Immortal Souls A Treatise on Human Nature (Edward Feser)

 CONTENTS 

Preface 

Part I: What is Mind? 

1. The Short Answer 

2. The Self 

3. The Intellect 

4. The Will 

Part II: What is Body?

5. Matter 

6. Animality 

Part III: What is a Human Being? 

7. Against Cartesianism 

8. Against Materialism 

9. Neither Computers nor Brains 

Part IV: What is the Soul? 

10. Immortality 

11. The Form of the Body 

Index

Preface The title of this book is bound to bring to mind two philosophers who are explicitly mentioned only here and there in what follows, but nevertheless loom large in the background throughout. The first is Plato (427-347 B.C.), whose dialogue Phaedo is the great work on the soul and its immortality in the history of Western philosophy. My longtime readers will not be surprised to find that the names of Aristotle (384-322 B.C.) and Thomas Aquinas (1225-1274 A.D.) appear with greater frequency in the book, and that I favor their position where it differs from Plato’s. All the same, though he was mistaken on crucial matters of detail, it was Plato who first got right the most important things – that the highest part of human beings is the intellect, that the intellect is incorporeal, that this entails that the soul survives death and is indeed immortal, that it will be rewarded or punished after death, and that all of this can be arrived at by philosophical reasoning independently of any special divine revelation. 1 It would be potentially misleading to describe the book’s aim as that of vindicating Plato, but I’ll risk doing so anyway (albeit with the qualifications one would expect a Thomist to make). 2 The truth is dearest, but Plato is still dear to me. The book is also intended to refute the other thinker its title will evoke, namely David Hume (1711-1776). Hume’s essay “On the Immortality of the Soul” is perhaps the most eloquent expression in Western thought of the falsehood that belief in life after death can find no rational support short of a special divine revelation (which, of course, he did not think has ever been given us). Hume’s Treatise of Human Nature, from which I borrow my subtitle, is the most influential source of the errors concerning substance, the self, the intellect, and the will that have led modern man radically to misunderstand his own nature. Clearing away this intellectual rubbish is a prerequisite to establishing that Plato was right and Hume wrong. As my subtitle indicates, the immortality of the soul is far from the only topic to be treated in the pages that follow. Indeed, immortality is addressed only at the end, in the last two chapters. Even the word “soul” will rarely be used until then. That is deliberate. Nothing I have to say in the first ten chapters strictly requires using the word. Also, “soul” has many connotations, not all of them relevant to the topic of a particular chapter, and some of them unfortunate in any case. Rather than repeatedly and needlessly risking misunderstanding and having to make tiresome qualifications, it seemed better to avoid the word until absolutely necessary. This is not the way discussions of the soul usually proceed. But given the word’s ambiguity, my view is that the least potentially confusing approach is first to give as thorough an account of human nature as is possible without using the word, and only then to explain where the notion of the soul fits in. The book is, then, a general treatment of the metaphysics of human nature. It addresses the main philosophical controversies concerning the self and personal identity, the nature of concepts, the relationship between thought and language, the freedom of the will, embodiment, animal intelligence, perceptual experience, innatism versus empiricism, dualism versus materialism, the philosophical implications of neuroscience and computer science, transhumanism, and so on. To be sure, the book is not exhaustive. Certain matters could have been treated in greater depth. For example, much more could be said about the nature and classification of the sensory powers and the appetites. Certain matters are not treated at all. For example, I say nothing about the distinction between the sexes, which is no small part of human nature. (I will treat that at length in a later book.) I had to draw the line somewhere. But the guiding criterion was to address every topic that needed to be addressed in order to defend the sober Aristotelian-Thomistic middle ground position between Cartesianism on the one hand and materialism on the other. 3 For that reason, the book’s length was unavoidable. Because human nature is complex, the ways we can go, and have gone, wrong about it are multifarious. I have tried to refute the errors that are most prevalent today. The book’s eleven chapters are grouped into four parts. Part I addresses the question “What is mind?” Chapter 1 gives a brief overview of what I will argue is the correct answer, which is that a mind is a self or persisting substance having, as its essential attributes, intellect and will. The next three chapters are then devoted to expounding and defending this answer in depth. Chapter 2 defends the reality and irreducibility of the self; chapter 3 explains the nature of the intellect and its irreducibility to any of the powers we share with other animals; and chapter 4 explains the nature of the will and establishes its freedom. Part II is devoted to the question “What is body?” Chapter 5 explains the nature of the material world in general and shows that the Aristotelian theory of form and matter not only has not been refuted by modern science, but is if anything vindicated by modern science. Chapter 6 addresses the nature of living matter, specifically, and in particular discusses what it is to be an animal. The book defends the traditional Aristotelian view that human beings are by nature rational animals, and these first two parts are essentially devoted to explaining what “rationality” and “animality” each amount to. Part III, which is labeled “What is a human being?”, refutes the two main modern misconceptions about the relationship between rationality and animality. Chapter 7 is a critique of Cartesian dualism, which radically sunders the human mind from the physical world, making the body something entirely extrinsic and inessential to us. Chapter 8 refutes materialism, establishing that though we are indeed bodily by nature, we are not entirely so, and that the intellect in particular is incorporeal or non-physical. Chapter 9 refutes currently popular claims to the effect that neuroscience has vindicated materialism, and that the mind is a kind of computer program implemented in the hardware of the brain. The upshot of this part of the book is that a human being is a single psychophysical substance with both corporeal and incorporeal properties and powers. Part IV turns finally to the question “What is the soul?” Chapter 10 argues that, since the intellect is an incorporeal part of a human being, it carries on beyond the death of the body, and that indeed on careful analysis this entails all the main elements of the traditional doctrine of the immortality of the soul. Chapter 11 explains how the position defended in the book relates to the Aristotelian thesis that the soul is the form of the body, and its implications for the origins of the soul and the prospects for its reunion with the body by way of a resurrection of the dead. Again, many other topics are treated along the way. Many of these are also topics I have addressed in earlier work, such as my book Philosophy of Mind, which first appeared almost twenty years ago.4 They are addressed in greater depth here, and in some cases my views have changed. For example, that earlier book is much too sympathetic to Cartesianism, especially in its treatment of perception. The approach of this book is much more consistently Aristotelian-Thomistic. Having said that, though Aristotle and St. Thomas are by far the greatest influences on my own views, this book is not concerned with exegesis of their work. The claims and arguments I defend in this book are mine, and not always necessarily theirs. If, on some topics, I say something that sounds very much like what they say, that is because I think they were right about it. If, on other topics, I say something that is different from anything they say, or even conflicts with what they say, that is because I think my approach is better. Like other Thomists, I am often accused of following Aquinas too closely, but also of not following him closely enough. In reality, I try only ever to follow an argument wherever it leads. I also emphasize, again, that this book is concerned only with what philosophy can, all by itself, tell us about its subject. It is not about theology, highly relevant though it is to theology. Many are bound to be surprised at just how much can be established by purely philosophical arguments where the existence and nature of the soul are concerned – just as, as I have shown in my book Five Proofs of the Existence of God, much can be established by purely philosophical arguments where the existence and nature of God are concerned.5 Immortal Souls, like Five Proofs, is intended as a contribution to understanding the praeambula fidei or “preambles of faith” – that is to say, the philosophical premises in light of which it can be rationally established that a special divine revelation really has occurred (though exactly how that can be established is a topic for another time). In this connection I will acknowledge one final lacuna that I think is unavoidable in a philosophical book with the particular purposes this one has, but has been increasingly palpable to me all the same. I refer to the absence of any treatment of the moral and spiritual ramifications of being a creature with an immortal destiny, yet deeply enmeshed in the material world, with the suffering and liability to death that that entails. The years during which I have worked on this book have been especially dark ones, personally as well as for the world in general. Working on the book has played no small part in my own dealing with this darkness, and it is my fervent hope and prayer that my readers will find in it something that is helpful to them. My usual debts to my beloved wife Rachel and our dear children Benedict, Gemma, Kilian, Helena, John, and Gwendolyn remain, and I thank them for their love and help as I labored over this book. And I’d like to acknowledge a further debt. In the face of the deaths of our father and our sister and the ailments of our mother, my brother Dan Feser has been an exemplar of loyalty and service. I dedicate this book to him, with love and affection.

Part I: What is Mind? 

1. The Short 

Answer What is the nature of the mind? The question is in one crucial respect different from questions we might ask about the natures of other things, such as stones, trees, or dogs. For in this case, the thing asked about and the thing asking are one and the same. This unique intimacy with its subject matter can hardly fail to facilitate answering the question. Indeed, as Augustine (354- 430 A.D.) argued, “the mind knows itself, even when it seeks itself.”6 It can discover something of its own nature merely by reflecting on the fact that it is trying to discover it. What it thereby discovers is that it is a thing that thinks and wills. For the very attempt to inquire into its own nature involves thinking and willing, and in knowing that it is doing these things, the mind knows that it is a thing of the kind that is capable of doing them. If it even tried to doubt this, the exercise in doing so would itself involve thinking and willing. Augustine puts the point by saying that the mind is a thing that knows and loves, and that we can be certain of this insofar as it is presupposed even in the act of questioning it. He elaborates as follows: 

For, we are, and we know that we are, and we love to be and to know that we are. And in this trinity of being, knowledge, and love there is not a shadow of illusion to disturb us. For, we do not reach these inner realities with our bodily senses as we do external objects… But, without any illusion of image, fancy, or phantasm, I am certain that I am, that I know that I am, and that I love to be and to know.

In the face of these truths, the quibbles of the skeptics lose their force. If they say; ‘What if you are mistaken?’ – well, if I am mistaken, I am. For, if one does not exist, he can by no means be mistaken. Therefore, I am, if I am mistaken… Nor, as a consequence, am I mistaken in knowing that I know. For, just as I know that I am, I also know that I know. And when I love both to be and to know, then I add to the things I know a third and equally important knowledge, the fact that I love. 

Nor am I mistaken that I love, since I am not mistaken concerning the objects of my love. For, even though these objects were false, it would still be true that I loved illusions. 7 

As Augustine points out, if the mind were to doubt its own existence, it would have to exist in order to do the doubting. So, its affirmation of its own existence is not something it can be wrong about. This would remain true even if it turned out that other things – such as the physical objects the mind takes itself to see, hear, taste, touch, and smell – were unreal (for example, if the mind were merely dreaming or hallucinating them). 

But more relevant to our present purposes is what the mind can know, not about its existence, but about its nature. In reflecting on its knowledge of its own existence, the mind comes to know that it knows this. It discovers that it is a knowing thing, a thing whose nature makes it capable of knowledge. When it reflects on the further circumstance that it loves the fact of its own existence and loves the fact that it has knowledge of its own existence, the mind thereby discovers that it is also a loving thing, a thing whose nature makes it capable of love. Even if these objects of its love had been unreal, it would nevertheless be a thing of the sort capable of love, insofar as the love itself would still be real even if its objects were not. 

In Augustine’s view, the mind and its capacities for knowledge and love are therefore parts of a metaphysical package deal, each inseparable from the other. He writes: “But just as there are two things, the mind and its love, when it loves itself, so there are two things, the mind and its knowledge, when it knows itself. Therefore, the mind itself, its love and its knowledge are a kind of trinity.”8 This gives us an answer, or at least a partial answer, to our question about the nature of the mind. The mind is a thing of the kind capable of knowing and loving. 

While this answer is stated in the third person, it will probably not have escaped the reader’s notice that Augustine writes in the first person. He talks not only of what his mind knows and loves, but of what he knows and loves, in a manner that implies that these ways of speaking are interchangeable. Of course, that reflects common usage. As the term is usually understood, your mind isn’t merely something you have. Your mind just is you, or at least it is you considered as a thing that knows, loves, and so forth. You might say, then, that to inquire into the nature of the mind is at the same time to inquire into the nature of the self. Thus, to say that the mind is a thing of the kind capable of knowing and loving is to say that you are a thing of the kind capable of knowing and loving. 

Does that entail that there is nothing more to your nature than being a thing of the kind that can know and love? Does it entail that having a body is not also part of your nature? No, those conclusions don’t follow, and we will see later that they are not true. The point for the moment is just to suggest that whether or not you are more than a thing with the capacities to know and to love, you are not less than that. The claim is that Augustine’s identification of knowing and loving as the characteristic activities of the mind or self gives us at least the rudiments of its nature, the bare minimum a thing must have in order to be a mind or self. 

Actually, Augustine’s thesis needs a little tidying up. Or at least, even if the terms “knowing” and “loving” are innocent in themselves, today they have connotations which can make his claims about the mind’s nature seem narrower than they really are. For one thing, Augustine needn’t confine himself to saying that the mind is a thing of the kind capable of knowledge. His argument justifies the broader claim that the mind is a thing of the kind capable of thinking, whether it has a thought of the sort we would ordinarily count as genuine knowledge, or whether instead it has a thought of the sort we would classify instead as a mere belief. For example, when I falsely believe that I see a tree in the distance, it is still the case that I am thinking that there is a tree there, even if I don’t in fact know that there is. 

