Rational Causation 9780674065338

Citation preview

Rational Causation

Rational Causation ERIC MARCUS

Harvard University Press Cambridge, Massachusetts, and London, England 2012

Copyright © 2012 by the President and Fellows of Harvard College All rights reserved Printed in the United States of America Library of Congress Cataloging-in-Publication Data Marcus, Eric, 1968– Rational causation / Eric Marcus. p. cm. Includes bibliographical references and index. ISBN 978- 0-674- 05990- 0 (alk. paper) 1. Causation. 2. Agent (Philosophy) 3. Act (Philosophy) BD530.M37 2012 122—dc23 2011021301

I. Title.

For my parents, Norman and Maria, and my wife, Lydia

Contents

Introduction

1

1

Rational Explanation of Belief 1. q, so p

14

18

1.1 A New Variant of Moore’s Paradox 18 1.2 The Difference between Assertion and Demonstration 20 1.3 Demonstrations as Expressions of Believing-for-a-Reason 22 1.4 The Source of Moore-Paradoxicality 24 1.5 Believing-for-a-Reason Is Not a Process 28 1.6 Defending the Expressibility Thesis from Purported Counterexamples 29 1.7 On the Significance of the Expressibility Thesis 32

2. S Believes That p Because q

36

3. S Believes That p Because S Believes That q 3.1 The Simultaneity Constraint 3.2 The Epistemic Character of Believing-for-a-Reason 45

4. Rational Abilities

42

43

50

4.1 Three Species of Disposition 51 4.2 Differences between the Species 54 4.3 The Ability Underlying Believing-for-a-Reason

58

viii

Contents

5. Anti-Psychologism about the Rational Explanation of Belief 61

2 Rational Explanation of Action

65

1. Acting-for-a-Reason as Practical Thought 2. Objections

68

75

2.1 Objections to the Expressibility of Actions and Reasons for Action 75 2.2 Objections to the Identification of Evaluation with Intending to Act 79 2.3 Objections to the Identification of Evaluation with Acting 86

3. Instrumental Teleological Explanation

92

4. Anti-Psychologism about the Rational Explanation of Action 100 4.1 The Equivalence Thesis 101 4.2 Explaining the Rational Role of the Psychological Non-Psychologistically 107

3 (Non-Human) Animals and Their Reasons 1. Animals Are Responsive to Reasons

116

119

2. Animal Responsiveness to Reasons Is Epistemic 3. Objects of Knowledge versus Objects of Belief

120 122

4. Evidence Supporting Animal Belief Better Supports Animal Knowledge 130 5. An Argument against Animal Belief 6. Animal Agency

134

137

7. Explaining Belief versus Explaining Knowledge

139

8. Aside on Why Human (But Not Animal) Perception Is Conceptual 141

4 Rational Explanation and Rational Causation 1. Causation and Rational Explanation 1.1 Synthesizing Causal Concepts 1.2 Rylean Conceptual Analysis 1.3 Davidson’s Central Argument

146 148 151

146

145

Contents

ix

1.4 Davidson on the Relation between Causation and Explanation 154 1.5 Steward on the Relation between Causation and Explanation 160 1.6 Kinds of Causation 164

2. Rational Causation

167

2.1 Rational Causation as the Manifestation of Rational Abilities 167 2.2 Objections 174

5 Events and States

183

1. Objects, Events, and Sortals 1.1 Object-Sortals 1.2 Event-Types

185

185 190

2. States and Events-in-Progress

196

2.1 States 196 2.1.1 States Are Dissective, Not Unitary 2.1.2 Two Kinds of Persistence 202 2.1.3 States Are Negatable 205 2.1.4 States Are Mass-Quantified 207 2.2 Events-in-Progress 212

197

6 Physicalism

222

1. Physicalist Arguments Foiled

224

1.1 The Causal Analysis of Mental Concepts 225 1.2 The Principle of the Nomological Character of Causality 229 1.3 The Causal Completeness of the Physical Realm

2. Physicalist Positions Refuted

237

2.1 Mental Events 238 2.2 Mental States and the Doctrine of Token-Identity 2.3. Mental States Are Not Physical 247 2.4 Mental Facts 252

3. Supervenience

231

246

253

Acknowledg ments

261

Index

263

Rational Causation

Introduction

This book examines how we explain what someone thinks or does by citing her reasons. How do such explanations work? And what does the fact that they work tell us about the nature of reality? Contemporary efforts to address these questions are often motivated in part by the concern that our ordinary conception of rationality contains at least a kernel of outdated science or supernaturalist ideology, as if a completely unguarded articulation of this conception would describe a ghostly presence animating our bodily machinery, meditating on the sensory messages received from perceptual organs and subsequently orchestrating physical motions on the basis of its ethereal calculations. The task of a philosophically sophisticated theory would then involve a kind of conceptual exorcism, a reconstruction of our responsiveness to reasons that is at least consistent with it being just another mechanical phenomenon. But I argue that the underlying metaphysical anxiety is rooted in a series of mutually reinforcing errors, ones whose correction makes the reconstruction unnecessary. Centrally, philosophers have failed to take seriously enough the idea that

2

Rational Causation

rational explanations postulate a distinctive sort of causation, which is neither efficient, physical causation nor some sort of telekinetic analogue. Recognizing it yields a theory of rational mindedness and agency that rejects naturalism, yet is untainted by supernaturalism. The specific forms that naturalist accounts of mind and agency take are determined to a large degree by two assumptions, which I consider here by way of introducing some of the book’s overarching themes. The first concerns the role that psychological concepts play in rational explanations; the second concerns the relation between rational explanations and the causal structure of the world. Together, these assumptions sustain the nearly universal opinion that believing and acting for a reason is fundamentally a matter of the efficacy of psychological states, states that are realized in some sense by physical states. I call this view ‘Psychologism.’ Philosophers of mind and action tend to treat as fi rmly grounded in ordinary practice the idea that giving someone’s reason is ultimately a matter of appealing to the fact that she is in a certain psychological state. Ordinarily, however, to cite someone’s reason by explicitly invoking psychological categories is to suggest that something is amiss. We would typically say that a man sinks his neighbor’s boat, for example, “because he believes that his neighbor is a spy,” as opposed to “because his neighbor is a spy,” if we were doubtful of the man’s epistemic credentials—if, say, we knew his neighbor was not in fact a spy. Free of such doubts, we dispense with psychological framing. We say simply: “She is setting glue-traps because her attic is infested with rats” or “He’s going to the veterinarian because his dog has been limping.” But if these examples are representative, why is the more psychologistic view of rational explanation so widely held? The argument from illusion provides a helpful model for understanding one source of the appeal of the orthodox view. Perceptual experience, that argument is thought to show, cannot consist in genuine cognitive access to the world, since a phenomenologically identical experience might not provide any such access. More broadly, many philosophers hold that the nature of a specific cognitive ability (e.g., a perceptual ability or an ability to reason) should be identified in terms of what all of its phenomenologically indistinguishable exercises— both successful and unsuccessful—have in common. Here is how the broader thesis applies in

Introduction

3

our case: because (a) we can always give someone’s reasons by appealing to her psychological states, (b) we cannot always do so by appealing to what she takes to be the facts, and (c) she cannot distinguish between believing p and believing p truly on phenomenological grounds, it is tempting to treat the psychological guises of rational explanations as more deeply revealing of the ability to believe or act for a reason. This is the first of the two deeply entrenched errors this book corrects; it is the topic of Chapters 1 and 2. My approach, by contrast, takes seriously the idea that the abilities exercised in individual cases of theoretical or practical reasoning should be understood primarily in relation to their successful exercises— exercises adequately characterized in non-psychological terms. Nonpsychological guises of rational explanations are, in an important sense, more fundamental than their psychological counterparts. Standing in the way of this approach is the thesis, prevalent since the publication of Donald Davidson’s “Actions, Reasons, and Causes,” that causal connections make rational explanations of belief and action true.1 Describing the specific character of this obstacle is tricky, for the substance of the thesis hinges crucially on what precisely causal connections are. Indeed, on one interpretation—viz., the unorthodox one developed here—I think that the thesis is true. As intended and as typically understood, however, Davidson’s thesis is thought to plant rational explanations firmly (albeit ‘non-reductively’) in the physical causal order, a causal order that Davidson, along with many others, identifies in terms of the idea of a natural law. I use the (admittedly dated) term ‘efficient’ to describe this sort of causation. The second deeply entrenched error I challenge is the assumption that effi cient causation makes rational explanations true. The thesis of “Actions, Reasons, and Causes,” so understood, makes it all but impossible to avoid thinking that it is the psychological rather than the non-psychological guises of rational explanation that are fundamental. Here is roughly how this line of thought progresses. Even if we cite events concerning the world ‘beyond the skin’ in rationally explaining a person’s thoughts and actions, such events must initiate causal sequences whose later members are bodily events. Near the end of such sequences, 1. Donald Davidson, “Actions, Reasons, and Causes,” in Essays on Actions and Events (Oxford: Clarendon Press, 1980), 3–20.

4

Rational Causation

we find the very psychological events (or their neural ‘realizers’) to which we might have appealed with the same explanatory accuracy in those very cases, and to which we must appeal if the relevant extra-bodily circumstance is a figment of the person’s imagination, as in the case of our spy hunter. The influence of the world beyond the skin seems both to be inessential to a belief or action’s being based on a reason and also to be dependent on psychological intermediaries, which thus appear as the crucial elements in the underlying causal story. This view of the nature of rational causal influence gives rise to a series of seemingly intractable problems. Here are two examples that loom large in the literature. First, there is the problem of epiphenomenalism about the mental. Since psychological events are thought not to be ‘type identical’ to physical events, it seems as if psychological events, or at least the psychological properties of those events, cannot be genuinely efficacious. The orthodox view, it is widely feared, does not credit the mental per se with any causal power. Second, there is the problem of deviant causal chains. Not every belief-desire pair that rationalizes an action corresponds to the reason the action was performed. “A climber,” Davidson observes, “might want to rid himself of the weight and danger of holding another man on a rope, and he might know that by loosening his hold on the rope he could rid himself of the weight and danger. This belief and want might so unnerve him as to cause him to loosen his hold.”2 Proponents of the orthodox view have struggled to fi nd a principled way to exclude such deviant cases. The portrayal of rational connections as efficient-causal fundamentally misconceives the nature of rational causation. This portrayal is appealing, firstly, because it fits with the conception of rational explanation alluded to above, according to which the central cases are those in which an action or belief is explained by explicitly citing the agent’s or believer’s state of mind. It is appealing, secondly, because it fits with the widely accepted idea that the only real principles corresponding to non-constitutive explanations of the way things are in the world are efficient-causal. Because this conception of rationality is fed from these two sources, dispelling it requires more than the de-psychologizing of rational explanation— 2. Donald Davidson, “Freedom to Act,” in Essays on Actions and Events, 79.

Introduction

5

the first goal of this book. It also requires allowing that there are other kinds of causation. The second goal is to say what, quite generally, it means to talk of different kinds of causation and, more specifically, what distinguishes efficient causation from the sort that I argue grounds rational explanation. Achieving the first goal would refute the ubiquitous ‘belief-desire’ model of understanding rational explanation. Achieving the second would amount to the end of physicalism in the philosophy of mind. My project here is ambitious. Its ambition is mandated by the symbiotic relation between the psychologistic approach to rationality and physicalism in the philosophy of mind. To attack the former while leaving the latter in place would come off as a metaphysically empty exercise in conceptual analysis. Even if the non-psychological guises of rational explanation are conceptually primary, they will seem to be metaphysically secondary so long as one holds a physicalist conception of causation. But a challenge to physicalism about causation must proceed from a specific kind of explanation that (a) is ill suited to naturalization and (b) can provide the resources for specifying another kind of causation. In what remains of this introduction I will sketch—in its broad, programmatic outlines and putting aside myriads of caveats, clarifications, and likely objections—the view of rational explanation that I argue satisfies these requirements. The guiding idea is that to believe or act for a reason is to represent the believed proposition or the performed action as inheriting a good-making status from another proposition or action. The basic elements of theoretical reasoning, propositions, are potential bearers of the status ‘to be believed’; the basic elements of practical reasoning are potential bearers of the status ‘to be done.’ Whereas theoretical rationality is the ability to believe what is to be believed on the basis of something else that is to be believed, practical rationality is the ability to do what is to be done on the basis of something else that is to be done. Notwithstanding this contrast, a higher level of abstraction yields an overarching similarity: successful exercises of each are the making of theoretical or practical inferences that preserve the relevant good-making status. These are the ideal cases of believing- and acting-for-a-reason. To identify believing- and acting-for-a-reason with subjects’ representing (in a sense to be articulated) inferential connections is to view such representings as essentially self-conscious. For example, Poirot’s believing

6

Rational Causation

that the butler did it because everyone else has an alibi is necessarily expressible by Poirot in something like this way: “Everyone else has an alibi, so the butler did it.” I call this sort of statement a demonstration. But one cannot, on pain of Moore-Paradoxicality, think this and yet disavow the thought underlying, “I believe the butler did it because I believe that everyone else has an alibi.” The latter is just an explicitly self-referential manner of expressing the same state, much as “the butler did it” and “I believe the butler did it” can express the same belief. Poirot represents the to-be-believed-ness of the butler’s having done it as following from the tobe-believed-ness of everyone else’s having an alibi. Because this representing is a way of thinking about the propositions and the relation between them, it is expressible as a demonstration. Because it is also a recognition of a specific doxastic requirement, it is expressible as a rational explanation of his own belief. The inextricability of these two aspects accounts for the ‘transparency’ of believing-for-a-reason. Ideally, this sort of representing amounts to a subject S’s knowing a fact p on the basis of her knowledge of another fact q. The ability underlying believing-for-a-reason is an epistemic one. In fully successful cases, we can explain why S believes that p by citing q itself. One sort of example of a less-than-fully-successful exercise involves thinkers who do not know that q. In such cases, one is limited to giving explanations that explicitly cite the subject’s mind, using forms of words such as “S believes that p because S believes that q,” which does not entail that S knows that q. The difference between this mind-citing explanation and the corresponding world-citing one (“S believes that p because q”) is typically just a difference in how far the explainer thinks S deviates from the epistemic ideal. The psychological guise of the explanation is thus not more metaphysically penetrating than the non-psychological guise but is instead parasitic on the latter, which is central. If the truth of rational belief-explanations consists in the sort of doxastic representing that I have sketched, what is the relation between S’s believing that p and S’s believing that q, where q is S’s reason for believing that p? Evidently, if someone asks why S believes that p, we can answer by citing the fact that she believes that q. The world is the way it is in the former respect as a result of its being the way it is in the latter respect. This reply does not give a constitutive account; the point is not that what it is for S to

Introduction

7

believe that p is for her to believe that q. Rather, a causal connection is thereby revealed to a comprehending inquirer. But this sort of causation is the manifestation of the rational ability just discussed, in terms of which it must therefore be understood. What it is for S’s believing that q to cause S’s believing that p is for S to exercise that ability. And this confers upon the causal connection a raft of distinctive features, including this one: S’s believing that q can only be a cause of S’s believing that p if those facts pertain to a single time. S’s believing that q is not part of the causal history of S’s believing that p, since what S represents (at a single time) is that the now-tobe-believed-ness of p is a consequence of the now-to-be-believed-ness of q. Given the conceptual kinship of abilities and dispositions, the thesis that the efficacy proper to believing-for-a-reason consists in the exercise of an ability invites a comparison to dispositional analyses of causation. According to the simplest version, “A causes B when A is the stimulus of some disposition and B is the corresponding manifestation.”3 Solubility is the disposition to dissolve when wet. Thus, putting a soluble tablet into water can cause its dissolution. On this model, a cause must be external to, since it is the cause of, the manifestation of the disposition. But in the case of rational abilities quite generally, it is, I argue, dubious whether anything plays the role of stimulus. More to the point, rational causation is internal to the manifestation of the ability under discussion here. The analogue to the tablet’s dissolving is a causal connection between q and S’s believing that p (when S knows that q). On the view I defend, the relevant ability’s manifestation is not the effect of a rational cause; rather, it is itself a rational-causal connection. The proposed account of rational action-explanation has the same general shape. To act for a reason is to represent the to-be-done-ness of an action as following from the to-be- done-ness of another action. (This is intended as a model for understanding instrumental action.) Because acting-for-a-reason consists in an agent’s so representing, agents can say, as Anscombe emphasized, what they are doing and why, yet not on the basis of observation or evidence. The ability to do what is to be done as a consequence of another action’s being to be done is, like the corresponding 3. Alexander Bird, “Antidotes for Dispositional Essentialism,” in The Metaphysics of Powers, ed. A. Marmodoro (New York: Routledge, 2010), 160–168, 161.

8

Rational Causation

theoretical ability, necessarily self-conscious. So part of what it is for S to f in order to y is to be able to ‘just give’ the following sort of explanation of her own action: “I am f -ing because I am y-ing.” This style of explanation— in which the unfolding of the means-action is explained in terms of the unfolding of the end-action—presupposes a degree of practical success, since an agent might represent, say, swimming to the moon as to-bedone, when no such action is in progress. To explain his behavior, we cannot say “He’s donning his bathing-suit because he’s swimming to the moon” but must instead switch to something like “He’s donning his bathing suit because he thinks he’s going to swim to the moon” or “because he intends to swim to the moon.” Here, as in the theoretical case, the point of invoking psychological vocabulary is to explain a defective exercise of a rational ability. But the psychological guises of rational action-explanation are for this very reason less revelatory of the underlying ability than the non-psychological guises. Central instances of acting-for-a-reason involve an agent’s performing one action because she is performing another, which can be captured by the following sort of explanation: “Dara is tasting all the desserts because she is reviewing the restaurant.” This explanation is, I argue, made true by a causal connection between the facts designated by the explanans and explanandum, a causal connection that consists in the exercise of the practical rational ability we have been discussing. The cause, then, is her reviewing the restaurant, and the effect is her tasting all the desserts. Rational causation has a teleological structure in the practical case, a thesis that requires careful articulation. To forestall the most likely misapprehension, the thesis does not, in my rendering, amount to the implausible idea that a future state of affairs somehow pulls the agent in its direction. Teleological explanations do not, even on the surface, postulate future states of affairs as action-explainers. We say that Dara was tasting all the desserts not because she later fi nished reviewing the restaurant but because she was, at the time of eating, engaged in the project of reviewing the restaurant. Similarly, we can explain why someone’s right ventricle was pumping some blood through her pulmonary artery by citing the fact it was, at the time of through-the-artery-pumping, also pumping the blood to the lungs. In instances of teleological causation, the unfolding of an event-whole is a cause of the concurrent unfolding of its parts.

Introduction

9

If the worldly facts designated by the explanantia of standard rational explanations are causes of what they explain, it is easy to account for the following widely accepted idea: giving someone’s reason for believing or acting explains by citing something that the individual took to cast the belief or action in a favorable light—to justify it, in a thin sense. This is the Equivalence Thesis. Intuitively, what justifies is precisely the sort of thing that rationally explains— e.g., the fact that everyone else has an alibi explains and justifies Poirot’s belief that the butler did it. But the Equivalence Thesis is a problem for Psychologism. Reasons, on the standard psychologistic model, are physically realized psychological states or facts; these are what cause and justify. But this seems wrong: Poirot does not infer the butler’s guilt from his own doxastic circumstances—i.e., it is not the fact that he believes that everyone else has an alibi that establishes the butler’s guilt but rather simply the fact that everyone else has an alibi. To accommodate this datum, some more sophisticated version of Psychologism is needed. There is, for example, the view that although we rationally explain by citing reasons, and reasons are (typically) non-psychological, it is not the reasons themselves that explain, but rather facts involving psychological attitudes toward reasons. Everyone else’s having an alibi is Poirot’s reason for believing that the butler did it, but the former does not, on this view, rationally explain the latter. There is also the view that although reasons are (typically) non-psychological facts that justify and explain, it is facts involving psychological attitudes toward reasons that cause. So our sample rational explanation does not posit the fact that everyone else has an alibi as a part of the origin of Poirot’s belief. Each of the psychologistic views is implausible in its own way. The standard view denies that worldly facts justify; the first sophisticated variant denies that they explain; the second sophisticated variant denies that the worldly facts that rationally explain are the ones that actually make a difference. Most, however, will be doubtful that there is a genuine alternative along the lines I am suggesting. “Sui generis non-physical causation,” at first blush, sounds like a fancy name for spiritual pushing, a new label on the same old Cartesian bottle. There are also the many arguments that philosophers take to establish physicalism in one form or another, arguments based on, among other things, causal analyses of mental concepts, the Nomological Character of Causality, the causal closure of the physical

10

Rational Causation

realm and the supervenience of the mental on the physical. For the most part, my replies to these arguments take the following form. There are, on the one hand, the considerations that have persuaded many to think of rational explanations as causal, thereby linking causation to the mental. And then there are, on the other hand, putatively general causal principles and theories of causation that link it to the physical. I argue that the features of rational explanation that provide the evidence for the first link are, just to the extent that they provide such evidence, also grounds for thinking that the putatively general principles and theories are not after all fully general. The second link fails for rational causation. For example, there may be some truth to the idea that what it is to be in a certain mental state is to be in a state that plays a certain causal role. According to a simple version of functionalism, mental concepts are just specifications of such roles. If mental state M is simply the state of being apt to be the effect of A, B, and C and the cause of X, Y, and Z, then the ontology of M is determined by the nature of the thing that occupies just that spot in the causal order. And we have learned that it is physical states that in fact play causal roles. Hence mental states are realized by physical states. However, the plausibility of causal analyses of mental concepts derives mostly from the roles that they play in rational explanations of thought and behavior. And these explanations postulate distinctively rational efficacy. To believe that the butler did it is to be in the sort of state that might be a rational consequence of something like this: believing that everyone else has an alibi. The obtaining of such a causal connection is just the subject’s representing the to-be-believed-ness of everyone else’s having an alibi as a source of the to-be-believed-ness of the butler’s having done it. But to say that Poirot believed the butler did it because (i.e., as a causal consequence of the fact that) he was in such-and-such a physical state is either a different and false rational explanation (Poirot did not draw any conclusions from the configuration of his “little grey cells”) or it is not a rational explanation at all—it is not, that is, an explanation that even purports to pinpoint the perceived source of the to-be-believed-ness of the butler’s having done it. Either way, the physical state does not play the role specified by a good causal analysis of believing that the butler did it. The fact, then, that causal specifications of mental concepts are possible (if it is a fact) confers no credence on, but (properly articulated) tells

Introduction

11

against, the idea that networks of (rationally) causally interrelated mental states are realized by networks of (efficiently) causally interrelated physical states. The view I defend is philosophically rather exotic, as the reader will no doubt have gathered from the foregoing summary. The more nuanced and circumspect elaboration that follows will not alter this impression, but it will, I hope, demonstrate that my account is comparatively faithful to common sense. Whereas the familiarity of psychologistic theories makes them easy to digest (philosophical comfort food, if you like), they suggest a profound gulf between ordinary understanding and reality. The whole point—the beauty, if you have a taste for it— of naturalistic approaches to apparently recalcitrant domains is that they show that (putatively) ontologically queer powers are nothing over and above the hidden operation of physical mechanisms. The operations are hidden in the sense that those who successfully employ explanations proper to those suspicious domains need have (and typically do not have) any knowledge of the ‘underlying’ causation. Although Plato might have known perfectly well why Socrates refused the opportunity to escape prison, he was utterly clueless as to the real causal mechanisms that linked Socrates’ beliefs with this refusal. The knowledge-why that we credit to Plato is at best highly superficial, reminiscent of the ‘knowledge’ of reality possessed by those mythical cave-dwellers who mistake shadows of objects for the real thing. I contend, on the contrary, that Plato’s grasp of the rational explanation for Socrates’ omission constituted his possession of a deep and full understanding of the efficacy that made that explanation true. This is not to deny, of course, that without neurophysiological causal processes, there would be no human thought. But rational explanations reveal a different sort of efficacy, efficacy that is perfectly natural even if it is not naturalistic. The fi rst two chapters develop and defend the conception of rational belief- and action-explanation as just sketched. Such explanations work by appealing to the exercise of theoretical and practical rational abilities, exercises in which the subject represents the to-be-believed-ness of a proposition or the to-be-done-ness of an action as derived (respectively) from that of another proposition or action. One is the ability to believe what is to be believed because it is to be believed; the other is the ability to do

12

Rational Causation

what is to be done because it is to be done. These accounts accommodate features of the phenomena with which psychologistic approaches tend to have trouble, establishing tight connections between them: the exercises of these abilities are self-conscious, in the sense that a person can selfascribe her believing- or acting-for-a-reason authoritatively, yet not on the basis of observation or evidence. World-citing guises of rational explanations are primary, even though mind-citing explanations are correct even when the corresponding world-citing explanations are not. Reasons can both justify and explain beliefs and actions. My emphasis on the essential self-consciousness of believing- and acting-for-a-reason might seem to render my view unable to account for the many similarities between the way we explain human thought and behavior and the way we explain the thought and behavior of non-rational animals. In Chapter 3 I respond to this charge. We can successfully and truly explain the thought and behavior of animals by using what are— at a certain level of abstraction, at least—the same explanation forms as in the human cases. Both my dog and I might, for example, open the pantry door “in order to eat some chips.” But that is consistent with the following claim: what it comes to for a ‘brute’ to think and act for reasons is different from what it comes to for a person to do so. The basic difference is that human capacities for rational thought and action involve representational powers that animals (as far as we know) lack. Humans represent propositions to themselves as to be believed and represent actions to themselves as to be done. Non-human animals do not. Hence the etiology of their thought and behavior—for all of its similarities to that of human thought and behavior—must be understood differently. Up to this point, I will have discussed only my first goal: to describe the fundamentally non-psychologistic character of rational explanations. My second goal is to show that such explanations are made true by distinctively rational causation. The analysis of rational explanation in the first half of the book should fit as snugly with the metaphysical outlook articulated in the second as the classic Humean theory of motivation fits with naturalistic metaphysics. And so I defend, in Chapter 4, the idea that there are different kinds of causation. I identify the minimal explanatory criteria associated with any causal claim and go on to show the possibility of distinct forms of causation, each associated with how the minimal explanatory

Introduction

13

criteria common to all are discharged. An appeal to natural law is one way in which this happens; but other principles can play the same specifying role. In particular, the notion of a rational ability can play such a role. Once this possibility is recognized, Davidson’s influential argument in “Actions, Reasons, and Causes” can be seen as trading on an ambiguity between a generic sense of ‘cause’ and a specific sense that is closely bound up with the idea of a natural law. Rational causation is distinct in virtue of the fact that it manifests not natural laws but the exercise of rational abilities. It is causal connections consisting in the exercises of these abilities that make rational explanations of belief and action true. My account of the character of rational explanation and causation is couched in the language of metaphysical categories, especially events and states. But my conception of these categories has very little connection to the way they are generally understood in the literature in the philosophy of mind. My view will thus not be fully intelligible until I have explained what I take events and states to be: the goal of Chapter 5. I also consider differences between the way issues of identity and individuation arise with respect to events and states, focusing especially on the question of whether events and states are particulars. This is essential preparation for the discussion of mind-body identity theories in Chapter 6. One cannot assess arguments for or against physicalism about mental states or events without clarifying the nature of events and states more generally. In Chapter 6 I argue, along the lines suggested above, that the rational etiology of acting- and believing-for-a-reason undermines the most powerful arguments for physicalism about the mind. I show, furthermore, that the mental is distinct from the physical across a broad range of metaphysical categories: events, states, and facts. Also, I contend that while my view is consistent with the supervenience of the mental on the physical, the significance of supervenience is overblown. Indeed the very fact of this consistency demonstrates that supervenience does not guarantee an ontology friendly to naturalistic requirements. The most common complaint made against philosophers who hope to resist the naturalistic approach to the mind is that they can succeed only at the cost of accepting dualism, epiphenomenalism, or eliminativism. I show, by example, that this is not so.

1 Rational Explanation of Belief

A rational explanation of belief cites a believer’s reasons and, in so doing, appeals to the exercise of a rational ability. To understand the way such explanations work, one must understand the ability. It is, I argue, a selfconscious ability to know facts on the basis of other known facts. To put it in the quasi-technical language that I develop over the course of this chapter, it is the ability to represent a known fact as a proposition whose to-be-believed-ness establishes the to-be-believed-ness of another proposition, which represents a further, thereby known fact. My overarching methodology—to study the form of explanation by investigating the underlying ability—proves itself by yielding a unified treatment of various phenomena central to believing-for-a-reason. My approach to the topic might seem counterintuitive. How could believing-for-a-reason fundamentally be a matter of a thinker’s epistemic engagement with the objective world? After all, beliefs based on reasons routinely fall short of this standard. Nothing could be more common than a person’s believing something on the basis of what is mistaken for a

Rational Explanation of Belief

15

fact— on the basis of what is merely believed. Building the epistemic dimension into an account of what it is to believe for a reason would seem to confuse the descriptive project of studying the rational etiology of beliefs and a normative inquiry into their epistemic properties. But to understand exercises of an ability, be they successful or unsuccessful, one must understand the ability itself—what it enables its possessor to do. Thus, those explanations that appeal to its successful exercises are conceptually central. The ability underlying believing-for-a-reason enables the thinking subject to know facts—typically facts about the extra-mental world— on the basis of other facts. So rational explanations that cite the world are the norm, whereas those that cite the mind— explanations that leave open the possibility of epistemic failure— are deviations from the norm.1 The former provide a better understanding of the phenomenon of believing-fora-reason than the latter. The centrality of world-citing (as opposed to mind-citing) rational explanations would seem to challenge (at least by undercutting an easy way of motivating) the psychologistic idea that believing-for-a-reason is, at bottom, a matter of goings-on inside the body of the reasoner. An accommodation is, however, not hard to envision: thinking subjects have a propensity to be influenced doxastically not just by proximate, mental causes but also by distal, worldly causes. In the right circumstances, the efficacy of the latter culminates in knowledge based on non-psychological facts. Perhaps this is paradigmatic of believing-for-a-reason. If so, world-citing rational explanations would be conceptually central. Nonetheless, the core phenomena of rationality would be the operations of the cognitive mechanisms that mediate the world’s influence. But I argue that this naturalistic accommodation misconceives in a basic way the relation between a worldly-reason and the belief on which it is based. The difference between world-citing and mind-citing explanations is not a matter of proximity. It’s not that the former explains something inner by appealing to something outer, whereas the latter explains some-

1. The mind is, of course, part of the world. Here I aim to invoke a familiar distinction between, for example, an explanation that appeals to the fact that it’s raining and one that appeals to the subject’s belief that it’s raining. I will say what this difference ultimately comes to in section 3.1 of this chapter.

16

Rational Causation

thing inner by appealing to something else that is inner. The point of rational explanations is not to explain internal states at all. What, then, is the right strategy for explaining the difference between the two guises of rational explanation? My alternative is based on the following idea. To say of someone that she holds a belief that p for a reason q is to say that she is in a state expressible in world-directed pronouncements of the form “q, therefore p,” “q, so p,” or “p, since q”—demonstrations. Someone who demonstrates expresses not just her believing p and her believing q, but her thinking of q as the source of p’s belief-worthiness. The rational explanation of someone else’s belief reports an inferential connection that the believer discerns between propositions she believes. To put in causal terms (our explicit topic in Chapter 4), q’s rational efficacy has nothing to do with a historical, perspective-independent causal tie between q and the relevant subject’s believing that p, but is rather a matter of the subject’s representing p as . q (as opposed to merely believing that q) is the cause of her believing that p just in case her representing q as to be believed amounts to knowledge that q. Psychologism is false because the rational-explanatory connection between a belief and the reason on which it is based has an ineliminable first-personal character. The scope of my investigation here is delimited by two aims. The first is constructive; the second is diagnostic. The constructive aim is to lay the groundwork for the metaphysical aspects of my view of rationality, which come to the fore in later chapters. The efficacy of reasons in relation to belief and action consists in the exercise of the abilities described in these first two chapters. The diagnostic aim is to present my view in a manner that makes plain its superior ability to account for the intuitive data that makes the orthodox, psychologistic approach to rational explanation seem correct. My alternative embraces the idea that rational explanations are psychological—their point, after all, is to tell us something about the mind of a particular thinking subject—but does so on the basis of a nonpsychologistic conception of the psychological. Here is how I shall proceed: I begin, in section 1, by identifying actualizations of the ability to believe for a reason through an examination of their characteristic linguistic ex-

Rational Explanation of Belief

17

pression in demonstrations. A rational explanation of S’s believing p for reason q attributes to S a state that she can express using a demonstration (e.g., “Everyone else has an alibi, so the butler did it”) and also by giving a rational explanation of her own belief (e.g., “I believe the butler did it because everyone else has an alibi”). This state consists in the speaker’s representing as to be believed one proposition as a consequence of the to-bebelieved-ness of another proposition. World-citing rational explanations (e.g., “Poirot believes that the butler did it because everyone else has an alibi”) are central. They are the topic of section 2. The epistemic requirement on such statements—Poirot must know that everyone else has an alibi— enables us to understand the character of mind-citing explanations, such as “Poirot believes that the butler did it because he believes that everyone else has an alibi.” I consider these in section 3. Rational explanations of belief that appeal to other beliefs are less epistemically demanding versions of rational explanations that appeal to worldly facts. The relation between these two guises of rational explanation reveals the normative character of believing-for-a-reason: to believe p for reason q without knowledge of q is an imperfect exercise of the underlying ability. In section 4 I elucidate the idea of a rational ability by comparing it to the general idea of a disposition. I adduce distinctive features of explanatory appeals to dispositions as evidence for my claim that rational explanations work by appealing to the exercise of a rational ability. An important part of the upshot of this discussion is that those guises of explanation that demand the successful exercise of an ability are more central to understanding it than those that are successneutral. Finally, in section 5, I show that these results undermine the orthodox, psychologistic view of rational explanation, which explains world-citing explanations in terms of mind-citing explanations, and the first-person point of view in terms of the third-person point of view. Before moving on, I note the following limitation of my discussion. My focus here is on cases of believing-for-a-reason that are inferential. There are other kinds of reason for belief. If a subject’s reason for believing that p is that she saw that p, remembers that p, or was told that p, then her belief that p is not (on my view) formed on the basis of an inference. As such, my account of believing-for-a-reason does not extend to these cases. Be that as it may, the inferential cases of believing-for-a-reason—‘reasoning’ in the narrowest sense of the term— are inarguably central to (and, as I argue in

18

Rational Causation

Chapter 3, distinctive of) human responsiveness to reasons. An attempt to extend the view propounded here to perception, memory, and testimony would divert us from this book’s central, already ambitious aims.

1. q, so p In this section, I isolate a kind of speech-act by means of which someone may express her taking some q as a reason to believe some other p. For example, “Everyone else has an alibi, so the butler did it.” Speech-acts of this form, demonstrations, are not, I argue, assertions and cannot be understood merely as the expressions of belief. They can be identified by the fact that they give rise to a variant of Moore’s Paradox. Unintelligibility threatens if one appends to the foregoing remark, “But I don’t believe that the butler did it because I believe that everyone else has an alibi.” Using this phenomenon as my lodestar, I argue that the underlying mental state, believing-for-a-reason, is constitutively tied to the possibility of expression both in the form of a demonstration and a rational explanation of one’s own belief. My goal in this section is not to give a reductive account of believing-for-a-reason—it is, on my view, sui generis—but to bring out some of its essential features. These features will loom large in later sections, in which I turn my attention to the attribution of believing-for-a-reason to others.

1.1 A New Variant of Moore’s Paradox Here are three examples of demonstrations: Everyone else has an alibi, so the butler did it. The lawn is healthy, so it gets watered frequently. The Steelers consistently win Super Bowls, so they are the best team.2

2. For the sake of simplifying my discussion, I stick to the “q, so p” construction. One equally well expresses this kind of thought using other constructions: e.g., “Everyone else has an alibi, therefore the butler must have done it”; “That lawn clearly gets watered frequently since it is in very good shape”; and “The Steelers are obviously the best team, because they consistently win Super Bowls.”

Rational Explanation of Belief

19

Unlike other statements of the form “q, so p” (e.g., “The plane was struck by lightning, so it crashed”), q in these statements is not what brought it about, or helped to bring it about, that p. Everyone else’s having an alibi did not make the butler do it. The lawn’s being healthy does not cause its being watered frequently. And consistently winning Super Bowls did not lead to the Steelers being the best team. These statements do not answer questions such as “Why is p true?,” or “How did it happen that p?,” or “What led to p?” Thus ‘q’ and ‘p’ do not stand in the relation of explanans to explanandum. A demonstration that takes the form “q, so p” is not an explanation of p. Still, it is plainly very closely connected to an explanation of why the speaker believes that p. And this is why it gives rise to a variant of Moore’s Paradox. Just as the original version of Moore’s Paradox sheds light on the nature of assertion and, ultimately, on the nature of belief, I will argue that my variant sheds light on demonstration and, ultimately, on the nature of believing-for-a-reason. Here is a Moore-Paradoxical statement: It is raining; but I don’t believe that it is raining. It would be hard to make sense of someone who said this. I do not mean simply that we would view it as a strange thing to assert and to think. I mean that it would take some work to discern an intelligible assertion at all and thus also to figure out what thought the utterance would express. A natural response would be brute incomprehension. Yet its absurdity does not seem to be the consequence of a contradiction between the conjuncts. Why should a conjunction of remarks, one about the weather and one about the speaker, pose such an interpretive challenge? The paradox is that sentences that are concerned with entirely different matters nonetheless conflict. Now consider: Everyone else has an alibi, so the butler did it; but I don’t believe that the butler did it because I believe everyone else has an alibi. This remark poses the same challenge. The first conjunct expresses the speaker’s taking it to be true that the butler did it because everyone else

20

Rational Causation

has an alibi. The speaker then denies, in the second conjunct, that her believing that everyone else has an alibi is among the reasons why she believes that the butler did it. It’s difficult to see what someone who utters this conjunction might be driving at, and yet, as in Moore’s original example, there is no contradiction between the conjuncts. The form of the original, then, is: Moore-Paradoxical Statement-Form (MP): p; but I don’t believe that p. And the form of the variant is: Variant of Moore-Paradoxical Statement-Form (VMP): q, so p; but I don’t believe that p because I believe that q.3 Note that there is nothing amiss about such statements where the first part is interpreted as an explanation of p. That is, there would be nothing absurd in someone saying, “This lawn is watered frequently, so it is healthy; but I don’t believe it is healthy because I believe it is watered frequently.” Thus, this variant of Moore’s Paradox can be used to distinguish demonstrations from explanations that use similar (or even the same) forms of words.

1.2 The Difference between Assertion and Demonstration Demonstrations are not assertions. To take issue with “Everyone else has an alibi, so the butler did it,” one says not that the whole thing is false, but that it’s bad reasoning or that the reasoning is sound but that it begins with a false premise. The statement as a whole does not express someone’s taking an affirmative stance toward a proposition. It might seem otherwise. It might seem as if the demonstration “q, so p” is just the same as asserting both that q and that the truth of q estab3. VMP must be distinguished from “p because q; but I don’t believe that p only because I believe that q.” Instances of that statement form are not absurd.

Rational Explanation of Belief

21

lishes (or helps to establish) the truth of p. But this is false. Consider this statement: “Everyone else has an alibi; and everyone else’s having an alibi entails that the butler did it; but I don’t believe that the butler did it because I believe that everyone else has an alibi; in fact I don’t believe that the butler did it.” This is, to be sure, odd. The speaker doesn’t believe something he knows that he is committed to believing. He cannot, perhaps, bring himself to believe that his close friend the butler did it. He’s admitting to being irrational. But there is no sense here, as there is in the case of MP and VMP, that the speaker is putting something forward and taking it back at the same time. Moore-Paradoxical utterances are not merely irrational or unusual; rather, they create the appearance of unintelligibility. Consider next the statement “Sam is unmarried, so available.” It seems this can be supposed—“Suppose Sam is unmarried, so available; what will you do then?”— and can form the antecedent of a conditional—“If Sam is unmarried and so available, then Mindy will be pleased.” This might be considered proof that the original statement is an assertion. But what is one supposing when one supposes that Sam is unmarried, so available? And what is the content of the conditional’s antecedent? This: (a) Sam is unmarried, (b) Sam is available, and perhaps also (c) Sam’s being available follows from his being unmarried. But to assert the conjunction of (a), (b), and (c) is not to demonstrate “Sam is unmarried, so he’s available.” For the following is not Moore-Paradoxical: “Sam is unmarried and available and his being unmarried entails his being available, but I don’t believe that he is available because I believe he’s unmarried.” Perhaps I am persuaded that Sam is available only by the fact that he is out with someone new every night. Again, this is odd but does not raise the specter of unintelligibility. By saying “Sam is unmarried, so he’s available,” I show not just that I believe Sam is unmarried, that I believe he is available, and that I believe there is an entailment relation, but that my belief that he is available is based (at least in part) on his being unmarried. This is more than another belief. It is something else, a state for which I use the quasi-technical expression ‘believing-for-a-reason.’ It stands to demonstration just as belief stands to assertion.

22

Rational Causation

1.3 Demonstrations as Expressions of Believing-for-a-Reason It is a truism that our utterances express our states of mind. Although no one thinks otherwise, there is substantial disagreement about precisely how a philosophical account of expression should go.4 I offer no account here, though what I say here may bear on the development of such an account. I begin by considering the phenomenon of expressive self-attribution— the way we can express our states of mind by explicitly stating that we are in them. Consider the case of beliefs and desires, for example. I can express my belief that the future is bleak by saying “I believe that the future is bleak.”5 Similarly, I can express my desire to eat a salami sandwich by saying “I want to eat a salami sandwich.” There is plainly a tie between the expressive character of these utterances and the fact that their subject is ‘I.’ “Arthur believes the future is bleak” counts as an expression of my belief regarding the future only if I know that I am Arthur. The essential first-personality of expression also emerges from the observation that self-attributions of mental states are not always expressions of those states. For example, if Harry believes he is jealous only because Audrey pointed out that he always tries to pick fights with Josh, then Harry does not express his jealousy by saying “It seems that I am jealous.” Here, he expresses his newfound belief that he is jealous but not his jealousy itself. When my self-attributions are expressions of the self-attributed state, I do not arrive at them merely by observing my own behavior, trusting my friends or psychoanalyst, or deducing them from generalizations about human nature. I can just say that I am in the state. This just saying is not a

4. For recent discussions of expression that are broadly congenial to my discussion here, see David Finkelstein, Expression and the Inner (Cambridge, MA: Harvard University Press, 2003), and Dorit Bar- On, Speaking My Mind (Oxford: Oxford University Press, 2004). 5. According to Higher Order Thought (HOT) theory, statements such as “I believe that the future is bleak” express second-order beliefs: the belief that I believe the future is bleak. David Rosenthal is most closely associated with this view. See his “Moore’s Paradox and Consciousness,” Philosophical Perspectives 9 (1995): 313–333. I outline my grounds for rejecting HOT theories in note 11.

Rational Explanation of Belief

23

matter of blurting something out that happens to be true. To just say that one is in a certain state is neither to blurt it out nor to say so on the basis of evidence or inference. Traditionally, the special status of this sort of pronouncement is characterized in terms of a distinctive authority or epistemic privilege associated with the fi rst-person point of view. I do not commit myself to any specific analysis of first-person authority here, but I take the phenomenon for which such theories are attempting to account to be familiar enough. To arrive at a self-attribution by way of observation or evidence is not to express the attributed mental state. For these assessments are equally available, in principle, to both the self-attributer and to anyone else. Where the self-attribution is expressive, it is arrived at in a fashion only open to the believer or desirer herself. No one else besides Harry can just speak about his jealousy. Mental states can also, of course, be expressed without any explicit selfreference. Thus I can express my belief by simply asserting that the future is bleak. Assertion is uniquely suited to belief in this respect: sincere assertion, as such, guarantees belief-expression. I put the point this way: assertions are canonical expressions of belief.6 Other mental attitudes that have propositional objects, e.g., doubt and denial, can be expressively self-ascribed using the relevant verb (e.g., “I doubt that . . .” and “I deny that . . .”), but they cannot be expressed by asserting the relevant contents. And it is not obvious that (nor is it relevant here whether) doubts and denials have their own kinds of speech-acts—speech-acts that are canonical expressions of doubt and denial. In the case of desire, what is desired is not propositional at all— at least if we take how desires are generally characterized at face value. For they are specified using infi nitival phrases, such as “to eat a salami sandwich.” In this case, not only can one not express one’s desire by simply

6. This raises a question of how precisely to characterize assertion—whether it can be characterized purely syntactically or in terms of the speaker’s aims or distinctive norms, or what have you. Here I employ the following operational defi nition: an assertion is anything that can be the fi rst conjunct in a Moore-Paradoxical statement of the form “p; but I don’t believe p.”

24

Rational Causation

asserting the content of the desire, but its content is simply not assertable at all.7 Of course all sorts of speech-acts can express my desire: “Bring me a salami sandwich!”; “Do you serve salami sandwiches?”; “This is the perfect moment for a salami sandwich.” But in these cases, it is not the kind of speech-act as such that guarantees desire-expression—in the way that (sincere) assertions, as such, guarantee belief-expression. Rather, the speaker expresses a desire to eat a salami sandwich by using a command, a question, or an assertion. My interest here is not in exploring whether there is any kind of speechact that stands to desire as assertion stands to belief. Nor am I interested in what mental states, if any, are canonically expressed by commands or questions. Rather, I am attempting only to make clear my own thesis regarding demonstrations. My claim is that just as assertion is the canonical expression of belief, demonstration is the canonical expression of believing-for-a-reason.8

1.4 The Source of Moore-Paradoxicality The existence of an analogy between assertion and demonstration is supported by the fact that both the original and my variant of Moore’s Paradox can be seen as the result of following this recipe: conjoin an instance of a speech-act type that is a canonical expression of mental type M with

7. According to the standard view, desires are ‘make-true’ attitudes toward propositions. The implicit propositional subject in “to eat a salami sandwich” is, on this view, just the subject in the larger sentence. But as Michael Thompson argues, no propositional substitution for “to eat a salami sandwich” in “I want to eat a salami sandwich” yields a statement with the same meaning as the original. (For example, the propositional substitution “I eat a salami sandwich,” on its natural interpretation, is a habitual and doesn’t say anything about the present at all.) See Michael Thompson, Life and Action (Cambridge, MA: Harvard University Press, 2008), 127–128. Following Thompson, I hold that desires relate a person not to a proposition but rather to a would-be action. I return to the topic of desire in Chapter 2, section 4.2. See also Matthew Boyle and Doug Lavin, “Goodness and Desire,” in Desire, Practical Reason, and the Good, ed. S. Tenenbaum (Oxford: Oxford University Press, 2010), 161–202. 8. This should not be confused with a performative view of believing-for-a-reason, according to which it is constituted by a demonstration or self-attribution. My view is that a demonstration expresses rather than constitutes the relevant state.

Rational Explanation of Belief

25

a statement that expressively denies the very same M. In the case of the traditional Moore-Paradoxical statement, the speech-act type is assertion, and so we have: “p; but I don’t believe p.” In the case of VMP, the speechact type is demonstration, and so we have “p because q; but I don’t believe p because I believe q.”9 This recipe makes it clear why the second conjuncts of MP and VMP are necessarily false. But that still doesn’t explain why the utterances as a whole are so hard to make sense of. I suggest the following: statements of these forms threaten unintelligibility because they seem to be expressions of mutually exclusive actualizations of the same abilities—the ability to take a stand on whether p is true (in the case of MP) and the ability to take a stand on whether q provides reason enough for believing p (in the case of VMP). Consider, in support of this hypothesis, the following fact. One very natural interpretation of the statements “I believe that p” and “I don’t believe that p” is that they express the speaker’s answer to the question: Is p true?10 The appearance of unintelligibility surrounding MP is dispelled if one interprets either conjunct as expressive of something other than the thinker’s stand on the question of p’s truth. Suppose that someone says “The subways are safe; but I don’t believe that the subways are safe.” The statement becomes intelligible if the speaker goes on to describe the first conjunct as the recitation of conventional wisdom or if she clarifies that the second conjunct is based on her observation of her own behavior (e.g., the fact that she never takes the subway) and so is expressive of her view of her own (lack of) belief. To be sure, it is, in the latter case, still unusual, bizarre, and even absurd in its own way—norms of speech and thought are flagrantly flouted, as many have observed. But it’s clear what she’s driving at.

9. Bar- On, in Speaking My Mind, discusses other “Moore-type confl icts” that seem to follow a very similar recipe: “p?; but I don’t wonder whether p”; “Ouch! This doesn’t hurt at all”; “[Yawn] this is really interesting”; “[Laughter] this joke is not funny at all.” See page 219. 10. Compare Arthur Collins, The Nature of Mental Things (Notre Dame, IN: University of Notre Dame Press, 1987), and Richard Moran, Authority and Estrangement (Princeton, NJ: Princeton University Press, 2001).

26

Rational Causation

However, if someone says “the subway is safe; but I don’t believe that the subway is safe,” and further interrogation reveals that the first conjunct is meant as a sincere assertion, and the second as a sincere rejection of a self-ascription, but not the sort that is based on evidence, and there is no equivocation or non-standard use of words, one simply cannot make anything of the utterance. The sentence, abstracted from the context of utterance, has non-contradictory truth-conditions. But every intelligible interpretation of what the speaker might mean by uttering the sentence has been ruled out. The appearance of unintelligibility surrounding VMP is dispelled in the same sorts of ways. If the speaker is simply repeating the reasoning— “There are cameras everywhere, so the subway is safe”—that she read on the Internet, there is nothing puzzling about her adding that she herself does not believe that the subway is safe because she believes there are cameras everywhere. “Cameras have nothing to do with it,” she might add. And if the second conjunct is an evidence-based rejection of a psychological explanation, we would fi nd this, as above, strange but comprehensible. If, for example, it were pointed out to the speaker that she thinks that the park is unsafe despite the ubiquity of cameras and reminded that she rode the subways worry-free prior to the installation of cameras, she might on the basis of the evidence deny that she believed the subway is safe because she believed that cameras are everywhere. So understood, “q so p; but I don’t believe that p because I believe that q” is no longer unintelligible. Beyond the analogy between the two absurd conjunctions, I hope that the following point has emerged. There are intelligible uses of the expressions “p, but I don’t believe that p” and “q, so p; but I don’t believe that p because I believe that q.” There are even intelligible uses in which the first conjunct expresses belief or reasoning and the second reports the absence of that very belief or reasoning. I call the intelligible (albeit bizarre) instances of these forms of words MP- and VMP-doppelgängers. But a genuinely Moore-Paradoxical utterance or an utterance of its variant, in which the second conjunct purports to disavow the very belief or reasoning expressed by the fi rst conjunct, is (and does not merely appear) unintelligible and does not express a thought at all. The central philosophical

Rational Explanation of Belief

27

interest of Moore-Paradoxical statements concerns how unintelligibility could be generated in this fashion.11 Here is the beginning of my explanation. Sincere assertion is the expression of belief. Sincere demonstration is the expression of believingfor-a-reason. Both states of mind have modes of expression in which the speaker self-ascribes the relevant state. These explicitly self-conscious expressions of belief and believing-for-a-reason employ forms of words that can also be used to report a belief or a believing-for-a-reason, i.e., to selfascribe on the basis of observation or evidence. When they are so used, there may be absurdity, but it is of a different kind. The unintelligibility associated with genuine instances of Moore-Paradoxicality is a consequence in part of expression’s being factive, in the following sense: one cannot express one’s being in a state (or not being in a state) unless one is actually in it (or not in it). And taking p to be true (or taking p to be true on the grounds that q) excludes simultaneously not taking p to be true (or not taking p to be true on the grounds that q). Hence, we don’t know what to make of someone who expresses a belief or a believing-for-a-reason and simultaneously seems expressively to deny believing the relevant p or believing the relevant p on the basis of the relevant q. Understanding such a

11. According to HOT theory, “I believe that p” expresses a second-order belief, viz., my belief that I believe that p. Its proponents must give a different sort of explanation of Moore-Paradoxical statements, since, on this view, ‘p’ is the expression of an ability to take a stand on whether p and “I don’t believe that p” is the expression of an ability to take a stand on whether I am in a certain psychological state. Their account of the Moore-Paradoxicality of the conjunction will reveal that it is highly bizarre or unnatural or irrational or in severe violation of some norm. Their defense of this interpretation is based on the irrationality-yet-intelligibility of what I’ve called the MPdoppelgängers. Since there are intelligible statements of the form “p, but I don’t believe that p,” the problem with Moore-Paradoxical statements cannot be that they simply fail to make sense. HOT-theorists proceed by considering the interpretations on which they do make sense, and then diagnosing what’s puzzling about them. (See also Mitchell Green, “Moorean Absurdity and Showing What’s Within,” in Moore’s Paradox: New Essays on Belief, Rationality, and the First Person, ed. M. Green and J. Williams [Oxford: Oxford University Press, 2007], 189–214.) But however it may be with MP-doppelgängers, genuine cases of Moore-Paradoxical utterances are worse off than this. They do not rise to the level of irrationality because they say nothing and express no thought. HOT-views cannot account for this.

28

Rational Causation

remark depends upon one or the other conjunct not being a genuine expression of the relevant state.

1.5 Believing-for-a-Reason Is Not a Process Demonstration is the canonical expression of believing-for-a-reason. ‘Reasoning’ may bring to mind a process that involves concentrating on the premise, then on the principle that connects premise to conclusion, finally focusing on the conclusion, and deeming it true. But believing-for-a-reason does not consist in a process at all. Rather, it is a state that can result from such a process. After I have considered (what I take to be) a true premise, and then perhaps a connecting principle, I make up my mind, after which point I believe p for reason q. On this proposal, believing-for-a-reason is not a mongrel state that includes someone’s believing that p and a bit of her psychological history, viz., her having focused on her belief that q. There is nothing MooreParadoxical about the following: “Everyone else has an alibi, so the butler did it; but I don’t believe that the butler did it because I had earlier decided that everyone else has an alibi.” Believing-for-a-reason is not simply believing that p, a state the obtaining of which is, as it happens, causally downstream from a mental affirmation of q. A belief and its ground have a different sort of tie, which I elucidate in what follows. Believing-for-a-reason need not be the result of any conscious process at all. In fact, much of what we take to be true for reasons is not obviously the result of conscious consideration of premises and connecting principles. I may believe that my neighbor has few friends because no one ever visits him. I may never have made this reasoning explicit, either to myself or to anyone else. Still, if asked the question “Why do you think he has few friends?,” I can reply, without any introspection or self-observation: “Because no one ever visits him.” That a subject is in the relevant state does not necessarily manifest itself in conscious review of the reasoning but does necessarily include the ability to express it both in the form of a demonstration and an expressive self-explanation, i.e., a rational explanation of one’s own belief that one can just give. We now turn to apparent counterexamples to this thesis.

Rational Explanation of Belief

29

1.6 Defending the Expressibility Thesis from Purported Counterexamples People often don’t know why they believe what they do. Consider this explanation: Billy believes that pork is healthier than poultry because his sister whispers it in his ear while he sleeps. The truth of this explanation does not depend on Billy’s ability to express it. But that is because his sister’s whispering it in his ear is not his reason. My claim is that if q rationally explains S’s believing p, then S is in a position to say authoritatively, and not on the basis of observation or evidence, that his reason for believing p is q. This is the Expressibility Thesis. It is much more difficult to find a clear-cut case of someone’s believing something for a reason, yet being unable to state the reason. Suppose Phil professes having a reason for thinking that more people eat baba ghanoush in Turkey than in any other country but then is unable to say what the reason is. At the very least, his apparent inability to produce the reason demands explanation. Let us consider a few possibilities. First, a thinker may be prevented from expressing what he is able, in an intuitively obvious sense of ‘able’ to express. For example, someone might be poised to administer a powerful electric shock as soon as Phil begins to give his reason, thereby rendering him unable to speak and hence to express his reason. Plainly, this kind of case calls for the disambiguation of the term ‘ability.’ Appropriately timed, powerful electric shocks can deprive someone, in one sense, of the ability to do almost anything that he is otherwise capable of doing: walking, whistling, multiplying four-digit numbers in his head, delivering eloquent speeches, etc. But even if shocks would prevent the exercise of these abilities, they do not erase the distinction between this man and someone who cannot walk, whistle, multiply, or speak eloquently even when circumstances permit. Intuitively, Phil lacks not the ability to say what his reason is but the opportunity to exercise this ability.12 Second, he may say that he cannot quite find the words at the moment, that he has difficulty articulating his reason, that it is too hard to articu12. Compare Anthony Kenny, Will, Freedom, and Power (Oxford: Blackwell, 1965), chapter 7.

30

Rational Causation

late or perhaps simply ineffable. The Expressibility Thesis does not say precisely where to draw the line on this list. If the first answer is really accurate, I would think that Phil can just say, in the relevant sense, what his reason for belief is. Just not at the moment. And the last answer suggests that the belief is held in place, so to speak, by something other than a reason, perhaps sheer stubbornness. The possibility that Phil might reply with one or another of these answers does not threaten the Expressibility Thesis so much as show that its application is not such an easy matter. Third, he may have completely forgotten his reason. Suppose, that is, that he has forgotten it and no amount of prompting can lead him to remember. In such a case, what is his reason for now thinking that more people eat baba ghanoush in Turkey than in any other country? Perhaps it is that he remembers once thinking he had a reason for believing it. In that case, his putative memory, not what he forgot, is his current reason. What he forgot used to be his reason for believing it, but no longer. This sort of case also poses no threat to the Expressibility Thesis. Fourth, he may admit that he cannot say why he believes it. Nonetheless, there may be an unconscious reason why he believes it. By this, I don’t mean a subpersonal ‘reason’; my topic in this book is states of the person, not of her parts. To explore the idea of a person’s unconscious reason, we need a more plausible candidate for repression than someone’s reason for thinking that more people eat baba ghanoush in Turkey than in any other country. Consider Dan, who believes that his wife’s new job bodes ill for their marriage and cites as his reason that it will substantially limit their ‘quality time.’ The real reason, however, is that her large new salary emasculates him (or so he feels). Perhaps he cannot stomach the idea that he would be moved by this consideration and so refuses to acknowledge his reason, even to himself. If I’m right, this must be more like a case of unwillingness, rather than genuine inability to say what his reason is. The point is difficult in part because we don’t always clearly separate these. We might say he won’t admit it, even to himself, which suggests he can. Or we might say he is unable to admit it, even to himself, which suggests he can’t. We also might say that he is unable to bring himself to admit it, a characterization that uses the

Rational Explanation of Belief

31

language of inability even as it suggests that admitting it is a genuine possibility for him. The best evidence in favor of the idea that we view people’s reason for belief as expressible— even in this difficult sort of case—is that we view it as a moral failing of Dan that he won’t own up to it. The fact that he would lobby against her taking the job because it would make him feel like less of a man is a failure of character. But the fact that he won’t acknowledge this reason is a further failure of character. And it is not simply the failure to draw an obvious conclusion from the evidence he is especially wellpositioned to appreciate. The failure consists in part in the fact that producing evidence is required to pressure him into owning up to it. But this is only a failure of Dan’s if just giving his reason is a live possibility for him. Ought implies can. To view it as a moral failure is thereby not to view it as, say, a neurological problem with his brain. Were it merely a mechanical defect or mental compulsion, it would be irrational and unfair to hold it against him. His failure to admit his reason is not a matter of his being unable to do so but rather of his being ashamed or embarrassed by believing for a reason that confl icts profoundly with his image of himself.13 It might be objected here that although it is certainly normal to be able to express one’s reason for belief, although we expect it and are rightfully suspicious of apparent exceptions, it is perfectly conceivable that someone might not have any such ability. However, careful consideration of our responses to actual and vividly imagined cases suggests otherwise. We are willing to accept that Dan is self-deceived and thus that his insistence on the ‘quality time’ story has whatever diminished form of sincerity separates the self-deceiver from the bald-faced liar. But self-deception is not simple ignorance. On account of its being the former, we expect him to be churlish when prodded on the connection between his wife’s prospective job and his pessimistic view of its consequences. Upon being reminded of

13. Here I contradict Finkelstein, who holds that “it is a defi ning characteristic of our unconscious mental states that we lack the ability to express them merely by selfascribing them” (Finkelstein, Expression and the Inner, 119). On my view, the psychological factors that render the belief unconscious do so by inhibiting the exercise of this expressive ability.

32

Rational Causation

his response to his wife’s long-standing complaints about his refusal to tear himself away from this laptop, he might feel compelled to recognize that the reduction of ‘quality time’ is not plausibly his reason for thinking that his wife’s new job bodes ill for their marriage. He might be thus led, in a similar fashion, to see correctly that his belief is in fact based on his felt emasculation by her new salary. But we do not view the change here as analogous to his discovery of a mole on his back. We view it rather as his having to face up to a fact of which he was aware all along. If we don’t view the matter this way, there is no longer any reason to view him as selfdeceived, as opposed to merely ignorant. Clearly, this is just an outline of an approach to a very difficult and important topic. My goal has been to show that there is a plausible strategy for response to each sort of purported counterexample to the Expressibility Thesis.

