E U NI T Y OF KA N T ’S D O CT R I NE Summary and Transition to Part Five The preceding chapter concludes the complicated dialectic that began in Chapter I. Let me briefly summarize the main aspects of my interpretation thus far before moving on to the final stage. For Kant the belief in free will is the metaphysical belief that we possess a supersensible causal power that is not determined by foreign causes. His transcendental idealism is a necessary but insufficient condition for justifying this belief: by restricting the scope of the deterministic causality of nature to sensible phenomena, Kant’s idealism makes ontological space for the existence of a supersensible free causality without showing that this space is (or really could be) filled. This raises the central question of how Kant seeks to justify the belief in free will. A highly distinctive feature of Kant’s view is that he does not try to construct a metaphysical theory for the metaphysical belief in free will: he does not give a theoretical account of how the causality of freedom operates, how it produces physical effects in the phenomenal world, or why free agents act and cause specific effects at particular places and times. Since the belief in freedom is theoretically incomprehensible to us, it is consigned to the normative standpoint from which we consider how we ought to act. This belief does not figure in the empirical standpoint which is focused on observing, explaining, and predicting the behavior of phenomenal objects (including ourselves). From the normative-­ practical standpoint, we can form a determinate representation of our supersensible (noumenal) causality because our pure self-­consciousness as autonomous moral agents provides a positive and objective, though only practical (thus theoretically fruitless) content or meaning for the idea of an atemporal causality. Because Kant’s doctrine seeks to vindicate the belief in a theoretically inscrutable type of causality, it must provide reasons for rejecting naturalistic, compatibilist conceptions of free will which avoid the cost of theoretical

336  Kant on Freedom and Rational Agency incomprehensibility. Kant gives two arguments to support his stance that if (as a naturalistic worldview contends) there were merely the causality of nature this would eliminate the only kind of free will worth having. First, the naturalistic claim that all our acts of practical judgment are empirically conditioned is incompatible with the unconditional objective necessity of our moral judgments. Second, the naturalistic claim that we are causally necessitated to make a particular choice on any given occasion is incompatible with the presumption that we are governed by rational, categorical moral oughts: a presumption which presupposes both the power to act from duty and the privative option to act contrary to duty. Freedom of will is not the only kind of freedom that plays a central role in Kant’s doctrine. In Kant’s considered view we possess a freedom of thought which yields a distinctive species of transcendental freedom even though it lacks the “true causality” to produce external objects as opposed to mere representations of existent objects. Our theoretical intellect is a source of autonomous cognitive laws and goals whose objective rational necessity is incompatible with the assumption that our theoretical representations are conditioned by foreign (natural or super-natural) causes. We must presuppose the objective rational necessity of our pure theoretical concepts (such as “causality”) and cognitive laws (such as the general causal principle) in our theoretical judgments, especially in judgments that seek to provide objective naturalistic explanations for sensible phenomena. Our empirical judgments about nature result from a sensibly affected but absolutely spontaneous (yet also non-voluntaristic) capacity for autonomous cognitive self-determination. Since our transcendental freedom of thought is a necessary condition for our actual objective cognition and knowledge of nature, we can know that our noumenal selves possess transcendental freedom of thought. This does not show, however, that our noumenal selves also possess the transcendental causality of a free will. Kant attempts to legitimize the belief in free will through a complex twopronged approach. In Kant’s view our ordinary (pre-philosophical) moral self-consciousness yields knowledge that we possess freedom of will: common agents possess certain knowledge of their moral duties and they (at least implicitly) recognize that they could not possess such normative-practical knowledge unless they had free will. This account of how our common moral self-consciousness justifies our belief in noumenal free will invites the objection that normative-­ practical premises cannot sufficiently justify a determinate yet theoretically inscrutable metaphysical belief in a supersensible causality that produces sensible effects. This objection is fueled by a naturalistic view which insists that we must subject every item of our conscious awareness, including our representation of the moral law, to the physiological explanations of theoretical reason. Rather than allowing our pure moral self-consciousness to justify our belief in noumenal free will, naturalists seek to debunk the alleged purity and rational necessity of our moral self-awareness by viewing our moral representations as empirically

KANT ’ S AESTHETICS AND THE UNITY OF KANT ’ S DOCTRINE  337 conditioned figments of the brain. Kant defends the integrity of our moral consciousness and thereby the epistemic basis of the belief in free will against this naturalistic challenge by invoking our freedom of thought. Naturalistic cognizers cannot presume that every aspect of our conscious awareness must be explicable via natural causes because the objective validity of their naturalistic explanations requires an empirically unconditioned, hence naturalistically inexplicable theoretical consciousness of a priori necessary cognitive laws. Since naturalistic cognizers must presuppose their freedom of thought, they cannot coherently debunk the belief in free will. This argument does not yield a theoretical proof of free will. Rather, it serves as a defense of the epistemically prior purely moral proof. This concludes my reconstruction of Kant’s doctrine as an account of our moral freedom of will and our epistemic freedom of thought. There is, however, one further important angle to Kant’s views on freedom that a systematic interpretation cannot ignore. I have focused on the two distinctive species of transcendental freedom which play a central role, respectively, in Kant’s moral philosophy and epistemology. In both cases, the relevant kind of free agency is a form of autonomous self-determination where the subject acts (chooses or thinks) under self-given rational laws. But Kant also brings up the idea of freedom in his aesthetics: he invokes the freedom of our imagination as enabling both our aesthetic production of beauty (in art) and our aesthetic experience of beauty (in art and nature). Can we fruitfully integrate Kant’s appeal to freedom of imagination with his views on freedom of will and freedom of thought? Or must we regard his appeal to aesthetic freedom of imagination as an aberration from his standard conception of freedom, a result which would disrupt the unity of his doctrine? The latter diagnosis may seem unavoidable in light of Kant’s insistence that freedom of imagination is a lawless freedom, namely, a freedom from the rational laws legislated by our higher intellectual faculties. In the concluding part and chapter of this book, I argue that despite this central difficulty we can find a place for aesthetic freedom of imagination in Kant’s systematic doctrine. If we take a certain (admittedly controversial) approach towards Kant’s aesthetics and if we allow ourselves some leeway in applying Kant’s strict definition of transcendental freedom as autonomy, then we can view freedom of imagination as a further (third) distinctive species of transcendental freedom whose lack of determination by intellectual laws uniquely captures the distinctive character of our aesthetic (as opposed to moral or epistemic) self-activity.

X Freedom of Imagination and the “Autonomy of Taste” Kant’s view that our aesthetic practices are based on our freedom of imagination suggests that moral choice, theoretical judgment, and aesthetic engagement have a common root in our free, spontaneous self-­activity. This would establish the idea of freedom as the anchor (“cardinal point”; Refl., 18:679) of all our truly meaningful endeavors. However, in light of what Kant deems essential to imaginative freedom it may seem that this is an altogether different concept than the canonical notion of freedom at issue in moral and epistemic contexts. If that were the case, Kant’s appeal to freedom of imagination would not systematically complete but rather fatally disrupt the unity of his doctrine of freedom. In this chapter I suggest how we might avoid this unhappy result and take aesthetic freedom of imagination as a distinctive species of transcendental freedom. If my proposal is acceptable, we can see how Kant’s appeal to freedom of imagination fruitfully enriches rather than disrupts his doctrine of freedom.

X.1.  The Problem with “Freedom of Imagination” I begin by considering Kant’s negative freedom-­condition: genuinely free agency must not be determined by foreign natural causes. As we saw, the free will and the free understanding both involve a “self-­activity” that is not determined by empirical laws. Likewise, if the imagination is considered in its “self-­activity,” “then it is in the first instance taken not as reproductive, as subjected to laws of association” (KU, 5:240). Since for Kant laws of association paradigmatically exemplify natural laws that determine phenomenal mental processes, he might be claiming here that the free imagination must not be determined by any natural or empirical laws. (I defend this point in Section X.4.) But Kant further characterizes imaginative freedom through a second negation: the aesthetic imagination relates to the understanding in a “play” that counts as “free” because no determinate concept restricts the imagination to a specific rule of cognition (KU, 5:217). Freedom of imagination “consists in . . . schematizing without concept” (KU, 5:287) and requires the absence of all constraint by concepts (KU, 5:242). Thus, it seems that in Kant’s aesthetics the “foreign causes” that hinder free imaginative activity include not only forces of nature with respect to which we are Kant on Freedom and Rational Agency. Markus Kohl, Oxford University Press. © Markus Kohl 2023. DOI: 10.1093/oso/9780198873143.003.0011

340  Kant on Freedom and Rational Agency passive but also conceptual rules that derive from our intellectual self-­activity. Since the imagination is determined by concepts in all its (objective or objectifying) cognitive functions,1 only the non-­cognitive, aesthetic production and experience of beauty leaves room for freedom of imagination.2 Kant’s main reason for requiring that the aesthetic imagination be free from determinate concepts or laws is that if aesthetic appreciation were a matter of grasping the extent to which a particular object conforms to some general conceptual rule, then the aesthetic function of the imagination would not be genuinely different from its role in ordinary cognition. This would invite the rationalistic view that aesthetic appreciation is but a confused, inferior way of representing intellectual perfection (KU, 5:228). Kant preserves the sui generis status of aesthetic appreciation by stressing that it relates to an object not discursively through concepts but intuitively through sensibility, namely through a disinterested feeling of pleasure that cannot be derived from intellectual rules (KU, 5:211–12).3 “If one judges objects merely in accordance with concepts, then all representation of beauty is lost. Thus there can also be no rule in accordance with which someone could be compelled to acknowledge something as beautiful” (KU, 5:216). But in denying that the free imagination can be determined by intellectual rules, Kant’s account seems to leave no room for a positive conception of how the imagination can be a “productive and self-­active . . . authoress of voluntary forms of . . . intuition” (KU, 5:240). This raises a serious problem. As we saw (in Chapter III), a negative characterization of freedom cannot stand on its own: the essential positive meaning of freedom signifies a capacity for spontaneous self-­ determination under autonomous laws of right reason that provide standards of 1  This assumption is famously denied by Heidegger 1991. However, Kant frequently emphasizes that the cognitive syntheses of the imagination are based on (or even belong to) the understanding (see, e.g., A119; B151–2; B161–2). For detailed criticisms of Heidegger’s view, see Banham 2006: 127–30. 2  Against this, one might argue that the cognitive synthesis of imagination in mathematical construction is similar to the free imaginative synthesis in the creation of art (Crawford 1982: 157–66; Gibbons 1994: 110–11). However, in mathematical construction the imagination is determined by objective rules: namely, by concepts determining the general properties of the particular to be constructed (e.g., a triangle as opposed to a circle) and by the axiomatic “universal conditions of the construction” (A716/B744) that are derived from the original intuition of space. One might further object to my claim that only the aesthetic imagination is free by invoking lecture notes where Kant (reportedly) says that our practical faculties stand under the free play of imagination when they are not subordinated to our law-­governed freedom of choice (V-­MO/Col, 27:362). However, here Kant is speaking loosely: since the imagination is not itself a source of desiderative striving, its practical employment must either be directed by moral laws for free choice or determined by sensible forces (inclinations, passions, or affects). In the same passage he (reportedly) says that without free choice a human being is “a plaything of other forces and impressions.” A different lecture transcript on the same topic has him saying that if we do not freely choose the direction of the imagination, “[t]he imagination directs itself according to the inclinations” (V-­Anth/Mron, 25:1260). 3  It is crucial to Kant’s argument here that aesthetic pleasure be disinterested; for the interested pleasure connected to representations of morally good actions can be derived from intellectual moral laws (KU, 5:211–12).

Freedom of Imagination and the “ Autonomy of Taste ”   341 correctness for free activity. This generic conception of free activity applies both in moral contexts, where we exercise our will as a capacity for free rational choice under the self-­legislated moral law, and in epistemic contexts, where we exercise our spontaneous “intellectual capacity of cognition” (V-­Met/Mron, 29:881) under “the general laws of the understanding” (KU, 5:186) that derive from the “autonomy . . . of the understanding” (EEKU, 20:225). Kant’s doubly negative account of imaginative freedom precludes an appeal to practical or theoretical laws. The imagination cannot provide its own laws since it is not a law-­giving faculty (KU, 5:241). While Kant does stress that aesthetic judgments are autonomous (KU, 5:282–3), the relevant notion of autonomy here again seems wholly negative: Kant characterizes the autonomy of taste merely as the absence of heteronomy when he says that autonomous judges of taste (e.g., a “young poet”) do not permit “foreign judgments” to determine their own aesthetic verdicts (KU, 5:282). But negative freedom without a positive counterpart, or spontaneous self-­determination without governing standards of correctness, seems to be an ab­surd­ity in the light of Kant’s concept of freedom as a rule-­governed power that is not a mere liberty of indifference (MS, 6:226). Since lawlessness is precisely the opposite of freedom (WDO, 8:145), “the randomly (regellos) roaming imagination” which “confuses the head through the change of representations that are not objectively attached to anything” (Anth, 7:177) does not deserve the honorific title of Kantian freedom. Moreover, in Kant’s view a purely negative concept violates the conditions for representing really possible objects or capacities (see Section III.2). As long as we operate with mere negations our thoughts lack reference to something: “Negation signifies a mere want, and, so far as it alone is thought, represents the abrogation of all thinghood” (A575/B603). A free will that could not be considered in positive terms as having some determinate law-­governed character would be nothing, ein Unding (GMS, 4:446). Accordingly, a purely negative conception of the free imagination would represent the mere want of a real power. I will argue that these problems are not, in the end, fatal to Kant’s account. Before developing my own constructive account in Sections X.2 and X.3, I examine some alternative attempts to vindicate Kant’s idea of imaginative freedom. Here I first consider a suggestion made by Paul Guyer.4 He argues that Kant never clarifies whether imaginative freedom is a negative freedom from conceptual constraint or rather a positive power to free itself from existing conceptual constraints via abstraction from available rules. For Guyer it must be the latter: typically, objects can be classified under some empirical concept or other, so if the mere availability of concepts were sufficient to block the imaginative freedom required for aesthetic appreciation such appreciation would be a very rare

4  See Guyer 1997a: 220–5.

342  Kant on Freedom and Rational Agency (perhaps impossible) occurrence. To avoid this implication, freedom of imagination must be conceived as a positive power to abstract from existing concepts and thereby to create the imaginative space that makes aesthetic appreciation possible. Although Guyer is not concerned with the problem I have posed, his appeal to freedom of imagination as a positive capacity to abstract from existing concepts suggests a way of avoiding the problematic notion of a purely negative freedom. Kant indeed says: “. . . to be able to abstract from a representation . . . proves a freedom of the capacity for thought and the power of the mind to control the state of its representations” (Anth, 7:131; cf. Chapter VI). Thus, Guyer’s suggestion meshes nicely with Kant’s general views on mental freedom. However, this is not sufficient to fill the gap left by Kant’s doubly negative characterization of imaginative freedom. The question is: how can there be a power of abstraction without rules governing the act of abstraction which prescribe that one should turn one’s imaginative focus away from certain features while staying focused on other features? In the above quote (Anth, 7:131) Kant is concerned with the “capacity for thought” whose acts of abstraction are governed by concepts. When I perceive a four-­legged, tail-­wagging, barking animal whose ears have a peculiar egg-­shaped form, I can abstract from that ear-­form because my thought is governed by the concept “dog,” a rule for conceiving different particulars as generically identical if they share the marks contained in the concept. Thus, freedom as a positive power of abstraction requires determinate conceptual rules governing the act of abstraction. But Kant’s negative characterization of imaginative freedom precludes the appeal to such rules.5 A second strategy for defusing this problem is to propose that the rule governing the free imagination derives from the reflective power of judgment. Unlike the imagination, this is a law-­giving faculty. One might suggest that its “heautonomous” principle lacks the kind of objective conceptual determinacy that would interfere with imaginative freedom. Textual support for this suggestion comes from a passage in the first introduction to the third Critique where Kant designates aesthetic judgments as judgments of reflection whose claim to universal validity is grounded in a subjective rule of the power of judgment (EEKU, 5:225). While I am not strictly opposed to this suggestion, I cannot endorse it as a (self-­standing) solution to the problem under consideration because it is fundamentally unclear which rule of reflective judgment is supposed to guide the aesthetic imagination. In the cited passage (EEKU, 5:225) Kant describes reflective judgment as a capacity for comparing and connecting concepts of objects, which 5  Crawford suggests that one analogy between the activity of imagination in aesthetic and mathematical creation is that one must abstract from irrelevant particular features to grasp the significance of the product (1982: 165). However, mathematical abstraction is governed by objective rules (concepts, axioms; see n. 2) that specify precisely what features one ought to abstract from and what other features one must keep in focus.

Freedom of Imagination and the “ Autonomy of Taste ”   343 are given to reflective judgment from elsewhere (namely, from the understanding), with “obtaining cases” where these concepts are instantiated. This abstract description is rather uninformative, but it raises a crucial worry concerning imaginative freedom: the act of applying given concepts to appropriate cases cannot exemplify how the aesthetic imagination freely schematizes without a concept. In the second introduction Kant specifies a concrete heautonomous rule of reflection that clearly pertains to the comparison and connection of concepts: namely, the principle of the formal purposiveness of nature as a rule for combining different empirical laws and concepts into a unified system (KU, 5:185–6). But since this principle directs the use of reflective judgment in the scientific classification of nature, it cannot govern the kind of freedom that our imagination enjoys in its aesthetic engagement with individual natural forms and artworks.6 A third proposal runs as follows. While Kant says that imaginative freedom consists in schematizing “without a law,” he also insists that the aesthetic imagination must schematize “lawfully” to be in free harmony with the understanding; it must accord with the conditions of cognition “in general” (KU, 5:217–18). One might suggest that the lawful accord of the imagination with the general conditions of cognition is sufficient to give imaginative freedom a positive direction.7 However, the idea that the imagination acts lawfully without being determined by any specific law is unclear, even paradoxical.8 I trust that this is a familiar source of frustration (and intrigue) for readers of the third Critique. But in the context of the problem under consideration, this worry has a special bite. For Kant, free agency must be directed by normative standards of correctness. Given this ­conception of free activity, the notion that freedom of imagination consists in “lawful” activity is insufficient unless one can specify certain laws that provide a positive normative direction for the imagination and thereby prevent it from collapsing into a random liberty of indifference. But this seems precluded by the negative condition that the free imagination must synthesize “without a law.”9 There is a cluster of readings which, despite central differences, share a central idea that they might offer in response to this worry: phrases like, “accordance with the general conditions or laws of cognition” designate a process of open-­ ended reflection that is indeterminately guided by the norms of empirical 6  See Allison 2001: 61–2. 7  On Allison’s view, the interplay between the faculties in aesthetic reflection “is ‘free’ insofar as it is not directed by a determinate concept, but it is still guided by the general conditions of cognition” (2001: 117–18). 8  Allison rephrases Kant’s formulations in various ways (2001: 49–50), but his reformulations seem just as paradoxical as Kant’s own phrases. Ginsborg raises this as a worry for Allison’s account (2015: 97–9). 9  Kneller suggests that the imagination “determines its object in accordance with feeling rather than objective law” (2007: 44). However, it is unclear how a feeling can govern or direct the imagination. Moreover, for Kant the free play of the imagination with the understanding precedes, or is the ground of, the feeling of aesthetic pleasure (KU, 5:216–19). Hence aesthetic pleasure cannot provide a direction for that activity.

344  Kant on Freedom and Rational Agency concept-­formation. For Allison the aesthetic imagination is focused on “a pattern or order . . . which suggests an indeterminate number of possible . . . conceptualizations, none of which is fully adequate, thereby occasioning further reflection.”10 Rachel Zuckert argues that the understanding sets an “indeterminate norm” for the aesthetic imagination by directing its free play towards an unreachable “ideal empirical conceptual content,”11 a “maximal, systematic unification of maximal diversity.”12 In Ulrich Schlösser’s view the aesthetic imagination has a positive orientation because it tends towards, but never fully realizes, the fitting combination of abundance and order that yields a fruitful empirical concept.13 This is an important line of interpretation which suggests an intriguing link between Kant’s aesthetics and his theory of empirical cognition. I cannot do full justice to this approach here, but I want to indicate two reasons why I am reluctant to accept it as a solution to our problem. First, the idea that the free imagination is governed by general norms of conceptualization still seems highly paradoxical in light of the fact that the aesthetic imagination is inherently opposed to conceptual classification. It is puzzling how the norms of empirical concept-­formation can provide positive guidance for an activity that resists compliance with these norms. An imaginative activity that defies the cognitive norms which direct it and whose products are never “adequate” relative to what these norms demand seems deficient: it is unclear why we should experience it as pleasant rather than frustrating.14 Moreover, if the free aesthetic imagination derives its positive direction from norms of empirical cognition, then it seems that freedom of imagination should be fully realized, not hampered, when the imagination successfully contributes to the formation of a determinate, fruitful cognitive-­empirical concept.15 The second problem with the idea that the free imagination is governed by norms of empirical conceptualization is that the aesthetic imagination seeks to (symbolically) exhibit supersensible ideas that transcend all empirical concepts (KU, 5:314). Since this aspect of Kant’s view will be crucial to my reading, I shall postpone expounding it until Section X.2. My present point is that since the aesthetic imagination strives towards exhibiting something supersensible that, as such, cannot be captured by any (however idealized) empirical concept, the

10  Allison 2001: 51. 11  Zuckert 2007: 291. 12  Zuckert 2007: 287–8. 13  Schlösser 2015: 216. 14  For a similar point, see Ginsborg 2015: 97–9. Allison replies (2003: 183) that since the free imagination does not aim at cognition there is no deficiency and frustration involved when that aim is not met. However, if the free imagination does not aim at cognition, it is unclear how it could be positively directed by the norms of cognition. 15  Schlösser 2015 suggests that the aesthetic imagination is free because its own tendency is not “dominated” by the understanding. But if the positive tendency of the imagination consists in striving towards the formation of fruitful empirical concepts, it is unclear why its freedom would be hampered if it were dominated by the rules of logical reflection that govern concept-­formation or by related cognitive (e.g., categorial) laws.

Freedom of Imagination and the “ Autonomy of Taste ”   345 positive guidance for our aesthetic imagination cannot derive from norms of empirical conceptualization. A further way of responding to the issue I have raised is to dismiss it as a pseudo-­ problem. Why, one might ask, should Kant’s aesthetics not happily employ a purely negative conception of unrestrained freedom? Perhaps we should welcome the possibility that Kant sacrifices a unified model of free activity, thus abandons his infamous systematic furor, to make room for a kind of freedom that fits his distinctive aesthetic subject matter. As we will see, I am sympathetic to the point that Kant’s idea of imaginative freedom must be understood in relation to the intricate demands he places on our distinctively aesthetic self-­activity. But it seems implausible that Kant would use his “cardinal” concept of freedom to designate a capacity which is entirely unlike freedom in its familiar senses. Moreover, Kant needs a unified model of free activity for his central claim that beauty is a symbol of morality (KU, 5:353). By this he means that there is an analogy between the mental state aroused by aesthetic reflection and moral consciousness. This analogy rests (partly, but essentially) on the connection between freedom of imagination and moral freedom of will: taste allows for a gentle transition from sensible pleasure to a pure moral interest because in taste the imagination can be regarded “even in its freedom,” which enables us to take a sensibly undetermined aesthetic pleasure in sensible objects (KU, 5:354). Importantly, Kant here does not posit a mere analogy between the concepts “free imagination” and “free will.” An analogy obtains when we can apply the same “rule of reflection” to two representations even though they refer to different objects (KU, 5:351), as in the case of “hand mill” and “monarchy.” The sensible representation of beauty can analogically symbolize the intellectual representation of the morally good even though these representations relate to separate objects, because both representations fall under the same “rule of reflection” (KU, 5:352). This requires that there be literal, non-­analogical connections between beauty and morality (just as there are literal connections between a hand mill and a monarchy, e.g., that both are determined by a single will). These are the connections that the analogizing rule of reflection picks up on: pleasure in beauty is pure and independent of sensible charms, like pleasure in the morally good; and the imaginative appreciation of beauty is free, like the will under moral laws. Thus, unless “freedom of imagination” and “freedom of will” are species of the same generic concept of freedom, this concept cannot support the rule of reflection that enables an analogical comparison between beauty and morality. Hence, there is a strong interpretive desideratum to integrate imaginative freedom with Kant’s general doctrine of freedom. I will try to develop an interpretation that meets this demand. In the next section I consider the imaginative freedom involved in aesthetic production. In Section X.3 I consider the imaginative freedom involved in the aesthetic experience of “given” beauty.