For another thing, these days the word “love” is commonly taken to connote a kind of emotion, but that is not how ancient writers like Augustine intend it. For these writers, what is essential to loving is to will something as good, and the affective state we associate with love can be absent even when we love something in this sense. Consider, for example, the way you might love an enemy. You would not have warm feelings for him; you might even have very negative feelings when you think of him. But if you willed what is good for him (such as his repentance of the evil he has done to you), you could still be said to love him. 

It would be less misleading for the modern reader, then, if Augustine’s position were formulated instead as the claim that the mind is a thing of the kind capable of willing. 9 When you deliberately carry out the exercise of thinking about whether you exist and whether you have knowledge, you are exercising your will. If you resolve to doubt that you have a will, that would itself be a further exercise of the will. Hence the very attempt to doubt that you have a will itself shows that you have one. So, you can conclude that your nature qua mind is to be a thing of the kind capable of willing, as well as of thinking. 

I would propose, then, that the lesson we can draw from the considerations raised by Augustine can be formulated by saying that the nature of the mind or the self is to be a thing of the kind capable of thinking and willing. To put it in yet other terms, we can say that it is to be a thing possessing intellect and will, where the intellect is the capacity to think and the will is (naturally) the capacity for willing. 

Readers having some familiarity with the history of philosophy will no doubt have been struck by the similarity of what Augustine says to views famously expressed by René Descartes (1596-1650), commonly regarded as the father of modern philosophy. Descartes’ famous Cogito argument (Cogito, ergo sum, which is Latin for “I think, therefore I am”) was intended to show that one cannot coherently doubt one’s own existence. 10 He also held that the mind or self is essentially a thing that thinks, and he included willing as a kind of thinking. But it is useful to make Augustine’s formulation rather than Descartes’ our take-off point, for a couple of reasons. For one thing, doing so helps to underline the point that there is nothing in the thesis that the mind or self is by nature a thing with the capacity to think and will that necessarily entails a commitment to all of Descartes’ distinctive views about human nature. For example, it does not entail that one has to agree with Descartes that mind and body are two distinct substances, or that the body is a kind of machine. Augustine would not have accepted either of those claims, and as we will see later, neither of them is true. For another thing, citing Augustine – a thinker who bridged the ancient and medieval worlds – makes it clear that there is nothing distinctively modern about that aspect of his conception of the mind or self that he shares with Descartes. 11 

Indeed, there is another thinker from Augustine’s era who, I suggest, can also help us to get an initial fix on the nature of the mind or self, namely Boethius (c. 480-524). We owe to him a definition of what it is to be a person that was enormously influential in medieval philosophy. A person, says Boethius, is an individual substance of a rational nature. 12 Let’s break down this definition. In the philosophical sense in which the term is being used here, a substance is to be contrasted with its attributes. 13 If we take some particular stone as an example, its solidity, color, texture, shape, and weight would be among its attributes, whereas the stone itself is the substance that has those attributes. Attributes exist only in a substance rather than in a freestanding way. For example, the color and weight of a stone exist only in the stone and cannot be pulled out and set alongside it, as it were, as separable entities. By contrast, a substance does not exist in another thing in the same sense. It is, again, that which bears attributes. It is precisely the sort of entity in which things of the kind that cannot exist in a freestanding way must inhere. 

What Boethius is saying is that a person is a kind of substance rather than an attribute or collection of attributes. If we think of particular thoughts and acts of will as attributes, then a person is not a collection of such thoughts and acts of will, but rather the substance in which those attributes inhere. Certainly Augustine thinks of a mind or self as a kind of substance. 14 When I first have the thought that I might not exist, and then have the thought that in fact I cannot fail to exist if I am even doubting that I do, Augustine would hold that it is the same one substance that persists from the one thought to the next. 

Aristotle distinguished between a primary substance, which would be an individual thing, and a secondary substance, which would be the species (more specific class) or genus (more general class) into which that thing falls. 15 For example, your pet Collie Fido would be a primary substance, and the species dog into which it falls, as well as the genus Canis into which that species falls, would be secondary substances. 16 When Boethius says that a person is an individual substance, he means that it is a primary substance, a particular thing rather than a class of things. (Of course, all individual persons fall under the class person, just as all individual dogs fall under the class dog. The point is just that any particular person, qua instance of the class, would be a primary substance or individual.) 

What it is to have a rational nature is just to be a thing of the kind possessing intellect or the capacity for thought. Now, as medieval thinkers like Thomas Aquinas argued, anything with an intellect must also possess a will. For a substance with an intellect acts by virtue of being inclined toward what its intellect takes to be good, and the will just is (on this analysis) an inclination of that sort. To be a person in the sense that medieval thinkers inherited from Boethius is, then, essentially just to be a mind or self in Augustine’s sense. 

Combining these ideas, we can say that to have a mind, to be a self, and to be a person are all essentially the same thing, and that what they amount to is being a substance which by nature possesses an intellect and a will. Naturally, this raises many questions, and we will address them in the pages to follow. For the moment let us note what this characterization includes and what it rules out. Non-human animals, plants, and inanimate objects would not count as having minds in this sense, nor as being selves or persons. (That is not to say that non-human animals are not sentient or conscious. They are. It is to say only that they lack intellects and wills, a claim I will defend later on.) Human beings, of course, count as having minds and thus as selves or persons. So too would angels, understood as entirely incorporeal or nonbodily rational substances; intelligent extraterrestrials, if there are any such things; and artificial intelligences or thinking machines, if those are possible. No one denies that extraterrestrial intelligences are at least possible in theory, even if it is controversial whether there really are any. But whether angelic minds or thinking machines really are possible are matters of controversy. We will see later on that in fact angelic minds are possible, whereas thinking machines are not possible. (I am well aware that that is the reverse of the conventional wisdom these days. The conventional wisdom is wrong.) The point for the moment is that the characterization of mind I have sketched out, though it clearly rules out quite a bit, is still broad enough to include, at least in principle, a fairly wide variety of possible kinds of mind.17 

It also leaves it open, at least for the moment, whether a purely material substance could have a mind (as materialists hold) and whether any mind, self, or person would instead have to be an entirely immaterial substance (as Descartes’ form of dualism holds). We will see that both materialists and Descartes are wrong – that a self or person, something with a mind, could be partly material and partly immaterial, but could never be entirely material. 18 But again, for the moment we do not need to settle those issues. 

What we do need to address in this first section of the book are three sets of questions and challenges that arise when considering more closely my proposed characterization of minds, selves, or persons. First, is a mind, self, or person really a substance? Does it really have the kind of stability or permanence that is traditionally attributed to substances? Might it not instead be a mere bundle of ever-changing attributes? Indeed, might a persisting self not be a kind of illusion? Second, exactly what is an intellect, and how does it differ from the sentience we share with non-human animals? Might the apparent rationality and meaningfulness of our thoughts not also be a kind of illusion, as eliminative materialists claim? Third, what exactly is the will, and why would something with an intellect have to have a will? Thinkers like Aquinas hold that the will’s freedom follows from its being a power of a rational substance, but might free will not be an illusion too? 

The answer is that none of these things is an illusion. You can know with certainty that you really are a persisting self, and that you possess rationality and free will. Indeed, none of that can coherently be doubted. The next three chapters are devoted to establishing these claims, and thereby expanding upon the short answer to the question “What is mind?” that I have given in this opening chapter.

2. The Self 

Its reality and irreducibility 

Many debates in the history of philosophy are about whether to accept a realist, reductionist, or anti-realist account of some phenomenon. A realist account of a phenomenon takes it to be real and to have precisely the nature that it appears to have. A reductionist account doesn’t quite deny the phenomenon’s reality, but holds that its true nature is different from what it seems to be. An anti-realist account takes the phenomenon to be altogether illusory, denying that it corresponds to anything in reality. 

For example, consider thoughts, such as the thought that the cat is on the mat or the thought that two and two make four. A realist account of thoughts would say that they really do exist, and that they are, as common sense supposes, very different in their nature from anything bodily – from muscular movements, chemical secretions, neural firing patterns, and the like. A reductionist account would not deny that thoughts are real, but would say that common sense is wrong about their nature. For instance, it might hold that, appearances notwithstanding, a thought really is nothing but a firing pattern in the neurons of the brain. An anti-realist account would hold that there really are no such things as thoughts, odd as that may sound. There are only the processes occurring in the brain, and what we take to be a thought is merely a kind of misperception of what is really going on there. 

Or consider free will. A realist account would say that free will is real, and that, as common sense supposes, our having it entails that what we do is up to us and not determined by forces outside our control. A reductionist account would not deny that free will is real, but would hold that its nature is different from what common sense assumes it to be. In particular, it might claim that an action of yours is free as long as you did it because you wanted to do it, even if what you wanted to do was not up to you and was entirely determined by forces outside your control. An anti-realist account, meanwhile, would flatly deny that there is any such thing as free will. 

We will have reason to consider the topics of thought and free will in detail in the pages to come, but for the moment I cite them merely as examples meant to illustrate the differences between realist, reductionist, and anti-realist approaches to a phenomenon. I should also note that we could make various further distinctions between approaches (for example, between different specific ways of spelling out a realist or reductionist account of the nature of some phenomenon). But our threefold distinction is sufficient for present purposes. 19 

Now, the self is another phenomenon to which these three approaches could be taken. Recall, from the previous chapter, the distinction between substances and their attributes. Whereas attributes exist only in a substance, a substance does not in turn exist in anything else in the same sense. For example, the color and weight of a dog don’t exist apart from the dog, in a freestanding way, but only in the dog. But the dog itself does not exist in anything else in the same sense in which its color and weight exist in it. A substance is thus what stands under or grounds attributes. It can also persist through the gain or loss of an attribute. For example, one and the same dog can at one time be black and at a later time gray, or at one time weigh twenty pounds and at a later time weigh thirty pounds. 

A realist view of the self holds that the self really exists, and that it has the nature that common sense takes it to have – the nature of a substance that underlies attributes such as intending to become a comic book artist or experiencing pain in one’s leg, and that persists through changes of those attributes. For example, when I was a teenager I intended to become a comic book artist, but a few years later decided to become a philosopher instead. And last night I experienced pain in my leg, but today it is gone. A realist view of the self would say, with common sense, that there really is some thing – namely, me – that underlies these different thoughts and experiences, and that it was the same one thing, the same self or person, that had the pain last night but lacks it now, and had the aim of becoming a comic book artist decades ago but decided on a different path (for better or worse) later on. 

A reductionist view of the self would agree that the self is real, but would say that common sense is mistaken to think that there is some further entity distinct from the self’s thoughts, experiences, etc., which stands under or grounds them. Rather, the self is, on this view, nothing more than the bundle or collection of its attributes. An anti-realist view of the self, meanwhile, would hold that, strictly speaking, there is no such thing. The self, on this view, is an illusion, and not something we can identify even with the fleeting bundle of thoughts, experiences, and the like which generate this illusion. 

The burden of this chapter is to defend the commonsense supposition that the self is real, and that it is not reducible to a mere bundle of thoughts and experiences (or to anything else for that matter). This will require defending the thesis that the self is a kind of substance. Now, the notion of substance is itself a matter of longstanding philosophical controversy. We will have reason in later chapters to address some of the broader issues that arise in connection with it. But a general treatment of substance is not necessary for present purposes, and is something I have in any event already provided elsewhere. 20 It will suffice for the moment to show that the self is a substance, whatever other substances there may or may not be. 

The basic argument is straightforward. First, each of us knows with certainty of his own existence simply by way of ordinary self-awareness. Second, reflection on the object of this knowledge makes it evident that it must have the nature of a substance. And third, none of the skeptical attempts to cast doubt on the reliability of self-awareness, nor any reductionist or eliminativist accounts of the self, can be made coherent. Naturally, each of these points needs spelling out. 

Self-awareness 

In an influential account of self-knowledge, Roderick Chisholm links such knowledge to the notion of a proposition that is “self-presenting” and thus “known directly.”21 A proposition is self-presenting to me, and thus directly known by me, when: (a) it is necessarily the case that, if the proposition is true, then I will know that it is true, and (b) it is in fact true. An example would be the proposition that I seem to see a door. For it is true at the moment that I seem to see one, and it is necessarily the case that when I seem to see a door, then I know that I seem to see one. It would make sense for me to say: “I seem to see a door, but in fact I do not really see one.” (For example, I might say this if I knew that what I was really looking at was a realistic painting of a door.) But it would not make sense for me to say: “I seem to see a door, but in fact I do not really seem to see one.” Even if there really is no door there, I can’t be wrong when I judge that it at least seems to me that there is one, if in fact that is how things seem. 