1.7 On the Significance of the Expressibility Thesis How is it possible that one and the same state could be expressible in both ways: as a demonstration and an explanation of one’s own belief? Here is my answer: to believe p for reason q consists in the thinker’s representing p to herself as . We can understand the relevant sense of ‘to be believed’ in relation to the notion of theoretical deliberation. When we consider whether to believe some proposition p, we weigh the evidence in favor and against its truth and perhaps come to some conclusion. In the end, we may make up our mind whether to believe it. To decide in the affirmative is thereby to represent it as to be believed, putting one in a position to say, authoritatively yet without self-observation, “I believe that p.” Similarly, consideration of whether p is to be believed as a consequence of q’s being to be believed just is making up one’s mind about whether to believe that p in light of q.14 If one decides the question in the affirmative, then one thereby

14. Compare Moran, Authority and Estrangement; Allan Gibbard, “Normative and Recognitional Concepts,” Philosophy and Phenomenological Research 64 (2002): 151–167; and Matthew Boyle, “ ‘Making up Your Mind’ and Mental Agency” (unpublished manuscript).

Rational Explanation of Belief

33

represents p as . The resultant state is a representing of a relation between doxastic deontic facts about propositions and at the same time also a representing of why one believes what one does. For it is (a) a way of thinking about the propositions and the relation between them that (b) puts the thinker under what she thereby recognizes as a binding doxastic obligation. Because of its (a)-aspect, believing-for-a-reason is expressible in the form of a demonstration. Because of its (b)-aspect, believing-for-a-reason is expressible in the form of a rational explanation of one’s own belief; to recognize the binding obligation is self-consciously to believe in conformity to it. Hence to believe p for reason q is to be in a position to say that q is one’s reason for believing p in either of these ways and to do so without any further investigation or introspection.15 But the connection between (b) and the possibility of just giving one’s reason for belief can seem dubious. For example, Isabella might, it seems, represent the to-be-believed-ness of there being cameras everywhere as having as a consequence the to-be-believed-ness of the subway’s being safe and yet she might nonetheless fail—knowingly even—to meet her doxastic obligations. There would thus be nothing Moore-Paradoxical about her acknowledging that she ought to but does not in fact believe the subway is safe because there are cameras everywhere. So, it seems that believing p for reason q cannot simply be the same as representing p as . This sort of representing cannot endow us with an ability to just say what our reasons in fact are, as opposed to what we think they should be. 15. Shah and Velleman argue that deliberating whether to believe p is considering whether p is true. See Nishi Shah and David Velleman, “Doxastic Deliberation,” Philosophical Review 114 (2005): 497–534. Here I make a similar point regarding believing that p for reason q. Shah and Velleman also argue, however, that to answer the question of whether I currently believe that p, either I must now make it the case that I believe that p by considering anew whether p or I must put the question to myself as a kind of stimulus and then catch myself as I react positively or negatively. But in most cases, I do neither of these things. If asked whether I believe, say, that Israel is part of Eu rope, I can answer in the negative straightaway, without either actively making up my mind or treating myself like an experimental test-subject. The same is true of believing-for-a-reason. A proper account of these states must explain how this is possible.

34

Rational Causation

To answer this objection, we can make a distinction between two sorts of deliberation: engaged and disengaged. On the one hand, we can take up the question of whether p is true or whether q establishes the truth of p in a spirit of doxastic openness; one can give oneself over to the results of one’s deliberation come what may. When one deliberates in this engaged manner, one’s beliefs are on the line. This strikes me as the typical case. If I am considering whether p, then upon encountering what I take to be decisive, indefeasible evidence— q—in favor of p, I thereby believe p in light of q. There is no extra step necessary, in which I decide whether to adopt the belief supported by the evidence, or whether to adopt it but not because of q. On the other hand, one can take up these sorts of questions in a more academic spirit, in a disengaged manner. One might be unable seriously to entertain the idea that one’s closest friend is a Russian spy, and yet one might nonetheless examine the evidence. Whether one will go along with the results of this inquiry may then not be settled simply by the power of the evidence; one’s loyalty may (perhaps misguidedly) trump what one acknowledges are decisive epistemic reasons for belief. Above I identified representing a proposition as ‘to be believed’ as the characteristic product of theoretical deliberation. More specifically, it is the characteristic product of engaged as opposed to disengaged deliberation (which is not to say that it must be the product of any deliberation). We might equally well use the same words to express the conclusion of either sort of deliberation: “I should believe p.” But only in the former case does this statement express the thinker’s representation of p as to be believed, in my quasi-technical sense. What Isabella expresses when she says she ought to believe the subway is safe because there are cameras everywhere is the sort of opinion that results from disengaged deliberation, in which a gap opens up between the result of the deliberation and what one in fact believes (which is not to say that Isabella must have engaged in any deliberation). It might seem as if this is just an attempt to stipulate the problem out of existence. Deliberation sometimes culminates in a judgment about what to believe and the corresponding belief. Sometimes it culminates in a judgment about what to believe without the corresponding belief. It seems as if I have, in effect, defined engaged deliberation as covering the former cases and disengaged deliberation as covering the latter cases. But that is not so.

Rational Explanation of Belief

35

There is, I submit, an intuitive difference between allowing one’s doxastic fate to be determined by the results of one’s consideration of the evidence and leaving it as an open question whether to accept the results of such consideration. Thus there are cases of disengaged deliberation that do eventuate in a judgment about what to believe and the corresponding belief. And that fact alone doesn’t show the deliberation to be engaged. In cases of engaged deliberation, the thinker’s thereby believing the relevant p is not a further fact beyond her recognition that she should believe it. In those cases, there is no extra step between recognizing what to believe and believing it. This is representing p as to be believed in my sense. Is Isabella lying, then, if she claims to have considered the question of the subway’s safety in an open spirit? Not necessarily. There are several other possibilities. One is that she is mistaken as to whether she really put her beliefs on the line when she weighed the evidence. This is a form of self-deception: she allows herself to believe that she considered the question of the subway’s safety in an open spirit, but in fact she was never open to the idea that it is safe. Another possibility is that her weighing of the evidence really did lead her to change her mind but that she backslid once she was sufficiently removed from the context of deliberation. The sort of backsliding is a common form of theoretical irrationality. And it is consistent with my central point here: there is nothing out-of-the-ordinary about the idea that a thinker’s viewing a proposition as having a certain normative status (to-be-believed-ness) could therein be her taking an affirmative psychological stance toward it (believing it). To sum up: consideration of the variant of Moore’s Paradox provides a window onto the special character of believing-for-a-reason. It has what seem to be distinct features: its directedness toward the relation between propositions and the availability of the state itself to be expressed by the reasoner. To say that believing-for-a-reason is the actualization of a selfconscious ability is to say that these two features of the state are in fact one. It is a kind of directedness toward propositions that, as such, makes for its own expressibility. To explain rationally why someone holds a certain belief is to attribute to her a state of believing-for-a-reason. The efficacy of a particular reason, I shall argue in Chapter 4, consists in a thinker’s exercise of the underlying ability.

36

Rational Causation

So far I have considered believing-for-a-reason from the point of view of the thinker— as it is expressed by a demonstration and a reflexive explanation. I start with the first-person perspective because even in explaining rationally why someone else believes what she does, one attributes to her a state that has the first-personal profile I’ve described. Over the next two sections I consider two ways of attributing believing-for-a-reason to others, viz., by citing the world and by citing the mind. Understanding the connection between these guises of rational explanation is, I argue, crucial to understanding the epistemic character of the underlying ability, as well as the kind of causation rational explanations reveal.

2. S Believes That p Because q Rational explanations attribute to a thinker the state of believing-fora-reason. The central difference between “S believes that p because q”— a world-citing rational explanation— and “S believes that p because S believes that q”— a mind-citing rational explanation—is that the former is more epistemically demanding than the latter.16 Ultimately, I argue that the more epistemically demanding guise of explanation is a better guide to the nature of the ability underlying believing-for-a-reason. In this section I focus on the epistemic requirement for the success of world-citing explanations of belief. What distinguishes reason-giving and non-reason-giving statements that use the form of words “S believes that p because q”? “John believes that Mary is fl irting with Peter because John is drunk” is an explanation of why John believes that Mary is flirting that does not give his reason. “John believes that Mary is flirting with Peter because she and Peter are thumb-wrestling” does give his reason. What is the difference?

16. Again, the labels ‘world-citing’ and ‘mind-citing’ are inexact. We might say “Orson is heading to the bunker because he believes that Martians are invading.” Presumably, that Orson believes that Martians are invading is a worldly fact, i.e., a feature of the world. Yet this is mind-citing. Also, not every rational explanation that appeals to the subject’s beliefs is a mind-citing explanation. “Natalie believes she was going crazy because she believed that doughnuts were communicating with her telepathically” is world-citing. I discuss this example in section 3.1.

Rational Explanation of Belief

37

Here’s something these examples have in common: both sorts of explanation require the truth of the explanans. It cannot possibly be true that John believes that Mary is flirting with Peter because John is drunk if, in fact, he is not drunk. Nor can it be true that John believes that Mary is fl irting with Peter because she and Peter are thumb-wrestling if, in fact, they are not thumb-wrestling. This seems to be a function at least in part of the factive character of explanations, which, as such, postulate some sort of dependence of what the explanandum designates on what the explanans designates. The obtaining of a fact or the occurrence of an event cannot depend on q if “q” is false. But in the case of rational explanations, there is a more illuminating explanation of the truth-requirement. We can get at that more illuminating explanation by noting the most salient distinguishing feature of the rational “S believes that p because q.” It entails “S believes that q.” Not so in the non-rational case. John need not believe that he is drunk in order for it to be true that he believes that Mary is flirting with Peter because he, John, is drunk. But John does need to believe that Mary and Peter are thumb-wrestling in order for it to be true that he believes that Mary is flirting with Peter because she and Peter are thumb-wrestling. The deeper explanation of the truth-requirement on the reason-giving “S believes that p because q” also explains the belief-requirement. It is the following thesis, recently defended by John Hyman: for it to be true that S believes that p because q, S must know that q.17 Suppose Louis, who is convinced that he has psychic powers, is being put to the test. He is asked to name the person sitting in the next room. The name ‘Emma’ comes to him. By sheer chance, this is the name of the person sitting in the next room. Louis decides ‘on that basis’ that there is someone in the next room whose name begins with ‘E’—i.e., he believes that there is someone in the next room whose name begins with ‘E’ for a reason he would express this way: “There is someone in the next room

17. John Hyman, “How Knowledge Works,” Philosophical Quarterly 49 (1999): 433– 451. In fact, Hyman argues for a stronger claim: to know that p just is to be able to do (or believe) something for the reason that p. I return to this stronger claim in Chapter 3. See also Peter Unger, Ignorance: A Case for Scepticism (Oxford: Clarendon Press, 1975).

38

Rational Causation

named ‘Emma.’ ” Still, it is not the case that Louis believes that there is someone in the next room whose name begins with ‘E’ because there is someone in the next room named Emma. Here is another example. Suppose Frank is convinced that all restaurants poison their hamburgers. Suppose, further, that Frank has been served a poisoned hamburger in a restaurant. Frank believes that this hamburger will kill him for a reason he would put this way: “The hamburger is poisoned.” It is not true, however, that Frank believes this hamburger will kill him because it is poisoned. To give those world-citing explanations is to credit Louis and Frank with genuine cognitive access to, respectively, the fact that there is someone in the next room named ‘Emma’ and the fact that the hamburger is poisoned. It is to say that these facts were ‘in view’ when they made up their minds. The perceptual metaphor, if indeed apt, already suggests what is missing in these cases. Knowledge is needed to validate rational explanations that appeal to the world. We can approach the point from a different direction. To say that S believes that p because q is to think of the fact that q as helping to underwrite or make true an explanation of why S is in a certain mental state. Many would put the point baldly: q is a cause of S’s believing that p. I argue in Chapter 4 that although this might be true, it is not an efficient cause. For the moment, however, let’s not worry about what kind of cause it is. In whatever sense q can be a cause of S’s believing that p, it seems clear that at least in the cases of Louis and Frank, the relevant facts are out of the causal loop. The fact that a woman named ‘Emma’ sits in the next room has no influence on Louis, and neither does the fact that the hamburger is poisoned have any effect on Frank. And so neither Louis’s nor Frank’s belief can be ‘becausally’ connected to the relevant fact. What would it take for Louis or Frank to be in the causal loop? Consider a case in which a worldly fact influences belief in a manner insufficient to guarantee that it is available to serve as a reason for holding that belief. Hunter’s having unwittingly ingested LSD does not rationally explain his believing that he is a giant centipede. To figure in such an explanation, such a fact, it seems, must influence the subject through some mode of awareness, such as perception, memory, testimony, or reasoning.

Rational Explanation of Belief

39

Not just any influence via one (or more) of these (and perhaps other) means will do. Hunter’s being on LSD might cause him to see everything as various shades of purple. His being on LSD thus might cause him to believe that his room has been painted purple, and it might cause him to believe this on the basis of perception. But his being on LSD is still not necessarily eligible to serve as his reason for belief. Here is a plausible lesson to draw from these considerations: in order for q to be the right sort of cause of S’s believing that p, q must influence S through instances of, among other methods perhaps, perception, memory, testimony, or reasoning that culminate in S’s knowing that q. We cannot use world-citing explanations in the cases of Louis and Frank even though they take the relevant facts to obtain, because these facts are not learned in any of the above (or any other) ways. It is true that there is someone in the next room named ‘Emma’ and that the hamburger is poisoned— but Louis and Frank don’t know it. One might wonder whether the knowledge requirement is too strong, whether what’s required is only that ‘q’ is true, that S believes that q, and that S is justified in believing that q. But even if S is justified in believing that q, so long as S does not know that q (supposing this is a genuine possibility), it will be wrong to say that S believes that p because q. Any discussion of the point comes with the threat of interminable digression regarding the precise nature of justification. But a brief consideration of a familiar example will, I hope, prove suggestive. Henry finds himself looking at the only real barn in the land of barn-facades. He believes that there is a barn in front of him, it’s true, and, arguably, he is justified. (If you think that he isn’t justified, then simply consider your own favored case of justified true belief without knowledge. If you think there are no such cases, then the present objection, presupposing as it does a gap between justification and knowledge, will not be an objection that you want to level.) But it is wrong to say that Henry believes that there is a shelter in front of him for his cows because there is a barn in front of him. And if you intuit differently, I predict that you also think that Henry does know that there is a barn in front of him. I don’t pretend that this is proof that what’s required for a fact to be a reason is knowledge, as opposed to mere justified true belief. But consideration of such cases does suggest that the

40

Rational Causation

burden of proof is on those who hold that what’s necessary is simply the latter. I take the point as settled in what follows.18 S’s knowledge that q is a necessary condition of q’s serving as a reason for S; but it is important to say what kind of necessary condition: is it a consequence or a precondition of being S’s reason? Perceiving and remembering that q entail knowing that q. But these are explanatorily prior to knowing that p. Statements such as “S saw that q” and “S remembers that q” entail “S knows that q” in virtue of explaining S’s knowing that q. Knowledge in these cases is a necessary consequence. Reason-giving statements such as “S believes that p because q” entail S’s knowing that q but do not explain S’s knowing that q. Rather “S believes that p because q” presupposes that S knows that q. S’s knowing that q enables S to base her belief on q. We will return to this idea in Chapter 3. It might be objected here that I am leaning too heavily on one form of world-citing rational explanation and that the epistemic requirement is peculiar to this form. Jonathan Dancy argues that locutions such as “His reason for doing it was that it would increase his pension” and “The ground on which he acted was that she had lied to him” are non-factive. If they were factive, then “His reason for doing it was that it would increase his pension, but in fact he was quite wrong about that” and “The ground on which he acted was that she had lied to him, though actually she had done nothing of the sort” would be self-contradictory.19 But, Dancy 18. It would be odd if reasons for belief were the only reasons that required knowledge. But nothing in the foregoing discussion relies on the explanandum’s citing a reason for belief, as opposed to action. Precisely the same considerations that support the knowledge-requirement on S’s believing that p “because q” also support this requirement on S’s ϕ-ing “because q.” Louis cannot send the hamburger back to the kitchen “because it’s poisoned” in the scenario envisaged. Nor can Henry corral his cows “because there’s a barn in front of him.” Where facts are reasons, whether they are reasons for believing or reasons for acting, the subject must know these facts. Indeed, the same can be said not just of actions and beliefs but of the whole range of attitudes and feelings to which the language of reasons apply, e.g., “S is suspicious because q,” “S is happy because q,” or “S is disgusted that q.” See Robert Gordon, “Emotions and Knowledge,” Journal of Philosophy 66 (1969): 408–413, and Unger, Ignorance. 19. Jonathan Dancy, Practical Reality (New York: Oxford University Press, 2000), 132. I should emphasize that Dancy raises this point in the ser vice of an attack on Psychologism. He argues that because rational explanations are not factive, they are

Rational Explanation of Belief

41

claims, they aren’t. And, of course, if such locutions do not require the truth of the explanans, then they do not require that the subject knows them to be true. Whether or not we use a factive (or, as I think of it, epistemically demanding) way of explaining the action “will be a comparatively arbitrary choice.”20 As he later puts it: “the fact that the explanation can be given in a non-factive form shows that it is a non-factive explanation, even though that very explanation can be given in ways that (for trivial reasons to do with the use of certain words) are themselves factive. In the latter cases, it is not the explanation itself that is factive, but the form of words that we use to express it.”21 I agree with Dancy that statements such as “The ground on which he acted was that she had lied to him, though actually she had done nothing of the sort” are not self-contradictory. Yet although they do not represent the same proposition as true and then as false, they do contain a milder sort of tension, just as “The ground on which he acted was that she had lied to him, and she had done exactly that” sounds only mildly redundant. In both cases, if the speaker had omitted the second clause, we would have taken the statement as an acceptance of the ground. The second clause prompts a double-take—a reinterpretation of the fi rst according to which it is more like a mind- citing explanation. This description is consistent with these styles of explanation being factive as they typically occur. I will not explore this phenomenon in any more detail here. At the very least, these putatively non-factive locutions have a cancelable implication that q is true and that the subject knows the relevant q. The question then remains why it should be that world-citing explanations entail or, in some cases, if Dancy is right, merely imply such knowledge. I shall argue that the epistemic (and so truth-involving) dimension of world- citing explanation is far from trivial, that it is in fact revelatory of the nature of having reasons for belief and action. I return to this issue in section 3.2. not causal. Hence to act (or believe) for a reason is not to be caused to act (or believe) by psychological states. I contrast Dancy’s Anti-psychologism with my own in Chapter 2, section 4. 20. Ibid., 135. 21. Jonathan Dancy, “Two Ways of Explaining Action,” in Agency and Action, ed. J. Hyman and H. Steward (Cambridge: Cambridge University Press, 2004), 25–42, 28.

42

Rational Causation

*

*

*

At the beginning of this section, I raised a question about why “S believes that p because q” requires the truth of q. I claimed that an answer to this question that appeals to the factive nature of explanation misses something important: the requirement that q is part of a more comprehensive requirement that S know that q. This requirement also explains, of course, why S must believe that q in order to believe that p “because q,” in the rational sense. But it also illuminates the character of mind-citing rational explanations. And the light it sheds helps to make clear the role of psychological categories in explanations that appeal to someone’s reasons.

3. S Believes That p Because S Believes That q Rational explanations have seemed to many to be proto-scientific— deployments of a ‘folk theory’ that postulate movements of unobserved entities to explain observed regularities. From this vantage point, mindciting explanation can seem to get at what’s crucial to rational explanations. World-citing explanations, by contrast, seem to add some information that is, strictly speaking, a separate matter from the rational explanation of belief. This is wrong, as I argue in the remaining sections of this chapter. To understand the character of mind-citing explanations, we need to understand them in relation to (a) the first-personal features of believingfor-a-reason discussed in section 1 and (b) the epistemic dimension of world-citing explanations discussed in section 2. The issues here will not be clear, however, until we improve upon our rough-and-ready understanding of the opposition between mind- and world-citing explanations. It is not a matter of their appealing to different portions of reality.22 Mind-citing explanations, in our sense, are distinguished by their being subject to the ‘Simultaneity Constraint’: the explanans and the explanandum designate facts that must be indexed to the same time. I discuss this principle in section 3.1. With this clarification in

22. This is not to say that we couldn’t use the labels ‘mind-citing’ and ‘world-citing’ to distinguish explanations on the basis of what portions of reality they cite. But this is not a philosophically interesting distinction, and is irrelevant to our diagnosis of Psychologism.

Rational Explanation of Belief

43

mind, I consider, in section 3.2, what to make of the fact that to give someone’s reason by explicitly citing her beliefs is to offer a hedged or guarded version of an explanation that we might otherwise have made by citing the world. I argue that it tells us something important about the nature of the ability underlying believing for a reason, viz., that it is epistemic.

3.1 The Simultaneity Constraint It is surprisingly difficult to find an explanation of someone’s holding a belief in which the explanans designates her holding another belief but doesn’t give her reason. The only sort of case I can think of involves successive believings. For example, suppose Sammy believes it’s time to walk Spot and so walks him, and then afterward notices that Spot is tired. We could speak of his belief that it’s time to walk Spot and his belief that Spot is tired as standing in a causal relation. We could put this—unnaturally, to be sure—in the form of an explanation: Sammy believed that Spot was tired because he believed that it was time to walk Spot. Obviously, this explanation does not give Sammy’s reason. Here’s how you can tell. It is not a past tense version of “S believes that p because S believes that q.” For there is no time at which it would have been true to say “Sammy believes Spot is tired because he believes Spot needs a walk.” Why does this matter? Because, as Matthew Boyle has pointed out, explanations that use the form of words “S believes that p because S believed that q” explain S’s believing that p in a very different manner than “S believes that p because S believes that q.”23 The former, unlike the latter, relates S’s believing that p to her psychological history. Because of this, they do not engender any VMP-style paradox. There is nothing paradoxical in saying, “Everyone else has an alibi, so the butler did it. But I don’t believe that the butler did it because I believed that everyone else has an alibi.” For the demonstration does not 23. Boyle, “ ‘Making up Your Mind.’ ” See also Sebastian Rödl, Self- Consciousness (Cambridge, MA: Harvard University Press, 2007), chapter 3.

44

Rational Causation

express anything regarding what I used to believe. And so the denial of a ‘becausal’ connection between what I used to believe and what I now believe does not produce any confl ict with my demonstration. Failing to appreciate the Simultaneity Constraint makes it tempting to think of S’s believing that q as chronologically prior to S’s believing that p, which would fit better an efficient-causal model.24 But any attempt to ‘correct’ the tensing by putting the explanans in the past and the explanandum in the present tense results in a different explanation. Since the ‘because’ in “S believes that p because S believes that q” is tied to the ‘so’ that S might employ in the corresponding demonstration “q, so p,” it still expresses a relation of inferential priority. According to a rational explanation of belief, S represents the to-be-believed-ness of q as establishing the to-be-believed-ness of p. There is thus no question of S’s believing that q obtaining at a different time from her believing that p.25 It might seem as if there are mind-citing rational explanations that violate the Simultaneity Constraint. There is, for example, the odd sort of case to which Dancy has drawn our attention (and to which we return in Chapter 2).26 We might say the following of Natalie, who is concerned about her own mental health because of the strange things she ‘fi nds herself believing’: Natalie believes she was going crazy because she believed that doughnuts were communicating with her telepathically. This is surely an explanation that gives Natalie’s reason and also one that cites the mind. But this belongs with the world-citing explanations. For 24. It is generally though not universally held that efficient causes must be prior to their effects. For an argument in favor of the existence of simultaneous efficient causation, see Myles Brand, “Simultaneous Causation,” in Time and Cause: Essays Presented to Richard Taylor, ed. Peter van Inwagen (Dordrecht, The Netherlands: D. Reidel Publishing, 1980), 137–153. For an argument against it, see D.  H. Mellor, The Facts of Causation (London: Routledge, 1995), 220–224. 25. It does not follow nor do I believe that “S believes that p because S believes that q” postulates a non-causal connection between psychological facts. But the Simultaneity Constraint does tell us something about the underlying causal connection. I discuss this feature of rational causation in Chapter 4, section 2.2. 26. See Dancy, Practical Reality, 125.

Rational Explanation of Belief

45

we do not view Natalie as in a position to demonstrate “Doughnuts were communicating with me telepathically, so I was going crazy.” Rather, we view her as in a position to demonstrate “I believed that doughnuts were communicating with me telepathically, so I was going crazy.” Thus, although the third-person version uses the form of words “S believes that p because S believed that q,” it is a world-citing explanation, albeit one in which the relevant facts concern the subject’s own beliefs.

3.2 The Epistemic Character of Believing-for-a-Reason In section 2 we saw that ‘q’ explains (becausally) why S believes that p only if S knows that q. But S need not know that q in order for “S believes that q” to explain (becausally) why S believes that p. Whereas “Frank believes that this hamburger will kill him because it is poisoned” requires that Frank know that the hamburger is poisoned, “Frank believes that this hamburger will kill him because he believes it is poisoned” does not. And this is the fundamental difference between the mind- and world-citing guises of rational explanation: knowledge of q is not required for the former. (Nor is there even an implication of knowledge, as there undeniably is with locutions such as “S’s reason for believing p is q.” If anything, “S believes that p because S believes that q” implies ignorance of q.) When a subject is in a position sincerely to say “q, so p,” she takes the fact that q to be cognitively available to her to serve as a guarantor of the truth of p. If we know (or perhaps even suspect) that it is not available, either because q is false or because she has no cognitive access to the fact that q, then we must cite her belief in order to explain her reasoning. The psychological frame thus reflects a modification of the way believing-fora-reason is attributed, a modification that can accommodate the explainer’s view that S may not know the relevant q. To some, this will suggest a hybrid view of world-citing explanations. First, there is believing-for-a-reason, which alone makes the relevant mind-citing explanation true. Second, there is the fact that S knows that q, which, in conjunction with believing-for-a-reason, makes the relevant world-citing explanation true. The absence of knowledge undermines one way of attributing believing-for-a-reason but not the state itself. An advocate of the hybrid view thus makes the following indisputable point. Since

46

Rational Causation

there is a ser viceable way of giving S’s reason for believing p without assuming knowledge, believing-for-a-reason does not necessarily involve S’s epistemic success. But the significance of this point is not so clear. The hybrid theorist sees things as follows: knowledge that q might be a normative requirement of believing-for-a-reason; but it is not—whereas belief that q is—a constitutive requirement. Believing-for-a-reason is thus ‘purely factual’ at its core. This is not to deny that reasoning can be good or bad, that, in general, belief, inference, argument, and judgment all are subject to various dimensions of normative (e.g., epistemic) assessment. It is just to deny that these features play a crucial role in explaining belief, as opposed to assessing it. Two philosophical tendencies support the hybrid view. There is, first, argument-from-illusion-style reasoning: the world can be more or less cooperative when our cognitive abilities are engaged, e.g., when we see (or seem to see) a grizzly bear or when we conclude that the library is unsafe because a grizzly bear is roaming the second floor (or so we believe). To study the ability itself, we factor out this sort of contingency and focus on seeming to see or reaching a conclusion in light of what we believe the facts to be. This thought is especially tempting in light of the fact that cases of success and cases of failure in the exercises of these abilities are phenomenologically indistinguishable from one another. The difference between the cases thus can seem to be extra-mental. We are thereby led to the idea that in order to understand reasoning, one should focus on rational explanation of beliefs in terms of other beliefs, as opposed to facts.27 Second, there is the tendency to separate factual and normative issues. On one side, we have questions about what there is and why things are the way they are; on the other side, we have questions about whether what there is is good or bad and whether things should be the way they are. Those who are sympathetic to this tendency, when confronted with a kind of object, act, state, or cause whose nature seems to include a normative property, look to factor it out, to explain it in terms of non-normative properties or simply to eliminate the category altogether.

27. The use of an argument from illusion to support the thesis that reasoning is based on beliefs, rather than facts, is discussed at length in Jonathan Dancy, “Arguments from Illusion,” Philosophical Quarterly 45 (2005): 421–438.

Rational Explanation of Belief

47

Much of what I say in the first half of this book is directed against these tendencies. In what remains of this section, I argue that it is constitutive of believing-for-a-reason that it is normatively diminished by the absence of knowledge: necessarily, any believing-for-a-reason is normatively diminished by failure to know the ground. I support this thesis by arguing that it provides the best explanation for a wide variety of phenomena associated with expressing and attributing believing-for-a-reason. Let us begin with the first-person perspective: Suppose that I do not know that everyone else has an alibi, and I know that I do not know it; but suppose that I believe that everyone else has an alibi. I do not, under these suppositions, take everyone else’s having an alibi to show that the butler did it. At best, I think that everyone else’s having an alibi would show—does show if I am correct— that the butler did it. But the demonstration “Everyone else has an alibi, so the butler did it” does express the speaker’s taking everyone else’s having an alibi to show that the butler did it. So someone who doesn’t know that q, and knows that he doesn’t know, is not in a position sincerely to demonstrate “q, so p.” Here’s another way of making the point. The same thought that one might express by saying “Everyone else has an alibi, so the butler did it” might equally well be expressed by saying “The butler must have done it, because everyone else has an alibi.” But there is no must about it if one knows that one doesn’t know that everyone else has an alibi. Of course, reasoning is still possible under what are recognized as less than ideal epistemic circumstances. I may believe, though not be sure, that Wilmer’s gun is loaded. Yet, on the basis of this belief, I might reason to the conclusion that we are going to get hurt if we do not flee. However commendable such reasoning may be in the envisaged scenario, it would nonetheless be more correct and, if time permitted, more natural to introduce various qualifications when promulgating it. Not: “That gun is loaded, so . . .” or “Since that gun is loaded, . . .” Rather, something more like: “That gun is, I believe, loaded, in which case . . .” or “If, as I think, the gun is loaded . . .” or “That gun is loaded, or so I think, so . . .” These last three ways of expressing the reasoning are hedged or guarded in relation to the first two. Does this tell us something about reasoning itself or just about the promulgation of reasoning?

48

Rational Causation

The same question arises in relation to the attribution of reasoning to others. We have already observed that “S believes that p because q” is epistemically demanding in a way that “S believes that p because S believes that q” is not. But this is just one example of a more general contrast. In addition to ‘because,’ we use in rational explanation phrases such as ‘in light of,’ ‘in view of,’ ‘since,’ ‘on account of,’ ‘on the grounds that,’ and ‘for the reason that,’ among others. No matter how we do it, there is a tendency to hear such attributions as knowledge-dependent, unless we use psychological framing. If I say “she believes that we should flee on the grounds that Wilmer’s gun is loaded,” then my statement gives rise to the same epistemic expectations as if I had said “she believes that we should flee because Wilmer’s gun is loaded,” even if Dancy is right (and I am wrong) that only the latter entails knowledge. Why do qualification-free attributions of believing-for-a-reason promote these expectations? A hybrid theorist is likely to appeal to some version of this thought. Rational explanations, absent psychological framing, contain (in some sense or another) assertions of the relevant p and q. And one should assert only what one knows.28 Thus it is typically impermissible to attribute believing-for-a-reason, absent psychological framing, unless one knows the relevant ground. Insofar as this strategy exploits a putative norm of speech, it points toward the idea that unqualified demonstrations and rational explanations of belief give rise to epistemic expectations not on the basis of any features of believing-for-reason per se but just insofar as it is expressed in an unguarded way. But even if I know that q, I may not explain someone else’s believing that p by citing simply “q” unless he also knows that q. So even if in saying “Poirot believes the butler did it because everyone else has an alibi,” I assert (in some sense or another) that everyone else has an alibi, and even if there is a knowledge-norm for assertion, this cannot explain why my attribution demands that Poirot know that everyone else has an alibi. Furthermore, the epistemic requirement for Poirot is not dependent upon his having actually articulated the relevant reasoning. We often ascribe reasoning to people who have not put it forward and so who have not (in 28. See Timothy Williamson, Knowledge and Its Limits (Oxford: Oxford University Press, 2000), chapter 11.

Rational Explanation of Belief

49

any sense) asserted the ground. And yet the epistemic requirement is still operative. The hybrid theorist might instead try something like the following strategy. The unqualified formulations are efficient ways of combining two pieces of information: that the thinker believes for a reason and that she knows the relevant ground. Perhaps it’s typical when someone believes that p for reason q that she knows that q. Thus the typical manner of attribution is non-psychological—the default mode. When circumstances are atypical, we use psychological framing. However, it is also typical that when someone believes that p for reason q that she possesses all of her toes. And yet we don’t have one family of expressions that mean roughly “believes that p for reason q and possesses a full complement of toes” and another family of expressions that mean roughly “believes that p for reason q and is missing at least one toe.” The existence of a more-than-contingent connection between believing that p for reason q and knowing that q better explains why the normal manner of attributing the former entails the latter. Here is my explanation: believing-for-a-reason is the actualization of a self-conscious rational ability to acquire knowledge by representing one proposition as to be believed on the basis of the to-be-believed-ness of another proposition. This ability is exercised perfectly only if it culminates in knowledge, which happens only if the premise-proposition corresponds to a fact known by the subject. When the subject lacks this knowledge, we introduce various qualifications. The purpose of these qualifications is precisely to attribute an exercise of the ability in a manner than does not give rise to the same epistemic expectations. This does not mean that one should never reason from q to p if one doesn’t know that q. But it does mean that in doing so, one’s reasoning is to that extent defective. So far I have discussed only ignorance of the premise. But ignorance of the principle connecting premise and conclusion also detracts from believingfor-a-reason. Note, first, that someone who concludes p on the grounds that q must at least believe that p follows from q. If Poirot doesn’t believe that everyone else’s having an alibi shows that the butler did it, then he can’t conclude that the butler did it on the grounds that everyone else has an alibi. But fully successful cases of believing-for-a-reason require more:

50

Rational Causation

the thinker must know that p follows from q. Suppose Fred Fabulist tells Gullible Gus that cumulonimbus clouds portend rain. Fred doesn’t know what he’s talking about, although he happens to be right. Gus believes him and for no other reason concludes, upon seeing an approaching cumulonimbus cloud formation, that it will soon rain. Although Gus may know that a cumulonimbus cloud formation approaches, he doesn’t know that it will soon rain. For he has a terrible reason for thinking that cumulonimbus clouds portend rain. If Fred were exposed to Gus as an ignoramus on meteorological matters, he would be obliged to abandon the judgment. My point does not depend on anything as strong as this principle: in order for S to know that p ‘because q,’ in the sense I have been discussing, S must know that q entails p.29 In extreme cases, where S’s reason for believing that p follows from q are truly atrocious, it seems clear that no new knowledge can arise as the result of the inference. But my point is just that to the extent that one falls short of knowing that p does indeed follow from q, the act of concluding that p on the basis of q is a less successful exercise of the ability underlying believing-for-a-reason. In this section I have argued that it is constitutive of believing-for-a-reason that ignorance of the premise or its inferential significance renders it deficient. But my argument for this claim rests crucially on the notion of a self-conscious rational ability. But what is a rational ability? In the next section, I answer this question.

4. Rational Abilities I have said that believing-for-a-reason is the exercise of a rational ability, in particular the ability to acquire knowledge by reasoning from known facts. In this section, I say what this claim amounts to and offer evidence for it. I begin by arguing (in section 4.1) that rational abilities are a species of disposition, albeit one that is distinctive in crucial respects (section

29. This is a stronger version of Fumerton’s Principle of Inferential Justification. See Richard Fumerton, Metaepistemology and Skepticism (Lanham, MD: Rowman & Littlefield, 1995), 85.

Rational Explanation of Belief

51

4.2). With this characterization in hand, I show (in section 4.3) that to give someone’s reason for belief is to appeal to the exercise of a rational ability.

4.1 Three Species of Disposition A rational power is a species of disposition. I distinguish three kinds of disposition: physical dispositions, organic functions, and rational abilities. To see how these are related compare:

(1) Physical Disposition (2) Organic Function (3) Rational Ability

(A) Event Predication

(B) Disposition Predication

The vase broke. The flower tracked the sun. James stood on his hands.

The vase was fragile. The flower is heliotropic. James can do handstands.

The idea that (1), (2), and (3) are species of a genus is not especially controversial, but it will nonetheless prove useful to consider the grounds for grouping them together. In each case, the statement in column (A) is in some sense or another explained by the statement in column (B). Although this much is now relatively uncontroversial (pace Moliere),30 there is little agreement on the way such explanations work. The (B) items explain the (A) items at least to this extent: it was no accident—no fluke—that the dropped fragile vase breaks, that the heliotropic flower turns toward the sun, or that James does a handstand when he tries. As we ordinarily use disposition-words, the fact that an object has a disposition does not entail that it exhibits the relevant behavior in the appropriate circumstances. A fragile vase might be dropped yet not break— if a sudden upward draft slows its fall, or if it is surrounded by the right 30. An exception is Elizabeth Prior, Robert Pargetter, and Frank Jackson, “Three Theses about Dispositions,” American Philosophical Quarterly 19 (1982): 251–257. Even though the authors deny that dispositions themselves explain their manifestations, they acknowledge that appealing to a disposition does provide at least a thin explanation. They argue that it is the ‘causal basis’ of the disposition that is doing the causalexplanatory work. For influential defenses of the causal-explanatory relevance of dispositions, see D.  H. Mellor, “In Defense of Dispositions,” Philosophical Review 83 (1974): 167–182, and George Molnar, Powers: A Study in Metaphysics (Oxford: Oxford University Press, 2003).

52

Rational Causation

packing material or, more fancifully, if an angel intervenes to prevent its breaking.31 A fragile object normally breaks when dropped,32 but it doesn’t follow from its being dropped that it breaks.33 Heliotropic flowers normally track the sun, but not always. A sunflower, for example, will not track the sun during the day if it has been covered the previous night by a transparent, blue-light-blocking cover. And if James is drunk, he might try and fail to do a handstand. (In this case, what James normally does is not at issue— a difference whose significance I touch on later in this section.) A statement about what an object that possesses a disposition normally does can be expressed in the form of a habitual, e.g., “A fragile object breaks when dropped,” or “A heliotropic flower tracks the sun.” Habitual statements are marked by the use of the present tense and by the fact that this use of the present tense does not describe any particular act by any specific subject at any time. As several authors have noted, a habitual tolerates exceptions.34 The claim that a fragile object breaks when dropped is consistent with the failure of some specific fragile object to break when dropped. Importantly, habituals tolerate only excused exceptions. To say that fragile objects break when dropped is to say that a fragile object’s failure to break when dropped demands explanation. Similarly, to say that a heliotropic flower normally tracks the sun is to say that the failure of a heliotropic 31. Mark Johnston, “How to Speak of the Colors,” Philosophical Studies 68 (1992): 221–263. 32. Others have attempted to formalize the relevant notion of ‘normal’ as it figures in discussions of dispositions. See, e.g., Michael Fara, “Dispositions and Habituals,” Noûs 39 (2005): 43– 82, and Daniel Bonevac, Josh Dever, and David Sosa, “The Conditional Fallacy,” Philosophical Review 115 (2006): 273–316. I do not need to commit myself to any formalization here. 33. The failure of this entailment scuttles standard conditional analyses of disposition-statements. An object can have the disposition, yet the relevant subjunctive conditional is false—hence, the conditional fails to provide a necessary condition of disposition-possession. See C. B. Martin, “Dispositions and Conditionals,” Philosophical Quarterly 44 (1994): 1– 8; Alexander Bird, “Dispositions and Antidotes,” Philosophical Quarterly 48 (1998): 227–234; and George Molnar, “Are Dispositions Reducible?,” Philosophical Quarterly 49 (1999): 1–17. 34. See Fara, “Dispositions and Habituals.” See also Thompson, Life and Action, chapter 4.

Rational Explanation of Belief

53

flower to track the sun demands explanation. Though talk of what an able handstander normally does, or does in ‘the appropriate circumstances,’ is not relevant to characterizing James’s ability to do handstands, there is nonetheless this important analogy. If an able handstander tries and fails to do a handstand, this demands explanation. In any of these cases, if no such explanation is forthcoming, then we have grounds for doubting whether the object has the relevant disposition. Physical dispositions, organic functions, and rational abilities are not necessarily actualized in the appropriate circumstances (or, in the case of rational abilities, when the subject engages the disposition), but when they are not, this fact demands explanation. As we ordinarily use disposition-words, the fact that an object exhibits, on a specific occasion, the behavior characteristic of a certain disposition does not entail that it has the corresponding disposition. Suppose I knock a vase to the ground, thereby breaking it. This does not prove that the vase was fragile. A non-fragile vase normally does not break when dropped but might do so after falling in just the right way.35 Similarly, a flower might continuously turn toward the sun despite not being heliotropic. A nonheliotropic flower does not normally do this, but it is possible: its motion might, for example, be manually and remotely controlled via some artificial implant. Similarly, James might, if he tried, have stood on his hands despite having no such ability—if, for example, he was underwater. In each of these cases, the relevant behavior is exhibited, yet there is no corresponding disposition. So the absence of a physical disposition, organic function, or rational ability makes the occurrence of the relevant behaviors not impossible but puzzling and thus especially in need of explanation. I claimed above that dispositions of all kinds offer a minimal explanation of their manifestations. To say that a dropped vase shattered because it was fragile is just to say it was not a fluke that it shattered, that it didn’t shatter despite not being disposed to do so when dropped. Dispositional

35. And so the standard conditional analysans’ of disposition-statements fails to provide a sufficient condition for disposition-possession: the relevant subjunctive conditional is true of the object, yet it lacks the disposition. Again, see Martin, “Dispositions and Conditionals”; Bird, “Dispositions and Antidotes”; and Molnar, “Are Dispositions Reducible?”

54

Rational Causation

explanations are necessarily limited because of the distinctive character of dispositional concepts, viz., they are only understood by reference to their manifestations and (in the case of non-rational dispositions) their conditions of manifestation. Because there is nothing more to having a disposition to X than normally to X in the appropriate circumstances (or, in the case of rational abilities, to X when one engages the disposition) to explain why an object Xs by appeal to its being so disposed can only convey that there was nothing abnormal about the occurrence. To summarize: I classify physical dispositions, organic functions, and rational abilities as species of a genus on the basis of the following similarities. In the case of each species, specific types explain the corresponding behaviors just in the sense that they show that it was no fluke that such behaviors occurred when they did. The presence of a disposition does not entail the occurrence of the behavior in the appropriate circumstances (or when a subject engages the disposition), though if there is no such behavior, this requires explanation. The occurrence, in the appropriate circumstances (or when the subject engages the would-be disposition), of the behavior that characterizes a disposition does not entail its presence, though if there is no such disposition, the occurrence requires explanation. There are also important similarities and differences between each species of disposition, which I discuss presently.

4.2 Differences between the Species A central difference between, on the one hand, physical dispositions and, on the other hand, organic functions and rational abilities, is that the actualization of a physical disposition does not admit of normative assessment. Vases do not break poorly or well. But individual heliotropic flowers do track the sun better or worse. And of course not every handstand is performed equally well.36 With respect both to organic functions and

36. It might be objected here that certain fragile objects break ‘better’ (into fewer pieces perhaps) than other fragile objects. But this ‘better or worse’ pertains not to the objects themselves but rather to their usefulness for certain purposes of ours (easier to clean up, less hazardous). In the case of a heliotropic flower, the standard is derived from it being the sort of flower it is.

Rational Explanation of Belief

55

rational abilities, we confront the same question concerning how precisely to describe cases in which the dispositions are not actualized properly. If a sunflower tracks the sun badly for a day, do we have an example of sun-tracking, albeit a bad one, or is the event not really an example of suntracking at all? Similarly, is a bad handstand still a handstand, or is it not really a handstand? However this issue is decided, we can say this much: the possibility that a heliotropic flower might track the sun badly on a particular day and that an able handstander might wobble and tilt during a specific attempt underscore the idea that a disposition can be activated yet fail to be successfully realized. Even if we prefer to describe such cases as ones in which a flower didn’t really track the sun or James didn’t really do a handstand, we plainly do not thereby forfeit our entitlement to view these events as manifestations, however subpar, of the relevant disposition. We saw above that when a physical disposition simply fails to be realized in the appropriate circumstances, this fact requires explanation. Given that the circumstances were right, why did it fail to materialize? Similarly, where an organic function or rational ability is actualized imperfectly, this also requires explanation. But there are differences between the cases. If a disposition simply fails to be actualized in the appropriate circumstances, what needs to be explained is the fact that an event of a certain sort does not occur. If a disposition is actualized imperfectly, what is explained is the occurrence of an event: e.g., a flower turns its face toward the sun, albeit in fits and starts or with unusual delay; a man uncertainly balances upside-down on his hands. The possibility of imperfect disposition-actualization is special to functions and rational abilities. And it gives rise to an interesting explanatory phenomenon, one that will be important in what follows. There is a striking asymmetry between explanations of successful and unsuccessful actualizations of dispositions. In the case of successful actualization, an appeal to a disposition can play an explanatory role that it can’t by itself play in the case of an unsuccessful actualization. If we watch a flower precisely maintaining its orientation toward the sun over the course of an afternoon, and you ask how this is possible (is it a trick? the wind? coincidence?), I can explain that it is heliotropic. That is what heliotropic flowers do. If we watch a flower do it more erratically, I can explain this in part by citing the fact that it is heliotropic. But the explanation is incomplete in this

56

Rational Causation

sense: there must be a further story as to why the flower failed to follow the sun more closely. Is there something wrong with the flower? Are the conditions abnormal in some way? The question “Why does this flower track the sun successfully?” is in this sense less demanding than the question “Why does this flower track the sun so badly?” The latter requires more explanation. Of course, in cases of successful and unsuccessful disposition-actualization, there are further questions: about the evolutionary history of this flower’s heliotropism and what mechanisms underlie it, for example. But in the imperfect case, the appeal to the disposition is itself unsatisfying in a way that it is not in the perfect case. There is more to explain, in cases of unsuccessful disposition-actualization than in cases of successful disposition-actualization. My aim here is ultimately to show that certain central features of the rational explanation of belief can best be understood when one sees believing-for-a-reason as the exercise of a rational ability. And this asymmetry (between the explanatory demands associated with perfect and imperfect disposition-actualization) is at the heart of this argument. We can see this asymmetry at work in the case of James and his ability to do handstands. James’s ability to do handstands is normally a part of a complete explanation of his doing a particular handstand. By citing this ability, we rule out that it was just chance or that he was supported by unseen strings, etc. But when the handstand is bad, we must also explain why he failed to do what he can, in fact, do. Is there something wrong with him? Are the conditions abnormal in some way? Again, there is more to explain in the case of failure than in the case of success. Whereas the normativity of organic functions and rational abilities provide a contrast with physical dispositions, rational abilities have their own distinguishing features. Intuitively we distinguish between a grown man’s ability to grow a beard (I don’t mean a man’s ability intentionally to let himself grow a beard) and his ability to do handstands. The former is something that happens to a person under certain conditions, if we can speak of the passage of a sufficient length of time as a ‘certain condition’; the latter is a product of his agency. Both manifest dispositions, but the former belongs with heliotropism—it is another organic function, a function of the human body. More generally, physical dispositions and organic

Rational Explanation of Belief

57

functions are tied to the circumstances in which their manifestations normally occur. Fragile objects break when dropped, poisons kill when ingested, current flows from a live wire if it is touched by a conductor, heliotropic flowers turn to face the sun, witch hazel blooms in winter, male humans grow facial hair starting at puberty, etc. In part because of the connection between what we think of as a person’s abilities and their agency, we see the following contrast between the traditional conditional analyses of physical dispositions and that of abilities. The antecedent of the conditional into which physical-dispositionstatements have traditionally been analyzed consists in the obtaining of an external condition (e.g., being dropped). The antecedent of the conditional into which some have analyzed rational-ability-statements is a volitional matter: an agent’s choosing or trying.37 Putting aside the reductive aspirations of this style of account, something about it seems right: special explanation is required not when a stimulus fails to bring about the exercise of a rational ability but rather when an agent’s attempt to exercise it is unsuccessful. This is in part why talk of what normally happens only finds a place in a characterization of the triggering-conditions of non-rational dispositions. The ability to do handstands does not bring with it a tendency to do handstands in some specific kind of circumstance. However, the abilities we possess qua rational beings are not limited to those whose manifestations are intentional actions. We have, in addition to practical abilities, theoretical abilities— e.g., the ability to remember the license-plate numbers of passing cars or the ability to use Bayesian reasoning to decide the likelihood of hypotheses. But the connection between volition and theoretical abilities is murky. The typical exercise of such an ability may not be an intentional action at all. Indeed, it is not entirely obvious that, say, remembering (as opposed to recalling) licenseplate numbers is the sort of thing that can be done intentionally. And even when it is, as in the case of making Bayesian inferences, a man may possess the ability even though he is unable to exercise it intentionally— perhaps he gets flustered when he sets his mind to it, as opposed to when he lets his thoughts take their own course. 37. See Roderic Chisholm, “Human Freedom and the Self,” in Free Will, ed. G. Watson (Oxford: Oxford University Press, 1982), 24–35.

58

Rational Causation

More to the point, S may know that p on the basis of an inference from q. His knowledge, on my view, is then constituted in part by the fact that he represents p as to be believed as a consequence of the to-be-believedness of q. But this is not an action at all—it is a state, one that need not be the result of any intentional act of deliberation on the question of p. Nonetheless, it is plausible to think that only beings capable of theoretical deliberation— creatures able to make up their minds about what to believe on the basis of rational considerations—believe for reasons. And making up one’s mind is, if not an intentional action, as much an expression of the freedom characteristic of rational creatures as are intentional actions. A full account of rational abilities—beyond what I can undertake here— would center on the intertwined practical and theoretical freedom distinctive of rational creatures, freedom that is inseparable from the sort of self-consciousness that looms large in my discussion of thought and action in this book.

4.3 The Ability Underlying Believing-for-a-Reason Believing-for-a-reason is the rational ability to extend one’s knowledge by representing some proposition as to be believed as a consequence of the to-be-believed-ness of some other proposition. In section 3 I discussed two ways in which believing-for-a-reason can go wrong: ignorance of the premise and ignorance of the fact that the premise establishes the truth of the conclusion. If, as I have been arguing in this section, imperfect exercises of rational abilities leave more to explain, then we would expect to see this borne out in the case of both kinds of imperfection. And we do see this. Suppose Poirot thinks that the butler must have done it because everyone else has an alibi, and he’s right to think so, and right to think so for that reason. Suppose further that I know all of this. In such a case, I know why Poirot thinks the butler did it, and I know it without qualification. Of course, there may still be much I do not know: how Poirot knew that everyone else had an alibi, or where he learned about the inferential significance of people having alibis, and so on. But this information goes beyond what “Why does Poirot believe the butler did it?” typically asks for. Consider by contrast the case of a rational explanation that appeals to an imperfect exercise of the ability underlying believing-for-a-reason. If I know that no one else has an alibi, I do not take myself fully to understand

Rational Explanation of Belief

59

why Poirot thinks the butler did it until I know why he ( falsely) thinks everyone else has an alibi. Similarly, suppose I know that Poirot believes that the butler did it because the guitar is a stringed instrument—i.e., he takes it to be believed that the butler did it as a consequence of its being to be believed that the guitar is a stringed instrument. In such a case, I am still very much in the dark about why Poirot thinks the butler did it until I know why he thinks the guitar’s being a stringed instrument shows that the butler did it. There are further questions in these cases because Poirot does not seem to me successfully to have exercised the ability appealed to in rational explanations. In these cases, too, there are those further questions about how he knows (or only why he believes) what he does or where he learned (or how he became misinformed) about the inferential significance of the premise. But, as with dispositional explanation more generally, there is more to explain in the cases of failure than in the cases of success. Since the ability underlying believing-for-a-reason is epistemic, actualizations of the ability that exhibit epistemic failure correspond to less illuminating and hence less satisfying explanations. The dispositional element in my account of rational explanation should not be overstated. The point is not that giving someone’s reason is like explaining a vase’s breaking by appealing to its fragility. The relevant analogue to explaining a vase’s breaking by appealing to its fragility would be explaining Poirot’s believing that the butler did it because everyone else has an alibi by citing his ability to believe for a reason. This is not a rational explanation of his belief. What rationally explains his belief, on my view, is that everyone else has an alibi. Yet part of what it comes to for a fact q to rationally explain a belief that p is, in ideal cases, for a believer who knows that q to recognize it as proof that p is true. And in so doing, the believer exercises the self-conscious, epistemic ability I have identified as underlying believing-for-a-reason. It is in this way that the notion of an ability enters into an account of what has happened when someone believes something for a reason. And it is the failure or less-than-perfect exercise of this ability that gives rise to the added explanatory burden. I would also deny that rational explanations explain the manifestation of a disposition by citing its triggering stimulus, in the way one might explain why a vase broke by citing its having been dropped. Physical dispositions are identified by their triggering circumstances and the eventtype (e.g., shattering, dissolving) that manifests the disposition. In a specific

60

Rational Causation

case, citing the trigger and citing the disposition itself are complementary parts of a more complete explanation of the manifestation. And so I can say the vase shattered because I dropped it and it was fragile. Here, the explanatory ‘because’ relates the manifestation to the trigger and the disposition. But the manifestation of the ability underlying believing-fora-reason is such that it is itself best described as an explanation. The manifestation of the relevant disposition is a fact such as that Poirot believed the butler did it because everyone else has an alibi. It is this sort of fact that is analogous to the vase’s breaking. To anticipate the discussion in Chapter 4, the exercise of this ability (and the corresponding practical ability) are themselves instances of rational causation. (For reasons alluded to above, I would deny that there is anything analogous, in the rational case, to the vase’s being dropped.) The relation between rational cause and effect, I will argue, can be modeled neither on the relation between trigger and manifestation nor on that between disposition and manifestation. Their relation must be understood instead in terms of the specific sort of doxastic representing I have described in this chapter (and the corresponding practical representing I describe in the next). This clarification should also quell any incipient suspicion that my project here is an attempt to revive Rylean dispositionalism about mental states. In arguing that rational explanations appeal to general rational abilities, I do not aim to show that believing-for-a-reason is dispositional as opposed to categorical. Believing-for-a-reason is a matter of how things are right now with the relevant subject, viz., his currently representing one proposition as to be believed as a consequence of the to-be-believed-ness of another proposition. His being this way right now no doubt has implications for how things would be were the subject different in this respect, but this does not distinguish believing-for-a-reason from any other property, mental or non-mental. And I am not, in any case, committed to these implications exhausting the nature of the relevant state. The rational explanation of belief appeals to a creature’s self-conscious rational engagement with the objective world. In ideal cases, a known fact is represented as a to-be-believed proposition, one that establishes the tobe-believed-ness of another proposition, which in turn also represents a thereby known fact. In the next section, I argue that that difference between rational explanations in their psychological and non-psychological

Rational Explanation of Belief

61

guises is just that the former leaves room for the possibility that the exercise of the ability underlying believing-for-a-reason is not fully successful. Because of this, such explanations are less explanatorily self-sufficient. The connection between world-citing rational explanation and successful believing-for-a-reason undermines a basic motivation for Psychologism.

5. Anti-Psychologism about the Rational Explanation of Belief This is Psychologism about believing-for-a-reason: P1: What it is for S to believe p for reason q is for S’s believing that p to be caused (in the right way) by S’s believing that q. P2: The causation that underlies believing-for-a-reason is efficient. I put off until Chapter 4 any discussion of causation. But these theses have implications about explanation. If the first thesis is true, then world-citing explanations should be explained in terms of mind-citing explanations. It is a central commitment of Psychologism that the latter is in this sense more fundamental than the former. Here’s how the view I’ve presented is anti-psychologistic. On my view, one cannot properly appreciate the role that appeal to the psychological plays except in terms of the potential failure of the corresponding nonpsychological explanation. S’s believing that p because he believes that q is also his believing that p because q, unless something has gone wrong. The efficacy of the subjective must be understood in terms of the capacity to engage with the objective. For this reason, I hold that world-citing explanations more perspicuously exhibit the character of the underlying rational ability than do mind-citing explanations. This reflects a more general truth. The nature of a particular ability resides in what it enables its possessor to do when it is exercised successfully.38 So, other things being equal, any style of explanation that appeals to a rational ability or function but that abstracts from its constitutive success-conditions must be less fundamental than one that does not so 38. Compare Rödl, Self- Consciousness, 151–158.

62

Rational Causation

abstract. Consider by way of comparison two possible explanations of a particular heliotropic flower’s movements over the course of the day: “It’s heliotropic,” and “it’s supposed to be heliotropic.” The latter will explain both the behavior of a sunflower that successfully tracks the sun as well as the behavior of one that does not. The former has a more problematic relation to cases of failure. If one appealed to heliotropism to explain poor sun-tracking, one would feel the pressure of an objection—“well, if it were really heliotropic, it wouldn’t still be facing the spot where the sun stood two hours ago.” But exactly because “it’s supposed to be heliotropic” abstracts from the question of whether the act explained was successful or not, it is less explanatorily fundamental than explanations that appeal simply to a flower’s heliotropism. Being heliotropic is having the disposition to track the sun. Being supposed-to-be-heliotropic is being an instance of a kind of thing that is heliotropic. The difference, if there is one, is that a defective instance of a kind might be denied the honorific ‘heliotropic,’ yet not denied the less-honorific ‘supposed to be heliotropic.’ Yet no one would argue that one should understand heliotropism in terms of should-beheliotropism. Any account of the latter must proceed by way of an account of the former. Mutatis mutandis, the same argument supports the thesis that mind-citing rational explanations are less fundamental than worldciting explanations. Let’s go back to the hybrid picture of world-citing rational explanations. On this view, to say S believes that p because q is to say, first, that S believes that p because S believes that q and, second, that S knows that q. The heart of believing-for-a-reason, on this view, is the first component, a study of which would reveal how reasoning really works. Just as the essence of visual perception lies in the component common to veridical experience and phenomenologically indiscernible hallucination—what the perceiver ‘can’t be wrong’ about—the essence of reasoning is the component common to believing-for-a-reason that culminates in knowledge and believing-fora-reason that doesn’t. The thinker ‘can’t be wrong,’ it seems, about her believing that p because she believes that q, though she can be wrong about her believing that p because q. Here is the problem with both views: the common component cannot be understood independently of the whole that it is supposed to explain.

Rational Explanation of Belief

63

Mind-citing rational explanations do not show how world-citing explanations ‘really work’—for the former say nothing more than the latter. They say less. For S to believe that p because S believes that q is for S to do something of a certain sort, perhaps successfully, perhaps unsuccessfully. The sort is: learn that p on the basis of q (as opposed to merely taking oneself to learn that p on the basis of q). The world-citing version requires that S knows that q. It is, in that sense, more demanding. And just because it comes closer to stating that the ability underlying believing-for-a-reason has been successfully employed, it exhibits that ability more perspicuously. It is thus wrong to understand world-citing in terms of mind-citing rational explanations of belief. P1 gets the order of explanation backward. Some of the appeal of the hybrid view derives from the neat way it fits with the physicalist sensibility of contemporary philosophy of mind and action. If rational explanation works by appeal to the psychological, and psychological states are physically realized, as many assume them to be, then the way is open to revealing the place of believing and acting for a reason in the causal processes that are describable from a scientific point of view. This motivation equally well explains the impulse to see the third-person point of view as more fundamental than the first-person point of view. This is also at odds with the approach I am developing here. Rational explanations on their face appeal to the perspective of the reasoner. Here is how Davidson puts it, in discussing the rational explanation of action: “A reason rationalizes an action only if it leads us to see something the agent saw, or thought he saw, in his action— some feature, consequence or aspect of the action the agent wanted, desired, prized, held dear, thought dutiful, beneficial, obligatory, or agreeable.”39 According to Psychologism, an explanation in terms of what the agent “saw, or thought he saw” can be presented with greater perspicacity when cast as an explanation in terms of the agent’s seeing or thinking he saw. Thus, the Psychologist takes explanations that appeal to the content of the reasoner’s perspective and interprets them as explanations that appeal to the perspective itself. He takes the third-person perspective to be explanatorily more fundamental than the first-person perspective. 39. Donald Davidson, “Actions, Reasons, and Causes,” in Essays on Actions and Events (Oxford: Clarendon Press, 1980), 3.