346  Kant on Freedom and Rational Agency

X.2.  Freedom of Imagination in Aesthetic Production For Kant the capacity to produce art that contains “spirit” (KU, 5:313) and originality (KU, 5:308) is genius. Spirit is one of the “faculties of the mind that constitute genius”; it is “the animating principle in the mind” which sets the artist’s mental powers into a free, self-­maintaining play (KU, 5:313). Hence, we can get a handle on what kind of freedom the imagination has in original artistic production by considering what Kant says about the imaginative activity of spirit and genius. Kant further explains spirit as “the faculty for the presentation of aesthetic ideas.” These are “representation[s] of the imagination” that occasion much thinking but are irreducible to any determinate concept because their rich sensible content outstrips our intellectual capacities (KU, 5:314).16 In creating aesthetic ideas “we feel our freedom from the law of association” that determines “the empirical use” of the imagination. The “productive” imagination transforms the sensible material that is lent to us by empirical nature “into something entirely different, namely, into that which steps beyond nature.” This transformation-­ process through which artistic spirit freely creates aesthetic ideas occurs “in accordance with” two types of principles (KU, 5:314). The first type of principle Kant mentions (at KU, 5:314) are “analogous laws,” “principles in accordance with which the understanding apprehends empirical nature.” This refers to the categorial principles of understanding that govern objective experience. I conjecture that Kant calls these principles “analogous laws” because in the context of aesthetic production they do not play their usual, determinately constraining role but still retain some relevance; hence their role in aesthetic production is (merely) analogous to their role as laws governing the objective experience of nature that “seems too mundane to us” in our aesthetic mindset. Categorial laws are relevant for the production of aesthetic ideas because such ideas have a representational content which contains a synthesis of sensible manifolds. Since this synthesis pertains to the supersensible, it yields a non-­literal, symbolic content which cannot be exhaustively captured by a priori categorial functions for combining sensible manifolds or by empirical concepts that exemplify these functions. Nevertheless, the synthetic content of aesthetic ideas must be partially articulable in some such conceptual terms. (I clarify this point below.) 16  This description of aesthetic ideas may seem reminiscent of Allison’s proposal that the aesthetic imagination pursues an open-­ended consideration of conceptual possibilities. However, for Allison this process is just like logical reflection aimed at empirical concept-­formation: the only difference is that the imagination forms no determinate concept (2001: 50–1). This invites the worries that I raised in Section X.1. Zuckert suggests that although the aesthetic imagination “transcends discursive conceptual cognition” (2007: 230), its aesthetic ideas can be viewed as “approximating the ideal of perfection . . . for . . . empirical conceptual content” (2007: 291). But if an aesthetic idea transcends discursive conceptual cognition and “steps beyond nature” (KU, 5:314), how can it pursue an ideal of perfection for essentially discursive empirical cognition that is strictly confined to the natural world?

Freedom of Imagination and the “ Autonomy of Taste ”   347 The second principles governing aesthetic production are ideas that “lie higher in reason” than concepts of understanding. Aesthetic ideas of imagination “strive toward something lying beyond the bounds of experience, and thus seek to approximate a presentation of ideas of reason (intellectual ideas)” such as the kingdom of the blessed. Even when artists illustrate “that of which there are examples in experience,” e.g., phenomena such as death, vice, love, or fame, they seek to magnify these sensible objects “beyond the limits of experience, with a completeness that goes beyond anything of which there is an example in nature, by means of an imagination that emulates the precedent of reason in attaining to a maximum” (KU, 5:314). The “precedent of reason” is set by reason’s regulative prescription to strive for a complete unified maximum of cognition. This totalizing model inspires the artistic attempt to enhance our experiences of ordinary sensible things so that these things no longer appear mundane and can arouse aesthetic pleasure. My key suggestion for salvaging a positive conception of imaginative freedom is that the principles of understanding and reason loosely govern or guide the imagination in aesthetic production without imposing a lawful constraint on free imaginative activity. Before I unpack this proposal, let me first clarify a potential ambiguity in Kant’s claim that the production of aesthetic ideas occurs “in ac­cord­ ance with” principles of reason. He mentions both specific rational ideas (e.g., the idea representing the kingdom of the blessed) and the general prescription of reason to strive for a complete unified maximum of representational content. Since Kant introduces these rules in the same context without indicating any separation, we should view them as closely intertwined (rather than as serving wholly distinct functions in aesthetic production). In its free creation of art, imaginative activity is always geared towards the sensible presentation of specific rational ideas. All specific rational ideas, as such, involve a complete maximum of representational content: e.g., the idea of the kingdom of the blessed represents a unified totality of virtue and happiness. Thus, the general prescription to strive towards a unified maximum of content is best understood as a sort of meta-­rule that directs all imaginative attempts to sensibly exhibit specific rational ideas. This is confirmed by the way in which the general prescription to attain a maximum of cognition is introduced in the first Critique (A508/B536): there it emerges as a constructive-­regulative (rather than a dialectical-­constitutive) way of employing the maximum of representational content that is contained in all specific metaphysical ideas of reason. Reason’s general call for maximization also indicates a meta-­rule for using concepts of the understanding in aesthetic production. Since such production strives for a presentation of rational ideas that involves a manifold of sensible intuitions, it must employ concepts of understanding as rules for combining sensible manifolds. But these concepts are not suitable for a symbolic sensible presentation of supersensible ideas unless their “mundane” cognitive content is magnified or

348  Kant on Freedom and Rational Agency elevated beyond ordinary sense-­ experience under the guidance of reason’s meta-­rule. To see how these rules govern the productive imagination, let us first consider how the imagination is determined by concepts in standard cases of empirical cognition. When I hear a barking noise around the corner my concept “dog” constrains me to imagine the source of this noise in certain ways rather than others, e.g., as having four legs and a tail but not as possessing wings. Or, suppose I perceive something whose sensible features (red, round, tiny black stem) call for the classification “apple.” Cognizing the object as an apple requires me to have implicit thoughts of other non-­actual perceptions of that same object; this means (in part) that my imagination must be disposed to construct images that present further potentially perceptible features of the apple, e.g., its white flesh.17 Here, again, concepts function as strict classificatory rules that determinately constrain my imaginative dispositions.18 By contrast, consider now a painter’s attempt to present the rational idea of the temptation to evil. This idea provides some direction as to what images might yield suitable models for its sensible presentation: the image of a blade of grass does not look like a promising choice but the image of an apple seems intriguing. If that image is selected as the primary vehicle of aesthetic presentation, the empirical concept “apple” provides some guidance: if an apple is to be depicted it ought to have certain general features rather than others, e.g., its shape ought to be round rather than triangular. But the function that the empirical concept serves in aesthetic production is merely “analogous” to the role that it plays in cognitive experience. In aesthetic contexts the concept does not prohibit—­as it would in ordinary cognition—­the imagination from emulating reason’s idea of complete totality by picturing the apple as shiny, ripe, and untainted in a magnified form that exceeds all actual experience (and thereby gives a new aesthetic dimension to the purely intellectual idea of temptation). Nor does the empirical concept prohibit the depiction of (say) a pitch-­black thorn that is growing from the apple and which always moves so that it is not visible to the observer. The rational idea of temptation to evil vaguely suggests some such sensible image, but without specifically demanding this (or any other particular) combination of intuitions. 17  I take these points from the seminal discussion of how the imagination contributes to cognition in Sellars 1978 and Strawson 1982. 18  This is only a sketch of how the imagination contributes to empirical cognition. For a complete account one would need to distinguish between: (1) the role the imagination plays in the formation of new concepts; (2) the role it plays in the above examples where one subsumes perceptions under a familiar concept; (3) the role it plays in empirical apprehension or perception. Some commentators hold that the imagination has a non-­conceptual character at least in (3) (Young 1988). Even if this were Kant’s view, it would not entail that the imagination in perception is free: qua non-­conceptual, the imagination in (non-­aesthetic) perception would be (as Young notes) the same as animal perception and would thus be determined by natural laws.

Freedom of Imagination and the “ Autonomy of Taste ”   349 These (admittedly schematic) cases illustrate the contrast between the exercise of imagination in empirical cognition that is determinately constrained by objective rules and the exercise of imagination in aesthetic production that is only loosely guided by rules. The guidance that the aesthetic imagination receives from empirical concepts and rational ideas leaves the subject with the negative freedom to ignore the constraints imposed by the unity of experience and to select from a vast range of possibilities for the aesthetic presentation of some rational content. But the looseness and vagueness of this conceptual guidance does not render it altogether negligible. The specific rational ideas that artists try to exhibit give some indication as to which sensible content is suitable for the presentation of specific rational themes.19 The rational (meta-) demand for a complete totality of cognition sets a precedent that the creative imagination can emulate in its attempt to reshape and magnify ordinary experiences into something non-­ mundane. And, although the creative imagination transcends the limits of empirical concepts, such concepts yield some direction to ensure that when artists transform a natural object into something that goes beyond nature the product is still recognizable as an object of a generic kind (e.g., an apple) so that the sensible content of the artwork can symbolize its theme(s) and stimulate our thinking.20 These guidelines can be regarded as rules governing the creative imagination: they give a positive direction to imaginative freedom and thereby prevent it from degenerating into an arbitrary liberty of indifference. Because these rules leave room for indeterminacy, subjective leeway and interpretation, they do not function as determinate, strict, or objective laws. This line between objectivity and arbitrariness might seem too fine. Does the looseness of the putative rules governing aesthetic production not undermine the authority that these rules would need to possess for preventing a liberty of indifference? The notion of a subjective rule that provides only a loose guidance is indeed peculiar. But artistic creativity is also a peculiar form of activity: it cannot be assimilated to law-­governed fields like morality or mathematics, and yet it does not collapse into an arbitrary anything goes-­stance. Artists cannot specify with mathematical precision why certain images or metaphors work in a particular aesthetic context but they can say things to justify, however inconclusively, their impression that they do work: they can indicate how these images interact with other components of the artwork or they can invoke the powerful cultural associations encouraged by certain images (as with the apple and temptation), etc. Thus, the peculiar, goal-­directed yet conceptually indeterminate focus of artistic creativity calls for a distinctive form of normative guidance.

19  For the importance of connecting the play of imagination to a rational “theme,” see Anth, 7:177. See also the helpful discussion in Rueger 2008: 306–9. 20  On this point, see Lüthe (1984: 73–4): genius must grasp which aesthetic attributes to incorporate into an aesthetic idea so that this idea can be seen as presenting specific themes.

350  Kant on Freedom and Rational Agency In Kant’s account, this guidance comes from rational ideas and concepts of understanding which provide intellectual input for aesthetic production without determinately constraining such production to take a particular course. A specific rational idea suggests certain sensible manifolds (images, sounds, colors, etc.), but it does not privilege a specific mode of presentation as the only possible or the most suitable one. Relatedly, in order for an aesthetic idea to “occasion much thinking” it must enable recipients to recognize the artwork under some empirical concept or other so that their mind is sufficiently stimulated to draw non-­ arbitrary connections to other concepts and themes. This imposes a normative constraint on the artistic imagination, a directive to incorporate into its aesthetic mode of presentation some of the marks contained in the relevant concepts. But this constraint differs notably from the constraint that an “empirical objective concept” (KU, 5:292) imposes on the imagination in ordinary cognition. The empirical concept (or the a priori category it exemplifies) does not lawfully determine how, precisely, artists must combine the various component parts of their symbolic sensible manifold; artists remain free to focus on just a few particular marks contained in the concept and ignore how these relate to further marks. Moreover, the artist is free to enhance the selected marks under the “meta”-rule that prescribes emulating reason’s totalizing paradigm in the generation of sensible content so that one can achieve a non-­mundane mode of sensible presentation. But this is only a vague rational directive that does not impose on the imagination a precise recipe for determining which particular conceptual marks must be enhanced or in what specific manner the sensible manifold must be shaped into a unified totality. Hence, reason’s general maximizing rule does not fix for the artist any one right way to proceed. By contrast, reason’s categorical moral imperative provides a lawful procedure for determining, with regards to every agent in relevantly similar circumstances, whether the agent is morally permitted to act in a certain way. To further clarify the contrast between indeterminate rules and strict laws, we can consider Kant’s claim that beautiful art has a “mechanical” component “which can be grasped and followed according to rules” of good taste. These are “determinate rules . . . from which one may not absolve oneself.” They set a lawlike paradigm for what counts as “academically correct” in art (KU, 5:310; Anth, 7:124–5). Presumably, Kant here refers to (e.g.) meter rules for poetry that specify with precision, thus without leaving any room for subjective leeway, how poets must (not) arrange their verses. Obedience to these laws of taste cannot be a sufficient condition for producing art that is original (spirited) and beautiful, but it is a necessary condition: taste “gives genius guidance as to where and how far it should extend itself . . . and by introducing clarity and order into the abundance of thoughts it makes the ideas tenable” (KU, 5:319). Obedience to these laws of taste conflicts with the “lawless freedom” of imagination (KU, 5:319) because such laws “permit damage to the freedom and richness of the imagination” (KU, 5:320).

Freedom of Imagination and the “ Autonomy of Taste ”   351 The phrase “lawless freedom” seems to violate Kant’s doctrine that a lawless freedom is a contradiction in terms, ein Unding. But my interpretation suggests a way of avoiding this consequence. Because freedom of imagination requires the absence of determination by strict lawful rules, such freedom is indeed “lawless”; hence, to the extent that the creative imagination is subjected to determinate academic laws of taste its distinctive freedom is hampered. But the imagination is also directed by rules that, due to their indeterminacy and looseness, do not govern aesthetic production as laws: namely, by cognitive concepts that function only “analogously” in aesthetic production and by rational ideas that provide a specific thematic content and a general totalizing orientation for the artistic process. These rules govern the productive imagination in its free activity. They ensure that the freedom of imagination, while indeed lawless, is not a purely negative capacity (ein Unding). This also means that if we are unhappy with Kant’s (perhaps anachronistic) appeal to strict lawlike canons of good taste, abandoning this component of his view would not force us into a conception where artistic creation is entirely random and rule-­less.21 One might object to my account that not all artistic creations which exhibit rational themes are beautiful: a poem or painting might symbolize a rational topic in an ugly (boring, trite, etc.) fashion. Hence, my proposal that the free imagination is governed by rational ideas (and other concepts) is not sufficient to explain how the imagination freely creates aesthetic value. But my account does not construe guidance from rational ideas as a sufficient condition for the creation of beauty. The exhibition of a rational idea may well fail to create something of aesthetic value; it may produce only an eclectic, boring conglomeration of commonplaces about (e.g.) moral evil. On my account the guidance that the productive imagination receives from rational ideas (and from other concepts) is merely necessary for the free creation of original, beautiful art. Thus, I concede that my account in this section does not fully (sufficiently) explain how artists can create an aesthetically valuable exhibition of rational themes. But I do not see this as a problem. My interpretation seeks to ensure that the lawless character of imaginative freedom does not collapse into a rule-­less, arbitrary liberty of indifference. This desideratum is met through the (necessary) condition that the imagination in its free, lawless creation of beautiful art must receive positive if indeterminate guidance from rational ideas (and from other concepts). In Sections X.4 and X.5, I consider what further conditions the imagination must meet if it is to freely create an original, beautiful artwork.

21  One might worry that Kant’s emphasis on objective canons of taste undermines his emphasis on the autonomy of art (see Guyer 1997b). But Kant can reply that compliance with such canons is never sufficient for original-­spirited art: objective laws taste may do some “damage” to imaginative freedom, but they cannot fully dissolve such freedom (or the autonomy of art) without thereby also destroying all aesthetic value.

352  Kant on Freedom and Rational Agency

X.3.  Freedom of Imagination in Aesthetic Experience For Kant the imagination is “productive and self-­active” not just in artistic creation but also in aesthetic experience. He admits that this seems counterintuitive because in its apprehension of a “given object” the imagination is “bound to a determinate form of this object, and to this extent [it] has no free play (as in invention [Dichten]).” Still, Kant argues, it is “quite comprehensible” that a given object “can provide” the imagination “with a form that contains precisely such a composition of the manifold as the imagination would design in harmony with the lawfulness of the understanding in general if it were left free by itself ” (KU, 5:240–1). Kant here treats artistic “invention” as the paradigm for the freedom of imagination and then characterizes imaginative freedom in aesthetic experience by reference to this paradigm: the aesthetic apprehension of a given form is free because it is “precisely” like a form of art that the imagination would freely invent. Significantly, this model also applies to the aesthetic experience of natural beauties: beautiful nature must seem like art (KU, 5:306). The imagination can “feel” its freedom in exploring a beautiful natural form because it can apprehend such a form like a product of its own artistic making, as if it had freely designed it.22 How can the aesthetic forms of beautiful flowers, birds, or snow-­covered mountains be “precisely” like the forms of intentionally designed artworks? We saw that the forms of beautiful artworks consist of aesthetic ideas. If the beautiful forms of natural objects are precisely like artforms, natural beauties must also express aesthetic ideas. This is indeed Kant’s view: “Beauty (whether it be beauty of nature or of art) can in general be called the expression of aesthetic ideas”; in the case of natural beauties “mere reflection on a given intuition, without a concept of what the object ought to be, is sufficient for arousing and communicating the idea of which that object is considered the expression” (KU, 5:320). Thus, the imagination is free in our aesthetic experience of nature because we can apprehend nature as if it was freely created by our artistic imagination; the reason why we can apprehend nature in this way is that we can imaginatively interpret beautiful nature as expressing those representations which characteristically result from our free artistic creativity: aesthetic ideas.23 22  One might object that the term “invention” need not have a purely artistic meaning; hence, the passages cited here do not show that artistic forms provide the only paradigm for the free aesthetic experience of nature. However, it is unclear what a non-­artistic manner of imaginative inventing (Dichten) is supposed to be. When Kant says that “given” natural objects can be apprehended aesthetically as sensible manifolds that the imagination would “design . . . if it were left free by itself,” the latter phrase refers to a process where the imagination produces a beautiful object that does not yet exist. I do not see that such production occurs anywhere other than in artistic creation. Moreover, in the present context “invention” must have an artistic meaning, for otherwise Kant would not be stressing that beautiful nature must seem, specifically, like art. 23  Rueger and Evren 2005 and Rueger 2008 also stress the importance of Kant’s appeal to aesthetic production in his account of free aesthetic experience. While I have learned from their helpful discussion, I do not agree with their view that aesthetic experience involves the following “counterfactual

Freedom of Imagination and the “ Autonomy of Taste ”   353 It is controversial whether Kant’s claim that all beauty expresses aesthetic ideas fits with his detailed analysis of aesthetic judgment.24 One might worry that assigning expressive content to all representations of beauty conflicts with the formalist strand of Kant’s aesthetics.25 I cannot fully address this worry here since I cannot pursue the difficult question of what Kant’s aesthetic formalism means or requires. I want to note, though, that Kant qualifies his formalist dismissal of sensations precisely when he argues that natural beauty has expressive content: “the charms in beautiful nature . . . are the only sensations which permit not merely sensory feeling but also reflection on the form of these modifications of the senses, and thus as it were contain a language that nature brings to us and that seems to have a higher meaning” (KU, 5:302). This “higher meaning” relates to transcendental ideas (that “lie higher in reason”; KU, 5:314). As we saw (in Section X.2), aesthetic ideas of imagination symbolically exhibit ideas of reason such as moral innocence. “Thus the white color of the lily seems to dispose the mind to ideas of innocence . . . . The song of the bird proclaims joyfulness and contentment with its existence. At least this is how we interpret nature, whether anything of this sort is its intention or not” (KU, 5:302). One might raise a related (though not necessarily formalist) concern here. My view that the aesthetic appreciation of both artistic and natural beauty involves conceptual, rational content may seem to conflict with two central features of Kant’s aesthetics: (1) Artistic beauty always involves concepts in the way that natural beauty does not; (2) Judgments of natural beauty are “without a concept.” However, (2) need not be taken to mean that such judgments have no conceptual content. We can plausibly read Kant as stressing that reflective judgments of natural beauty employ concepts only in a symbolic-­indeterminate way rather than in the standard, determinative-­ objectifying function they serve in empirical ­cognition.26 Moreover, while this point applies to both natural and artistic beauty, test”: “we . . . can compare the given form with a form that the imagination could have produced freely . . . if this counterfactual test comes out positive—­the given form and the counterfactual free form coincide—­then we say that the given object is beautiful” (Rueger 2008: 301–2). I do not quite see how this test is meant to work: how does one compare an actual with a “counterfactual form”? Moreover, for Kant ordinary aesthetic experience does not involve any such sophisticated decision-­ procedure that seems detached from subjective immersion in beauty. Ordinary aesthetic experience imagines a given form as an expression of aesthetic ideas; about these ideas one can make the true counterfactual statement that they could have been produced by the artistic imagination, but that counterfactual belongs to philosophical analysis and thus need not enter into the content of ordinary aesthetic experience. 24  Lüthe claims that aesthetic ideas do not figure in Kant’s doctrine of taste but belong solely to his Produktionsästhetik (1984: 67: cf. Kuhlenkampf 1978: 154). Such a reading is understandable given the absence of explicit references to aesthetic ideas in Kant’s “Four Moments,” but it cannot be squared with passages such as KU, 5:302, 320. Those who hold (like I do) that aesthetic ideas are central to Kant’s doctrine of taste include Allison 2001: 288; Chignell 2007c; Rogerson 2008; Rueger and Evren 2005; Rueger 2008. 25  For a seminal treatment of this issue, see Guyer 1977. 26  Here I disagree with Ginsborg’s (2015) view that judgments of taste involve a “primitive normativity” that abstracts from any conceptual basis. On her account a judgment of taste claims nothing

354  Kant on Freedom and Rational Agency we can accommodate (1) by indicating a further sense in which the appreciation of beautiful nature proceeds without certain concepts that must figure in the reception of art. Kant indicates this sense when he says (emphasis mine): “in beautiful nature” unlike in beautiful art, “the mere reflection on a given intuition, without a concept of what the object ought to be, is sufficient for arousing and communicating the idea” that the object expresses (KU, 5:320). In aesthetic production the artist is operating with a deliberate plan of what the artwork ought to signify (e.g., a sensible presentation of the rational themes of hell and envy). The receptive appreciation of art also involves some (perhaps implicit) awareness of the fact that the artwork was produced via rule-­governed intentional design. By contrast, the thought that a given sensible form expresses intentional design is characteristically absent from the aesthetic experience of natural beauty (cf. KU, 5:302). A further objection to my account is that it does not adequately capture the sense in which judgments of natural beauty involve a free play of imagination and understanding: if our imagination interprets the sensibly given as symbolizing transcendental ideas, it rather seems to be playing with reason. However, since the aesthetic imagination is focused on the sensible presentation of intellectual ideas it must also be “playing with” concepts of the understanding qua (symbolic, analogical) rules for combining sensible manifolds into a unified maximum of representational content. Thus, we should take Kant to hold that the free play in aesthetic experience involves both understanding and reason. This follows from his view that beauty “in general” (KU, 5:320), “at bottom” (KU, 5:356), thus not merely occasionally or in select cases, is an expression of aesthetic ideas that strive towards presentations of rational ideas via syntheses of sensible manifolds that are governed by analogical rules of understanding.27 The fact that our aesthetic apprehension of nature involves interpreting nature as a sensible sign of intellectual ideas is congenial to my key claim that the imagination in its free activity is guided by indeterminate conceptual rules. In aesthetic experience these rules do not dictate specific aesthetic responses with law-­like objectivity, but they provide a conceptual framework for our subjective aesthetic interpretation of nature. For instance, cognitive concepts direct us to combine but its own universal validity. Like others (Dohrn 2006: 206; Schlösser, 2015: 211), I have trouble seeing how simply asserting that my judgment has universal validity suffices to make it the case that the judgment has this property. In accepting that judgments of taste have a non-­primitive conceptual basis which (however) does not allow for complete determination of aesthetic properties, I am in general agreement with Ameriks, although I do not accept his further point that this conceptual basis suffices to ground the objectivity of taste (2003: 312–18, 338–41). 27  One might hold that the imagination plays with reason only when it apprehends the sublime. I cannot consider Kant’s intricate views on the sublime here. But we can take him to hold that the “play” which occurs in the case of the sublime is specifically different from the way in which the imagination plays with rational ideas in its apprehension of beauty. To mention just two central differences: first, much unlike the experience of beauty the encounter with the sublime involves a (partly displeasing) conflict between imagination and reason; second, this encounter includes an awareness of the imagination’s inferiority as compared to reason’s powers (KU, 5:258).