Now, if I know this proposition, argues Chisholm, then I also thereby have knowledge of myself. After all, what I know when I know that I seem to see a door is an attribute of me – namely, my attribute of seeming to see a door. So, if I know that I have this attribute, then I can know that I exist. I could hardly have the attribute otherwise. In this way, knowledge of a selfpresenting and thus directly known proposition affords one knowledge of the self which knows it. (We’ll consider a possible objection to this in a moment.) 

The reader may have noticed that this is similar to the argument from Augustine with which we began the previous chapter. Indeed, I think it is essentially a variation on Augustine’s point. 22 Augustine argued that, even when he is mistaken when making some judgment, he must exist in order to be mistaken. Knowledge of the judgment, whether that judgement is mistaken or not, thus suffices to give one knowledge of the self which makes the judgment. Of course, this is also reminiscent of Descartes’ Cogito, ergo sum (“I think, therefore I am”). What Chisholm is doing is restating the thrust of Augustine’s and Descartes’ point in a way that makes it explicit that selfknowledge is a byproduct of knowledge of a certain kind of proposition. 

However, there is a way of developing this idea which is familiar from the Augustinian and Cartesian traditions and which has evidently influenced Chisholm, but which is not essential to the argument, and which one need not endorse in order to accept it. Augustine holds that, though the mind knows the physical world through sense organs, it knows itself through itself. 23 This might seem to imply that knowledge of the self is direct, and in no way dependent on sensory experience. The Cartesian tradition is associated with the idea of a special faculty of introspection or inner sense, by which we know the self in a manner that parallels the way perception affords us knowledge of material objects. 24 Chisholm speaks of “direct awareness of the self” or “direct acquaintance” with it. 25 

But one may agree that certain knowledge of the self follows from knowledge of propositions of the kind to which Chisholm appeals, without holding that we are directly aware of or acquainted with the self, or that selfawareness involves a special faculty of introspection or inner sense that makes sensory experience unnecessary. An alternative position is that of Thomas Aquinas, who holds that the mind cannot know anything at all, not even itself, without sensory experience. He holds also that when the mind does know itself, this is not a matter of its having direct introspective acquaintance with itself, but rather of knowing its own acts and reflecting on their implications. Aquinas writes:

The first object of its act of understanding… is the nature of a material thing. And therefore that which is first known by the human intellect is an object of this kind, and that which is known secondarily is the act by which that object is known; and through the act the intellect itself is known. 26

For example, I look at what is there in front of me and judge that that is a door. Then my attention turns from this object of my perceptual experience to the perceptual experience itself, and I judge that I seem to see a door. Finally, reflecting on the fact that this is how things seem to me, I judge that there must therefore be a self or mind to whom things seem this way. 

Expounding Aquinas’s position, Robert Pasnau aptly characterizes this as a process of “cognitive ascent.”27 It is not a matter of perceiving the self by way of a special introspective faculty that is analogous to the visual faculty by which we perceive the door. Rather, “what happens is simply that one thought gives rise to another, in the familiar way we constantly experience.”28 Just as the perceptual experience of the door can trigger the thought that that door needs to be repainted, so too it can lead to the thought that I seem to see a door, which might in turn lead to the further thought that there is a self which is having that thought. Following Aquinas, Pasnau appeals to the analogy of an image in a mirror. 29 When looking at a mirror in which a door is reflected, I may think either “That is a door” or “That is a mirror image of a door,” depending on whether my attention is focused on the object being reflected, or on the mirror image itself. But the difference is a difference in what my mind attends to, not a difference in mental faculties being deployed. The same is true, on Aquinas’s account, when the mind knows an object outside it, or knows the mental act by which it knows that object, or knows the self which performs that act. 

Because Aquinas’s position posits no special introspective faculty, we might, following Therese Scarpelli Cory, speak of it as an account of how “self-awareness” works. 30 She opts for that expression because it has a familiar everyday usage and is not widely associated with a single technical philosophical sense. Now, my aim for the moment is not to argue that the notion of a special introspective faculty is dubious and that Aquinas’s alternative account of self-awareness is correct (though that is in fact what I think). 31 The point is merely that arguments of the kind that Augustine and Chisholm give for the certainty of our knowledge of the self do not require positing a special faculty of introspection or inner sense. One could instead adopt Aquinas’s account. Hence objections to the notion of a special introspective faculty are not per se objections to Augustine’s or Chisholm’s reasons for claiming that the reality of the self can be known with certainty. 

The substantiality of the self 

At this point, however, one might raise an objection associated with the 18th - century thinker Georg Christoph Lichtenberg, who directed it against Descartes’ Cogito. The most Descartes is entitled to claim, held Lichtenberg, is that there is thinking going on, but not that there is an I or self who is doing the thinking. Similarly, it might be claimed, the most that Augustine or Chisholm is entitled to say with certainty is that it seems that there is a door, but not that there is an I or self to whom it seems this way. Thus, whether we conceive of self-knowledge on the model of introspection or on the model of cognitive ascent, arguments like those of Augustine and Chisholm do not really establish the reality of the self. For its reality simply doesn’t follow from the proposition that it seems that there is a door, or the proposition that there is thinking going on, or any other such proposition. 

But Lichtenberg, I would suggest, is on closer inspection not in fact putting forward a clear or even coherent alternative, as can be seen from some considerations raised by Bernard Williams. 32 Consider the proposition that it seems that there is a door and the proposition that it seems that there is a dog. Does it follow from these two propositions that it seems that there is a door and a dog? No. For the first and second propositions could be true even if the third is not. For example, if I was looking at a door in a room where no dog was present, and you were looking at a dog in the middle of a park where no door was present, the first two propositions would be true, but the third would not. 

To explain this lack of entailment, though, I had to make reference to two different minds or selves – yours and mine – and the different contexts in which they find themselves. And in general, there is no way to make sense of the logical relationships between propositions about how things seem, or about what is being thought, or the like, without making reference to minds or selves to whom things seem a certain way, or who have the thoughts in question. 

Note that it would not help Lichtenberg if we added some further element to each proposition in order to capture the difference in context, but without making reference to minds or selves. For example, suppose we added the word “here” to our three propositions. Would the proposition that it seems that there is a door here and the proposition that it seems that there is a dog here entail the further proposition that it seems that there is a door and a dog here? No, but again, the reason is that the word “here” might in the first proposition indicate the context in which one mind or self finds itself (such as the room with the door in it), and might in the second proposition indicate the different context in which a different mind or self finds itself (such as the park with the dog in it). The reference of the word “here” (which is an example of what philosophers call an indexical term) cannot be known until we know which speaker or thinker is using it. Again, reference to minds or selves is essential in order to make sense of the logical relationships in question. 

The problem with Lichtenberg’s view, however, does not arise merely because of the special natures of minds and their thoughts. It runs deeper than that. Having things seem a certain way is a kind of attribute, as are thoughts in general. They are, in all respects relevant to the present point, metaphysically on a par with non-mental attributes such as being red, being cold, being round, and so on. And what all attributes have in common is that they cannot intelligibly be identified or individuated except by reference to a substance whose attributes they are. For example, I cannot identify or pick out the particular whiteness of the door in front of me without reference to the door whose whiteness it is. And I cannot individuate or distinguish that whiteness from the whiteness of the wall next to the door without reference to the wall in which that separate instance of whiteness inheres. Similarly, we cannot identify or individuate the thought that it seems that there is a door without reference to some substance in which that thought, qua attribute, inheres. 33 And to be a substance in which thoughts inhere just is to be a mind or a self. So, if I can know with certainty that it seems that there is a door, then I can go on to know with certainty that there is a mind or self to whom it seems that way. 

But might there be some alternative way to identify and individuate thoughts, without reference to the self? One proposal is that we can instead identify and individuate a thought by reference to its causes and effects. 34 The idea would be that, for example, the thought that it seems that there is a door can be picked out and distinguished from other thoughts by reference to the fact that the thought was caused by light from the door striking a certain pair of eyeballs, and that it went on in turn to generate the effect of the writing of the sentence that you are now reading. 

But the problem with this might be obvious. For we need to ask: Exactly whose eyeballs were struck by the light from the door? And exactly who wrote the sentence in question? The answer, of course, is that the eyeballs and act of writing the sentence were mine. And I am a self. It is, after all, substances that act as causes, and that are affected by other causes. The light from the door affected the organs of a certain substance, namely me; and it was a certain substance, namely me, who wrote the sentence. So, the thought that it seems that there is a door, which was generated by the light and produced the writing of the sentence, is an attribute of a substance, namely me. And a substance which has thoughts as attributes just is, again, a mind or self. 

The problem, in short, is that in trying to disassociate thoughts from substances, it won’t do to appeal instead to causes and effects, because the relevant causes and effects cannot themselves be made sense of without reference to precisely the sorts of substances that the view in question is trying to avoid.35 Of course, those who deny that there is such a substance as the self will disagree, but the problem is that the alternative proposal under consideration fails to provide a non-question-begging reason for the disagreement. 

Now, the critic might at this point abandon Lichtenberg’s view but still argue that it is possible to avoid the conclusion that the self is a substance. He might concede that there must be a mind or self which has whatever thoughts, experiences, etc. that there may be, but claim that this mind or self is nothing more than a collection or bundle of the thoughts, experiences, etc. It isn’t a substance underlying the collection. This is the “bundle theory” of the self famously associated with David Hume. 36 Whereas Lichtenberg’s position suggests an anti-realist view of the self, Hume’s amounts to a reductionist view. 

The first thing to say in reply is to reiterate the point that thoughts, experiences, and other mental states are attributes, and attributes cannot be identified or individuated apart from the substance in which they inhere. This objection stands whether we are talking about a particular thought or a collection of thoughts. Nor could Hume solve the problem by identifying and individuating each mental state by reference to the bundle of which it is a part, rather than by reference to a substance in which it inheres. For that would only yield a viciously circular explanation. He would be identifying individual mental states by reference to the bundles of which they are members, and identifying the bundles of which they are members by reference to the individual mental states. 

There are yet other problems with identifying and individuating any bundle or collection of mental states (as opposed to an individual mental state). Suppose that there are currently three mental states associated with me: (a) the perceptual experience of seeing a door, (b) the memory of drinking coffee this morning, and (c) the thought that it is almost 2:30 pm. And suppose that there are currently three mental states associated with you: (d) the perceptual experience of seeing a dog, (e) the memory of having pizza for dinner last night, and (f) the thought that you have a flight to catch this evening. On the bundle theory of the self – to oversimplify it a bit in a way that does not affect the present point – I am nothing more than the collection made up of (a), (b), and (c), and you are nothing more than the collection made up of (d), (e), and (f). But exactly what makes it the case that the mental states in question make up just these two discrete bundles? Why isn’t (d), say, also part of my bundle rather than yours? Or why don’t (c) and (d) make up a third bundle of their own, and thus a distinct self from either you or me, rather being parts of our bundles? 

In response, the bundle theorist might point out that from the first-person point of view of my current stream of consciousness, I am aware of seeing a door, of remembering drinking coffee, and of thinking that it is almost 2:30, but am not aware of seeing a dog. That is true, but that does not explain why (d) is not a part of my bundle in a way that will save the bundle theory. For exactly what does this first-person point of view amount to? Is it the point of view of a substance which has (a), (b), and (c) as its attributes, but not (d)? The bundle theorist can’t say that, because his whole point is to avoid having to posit any substance distinct from the collection of mental states. Does this first-person point of view instead reflect some further mental state (g)? In that case, the same problem just arises again: Why is (g) part of my bundle rather than yours, or rather than part of some third bundle? 

Hume himself held that we take mental states to be parts of the same one bundle because of relations of resemblance and causation that hold between them. But this is no help at all. My memory of drinking coffee this morning and your memory of eating pizza last night resemble one another insofar as they are both memories, but they are not parts of the same bundle. Suppose the bundle theorist responds by pointing out that it matters to whom two mental states resemble one another. If in fact I had these two mental states – (b) the memory of drinking coffee this morning, and (e) the memory of eating pizza last night – and found that they resembled one another, then I would regard them as part of the same bundle. But, as it happens, I don’t have both of them. But the problems with such a response should be obvious. 

For we need to ask what this self is to whom two mental states might resemble one another. If we say that it is a substance, then that will defeat the whole purpose of the bundle theory, which was to avoid having to posit any substance underlying the mental states that make up the bundle. Should we say instead that the self in question is just the bundle itself? Then we will be stuck once again with a viciously circular explanation. We will be purporting to explain why different mental states are part of the same bundle in terms of their resemblance to one another – and then purporting to explain the sense in which two different mental states might resemble one another in terms of their being part of the same bundle. 