64

Rational Causation

I argue instead that appeal to the third-person in rational explanation, properly understood, implicitly makes reference to the first person. Giving someone else’s reasons does its explanatory work by situating the explained belief in a more complex rational state, viz., believing-for-a-reason. And part of the very idea of this state is that it is expressible in the form of a demonstration and reflexive explanation. Even the explicitly psychological forms of rational explanation are not fully third-personal. For they do not confer on someone who understands them knowledge of why a certain state obtains that abstracts from the believer’s point of view. Such explanations do not purport to reveal between explanans and explanandum a connection that is graspable apart from taking up the subject’s perspective. The rational-explanatory tie between a person’s believing that p and believing that q is a shadow—perfectly real but metaphysically secondary— of the inferential connection the subject finds (or merely takes herself to find) between p and q. Postulating such a psychological tie thus does not— on the surface at least— commit one to the existence of an efficient causal relation between psychological items, a relation that could obtain only if the psychological items were themselves physical. I return to the metaphysical upshot of this discussion beginning in Chapter 4. My point, I should stress, is not to deny that rational explanations are psychological. In emphasizing the point of view of the thinking subject, we cannot lose sight of the fact that what we are explaining is the subject’s view of the world. But if we treat the psychologically framed explanations as primary, we are led to a distorted conception of rational explanation. And this, in turn, also amounts to a distorted conception of psychology itself, one that founds a fundamental misconception of the place of the mind in nature. Once the psychologistic distortion is corrected, one can allow that rational explanations (whether they feature psychological vocabulary or not) are psychological—just on a non-psychologistic understanding of the psychological. In this chapter I have presented an anti-psychologistic view of the rational explanation of belief. Mind-citing explanations must be understood in terms of world-citing explanations; and third-personal rational explanations cannot be extricated from the perspective of the thinking subject. In Chapter 2 I offer an account of the rational explanation of action in the same spirit.

2 Rational Explanation of Action

The rational explanation of action is often referred to as belief-desire explanation. Although the label is generally viewed as specifying a topic, as opposed to a theoretical orientation, those who employ it tend to agree that to give someone’s reason for acting is to cite what are ultimately internal, physically realized psychological triggers of bodily movement. Our ‘folk theories’ of action (perhaps refined by a scientific psychology) specify the causal profile of the triggers that constitute our reasons for action, as opposed to mere causes of action. But this psychologistic approach to practical rationality is based on a misconstrual of the role played by psychological categories in the explanation of action. More specifically, it fails to appreciate that psychological terms are used not to postulate unseen internal items but rather to modify the manner in which we appeal to the exercise of the self-conscious, rational ability underlying acting-for-a-reason. The goal of this chapter is to describe that ability in enough detail both to account for the deceptive appeal of the orthodox view and to provide the conceptual basis for an alternative metaphysics, which I shall develop in subsequent chapters.

66

Rational Causation

Acting-for-a-reason, like believing-for-a-reason, is just someone’s making an inference— extending a normative status. The successful theoretical reasoner (in a central class of cases) recognizes that the alreadysettled-upon belief-worthiness of a proposition bears decisively on the belief-worthiness of another proposition. In an analogous fashion, the successful practical reasoner (in a central class of cases) recognizes that the already-settled-upon performance-worthiness of an action bears decisively on the performance-worthiness of another action. The basic elements of the inference are different: propositions in the theoretical case, actions in the practical case. And the statuses are also different: propositions can bear the status to be believed, and one proposition’s bearing that status can be the consequence of another’s; actions can bear the status to be done, and one action’s bearing that status can be the consequence of another’s. In ideal cases, a theoretical inference culminates in knowledge; and a practical inference culminates in right action. These important differences must be understood, however, in terms of overarching similarities. To give someone’s reason for belief or action is to explain by appealing to the exercise of the ability to make the relevant inference. Properly understood, this inferential view of practical reasoning also explains our ability to just say what we are doing and why, a phenomenon explored most famously by Elizabeth Anscombe in Intention, to which my thinking is deeply indebted.1 Just as a rational explanation of what one believes is an expression of one’s believing-for-a-reason, a rational explanation of what one is doing is an expression of one’s acting-for-a-reason. A person who is acting for a reason can explain her action by citing that reason not on the basis of evidence or observation but rather simply because she is intentionally performing the action for that reason. In both the case of believing- and acting-for-a-reason, the possibility of expression is due to the fact that what is expressed is nothing beyond the subject’s representing an item of the relevant sort (proposition or action) as 1. G.  E.  M. Anscombe, Intention (Cambridge, MA: Harvard University Press, 2000). However, whereas she speaks of this ability as involving a kind of knowledge (practical knowledge), I put the point differently. My reason is chiefly rhetorical: so labeled, it tends to be interpreted (wrongly, I would argue) as the thesis that we have a belief about what we are doing that we are justified in holding (albeit not on the basis of observation or evidence).

Rational Explanation of Action

67

inheriting a good-making status (to-be-believed-ness or to-be-done-ness) from another item of the same category. When we opt to explain someone’s performing an action (e.g., Dara’s tasting all the desserts) by explicitly appealing to her state of mind (e.g., her intending to review the restaurant) as opposed to her further action (e.g., her reviewing the restaurant), it is generally either because we doubt that the intended end-action is really in the offing or because we do not view the agent as having yet begun to perform the end-action. The mindciting guise of rational action-explanation thus takes on fewer ontological and epistemic commitments than the world-citing guise. Precisely for this reason, the former tells us less about the nature of the underlying practical ability and so is ill suited to serve as a basis for understanding practical rationality. However, each guise has the same purpose, viz., to pinpoint the source of the performance-worthiness of the explained action, from the agent’s point of view. Advocates of the proto-scientific interpretation of reasons-explanation strain to get this right but cannot. I defend this idea in four stages. I sketch the basic thesis that acting-fora-reason is the agents’ representing one action as to be done as a consequence of the to-be-done-ness of another action (section 1) and reply to objections (section 2). I argue that this conception of rational actionexplanation fits with a plausible account of teleological explanation more generally (section 3). Finally I argue that my non-psychologistic conception of action-explanation can, whereas psychologistic approaches cannot, vindicate the intuitive idea that what justifies an action can be exactly what explains it (section 4). This chapter is tightly focused on the conceptual issues pertinent to my diagnostic and constructive aims. There is, of course, much more to say about the topics in the philosophy of action that I address, not to mention the ones that I do not. Perhaps the most salient limitation of my discussion is the restriction to instrumental action-explanation. Just as the tobe-believed-ness of propositions derives not just from the to-be-believedness of other propositions, the to-be-done-ness of actions derives not just from the to-be-done-ness of other actions. Although I think that an account of the sort I offer could, in both the theoretical and practical spheres, be made more comprehensive, such an extension is not required

68

Rational Causation

to accomplish my objectives here. If my analysis of the point of deploying psychological vocabulary in rational explanations is correct in the domain of instrumental action, there is no reason why it should not hold, too, in the domain of non-instrumental action. Limiting the scope of my discussion enables me to convey the basic shape of the account of actionexplanation I favor while avoiding a variety of tangential complications.

1. Acting-for-a-Reason as Practical Thought The expressibility of believing-for-a-reason—the fact that we can just say what our reasons for belief are—is a part of its nature, I have argued. It is not far-fetched to postulate a connection between the possibility of articulate expression and belonging to the category of the mental; perceptual experiences, moods, emotions, and sensations all seem to be expressible in this sense. And yet actions and reasons for action, too, have this feature. As Anscombe argued, agents can speak with authority about what they are doing and why, yet not on the basis of observation or evidence. Since it is only the agent herself who can do this, it falls under the rubric of first-person authority. Statements made with first-person authority have the following distinctive feature. The person who makes them rejects the “how do you know?” question. If Nate says he has a pain in his leg or that he wants to fight Mike Tyson or that he is unsure who will be the next president, we recognize that, although the statement describes a contingent matter of fact, it would be inapposite to ask how he knew he was in the relevant mental state. If anyone else attributes these states to Nate, the “how do you know?” question is perfectly apt. It would be inapposite to ask Nate because we think that the particular sort of authority Nate possesses with respect to the relevant statements obviates his looking for evidence.2 But this also seems true of what he is doing or, in any case, what he is doing intentionally— our sole topic here. If I say that Nate is walking to the zoo, and you ask me how I know, the question is apt, and I can answer it

2. Compare David Finkelstein, Expression and the Inner (Cambridge, MA: Harvard University Press, 2003), and Dorit Bar- On, Speaking My Mind (Oxford: Oxford University Press, 2004).

Rational Explanation of Action

69

by producing evidence. If you call Nate and ask him what he is doing and he says that he is walking to the zoo, your question “how do you know?” is decidedly odd. Nate, if he feels like indulging you, may reply by producing evidence. He may mention visually apprehending his increasing proximity to the zoo, or if he is a philosopher of a certain stripe, he might invoke his intentions and what they portend. But however good these pieces of evidence might be, we do not ordinarily view the authority associated with his original statement as in any way derived from them. (And to the extent that we do think it is so derived, his action takes on a peculiar cast, as if he were just along for the ride in a body set on its own course.) We do not think that his statement is based on his recognition that the path he is taking leads right to the zoo. Nor do we think that his statement is expressive of a sophisticated but phenomenologically inconspicuous inference made on the basis of an introspective examination of his mental states. One might argue that our ordinary intuitions here are wrong, that, in the end, we will have to postulate some such mental arithmetic to explain his ability to ‘just make’ the statement. But the same could be said of our ability to just say that we believe that it is raining or that we are feeling pain. My point for the moment is that intentional action falls into the class of things that we ordinarily treat people as capable of speaking authoritatively about, yet not on the basis of any evidence or inference. This practice coexists peacefully with an understanding that an agent sometimes means to express what he is doing by saying “I am f -ing,” even though he is not f -ing. If, unbeknownst to Nate, there is no zoo, he cannot be walking there; and so there is no walking-to-the-zoo about which Nate can speak authoritatively. Our thinking of agents as speaking authoritatively about what they are doing is thus not a matter of their being infallible detectors of their own action. A question to which we shall return is why, if such errors are possible, we allow that an agent who is f -ing can just say so. One way that the expressibility of belief manifests itself is in the unintelligibility of a person whom we suppose to believe something, yet who is capable of being informed that he believes it. If you suppose me to believe that it is raining out, it makes little sense for you to tell me that I believe that it is raining out. And if I were to reply, umbrella in hand, “Do I? I had no idea,” your supposition would be difficult to sustain. Similarly,

70

Rational Causation

if you suppose me to be intentionally chopping down a tree, it makes little sense for you to tell me that I’m chopping down the tree. And if I were to reply, while swinging my axe, “Am I? I had no idea,” you would be hugely perplexed. The source of your perplexity in both cases is not a matter of wondering how such a typically reliable skill—telling what one believes or is doing—might have misfired or failed to engage. Rather, the very idea of such obliviousness is conceptually challenging. Of course, you could inform me that I was succeeding in performing the relevant action, that, e.g., the tree looked ready to fall. But to be informed of this presupposes that I am already in the process of chopping down the tree. If this is a typical case, then I was doing it even before I was certain I would succeed. Also, you can tell me what I did, which I may have forgotten. I speak authoritatively (in the relevant sense) about what I am doing, not about whether what I am doing will culminate in something I’ve successfully done, and not about what I have already finished doing. Having the right grammatical construction in mind helps to avoid confusion. The issue does not concern what one does. For a statement such as “I bicycle to work” does not say about oneself with respect to action what “I believe that it’s raining” says about oneself with respect to belief. For “I bicycle to work” self-ascribes not an action currently being performed but rather a habit. The practical analogue to what one believes is not what one does but rather what one is doing. I can, for example, express what I am doing by saying “I am bicycling to work.” Actions are often expressed in the imperfective form: “I am f -ing.” This is not to say that “I am f -ing” is always an authoritative expression of what one is doing. A late-night reveler may suspect that the loud music he is playing is disturbing his neighbors. He believes it after hearing pounding on the wall and cries out triumphantly “I am disturbing my neighbors!” But this expresses not what he is intentionally doing but rather his belief that he is (unintentionally) disturbing his neighbors. Where “I am f -ing” is an action-expression, it is an expression not of one’s belief that one is f-ing, but of one’s f-ing itself. This claim is analogous to my claim in Chapter 1 that a statement “I believe that p” can be an expression not of one’s believing that one believes that p, but of one’s believing that p itself. A passage from Anscombe’s Intention is suggestive of the special character of the statements that express actions:

Rational Explanation of Action

71

Now the statement: ‘The water is running out of a pipe round the corner’ stands in the same relation to the statement ‘I’m replenishing the house water-supply’ as does ‘My teeth are false’ to the order ‘Clench your teeth’; and so the statement (on grounds of observation) ‘You are not replenishing the house water-supply’ stands in the same relation to the description of intentional action ‘I am replenishing the house-water supply’ as does the well-founded prediction ‘The man isn’t going to clench his teeth, since they are false’ to the order ‘Clench your teeth’. And just as the contradiction of the order: ‘Clench your teeth’ is not ‘The man, as is clear from the following evidence, is not going to do any clenching of teeth, at least of the sort you mean’, but ‘Do not clench your teeth’, so the contradiction of ‘I’m replenishing the house water-supply’ is not ‘You aren’t, since there is a hole in the pipe’, but ‘Oh, no, you aren’t’ said by someone who thereupon sets out e.g. to make a hole in the pipe with a pick-axe. And similarly, if a person says ‘I am going to bed at midnight’ the contradiction of this is not: ‘You won’t, for you never keep such resolutions’ but ‘you won’t, for I am going to stop you’.3 Anscombe claims here that there is a difference between action-expressions and other self-predications that utilize the word-schema “I am f -ing,” and that this difference is a reflection of a difference in the underlying thoughts. The action-expression is the verbalized form, so to speak, of the action itself. Or rather, since what we are dealing with are not completed actions but actions-in-progress: the action-expression is the verbalized form of the action’s being under way. As such, its ‘contradiction’ is the verbalized form of another action’s being under way, viz., the thwarting of the original action. An action-expression is not the agent’s attempt to describe the world—it is not an assertion, which expresses belief—but rather something else. What else? And how does this help us understand how it is possible that someone can speak authoritatively about what they are doing, yet not on the basis of observation or evidence? Here is how it is possible in the case of belief: “I believe p” can express one’s stand on the question of whether p is to-be-believed and hence need 3. Anscombe, Intention, 55.

72

Rational Causation

not be based on any behavioral evidence or introspection. Similarly, here is how “I am f -ing” can be an expression of an action itself and not my belief that I am so acting: “I am f -ing” can express one stand on the question of whether f -ing is to-be-done and hence need not be based on any observational evidence concerning my progress toward a completed act of f -ing. Because actions just are agents’ representings of to-be-done-ness, they are also expressible. To be f -ing just is to represent f -ing as to be done. This is why the uneliminated possibility that the zoo might have closed, for all Nate knows, does not undermine his first-person authority. For that authority is grounded not in his exclusive access to a justification for believing a certain description of reality but rather in his ability to decide for himself what the thing to do is. My further contention is that just as “I believe that p because q” can be the verbal incarnation of believing-for-a-reason, “I am f -ing because I am y-ing” can be the verbal incarnation of acting-for-a-reason. Just as believing-for-a-reason is expressible because it is the believer’s representing one proposition as , acting-for-a-reason is expressible because it is the agent’s representing an action4 as .5 There is, then, a single sense in which both theoretical and practical inferences are valid: they preserve the good-making quality of the premise in the conclusion. It will be helpful to introduce this core thesis by comparing the role played by worldly facts in rational belief- and action-explanation. Here is

4. What agents represent as to be done are action-types (e.g., going to Yankee Stadium, which many people can do separately or en masse, a single time or repeatedly), not action-tokens (e.g., a completed trip to the stadium, with all the details fully determinate). I discuss this distinction at length in Chapter 5. 5. It is perhaps worth emphasizing that my view is not constitutivism. It is not Jones’s giving the explanation or his propensity to give the explanation that makes it true. For a constitutivist, this would be the order of explanation: S believes that he is f -ing because he is y -ing, so S is f -ing because he is y -ing. (Similarly, the order of explanation for the constitutivist in the theoretical case is: S believes he believes that p because q, so he believes that p because q.) I say, on the contrary: making a practical inference is a matter of representing f -ing as . His ability to just give the explanation is a consequence of his having made the inference.

Rational Explanation of Action

73

a point of similarity: actions, like beliefs, can be rationally explained by appealing to such facts, and both kinds of explanation presuppose the subject’s knowledge of the relevant facts. For example, if Jones knows that the 4 train goes to Yankee Stadium, this fact is available to serve as his reason for acting. If an observer credits him with this knowledge, then, upon seeing him board the 4 on a game day, bedizened with Yankee paraphernalia, she might say: “Jones is taking the 4 train because it goes to Yankee Stadium.” Such an explanation attributes a thought to Jones that he can express by saying “I am taking the 4 train because it goes to Yankee Stadium.” If the 4 train did not go to Yankee Stadium, we could not explain his taking it by appeal to the fact that it goes there but would instead have to revert to an explanation that explicitly cites his beliefs: “Jones is taking the 4 because he believes it goes to Yankee Stadium.” In ideal cases of believing-for-a-reason, the fact that explains why S believes that p—viz., q—is represented by a proposition that entails p. And precisely to the extent that q falls short of entailing p (i.e., failing to show that it’s true, failing to show that it’s very likely, failing to show that it’s even remotely likely), q fails to explain rationally why S believes that p. The point is not that the thinker makes a ‘step in logic.’ Rather the point is, as we saw in Chapter 1, that the degree of light shed by a rational explanation of belief is dependent on the strength of the inferential connection between the corresponding propositions. The weaker the connection, the more “S believes that p because q” leaves unexplained. The ability to reason theoretically, when exercised fully successfully, involves a grounding belief that q and a grounded belief that p, such that q entails p. The role of a worldly fact q that explains why S is f -ing is fundamentally different. Such explanations do not depend on q’s entailing the truth of “S is f -ing.” Nor do they depend on “S is f -ing” being entailed by the conjunction of “q” (or “S believes that q”) and an instance of “S wants to y.” In fact, the very idea that statements such as “S wants to y ” and “S is f -ing” figure in practical reasoning in the way that “q” and “p” figure in the thought expressible by the demonstration “q, so p” rests on a mistake.6

6. Anscombe explicitly makes this point in G. E. M. Anscombe, “Practical Inference,” in Virtues and Reasons, ed. R. Hurst house, G. Lawrence, and W. Quinn (Oxford: Oxford University Press, 1995), 1–34. However, see page 7 for a Dancy-like exception.

74

Rational Causation

When a thinker believes-for-a-reason, the thinker extends the status of to-be-believed-ness from q to p. In the case of theoretical reasoning, this status is thus transmitted through and to what is believed, viz., propositions. As we have seen, such theoretical reasoning also puts the thinker in a position to express a rational explanation of her own belief: “I believe that p because I believe that q.” But the primitive elements of the inference are simply the propositions p and q. In ideal cases of practical reasoning, it is not to-be-believed-ness but to-be- done-ness that is preserved: an agent extends the status of to-be- done-ness from y -ing to f -ing, i.e., through and to what is done. To make this inference is at least to believe— in ideal cases, to know— that f -ing constitutes progress made toward completing an act of y -ing. The instrumental fact thus plays a merely facilitating role in the practical inference. For Jones to take the 4 train because it goes to Yankee Stadium— i.e., because of an instrumental fact— is just for him to take it because he represents going to Yankee Stadium as to be done— i.e., because of an end-action. Thus, we must understand the role of worldly facts very differently in theoretical and practical cases. I will say more about this in section 4. Phrases such as ‘to be done’ or ‘to-be-done-ness’ are sometimes used by philosophers to express a status that is normatively neutral. On such uses, to say that one regards f -ing as to-be-done is just to say that one has an attitude with a certain direction of fit toward a proposition describing my f -ing, viz., the attitude is satisfied iff the proposition becomes true. This is not how I use these terms. On my use, there is a distinction between an agent’s regarding an action as to be done and its really being to be done. To act for a reason is to exercise an ability that, when exercised successfully, is the agent’s doing something that is in fact to be done. If an agent should not f, all things considered, then f -ing is not to be done, regardless of how she may represent it, and the attitude is to that extent defective. This standard is, furthermore, not just internal to the attitude’s being what it is—viz., an attitude that is defective to the extent that it is directed toward what should not be done—it is also internal to how the agent understands the attitude. It is a condition of its possession that the agent recognize the relevance of considerations that weigh in favor or against its object’s being what she should do. To put it a different way, an

Rational Explanation of Action

75

agent understands the attitude as answering the question “what should I do?” and not just “what shall I do?”7 In the next section I consider three sorts of objection to the view of action just sketched. First, there are, it seems, counterexamples to the thesis that we can just say what we are doing. Specifically, there are (a) cases of self-deception in which people sincerely deny doing what they are, in fact, intentionally doing and (b) cases of action-types (e.g., winning a race) such that people are generally not in a position to just say whether they are performing the relevant action or not. Second, my proposal seems to assume that recognizing the goodness of an action guarantees the motivation to perform it and also that actions are performed ‘under the guise of the good.’ Both theses are highly controversial. Third, even if representing an action as to be done guarantees an intention to perform the action, it surely does not guarantee the actual per for mance of the action. Hence performing an action cannot possibly be the same as representing it as to be done. Beyond these objections, some will view the resultant notion of action as intolerable, because it is inconsistent with (1) the fact that non-human animals act, (2) the metaphysical structure of events, and/or (3) a plausible understanding of the place of action in the causal structure of the world. I will address these issues (among others) in Chapters 3, 5, and 6, respectively.

2. Objections 2.1 Objections to the Expressibility of Actions and Reasons for Action I start from the premise that actions are as such expressible. But what of cases involving self-deception? People are often very much in the dark about what they are doing and their reasons for doing it. Take, for example, Joe, who is wedded to an image of himself as a loving sibling, one who

7. Matthew Boyle and Doug Lavin, “Goodness and Desire,” in Desire, Practical Reason, and the Good, ed. S. Tenenbaum (Oxford: Oxford University Press, 2010), 161– 202, 169. To say that an agent regards f -ing as to be done, in my sense, is thus also not merely to say that it is expressive of a plan to f. Thus my use of the phrase differs from Allan Gibbard’s in his Thinking How to Live (Cambridge, MA: Harvard University Press, 2003).

76

Rational Causation

would seek to do whatever is necessary to contribute to the happiness of his brother, Frank. But suppose further that at every turn, under a thin cover of helpfulness and sympathy, Joe subtly and diabolically diminishes Frank’s self-esteem and thereby his chances of success. The manner in which this is carried out relies, we can imagine, on a very high degree of sensitivity to Frank’s psychological vulnerabilities. We accept the possibility that when Joe denies that he is sabotaging Frank, he is sincere—in the same diminished sense that Dan, our selfdeceived, emasculated husband from Chapter 1, is sincere when he bemoans that his wife’s new job would reduce their quality time. We ordinarily deploy some not-especially-illuminating expressions to describe this situation. We might say that Joe refuses to acknowledge his malicious behavior, that he is hiding it from himself, that he is in denial, that ‘on some level’ he knows what he is doing, even if ‘on another level’ he does not. More telling than our uncertain and easily shifting vocabulary for describing such cases is the fact that we view his treatment of his brother as utterly damning, an attitude that would be irrational if Joe had no better than an observer’s relation to his actions. His being self-deceived is not any sort of mitigating factor when it comes time to judge him. It is not only not exculpatory (as it would be if Joe did not know ‘on any level’ that he was sabotaging Frank); it is not even an excuse. If anything, it is an exacerbating circumstance, adding cowardice and dishonesty to cruelty. We judge him for his failure to acknowledge what he is doing. This judgment makes sense only if he has the ability to fess up. Ought implies can. The case is surely puzzling. The very concept of self-deception has a whiff of paradox about it.8 To say that Joe is self-deceived requires that he is the deceived and the deceiver, a peculiar position that involves thinking of him as playing what seem to be incompatible roles. And there is a tendency to try to make this easier than it is by distinguishing between roleplayers.9 But this is not a mild case of multiple personality disorder; it is

8. See Stanley Paluch, “Self-Deception,” Inquiry 10 (1967): 268–278. 9. See, for example, Donald Davidson, “Paradoxes of Irrationality,” in Philosophical Essays on Freud, ed. R. Wollheim and J. Hopkins (Cambridge: Cambridge University Press, 1982), 289–305.

Rational Explanation of Action

77

not Joe simply forgetting what he knew at an earlier time; it is not a case of Joe’s brain possessing a piece of information that Joe himself lacks. Were it any of these circumstances, Joe would be as much a victim as Frank. What distinguishes actions that are performed intentionally yet unconsciously from what is done unintentionally is just that the subject of the former, unlike the subject of the latter, is in a position to speak about the action from the agent’s point of view. Of course, there are, as was discussed in Chapter 1, circumstances that might prevent the exercise of the ability: e.g., physical incapacity or trouble finding the right words. There are also psychological pressures that can inhibit the exercise of the ability, as in Joe’s case. But putting these aside, a case in which someone simply cannot fess up to what he is intentionally doing is not just abnormal; it is inconceivable. To the extent that we become convinced that someone is entirely incapable of just saying what he is doing, we do not treat the relevant action as intentional. And to the extent that we are convinced that an action is intentional, we do not give up on the idea that the agent could just say what he is doing, even if he cannot bring himself to admit it. If his only path to speaking authoritatively about the ‘action’ is the same as an observer’s, then it is not an intentional action at all. We do not conceive of a person’s doing something as fully separable from his ability to say what he is doing. I argued in Chapter 1 that when we take someone to believe that p because q, in the rational sense, we thereby also take him to be in a position to express this in the form of an explanation of his own belief. If someone denies that he believes that p because q, then, unless we take him to be insincere or deeply self-deceived, we abandon the explanation. And if we do take him to be insincere or self-deceived, we do think that it is within his power to ‘just give’ his reasons. But nothing in that discussion turned on the reasons being reasons for belief, as opposed to action. And so the same conceptual difficulty that arises when trying to co-conceive Joe’s intentionally sabotaging Frank with his not being in a position to just say so arises equally in trying to co-conceive his doing something else (e.g., finishing one of Frank’s sentences) in order to sabotage Frank and his being unable to say that this is his reason for doing it. If someone denies that he is f -ing because ____, then, unless we take him to be insincere or deeply self-deceived, we abandon the explanation.

78

Rational Causation

* * * But there is still a potential hitch. Even putting cases of self-deception aside, there is another class of would-be counterexamples to the thesis that agents can speak authoritatively about what they are doing, yet not on the basis of observation or argument. If someone has won a marathon, earned an ‘A’ in Latin, and bowled a perfect game, we would (typically) credit him with having done what he meant to do. Looking back, we might describe what was happening in medias res using these verb-phrases in the progressive: he was winning the marathon, earning an ‘A’ in Latin, and bowling a perfect game. Yet at those times, without the benefit of hindsight, it is not at all clear that he was in a position to say one way or the other whether he was, in fact, achieving any of these things. During those times, the agent can, it seems, be informed that he is intentionally doing these things. Thinking about these cases demands a certain amount of care, however. Is it so clear that we are dealing with intentional actions? In fact, it is rather odd to think of someone performing an action intentionally under one of these descriptions. Consider by way of evidence the peculiarity of applying the ‘why?’ question that asks for a reason in these cases: “Why is he winning a marathon?”; “Why is he earning an ‘A’ in Latin?”; “Why is he bowling a perfect game?” It is natural to hear these questions as asking for what explains his success, not what his reasons are. Contrast these with “Why is he running a marathon?”; “Why is he taking Latin?”; and “Why is he bowling?” The fact that it is, at the very least, awkward to think of the counterexamples as performed for reasons should give us pause in simply accepting that they are performed intentionally. Surely there is a connection between the fact that an agent seems to need evidence in order to speak authoritatively about performing an action under one of these descriptions and the fact that demands for their explanation seem to ask for evidence of success, rather than reasons. But what is the connection? It is clear what would need to be added to these examples in order to hear the questions as asking for reasons. In each case, the more that we view it as within the man’s power to achieve the relevant result, the easier it is to hear a “why?” question as asking for a reason, and exactly to that extent we are willing to credit him with being in a position to speak authoritatively about performing the action. To say that it is within someone’s power to do something is to say that his success in performing the

Rational Explanation of Action

79

action needs no external explanation. A man who can run twenty-five miles in an hour can win or lose a marathon at will. Only his failure to win a marathon would need to be explained. In his case, it’s easy to hear the question “Why is he winning the marathon?” as asking for a reason. And it is precisely to this extent easy to hear him as speaking authoritatively about what he’s doing when he says “I am winning a marathon.” But then mustn’t a man know that he is able to f in order to speak authoritatively about a specific instance of his f-ing? And doesn’t this mean that his so speaking is based at least in part on his evidence that he is capable of f-ing? This line of thought is tempting only if we think that what is at issue is the agent’s estimation of whether a certain proposition is to be believed or not, and that when he says “I am f-ing,” this statement expresses the resultant judgment. In fact, it is simply an expression of his representing f-ing as to be done—this is, on my view, his f-ing. Knowledge that one is capable of f-ing may enter into whether one can f intentionally, but the evidence on which such knowledge is based is not what underwrites the authoritativeness of my statement that I am f-ing. This authority is rooted in my representing an action as to be done, not in my representing a proposition as to be believed.

2.2 Objections to the Identification of Evaluation with Intending to Act How is it possible that we can just say what we are doing? On my view, part of what it is to be f -ing intentionally is to be able to just say that one is f -ing, and not on the basis of observation or evidence. This pushes the question back to: How is it possible that we intentionally f, for any f, given that it consists in part in the ability to just say that one is f -ing, and not on the basis of observation or evidence? My answer: because intentionally f -ing is representing f -ing as to be done. Specifically, it is the sort of representing as to be done that is the characteristic product of making up one’s mind about what to do. Speaking with authority about what one is doing (and why) is no more mysterious than speaking with authority about what one has made up one’s mind to do (and why).10

10. See Matthew Boyle, “ ‘Making up Your Mind,’ ” for an insightful examination of the question of how we can speak authoritatively about what we have made up our minds to do.

80

Rational Causation

It is not that I can speak authoritatively about what I am doing because I know what I intend to do and also something else, e.g., that my body seems to be headed toward my destination. Rather, my ability to speak authoritatively about my intentions and my ability to speak authoritatively about my actions are a single ability. And that is because intending to f and f -ing are not, I argue in section 2.3, fundamentally different sorts of things. Putting this issue aside for the moment, however, other difficult issues cluster in the vicinity of any attempt to identify evaluation with doing something or intending to do it. My proposal seems at home in the neighborhood of Motivational Internalism, according to which a judgment of the good necessarily motivates, and the ‘guise of the good’ thesis, according to which intentional actions necessarily aim at the good. My view thus appears vulnerable to counterexamples to this theses: e.g., accidie, in which an agent fails to pursue or even to be motivated by recognized goods, and akrasia, in which an agent acts against her better judgment. Given the larger aims of this book, a provisional discussion of these matters will have to suffice.11 My aim in discussing them is just to show that there are plausible reasons to doubt whether my view entails the impossibility of any ordinarily recognized forms of practical irrationality. First, let me be clear about what versions of these theses are in play here. Above I identified the relevant sense of to-be-done-ness by reference to its role in answering the question “what should I do?” This question can be heard as asking about the morally right thing to do. So understood, my identification of acting with representing as to be done would be a nonstarter, since relatively few of our actions have moral import. I had no moral motivation for eating strawberry ice cream last night, nor did I view eating it as morally good. It is rather a broader sense of “what should I do?” that helps define the relevant sense of ‘to be done.’ To-be-done-ness is what Sergio Tenenbaum calls a formal end of an activity, “the end one must ascribe to an agent insofar as she is engaged in that activity.”12 He cites, by way of example, winning as the formal end of

11. For an especially searching examination of these issues, one whose upshot is broadly congenial to my approach here, see Sergio Tenenbaum, Appearances of the Good (Cambridge: Cambridge University Press, 2007), especially chapters 7 and 8. 12. Ibid., 6.

Rational Explanation of Action

81

engagement in a competitive game. If someone’s activity is not rightly judged according to the contribution it makes to winning, then that person is not playing in a competitive game. Furthermore, part of playing a game is recognizing the relevance of assessments of how successfully one’s activity is contributing to winning. Tenenbaum argues that ‘the good’ plays an analogous role in practical reasoning. Here is how Boyle and Lavin make the point: Just as a rational believer is one who can reflect on his grounds for belief by putting to himself the question “Why p?” so a rational agent is one that can reflect on his grounds for action by putting to himself the question “What speaks in favor of doing A?” Just as “truth” names the standard we apply in answering the former question, so “goodness” names the standard for answers to the latter: It specifies the topic on which a consideration must bear in order even to be a candidate answer to the question.13 To say that intentionally f -ing is representing f -ing as to-be-done, in my sense, is not to link acting with ‘doing good,’ on any substantive construal of that phrase. It is to say that an agent’s representing an action that is in fact to be done as to be done is the standard against which the outcome of practical deliberation is measured. Ideally, deliberation culminates in the performance of an action that is to be done, not from the perspective of justice or pleasure or eschatology, but all things considered. The relevant senses of ‘should,’ ‘good,’ or ‘to be done’ that characterize the point of practical deliberation are no more tied to a specific standard of conduct than are subsequent judgments about whether the action should have been performed. Someone who rejects a criticism of her action of the form “in light of p, you shouldn’t be f -ing” does so because she disputes p’s truth or the specific standard of conduct that citing p implicitly brings to bear or the relevance or salience of that standard. She does not do so because she doesn’t care about what she should do, unless ‘should’ is understood as tied to a specific standard: moral, prudential, or whatever. 13. Boyle and Lavin, “Goodness and Desire,” 191.

82

Rational Causation

Even with this qualification, however, it will seem doubtful that merely viewing an action as to be done is sufficient for performing it. My thesis still requires a very strong version of Motivational Internalism, even if it is not moral-motivational internalism. Since representing an action as to be done is performing it, it necessitates possession not just of some motivation for performing it but of suffi cient motivation. And this entailment seems to fly in the face of everyday experience, in which we fail to act or even to be motivated by what we judge to be the right thing to do, all things considered. My reply to this objection will turn in part on a distinction I made in Chapter 1 between engaged and disengaged deliberation. The distinction is not, I emphasize, meant to bear the full weight of the relevant version of Motivational Internalism. This book as a whole is a defense of an approach to a family of interconnected conceptual and metaphysical issues; this version of Motivational Internalism is a consequence of, and derives it plausibility from, the whole theory of which it is a part. But I contend, further, that my quasi-technical expression ‘representing as tobe-done’ is well grounded in an ordinary evaluative concept and that my identification of acting with its application does not prevent us from recognizing that people are often not motivated to do what they think they ought. Consider a man who examines the t-shirts in his closet with a view to deciding what to wear to the World Trade Organization protest. Certain possibilities— e.g., Dr. Dog, The Strand—are ruled out as irrelevant, some— e.g., Che, Warhol’s Mao— as too trendy, others— e.g., Kropotkin, Goldman— as too obscure. Finally he sees “Cops Hate Freedom”: fitting, bold, and likely to attract all of the right kinds of attention. He concludes he should wear it and does. Is his deciding to wear it a separate step from his concluding he should wear it? Must there be, for him, a further question of whether he shall do what he should? An affirmative answer seems obligatory in light of the fact that these questions can come apart. A man may say sincerely that he ought, all things considered, to fi x his faulty brakes as soon as possible. He comes to this conclusion, we can suppose, after contemplating the danger of continuing to drive and the relative importance of his health over any occupation or hobby. And yet instead of driving to the mechanic, he drives to the casino. For him, in this situation, deciding what he should do is one thing, decid-

Rational Explanation of Action

83

ing what he shall do is another. Does the fact that these questions can come apart show that they must come apart? I say “no.” My description of these two cases leaves open the possibility that the protester and the gambler deliberate with very different attitudes. The protester, we can imagine, has simply given himself over to the results of his deliberation about what to wear. His will is on the line as he assesses his options. It is engaged deliberation. The gambler, on the contrary, is disengaged from his deliberation. He is not making up his mind about what to do when he considers the case in favor of fi xing his car. His conclusion is just one factor when it comes time to decide. When I say that to f is to represent f -ing as to be done, I mean that it is to regard f -ing the way an engaged practical deliberator regards her conclusion, which she can express equally either by saying “I should f” or “I shall f.” Someone who believes that she ought to f, yet has not made up her mind to f, will regard f -ing the way a disengaged deliberator does. She may believe that f -ing is the thing to do, she may desire to f, but she has not represented f -ing as to be done in my sense. For representing f -ing as to be done is not believing or desiring (or ‘besiring’), it is f -ing—or, to adopt the more careful formulation to be discussed in section 2.3, it is either f -ing or something that must be understood as a qualified form of f -ing, e.g., being in an early phase of f -ing, or highly defective f -ing. It is worth emphasizing that my claim is not that representing an action as to be done is necessarily the outcome of engaged deliberation, rather just that it is the characteristic product of such deliberation. Picking up a $100 bill off the street does not require deliberation, but the agent does regard picking it up in the same way as someone who deliberates (in an engaged manner) to that conclusion. It might be argued that it is question-begging to defend my view by appealing to the distinction between engaged and disengaged deliberation, since it simply presupposes just what is at issue, viz., that there are cases in which what one shall and what one should do are not separate questions. But although this distinction does settle what is at issue, invoking it is not ad hoc. For we are all perfectly familiar with the phenomenon of engaged deliberation; it is as psychologically real and no less ‘ordinary’— and thus no less deserving of philosophical deference— than believing one ought to do something that one then fails to do. All other things being equal, we should accommodate both phenomena. And we can.

84

Rational Causation

The fact that statements of the form “I should f” can express either a mental state characteristic of engaged or disengaged deliberation accounts for why such statements can give rise to a familiar sort of miscommunication. Consider Graham and Valerie, a married couple who decide to save money by ditching Netflix and their satellite dish in favor of illegally downloading the films and television shows that they like to watch. Graham, however, thinks it is wrong to do, not just morally, but all things considered. Despite being the impetus behind the original decision and the chief implementer of the plan, he has felt this way all along. Until Graham explains his misgivings, Valerie hasn’t given it much thought. Upon hearing his explanation, Valerie, convinced, says “We should stop.” “I agree. That’s my point—we should never have started,” Graham replies. Valerie, irritated: “No, I mean we really should.” For Valerie, the decisiveness of the case against illegal downloading settles the practical question. This fact determines what she ‘means’ when she says “we should stop.” Graham ‘means’ something else by “we should stop”: for him, the practical question is not settled. My view, then, is that one is ipso facto sufficiently motivated to f when one represent f -ing as to be done, since f -ing is representing f -ing as to be done. The argument in favor of the relevant version of Motivational Internalism (and, similarly, for the ‘guise of the good’ thesis) is just the argument in favor of my view as a whole. Examples of weakness of the will and the like show that we have a concept of ‘the thing to do’ such that (a) its application to an action is consistent with not being motivated to perform it and (b) its non-application to an action is consistent with performing it. But these examples are consistent with our also having a (surely not unrelated) concept of ‘the thing to do’ such that (a) and (b) are false. It’s this latter concept that I use to help define ‘representing as to be done.’ One strategy (Aristotelian) for defending internalism invokes some version of this idea: an agent who judges a certain course of action to be good with a full understanding of its grounds is necessarily motivated to pursue it.14 Failure to be properly motivated is a consequence of a less than complete grasp of why he should perform the relevant action. According to 14. See John McDowell, “Virtue and Reason,” The Monist 62 (1979): 331–350; and Tenenbaum, Appearances of the Good.

Rational Explanation of Action

85

another strategy (Kantian), it is to the extent, and only to the extent, that our will is rational that “should” and “shall” converge.15 Both strategies are congenial to my approach here. My aim has only been to remind the reader that the limited sort of internalism I advocate here has a firm basis in our experience of agency. It should be clear now how I handle challenges to the ‘guise of the good thesis.’ The version of that thesis to which I am committed says that if someone is intentionally f -ing, then she represents f -ing as to be done. The classic counterexamples to this thesis involve akrasia: cases in which someone f s instead of y-ing, despite thinking that y-ing is, all things considered, better than f -ing. There are also cases of perversion and cussedness, in which agents do what they know that they shouldn’t just because they shouldn’t.16 My strategy in each sort of case is to argue that the agent does not represent the un-chosen action as to-be-done in my sense. Here is a case that combines akrasia and perversion: a man attending the funeral of a cherished friend might be seized by a desire to do the absolute wrong thing— say, make a pass at the deceased’s widow. He might do it, and do it just because he felt he shouldn’t, despite believing that consoling her is the right thing to do. On my view, when he decides to act on his desire to do the wrong thing, he represents doing the wrong thing—the thing that egregiously violates the situationally salient principles of morality and etiquette—as to be done. And so if someone advised him that “around these parts,” people are supposed to make passes at the widows of cherished friends, he would view this fact as establishing that he shouldn’t do it after all. The to-be-done-ness of making a pass at the widow is, from this man’s point of view, inherited from the to-be-done-ness of doing just the wrong thing. There is no contradiction in knowingly representing doing the wrong thing as to be done, for what’s to be done is not, as I said above, any kind of substantive classification. 15. See Christine Korsgaard, “Skepticism about Practical Reason,” Journal of Philosophy 83 (1986): 5–25. 16. For a detailed discussion of these and other putative counterexamples to the ‘guise of the good’ thesis, see Michael Stocker, “Desiring the Bad: An Essay in Moral Psychology,” Journal of Philosophy 76 (1979): 738–753. See also David Velleman, The Possibility of Practical Reason (Oxford: Oxford University, 2000), chapter 5.

86

Rational Causation

I claimed above that my view does not commit me to denying any phenomenon that we ordinarily recognize. Yet it seems to rule this out: our gambler sees getting his brakes fixed in the way that an engaged deliberator sees the conclusion of her practical reasoning and yet fails to follow through, despite no external impediment or perceived change in the balance of considerations. In fact, this is perfectly consistent with my view. The gambler in this scenario goes from regarding getting his brakes fixed as to be done to regarding it merely as what he should do, all things considered. On my view, this is analogous to the sort of doxastic backsliding described in Chapter 1.17 Of course, how one interprets any of these examples will depend in large part on one’s antecedent philosophical commitments. If recent discussion has shown anything, it is that both defenders and critics of the ‘guise of the good’ thesis have the resources to explain any example in a manner that renders it consistent with their own views.18 Indeed, that is all I have tried to do here.

2.3 Objections to the Identification of Evaluation with Acting According to the deeply entrenched orthodoxy, actions are bodily events that are distinguished from non-actions by their causal history. So conceived, it would be natural to think that we are at best in a position to speak authoritatively (yet not on the basis of observation) only about the mental causes of action— about our intentions, say— and that we find out what we are actually doing by observation.19 Failing that, one might argue instead that our authority is expressive of a (non-evidence-based) belief

17. Richard Holton argues that this sort of failure of resolution is what the ordinary expression ‘weakness of the will’ denotes. See his “Intention and Weakness of the Will,” Journal of Philosophy 96 (1999): 241–262. 18. This point is made by both Boyle and Lavin in “Goodness and Desire” (a defense of the thesis) and Kieran Setiya in “Sympathy for the Dev il” (a critique of the thesis), also in Desire, Practical Reason, and the Good, ed. Tenenbaum. 19. Kevin Falvey calls this the ‘two factor’ thesis. See his “Knowledge in Intention,” Philosophical Studies 99 (2000): 21–44. He rejects it, invoking many of the same considerations in Anscombe’s defense that I adduce here. But he errs in thinking of our stating what we are doing as our offering a description, which would make it the expression of a belief about what one is doing.

Rational Explanation of Action

87

about what we are doing that is itself the cause of what we are doing,20 or that we infer what we are doing on the basis of the evidence provided (chiefly) by our intention to do it.21 If to f is nothing besides representing f -ing as to be done, however, then, there need be no great mystery about how we can say what we are doing. But that is a big ‘if.’ Representing an action as to be done may be necessary and sufficient for intending to perform an action, but it is plainly insufficient for actually performing the action. There are, first, intentions for the future, which are distinguished from intentions in action exactly by the fact that one is not yet performing the relevant action. Second, there are cases in which one fails to be doing what one intends to be doing on account of a gross mismatch between one’s intention and the world. How do we get from authoritatively speaking of what one intends to do (because it is our representing the intended action as to be done) to authoritatively speaking about what one is, in fact, doing? I will discuss both kinds of case in what follows. But it is worth sketching at the outset the gist of my reply. The fact that an agent might represent a course of action as to be done and yet not be doing it makes it seem as if actually doing it (i.e., being in the midst of doing it) is something beyond representing it to be done, something extra about which the agent may or may not be in position to speak authoritatively. Yet the two justmentioned kinds of cases are consistent with there being nothing extra in this sense. To see this, consider the relation between an English oak tree and the species Quercus robur. There are instances of the latter that are not instances of the former—acorns, for example. But this does not tempt anyone into thinking that an English oak is something beyond an instance of Quercus robur, an instance plus some extra-Quercus-robury thing. There is no prising apart its maturity from its oakishness, no ‘bare’ adulthood that seeps with age into its otherwise vegetable nature. The nature of action, I contend, is exhausted by its being the representing of to-be-done-ness,

20. See David Velleman, Practical Refl ection (Princeton, NJ: Princeton University Press, 1989), and Kieran Setiya, “Explaining Action,” Philosophical Review 112 (2003): 339–393. 21. See Sarah Paul, “How We Know What We’re Doing,” Philosopher’s Imprint 9 (2009): 1–24.

88

Rational Causation

even if some such representings neither develop nor even have a chance of developing into fully fledged doings. The analogy is not exact; but it highlights the conceptual space within which I shall maneuver. A week from today, Geoff will climb into the cockpit of a small airplane and pilot it from Newark to London. That, in any case, is his plan. Let us stipulate that he represents both climbing into the cockpit and flying the plane to London as to be done, and that he represents the former as to be done as a consequence of the to-be-done-ness of the latter. At the moment, he is walking south on Broadway between Seventy-First and Seventieth Streets. Climbing into the cockpit and flying to London would seem to be paradigmatic future intentions, and not intentions in action. And yet this is not so clear. For there are no hard and fast rules delineating when we may describe someone as f -ing and when we may describe someone as merely having an intention to f in the future. To see this, it helps to consider first a related point, regarding two uses of the progressive. Here is how Kevin Falvey makes the distinction: A person may be doing something, in a suitably broad sense, when at the moment she is not doing anything, in a more narrow sense, that is for the sake of what she is doing in the broad sense. Suppose a friend stops by my house and wants to go for a walk, and I say, “I can’t; I’m making bread.” This could be true even if as I say it I’m sitting on the couch reading the newspaper—perhaps I’m waiting for the bread to rise before putting it in the oven. The question what someone is doing sometimes calls for a more narrow answer, and context determines how narrow an answer is indicated.22 It is not entirely obvious that what’s at issue here are difference senses of the progressive. But this much is incontrovertible. A single description (“I’m making bread”) that in one context would be perfectly acceptable (e.g., in response to “How about a walk?”) might not in another (in response to “What are you doing on the couch?”). One might think that

22. Falvey, “Knowledge in Intention,” 22. The distinction between broad and narrow senses of the progressive originates from Anthony Galton, The Logic of Aspect (Oxford: Clarendon Press, 1984), 84.

Rational Explanation of Action

89

what we are really doing corresponds to what would be acceptable in any context, but this goes entirely against common usage and would require the support of a powerful argument. We regularly use the progressive quite permissively to characterize and explain what we are doing— e.g., “I can’t go on a walk because I’m making bread.” And we would regard someone who routinely ‘corrects’ such statements— e.g., “You’re not making bread, you’re standing here at the door talking to me!”—not as a stickler but as a jackass. The slipperiness in the progressive’s application-conditions extends to the question of when an action begins. Suppose Geoff stops to buy some aviator glasses. He is picking them out, standing in line for the cashier, etc. When does it become true that he is buying them—i.e., when does his future intention end and his intentional action begin? When he enters the store? When he heads to the cashier’s desk, glasses in hand? Or after he actually gets in line? Or when he’s handing over his credit card? Or only after the credit card is being swiped? Or only when MasterCard begins to process the request? Or when he starts signing the receipt? One can imagine various conversational settings that would draw us toward one or another of these answers. It is not clear why we must identify a clean break between before and after the buying has really commenced.23 If asked why he is buying the glasses, Geoff might say “because I’m flying to London.” Only a jackass would think it clever to ‘point out’ that the explanans is not literally true. But a change in conversational context might lead to a principled rejection of the progressive. If Geoff’s mother mistakenly thinks his fl ight is today, she might be surprised to receive his call. “Aren’t you flying to London?” “No, that’s next Tuesday.” It is difficult, if not impossible, to imagine a setting in which it would be natural for Geoff to say (today), “I am climbing into the cockpit.” But it is not clear what lesson we should draw from this fact. Even if it turned out—as seems unlikely—that there is some rigid codification of English usage for the progressive and that every appearance of exception could be explained away, what reason is there for thinking that this would be more than an idiosyncratic feature of English? Why should we treat this hypothetical English code as metaphysically deep? My contention is that the interesting 23. Compare Anscombe, Intention, section 23.

90

Rational Causation

philosophical distinction is just between what an agent does and does not represent as to be done. (Another important distinction is between what an agent represents as to be done and what an agent actually succeeds in doing, i.e., what actions she completes. But that is not relevant here, since my claims regarding what agents can just say is limited to actions-in-progress.) We use the category of ‘intention’ to make various distinctions within the class of things represented by an agent as to be done. Two are of interest here. We have just been discussing the first: we say a man merely intends to f and is not yet f -ing when he is at a relatively early stage in the process of fulfilling the intention. As we have seen, conversational developments can ‘move the line,’ a phenomenon that we accommodate by shifting between the progressive and the language of intention. Second, we say a man merely intends to f and is not f -ing when there is a gross mismatch between the man’s plan and the way the world is. Geoff may intend to walk to the Regency Theater; but it closed years ago. Although we might say “Geoff turned south on Broadway at Seventy-First Street because he intended to walk to the Regency,” we cannot say “Geoff turned south on Broadway at Seventy-First Street because he was walking to the Regency.” Both describe Geoff as making the same practical inference, but the former presupposes a greater degree of practical success. Like the fi rst sort of example, this shows that someone might represent f -ing as to be done yet not be f -ing. But it does not follow that f -ing is something beyond representing f -ing as to be done. It is consistent with a different picture: actions are exemplars of representings-to-be-done. Mere intentions are such representings in their earliest (according to flexible standard) phases, or highly defective representings. The best evidence for this controversial view is that it squares well with the correct account of teleological explanation more generally. So far I have focused on the first-personal aspect of acting-for-a-reason, especially on its classification as a special sort of thought. Rational actionexplanation, in turn, is the attribution of a thought. But whatever else it may be, it is also a form of teleological explanation. In the next section, I will show that my view of the role of psychological categories in rational explanation fits nicely with a plausible theory of teleology.

Rational Explanation of Action

91

It might be argued, however, that even if it were true that there is no deep difference between f -ing and intending to f, it is still not clear how someone can speak authoritatively about what he is doing, yet not on the basis of evidence, given that evidence would seem to be required to know nothing serious is amiss. Mustn’t one know that there is no severe mismatch between one’s plan and the world in order to say authoritatively that one is f -ing? No, for just saying what one is doing is not a matter of being justified in believing the proposition that one is f -ing. It is a matter of an evaluation—more precisely of a certain representational attitude toward f -ing. If it turns out that one isn’t f -ing, then nothing one says is the verbalized form of the action. But since action-expressions are not assertions, their authoritativeness is not based on theoretical knowledge that the conditions for executing the plan to f are favorable. In order to be in a position to just say that one is f -ing, one must know whatever is required in order to perform the relevant action intentionally. This constraint should not, however, be overstated. Consider Davidson’s famous example of a man who is intentionally making ten carbon copies as he signs a sheet of paper. Plausibly, he must know roughly the way carbon paper works in order to be signing all ten intentionally. But he does not need to know that he will succeed. Even though he is not certain of success, he is nonetheless able to just say that making ten copies is what he is doing. For authoritatively saying what one is doing is not issuing a guarantee of eventual success. He may not even believe that the pressure he is currently applying is making it all the way down to the last copy. Still, his  expectation of failure need not— has not, in Davidson’s example— undermined his representing signing all of them as to be done. And this is his intention, an intention that is, contrary to his belief, coming to fruition in action. So the signer can say “I am making ten copies” with the authority distinctive of saying what one is doing. (Thus I would argue that Davidson’s example does not, after all, refute the central thesis of Intention.) In this section I have replied to three objections to my thesis that actingfor-a-reason is a distinctive sort of thought. But as much as anything else, this proposal is likely to be met with incredulity on purely metaphysical grounds. Bicycling to work may causally depend on thought; perhaps it is

92

Rational Causation

even partly constituted by thought. But identical to thought? Thinking one is doing something is one thing, actually doing it is another. Acting, in typical cases, is a matter of a material object moving through space. How could that be a mental affair? No view of the sort that I’m advocating will be acceptable until this question is satisfactorily answered. In Chapter 6 I will do so. Bicycling to work is both a matter of a corporeal being traveling from one place to another and nothing over and above that being’s representing bicycling to work as to be done. The thesis that action is a kind of thought does not amount to the interiorizing of action but rather something more like the exteriorizing of thought.

3. Instrumental Teleological Explanation Teleological explanation is not, as such, explanation by appeal to psychological states such as belief and desire. Following Michael Thompson’s groundbreaking work, Life and Action, I will argue here that it is explanation of a part of an event by appeal to the whole of which it is a part, where the former is for the sake of the latter. 24 Whereas rational teleological explanations pertain to the thoughts of sapient beings, and are in that sense psychological, organic teleological explanations pertain to the biological processes of living beings and do not depend on conceiving of, for example, hearts as believing or wanting anything. By seeing what these two species of teleological explanations have in common, we arrive from a different vantage point on the claim made at the very end of the last section: a central role of psychological concepts in rational explanation is to salvage explanations in those cases in which the wouldbe event for the sake of which an action is being performed is not, in fact, occurring. Before I get into the substance of the section, it will be useful very briefly to situate my discussion in relation to recent philosophical work on teleology. Much of this work revolves around whether and how to fit teleological explanations into a mechanical view of the natural world. More specifically, philosophers have attempted to understand biological func24. See Michael Thompson, Life and Action (Cambridge, MA: Harvard University Press, 2008), especially part 2.

Rational Explanation of Action

93

tions in purely mechanistic terms. There has been much debate in the philosophy of biology about which such account of functions is best,25 and some have attempted to extend these accounts to provide a biologically grounded or inspired view of mental content.26 Nothing I say here is directly relevant to either of these projects. Insofar as it is a goal of this book to articulate a notion of rational, teleological, non- efficient causation, and insofar as I take teleological causation, so understood, to be a natural phenomenon in the actual world, I do not accept the idea that every teleological explanation must be ‘naturalized’ or rejected. But in this chapter, I have tried to abstract entirely from the question of whether the naturalization of teleology quite generally is necessary or possible. Rather, my focus will be just on the structure of teleological explanation. In particular, I discuss the form of the teleological explanation of individual events (such as actions), rather than, as is typical, specific traits of species or members of species. In Chapter 4 I use the unorthodox characterization of these explanations to articulate a conception of teleological causation. When events are explained teleologically, the explanans is traditionally thought to designate a future state of affairs. This conception of teleology is common to both those who doubt that there is any fundamental teleology in the world27 and those who are realists about unreduced teleological explanation.28 But it rests on a mistake. Teleological explanations of specific unfolding events, as we shall see, typically appeal not to future states

25. See Robert Cummins, “Functional Analysis,” Journal of Philosophy 72 (1975): 741–760; Larry Wright, Teleological Explanations (Berkeley: University of California Press, 1976); Karen Neander, “The Teleological Notion of ‘Function,’ ” Australasian Journal of Philosophy 69 (1991): 454–468; and Paul Griffiths, “Functional Analysis and Proper Functions,” British Journal for the Philosophy of Science 44 (1993): 409–422. 26. See, e.g., Ruth Millikan, Language, Thought and Other Biological Categories (Cambridge, MA: MIT Press, 1984), and Fred Dretske, Explaining Behavior (Cambridge, MA: Bradford Press, 1988). 27. For a recent example, see John Hawthorne and Daniel Nolan, “What Would Teleological Causation Be?,” in Hawthorne, Metaphysical Essays (Oxford: Oxford University Press, 2006), 265–284. 28. For a recent example, see Scott Sehon, Teleological Realism (Cambridge, MA: MIT Press, 2005).

94

Rational Causation

of affairs but rather to concurrently occurring larger event wholes, of which the explained events are parts. Here is a very basic schema of teleological explanation, one by which we explain the occurrence of some particular past event in terms of an end: o was f -ing in order to y. Here are two instances: Jones’s right ventricle was pumping some blood through his pulmonary artery in order to pump it to his lungs. and Dara was tasting all the desserts in order to review the restaurant. Some hold that statements like the former, in which the teleological form is applied to non-rational organic substances, cannot literally be true. Nevertheless, no one would doubt that the above example is superior to Jones’s right ventricle was pumping some blood through his pulmonary artery in order to pump it to his nose. The original example, by comparison, is not just intelligible but positively successful as an explanation. The grammatical form of the explanans in these examples is an unconjugated verb phrase—here, ‘to pump some blood to the lungs,’ and ‘to review the restaurant.’ Whereas ‘in order to’ joins a sentence fragment to an unconjugated verb phrase, ‘because’ is a sentential connective. Thus, when we switch to a style of explanation that introduces its explanans with ‘because’ rather than ‘in order to,’ the explanans-expression takes the form of a sentence, yielding ‘it was pumping blood to the lungs’ and ‘she was reviewing the restaurant.’ The name for the object is thus joined to the verb phrase in a determinate manner, thereby settling questions of the verb’s mood, tense, and aspect.

Rational Explanation of Action

95

In central cases of such ‘in order to’ explanations, the object’s ends are in the offing: when o was f -ing, o was also y-ing. We can then becausally explain the relevant f -ing by appeal to another event. When the ends are not in the offing, we resort to other measures. Let’s look first at the former sort of case: o was f -ing because it was y-ing. Using our examples: Jones’s right ventricle was pumping some blood through his pulmonary artery because it was pumping the blood to his lungs. Dara was tasting all the desserts because she was reviewing the restaurant. Note that the verb-aspect in the explanantia-expressions is imperfect, as Thompson has emphasized. Dara’s dessert-tasting is explained by the fact that she was reviewing the restaurant, not by the fact that she reviewed the restaurant. In “Dara reviewed the restaurant,” the verb phrase has perfect verb-aspect and, as such, requires that Dara actually finished reviewing the restaurant. “Dara was reviewing the restaurant” requires no such thing, as this sentence may be true even if she died in mid-review. One might think that “o was f -ing because o was y-ing” is just slightly more handy than “o was f -ing because o y-ed,” insofar as the latter applies only in cases where the y-ing was completed. But this misses the real importance of the imperfective in such explanations. The problem with perfect verb-aspect in the explanans portion of teleological explanations is not that it would require the explanantia to designate completed events and as such would apply to fewer cases. Rather, the problem is the very fact that the explaining events are represented as completed—in Dara’s case, a completed act of restaurant-reviewing. It is not because she finished reviewing the restaurant that she was tasting all the desserts. To say “Dara was tasting all the desserts because she reviewed the restaurant” would be to say, perhaps, that tasting all the desserts was her reward for having reviewed the restaurant. This way of trying to put the explanans fatally obscures the fact that tasting all the desserts was a part of

96

Rational Causation

her reviewing the restaurant and, as such, must have been occurring while the restaurant-reviewing was in progress or incomplete. But there is nothing special about Dara’s case. The same holds for Jones’s right ventricle. It is not because Jones’s right ventricle had finished pumping the blood to the lungs that it was pumping the blood through the pulmonary artery—rather, it is because it was in the midst of pumping the blood to the lungs. Pumping the blood through the pulmonary artery was a part of pumping it to the lungs. It is crucial to a proper understanding of the teleological explanation of one event by another that the explanansevent is represented using imperfect rather than perfect verb-aspect.29 The sorts of rational and organic explanations that I have been considering are teleological. We may use ‘in order to’ versions of these explanations in each sort of case. In each sort of case, the ‘because’ in the corresponding becausal explanation introduces the teleological explanans propositionally, rather than with an infinitival verb-phrase. But organic and rational explanations are formally different: the expressions ‘in order to’ and ‘because’ have different underlying structures as they occur in these different domains. This becomes apparent upon considering cases in which the relevant end-events are not in the offing. These are not the cases in which o simply never fi nished y -ing, i.e., cases in which ‘o y -ed’ is false. Rather, they are cases in which o never even started y-ing, i.e., cases in which ‘o was y-ing’ is false. Perhaps Dara was simply delusional and had never been a restaurant critic; perhaps Jones’s lung had suddenly disintegrated, prior to that particular quantity of blood being pumped anywhere. Dara was thus not in fact reviewing the restaurant, and Jones’s right ventricle was not in fact pumping the blood to the lungs. And so it was not because either of these

29. It might be objected here that a teleological explanation does not require an explanans with imperfective verb-aspect unless the explanandum has imperfective verb-aspect. There is no problem with explanations in which both verbs have perfect aspect. So “Dara tasted all the desserts because she reviewed the restaurant” is OK. This sounds off to my ears. But be that as it may, this teleological explanation depends on it also being true that Dara tasted all the desserts because she was reviewing the restaurant. There is no dependence in the other direction. This is a metaphysical necessity: the means-event occurs because the end-event was in the midst of occurring, and for the sake of helping to bring it to completion. The imperfective explanans is thus fundamental. I return to this point in Chapter 4, section 2.2.