Freedom of Imagination and the “ Autonomy of Taste ”   355 certain aspects of our perceptual field, such as the blossoms, stem, and leaves of a lily but not the mosquito flying around it, as parts of one organism. On that basis we can interpret the graceful purity of the lily’s color as symbolizing the idea of a living being’s moral innocence. Here one may press worries similar to those which arose in Section X.2: is the imagination truly governed by rules when these rules do not impose objective (lawlike) constraints? My reply is, as before, that the appeal to a subjective, indeterminate kind of rule-­governedness is legitimate precisely because it uniquely fits the character of aesthetic experience. Such experience lends itself to a special kind of discourse about the aesthetic merit of (natural or designed) objects whose possibility rests on two conditions. On the one hand, aesthetic discourse is structured by rules that allow for genuine discussion and disagreement; thus it differs from pseudo-­debates about arbitrary sensual pleasures (“is hummus delicious or disgusting?”; cf. KU, 5:211–13). Since aesthetic judgments that are based on the free imaginative apprehension of natural forms go beyond an enumeration of idiosyncratic sense impressions, they must be expressible in terms of shared conceptual presuppositions which distinguish genuine discourse from mere emoting. On the other hand, aesthetic discourse characteristically cannot be settled via objective arguments or proofs that subsume particulars under determinate concepts or laws; it always contains an irreducible degree of subjective interpretation that leaves some conceptual indeterminacy. Let me illustrate this point by elaborating a bit on Kant’s lily example. The fact that a given object instantiates the general marks which allow us to classify it as a white lily does not entail that this object yields a beautiful, aesthetically pleasant expression of moral innocence. The verdict that this particular flower can be taken as a beautiful expression of moral innocence requires actual perception.28 An imaginative response assessing the lily’s aesthetic quality must dwell on perceptible features such as the relations that obtain between the different parts of the plant or between the plant and the larger natural scenery. Another observer might challenge my assessment that this flower is beautiful by pointing to what they see as the oddly disfigured shape of the petals which prevent the flower from expressing admirable moral traits. Here I can ask for clarification: why do the petals strike the other observer as disfigured? The observer’s response may spark further discussion: perhaps the petals are indeed oddly shaped but the oddness should not be seen as a disfigurement because it makes sense in the context of the whole—­the unusual shape of the petals complements their untainted whiteness and makes the lily stand out vis-­à-­vis other, similarly white but uniform and boring looking flowers. In the context of interpreting the lily as a sensible sign of moral ideas, this might mean: the lily symbolizes a character whose virtue is 28 For helpful discussion of why aesthetic appreciation requires perception, see Ginsborg 2015: 28–9.

356  Kant on Freedom and Rational Agency pronounced by an eccentric form that parts with ordinary conventions. While this debate cannot be settled through objective proofs, the (admittedly, still schematic) moves I have sketched are clearly above the level of “Mmh, hummus is delicious!” versus “Bah, hummus grosses me out.” One might object that the conceptual presuppositions underlying such debates, e.g., the notion of a link between whiteness and moral innocence or (in my earlier example) between an apple and moral temptation, depend on particular cultural (religious, literary, etc.) traditions. Hence, one might argue, they cannot support any claim to (subjective) universal validity. However, as long as the relevant concepts and the associations they encourage are part of a public rule-­governed culture, they can guide the aesthetic imagination in a universally valid fashion: everyone can grasp the relevant conceptual links by familiarizing themselves with the relevant cultural framework. There need not be any one cultural tradition that is uniquely relevant for aesthetic appreciation: Western Europeans might increase their aesthetic sensitivity by immersing themselves into (say) Indigenous customs and traditions. For Kant, the faculty of taste cannot simply be presupposed in subjects regardless of whether they are “prepared through culture” (KU, 5:265). While judgments about the sublime require culture more than judgments about beauty (KU, 5:265), this implies that judgments about beauty presuppose a degree of culture as well: we ascribe the faculty of taste to people only if they have “some culture” (KU, 5:265–6). The examples I have discussed suggest that the rational themes we apprehend in our aesthetic experience of nature via aesthetic ideas are moral themes. This is controversial, even among those who accept that for Kant all beauty involves aesthetic ideas that express some thematic rational content.29 I do not want to rule out that the experience of beauty might also involve non-­moral rational themes. However, I am inclined towards the moralistic interpretation, partly because of Kant’s exclusive focus on moral themes in his own examples but chiefly because he concludes the Critique of Aesthetic Judgment by declaring that “taste is at bottom a faculty for the judging of the sensualizing of moral ideas” (KU, 5:356).30 I shall now further clarify and defend my account of imaginative freedom in aesthetic experience by showing how it relates to some important issues that came up in Section  X.1, where I critiqued other strategies for showing how we can regard the imagination as (positively) free. In Section X.1 I argued that we can only regard freedom of imagination as a mental power of abstraction if aesthetic abstraction is governed by rules. On my reading this condition is satisfied: since the aesthetic interpretation of natural

29  See Chignell 2007a; Rogerson 2008; Rueger and Evren 2005; and especially Matherne 2013. 30  Proponents of the non-­moralistic reading often stress religious ideas as an example of aesthetically relevant rational but non-­moral ideas. However, for Kant morality is the essence of any rational religion (Anth, 7:192, 200).

Freedom of Imagination and the “ Autonomy of Taste ”   357 objects consists in imagining how such objects “sensualize” rational ideas, those ideas direct the imagination to free itself from representations that interfere with its interpretive effort. For example, when I interpret a lily as a symbol of innocent moral character this moral idea directs my imagination to abstract from: (i) my impression of a pleasant smell and my associated desire for white chocolate that would distort my imaginative focus on the purity and self-­sufficiency of an innocent moral character; (ii) cognitive concepts such as the principle that the lily is subject to mechanistic laws which would disrupt my imaginative focus on the moral quality of a free living being; (iii) concepts of “external [purposiveness], i.e., utility” (KU, 5:226), e.g., the knowledge that lilies sell well as ornamental plants which would inhibit my representation of non-­instrumental moral status. One might object that my emphasis on the central role that moral ideas play in aesthetic interpretation cannot explain Kant’s insistence that aesthetic experience must also abstract from the concept of “objective inner purposiveness, i.e., perfection” (KU, 5:227). The moral ideas that, on my reading, guide imaginative aesthetic experience designate the highest kind of objective perfection or inner purposiveness. My response involves a distinction (inspired by KU, 5:228) between two ways in which concepts of moral perfection can influence our responses to given perceptions. First, suppose I perceive a human agent who refrains from lying under conditions of duress; my cognition that the agent performed a good action under adverse circumstances produces in me a moral feeling of pleasurable admiration. Here the determining ground of both my practical cognition and the resulting (interested) feeling of pleasure is an objective concept of inner purposiveness (the moral law). Second, suppose I perceive a lily whose sensible form enables me to imagine it as a symbol of moral ideas, where this process of imaginative interpretation produces or constitutes a (disinterested) feeling of aesthetic pleasure. This feeling is not based on my cognition that the lily falls under a concept of moral perfection or goodness. Rather, to feel the pleasure that lies in imagining the flower as “sensualizing moral ideas” I must abstract from the constraints of objective practical cognition: I must bracket my certain knowledge that flowers are not proper subjects of moral predicates. Or, suppose (cf. KU, 5:204) that the lily is on a plantation maintained by slave labor. Here I can only initiate the free play of the faculties that imagines the lily as expressing positive moral themes if I abstract from my “interested” judgment that the lily morally speaking ought not to exist. I must abstract from my practical cognition that the existence of the flower falls under objective moral concepts to achieve a “disinterested” form of appreciation where moral concepts yield merely subjective guidelines for my aesthetic imagination of nature. Another important issue that came up in Section X.1 is Kant’s characterization of the aesthetic imagination as involving a “lawfulness without a law” or an “accordance with conditions of cognition in general.” While these phrases are

358  Kant on Freedom and Rational Agency clearly central to Kant’s analysis of taste, they seem uninformative, even ­paradoxical.31 To some extent, this allows every interpretation to put its own spin on these phrases. On interpretations like mine, these phrases may be (somewhat) illuminated through Kant’s doctrine of aesthetic ideas. In our aesthetic experience of nature, we apprehend sensibly given forms via aesthetic ideas, as symbolic exhibitions of rational themes. This process involves a purposive, rule-­governed, and pleasant interplay of all our cognitive (conceptual and sensible) faculties. This interplay is “lawful” because the aesthetic exploration of sensible manifolds is directed by intellectual rules that yield lawful conditions of all (practical and theoretical) cognition in general. This is a “lawfulness without a law” because in our aesthetic experience these conditions do not function as laws, i.e., they do not objectively determine aesthetic responses. Free aesthetic experience is not determined by the moral law (e.g., the moral law does not dictate that I must take pleasure in this lily as the expression of innocent character), but it can still be regarded as morally “lawful” and as being “in accordance with conditions” of practical cognition because it is loosely guided by rules for aesthetic interpretation that are drawn from our objectively lawful moral framework. Likewise, although free aesthetic appreciation is not determined by objective concepts of understanding (e.g., the concept “songbird” does not prevent my aesthetic interpretation from ignoring conceptual marks such as “determined by instinct” or “eats maggots”), it does accord with the general conceptual conditions of empirical cognition insofar as these conditions provide a general discursive framework for making aesthetic sense of things. For instance, conceptualizing a songbird as a type of substance that persists through time and that has causal powers enables me to represent the bird as a source of distinctive sound patterns and (on that basis) to imagine it as the organic seat of character traits such as joyful contentment.32 In roughly this vein, my reading might be able to account for Kant’s puzzling notion that aesthetic experience involves a “lawfulness without a law.” However, here my account invites an objection which arises from my earlier critique of the 31  To clarify: there is nothing inherently paradoxical in saying that the imagination as such accords with the general conditions of cognition. In §9 of the third Critique Kant indicates that the imagination contributes to cognition via its rule-­governed composition of sensible manifolds; this is line with his account in the first Critique. But he denies that the free, aesthetic imaginative synthesis of sensible manifolds is governed by the categories qua necessary conditions of objective cognition. This denial raises the puzzle of how the aesthetic imagination still accords with the conditions of cognition in some elusive “general” sense. 32  In supposing that the free imagination is governed by categories such as “substance” and “causality,” I follow Makreel (1990: 49–59). By contrast, Zuckert argues that aesthetic judging does not even implicitly employ the categories (2007: 292–5). But if the representation of objects in aesthetic experience were completely devoid of categorial content, it would be a wholly private mental episode that, as such, could not lend itself to intersubjective discourse. However, Zuckert is right in the sense that the categories do not figure in aesthetic responses as objective rules: categorial constraints on symbolic aesthetic representation are considerably more flexible than (merely “analogous” to) categorial constraints on objective, cognitive experience.

Freedom of Imagination and the “ Autonomy of Taste ”   359 proposal that aesthetic experience involves a “lawfulness without a law” insofar as it is an open-­ended process guided by the norms of empirical concept-­formation (cf. Section X.1). I argued that since aesthetic experience defies empirical conceptualization, this proposal turns aesthetic experience into a deficient, hence frustrating process. One might hold that my reading fares no better here: if aesthetic experience involves aesthetic ideas that strive for a presentation of supersensible objects by displaying sensible features which, despite all maximization, cannot truly capture the supersensible, then aesthetic experience is inherently deficient and frustrating on my reading as well. My response has two parts.33 First, even if my reading had to concede a final disconnect between what the imagination strives for and what it can achieve in aesthetic experience, my reading would still allow free imaginative activity a constructive teleological structure that it lacks on the alternative proposal. Although the imaginative apprehension of sensibly given forms cannot attain a complete presentation of supersensible content, we can at least regard the imagination as striving towards such a presentation: it seeks (“as it were”) to answer the call from the supersensible. By contrast, the imagination in its freedom actively resists compliance with the conditions for subsuming sensible objects under empirical concepts, conditions which supposedly guide the imagination on the alternative reading. The suggestion that we take pleasure in an imaginative process which (at least) strives towards its telos does not have the same paradoxical upshot as the alternative suggestion that we take pleasure in an imaginative process which actively defies the norms that supposedly give it a positive goal or direction. But secondly, there is no real disconnect, on my reading, between the positive orientation and the capabilities of the aesthetic imagination. Here we must note the subtle but importance difference between (1) the claim that the imagination seeks to present unpresentable ideas and (2) the claim that the imagination seeks to approximate a presentation of unpresentable ideas. Claim (2) does not imply a sense of failure or frustration because a combination of sensible intuitions can truly approximate a presentation of rational ideas. “Approximation” here is a success term designating a fitting, pleasing sensible form that may well be inherently complete qua sensible approximation of supersensible content.34 (Though we should leave room for degrees here, i.e., for greater and lesser, deeper and shallower approximations, if we want to allow that things can have more or less aesthetic value.) Likewise, the maximizing orientation that rational ideas provide for aesthetic interpretation should not be understood as a potentially frustrating demand to look for ever more and more ad infinitum but rather as a directive to 33  The following remarks also apply, mutatis mutandis, to imaginative activity in artistic creation (cf. Section X.2). 34  Here I do not follow Rueger and Evren (2005: 245) who argue that a symbolic aesthetic presentation can never fully succeed; if I understand them correctly, they take this to show that we can never truly experience beauty.

360  Kant on Freedom and Rational Agency grasp something remarkable which fittingly surpasses our mundane, ordinary experience (e.g., an extraordinarily pure form of whiteness); this, too, is a success condition that can be met. Another crucial worry one might raise for my account of aesthetic experience is that it fails to connect with Kant’s deduction of judgments of taste. Kant argues that such judgments can make a claim to universal validity because aesthetic experience accords with subjective conditions of cognition (KU, 5:289–92). On my reading, this accordance centrally involves aesthetic ideas. However, aesthetic ideas are not mentioned in Kant’s deduction. One might hold that Kant here focuses solely on the subjective conditions of empirical cognition, which have little if anything to do with aesthetic ideas qua representations of transcendent themes.35 But Kant’s deduction is so puzzling and its prospects of success are so unclear that we should perhaps welcome the possibility to utilize considerations that may help to defend the intersubjectivity of taste even though they do not figure explicitly in the deduction.36 This response gains further support from the fact that there are striking problems for views which seek to defend the intersubjectivity of aesthetic experience solely via the claim that such experience accords with the subjective conditions of empirical cognition.37 I do not want to rule out that these conditions play an important role in Kant’s deduction of taste (and more generally in his aesthetics). My account requires only that the doctrine of aesthetic ideas is also relevant to the deduction. While I cannot address Kant’s philosophically complex and textually challenging deduction argument here, I want to consider two passages in the third Critique which suggest that aesthetic ideas have an important role to play in this argument. In the first introduction Kant makes two points that call to mind his argument in the deduction. First, he explains that an aesthetic judgment is based on “three actions of the self-­active faculty of cognition” (EEKU, 5:220) that are presupposed by every empirical concept: apprehension of a sensible manifold; comprehension of that manifold into the concept of an object; intuitive exhibition of an object corresponding to that concept. Second, he argues that these acts can succeed only if a relation of fitting agreement obtains between imagination and understanding. 35  This worry arises for all readings that assign a central role to aesthetic ideas in Kant’s aesthetics. 36  See Rueger 2008: 312 for a related point. Chignell 2007a argues in a similar fashion. 37  For instance, if accordance with subjective conditions of empirical cognition were sufficient for the universally valid experience of beauty, then every ordinary empirical object would be beautiful since every such object accords with subjective condition of empirical cognition. (This problem does not arise on my reading because not all ordinary empirical objects can be taken to symbolize the supersensible.) One might respond that only some empirical objects qualify as beautiful because we can apprehend them in a distinctively easygoing manner, when our faculties display a special proportion of mood, whereas in standard cases of cognition our “cognitive investment” is more cumbersome (see Schlösser 2015: 212–16). But (cf. Guyer 1997a: 263) talk about proportions of mood and ease seems like a digression into contingent aspects of our individual empirical psychology. It is hard to see how such talk can carry the normative-­intersubjective weight required for the deduction of taste.

Freedom of Imagination and the “ Autonomy of Taste ”   361 In a reflective (as opposed to determining) judgment, such fitting accord requires that these faculties mutually stimulate their respective activities. An object which enables such stimulating accord is subjectively purposive for the power of judgment. Here Kant invokes his oft-­repeated suggestion that judgments of taste are grounded in a play of imagination and understanding which involves the general conditions of cognition in a merely subjectively purposive fashion. He adds that in this subjectively purposive play “no determinate concept of the object is being required nor generated,” so that the aesthetic judgment which is grounded in the free play of the faculties “is not itself a judgment of cognition” (EEKU, 20:221). The imaginative apprehension of a sensible manifold here agrees with the exhibition of a concept but leaves indeterminate which concept is at issue (unbestimmt welchen Begrifs). This claim may seem intractably paradoxical, but it may be illuminated by appeal to aesthetic ideas qua symbolic exhibitions of rational ideas. Rational ideas are not theoretically determinate concepts of objects; they cannot figure (directly) in theoretical cognition because their intellectual content transcends all sensibility. Accordingly, the inexhaustibly rich content of aesthetic ideas which (approximately) sensualizes rational ideas also resists subsumption under determinate concepts (KU, 5:343). Thus, Kant’s claim (C) that the aesthetic apprehension of sensible manifolds involves a subjectively purposive play of the faculties which exhibits no determinate concept might refer to the special kind of apprehension that the imagination achieves when it sensualizes (in a cognitively stimulating fashion) cognitively indeterminate ideas of reason via aesthetic ideas.38 Since (C) is central to Kant’s deduction of taste, this would mean that aesthetic ideas play an important role in this deduction. In another passage Kant stresses that aesthetic ideas provide a “rich undeveloped” material which the understanding applies subjectively for the animation of the powers of cognition and thus “indirectly also for cognitions” (KU, 5:317). The appeal to a “subjective” animation of the powers of cognition that has some “indirect” implications for cognitions echoes Kant’s appeal to the subjective conditions of cognition in the deduction. Hence, what he says here (at KU, 5:317) suggests that aesthetic experience satisfies the subjective conditions of cognition (at least partly) because aesthetic ideas positively relate to (i.e., provide stimulating input for) the cognitive power of understanding. This point secures a central relevance for aesthetic ideas in the deduction of taste. To conclusively develop this point, one would have to give a detailed account of the various ways in which aesthetic ideas might stimulate our (theoretical, practical) cognitive powers and thereby contribute “indirectly” to cognition.

38  Similarly, Rueger and Evren argue that the notion of “schematizing without a concept” refers to the act of symbolizing indeterminate ideas of reason (2005: 230, 241). However, I cannot follow their further point that natural beauty symbolically presents the systematicity of nature with respect to its empirical laws (2005: 232, 238–41).

362  Kant on Freedom and Rational Agency

X.4.  The Metaphysics of Imaginative Freedom In this section I examine whether Kant’s appeal to freedom of imagination accords with the metaphysical conditions that he imposes on genuinely free activity. As we saw, freedom generally requires independence from the mechanism of nature. A free subject can be influenced (“affected”) by natural conditions but this influence must not determine the subject’s activity (MS, 6:213). Determination by natural causes permits no more than a relative sense of freedom or spontaneity where the actions of a subject are (proximately) caused by the subject’s inner states but those states and the laws that govern them are ultimately the product of foreign causes. This is the mere “freedom of a turnspit” (KpV, 5:95). Here again, Kant’s conception of imaginative freedom, specifically his notion of genius, raises a problem. In the creation of art, “the nature in the subject” gives a rule to art (KU, 5:307). Through genius “nature . . . prescribe[s] the rule . . . to art” (KU, 5:308); genius can be considered “a favorite of nature” (KU, 5:318). For some, the natural origin of art is a welcome sign that nature is suited to our moral intentions.39 But Kant’s anti-­naturalistic doctrine of freedom cannot tolerate the claim that the free creation of aesthetic ideas is due to natural causes. One might suggest that Kant’s negative characterization of imaginative freedom requires only the absence of determination by laws of associative reproduction (KU, 5:240, 314). This would leave room for natural determination in other, non-­associative ways. However, any natural cause as such must operate according to some empirical law and must be determined to its causality by some preceding cause (cf. Chapter I). So, if the creative imagination really operates as “the nature in the subject,” this nature must be a law-­governed mechanism that is causally necessitated by foreign causes. In that case the aesthetic imagination is a mere turnspit lacking proper self-­activity. Whether or not the determining mechanism is associative seems irrelevant as far as this implication is concerned. Moreover, permitting that artistic creation is determined by some natural causes would undermine Kant’s point that there can be no scientific description of how genius gives the rule to art (KU, 5:308): he would have to concede that such a description is possible in principle even if we currently lack the requisite psychological or neurological insight.40

39  See Guyer 2007b: 290–1; Rogerson 2008: 50. Both suggest that if art qua product of nature symbolizes moral ideas, this reveals that nature is amenable to our moral purposes. But Kant does not argue in this way: he denies the claim that “artistic beauty is a special case of natural beauty” (Rogerson 2008: 97). For Kant natural beauty is morally privileged over artistic beauty because only natural beauty can reveal that nature is morally purposive for us apart from our intentional design (KU, 5:299–300). If artistic beauty were likewise (ultimately) a product of natural causes, this argument would be fallacious since natural and artistic beauty would be morally on par. 40  For a similar point, see Heidemann 1968: 178.