Suppose the defender of the bundle theory opts instead to posit some further mental state: (h) the experience of finding that (b) and (e) resemble one another. The idea here would be that (b) and (e) will be part of the same bundle if they are associated with the further mental state (h), but not otherwise. But this obviously won’t work. For now we need to know how this new mental state (h) does or does not come to be part of the same bundle as (b) and (e). We will have just kicked the problem back a stage, rather than solving it. 

Nor does appeal to causal relations between mental states help. For one thing, they have to be causal relations of the right kind. Suppose my memory that I drank coffee this morning leads me to tell you about it, and this in turn triggers in you a memory of having eaten pizza last night. Then there will be a causal relation between (b) and (e), but obviously that would not be enough to make them members of the same bundle. (b) is still my memory and not yours, and (e) is still your memory and not mine. So, suppose we say that two mental states will be part of the same bundle if the specific causal relation that holds between them has to do with their being generated by the same brain, or associated with the same soul. The problem here is that we’re now bringing back in the idea of an underlying substance, when the whole point of the bundle theory was to get rid of it. Suppose, then, that we say instead that two mental states will be part of the same bundle if they are causally related to some third mental state. The problem now is that we will once again merely have kicked the problem back a stage rather than solved it. For exactly what makes their causal relation to this third mental state the right kind – the kind that will suffice to make them part of the same bundle (whereas other causal relations do not suffice to do so)? 

These problems with Hume’s bundle theory of the self reflect problems afflicting reductionist theories of substance in general. 37 There simply is no coherent way to reduce substances to collections of attributes, whether the attributes in question are thoughts, experiences, and the like, or are of some other type. Just as we cannot coherently doubt the reality of the self, neither can we coherently doubt that the self is a substance. 

The points made so far establish the incoherence of eliminative and reductionist views of the self considered synchronically or at a particular moment of time. Similar points apply to the self considered diachronically or over a span of time. Suppose I reason from the premises that all men are mortal and that Socrates is a man to the conclusion that Socrates is mortal. I cannot make sense of this sequence as an inference or chain of reasoning without supposing that there is a single self that had all three thoughts, persisting over time from the first through the second and on to the third. Certainly there would be no inference if there were instead three distinct ephemeral selves succeeding one another, any more than there would be if Tom had the thought that all men are mortal, Dick had the thought that Socrates is a man, and Harry had the thought that Socrates is mortal, but none of them had all three thoughts. 

It would be even less plausible to say that an inference had occurred if there were no selves at all, not even of an ephemeral sort, but merely three ownerless thoughts somehow occurring one after another in that sequence. That would be like saying that an inference had occurred if random spills of ink had by chance produced strings of marks that looked vaguely like the sentences “All men are mortal,” “Socrates is a man,” and “Socrates is mortal” on three adjacent pieces of paper, but with there being no person around to read them much less write them. A sequence of thoughts without a thinker, of the kind posited by Lichtenberg, could never amount to thinking in the sense of working through a chain of reasoning – as Lichtenberg himself did in the course of developing his critique of Descartes’ Cogito. 

Indeed, the problem arises even if we consider just a single thought, let alone a sequence of thoughts. In order even to have the thought that all men are mortal, the self needs to persist at least through the short amount of time it takes to think through it. 38 If there is no self that persists through that time, there will be no thought, but at most a sequence of fragments of a thought. And once again, the situation is even more problematic if, following Lichtenberg, we suppose that there are no selves at all, not even a succession of fleeting ones. In the absence of a self, the sequence all men are mortal wouldn’t constitute a thought any more than a pile of Scrabble letters spilled onto the floor would.

Perhaps, in his defense, Lichtenberg would bite the bullet and suggest that, for all we know, there really is nothing that constitutes a thought any more than spilled Scrabble letters do. But this would merely add incoherence to incoherence. To make such a suggestion is itself to express a thought and to invite us to agree with it. Or if it isn’t, then what exactly would Lichtenberg be doing in saying such a thing? Making noises with no more meaning than a pile of spilled Scrabble letters? If so, then Lichtenberg wouldn’t be saying anything true, any more than a grunt or a cough is true – in which case, he wouldn’t really have offered any defense of his position at all. (We’ll have reason later on to consider in greater depth the incoherence of denying the reality of thoughts.) 

What about a view that doesn’t eliminate the persisting self but reduces it to a collection of parts succeeding one another in time? This is the thesis that a person is really nothing but a collection of person-stages. 39 There is the stage of me that exists at time t1 , the stage of me that exists at time t2 , the stage of me that exists at time t3 , and so on, and what I am is just the collection of these stages. But this won’t work, for the same reason that Hume’s “bundle theory” doesn’t work. For consider a further sequence of person-stages that might be attributed to you, such as the stage of you that exists at time t4 , the stage of you that exists at time t5 , the stage of you that exists at time t6 , and so on. Exactly what is it that makes it the case that all and only stages t1 through t3 are stages of me, and all and only stages t4 through t6 are stages of you? Why not instead say that t4 is one of my stages too, or that stages t3 and t4 are stages of some third person, or posit any number of other possible combinations? 

However we answer such questions (by appealing to causal relations, or whatever), they are going to face exactly the sorts of problems Hume’s bundle theory faces. Just as we cannot identify and individuate mental states except by reference to the self whose mental states they are, so too we cannot identify and individuate person-stages except by reference to the person of whom they are stages. 40 Hence the attempt to reduce persons to collections of person-stages, like the attempt to reduce the self to a bundle of mental states, simply gets things the wrong way around.41 

When considering the self diachronically no less than synchronically, then, there is no way coherently to deny its substantiality and irreducibility. Note that nothing said so far depends on what kind of substance the self is. The same points could be made whether we think of the self as essentially bodily, or as essentially incorporeal, or as a mixture of the bodily and the incorporeal. I will return later on to the question of which of these views of the self is correct, but for the moment we can leave it to one side. 

Contrary arguments 

The case is not complete, however, until we see what is wrong with all the competing views on offer. There have, in the history of philosophy, been three general sorts of challenges to the reality and irreducibility of the self: metaphysical, phenomenological, and scientific. Let’s consider each in turn. 

Metaphysical challenges 

The most influential metaphysical argument against the substantiality of the self in the history of Western philosophy derives from John Locke (1632- 1704). The thrust of the argument is very simple. Locke holds that the same self that at one time is associated with one substance can in principle come to be associated with a different substance. He has in mind scenarios like the one in the movie Freaky Friday, wherein a mother and daughter magically exchange bodies. (Locke’s own famous example is that of a prince whose consciousness comes to inhabit the body of a lowly cobbler.) Of course, such scenarios are fantasy, but Locke thinks that they are possible at least in theory, and that since a person who first inhabits one body can in such a scenario later come to inhabit another body, it follows that the self cannot be identified with a physical substance. But he holds that a person’s consciousness might in principle jump from one soul to another no less than from one body to another, so that the self cannot be identified with an incorporeal substance either. Therefore, Locke concludes, a person or self cannot be identified with a substance of any kind. Rather, a person is to be identified with that which he imagines jumping from one substance to another – the person’s consciousness, the content of which might include memories of having been in one body and awareness of now inhabiting another. 42 

A standard amendment to this account is thought necessary in order to save it from an objection raised by Thomas Reid (1710-1796). 43 Reid asks us to imagine a brave military officer who recalls being flogged as a boy for stealing, and who at an advanced age becomes a general who remembers his exploits as a young officer but has forgotten the boyhood flogging incident. Since the aged general is conscious of having been the young officer, by Locke’s criterion he is the same person as the young officer; and since the young officer is conscious of having been the boy, he is the same person as the boy. Now, identity is a transitive relation. That is to say, if A is identical to B and B is identical to C, then A is identical to C. Hence, if the aged general is identical to the young officer and the young officer is identical to the boy, then the aged general is identical to the boy. But since the aged general does not recall having been the boy, it would follow by Locke’s criterion that he is not the same person as the boy. Hence, Locke’s position entails a contradiction. 

Commentators on Locke routinely point out that he can get around this problem by simply allowing that the connection of memory between stages of the self need not be direct. He can say that a person A is identical to an earlier person C if either A is conscious of having been C, or A is conscious of having been some earlier person B and B was conscious of having been C. With this emendation, the aged general in Reid’s example would be the same person as the boy by Locke’s criterion, and the contradiction disappears. 

A far more serious problem, though, is posed by the circularity objection commonly attributed to Joseph Butler (1692-1752). 44 Memory can be unreliable. An insane person living in the twenty-first century who claims to recall the sting of defeat at Waterloo does not really remember being Napoleon, even if he thinks he does. So, in order for a person A to be identical to an earlier person B, it is not sufficient that A seems to remember being B. It has to be the case that A really does remember being B. But for A really to remember being B presupposes that A and B are in fact identical. Hence, Locke’s account is, implicitly, viciously circular. It purports to analyze personal identity in terms of memory, but memory in turn must be analyzed in terms of personal identity. 

In an influential analysis, Sydney Shoemaker proposed that Locke’s account can be salvaged by appeal to what he calls “quasi-memory.”45 A quasi-memory is a genuine recollection of an experience of some past event or action, but where (to avoid circularity) this does not imply that the person having the quasi-memory is the one who witnessed the event or carried out the action. For instance, if Reid’s young officer remembers an experience of being flogged as a boy and there really was such an experience, this could count as a “quasi-memory” even if the officer was not in fact the one who had that boyhood experience. Memories in the ordinary sense are, on this account, treated as a species of quasi-memories. In particular, something will count as a genuine memory in the ordinary sense when it is a quasi-memory and the person having the quasi-memory is indeed the one who experienced the event or carried out the action. 

Now, consider that, in situations where a person falsely seems to remember doing something that he did not in fact do (as in the case of the insane person who thinks he is Napoleon), the pseudo-memory typically has a cause that is very different from the cause of a genuine memory. For example, it might result from a blow to the head, brainwashing, drug abuse, or mental illness. Crucial to genuine memory, then, is the existence of the right sort of causal chain linking one’s apparent memory to the event or action that one seems to remember. With these ideas in hand, Shoemaker suggests, we can update Locke’s account by saying that a person A will be identical to some earlier person B if A has a quasi-memory of being B and there exists the right kind of causal chain connecting this quasi-memory to B. This, he holds, preserves the spirit of Locke’s analysis while avoiding the circularity problem. 

Now, one problem with Shoemaker’s “quasi-memory” analysis is that it is questionable whether one can coherently be said to have a genuine memory of an experience that was not one’s own. 46 But even putting that aside, there is a deeper problem. For what, exactly, will make a causal chain between A and B the “right” kind? It is hard to see how to answer this question in a way that doesn’t presuppose a judgment about whether A and B are identical. Shoemaker himself deliberately tailors his account of quasi-memory and its causes so that it will be as close as possible to our ordinary conception of how memory works in cases where A and B are identical. Hence, consider a case where two later people seem to remember doing what B did. A famous example is provided by Derek Parfit’s “teletransportation” scenario, in which a machine scans a person’s body and brain, destroys them, then sends the information to Mars, where a second machine generates an identical copy of the original person. 47 Since the person who steps out of the machine on Mars has the memories of the original, it might seem that by Locke’s criterion, he is the same person as the original. But now imagine that the first machine does not destroy the original person’s body and brain before sending the information to Mars. Then we’d have two people with the memories in question. On our ordinary notion of memory, only one person can remember doing what an earlier person did. Shoemaker thus stipulates that the “right” kind of causal chain will be one that does not involve “branching” of the kind illustrated by Parfit’s example. But as Harold Noonan points out, the problem with this procedure is that it gives the lie to the idea that Shoemaker’s “quasi-memory” analysis avoids circularity. 48 Shoemaker purports to analyze personal identity in terms of the “right” sort of causal connection, but a causal connection is not allowed to count as the “right” sort if it would entail departing too far from what we would ordinarily count as personal identity. 