Rational Explanation of Action

97

events were occurring that some other events were occurring. Nonetheless the following becausal explanations are available: Jones’s right ventricle was pumping some blood through his pulmonary artery because its function is to pump blood to the lungs. and Dara was tasting all the desserts because she intended to review the restaurant. To say that Jones’s right ventricle’s function is to pump blood to the lungs is to say that it is an instance of a kind to which pumping blood to the lungs belongs as a standard. Pumping blood to the lungs is thus something that right ventricles might fail in regard to, in a way that left ventricles cannot. The truth of the explanantia in our sample organic explanation thus depends upon the fact that Jones’s right ventricle is an instance of a kind that has a certain feature. The relationship between kinds of this sort and their features can be represented as habituals, e.g., “The right ventricle pumps blood to the lungs.” As was discussed in Chapter 1, sentences of this form are general and atemporal. They are general, in the sense that they refer not to particular Os, but to the kind, O. And they are atemporal in the sense that although the verb is conjugated in the present tense, it doesn’t refer to any ongoing act of f -ing. So, to say “the O fs” is not to say of any particular o of kind O that it is currently f -ing. It is rather to say that it is normal (in the normative, not statistical, sense) for an O to f. Thus, Jones’s right ventricle was pumping some blood through his pulmonary artery because its function is to pump blood to the lungs entails Jones’s right ventricle was pumping some blood through the pulmonary artery because the right ventricle pumps blood to the lungs.

98

Rational Causation

In cases in which a particular O f -ed because the O ys, it is also true that the O f s because the O ys. And so both examples entail: The right ventricle pumps blood through the pulmonary artery because it pumps blood to the lungs. The points of similarity and difference between organic and rational explanations are highly instructive. Where a teleological explanation is organic, the truth of o was f -ing in order to y logically depends on the truth of The O f s because the O ys [where o is an O]. But a rational teleological explanation does not depend on any substitution-instance of this schema. Consider again “Dara was tasting all the desserts in order to review the restaurant.” If this statement depends on some substitution instance of the above schema, we need a substitution-instance for O. The most plausible candidates are ‘restaurant critic’ and ‘person.’ But they fail miserably. Dara is not, we can suppose, a restaurant critic at all (she is delusional), so her tasting all the desserts in order to review the restaurant cannot have anything to do with her belonging to the supposed kind, ‘restaurant critic.’ And although Dara is a person, it is neither true that ‘the person’ tastes all the desserts nor that ‘the person’ reviews the restaurant; and so it cannot possibly be true that ‘the person’ tastes all the desserts because ‘the person’ reviews the restaurant. Dara’s intention to review the restaurant is not a logical consequence of her being the sort of thing she is, although it is a manifestation of her being the sort of thing she is. Dara’s intention to review the restaurant is a manifestation of her being an agent, a fact that reveals itself in the battery of other judgments that we must be prepared to make about her when we say that Dara was tasting all the desserts because she intended to review the restaurant. This statement has been taken to entail the following:

Rational Explanation of Action

99

Dara knew that she was tasting all the desserts in order to review the restaurant; Dara believed that tasting all the desserts was a means to reviewing the restaurant; Dara ought to have tasted all the desserts in light of the (putative) fact that she was reviewing the restaurant; and Dara intended . Perhaps not all of these entailments go through. But they at least provide a reasonable starting point for a discussion of the logical significance of the category of intention. It is, on the contrary, not remotely plausible that the success of Jones’s right ventricle was pumping some blood through his pulmonary artery because its function is to pump blood to the lungs depends on the ventricle’s knowing, believing, or intending anything. The category of ‘function’ operates very differently. These data provide good reason for taking teleological explanations to have different underlying structures as they are deployed in rational and organic cases. This is surely a more satisfying result than to say that organic teleological explanations are never literally true. In other words, what it is for a ventricle to f in order to y is different than what it is for a person to f in order to y. And what it is for a ventricle to f because it is y -ing is different than what it is for a person to f because he is y -ing. The rational and organic forms are differentiated in virtue of the former entailing o was f -ing because she intended to y and the latter entailing o was f -ing because its function is to y.

100

Rational Causation

Here we see the analogous role played by an individual’s intention in rational explanation and an object’s function in organic explanation. One central point of invoking these concepts is to preserve ‘becausal,’ as opposed to ‘in order to,’ teleological explanations in those cases in which the relevant y-ing is not in the offing. This does not reflect a fetish for sentential connectives. Our interest in preserving explanations of this form is due to our interest in knowing on what facts the occurrence of the explained event depends. The ‘in order to’ forms do not quite provide this information. Pulling together strands from this and previous sections: to say someone acts for a reason is to ascribe to them a thought. Specifically, the agent represents the to-be-done-ness of the end-action as conferring to-be-doneness on the means-action. Thinking this thought is performing an action for a reason and is also, as such, being in a position to just give the explanation. Unlike in the case of organic teleological explanation, the ‘because’ of rational action-explanation ultimately relates back to a subject’s making an inference: this is the thing to do because that is the thing to do. Rationally explaining someone else’s action, whether or not we explicitly invoke psychological categories, is attributing to them a thought that features such a ‘because.’ But explicitly invoking such categories enables us to preserve what is in some sense the same explanation in cases of radical failure. In the next section, I turn to the anti-psychologistic upshot of this idea.

4. Anti-Psychologism about the Rational Explanation of Action This is Psychologism about acting-for-a-reason: P3: For S to f for a reason is for S’s f -ing to be caused (in the right way) by S’s beliefs and desires. P4: The causation that underlies acting-for-a-reason is efficient. Acting-for-a-reason, on this view, is accounted for in terms of the efficacy of the agent’s psychological states. World-citing guises of rational actionexplanation must thus be understood in terms of mind-citing guises, since the latter cut closer to the metaphysical bone. If the analysis I have pre-

Rational Explanation of Action

101

sented is correct, Psychologism gets things backward. Rational actionexplanations that explicitly cite the mind must be understood in terms of a practical ability to be responsive to worldly facts. As in Chapter 1, worldciting explanations, insofar as they presuppose a greater degree of success, are more central to understanding the underlying ability. The psychologistic approach derives its appeal in part from its assimilation of rational efficacy—i.e., the efficacy underlying acting for a reason— to the efficacy of mechanisms, which is, if not exactly well understood, at least beyond suspicion. To sever the tie between reasons and the efficient causal order would, in the eyes of many, undermine our warrant for treating reason-citing explanations as genuinely illuminating. This impression is due in part to the perceived absence of a viable alternative—how else to ground the power of reasons in the material world? Part of the point of the book as a whole is to offer an alternative. In what remains of this chapter, I put aside this issue and focus on an intuitively dissatisfying implication of the psychologistic analysis of rational explanations, viz., that the very thing that justifies the action cannot explain it. Broadly speaking, this, too, is a worry about whether reasons are consequential, now turned against Psychologism. The question is whether Psychologism can account for the power of reasons themselves, as opposed merely to attitudes toward reasons. I argue that my version of Anti-Psychologism vindicates the intuitive idea that what explains an action can be precisely what justifies it. Unlike Dancy’s anti-psychologistic approach, it does so while also providing the resources to explain why, when one is f -ing because p and in order to y, one must believe that p and both intend and desire to y.

4.1 The Equivalence Thesis Davidson emphasizes that the notion of a reason, as it figures in phrases like ‘reason for acting’ has both a justificatory and an explanatory dimension. When considering a prospective course of action, a reason for performing the action is something that— and here there are a variety of ordinary expressions that convey the gist—favors it, recommends it, casts it in a favorable light, supports it, justifies it, makes it a sensible or good thing to do, etc. Someone might say “You should take your umbrella to work today.” I reply “Why?” “Because it is supposed to rain.” “Thanks for

102

Rational Causation

the tip!” Later, if asked why I brought my umbrella to work, I might state the same fact: “It was supposed to rain.” The very facts that one views, initially, as supporting a course of action can also be that in light of which one ultimately acts. Citing such facts then explains why I, for example, took the umbrella to work. Of course, reasons often fail to justify; but still, an agent who acts for a reason must (one would naturally think) have at least taken the putative fact to cast the act in a favorable light—to justify it, in Davidson’s ‘anemic’ sense.30 This is the Equivalence Thesis.31 An agent’s reasons, according to this thesis, rationally explain an action insofar as they are the source of the favorable light in virtue of which the action seemed to the agent as the thing to do. There are, of course, uses of the word ‘reason’ that are disconnected from justification. To speak, for example, of the reason that the bridge collapsed is not to speak of a reason the bridge had for collapsing. The ‘reasons’ at issue in the Equivalence Thesis are distinctive to rational explanation. Also, the thesis does not prevent us from acknowledging that there are two different senses in which an agent may ‘have’ a reason. In one sense, to have a reason is a purely normative, explanatorily inert matter. To take Michael Smith’s example, a man may have, in the normative sense, a reason to buy a painting—it’s a Picasso—but because he fails to see the good in buying it—he doesn’t know that it’s a Picasso—he does not have, in the motivating sense, a reason to buy the painting.32 My concern here is just with having reasons in the motivating sense. The point of 30. An action is justified in this sense if “from the agent’s point of view there was, when he acted, something to be said for the action.” See Donald Davidson, “Actions, Reasons, and Causes,” in Essays on Actions and Events (Oxford: Clarendon Press, 1980), 9. 31. The Equivalence Thesis is closely related to Dancy’s ‘normative constraint’ (“that in the light of which one acts must be the sort of thing that is capable of being among the reasons in favor of so acting”; see Jonathan Dancy, Practical Reality [New York: Oxford University Press, 2000), 103]) and his ‘explanatory constraint’ (“any normative reason is capable of contributing to the explanation of an action that is done for that reason”; ibid., 101). See also Joseph Raz, “Reasons: Explanatory and Normative,” in New Essays on the Explanation of Action, ed. C. Sandis (London: Palgrave Macmillan, 2009), 184–202, 195: “We can baptize this thought ‘normative/ explanatory nexus,’ namely that regarding every normative reason, it is possible for it to feature in an explanation of the action for which it is a reason as a fact whose recognition motivated the agent to perform it, and guided him in its per for mance.” 32. Michael Smith, The Moral Problem (Malden, MA: Blackwell Publishing, 1994), section 4.2.

Rational Explanation of Action

103

the Equivalence Thesis is that there is a normative dimension to having (in the motivating sense) a reason: reasons motivate insofar as they make potential actions appear sensible or justified. Finally, the Equivalence Thesis does not say that an agent who acts for a reason does so with a mind to ‘doing good,’ on any specific substantive conception of this expression. As David Velleman puts it, I cannot act for reasons if I don’t care about doing what’s justified or (as I would prefer to put it) what makes sense. But I can still care about doing what makes sense even if I don’t care about the good. This possibility is demonstrated by my capacity to be guided by what makes sense in light of a counter-evaluative mood such as despair, since what makes sense in light of such a mood just is to do what’s bad rather than what’s good.33 Can Psychologism account for the way explanation and justification are tied together in the rational explanation of action? Psychologistic philosophers have traditionally held that reasons just are psychological states and that psychological states do indeed play both causal- explanatory and justificatory roles. Velleman describes the Standard Model this way: We want something to happen, and we believe that some behavior of ours would constitute or produce or at least promote its happening. These two attitudes jointly cause the relevant behavior, and in doing so they manifest the causal powers that are partly constitutive of their being, respectively, a desire and a belief. Because these attitudes also justify the behavior they cause, that behavior eventuates not only from causes but for reasons.34 Dancy has argued, however, that Psychologism is impossible to square with the Equivalence Thesis. The intuitive basis for his contention can be gleaned from a simple example. Suppose my drain is clogged. What should I do? As I try to answer this question, my concern is not with my own 33. Velleman, Practical Reason, 121–122. 34. Ibid., 5.

104

Rational Causation

psychological states but with the sink, what’s clogging it, whether Drano is likely to work, whether plunging might be required, etc. Whatever I ultimately decide, my reasons will be facts about the world, not about myself. Once I have acted, these facts are both what made doing what I did seem like the thing to do and also what explain my having done it. The drain-related facts explain my action just because I took them to justify it. It is not my beliefs that cast a favorable light on prospective actions but rather worldly facts. Dancy’s point can be thrown into greater relief by considering again the unusual sort of case in which a reason does concern the agent’s own psychological states. He cites the following example: “Someone who believes that there are pink rats living in his shoes may take that he believes this as a reason to go to the doctor or perhaps a psychoanalyst. This is quite different from the person who takes his belief that there are pink rats living in his shoes as a reason to call in the pest control officer.”35 Dancy characterizes the difference between the two in this way: the first person takes the fact of his believing as a reason (to go to the doctor), whereas the second takes what he believes as a reason (to call the exterminator). The former is the highly atypical case in which a world-citing explanation cites a psychological fact. Although Dancy’s focus is exclusively on reasons specified using ‘that’ clauses, the same point can be made about reasons we specify using infinitival verb phrases as well—those that figure in the specification of intentions and desires, among other psychological attitudes.36 For example, “I mixed the mortar to bind the tile to the floor.” What recommended mixing the mortar to me and what ultimately explains my mixing the mortar are not features of my mind but rather of floor-tiling. However, several psychologistic philosophers have lately argued that their basic thesis is not ultimately incompatible with Dancy’s point. The most important thing, as far as Psychologism is concerned, is not what reasons themselves are but rather that to act for a reason is to be caused to

35. Dancy, Practical Reality, 125. See also John Hyman, “How Knowledge Works,” Philosophical Quarterly 49 (1999): 433–451, 444. 36. See also Philip Pettit and Michael Smith, “Backgrounding Desire,” Philosophical Review 99 (1990): 565–592.

Rational Explanation of Action

105

act by beliefs and desires. To count as a view of reasons-explanation, a psychologistic theory must just say how the distinctive features of acting for reasons (as opposed, say, to being pushed) can be explained in terms of the theory’s preferred causal schema. Standard Psychologism meets this obligation in part by identifying that in light of which the agent acts with the very psychological states that cause the action.37 But that is not the only psychologistic option. A Sophisticated Psychologist instead argues that whereas reasons themselves, i.e., considerations salient in deliberation, are typically worldly, it is nonetheless psychological states—in particular, attitudes toward those considerations—that bring about the action. My reason for pouring Drano into my sink is that doing so will unclog it. But my action is caused by my belief that Drano will unclog my sink. Evidently, the Equivalence Thesis is the cost of Sophistication. The Psychologist can maintain a commitment both to an intuitive conception of action-justifiers and to the fundamental causal importance of the psychological, but only at the expense of denying that the same things can play both roles. Of course Sophisticated Psychologists can simply deny that the rejection of the Equivalence Thesis is a cost. Wayne Davis defends (as non-revisionary) the view that “actions can be explained by stating the reasons for which they are performed” even though “the reasons themselves are [not] what explain the actions.”38 On his view, what I believe, intend, and want can stand in justificatory relations to my action. They are my reasons. But my psychological states (or that fact that I am in them) are really what explain my performing the action. On this view, we are not exactly responsive to reasons themselves, which are intensional and so inefficacious items. Instead we are moved to act by attitudes toward reasons. Kieran Setiya’s view is more sophisticated still. He agrees with Dancy that reasons are (typically) worldly (as opposed to psychological) and, unlike Davis, also accepts that reasons really do explain actions. But for him 37. Dancy rightly complains that Psychologists often fail properly to distinguish the view that psychological states are causes of actions from the view that psychological facts—facts of my being in certain psychological states— are causes of actions. Nothing I say here against Psychologism will turn on the difference between the two views. 38. Wayne Davis, “Reasons and Psychological Causes,” Philosophical Studies 122 (2005): 51–101, 57.

106

Rational Causation

the important point is that the truth of a rational explanation of action consists in the action’s being made to happen via the efficacy of the psychological. That Drano will unclog my drain rationally explains why I poured it into my sink. But it is not the putative worldly fact that led to my performing the action but rather a complex psychological state involving the belief itself.39 Reasons themselves do explain action; but they do not, on his view, cause action. This strategy removes the wedge between reason and explainer of action and places it instead between rational-explainer of action (worldly) and causal-explainer of action (psychological). Insofar as the latter is identified as what ultimately makes the action happen, this strategy violates the spirit if not the letter of the Equivalence Thesis.40 What ultimately explains the action, on this view, is never a reason. Standard Psychologism is dubious because it misidentifies that in the light of which I act. Sophisticated Psychologism is implausible (by my reckoning, at least) because it rejects, or at best hedges on, the Equivalence Thesis. That in light of which I act is not, according to Sophisticated Psychologism, what ultimately explains my action. My view, recall, is that the difference between mind- and world-citing guises of rational explanation is not, as the Sophisticated Psychologist would have it, that the former makes explicit the real or ultimate action-explainer, which was only implicit in the latter. Rather, the former invokes the ability underlying acting-for-a-reason in such a way as to make it possible to rationally explain the action even when the agent’s reasons are unsound. Like Dancy, I hold that world-citing explanations are of primary importance in understanding the interconnected justificatory and explanatory dimensions of rationality. But unlike Dancy, I reject the idea that what really explains an action is worldly as opposed to mental. The very idea of such an opposition reflects a kind of category mistake, thereby disguising the nature of the rational-explanatory work done by appeal to the psychological. Further39. Setiya, “Explaining Action.” In discussing the views of others, I often follow them in speaking of psychological states as if they were particulars, as in “the belief itself.” I argue in Chapter 5, however, that the very idea of a par tic u lar psychological state is confused. 40. In fact, Setiya rejects outright the intuition behind the Equivalence Thesis. He argues that ‘agent’s reasons’—the considerations salient in deliberation— are what leads the agent to act but need not even seem to cast the action in a favorable light.

Rational Explanation of Action

107

more, it also makes it difficult for Dancy to explain why it is that when we act for reasons, we must have the relevant psychological states, a difficulty upon which the Sophisticated Psychologists have rightly seized. My version of Anti-Psychologism is consistent with the Equivalence Thesis but does a better job accounting for the role of the psychological in reasoning.

4.2 Explaining the Rational Role of the Psychological Non-Psychologistically Rational action-explanations appeal to the exercise of an ability to do what is to be done because something else is to be done. To say that someone is f -ing because she is y-ing is to say that she represents the to-be-done-ness of f -ing as a consequence of the to-be-done-ness of y-ing. When this ability is exercised successfully, the agent’s representing the relevant actions as to be done is also her performing them. And in such cases, both “S is f -ing because S is y-ing” and “S is f -ing because S intends to y ” will provide good explanations. S’s f -ing will, in this sort of case, also manifest instrumental knowledge, and so something in the vicinity of “S is f -ing because f -ing is a means to y-ing,” in addition to “S is f -ing because S believed that f -ing is a means to y-ing” will also be correct. But the rational ability to which such explanations appeal is not always exercised successfully. S may not be y-ing. And y-ing may not be the thing to do. Also, she may lack the requisite instrumental knowledge, either because f -ing is not a means to y-ing, or because she is ignorant of this fact, or because the to-be-done-ness of y-ing fails to confer to-be-done-ness on f -ing. We use psychological vocabulary to signal any of these sorts of failures. We appeal to intentions, beliefs, desires, etc. One point of shifting to talk of the mind from talk of the world is to frame the explanans in such a manner that it is not threatened by the possibility of failure. It is because the world-citing explanations presuppose a greater degree of success than the corresponding mind-citing explanations that instances of the former are more explanatorily self-sufficient than instances of the latter. Citing Dara’s reviewing the restaurant perfectly well explains her trying all the desserts. An explanation citing her belief that she is reviewing the restaurant or her intention to review the restaurant suggests that her belief is mistaken, that she is not doing what she intended. Intuitively,

108

Rational Causation

an explanation in terms of mere belief or intention (or desire, to which we turn shortly) does not explain as satisfactorily why Dara tasted all the desserts. For it raises the question of the source of her error. Just as more explanation is required to understand why a heliotropic flower tracks the sun unsuccessfully than to understand why it does so successfully, mind-citing explanations typically require supplementation in a respect that world-citing explanation do not. To make up the difference, we account for her mistake. This was a central point of Chapter 1. An explanation that works by appeal to the exercise of an ability is less explanatory to the extent that the postulated exercise is less successful. This approach would seem to provide the materials to answer the following question of Dancy’s: “How should we explain the fact that, where the agent’s reason for acting is that p, the agent must believe that p, if not that the agent’s reason for acting is ‘really’ that he believes that p?” 41 The analogous question arises for intentions and desires. If an agent is f -ing because she is y-ing, the agent must intend and desire to y. Why must this be so if not because her intention and desire to y are her real reasons? The foregoing sketch suggests a quick answer. An explainer may always refrain from committing herself to the agent’s success, to her knowing the relevant p or her in fact y-ing. Thus, wherever it is correct to explain S’s f -ing by saying “because p” or “because she is y-ing,” it is also correct (even if misleading) to explain by saying “because she believes that p” or “because she intends to y ” or “because she desires to y.” And since ‘because’ is factive, it follows that she is in the relevant psychological state. But Dancy doubts that his question can be answered along these lines. After considering an answer inspired by Arthur Collins’s similar view of the point of deploying psychological terms,42 Dancy says: “The question is not one about when we do and when we don’t adopt the psychologized form of explanation. It is about the role of the belief in the story as a whole if it is not playing the focal role normally attributed to it.” 43 The “focal

41. Dancy, Practical Reality, 126. 42. See Arthur Collins, The Nature of Mental Things (Notre Dame, IN: University of Notre Dame Press, 1987), and also his “The Psychological Reality of Reasons,” Ratio 10 (1997): 108–123. 43. Dancy, Practical Reality, 126.

Rational Explanation of Action

109

role normally attributed to it” is, presumably, its role on the Standard picture: beliefs are reasons for action. Given that Dancy thinks that reasons are not (typically) psychological, he must say what role beliefs play. He cannot allow that beliefs are causes of action, because then actionexplanations in terms of psychological causes would seem to be rivals to (and potential swallowers-up of) rational action-explanations.44 And yet we must, it seems, assign beliefs some role. After all, f -ing because p depends upon believing that p. Dancy ultimately settles on the view that believing that p is an enabling condition of someone’s f -ing for the reason that p.45 If I were not alive, I could not be baking bread, and so could not be doing anything (e.g., turning on the oven) because I’m baking bread. Being alive is an enabling condition of my acting in light of the fact that I’m baking bread. But it is not my reason for turning on the oven. Similarly, believing that p is necessary for me to be f -ing because p, yet it is not my reason for f -ing. But the ‘enabling condition’ view sheds little if any light on Dancy’s original question, a fact that has not been lost on his Sophisticated critics. Davis offers his own answer to the question before assessing Dancy’s: My answer is simple: acting because one believes that p is part of what it is to act for the reason that p. Dancy’s answer is that the belief is an “enabling condition.” . . . Believing that p enables one to act for the reason that p. This is true, but it seems to be just a restatement of the fact we wish to explain, namely, that we cannot act for the reason that p unless we believe that p.46

44. Dancy discusses the ‘rival explanation’ concern in his “Two Ways of Explaining Action,” in Agency and Action, ed. J. Hyman and H. Steward (Cambridge: Cambridge University Press, 2004), 25–42. 45. I do not consider here his second suggestion, according to which we should understand the reference to belief in “S is f -ing because she believes that p” appositionally, as equivalent to “S is f -ing because p, as he believes.” (See Dancy, Practical Reality, 127–130). In the end, he allows that this may simply be another way of putting the enabling-conditions account. 46. Davis, “Reasons and Psychological Causes,” 77. Setiya elaborates on this objection in his “Reasons and Causes,” European Journal of Philosophy 19 (2011): 129–157.

110

Rational Causation

There is a further problem. The fact that S’s believing that p is an enabling condition of S’s f -ing because p leaves it mysterious why we can also explain her f -ing by citing her believing that p—whereas we cannot explain her f -ing by citing other enabling conditions, e.g., that she is alive. Dancy might be leery of this way of putting it—his criticism of Standard Psychologism, after all, is that people rarely act in light of their own psychological states. But no one can dispute that in cases in which p is false, we do in fact routinely and successfully convey the reason why an agent performed an action by citing her believing that p, something we sometimes also do without incident even when the agent knows that p. No one ever thinks, upon hearing these mind-citing explanations, that the agents are being said to act in light of their own beliefs, as opposed to (perhaps merely putative) worldly facts. My point is that there is plainly a connection between the fact that “S is f -ing because p” entails—indeed gives the same explanation as—“S is f -ing because S believes that p” and the fact that “S is f -ing because p” entails “S believes that p.” One would think that a satisfactory explanation of the latter entailment would (at the very least) also shed light on the former. But the “enabling condition” account does not. The answer I gave above to Dancy’s question attempted to explain the latter entailment by explaining the former entailment. This answer was too quick. But by bringing together other elements of the story that I have been telling, we arrive at a satisfying answer that explains both. Let us begin by considering theoretical reasons. I identify believing-for-a-reason with representing p as . In successful cases, the proposition that q corresponds to the known fact that q, which is both that in the light of which the agent believes that p and what rationally explains her believing that p. Knowledge is thereby extended from q to p. It is q, rather than S’s believing that q, that is the focal point in such explanations: it is from the to-be-believedness of q that S infers p. Nonetheless, q is thereby represented as to be believed. And on the view defended in Chapter 1, to represent a proposition as to-be-believed is to believe it. Thus, S must believe that q in order to believe that p for the reason that q. Furthermore, appealing to q and appealing to S’s believing that q explain S’s believing that p in the same way, viz., by invoking S’s representing q as to be believed. Both the mind-

Rational Explanation of Action

111

citing and world-citing guises appeal to the same exercise of the underlying theoretical ability. Successful explanations of each sort reveal, respectively, the mental and extra-mental facts on which S’s believing that p rationally depends. The same strategy can be applied in the practical sphere. Acting-fora-reason is representing f -ing as to be done as a consequence of the to-bedone-ness of y-ing. In successful cases, she is doing what is to be done because it is a means to doing something else that is to be done. This is reasoning from the to-be-done-ness of y-ing. As such, she acts in light of her y-ing, rather than in light of any attitude toward y-ing. And so long as it is a fact that she is y-ing, it is precisely this fact that explains her f -ing. Still, y-ing is thereby represented as to-be-done, and so she must intend to y. For to say that someone intends to y is just to say that she represents y-ing as to be done while leaving it open whether such representing is a fully fledged y-ing, or y-ing in its earliest stages, or a completely botched instance of y-ing. Furthermore, appealing to her y-ing and appealing to her intending to y explain her f -ing in the same way, viz., by invoking her representing y-ing as to be done. The mind-citing and world-citing guises appeal to the same exercise of the underlying practical ability. The psychological and non-psychological portions of reality on which S’s f -ing rationally depend are thereby revealed. If S does not know that q, she cannot believe anything because q. And if S is not y-ing, she cannot do anything because she is y-ing. It might be objected here that in these less-than-ideal cases I too am forced to abandon the Equivalence Thesis, since the psychological explanantia to which we must then resort do not pick out facts in light of which S believes or acts. But this is, if anything, an advantage of my view. When that in light of which S believes or acts is not a fact at all, but rather merely a putative fact, we do not view her belief or action as dependent on it. This is precisely what is signaled by our rejection, in those cases, of the world- citing explanations. I speak very little about desire in this chapter. Here’s why: the role of desire in rationally explaining what someone actually does is subsumed by the role of an intention to do the same thing. Tenenbaum has argued that the relation between a desire to y (in the broad, philosophical sense of ‘desire’) and an intention to y is the difference between y-ing appearing

112

Rational Causation

to the agent as good and the agent’s judging that y-ing is good.47 In my terms, an agent who merely desires to y does not represent y-ing as to be done; but y-ing does, from a certain perspective, appear to her as to-bedone, and thus as a potential source of the to-be-done-ness of other actions. To represent y-ing as to be done is to accept the appearance of its to-be-done-ness. y-ing must appear as to be done to someone who is f -ing because she is y-ing, since y-ing just is the endorsement of such an appearance. Thus, to be anyone’s reason for acting, y-ing must be both intended and desired. Part of the reason desire has figured so prominently in psychologistic conceptions of rationality is that an action-explanation is, on this sort of view, true in virtue of efficient causal connections between the action and certain psychological states. Typically, those states are thought of as part of the causal history of the action. So to give someone’s reason is, ultimately, to explain what she is doing in causal-historical terms. And one might expect the psychological history of an action to include the agent’s antecedently wanting to perform it. But rational action-explanation does not purport to describe the causal history of an action any more than rational belief-explanation purports to describe the causal history of a belief. The issue is not, in the latter case, how someone came to believe something, but rather what, at the time of belief, makes it seem like the thing to believe. Similarly, the issue in the former case is what, while the agent is performing the action, makes it seem like the thing to be doing. Once antecedent psychological states are recognized as falling outside of the scope of such explanation, one sees that merely desiring to y, in the absence of intending to y, cannot rationally explain why someone is f -ing. (This is not, of course, to deny the philosophical importance of desire in a broader treatment of practical rationality.) Finally, let us turn to the question of why f -ing because it is a means to y-ing entails believing that f -ing is a means to y-ing, i.e., why performing an action in light of the instrumental fact that p requires that the agent believe that p. Suppose that every four months Esmeralda sprinkles what she thinks of as ‘magical lucky powder’ on her lawn in order to stimulate growth of the grass. Suppose, further, that the powder is in fact lawn fer47. See Tenenbaum, Appearances of the Good, especially chapter 1.

Rational Explanation of Action

113

tilizer. Sprinkling the powder is a means of stimulating lawn growth. And Esmeralda is sprinkling the powder in order to stimulate the growth of her lawn. But we cannot say, without qualification, that Esmeralda is sprinkling the powder on her lawn because she’s stimulating its growth, even if she is sprinkling the powder because she intends (and desires) to stimulate lawn growth. For to say that she’s doing it because she’s stimulating lawn growth entails or at least suggests that she knows that it does. This example brings out the appeal of the idea that to say that someone is f -ing because she is y-ing is to attribute to her instrumental knowledge of the connection between f -ing and y-ing. Intuitively, “Esmeralda is sprinkling the powder on her lawn because she’s stimulating its growth” (which appeals to another unfolding action) and “Esmeralda is sprinkling the powder on her lawn because doing so stimulates lawn growth” (which appeals to an instrumental fact) are two different versions of the same rational explanation. Furthermore, the same ignorance requires us to retreat to the success-neutral forms “Esmeralda is sprinkling the powder because she intends to stimulate lawn growth” and “Esmeralda is sprinkling the powder because she believes that doing so stimulates lawn growth.” This dependence of successful exercises of the ability underlying actingfor-a-reason on instrumental knowledge is easy to understand, given the role that instrumental facts play on the picture of practical reasoning that I have sketched here. The to-be-done-ness of f -ing is only a consequence of the to-be-done-ness of y-ing if f -ing is a means to y-ing. And so to represent the to-be-done-ness of f -ing as a consequence of the to-bedone-ness of y-ing is at the same time to represent it as to be believed that f -ing is a means to y-ing. In ideal cases, the agent knows that f -ing is a means to y-ing. Dancy’s question is “How should we explain the fact that, where the agent’s reason for acting is that p, the agent must believe that p, if not that the agent’s reason for acting is ‘really’ that he believes that p?” The answer to this question should also answer the question why, whenever we can give someone’s reason by citing the instrumental fact that p, we can also do so by citing the fact that she believes that p. My account of actingfor-a-reason answers both. In viewing y-ing as casting a favorable light on f -ing, the agent thereby thinks of the latter as a means to the former.

114

Rational Causation

She acts in light of this instrumental fact. Dancy is right to insist that agents rarely act in light of their own beliefs. But explanations that appeal to p are just more central than those that appeal to S’s believing that p, more central because they presuppose a more successful exercise of the underlying ability— one that harnesses knowledge that p. But part of representing the means as inheriting the to-be-done-ness of the end is representing as to be believed (and so believing) the proposition that f -ing is a means to y-ing. Furthermore, appealing to p and appealing to her believing that p are different guises of the same explanation, each postulating the very same exercise of the underlying practical ability. The argument from illusion involves the idea that we should base an understanding of (representational) experience on the phenomenologically indistinguishable item common to both veridical and non-veridical cases of experience. The appeal of Psychologism depends in part on the idea that we should base an understanding of what it is to act for a reason on  the phenomenologically indistinguishable items common to cases in which the agent’s reasons are sound and those in which they aren’t. Critics of the argument from illusion contend that veridical experience is of primary importance in understanding experience tout court.48 I believe we should adopt the same strategy for understanding believing- and actingfor-a-reason. And so my burden in this book is to articulate a conception of the place of rationality in the world based on the idea that the central cases of rational thought and action are those in which the agent’s reasons are good. But an analysis of the character of rational explanation is not sufficient for this purpose. Indeed, the Psychologist’s hold on contemporary philosophy has as much to do with the metaphysical difficulties faced by its opponents as the intuitiveness of its own preferred analysis. In order to transform this analysis of rational explanation into a full-blown metaphysical picture, I must answer the following questions: (1) What is the causal role of those facts and ends that are an agent’s reason for acting? I have argued that explanation by facts and ends is, in a sense, primary. But many will boggle at the idea that facts and ends could be 48. See, e.g., John McDowell, Mind and World (Cambridge, MA: Harvard University Press, 1994).

Rational Explanation of Action

115

direct causes of action, i.e., causes not via psychological intermediaries. Yet if facts and ends are causes only via psychological intermediaries, the line between my view and Psychologism starts to blur—facts and ends seem to be causes of action only derivatively and contingently. And if one instead denies that such facts are causes at all, it’s not clear how we can consistently view them as making a difference to what people think and do. (2) If being a difference-maker is sufficient for being a cause, then psychological states, or our being in them, would seem to be causes of action. But how should we understand the causal role of psychological states in the performance of action if not as inner items, efficiently causing observable behavior? I answer these questions in Chapter 4. In the meantime, I defend the analysis that I’ve presented so far from what might seem to be a devastating objection, one that threatens any view that makes self-consciousness an essential feature of rationality. The objection begins with the observation that we also use what seem to be the same forms of explanation to account for the behavior of non-human animals, despite their inability to just say why they are behaving as they are. In the next chapter, I reconcile this fact with my theory.

3 (Non-Human) Animals and Their Reasons

‘Naturalism’ is a disputed label in philosophical taxonomy. Few aspire to a view conceived of as ‘supernaturalist.’ But in the philosophy of mind, ‘naturalism’ has come (albeit not without resistance)1 to encompass both (a) the thesis that human beings are no less a product of the natural development of the world than ants, alligators, and apes and (b) the thesis that there are differences only of degree between the cognitive abilities of human beings and those of intelligent non-human animals (from now on, just ‘animals’). Through this tendentious labeling, philosophers who are happy to accept (a) but who balk at (b) are categorized as anti-naturalists, along with the fundamentalists, astrologists, and mystics. Underlying the rhetorical dispute, there is a serious philosophical conundrum. On the one hand, we recognize the perceptual, cognitive, and behavioral conti-

1. See especially John McDowell, Mind and World (Cambridge, MA: Harvard University Press, 1994), and Jennifer Hornsby, Simple Mindedness (Cambridge, MA: Harvard University Press, 1997).

(Non-Human) Animals and Their Reasons

117

nuities between humans and animals, continuities subserved by significant anatomical and genetic overlap and ultimately explained by a shared evolutionary history. On the other hand, there are aspects of our categorically divergent treatment of humans and animals that seem not just to reflect species-chauvinism or a lingering intellectual bias from a pre-scientific era, but to have a firm basis in reality—think, for example, of the differing institutions designed to address ‘misbehavior’ by a dog and normal human adult. It would be nice to fi nd a way to affi rm the truth in both sides of the argument. But the view that I have proposed seems weighted heavily on the side of discontinuity. Our ability to explain ourselves is at the heart of my conception of rationality: to believe or act for a reason is to be in a position to give an account of one’s belief or action. Since animals cannot do this, they do not believe or do things for reasons. I also have argued that theoretical and practical rationality have an essential epistemic dimension. To be thinkers and agents, animals must be knowers—but it is not obvious that they are. One doesn’t need to accept (b) to be troubled by these consequences. For it seems to be at odds with our very illuminating ordinary and scientific explanations of animal behavior, which are, to a significant degree, isomorphic to rational explanations of human behavior. But I argue in this chapter that my view does not, after all, require us to surrender the idea that animals, like people, are agents and knowers. It does not require pet owners and ethologists to back off from self-evident or hard-won explanatory truths about animal behavior. But what it is to act for a reason and what it is to know a fact differ, it turns out, in the cases of humans and animals. And this difference accounts for why human agency essentially involves self-consciousness, whereas the agency of animals—animal agency, as I’ll call it— does not. My response to the objection from continuity, in other words, is to concede that my account of responsiveness to reasons cannot be applied to animals, but to argue that this does not show that the account is inconsistent with their being responsive to reasons. Rather, it shows only that animal agency must be understood differently than human agency. Similarly, animals do have knowledge of facts. But what it comes to for an animal to know is different than what it comes to for a human to know. The underlying

118

Rational Causation

explanation for these differences, it turns out, is the same. Animals cannot take attitudes toward propositions. Human responsiveness to reasons and human knowledge essentially require the exercise of an ability to take such attitudes. And in part because of this difference, an animal neither needs to be nor can be in a position to explain itself when it acts for a reason. I begin, in the first few sections, by discussing the following inconsistent triad: (A) Animals are responsive to reasons. (B) Responsiveness to reasons requires knowledge of facts. (C) Animals do not have factual knowledge. Sections 1 and 2 establish the plausibility of (A) and (B), respectively. I adopt the following strategy for undermining the allure of (C). Although the explanation of animal action in terms of reasons does indeed depend on the animal’s knowledge of facts, this is not as high a hurdle as it might seem. Less is required for an animal than for a human to know a fact. I approach this thesis in stages. Knowledge, I argue, relates a knower not to a proposition but rather to a fact (section 3). All of the evidence explained by the hypothesis that animals are believers is better explained by the hypothesis that they are knowers (section 4). Furthermore, there is a positive reason for doubting that animals can believe anything (section 5). According to the view I defend, there are two forms of knowledge: human and animal. The former entails belief; the latter does not. Human knowledge does, whereas animal knowledge does not, require the conceptual wherewithal to grasp propositions. Hence human knowledge is a higher cognitive achievement than animal knowledge. In section 6 I attempt to say why this should be so by appealing to a general difference between the character of human and animal agency. Because animals do not believe, there are no explanations of why animals believe. There are, however, explanations of how animals know. I discuss this difference in section 7. Finally, I argue that the view that I defend in this chapter may help to shed light on a related dispute in the philosophy of perception, viz., over whether the content of human perception is conceptual in a sense in which animal perception is not (section 8). Before moving on to the first section, I make the following disclaimer: throughout this chapter I distinguish between human and animal knowl-

(Non-Human) Animals and Their Reasons

119

edge and agency. I speak generally and uncritically about what I describe as distinctively human abilities. But what is crucial for my purposes is just that there are two kinds of knowledge, and two kinds of responsiveness to reasons, and that creatures capable of having attitudes toward propositions have one kind of knowledge and responsiveness to reasons, and creatures not capable of having such attitudes may yet have the other. My use of the labels ‘human’ for the former and ‘animal’ for the latter reflects my view of which beings in fact have the relevant abilities and which do not; but that view does not play any role in the argument. I do argue that the typical grounds for thinking that animals must have propositional attitudes are faulty. But if there are good reasons for thinking that some of the ‘higher’ non-human species of animal do in fact have attitudes toward propositions, then I will be among the first to welcome them into the ranks of what I here call the ‘human,’ perhaps with a copy of this book. And though I then might owe them an apology for my chauvinistic terminology, the underlying thesis would not be threatened.

1. Animals Are Responsive to Reasons Just as we make a distinction between what a person does and what happens to a person, we make a distinction between what an animal does and what happens to an animal. If a man falls down the stairs, and a child were to ask, “Why did he do that?,” we are likely to reply with something like the following: “He didn’t do it on purpose. It was an accident. He tripped.” In so doing, we reinforce the idea that on some occasions of human movement, the question that asks for a man’s reason has no application. A very similar pattern of question and answer arises in connection with a dog’s falling down the stairs. In both cases, a reply that doesn’t reject the question is possible: “He’s practicing his tumbling,” we might say of the human, or “It’s a new trick we’ve taught him,” we might say of the dog. When a human or a dog tumbles on purpose, we can speak of the reason why the tumble was taken. The guises of animal action-explanation mirror those of human action. Here are examples: “My cat Opie was pawing the pantry door open in order to retrieve his treats”; “Opie was pawing the pantry door open because he was retrieving his treats”; “Opie was pawing the pantry door open because

120

Rational Causation

he intended to retrieve his treats”; “Opie was pawing the pantry door open because the treats were in the pantry.” As in the human case, an explanation in which we appeal to the unfolding of an event is more demanding than one in which we appeal to an intention. If Opie has no more treats, then he is not doing anything because he is retrieving his treats. Nonetheless we can explain his action by appealing to what he intended to do. Similarly, explanations that appeal to facts are more demanding than those that appeal to what the animal took to be the facts. Supposing that there are no treats in the pantry, Opie is not pawing the pantry door open because there are treats in the pantry. At best, Opie is pawing the pantry door open because he expects them to be there, or because they are generally in the pantry, or because the pantry still smells of treats. (A), then, is plausible for these reasons: First, there is a distinction between what an animal does and what happens to it. Second, there are explanations of animal behavior that exploit that distinction and that are isomorphic in obvious respects to rational explanations of human action.

2. Animal Responsiveness to Reasons Is Epistemic The patent success of reason-citing explanations of non-human animal behavior would appear to threaten any theory of acting for reasons that places the bar so high that only humans can surmount it. And it might seem as if this is just what I have done. According to the view propounded in the last chapter, Opie must know that the treats are in the pantry and that by pawing the pantry door open he can retrieve them. Must animals really know such things? Yes. An animal cannot be oblivious to them, on pain of denying that the animal is engaging in any purposeful behavior at all. If he is unaware that the treats are in the pantry, then his pawing the door open does not have the right connection to their presence. If it has not dawned on him that he might open the door by using his paws, then we could hardly think of his pawing as being a means to his opening the door. The language we use to describe Opie’s relation to his circumstances is unmistakably epistemic. The failure of it to have dawned on him that p just is his ignorance of the fact that p—the same for his obliviousness and unawareness.

(Non-Human) Animals and Their Reasons

121

This is not just a matter of our intuitions regarding when to use the word ‘knowledge.’ The epistemic dimension of animal explanation emerges from considering the epistemic requirement on world-citing explanations. Suppose that we have decided to move Opie’s treats to the shelf above the pantry. The treats then happened to fall behind the shelf and back into their usual position in the pantry. Opie, having not observed the switch and suffering from a stuffy nose, is completely unaware of these developments. Perhaps Opie paws the door open because the treats are usually in the pantry or because he expects them to be in the pantry. But he doesn’t paw the door open because they are in the pantry full stop. His expectation’s being met was merely fortuitous. Thus I argue that (B) is plausible in the case of animals for the same reasons it is plausible in the case of humans. First, animals act only in light of those facts of which they are aware. And second, the relevant sort of awareness is epistemic in character, arrived at by means of knowledgeconferring abilities such as perception and memory. For these reasons, epistemic failures force us to abandon world-citing explanations in the case of animals no less than in the case of humans. Some will no doubt object that all talk of attributing knowledge to animals is just anthropomorphizing—useful, perhaps, but not literally true. But (C) is not as easy to defend as one might at first think. Suppose an asteroid is about to strike Opie’s house. Barring Opie’s having very special cognitive powers, there is nothing that Opie can do because an asteroid is about to strike the house. Yet there is something that Opie can do because there are treats in the pantry. Someone who accepts (C) denies that the difference is that Opie is ignorant in the first case yet knows in the second case. Presumably a supporter of (C) would explain the difference in terms of a concept like belief. An epistemic internalist might reason: perhaps Opie believes that the treats are in the pantry, but he cannot justify his belief; hence, he doesn’t know. I argue, on the contrary, that to say that Opie believes that there are treats in the pantry is to credit him with a higher achievement than to say that he knows that they are in the pantry. To appreciate this, one must recognize the difference between objects of belief and objects of knowledge.

122

Rational Causation

3. Objects of Knowledge versus Objects of Belief According to Zeno Vendler, ‘that’ clauses are systematically ambiguous between a fact-interpretation and a proposition-interpretation. He argues, further, that ‘that p’ in “S believes that p” refers to a proposition, whereas ‘that p’ in “S knows that p” refers to a fact. 2 I now review some of the evidence for this claim. There is little consensus regarding the precise nature of propositions and facts. With respect to propositions, this much is relatively uncontroversial. They are the kinds of things that are true or false, that stand (or fail to stand) in logical relations such as entailment and consistency, and are the objects of certain mental attitudes, paradigmatically belief, doubt, and denial. For our purposes here, it is not necessary to decide the question of whether these roles are played by the senses of sentences, as Frege held, or are analyzable in terms of possible worlds,3 or in some other way. Regarding facts, few would contest this: to say a proposition is true is to say that it fits the facts; to say a proposition is false is to say that it fails to fit the facts. Frege famously held that facts just are true propositions,4 but most later philosophers have demurred.5 Just because propositions are the sorts of things that are true or false, we must understand them as having a referential character—they are faithful or unfaithful representations of the way things are. The way things are—the facts—are not representations of anything. In one important respect, however, the debate about the relation between true propositions and facts is different from other common debates about the identity or non-identity of metaphysical categories. Here are other examples: some hold that objects (or statements about objects) are actually events (or statements about events);6 some hold that events

2. Zeno Vendler, Res Cogitans (Ithaca, NY: Cornell University Press, 1972). 3. See Robert Stalnaker, Inquiry (Cambridge, MA: MIT Press, 1984). 4. See Gottlob Frege, The Foundations of Arithmetic, trans. J.  L. Austin (Oxford: Basil Blackwell, 1974). 5. An exception is Timothy Williamson, Knowledge and Its Limits (Oxford: Oxford University Press, 2000). 6. This is one way of understanding the perdurantist tradition in ontology. See David Lewis, On the Plurality of Worlds (Oxford: Blackwell, 1986); Mark Heller, The

(Non-Human) Animals and Their Reasons

123

(or statements about events) are actually states of affairs (or statements about states of affairs);7 and according to the various traditions of nominalism, abstract entities (or statements about abstract entities) are actually concrete entities (or statements about concrete entities).8 Debates about the relation between facts and propositions might seem simply to be another example. But this dispute is different in the following respect: there is no question that there are grammatical/conceptual distinctions between object- and event- expressions, between event- and stateexpressions, between abstracta- expressions and concretia-expressions. But philosophical considerations— e.g., considerations of metaphysical parsimony, the alleged ‘queerness’ of entities of certain kinds, the demands of a favored semantic theory— can make the reduction seem necessary or at least desirable. In the case of facts and propositions, however, their identification is tempting not solely as a consequence of external philosophical pressure, and despite a prima facie case to the contrary. Rather, their identification is tempting at least in part because of a prima facie case, which is itself an additional source of philosophical pressure. For we use the same grammatical constructions to talk about facts and propositions. If John utters the sentence ‘Grass is green,’ and Johann utters the sentence ‘Das Gras ist grün,’ we say that the sentences they uttered express the same proposition. To say which proposition they have uttered, we utter one of those same sentences again, or perhaps a synonymous expression in a third language. Or we might try to say which proposition by giving a nominalization of one or another of these sentences, e.g., that grass is Ontology of Physical Objects (Cambridge: Cambridge University Press, 1990); and Theodore Sider, Four-Dimensionalism (Oxford: Oxford University Press, 2001). 7. See Alvin Goldman, A Theory of Human Action (Princeton, NJ: Prentice-Hall, 1970), and Jaegwon Kim, “Events as Property Exemplifications,” in Action Theory, ed. M. Brand and D. Walton (Dordrecht: Reidel, 1976), 159–177. 8. Perhaps the most currently popu lar form of nominalism is the trope-theoretic account of universals. See G.  F. Stout, “Are the Characteristics of Par tic u lar Things Universal or Par tic u lar?,” Aristotelian Society Supplementary Volume 3 (1923): 114–127; D.  C. Williams, “On the Elements of Being: I,” Review of Metaphysics 7 (1953): 3–18; and Keith Campbell, “The Metaphysics of Abstract Particulars,” in Properties, ed. D. H. Mellor and A. Oliver (Oxford: Oxford University Press, 1981), 125–139.

124

Rational Causation

green. Or we may adopt a more heavy-handed approach and say “the proposition that grass is green.” Grass is green. That’s a fact. Das Gras ist grün. That is also a fact. Wait . . . is that also a fact? At best, the ‘also’ is misleading, for it is not an additional fact. The second and fourth sentences of this paragraph use a demonstrative ‘that,’ which designates one fact. And the fact in question is the one designated by the first and third sentences. Let us try to say which fact this is. Our options are the same as above. We can repeat one or another of the original sentences. We can nominalize one or another of these sentences. Or we can say simply “the fact that grass is green.” There is, then, an interesting grammatical isomorphism between factand proposition-talk. Perhaps this helps to explain why Vendler’s thesis has been so thoroughly neglected.9 Still, there are well-known reasons for resisting the prima facie grammatical case. Perhaps the most well-known is sometimes put this way: the fact that grass is green would not ‘exist’ if grass were not green, but the proposition that grass is green would.10 Put differently, it would not be a fact that grass is green if grass were not green. But ‘grass is green’ would still express a proposition if grass were not green. This is related to the difference of which I spoke earlier: that grass is green, understood as a proposition, is the sort of thing that is true or false. But a fact is not the sort of thing that is true or false. Either it is a fact, or it is not. Frege’s influence notwithstanding, then, it is generally accepted that ‘that’ clauses refer to both facts and propositions. So let us consider Vendler’s proposal regarding the differing complements of ‘S knows ____’ and ‘S believes ____.’ Certainly, if ‘that’ clauses are ambiguous, then we cannot take it for granted that the two verb-phrases are directed at the same kinds of objects. What is Vendler’s evidence? I now go through some examples. (1) John believes that grass is green.

9. But see Kenneth Sayre, Belief and Knowledge: Mapping the Cognitive Landscape (Lanham, MD: Rowman and Littlefield, 1997), and Gilbert Harman, “Category Mistakes in M&E,” Philosophical Perspectives 17, no. 1 (2003): 165–180. 10. G. E. Moore, Some Main Problems of Philosophy (London: Unwin, 1953).

(Non-Human) Animals and Their Reasons

125

(2) I know what John believes. (3) I believe what John believes. Vendler’s thesis, recall, is that ‘S knows ____’ takes facts as objects, and ‘S believes ____’ takes propositions as objects. If that’s right, then ‘what John believes’ in (2) cannot refer to the object of John’s belief—the proposition on display in (1)— since that object is not the right kind of thing. It is not a fact. And Vendler’s view predicts accurately that ‘what John believes’ doesn’t refer to the object of John’s belief. (2) doesn’t mean that I know that grass is green. Rather it means that I know that John believes grass is green, which is the fact on display in (1). Vendler holds that (2) is similar to (4) I know what he lost in which ‘what he lost’ is a new sentence-nominalization—a ‘whnominalization,’ as Vender puts it— of ‘that he lost a watch.’ (4) is thus a transformation of: (4a) I know that he lost his watch. Similarly, (2) I know what John believes is a transformation of (2a) I know that John believes grass is green. Since, according to Vendler, it is propositions that are believed, ‘what John believes’ in (3) must be analyzed differently. His view predicts, correctly, that ‘what John believes’ refers back to the proposition on display in (1), viz., the proposition that grass is green. Vendler holds that (3) is similar to (5) I found what he lost

126

Rational Causation

which is a transformation of (5a) I found that which he lost [namely, the watch]. (5a) in turn derives from (5b) I found the watch, which he lost. Here, ‘what he lost’ is ultimately derived from a relative clause. Similarly, (3) I believe what he believes is, according to Vendler, a transformation of (3a) I believe that which he believes [namely, that grass is green]. The attempt to give a relative-clause reading of the what-clause in (4) yields the obvious misinterpretation (4b)* I know his watch. And the attempt to do the same for (2) yields the obvious misinterpretation (2b)* I know [the proposition that] grass is green. The attempt to give a wh-nominalization reading of the ‘what’-clause in (5) yields the obvious misinterpretation (5c)* I found that he lost the watch. And the attempt to do the same for (3) yields the obvious misinterpretation (3b)* I believe that he believes that grass is green. Consider now:

(Non-Human) Animals and Their Reasons

127

(6) John knows that grass is green. (7) I know what John knows. (8) I believe what John knows. (*?) Since there are two facts on display in (6)—the fact that John knows that grass is green, and the fact that grass is green—Vendler’s theory predicts that (7) is ambiguous. It could derive from (7a) I know that which he knows Or (7b) I know that he knows that grass is green. And in fact (7) is ambiguous in just this way. Someone who wants to learn everything that John already knows about grass is looking for the man who says (7a). Someone who wants to know just how much John has found out about the color of grass is looking for the man who says (7b). Vendler calls (8) deviant.11 I have found that those who have no acquaintance with contemporary epistemology tend to agree with Vendler. Those with a background in epistemology, however, tend to hear the pragmatically infelicitous yet intelligible (8a) I (merely) believe what John (truly) knows. Since it is a part of basic training in epistemology to allow that to know that p is to believe (the very same) p and also to satisfy certain other requirements, this discrepancy is not surprising. But even so, what proposition is grasped by philosophers who understand (8a)? To answer this question, I first circle back and discuss the pronoun ‘it,’ where ‘it’ is anaphorically connected to sentences containing ‘that’ clauses. According to Vendler, ‘S says ____,’ like ‘S believes ____’ takes propositions as objects. Consider now: 11. Vendler, Res Cogitans, 111.

128

Rational Causation

(9) John said that the French would attack by nightfall, and Frank believed it. Since, if Vendler is right, the only proposition on display in (9) is the proposition that the French would attack by nightfall, his theory predicts that the ‘it’ refers back to what John said. And this is by far the most natural reading. Could the ‘it’ refer back to John’s having said that the French would attack by nightfall? Under pressure, perhaps. We can be very resourceful interpreters, as we shall see shortly. Consider: (10) John said that the French would attack by nightfall, and Frank knew it. If Vendler is right, then we would expect the ‘it’ in (10) to refer to the fact that John had said that the French would attack by nightfall—for this is, according to Vendler, the only fact on display in (10). And indeed this is by far the most natural interpretation. Let’s now consider one of those contexts in which there is pressure  to  abandon the natural interpretations, which seem to support Vendler. (11) John said that the French would attack by nightfall. Frank believed it—in fact, he knew it. In this case, which is similar to (8a), both uses of ‘it’ are obviously anaphorically connected to the same expression, viz., ‘that the French would attack by nightfall.’ And since, according to Vendler, the expression in this context designates a proposition, it would appear that the object of ‘S knows ____’ can indeed be a proposition. But another interpretation is available. Perhaps the second ‘it,’ under threat of meaninglessness, exploits an ambiguity in the ‘that’-clause to which it harks back. Depending on the context, ‘that the French would attack by nightfall’ can refer to a fact or a proposition. And making sense of (11) requires that we hear it first as the latter and then as the former. This explanation assimilates (11) and (8a) to a sentence such as

(Non-Human) Animals and Their Reasons

129

(12) Ron was still off his rocker and his medications. We know what (12) means, but its intelligibility depends on our making ‘off’ do double-duty—syllepsis. Consider these: (13) They were going to kill George, and he knew it. (14) They were going to kill George, and he believed it. (*?) The ‘it’ in (13) refers to the fact that they were going to kill George. The ‘it’ in (14), on Vendler’s view, is rootless, leaving the listener searching for a proposition that is not available. Having nowhere else to go, she settles on the other meaning of ‘they were going to kill George,’ according to which it designates a proposition. The fact that the ‘it’ in (13) designates a fact makes (15) go down very smoothly: (15) They were going to kill George. He knew it, and knew that it would break his wife’s heart. What George knew—the fact that they were going to kill him—is the sort of thing that can break someone’s heart. But now look at: (16) John said that they were going to kill George. George believed it, and believed that it would break his wife’s heart. What George believed—the proposition that they were going to kill him—is obviously not the heartbreaker. The second ‘it,’ then, doesn’t refer to the same thing as the first ‘it.’ Still, (16) is perfectly intelligible. And that’s because the second ‘it’ relies on a different meaning of ‘they were going to kill George’—the meaning that ‘it’ has when it is the object of a verb phrase like ‘S knows ____.’ And so we see that the fact that two occurrences of ‘it’ refer back to the same ‘that’-clause does not entail that this clause has the same interpretation in the case of each ‘it.’ I will take it as established in what follows that knowledge relates a subject to a fact and that belief relates a subject to a proposition.

130

Rational Causation

4. Evidence Supporting Animal Belief Better Supports Animal Knowledge Here is why Vendler’s thesis is important for our discussion: since the ‘that’-clause in ‘S knows ____’ designates a fact, whereas the ‘that’ clause in ‘S believes ____’ designates a proposition, it is no longer obvious that knowing that p requires everything that believing that p requires, only more. This will strike the reader as a peculiar claim. The view that believing that p is a necessary condition of knowing that p garners perhaps as much consensus as any other substantive philosophical claim. Even those who abjure the tradition of analyzing knowledge in terms of belief tend to accept that believing that p is a requirement of knowing that p.12 And this view is, I think, unassailable in its application to humans. In its application to animals, however, it is false. For animals are incapable of having any propositional attitudes. There is a simple argument to that effect, which I shall present shortly. But I have found that the argument is more effective rhetorically if I prepare the way by first dispelling two sources of prejudice against its conclusion. Many think that the idea that animals have propositional attitudes is obviously sound. There are two central reasons: First, it seems to be entailed by the fact that animal behavior admits of intentional psychological explanation.13 Second, the attribution of propositional attitudes to animals is part of the best scientific theories of animal cognition.14 I take up these points in turn. Animals, it will be argued, must have attitudes toward propositions or else our purposive explanations of their behavior are not literally true. If Opie doesn’t believe that the treats are in the pantry, for example, then he cannot paw the door open because the treats are in the pantry. I reply that although we surely must appeal to Opie’s cognitive abilities if we are to

12. E.g., Williamson, Knowledge and Its Limits. 13. See Jerry Fodor, “Propositional Attitudes,” Monist 61 (1978): 573–591, and Daniel Dennett, The Intentional Stance (Cambridge, MA: MIT Press, 1987). 14. See Richard Routley, “Alleged Problems in Attributing Beliefs and Intentionality to Animals,” Inquiry 24 (1981): 385–417.

(Non-Human) Animals and Their Reasons

131

account for his behavior, these abilities put him in touch not with propositions but rather with facts. Opie knows that the treats are in the pantry because he saw them in the pantry, or because he remembers that they were in the pantry, or because he’s learned that when Bruno (my dog) sniffs around the pantry, the treats are generally inside. It is knowledge acquired through some such means that explains how he can do something because the treats are in the pantry. At no point in this story need we bring in an animal’s attitude toward a proposition or any other representational entity. But, it will be replied, we can still make rational sense of an animal’s behavior even when it does not know, e.g., “Opie pawed the door open because he believed that there were treats in the pantry.” Even when an animal isn’t misled, we can explain its behavior in this epistemically neutral manner. The point of these explanations is not, however, to cite mental contact between an animal and a proposition. Rather, their point is to explain the action in a manner that abstracts from the animal’s epistemic success. In the case of persons, citing what they merely believe serves this purpose (for reasons we shall consider below). But it can also be served by explanations that do not appeal to any propositional attitudes. If the treats are no longer in the pantry, we can explain his behavior by saying that Opie expected the treats to be in the pantry. If Opie runs away from a non-existent threat, we can say that Opie ran under the bed because he was afraid that the neighbor’s dog was after him again. The object of expectation and fear are not (or, in any case, not typically) such truth-valued entities as propositions but rather facts, events, and objects. More must be said here. If animals lack the cognitive sophistication to grasp propositions, they surely can fi nd no purchase on representations of non- obtaining states of affairs, non- occurring events, and non- existent objects. It will also not do to credit animals with cognitive  access to non- existent worlds. But to say that an animal expects some event to occur is not to attribute to an animal assent to the accuracy of a representation, even if to expect something is to be in a representational state. An animal can be in a representational state without there being a representational entity that functions as a representation for the animal, in the way sentences, the propositions that sentences

132

Rational Causation

express, and pictures, among other things, function as representations for us.15 My approach to animal expectation is based on the idea that it is the sort of intentional directedness toward the world that culminates, in successful cases, in animal knowledge of a fact, e.g., Bruno’s knowledge that I am about to walk him. Were Bruno’s expectation misplaced, Bruno would nonetheless represent an occurrence of my walking him as immanent. But this is not to say that he has an attitude toward a representation of a possible event. An animal’s expectation, we should say, is not an attitude toward an entity that represents things as thus-and-so, but rather it is the animal itself representing things as thus-and-so. My argument is perfectly consistent with animal representation in this latter sense. Nonetheless animals cannot believe, for belief is an attitude toward a representation. Animals, I contend, are knowers but not believers. They are knowers in virtue of perceiving, remembering, and figuring stuff out. I don’t propose here to investigate precisely which sorts of facts various animals can and cannot know— a topic of contemporary empirical research. Obviously, the perceptual and cognitive powers of different animals differ, putting them on to different ranges of facts (and thereby also delimiting different ranges of possible expectation). It is consistent with my view that the extent of their knowledge is considerable, including such things as what Bermudez calls ‘proto-causal conditional states of affairs’16 (e.g., that the sound of water running in the morning is a result of the humans’ being awake), and also of the epistemic and perceptual states of other animals (e.g., knowledge that the dominant chimp doesn’t see hidden food).17 The important point is this: in giving an animal’s reasons for doing what it does, we appeal to such epistemic abilities. Just as in the human case, where an

15. Some may fi nd it tempting to invoke sub-animal-level representational states as a way of characterizing what it is for an animal to represent (by expecting, for example) without having an attitude toward a representation. Expectations and the like, it might be thought, should be identified with informational states of the animal’s cognitive subsystems. I am doubtful of this, but will not pursue the issue any further here. 16. J. L. Bermúdez, Thinking without Words (Oxford: Oxford University Press, 2003). 17. See Brian Hare, Josep Call, Bryan Agnetta, and Michael Tomasello, “Chimpanzees Know What Conspecifics Do and Do Not See,” Animal Behaviour 59 (2000): 771–785.