Freedom of Imagination and the “ Autonomy of Taste ”   363 One might suggest that we can avoid these implications if we take “nature” in the broader sense that Kant envisages in the third Critique, namely, as involving teleological rather than mechanistic principles. However, the idea of teleological causality entails an objective purposiveness that is incompatible with Kant’s aesthetic anti-­perfectionism. If the “nature in the subject” that gives a rule to art were a teleological natural cause, then the creative activity of imagination would be directed towards an end that can be assessed according to an objective standard of perfection. In that case the free play of the aesthetic imagination would not exhibit a merely subjective purposiveness (KU, 5:350). I propose that we can solve this problem by showing that when Kant refers to “the nature in the subject” as the origin of rules for art, he means something unlike nature in the standard sense that is determined by mechanistic or teleological principles. That Kant might have a non-­literal, merely analogical concept of nature in mind here is suggested by his remark that the aesthetic imagination creates “as it were” “another nature” from the sensible manifold that the “real” nature gives us through sense-­perception (KU, 5:314). But the most direct textual support for my proposal stems from his first “remark” on his solution to the antinomy of taste. This antinomy concerns a seeming clash between the claim that a judgment of taste cannot be grounded in concepts (because otherwise disagreements about matters of taste could be settled via determinate proofs) and the claim that a judgment of taste must be grounded in concepts (because otherwise the judgment could not lay claim to intersubjective-­universal validity) (KU, 5:338–9). Kant argues that we can preserve both claims if judgments of taste are grounded in an  indeterminate concept. In his view, the only concept that fits this bill is the ­ra­tional idea of the supersensible substratum underlying sensible appearances (KU, 5:339–40). In his “remark” (KU, 5:343–4) Kant elaborates on this solution. He says that both rational and aesthetic ideas must have their “principles . . . in reason”: “the former [ideas of reason] in the objective and the latter [aesthetic ideas] in the subjective principles of its [reason’s] use.” He further claims that “as a result of this, one can also explain genius in terms of the faculty of aesthetic ideas.” This supposedly indicates “the reason why in products of genius nature (that of the subject), not a deliberate end, gives the rule to art (the production of the beautiful).” The rule that genius gives to art does not derive from a determinate “precept” but from “that which is merely nature in the subject, i.e., the supersensible substratum of all our faculties (to which no concept of the understanding attains).” Since Kant here equates “nature in the subject” with “supersensible substratum of all our faculties,” we can read his claim that “the nature in the subject” gives a rule to art as employing the term “nature” in a non-­standard sense that does not imply the causal determinism which reigns in the sensible world (i.e., in nature taken in a literal, phenomenal sense).

364  Kant on Freedom and Rational Agency But this solution to the problem of how nature can give rules to free artistic creation raises further puzzles. Why does Kant attribute aesthetic ideas to the “nature” in the artist? Why does he identify this “nature” with the supersensible substratum of our faculties? More generally, what does Kant’s extremely difficult, terse argument in the first remark amount to? I can here only suggest a partial reconstruction of Kant’s challenging view, as it relates to imaginative freedom. One puzzling aspect of Kant’s argument in the first remark is his explanation of genius: because (“as a result of ” the fact that) aesthetic ideas have subjective principles in reason, one can explain genius as a faculty of aesthetic ideas. To understand this explanation, we must remember a fundamental desideratum of Kant’s aesthetics: he seeks to make room for the notion that judgments of (here, specifically artistic) taste have a “subjective standard” (KU, 5:344) of correctness. This standard must satisfy two criteria. On the one hand, it must ground a claim to universal validity (“a rightful claim to please everyone”; KU, 5:344). On the other hand, it must not enable objective proofs of validity. Since art is the product of genius, Kant needs an account of genius (more generally, of the artistic imagination) that explains how its products can be judged according to a standard of correctness that meets both of these criteria. He gives an account where the aesthetic ideas which arise from genius and which provide the content, meaning, and value of artworks are rooted “in the subjective principles of its [reason’s] use.” Since aesthetic ideas are based on principles that belong to our shared rational faculties, judgments about art can lay claim to universal validity (this satisfies the first criterion). Since these principles are (unlike, say, moral principles of reason that objectively determine how we must act) subjective and indeterminate, the truth or falsity of judgments about art cannot be established via objective proofs (this satisfies the second criterion). I propose that Kant’s attempt to combine the subjectivity and the universal validity of aesthetic judgments leads him to claim that “the nature in the subject” qua “supersensible substratum of all our faculties” gives rules to art. Here we need to understand two points. First, why does Kant call the supersensible rule-­giving component in artistic creation “nature in the subject”? Second, why does he insist that this “nature” involves our supersensible faculties rather than our sensible “nature” in the literal, empirical sense of that term? Regarding the first point, I suggest that when Kant attributes the origin of artistic rules to nature, he is playing on the fact that the term “nature” connotes fortuitous, intangible forces that are not under the control of consciously deliberative thought processes. His reason for tracing the rule for art to “nature” is that this rule is not a “precept” that one could devise via deliberate planning: “in products of genius nature (that of the subject), not a deliberate end, gives the rule to art” (KU, 5:344). An artist cannot intentionally devise this rule because she “does not know how the ideas” for her product arise in her mind (KU, 5:307–8). The supersensible substratum of our faculties that provides the rule for art can be

Freedom of Imagination and the “ Autonomy of Taste ”   365 called “nature” because it gives this rule in a manner that is analogous to the efficient causality of real natural forces: such forces are often unpredictable; they (or their effects) strike us fortuitous; and they are not under our deliberate control.41 If the production of beautiful art were a fully deliberate process where the artist has discursive knowledge of how to proceed at every step, then artistic creation would be based on determinate, lawful rules. Those same rules would then also enable a conceptually determinate assessment of art which would consist in discerning whether the artist has correctly applied the objective rules that are necessary and sufficient for creating a beautiful artwork. But this is incompatible with the merely subjective validity of judgments about beautiful art (KU, 5:307). Thus, Kant’s desideratum that aesthetic judgment does not admit of objective proof leads him to claim that (original) art is not (solely) the result of conscious, rule-­governed deliberative planning: it is (partly) due to a “nature” that provides the artist with a fortuitous, non-­deliberate moment of inspiration. Conscious reflection on the concepts and ideas governing free artistic production is insufficient to discern how one must combine sensible manifolds under these rules to create a spirited, beautiful artwork. The fact that one cannot determine via ra­ tional deliberation how to apply these rules in concrete instances of aesthetic creation explains why they cannot be transformed from loose, indeterminate guidelines into objective laws of taste (cf. Section X.2).42 But why does Kant identify the inspirational “nature in the subject” with “the supersensible substratum of all our faculties”? I suggest that for Kant aesthetic ideas must be traced to our supersensible free “nature” because he views this non-­ empirical origin as a necessary condition for the intersubjective validity of aesthetic ideas. This suggestion should not come as a surprise: in preceding chapters we have repeatedly encountered Kant’s view that our freedom from “the mech­an­ ism of subjectively determining” sensible-­natural causes is a necessary condition for the apriority, necessity, and universal validity of our representations—­in both moral and epistemic contexts. If Kant invokes the same consideration with regards to our aesthetic representations, he thereby forges a highly significant

41  Kant often assigns a non-­standard, analogous meaning to the term “nature.” One prominent example is his claim that people are evil “by nature”: this is meant to capture the idea that evil is a fundamental property of the human character as we know it through experience, which (just like a basic natural Grundkraft, e.g., repulsive material force) cannot be explained via some further underlying cause. But since this character is freely chosen, Kant expressly rules out that “nature” here refers to a natural disposition in the literal sense (RGV, 6:31–2). 42  The issue here is not merely that applying a rule is not governed by further rules. This is (on pain of an infinite regress) true for every rule, including fully objective concepts (A132–4/B171–3). The application of objective concepts is a task for the faculty of judgment which allows thinkers to ra­tion­ ally determine whether a particular is to be subsumed under the concept. By contrast, because the inspiration for artistic production stems from fortuitous forces of “nature,” no amount of teaching or attention to examples (KU, 5:308–9) can ever fully determine how one must apply the general discursive rules for aesthetic production when one creates a specific artwork.

366  Kant on Freedom and Rational Agency unifying link between aesthetic freedom of imagination, moral freedom of practical reason, and epistemic freedom of theoretical reason.43 To confirm that this consideration is indeed central to Kant’s aesthetics, let us examine the implications of the naturalistic claim that the inspiration for original aesthetic ideas derives from the sensible natural causes that make up the artist’s empirical psychology. Kant states these implications when he criticizes Burke’s psychological explanation of judgments of taste (EEKU, 20:238). His argument here is as follows. (1) Judgments that express an ought make a normative claim to necessity and universal validity; as such, they require some a priori principle as their foundation. (2) If a judgment can be explained as the result of phenomenal causes that belong to the nexus of empirical psychology, then the judgment lacks an a priori foundation and is merely contingent. (3) An empirically conditioned judgment (of the sort mentioned in (2)) cannot express an ought or make a legitimate claim to universality and necessity (cf. Refl., 18:182–3). (4) Aesthetic judgments do make a legitimate claim to universality and necessity; hence, they cannot be of the sort mentioned in (2): rather, must be of the sort mentioned in (1). This argument fruitfully applies to our present context as follows. The dispositions that constitute our empirical psychology are wholly contingent and, ra­tion­ ally speaking, arbitrary. If the imagination in its creation of aesthetic ideas were determined by such dispositions, aesthetic ideas would not have “principles . . . in reason,” i.e., in our shared rational faculties. Rather, the rule that artists give to their creation would be “subjective” in a pejorative-­ idiosyncratic sense (cf. Chapter  IV). Accordingly, whether or not recipients can appreciate an artwork would then also depend on the egocentric (“private”) psychological conditions that determine whether they happen to find the artwork’s sensible components pleasing or exciting: conditions that are no different from the psychological causes which determine whether one takes pleasure in Nutella-­dipped cucumber. As a result, aesthetic responses to art could not lay claim to the kind of subjective validity that is distinctive of aesthetic judgments: a type of “subjectivity” that is also intersubjective and hence can make a normative demand on the agreement of others. To account for the universal validity of responses to art, Kant argues that “the nature in the subject” which inspires artistic creation is not the artist’s arbitrary empirical psychology but rather “the intelligible” or “supersensible substratum of all our” shared cognitive faculties (KU, 5:344). Since this non-­natural substratum is free from empirical determination by contingent, egocentric psychological conditions, it can spontaneously supply an a priori necessary, rational 43  This explains why Kant insists that our aesthetic activity must be tied to the supersensible just like other (moral, epistemic) forms of activity: in all these cases, our activity generates representations that lay claim to (objective or subjective) universal-­intersubjective validity. This yields a partial response to Guyer’s claim (1997a: 294–306) that Kant’s appeal to the supersensible in the third Critique is merely a contrived attempt to connect his aesthetics with his idealism and to correct the flaws in his deduction of taste.

Freedom of Imagination and the “ Autonomy of Taste ”   367 foundation for artistic creation. This, in turn, provides art with an intersubjective value and meaning that all human subjects can appreciate regardless of their contingent empirical differences. Thus, judgments that assess the value of art can make a normative demand on the agreement of others. To summarize: because aesthetic ideas originate in our shared rational nature rather than in our contingent, idiosyncratic empirical psychology, the pleasure that these ideas induce can be universally “imputed” to everyone (KU, 5:344). Since the rational ground of aesthetic ideas is a supersensible aspect of ourselves that we cannot capture via determinate, objective cognitive concepts, we cannot understand or explain how or why these ideas (fail to) arise in our productive imagination. Hence, there can be no determinate rule for genius, one cannot learn how to invent spirited art (KU, 5:307–9), and (accordingly) one cannot assess artworks via objectively valid criteria.

X.5.  Freedom of Imagination in Aesthetic Production Revisited In this section I clarify how Kant’s appeal to the fortuitous supersensible origin of aesthetic ideas fits into the account of rule-­governed artistic creativity that I sketched in Section X.2. According to my argument in Section X.2, the imagination is governed in its free artistic creation by ideas and concepts that function here as indeterminate rules. These rules give positive guidance to aesthetic production and thereby prevent the lawless freedom of imagination from degenerating into a random liberty of indifference. But this does not sufficiently account for the spirited creation of aesthetic value, because a successful exhibition of the rules governing artistic production might yield only a boring, trite, eclectic (etc.) product. My argument in Section X.4 shows what further condition is required for the creation of beautiful art: the “nature in the subject” qua “supersensible substrate of all its . . . faculties of cognition” (KU, 5:344) must provide the artist’s imagination with a fortuitous inspiration that enables the free creation of a beautiful, original, spirited artwork. Thus, the free creation of beautiful art has two central aspects: first (i), a deliberate, intentional, rule-­governed aspect; second, (ii) a non-­deliberate fortuitous aspect (the moment of inspiration). As we saw in Section  X.2, Kant imposes a further constraint upon artistic creation, namely (iii) that it conforms to academic laws of taste. However, since these laws “permit damage to” the free exercise of imagination, they do not govern artistic creation insofar as it is free. One might wonder why an account of free aesthetic production must include the deliberative-­conscious aspect (i). Could such an account not focus exclusively on aspect (ii), on the inspiration that comes from the artist’s supersensible “nature” and that enables artists to invent aesthetic ideas which elicit a universally

368  Kant on Freedom and Rational Agency valid pleasure? There are two reasons why (i) must figure in a viable conception of how our imagination freely creates spirited artistic beauty. First, since (ii) concerns a moment of inspiration that is not even implicitly deliberate, receiving such inspiration is not directed by any (however indirect or implicit) consciousness of rules. Thus, an exclusive focus on (ii) ignores the idea that artistic creation is a rule-­governed conscious activity. But this is tantamount to abandoning Kant’s generic concept of freedom (cf. Section  X.1). Likewise, artistic freedom cannot be conceived as a free purposive play between the imagination and our higher intellectual faculties (KU, 5:240–1) unless this play consists in a conscious, rule-­governed activity. While Kant denies that deliberate activity is sufficient for creating original art, he insists that free artistic production requires a conscious effort through which the artist “dares to . . . sensualize ra­ tional ideas of invisible beings” (KU, 5:314). It is only by making a suitable impact on this “daring” intentional effort that fortuitous inspiration can meaningfully contribute to artistic production. The non-­deliberate inspiration for inventing aesthetic ideas and (thereby) giving a rule to art can occur only in relation to a deliberate plan to form such a rule, i.e., to create an aesthetically pleasant symbolization of specific rational themes. The artist’s attempts to devise and realize this plan must involve her conscious imagination of ways in which the specific themes that she has deliberately selected might be sensibly presented. As a free and conscious activity of imagination, this process must be positively directed by the artist’s awareness of norms. This norm-­governed awareness cannot derive from (ii) an entirely non-­deliberate fortuitous moment of inspiration. It also cannot derive from (iii) academic canons of taste that govern only the “mechanical” component of artistic production and thereby hamper “the freedom and richness of the imagination” (KU, 5:312). Hence, to make room for norms that positively guide (rather than restrict) the free deliberate use of the imagination, we must posit rules that indeterminately, un-­lawfully direct the artist’s conscious imaginative efforts to sensualize specific rational ideas in a beautiful manner. (This point also applies, mutatis mutandis, to free aesthetic experience.) Second, the appeal to (i) a norm-­governed deliberative aspect of imaginative freedom is needed to secure another important dimension of human freedom: as imperfect agents we are under the influence of conditions that induce us to make avoidable mistakes which call for reproach or blame (cf. Chapters  III,  V). Components (ii) and (iii) do not account for this dimension of our free aesthetic activity. While the claim that an artist lacked (ii) inspiration implies that her creation is aesthetically unremarkable or without spirit (geistlos), these aesthetic verdicts do not concern a fault in the creative process that was under the artist’s deliberate control and that she could have avoided. Arguably, some aesthetic criticisms do imply a (to be sure, non-­moral) reproach or blame and thereby betray our conviction that the artist was free to do better, but without invoking (iii) objective canons of taste. Consider, for example, the verdict that for all the genius

Freedom of Imagination and the “ Autonomy of Taste ”   369 we can witness in the Sopranos, David Chase and his writers should have seen that the plot line where Tony and Adriana verged on having an affair is rather clumsy and contrived. This criticism requires an appeal to rules governing creative activity that satisfy two conditions. First, these rules must be consciously accessible to artists as action-­guiding oughts, so that we can we truly impute to artists the aesthetic flaws which result from their failure to comply with these rules. Second, these rules cannot have lawlike objectivity since that would preclude the characteristically aesthetic debate between those who consider the abovementioned plot line aesthetically flawed and those (including presumably Chase himself) who believe that it is effective. The fact that Kant’s account employs both (i) and (ii) raises a further question. The rules that (indeterminately) govern deliberate artistic creation include transcendental ideas of reason. In addition, “Subjective principles of the use of ­reason” (KU, 5:344) provide non-­deliberate, fortuitous artistic inspiration. What is the relation between these subjective principles of the use of reason (in (ii)) and transcendental ideas of reason (in (i))? One might suggest that the subjective principles of the use of reason just are transcendental ideas of reason. After all, Kant often stresses that reason’s transcendental ideas are subjectively necessary for the use of our rational faculties (cf. A648/B678; KpV, 5:4; KU, 5:450). However, at KU, 5:344 Kant distinguishes between the “subjective principles of the use of reason” that yield fortuitous inspiration for aesthetic ideas and objective ideas of reason. As we saw (in Section IV.2), Kant’s “objective-­subjective” terminology is ambiguous. There are two ways in which transcendental ideas of reason can be regarded as objective. First, ideas of practical reason yield objective (moral) laws. Second, even the transcendental ideas of theoretical reason are “subjective” only in comparison to the constitutive categories of understanding. They may be regarded as “objective” when compared to aesthetic ideas because unlike the latter they make a regulative contribution to the objective cognition of nature. At KU, 5:344 Kant argues that because both moral and theoretical ideas of reason have objective import, these ideas must be distinguished from the merely “subjective principles of the use of reason” that non-­deliberately inspire the creation of aesthetic ideas (in (ii)). I suggest that when Kant traces aesthetic ideas to “subjective principles of the use of reason,” he is not referring to the faculty of reason in the narrow sense as the seat of (“objective”) transcendental ideas. Rather, he designates the faculty of reason in a broad sense as involving all our higher faculties of cognition (cf. A835/B863).44 This aligns with his further claim that aesthetic ideas derive from a “supersensible substrate . . . in relation to which all our cognitive faculties agree.”

44  I owe this suggestion to Ulrich Schlösser.

370  Kant on Freedom and Rational Agency Aesthetic ideas derive not directly from “reason” in the narrow sense (qua seat of specific, comparatively “objective” principles) but from a supersensible common ground of all our shared cognitive faculties. This enables the distinctive way in which these faculties are exercised in aesthetic activity. Because aesthetic ideas arise from a common root of our shared cognitive capacities, they can elicit a pleasantly harmonious, universally valid play among those faculties. There is, however, a tight connection between these two different aspects of “reason.” Aesthetic ideas elicit a pleasant harmony among our cognitive faculties when they fittingly approximate a sensible presentation of morally objective ideas of reason. Thus, the “subjective” principles of reason (in the broad sense) that inspire the invention of aesthetic ideas must be directed towards the presentation of objective moral ideas of reason (in the narrow sense). Accordingly, Kant refers to “the ultimate end given by the intelligible in our nature to make all our cognitive faculties agree, which is to serve as the subjective standard” for aesthetic appreciation (KU, 5:344). This “ultimate end” is our highest moral vocation for the use of all our faculties (cf. A840/B868). Perhaps Kant here seeks to make roughly the following (admittedly, vague) suggestion: because our ultimate moral end derives from a supersensible aspect of our rational selves (GMS, 4:457–8; KpV, 5:86–7), the supersensible substratum underlying all our shared cognitive faculties is uniquely suited to provide artists with inspiration for symbolizing moral ideas in a beautiful form which enables a pleasant harmony among these faculties.

X.6.  Conclusion: Freedom of Imagination and the Unity of Kant’s Doctrine I have given an account of the freedom that our imagination exhibits in our production of art and in our experience of artistic and natural beauty. The question is whether this account truly solves the puzzle I presented at the beginning. Can we regard freedom of imagination as a genuine type of Kantian freedom in a way that preserves the unity of Kant’s doctrine? My argument in Section X.4 shows that freedom of imagination satisfies the negative metaphysical definiens of Kant’s transcendental concept of freedom: in its free activity the imagination is not determined by foreign natural causes. Moreover, Kant emphasizes this metaphysical aspect of imaginative freedom for the same systematically important reason that he also invokes in moral and ep­i­ ste­mic contexts: he argues that the freedom of our imagination from “the mech­ an­ism of subjectively determining” empirical causes is a necessary condition for the a priori universal validity of our aesthetic representations. This establishes a significant point of continuity between freedom of imagination and the two paradigmatic kinds of Kantian freedom, moral freedom of will and epistemic freedom of thought.

Freedom of Imagination and the “ Autonomy of Taste ”   371 Establishing such continuity is more challenging with regards to Kant’s positive definiens of transcendental freedom. This definiens includes two conditions: first, free agency must be governed by rational standards of correctness; second, these standards must be autonomous norms that arise from rational self-­legislation. In his account of moral freedom of will and epistemic freedom of thought, Kant further identifies these norms with determinate, objective intellectual laws. Since my account in this chapter concedes that the aesthetic imagination satisfies these conditions only in a somewhat looser sense, it may seem that my account provides merely an analogy between freedom of imagination and the genuine forms of (moral and epistemic) freedom which strictly exemplify the positive definiens. In that case, we could speak of imaginative “freedom” only in an analogical or metaphorical sense that has no proper place in Kant’s doctrine of truly (literally speaking) free agency. But instead of demoting the powers of imagination as being merely analogous to genuine moral or epistemic freedom, we should rather say that the imagination genuinely exemplifies Kant’s conception of transcendental freedom in a highly distinctive manner that accords with the peculiar character of aesthetic as opposed to moral or epistemic activity. Consider here the merits of my key proposal (call it (P)) that the free imagination is loosely governed by indeterminate rational norms. On the one hand, (P) defuses the worry that the free imagination stands under no rules whatsoever and must thus be viewed as an arbitrary liberty of indifference whose products lack universal-­intersubjective validity. This worry is clearly the central motive for Kant’s positive definition of freedom as a rule-­ governed power. On the other hand, (P) also prevents an objectivist, rationalistic over-­intellectualization of imaginative freedom: it avoids the implausible assimilation of free aesthetic production and experience to other (moral, epistemic) types of free agency which have a squarely cognitivist-­intellectual character that requires rigid, determinate governance by objective rational laws. Since (P) accounts for both the universal validity and the irreducible subjectivity of our aesthetic representations, it yields a philosophical analysis of taste which (unlike purely subjectivizing or overly intellectualizing aesthetics) remains true to its authentic character. Admittedly, there is no way of getting around the fact that the imagination lacks autonomy in the precise sense that goes into Kant’s positive definiens. This is because the free imagination does not give its governing rules to itself but takes them from our higher intellectual faculties. For some, this might be enough to show that we cannot speak of genuine freedom here. However, we might also say that the free imagination has its own peculiar, sui generis form of autonomy, “the autonomy of taste” (KU, 5:282). Imaginative activity (in aesthetic production and experience) can be understood as a special form of autonomy for two reasons. First, the aesthetic imagination positively relates itself to, i.e., invites guidance and direction from norms of understanding and reason that are autonomous in the

372  Kant on Freedom and Rational Agency strict or primary sense. Although these rules do not arise from the imagination itself, they are still not “foreign” to the imagination since they originate in the free unified human mind (Gemüt) that includes the imagination as an integral part. Against this, one might insist that the rules governing the aesthetic imagination are detached from, hence foreign to, its own mode of operation. This is because, one might hold, the aesthetic value created by free imaginative activity is independent from the rules that direct such activity: the greatest expressions of moral or epistemic value result from our strict conformity to the autonomous laws ­governing our free will or understanding, whereas original beauty arises only when our free imagination does not fully conform to the rules that govern it. However, if the imagination creates or appreciates something beautiful, then its activity does comply with the norms governing such creation or appreciation. To be sure, this sense of compliance is not the same as in moral or epistemic cases, but that is just because the corresponding type of rule-­governedness is different: if the aesthetic imagination operates successfully, it satisfies the norms which direct it in the one and only, indeterminate sense that loose (as opposed to strict law-­ like) norm-­governedness requires (or enables). For instance (cf. Section X.3), the imagination successfully complies with reason’s vague directive to obtain a maximum of representational content if it fittingly combines and magnifies given sensible manifolds in a non-­mundane, aesthetically pleasing manner. Moreover, the aesthetic value arising from free imaginative activity is essentially tied to the rules that govern such activity: since all beauty involves a “sensualizing of moral ideas” (KU, 5:356), and since (as we saw) moral ideas play a key role in guiding aesthetic creation and appreciation, the moral value contained in these ideas is integrally connected (though not identical) to the aesthetic value of beauty. Hence, the intellectual norms governing the free imagination are not foreign to the distinctive value it seeks to create via its aesthetic mode of operation. The second reason why the free imagination can be considered autonomous is that it allows rational directives to guide its activity while at the same time refusing to be dominated by them. Free imaginative activity develops the input it receives from intellectual rules in a way that is geared towards its non-­moral, non-­epistemic, specifically aesthetic telos. It thereby exemplifies the notion of self-­determination that is central to Kant’s notion of autonomy. The free imagination determines itself by taking a cue from objective intellectual norms but then interpreting that cue in its own characteristic, subjectively playful manner. The special way in which the aesthetic imagination takes up discursive intellectual content is reflected in the special character of its products: unlike other (moral, epistemic) species of spontaneous activity, the imagination spontaneously determines itself to a symbolic sensible (rather than purely intellectual-­conceptual) (re)presentation of the supersensible. This special kind of self-­determination

Freedom of Imagination and the “ Autonomy of Taste ”   373 allows the imagination to make supersensible rational content amenable to the emotive, sensible side of human nature: to our faculty of feeling.45 If this account (or something akin to it) is on the right track, we can conclude that freedom of imagination qualifies as a genuine species of transcendental freedom: the imagination exemplifies the only type of spontaneous self-­determination that fits the aesthetic dimension of human life. My overall interpretation in this book thus establishes the (generic) idea of transcendental freedom as the cardinal point of all three major areas that constitute Kant’s systematic philosophy. Accordingly, Kant can view such freedom as the proper anchor of all meaningful human activity, in moral, epistemic, and aesthetic contexts. 45  Because the aesthetic imagination acts in a conscious manner which takes cues from intellectual norms, it is intellectually instructed (albeit not determined) rather than “blind.” Accordingly, it also qualifies as a free faculty, unlike the unfree lower faculties of sensibility whose exercise “cannot be instructed” by norms “because they are blind” (V-­MO/Col, 27:244).