In response to Noonan, Sven Bernecker suggests that it is because our ordinary understanding of memory and personal identity is more familiar than Shoemaker’s “quasi-memory” analysis that the latter seems parasitic on the former. If instead we lived in a world in which “quasi-memories” were common and ordinary memories rare, Shoemaker’s analysis would seem natural and our ordinary understanding would seem to be parasitic on it. 49 Now, whether “quasi-memory” is a coherent notion in the first place is part of what is at issue. If it isn’t, then Bernecker is not in fact describing a genuine possibility. But even apart from that, his reply misses the point. It is comparable to saying that in a world where decorative faux fruit was common and real fruit rare, we would take our conception of the latter to be parasitic on the former. Or it is like saying that in a world where heart disease was so common that healthy hearts were rare, we would take our conception of the latter to be parasitic on the former. In both cases, we probably wouldn’t make these judgements, but if we did, we would simply be confused, because such judgments would clearly be false. The notion of decorative faux fruit is parasitic on the notion of real fruit, whereas the notion of real fruit is entirely independent of that of decorative faux fruit. We would have the notion of real fruit whether or not we ever started the practice of making decorative faux fruit, but we would never have formed the concept of decorative faux fruit if we hadn’t first understood what real fruit is. Similarly, we would have no notion of what heart disease is except by contrast with what we know to be true of healthy hearts, but we would know what a healthy heart is even if there were no such thing as heart disease. There is an asymmetry in both cases, and a comparable asymmetry exists between the notion of memory and the notion of “quasi-memory.” The latter, if it makes sense at all, cannot be understood except by contrast with the former, whereas the former can be understood whether or not we ever conceive of the latter. 50 

But Parfit’s “teletransportation” example points to what is in my view the fundamental problem with Locke’s account, though not for the reasons usually supposed. Consider a further variation on the scenario. Imagine that after the first machine scans a person’s body and brain, destroys them, and sends the information to Mars, the second machine generates two identical copies of the original person. Now, since neither copy has the original body, there can be no question of one of them having a more favorable claim to being the same person as the original on that score. By Locke’s standard, that doesn’t matter anyway. That A is conscious of having been B is sufficient for A to be the same person as B. The trouble is that in this case, two later people are conscious of having been the earlier person. Now these two people cannot be identical to each other. After all, if one of them killed the other, it would be absurd to say that the same person would in that case be both dead and alive. But if they are not identical to each other, then neither can either one be identical to the original person. For as noted when discussing Reid, identity is a transitive relation. Hence, if either of the two copies really were identical to the original person, they would have been identical to each other. 

This “reduplication” objection might seem to be a decisive refutation of Locke’s account. 51 But many philosophers have claimed that with further tinkering, Locke’s alleged insight can still be preserved in the face of this and other problems facing it. Robert Nozick argues that even where two or more later persons have, by Locke’s criterion, some claim to being identical to some earlier person, if one of them has at least a somewhat greater degree of continuity with the earlier person than the others do (such as a more complete memory), then this “closest continuer” is the one we should judge identical with the original. 52 Parfit, by contrast, argues that the lesson we should draw from the difficulties facing reductionist theories of the self like Locke’s is not that the self is irreducible, but rather that there is no self over and above whatever can be captured in a reductionist analysis. 53 A Lockean, on this view, should simply note that sometimes a single later person A will have quasi-memories of doing what an earlier person B did that are caused in the right kind of way; sometimes several later people might have such quasimemories, but one will be more closely continuous with B than the others are; and sometimes all of them will be equally continuous. But that is all that can be said. There is no “further fact” of the matter, over and above these various degrees of continuity, about which is identical to the original person. If the self is supposed to be something over above them, then there is no such thing. 

Naturally, the various ways of responding to the problems facing analyses like Locke’s are controversial. But I submit that the whole debate misses the point, or at least the deepest point. To see what I have in mind, forget about the self for a moment and consider a different example, such as a carrot. Suppose I suggested that the attributes of some particular carrot, such as its orange color, shape, and size, might in theory jump to the substance that had been associated with a different carrot. Suppose I spelled this out by describing carrot A as being of one shade of orange, shape, and size and carrot B has having a slightly lighter shade of orange and slightly larger shape and size, and then proposing that we might find one morning that the attributes that the night before had existed in substance A were now to be found in B, and vice versa. Suppose I then suggested that the carrot itself had jumped from A to B, so that what a carrot is is not really a kind of substance after all, but rather a collection of attributes that might be associated with different substances at different times. Suppose someone pointed out in reply that we could imagine two carrots, B and C, having the attributes A had the day before, and that this posed a problem for my theory. Suppose I responded by saying that if either B or C was at least slightly more similar to A, then this “closest continuer” of A was the true original carrot. Suppose someone else suggested instead that the real lesson we should draw is that there is no such thing as a carrot over and above whatever degrees of continuity of attributes like color, size, shape, etc. one object has with an earlier one. 

This whole discussion would, of course, be silly, but not just because the metaphysics of carrots seems unworthy of such heavy going. The reason it would be silly is that attributes are simply not the sorts of thing that can jump from one substance to another. For example, there is no such thing as the orangeness of a particular carrot jumping to another carrot. To be sure, you could peel off the surface of one carrot and affix it to that of another carrot. But that’s not what I’m talking about. That would involve taking a piece of the first carrot and moving it, which can of course happen. And of course, you could alter the shade of the second carrot so that it looks more like the first. But that’s not what I’m talking about either. What I’m saying is that the first carrot’s orangeness itself – not a piece of the carrot but the orangeness of the piece – cannot be separated out and moved from one substance to another. That’s just not how attributes work. So, whatever we’re imagining when we imagine carrot B now looking the way carrot A did, it is not a matter of attributes jumping from one substance to another. It is at most a case of carrot B now having similar attributes to those that A had. It is a case of B having somehow been transformed into a replica of A, that’s all. And the scenario where two carrots, B and C, come to look exactly like A did, reinforces the point. The natural thing to say is that we’ve now got two replicas of A, neither of which (being mere replicas) is actually the same carrot as A. All the chin-pulling about which of them is the “closest continuer” of A, and whether there really is such a thing as a carrot over and above the degrees of continuity, would be for nothing, since it would be based on the erroneous assumption that attributes can jump from one substance to another. 

Am I saying that Locke’s theory and the entire centuries-long debate it has spawned is based on a similar mistake? Yes, that is exactly what I am saying. Consciousness, memories, personality traits, and the like are attributes. Hence they cannot jump from one substance to another, and the kinds of scenarios Locke and his followers imagine in no way show what they think they do.54 If someone wakes up in the cobbler’s body talking like the prince and seeming to remember what the prince did, then whatever is going on, it is not a matter of the prince’s consciousness, memories, and personality traits jumping from the prince’s body or soul to the cobbler’s, because attributes can’t do that. It would instead be a matter of the cobbler now somehow having very similar memories and personality traits to those of the prince. Similarly, whoever steps out of the machine on Mars in Parfit’s teletransportation example, it is not the original person, because memories, being attributes, can’t jump from one substance to another. It would merely be a replica of the original person, and thus (qua replica) not the original person. The scenario where two exactly similar people walk out of the machine on Mars should make that obvious. For clearly they are mere replicas, and since they came about in exactly the same way that the single person walking out of the machine does, he is a mere replica too.55 

The fundamental problem with Locke’s position, then, is that it argues for the conclusion that the self is an attribute or collection of attributes, by way of an argument that describes the self doing something (jumping from one substance to another) that an attribute cannot intelligibly be said to do.56 It also presupposes that we can identify and individuate attributes apart from the substance in which they inhere, which, as I have already argued, we cannot coherently do.57 There are other problems too. For example, if the self is to be analyzed in terms of the continuity of its conscious experiences, what happens to it when a person is entirely unconscious? Barry Dainton proposes dealing with this problem by arguing that the self can be identified with the capacity for conscious experience, since a capacity can exist even when it is not exercised.58 But a capacity too is an attribute, and like other attributes it can neither jump from one substance to another nor be identified and individuated apart from the substance whose capacity it is. The Lockean approach to thinking about the nature of the self simply cannot be extricated from this fundamental and fatal conceptual muddle. 

Another famous metaphysical challenge to the reality and irreducibility of the self comes from Buddhism. 59 Whereas the Lockean approach is developed primarily in terms of considerations having to do with the nature of the self specifically, the Buddhist “no-self” doctrine reflects a more general critique of the very notion of substance. 60 Common sense regards tables, chairs, rocks, trees, and other ordinary objects as persisting entities each of which has its own nature or essence sharply demarcating it from other things. Buddhism takes this to be an illusion. In reality, all such entities are entirely reducible to their parts and the causal relations between the parts. For example, what we think of as a tree is really nothing more than a collection of parts causally related in such a way that the resulting composite gives the temporary appearance of corresponding to our concept of a tree. But all that really exists are the parts themselves, and there is in reality nothing corresponding to our concept, or to the purported essence of a tree that that concept captures. The parts that make up the collection are also continually changing, so that the collection corresponding to what we think of as the tree at one moment is not exactly the same as the one that exists at any other moment. Moreover, the character of the composite is ultimately determined by the causal relations its parts bear to everything else that exists, so that a complete description of any one thing would make reference to everything else. Hence there are no essences by which one thing might be neatly distinguished from other things, and no substances that endure over time. This picture of reality is known as the doctrine of “dependent coorigination,” since it takes the things that make up the world of everyday experience to be interrelated aspects of one whole rather than the discrete entities they appear to be. 61

Now, I would argue that Buddhism’s general metaphysical picture of the world is false. Substances and essences, I maintain, are real, and most ordinary material objects are not in fact reducible to the sum of their parts. Defending such general metaphysical claims is beyond the scope of a book about human nature, but I have done so at length elsewhere. 62 Fortunately, it will suffice for present purposes to note the insuperable difficulties facing attempts to apply the general Buddhist metaphysical picture to an analysis of the self. 

Buddhism holds that the self too is entirely reducible to its parts, and indeed that strictly speaking it cannot be said to exist at all. For none of the parts count as the self, and there is nothing over and above the parts that could count as the self either. The parts in question are referred to as the “five aggregates,” and comprise bodily processes, sensation, perception, impulses to action, and consciousness. Causal relations between parts in these five classes give rise to the construct we call the “self” in such a way that it seems to have a distinctive essence and to be more than the sum of these parts. But this, it is claimed, is an illusion. 

The reader may have noticed that this position is very similar to the ones Locke and Hume would defend more than two millennia after the time of the Buddha. Not only is it open to the same objections that Locke’s and Hume’s views face, but such objections were raised fifteen centuries ago by critics of Buddhism associated with the Nyāya philosophical tradition within Hinduism. 63 For example, the Nyāya thinkers note that mental states are attributes, and attributes cannot be identified or individuated except by reference to a substance in which they inhere. 64 They also note that there is no way to identify the relevant causal relations between mental states without reference to a self. 65 For example, suppose I have the thought that the cat is on the mat and this triggers in me the desire to utter the sentence “The cat is on the mat.” The Buddhist will say that this causal relation between the thought and the desire is what makes them elements of what I take to be the same one self. But suppose my utterance in turn triggers in you the thought that someone should feed the cat. Then we have a further mental state that is causally related to the first two. But that mental state is not part of the same self with which the first two are associated. So, though the Buddhist claims that we can dispense with reference to selves and replace it with reference to causal relations between mental states, in fact we first have to know which mental states are associated with which selves before we can determine which causal relations are relevant. The analysis implicitly presupposes the self and thus hardly eliminates the self. 

The Buddhist responds that “mental perception” is what binds the belief that the cat is on the mat and the desire to utter the sentence together so that they, but not your belief that someone should feed the cat, come to be associated with the same self. But this just raises the same problem over again. 66 For now we need to know what makes this event of “mental perception” part of the same self as the belief that the cat is on the mat and the desire to utter the sentence, whereas your belief that someone should feed the cat is not part of it. Just as we cannot identify the mental states or relevant causal relations without first knowing which self they belong to, so too we cannot identify the relevant event of “mental perception” without first knowing which self it belongs to. Once again, the analysis presupposes the self and thus cannot be said to eliminate the self. 

In order to deal with such problems, the Pudgalavādin or “Personalist” school of thought within Buddhism posited the “person” as that which individuates the relevant mental states and causal relations. But the trouble with this approach is that spelling out what it is to be a “person” puts Personalist Buddhism in a dilemma. If a “person” is nothing more than a collection of causally related mental states, then we are back where we started and the Personalist has not added anything to the analysis except some new jargon. But if a “person” is something over and above a collection of causally related mental states, then it appears to be a self – in which case, the Personalist will have abandoned rather than vindicated the Buddhist no-self doctrine. 67 

The other points I have already made above in defense of the reality and irreducibility of the self also entail the falsity of the Buddhist analysis. Hence Buddhism does not add anything new to what we have already seen. To be sure, the Buddhist position is older than the views we’ve considered up to now, and was worked out in detail by thinkers of genius over the course of centuries. For that it deserves our respectful consideration. But that does not entail that Buddhism does not make the same mistakes that Lichtenberg, Locke, Hume, Parfit and company made. It just made them earlier. 

Phenomenological challenges 

The second of the main challenges to the reality of the self is phenomenological. That is to say, it appeals to what we know from the firstperson point of view of conscious experience (as opposed to the third-person point of view of metaphysical inquiry or scientific investigation). It claims that the self is not revealed to us from this first-person point of view, and that we therefore have reason to doubt its reality. 