(Non-Human) Animals and Their Reasons

133

animal f s because p, we can ask: how does it know that p? And if no good answer is forthcoming— either because p is false, or because the animal doesn’t know that p—we must revert to a form of explanation that is epistemically neutral. And although we sometimes use propositional-attitudetalk to frame such explanations, such talk, which I do regard as anthropomorphic, is dispensable. This error theory of explanations such as “Opie pawed the door open because he believed that there were treats in the pantry” is, qua error theory, charmless. But what it lacks in charm it makes up for in avoiding the absurdity of the opposing view, which I hope to demonstrate after considering a second ground for thinking that animals must take attitudes toward propositions. Here is a second rationale for the view that I oppose: in giving a scientific explanation of how it is possible that an animal sees, remembers, and learns, we invoke internal representations of events and states of affairs. And a belief just is (according to a widely held philosophical view) an internal state that represents propositionally the world as being a certain way— a state whose content is satisfied if and only if the world is that way— and that helps to explain how the believer manages to achieve relevant goals. The question of whether animals have beliefs, and hence take attitudes toward propositions, is thus simply the question of whether there exists inside their brains a network of neural representations of the world outside their bodies that meets certain basic conditions— centrally, whether individual representations have parts that are combined and recombined in the manner of words, as opposed, e.g., to images— and whether these representations play the appropriate causal role in the mediation of, among other things, perception and action. And there is, according to this line of thought, good evidence for believing that animal brains do implement such networks. But the issue raised by my argument is prior to the question of whether animal beliefs can be identified with internal representational states. We do not attribute beliefs to humans on the basis of knowledge of or speculation about the neural architecture of their brains. Rather, we are moved to look for neural ‘correlates’ or ‘realizers’ of belief in part because we come to philosophy already attributing beliefs to humans. My argument, if successful, undercuts the motivation for trying to find the ‘correlates’ or

134

Rational Causation

‘realizers’ of animal beliefs. My conclusion would still, in principle, leave room for the idea that such internal states correspond to other cognitive states of animals—just not beliefs.

5. An Argument against Animal Belief Here is my argument:18 (1) Having an attitude toward a proposition requires understanding it. (2) Animals cannot understand propositions. Therefore, (3) Animals have no attitudes toward propositions. In defense of (1): attitudes toward propositions involve taking a stance about the truth of a proposition. One thinks that it is true or false, or one withholds judgment. One suspects or wishes it were true or false. One holds that it is probably or possibly or almost certainly or certainly not true. It is utterly mysterious how a creature incapable of understanding a proposition could have any view of such matters. We don’t count any person as having any kind of opinion about the truth or falsehood of a proposition whom we don’t take to be capable of understanding it. If James lacks the cognitive wherewithal to grasp propositions concerning the housing bubble, he cannot have any sort of view about the truth of the proposition that the Federal Reserve is responsible for the

18. My argument bears some similarity to those of René Descartes, Discourse on Method, in Descartes: Selected Philosophical Writings, trans. John Cottingham, Robert Stoothoff, and Dugald Murdoch (Cambridge: Cambridge University Press, 1988); Stephen Stich, “Do Animals Have Beliefs?,” Australasian Journal of Philosophy 57 (1979): 15–28; and Donald Davidson, “Rational Animals,” Dialectica 36 (1982): 317– 327. But it is more narrowly focused on the question of whether animals have attitudes toward propositions. The conclusion of my argument is consistent with what those earlier arguments were in part designed to rule out: animal knowledge. (See also the argument against animal’s possessing ‘higher-order’ propositional attitudes in Bermúdez, Thinking without Words.)

(Non-Human) Animals and Their Reasons

135

housing bubble. He may have a view about a sentence to which he has perceptual access but not to the proposition it expresses. There is no reason to multiply examples here. This standard works the same way in every case. This thesis is not the consequence of any specific theory of what it is to understand a proposition. Rather, specific theories are less plausible to the extent that they unmoor taking stances toward propositions from understanding them. The intuitive plausibility of (1) puts the burden of proof on those who would reject it. It might be argued, in the fashion of the second objection above, that having an attitude toward a proposition amounts to one’s body containing the right internal representation in the right spot in the efficient-causal order. An animal does not need to understand internal representations in order to believe the propositions they express; it just needs to contain them. This sort of reductionism is often simply taken for granted. But the issue here is whether, in the relevant sort of case, there is any reason for thinking that there is a propositional attitude to reduce. According to the first premise, it is only the states of beings that are capable of understanding propositions that are the appropriate candidate for propositionalattitude reductions. Perhaps there are proposition-expressing states in the brains of animals, but without further argument, these cannot be identified with the animal’s own propositional attitudes. We can, of course, use the phrase ‘propositional attitude’ to refer both to the understanding-requiring states of persons and the understandingindifferent states of animals and their parts. But in so doing we bury a crucial distinction, one that is vital for a proper account of the difference between human and animal minds. It would be perverse to coin technical terms for the purpose of grouping together what are fundamentally distinct. It also might be argued, in the fashion of the first objection above, that a creature counts as having an attitude toward a proposition in virtue of behaving as if the proposition is true, likely, possible, unlikely, or false. But as I argued above, rational explanations of animal behavior invoke epistemic and not doxastic abilities. We can account for the purposive behavior of animals without ascribing any attitudes toward propositions. Such explanations require the (fallible) capacity to be aware of facts as well as other attitudes (e.g., expectation and fear) toward non-representational

136

Rational Causation

entities (e.g., events and objects). But they do not require a grasp of propositions. Let us move now to (2), the thesis that animals cannot understand propositions. The truth is that I don’t care much about it. I suspect that it is true. But if there are some animals that can, in fact, understand propositions, more power to them. My central goal is ultimately to draw a line, not to say precisely which creatures fall on which side of the line. On one side of the line are creatures that do understand propositions and act for reasons; on the other side are creatures that cannot understand propositions but do still act for reasons. Ultimately, I argue here that not only is the nature of the factual knowledge possessed by beings of the two sorts fundamentally different, but so is the nature of their abilities to act for reasons. Still, it will be useful to consider the prima facie case against the idea that animals understand propositions. A natural test for whether a being grasps a proposition is whether it is capable of asserting or denying the proposition. Failing that, there ought to be some evidence that it is sensitive to the syntactic and semantic significance of those parts of sentences that determine (in a specific context) what propositions sentences express. Certainly, we don’t count people as understanding propositions without some such evidence. A signature at the bottom of a contract means nothing unless there is some evidence that the signatory understood or was at least capable of understanding what the contract said—unless she understood or was capable of understanding the sentences, in the sense that requires knowing or potentially knowing what propositions the sentences expressed. As far as I know, there is no decisive evidence in favor of animals satisfying either of these requirements. Once again, the same objections reappear. Perhaps understanding is a matter of internal representations. Perhaps understanding is just a matter of behaving as if one thinks certain propositions are true or false. As above, I reply: the animal’s inability to manifest the requisite sensitivity to the meaning of the linguistic items that express propositions preempts any investigation of what understanding such expressions ‘consists in’ at the neurological level. And purposeful behavior can be explained without recourse to the ability to understand propositions.

(Non-Human) Animals and Their Reasons

137

6. Animal Agency Why should we accept that there are two species of knowledge, human and animal, distinguishable by the fact that only the former sort requires belief? And if knowledge is a relation to a fact, why should knowledge have any implications for belief even in the case of humans? In this section, I sketch answers to these questions. In Chapter 1 I argued for the idea that knowledge that p is necessary for acting because p (in the rational sense of ‘because’). Hyman has argued— persuasively, by my lights—for a stronger claim: “a person knows that p if and only if . . . the fact that p can be among her reasons for performing a certain kind of action or for refraining from performing a certain kind of action; for believing or doubting or hoping or wanting something; or for not believing or doubting or hoping or wanting something; and so on.”19 Knowledge that p amounts to or is in any case necessary and sufficient for an ability to f because p. My explanation of the way knowledge speciates builds on Hyman’s idea. If knowledge and rationality were connected as Hyman envisages, one would expect that any difference between human and animal knowledge should correspond to a difference between human and animal agency. According to the account of Chapter 2, an action is a sort of event that is characterized by the fact that its subject can say that it is occurring and not on the basis of observation or evidence. An agent can just say she is performing the action. Though not every action is performed for a reason, in the case of actions that are performed for reasons, an agent can just say not only what she is doing but also why. Often, and in the central cases, the answer to the reason-demanding why-question directs one to a further action, e.g., “I am tasting all the desserts because I’m reviewing the restaurant.” But sometimes an answer directs one to a fact that bears on such an instrumental connection: “because I am the restaurant critic for the City Pages.” Here, p is S’s reason for f -ing only if S can just say, in reply to a query, “because p” (or something to that effect). As we have seen, there is quite a lot to say about what is going on when an agent gives such an answer. My focus in this chapter is narrow. In saying 19. John Hyman, “Knowledge and Evidence,” Mind 115 (2006): 891– 916, 894.

138

Rational Causation

“because p” or “for the reason that p” or “on account of p,” an agent gives her reason and makes explicit what she (purportedly) knows by, among other things, expressing an affirmative attitude toward the proposition that p. Perhaps it is wrong to say that she has asserted the proposition, but she has nonetheless revealed that she believes it. The ability to give one’s reason is part of what it is to act for that reason. And so part of what it is to act on the basis of a fact is to believe the corresponding proposition. Earlier I asked this question: if knowledge is a relation to a fact, why should knowledge have any implications for belief, even in the case of humans? Putting Hyman and my Anscombe-inspired account together, we get this answer: knowledge that p is the ability to f because p; and f -ing because p requires, in humans, taking a propositional attitude toward p. Hence knowing (the fact that) p entails believing (the proposition that) p in the case of humans. This brings us to the other question with which I began this section. Why should knowledge speciate into human and animal varieties? A preliminary answer: because acting for reasons does, too. The account in Chapter 2 is not designed to shed any light on, e.g., what it is for a cat to paw a door open because there are treats inside. To say that Opie is pawing the door open because there are treats in the pantry in no way depends on his ability to say, in response to a request for an explanation, “because the treats are in the pantry.” According to my view, then, believing p is a requirement of a person’s knowing that p not because knowledge as such requires belief but because the human ability to act in light of a fact depends upon having an attitude toward the relevant proposition, whereas the corresponding animal ability does not. For a human but not for an animal, acting on the basis of a fact consists partly in representing a proposition designating the fact as to be believed, and this just is to be in a position to state the fact in response to a request for one’s reason. To state such a fact is both to give an account of one’s behavior and to say what one knows. In saying what one knows, one makes a statement that expresses a proposition that one believes.

(Non-Human) Animals and Their Reasons

139

7. Explaining Belief versus Explaining Knowledge It has been a commonplace since at least the time of Ryle that there is a fundamental difference between “Why does he believe?” and “How does he know?”— even if we confi ne our attention to cases of knowledge that.20 The former question seems to ask for something more like a cause, the latter for something more like an entitlement. While I do not deny that there is some truth to this, and that there are a number of other significant differences as well, I also think there is an important connection between the questions, one that will shed light on the specific kind of agency possessed by animals. In par ticu lar, I hope it will help to shed light on the similarities and dissimilarities between rational explanations of what people believe and reasons-based explanations of what animals know. “Why does he believe?” sounds like a request for a causal explanation, whereas “How does he know?” sounds like a request for a justification. Why is this? In part because, often, we ask why someone believes something in order to find out either his bad reason for believing or the aspect of his psychological history that explains his epistemically questionable belief. If you take Ray to know that Marlo killed Joe, then you would not ask why Ray believes that Marlo killed Joe—unless, perhaps, you were deliberately attempting to mislead. If you ask why Ray believes, you thereby suggest or at least leave open the possibility that he doesn’t know. Perhaps you do this because you know he doesn’t know, or you don’t know he does know, or you don’t want it to seem as if you know that he knows. To ask why he believes can thus be a request for the genesis of the error—the mendacious witness, the bogus evidence, the hasty assumption, the lying eyes, the pre-existing animosity toward Marlo, etc. In some of these cases, the answer is a mere cause (e.g., the last two); in others the answer gives a (bad) reason (e.g., the first two). But answers to “How does he know?” are also answers to “Why does he believe?” Any of the following answers would do just as well for both questions: “He saw it with his own eyes,” “He read it in the paper,” “He himself hired Marlo to do it,” or “Marlo was the only suspect who lacked an alibi.” 20. Gilbert Ryle, The Concept of Mind (New York: Barnes and Noble, 1949), chapter 5.

140

Rational Causation

What does any of this have to do with animals? Animals do not have attitudes toward propositions and thus there are no rational explanations of their holding such attitudes. Yet we can say how they know what they do: we can cite the specific exercise of the relevant epistemic ability, e.g., because Opie remembers that the treats are kept in the pantry. But like all abilities, an animal’s epistemic abilities sometimes fail. And when they do, the result is an animal that is not in touch with the facts. And in such cases, we can explain their failure to be in touch with the facts by appealing to the failure of the ability and whatever explains that failure. Consider the following example. Typically when I pick up Bruno’s leash, I am taking him on a walk. He has learned this. And so when I pick up his leash, he becomes excited. Now consider a particular occasion. I have picked up the leash; Bruno is excited. A guest asks: “Why is he so excited?” “Because I’m taking him on a walk,” I answer. This ‘why’ question asks for Bruno’s reason. You can tell by what happens when Bruno does not know that we are going on a walk. If I have only picked up the leash in order to strike Opie with it ( just an example, I promise), then my answer would have to change. I might say instead, “he’s excited because he expects me to walk him” or “he’s excited because he thinks that I’m going to walk him.” I argued above that even in the latter case, we must not think that what explains his excitement is his attitude toward a certain proposition. Rather, it is his expectation of a certain event. Suppose the guest asks: “Why does he think you’re going to walk him?” My answer: “Because I picked up his leash—typically I only do that when I’m taking him on a walk.” Here, we give an account of Bruno’s thought— not of his belief that a certain proposition is true but rather of his expectation that a certain event will soon commence. It’s an account that appeals to an epistemic ability: his expectation would have constituted knowledge that I would walk him had circumstances been typical. And his knowledge would have been based on his having learned that my picking up his leash is an indication that I am going to walk him. But it does not depend on an animal’s representing any proposition as to be believed. The explanation of animal thought, then, should be understood as follows. In cases of success, we say how an animal knows. In cases of failure, we explain what it mistakenly thinks by appealing to the failed exercise of an ability whose successful exercises culminate in knowledge. Much of

(Non-Human) Animals and Their Reasons

141

the same explanatory apparatus is available to us in explaining human beings. But in the case of humans, these explanations account not just for what is known (or what fails to be known) but also what is believed. In the case of humans, if q is how S knows that p, then, necessarily, S views p as to be believed on the basis of q. In successful cases of inference, q is already represented as to be believed; and the thinker extends this status to p and thereby knows that p. The more general point, the one relevant to the current discussion, is this: in the case of human beings, how one knows that p—at least when this concerns the rational basis of knowledge—is necessarily represented by the knower as how he knows. And the same, as we have seen, goes for why S believes that p, where what is in question is a rational explanation of belief.

8. Aside on Why Human (But Not Animal) Perception Is Conceptual There is a debate in the literature on perception about whether a person must have the concepts required to think that p in order to perceive that p—whether human perception is conceptual. Among the most influential reasons for holding that human perception is non-conceptual is that animals, who are thought to lack concepts, can perceive.21 Non-conceptualists about human perception find in some feature of animal perception reason to deny a categorical difference in kinds of perception. Conceptualists hold that there are overriding reasons to insist upon such a difference. But the common assumption that animals lack concepts sits uncomfortably with the equally common view that animal behavior can be explained by appealing to what they believe. For the sort of concepts that are thought to distinguish humans from animals are just those cognitive abilities paradigmatically exercised in belief. If animals can believe, it is at least a little puzzling that they should be the locus of discussions of nonconceptual perception. This puzzle reflects the powerful attraction of contradictory

21. See Gareth Evans, The Varieties of Reference (Oxford: Oxford University Press, 1982); Fred Dretske, Naturalizing the Mind (Cambridge, MA: MIT Press, 1995); Christopher Peacocke, “Does Perception Have a Nonconceptual Content?,” Journal of Philosophy 98 (2001): 239–264; and Bermúdez, Thinking without Words.

142

Rational Causation

views of animals. The approach I have defended in this chapter reveals a strategy for diffusing this tension. The strategy is based in part on denying that animals believe, and hence that animals possess the cognitive repertoire that belief requires. But the foregoing discussion should also render less implausible the idea that human and animal perception should differ so fundamentally. What differentiates human knowledge and agency from animal knowledge and agency is the former’s self-consciousness: the ability of a person to say what she is doing and why, and so to make explicit what she knows. Because human knowledge is articulate, humans are believers. As a consequence, whereas there are explanations—inter alia, explanations that appeal to perception— of how both animals and humans know what they do, only humans have reasons for belief. John McDowell has long argued that the distinctively conceptual character of human perception is bound up with the latter fact.22 My discussion here points toward the same idea, albeit from a slightly different vantage point. Suppose Bruno and I see Opie knock over a vase. In mine but not Bruno’s case, seeing this confers upon me the ability to say that I saw it. My perception furnishes me with an awareness of my own perceptions, an awareness constituted by or manifested in the ability to just say (authoritatively but not on the basis of observation or evidence) such things as “I saw such and such.” No state that fails to confer such an ability on a person is her perception.23 We would not count a person as having seen something if she at no time had the ability to report her having seen it. But animal perception is not self-conscious in this sense. If part of having seen something is the expressibility of my having seen it, then I see only that which I am able to say that I see. But the ability to say that I saw Opie knock the vase over presupposes having the conceptual repertoire to take an affirmative stance toward the proposition that Opie knocked over the vase. For one cannot have the cognitive wherewithal to state that one saw Opie knock over the vase without the ability to believe that he knocked it over. The power of perception in humans is inseparable

22. McDowell, Mind and World. 23. My thinking here is indebted to chapter 5 of Sebastian Rödl, Self- Consciousness (Cambridge, MA: Harvard University Press, 2007).

(Non-Human) Animals and Their Reasons

143

from the ability to believe and hence depends on the subject’s possession of the concepts necessary to grasp the relevant propositions and judge them true. Animal perception makes no such requirements. If my reason for believing that Opie knocked the vase over is that I saw him do it, the content that matters as far as my ability to justify my belief goes is not that of a perception but that of a proposition that describes a perceptual fact. There is no need to bring in the notion of perceptual content at all in order to establish the distinctively conceptual character of human perceptual experience. This is not to say that my reason for believing that Opie knocked over the vase is that I believe that I saw him do it. My reason is the fact that I saw him do it. But for this to be my reason, I must be able to cite it as my reason, and that depends upon conceptual abilities that are actualized in believing that I saw him do it. It is precisely these abilities that are actualized in perception itself, on account of the self-consciousness of human perception. In humans, perception is an ability whose exercises are themselves as such fit to serve as reasons for belief. And part of this fitness is the subject’s consciousness of her own perceptions. Animals, too, know on the basis of perception. Opie might know that Bruno was darting around the room because Opie can see Bruno doing it. Such knowledge enables Opie to hide under the couch because Bruno is darting around the room. If I were asked how Opie knew that Bruno was darting around the room, I could say: “because he saw Bruno doing it.” But the truth of this explanation, unlike the truth of the corresponding explanation in the case of humans, does not depend on the animal’s ability to advert to its own perception in explanation of its knowledge. Thus, the “how does he know?” question has application; the “why does he believe?” question does not. (For the same reason, it makes no sense to address the “how do you know” question to Opie.) This ought to render conceptualism about human perception at least a little less strange. This chapter answers the following objection: since the actions of animals can be explained by appeal to reasons despite animals’ being unable to explain their own behavior, the view propounded in Chapters 1 and 2 must be wrong. It is wrong, I concede, if construed as an account of responsiveness to reasons tout court. But agency takes a different form in humans than in

144

Rational Causation

animals. In answering this objection, I hope to have arrived at something independently worthwhile: a sketch of a strategy for revealing a network of interrelated similarities and dissimilarities in human and animal life. Humans and animals are agents. Perception furnishes us both with knowledge, which in turn makes possible acting in light of facts. Animals, no less than humans, learn from what they see and act on the basis of what they learn. But responsiveness to reasons in the human case is tied to a capacity for selfexplanation that depends upon the conceptualization of perception, action, and knowledge. What are at one level of abstraction the same faculties in humans and animals thus demand very different accounts.

4 Rational Explanation and Rational Causation

Rational explanation is a kind of causal explanation, according to the orthodox, psychologistic view. Indeed, many prominent critics of Psychologism also embrace causalism about rational explanation, as I will call it.1 Prior to Davidson’s “Actions, Reasons, and Causes,” the issue was treated, much more so than now, as an open question. This development has been remarkable in part because it is somewhat mysterious. As I shall show here, it is surprisingly difficult to say how precisely that paper’s most influential argument is supposed to work. When the dust settles, we shall see that his argument rules out much less than typically thought—indeed,

1. See John McDowell, Mind and World (Cambridge, MA: Harvard University Press, 1994); Jennifer Hornsby, Simple Mindedness (Cambridge, MA: Harvard University Press, 1997); and Helen Steward, The Ontology of Mind (Oxford: Oxford University Press, 2000). Other prominent critics reject causalism. See, e.g., George Wilson, The Intentionality of Human Action (Stanford, CA: Stanford University Press, 1989), and Jonathan Dancy, Practical Reality (New York: Oxford University Press, 2000).

146

Rational Causation

its conclusion may even be consistent with the views of some of the avowed anti-causalists against whom it was targeted. The point of this chapter is to defend the idea that the kinds of rational explanation that I have discussed in Chapters 1 and 2 are made true by rational causation. The explanantia of rational explanations designate causes of belief and action. Further, the specific form of understanding conveyed by giving someone’s reasons for belief or action constitutes a grasp of the nature of the causation that links the reason and its effect. But the meaning of these claims will be obscure until we have become clear about certain basic terminological and methodological issues pertinent to the study of the metaphysics of causation—issues that come to the fore in gauging the power of Davidson’s argument in “Actions, Reasons, and Causes.” In section 1 of this chapter I raise these issues with an eye toward achieving the necessary measure of clarity. I will then be in a position, in section 2, to articulate and defend my view of the relation between rational explanations and rational causation.

1. Causation and Rational Explanation 1.1 Synthesizing Causal Concepts In the introduction to their anthology, Causation and Counterfactuals, the editors, John Collins, Ned Hall, and L. A. Paul, raise a basic question: “What sort of project are philosophers of causation engaged in, that consulting intuitions about cases would be an appropriate way to pursue it?”2 This question is especially pressing in light of the fact that most of the contemporary literature on causation (and much of the contents of their volume) consists simply of discussion of potential counterexamples to theories of causation: e.g., cases of overdetermination, preemption (early, late, double, or trumping), omission, prevention (single and double), transitivityfailure, fizzling, and misconnection. The authors then contrast two aims that a philosopher of causation might have, aims that would determine very different projects:

2. John Collins, Ned Hall, and L. A. Paul, eds., Causation and Counterfactuals (Cambridge, MA: MIT Press, 2004), 30.

Rational Explanation and Rational Causation

147

On the one hand, she might, like Lewis, take herself to be providing a good old-fashioned conceptual analysis of causation— a detailed explanation, that is, of how our ordinary concept works. . . . On the other hand, she might view her account as at least partially stipulative— that is, as providing a cleaned up, sanitized version of some causal concept that, though it may not track our ordinary notion of causation precisely, nevertheless can plausibly be argued to serve some useful theoretical purpose. . . . This tolerant methodological perspective [i.e., one that allows for projects with the second aim] gains further support from the observation that causal concepts are used all over the place, both in philosophy and in the sciences; indeed, just sticking to philosophy, it often seems that for any philosophically interesting X, there is at least one “causal theory of X” on the market. It would be hasty to assume that the causal concept or concepts at work in any such theory is just our plain old ordinary one. In short, then, there is good reason to think that there is plenty of work available for those philosophers of causation who take themselves to be in the business of, as it were, “conceptual synthesis,” rather than oldfashioned conceptual analysis. What we wish to emphasize is that even someone interested in “synthesizing” a new and potentially useful causal concept needs to heed these intuitions, else she risks cutting her project free of any firm mooring. More specifically, a reasonable and cautious approach for her to take is to treat intuitions about cases as providing a guide to where interesting causal concepts might be found. Thus, although the account can selectively diverge from these intuitions, provided there are principled reasons for doing so, it should not diverge from them wholesale.3 According to the first option, all of the specific, substantive constraints on a philosophical theory of causation ultimately derive from our ordinary concept. We use technical philosophical resources (e.g., possible world semantics for counterfactual statements) and apply generic theoretical values (e.g., simplicity and elegance). But the correct theory, constructed using these resources and guided by these values, will be the one 3. Ibid.

148

Rational Causation

that answers to our ordinary concept. The test for how well a theory answers to our ordinary concept is how closely its verdicts about specific cases match our intuitions. The second option sees an additional source of data: ser vice to some other theoretical purpose, which might justify a motivated, quasi-stipulative elaboration of an ordinary causal notion. Presumably, the methodological allowance for such an elaboration is predicated on the idea that our ordinary notion of causation is only an inchoate proto-notion, or a cluster of closely connected notions, or perhaps a cluster of inchoate proto-notions. This incompleteness and/or multiplicity at the conceptual level would leave open the possibility that the process of philosophical refinement and elaboration would yield several legitimate concepts of causation, each equally faithful to our ordinary conceptual practice. Indeed, if the word ‘cause’ and its cognates are equivocal—if, i.e., the “very basic principles governing our application of ‘cause’ have been shown to come into conflict”4 —then the best a metaphysician of causation can do is to produce several theories, each one grounded in a distinct sense of ‘cause,’ each perhaps suited to a different theoretical purpose. If this is right, then one source of evidence in favor of a particular conception of causation is that it makes it possible to formulate the best theory of some concept or phenomenon of philosophical interest. Of course, to recognize the relevant theory as causal, the theory’s use of causal concepts must be continuous with (i.e., not “unmoored” from) ordinary practice. A theory of X thus cannot introduce a completely novel conception of causation, but it can refi ne an existing conception, even if the refinement is one that emerges neither from any other theoretical pursuits nor from mere ‘good old-fashioned conceptual analysis.’ My view is just such a refinement.

1.2 Rylean Conceptual Analysis Let us consider the recent history of the debate over the causal status of reason-giving explanations in light of this methodological overview. 4. Ned Hall, “Two Conceptions of Causation,” in Collins, Hall, and Paul, Causation and Counterfactuals, 225–276, 255.

Rational Explanation and Rational Causation

149

“Actions, Reasons, and Causes” was framed as a response to the Rylean tradition in the philosophy of mind, and especially to the Routledge and Kegan Paul series of ‘little red books,’ whose authors included A.  I. Melden, Peter Winch, and Anthony Kenny.5 It is perhaps not too much of a stretch to describe these authors as engaged in ‘good old-fashioned conceptual analysis.’ Certainly Ryle, in The Concept of Mind, took the clarification of the distinction between mental and physical concepts to be the key to dissolving the appearance of a mind-body problem. But Rylean conceptual investigations look very different from those conducted by contemporary philosophers of causation. The latter have typically been in the business of articulating necessary and sufficient conditions for the application of the ordinary word ‘cause.’ Cases are then imagined that test the theory by would-be counterexample. Would we really (or wouldn’t we in fact) use the word ‘cause’ to describe this? Rylean conceptual investigations, by contrast, involve articulating the conceptually distinctive features of rational as opposed to causal explanations of action, features revealed by examining broadly the practice of giving and asking for reasons. Very briefly, here are some examples: Ryle argued that to explain an action by giving a motive is to “subsume it under a propensity or behavior trend,”6 and not to cite a “happening.” This distinction is reflected in the grammatical difference between the sorts of mental terms that figure in rational explanations and the sorts of physical terms that figure in causal explanations. The former are dispositional, “elliptical expressions of general hypothetical propositions of a certain sort”;7 the latter are categorical. To say that a man boasts from vanity, Ryle (infamously) argued, “is to be construed as saying ‘he boasted

5. See, e.g., A. I. Melden, Free Action (London: Routledge and Kegan Paul, 1961); Anthony Kenny, Action, Emotion and Will (London: Routledge and Kegan Paul, 1963); and Peter Winch, The Idea of a Social Science and Its Relation to Philosophy (London: Routledge and Kegan Paul, 1995). See also G. H. Von Wright, Explanation and Understanding (London: Routledge and Kegan Paul, 1971), which came later than Davidson’s essay, but which offers anti-causalist arguments in a similar spirit. Davidson also includes G. E. M. Anscombe, Intention (Cambridge, MA: Harvard University Press 2000), as representative of the Rylean anti-causalist movement. 6. Gilbert Ryle, The Concept of Mind (New York: Barnes and Noble, 1949), 110. 7. Ibid., 85.

150

Rational Causation

on meeting the stranger and his doing so satisfies the law-like proposition that whenever he finds a chance of securing the admiration and envy of others, he does whatever he thinks will produce this admiration and envy.’ ”8 Whereas rational explanations cite behavioral tendencies, causal explanations cite happenings in the here-and-now. The cause, on Ryle’s view, of the window’s breaking is the ball’s having struck it, not the window’s fragility. But what rationally explains action is more like the latter than the former. On the basis of considerations of this sort, Ryle concludes that rational explanations are not causal. Melden argues as follows: the concept of cause is tied to that of an empirical law—a generalization discoverable on the basis of induction from cases of the co-occurrence of independently intelligible event-types. But in the case of rational explanations, the tie between explanans and explanandum is conceptual or logical. This disqualifies the reason-action relation from being causal: “the very notion of a causal sequence logically implies that cause and effect are intelligible without any logically internal relation of the one to the other.”9 Winch held that whereas causal explanations must be understood by way of the generality of an empirical generalization, rational explanations must be understood through the generality of a rule—in particular, the sort of rule that is intelligible only against the background of shared social institutions and ways of life. Rational explanations attribute to the agent a thought of this form: “In view of such and such considerations this will be a reasonable thing to do.” What’s reasonable, in turn, depends on “the accepted standards of reasonable behavior current in his society.”10 Rational and causal explanations are distinct in virtue of their being based on different sorts of principles. Anscombe argued that rational explanations of actions are characterized by a special kind of knowledge possessed by the agent. Both the action and what rationally explains it are, as such, practically known by the agent. Practical knowledge, in turn, is identified in part by the contrast with causal knowledge. Since no cause can be known practically, yet what

8. Ibid., 89. 9. Melden, Free Action, 52. See also Kenny, Action, Emotion and Will, 109. 10. Winch, The Idea of Social Science, 81.

Rational Explanation and Rational Causation

151

rationally explains actions are so known, one cannot rationally explain an action by appeal to causes. Rylean conceptual investigations work by contrasting the conceptual character of two sorts of explanations. The method is not to set before the reader a case in which, say, Poirot believes that the butler did it because everyone else has an alibi, and then exhort the reader to agree that it would be odd-sounding to call everyone else’s having an alibi a cause of Poirot’s believing. It is rather to argue that the concept of a cause is bound up with certain grammatical, logical, or epistemic features of explanations of a certain sort, features that distinguish them from explanations in which we appeal to reasons. Thus, they convey a different sort of understanding than do causal explanations. As we shall see, Davidson’s argument has been so influential in part just because of doubts about whether the grammatical, logical, and epistemic features of an explanation are metaphysically significant, doubts about whether different sorts of understanding need correspond to different sorts of objects of understanding.

1.3 Davidson’s Central Argument One way of replying to Ryle and the other post-Rylean anti-causalists is simply to argue that rational explanations, despite their distinctive features, can nonetheless be analyzed in causal terms. This is, in effect, the strategy of analytical functionalists such as Lewis and Armstrong, but it is not Davidson’s. Davidson agrees with the anti-causalists that rational explanations are conceptually distinctive and hence that there cannot be a reduction of the psychological to the physical. This makes Davidson’s anti-causalist argument more potent, since it does not depend on the success of the analytic project. According to Davidson, the grammatical, logical, and epistemic features of rational explanation that Ryle and his successors sought to highlight in order to shed light on the mind-body problem are, in an important sense, irrelevant to the metaphysics of mind. Rational explanations are conceptually distinctive, on this view, yet they are made true by precisely the sort of causal processes that make true ordinary causal explanations. There are several arguments in “Actions, Reasons, and Causes” directed against the Rylean anti-causalists. One has been especially influential: “Noting that nonteleological causal explanations do not display the element

152

Rational Causation

of justification provided by reasons, some philosophers have concluded that the concept of cause that applies elsewhere cannot apply to the relation between reasons and actions, and that the pattern of justification provides, in the case of reasons, the required explanation.”11 Davidson understands the anti-causalists as saying that the relevant explanationform is justificatory as opposed to causal. He puts the problem with the proposal this way: “But suppose we grant that reasons alone justify actions in the course of explaining them; it does not follow that the explanation is not also— and necessarily— causal.”12 Rather, rational causes are distinguished by their also serving as the agent’s justification. As Davidson understands the anti-causalist, she holds that citing a justification that an agent had for performing an action is by itself a way of explaining an action. This view is hopeless: “But then something essential has certainly been left out, for a person can have a reason for an action, and perform the action, and yet this reason not be the reason why he did it. Central to the relation between a reason and an action it explains is the idea that the agent performed the action because he had the reason.”13 I am not aware that anyone ever held the view that to say what reason someone had to perform an action is by itself to explain her performance of it. Indeed, giving a reason someone had to perform an action is on its face not a kind of explanation at all. Justificatory relations are inert. They make a difference to what happens by way of a person who believes and acts in light of them. To put it another way: action-explanations purport to reveal genuine worldly dependence. Merely justificatory relations cannot by themselves account for such dependence. This is Davidson’s point. It remains to be seen, however, why he thought his opponents held the view he refuted and, more importantly, how his argument came to be treated as a devastating objection to the views that the anti- causalists did hold. If no one ever held that “John had a reason to flip the switch, viz., to illuminate the room; and John fl ipped the switch” captures the explanatory content of “John fl ipped the switch because he wanted to illuminate the

11. Donald Davidson, “Actions, Reasons, and Causes,” in Essays on Actions and Events (Oxford: Clarendon Press, 1980), 9. 12. Ibid. 13. Ibid.

Rational Explanation and Rational Causation

153

room,” why did refuting this equivalence become a watershed in the philosophy of mind and action? Let us consider in slightly more detail the case of Anscombe. She holds that to give someone’s reason for performing an action is to comply with a special sort of explanatory request, a special sense of the question “Why?” Much of Intention is a painstaking elucidation of the character of this sort of explanation through a detailed consideration of various kinds of responses to the relevant why-question: ways of answering, rejecting, and neither answering nor rejecting the question. Ultimately, as I noted above, she ties this distinctive form of explanation to a special kind of reasoning and a special kind of knowledge. Felt desires are part of the causal history of some actions, but insofar as the citing of intentions and desires answers the relevant why-question, it explains the action in a non-causal way, viz., by invoking the agent’s practical reasoning. To give a cause of an action is not, according to Anscombe, to comply with a request for an explanation of the relevant sort. Davidson’s fictitious anti-causalist offers the non-explanatory relation of justification as a substitute for an explanatory relation. But Intention is an attempt directly to elucidate a non-causal sense of ‘because,’ a sense that figures in the distinctive explanations answering the special why-question. In the same vein, when Winch highlights the role played by justification in rational explanation, the point is not that the justificatory relation is itself an explanatory relation but rather that the concept of justification plays a crucial role in elucidating a non-causal form of explanation.14 Davidson himself did as much as anybody to articulate a distinctive, rational form of explanation. Physical and mental explanations, he argued, cannot be tightly connected if each is to “retain allegiance to its proper source of evidence.” He famously describes their differing allegiances this way: It is a feature of physical reality that physical change can be explained by laws that connect it with other changes and conditions physically described. It is a feature of the mental that the attribution of mental phenomenon must be responsible to the background of reasons, beliefs, 14. See Winch, The Idea of Social Science, 81.

154

Rational Causation

and intentions of the individual. . . . [W]hen we use the concept of belief, desire, and the rest, we must stand prepared, as evidence accumulates, to adjust our theory in the light of considerations of overall cogency: the constitutive ideal of rationality partly controls each phase in the evolution of what must be an evolving theory.15 Davidson’s commitment to the distinctive character of rational explanation raises an interpretive puzzle. Davidson chose the inert relation of mere justification to represent the anti-causalists gloss on the rationalexplanatory ‘because.’ But why not view the entire distinctive apparatus of rational explanation—not just the inert relation of justification— as revelatory of the reality corresponding to the ‘because’ of rational explanation? Why not, in other words, view the conceptual character of rational explanation as transparent to the metaphysical structure underlying such explanations? To answer this question, both as it pertains to Davidson and to many of those who have been persuaded by his argument, we must consider how Davidson and others have understood the distinction between causation and explanation. If one sees this distinction roughly as Davidson does, it will seem as if the special features of rational explanation are of no help in saying what it is to act for a reason.

1.4 Davidson on the Relation between Causation and Explanation In “Causal Relations,” Davidson emphasizes the distinction between two kinds of causal statement. First, there are singular causal claims, e.g., “The short circuit caused the fire,” in which “the relation of causality between events can be expressed . . . by an ordinary two-place predicate in an ordinary, extensional, first order language.”16 Second, there are causal explanations, e.g., “The boat sank quickly because it was full of holes,” in which ‘because’ is a non-truth-functional connective that relates statements or facts or propositions. P. F. Strawson, in an explicitly Davidsonian moment, puts the idea this way: 15. Donald Davidson, “Mental Events,” in Essays on Actions and Events, 222–223. 16. Donald Davidson, “Causal Relations,” in Essays on Actions and Events, 161.

Rational Explanation and Rational Causation

155

We sometimes presume, or are said to presume, that causality is a natural relation which holds in the natural world between particular events or circumstances, just as the relation of temporal succession does or that of spatial proximity. We also, and rightly, associate causality with explanation. But if causality is a relation which holds in the natural world, explanation is a different matter. . . . It is an intellectual or rational or intensional relation. It does not hold between things in the natural world, things to which we can assign places and times in nature. It holds between facts or truths.17 This way of describing the causation/explanation distinction has a very innocuous reading, one that does not go as far as Davidson and Strawson want to take it. An explanation, it might be fairly thought, consists of utterances, statements, or propositions. As such, explanations belong to the realm of the representing. But causation belongs to the realm of the represented. Causal explanations are, whereas causal sequences are not, as such representations of something. Explanation thus seems mind-dependent in a way that causation does not. An explanation is good qua explanation to the extent that it can convey the relevant sort of grasp of what is explained to someone who understands it, where the “relevant sort” of grasp is a function, among other things, of what sort of explanation it is: causal, mathematical, ethical, etc. There is no corresponding goodness in causation. I have no quarrel with this innocuous rendition of the distinction. My quarrel is with a less innocuous reading of the passage from Strawson and the distinction it addresses, according to which the point is not just that explanans and explanandum are representational items but that what they pick out—the facts—are not real in the way that items in other categories— e.g., events—are real. The less innocuous interpretation postulates a pair of related distinctions. On the one hand, there are real items, such as events or circumstances, and real connections between them, such as causation, and, on the other hand, there are purely intellectual items, such as facts or truths, and the merely intellectual connections between them, such as explanation. The first distinction (real versus merely intellectual 17. P. F. Strawson, “Causation and Explanation,” in Essays on Davidson: Actions and Events, ed. B. Vermazen and M. Hintikka (Oxford: Clarendon Press, 1985), 115–135, 115.

156

Rational Causation

items) lies in the background of Davidson’s view that the designata of causal explanans and explanandum are not causes and effects, that only singular terms pick out causes and effects. The second distinction (real versus merely intellectual connections) is crucial to his view that the explanatory connection conveyed by an explanation is not necessarily (and perhaps not ever) the underlying real connection. Let us briefly discuss these views. Davidson holds that it is events that are causes and effects. Sentences designate facts. Hence, the expressions flanking the sentential connective ‘because’ do not designate causes or effects. Rather, they designate the fact that, e.g., events of the relevant types occurred. This is the received view in the metaphysics of causation.18 Notice, however, that it goes beyond the more innocuous rendition of the causation-explanation distinction. On the more innocuous view, the point is just that sentences or propositions (of which explanations consist) are not causes and effects. On its less innocuous interpretation, the point is that sentences and propositions do not even designate causes and effects. For Davidson, the source of the illumination provided by a causal explanation need not correspond to the real-world principle that links cause to effect. Davidson holds, for example, that although knowledge of highlevel causal generalizations (e.g., softwoods make poor upholstery frames) may help constitute someone’s understanding of why a pair of events (Billy’s leaping on the chair, the chair’s collapse) stand in the relation of cause and effect, such generalizations are at best evidence for the existence of a strict causal law covering the events in question.19 Such high-level generalizations are not the actual laws that connect cause and effect. I raise the less innocuous interpretation of the causation-explanation distinction because I think it helps to explain widespread obliviousness to the possibility of an easy anti-causalist reply to Davidson’s argument.20 The form of that reply is to argue that the special conceptual features of rational

18. It is propounded in several of the essays in David Lewis, Philosophical Papers, vol. 2 (Oxford: Oxford University Press, 1986). 19. Davidson, “Actions, Reasons, and Causes,” 16. 20. Widespread but not universal: see Wilson, The Intentionality of Human Action, 177, and Dancy, Practical Reality, 163, for versions of this response.

Rational Explanation and Rational Causation

157

as opposed to causal explanation help to define a conception of non-causal explanation that tells us what it is for a man to do something because of his reason. This is, as we saw, the basic Rylean strategy. But under the sway of the less innocuous reading, the features of such explanations that an anticausalist is apt to rely on— e.g., conceptual or epistemic features—belong with the unreal connections that characterize explanation, as opposed to the real connections that characterize causation. Thus the anti-causalist does not seem to have the right material with which to build a conception of a kind of explanation that refutes Davidson’s argument—the argument that only causation can properly account for the real connection between reasons and actions. But the less innocuous interpretation of the causation-explanation distinction is dubious. First, there are good reasons to be suspicious of the size of the gap that it opens between explanation and reality, reasons to be suspicious of just how little Davidson thinks non-fundamental explanations penetrate into reality.21 Second, the plausibility of the less innocuous reading is due at least in part to a focus on causation and causal explanation, to the exclusion of other sorts of explanation. The first point splits into two: one might argue that causal explanations refer directly to causes and effects and not just to facts involving them; and one might argue that someone who understands a causal explanation thereby grasps the real, causal connection that links cause and effect. I am concerned chiefly with the latter, but since I return to the former gap below, it merits provisional discussion. There is a long-standing dispute in the metaphysics of causation about whether causes and effects are events, as Davidson holds, or whether they are facts, as Jonathan Bennett and D.  H. Mellor, among others, hold. If partisans of the fact-view are correct, then causes and effects are what explanantia and explananda of causal explanations designate.22 Partisans of the event-view are in the majority. It is, however, a fractious majority. Kim and Lewis are said to agree with Davidson that events

21. This point is well made in Steward, The Ontology of Mind (especially chapters 5 and 6), to which my discussion here is indebted. 22. One might also follow Steward and allow that causes are not tied to any one metaphysical category, that, e.g., both events and facts are causes. See Steward, The Ontology of Mind, chapter 6.

158

Rational Causation

are true causes and effects, but the agreement is hollow, given that their views on what events are differ so fundamentally from Davidson’s own. More recently, new candidates for the role of causes and effects have received more attention, especially ‘tropes’ and ‘aspects.’ Ultimately, I argue that there are powerful reasons for denying that the relata of rational causation are particulars of any sort, but more stage-setting is required to appreciate these reasons. Let us now turn to the size of the gap Davidson finds between explanatory and causal connections. Prior to doing philosophy, we do not discern logical space between an explanation’s being correct and its being a window onto the reality that makes it correct. Part of what this means is that we do not ordinarily distinguish between the principles that we grasp in order to understand a true explanation and the principles that make it true. Someone who knows nothing about the ‘folk physics’ of solidity and fragility would be in a poor position to appreciate the explanation “the window broke because it was struck by a baseball.” According to Davidson, what separates such a person from someone who possesses the relevant folk-physical knowledge is not that the latter understands the real principles that underlie the explanation but rather that she better understands the evidence in favor of there being some such principles. A man who knows that the window broke because windows are fragile, baseballs are hard, and fragile objects break when forcefully struck with hard objects knows why it broke in the sense that he knows a causal explanation. But he doesn’t know, according to Davidson, why it broke in the sense that he doesn’t know any of the actual principles (for Davidson, the strict laws) that, in reality, link the rock-throwing and the window-breaking. This distinction among senses of ‘knowing why’ does not exist except in philosophy. Ordinarily, to the extent that it can be shown that one doesn’t understand any real causal principles that connect cause and effect, there is pressure to confess that one really doesn’t know why the relevant effect occurred or obtained at all. The oddness of Davidson’s less-than-innocuous way of understanding the causation-explanation distinction does not, of course, show that there is no such distinction. But I do think it requires us to be very cautious about declaring as purely intellectual or unreal the principles grasped by someone who understands a good, non-fundamental causal explanation.

Rational Explanation and Rational Causation

159

My second objection is based on a simple observation: causation is not what makes explanations in general true but rather what makes causal explanations true. One might, for example, attempt to explain a mathematical theorem by presenting several different methods of proving it true. But the reality that makes these proofs correct (assuming there is such a reality) is not obviously a matter of causal connections. Similarly, one can explain the meaning of an utterance by appeal to linguistic conventions, the morality of an action by appeal to ethical principles, or the prudence of a betting strategy by appeal to game-theoretic models. These explanations do not represent causal dependencies between facts or anything else. So it is false that every explanation of the form ‘p because q’ postulates a causal connection, though it may postulate another kind of potentially truthmaking—i.e., real—connection: mathematical, conventional, ethical, gametheoretic, etc. Presumably, the anti-causalist will argue that explanations of actions that appeal to an agent’s reasons postulate not causal connections but a different kind of real connection. And the fact that such a connection may be inextricably bound up with rational agents is no more a reason to regard it as a metaphysical fiction as it would be to so regard a man-made traffic ordinance, which we (most of us, anyway) accept as making-true an explanation of why a certain driver is speeding. There is, of course, the remaining question of whether the anti-causalists ever presented a fully satisfactory account of the real non-causal connection that makes-true rational explanation. But my point is just that one cannot preempt investigation into the possibility of such accounts simply by invoking the distinction between causation and explanation. The most charitable interpretation of Davidson’s central argument against the anti-causalists treats it not as a refutation at all but rather as a challenge, viz., to isolate a real non-causal connection that could make rational action-explanations true. In section 1.3 I argued that no anti-causalist was attempting to make the notion of mere justifi cation by itself do explanatory work. Rather, anticausalists were attempting, in various ways, to use rational connections to formulate a conception of non-causal explanation. So it is no argument against the anti-causalist that mere justification cannot account for the reality behind the rational-explanatory ‘because.’ In this section I have considered how the explanation-causation distinction itself, on a certain interpretation, might motivate the view that rational explanation depends

160

Rational Causation

on causal connections. But the interpretation is dubious. The mere fact that a rational explanation postulates a real (and not merely intellectual) connection between an action and the reason it was performed does not prove it to be causal explanation. The non-causalist thinks that there are real, non-causal connections that help make-true rational explanations of action and that the metaphysical character of the former are revealed by the conceptual character of the latter. Ultimately, however, I think that the insights variously expressed by the anti-causalists are best served by recasting the approach as a form of causalism. There is a plausible rationale for using causal language in speaking of the reality corresponding to rational explanation, but it is in the spirit neither of Davidson’s view nor of the view of many of his even more naturalistically inclined philosophical descendants. This rationale is rooted in a very different conception of the causation-explanation distinction, one articulated by Helen Steward.

1.5 Steward on the Relation between Causation and Explanation Steward, while eschewing a reductive analysis of causation, argues that we can explicate the notion of cause by considering “our purpose in providing specifically causal explanations of events and circumstances, and of producing generalizations on the basis of the specific instances of causality which we encounter.”23 To see this purpose clearly, we look to the sort of understanding of the world that causal explanations provide: “A causal understanding of something or other is an understanding of why the world is as it is in a certain respect—either why it is or why it is not a certain way—which appeals either to how it came about that it is as it is in the first place or to what maintains it as it is.”24 Causal explanations provide us with an understanding that thereby enables us to “bring things about by bringing other things about, or maintain things in a certain way by ensuring that conditions remain a certain way, or prevent things from happening by preventing certain other things from happening.”25 Such explanations, when success-

23. Ibid., 182. 24. Ibid., 183. 25. Ibid.

Rational Explanation and Rational Causation

161

ful, put us in touch with “dependencies which are objective, rooted in reality,” and do not “hold only in virtue of the propensity of human beings to find the mention of these facts explanatorily compelling.”26 I do not intend to do full justice to Steward’s view here, which is rich and in many ways (though, as we shall see, not all) congenial to my own. But three elements of her view come into focus in these passages and are of particular relevance to our discussion. First, she proposes, in effect, that a philosophical account of causation should approach its topic by way of a certain kind of understanding. Causation is an aspect of the world that is revealed to someone by virtue of her understanding an explanation of a certain sort. Second, she says what such explanations explain: “why the world is as it is in a certain respect— either why it is or why it is not a certain way.” Causal understanding is thus what results from grasping a correct answer to a question about why an event did or did not occur or about why certain circumstances do or do not obtain. Answers to such questions—which I’ll call worldly explanations—reveal the dependence of facts and particulars on other facts and particulars. But not every kind of worldly dependence, as I’ll call it, is of the causal sort. Steward’s formulation is, third, tailored in part to rule out various kinds of non-causal answers to questions about “why the world is as it is in a certain respect.” (Again, this is not to suggest she aims for a reductive account of the concept of cause in terms of other, more primitive notions. She explicitly disavows this ambition.) Suppose a man who is shooting a rifle in the air asks, “why is what I’m doing illegal?” One might answer by appealing to the law against discharging firearms in a nature preserve. But this statute does not help to explain why the man’s shooting came about or what kept it going or prevented it from stopping. As Steward’s gloss on the notion of cause predicts, we would not say that the law is a cause of its being illegal. Rather the law is that in virtue of which it is illegal—the law constitutes its illegality. Similarly, a girl’s having no living parents is not a cause of her being an orphan but rather that in which her being an orphan consists. The cause of her being an orphan is whatever explains the death of her parents. Steward gives another example of a constitutive explanation: “My porridge 26. Ibid., 185.

162

Rational Causation

is warm because the particles in it have a mean kinetic energy of value V.”27 Here, too, the explanation does not say how the porridge came to be warm or what kept it warm or what prevented it from cooling or heating. The penal code, the meanings of terms, and the laws of nature are sources of constitutive worldly dependencies. I will argue, in section 2, that although rational causation is a kind of non-constitutive worldly dependence, it is not one that is properly described as answering to the “what brought it about?” or “what maintained it?” questions. But putting aside for the moment whether Steward’s intuitive formulation excludes only constitutive worldly explanation, let us consider the ramifications of an account like Steward’s for the debate between causalists and anti-causalists. A causal explanation is, on this hypothesis, a non-constitutive explanation of why the world is the way it is such that understanding the explanation amounts to knowledge of a real connection between facts. This would make it an analytic truth that the real connection that non-constitutive worldly explanations represent as obtaining between explanans-designatum and explanandum-designatum is causal. Since rational explanation is self-evidently not constitutive, it would follow that an account of rational explanation that disclaims the ambition of being a theory of rational causation is fundamentally confused. Thus, if one thinks of causation roughly the way Steward does, it is tempting to hear Davidson’s central argument in “Actions, Reasons, and Causes” as making a trivial point: an explanation that is neither causal nor constitutive cannot explain someone’s action or belief, since it does not even purport to explain why the world is as it is. But of course this is neither Davidson’s nor the anti-causalist’s conception of causation. For Davidson, as we have seen, successful causal explanations typically shed very little light on genuine causal connections. Only singular causal claims pick out causes and effects, and the truth of such claims consists in the subsumability of those events under strict laws of nature. This is to say that causation, on Davidson’s view, just is efficient causation. Davidson’s argument is thus not that the anti-causalist has somehow lost track of the fact that to give someone’s reason for acting is to explain why a certain event occurred. (Although he does give this im27. Ibid., 180.

Rational Explanation and Rational Causation

163

pression by having his anti-causalist offer justifi cation as a truth-maker for rational explanations.) His argument, as I said above, is best thought of as a challenge to fi nd some relation ‘in the world’ besides efficient causation that could make-true rational explanations. And more sophisticated contemporary defenders of Davidson’s argument hold that no one has yet produced any such relation.28 For the anti-causalists, the claim that rational explanations are causal is a substantive thesis about the particular sort of (non-constitutive) worldly connections that make those explanations true. Causation is thus not, on their view, the name for the real relation underlying all (non-constitutive) worldly explanation. It is, as for Davidson, efficient causation. But the lasting taint on anti-causalism is due at least in part to a failure to see this point clearly, a failure that Winch, in the second edition of The Idea of a Social Science and Its Relation to Philosophy, owns up to. In the preface Winch writes of the first edition: “The result was that I found myself at times denying that human behavior can be understood in causal terms, when I should have been saying our understanding of human behavior is not elucidated by anything like the account given of ‘cause’ by Hume (and Mill).”29 His point is to make clear that in rejecting that rational explanations postulate causes, he was not relinquishing the idea that such explanations postulate real, worldly connections. Rather, he was rejecting the idea that such connections could be understood in terms of natural laws. The real dispute between Davidson and the anti-causalists is whether rational explanations are made true by efficient causal connections or by some other kind of worldly relation. Framed thusly, I take the part of the anti-causalist, defending what I take to be a broadly Anscombian position. However, as Winch ultimately realized, the best way to make this point is to argue that there are other kinds of causation besides efficient causation. My position that there is sui generis rational causation has both a substantive and a terminological dimension. The substantive point is that whereas some (non-constitutive) worldly explanations are made true by efficient causal connections, rational explanations are made true by a dif-

28. See Kieran Setiya, “Reasons and Causes,” European Journal of Philosophy 19 (2011): 129–157. 29. Winch, The Idea of Social Science, xii.

164

Rational Causation

ferent sort of connection. The terminological point is this: because this different sort of connection is just what is revealed to someone in virtue of their understanding an explanation of why a person believed or acted as she did—a non-constitutive explanation of why the world is as it is in a certain respect—rational explanation should be included under the rubric of ‘cause.’ I began this chapter with a discussion of methodology in the study of the metaphysics of causation. Collins, Hall, and Paul argue that there need not be just one concept or kind of causation, that there might be several equally valid conceptions of cause. They claim that one source of evidence in favor of a conception of cause is the role it can play in the development of a philosophical theory of some X, even if a different notion is required for the best philosophical theory of Y. The best theory of rationality, I contend, is one that employs a distinctive causal notion. Meanwhile, Steward has given a plausible overarching conception of the more general category of cause, of which, I argue, efficient and rational causation are two kinds.

1.6 Kinds of Causation It is one thing to point out that Davidson’s argument against the anticausalist is really just a challenge. It is another to meet that challenge: my aim in section 2. In what remains of this section, I make clear what it means to distinguish between kinds of causes in the general sense. We will then be ready, in the next section, to appreciate distinctively rational causation. Davidson is that rare thinker who is hailed as an inspiration by mutually hostile philosophical movements. He is honored by naturalists in the philosophy of mind and action for both his powerful arguments in favor of physicalism and the conceptual ground-clearing that facilitated the naturalistic approach to philosophical psychology. But Davidson’s conception of distinctively rational explanation has also been profoundly influential in an anti-physicalist tradition in the philosophy of mind and action. Philosophers such as John McDowell and Jennifer Hornsby have argued that when one carries this conception to its logical conclusion, one arrives at the idea of an aspect of the spatiotemporal world that is

Rational Explanation and Rational Causation

165

governed not by nomological but rather rational causal principles.30 I follow in this second tradition. The received view—the view that I, following McDowell and Hornsby, dispute—is that the concept of natural law should occupy a central place in a general theory of causation. On Davidson’s view, it is part of the very idea of a causal connection that causally connected events have descriptions under which they can be subsumed under strict laws.31 Natural laws also play a crucial role in most reductive accounts of causation. According to classic regularity accounts, c causes e just in case there is a natural law according to which the occurrence of a c-type event (together with other conditions) entails the occurrence of an e-type event.32 Accounts based on the idea that cause and effect are connected by physical processes typically rely on laws to specify what counts as a process of the relevant sort— e.g., by specifying causal processes in terms of a quantity that is conserved according to physical laws.33 And in counterfactual analyses of causation, natural laws play a crucial, albeit more indirect, role. They are a primary determinant of the metric of similarity between possible worlds, which in turn determines the truth-conditions for the counterfactual statements into which singular causal claims are analyzed. That the most prominent theories of causation orbit around the notion of a natural law is a sign of the widespread, albeit not universal,34 view that the two notions are fundamentally and inextricably linked. To be sure, laws figure differently in 30. See John McDowell, “Functionalism and Anomalous Monism,” in Action and Events: Perspectives on the Philosophy of Donald Davidson, ed. B. McLaughlin and E. LePore (Oxford: Blackwell, 1985); and Hornsby, Simple Mindedness, especially chapter 8. 31. For another non-reductive account of causation that gives natural laws pride of place, see D. M. Armstrong, “Going through the Open Door Again: Counterfactual versus Singularity Theories of Causation,” in Collins, Hall, and Paul, Causation and Counterfactuals, 445–458. See also Nancy Cartwright, “Causation: One Word, Many Things,” Philosophy of Science 71 (2004): 805– 819. 32. David Hume, An Enquiry Concerning Human Understanding (1748), and John Mackie, The Cement of the Universe (Oxford: Clarendon Press, 1974). 33. See Phil Dowe, “Wesley Salmon’s Process Theory of Causality and the Conserved Quantity Theory,” Philosophy of Science 59 (1992): 195–216; and Wesley C. Salmon, “Causality and Explanation: A Reply to Two Critiques,” Philosophy of Science 64 (1997): 461–477. 34. See, e.g., G.  E.  M. Anscombe, “Causality and Determination: An Inaugural Lecture,” (Cambridge: Cambridge University Press, 1971).

166

Rational Causation

these accounts. But they all seem to be motivated by the idea that to understand fully how a cause brings about an effect, one must understand the way in which that specific instance of causation manifests natural laws. It is the nomologically significant properties of the cause that— together with the laws themselves and the nomologically significant background conditions— explain its having the relevant effect.35 I argue, however, that although this may be the right approach to efficient causation, it is the wrong approach to another kind of causation. A distinct kind of causation would be captured by an account in which the laws of nature do not play this central role. It would be an account of causation if it described real, non-constitutive connections that could be appealed to in order to convey an understanding of why the world is as it is in a certain respect. It would be an account of rational causation if the characteristic activities of rational beings played the sort of fundamental role that appeals to laws do in traditional accounts of efficient causation. My positive view, which I spell out in the next section, follows along these lines. My proposal that there is a distinct kind of rational causation is, I have found, easily misunderstood. Some construe it as postulating a supernatural, telekinetic force in the universe, one that would, if it existed, foil any attempt by physicists to produce a theory that was fully successful by their own lights. But to think of rational causation as involving spooky forces is just to give it an efficient-causal interpretation. (This error is an important topic in Chapter 6.) Some see me as butchering ordinary language by misrepresenting how we typically use the word ‘cause’ and its cognates. But my intention is not to match exactly how we ordinarily use the word, though my use is also not ‘unmoored’ from ordinary causal talk. People often, perhaps typically, use ‘cause’ to refer to what I here call ‘efficient causes.’ A non-philosopher is likely to furrow her brow if one says, as I do, such things as “The fact that

35. See also John Hawthorne and Daniel Nolan, “What Would Teleological Causation Be?,” in J. Hawthorne, Metaphysical Essays (Oxford: Oxford University Press, 2006), 265–284, in which the authors invoke laws to distinguish backward efficient causation from teleological causation (or in any case, from what teleological causation would, on their view, be).