Bibliography Abaci, Uygar (2019). Kant’s Revolutionary Theory of Modality. Oxford: Oxford University Press, 2019. Abela, Paul (2002). Kant’s Empirical Realism. Oxford: Oxford University Press, 2002. Adams, Robert (1994). Leibniz: Determinist, Theist, Idealist. Oxford: Oxford University Press, 1994. Adams, Robert (1997). “ ‘Things in Themselves.’ ” Philosophy and Phenomenological Research 57 (4), 801–25, 1997. Adickes, Erich (1924). Kant und das Ding an sich. Kassel: Pan Verlag, 1924. Allais, Lucy (2010). “Transcendental Idealism and Metaphysics: Kant’s Commitment to Things as They Are in Themselves.” D. Heidemann (ed.), Kant Yearbook 2010: Metaphysics, 1–33. Berlin: De Gruyter, 2010. Allais, Lucy (2015). Manifest Reality. Kant’s Idealism and His Realism. Oxford: Oxford University, 2015. Allison, Henry (1984). Kant’s Transcendental Idealism. New Haven: Yale University Press, 1984. Allison, Henry (1990). Kant’s Theory of Freedom. Cambridge: Cambridge University Press, 1990. Allison, Henry (1996). Idealism and Freedom. Cambridge: Cambridge University Press, 1996. Allison, Henry (2001). Kant’s Theory of Taste. Cambridge: Cambridge University Press, 2001. Allison, Henry (2003). “Reply to the Comments of Longuenesse and Ginsborg.” Inquiry 46, 182–94, 2003. Allison, Henry (2004). Kant’s Transcendental Idealism. 2nd edition. New Haven: Yale University Press, 2004. Allison, Henry (2011). Kant’s Groundwork for the Metaphysics of Morals: A Commentary. Oxford: Oxford University Press, 2011. Allison, Henry (2013). “Kant’s Practical Justification of Freedom.” S. Baiasu, M. Timmons (eds.), Kant on Practical Justification, 284–99. Oxford: Oxford University Press, 2013. Allison, Henry (2020). Kant’s Conception of Freedom. A Developmental and Critical Analysis. Oxford: Oxford University Press, 2020. Ameriks, Karl (2000a). Kant’s Theory of Mind. Oxford: Oxford University Press, 2000. Ameriks, Karl (2000b). Kant and the Fate of Autonomy. Oxford: Oxford University Press, 2000. Ameriks, Karl (2002). “ ‘Pure Reason of Itself Alone Suffices to Determine the Will’ (42–57).” O.  Höffe (ed.), Klassiker Auslegen. Immanuel Kant: Kritik der praktischen Vernunft, 87–100. Berlin: Akademie Verlag, 2011. Ameriks, Karl (2003). Interpreting Kant’s Critiques. Oxford: Oxford University Press, 2003. Ameriks, Karl (2013). “Is Practical Justification in Kant Ultimately Dogmatic?” S. Baiasu, M. Timmons (eds.), Kant on Practical Justification, 153–75. Oxford: Oxford University Press, 2013. Ameriks, Karl (2019). Kantian Subjects: Critical Philosophy and Late Modernity. Oxford: Oxford University Press, 2019. Anderson, Lanier R. (2001). “Synthesis, Cognitive Normativity, and the Meaning of Kant’s Question, ‘How Are Synthetic Cognitions A Priori Possible’?” European Journal of Philosophy 9 (3), 275–305, 2001. Anscombe, G. E. M. (1958). “Modern Moral Philosophy.” Philosophy 33 (124), 1–19, 1958.

376 Bibliography Anscombe, G. E. M. (1993). “Causality and Determination.” M. Tooley and E. Sosa (eds.), Causation, 88–104. Oxford: Oxford University Press, 1993. Aquila, Richard (1979). “Things in Themselves and Appearances: Intentionality and Reality in Kant.” Archiv für Geschichte der Philosophie 61, 293–308, 1979. Aristotle (EE). Eudemian Ethics. In J. Barnes (ed.), The Complete Works of Aristotle. Volume 2: The Revised Oxford Translation. Princeton: Princeton University Press, 1984. Aristotle (EN). Nicomachean Ethics. In J.  Barnes (ed.), The Complete Works of Aristotle. Volume 2: The Revised Oxford Translation. Princeton: Princeton University Press, 1984. Armstrong, David (1983). What Is a Law of Nature? Cambridge: Cambridge University Press, 1983. Ayer, A. J. (1963). The Concept of a Person and Other Essays. New York: Macmillan, 1963. Baiasu, S. (2013a). “Introduction: Practical Justification in Kant.” S. Baiasu, M. Timmons (eds.) Kant on Practical Justification, 1–21. Oxford: Oxford University Press, 2013. Baiasu, Sorin (2013b). “Kant’s Rechtfertigung and the Epistemic Character of Practical Justification.” In S.  Baiasu, M.  Timmons (eds.), Kant on Practical Justification, 22–41. Oxford: Oxford University Press, 2013. Banham, Gary (2006). Kant’s Transcendental Imagination. London: Palgrave Macmillan, 2006. Baron, Marcia (1999). Kantian Ethics Almost without Apology. Ithaca: Cornell University Press, 1999. Baum, Manfred (1986). Deduktion und Beweis in Kants Transzendentalphilosophie. Hain: Königstein/Ts, 1986. Baum, Manfred (2014). “Sittengesetz und Freiheit. Kant 1785 und 1788.” H.  Puls (ed.), Kant’s Rechtfertigung des Sittengesetzes in Grundlegung III: Deduktion oder Faktum?, 209–26. Berlin: De Gruyter, 2014. Baumgartner, Michael (2008). “Regularity Theories Reassessed.” Philosophia 34, 327–54, 2008. Beck, Lewis White (1960). A Commentary on Kant’s “Critique of Practical Reason.” Chicago: University of Chicago Press, 1960. Beck, Lewis White (1987). “Five Concepts of Freedom in Kant.” J.  T.  J.  Srzednick (ed.), Philosophical Analysis and Reconstruction: A Festschrift to Stephan Körner, 35–51. Leiden: Martinus Nijhoff, 1987. Beebee, Helen (2000). “The Non-Governing Conception of Laws of Nature.” Philosophy and Phenomenological Research 61 (3), 571–94, 2000. Beebee, Helen (2003). “Local Miracle Compatibilism.” Nous 37 (2), 258–77, 2003. Bennett, Jonathan (1974). Kant’s Dialectic. Cambridge: Cambridge University Press, 1974. Bennett, Jonathan (1984). “Commentary: Kant’s Theory of Freedom.” A. Wood (ed.), Self and Nature in Kant’s Philosophy, 102–12. Ithaca: Cornell University Press, 1984. Bird, Graham (1962). Kant’s Theory of Knowledge. London: Routledge & Kegan Paul, 1962. Bird, Graham (2006). The Revolutionary Kant. A Commentary on the Critique of Pure Reason. Chicago: Open Court, 2006. Bittner, Rüdiger (1983). Moralisches Gebot oder Autonomie. Freiburg: Karl Alper Verlag, 1983. Bojanowski, Jochen (2006). Kant’s Theorie der Freiheit. Berlin: De Gruyter, 2006. Bojanowski, Jochen (2015). “Die Deduktion des kategorischen Imperativs.” D. Schönecker (ed.), Kants Begründung von Freiheit und Moral in Grundlegung III. Neue Interpretationen, 83–108. Paderborn: Mentis Verlag, 2015. Bok, Hilary (1998). Freedom and Responsibility. Princeton: Princeton University Press, 1998. Boyle, Matthew (2009). “Active Belief.” Canadian Journal of Philosophy Supplementary Volume 35(S1), 119–47, 2009. Boyle, Matthew (2011). “ ‘Making up Your Mind’ and the Activity of Reason.” Philosophers’ Imprint 11 (17), 1–24, 2011.

Bibliography  377 Brandt, Reinhard (1988). “Der Zirkel im dritten Abschnitt von Kants Grundlegung zur Metaphysik der Sitten.” H.  Oberer, G.  Seel (eds.), Kant: Analysen-Probleme-Kritik, 169–91. Würzburg: Königshausen und Neumann Verlag, 1988. Brewer, Bill (1995). “Mental Causation: Compulsion by Reason.” Proceedings of the Aristotelian Society Supplementary 69, 237–53, 1995. Brink, David (1989). Moral Realism and the Foundations of Ethics. New York: Cambridge University Press, 1989. Brook, Andrew (1994). Kant and the Mind. Cambridge: Cambridge University Press, 1994. Brueckner, Anthony L. (1984). “Transcendental Arguments II.” Nous 18 (2), 197–225, 1984. Buroker, Jill Vance (2017). “Kant on Judging and the Will.” A.  Gomes, A.  Stephenson (eds.), Kant and the Philosophy of Mind: Perception, Reason, and the Self, 189–203. Oxford: Oxford University Press, 2017. Burge, Tyler (1998). “Reason and the First Person.” C.  Wright, B.  Smith, C.  McDonald (eds.), Knowing Our Own Minds, 243–70. Oxford: Oxford University Press, 1998. Carnois, Bernard (1986). The Coherence of Kant’s Doctrine of Freedom. Chicago: Chicago University Press, 1986. Chignell, Andrew (2007a). “Belief in Kant.” Philosophical Review 116, 323–60, 2007. Chignell, Andrew (2007b). “Kant’s Concepts of Justification.” Nous 41 (1), 33–63, 2007. Chignell, Andrew (2007c). “Kant on the Normativity of Taste: The Role of Aesthetic Ideas.” Australasian Journal of Philosophy 85, 415–33, 2007. Chignell, Andrew (2012). “Kant, Real Possibility and the Threat of Spinoza.” Mind 121 (483), 635–75, 2012. Chignell, Andrew (2014). “Modal Motivations for Noumenal Ignorance: Knowledge, Cognition, and Coherence.” Kant-Studien 105 (4), 573–97, 2014. Chisholm, Roderick M. (2003). “Human Freedom and the Self.” G. Watson (ed.), Free Will, 26–37. Oxford: Oxford University Press, 2003. Churchland, Patricia (1987). “Epistemology in the Age of Neuroscience.” Journal of Philosophy 84, 546–83, 1987. Collins, Arthur (1999). Possible Experience: Understanding Kant’s Critique of Pure Reason. Berkeley: University of California Press, 1999. Crawford, Donald W. (1982). “Kant’s Theory of Creative Imagination.” T. Cohen, P. Guyer (eds.), Essays in Kant’s Aesthetics, 151–78. Chicago: University of Chicago Press, 1982. Darwall, Stephen (2006). The Second-Person Standpoint: Morality, Respect, and Accountability. Cambridge, MA: Harvard University Press, 2006. Darwin, Francis (1887). The Life and Letters of Charles Darwin, Including an Autobiographical Chapter. Volume 1. London: John Murray, 1887. De Sousa, Ronald (1984). “The Natural Shiftiness of Natural Kinds.” Canadian Journal of Philosophy 14 (4), 561–80, 1984. Della Rocca, Michael (2008). Spinoza. London: Routledge, 2008. Dennett, Daniel (1984). “I Could Not Have Done Otherwise—­So What?” Journal of Philosophy 81, 553–65, 1984. Dohrn, Daniel (2006). Konzinnität und Kohärenz. Naturschönheit und Natursystem in Kants Kritik der Urteilskraft. Berlin: De Gruyter, 2006. Dretske, Fred (1977). “Laws of Nature.” Philosophy of Science 44 (2), 248–68, 1977. Ducasse, C.  J. (1993). “On the Nature and the Observability of the Causal Relation.” M. Tooley, E. Sosa (eds.), Causation, 125–36. Oxford: Oxford University Press, 1993. Earman, John (2004). “Determinism: What We Have Learned and What We Still Don’t Know.” J.  Campbell, M.  O’Rourke, D.  Shier (eds.), Freedom and Determinism, 21–46. Cambridge, MA: MIT Press, 2004.

378 Bibliography Engstrom, Stephen (1994). “The Transcendental Deduction and Skepticism.” Journal of the History of Philosophy 32, 359–80, 1994. Engstrom, Stephen (2009). The Form of Practical Knowledge. Cambridge, MA: Harvard University Press, 2009. Enoch, David (2010). “The Epistemological Challenge to Metanormative Realism: How Best to Understand It, and How to Cope with It.” Philosophical Studies 148 (3), 413–38, 2010. Fales, Evan (1990). Causation and Universals. London: Routledge, 1990. Foot, Philippa (1972). “Morality as a System of Hypothetical Imperatives.” Philosophical Review 81 (3), 305–16, 1972. Forster, Michael (2010). Kant and Skepticism. Oxford: Oxford University Press, 2010. Frankfurt, Harry (1988). The Importance of What We Care About. Cambridge: Cambridge University Press, 1988. Frankfurt, Harry (2003). “Alternate Possibilities and Moral Responsibility.” G.  Watson (ed.), Free Will, 167–76. Oxford: Oxford University Press, 2003. Friedman, Michael (1992). “Causal Laws and the Foundations of Natural Sciences.” P.  Guyer (ed.), The Cambridge Companion to Kant, 161–99. Cambridge: Cambridge University Press, 1992. Friedman, Michael (1996). “Exorcising the Philosophical Tradition: Comments on John McDowell’s ‘Mind and World.’ ” Philosophical Review 105, 427–67, 1996. Friedman, Michael (1999). Reconsidering Logical Positivism. Cambridge: Cambridge University Press, 1999. Frierson, Patrick (2003). Freedom and Anthropology in Kant’s Moral Philosophy. Cambridge: Cambridge University Press, 2003. Frierson, Patrick (2011). “Two Standpoints and the Problem of Moral Anthropology.” B. Lipscomb, J. Krueger (eds.), Kant’s Moral Metaphysics, 83–110. Berlin: De Gruyter, 2011. Frisch, Mathias (2005). Inconsistency, Asymmetry, and Non-Locality: A Philosophical Examination of Classical Electrodynamics. Oxford: Oxford University Press, 2005. Freudiger, Jürg (1993). Kant’s Begründung der praktischen Philosophie. Systematische Stellung, Methode und Argumentationsstruktur der “Grundlegung der Metaphysik der Sitten.” Bern: Haupt Verlag, 1993. George, Rolf (1982). “Vorstellung and Erkenntnis in Kant.” M. S. Gram (ed.), Interpreting Kant, 77–115. Iowa City: University of Iowa Press, 1982. Gibbard, Allan (1990). Wise Choices, Apt Feelings. Cambridge, MA: Harvard University Press, 1990. Gibbons, Sarah L. (1994). Kant’s Theory of Imagination. Oxford: Oxford Clarendon Press, 1994. Ginsborg, Hannah (2015). The Normativity of Nature: Essays on Kant’s Critique of Judgment. Oxford: Oxford University Press, 2015. Goldman, Alvin (1967). “A Causal Theory of Knowing.” Journal of Philosophy 64, 357–72, 1967. Goodman, Nelson (1955). Fact, Fiction and Forecast. Cambridge, MA: Harvard University Press, 1955. Graham, Peter (2011). “Ought and Ability.” Philosophical Review 120, 337–82, 2011. Grenberg, Jeanine (2014). Kant’s Defense of Ordinary Moral Experience. Cambridge: Cambridge University Press, 2014. Grier, Michelle (2001). Kant’s Doctrine of Transcendental Illusion. Cambridge: Cambridge University Press, 2001. Grüne, Stefanie (2007). Blinde Anschauung. Frankfurt: Vittorio Klostermann, 2007.

Bibliography  379 Gunkel, Andreas (1989). Spontaneität und moralische Autonomie: Kants Philosophie der Freiheit. Bern: Haupt Verlag, 1989. Guyer, Paul (1977). “Formalism and the Theory of Expression in Kant’s Aesthetics.” KantStudien 68, 46–70, 1977. Guyer, Paul (1987) Kant and the Claims of Knowledge. Cambridge: Cambridge University Press, 1987. Guyer, Paul (1990). “Reason and Reflective Judgment: Kant on the Significance of Systematicity.” Nous 24 (1), 17–43, 1990. Guyer, Paul (1995). “The Possibility of the Categorical Imperative.” Philosophical Review 104 (3), 353–85, 1995. Guyer, Paul (1997a). Kant and the Claims of Taste. Second Edition. Cambridge: Cambridge University Press, 1997. Guyer, Paul (1997b). Kant and the Experience of Freedom. Cambridge: Cambridge University Press, 1997. Guyer, Paul (2003a). “Kant on the Systematicity of Nature: Two Puzzles.” History of Philosophy Quarterly 20 (3), 277–95, 2003. Guyer, Paul (2003b). “Kant on Common Sense and Skepticism.” Kantian Review 7, 1–37, 2003. Guyer, Paul (2005). Kant’s System of Nature and Freedom. Oxford: Oxford Clarendon Press, 2005. Guyer, Paul (2007). “Naturalistic and Transcendental Moments in Kant’s Moral Philosophy.” Inquiry 50: 444–64, 2007. Guyer, Paul (2010). “The Deduction of the Categories.” P.  Guyer (ed.), The Cambridge Companion to Kant’s Critique of Pure Reason, 118–50. Cambridge: Cambridge University Press, 2010. Hacking, Ian (1991). “A Tradition of Natural Kinds.” Philosophical Studies 61, 109–26, 1991. Haji, Ishtiyaque (2002). Deontic Morality and Control. Cambridge: Cambridge University Press, 2002. Haji, Ishtiyaque (2012). Reason’s Debt to Freedom. Oxford: Oxford University Press, 2012. Hampshire, Stuart (1971). Thought and Action. New York: Viking Press, 1971. Hanna, Robert (2005). “Kant and Nonconceptual Content.” European Journal of Philosophy, 13 (2), 247–90, 2005. Hanna, Robert (2009). “Freedom, Teleology, and Rational Causation.” Kant-Yearbook 1 (1), 99–142, 2009. Hatfield, Gary (1990). The Natural and the Normative: Theories of Spatial Perception from Kant to Helmholtz. Cambridge, MA: MIT Press, 1990. Hatfield, Gary (1992). “Empirical, Rational and Transcendental Psychology: Psychology as Science and as Philosophy.” P. Guyer (ed.), The Cambridge Companion to Kant, 200–27. Cambridge: Cambridge University Press, 1992. Hawthorne, John (2003). Knowledge and Lotteries. Oxford: Oxford University Press, 2003. Heal, Jane (1996). “Simulation, Theory and Content.” P.  Carruthers, P.  K.  Smith (eds.), Theories of Theories of Mind, 75–89. Cambridge: Cambridge University Press, 1996. Heidegger, Martin (1991). Kant und das Problem der Metaphysik. Gesamtausgabe I. Abteilung, Band 3. Frankfurt: Vittorio Klostermann, 1991. Heidemann, Ingeborg (1968). Der Begriff des Spieles. Berlin: De Gruyter, 1968. Heimsoeth, Heinz (1956). Studien zur Philosophie Immanuel Kants. Bonn: Bouvier Verlag, 1956. Henrich, Dieter (1954–5). “Das Prinzip der kantischen Ethik.” Philosophische Rundschau 2, 29–34, 1954–5.

380 Bibliography Henrich, Dieter (1973). “Der Begriff der sittlichen Einsicht und Kants Lehre vom Faktum der Vernunft.” G.  Prauss (ed.), Kant: Zur Deutung seiner Theorie vom Erkennen und Handeln, 77–115. Cologne: Kiepenhauer & Wietsch, 1973. Henrich, Dieter (1975). “Die Deduktion des Sittengesetzes.” A.  Schwan (ed.), Denken im  Schatten des Nihilismus, 55–112. Darmstadt: Wissenschaftliche Buchgesellschaft, 1975. Henrich, Dieter (1989). “Kant’s Notion of a Deduction and the Methodological Background of the First Critique.” E. Förster (ed.), Kant’s Transcendental Deductions, 29–46. Stanford: Stanford University Press. Henson, Richard  G. (1979). “What Kant Might Have Said: Moral Worth and the Overdetermination of Dutiful Action.” Philosophical Review 88, 39–54, 1979. Herman, Barbara (1996). The Practice of Moral Judgment. Cambridge, MA: Harvard University Press, 1996. Hieronymi, Pamela (2009). “Two Kinds of Agency.” L. O’Brien, M. Soteriou (eds.), Mental Actions, 138–62. Oxford: Oxford University Press, 2009. Hill, Thomas (1973). “The Hypothetical Imperative.” Philosophical Review 82, 429–50, 1973. Hill, Thomas (1992). Dignity and Practical Reason in Kant’s Moral Theory. Ithaca: Cornell University Press, 1992. Hill, Thomas (2002). Human Welfare and Moral Worth. Kantian Perspectives. Oxford: Oxford University Press, 2002. Hill, Thomas (2011). Virtue, Rules and Justice. Oxford: Oxford University Press, 2011. Hogan, Desmond (2009a). “Three Kinds of Rationalism and the Non-Spatiality of Things in Themselves.” Journal of the History of Philosophy 47, 355–8, 2009. Hogan, Desmond (2009b). “Noumenal Affection.” Philosophical Review 118 (4), 501–32, 2009. Hogan, Desmond (2010). “Kant’s Copernican Turn and the Rationalist Tradition.” P. Guyer (ed.), The Cambridge Companion to Kant’s Critique of Pure Reason, 21–40. Cambridge: Cambridge University Press, 2010. Hoppe, Hansgeorg (1983). Synthesis bei Kant: Das Problem der Verbindung von Vorstellungen und ihrer Gegenstandsbeziehungen in der “Kritik der reinen Vernunft.” Berlin: De Gruyter, 1983. Horn, Christoph (2015). “Das Bewusstsein, unter dem moralischen Gesetz zu stehen. Kants Freiheitsargument in GMS III.” D.  Schönecker (ed.), Kants Begründung von Freiheit und Moral in Grundlegung III, 137–56. Paderborn: Mentis Verlag, 2015. Hudson, Hud (1994). Kant’s Compatibilism. Ithaca: Cornell University Press, 1994. Hume, David (1993). An Enquiry Concerning Human Understanding. Indianapolis: Hackett Publishing, 1993. Hume, David (2009). A Treatise of Human Nature. Oxford: Oxford University Press, 2009. Indregard, Jonas (2018). “A Gradual Reformation: Empirical Character and Causal Powers in Kant.” Canadian Journal of Philosophy 48 (5), 662–83, 2018. Irwin, Terence (1984). “Morality and Personality: Kant and Green.” A. Wood (ed.), Self and Nature in Kant’s Philosophy, 31–56. Ithaca: Cornell University Press, 1984. Jauernig, Anja (2008). “Kant’s Critique of the Leibnizian Philosophy: Contra the Leibnizians, but Pro Leibniz.” D.  Garber, B.  Longuenesse (eds.), Kant and the Early Moderns, 41–63. Princeton: Princeton University Press, 2008. Jauernig, Anja (2021). The World According to Kant. Oxford: Oxford University Press, 2021. Johnson, Andrew (2005). “Kant’s Empirical Hedonism.” Pacific Philosophical Quarterly 86 (1), 50–63, 2005. Jolley, Nicholas (2005). Leibniz. London: Routledge, 2005. Joyce, Richard (2001). The Myth of Morality. Cambridge: Cambridge University Press, 2001.