The most famous argument to this effect in the history of Western philosophy comes from Hume. The self, he says, is supposed to be something that underlies the different “perceptions” that we have and that persists through changes from one perception to another. Hence, to be aware of the self from the first-person point of view would entail being aware of something distinct from our perceptions and persisting in this way. Yet the examination of conscious experience reveals no such thing. Hume writes: 

For my part, when I enter most intimately into what I call myself, I always stumble on some particular perception or other, of heat or cold, light or shade, love or hatred, pain or pleasure. I never can catch myself at any time without a perception, and never can observe any thing but the perception. 68

 Hume suggests that when others carry out this exercise, they will get similar results. For example, when you look within your own mind, you might encounter the thought that the cat is on the mat, the experience of seeming to see a door, and so on. But this or that particular thought, experience, or other mental items is all you will encounter, and you will never encounter any self in which they inhere. 

The first indication that there is something wrong with this argument is that Hume himself reports what is going on by saying things like “I always stumble on some particular perception or other.” To what does this “I” refer if not the self? It will not do to suggest, as Lichtenberg might, that we can eliminate the first-person reference and instead say something like “Some particular perception is all that is ever observed.” For in that case, Hume will be saying something that neither he nor anyone else can support merely from an examination of conscious experience. If an examination of your conscious experience fails to turn up anything but particular perceptions, it hardly follows that particular perceptions are all that is ever observed anywhere. That would like saying that since I do not ever observe within my own conscious awareness the experience of skydiving, it follows that “The experience of skydiving is never observed.” Of course, it is in fact observed – by introspective skydivers, even though never by me. Note that the point is not that this by itself shows that there is in fact a self. The point is rather that the force of Hume’s argument rests on what he has been able to determine from the first-person evidence. Hence, if he has to reformulate his conclusion in a way that eliminates reference to the first person, he will end up stripping the argument of any force. 69 

There is also the question of what demarcates the perceptions I am aware of from the perceptions you are aware of if it is not the fact that there are two distinct selves who have the respective perceptions. Now, Hume, as we have seen, will say that what demarcates these perceptions is that they are parts of two distinct bundles. This might also seem to legitimate his use of “I” in the passage quoted. It refers to a particular bundle of perceptions, even if not to anything distinct from the bundle in which the perceptions inhere. But this will be an adequate response only if Hume’s “bundle theory” of the self is coherent, and as we have seen, it is not. 

The deeper problem with Hume’s argument, however, is that it presupposes that the self, if it is real, ought to be knowable to itself in just the way that the other things the self knows are. This is a little like saying that if the eye is real, then it ought to be able to see itself in just the way it sees other things. In fact, of course, while the eye can see other things directly, it can see itself only indirectly, through reflections. What Hume fails to consider is that, like the eye, the self can know itself only indirectly. 70

Here is how the mistake arises. When you perceive a door, the door is the object of your perception. It is tempting to think that when you are aware of the fact that you are perceiving the door, that is because the perception of the door itself becomes, in turn, the object of something like an inner perception or introspective state. “Higher-order” theories of consciousness are based on this supposition, and hold that a mental state like perceiving the door will be unconscious unless it becomes the object of a higher-order mental state. 71 Our knowledge of the internal world of the mind is thereby modeled on our knowledge of the external physical world. With this model in mind, it can seem that the self ought to be directly observable in something like the way the door is, so that when we do not in fact observe it, doubts about its reality seem warranted. But in fact this model of self-knowledge is wrong, so that the Humean skepticism about the self that it implies is without foundation. 

As philosophers in the phenomenological tradition argue, the self’s awareness of itself and its mental states is, in fact, first and foremost “prereflective” in nature. 72 Pre-reflective awareness is to be distinguished from reflective awareness on the one hand, and complete unconsciousness on the other. It is clear enough what it is for something to be unconscious. For example, that blood is passing through your veins and arteries, that food is passing through your intestines, and similar facts are things of which you are usually totally unconscious. Unless something goes wrong, you would normally not be experiencing or thinking about them at all. Reflective awareness, meanwhile, is the sort of thing manifest when you deliberately turn your attention to what you are currently thinking or feeling, when you become self-conscious upon seeing yourself in a mirror or embarrassing yourself, when you rehearse the events of your life to yourself or others, and so on. In such cases, awareness of yourself and your mental states has an explicit character. 

Pre-reflective awareness, qua awareness, is not unconscious. But neither is it explicit. Suppose that I am driving but have stopped for a red light. Then, while trying to remember what I need to pick up at the supermarket, I notice that the car in front of me has started to move after the light has turned green, and I proceed to drive straight ahead as it turns the corner. A moment later my passenger asks: “What color was that car?” and I pause and say that I think it was silver. In this sequence, a visual experience that had as its object the red light is followed by a visual experience that has a different object, the car. That in turn is followed by a recollection which also has the car as its object. We have a mental state of one type followed by a distinct mental state of the same type followed by a third mental state of a different type, where the object of the third mental state is the same as the object of the second one but different from the object of the first. So, neither any particular mental state, nor any one type of mental state, nor any one object of a mental state persists through the sequence. 73 

Now, let’s notice a few things about this example. First, some of what is going on are things I am aware of rather than unconscious of, but not in a reflective way. For example, I was aware of the car in front of me, which is why I could later report on its color. But when I saw it, I was lost in thought trying to remember what I needed to buy, and thus was not paying much attention to it. Had my passenger asked me the question ten minutes later, I may well have completely forgotten that there was a car, so little did it make an impression on me. In fact, I may remember it later on only precisely because I was asked the question and thereby made to reflect on the experience. But the experience itself was “pre-reflective” rather than reflective, even if it was, again, not unconscious. In fact, this pre-reflective awareness is a precondition of there being anything to reflect on. If I had been entirely unconscious of the car, I wouldn’t have been aware of it, and thus could hardly reflect later on its color. But had my initial awareness been reflective rather than pre-reflective, I wouldn’t have had to think about it for a moment in order to recall the color. 

A second thing to note is that despite the fact that no particular mental state, nor any type of mental state, nor any object of a mental state persists through the sequence, they nevertheless are all part of one and the same sequence. What gives the sequence this unity is that each member of it and the sequence as a whole involve the way things are “for me.” It first seemed to me that there was a red light, then it seemed to me that the car ahead of me has moved, then it seemed to me that the car had been silver, and then as I reflected on its color it seemed to me that it was I who had seen the red light and then the car. Moreover, the nature of what was experienced, and not merely who was experiencing it, reflect this first-person character. The red light and the car appeared the way one would expect them to appear to someone perceiving them from my perspective (from inside the automobile I was driving, given the distance from them that I happened to be at, traveling at the particular speed I was moving at, etc.). The natural way to describe the sequence and each member of it is by reference to the self to whom things appear the way they do, and (as we have seen) it is at best extremely difficult to describe them without reference to a self. Here too, though, all of this is typically pre-reflective. I do not explicitly think “It seems to me that there is a red light,” then “It seems to me that the car has moved,” and so on. I may go on to think such things later on if I reflect on the sequence, but even if I do not, it is still the case that it involved a series of ways that things appeared for me. 

Writers in the phenomenological tradition emphasize that consciousness is neither something additional to pre-reflective awareness nor a matter of taking such awareness as an object of thought or perception that would otherwise remain unconscious, as “higher-order” theories of consciousness suppose. Rather, consciousness is constitutive of pre-reflective awareness. Nor, they argue, is it merely that the account I have been summarizing happens to correspond to the character of the human mind. Rather, our fundamental mental states must involve pre-reflective awareness. For on the one hand, they cannot be unconscious. If a mental state is unconscious, it is hard to see why it would become conscious simply by virtue of becoming the object of some further mental state that is itself unconscious. But on the other hand, if we suppose that our fundamental mental states involve reflective awareness, then we will be led into a vicious regress. If a certain mental state is conscious only if it is the object of explicit reflection, then that act of explicit reflection will be conscious only if it is in turn the object of a higherorder act of explicit reflection, which will in turn be conscious only if it is the object of a yet higher-order act, and so on ad infinitum. 74 

Now, how does all of this relate to the challenge raised by Hume? In the following way. Hume supposes that in order to be aware of itself, the self would have to become directly perceived in an act of explicit or reflective awareness. But that is not in fact how all awareness works, and indeed it is not how our fundamental mode of awareness works. (Hume, alas, was no better a phenomenologist than he was a metaphysician.) We can be, and usually are, aware of things only implicitly and pre-reflectively, and sometimes (as with the eye which sees itself in a mirror) we can know them only indirectly. In the case of the self, we are initially aware of it prereflectively in the manner described by modern phenomenologists. Then we reflect on it by what I referred to earlier as “cognitive ascent” and, while not directly perceiving it, infer that it must not only be real but must have the nature of a substance, for the reasons given above. 

Hume, of course, thinks this is all illusory. But the point is that he has given us no good reason to agree with him. His account of what selfawareness would be like is simplistic, and in trying to show that we are not aware of the self even he cannot entirely avoid reference to it, so tightly is the notion of the self tied to our very conception of awareness. Moreover, Hume’s alternative reductionist analysis of the self falls apart upon inspection, as we have seen. 

Phenomenological investigation arguably reveals our ordinary notion of the self to be even richer than the points made so far would indicate. Galen Strawson argues that the self is ordinarily experienced or conceptualized as: a thing, in some robust sense; mental in its nature; single or unified both synchronically and diachronically; distinct from other things; a subject of experience; an agent; and having a distinctive character or personality. 75 He holds that these aspects of the ordinary phenomenology of the self are so basic that they “are situated below any level of plausible cultural variation.”76 Significantly, he makes these claims despite the fact that he not only does not think that there is anything in objective reality that actually has all of these features, but also denies that he finds all of them in his own introspective awareness of himself. For example, he says that he does not personally have a strong sense of himself as being continuous over long periods of time, even though most other people do.77 He merely holds that these are in fact aspects of most people’s experience of the self, whether or not that experience is delusory. 

Naturally, one might dispute Strawson’s account of the phenomenological facts. But that even a thinker sympathetic to the reductionist position could concede that that position conflicts with the phenomenology shows that the appeal to introspection simply does not by itself have the force that Hume thinks it does. For one thing, the phenomenology gives much stronger support to the view that the self is real and irreducible than Hume supposes. For another, even if Hume had gotten the phenomenology right, it simply wouldn’t follow that the self is not real and irreducible. For two can play at Strawson’s game. If Strawson can allow that there appears to be a self having all the aspects he identifies but still argue that the appearance is illusory, then one could also allow, with Hume, that there appears not to be a self, and yet still argue that that appearance is illusory. At the end of the day, what is decisive is not what the phenomenology reveals, but which view has the stronger metaphysical considerations in its favor. By that standard, as I have shown, Hume’s critic has the better of the argument. 

Scientific challenges 

It is sometimes claimed that modern science has shown that the irreducibility or even the reality of the self is illusory. Some of these arguments are remarkably shoddy. For example, psychologist Bruce Hood appeals to considerations such as the fact that brain damage can dramatically alter personality, that the neural processes underlying various mental phenomena are distributed throughout the brain rather than centralized in any one part of it, that the brain fills in gaps in the meagre information it receives from the environment when generating conscious awareness of the world and of ourselves, and that the roles we play as men, women, parents, workers, and so on are shaped by our social interactions with others. 78 Such facts, he claims, vindicate the Humean and Buddhist “bundle” theory of the self, and indeed reveal that the self is an “illusion.” 

But this is a blatant non sequitur. The thesis that the self is real and irreducible, you will recall, entails that it is a substance that persists through changes of its attributes and cannot coherently be identified with the collection of attributes themselves. How does the falsity of this claim follow from considerations like those adduced by Hood? Yes, damage to the brain can result in changes in personality, memory, and the like that are so dramatic that the person may even seem unrecognizable. How does it follow that there isn’t a single substance in which both the mental attributes that existed before the brain damage, and those that exist after it, inhere? Yes, the neural processes associated with various mental activities are distributed throughout the brain rather than located in one region of it. How does that show that there isn’t a single substance to which these various processes themselves, along with the mental states associated with them, can be attributed? Remember that I am not at the moment addressing the question whether the self is an immaterial substance, or a material one, or possesses a mixture of material and immaterial attributes. I will return to that later on. The claim is simply that it is a substance of some kind, and considerations like those cited by Hood do not show otherwise. (Even Hood, when describing a person’s neural processes and mental states, associates them with the same one brain.) 

If, in generating conscious experiences of the world around us and of our own minds, the brain fills in gaps in the information supplied by the senses, how does that show that there isn’t a persisting substance in which the experiences inhere? If your understanding of what it is to be a man or a woman, a good parent, or a dutiful employee is deeply influenced by the expectations of the fellow members of your society, how does that show that there isn’t a single abiding substance to which that understanding can be attributed? Wolves learn how to hunt while still cubs, through play with other cubs. Chimpanzees learn grooming from their mothers. Should we conclude that wolves and chimpanzees are also “illusions”? Again, the conclusion Hood draws doesn’t follow from his premises. To identity various internal and external factors that influence how a self comes to conceptualize itself simply does not entail that there is no self, or that the self is nothing but a collection of mental states. 