Rational Explanation and Rational Causation

167

everyone else has an alibi is the cause of Poirot’s believing that the butler did it.” But philosophers have come to use the expressions ‘cause’ and ‘causal connection’ to signal something more general: the real items and non-constitutive relations that explain why the world is as it is. Reasons are causes in this sense.

2. Rational Causation My theory of rational causation amounts to taking rational explanations of thought and action (as I have characterized them) at face value. Their conceptual structure is, I propose, transparent to their metaphysical structure. Rational explanantia and explananda designate rational causes and effects; and rational causal connections consist in the exercise of inferential theoretical and practical rational abilities. All of the materials necessary to reply to Davidson’s challenge are already implicit in the form of rational explanation. Unsurprisingly, the resultant conception of causation is fundamentally different from what has been proposed by philosophers who study causation using mechanistic examples as a guide.

2.1 Rational Causation as the Manifestation of Rational Abilities To say that S believes that p because q or that S is f -ing because she is y -ing is to explain why the world is the way it is in a certain respect. Specifically, it is to say that S’s believing that p is the result of q or that S’s f -ing is the result of her y -ing. The obtaining of the facts designated by the explananda depends on the obtaining of the facts designated by the explanantia. This is not constitutive dependence, as it is not a claim about what S’s believing that p or S’s f -ing consists in. Rather, it is causal dependence. One mark of the distinctive character of this causal dependence is its self-consciousness. To be my reason for belief or action is for me to be in a position to say authoritatively that it is, yet not on the basis of observation or evidence. Part of what it is to be a rational cause or effect is to be avowable as such by the subject. This expressibility is due to the fact that to believe for a reason is to represent the to-be-believed-ness of the one proposition as conferring to-be-believed-ness on another; and to act for a

168

Rational Causation

reason is to represent the to-be- done-ness of one action as conferring to-be- done-ness on another. I argued in Chapter 2 that this helps us understand why “S believes that p because q” gives the same reason for believing as “S believes that p because S believes that q”; and why “S is f -ing because she is y-ing” gives the same reason for acting as does “S is f -ing because she intends to y.” The rational-causal dependence of S’s believing that p on q and S’s f -ing on her y-ing is inextricable from their dependence on S’s believing that q and S’s intending to y. This is not because these psychological facts are necessary causal intermediaries but rather because the psychological and non-psychological guises of a rational explanation attribute to the subject a single exercise of the relevant rational ability. Rational causation consists in the exercise of these abilities—the ability, in the theoretical case, to believe what is to be believed on the basis of something else that is to be believed; and the ability, in the practical case, to do what is to be done on the basis of something else that is to be done. Since the rational efficacy of q depends upon its being known by S, it follows that rational causation is, in an important sense, normative. If S merely believes that q, then the fact that q is not a rational cause of S’s believing that p; q’s efficacy depends on S’s epistemic success. Thus, one cannot neatly separate ‘purely factual’ questions regarding the causal order from normative questions regarding the quality of S’s grounds for belief. Similarly, the efficacy of S’s y-ing on her f -ing is dependent upon (a) her representing the to-be-done-ness of y-ing culminating in actual y-ing (as opposed to a mere intention to y) and (b) her knowledge of the instrumental connection between f -ing and y-ing. The mind-citing versions of these theoretical and practical explanations are more forgiving. S’s believing that q and her intending to y rationally cause (respectively) her believing that p and her f -ing even when epistemic and practical failures have undermined the corresponding worldly causes. This might raise doubts about the worldly causes. Let’s focus on the theoretical case first. Given that the rational efficacy of a fact is—whereas the rational efficacy of someone’s believing the corresponding proposition is not—norm dependent, and given that causation is not generally considered a normative affair, it is tempting to think that believingfor-a-reason must, at bottom, be a matter of, so to speak, a believing-to-

Rational Explanation and Rational Causation

169

believing causal connection, rather than a fact-to-believing causal connection. The latter, one might think, just is the former, plus something extra. This is the causal version of the hybrid view discussed in Chapter 1. According to that view, a rational world-citing explanation of someone’s belief is made true in part by believing-for-a-reason itself, and in part by something external to it, viz., the fact that S knows that q. I argued, in reply, that it is constitutive of believing-for-a-reason that it is diminished by the absence of knowledge, and thus whether S knows the relevant q is not after all external to her believing-for-a-reason. My reply here follows the same course: the causal connection between S’s believing that p and S’s believing that q consists in S’s being in a certain state. S’s being in the state is the exercise of a specific cognitive ability, viz., the ability to know one thing on the grounds of another. This ability is identified in terms of its successful exercise. As I emphasized in Chapter 1, this is how one identifies abilities in general. We do not identify the ability to walk by citing instances of stumbling. To understand the rational causal connection between S’s believing that p and S’s believing that q, one must thus look to the ability whose successful exercise would be to know that p on the basis of q. And to have done this would amount to q rationally causing S’s believing that p. The point is not that S’s believing that p for reason q consists in a causal tie between q and S’s believing that p, as opposed to one between S’s believing that q and S’s believing that p. The point is rather that the nature of the relevant sort of believing-to-believing causal connection—the nature of the psychological efficacy of the belief that q— can be understood only by seeing it as the manifestation of a theoretical ability whose successful exercise constitutes a fact-to-believing causal connection. Thus I only reject the thesis that rational explanations are psychological on its psychologistic interpretation. The same goes for practical rational causation. Representing y-ing as to be done, when successful, just is y-ing; no matter what, it is intending to y. For this reason, when “S is f -ing because she is y-ing” is false solely due to her failure to be y-ing, she is nonetheless f -ing because she intends to y. Although this shows that rational action-explanation is psychological, the causal connection between actions is more fundamental— its greater fragility notwithstanding—than the connection between the explained action and the intention. The latter can be understood only as

170

Rational Causation

the exercise of an ability identified via its manifestation in the former. There is thus no way to get an independent foothold on the intention-toaction causal tie, a foothold that might then be used to reveal the nature of the action-to-action tie. Another source of doubt about the viability of my notion of rational causation, in addition to its normativity, is what might be called third-personalism about causation. The efficacy of q, on my view, is not separable from its being represented by someone as to be believed. One is then tempted to say that q by itself is not really doing any causal ‘work,’ since its causal power is dependent upon recognition by the believer. S’s believing that q can, by contrast, stand on its own and seems to absorb the efficacy we might be inclined to attribute to the fact that q. The perspective-dependence of the efficacy of q may thereby seem to make the worldly cause somehow secondary to the perspective-independent psychological cause. The appeal of this thought is tied to the following reasoning: both S’s believing that p and S’s believing that q involve, in an obvious way, S’s take on things. It would be absurd to try to eliminate S’s perspective from these facts, since these are just facts about S’s perspective. But the causal connection between the psychological facts is not another fact about S’s perspective but rather a perspective-independent fact about how the psychological facts are objectively related. However plausible sounding this may be, it is wrong. The ‘because’ in “S believes that p because S believes that q” must be understood as the shadow cast by the ‘so’ in “q, so p,” the demonstration by which S can express the underlying state. The rationalcausal tie between S’s believing that p and S’s believing that q is just as much a matter of S’s perspective as the beliefs themselves, as it consists in S’s representing an inferential connection between q and p. It is only because this is so that S is in a position authoritatively to speak about her reasons for belief, yet not on the basis of observation or inference. One will view the causal power of q as secondary if one thinks that rational explanations describe perspective-independent efficacy—but they do not. This point comes into focus if we consider what would otherwise be an odd feature of rational explanation. Steward, in the passage cited above, states that an answer to a causal question “appeals either to how it came about that it is as it is in the fi rst place or to what maintains it as it is.” And this certainly seems true of answers to effi cient-causal questions. Such

Rational Explanation and Rational Causation

171

answers, we could say, cite either change-prompters or change-preventers. Under the heading of change-prompters, I include the causes of an object’s acquiring or losing properties (changing states) and being created or destroyed. Under the heading of change-preventers, I include causes of an object’s retaining a property (remaining in a state) and an object’s being preserved. But a rational cause is neither a change-prompter nor a change-preventer. To understand it as a change-prompter, one would presumably think of it as having caused a change in what the thinker believes. But the rational cause of belief is not what made a believer out of the relevant subject but rather what makes a believer out of him. Poirot may have come to believe that the butler did it on the basis of an anonymous tip, only later learning that everyone else had an alibi. His reason for believing— a rational cause of belief—is still that everyone else has an alibi, even though something else prompted Poirot to believe that the butler did it. To understand a rational cause as a change-preventer is to think of it as what causes the subject not to change what he believes. However, it is false that if S believes that p because q, then S’s belief that q causes her not to change her mind. It is consistent with her believing for that reason that nothing causes her not to change her mind, that S will at the very next moment abandon her belief that p— even despite continuing to believe that q— or retain her belief yet for a different reason. Both the change-prompting and change-preventing models ignore the following basic fact: to give S’s reason for belief is primarily to state how things are now and only derivatively how they were earlier or will be in the future. Conceiving of change-prompting and change-preventing as exhausting the possibilities reflects a failure to see the phenomenon of rational causation as fundamentally first- rather than third-personal. Neither model registers the distinctive character of the tie between S’s believing that p and S’s believing that q. That tie consists in S’s representing p as . To say that S believes that p because she believes that q at a particular time is just to say that she is at that time in a state of representing a certain inferential connection between p and q. Similarly, S’s representing the end-action as to be done does not explain her performance of the means-action by describing what initiated it. A

172

Rational Causation

man may have started swimming the Amazon in order to prove something to his father, but that may no longer be his reason. Perhaps he no longer cares what his father thinks and now has decided to make a name for himself by swimming all of the world’s great rivers. To say that this is his reason for swimming the Amazon does not explain what prompted him to do it. To give someone’s reason for performing the action is also not to say what causes him not to stop. Consistent with the explanation, he may continue to perform the action even as his reasons change, or stop performing it even though he has not abandoned the larger project of which the explained action was a part. Practical causes are neither action-initiators nor action-perpetuators. Just as a subject’s believing a certain proposition can only be a rational cause of her believing another proposition at that very time, an agent’s performing a certain action (or intending to perform it) can only be a rational cause of her performing another action at that very time. This is a consequence of the fact that the ‘because’ in “S is f -ing because S is y-ing” must be understood in terms of the practical-inferential tie that S represents as holding between f -ing and y-ing. Perhaps the most salient difference between the practical and theoretical cases is that since practical causation is teleological, the cause is a whole of which the effect is a part. But can a whole really be a cause of its parts? Many will view this as disqualifying the relation from being genuinely causal. In fact, this is just one of a series of apparently disqualifying features. In the next section I rebut an objection from weirdness, as well as an objection from the opposite point of view— an objection that rational causation is not sufficiently distinctive to count as another kind of causation. Let us return to the debate about whether causes and effects are events or facts or something else. Although I think that Steward mischaracterizes the shape that answers to causal questions must take, I rely on the other key elements of her approach mentioned above: causal questions ask why the world is as it is in a certain respect but not for what the world’s being that way consists in. And causation is just what is revealed to someone by virtue of her understanding answers to those questions. My account of rational causation is simply a consequence of combining this view of the relationship between causation and explanation with the conception of rational explanation detailed in the fi rst two chapters. Un-

Rational Explanation and Rational Causation

173

surprisingly, then, I hold that rational causes and effects are facts, even if (and I take no stand on it here) efficient causal relata are events or perhaps something else. The debate over causal relata, recall, arises in part because efficientcausal statements take two forms: “p because q,” in which ‘because’ is a sentential connective and ‘p’ and ‘q’ pick out facts; and “e caused f,” in which ‘caused’ is an ordinary two-place predicate and ‘e’ and ‘f’ are singular expressions that pick out events. Metaphysicians of causation have tended to choose sides as to which kind of sentence provides the most perspicuous representation of genuine causal relata on the basis of a welter of semantic and metaphysical considerations. Hornsby and Steward argue that, in the case of rational causal explanations, there are no mental particulars to play the role of the referents in singular causal claims.36 And they conclude, in part on this basis, that rational causation must be understood in terms of causal connections between facts. In the next chapter, I give an account of the distinction between particulars and non-particulars that supports the Hornsby-Stewart line. According to that account, states do not stand to any particulars as object-sorts stand to their instances and event-types stand to their tokens. Beliefs simpliciter are general things and thus not what people have in mind when they envision beliefs ‘themselves’ as causes. A singularist approach to rational explanations, I shall argue, faces the fundamental obstacle of there being no singularities to play the relevant causal roles. I argue, furthermore, that events-in-progress are also not particulars, and so to explain an action by the fact than another was under way is also not to explain by appeal to a particular. But even before we delve into the metaphysics of particularity, there is already a good reason for taking rational causes—at least in the theoretical case—to be non-particular. The engine of theoretical rational causation is inference from and to the to-be-believed-ness of propositions. To say of S that she believes that p because q is to say that she represents p as to be believed on the basis of the to-be-believed-ness of q. Since rational-causation is fundamentally firstpersonal and since theoretical reasoners have in view (would-be) facts, facts are rational causes. The event-partisans in the debate over the nature of the efficient-causal relata contend that the role of ‘becausal’ explanatory 36. Hornsby, Simple Mindedness, and Steward, The Ontology of Mind.

174

Rational Causation

statements (as opposed to singular causal statements) is not to represent directly the efficient causal chain but to convey information about it. The nomological character of efficient causation (however it is ultimately understood) may be neutral in the debate about what efficient causal relata are. But in the case of rational causal explanations, the fundamentally inferential character of rational causation favors factual relata, which have a propositional structure. Given this difference, whatever theoretical pressure is exerted by the goal of having a uniform account of the causal relata across causation of all kinds has a counterweight in the value of having accounts that are responsive to the most basic differences between these kinds.

2.2 Objections Causation is generally thought to be a relation between ‘distinct existences,’ traditionally understood as events. But if one event is a part of another, they are not distinct existences and hence cannot, it seems, stand in the relation of cause and effect. This is why counterfactual definitions (or attempts at definitions) of causal dependence are intended to apply only in cases in which the candidates are distinct events. Suppose a ball falls from A to B to C. Its fall from A to B might be counterfactually dependent on its fall from A to C—if the closest possible world in which it doesn’t fall to C is one in which it doesn’t fall at all. In that case, if it had not fallen from A to C, it would not have fallen from A to B. But this dependence cannot be causal: the ball’s falling (or fall) from A to C cannot have caused its falling (or fall) from A to B. For the latter is a part of the former. The relation of part-hood rather than the relation of causation must explain the counterfactual dependence of the A-to-B fall on the A-to-C fall. We can bring my thesis out from under the shadow of paradox by observing a crucial disanalogy between this sort of case and that of practical causation. The mere relation of part-hood between the A-to-B fall and the A-to-C fall does not supply the material for an explanation of the former in terms of the latter. If someone asks, “Why did the ball fall from A to B?,” it is no answer to say “because it was falling from A to C,” counterfactual dependence notwithstanding. But “because he was walking from 100th Street to Tenth Street” is a perfectly good answer to the question

Rational Explanation and Rational Causation

175

“why did he walk from 100th Street to Fiftieth Street?” The answer conveys that he walked to Fiftieth Street in order to get to Tenth Street. And it is precisely on account of this explanatory connection that it makes sense to speak of causation in this case, yet not the other. It is easy to lose track of this point if one has in mind as a would-be teleological cause a completed event, with all of its parts already fully determinate. It would be plain nonsense to speak of the completed event that is Dara’s reviewing of the restaurant as causing (qua completed event) the occurrence of one of its parts. The part, after all, must already have occurred in order for the whole to be complete. But the point of adopting the means is that it (ideally) brings the end closer to completion. The teleological explainer is an unfolding end-event (more carefully, the fact that an event of the end-sort is unfolding), and so the verb-aspect of the explanans is imperfective.37 A teleological explanation is a characterization of the process by which an event becomes fully determinate. The fact that the means-event is occurring is a result of the fact that the end-event is unfolding, where ‘result’ has a teleological-causal sense. The oxymoronic feel to the thesis that wholes cause their parts derives from a failure to think through what it means in the context of specifically teleological causation. But my reply here brings to the fore other seemingly problematic aspects of rational causation. In a mind-citing explanation, the postulated cause and effect are necessarily indexed to the same time. To say that S believed that p because S believed that q (or that S was f -ing because she intended to y), in the rational sense, is to say that there was some one time at which S believed that p and believed that q (or at which she was f -ing and intended to y) and that these two facts concerning S at that time stand in a rational causal relation. But most reject the idea of simultaneous causation.38 There is also a temporal peculiarity associated with world-citing rational explanations: there are no constraints on the temporal connections between q and S’s believing that p except for those imposed by the requirement that S know that q. If S can know that dinosaurs roamed the earth millions of years ago, then S can (now) believe that the world is more than 10,000 years old because dinosaurs roamed the earth millions of years ago. 37. Or at least it is imperfective in the fundamental case— see Chapter 2, note 29. 38. See Chapter 1, note 24.

176

Rational Causation

And if S can know that it will rain tomorrow, then she can (now) believe that she needn’t water her plants today because it will rain tomorrow. The idea that causes can be later than their effects—or, to put it in more factappropriate language, the idea that causes can be indexed to later times than their effects—is even more controversial than simultaneous causation.39 But I have shown that what ties a rational cause to its effect is an agent’s representing the obtaining of an inferential connection. Since the premise of the inference must be believed or intended at the very same time as the conclusion, the relevant believings and intendings are simultaneous with their effects. And since (or, in any case, if) knowledge of the future is possible, a future fact can (or could) be that in light of which a person believes or acts now. And this just is for that fact to be rationally efficacious. It might seem as if my reply to the general objection misses the point. The point, some would argue, is that my own analysis of rational explanation proves that it does not after all postulate a causal connection between the explanandum-fact and the explanans-fact (or between events in their respective neighborhoods). I say, on the contrary, that this is evidence that rational causation is a different kind of causation. How should this dispute be settled? If rational causation really is a species of causation, then there must be something that unites, under a single metaphysical category, a window’s shattering because it was struck by a ball and a man’s shattering a window because he was terrorizing the neighborhood. On my view, what unites them is at least the following: the becausal tie is a real (and not merely intellectual) relation that could be revealed to someone in virtue of understanding a non-constitutive explanation of why the world is as it is in a certain respect. Although nothing that I say here precludes a more illuminating approach to the issue—more on this shortly—there is at least some reason to expect a multiplicity of kinds of cause. To see this point, it will be helpful to consider an analogy. One might think that there is something that unites objects under one heading and events under another. Traditionally, objects and events are thought to have different relations to time. Objects persist; events unfold and so have begin39. Backward causation is thought to give rise to paradoxes. See D.  H. Mellor, “Causation and the Direction of Time,” Erkenntnis 35 (1991): 191–203.

Rational Explanation and Rational Causation

177

nings, middles, and ends. One can consistently suppose that something like this is right and yet deny that there is just one kind of object and one kind of event. As I will put it in Chapter 5, one can deny that there is just one principle of identity for objects in general and another for events in general. One might hold that there are different kinds of objects and that these differences in kind are constituted by their differing principles of identity. Because those principles apply to persisting things, they are all, despite their differences, principles of identity for objects. One might hold the analogous view of events, too. Indeed, these are just the views I defend in the next chapter. The point here is not that we should deploy the concept of a principle of identity in a new way to distinguish between different kinds of causes. Rather, the idea is that when we are dealing with metaphysical categories at a very high level of abstraction (such as object, event, cause, and property), we should expect to find deep differences in the natures of their instances. Because the causal status of reasons and reasons-explanation is controversial, instances of rational explanation are not part of the diet of examples on which the debates about the nature of causation subsist. If these debates result in consensus on principles constitutive of causation— e.g., “no part-whole causation” or “no simultaneous causation”—it does not follow that any relation between reasons and beliefs or actions that fails to satisfy these principles is not causal. It may be that those principles are just constitutive of the kind of causation exhibited in the examples, which deliberately exclude rational explanations. Failure to appreciate this point leads philosophers to bring their favored theory of causation to the question of the efficacy of reasons: Do beliefs and desires stand in the already decided-upon relation to actions? Or how can we fit beliefs and desires into the settled-upon framework? My contention is that rational explanations already embody a distinctive, substantive conception of causation, one that renders otiose and misguided any attempt to impose upon the rational domain an exogenous causal theory. Doing so produces unsurprising failures of fit and gives rise to an endless series of pseudo-problems and ever more resourceful (or desperate, depending on your point of view) attempts to avert philosophical disaster—in the form, for example, of epiphenomenalism or eliminativism. There is also this familiar conundrum: A man is a guest at a party hosted by a woman toward whom he has an undeclared antipathy. While

178

Rational Causation

drinking a glass of merlot on a white shag rug, he fantasizes about ‘accidentally’ staining the rug by spilling the wine. He is overcome with excitement and consequently spills his wine on the rug. And so a man is caused to spill his wine, one might say, by his desire to stain the rug and his belief that by spilling the wine, he will stain the rug. But he did not spill the wine for any reason. The supposed puzzle concerns how to characterize the efficacy of the belief-desire pairs that serve as an agent’s reason in such a manner as to exclude this sort of case.40 On the view I have presented here, there is no such puzzle. The causal chain between the agent’s believing and his action is special in virtue of being a special kind of causation.41 The problem cases are excluded because they involve the wrong kind of causation. And so the effects—the so-called ‘actions,’ which are really no such thing—are also not of the right sort. Spilling the wine is not an exercise of a rational ability at all. Suppose that there were some overwhelming lexicographical reason to withhold the label ‘causal’ from the relation I have described. Still, it would do nothing to show that rational explanations should not be construed as postulating it. It would not show, in particular, that the truthmakers for rational explanations are efficient-causal relations. Rational explanation would have turned out to be non-causal—not in the exciting sense of being in some business other than saying why things in the world are the way they are, but in the boring sense of not falling under the scope of the same English word. So much the worse for the word. So far I have considered an objection on the grounds of difference—the relation of rational causation is too weird to be grouped with the uncontroversial instances of causation. But it might also seem that I am vulnerable to an objection from sameness. Where is the argument that rational causation, as I understand it, cannot in the end be assimilated to one of the going theories of causation? Perhaps some version of the counterfactual theory could do the trick. After all, rational explanations do imply

40. Davidson poses this problem for his own view and considers various replies in his Essays on Actions and Events. See also Alfred Mele, Motivation and Agency (Oxford: Oxford University Press, 2003). 41. Compare Sebastian Rödl, Self- Consciousness (Cambridge, MA: Harvard University Press, 2007), 46–49.

Rational Explanation and Rational Causation

179

roughly the same sorts of counterfactuals as do explanations of other sorts. All of these statements seem relatively safe conclusions to draw from our examples: If Dara had not intended (or desired) to review the restaurant, she would not have tasted all the desserts. If Dara had not in fact been reviewing the restaurant, she would not have tasted all the desserts. If Poirot had not believed that everyone else had an alibi, he would not have believed that the butler did it. If it were not true that everyone else had an alibi, Poirot would not have believed that the butler did it. These conclusions are only relatively safe, since it is always possible that the believer or agent might have reached the same conclusion on different grounds. But of course possibilities of roughly the same sort equally pose an obstacle for counterfactual analyses of instances of non-rational causation. But even if to give someone’s reason for action or belief is to commit oneself to counterfactuals of the above sort, and even if this commitment is constitutive of such statements being causal, one is equally thereby committing oneself to a certain sort of explanation of why such counterfactuals hold. To say that Poirot believed that the butler did it because everyone else had an alibi specifies the character of the causal connection between cause and effect, just as does the statement that a sudden gust of wind knocked over a paper cup. In the case of rational explanations, what explains the truth of the entailed counterfactuals are the actualizations of rational abilities. In the case of the wind and the cup, it is natural laws that do so. There is thus no prima facie inconsistency between the thesis that all causal claims can be analyzed counterfactually and the idea that there are different kinds of cause. The differences in kinds of cause would then simply correspond to differences in what explains the truth of the relevant counterfactuals.42 Since the notion of a disposition—more specifically that of a rational ability—plays a central role in my account, it is also worth considering the contrast between rational and efficient causation from the perspective of a dispositional account of causation. According to what Alexander Bird calls “The Simple Dispositional Analysis of Causation,” “A causes B when

42. But it will be asked: don’t exercises of rational abilities supervene ultimately on the distribution of physical properties, thereby bringing rational causation into the ambit of natural laws? I discuss the relevance of supervenience to my view in Chapter 6.

180

Rational Causation

A is the stimulus of some disposition and B is the corresponding manifestation.”43 A fragile vase’s being dropped is a cause of its breaking, since fragility links stimuli of the dropping-sort with breaking, the manifestation of fragility. On this view (and any more sophisticated variant in the same spirit), the role of a disposition is that of a condition for the causal connection between events of one sort and events of another. But this is not the role of a disposition in my account. To begin with, rational abilities are not associated with stimuli at all. To think of someone as able to do handstands is not to think that he would do handstands if an event of a certain type were to occur. Part of what distinguishes abilities from other dispositions is that we do not think of their manifestations as automatic, given the right external circumstances. Their exercise is typically at the discretion of their bearers. There is some temptation to view choosing to f as the internal stimulus that activates the ability to f, but the temptation is based chiefly on an antecedent theoretical investment in the success of analyses of this kind for dispositions across the board. I do not intend, in any case, to argue against the choice-as-internal-stimulus idea. I will assume that it is false in what follows, that nothing plays the role of stimulus in the case of rational abilities. But this assumption is not integral to identifying the contrast between efficient and rational causes. It is also worth noting, however, that (as discussed in Chapter 1) rational abilities include theoretical as well as practical abilities, and exercises of the former are not (or not always) volitional matters. A man may be able to remember all of the license plates that he sees over the course of a day, but this feat of memory may not be performable at will (perhaps his trying interferes with the ability), nor does its accomplishment amount to the performance of an action. There should be no temptation at all to analyze such an ability as something that a thinking subject intentionally does in response to a volitional stimulus. If nothing plays the role of stimulus to rational abilities, then either the Simple Analysis fails for certain kinds of causation or causation plays no role in the manifestation of rational abilities. Unsurprisingly, I reject the latter. No stimulus stands to “S is f -ing because she is y-ing” or “S be43. Alexander Bird, “Antidotes for Dispositional Essentialism,” in The Metaphysics of Powers, ed. A. Marmodoro (New York: Routledge, 2010), 160–168, 161.

Rational Explanation and Rational Causation

181

lieves that p because q” as vase-dropping does to vase-breaking. But S’s f -ing and her believing p—both of which individually manifest S’s agentive and reflective abilities— are caused by the facts cited in these explanations. The cited facts, however, do not have the efficacy of stimuli that engage a disposition. Rather, the entire causal transactions themselves are manifestations of rational abilities. “S is f -ing because she is y-ing” gives the form of a central manifestation of practical rationality. “S believes that p because q” gives the form of a central manifestation of theoretical rationality. Rational efficacy, unlike efficient efficacy, is internal to the manifestation of an underlying disposition. It might be objected here, both to the simple dispositional account and to my own view, that dispositions are causes of their own exercise. This would seem to fit with a central thesis of my discussion of dispositions from Chapter 2: citing a disposition gives a thin explanation of its manifestations. And an advocate of a dispositional analysis of causation need not say otherwise. Bird suggests that it is an advantage of this sort of analysis that it preserves the ordinary distinction between cause and condition: the stimulus being the former, the disposition being the latter. And I tend to think that he is right about ordinary English. Be that as it may, even if a rational ability is a cause of its acts in the thin sense, it is not the rational cause of its acts. People rarely if ever act for the sake of or believe in light of their rational abilities to act and believe for reasons. Still, we might say that rational abilities are rational causal conditions in the sense that they are conditions for the obtaining of rational causal connections. At the end of Chapter 2 I identified two questions that my account of rational causation would have to answer: (1) What is the causal role of those facts and ends that are an agent’s reason for acting, if they do not operate via the efficacy of internal, psychological items? (2) How should we understand the causal role of psychological states in the performance of action if not as internal items, efficiently causing observable behavior? These questions were pressing insofar as they highlighted the absence of a fullyfledged metaphysical alternative to Psychologism. My answers to these questions should now be clear. Rational causation consists in the agent’s representing an inferential connection between actions. The rational efficacy of the world and the mind are two aspects of her representing the

182

Rational Causation

premise-action as to be done. No rational ‘force’ is thereby propagated through space. There are no intermediaries (psychological or otherwise) through which the rational influence of the world must travel in order to impact what an agent does. And there is nothing spatially inner about the psychological causes of action. The above questions embody misconceptions about rationality that derive from a failure to recognize and take ontologically seriously the distinctive character of rational explanations.

5 Events and States

A central ambition of phi losophers of mind is to show both that and how  the mind matters to the way the world is. Events and states are considered vehicles of the mind’s mattering, in the sense that the ambition is thought likely to be achieved by establishing the efficacy of mental events and states. The orthodox, naturalistic presumption is that establishing their efficacy will amount to revealing the intimate relation that they stand in to physical events and states: identity, realization, or supervenience of some strong variety. The horizontal project of tracking the mental’s influence is thereby joined, via these metaphysical categories, to the vertical project of situating the mental in the hierarchy of being. But this strategy is problematic both in its basic physicalist conception and in the way its execution relies on various misrepresentations of the event-state distinction, misrepresentations that muddy the issues on which the legitimacy and feasibility of naturalistic projects turn. Physicalism is my topic in the next chapter; here I focus on events and states.

184

Rational Causation

My motive for discussing these categories in detail is not, however, solely critical. I have so far made free use of categorical terms, especially ‘event,’ ‘state,’ and ‘fact.’ I have argued that beliefs and actions based on reasons are caused by worldly facts: by an agent’s being in certain mental states and, in the case of action, by the fact that certain events are unfolding. But the substance of these accounts— especially the account of reasons for acting— depends crucially on a rather idiosyncratic understanding of events, one that consequently requires a scrupulous defense. I rely especially on the idea that events can either be in progress or complete, a distinction that is familiar enough to those who think philosophically about the language of events—particularly those who follow in the RyleVendler-Kenny tradition—yet has played virtually no role in the metaphysical literature on events. The distinction is crucial, I argue, not just for understanding the way rational causation works but also for recognizing the distinctive kind of particulars that completed events are and, further, for realizing that that events-in-progress are not particulars at all. Ultimately, I argue in section 1 that part of being a particular—whether an object or an event—is being an instance of a universal that possesses a principle of identity. The object-sortal goat and the event-type tennis match determine conditions under which an object o or event e is, respectively, the same goat or tennis match as object p or event f. Philosophers of mind have largely ignored the way such principles bear on the question of physicalism, but in fact they are central to understanding the relation between mental and physical particulars, as I show in the next chapter. Furthermore, universals that do not possess principles of identity do not have par ticular instances in the sense that object-sortals and event-types do. So, for example, there are no reds, as there are goats (even red goats) and tennis matches. States, I argue in section 2, do not have tokens and thus, like eventsin-progress, are not particulars.1 There is thus no question of whether, say, beliefs are type-distinct but token-identical to physical states.

1. This thesis has been defended previously in Jennifer Hornsby, Simple Mindedness (Cambridge, MA: Harvard University Press, 1997), and Helen Steward, The Ontology of Mind (Oxford: Oxford University Press, 2000). Nonetheless, it is, to say the least, highly controversial—it is hard to crack a book on the philosophy of mind or action that does not assume at some point that there are particulars that fall into the category of state.

Events and States

185

My foray into categorical metaphysics, then, serves two functions. First, it provides a necessary backdrop for the positive claims I have made in previous chapters concerning the nature of rational causation. Second, it yields the materials essential for the anti-physicalist argument to follow in Chapter 6.

1. Objects, Events, and Sortals Objects and events are different in crucial respects, but they have this much in common. They are particulars. What exactly does their particularity come to? Three common characteristics are salient. First, objects and events are countable. We can intelligibly ask how many jelly beans there are in a jar or how many times Jones has crossed the street. Second, they exist in space and time, albeit, as we shall see, rather differently. We can intelligibly ask where the Stanley Cup is currently located and how long it has been there, and also where Sarah ran the marathon and how long it took her. Because, third, objects persist 2 and events take time,3 questions can intelligibly be asked about whether, e.g., the saguaro in your living room today is the same cactus as used to sit in your garden last spring, and whether, e.g., Rob and Hal are playing the same game of chess this morning as they were playing last night. In this section, I make some progress toward explaining these commonalities by appeal to a fourth: objects and events fall into sortals (as they are called in speaking of objects) or types (as they are called in speaking of events). I discuss this idea first in relation to objects (section 1.1), then events (section 1.2).

1.1 Object-Sortals What does it mean to say that objects belong to sortals? In order to answer this question, it will help to distinguish between classes of properties 2. I assume here that objects are the three-dimensional enduring things we ordinarily take them to be. This view is, to be sure, controversial. For a congenial defense of the framework that I take for granted here, see Amie Thomasson, Ordinary Objects (New York: Oxford University Press, 2007). 3. There are instantaneous events, too. I will bring these into our discussion in section 2.2. Until then I focus on events that have duration.

186

Rational Causation

and to make parallel distinctions between classes of general terms. Strawson distinguishes between material or stuff (e.g., gold, water), sortals, (e.g., chair, dog), and states (e.g., red, angry).4 Here I am concerned with the distinction between sortals and states. We can get a feel for this difference by observing that sortals answer what-questions. If I say that I have something that I want you to see, the question “What?” asked in response often has a straightforward answer— for example, “a rocking chair” or “a beagle.” If I respond to the whatquestion by saying “a red thing” or “something heavy,” I characterize the object, but I do not answer the what-question. To give these responses is to avoid answering the question. Being red or heavy are states of things, not sorts of things. We can explain the distinguishing characteristic of sortals by appeal to the notion of a principle of identity. The principle of identity for a property determines the conditions under which a particular instantiating the property persists. As Anil Gupta puts it, “the principle of identity for ‘river’ is the rule in virtue of which an object at a time (and a world) is the same river as an object at another time (or world).”5 The river we now call ‘the Mississippi’ is the same river that Mark Twain called ‘the Mississippi’ in virtue of such facts as the following: the earlier river runs through roughly the same places, is fed from roughly the same sources, empties into the same ocean as the current river, has done so continuously since the time of Huckleberry Finn, and so on. To make a sortal-predication of an object is to attribute to the object a property that determines a principle of identity for its instances. To say that the sortal cat determines the identity conditions for its instances is to say that it follows from something’s being a cat that it can persist through specific changes but not others. It does not, however, follow from something’s being red that it can persist through specific changes but not others. The quality red determines the conditions under which an object is red, but not the conditions under which an object at a

4. P. F. Strawson, “Par tic u lar and General,” Proceedings of the Aristotelian Society 54 (1954): 233–260. Strawson, however, calls sortals ‘substances’ and states ‘qualities.’ 5. Anil Gupta, The Logic of Common Nouns (New Haven, CT: Yale University Press, 1980), 2.

Events and States

187

time is the same red thing as an object at another time. Following Dummett (albeit loosely), we can put this by saying that states have no principles of identity, but only principles of instantiation. The principle of instantiation for a property determines whether an object at a time instantiates the property. The difference between sortals and states is thus that the former have principles of identity, whereas the latter have principles of instantiation.6 I hold, then, that objects instantiate sortals and that to instantiate a sortal is at least in part for there to be a principle of identity that determines the conditions under which the object persists. Objects, however, do not instantiate principles of identity as such, but rather only insofar as they are par ticu lar sorts of objects. Here is how Wiggins puts the idea: “If a is the same as b, then it must also hold that a is the same something as b,”7 where something is a quantifier ranging over determinate sortals. Or, if a is the same as b, there must be an answer to the ‘same what?’ question. To say that object is not a true sortal is to say that it is not a proper answer to the ‘same what?’ question. In saying that a is an object, we do not say what a is. Wiggins thus distinguishes ‘dummy sortals,’ such as object and thing from genuine sortals such as dog and table. Terms for dummy sortals share the grammar of terms for true sortals (e.g., they are modified by articles and quantifiers), but are not associated with a principle of identity.8 6. Dummett makes this point in slightly different terms. He says that someone who grasps the sense of a ‘substantive general term’ (a term for a sortal) must know both the associated criterion of application and the associated criterion of identity; whereas someone who grasps the sense of an ‘adjectival general term’ must know only the associated criterion of application, since there is no associated criterion of identity. See Michael Dummett, Frege: Philosophy of Language (Cambridge, MA: Harvard University Press, 1973), 547–548. Whereas Dummett speaks of criteria, I (following Gupta) speak of principles. A criterion of identity is (at least in Dummett’s usage) primarily epistemic in character. A principle of identity, as Gupta puts it, is “the metaphysical counterpart of such an epistemic rule” (Gupta, The Logic of Common Nouns, 2). Since I am concerned with metaphysical issues and not with epistemological ones, I have recast Dummett’s point. Also, I am not committed to the idea that there are principles of instantiation (the metaphysical counterparts of criteria of application) for sortals. 7. David Wiggins, Sameness and Substance (Oxford: Basil Blackwell, 1980), 47. 8. Dummett makes a parallel point when he says that expressions such as ‘. . . is the same red thing as . . .’ have “no univocal sense” (Dummett, Frege, 548).

188

Rational Causation

Identity statements, on this view, are sortal-dependent, an idea that figures importantly in the argument in Chapter 6.9 Since it is in virtue of some principle of identity that object a at a time is identical to object b at some other time, and since object does not supply one, a determinate sortal is required.10 The most compelling argument in favor of the view that object is a dummy sortal is the one that supports the view that constitution is not identity.11 That argument, stripped of its bells and whistles, is this: the collection of molecules of which a goat is composed and the goat itself have different properties—the collection, for example, might have no future while the goat does.12 These differences highlight differences in the sorts of object that goats and collections of molecules are. What matters to the survival of a collection of molecules is different from what matters to the survival of a goat, i.e., the principle of identity for collections of molecules is different from the principle of identity for goats. Simply being an object does not settle an object’s identity conditions. This kind of argument thus points toward the conclusion that there are no principles of identity for objects as such, and thus that object is not a true sortal.13

9. I would not endorse a stronger and more controversial thesis, what Wiggins calls the Sortal Relativity of Identity. According to this stronger thesis, there can be sortals F and G, such that x is the same F as y but not the same G as y. See P. T. Geach, “Identity,” Review of Metaphysics 21 (1967): 3–12; Nicholas Griffi n, “Ayers on Relative Identity,” Canadian Journal of Philosophy 6 (1976): 579–594; and Harold Noonan, Objects and Identity (Boston: Kluwer Academic Publishers, 1980). 10. For an argument that identity does not require sortals at all, see Michael Ayers, “Individuals without Sortals,” Canadian Journal of Philosophy 4 (1974): 113–148, and Michael Ayers, “Is “Physical Object” a Sortal Concept? A Reply to Xu,” Mind and Language 12 (1997): 393–405. For reply, see Griffi n, “Ayers on Relative Identity.” 11. For defenses of this view of constitution, see David Wiggins, “On Being in the Same Place at the Same Time,” Philosophical Review 77 (1968): 90– 95; Mark Johnston, “Constitution Is Not Identity,” Mind 101 (1992): 89–105; and Lynne Rudder Baker, “Why Constitution Is Not Identity,” Journal of Philosophy 94 (1997): 599– 621. 12. A sortalist need not take this view of constitution. For one who doesn’t, see Edmund Runggaldier, “Sortal Continuity of Material Things,” Erkentnis 48 (1998): 359–369. 13. For an argument that object is a true sortal, see Fei Xu, “From Lot’s Wife to a Pillar of Salt: Evidence that Physical Object Is a Sortal Concept,” Mind and Language 12 (1997): 365–392, with help from Eli Hirsch, The Concept of Identity (Oxford: Oxford

Events and States

189

The idea that object cannot supply a principle of identity for particular objects is plainly related to Frege’s observation that object does not supply a principle of individuation.14 He puts the point this way: “Only a concept which isolates what falls under it in a definite manner, which does not permit any arbitrary division of it into parts, can be a unit relative to a finite Number.”15 Object does permit arbitrary division into parts: no matter how one divides a cat, each resultant part will itself be an object. And so there is something fundamentally amiss about the question “how many objects are on the mat?” The sortal cat does not, in this sense, permit arbitrary division into parts. Very few ways of dividing a cat will result in a multitude of cats. Before moving on to our discussion of events, let us settle on some terminology for different types of sortal. If a sortal is such that the same object cannot at one time instantiate it and at another time not instantiate it, I call it (following Wiggins) a substance sortal. Lion, chair, and painting are substance sortals.16 Besides substance sortals, there are also phase

University Press, 1982). For a trenchant reply, see David Wiggins, “Sortal Concepts: A Reply to Xu,” Mind and Language 12 (1997): 413–421. 14. The connection between identity and individuation is, however, complex. For discussion, see Hector Neri Castañeda, “Individuation and Non-Identity: A New Look,” American Philosophical Quarterly 12 (1975): 131–140, and Elias Savellos, “Criteria of Identity and the Individuation of Natural-Kind Events,” Philosophy and Phenomenological Research 12 (1992): 807– 831. 15. Gottlob Frege, The Foundations of Arithmetic, trans. J. L. Austin (Oxford: Basil Blackwell, 1974), 66. Though Frege typically uses examples of objects that fit with our use of the term here (according to which there is a contrast with ‘event’), it is worth noting that Frege’s purely formal conception of ‘object’ would cover (token) events as well. 16. It is very natural to think of substance sortals as essential properties of the particulars that subsume them. This is how Wiggins thinks of them. (See Wiggins, Sameness and Substance, 122.) But, as Penelope Mackie has argued, one can consistently hold both of the following: First, it’s true that “an individual cannot change its principle of identity over time, nor can it have two different principles of identity simultaneously.” Second, it’s false that “if an individual x has a principle of identity P, then x could not have existed without having P.” (Penelope Mackie, “Sortal Concepts and Essential Properties,” Philosophical Quarterly 44 [1994]: 311–333, 321. I have altered the passages slightly. Where she speaks of principles of individuation, I speak of principles of identity. This change is of merely terminological significance.) I leave it as an open question here whether object and event sortals are essential to the par tic u lar objects

190

Rational Causation

sortals. Syntactically, names for phase sortals can either be simple expressions (e.g., boy, soldier) or complex expressions made up of modified sortalterms (e.g., skinny beagle, new car). The difference between substance and phase sortals is this: a boy can survive without being the boy he was; a skinny beagle can survive as a fat beagle. But a man cannot survive without being the man he was; a car cannot survive as a grasshopper. Thus, from the fact that an object instantiates a phase sortal at a given time and that it exists at a later time, it does not follow that the object will instantiate that phase sortal at the later time. As we shall see, there are no phase sortals for events— a clue to the fundamentally differing natures of objects and events.

1.2 Event-Types The notion of an event-type is, at least when clear-headedly applied, analogous to that of an object-sortal. And the thesis that events necessarily belong to types can be supported along the very same lines as the thesis defended in section 1.1. Consider by way of evidence, fi rst, that one can, just as in the case of objects, distinguish between two sorts of predications of events. On the one hand, one can say of an event that it was a crossing of Columbus Avenue or the making of a loaf of bread. On the other, we can say of the crossing that it was brisk or hazardous or drunken; we can say of the making of the loaf that it was inept or masterful. In the fi rst sort of case, we say what an event was; in the second sort of case, we do not. The fi rst sort of predication subsumes a par tic u lar or token event under an event-type. I call predications of the second sort quality-predications of events. My interest in event-qualities here does not extend beyond this: qualities of events are not associated with principles of identity. What makes an event a brisk or hazardous and what makes an event b inept or masterful does not determine whether a is the same or different from c, or

and events they subsume. What is crucial to my point here is just that an individual instantiating a substance sortal in a par tic u lar world instantiates it in every possible development of that world.

Events and States

191

whether b is the same or different from d.17 This is what the event-types crossing of Columbus Avenue and making a loaf of bread do. To say that events fall into types is thus to say that every event must instantiate a type-property P that determines the conditions under which an event e is the same P as event f.18 I motivated the sortalist view of objects by focusing on the fact that different sorts of objects (goat, collection of goat-composing molecules) have different principles of identity. I motivate the corresponding view of events by showing that there are no principles of identity for events as such. According to this view, event and happening are dummy-types, whereas crossing of Columbus Avenue and making of a loaf of bread are genuine types. To show this, however, one must overcome skepticism about the very idea that the notion of a principle of identity gets a grip in the case of an event. After all, the principle of identity associated with a particular object determines (among other things) the conditions under which it persists. And arguments in favor of the sortalist view of objects depend on the differing conditions under which different sorts of objects persist. But there are many who would argue that the notion of persistence does not so much as get a grip on events, that they are not the right sorts of things to persist. To say that objects but not events persist is to say in part that objects but not events are fully present over any portion of the time during which they exist. Thus, if Jones crossed Columbus Avenue yesterday, the entire walk was not present as he crossed. For the entire walk consists in part of the beginning and in part of the end of the walk. And the end of the walk is not present at the beginning of the walk. To say that events do 17. This is not to deny a relationship between qualities and types: what it is for e to be brisk might well depend on what type of event e is. Nonetheless, what makes e brisk does not determine e’s type, minimally, because there might be another type for which being brisk comes to the same thing. 18. The view that events must belong to types is inconsistent with the thesis that events are tropes. That thesis, as typically understood, involves the idea that an event is the bare instantiation of a property— any property, including qualities, which are the standard examples. For a defense of this view of events, see Jonathan Bennett, Events and Their Names (Indianapolis: Hackett Publishing Company, 1988). On the view of events sketched here, every event must instantiate an event-type. There can be no event that is the bare instance of any other sort of property.

192

Rational Causation

not persist is thus to elaborate on the notion that events have temporal parts. Consider a related point. Fred Dretske argues that events do not, strictly speaking, move.19 Movement requires something to be at one location and then at another. If a car is first in the garage and later in the swamp, then it will be correct still later to say that it was in the garage and also correct to say that it was in the swamp. But in the case of a race that began on land but fi nished in water, it is correct to say neither that it took place on land nor that it took place in water. Rather, it took place partly on land and partly on water. (Some of the parts of the race took place in water; some of its other parts took place on land.) The contestants moved from land to water; the race did not move; it is not the sort of thing that moves. An event does not move because in considering the location of the whole event, we must include the locations of all of its parts.20 The whole race does not exist first on land, later in water. For an event is complete only when it is over: until Jones has finished crossing Columbus Avenue, it would be incorrect for him to say “I’ve walked across Columbus Avenue” (if he has never crossed it before). There is no token event of the relevant type until after the event is over. Hence, objects but not events are fully present over the entire time during which they exist. An event is never fully present in this sense. Either it has not yet occurred, or it is over. Because of this difference, there is no distinction between substance and phase event-types.

19. Fred Dretske, “Can Events Move?,” Mind 76 (1967): 479–492. See also P. M. S. Hacker, “Events and Objects in Space and Time,” Mind 91 (1982): 1–19. 20. It might be conceded that although the race has not moved from land to water in the case above, we could easily construct a case where this description would fit. Say, after about half of the contestants have moved from land to water, subsequent contestants are diverted to a different path of the same distance through the water (on account of sharks). Here, an opponent of Dretske might argue, the race really has moved. I would propose an alternative description of this scenario: the path of the race has been moved, i.e., the ground slated to be covered by the quickest contestants is not the ground slated to be covered by the others. The race, understood as an event, did not move. The whole race could not have moved, since its path was altered prior to its being over, i.e., prior to there being a whole race. I do not contend that my redescription here constitutes an argument, only that it suggests a way of capturing the intuitive appeal of the alleged counterexample without giving up on Dretske’s claim.

Events and States

193

It might be thought otherwise. A football game might be close for three quarters but turn into a rout in the fourth. There is a close game and later a drubbing—the game continues but not as close. There is a raucous, outdoor party (a ‘rager’). It begins to rain; people move indoors; many leave. Now there is a tame, indoor party (a ‘snoozer’)—the party continues, but indoors, and not as raucous. Should we think of the rager and the snoozer as we think of Lincoln the boy and Lincoln the man? No. For the phases of events are the parts of the events, whereas the phases of objects are not the parts of objects. And that is why the concepts we use to refer to the phases of objects pick out the objects themselves, whereas the concepts we use to refer to the phases of events pick out their parts. So whereas the adolescent is the person, the fi rst inning is not the game. The party in the above example is neither a rager nor a snoozer. For if a certain party p is a rager and p is identical to q, then q must also be a rager. If p begins loud and raucous and later becomes quiet and dull, then although a rager appeared to be in the offi ng, none materialized. The whole party was not fi rst a rager, then a snoozer. For the whole party (considered as an event, with a temporal beginning, middle, and end) did not yet exist when it was raging. Thus the whole party does not change from rager to snoozer. Events, properly speaking, do not change. We can still make sense of a sentence that appears to ascribe change to an event, e.g., “the party used to be a rager, but now it’s a snoozer.” It means that the first portion of the party was a rager, and the remaining portion a snoozer. But now we are no longer classifying the same event under two types. ‘Snoozer’ and ‘rager’ are thus not phase-types but rather contradictory substance-types. Either they refer to parts of the party, or one or both misdescribes the whole party. Thus, “the boy became a man” has to be analyzed rather differently than “the rager became a snoozer.” In the first case, a particular changes, first exemplifying one, then another phase-sortal. In the second case, a particular is divided into two parts that exemplify contradictory substance-types. Much of what is challenging in the last few paragraphs relates to the peculiar temporality of events, a feature of event-ontology that has been overlooked by many who write on the topic. I say much more about the temporal characteristics that distinguish events from objects in the next

194

Rational Causation

section. But for present purposes, this further elaboration is not necessary. For I am in the midst of defending a proposed analogy between objects and events—that they both fall into sortals or types—from the perceived threat posed by their very different underlying natures. My point is that these profound differences provide no reason to deny that events must be associated with properties (viz., event-types) that provide principles of identity. For nothing in the difference between objects and events just sketched obviates the need for a principle of identity. In the case of both objects and events, a sortal or type is a property in virtue of which it is correct to say that a particular object or event m at one time is the same as a particular object or event n at another time; the principle of identity associated with a particular sort or type is that in virtue of which an object or event has the unity it does over time—however different diachronic objectunity may be from diachronic event-unity. The same kinds of considerations that support the sortal-dependence of object-identity also support the type-dependence of event-identity. As we saw above, the view that objects as such do not have a principle of identity (and thus that object is not itself a proper sortal) is supported by the fact that different sorts of objects are associated with divergent principles of identity. What one sort of object (say, a collection of molecules) can survive (say, the arbitrary rearrangement of its parts), another sort of object (say, a painting) cannot. If it is hard to see that there are analogous considerations with respect to event, it is in part because they are so obvious. In the object case, there is a pair of sortal-concepts (e.g., ‘goat,’ ‘collection of goat-composing molecules’) that pick out objects occupying the same place at the same time. The fact that they occupy the same place at the same time makes it tempting to identify them, a temptation that can be resisted by considering their differing temporal and counterfactual properties. The possibility of such a divergence becomes clear when we reflect on their divergent principles of identity, typically through the use of thought experiments. If anything, the need for principles of identity in the case of events should, on reflection, be even more apparent than in the case of objects. For in the latter case, there remains a strong temptation to identify objects that reside at the same place at the same time. But there is very little

Events and States

195

temptation to do so in the former case.21 Consider, for example, the relationship between the second set of a tennis match— say, the men’s singles finals at the 2008 U.S. Open— and the match itself. The set was occurring over some portion of the time during which the match as a whole was occurring, and they were occurring at the same place, in whatever sense events can be said to occur at places. Just as in the case of objects, the explanation of their non-identity involves a difference in their principles of identity. Their different principles of identity determine the following facts: the second set is a part of the entire match. The second set begins with the first game after the end of the first set and ends when one player has won six games and his opponent has won four games or fewer. If his opponent has won five games the player must win seven games, and if his opponent has also won six games there will be a tiebreaker. The match begins with the first game of the first set and ends when one player has won three sets.22 Further support for the view that there is no principle of identity for events as such can be found in the generally accepted view that there are no principles of individuation for events as such. The point is analogous to the one I raised in the discussion of objects. There is a determinate answer to

21. As far as I know, only E. J. Lemmon, “Comments,” in The Logic of Decision and Action, ed. N. Rescher (Pittsburgh: University of Pittsburgh Press, 1967), 96–103, argues for the view that two events cannot occupy the same place at the same time. For a convincing reply, see Donald Davidson, “The Individuation of Events,” in Essays on Actions and Events (Oxford: Clarendon Press, 1980), 178–179. 22. Someone might object here that two people (e.g., J.  J. Thompson and Anscombe), neither of whom could be said to grasp the ordinary concept of a murder better than the other, might disagree about the temporal boundaries of a murder and thus disagree about the principle of identity for murders. The concept of a murder, it will be said, is inherently murky. One might thus conclude that there is no determinate principle of identity for murder and perhaps for many other event-types as well. This might well be true; but if it is, it only strengthens the analogy with the principles of identity associated with different kinds of object. For one can analogously argue that an ordinary grasp of such concepts as ship, caterpillar, and person does not furnish one with defi nite answers to every question that might arise over the survival of a par tic u lar ship, caterpillar, or person. To argue that such murkiness undermines principles of identity for object-sortals and event-types across the board would, in effect, be to argue that there are no facts about the identity of objects and events. I take this to be an unacceptable result.

196

Rational Causation

the question “how many?” asked in relation to events only where a genuine event-type is in play. Thus “how many firings of Coach Knight took place in 2000?” has a determinate answer, but “how many events occurred in 2000?” does not. The following point plays an important role in the argument in Chapter 6: according to my view of objects and events, there are two kinds of grounds for rejecting an identity-statement between objects or events a and b. First, a and b might fail to share a sortal or type and thus have different identity-conditions. Second, a and b might share a sortal or type but might be distinct according to that sortal’s or type’s principle of identity. So far I have argued that there is a connection between particularity and belonging to a kind that determines principles of identity and individuation for its instances. This leads to the following conclusion regarding states and events-in-progress: nothing in either category is a particular.

2. States and Events-in-Progress Events and objects are particulars par excellence. In this section, I make the case that states are not particulars by showing that they lack the features of events and objects that plausibly explain their claim to particularity. Specifically, I argue (in section 2.1) that states do not have principles of identity and individuation and that only universals with such principles have instances or tokens.23 I then extend this argument (in section 2.2) to events-in-progress, which, we have seen, play a crucial role in the rational etiology of action.

2.1 States I begin (in section 2.1.1) by reviewing a way of distinguishing states and events most clearly articulated by Anthony Galton in The Logic of Aspect.24 23. As we shall see in section 2.2, one kind of partic u lar calls for a slight revision of this statement. Principles of identity are necessary for any kind of partic u lar that lasts. Thus instantaneous events need and have no principles of identity. They do, however, have principles of individuation. States, as we shall see, are not instantaneous events. Furthermore, they lack principles of individuation. 24. See Anthony Galton, The Logic of Aspect (Oxford: Clarendon Press, 1984), chapter 2. Broadly construed, this way of distinguishing between states and events has

Events and States

197

It will emerge from this discussion that whereas events are unitary, states are dissective. I continue (in section 2.1.2) with a discussion of a metaphysically salient difference between states and objects: the persistence of an object is a matter of the continued presence of a particular over time; the persistence of a state is a matter of a particular continuing to exemplify a universal. There are two further reasons for rejecting the very idea of particular states: (section 2.1.3) states but neither events nor objects are negatable; and (section 2.1.4) event-nominalizations are count-quantified, whereas state-nominalizations are mass-quantified. Each of these points is ultimately tied to the fundamental reason why the type-token distinction has no application to states: states lack principles of identity and individuation. 2.1.1 States Are Dissective, Not Unitary According to one recent and plausible way of thinking, event-talk arises because there is a distinction between perfective and imperfective verb aspect.25 A verb-predication with perfective aspect has a sense of completion; a verb-predication with imperfective aspect does not. Consider the verb-phrase ‘to cross the street.’ We can say both

been discussed in Gilbert Ryle, The Concept of Mind (New York: Barnes and Noble, 1949); Anthony Kenny, Action, Emotion and Will (London: Routledge and Kegan Paul, 1963); Zeno Vendler, Linguistics and Philosophy (Ithaca, NY: Cornell University Press, 1967); Alexander Mourelatos, “Events, Processes, and States,” Linguistics and Philosophy 2 (1978): 415–434; David Dowty, Word Meaning and Montague Grammar (Boston: Reidel, 1979); Emmon Bach, “On Time, Tense, and Aspect: An Essay in English Metaphysics,” in Radical Pragmatics, ed. Peter Cole (New York: Academic Press, 1981), 63– 81; Barry Taylor, Modes of Occurrence: Verbs, Adverbs, and Events (Oxford: Blackwell, 1985); M. J. Cresswell, “Why Objects Exist but Events Occur,” Studia Logica 45 (1986): 371–375; and Terence Parsons, Events in the Semantics of English (Cambridge, MA: MIT Press, 1990). For an illuminating discussion of its roots in Aristotle, see Terence Penner, “Verbs and the Identity of Actions: A Philosophical Exercise in the Interpretation of Aristotle,” in Ryle: A Collection of Critical Essays, ed. O. Wood and G. Pitcher (Garden City, NY: Doubleday, Anchor Books, 1970), 393–460. 25. My discussion follows Galton, The Logic of Aspect. For some other important discussions of verb-aspect, see Mourelatos, “Events, Processes, and States”; Dowty, Word Meaning; and Michael Bennett, “Of Tense and Aspect: One Analysis,” in Syntax and Semantics, vol. 14, ed. P. Tedeschi and A. Zaenen (New York: Academic Press, 1981), 13–30; and Parsons, Events in the Semantics of English.

198

Rational Causation

(A) Jones crossed Columbus Avenue and (B) Jones was crossing Columbus Avenue. In (A), ‘to cross the street’ has perfective aspect (there is a completed streetcrossing registered); in (B), ‘to cross Columbus Avenue’ has imperfective verb aspect (there is no completed street-crossing registered). (A) does not follow from (B), for Jones might have been trampled fatally by a bull before he finished his crossing. Events are the kinds of things that finish and, as such, are the kinds of things that can remain unfinished. Wherever there are events that take time, we can always distinguish between perfective and imperfective verb aspect. Now in the case of states, no such distinction arises. Let us try to formulate a parallel distinction: (C) Jones believed that the end was nigh. (D) Jones was believing that the end was nigh. Here (C) and (D) do not have the same relationship as (A) and (B). For (B) does not entail (A). But it’s not clear how to interpret (D) so that it does not entail (C). Could Jones have been believing that the end was nigh yet never have finished believing it? Even if we could somehow make sense of this form of words, it would turn out that for Jones to have believed this, in the sense of having fi nished believing it, will be something quite different than for Jones to have believed it in the ordinary sense. For ‘having believed,’ in the ordinary sense, does not involve having fi nished doing or finished undergoing anything—believing is not the kind of thing that finishes.26 26. Also, as Galton persuasively argues, state-predications, such as “Jones believed that p” (in the ordinary sense), have a present-tense version: “Jones believes that p.” But event-predications such as “Jones crossed the street” have no present-tense sense. “Jones is crossing the street” is not the present-tense version of “Jones crossed the street;” rather, it is the present-tense version of “Jones was crossing the street.” And although “Jones crosses the street” has a use, it is not used to say about the present

Events and States

199

This is the important point: although we can use what are ordinarily stative verb-phrases to pick out events (in such a way that there is an analogue to the relationship between (A) and (B)), in so doing we change what it is that we are talking about. We are no longer speaking of states of belief but rather of certain (still obscure) events, to which we have extended the use of the word ‘believe.’ We are now talking perhaps about a certain contemplative episode in Jones’s mental life. But the connection between this episode and his beliefs (in the ordinary sense) is tenuous. As will become clear, it is of the utmost importance in considering how correctly to apply categories such as ‘state’ and ‘event’ that one not change the subject. Here, changing the subject amounts to changing the semantics of belief-talk so as to make it conform to the model of event-talk. But this doesn’t show that beliefs are events, only that we can use the term ‘belief’ to pick out events. In the case of event-predications, then, there is a distinction between perfective and imperfective verb-aspect; in the case of state-predications, there is not. This formal distinction has a material corollary. States are dissective; events are not. Galton puts the point in roughly the following way: whenever a state obtains for a stretch of time, it also obtains for every substretch of that time. Even though an event has phases, these phases are not events of the same type as the whole.27 If Jones believes that p throughout interval i, then at any time during i, it will be true that Jones believed that p. But if Jones crossed the street during a thirty-second period, it is not true until after this period is over that Jones crossed the street. Events take time; states do not. And this is why (A) does not follow from (B) in the case above, and also why (C) does follow from (D)— so long as ‘believing’ is stative. But this way of distinguishing between states and events also provides reason for thinking that event-types have tokens, whereas states do not. Since events are unitary, they are countable. They have beginnings and ends; and so it makes sense to ask: “how many?” We can ask: “How many times did Jones cross Columbus Avenue during September 2008?” Because states are dissective, we ask not “how many?” but “for how long?”

what “Jones crossed the street says about the past.” See Galton, The Logic of Aspect, 11–16. See also Chapter 1, note 7. 27. Galton, The Logic of Aspect, 26.