Bibliography  381 Joyce, Richard (2006). The Evolution of Morality. Cambridge, MA: MIT Press, 2006. Kahane, Guy (2011). “Evolutionary Debunking Arguments.” Nous 45 (1), 103–25, 2011. Kain, Patrick (2010). “Practical Cognition.” B.  Bruxvoort, J.  Kruger (eds.), Kant’s Moral Metaphysics, 177–209. Berlin: De Gruyter, 2010. Kane, Robert (1996). The Significance of Free Will. Oxford: Oxford University Press, 1996. Kant, Immanuel: Religion within the Limits of Pure Reason. Translated by T. M. Greene and H. H. Hudson. New York: Harper and Brothers, 1960. Kant, Immanuel: First Introduction to the Critique of Judgment. Translated by W. Pluhar. Indianapolis: Hackett, 1989. Kant, Immanuel: Critique of Practical Reason. Translated by T.  K.  Abbott. Buffalo: Prometheus Books, 1996. Kant, Immanuel: Critique of Pure Reason. Translated by P. Guyer, A. Wood. Cambridge: Cambridge University Press, 1998. Kant, Immanuel: Review of Schulz’s Attempt at an Introduction to a Doctrine of Morals for All Human Beings Regardless of Different Religions. Translated by M.  Gregor. The Cambridge Edition of the Works of Immanuel Kant: Practical Philosophy. Cambridge: Cambridge University Press, 1999. Kant, Immanuel: Groundwork for the Metaphysics of Morals. Translated by T.  K.  Abbott. Peterborough: Broadview Press, 2003. Kant, Immanuel: Lectures on Logic. J.  Michael Young (ed.), Cambridge: Cambridge University Press, 2004. Kant, Immanuel: Metaphysical Foundations of Natural Science. Translated by M. Friedman. Cambridge: Cambridge University Press, 2004. Kant, Immanuel: Prolegomena to Any Future Metaphysics. Translated by G.  Hatfield. Cambridge: Cambridge University Press, 2004. Kant, Immanuel: Anthropology from a Pragmatic Point of View. Translated by R. B. Louden. Cambridge: Cambridge University Press, 2006. Kant, Immanuel: Critique of the Power of Judgment. Translated by P. Guyer, E. Matthews. Cambridge University Press. Keller, Pierre (1998). Kant and the Demands of Self-Consciousness. Cambridge: Cambridge University Press, 1998. Kim, Jaegwon (2003). “The American Origins of Philosophical Naturalism.” Journal of Philosophical Research, APA Centennial Volume, 83–98, 2003. Kitcher, Patricia (1984). “Kant’s Real Self.” A.  Wood (ed.), Self and Nature in Kant’s Philosophy, 113–47. Ithaca: Cornell University Press, 1984. Kitcher, Patricia (1990). Kant’s Transcendental Psychology. Oxford: Oxford University Press, 1990. Kitcher, Patricia (1991). “Kant’s Dedicated Cognitivism.” J.  C.  Smith (ed.), Historical Foundations of Cognitive Science, 189–210. Amsterdam: Kluwer Academic Publishing, 1991. Kitcher, Patricia (1999). “Kant’s Epistemological Problem and Its Coherent Solution.” Nous 33, 415–41, 1999. Kitcher, Patricia (2011). Kant’s Thinker. Oxford: Oxford University Press Kitcher, Philip (2006). “Biology and Ethics.” D. Copp (ed.), The Oxford Handbook of Ethics, 163–85. Oxford: Oxford University Press, 2006. Kleingeld, Pauline (2010). “Moral Consciousness and the ‘Fact of Reason.’ ” A.  Reath, J.  Timmermann (eds.), Kant’s Critique of Practical Reason: A Critical Guide, 55–72. Cambridge: Cambridge University Press, 2010. Klemme, Heiner  F. (2014). “Is the Categorical Imperative the Highest Principle of Both Pure Practical and Theoretical Reason?” Kantian Review 19 (1), 119–26, 2014.

382 Bibliography Kneller, Jane (2007). Kant and the Power of Imagination. Cambridge: Cambridge University Press, 2007. Kohl, Markus (2014). “Transcendental and Practical Freedom in the Critique of Pure Reason.” Kant-Studien 105 (3), 313–35, 2014. Kohl, Markus (2015a). “Kant on the Inapplicability of the Categories to Things in Themselves.” British Journal for the History of Philosophy 23 (1), 90–114, 2015. Kohl, Markus (2015b). “Kant on Determinism and the Categorical Imperative.” Ethics 125 (2), 331–56, 2015. Kohl, Markus (2015c). “Kant on Freedom of Empirical Thought.” Journal of the History of Philosophy 53 (2), 301–26, 2015. Kohl, Markus (2015d). “Kant and ‘Ought Implies Can.’ ” Philosophical Quarterly 65 (261), 690–710, 2015. Kohl, Markus (2016). “Kant on Idealism, Freedom, and Standpoints.” Archiv für Geschichte der Philosophie 98 (1), 21–54, 2016. Kohl, Markus (2017a). “Radical Evil as a Regulative Idea.” Journal of the History of Philosophy 55 (4), 641–73, 2017. Kohl, Markus (2017b). “The Normativity of Prudence.” Kant-Studien 108 (4), 517–42, 2017. Kohl, Markus (2018a). “Kant’s Standpoint Distinction.” Kantian Review 23 (2), 229–55, 2018. Kohl, Markus (2018b). “Kant’s Critique of Instrumental Reason.” Pacific Philosophical Quarterly 99 (3), 489–516, 2018. Kohl, Markus (2018c). “The Metaphysical Deduction and the Shadow of Humean Skepticism.” Kant-Studien 109 (3), 367–94, 2018. Kohl, Markus (2020a). “Kant on Cognizing Oneself as a Spontaneous Cognizer.” Canadian Journal of Philosophy 50 (3), 395–412, 2020. Kohl, Markus (2020b). “Spontaneity and Contingency: Kant’s Two Models of Rational SelfDetermination.” M.  Kissner, J.  Nöller (eds.), The Concept of Will in Classical German Philosophy, 29–48. Berlin: De Gruyter, 2020. Kohl, Markus (2020c). “Kant’s Conception of Freedom. A Developmental and Critical Analysis, Henry Allison.” European Journal of Philosophy 28 (3), 805–9, 2020. Kohl, Markus (2021a). “Kants subjektivistische Begründung von Moral und Freiheit im Naturrecht Feyerabend.” D. Hüning, S. Klingner, G. S. Bordoni (eds.). Auf dem Weg zur kritischen Rechtslehre? Naturrecht, Moralphilosophie und Eigentumstheorie in Kants “Naturrecht Feyerabend,” 150–71. Leiden: Brill, 2021. Kohl, Markus (2021b). “Kantian Subjects: Critical Philosophy and Late Modernity, Karl Ameriks.” Kantian Review 26 (2), 335–40, 2021. Kohl, Markus (2021c). “A Priori Intuition and Transcendental Necessity in Kant’s Idealism.” European Journal of Philosophy 29 (4), 827–45, 2021. Kohl, Markus (forthcoming1). “Ist die Lehrbarkeit der Tugend vereinbar mit Kant’s Theorie der Willensfreiheit?” B. Dörflinger, D. Hüning (eds.). Kant als Tugendethiker?, Studien und Materialien zur Geschichte der Philosophie. Hildesheim: Georg Olms Verlag, forthcoming. Also available at: https://philpapers.org/rec/KOHIDL. Kohl, Markus (forthcoming2). “Die radikale Unbegreiflichkeit von Gott für den menschlichen Verstand.” B.  Dörflinger, D.  Hüning (eds.). Die Gottesidee in Kants theoretischer und praktischer Philosophie, Studien und Materialien zur Geschichte der Philosophie. Hildesheim: Georg Olms Verlag, forthcoming. Also available at: https://philpapers.org/ rec/KOHDRU. Kohl, Markus (forthcoming3). “Kant on Cognition and Knowledge.” A.  Gomes, A. Stephenson (eds.). The Oxford Handbook of Kant. Oxford: Oxford University Press, forthcoming.

Bibliography  383 Kohl, Markus (forthcoming4). “Nietzsche versus Kant on the Possibility of Rational SelfCritique.” E. Valdez (ed.). Rethinking Kant. Volume 7. Newcastle: Cambridge Scholars Publishing, forthcoming. Korsgaard, Christine (1996a). Creating the Kingdom of Ends. Cambridge: Cambridge University Press, 1996. Korsgaard, Christine (1996b). The Sources of Normativity. Cambridge: Cambridge University Press, 1996. Korsgaard, Christine (1997). “The Normativity of Instrumental Reason.” G. Cullity, B. Gaut (eds.), Ethics and Practical Reason, 215–54. Oxford: Oxford University Press, 1997. Kosch, Michelle (2006). Freedom and Reason in Kant, Schelling, and Kierkegaard. Oxford: Oxford University Press, 2006. Kuehn, Manfred (2001). Kant. A Biography. Cambridge: Cambridge University Press, 2001. Kuhlenkampf, Jens (1978). Kants Logik des ästhetischen Urteils. Frankfurt: Vittoria Klostermann, 1978. Kvist, Hans-Olof (1978). Zum Verhältnis von Wissen und Glauben in der kritischen Philosophie Immanuel Kants. Turku: Abo Akademi, 1978. Lange, Marc (2013). “Grounding, Scientific Explanation, and Humean Laws.” Philosophical Studies 164, 255–61, 2013. Langton, Rae (1998). Kantian Humility: Our Ignorance of Things in Themselves. Oxford: Oxford University Press, 1998. Lavin, Douglas (2004). “Practical Reason and the Possibility of Error.” Ethics 114 (3), 424–57, 2004. Lehrer, Keith (1968). “Cans without Ifs.” Analysis 29 (1), 29–32, 1968. Leibniz, Gottfried Wilhelm (1988). Theodicy. Translated by E. M. Huggard. Chicago: Open Court, 1988. Leiter, Brian (2015). Nietzsche on Morality. London: Routledge, 2015. Lewis, David (1973a). “Causation.” Journal of Philosophy 70, 556–67, 1973. Lewis, David (1973b). Counterfactuals. Oxford: Basil Blackwell, 1973. Lewis, David (1983). “New Work for a Theory of Universals.” Australasian Journal of Philosophy 61, 343–77, 1983. Lewis, David (1994). “Humean Supervenience DeBugged.” Mind 103, 473–90, 1994. Lewis, David (2000). “Dispositional Theories of Value.” D.  Lewis, Papers in Ethics and Social Philosophy, 68–94. Cambridge: Cambridge University Press, 2000. Lewis, David (2003). “Are We Free to Break the Laws?” G. Watson (ed.), Free Will, 122–9. Oxford: Oxford University Press, 2003. Locke, Don (1973–4). “Natural Powers and Human Abilities.” Proceedings of the Aristotelian Society 74, 171–87, 1973–4. Loeb, Louis  E. (2002). Stability and Justification in Hume’s Treatise. Oxford: Oxford University Press, 2002. Loewer, Barry (1996). “Freedom from Physics: Quantum Mechanics and Free Will.” Philosophical Topics 24 (2), 91–112, 1996. Loewer, Barry (2004). “Humean Supervenience.” J.  Carroll (ed.), Readings on Laws of Nature, 176–206. Pittsburgh: University of Pittsburgh Press, 2004. Longuenesse, Béatrice (1998). Kant and the Capacity to Judge. Princeton: Princeton University Press, 1998. Longuenesse, Beatrice (2017). I, Me, Mine. Oxford: Oxford University Press, 2017. Lüthe, Rudolf (1984). “Kants Lehre von den ästhetischen Ideen.” Kant-Studien 75, 65–74, 1984. Lyons, T. D. (2003). “Explaining the Success of a Scientific Theory.” Philosophy of Science 70, 891–901, 2003.

384 Bibliography MacKay, D.  M. (1960). “On the Logical Indeterminacy of a Free Choice.” Mind 69, 30–40, 1960. Mackie, J. L. (1991). Ethics. Inventing Right and Wrong. London: Penguin Books, 1991. Makreel, Rudolf A. (1990). Imagination and Interpretation in Kant. Chicago: University of Chicago Press, 1990. Marshall, Colin (2010). “Kant’s Metaphysics of the Self.” Philosophers’ Imprint 20, 1–21, 2010. Marshall, Colin (2013). “Kant’s One Self and the Appearance/Thing-in-Itself Distinction.” Kant-Studien 104 (4), 421–41, 2013. Marshall, Colin (2018). “Never Mind the Intuitive Intellect: Applying Kant’s Categories to Noumena.” Kantian Review 23 (1), 27–40, 2018. Martin, Gottfried (1969). Immanuel Kant. Berlin: De Gruyter, 1969. Matherne, Samantha (2013). “The Inclusive Interpretation of Kant’s Aesthetic Ideas.” British Journal of Aesthetics, 53, 21–39, 2013. Matthews, H.  E. (1969). “Strawson on Transcendental Idealism.” Philosophical Quarterly 19, 204–20, 1969. McCarthy, Richard (2009). Kant’s Theory of Action. Oxford: Oxford University Press, 2009. McDowell, John (1978). “Are Moral Requirements Hypothetical Imperatives?” Proceedings of the Aristotelian Society 52, Supplementary Volumes, 13–29, 1978. McDowell, John (1994). Mind and World. Cambridge, MA: Harvard University Press, 1994. McDowell, John (1998a). “Having the World in View: Sellars, Kant, and Intentionality.” Journal of Philosophy 1998, XCV (9), 431–91, 1998. McDowell, John (1998b). “Might There Be External Reasons?” McDowell, Mind, Value and Reality, 95–111. Cambridge, MA: Harvard University Press, 1998. McLear, Colin (2020). “On the Transcendental Freedom of the Intellect.” Ergo 7 (2), 35–104, 2020. Meerbote, Ralf (1984). “Kant on the Nondeterminate Character of Human Actions.” W. L. Harper, R. Meerbote (eds.), Kant on Causality, Freedom, and Objectivity, 138–63. Minneapolis: University of Minnesota Press, 1984. Meerbote, Ralf (1991). “Kant’s Functionalism.” J. C. Smith (ed.), Historical Foundations of Cognitive Science, 161–88. Amsterdam: Kluwer Academic Publishing, 1991. Melnick, Arthur (1973). Kant’s Analogies of Experience. Chicago: University of Chicago Press, 1973. Moore, Adrian W. (2013). “Freedom, Temporality, and Belief: A Reply to Hare.” S. Baiasu, M. Timmons (eds.), Kant on Practical Justification, 315–18. Oxford: Oxford University Press, 2013. Moore, G. E. (1912). Ethics. Oxford: Oxford University Press, 1912. Moore, G. E. (1993). Selected Writings. London: Routledge, 1993. Moran, Richard (2001). Authority and Estrangement. An Essay on Self-Knowledge. Princeton: Princeton University Press, 2001. Morrisson, Iain (2005). “On Kantian Maxims: A Reconciliation of the Incorporation Thesis and Weakness of the Will.” History of Philosophy Quarterly 22 (1), 73–89, 2005. Nagel, Thomas (1989). The View from Nowhere. Oxford: Oxford University Press, 1989. Nehamas, Alexander (1987). Nietzsche: Life as Literature. Cambridge, MA: Harvard University Press, 1987. Nelkin, Dana (2000). “Two Standpoints and the Belief in Freedom.” Journal of Philosophy 97 (10), 564–76, 2000. Nelkin, Dana (2011). Making Sense of Freedom and Responsibility. Oxford: Oxford University Press, 2011.

Bibliography  385 Neta, Ram (2003). “Contextualism and the Problem of the External World.” Philosophy and Phenomenological Research 66, 1–31, 2003. Nietzsche, Friedrich (1966). Beyond Good and Evil. Translated by Walter Kaufmann. New York: Random House, 1966. Nietzsche, Friedrich (NF 1888: Gruppe 14, 109). Nachgelassene Fragmente. http://www. nietzschesource.org/#eKGWB/NF-1888, 14. Nozick, Robert (1983). Philosophical Explanations. Cambridge, MA: Harvard University Press, 1983. O’Neill, Onora (1989). Constructions of Reason. Cambridge: Cambridge University Press, 1989. Owens, David (2000). Reason without Freedom: The Problem of Epistemic Normativity. London: Routledge, 2000. Papineau, David (1992). “Reliabilism, Induction, and Skepticism.” Philosophical Quarterly 42 (66), 1–20, 1992. Pasternack, Lawrence (2011). “The Development and Scope of Kantian Belief: The Highest Good, the Practical Postulates, and the Fact of Reason.” Kant-Studien 102 (3), 290–315, 2011. Paton, H. J. (1958). The Categorical Imperative. A Study in Kant’s Moral Philosophy. London: Hutchinson & Co., 1958. Pereboom, Derk (2005). “Kant on Transcendental Freedom.” Philosophy and Phenomenological Research 73 (3), 537–67, 2005. Pereboom, Derk (2006). Living Without Free Will. Cambridge: Cambridge University Press, 2006. Pettit, Philip and Smith, Michael (2003). “Freedom in Belief and Desire.” G. Watson (ed.), Free Will, 388–407. Oxford: Oxford University Press, 2003. Pippin, Robert (1987). “Kant on the Spontaneity of Mind.” Canadian Journal of Philosophy 17, 449–75, 1987. Plantinga, Alvin (1993). Warrant and Proper Function. Oxford: Oxford University Press, 1993. Pogge, Thomas (1991). “Erscheinungen und Dinge an sich.” Zeitschrift für Philosophische Forschung 45 (4), 489–510, 1991. Pollok, Konstantin (2007). “ ‘Wenn Vernunft volle Gewalt über das Begehrungsvermögen hätte’—Über die gemeinsame Wurzel der Kantischen Imperative.” Kant-Studien 98 (1), 57–80, 2007. Pollok, Konstantin (2017). Kant’s Theory of Normativity. Oxford: Oxford University Press, 2017. Pothast, Ulrich (1980). Die Unzulänglichkeit der Freiheitsbeweise. Zu einigen Lehrstücken aus der neueren Geschichte von Philosophie und Recht. Berlin: Suhrkamp, 1980. Prauss, Gerold (1974). Kant und das Problem der Dinge an sich. Bonn: Bouvier Verlag, 1974. Prauss, Gerold (1983). Kant über Freiheit als Autonomie. Frankfurt: Vittorio Klostermann, 1983. Proops, Ian (2003). “Kant’s Legal Metaphor and the Nature of a Deduction.” Journal of the History of Philosophy 41, 209–29, 2003. Proops, Ian (2021). The Fiery Test of Critique: A Reading of Kant’s Dialectic. Oxford: Oxford University Press, 2021. Pryor, Jim (2000). “The Skeptic and the Dogmatist.” Nous 34 (5), 517–49, 2000. Puls, Heiko (2015). “Kant über die äußerste Grenze aller praktischen Philosophie—­Ein Kommentar zur Sektion 5 der Grundlegung.” D.  Schönecker (ed.), Kants Begründung von Freiheit und Moral in Grundlegung III, 157–214. Paderborn: Mentis, 2015. Quarfoot, Marcel (2006). “The Circle and the Two Standpoints.” C. Horn, D. Schönecker (eds.), Groundwork for the Metaphysics of Morals, 285–300. Berlin: De Gruyter, 2006.

386 Bibliography Quine, W. V. (1951). “Two Dogmas of Empiricism.” Philosophical Review 60, 20–43, 1951. Quinn, Warren (1994). Morality and Action. Cambridge: Cambridge University Press, 1994. Railton, Peter (1984). “Alienation, Consequentialism, and the Demands of Morality.” Philosophy and Public Affairs 13, 134–71, 1984. Railton, Peter (1986). “Moral Realism.” Philosophical Review 95, 163–207, 1986. Rauscher, Frederick (2002). “Kant’s Moral Anti-Realism.” Journal of the History of Philosophy 40 (4), 477–99, 2002. Rawls, John (2003). Lectures on the History of Philosophy. Cambridge, MA: Harvard University Press, 2003. Reath, Andrews (1989). “Hedonism, Heteronomy, and Kant’s Principle of Happiness.” Pacific Philosophical Quarterly 70, 42–72, 1989. Reath, Andrews (2006). Agency and Autonomy. Oxford: Oxford University Press, 2006. Reginster, Bernard (2009). The Affirmation of Life: Nietzsche on Overcoming Nihilism. Cambridge, MA: Harvard University Press, 2009. Rickert, Heinrich (1921). System der Philosophie. Erster Teil: Allgemeine Grundlegung der Philosophie. Tübingen: Mohr Verlag, 1921. Riehl, Alois (1894). The Principles of the Critical Philosophy. Translated into English by A. Fairbanks. London: Kegan Paul, 1894. Rogerson, Kenneth A. (2008). The Problem of Free Harmony in Kant’s Aesthetics. Albany: State University of New York Press, 2008. Rosen, Gideon (1994). “Objectivity and Modern Idealism.” M.  Michael and J.  O’LearyHawthorne (eds.), Philosophy in Mind, 277–319. Alphen aan den Rijn: Kluwer, 1994. Rosefeldt, Tobias (2000). Das logische Ich. Kant über den Gehalt des Begriffs von sich selbst. Berlin: Philo, 2000. Rueger, Alexander (2008). “The Free Play of the Faculties and the Status of Natural Beauty in Kant’s Theory of Taste.” Archiv für Geschichte der Philosophie 90, 298–322, 2008. Rueger, Alexander and Evren, Sahan (2005). “The Role of Symbolic Presentation in Kant’s Theory of Taste.” British Journal of Aesthetics 45 (3), 229–47, 2005. Russell, Bertrand (1992). “On the Notion of Cause.” J. Slater (ed.), The Collected Papers of Bertrand Russell v6: Logical and Philosophical Papers 1909–1913, 193–210. London: Routledge, 1992. Ryle, Gilbert (1949). The Concept of Mind. London: Hutchinson’s University Library, 1949. Salmon, Wesley (1984). Scientific Explanation and the Causal Structure of the World. Princeton: Princeton University Press, 1984. Scanlon, Thomas (2000). What We Owe to Each Other. Cambridge, MA: Harvard University Press, 2000. Scanlon, Thomas (2003). “The Significance of Choice.” G. Watson (ed.), Free Will, 352–71. Oxford: Oxford University Press, 2003. Scarano, Nico. “Moralisches Handeln. Zum dritten Hauptstück von Kants Kritik der praktischen Vernunft.” O.  Höffe (ed.), Immanuel Kant. Kritik der praktischen Vernunft, 117–31. Berlin: Akademie Verlag, 2002. Schafer, Karl (2010). “Evolution and Normative Skepticism.” Australasian Journal of Philosophy 88 (3), 471–88, 2010. Schafer, Karl (2017). “Kant’s Conception of Cognition and Our Knowledge of Things-inThemselves.” K.  Schafer, N.  Stang (eds.), The Sensible and Intelligible Worlds. Oxford: Oxford University Press, 2017. Schafer, Karl (2018). “Practical Cognition and Knowledge of Things-in-Themselves.” D.  Heide, E.  Tiffany (eds.), The Idea of Freedom. New Essays on the Significance and Interpretation of Kant’s Theory of Freedom. Oxford: Oxford University Press, 2018.