There is also the fact that in spelling out the details of the mechanisms by which the purported illusion of the self is generated, Hood cannot avoid making reference to the very self he says is an illusion. Hood says that it is your brain in which can be found the neural activity that generates your experiences of the world and your sense of self, thereby generating your purported “self illusion.” So who exactly is Hood talking to, if “you” are only an illusion? How could there fail to be a substance underlying the neural processes and mental states in question, if Hood cannot even identify the latter without making reference to the former? Hood is aware of the problem and responds with hand-waving to the effect that it merely reflects the limits of language. 79 But the problem has nothing to do with the limits of language, any more than the impossibility of round squares has anything to do with the limits of language. Rather, the problem, with the denial of the self as with nonsensical talk about round squares, is conceptual incoherence. For the reasons I’ve now spelled out at length in this chapter, there is simply no way coherently to articulate the claim that thoughts, experiences, and the like might exist without a substantial self or that the self is nothing but a bundle of such thoughts and experiences. 80

 The weakness of such pop science arguments notwithstanding, it might seem that psychiatry and neuroscience have provided the ingredients for more serious challenges to the reality and substantiality of the self. For example, consider dissociative identity disorder (DID) (formerly known as multiple personality disorder or MPD). Patients with this disorder appear to shift between several different identities or personalities, each of which has its own distinctive character traits, memories, and the like. Each personality appears to regard the others as distinct persons rather than as roles played by the same one person, and in some cases one personality might have no knowledge of another. Some researchers suggest that different streams of conscious awareness are associated with each personality. All of this might appear to support the Humean claim that the self is really nothing more than an aggregate of mental elements which gives the false appearance of being a single unified substance, where this appearance breaks down in DID cases. Kathleen Wilkes, who is critical of Lockean appeals to bizarre thought experiments, and who judges that less extreme dissociative phenomena like schizophrenia and fugue states do not pose a serious challenge to the reality and unity of the self, nevertheless concludes that “the concept ‘person’ has fractured under the strain” of the evidence concerning dissociative identity disorder. 81 

However, on closer inspection this argument can be seen to have no force. First of all, as even Wilkes acknowledges, the evidence concerning DID can be interpreted in more than one way, and is problematic insofar as at least some of the more exotic DID phenomena can be attributed to roleplaying on the part of patients and ideologically-inspired wishful thinking on the part of those treating and studying them. 82 Indeed, some of the more spectacular purported examples of dissociative identity disorder, such as the famous case of “Sybil,” have been exposed as fraudulent. 83 Accordingly, whether DID ought still to be included in the American Psychiatric Association’s Diagnostic and Statistical Manual of Mental Disorders is a matter of longstanding controversy. 

But even if we allow that what are characterized as “dissociated identities” or “multiple personalities” are not always the artifacts of a patient’s deception or a therapist’s prompting but result from a genuine and distinctive disorder of some kind, it simply does not follow that there is not a single persisting substance – and indeed, a single self – underlying these unusual psychological phenomena. For one thing, none of the phenomena actually entails the existence of distinct streams of consciousness, as opposed to a single stream of consciousness that has become delusional. We are all familiar with the psychological discontinuity and incoherence associated with everyday circumstances like sleep, dreams, drunkenness, senility, and the like. More dramatic disruptions to normal functioning attend disorders like amnesia; schizophrenia, which can involve the experience of hearing thoughts that seem to come from outside oneself; and anosognosia, in which patients seem not to be aware that they are suffering from ailments like blindness or paralysis. In neither these ordinary cases nor the extraordinary ones are we usually tempted to say that there is more than one self present. But as Tim Bayne argues, DID phenomena seem to be more extreme instances of the sorts of discontinuity and incoherence associated with these less dramatic sorts of disorder, so that the most natural interpretation of them is that they simply involve a more severe breakdown of continuity and coherence in a single self. 84 

As Bayne suggests, the strange phenomena associated with anosognosia and schizophrenia can be explained in terms of dysfunctions in introspective access to one’s conscious states, the mistaking of vivid mental imagery for perception, incapacity for certain kinds of inference, and the like. It is not that there is more than one stream of consciousness, but that the patient is unable to detect or properly interpret everything that is going on in the one stream. Dissociative identity disorder plausibly involves dysfunctions of these sorts together with other factors. For example, there is a more extreme version of the incoherence that results from ordinary self-deception. (Consider the example of a woman who at some level knows that her husband is committing adultery, but will not let herself entertain the thought and instead rehearses to herself other explanations of his suspicious behavior.) There is an exaggerated version of the struggle we often engage in to control our emotions or impulses. (Think of St. Paul’s famous characterization of sin as something analogous to an alien force within that leads him to act contrary to the good that he wants to do.85 ) There is a more melodramatic version of the role-playing we engage in when we present ourselves to others in different ways in different contexts, or fantasize about living different sorts of lives. (Think of an actor who can play radically different sorts of character equally convincingly, or even rapidly shift back and forth between different characters as he carries on a dialogue with himself. 86 ) Again, one need not hold that DID patients are knowingly engaging in deception or make-believe. The confabulation might instead be a consequence of severe dysfunction and delusion. In any event, it would involve a single self mistakenly, if sincerely, believing that it had fragmented into several selves, or even that the several selves had been present all along. 

Stephen Braude, who, unlike Bayne, does believe that there are distinct streams of consciousness in cases of dissociative identity disorder, nevertheless rejects the conclusion that there are distinct selves, and indeed argues that there must be a single underlying self. 87 He notes, first, that it would be fallacious to infer from the thesis that a person has dissociated into several identities to the conclusion that there must have existed several discrete elements prior to the dissociation of which the person was composed as an aggregate. This is to assume what Braude labels the “principle of compositional reversibility,” to which he says there are clear counterexamples. 88 For instance, if you break a table in half with an axe, it doesn’t follow that those two halves existed prior to the table and that the table was made by combining them. Rather, the division creates parts that didn’t previously exist, and something similar can be said to happen with the traumatic events that typically precede dissociative identity disorder. There is also the fact that the character of the identities that result closely reflect the contingent facts about the trauma. For example, sexual abuse might result in the creation of an identity that is either highly prudish or highly sexual, physical abuse might give rise to an identity with a very protective personality, and so on. It is hardly plausible to suppose that identities corresponding to traumatic events in this way just happened to exist already in the person prior to the trauma. 

Furthermore, argues Braude, we cannot make sense of the distinct identities that are exhibited in DID cases unless we take them to reside in a single underlying self. For one thing, the multiple personalities tend to share the same language and certain linguistic idiosyncrasies, a common stock of basic abilities (reading, doing arithmetic, understanding social cues, etc.), and certain memories of the pre-dissociated life of the person. This is hard to make sense of unless we suppose that there is a single self which possesses these features, and of which the distinct identities are all manifestations. 89 These fundamental features are so tightly integrated into the characters and behaviors of the distinct identities that it is implausible to suppose we could isolate the latter from the former, and treat them as distinct freestanding selves rather than variations on the same one underlying self. 90 

For another thing, the dissociated identities typically arise, again, in response to some kind of trauma. This is hard to make sense of unless we suppose there is a single self which finds itself conflicted and under stress, and generates the identities as a coping mechanism. Such inner psychological conflict could hardly exist if there were two or more irreducible selves present, any more than the existence of two people in the ordinary sense who disagree about something entails inner psychological conflict. The fact that the vast majority of people who go through traumatic events like the ones in question do not exhibit DID indicates that DID is simply an extreme means by which a single self might try to cope with a bad situation, where others might opt for other strategies (therapy, fantasy, drugs, or whatever).

 We might also note that the distinct identities manifest in DID patients often have access to each other’s thoughts (though sometimes they claim not to). It is sometimes suggested that what makes this possible is a kind of telepathy, analogous to the ability to know others’ thoughts that mind-readers claim to possess. 91 But surely the more obvious and simpler explanation is that the identities are all simply different manifestations of the same one self! 

On closer inspection, then, dissociative identity disorder does not cast any serious doubt on the reality and irreducibility of the self. But there is another, physiologically grounded sort of dissociation that might seem to do so, namely the strange behavior exhibited by “split-brain” or commissurotomy patients. The procedure in question involves a partial severing of the fibers connecting the two hemispheres of the brain, usually in the corpus callosum. It is sometimes used as a means of alleviating the more severe symptoms of epilepsy. Experiments with these patients have in some cases given rise to behavior indicative of the hemispheres functioning at cross-purposes. For example, a patient’s right hand might repeatedly interfere with what his left hand is trying to write, even crossing out and changing it. Or, when a stimulus is hidden from the hemisphere of the brain governing speech, but nevertheless produces an appropriate behavioral reaction (such as laughter) when presented to the other hemisphere, the patient will, when asked, confabulate a sincere but false explanation for this behavior. 92 Such results are sometimes claimed to indicate that what had once been a unified stream of consciousness has divided into two streams, each associated with one hemisphere of the brain. It has even been argued that there are really two persons associated with a human brain even before commissurotomy, and that the procedure simply reveals this bifurcation rather than creating it. 93 

Here too, though, closer inspection undermines the more sensational interpretations of the evidence, and certainly gives no grounds for doubting the reality of a single irreducible self. The first thing to note is that commissurotomy patients rarely exhibit odd behavior of the kind in question outside of the highly artificial experimental contexts that researchers have designed in order to elicit it. Typically patients do not like being put in such contexts, and they do not regard themselves as being two persons or otherwise dissociated in the extreme ways some researchers claim they are. 94 Their everyday behavior is largely normal and would not lead the average observer to judge that there is anything unusual about them. 95 

Commissurotomy patients are also typically already brain-damaged from epilepsy, so that it is not clear that the odd behavior some of them exhibit in experimental circumstances stems from the commissurotomy rather than from the pre-existing brain damage. 96 The two hemispheres of their brains are also never literally entirely split off from one another. For one thing, sometimes the corpus callosum is not entirely cut. But even when it is, there remain many other connections between the two hemispheres at other places in the brain. Moreover, there is nothing necessarily special about the fact that two brain states or processes exist in different hemispheres in the brain. Neural states and processes in different parts of the same hemisphere can be more remotely connected to one another than neural states and processes in different hemispheres sometimes are. 97 Hence it is somewhat arbitrary to posit exactly two streams of consciousness, each associated with one of the two hemispheres of the brain – as opposed, say, to many streams of consciousness associated with the many different sub-hemispheric areas of the brain. But if we are not going to attribute different streams of consciousness to different parts of the same hemisphere, why attribute different streams to the different hemispheres? 

Even Wilkes, despite being overly impressed by the evidence from dissociative identity disorder, urges a more cautious interpretation of “splitbrain” behavior. As she points out, there are many phenomena of both an everyday and an abnormal sort that involve a greater degree of dissociation than commissurotomy patients usually exhibit, but where we would not be tempted to posit more than one person. 98 For example, there is the fact that it often takes a person time fully to realize or admit to himself the reality of some unpleasant circumstance, so that he may act in a manner inconsistent with what at some level he knows to be true. There is the fact of divided attention, wherein a person can carry out two very different complex tasks at the same time (such as negotiating traffic while carrying on a philosophical conversation). There is the phenomenon of “being of two minds” about some matter. There is weakness of will, wherein we do things that in some sense we do not want to do. There is outright self-deception, wherein one gets himself to believe something that at some level he knows to be false. Among abnormal cases, there is the amnesia associated with fugue states and the odd behavior that can be elicited via hypnosis. Now, in all of these normal and abnormal cases, we have a person who acts in a unified way overall, and simply exhibits a certain degree of dissociation under special circumstances. We don’t treat any of it as evidence of more than one self, but rather interpret it as the multitasking of, or confusion within, a single self. There is no reason not to interpret the behavior of “split-brain” patients and simply an extension of the same sort of thing. We don’t have more than one self in these cases. We simply have behavior on the part of a single self, the oddity of which is perfectly understandable given that the self in question is, after all, braindamaged. The oddity of the behavior no more gives us reason to deny a single self then does the oddity of the gait of someone whose legs are damaged, or the oddity of the speech of someone whose teeth, jaw, or facial muscles have been damaged. 

Then there is the fact that the two hemispheres of the brain, just as any other two parts of the brain, are built to complement one another’s functioning. They are not built to operate independently, but rather work as a unit. 99 Moreover, the rest of the human body is simply not constructed in a way that would make it suited to be operated by two minds, but also functions as a unit. 100 All things considered, then, attributing two selves (as opposed to a single confused self) even to a commissurotomy patient (let alone to the pre-commissurotomy person) is by no means the natural interpretation of the behavioral and physiological evidence, but on the contrary is a highly unnatural interpretation. 