200

Rational Causation

To the question “for how long has Jones believed that Nixon was a genius?,” we might give an answer like “for two years.” Along the same lines, we ask, “for how long has the Jell-O been liquid?” and not “how many times has the Jell-O been liquid?” The applicability of the “how many?” question is tied to the countability of that to which it is applied. But countability and particularity go hand in hand. If “how many?” questions cannot be asked of states, then states are not particulars. It might be countered that we can meaningfully ask how many times the Jell-O has been liquid. Similarly, we can ask: “How many times has your hair been blond?”; “How many times have you weighed over 200 pounds?”; “How many times have you wanted to punch that guy in the nose?”; “How many times have you thought that your grandmother needs glasses?” But every such case falls into one of two categories. Either it is a case in which what is being counted is not really the state in question or it is one in which a word that typically is used to refer to a state instead refers to an event. I discuss each sort of case in turn. The Jell-O, hair, and weight examples are instances of the first sort of case. In these cases, what is being counted is not being Jell-O, having blond hair, or weighing over 200 pounds. It is rather gaining or losing (or perhaps gaining and losing) the relevant state. But I do not deny that one can count how many times the Jell-O has become (and then ceased to be) solid, or how many times Harlow dyed (and then dyed back or let grow out) her hair, or how many times Zero reached (and then fell below) 200 pounds. For these are events, not states. It might be countered that enumerating such events just is how (supposed) token states are counted. But someone who replies in this way must now explain why it is that event-types, such as dying one’s hair or becoming solid, supply their own principles of identity and individuation to their instances, whereas supposedly analogous state-types, such as having blond hair or being solid, only supply principles of identity and individuation via some associated event-type. Here, again, is what it means to say that event-types have principles of identity: a specific event-type E determines the conditions under which it is correct to say that an event e1 at one time is the same E as event e2 at another time. This is what it means to say that event-types have principles of individuation: a specific event-type E determines the conditions under

Events and States

201

which it is correct to say that the number of instances of E that have occurred (within indicated spatiotemporal parameters) is X. We implicitly or explicitly rely on such principles when we try to figure out whether the solidification of the Jell- O that occurred today (an event) is the same as the solidification of the Jell-O that occurred yesterday, or just how many times the Jell-O has solidified. States do not have such principles; and this is why a defender of state-tokens is forced to look to event-types to settle questions regarding the counting of supposed state-tokens. We ask how many times the Jell-O has been solid. To answer, we count events: how many times it has become solid. But this serves only to undermine the case for the intelligibility of the idea of a token state. Furthermore, the question, “These two times at which the Jell-O is solid, both of which obtain during a stretch of unbroken solidity, do they count as two states or only one?” seems to have no principled answer. I expand on this point in section 2.1.2. States do not have principles of identity and individuation. Events (and objects) do. And because states do not have such principles, they are not particulars. Let us return to the second sort of case and these examples: wanting to punch someone in the nose and thinking that one’s grandmother needs glasses. If asked how many times I’ve wanted to punch that guy in the nose, I count certain episodes in my conscious life—for example, how many times I have derived pleasure from envisioning the prospect. If asked how many times I have thought that my grandmother needs glasses, I count how many times I have said it to myself, or how many times I’ve taken note of it, or how many times it has occurred to me, etc. But these kinds of mental episodes are not the topic of statements like “I want to punch that guy in the nose” and “I think that my grandmother needs glasses.” I might want to punch someone in the nose because I have been paid to. If asked how many times I have wanted to, I may simply not find any instance of the sort of mental event-type that the question is asking me to count. ‘Wanting’ ordinarily figures in state-predications; but the question “how many times have you wanted to . . . ?” asks about events. Similarly, when we speak of thinking that p, we are typically thinking of a state, one that need not manifest itself in any conscious episode of inner-speaking. But if asked to count how many times I have thought that p, this is precisely the sort of episode I look to. Both words are ambiguous between state and event readings.

202

Rational Causation

In the case of event-predications but not state-predications, there is a distinction between perfective and imperfective verb-aspect. This fact is significant because it helps us to see the sort of particulars that events are. They are the sort of particulars that have beginnings, middles, and ends. And it is this fact about them that accounts for their specific form of countability.28 My claim is not that the particularity of every particular reveals itself in a distinction between predications with perfective and imperfective verb-aspect. But this distinction in the case of event-predication points toward the characteristic of types that explains their having tokens. Universals, such as object-sortals and event-types, that are instantiated, as opposed to merely exemplified, have principles of identity and individuation. 2.1.2 Two Kinds of Persistence One cannot make a case for treating states as particulars by assimilating them to events. Can the particularity of states be established by assimilating them to objects? Few would argue that states are objects. First, the concepts of part-hood and composition have no univocal application to objects and states. Second, objects and states do not have the same relation to space. Objects compete for space with other objects. With the exception of the objects it composes or is composed by, an object cannot be at the same place and time as another object. But states do not compete for space with other states. We can put this point by noting that objects but not states take up space. Though these differences may well shed light on the non-particularity of states, my focus is on a different difference, one that I explain by way of a similarity: objects and states can both be said to persist. But what this comes to in each sort of case is fundamentally different. We can begin to

28. It might be argued that not all events are countable. One could ask, “how many times has the universe expanded?” Perhaps the answers “just once” and “more than once” seem equally good. But expanding, like running, is not an event—rather, it is a process. To specify something countable, the verb-phrase must provide a telos: e.g., Jones ran a mile, Arthur crossed the street, the Colts defeated the Bears, etc. Once such a telos is provided, our initial question can be answered. If the question asks how many times the universe has expanded amount a, then if it hasn’t yet expanded amount a, the answer is “no times,” and if it has already expanded amount a, then the answer will be “one or more times.” We will discuss processes and teloi in more detail in section 2.2.

Events and States

203

appreciate this point by first considering the case of events, in which the notion of persistence fails to gain a foothold. Persistence requires something fully present at one time and still fully present at a later time. But, as I have argued, events (at least non-instantaneous events) are not fully present as they occur. The whole walk across the street does not persist from the beginning of the walk to the end of the walk; for the walk across the street is not whole until the street has been crossed. Both objects and states are, however, fully present when they exist or obtain, and so we can talk of persistence in both sorts of case. But because existing and obtaining are distinct, the sorts of presence and persistence that characterize objects and states are also distinct. Here, on my view, is the difference. In the case of a state, e.g., Jones’s believing that Nixon was a genius, the presence of the belief is Jones’s exemplifying a universal; and its persistence is his continuing to do so. In the case of an object, e.g., Jones himself, presence is just the existence of a particular at a time; and its persistence is its continued existence over time. Let us consider the opposing view—that the persistence of a state is the persistence of a particular, viz., a token that stands to a state-universal as an event stands to a type of event. Let us say that Jones believes at time t that Nixon was a genius. Call this mental state-token n, and its type N. Now if n is an object-like particular, then there must be something that separates a case where n is replaced by another mental state token m also of type N at time t + 1, and a case where n persists through t + 1. But our ordinary grasp of mental states gives us no basis to distinguish between these scenarios. (Nor is there any science of states to help settle the matter.) Thus it is difficult even to understand a person who says “Yes, Jones has always believed that Nixon was a genius and in that sense Jones still believes what he used to; but I wonder whether his current belief is numerically identical with his former belief.” The discombobulating effect of this question stands in stark contrast with the corresponding, user-friendly question regarding objects. For example: “I know you used to have kidneys; and you still have kidneys; but are those the very same kidneys with which you were born?” The point is not that there would be no phenomenological difference between a scenario in which a token state was replaced by another and a scenario in which a token state persisted. The perplexing character of the question regarding the numerical identity of Jones’s belief casts doubt on

204

Rational Causation

whether there is any prima facie reason for thinking we are even entertaining genuine possibilities when we attempt to contrast a token belief that Nixon was a genius persisting and such a token being replaced by another, type-identical token. If we know what a kidney is, then we know the difference between someone who has the same kidneys she was born with and someone who has new kidneys. Knowing what a belief is provides no corresponding understanding of the allegedly intelligible analogous contrast. Again, I have an explanation for this difference: states are universals that have neither principles of identity nor principles of individuation. There are a few objections that might be raised here. Someone might have the following thought: “Perhaps the question regarding the numerical identity of a person’s belief over time is odd. But we can ask about the numerical identity of the beliefs of several people. When Jones goes out for dinner with his friends, we can ask ‘how many instances of the belief that Nixon was a genius are present?’ ” But we answer this bizarre question, if at all, by counting the number of people present who believe Nixon was a genius. People are, of course, countable. The answerability of this question provides no reason for thinking that there are individual belief-states. On the view I’ve outlined, there are several people, each of whom are in the same state: that of believing Nixon was a genius. The example provides no reason for thinking that the states themselves, as opposed to the people, are countable particulars. If states had no principles of identity, it might be objected, then there would be no answer to the question “what is the difference between the state of being red and the state of being blue?” Yet there plainly is an answer. Similarly, there must be principles of individuation for states, it might be thought, since the question “how many colors are there in the American flag?” plainly has an answer. But the difference between red and blue is not a difference in principles of identity but rather a difference in principles of instantiation—a difference in the principles in virtue of which an object at a time instantiates the state of being red or being blue. And the countability of colors was never in question here. Principles of individuation provide individuation-conditions for particulars. Colors are universals. Talk of numerical identity in conjunction with the obtaining of a state is strained. Here is why: such talk has no application to states, since they lack principles of identity and individuation.

Events and States

205

2.1.3 States Are Negatable There are two further reasons for denying that states are particulars of any kind. First, in the case of objects and events, their absence is not itself an object or event. The absence of a cupcake is not itself a particular on par with a cupcake. There is a sense in which there are non-cupcakes in the world, viz., all of the objects that are not cupcakes. But the absence of a cupcake is not itself an object. Thus we cannot ask the same questions about cupcake-absences as we can about cupcakes. “How many cupcakes are there in the box?” makes sense. Barring a very creative interpretation “how many cupcake-absences are there in the box?” does not. We can make the same point about events. Take the case of an eventtype, say crossing of Columbus Avenue. A token of this event-type is a particular crossing of Columbus Avenue. But the absence of such an event is not a particular event. If it were, then the question “how many times has Jones not crossed Columbus Avenue” would make sense. But this question does not make sense. How would one go about counting these supposed particulars? Say, over the course of a five-minute interval, Jones crossed Columbus Avenue twice. Now, how many times over the interval did he not cross it? There are many Columbus-Avenue-crossings that he might have made during this interval but didn’t. He might have started a few moments earlier or later than he did; he might have crossed while sleeping, walking on his hands, rolling, zigzagging, staggering, or skipping. These are not the crossings he made. Are we to say that while he crossed the street twice, he also didn’t cross the street many more times? As in the case above, there is a sense in which there are non-crossings, viz., all of the events that are not crossings. But the non- occurrence of a crossing is not itself a par tic u lar. There are many non-actual crossings. But there are no actual non-crossings—or rather, ‘non-crossing’ does not pick out a particular. So, as above, ‘how many?’ questions make sense as asked about particular events but not as asked about their absences. If states were particulars, then one might expect to find the same asymmetry between presence and absence. But this is not what one fi nds. States are negatable.29 Consider the following pair of sentences, the first of which

29. Galton, The Logic of Aspect, 26–27.

206

Rational Causation

reports the presence of a state, the second of which reports the absence of a state: (E) The table is red. (F) The table is not red. Unlike in the case of objects and events, any question that we can ask with respect to (E), we can intelligibly (if a bit awkwardly) ask of (F). Was the table red/not red in 1998? For how long has the table been red/not been red? Is the table still red/not red? It does not matter whether we are speaking of a state or the absence of a state. This suggests that the absence of a state is, if not itself a state, in any case ontologically on par with a state. The absence of objects and events are not object-like and event-like; but the absence of a state, it seems, is state-like. It might be objected here that state-absences, such as being not-red, are not genuinely states at all. The reality of these ‘negative properties,’ as they are sometimes called, is controversial. And if there are no such properties, then I cannot argue from a disanalogy between negative- states on the one hand and event-non- occurrences and object-absences on the other. But my point here does not depend on the reality of negative properties. It relies only on the fact that the very same questions that can be  intelligibly asked and answered regarding the obtaining of a state can also be asked and answered regarding the failure to obtain of a state. Perhaps this is because the failure to obtain of a state is itself a state, but I am not committed to this view. Either way, there is a striking disanalogy with events and objects, where the very same questions that can be intelligibly asked and answered regarding the presence of an object or the occurrence of an event cannot intelligibly be asked and answered regarding the absence of an object or non- occurrence of an event. I contend that the best explanation of this discrepancy is that objects and events are particulars, whereas states are not. As above, the explanation exploits the notion of a principle of identity: particulars that persist through or take up time (objects and events, respectively) must be instances of kinds or types that determine principles of identity for these

Events and States

207

instances.30 In the case of both a goat and a tennis match, there are principles of identity associated with goats and tennis matches that determine the conditions under which goat g or tennis match t at one time is the same goat or tennis match as goat h or tennis match u at some other time. The difference between a collection of goat-composing molecules and g, or between a set (in the tennis sense) of t and t itself, can best be captured by appealing to their different principles of identity. But goat-absences are not associated with a principle of identity, for there is no such kind of thing as a not-a-goat. Similarly, there is no such kind of thing as a not-a-tennis-match. Among the non-goats there are numbers, giraffes, collections of molecules, sets, galaxies, continents, events, and works of art, none of which share a principle of identity. “Is a the same goat-absence as b?” does not ask a determinate question because there is no principle of identity for goat-absences that would determine the answer to such a question. “How many goat-absences are there in the room?” fails to ask a question for the same reason. Particularity requires principles of identity and individuation. Particular-absences do not have such principles (i.e., goat-absences don’t, tennis-match-absences don’t, continent-absences don’t, etc.). Hence, particular-absences are not particulars—hence the non-negatability of objects and events. Events and objects are not negatable because they have principles of identity and individuation. They have these principles because they are particulars. States are negatable because they have neither principles of identity nor individuation. And so they are not particulars. The negatability of states, then, is deeply connected to the fact that states are not particulars. The non-negatability of events and objects, in turn, is deeply connected to the fact that events and objects are.

2.1.4 States Are Mass- Quantified A final reason for rejecting the thought that there are token or par ticular states comes from a consideration raised by Mourelatos in his seminal

30. For illuminating discussions of the relationship between kind-membership and identity questions, see Wiggins, Sameness and Substance, and E. J. Lowe, Kinds of Being (Oxford: Basil Blackwell, 1989).

208

Rational Causation

“Events, States, and Processes.”31 There he contrasts event-nominalizations and state-nominalizations. The former are count-quantified; the latter are mass-quantified. For example, the event-predication (G) Jones crossed Columbus Avenue is nominalized as (H) There was a crossing of Columbus Avenue by Jones. The state predication (I) Jones hates Smith is nominalized as ( J) There was hating (or hate) of Smith by Jones. Salient in this contrast is the presence of an article in the former case and the absence of an article in the latter case. The presence of the article in the event-nominalization signals that the sentence is governed by an existential quantifier ranging over countables, i.e., particulars. Thus (K) Jones crossed Columbus Avenue three times is nominalized as (L) There were three crossings of Columbus Avenue by Jones. The absence of the article in the state-nominalization signals that there is no quantifier ranging over countables at all. (J) is, in this respect, like

31. Mourelatos, “Events, Processes, and States.” Steward, in The Ontology of Mind, also brings the argument of Mourelatos’s paper to bear in her discussion of states, albeit to different effect.

Events and States

209

(M) There was water in the pot We say neither (N)* There were three hatings (or hates) of Smith by Jones nor (O)* There were three waters in the pot but rather (P) There was much hating (or hate) of Smith by Jones and (Q) There was much (or little) water in the pot. State-nominalizations are mass- rather than count-quantified. It might be objected that although we have not heretofore count-quantified state-nominalizations, we could begin to do so. Our ordinary talk about states, the objection might go, is primitive; but we could gradually change that, developing a more sophisticated language that would pick out these neglected particulars. There is some truth to this objection. We could undoubtedly give a meaning to sentences such as (N) and (O). But in so doing, we have not made states or stuff countable; rather we have once again changed the subject. To say that there are three waters in the pot might be to say, for example, that there is some Perrier, some East River, and some Poland Springs in the pot. It also might be to say that there are three units of water of some determinate size. But in both cases we have stopped talking about stuff and have instead begun to talk about something else: kinds in the first case and objects in the second. A hating of Smith, in turn, might be something like a mental episode in which I focus on the qualities of Smith I hate, during which feelings of anger and the like pass through my consciousness. How many hatings of

210

Rational Causation

Smith I have undergone in the last month could then be understood as the number of such episodes that have occurred. But now ‘hating’ has taken on an event use (as ‘believing’ did in section 2.1.1). For hatred, the state, is not an episode at all. It is not required for the truth of “Jones has always hated Smith” that Jones has been undergoing one long episode of this sort—it may be that Jones thinks of Smith very rarely. That we can use the word ‘hate’ to pick out a particular thus does not show that what we have heretofore called ‘hate’ is a particular or has particular instances. Rather, when we use the word ‘hate’ to pick out a particular we have, as above, simply begun to use ‘hate’ in order to talk about a certain kind of event and its instances. We have changed the subject. Thus, although we are not forbidden from inventing a token sense for ‘hate,’ it is just that—an invention. It does not show anything about the metaphysics of hate, the state. In appreciating this point, it is important to keep this in mind: I do not hold that no mental item is count-quantified. My topic in this section is states, not the mental. Smith might have experienced a longing for peas today several times. He may have felt a pinprick of jealousy six times and each time been embarrassed at his own feelings. But these experiences, feelings, and reactions are mental events. As above, we can also make sense of a question such as “how many times has Smith been gloomy today?,” even though being gloomy is a state. In such a case, we count how many times he has become (and then later ceased to be) gloomy. Becoming (and later ceasing to be) gloomy are countable types. Being gloomy is not. Event-nominalizations are count-quantified; state-nominalizations are not count-quantified. A plausible explanation for this discrepancy is that events, because they have principles of identity and individuation, are countable, and states, because they have no such principles, are not. But if states are not countable, then they are not particulars. We have seen that states are not particulars of any familiar type and that there are good reasons for thinking that states are not particulars of any type at all. 1. Whereas events are unitary, states are dissective. 2. The persistence of an object is a matter of the continued presence of a particular over time; the persistence of a state is a matter of a particular continuing to exemplify a universal.

Events and States

211

3. States are negatable; events and objects are not. 4. Event-nominalizations are count-quantified; state-nominalizations are mass-quantified. These points can all be explained by the fact that event-types and objectsortals have principles of identity and individuation, whereas states lack them. And lacking such principles, states do not have instances, whereas object-sortals and event-types do. Thus there are no things that are rightly called states and in the category of particulars. This is important for at least these two reasons. First, if states are not particulars, then someone who holds that causal relata are particulars must abandon the idea that states are causal relata. This would be an uncomfortable concession for the naturalist who makes extensive use of the idea of causally efficacious token states. Second, since there are no token states, there are no token mental states or token physical states. We can intelligibly raise questions about identity, ‘realization,’ or ‘supervenience’ with respect to mental and physical states only to the extent that these relations can connect universals. And so if, for example, mental and physical universals are not identical, there is no sense in which mental states are identical to physical states. There is no intelligible ‘doctrine of state-token identity’ to provide a middle ground for physicalists who reject state-type-identity.32 Those who hold that beliefs, desires, and states more generally are efficacious and also that only particulars are efficacious may try to shrug this result off. Davidson, in the following passage, suggests a strategy for shifting causal responsibility from states to events: “In many cases, it is not difficult at all to find events very closely associated with the primary reason. States and dispositions are not events, but the onslaught of a state or disposition is. A desire to hurt your feelings may spring up at the moment you anger me. I might start wanting to eat a melon just when I see one; and beliefs may begin at the moment we notice, perceive, learn, or remember something.”33

32. ‘Direction of fit’ analyses of beliefs and desires also typically rely on token interpretations of these states. 33. Donald Davidson, “Actions, Reasons, and Causes,” in Essays on Actions and Events, 12.

212

Rational Causation

This is hardly satisfying, however. The primary reason, Davidson seems to concede here, is not really the cause of the action; the causes are “closely associated” events. He thereby seriously compromises a chief selling point of his view, the idea that reasons are causes—i.e., more or less what I called, in Chapter 2, the Equivalence Thesis. Furthermore, the events in question are ones of which we almost never think or speak: “onslaughts” of states or dispositions, startings to want, and beginnings to believe. The resultant view has become rather far removed from the ordinary conception it is intended to vindicate.

2.2 Events-in-Progress In Chapter 4 I argued that an action performed for an instrumental reason is the effect of another event’s unfolding, such that the means-event is a part of the whole-event. I considered an objection from the principle that effects cannot stand to causes in the relation of part to whole. My reply depended on the fact that the explanandum-action is a part of an unfolding and so incomplete whole. It is not, in other words, a matter of the efficacy of a particular. The efficacy of the unfolding event can be nothing other than that of the fact that an event of a certain type was unfolding. I now defend the claim that events-in-progress are not particulars and explain how they fit into the taxonomy I have so far developed in this chapter. You may think that it is obvious that events-in-progress are particulars. Suppose Jones was crossing Columbus Avenue when he was ‘beamed up’ to an alien spacecraft. Even though he never fi nished crossing the street, there was an event of a certain type, crossing of Columbus Avenue, unfolding. It may seem to fly in the face of this reality to deny the existence of such an event until after it is already fi nished. Furthermore, if Jones was crossing the street briskly and at night, then it seems we can predicate ‘brisk’ and ‘at night’ of the unfolding event, just as we can in the case of a completed event. Finally, if Jones stopped after a few steps to tie his shoe and then resumed (before being abducted), we would know to say that he was still crossing the street and not crossing the street again. Such knowledge may seem to bespeak some grasp of individuation conditions for events-in-progress. I revisit these considerations after saying why the thesis they appear to support cannot be true.

Events and States

213

The traditional approach to the distinction between sentences with perfective and imperfective verb-aspect has been to analyze the latter in terms of the former,34 a project that suggests the possibility of an ontological reduction of events-in-progress to other entities. How precisely this would shape our understanding of causal explanations that appeal to facts about unfolding events is not a topic into which we need to delve here. For the traditional approach has been convincingly refuted, chiefly by Terrence Parsons and Zoltán Gendler Szabó.35 My concern here is rather with the manner in which statements with imperfective verb-aspect are thought best to be represented. Though Parsons and Szabó have significant disagreements, they share a commitment to understanding statements with either perfective or imperfective aspects as quantifying over particulars. Here is Szabó’s analysis of statements with perfective verb aspect: “Jones crossed Columbus Ave” is true iff (i) “Jones was crossing Columbus Ave” is true of some event e (ii) “Jones is across Columbus Ave” is true of some state s (iii) e causes s and (iv) If e causes e' and e' causes s then “Jones was crossing Columbus Ave” is true of e'.36 This analysis suggests the following picture: as soon as the event is in progress, there is a particular event of the type crossing of Columbus Avenue. It is not a completed event of this type; rather it is an in-progress event of this type.37 The event is completed when an event-in-progress ‘causes,’ 34. See, e.g., Dowty, Word Meaning. 35. See Parsons, Events in the Semantics of English; Zoltan Gendler Szabó, “On the Progressive and the Perfective,” Nous 38 (2004): 24–59; and Zoltan Gendler Szabó, “Things in Progress,” Philosophical Perspectives 22 (2008): 499–525. 36. Szabó, “On the Progressive and the Perfective,” 45. 37. See also Szabó, “Things in Progress,” for a more explicit discussion of this aspect of the analysis.

214

Rational Causation

in a quasi-technical sense,38 the ‘target state,’ which is Jones’s being across Columbus Avenue. But this analysis is deeply puzzling, even putting aside the odd use of ‘causes’ and the reification of states. For it is not clear that there is any satisfying way of understanding what sort of thing the relevant es and e's are supposed to be. To see this, let us ask whether ‘in-progress’ is an ordinary or an alienating adjective. It cannot be ordinary, for it would then follow from the existence of an in-progress Columbus Avenue-crossing that there is (or was) already an instance of the event-type crossing of Columbus Avenue. But there isn’t (or wasn’t). For Jones (we can suppose) has yet to cross the street. The answer to the question of how many times Jones has crossed the street is “none.” Before he crosses the street, there is no instance of the type. But if ‘in-progress’ is an alienating adjective, as Szabó seems to think,39 we then are left with a question about what, in the most general sense, events-in-progress are. A basic question about them is raised by clause (iv) above. It suggests that there may be many in-progress Columbus Avenue street-crossings leading up to the target-state, each of which is ‘causally’ connected to the other. But it is consistent with there being only one such event-in-progress, presumably one that lasts as long as it takes Jones to cross the street. Which option is correct: one or many? Putting aside the constraints imposed by Szabó’s specific theoretical interests and commitments (whose assessment lies well beyond the scope of the present inquiry), we know perfectly well what determines the answer to such questions in the case of events: principles of identity and individuation for specific event-types. To know what it is to cross Columbus Avenue is to know how to identify and individuate crossings. But nothing about the idea of being in the midst of crossing Columbus Avenue tells us how many in-progress crossings are in a whole crossing. This supposed categorization does not help to answer the “one or many?” question. We saw earlier in this section that the sort of quantitative questions that apply to completed events do not apply to states. It makes no sense to ask (about a state) how many times Jones hated Smith during a five38. Ibid., 513. 39. Ibid., 515–518.

Events and States

215

minute span, though it does make sense to ask how many times Jones crossed Columbus Avenue during a five-minute span. Similarly, it seems, it makes no sense to ask (about an event-in-progress) how many times Jones was in the midst of crossing Columbus Avenue during a five-minute span. In the case of states, we can ask for how long the object was in the state: “for how long did Jones hate Smith?” Similarly, this question is perfectly intelligible: “for how long was Jones crossing Columbus Avenue?” In the case of both states and events-in-progress, we cannot ask how long it took: e.g., “how long did it take Jones to hate Smith?” or “how long did it take for Jones to be crossing the street?” But we certainly can ask the question about events: “how long did it take for Jones to cross the street?” These data suggest that events-in-progress are, like states, non-particular. It might be countered that events-in-progress are more like what Vendler called processes. According to the fourfold Ryle-Vendler taxonomy, process-predications are distinguished from state-predications by the fact that the former use verb-phrases with continuous or progressive tenses, whereas the latter do not. Process-predications, unlike what Vendler called accomplishments—paradigmatic events in my sense40 — exhibit the following pattern of inference. If Jones is running, it follows that he has run. But if Jones is running a mile, it does not follow that he has run a mile. As Vendler puts it: “Running a mile and drawing a circle have to be finished, while it does not make sense to talk of finishing running or pushing a cart. Thus we see that while running or pushing a cart has no set terminal point, running a mile and drawing a circle do have a ‘climax,’ which has yet to be reached if the action is to be what it is claimed to be.” 41 Processes, unlike accomplishments, are atelic. Vendler goes on to point out that it makes no sense to ask how many times Marion ran over a ten-minute stretch during which the statement “Marion is running” was true or to ask how long it took for her to be running. Yet it makes perfect sense to ask for how long she was running.

40. Events, on my view include Vendler’s ‘accomplishments’ and also his ‘achievements,’ to be discussed shortly. 41. Vendler, Linguistics and Philosophy, 100.

216

Rational Causation

Someone might argue that events-in-progress are more like processes and less like states for the following reason. There is nothing about the category of a state that demands that a state obtain for more than an instant, even if there are durative requirements on the obtaining of specific kinds of state. If a man cannot be red for a mere instant, it is not because of the requirements of statehood but rather being red. But processes must, it seems, have temporal extension. A man cannot be swimming or falling or adding for a bare instant. Similarly, a man cannot be in the midst either of swimming a mile or falling down the stairs for a bare instant. The assimilation of events-in-progress to processes is of no help, however, to the advocate of the particularity of events-in-progress. The principles of identity and individuation for events derive in part from their teloi. It is reaching the telos that makes it true that there is a new instance of the type. It is because one can finish performing an event of a certain type—by reaching its telos— and start performing another event of that same type that there can be a question about whether one is still performing the same E as one was performing earlier. But a process (in Vender’s sense) has no telos. Furthermore, there is something fishy about the whole idea that eventsin-progress are a distinct metaphysical category alongside events, states, and processes. Consider again the question of the nature of their relationship to time. Objects are the sorts of things that can be fully present at a given time and that can persist or endure through time. Events have beginnings, middles, and ends. They don’t exist all at one time but rather are whole only after they have come to completion. Events, in other words, take time. States obtain, or not, at specific times and over specific intervals. Events-in-progress, in one way, seem like processes, which go on for at least a little while. But a process-type, such as running, is not the sort of thing that is completed, for it has no telos. Crossing of Columbus Avenue, the event-type, does have a telos. But crossing of Columbus Avenue, the event-in-progress-type, cannot, it seems, have a telos. Or at least it cannot have a telos in the way the corresponding event-type does, or else it would simply be the event-type. If events-in-progress are a distinct metaphysical category from event—if, that is, ‘in-progress’ is an alienating adjective—then they would have to be the kinds of things that

Events and States

217

do not come to completion in the way events do. But isn’t the whole idea of an event-in-progress that it can culminate in a completed event?42 It might be thought that we abandoned the first option—that events-inprogress are events—too quickly. Let us try it out again. Perhaps we should say that when Jones begins to cross Columbus Avenue, there is a new particular on the scene— an instance of the event-type, crossing of Columbus Avenue. Until it is complete, the particular is an incomplete instance of the type. After it is over, it is a complete instance. This way of looking at the matter is appealing only insofar as one fails to keep the temporal features of events clearly in view. We imagine an event as something that persists through time and completeness as a property that the event eventually acquires, the way a grizzly bear acquires an extra layer of fat before a long hibernation. But events are not objects. Events, as I argued in section 1.2, do not change. There are no such particulars that go from being incomplete to complete. Completeness is not a final property that a particular event acquires; it is that in virtue of which there is a particular event. If no one has crossed Columbus Avenue, there is not yet a crossing of Columbus Avenue, i.e., the answer to the question “how many times has Columbus Avenue been crossed?” is “none.” To abandon this principle would be to surrender our grasp of the identity and individuation conditions for this event-type. It would be, I contend, to lose contact with the basic idea of Columbus-Avenue-streetcrossings as particular events. The same point applies, mutatis mutandis, for all event-types. This is just to say, again, that events are not present over the whole time they occur. Let’s return to the intuitions that seem to support the view of eventsin-progress as particulars. (a) Denying the existence of a par ticular eventin-progress sounds like denying that an event was, in fact, in progress. (b) We can apparently make predications of unfolding events. (c) There is a distinction between the continued unfolding of an event and the inau-

42. Of course, part of the point of Szabó’s theory is (transposing to the material mode) to understand completed events in terms of sui generis par tic u lar in-progress events. But part of my point is that we would then be explaining the perfectly familiar in terms of the deeply obscure. The fundamentally dual-aspect nature of events must be the starting point for theoretical reflection.

218

Rational Causation

guration of the unfolding of another event of the same type. I take up each of these points in turn. It is of course often true that an event of a certain type was unfolding. Jones, we can suppose, was indeed crossing Columbus Avenue. In this sense, the event-type crossing of Columbus Avenue has a kind of actuality—a real presence in the spatiotemporal world—that it lacks when no one is crossing that street. But this actuality consists not in the existence of a particular but rather only in the obtaining of a fact. And I am hardly suggesting any sort of eliminativism about such facts. Still, they are just that: facts. A perfective event-predication says that an event of a certain type occurred—in the case of our example, an event of the type crossing of Columbus Avenue. An imperfective event-predication states that an event of that same type was occurring. These statements are equally true, but only the former entails the existence of a token event of the described type. Like state- and process-predications, then, imperfective predications do not postulate the existence of a new particular. But this similarity should not be overstated. Imperfective event-predications involve a kind of forecast for new particulars. The imperfective tells us that a particular is in-themaking, where ‘in-the-making’ does not refer to an existing particular’s peculiar way of being but rather expresses, among other things, that the failure to emerge of a not-yet-existing particular of the relevant type would necessarily involve an interruption. Under the influence of the Davidsonian view of the semantics of statements regarding events, it is apt to seem to a philosopher as if statements such as “Jones was briskly crossing Columbus Avenue at night” represent Jones’s (incomplete) street-crossing as a particular and ‘brisk’ and ‘at night’ as its properties. Davidson famously argues that his style of analysis can best explain why, for example, “Jones briskly crossed Columbus Avenue at night” entails “Jones crossed Columbus Avenue.” He never attempts to extend the argument to state-, process-, or imperfective-event-predications;43 but the idea that adverbs and adverbial phrases function logically in the way adjectives and adjectival phrases do is now well entrenched.

43. Parsons endorses the analogous arguments for state predications, though he admits they are far less convincing. See Parsons, Events in the Semantics of English, chapter 10.

Events and States

219

My reply is that any semantics for state-, process- or imperfectiveevent-statements that would quantify over particular states, processes, or events-in-progress is refuted by the arguments already given. Sadly, I have no alternative analysis to offer by way of explanation of the inferences at issue. But it is worth noting that the same problem arises concerning mass-quantified ‘stuff.’ “There is boiling water in the pot” entails “There is water in the pot,” but in this case, too, there is no par ticular over which to quantify. Perhaps the semantics that could explain stuffy inferences would provide some clue to the others. In the meantime, I suggest continuing to think of adverbs as we ordinarily do: as modifying what verb-phrases, not what noun-phrases, pick out. Finally, there is the idea that the intelligibility of the difference between continuing an already-in-progress event and beginning a new event of the same sort bespeaks identity and individuation conditions for eventsin-progress. This is wrong. The difference in question bespeaks only individuation conditions for startings and stoppings, which are what Vendler called achievements. Suppose Jones is crossing Columbus Avenue when he notices that his shoe is untied. He bends down to tie it and then proceeds to the other side of the street. We think of the second half of his journey as a resumption of his street-crossing and not as starting over. Suppose, however, he is abducted by aliens and, after several months of interrogation, is ‘beamed’ back down to the middle of the street. Suppose, further, that upon being ‘beamed’ back, he takes his original course. In such a case we would say that he started again, not that he resumed crossing. No doubt there are many factors that influence our judgments about what to say in such cases; but they do so, I submit, by influencing whether we think that someone has stopped performing the action that she was in the midst of performing. Even when Jones is tying his shoe, he is crossing the street—the broadness of the progressive (discussed in Chapter 2, section 2.3) permits this. That is why we say he is still crossing the street once he resumes his street-crossing after tying his shoe. While undergoing alien interrogation, he is no longer crossing the street. Hence, when he returns, he is crossing the street again. This opposition of ‘still’ and ‘again’ does indeed imply the applicability of principles of individuation, but only to startings, a species of achievement.

220

Rational Causation

According to Vendler, to say that one “reaches the hilltop, wins the race, spots or recognizes something” is to say something about “a defi nite moment.”44 Achievements must, according to Vendler, be distinguished from accomplishments—which together exhaust the category of ‘events’ as I use the term: “When I say that it took me an hour to write a letter (which is an accomplishment), I imply that the writing of the letter went on during the hour. This is not the case with achievements. Even if one says that it took him three hours to reach the summit, one does not mean that the ‘reaching’ of the summit went on during those hours.”45 Even without delving too deeply into the nature of achievements, this much is safe to say: someone who goes from being on one side of the street to being in the midst of crossing it must at some point have started crossing it. And someone who crossed the street must at some point have fi nished crossing it. Also, someone who began to cross the street but never fi nished crossing it at some point stopped crossing it. The achievements relevant to our discussion are the boundaries of processes and accomplishments. Kinds of boundary-achievements (e.g., starting to cross Columbus Avenue, finishing cooking breakfast) do determine principles of individuation for the particulars falling under them, though not principles of identity. For achievements have no duration. Principles of identity concern the identity over time of particulars. Particulars, like achievements, that do not last thus have no principles of identity. How many times S has begun to f is determined by how many times it went from being false to being true that S was f -ing. This works whether ‘f’ is an accomplishment or a process verb or verb-phrase. Similarly, how many times S has become F, where F is a state-term is determined by how many times it went from being false to being true that S was F. How many times S finished f -ing or stopped being F is determined by how many times it went from being true to false that S was f -ing or F. S’s f -ing again, as opposed to continuing to f, is not a mater of the emergence of a new particular event-in-progress but of S’s having previously ceased f -ing. S’s continuing to f, as opposed

44. Vendler, Linguistics and Philosophy, 102. 45. Ibid., 104. See also Szabó, “On the Progressive and the Perfective,” 43, for discussion of this point.

Events and States

221

to f again, is not a matter of the persistence of a particular event-inprogress, but of S’s not having ceased f -ing. It might be objected that there is another use of quantitative language for which I have yet to account. It is not hard to imagine a circumstance in which the following would be perfectly intelligible: “What’s going on over there? It looks like a crossing. Maybe two or three. Yes, the whole family is going for it.” Three crossings have begun, yet none is complete. These are not, on my view, three token crossings. So what exactly are there three of? Mom, Dad, and Junior are all crossing. There are three people in the midst of crossing; and if they are all successful, there will be three token crossings. As above, this is all there is to there being three crossings in-progress: not three currently existing token crossings with the property of being under way but rather the imperfective promise of three such tokens. According to the third objection to the thesis that events-in-progress are not particulars, it is refuted by the applicability of quantitative language to what was happening and not just to what happened. I have argued that such language does not show that being in the midst of crossing Columbus Avenue is itself a countable but rather only that startings of, stoppings of, and completed Columbus Avenue street-crossings are countable. This both explains the relevant uses of quantitative language and avoids postulating the existence of entities about which basic individuative questions cannot intelligibly be asked. In this chapter, I have sketched a metaphysical framework for understanding a host of categories, centrally states and events. I had a backward-looking and a forward-looking reason for doing so. The former was to show that the categories of state and event can bear the burden I have assigned them in previous chapters. The latter was to set the stage for Chapter 6, in which they will play an essential role in defending my anti-physicalist view of mind and action.

6 Physicalism

Physicalism has been dominant in the philosophy of mind for at least the past fifty years. Most philosophers of mind consider themselves physicalists of one sort or another. Many treat it simply as a rule of ‘the game’ that any reason for thinking that one view is somehow more entitled to the label ‘physicalism’ than another is ipso facto a reason for favoring the one over the other. A more subtle manifestation of this mood is the manner in which views of the mind are classified, viz., in terms of how they represent the relation between the mind and the physical world: dualism, occasionalism, panpsychism, behaviorism, identity theories, non-reductive materialism, emergentism. It is thought that to describe the relation between the mental and the physical is to articulate the nature of the mind. But this is itself a physicalist prejudice. I have already stated the view of the mind on offer in this book— a view, more specifically, of the human mind in the active exercise of its rationality. My goal in this chapter is not to characterize in precise detail the relationship between the mental and the physical. Rather, it is just to show

Physicalism

223

that my view of the conceptual character and metaphysical underpinnings of rational explanation undermines the central motivations and arguments for many physicalist doctrines. These results are, of course, of independent interest. But they also serve as a crucial part of my defense of the view against a variety of objections that will no doubt have cropped up in the reader’s mind over previous chapters. I show in section 1 that typical arguments in favor of a whole range of physicalist positions regarding events, facts, and states are mistaken. And the fact that rational causation is, in the sense that I have articulated, distinct from efficient causation, points us in the direction of positive, novel arguments for thinking that these forms of physicalism are false, as I show in section 2. The versions of physicalism that I discuss in these first two sections are reductionist, in the broadest sense of the term. According to these views, mental events, states, and/or facts stand in the relation of identity or realization to physical events, states, or facts. But I do not deny that there are physical constraints on rational activity. My position is, for example, consistent with global supervenience: the thesis that physically identical worlds are identical tout court. As I discuss in section 3, some might take this as a reason for describing my own view as physicalist or reductionist. It is, after all, not dualist, as generally understood. But it is silly to argue about the ‘true’ extension of technical philosophical terms. If someone wants to describe my view as a form of physicalism, fine. Physicalism is then consistent with a view of the mind according to which it is a field of non-physical and non-physically realized events, states, and facts and is governed by a kind of non-physical causation. Whatever one calls it, my position is such that no one who in fact calls himself a physicalist would be willing to endorse it. Before I begin, an important caveat. This book concerns a relatively narrow (albeit central) subclass of mental states: beliefs, desires, and intentions. The arguments in favor of physicalism I consider in the first section are not so narrowly focused. My replies, which are based just on the analysis of rational explanations of belief and action, have no direct implications for these same physicalist arguments as applied to kinds of mental states that I have not discussed here. It remains to be seen whether anything like my strategy for refuting these arguments has a legitimate extension to other kinds of mental states. Similarly, my arguments against

224

Rational Causation

physicalism in the second section at best yield the conclusion that beliefs, intentions, desires, and actions are not physical or physically realized. Physicalism, as generally understood, would thereby be refuted. But basic metaphysical questions remain concerning sensations, feelings, and the like. (It is worth pointing out, however, that the case for physicalism has generally been thought to be weakest in relation to these ‘qualitative’ mental states.)

1. Physicalist Arguments Foiled All of the influential arguments in favor of physicalism are based in one way or another on causal considerations. Unsurprisingly, then, the introduction of a novel form of causation throws a wrench into the works. In this section, I consider three physicalist arguments and show how each relies on an attempt to exploit the causal character of rational explanations, an attempt that in each case must fail if rational causation is, as I have argued, fundamentally non-efficient. The first argument (section 1.1) is based on ‘the causal analysis’ of mental states. It is the foundation of all of the broadly functionalist doctrines. Next (section 1.2) I discuss Davidson’s argument for anomalous monism, which attempts to identify the mental and the physical through the principle that causation has a fundamentally nomological character. Finally, I argue (section 1.3) that physicalist arguments based on the ‘principle of the causal completeness of the physical’ also fail, in part because the kind of causation at issue in this principle is efficient, not rational. Throughout this discussion, I make frequent use of the term ‘physical’ and its cognates. As to the question of what, precisely, these terms mean— whether, e.g., the physical is a realm to which only current or ideal physics directly refers, or whether it also includes as such some of the states, events, and objects that figure in common speech and experience—I defer to the physicalist herself. Arguments in favor of physicalism, as we shall see, have often relied on principles, such as Completeness, the Nomological Character of Causality, and Basic Global Supervenience, that would not be plausible on more inclusive interpretations of ‘physical.’ But my arguments against these principles do not depend on the stricter understanding of this language. Ultimately, I shall argue for a position that no member of the physicalist family would claim as kin.

Physicalism

225

1.1 The Causal Analysis of Mental Concepts The causal analysis of mental concepts, as D. M. Armstrong called it, is the source of a great variety of physicalist doctrines about the mind. According to this analysis, “the concept of a mental state essentially involves and is exhausted by, the concept of a state that is apt to be the cause of certain effects or apt to be the effect of certain causes.”1 The concept of belief, on this view, is the concept of that which is produced by certain causes and which brings about certain effects, “whatever it is.” Mental terms pick out entities extrinsically, in terms of their relational (causal) properties, rather than in terms of the underlying nature (or natures) of these entities. For Armstrong, Lewis, and others, it is a very short step from the causal analysis of mental concepts to the identity of mental and physical states.2 For it is states whose nature is intrinsically physical that are thought, ultimately, to be causes and effects. Hence, mental states just are physical states. Or, to use the preferred term in these discussions, mental properties are just physical properties. This view, first-order or realizer functionalism, must be distinguished from another: second-order or role functionalism. Here is what the views have in common: according to both, mental concepts are specifications (implicit in ordinary talk and perhaps refi ned by a scientific psychology) of causal roles. Here is the difference: according to realizer functionalism, the physical properties that play the specified roles just are mental properties.3 This is the Armstrong-Lewis view. According to role-functionalism, mental properties are the properties, possessed by physical properties, of playing the right roles, i.e., mental properties are second-order properties possessed by physical properties.4 1. D. M. Armstrong, “The Causal Theory of the Mind,” in Mind and Cognition, ed. W. Lycan (Oxford: Blackwell, 1990), 20–26, 22. 2. See David Lewis, “Psychophysical and Theoretical Identifications,” Australasian Journal of Philosophy 50 (1972): 249–258. 3. This would be type-physicalism if and only if, for each functionally characterized mental role, there were just one physical property that played the role. I don’t discuss type-physicalism about the psychological states at issue in this book, since it is not a position anyone holds or has held for a very long time. 4. See Ned Block, “Troubles with Functionalism,” in Perception and Cognition: Issues in the Foundations of Psychology, ed. C. W. Savage (Minneapolis: University of Minnesota Press, 1978), 261–325.

226

Rational Causation

The advantage of the latter view is that it accounts for the idea that two beings, both of whom, say, believe that the universe is expanding, have something genuinely in common. On this view, despite possibly having nothing physically in common, they share a mental property. It is this version of functionalism that seems best to account for the intuition that mental states are ‘multiply realizable.’ But role-functionalism is charged with being insufficiently physicalist. For some, it is enough to point out that it countenances non-physical properties, viz., second-order functional properties, which can be shared by beings with nothing physically in common. Others charge that the problem with these properties is not simply that they aren’t identical to any physical properties but that they either possess spooky causal powers or that they spookily possess no causal powers. The causal power possessed by such a second-order property would be spooky just because the property itself is not physical. And their failure to possess causal powers would be spooky because to be real is, according to the popular dictum of Samuel Alexander, to be efficacious. The central line of defense against the ‘not physicalist enough’ objection to role-functionalism is to argue that functional properties, while not identical to physical properties, are nonetheless so closely related to physical properties that the resultant view still has some claim on the labels ‘physicalism’ or ‘materialism’ or ‘naturalism.’ The word ‘realize’ is generally deployed to convey the closeness of this relation, though its meaning varies rather widely. I make the distinction between these two kinds of functionalism to show how wide-ranging the influence of the causal analysis of mental concepts is, encompassing both identity and realization brands of physicalism. I rebut both in section 2. In this section my aim is more limited. I show here that to the extent that something true can be salvaged out of the causal analysis of mental concepts, it is not a truth that supports physicalism. The causal analysis of concepts like belief and desire are plausible for two reasons. First, unlike in the case of pain or pleasure, there is no temptation to identify believing and desiring with some characteristic feeling—they have no phenomenology. It is for this reason that whereas functionalism is back on its heels on the topic of consciousness, it remains widely popular as an approach to understanding ‘propositional attitudes.’ Second, explain-

Physicalism

227

ing the thought and behavior of others accounts for one central use of these concepts, and such explanations are, as I have argued in Chapter 4, causal in the broad sense. I have three objections to the argument for physicalism based on the causal analysis. The fi rst is a cavil; the second is deeper; the third is crushing. First, the analysis, as formulated, seems to rely on the idea that beliefs and desires are a class of particulars that stand to belief- and desire-types as token events stand to event-types. This picture is, as I have argued, fundamentally confused. There are no token states. The idea of beliefs and desires as causally efficacious is at best the idea of states such that someone’s being in the states is apt to cause and be caused by such-and-such. Second, the analysis is based on the use of mental concepts in the explanation of the thought and behavior of others. When I say “I believe Missouri extends farther east than Minnesota,” I do not take myself primarily to be self-ascribing a state that is “apt to be the cause of certain effects or apt to be the effect of certain causes”— as I would in the case of, e.g., “I have contracted a contagious disease.” This is not to say that someone else couldn’t appeal to this belief to explain what I do or what else I think. But what I am saying about myself when I use the word is not that I am in a state that has such-and-such a causal profile—I am not or in any case am not merely giving you permission to draw various conclusions about what happened before and what is likely to happen next—but rather simply that I take a certain proposition to be true.5 (Hence the Moore-Paradoxicality of “p; but I don’t believe p.”) Such statements are paradigmatically uttered for the purpose of declaring one’s allegiance to a specific answer to a question at hand. That fact must be front and center in any good analysis of the self-ascriptions of belief. Before you object that I am confusing conceptual analysis with metaphysics, recall that I am making an objection to the causal analysis of mental concepts. My point here is not that this fact about meaning establishes that mental states are not individuated causally. This is my complaint: the idea

5. Compare Arthur Collins, The Nature of Mental Things (Notre Dame, IN: University of Notre Dame Press, 1987), and Richard Moran, Authority and Estrangement (Princeton, NJ: Princeton University Press, 2001).

228

Rational Causation

that the traditional causal analysis of mental states could be exhaustive wildly overstates its prima facie plausibility, which is tied solely to its third-person uses. Once the weakness of the causal analysis in its application to firstperson statements is recognized, the attraction of functionalist analyses of third-person belief-attributions also begins to wane. After all, to say of another that she believes that p is just to say of her what one says of oneself using the first person. This leads us to the third and most telling objection. Armstrong’s approach is often described as topic- neutral. The thought behind this label is that in giving a functional analysis of X, one does not commit oneself to any specific view of the nature of the properties that play the X-role. But functional analyses of concepts are only as topicneutral as the conceptions of causation on which they rely. To the extent that ‘cause’ and ‘effect’ are understood as efficient, an analysis that employs them is not topic-neutral after all. For it prejudices subsequent discussion of the ontology of the relevant kind of thing in favor of physicalist accounts, since efficient causation is linked especially to physical laws. But the plausibility of any causal analysis of concepts such as belief, desire, and intention derives from our grasp of their essential role in rational explanations. And the causation that makes these explanations true is rational, not efficient. There is in fact a very long step from even a partial causal analysis of these concepts to the idea that physical states play the corresponding roles. To say that Poirot believes that the butler did it because he believes that everyone else has an alibi is not to explain by appeal to his being in a state that is apt to be the efficient effect of his believing that Colonel Mustard has an alibi, that Professor Plum has an alibi, etc., and that is apt to be the efficient cause of, say, his locking the butler in the basement. Rather, it is to explain by appeal to his being in a state at which he might arrive by an inference from the fact that Colonel Mustard has an alibi, that Professor Plum has an alibi, etc., and that is apt to be the grounds for his performing an action such as locking the butler in the basement. These causal connections are the manifestations not of physical laws but of the exercises of Poirot’s rational abilities. Good analyses of mental concepts such as belief, desire, and intention provide a route neither to the thesis that mental properties are identical to physical properties (as the realizer-functionalist

Physicalism

229

would have it) nor to the thesis that they are second-order properties possessed by physical properties (as the role-functionalist would have it).

1.2 The Principle of the Nomological Character of Causality The only other physicalist argument comparable in influence to the functionalists’ is Davidson’s argument for Anomalous Monism. Davidson’s project is explicitly anti-analytic. He famously argues for a kind of holism about the mental: “Beliefs and desires issue in behavior only as modified and mediated by further beliefs and desires, attitudes and attendings without limit.”6 Although the only analytic project he explicitly targets in “Mental Events” is ‘definitional’ behaviorism, his remarks apply equally well to causal analyses of mental concepts that (unlike behaviorist analyses) allow other mental terms to occur in the definientia: “It is a feature of the mental that the attribution of mental phenomena must be responsible to the background of reasons, beliefs, and intentions of the individual. . . . [W]hen we use the concepts of belief, desire, and the rest, we must stand prepared, as evidence accumulates, to adjust our theory in the light of considerations of overall cogency: the constitutive ideal of rationality partly controls each phase in the evolution of what must be an evolving theory.”7 In part on the basis of such holism, Davidson also rejects the idea that there could ever be strict laws linking mental and physical predicates or laws linking mental predicates to one another. Nonetheless, he argues that particular mental events are identical to particular physical events on the basis of the Principle of the Nomological Character of Causality: “events related as cause and effect fall under strict deterministic laws.” In combination with the Principle of Causal Interaction—“at least some mental events interact causally with physical events”— and the Anomalism of the Mental—“there are no strict deterministic laws on the basis of which mental events can be predicted and explained”8 —he derives the

6. Donald Davidson, “Mental Events,” in Essays on Actions and Events (Oxford: Clarendon Press, 1980), 217. 7. Ibid., 222–223. 8. Ibid., 208.

230

Rational Causation

conclusion that at least some mental events fall under strict deterministic laws and are as such physical. But the appeal of the Principle of Causal Interaction (at least in its application to the sorts of mental states that are the focus of our inquiry) is based on rational explanations. The examples that figure most prominently in Davidson’s discussions are of rational action-explanation. People do things as a result of what they want and think. And what they do often involves their moving stuff around. Hence the nearly inescapable conclusion that mental states and events affect stuff. But the efficacy of mental states in central cases of action-causation is rational efficacy. A man may push a sofa down a hallway because he wants to move it into the living room. His moving of the sofa is thus a consequence of his desire. But, as I have argued, it is a rational consequence. And the effects of rational causation are not physical events at all. They are, as I argued in Chapter 2, thoughts. I will say more in section 2.1 about what it means to say that an agent’s movement is a thought. Davidson’s argument fails because the plausibility of the Principle of Causal Interaction derives in large measure from the rational efficacy of the mental, whereas the Principle of the Nomological Character of Causality owes its plausibility to an analysis of the idea of effi cient causation.9 There is no single sense of ‘cause’ on which both principles turn out true. Hence they do not combine to help establish that mental events fall into the extension of physical predicates.

9. Davidson defends the principle in “Law and Cause,” Dialectica 49 (1995): 263– 279, by arguing that what we count as a change and thus what is in need of causal explanation is determined in part by what laws are already available to explain change. But the examples on which he focuses are from scientific inquiries, in which the cause-law connection is not in dispute (at least not by me). By Davidson’s own lights, however, we are guided in rational explanation by ‘the constitutive ideal of rationality’ and not by the ideal of a law. If this is right, it remains unclear why we should think that the lesson from science should carry over to the realm of reason-giving explanations. Compare John McDowell, “Functionalism and Anomalous Monism,” in Action and Events: Perspectives on the Philosophy of Donald Davidson, ed. B. McLaughlin and E. LePore (Oxford: Blackwell, 1985), and Jennifer Hornsby, Simple Mindedness (Cambridge, MA: Harvard University Press, 1997), especially chapter 8.

Physicalism

231

1.3 The Causal Completeness of the Physical Realm Another motivation for physicalism about the mind derives from what is called alternatively the completeness or closure of the physical realm. Very crudely, the idea is that the self-sufficiency of physical causal processes leaves no room for non-physical particulars or properties to get in on any of the world’s causal action. Before replying, I make this thought a little more precise. I have found it useful to employ ‘Completeness’ and ‘Closure’ to denote different ideas. What I call Completeness can be understood as the ontological version of the thesis that physics is in-principle completable, that there is some true physical theory capable of fully explaining why physical processes unfold in precisely the way they do. To say that a theory fully explains why physical processes unfold in the way that they do is to say that at each stage in a physical causal chain, the causal connection between it and earlier and later stages can be completely accounted for by a true physical theory.10The physical causal histories of physical events are complete in virtue of this fact about them; and as far as we know, this property of physical causal histories is unique to them. According to the principle of Closure, physical events cannot interact causally with non-physical events or with physical events in virtue of their non-physical properties. Closure represents the physical world as cloistered from the influence of anything non-physical: nothing non-physical can affect the physical. Sometimes, these principles are simply confused. The premise of Lewis’s argument for identity theory is “the plausible hypothesis that there is 10. This does not mean that every physically describable state of affairs is causally determined by prior physical states of affairs, nor does it mean that every feature of every physical event has a cause. Famously, there are significant correlations between the events in spatially disparate quantum mechanical systems that have no common cause. Completeness says only that one need never depart from descriptions of the physical in order to account for what happens there. The just-mentioned correlations would only violate the causal completeness of the physical if there were some reason to think that they were the result of non-physical intervention. Also, it may be that there are some physical events that have no causes at all. If so, we could modify Completeness slightly: all physical events either have complete physical causal histories or have no causal histories at all.

232

Rational Causation

some unified body of scientific theories, of the sort we now accept, which together provide a true and exhaustive account of all physical phenomena (i.e., all phenomena describable in physical terms).”11 This premise, Lewis says, “does not rule out the existence of nonphysical phenomena: it is not an ontological thesis in its own right. It only denies that we need ever explain physical phenomena by non-physical ones. . . . All manner of nonphysical phenomena may coexist with them, even to the extent of sharing the same space-time, provided only that the nonphysical phenomena are entirely inefficacious with respect to the physical phenomena.”12 In the first part of this passage, this premise states that we never need to explain physical phenomena by non-physical ones. But to say that we don’t ever need to causally explain physical phenomena by appealing to nonphysical phenomena is not to say that we cannot causally explain the former by appealing to the latter. The collision of a pair of molecules may well be traceable back via the laws of physics to earlier physical events, but this fact does not on its face rule out that the very same collision might also be traced back, via different kinds of causal principles, to a man’s desire to eat lunch, or to the fall of the Dow Jones Industrial Average, or to any number of other kinds of factors. The view that we don’t need to depart from physical explanations to explain physical events is Completeness; the view that we necessarily go wrong in so departing is Closure. More recently, philosophers have been careful to avoid simply confusing the principles. But there is still widespread agreement that Completeness provides a powerful reason for accepting physicalism. Jaegwon Kim argues, in effect, that Completeness, together with other plausible principles, entails Closure. The overdetermination that would result from denying the inference, he contends, is unacceptable.13 Others have rejected that inference but nonetheless hold that for the resulting overdetermination to be acceptable—for mental events or states that are not identical to physical events or states to have causal power despite Completeness—they must be realized in some sense by physical events or states. They must, for exam-

11. David Lewis, “An Argument for the Identity Theory,” Journal of Philosophy 63 (1966): 17–25, 23. 12. Ibid., 23–24. 13. Jaegwon Kim, Mind in a Physical World (Cambridge, MA: MIT Press, 1998).

Physicalism

233

ple, stand to the latter as determinable to determinate,14 or they must be mechanistically explainable in terms of the latter.15 There is consensus, then, that even if the efficacy of the mental is not ruled out by Completeness, its own efficacy must piggyback somehow on the efficacy of the physical. And so there is a cottage industry of finding just such an acceptable piggybacking story. Were rational causation simply efficient causation ‘at a higher level’ than physical causation, then perhaps we would be obliged to accept the ‘piggybacking’ requirement for the mental. If a falling rock shatters a window, then even though rocks and windows are not the sorts of objects mentioned by the physical theory that figures in Completeness, the falling of the rock and the shattering of the window must be manifestations of the billions of events occurring amongst the baseball’s and window’s physical parts. As the vast literature on the argument from overdetermination demonstrates, it is surprisingly difficult to make this thought precise and, as a consequence, very difficult to know how either to defend it or to determine its consequences for the exact status of ‘higher-level’ causation. But none of that is my concern here. My point is that even if we accept that something of the sort is true, we are not thereby obliged to view instances of rational causation as manifestations of the billions of causal interactions occurring amongst the relevant person’s physical parts. For rational causation is not physical causation viewed from a ‘high level.’ It is not the same sort of causation at all. The rational ‘because,’ in the first instance, links the elements of theoretical or practical inference. In central theoretical cases, a thinker extends the status ‘to be believed’ from one or more propositions to another proposition. In practical cases, an agent extends the status ‘to be done’ from one action to another. The ‘because’ in a rational explanation refers to this inferential nexus. To say S was f -ing because q is to say that S made a practical inference, the conclusion of which was her action, and

14. Stephen Yablo, “Mental Causation,” Philosophical Review 101 (1992): 245–280. 15. See Jerry Fodor, “Making the Mind Matter More,” Philosophical Topics 17 (1989): 59– 80, and Terrence Horgan, “From Supervenience to Superdupervenience: Meeting the Demands of a Material World,” Mind 102 (1993): 555–586.