Bibliography  387 Schaffer, Jonathan (2001). “Causes as Probability-Raisers of Processes.” Journal of Philosophy 98 (2), 75–92, 2001. Schapiro, Tamar (2009). “The Nature of Inclination.” Ethics 119, 229–56, 2009. Schlösser, Ulrich (2015). “Kants Konzeption der Mitteilbarkeit.” Kant-Studien 106 (2), 201–33, 2015. Schönecker, Dieter (1999). Kant: Grundlegung III. Die Deduktion des kategorischen Imperativs. Freiburg: Karl Alber Verlag, 1999. Schönecker, Dieter (2005). Kants Begriff transzendentaler und praktischer Freiheit. Eine entwicklungsgeschichtliche Studie. Berlin: De Gruyter, 2005. Schönecker, Dieter (2013). “Kant’s Moral Intuitionism: The Fact of Reason and Moral Predispositions.” Kant-Studies Online, 1–38, 2013. Schönecker, Dieter and Wood, Allen (2002). Kants “Grundlegung zur Metaphysik der Sitten.” Ein einführender Kommentar. Paderborn: Schöningh, 2002. Schroeder, Mark (2007). Slaves of the Passions. Oxford: Oxford University Press, 2007. Scriven, Michael (1965). “An Essential Unpredictability in Human Behavior.” B. Wolman, E.  Nagel (eds.), Scientific Psychology: Principles and Approaches, 411–25. New York: Basic Books, 1965. Sellars, Wilfrid (1970). “. . . This I or He or It (the Thing) Which Thinks . . . .” Proceedings of the American Philosophical Association 44, 5–31, 1970. Sellars, Wilfrid (1978). “The Role of Imagination in Kant’s Theory of Experience.” http:// www.ditext.com/sellars/ikte.html, 1978. Sgarbi, Marco (2012). Kant on Spontaneity. London: Bloomsbury Publishing, 2012. Shah, Nishi (2003). “How Truth Governs Belief.” Philosophical Review 112, 447–82, 2003. Sider, Theodore (1992). “Tooley’s Solution to the Inference Problem.” Philosophical Studies 67, 261–75, 1992. Smith, Norman Kemp (2003). A Commentary to Kant’s “Critique of Pure Reason.” London: Palgrave Macmillan, 2003. Srinivasan, Amia (2019). “Genealogy, Epistemology, and Worldmaking.” Proceedings of the Aristotelian Society 119 (2), 127–56, 2019. Stang, Nicholas (2010). “The Non-Identity of Appearances and Things in Themselves.” Nous 47 (4), 106–36, 2010. Stang, Nicholas (2016). Kant and the Metaphysics of Modality. Oxford: Oxford University Press, 2016. Steigleder, Klaus (2002). Kants Moralphilosophie. Die Selbstbezüglichkeit reiner praktischer Vernunft. Stuttgart: Metzler, 2002. Stevenson, Leslie F. (2003). “Opinion, Belief or Faith, and Knowledge.” Kantian Review 7, 72–101, 2003. Stich, Stephen (1990). The Fragmentation of Reason: Preface to a Pragmatic Theory of Cognitive Evaluation. Cambridge, MA: MIT Press, 1990. Strawson, Galen (2003). “Mental Ballistics or the Involuntariness of Spontaneity.” Proceedings of the Aristotelian Society 103 (1), 227–56, 2004. Strawson, P.  F. (1966). The Bounds of Sense. An Essay on Kant’s Critique of Pure Reason. London: Methuen, 1966. Strawson, P. F. (1982). “Imagination and Perception.” R. Walker (ed.), Kant on Pure Reason, 82–99. Oxford: Oxford University Press, 1982. Strawson, P.  E. (2003). "Freedom and Resentment." G.  Watson (ed.), Free Will, 72–93. Oxford: Oxford University Press, 2003. Street, Sharon (2006). “A Darwinian Dilemma for Realist Theories of Value.” Philosophical Studies 127 (1), 109–66, 2006.

388 Bibliography Stroud, Barry (2006). “The Constraints of Hume’s Naturalism.” Synthese 152, 339–51, 2006. Stroud, Barry (2011). Engagement and Metaphysical Dissatisfaction: Modality and Value. Oxford: Oxford University Press, 2011. Sussman, David (2008). “From Deduction to Deed: Kant’s Grounding of the Moral Law.” Kantian Review 13 (1), 52–81, 2008. Timmermann, Jens (2010). “Reversal or Retreat? Kant’s Deductions of Freedom and Morality.” A. Reath, J. Timmermann (eds.), Kant’s Critique of Practical Reason: A Critical Guide, 73–89. Cambridge: Cambridge University Press, 2010. Timmermann, Jens (2022). The Will at the Crossroads. An Essay on Kant’s Theory of Practical Failure. Oxford: Oxford University Press, 2022. Tolley, Clinton (2006). “Kant and the Nature of Logical Laws.” Philosophical Topics 34, 471–506, 2006. Tooley, Michael (1987). Causation: A Realist Approach. Oxford: Oxford University Press, 1987. Tooley, Michael and Sosa, Ernest (1993). “Introduction.” M.  Tooley and E.  Sosa (eds.), Causation, 1–32. Oxford: Oxford University Press, 1993. Van Cleve, James (1999). Problems from Kant. Oxford: Oxford University Press, 1999. Van Fraasen, Bas C. (1989). Laws and Symmetry. Oxford: Oxford University Press, 1989. van Inwagen, Peter (1983). An Essay on Free Will. Oxford: Oxford Clarendon Press, 1983. van Inwagen, Peter (1993). Metaphysics. Boulder: Westview, 1993. Vilhauer, Benjamin (2010). “The Scope of Responsibility in Kant’s Theory of Free Will.” British Journal for the History of Philosophy 18 (1), 45–71, 2010. Walker, Ralph (1978). Kant. The Arguments of the Philosophers. London: Kegan & Paul, 1978. Walker, Ralph (2010). “Kant on the Number of Worlds.” British Journal for the History of Philosophy 18 (5), 821–843, 2010. Wallace, R. Jay (1994). Responsibility and the Moral Sentiments. Cambridge, MA: Harvard University Press, 1994. Wallace, R. Jay (2006). Normativity and the Will. Oxford: Oxford Clarendon Press, 2006. Walsh, W.  H. (1975). Kant’s Criticism of Metaphysics. Chicago: University of Chicago Press, 1975. Ware, Owen (2014). “Rethinking Kant’s Fact of Reason.” Philosophers’ Imprint 14 (32), 1–21, 2014. Ware, Owen (2017). “Kant’s Deductions of Morality and Freedom.” Canadian Journal of Philosophy 47, 116–47, 2017. Warren, Daniel (1998). “Kant and the Apriority of Space.” Philosophical Review 107, 179–224, 1998. Warren, Daniel (2001). Reality and Impenetrability in Kant’s Philosophy of Nature. London: Routledge University Press, 2001. Warren, Daniel (unpublished). “Kant on Inner Determinations and Things in Themselves.” Watkins, Eric (2002). “Kant’s Transcendental Idealism and the Categories.” History of Philosophy Quarterly 19 (2), 191–215, 2002. Watkins, Eric (2004). Kant and the Metaphysics of Causality. Cambridge: Cambridge University Press, 2004. Watkins, Eric and Willaschek, Marcus (2017a). “Kant’s Account of Cognition.” Journal of the History of Philosophy 55 (1). 83–112, 2017. Watkins, Eric and Willaschek, Marcus (2017b). “Kant on Cognition and Knowledge.” Synthese 197 (8), 3195–213, 2017. Westphal, Kenneth (2005). Kant’s Transcendental Proof of Realism. Cambridge: Cambridge University Press, 2005.

Bibliography  389 White, Roger (2010). “You Just Believe That Because . . . .” Philosophical Perspectives 24 (1), 573–615, 2010. Widerker, David (2003). “Libertarianism and Frankfurt’s Attack on the Principle of Alternative Possibilities.” G. Watson (ed.), Free Will, 177–89. Oxford: Oxford University Press, 2003. Wiggins, David (1969). “Freedom, Knowledge, Belief and Causality.” Royal Institute of Philosophy Lectures 3, 132–54, 1969. Wiggins, David (2003). “Towards a Reasonable Libertarianism.” G. Watson (ed.), Free Will, 94–121. Oxford: Oxford University Press, 2003. Willaschek, Marcus (1991). “Die Tat der Vernunft. Zur Bedeutung der Kantischen These vom ‘Faktum der Vernunft.’ ” G.  Funke (ed.), Akten des Siebten Internationalen KantKongresses, 455–66. Bonn: Bouvier Verlag, 1991. Willaschek, Marcus (1992). Praktische Vernunft. Handlungstheorie und Moralbegründung. Stuttgart: Metzler, 1992. Willaschek, Marcus (2010). “The Primacy of Practical Reason and the Idea of a Practical Postulate.” A.  Reath, J.  Timmermann (eds.), Kant’s Critique of Practical Reason: A Critical Guide, 168–96. Cambridge: Cambridge University Press, 2010. Willaschek, Marcus (2018). Kant on the Sources of Metaphysics. The Dialectic of Pure Reason. Oxford: Oxford University Press, 2018. Williams, Bernard (1973). Problems of the Self. Cambridge: Cambridge University Press, 1973. Williams, Bernard (1986). Ethics and the Limits of Philosophy. Cambridge, MA: Harvard University Press, 1986. Williams, Bernard (1997). “Internal and External Reasons.” S.  Darwall, A.  Gibbard, P.  Railton (eds.), Moral Discourse and Practice, 363–72. Oxford: Oxford University Press, 1997. Williamson, Timothy (2000). Knowledge and Its Limits. Oxford: Oxford University Press, 2000. Windelband, Wilhelm (2015). “Critical or Genetic Method?” S. Luft (ed.), The Neo-Kantian Reader, 271–86. London: Routledge, 2015. Wolf, Susan (1980). “Asymmetrical Freedom.” Journal of Philosophy 77, 151–66, 1980. Wolf, Susan (1990). Freedom Within Reason. Oxford: Oxford University Press, 1990. Wolf, Susan (2003). “Sanity and the Metaphysics of Responsibility.” G.  Watson (ed.), Free Will, 372–87. Oxford: Oxford University Press, 2003. Wolff, Michael (2009). “Warum das Faktum der Vernunft ein Faktum ist. Auflösung einiger Verständnisschwierigkeiten in Kants Grundlegung der Moral.” Deutsche Zeitschrift für Philosophie 57 (4), 511–49, 2009. Wolff, Robert Paul (1963). Kant’s Theory of Mental Activity: A Commentary on the Transcendental Analytic of the Critique of Pure Reason. Cambridge, MA: Harvard University Press, 1963. Wood, Allen (1974). Kant’s Moral Religion. Ithaca: Cornell University Press, 1974. Wood, Allen (1978). Kant’s Rational Theology. Ithaca: Cornell University Press, 1978. Wood, Allen (1984). “Kant’s Compatibilism.” A.  Wood (ed.), Self and Nature in Kant’s Philosophy, 73–101. Ithaca: Cornell University Press, 1984. Wood, Allen (1999). Kant’s Ethical Thought. Cambridge: Cambridge University Press, 1999. Wood, Allen (2002). “Preface and Introduction (3-16).” O. Höffe (ed.), Klassiker Auslegen. Immanuel Kant: Kritik der praktischen Vernunft, 21–36. Berlin: Akademie Verlag, 2002. Wood, Allen (2007). Kantian Ethics. Cambridge: Cambridge University Press, 2007.

390 Bibliography Wood, Allen (2013). “Kant on Practical Reason.” S. Baiasu, M. Timmons (eds.), Kant on Practical Justification, 57–86. Oxford: Oxford University Press, 2013. Woodward, James (1990). “Laws and Causes.” British Journal for the Philosophy of Science 41, 553–73, 1990. Woodward, James (2003). Making Things Happen: A Theory of Causal Explanation. Oxford: Oxford University Press, 2003. Wuerth, Julian (2014). Kant on Mind, Action, and Ethics. Oxford: Oxford University Press, 2014. Young, J. Michael (1988). “Kant’s View of Imagination.” Kant-Studien 79, 140–64, 1988. Zuckert, Rachel (2007). Kant on Beauty and Biology. Cambridge: Cambridge University Press, 2007.

Index For the benefit of digital users, indexed terms that span two pages (e.g., 52–53) may, on occasion, appear on only one of those pages. Abaci, Uygar  77n.6, 78n.7, 84nn.16, 18, 88n.22, 89n.23, 331n.50 Abela, Paul  40n.68, 164n.10 accountability  32, 64–5, 72, 84n.17, 88–9, 95, 133–4, 165, 202n.83, 206–7, 211, 247–8, 295, 300–1, see also imputation, responsibility Adams, Robert  35n.55, 76n.2, 223–4 affection as distinct from causal determination  65–6, 195n.72, 200–1 noumenal 29n.41 sensible  90–1, 94–6, 156–7 akrasia  148–52, 175–6, 207 Allais, Lucy  32n.45, 33nn.48, 50, 34n.53, 35n.54, 36nn.59–60, 57n.11, 185n.54, 245n.80, 253, 255–6 Allison, Henry  19–25, 27n.38, 30n.42, 31nn.43, 44, 32n.45, 33n.48, 37n.62, 49nn.84, 86, 50n.87, 52n.2, 56n.10, 66n.31, 81nn.13, 15, 93n.27, 95n.31, 96n.33, 34, 97n.35, 103–4, 131–2, 143n.29, 147n.39, 149–51, 154–6, 161–2, 175n.30, 181, 184n.51, 185n.55, 186n.60, 187, 196–7, 199, 208n.5, 243n.74, 255, 259n.17, 267–71, 274n.48, 279n.54, 311n.15, 331n.49, 333n.51, 343–4, 346n.16 alternative possibilities  5–6, 55, 58, 64, 78, 88–91, 131–48, 153–8, 174–80, 205–10, 237–8, 329–34, 368–9 Ameriks, Karl  5, 8n.10, 20n.16, 37n.62, 49n.86, 50n.87, 69–70, 76n.4, 79n.10, 96n.33, 109n.17, 124n.39, 131, 186n.58, 187nn.61, 63, 190n.65, 193n.69, 235–6, 245n.76, 249n.87, 258n.14, 261n.20, 271n.42, 273nn.45–46, 274n.48, 280n.55, 284n.62, 296n.85, 299n.88, 307n.9, 353n.26 Anderson, Lanier  231n.61 Anscombe, Elizabeth  147n.38, 215n.23 apperception  28, 162, 180–93, 304 see also cognition of the self appearances and things in themselves (noumena, intelligible objects)  2, 17–18, 22–3,

28–37, 40–1, 51–5, 77n.5, 82, 180–1, 184–6 as perceptual representations  162–6, 177, 211 as spatial objects  193–6 Aristotle 146–8 Armstrong, David  218nn.37–38, 227n.52 assent (belief, Fürwahrhalten)  19–20, 161, 262 autocracy  259n.17, 289n.76 automaton spirituale  126, 161 autonomy of aesthetic judgment (art, taste, imagination)  340–1, 351n.21, 371–3 of practical reason (will)  50, 78–80, 97n.35, 155, 259n.17, 289–90, 298, 327n.38 of theoretical reason  198–9, 326–7 of understanding  7, 162–3, 197–9, 315–16 axiom as postulate  264–5 mathematical  264–6, 320–1, 340n.2 moral law as  10–11, 258–9, 264–6, 274–5, 279, 281–2, 284, 305–7, 311–12 Baiasu, Sorin  262n.23, 264n.25 Banham, Gary  340n.1 Baron, Marcia  147n.39 Baum, Manfred  299n.88 Baumgartner, Michael  216nn.27–28, 217n.32 beauty artistic  346–53, 363–9 as a symbol of morality  345 judgments of  340–1, 355–7, 360–7 natural 352–62 Beck, Lewis White  29n.41, 30n.42, 32n.47, 51n.1, 93n.28, 95n.31, 96nn.34–35 Beebee, Helen  138n.19, 216n.30 Bennett, Jonathan  3nn.3–4, 32n.47, 51n.1, 58–9, 73n.44, 133n.8 Bird, Graham  8n.12, 103–4, 262n.22 blame (blameworthiness) aesthetic 368–9 epistemic  166n.11, 202n.83, 206–7, 247–8 moral  2, 32n.47, 64–5, 95, 134, 158n.66, 202n.83, 206–7, 301n.90

392 Index Bojanowski, Jochen  22n.26, 30n.42, 93n.28, 97n.35, 306n.8 Bok, Hilary  58–9, 60n.17, 132–3, 153n.52 Boyle, Matthew  169n.17, 172n.21, 175n.30 Brewer, Bill  176n.34, 210–11 Burge, Tyler  175n.30 Canon of Pure Reason  48–50, 254n.3, 259n.15, 271–3, 297–300 Carnois, Bernard  150n.44, 255n.6, 259n.17, 289n.75 categorical imperative and hypothetical imperatives  110–12, 116, 153–4 and moral law  134–6, 267–73 as axiom  265 See also alternative possibilities, determination of finite wills by reason, moral duty, rational necessitation categories aesthetic meaning or application of  346–9, 358n.32 and the intuitive or divine intellect  77–8, 201 extension of to the supersensible  80–6, 191 origin of  104, 190, 198, 222, 229, 236–41, 243 pure or unschematized  75–8, 80–2, 247n.82 practical meaning or application of  81–6, 295 schematized  75, 164 See also causality, substance causality contemporary theories of  215–19 genuine or true  193–9, 317, 325–6 of freedom (noumenal, supersensible)  2–3, 16–18, 25–32, 41–8, 51–5, 65n.27, 67, 70–86, 197, 300–2 of nature (empirical, phenomenal)  5–6, 8–9, 15–18, 25–31, 48, 51–8, 66–7, 70–2, 97–8, 111–16, 124–6, 131–2, 362–3 of reason see causality of freedom practical concept of  80–6, 91–3, 255–6, 295 theoretical (schematized) concept (category) of  164n.9, 174–5, 198, 201–2, 214–15, 219–22, 229–42, 314–15 cause see causality certainty and knowledge  239–40, 243, 259–60, 263–5, 274–7, 298n.87 empirical 321n.25 moral 271–3 predictive  55–6, 63 character empirical (phenomenal, spatiotemporal)  29–30, 39, 51–6, 62–8, 70–3

intelligible (non-sensible, moral, noumenal)  28–30, 35, 39, 51–5, 67–73, 83, 89n.23, 149–51, 193–4, 246 Chignell, Andrew  161, 168, 177n.35, 240n.69, 262n.23, 360n.36 Churchland, Patricia  227–8 cognition and certainty  262, 264–5 and knowledge  255–6, 263–4, 295 divine  28n.40, 61n.20, 77–8, 83–5, 209–10, 329–32 in a narrow sense (cognitionN)  181–95, 242–3, 246, 248–9, 283–4, 295 in a wide sense (cognitionW)  181–7, 246, 248–9 of the self  180–9, 192–3, 244–8 of the supersensible  80–6, 295–6 practical  85–6, 255–8, 260, 262, 265–8, 274–80, 288–9, 295–6, 305–7, 357 theoretical  28, 31, 75–6, 81–2, 163–6, 174–5, 219–20, 249–50, 262, 266, 348 common human reason (common human sense, common understanding)  38, 42–3, 64, 71–2, 74, 281–2, 286, 306, 320–3 compatibilism  4–6, 16–17, 20–2, 37–8, 40n.67, 46–50, 60–1, 98–9, 131–3, 137–9, 153–5, 158, 235–6 contingency empirical (relative)  6, 97–9, 103, 105, 108–9, 111–21, 123–5, 220–2, 225, 227n.52, 230–1, 233, 276, 366–7 noumenal (absolute, moral)  10, 32n.46, 78, 84–91, 94–5, 97–9, 135n.14, 153, 157–8, 199, 206, 237–8, 268–9, 331, see also alternative possibilities, rational necessitation practical concept of  83–6 theoretical concept of  78, 80, 83–4 Crawford, David  340n.2, 342n.5 Darwin, Charles  228 deflationary (anti-metaphysical) interpretations  2–3, 18–25, 31–3, 40, 71, 184 deduction of categories  104, 194, 236–40 of freedom  256, 299 of moral law  194, 264, 274n.48, 299 of taste  360–2 deed as act of willkür 95 legislative  93–6, 138–9, 267–8 Dennett, Daniel  4, 138n.19 de Sousa, Ronald  212n.16