234

Rational Causation

which she could express by saying “I am f -ing because q” or “I’ll f because q.” This is the way rational explanations explain and the kind of causation they postulate. It has nothing to do with placing events, states, or facts under the rubric of natural laws. It is not, in other words, efficient causation. And so it is slightly misleading to speak (as I have16) of rational causes as overdetermining causes at all.17 Let us approach the point from a different direction. Consider the causal chain from the falling of the rock to the shattering of the window. According to physicalist arguments of the sort we are considering, Completeness, together with other plausible assumptions, entails (though unlike Closure does not simply state) that this causal chain consists in, is realized by, or is nothing over and above a physical causal chain. No one finds this puzzling or aspires to discover a principled reason for viewing rocks or the property of being a rock as ‘filling in the gaps’ in physical causal chains. But such an idea is not absurd. The attribution of magical, rather than physical, forces to rocks is simply far-fetched. Consider now the rational causal ‘chain’ from wanting to say hello to a friend across the street to the action of crossing the street. According to one central argument for physicalism, Completeness entails that this chain piggybacks on a physical causal chain. People have found this thought puzzling, and many sense that something important has been left out. But how should one go about rectifying this picture? For lack of any better options,

16. I made this mistake for many years—for example, in “Mental Causation: Unnaturalized, but Not Unnatural,” Philosophy and Phenomenological Research 63 (2001): 57– 83. 17. In “The Mental Causation Debate (Mental Causation I),” Proceedings of the Aristotelian Society Supplementary Volume 69 (1995): 211–236, Tim Crane makes a related point in a discussion of arguments for physicalism from Completeness: “There is an important assumption hidden in these arguments. I call it the ‘homogeneity’ of mental and physical causation. That is, if the arguments are going to work, there must be a confl ict between mental causation and the completeness of physics. But if this is so, the notion of causation is the same notion applied to the physical and the mental alike. . . . There is no confl ict— and thus no need for an identity thesis—if the notions of causation employed are so different. Another way of putting this point is that the arguments for physicalism must assume that the labels ‘mental’ and ‘physical’ as applied to causation are really transferred epithets—what is mental and physical are the relata of causation, not the causation itself” (219).

Physicalism

235

philosophers have tended to assume that Completeness itself would have to be rejected if the ordinary view were to be fully vindicated. Mental events or states, the thought goes, would be causal powers that fill in the gaps in physical causal chains.18 It goes unnoticed in these discussions that this picture of mental causal powers is no closer to the one with which we normally operate than the physicalist picture it is supposed to correct. Mental causation, according to this common misrepresentation of our ordinary view, crucially involves the causation of physically inexplicable physical motion. This suggests that the only thing we ought to count as a distinctively mental cause is something that impinges on physical processes in the way physical forces do but that is after all non-physical. By thinking of mental causes as picking up the causal slack left by physical causes, we adopt what I have called the telekinetic view of mental causation: i.e., mental causation as some kind of spiritual pushing. On this view, we stand to our bodies, or perhaps just to our brains, in the same relation that Yuri Geller purports to stand to his fabulous bending spoons. Take, for example, the motion of a particle of my thumb. On the telekinetic view, when I intentionally raise my hand there are components of the etiology of this particle’s rise that are (at least in part) mysterious from the perspective of the physical sciences. Their best theories may be (if indeterministic) neutral on the question of which of several paths a particle will travel along or (if deterministic) make predictions about its trajectory that prove to be false. The mental cause thus alters the particle’s trajectory, as perhaps the wind might alter the course of a soundly hit tennis ball. This picture of mental causation is not a faithful representation of what is implicit in our ordinary explanatory practice but rather a philosophical distortion. The distortion is a consequence of the failure to recognize the distinctive character of rational causation. Under the influence of this distortion, we expect to fi nd the same kind of evidence for the presence of mental causes as we find for the presence of physical causes. If the operation

18. For defenses of something like this view, see E. J. Lowe, “The Problem of Psychophysical Causation,” Australian Journal of Philosophy 70 (1992): 263–276; E. J. Lowe, “The Causal Autonomy of the Mental,” Mind 102 (1993): 629– 644; and Lynne Rudder Baker, “Metaphysics and Mental Causation,” in Mental Causation, ed. J. Heil and A. Mele (New York: Oxford University Press, 1993), 75– 95.

236

Rational Causation

of physical causation is detectable on the basis of the behavior of lone molecules, flitting one way as opposed to another, so should the operation of mental causation. But there is no commonsensical reason to think that we can distinguish genuine from merely apparent mental causation by attempting to discern physically inexplicable atomic motion. Were I interested in scrutinizing Yuri Geller’s supernatural pretensions, I would devise and conduct experiments that test for the occurrence of this motion. But such tests do not bear at all on the question of the existence of (ordinary) mental causation. As far as I know, no one has ever thought to test for the existence of mental causation using such an experiment. And that is because when I think, for example, that someone crossed the street because he wanted to talk to a friend, I do not see his body as having been blown into the street by some kind of ghostly wind. But if such experiments are commonsensically thought to be irrelevant to mental causation, then the commonsense view of mental causation does not require violations of Completeness. Here is the central point. In the case of the rock, the attribution of nonphysical efficient power to its fall does not get the kind of causation wrong, it just gets the manner of its realization wrong: magic instead of physics. In the case of action, the attribution of non-physical efficient power to mental states results in nothing recognizable as the efficacy of a mental state in relation to an action. More specifically, it does not restore to the mind powers that we intuitively confer upon it. This shows that it is not Completeness that really threatens our ordinary view of the mind. The physicalist picture, according to which the causal power of beliefs and desires consists in physical efficient causation, and the telekinetic picture, according to which the causal power of beliefs and desires consists in magical efficient causation, each reflects obliviousness to the sui generis character of rational causation. This obliviousness is the real threat to our ability to recognize the special powers of minded beings. Here is a Completeness-related question that might motivate physicalism about the mind. If every physical event has a complete causal history, then how else can one move an atom except by figuring as part of the physical causal history of that atom’s movement? Questions of the form “how did S f?” typically mean the same as “by doing what did S f?” or “what did S do in order to f?” The answer to such questions is thus typi-

Physicalism

237

cally something else that the agent did.19 If the question is “how did Albert move atom o from A to Z?,” a perfectly adequate answer might be “by first moving it from A to B, and then moving it from B to C, . . .” But answers such as these do not place the agent in a causal relation to events that are part of the complete physical causal history of the atom’s motion. Rather, they ascribe rational power to the agent, the power to bring about an action performed for a reason. “How did S f?,” as ordinarily understood, does not direct us to a different kind of efficacy, a kind in which rational efficacy ‘ultimately’ consists. But how, the point will be pressed, could a physical event, such as an atom’s motion, be the product of a kind of causation that is genuinely distinct from physical causation, given that physical events can be completely accounted for by appeal to their physical causal history? Does this not require that we view mental efficacy in some sense or another as supervenient on physical efficacy? In other words, does the problem that Completeness seems to pose really depend on the homogeneity of the causal relation, or does it merely depend on the identity of effects? This question begins to loom: do beliefs, desires, and intentions have physical effects? I return to it in section 3.

2. Physicalist Positions Refuted In the last section I argued that the most powerful arguments and motivations for physicalism about the mind are undermined once one recognizes the distinctive character of rational causation. In this section I argue that the reality of rational causation, together with the view of states and events advanced in the last chapter, provides positive grounds for rejecting each of these forms of physicalism: 1. Mental events are identical to physical events. 2. Mental states are token identical to physical states.

19. Many philosophers have held that sometimes this question has no answer, or rather that it can be answered by saying something like “she just did it, that’s all.” But Doug Lavin, in “Must There Be Basic Action?,” Nous (forthcoming), argues that this is wrong. If S is f -ing, then there is some y, such that S is y -ing in order to f.

238

Rational Causation

3. Mental states are identical to physical states. 4. Mental states are realized by physical states. 5. Mental facts are identical to physical facts. I extract my arguments against these positions from a picture of the way kinds of causation must enter into an account of the nature of various kinds of events, states, and facts.

2.1 Mental Events In Chapter 5 I argued that event-identity, like object-identity, is sortal dependent. Since it is in virtue of some principle of identity that event a at one time is identical to event b at some other time, and since event does not supply such a principle, a determinate sortal is required. Following Wiggins, I put the point this way: if a is the same as b, then it must be the same something as b, where something is a quantifier ranging over determinate sortals. Whereas Wiggins applied this principle to objects, the argument of Chapter 5 shows we must extend it to events as well. Here is a more precise statement of the idea: The Type Dependence of Event-Identity (TDE): for any events a and b, a is identical to b if and only if there is some event-type F such that (a) there is a principle of identity for Fs; (b) a and b are Fs; (c) according to the principle of identity for Fs, a is the same F as b. The event-version of the doctrine of token-identity says that for any token mental event m, there is some token physical event p to which it is identical. And according to TDE, if m is identical to p, then there is some event-type F such that m is the same F as p. So there would have to be a type that subsumes m and p. What reasons are there for or against thinking there is any such type? Suppose m is of mental sort M and p is of physical sort P, and m is identical to p. How must M and P be related if TDE is to be satisfied? Consider a parallel example involving objects. Let’s say an object o is of sort O and an object n is of sort N, and o = n. How must O and N be related if

Physicalism

239

object-identity is sortal dependent? Since there are no phase types for events, I restrict my attention to substance sortals. So o is an O throughout its existence, and n is an N throughout its existence. O and N must have the same principle of identity—it is impossible that O and N deliver different verdicts as to the survival of o and n, on pain of contradiction. O and N must thus be related in one of three ways. Either (1) O stands to N as animal to critter (identity); (2) O stands to N as animal stands to lion (determinate to determinable); or (3) O stands to N as lion to animal (determinate to determinable). Sortals that stand in one of these relations overlap. Such overlapping is a necessary and sufficient condition for o and n to be subsumable by the same sortal. Returning to the case of events, there are three ways that the doctrine of token-identity can be saved from the threat posed by TDE. In the case of every token-identity claim between mental and physical events, there must be types M and P such that M and P overlap. Without such overlap, there could not be a common type subsuming token mental and physical events. But there are good reasons for thinking that mental and physical eventtypes do not overlap. For overlapping, I argue, requires that event-types are examples of the same kind of causation. Events, according to the characterization developed in Chapter 5, are changes. They are not the changes philosophers have sometimes identified them with—the adding or losing of a state.20 Rather they are an object’s moving toward a telos (when the event is in progress) or an object’s having reached a telos (after the event is complete). That events are changes is the material correlate of the formal fact that imperfective event-predications ascribe to an object progress toward some telos, and perfective eventpredications say that this progress has culminated in the object’s reaching the telos. Some changes are possible for certain kinds of objects but not for others. A large boulder can flatten a house; a small pebble cannot. In this sense, flattening a house is not a possibility for the boulder. But objects can be distinguished not just by what specific changes they can bring about or undergo but also by what kinds of changes they can bring about or undergo. 20. See Lawrence Brian Lombard, Events: A Metaphysical Study (London: Routledge and Kegan Paul, 1986).

240

Rational Causation

Flattening a house is not a possibility for a pebble. But pebbles possess the kind of power that, in sufficient quantities, would flatten a house. Flattening a house is not an incongruous kind of change to attribute to a pebble. Flowering, giving birth, and jogging a mile, on the contrary, are incongruous kinds of changes to attribute to the pebble. For the pebble does not possess the kind of powers that would make figuring in these events possible. Plants, animals, and people are, respectively, the right sorts of objects to undergo those changes. I put the point this way: flowering, giving birth, and jogging a mile are manifestations of distinct kinds of causation, kinds of causation that must enter into any account of the nature of plants, animals, and people.21 I illustrate this idea by discussing actions, which are, on my view, a kind of mental event. This should not prompt any objection from the physicalist who, if anything, is likely to view actions as more obviously physical then other mental events, e.g., a panic attack or doing a problem of multiplication ‘in one’s head.’ Action should, by the lights of the physicalist, be a very hard case for the anti-physicalist. This will also provide an occasion to elaborate upon and defend an idea, glossed in Chapter 2, which at first seems rather difficult to understand: that a man’s motion through space is itself a mental event. The thesis that actions are thoughts is not a matter of pushing actions back into the head, so to speak, but rather of pushing thoughts out into the world. Let’s take crossing of Columbus Avenue as our sample event-type. For Jones to be crossing the street is for him to represent the crossing of Columbus Avenue as to be done. A chicken can also cross Columbus Avenue. A chicken can even cross Columbus Avenue for a reason, e.g., to get to the other side. But in saying a chicken is crossing Columbus Avenue, one is not, on my view, saying the same thing about the chicken as one says about Jones. This was the point of Chapter 3. For one is not saying that the chicken is representing getting to the other side as to be done, nor is one saying that the chicken is inferring the to-be- done-ness of crossing the street from the to-be-done-ness of getting to the other side. Thus, the fact that the chicken cannot explain herself does not cast any doubt on 21. This discussion is indebted to Michael Thompson, Life and Action (Cambridge, MA: Harvard University Press, 2008), especially part 1.

Physicalism

241

the animal explanation. “. . . is crossing Columbus Avenue” in the sense in which this expression can be affi xed to the name of an agent thus does not signify the same kind of event as when it is affi xed to the name of an animal. Similarly, “. . . is crossing Columbus Avenue,” in the sense in which the expression is applied to an agent does not express the sort of thing that a physical object can do. However, even if one is sympathetic to the idea that the cognitive aspects of human and animal action differ in the way I have suggested, there is something hard to resist about the idea that there is nonetheless a common element in their actions—the plain bodily movement, which we can separate (in thought, even if not in reality) from the cognitive activity that surrounds it. The mental contribution, in both cases, seems as if it must be something separable from the body’s locomotion. But the idea of such a common, extra-mental bodily element is confused. The notion that the identity of an event-type is determined in part by how the object reaches the telos is not so foreign. It is uncontroversial that walking a mile and running a mile are different actions. There is no reason to think this cannot happen under the cover of a single expression. Suppose Greg has dented Jonah’s car with a hammer in order to exact revenge for some insult. Each of these statements will then be true: “Greg dented Jonah’s car” and “The hammer dented Jonah’s car.” But these sentences don’t say that Greg and the hammer did the same thing, i.e., that they instantiate the same event-type. The sentence “Greg dented Jonah’s car and so did the hammer” sounds like “You are free to execute your laws and your citizens,”22 an example of syllepsis. Could Greg and the hammer ‘dent the car’ in the very same sense? It might seem as if the sylleptic reading is an artifact of the hammer’s serving as an instrument. “O dented P” is ambiguous between, roughly, O’s being the agent behind the denting and O’s being that instrument through which the agent accomplished the denting. Or it might be thought that the oddness of the sentence is not a matter of syllepsis at all but rather double-counting: stating or suggesting that there were two dentings. Let’s consider instead a case in which the hammer figures in an event other than as an instrument. 22. The example is, I’m afraid, from Star Trek: The Next Generation, season 1, episode 14.

242

Rational Causation

Suppose a hammer fell from the sky (out of the wheel well of a passing jetliner), denting Jonah’s car, and then Greg dented the car further with his fists. In this case, the hammer isn’t anyone’s denting utensil. And there really are a pair of distinct events. Still, the following sounds sylleptic: “First a hammer and then Greg dented the car.” In the agentive sense, “O dents P” can intelligibly enter at either side of ‘in order to’ constructions; in the non-agentive sense, it can’t. Greg, for example, can swing his arms as fast as possible in order to dent the car, and he can dent the car in order to exact revenge. ‘Dent the car,’ in the sense that applies to Greg, designates the sort of event that can enter into intentionalteleological relationships to arm-swinging and revenge-exacting. The teleological character of this sort of denting is internal to it. Each successive temporal phase of the arm-swinging occurs for the sake of the denting and by virtue of that fact is part of Greg’s car-denting.23 ‘Dent the car,’ in the sense that applies to the hammer, does not designate that sort of event. It is not the sort of event whose occurrence teleologically explains or is teleologically explained by other events. It does not have an internal teleological structure. The temporal phases that constitute the hammer’s denting of the car do not occur for the sake of the result. My claim is not that events with this sort of internal teleological structure only have agents as subjects: non-human animals, plants, and certain artifacts also act (in a broad sense) for the sake of ____. But hammers do not possess even the ability to act for the sake of ____ in the broad sense. To say, in the normal case, that a hammer dents a car is to portray the hammer as causally involved in the car-denting by way of some physical causal process. To say, in the ordinary sense, that Greg dented the car is not simply to say that he was the subject of an event one upshot of which was the car’s being dented. It is to specify that the car’s being dented was an actualization of a kind of ability that issues in events exhibiting self-conscious teleology. A hammer, which possesses no such ability, dents the car in a different sense. (Of course Greg and the hammer can dent the car in the same sense. Suppose the hammer fell from the sky onto the car, thereby denting it, and then soon after Greg fell from the sky denting it further. In such a case, we 23. See Lavin, “Must There Be Basic Action?,” for extensive discussion of this sort of internal teleological structure.

Physicalism

243

could say non-sylleptically: “First a hammer and then Greg dented the car.”) Consider now movement, in the most general sense. Following Hornsby, we can distinguish between two senses of ‘move’: in “Fred moved his foot from A to B” on the one hand and in “Fred’s foot moved from A to B” on the other. ‘Move’ is a transitive verb in the former case (nominalized as ‘movement t’) and an intransitive one in the latter (nominalized as ‘movement i’). The events designated by the above statements cannot, it seems, be the same, since the foot’s having movedi seems to be something like the upshot of Frank’s having movedt it. What is the relation between the action reported by “Fred movedt his foot from A to B” and the event reported by “Fred’s foot movedi from A to B”? Here are a pair of options: (1) The movementt is a cause of the movement i.24 (2) The movement t is the agent’s causing the movement i.25 On both views, the movement i’s relation to the efficacy of the agent is purely external: that very event might have occurred even if the action or agent had not caused it. The movement i, so conceived, is an event that might occur without the intervention of agency. But what type of event is the movement i? It would seem to be a very general type— one might say a pure movement. According to what I call  the doctrine of pure movement, there are individual events that are movements in the most general sense but are not any particular sort of movement, e.g., not a falling or being pushed or a reflex or paratonic movement. A foot-movement i cannot be any of these sorts of movements, since its being so would exclude its being the upshot of an action. (No one thinks that an action or an agent causes a fall or a being pushed or a reflex.) And it cannot be an action, since these movementsi are, by hypothesis, distinct from actions. It must be a movement i pure and simple. But pure movement is a myth. It seems fairly straightforward to define movement-in-general, at least to a crude approximation: for an object to move from A to B is just for it 24. See Jennifer Hornsby, Actions (London: Routledge and Kegan Paul, 1980), and Timothy O’Connor, “Agent- Causation,” in Agents, Causes, and Events: Essays on Indeterminism and Free Will, ed. T. O’Connor (New York: Oxford University Press, 1995), 173–200. 25. See Maria Alvarez and John Hyman, “Agents and Their Actions,” Philosophy 73 (1998): 219–245.

244

Rational Causation

to be at A at one time and B at a later time, and to be at a progression of spots on a spatial path from A to B during the intervening time. The fact that such a defi nition appears to be adequate can seem to render unproblematic the notion of a movement of no par ticu lar kind. But this is a mistake. The fact that we can accurately report the occurrence of the event by saying “an object moved from A to B” no more entails the existence of pure movement than it does the existence of pure objects, i.e., objects of no particular kind. The notion that a ball’s movement from A to B is something distinct from, say, its fall from A to B is completely implausible. What makes a ball’s fall from A to B a fall is the fact that the movement has a certain explanation—viz., that it was perpetuated merely by the ball’s own weight. Any movement that can be explained this way is a fall. The ball’s supposed pure movement would have to be a movement that could not be explained this way. But it is also not a movement that could be explained any other way, in any of the ways suitable for the other sorts of just-listed movement-types. For then it would be an instance of one of those types and not a pure movement after all. If, e.g., the ball’s moving from A to B could be explained by the continued force exerted by a contiguous object, then it would be a being pushed from A to B. The idea of a pure movement is a myth because there are no unexplainable movements. These reflections suggest that the relation between statements such as “The ball moved from A to B” and “The ball fell from A to B” is analogous to the relation between “There is an object on the table” and “There is a ball on the table.” (Just as object is not a true substance sortal, movement from A to B is not a true event sortal.) A ball’s being on the table guarantees the truth of “There is an object on the table” not by guaranteeing the existence of a distinct ‘pure’ object on the table but simply by being an object on the table. Similarly, a ball’s falling from A to B makes “The ball moved from A to B” true not by guaranteeing the occurrence of a ‘pure’ movement but by being a movement. Similarly, action is not the cause or the causing of a pure movement—it can’t be, since there is no pure movement—but rather a form of movement, like falling or being pushed. If someone asks, “why did the foot movei from A to B?,” “he movedt it” is illuminating insofar as it conveys that he moved

Physicalism

245

the foot himself—it did not fall, it was not nudged, etc. This answer conveys that the movement was an action. The movement i of the finger has been illuminatingly reclassified as a movement t. This illuminating reclassification can, however, be mistaken for a causal explanation, thus giving rise to the false impression that the agent or something the agent did caused the pure movement, and thus that the action is one thing, and the movement i is its upshot. My claim, then, is that Jones’s crossing Columbus Avenue and the chicken’s crossing Columbus Avenue are different forms of movement. It might be countered that fallings are just pure movements characterized in terms of their causes, that the idea of a falling is that of pure movement plus some externally related cause. But what explains why a certain movement is a falling is not an externally related cause. Causes of falling include such factors as having been dropped or tripped or losing support. It is a mark of their being causes of movement that no par tic ular event of falling is necessarily linked with any par tic u lar cause. But the relation between a movement that is a fall and its having a fallingappropriate explanation is constitutive. To be a fall, it is constitutively necessary that the relevant object’s motion is perpetuated merely by its own weight. One cannot separate, in thought or in reality, a specific fall from the movement-perpetuation that characterizes fallings as such. More generally, any par tic u lar movement is inseparable from how the movement transpires. Why does nothing count as an action that an agent is not thereby in a position to just say she is doing? Because action is thought— even when, as is typical, it is also a material object moving through space. Actions are marked off from other events (including other movements) by virtue of the fact that they are the sorts of events that are premises and conclusions of practical reasoning. As such, they are distinguished by the fact that they have rational explanations, i.e., by virtue of the way they are enmeshed in rational causation. Action is (typically) an object’s moving through space according to rational-causal principles. Let’s return to the issue of event-identity. A principle of identity determines whether some event a that was occurring at t1 is the same as some event b that was occurring at t2. If a = b, there is a subsuming type, something that answers the ‘same what?’ question. Since an event is, in the

246

Rational Causation

simplest cases, an object’s reaching a telos, we can represent it using a statement of the form ‘o f -ed.’ The question of whether a is identical to b is just the question of whether a is the same f -ing as b. If I am told by Linda that Jones sheepishly crossed Columbus Avenue this morning and by Eleanor that Jones quickly crossed Columbus Avenue this morning, I might wonder whether Jones’s sheepish crossing was the same as his quick crossing. Quick crossing of Columbus Avenue and sheepish crossing of Columbus Avenue stand to crossing of Columbus Avenue as determinates to determinable. They all involve an object’s reaching the other side of Columbus Avenue. This supplies a condition that any event that is an instance of one of these types must meet. But the relevant condition does not just require the arrival of the object at its destination. If Jones is hurled across the street, the result is not the occurrence of an event of the type crossing of Columbus Avenue. To be an instance of this type, an object must arrive at the other side by way of intentional action. Similarly, for a suitcase to instantiate the sortal fall from the top of Haley Center is for it to reach the ground by way of gravity and not, say, by dumbwaiter or by Jeeves’s lugging. A basic necessary condition for a pair of events being instances of the same type is their being manifestations of the same causal order. Crossing of Columbus Avenue does not overlap with any physical event sortal because they are manifestations of different kinds of causation. Physical eventtypes are bound to have different principles of identity than action-types. What makes it true that a physical event has occurred is that an object has reached a certain telos by way of efficient causation. What makes it true that an action has occurred is that an agent reaches a certain telos by way of rational causation. There is thus no type that subsumes both a particular action and a token physical event. Hence actions are neither type- nor tokenidentical to physical events.

2.2 Mental States and the Doctrine of Token-Identity Many simply assume that mental states are token-identical to physical states. Virtually everyone thinks that the doctrine is an intelligible middle ground between the view that mental state-universals are identical to physical

Physicalism

247

state-universals and the view that mental and physical states are distinct tout court.26 But it is not. For, as the argument of Chapter 5 shows, there are no token states; more carefully, there are no things that are both rightly called states and in the category of particulars. The type-token distinction, as it is generally understood by philosophers of mind, has no application to states. Token states are a philosopher’s fiction. This is not to say that we cannot debate the question of whether mental states are identical to physical states. We can. But there are no such particulars as individual beliefs, desires, and intentions about which we can then ask whether they are identical to individual physical states. States are universals, not particulars. And mental states are neither identical to nor realized by physical states. Or so I will now argue.

2.3. Mental States Are Not Physical Physicalism about states is the view that mental states are identical to or realized by mental states. Switching to the preferred term in these discussions: mental properties are identical to or realized by physical properties. One popular conception of property-individuation is causal: property P is identical to property Q if and only if P and Q “contribute the same causal powers” to the objects that exemplify them. To say that a property contributes a causal power to an object is to say, roughly, that an object is disposed to have certain effects because it possesses the property. Property P realizes property Q if and only if Q contributes to objects that possess Q a subset of the causal powers contributed by P.27 Mental properties, 26. For exceptions see especially Hornsby, Simple Mindedness, and Helen Steward, The Ontology of Mind (Oxford: Oxford University Press, 2000), but also Daniel C. Dennett, The Intentional Stance (Cambridge, MA: MIT Press, 1987), 52; Collins, The Nature of Mental Things; David Hunter, “Mind-Brain Identity and the Nature of States,” Australasian Journal of Philosophy 79 (2001): 366–376; and, in a different vein, William Robinson, “States and Belief,” Mind 99 (1990): 33–51. 27. These conceptions of property identity and property realization are employed in William Lycan, Consciousness (Cambridge, MA: MIT Press, 1987); Ernest LePore and Barry Loewer, “More on Making the Mind Matter,” Philosophical Topics 18 (1989): 175–191; Kim, Mind in a Physical World; Lawrence Shapiro, “Multiple Realizations,” Journal of Philosophy 97 (2000): 635– 654; Sydney Shoemaker, “Realization and Mental Causation,” in Physicalism and its Discontents, ed. C. Gillett and B. Loewer (Cambridge:

248

Rational Causation

I argue, are neither identical to nor realized by physical properties in this sense. I will refute the identity thesis by refuting the (weaker) realization thesis. A person’s physical properties do not contribute, in the intended sense, a superset of the causal powers contributed by someone’s believing, desiring, or intending what they do. The idea is that the putative efficient-causal profile of an instantiated mental property is just a fragment of the fuller efficient-causal profile of the instantiated physical property. But the efficacy of mental states must be understood in terms of rational causation. Even in cases in which people act or believe in light of their physical properties, they do not act or believe because, in the efficient causal sense, of their physical properties. An agent’s physical properties do, of course, explain the agent’s causing (in the efficient sense) certain physical events and states. For example, it is because Benjamin weighed more than fifty pounds that the rock bounced off him, as opposed to knocking him over. But an agent’s power to rationally bring about actions is not a matter of efficient causation at all. Let us put aside, however, this book’s central theses for the moment and consider in its own terms an argument that the power of a mental property to bring about an action just is the power of some physical property to bring about a change in a physical object. The argument that I envisage assumes neither that persons are identical to their bodies nor that actions are identical to physical events. If successful, it would establish that the causal contributions made by a mental property can be fully accounted for in terms of the causal contributions made by physical properties. Suppose S was f -ing because she wanted to y. The physicalist thinks the rational causation described by an instance of that schema is realized by the efficient causation described by this schema: o was e-ing because o had property P, where o is a physical object, e-ing is a physical event, and P is a physical property. This physicalist supposes that the presence of some o was (in the right context) sufficient for the presence of S, that o’s possessing some P (in the right context) is sufficient albeit not necessary

Cambridge University Press, 2001), 74– 98; and Carl Gillett, “The Metaphysics of Realization, Multiple Realizability, and the Special Sciences,” Journal of Philosophy 100 (2003): 591– 603.

Physicalism

249

(on account of the ‘multiple realizability’ of mental properties) for S’s wanting to y, and that o’s e-ing (in the right context), for some e, is a sufficient but not necessary condition for S’s f -ing. Assuming that there are physical objects, events, and properties that stand in these relations to people, actions, and their intentional states, the argument proceeds as follows. The causal power of S’s wanting to y in relation to S’s f -ing reveals itself in counterfactuals such as this: if S had not wanted to y, then S would not have f -ed. And, according to this line of thought, that counterfactual statement is true because this one is true: if o had not possessed P, then o would not have e-ed. According to this strategy, then, the reality behind any specific instance of a rational causal connection is just whatever makes a certain non-rationally articulable counterfactual statement true. If the agent-necessitating physical object had not been in the desire-necessitating physical state, then the actionnecessitating physical event would not have occurred. Since, presumably, it is physical laws that explain the truth of such counterfactual statements, the rational causal connection is, after all, just a manifestation of those same physical laws. The suggestion requires that we take a lot on faith: e.g., that there are P-properties and e-events and o-objects that stand in the right relations to every person, action-type, and mental state and that they stand in matching causal relations with one another. The central reason for thinking that there must be some such properties, events, and objects has always been that it provided the best way of accounting for the causal role of mental states in the etiology of action. Since we have already pulled the rug out from under this rationale, it is no longer clear why we should take the suppositions it motivates seriously. But even if we do, the argument still does not work. For the counterfactual significance of “S was f -ing because she wanted to y ” is not captured by “if o had not possessed P, then o would not have e-ed.” Since o’s possessing P is not necessary for S to want to y and e-ing is not necessary for f -ing, then we can suppose that S might have wanted to y even if o had lacked P, and that S might have been f -ing even though o was not e-ing. If o hadn’t possessed P and yet S had wanted to y, then, it seems plausible to suppose, S would still have been f -ing. Thus there is a counterfactual tie between S’s wanting to y and S’s f -ing that is indepen-

250

Rational Causation

dent of the counterfactual tie between o’s possessing P and o’s e-ing. If Jones had wanted to talk to his friend, then even if his underlying physical state had been different (if, say, his brain had been ‘wired’ a bit differently), he would have crossed the street, though perhaps slightly differently. The counterfactual tie between Jones’s mental state and his action in this sense goes beyond the counterfactual tie between the underlying physical state and event.28 At this point, the physicalist might suggest that the causal contribution that P makes to S’s f -ing consists in its being one of many possible ‘realizers’ of S’s wanting to y. The counterfactual significance of “S was f -ing because she wanted to y ” is captured not by “if o had not possessed P, then o would not have e-ed” but something more like this: “if o had not possessed one or another of the possible wanting-to-y-necessitating physical properties, then o would not have been the subject of one or another of the f -necessitating physical events.” I cannot see how the identification of the causal power of a mental property with that of a disjunction of this sort would count as any kind of vindication for physicalism. For these disjunctions have no unity from the perspective of physical theory. They are open-ended disjunctions, entry to which is priced in the currency of ordinary psychological rather than physical concepts. They include basically whatever it takes to make sure that all of the exemplifications of the relevant mental properties have something physically in common and are in that sense entirely ad hoc. The order that counterfactual relations between such disjunctions reveal in the causal structure of the world is not physical order at all but rather rational order.29 It is important, I would think, to being a physicalist that one holds that physical theories and explanations ‘carve nature at the joints’: it is only by describing the world using the resources of the physical sciences that we can limn the nature of reality. For it is only the physical sciences that re28. I take Yablo to make a similar point, albeit in different language, in “Mental Causation.” 29. Kim makes the same complaint, though to very different effect, in “Multiple Realizability and the Metaphysics of Reduction,” Philosophy and Phenomenological Research 52 (1992): 1–26. For reply see Ned Block, “Anti-Reductionism Slaps Back,” in Philosophical Perspectives, vol. 11, ed. J. Tomberlin (Boston: Blackwell, 1997).

Physicalism

251

veal true causal order, and there is no way of being in the world that is not a way of being part of its causal order. It is the thought behind this slogan that gives impetus to the attempt to discover how the (apparently) nonphysical properties that we would like to find a place for in the world are realized by (or identical to) physical properties. But in attempting to substantiate the realization thesis, one is led to that aspect of the causal structure of the world that is populated by mental facts and their constituent properties. The proponent of property-physicalism attempts to find a foothold there for properties whose discovery and investigation is the job of the physical sciences. To gain that foothold, however, she must see the reality of mental properties as consisting only in grossly contrived physical properties that play no role in scientific explanations. These properties are physical in the sense that they are disjunctively composed of only physical properties. But they are rescued from obscurity in the infinity of disjunctive physical properties only by dint of our outlining the shadows cast by mental properties. The disjunctive properties, as such, are reached only by human language through applications of ordinary psychological concepts together with technical philosophical machinery. These properties, as such, play a role neither in scientific explanations nor in the ordinary practice of understanding and explaining people and their actions. They are of use only to physicalists anxious to find a way out of the epiphenomenalist or eliminativist corner in which they have painted themselves. But we should not let the physicalist paint us into this corner. In typical cases, counterfactuals of the form “if S had not wanted to y, then S would not have f -ed” are true in virtue of the fact that S inferred the to-be-done-ness of f -ing from the to-be-done-ness of y-ing. Rational counterfactuals are explained in terms of the exercise of rational abilities. As I have argued, to say that S’s wanting to y causes S’s f -ing is not to say the same thing about the relation between S’s wanting to y and S’s f -ing as we say of the relation between a match being struck and its lighting. Given the distinction between rational and efficient causation, the very idea that physical states make the same causal contributions to the objects that possess them as rational states do is a non-starter. Nonetheless, we have entertained one possible route for showing that such contributions might overlap: if the counterfactual significance of rational causal statements could somehow be captured by physical counterfactuals, which

252

Rational Causation

in turn could be explained by physical laws. But now we have seen that such a strategy is burdened with questionable assumptions and methods and achieves a result that approximates success only by violating the spirit of the physicalist doctrine it aims to support. Mental properties, I conclude, are neither identical to nor realized by physical properties. And as we shall now see, this shows that mental facts are not identical to physical facts.

2.4 Mental Facts Sometimes physicalism is framed as a doctrine about the identity of physical and mental facts.30 Are all of the mental facts also physical facts? To answer this question, we must first observe that the category of fact is sui generis. Facts are neither event-tokens nor types: one can sensibly speak neither of a fact occurring nor of one having occurred, nor is a fact a type whose instances are occurrences. Facts are also neither objects nor properties. A simple fact ‘involves,’ in a sui generis sense, an object’s having a property or figuring in an event or process. The identity of physical and mental facts thus turns on the identity of fact-constituents: mental and physical particulars, properties, and events. A physicalist might object here that a mental particular’s exemplifying a mental property or figuring in a mental event might be identical to a physical particular’s exemplifying a physical property or figuring in a physical event, even if the particulars and the properties or events are distinct. It’s unclear how this could be so. Consider a pair of distinct properties P and Q. To say that these properties are distinct is to say, among other things, that what it is for an object to be P is different than what it is for an object to be Q. But one cannot maintain this while also insisting that an object’s exemplifying P could be the same as its exemplifying Q. Similarly, to say that f -ing is distinct from y -ing is to say that what it is for an object to be f -ing is different from what it is for an object to be y-ing. An object’s

30. The Knowledge Argument, as traditionally understood, relies on this conception of physicalism. See Frank Jackson, “Epiphenomenal Qualia,” Philosophical Quarterly 32 (1982): 127–136. Some, however, take physicalism about facts to depend just on the supervenience of mental upon physical facts. I discuss supervenience in the next section.

Physicalism

253

f-ing could thus not simply be the same as its y-ing. Mental events and properties are not physical, so mental facts are not physical facts. In this section I have sketched a metaphysics of mind according to which the mental is distinct from the physical across a broad range of categories. In the next section I argue that this view—together with its broadly antiphysicalist upshot—is consistent with the mental’s depending on the physical in a weak sense.

3. Supervenience I have up to this point postponed discussion of what many take to be the central tenet of physicalism, the supervenience of the mental on the physical: “no mental difference without a physical difference.” Supervenience physicalism is, according to Lewis, minimal physicalism.31 He has in mind something like the following relatively simple formulation, which is minimal insofar as it is accepted by everyone who calls herself a physicalist: “Any world which is a minimal physical duplicate of our world is a duplicate simpliciter”—Basic Global Supervenience.32 I argue here that my view is consistent with this thesis and moreover that there is no independent reason for accepting a stronger supervenience principle that is not consistent with my view. “No mental difference without a physical difference,” far from implying that mental events, states, etc., are identical to their physical counterparts, suggests distinctness. It suggests, that is, that there can be no difference of one sort without a difference of another sort. And recall that Davidson originally invoked supervenience to show that one could be a physicalist about the mental without being a reductionist. My view is just another example of a non-reductionist view that is consistent with Basic Global Supervenience.

31. David Lewis, “New Work for a Theory of Universals,” Australasian Journal of Philosophy 61 (1983): 343–377. 32. Frank Jackson, From Metaphysics to Ethics (Oxford: Oxford University Press, 1998), 12.

254

Rational Causation

It is easy enough to dispel the appearance of inconsistency: Jones’s crossing Columbus Avenue is not identical to any physical event, but had he not crossed the street, things would not have been physically just as they were. The atoms out of which his body is composed would not have been where they in fact were. Jones’s wanting to speak to his friend is not a physical state, but had he not wanted to talk to his friend, the atoms composing his brain may well not have been organized in precisely the same way. The fact that Jones’s crossing the street is a rational consequence of the fact that he desires to speak to a friend does not entail or even suggest that in a physically identical world, he might have crossed the street for some other reason or for no reason. It is thus not obvious why the existence of rational causation should be thought to threaten Basic Global Supervenience. But the supervenience of the mental on the physical is sometimes thought to drain the mental of any efficacy of its own. If this ‘draining away’ concern were legitimate, then my view would indeed be at odds with supervenience. Here, in brief form, is how the worry might arise. Mental events and states have physical effects. According to Completeness, these physical effects ‘already’ have complete physical causal histories. If that is so, mental events and states must themselves supervene on this physical causal history. Furthermore, the efficacy of mental events and states must consist just in the efficacy of the physical events and states upon which they supervene. Recall that in section 1 I argued that because mental causation is of a different sort than physical causation, Completeness is no threat to the efficacy of the mind. I tried to make this fact plain by showing the bizarre result of trying to remove the alleged threat by picturing mental causes as filling in the gaps in physical causal chains. Here we see that the threat seems to survive so long as we think of mental causation as culminating in physical events. It is worth being clear about why (and whether) we should think that beliefs, desires, and intentions have physical effects. Here’s one reason for thinking we should: people can perform actions describable in physical terms. People can move sofas, even individual atoms, because of what they want or believe or intend. Isn’t this sort of moving of an atom a physical event? No. It’s an action, the manifestation of a rational practical power. Its being such a manifestation is constitutive of its being the form of movement that it is. An atom’s being ‘pushed’ by an electric field from A

Physicalism

255

to B is a physical event. It has physical causes— complete physical causal histories. It is a manifestation of physical laws. Just because we use physical terms, such as ‘atom,’ to characterize an action does not mean that the action is physical. For whether an event is physical is a matter of its exemplifying distinctively physical causation. This is all in line with the argument of section 2.1. I argued there that actions are not identical to physical events, for they are examples of a different kind of causation. Still, it might be held that beliefs, desires, and intentions must have physical effects if they are rationally to cause actions, even if these effects are not manifestations of rational causation. After all, for a person to have moved an atom from A to B, the atom must, one might think, have moved (in some determinate, physical sense) from A to B. To be a cause of the former just is to be a cause of the latter. And so any view that is inconsistent with the physical efficacy of rational states must be false. If beliefs, desires, and intentions cannot have physical effects, then they cannot really cause actions. Hence, they must cause physical events. Even if that is so, however, it would not lend any support to the idea that the rational causal power of mental events and states is drained away by their supervenience bases. At best it would show only that their physical causal power (i.e., their power to cause physical events) is drained away by their supervenience bases. But there is, in any case, no reason to accept the inference from “S’s belief that q is a rational cause of her f -ing” and “S’s f -ing entails the occurrence of physical event p” to “S’s belief that q is a cause of p.” That inference is no better than this one: “The cold weather is a cause of the Miami Dolphins’ defeat” and “the Miami Dolphins’ defeat entails the earth’s continuing to revolve around the sun,” therefore “the cold weather is a cause of the earth’s revolving around the sun.” If the former inference seems to be more plausible, it is only because (1) we tend to conflate actions and physical events— and so the causing of an action just seems the same as the causing of a physical event— and (2) we tend to conflate efficient and rational causation— and so the only way of causing the action would seem to be by efficiently causing something. Perhaps the most compelling reason for thinking that mental states have physical effects is that the actions they cause can have physical effects. I might, for example, cause a man (and the molecules of which he is composed) to fall into a ravine by pushing him off a cliff. Let’s grant, for

256

Rational Causation

the sake of argument, that the mental states that rationally cause my action are also causes of that action’s physical effects. Still, it is not true that they have those effects in virtue of the efficacy of their supervenience bases. Rather, they have them in virtue of their own efficacy, insofar as they are rational causes of action. This would mean that a physical event does have (rational) causes over and above its complete physical causal history. But there should no longer be any temptation to view the efficacy of rational causes as consisting in that of the underlying physical processes. Here is another strategy for trying to capture the concern that the supervenience of the mental on the physical entails the draining away of mental causal powers, a strategy that mirrors the physicalist argument for property-realization discussed in section 2.3. If the physical facts about the world entail the mental facts, then things are the way they are mentally because of the way they are physically. S’s f -ing may be occurring because (in the rational-causal sense) S wanted to y, but the action also is occurring because (in the supervenience sense) the world is physically laid out in the way it is. And those physical facts about the world that entail (on account of some variety of supervenience) the occurrence of the action are, it might be suggested, efficient effects of those physical facts about the world that entail (again, on account of some variety of supervenience) the agent’s wanting to y. (I am supposing for the sake of this argument that causal relata are facts. The argument could be reformulated, albeit more torturously, in terms of events.) But even if this suggestion is correct, it simply does not provide, without further argument, any reason for thinking that the rational relation between S’s wanting to y and f -ing is a casualty of supervenience. An action may supervene on physical events that are themselves efficient effects of earlier physical events, yet it may nonetheless occur for the sake of y-ing. Supervenience, even of this more local variety, does not by itself provide grounds for denying that it was with a view to y-ing that S was f -ing—that it was with a view to something else or nothing at all. The rational etiology of action no more requires violations of supervenience than it does violations of Completeness. Doing something for a reason, on my view, does not entail or even make remotely likely that there could be physically identical circumstances that are mentally different.

Physicalism

257

Since my view is consistent with supervenience, some may charge that I am, after all, a physicalist. But charges such as “your view is too physicalist”— or its opposite number, “your view isn’t physicalist enough”—are, in light of the fact that ‘physicalism’ is a purely technical term that’s been assigned many different meanings, simply vapid. Furthermore, these sorts of complaints tend to be motivated by the assumption that views of the mind should be categorized solely on the basis of how they portray the relation between the mental and the physical—a physicalist assumption this book aims to challenge. To the extent that I can tap into the animating spirit behind the quest for the ‘correct’ defi nition of physicalism, I would judge that my view is quite hostile to it: rational beings are the subjects of events that are not physical, for these events do not instantiate physical event-sortals. Rational properties, such as beliefs, desires, and intentions, are not physical properties, for their ‘causal contributions’ to rational beings are neither identical to nor a subset of the causal contributions made by any physical properties. Rational facts are not the same as physical facts, since they have distinct constituent properties and/or events. Rational causation is sui generis and is fundamentally different from physical (i.e., efficient) causation. I describe my view as anti-physicalist in virtue of these facts about it. It is true that my view is not dualist on the standard meaning of the term. For a dualist rejects Completeness and perhaps Basic Global Supervenience as well. This does not show that I’m ‘really’ a physicalist after all but rather only that dualism does not exhaust the alternatives to physicalism in the philosophy of mind. Of course to say that my view is consistent with Basic Global Supervenience is not to say that it is compatible with every variation on the basic theme of supervenience.33 For example, some have argued that a truly physicalist view must explain in physical terms why the relation of supervenience holds (‘superdupervenience’34) or must apply the basic principle

33. For seminal essays on the varieties and significance of supervenience, see Jaegwon Kim, Supervenience and Mind (Cambridge: Cambridge University Press, 1993). For some more recent essays, see Elias Savellos and Umit Yalcin, Supervenience: New Essays (Cambridge: Cambridge University Press, 2007). 34. See Horgan, “From Supervenience to Superdupervenience.”

258

Rational Causation

of supervenience to individuals,35 or to space-time regions,36 rather than simply to whole worlds. But my project does not hinge on consistency with stronger forms of supervenience. Stronger versions of supervenience are sometimes formulated with an eye toward ruling out far-out beings or circumstances. But if my commitments turn out to be inconsistent with one of these stronger formulations, it has nothing to do with my interest in defending ‘epiphenomenal ectoplasm.’37 Nor is it my job to formulate a supervenience-thesis that rules out epiphenomenal ectoplasm without ruling out my own view. The point of a search for a more stringent formulation is typically just to articulate a requirement the demonstrated satisfaction of which would amount to the reduction of everything mental to the physical—to the naturalization of the mind. Since the point of this book is to call that ambition into question, it can hardly be counted against my view that it violates a supervenience principle that is thusly motivated. Some formulations of supervenience may be inconsistent with my view, but they do not provide any independent reasons for rejecting it. The current popularity of physicalism cannot be understood on the model of a consensus about any specific thesis. For its popularity far outstrips any agreement about how precisely it should be understood. This makes it a frustratingly elusive target. Since a physicalist is not typically committed to any particular statement of physicalism, no argument against a particular thesis can defeat it. What I called above ‘the animating spirit of physicalism’ does not draw its power from any one master argument, so it cannot be put to rest by refuting a specific argument. There is a well-entrenched tendency to view any position that is “less physicalist” in some respect than another as being dualist in that respect. And our formative philosophical experiences—typically, as students in introduction to philosophy classes— have left most of us with a not entirely rational fear of being seen with any member of the dualist family, no matter how distantly related.

35. This is how Davidson originally framed the principle. See “Mental Events,” 214. 36. See Terrence Horgan, “Supervenience and Microphysics,” Pacifi c Philosophical Quarterly 63 (1982): 29–43. 37. Ibid.

Physicalism

259

In light of the challenges facing the anti-physicalist, I have adopted a multipronged approach. I have attacked what I take to be a central motivation behind physicalism: the mistaken assumption that every worldly causal explanation is made true by efficient causation. All of the major arguments in favor of physicalism are based on this assumption. There are, however, other forms of causation, one of which I have explored in this book. I have also defended a non-dualist metaphysics of mind according to which the most prominent physicalist theses are false. In the process, I hope to have sown seeds of doubt among the physicalist faithful and perhaps to have caught a few still-impressionable minds unaware.

Acknowledgments

I was profoundly influenced by John McDowell’s classes at the University of Pittsburgh in the 1990s. The “space of causes,” he taught, straddles the non-overlapping “space of reasons” and “space of laws,” a metaphor that fits the central thesis of this book. My conception of the point and nature of philosophy was fundamentally shaped by classes and conversations with Jim Conant. I was lucky enough to have the opportunity to talk through many of my initial forays on these topics with my mentor and advisor, John Haugeland. It was only years after my initial encounters with Michael Thompson that I came to realize the bearing of his work on my own projects, but its impact, especially on Chapter 2, is, if anything, greater than what the footnotes convey. Sebastian Rödl visited Auburn University in 2005 and presented material from his then forthcoming book, Self- Consciousness; in it I discovered insights that sparked some of the central themes of Rational Causation, especially the first two chapters. Many friends and colleagues were kind enough to read and discuss portions of the book. Thanks to Matt Boyle, Carl Craver, Anton Ford, Keren Gorodeisky, Matthias Haase, Arata Hamawaki, Kelly Jolley, Doug Lavin, and also to Ram Neta, without whose encouragement I might never have set out to write Rational Causation. The input of Lindsay Waters and two superb anonymous referees at Harvard University Press led to substantial improvements. And I owe a special debt of gratitude to Guy Rohrbaugh, whose generous comments on every draft dramatically strengthened the final product. I am grateful for permission to use revised portions of some previously published work. Chapter 5, section 1, consists partly of material adapted from “Events, Sortals, and the Mind-Body Problem,” Synthese 150 (2006): 99–129. A few paragraphs in Chapter 6, section 1.3, are modified from “Mental Causation in a Physical World,” Philosophical Studies 122 (2005): 27–50. Passages from both are reprinted here with kind permission from Springer Science + Business Media. Chapter 5, section 2.1, contains material culled from “Why There Are No Token States,” Journal of Philosophical Research 34 (2009): 215–241; thanks to the Philosophy Documentation Center.

Index

Accidie, 80 Accomplishments, 215, 220 Achievements, 219–220 Acting-for-a-reason, 7, 65–115; and Equivalence Thesis, 101–107; evaluation identified with acting, 86– 92; evaluation identified with intending, 79– 86; and expressibility of actions and reasons for action, 75–79; objections to account of, 75– 92; as practical thought, 68–75; and Psychologism, 100–115; and teleology, 92–100 “Actions, Reasons, and Causes” (Davidson), 3, 13, 145, 149, 151, 162–163 Action-types, 72n4, 238–246 Akrasia, 80, 85 Animal agency, 117, 137–138. See also Non-human animals Anomalous Monism, 223, 229–230 Anscombe, Elizabeth, 7, 66, 66n1, 68, 70–71, 73n6, 138, 153 Argument from illusion, 2, 46, 46n27, 114 Armstrong, D. M., 151, 225, 228 Aspect, 94– 96, 175, 196–202, 213 Assertion, 18–24, 23n6, 27 Atelic character of processes, 215 Bar- On, Dorit, 22n4, 25n9 Behaviorism, defi nitional, 229 Believing-for-a-reason, 14– 64; assertion differentiated from demonstration, 20–21; demonstrations as expressions of, 18–35; dispositional character of, 51– 60;

epistemic character of, 6, 17, 45–50, 59; and Expressibility Thesis, 29–36; and mind-citing explanations, 42–50; and Moore’s Paradox, 18–20, 24–28; not a process, 28; and Psychologism, 61– 64; and rational abilities, 50– 61; and variant of Moore’s Paradox, 18–20, 24–28; and world-citing explanations, 36–42 Bennett, Jonathan, 157 Bermudez, J. L., 132 Biological functions, 92– 93 Bird, Alexander, 179–181 Boyle, Matthew, 43, 81 Causal completeness of physical realm, 223, 231–237, 231n10, 254 “Causal Relations” (Davidson), 154 Causal understanding, 160–164 Causation, 145–182; and the analysis of mental concepts, 225–229; Davidson on, 151–160; explanation’s relation to, 154–164; and explanatory connections, 158; kinds of, 12, 164–167; and movement, 245–246; Nomological Character of Causality, 223; and rational explanation, 146–167; Rylean conceptual analysis, 148–151; singular causal claims, 154, 173; synthesizing causal concepts, 146–148. See also Rational causation Causation and Counterfactuals (Collins, Hall, and Paul), 146 Changes, events as, 239 Collins, Arthur, 108

264

Collins, John, 146, 164 The Concept of Mind (Ryle), 149 Conditional analyses of dispositionstatements, 52n33, 53n35, 57 Crane, Tim, 234n17 Dancy, Jonathan, 40–41, 40–41n19, 44, 101, 102n31, 104–105, 105n37, 108–110, 109n44, 113–114 Davidson, Donald, 3–4, 13, 63, 91, 101, 145, 151–160, 211–212, 223, 230n9, 253 Davis, Wayne, 105, 109 Defi nitional behaviorism, 229 Deliberation, 34–35, 82– 84 Demonstrations: assertions differentiated from, 20–21; as expressions of believingfor-a-reason, 16, 18–24, 24n8, 28, 36 Descartes, René, 134n18 Desire, 22, 24n7, 111–112, 153 Disengaged deliberation, 34–35, 82, 83– 84 Disjunctive properties, 250, 251 Dispositions; imperfect actualizations of, 55–56; ‘normal,’ 52n32; and rational causation, 179–180; species of, 51–58; underlying believing-for-a-reason, 58– 61 Dissective character of states, 197–202 Doctrine of Pure Movement, 243 Doctrine of Token-Identity, 246–247 Dretske, Fred, 192, 192n20 Dualism, 13, 222, 223, 257, 258 Dummett, Michael, 187, 187n6, 187n8 Dummy sortals, 187, 188 Efficient causation: in causation typology, 166; nomological character of, 174; rational causation distinct from, 3, 170–171, 223, 251, 257; and rational explanation, 163; simultaneous, 44n24 Eliminativism, 13, 177 Engaged deliberation, 34–35, 82, 83– 84 Epiphenomenalism, 4, 13, 177, 258 Equivalence Thesis, 9, 101–107, 102n31, 106n40, 212 Evaluation: identification with acting, 86– 92; identification with intending to act, 79– 86. See also Deliberation Event-nominalizations, 208–210 Events, 183–221; as causes and effects, 156; as changes, 239; events-in-progress, 212–221; event-tokens, 190, 196, 227; event-types, 4, 190–196, 216; principles of identity for, 191–195, 201; principles of

Index

individuation for, 195–196, 200–201; and states, 196–212; as tropes, 191n18; type-dependence of, 191–196, 238–247; unitary character of, 197 Explanation: of action, 65–115; of belief, 14– 64; causal vs. constitutive, 161–162; causation’s relation to, 154–164; explanatory connections, 158; Rylean conceptual analysis, 148–151 Expressibility: of actions and reasons for action, 75–79; of belief and reasons for belief, 22–28; and self-attributions, 22–23 Expressibility Thesis, 29–36; defense of, 29–32; significance of, 32–36 Facts: and acting-for-a-reason, 72–74, 112–114; and believing-for-a-reason, 36–42; and knowledge by non-human animals, 118, 122, 123; mental, 252–253; and propositions, 122–129; and rational causation, 172–174 Falvey, Kevin, 86n19, 88 Finkelstein, David, 22n4, 31n13 Forms of movement, 243–246 Frege, Gottlob, 122, 124, 189 Fumerton, Richard, 50n29 Functional properties, 225–226 Galton, Anthony, 196–197, 198n26, 199 Gibbard, Allan, 75n7 ‘Guise of the good,’ 80, 85n16 Gupta, Anil, 186, 187n6 Habituals, 52 Hall, Ned, 146, 164 Higher Order Thought (HOT) theory, 22n5, 27n11 Holton, Richard, 86n17 Homogeneity of mental and physical causation, 234n17, 237 Hornsby, Jennifer, 164–165, 173, 243 Hyman, John, 37, 37n17, 137, 138 The Idea of a Social Science and Its Relation to Philosophy (Winch), 163 Identity: criterion of identity, 187n6; of events and states, 13; and individuation, 189n14, 189n16; numerical, 204; and physicalism, 226; sortal-dependence of, 185–190, 194, 238; token-identity, 246–247; type-dependence of, 190–196, 238–247. See also Principles of identity

Index

Instantaneous events, 185n3, 196n23, 219–221 Intellectual vs. real connections, 155–156 Intention: acting identified with, 86– 91; and desire, 111–112; evaluation identified with, 79– 86; and fi rst-person authority, 86– 91; Intention (Anscombe), 66, 70–71, 153; and non-human animals, 132; and teleological explanations, 97–100 Jackson, Frank, 51n30 Justification, 39, 152, 153, 159, 163 Kenny, Anthony, 149, 184 Kim, Jaegwon, 157–158, 232, 250n29 Knowledge: and acting-for-a-reason, 66n1, 72– 74, 112–114; and believing-fora-reason, 36–42, 45–49; in non-human animals, 118, 130–134, 137, 139–141; objects of, 122–129; and world- citing explanations, 38, 39, 40–41, 40n18 Knowledge Argument, 252n30 Lavin, Doug, 81, 237n19 Lewis, David, 151, 157–158, 225, 231, 253 Life and Action (Thompson), 92 The Logic of Aspect (Galton), 196–197 Mackie, Penelope, 189n16 Mass-quantified character of states, 207–212, 219 McDowell, John, 84n14, 114n48, 142, 164–165, 230n9 Melden, A. I., 149, 150 Mellor, D. H., 157 Mental events, 235, 238–246, 254 Mental facts, 252–253 Mental properties, 225, 247–252 Mental states, 246–252; causal analysis of, 223; and Doctrine of Token-Identity, 246–247; efficacy of, 230; expressions of, 23; non-physical character of, 247–252; and physical causation, 254–256 Mind-citing explanations: and acting-fora-reason, 67, 100–114; and believing-fora-reason, 15, 36, 36n16, 42–50, 61– 64; and Psychologism, 61– 64, 100–114; and simultaneity constraint, 43–45 Minimal physicalism, 253 Moore’s Paradox, 18–28, 32–35 Motivational Internalism, 80– 84

265

Mourelatos, Alexander, 207–208 Movement, 192, 192n20, 243–246 Negatable nature of states, 205–207 Negative properties, 206 Nomological Character of Causality, 165, 229–230 Non-human animals, 116–144; and agency, 137–138; argument against belief in, 134–136; evidence supporting belief supports knowledge better, 130–134; and explanations of belief, 139–141; and objects of knowledge vs. objects of belief, 122–129; and perception, 141–144; responsiveness to reasons, 119–121 Object-sortals, 185–190 Organic functions, 51–58; imperfect actualization of, 55–56; normativity of, 56; and teleological explanations, 94–100 Pargetter, Robert, 51n30 Parsons, Terrence, 213, 218n43 Particulars, 184, 196–211 Paul, L. A., 146, 164 Perception, 141–144 Persistence, 185, 191–192, 202–204 Phase sortals, 189–190, 192, 193 Physical dispositions, 51–58 Physicalism, 222–259; and causal analysis of mental concepts, 164, 225–229; and causal completeness of physical realm, 231–237; and Doctrine of Token-Identity, 246–247; and mental events, 238–246; and mental facts, 252–253; and mental states, 226, 246–252; and Principle of the Nomological Character of Causality, 229–230; refutation of positions, 237–253; and supervenience, 253–259; undermining arguments for, 13, 224–237 Physical properties, 247–251 Practical knowledge, 66n1, 150 Practical thought, 68– 92 Principles of identity: and negative properties, 206; for objects and events, 185–196, 245–247; and rational explanation, 177; and teloi, 216 Principles of individuation: for events, 195–196, 200–201; for objects, 189; for particulars, 204; and teloi, 216 Prior, Elizabeth, 51n30 Processes, 215–216

266

Properties, mental vs. physical, 225, 247–252 Property realization, 247–249, 247n27 Propositions: attitudes toward, 130–136, 226; and facts, 122–129; and non-human animals, 122, 123, 130–136 Psychologism: and acting-for-a-reason, 100–115, 105n37; and believing-fora-reason, 16, 61– 64; defi ned, 2, 61, 100; and Equivalence Thesis, 9, 101–107; Sophisticated, 9, 105–107; Standard, 106, 110 Rational abilities, 50– 61; imperfect actualization of, 55–56; normativity of, 56; and rational causation, 167–182; and as a species of disposition, 50–58; underlying acting-for-a-reason, 100–101; underlying believing-for-a-reason, 58– 61 Rational causation, 167–182; inferential character of, 174; as manifestation of rational abilities, 167–174; objections to, 174–182. See also Acting-for-a-reason; Believing-for-a-reason Raz, Joseph, 102n31 Realization, 211, 226, 247–252 Realizer functionalism, 225, 228–229 Role functionalism, 225, 226 Rosenthal, David, 22n5 Ryle, Gilbert, 139, 148–151, 184, 215 Rylean conceptual analysis, 148–151 Self-consciousness, 58, 117, 142, 167 Self-deception, 31, 75–77 Setiya, Kieran, 105, 106n40 Shah, Nishi, 33n15 Simple Dispositional Analysis of Causation, 179–180 Simultaneity Constraint, 42–45 Singular causal claims, 154, 173 Smith, Michael, 102

Index

Sortals: dummy sortals, 187, 188; object-sortals, 185–190; phase sortals, 189–190, 192, 193; substance sortals, 189–190n16, 192, 193 State-nominalizations, 208–210 States, 196–212; dissective character of, 197–202; mass-quantified character of, 207–212; mental, 246–252; negatable nature of, 205–207; and persistence, 202–204; present-tense predication of, 198n26 Steward, Helen, 160–164, 170, 172, 173 Strawson, P. F., 154–155, 186, 186n4 Substance sortals, 189–190n16, 192, 193 Supervenience, 211, 223, 253–259 Szabó, Zoltán Gendler, 213, 214, 217n42 Telekinetic view of mental causation, 235–236 Teleological explanation, 8, 92–100, 96n29 Teloi, 202n28, 215–216, 239 Tenenbaum, Sergio, 80– 81, 111 Thompson, Michael, 24n7, 92, 95 Tropes, 158, 191n18 Type-physicalism, 225n3 Velleman, David, 33n15, 103 Vendler, Zeno, 122, 124–125, 127–129, 130, 184, 215, 219, 220 Verb aspect, 94– 96, 175, 196–202, 213 Weakness of the will, 84, 86n17 Wiggins, David, 187, 188n9, 189n16 Winch, Peter, 149, 150, 153, 163 World- citing explanations: and acting-fora-reason, 72–73, 107–108; and believingfor-a-reason, 15, 17, 36–42, 36n16, 42n22, 45, 64; and Psychologism, 61– 64, 100–101; and rational causation, 175–176 Yablo, Stephen, 250n28