Index  393 determination absolutely necessary inevitable rational self  89, 89n.23, 91, 94, 135, 174–5, 197n.76, 207 by foreign causes  10, 30, 78–80, 86–7, 93–4, 96–8, 125–6, 138–9, 154–5, 167, 197–202, 215, 239, 243, 253, 298, 304, 326, 329–30, 339–40, 362, 370–2 by the truth  209–48, 313–14 free (absolutely spontaneous) self  79–80, 86–9, 89n.23, 91, 94, 96, 126, 158n.66, 174–5, 181, 187–96, 201, 246, 315–16, 372–3 of finite wills by reason  135, see also alternative possibilities, rational necessitation practical-causal (PCD)  188–97 time  29–30, 52–3, 56–7, 173n.25, 234–5 theoretical-cognitive (TCD)  188–200, 246, 248–9 Descartes, René  113, 177n.35, 236–7, 241, 261n.19 determinism  5–6, 9, 15–16, 20–1, 30–1, 39–40, 52–3, 56–7, 87–9, 132–3, 136, 153, 157–8, 208–10, 219–22, 313–14, 363 dogmatic as a pejorative term for Kant’s view  5, 7, 131, 273–5, 282–3, 297n.86, 303, 309–10 as a technical term  273 empiricism (naturalism)  17, 22n.25, 97, 124–5, 221n.46, 309–10 rationalism  17, 186–7, 223, 244–5, 297n.86 Dohrn, Daniel  353n.26 doxastic voluntarism  163, 167–9, 286–7, 325–6 Ducasse, C.J.  215n.23 duty acting from  135, 291–2, see also rational necessitation moral  68, 90–1, 206–7, 254–6, 262–3, 265, 268–72, 288, 297–8, 310–11, 332–3, see also alternative possibilities, categorical imperative of speculative philosophy  40, 64, 71–2, 74 Earman, John  9n.13 empirical standpoint  8, 18–25, 39–46, 59–60, 70–4 empiricism  8, 17, 19n.12, 22–3, 97, 124–5, 178, 213, 239, 242–3, 249, 279–81, 315, 328–9 Engstrom, Stephen  95n.30, 143n.29, 261n.20, 263n.24, 283n.60 Enoch, David  107nn.13–14, 126n.40 evil  67–70, 79–80, 84n.17, 87, 89n.23, 95, 107, 149–52, 202–3, 329–31, 365n.41

experience aesthetic  339–40, 345, 352–61 judgment of  164–5, 169n.17, 170–1, 173–4 outer  193–5, 320–1 practical  94, 140–1, 346–9 spatiotemporal  2–3, 16–18, 22–3, 29, 31, 42–3, 49, 57, 70–1, 104, 229, 307, 309–10 fact of reason  7, 93–4, 272n.44, 273–4, 279–84, 286–7, 291, 301, 306–7, 311–12 faith (Glaube)  1, 7, 19–20, 23–4, 31, 75, 161, 203, 242, 248, 253–6, 258–60, 262–4, 271–2, 283–4, 292, 299–300, 320–1, 329–30, 332–3 Fales, Evan  215–16, 217n.34 feeling aesthetic  195n.73, 340, 343n.9, 357, 372–3 empirical  45, 65–6, 80, 107–9, 113–17, 143–6, 151–2, 187, 215–16, 276–7, 279–80, 305–9, 317 moral  65n.27, 107, 266, 268n.35, 357 Frankfurt, Harry  119–20, 158n.66 frailty see akrasia freedom of imagination  11, 337, 339 freedom of thought (epistemic freedom) and divine determination  236–40 as autonomy  197–9 as distinct from freedom of will  188–9, 191, 193–6, 201–3, 253, 295, 304–5, 317–18 as legislative freedom  208, 220–2 as transcendental freedom  196–203 knowledge of  242–9 negative  163–8, 202–3 positive  167–76, 202–3 freedom of will and divine freedom  86–93 and divine foreknowledge  329–34 as autonomy  50, 78–80, 96–7, 138–9, 288–90, 297–9 executive  93–9, 131, 205–8, 289–90 knowledge of  248–9, 253–61, 285–323 legislative  93–9, 103, 138–9, 208, 220, 289–90 practical versus transcendental  48–50 See also absolutely necessary inevitable rational self determination, causality of freedom (noumenal, supersensible), holy (perfectly rational) will, noumenal (absolute, moral) contingency, practical concept of causality, rational necessitation freedom of a turnspit  78–9, 86–7, 133, 154n.54, 255, 362 Friedman, Michael  52n.2, 176n.33, 213n.19 Frierson, Patrick  19n.12, 23–5, 57–8, 65n.27, 154n.55, 286n.67

394 Index Frisch, Mathias  219n.42 Forster, Michael  196n.74 genetic fallacy  103–4, 122–4, 222–6 genius  346–52, 362–7 Gibbons, Sarah  340n.2 Ginsborg, Hannah  343n.8, 344n.14, 353n.26, 355n.28 Gesinnung  67–8, 149–50, 271–2 God deistic (theoretical) concept of  332 theistic (practical) concept of  332–3 See also absolutely necessary inevitable rational self determination, categories and the intuitive or divine intellect, divine cognition, holy (perfectly rational) will, faith, freedom of will and divine freedom, freedom of will and divine foreknowledge Goldman, Alvin  222n.48 good as non-imperatival deontic concept  84n.17, 268–70 guise of the  151–3 will  43, 67–70, 83–6, 135–6 Goodman, Nelson  212 Graham, Peter  138n.20 great reversal  190n.65, 258n.14, 274n.48, 299, 334 Grenberg, Jeanine  264n.25, 271n.42, 281n.57 Grier, Michelle  178n.37, 183–7 Grüne, Stephanie  181n.43, 192n.67, 214n.20 ground epistemic (ratio cognoscendi)  162, 244, 260–1, 306–7, 315–16 real (ratio essendi)  260–1, 288–90, 294n.82, 301, 315–16, 318, 332–4 Guyer, Paul  33n.49, 34n.51, 46n.78, 52n.2, 106n.11, 115n.27, 179n.38, 243n.74, 273–4, 296n.85, 341–2, 351n.21, 353n.25, 360n.37, 362n.39, 366n.43 Hacking, Ian  212n.16 Haji, Ishtiyaque  138n.20, 139n.24 happiness  274–9, 292–3 Hatfield, Gary  56n.10, 132n.5 heautonomy 342–3 Heidemann, Ingeborg  362n.40 Heimsoeth, Heinz  196n.74 Henrich, Dieter  103n.2, 199, 202n.83 Henson, Richard  147n.39 Herman, Barbara  144n.30, 147n.39 heteronomy  50, 110–12, 126, 138–9, 189–90, 297–9, 305, 340–1 highest good  254–6, 290–4, 332–4 Hampshire, Stuart  175n.30

Hanna, Robert  161, 199n.81 Heals, Jane  213 Hill, Thomas  18–22, 38, 78–9, 150n.43, 155n.61, 286n.69 Hieronymi, Pamela  169n.17 Hogan, Desmond  27n.39, 28nn.40–41, 32n.46, 41n.69, 255n.6, 303 holy (perfectly rational) will  83–4, 88–9, 135–6, 139–42, 139n.24, 156–7, 267–71, see also absolutely necessary inevitable rational self determination, God, good as non-imperatival deontic concept Horn, Christoph  208nn.1, 5, 304n.6, 311n.14 Hudson, Hud  56–7 Hume, David  4, 21, 22n.25, 44n.76, 52, 80–1, 104, 112–13, 114n.26, 154n.54, 166, 168, 174–5, 211–12, 215–23, 227n.52, 229–31, 235–6, 239–42, 279–80, 283, 327–9 I see apperception, cognition: of the self idea aesthetic  346–50, 352–4, 356, 358–67 of freedom  1, 3–4, 11, 20–1, 27, 41–2, 55–6, 64, 87, 92–3, 161, 167–8, 255–9, 256n.10, 259n.17, 339 of reason (transcendental)  1, 256n.10, 259, 287, 344–5, 347–51, 353, 356–60, 369–70 illusion empirical (perceptual)  163–7, 169–70, 177–8 transcendental 178–80 immortality  1, 246, 253–7, 259–60, 285, 288–90, 292–4, 298–300 imputation  32, 35–6, 73n.45, 134n.11, 161, 165, 186–7, 206–7, 247–8, see also accountability, responsibility incompatibilism about freedom of imagination  339, 362–7 about freedom of thought  203, 205–10, 219–22, 235–42 about freedom of will  4–6, 16–17, 20–2, 37–40, 46–7, 60–1, 97–9, 129, 131–9, 153–7 incorporation thesis  131–2, 149–51 Indregard, Jonas  67n.37 Irwin, Terence  5n.7, 131n.2 Jauernig, Anja  37n.63 Joyce, Richard  106, 109–10, 116 Kim, Jaegwon  8n.11 Kahane, Guy  107n.13, 112n.23 Kain, Patrick  255n.6, 257n.11, 259n.17, 289nn.75, 76, 296n.84 Kane, Robert  9n.15 Keller, Pierre  199n.81

Index  395 Kemp Smith, Norman  261n.20 Kleingeld, Pauline  93n.29, 274n.48, 279n.54, 280n.55 Kneller, Jane  343n.9 knowledge empirical 320–1 general account of a priori  242, 263–6 Kant’s internalism about  177n.35, 231n.61, 240n.69, 287n.70 moral 261–85 of transcendental epistemic freedom  242–9 of transcendental freedom of will  248–9, 253–64, 285–302, 305–6, 316, 318–23 Kitcher, Patricia  161, 167n.12, 169n.17, 175n.30, 183n.44, 185n.56, 187n.63, 192–3, 196–7, 197n.76, 202n.83, 229–36, 241, 248–9, 326n.35 Klemme, Heiner  324n.26, 327n.38 Korsgaard, Christine  18–22, 24n.30, 41n.71, 58–60, 60n.17, 103–4, 111n.22, 121, 131n.1, 132–3, 155n.61, 156–7, 258n.14, 269n.38 Kuehn, Manfred  1 Langton, Rae  33n.48, 37nn.61–62 Lange, Marc  216n.30 Lavin, Douglas  136n.15, 139–40, 144–5, 156–7, 269n.38 Leibniz, Gottfried Wilhelm  5, 86–7, 91, 154n.54, 223, 237 Leiter, Brian  123nn.35–36, 138n.19, 309n.11, 327–9 Lewis, David  4, 117–18, 138n.19, 142n.27, 212n.17, 216n.28, 217n.34, 218n.38, 219n.45, 223, 229n.59, 233n.62 libertarianism  2, 9n.15, 39–40, 46–7, 51, 209 liberty of indifference  11, 140n.25, 202, 340–1, 343, 349, 351, 367, 371, see also alternative possibilities, contingency: noumenal (absolute, moral) Locke, John  104–5, 177n.35 Loeb, Louis  175n.29 Loewer, Barry  9n.15, 47n.80, 214n.21, 218–19, 219n.45, 229n.59, 233 Longuenesse, Béatrice  33n.50, 47n.81, 164n.9, 174n.28, 185nn.55, 56, 186n.58, 231n.61, 248n.83 Lüthe, Klaus  349n.20, 353n.24 Mackie, J.L.  110n.20, 111n.22 Makreel, Rudolph  358n.32 Marshall, Colin  34nn.52–53, 36nn.57–58, 77n.6, 185n.54 Martin, Gottfried  184n.53, 196n.74 Matherne, Samantha  356n.29

maxim highest (life-governing) practical  67–70, 149–50 regulative of theoretical reason  172n.23 subjective epistemic  171–2, 230 subjective practical  66, 91–5, 131–2, 147–51 McCarthy, Richard  56n.10, 64–70, 149n.41, 151n.45 McLear, Colin  88nn.22–23, 169n.17 McDowell, John  140n.25, 147n.38, 170n.19, 172n.21, 175n.30 materialism  9, 132n.5 mathematics and aesthetics  340n.2, 342n.5 and etiological skepticism  224n.50 and transcendental idealism  320–1 as analogous to morality  262–6, 272n.44, 279–81 mechanism of nature (of merely subjectively determining causes)  1, 18, 25–7, 56–7, 105, 110–14, 128–9, 220–1, 230–1, 310–11, 313–14, 322–3, 362, 365–6, 370 thinking  10, 161, 165, 180, 192, 197–8, 205–6, 244–5 See also automaton spirituale metaphysical reading of Kant’s doctrine of freedom  2–4, 18–25, 29–37, 40–2, 44–6, 296–7 reading of Kant’s idealism  29–37, 40–2, 184–6, 296–7 Moran, Richard  172n.22, 173n.24 Moore, Adrian  69n.43 Moore, G.E.  4, 60n.17, 137n.17, 173n.24 moral law see autonomy of practical reason (will), categorical imperative versus moral law, deduction of moral law, fact of reason, legislative freedom of will, moral law as axiom moral (purely rational) motive (motivation)  45, 50, 65n.27, 80, 83, 114, 131, 138–9, 143–9, 153, 155n.60, 156, 266–8, 289–94, 297–9, 305–6, see also acting from duty Morrisson, Iain  150n.43 Nagel, Thomas  3nn.3–4, 46nn.78, 79, 209–11, 224, 313–14 naturalism as a threat to morality  5–6, 45, 106–7, 109–19, 133–9, 307–10, 312–13, 317–18, 322–3 as a threat to theoretical cognition  126–9, 210–36, 240–2, 249–50, 313–16 as priority of common understanding over science 8n.12

396 Index naturalism (cont.) as transcendental realism  8n.12, 17–18, 22–3 definition of  8–9 See also dogmatic empiricism (naturalism) nature see appearances, causality of nature, empirical character, naturalism, spatiotemporal experience necessitation natural (empirical, hypothetical, nomological)  5–6, 8–9, 15–16, 29–30, 30n.42, 38, 57–8, 64, 134–5, 137 of assent (judgment)  161, 180, 200–1, 217–18, 221–2, 225, 314–15, 328–9, 362 psychological 56–8 rational  94–6, 135n.14, 268–9, see also determination of finite wills by reason necessity natural (conditional, empirical, hypothetical, nomological)  5–6, 15–17, 28–32, 36–8, 52–4, 56–7, 71–2, 97–9 practical concept of  83–6 rational (moral, objective, unconditional)  6, 10, 83–9, 92, 95, 98–9, 104–19, 125, 129, 143–6, 156, 225, 263–4, 266, 268–9, 279–82, 291–4, 305–6, 322–3, 365–6 rational (objective, epistemic)  219–22, 225, 311, 313–16, 322–3, 365–6 subjective aesthetic  365–7 subjective non-rational  107–8, 111–19, 220–2, 236–8, 314–15, 322–3, see also mechanism of nature (of merely subjectively determining causes) subjective rational  108, 254–6, 259–60, 292–4, 369 theoretical concept of  78, 80, 83–4 Nehamas, Alexander  123–4 Nelkin, Dana  42n.74 Nietzsche, Friedrich  123–4, 228n.56, 327–9 normative laws  39, 78–80, see also moral law standpoint  3–4, 18–25, 39–46, 59–60, 70–4, 78–9 noumenal causality see causality of freedom noumena and contingency see noumenal (absolute, moral) contingency character of see intelligible character cognition of see cognition of the supersensible negative concept of  28–9, 81–2, 162 positive concept of  77–8, see also practical concept of causality versus appearances see appearances and things in themselves

objectivity epistemic see rational (objective, epistemic) necessity practical see rational (moral, objective, unconditional) necessity objective validity  22n.25, 27, 76n.4, 104, 112–13, 117, 124, 143–4, 172, 201–2, 221–2, 224, 236–44, 256, 275n.52, 306–7, 314–16 obligation see imperative versus hypothetical imperatives, categorical imperative versus moral law, moral duty, rational necessitation O’Neill, Onora  57n.11, 66, 271n.42, 284n.63, 323–7 opinion  19–20, 161, 170–2, 240n.69, 242, 243n.74, 255–6, 259, 262–3, 277 ought aesthetic 368–9 epistemic see rational (objective, epistemic) necessity moral see categorical imperative, moral duty, rational necessitation, rational (moral, objective, unconditional) necessity Owens, David  121–2, 169–70, 177n.35 Pasternack, Lawrence  262nn.21, 22, 264n.25, 271–3 Pereboom, Derk  58n.12, 67n.38, 88n.22, 134n.10, 286n.67, 300–1 personal identity  35n.54, 247–8, 295 Pettit, Philip and Smith, Michael  175n.31 phenomena see appearances and things in themselves, empirical (phenomenal, spatiotemporal) character, spatiotemporal experience Pippin, Robert  248–9 Plantinga, Alvin  228n.57, 241 pleasure aesthetic  195n.73, 340, 343n.9, 345, 347, 357–8, 366–8 empirical feelings of  117, 131, 143–4, 146n.37, 151–2, 181–2, 276–7, 307–8, 355, 366–7 moral 357 Pollok, Konstantin  156n.62 possibility logical  22–3, 25–6, 72, 257 noumenal 84–6 real  17–18, 27–8, 54–5, 257–8 postulate as axiom  264–5 mathematical  259–60, 264–6 moral law as  259–60, 264–6 of practical reason  259–60

Index  397 Prauss, Gerold  7n.9, 31n.44, 297n.86, 310n.12 privation  89–90, 132–3, 135 proof of empirical determinism  16n.2, 52–3 of free will by practical reason  253–61, 285–302 of moral law  273–85 of free will by theoretical reason  303–5 Proops, Ian  26n.36, 40n.67, 61n.20, 68n.39, 73n.44, 181n.42, 185n.55, 186n.58, 245n.77, 261n.19, 281n.57, 294n.82 prudence 274–9 Pryor, Jim  320n.23 Puls, Heiko  78n.8, 183n.45 Putnam, Hilary  227n.52 Quarfoot, Marcel  304n.6, 311n.14 Quine, W. V.  213 Quinn, Warren  151–2 rationality see autonomy, normative laws, normative standpoint, rational necessitation, rational (objective, epistemic) necessity, rational (moral, objective, unconditional) necessity, subjective rational necessity Rickert, Heinrich  208–9 Railton, Peter  111n.22, 143n.29 Rawls, John  289n.75 reality objective  17–18, 22–4, 27, 76n.4, 89–90, 257–9, 289–90, 295, see also objective validity of freedom  256, 259 reason see autonomy, normative laws, normative standpoint, rational necessitation, rational (objective, epistemic) necessity, rational (moral, objective, unconditional) necessity, subjective rational necessity, unity of reason Reath, Andrews  65n.27, 80n.12, 95n.30, 137n.18, 143n.29 Reginster, Bernard  123 respect for moral law  65n.27, 69, 107, 147n.39, 148, 266, 291–2 for persons  202n.83, 292n.79 Riehl, Alois  103–4 rigorism 149–51 Rogerson, Kenneth  353n.24, 362n.39 Rosefeldt, Tobias  183n.46, 184n.51, 208–9, 248n.86 Rosen, Gideon  48n.82 Rueger, Alexander  349n.19, 352n.23, 353n.24, 360n.36

Rueger and Evren, Sahan  352n.23, 353n.24, 356n.29, 359n.34, 361n.38 Russell, Bertrand  218–19 Ryle, Gilbert  58–61 Salmon, Wesley  216n.31, 217n.35 Scanlon, Thomas  120, 137n.18, 152n.50 Schapiro, Tamar  141n.26, 152n.50 Schlösser, Ulrich  343–5, 353n.26, 360n.37, 369n.44 Schönecker, Dieter  183n.45, 266n.31, 267n.33, 268n.35 Schönecker, Dieter and Wood, Allen  267n.33, 268n.36 Schulz, Johann Heinrich  208, 208n.5, 220–2, 304, 308–16 Schafer, Karl  35n.54, 107nn.13–14, 112n.23, 126–8, 249–50, 255n.6, 287–9 Schaffer, Jonathan  9n.14, 217n.33 science and knowledge  258, 265 mathematics as  258, 262, 265–6, 279–80, 320–1 morality as  258, 262–6, 269–70, 274–5, 279–81 psychology as natural  8–9, 55–7, 61n.20, 63n.21 Second Analogy  15–16, 52–3, 66–7, 167, 230, 242–3, 243n.74, 284, 324, 331–2 self see absolutely necessary inevitable rational self determination, apperception, cognition of the self, empirical (phenomenal, spatiotemporal) character, free (absolutely spontaneous) self determination, intelligible (non-sensible, moral, noumenal) character self-deception  147–50, 278–9, 283, 286n.69 Sellars, Wilfrid  113, 161, 165, 186–7, 197–202, 211, 244–5, 348n.17 Sgarbi, Marco  199n.81 Sider, Theodore  218 skepticism about morality  7, 109–19, 122–4, 126–9, 273–85 and philosophical methodology  283–4, 318–23 Cartesian  56n.8, 283–4, 319–21 genealogical (etiological)  106, 109–19, 122–4, 126–9, 222–5, 249–50 Humean  215–20, 237, 239–43, 279–80 soul as immortal  1, 253–5, 259–60, 288–90, 292–5, 298–300 as interacting with body  44n.76, 195n.72

398 Index soul (cont.) as subject to natural necessity  25–6, 56–7 as topic of rational psychology  183–7 See also apperception, cognition of the self spontaneity absolute see absolutely necessary inevitable rational self determination, autonomy, causality of freedom (noumenal, supersensible), free (absolutely spontaneous) self determination, freedom of imagination, freedom of thought (epistemic freedom), freedom of will relative see freedom of a turnspit standpoint see deflationary (anti-metaphysical) interpretations, empirical standpoint, normative standpoint Stich, Stephen  228 Srinivasan, Amia  103, 110n.21, 112n.23, 127n.41, 249 Stang, Nicholas  84nn.18–19, 240nn.68, 70 Strawson, Galen  210n.10 Strawson, P.F.  4, 46n.78, 194–5, 348 Street, Sharon  106, 109–10, 112n.23 Stroud, Barry  240–1 sublime  354n.27, 356 substance as topic of rational psychology  185–7, 246–8 metaphorical (analogical) concept of  69–70, 92, 248, 295, 357–8 objective practical concept of  295 theoretical (schematized) concept of  69, 214–15, 229, 279–80 See also categories Sussman, David  280n.55 teleology  124–5, 146–7, 363 thing in itself see appearances and things in themselves (noumena, intelligible objects), noumena, transcendental idealism Third Antinomy  15–18, 22–32, 38–42, 48–55, 257 Timmermann, Jens  150n.43, 156n.62, 282n.58 Tolley, Clinton  39n.66 Tooley, Michael  218nn.37–38 Tooley, Michael and Sosa, Ernest  218n.38 transcendental idealism as connected to epistemic incompatibilism 229–36 as necessary but insufficient for freedom  2, 15–18, 22–37, 48, 51–5, 257, 296–7, 301–2, 306, 309–10 See also appearances and things in themselves (noumena, intelligible objects), noumena, transcendental realism

transcendental realism see appearances and things in themselves (noumena, intelligible objects), dogmatic empiricism (naturalism), dogmatic rationalism, naturalism as transcendental realism, transcendental idealism understanding as discursive intellect  75–8 See also autonomy of understanding, categories, freedom of thought, rational (objective, epistemic) necessity unity of reason  323–4 Van Cleve, James  24n.31, 33nn.49–50, 46n.78 van Inwagen, Peter  137n.18, 153n.51 vice epistemic  172, 176–80, 202n.83, 206–7, 247–8 moral  64–5, 67–70, 142–53, 202n.83, 206–7, 283 Vilhauer, Benjamin  73n.44 virtue epistemic  172, 176–7 moral  64–70, 146–53, 254–5, 259n.17, 292–3, 332, 355–6 Walker, Ralph  73n.44, 211–12, 244–7 Wallace, R. Jay  4, 59n.14, 60n.17, 137–8 Walsh, W.H.  261–2 Ware, Owen  282n.59 Warren, Daniel  33n.48, 37n.61, 103–4, 258n.12 Watkins, Eric  3n.4, 24n.32, 36n.56, 40–1, 41n.72, 42n.74, 64, 66–7, 266n.31 Watkins, Eric and Willaschek, Marcus  181n.43, 189n.64 weal/woe 151–3 White, Roger  113n.25, 128, 209–11, 222–6, 249–50, 313–14 Widerker, David  158n.66 will as pure practical reason and willkür 93–7 See also absolutely necessary inevitable rational self determination, autonomy of practical reason (will), causality of freedom, deed, determination of finite wills by reason, free (absolutely spontaneous) self determination, freedom of thought as distinct from freedom of will, freedom of will, good will, holy (perfectly rational) will, incompatibilism about freedom of will, knowledge of transcendental freedom of will, moral

Index  399 (purely rational) motive (motivation), noumenal (absolute, moral) contingency, practical-causal determination (PCD), practical concept of causality, proof of free will by practical reason, subjective practical maxim Williams, Bernard  15–16, 114n.26, 119n.30, 153n.53, 168 Williamson, Timothy  320n.23 Willaschek, Marcus  29n.41, 77n.6, 93n.29, 96n.33, 150n.43, 155n.61, 253, 254nn.3, 4, 255n.6, 258n.13, 259, 264–5, 268n.34, 269n.37, 274n.49, 279n.54 Windelband, Wilhelm  103–4 Wiggins, David  167–9, 175, 209–11, 224

Wolf, Susan  137nn.17, 18, 140n.25, 145–6, 156–7, 175n.31 Wolff, Michael  260n.18, 264n.27, 266n.30, 281n.57 Wolff, Robert Paul  167n.12, 234n.64 Wood, Allen  5n.7, 17n.3, 55n.5, 67n.38, 131n.2, 143n.29, 146n.35, 152n.49, 154n.56, 155n.61, 208–9, 253, 259, 273–4, 276n.53, 285–6, 288–9, 292n.79, 306n.7 Woodward, James  217n.36, 315n.22 Wuerth, Julian  96n.33, 186n.58, 245nn.77, 78 Young, J. Michael  348n.18 Zuckert, Rachel  343–5, 346n.16, 358n.32