Finally, as Braude points out, even if we were to go along with the suggestion that two selves come to exist post-commissurotomy, it simply wouldn’t follow that there was not a single unified self pre-commissurotomy, or that the single self that then existed has now divided into two. We could say instead that the previous self simply went out of existence after the commissurotomy and was replaced by two new selves, in a manner comparable to cell division or the slicing of a flatworm into two new worms. 101 The point isn’t that this is plausibly what actually happens in commissurotomy (I don’t think it is), but rather to reinforce the lesson that there is nothing in the scientific evidence itself that forces on us an interpretation contrary to the reality and irreducibility of the self. There are alternative ways to interpret it, and given that, as we have seen, there is independent reason to affirm the reality and irreducibility of the self, we ought to opt for one of those alternatives. 

Reductionists and eliminativists might insist that the most I have shown is that we have to think about the self as if it were real and irreducible, but not that there actually is such a self. Contemporary naturalist philosophers commonly assume that if a metaphysical claim is not grounded in natural science, then the only thing left for it to be is a deliverance of “conceptual analysis.” But such analysis, claims the philosophical naturalist, can only ever tell us about how we conceptualize reality, and not how reality is in itself. 

The fundamental problem with this objection is that it is simply false to suppose that metaphysical reasoning cannot be a third source of knowledge distinct from either natural science or conceptual analysis. This is a large issue that I cannot settle in a book that is about human nature, specifically, but I have addressed it at length elsewhere. 102 In any event, it will suffice for present purposes to make a narrower point. 103 It is not enough for the reductionist or eliminativist merely to assert that there may be no irreducible self despite our having to think as if there is. In order for us to take this assertion seriously, we need some coherent account of what it would be for there to exist no irreducible self even though we have to think as if there is. But the arguments I have given in this chapter claim precisely to show that we cannot give a coherent account of such a scenario. Hence, to respond to these arguments by suggesting that reductionism or eliminativism might still be true, even if we cannot see how, merely begs the question. 

So much, then, for challenges to the reality and irreducibility of the self. Let us turn now to challenges to the reality of the self’s thoughts and free choices.

3. The Intellect 

Intellect versus sentience 

The intellect is that power by which we are able to entertain concepts, to affirm the truth or falsity of propositions, and to assess the cogency of arguments. For example, we can entertain the concept of being a man; we can affirm the truth of the proposition that all men are mortal; and we can determine that from the propositions that all men are mortal and that Socrates is a man, the conclusion that Socrates is mortal logically follows. Traditionally, the act of entertaining a concept is known as simple apprehension; the act of affirming or denying the truth of a proposition is known as judgment; and the act of inferring a proposition from other propositions is known as reasoning. Simple apprehension, judgment, and reasoning are the intellect’s basic operations. 104 Being rational or capable of thought is essentially a matter of having the capacity for these operations of the intellect. 

To be sure, things are more complicated than this summary indicates. In Aristotelian-Thomistic philosophy, a distinction is drawn between active intellect, which is the power by which we first form a concept, and passive intellect, which stores concepts once they are formed. In addition to judgment, contemporary philosophers of mind typically distinguish a variety of further propositional attitudes, so called because they involve a thinker’s taking a certain attitude toward a proposition. For example, you might believe that the cat is on the mat, desire that the cat is on the mat, fear that the cat is on the mat, and so on. These would be different possible attitudes one could take toward the same proposition. Or you might believe that the cat is on the mat, believe that the dog is on the log, believe that all men are mortal, and so on. These would be different propositions toward which one might take the same attitude. In contemporary philosophy of mind, the capacity for thought is usually analyzed primarily in terms of this notion of the propositional attitudes. 105 

We will return to such details later on, but for the moment we can focus on the capacity to grasp concepts, propositions, and arguments as what is fundamental to having an intellect or being a rational or thinking being. Now, of these three, concepts are the most basic. For a proposition involves combining concepts. For instance, in order to formulate the proposition that all men are mortal, we need first to have the concept of being a man and the concept of being mortal. Arguments, in turn, are built up out of propositions, such as the propositions that all men are mortal, that Socrates is a man, and that Socrates is mortal. In order to understand the intellect, then, the first thing we need to understand is what a concept is. 

The easiest way to begin is to contrast what the intellect does when entertaining concepts with what the mere sentience that we share with other animals makes possible. So, a brief account of sentience is in order. A further distinction traditionally drawn in Aristotelian-Thomistic philosophy is that between the external senses and the internal senses, both of which are typically possessed by sentient organisms. The external senses are the familiar ones: sight, hearing, taste, smell, and touch. Each of them is associated with its own characteristic range of sensations. For example, sight might yield a sensation of a rectangular red expanse; hearing might yield a sensation of a loud bang; taste, a sensation of bitterness; smell, a sensation of a sweet fragrance; touch, of warmth. 106 Obviously, this is very elementary. Much more could be said about the external senses at the neurological, psychological, and philosophical levels of description, and we will have more to say later on. But this will suffice for present purposes. 

The internal senses are the synthetic sense, instinct, the imagination, and sensory memory. Some of the jargon may be unfamiliar, but the capacities it refers to are not. The synthetic sense is what unifies the deliverances of the five external senses into a coherent conscious experience. For example, when sight yields a sensation of an elongated yellowish expanse, touch a sensation of a smooth and solid surface, taste a sensation of sweetness, and so on, the synthetic sense combines these in a way that gives rise to the conscious awareness of a banana. It presents the diverse qualities revealed by sight, touch, taste, etc. not as a mere aggregate, but rather as aspects of one and the same object. 107 

Instinct is also known as the “estimative power,” though while that expression might be at least a little more informative, it is less familiar. Instinct in the sense I am talking about is the capacity of an organism to apprehend something sensed as either beneficial or harmful, and to be inclined, accordingly, either to pursue it or to avoid it. For instance, once you have a sensory experience of a banana, you might “estimate” it to be desirable and therefore be at least somewhat inclined to get hold of it so as to eat it either at that moment or later on (though, of course, competing inclinations might override that one, or you might happen to be someone who doesn’t like bananas). By contrast, sensory experience of a snake might lead you to estimate or apprehend it as dangerous, and therefore to be at least somewhat inclined to flee from it (though of course, that inclination too may be overridden by a competing one). 

The imagination, in the sense we are concerned with, is the capacity to form images (also traditionally known as “phantasms”), understood as copies of what we experience in sensation. For example, we can not only have a visual sensation of an elongated yellow expanse, but, after it is no longer present, can also bring to mind a mental image of what that yellow expanse looked like. We can not only have a tactile experience of a smooth and solid surface, but can also later bring to mind what that surface felt like. We can not only taste sweetness, but can call to mind what tasting sweetness is like even when we are not currently eating anything sweet. We can not only have a sensory experience of a banana, but can form a mental image of a banana. And so on. As these examples indicate, there is mental imagery associated with all five external senses and with the synthetic sense. We can also form images of things we have never had sensory experience of, by combining aspects of things we have experienced. For example, we can combine the image of a banana with the image of redness, and thereby imagine what a red banana would look like. 

Sensory memory retains what is known via instinct or the estimative power, just as imagination retains what is sensed. In this way, sensory memory goes beyond imagination. For example, merely to imagine a yellow banana is not necessarily to remember it. (Suppose I had only ever seen green bananas and yellow things of other kinds, and form the mental image of a yellow banana by combining images of these other things.) In sensory memory, the triggering of such an image will be causally linked to the past event of actually having experienced a yellow banana, and will be associated with an inclination toward bananas that have not been experienced (say, an inclination to seek out other bananas in the future given that the one experienced before tasted good). 

Much more could be said about these “internal senses” as well, but here too what has been said will suffice for present purposes. In any event, it is not terribly controversial that sentient organisms possess something like the capacities just described, even if contemporary philosophers and scientists would not use the jargon that Aristotelian-Thomistic philosophers traditionally do, nor carve the conceptual territory up in precisely the same way. 

The point is that organisms without concepts, and thus without intellects, can and do possess the external and internal senses. For instance, a dog can see a triangle, but it does not conceptualize it as a closed plane figure with three straight sides. A dog can see, hear, smell, touch, and (if it is an angry mood!) taste a human being, but it does not conceptualize a human being as a rational animal. 108 For that reason, it does not entertain propositions about triangles or human beings, much less reason from one proposition to other propositions. But a dog can do things like instinctively estimate the piece of food a particular human being is holding out to it as desirable, and thus be inclined to approach it. It can have images of sights, odors, tastes and the like that it has experienced in the past triggered in its consciousness. It can by virtue of such experience come to associate a certain human being with food, and thereby be moved in future to seek that human being out. 

Of course, human beings possess these external and internal senses too, but on top of that, we do conceptualize the deliverances of these senses. We can see a particular triangle and can form mental images of that triangle and of other possible particular triangles. But in addition, we can grasp the concept of a triangle as a closed plane figure with three straight sides, and are enabled thereby to entertain propositions about triangles and to engage in geometrical reasoning that reveals to us further things about their nature. We can see, hear, and touch a particular human being, and can form mental images of other actual and possible human beings. But in addition, we can grasp the concept of a human being as a rational animal, and are enabled thereby to entertain propositions and to reason about human beings. And so on. 

Against imagism 

But might our ability to entertain concepts not be merely an extension of what other sentient organisms are capable of, differing from it in degree rather than kind? That is the implication of a view known as “imagism,” which holds that concepts are just mental images. It is commonly associated with the early modern empiricist tradition. 109 To see why imagism cannot be true illuminates the nature of concepts and of how the intellect differs from mere sentience. To be sure, one does not need to be an empiricist or a philosopher of any other school of thought in order to suppose that a concept is a kind of mental image. The average person may think that to have the concept of a triangle, for example, is to have a picture of a triangle before the mind’s eye. Certainly it is true that when we think about triangles, we tend to form such imagery, and that when we think about trees, dogs, or whatever, we tend to form images of those things. All the same, the concept is not and cannot be the same thing as any image, for several reasons. 110 

Consider first that concepts are universal. For example, the concept of a triangle as a closed plane figure with three straight sides applies to all triangles, whether right, acute, or obtuse, whether isosceles, scalene, or equilateral, whether drawn in ink, pencil, or chalk, whether black, red, or green in color, and so on. The concept abstracts from these varying features that particular triangles might possess and captures only what is common to every member of the class. But no mental image of a triangle can be similarly abstract and universal. Any image will be, say, of a green equilateral triangle, or a red right triangle, or what have you. It will always be an image of some particular triangle, having concrete features that not all triangles have in common. Hence the concept cannot be identified with any image. The point is even more obvious when we consider more abstract concepts – that is to say, concepts applying to even wider varieties of thing. For example, the concept of being a living thing applies to bacteria and algae, dandelions and oak trees, snails and spiders, mice and lions, frogs and sharks, starfish and Tyrannosaurus Rex, human beings and (if there are any) extraterrestrials. But there is no image one can form that comes remotely close to resembling all of these things. For any mental picture of a living thing will have too many concrete features unique to living things of the kind pictured, and will be missing features possessed by living things of other kinds. 111 

A second difference between concepts and mental images is that many concepts have a clarity and distinctness that no image can possess. Consider, for example, the mental pictures you might form of a crowd of five hundred people, a crowd of seven hundred and fifty people, and a crowd of one thousand and twenty-four people. The images are all bound to be roughly the same. But the concept of a crowd of five hundred is clearly and distinctly different from the concept of a crowd of seven hundred and fifty, which is clearly and distinctly different from the concept of a crowd of one thousand and twenty-four. Or consider an example famously used by Descartes to make the same point. 112 A chiliagon is a geometrical figure with one thousand sides, and a myriagon is a geometrical figure with ten thousand sides. A mental picture you might form of one of these will look more or less the same as the image you form of the other, and they will resemble too any mental image you might form of a one thousand and twenty-four sided figure, or indeed of a circle. There is no clear and distinct difference between any of these images. But there is a clear and distinct difference between the concept of a chiliagon, the concept of a myriagon, the concept of one thousand and twenty-four sided figure, and the concept of a circle.

 So far we have been considering examples of things of which we can form at least some image, however limited its applicability to all the members of a class. But a third problem with imagism is that there are some things of which we have concepts, but of which we can form no image at all. Consider, for example, the notion of a geometrical point, which has a location in space but no length, width, or depth. When we try to form an image of a point, we are bound to picture a dot, like the period at the end of this sentence. But that is not, strictly speaking, a point, since it has length and width and a point does not. Or consider the geometrical notion of a line, which has length but no width or depth. This too cannot strictly be imagined, for when we try we always end up picturing something with length and some minimal width. 

Consider also abstractions like the concept of law. Law is not a physical object you can see or otherwise perceive, and thus it is not something of which you could form an image. Of course, you might picture in your mind a courtroom, jurors, and someone in a judge’s robe. But that is merely to imagine things that are contingently associated with law, not law itself. Law might exist even if these things did not. Or you might call before your mind the word “law,” either its appearance as written or the sound you hear when it is spoken. But that too is not the same thing as law itself. Law existed in ancient China and ancient Rome, but the English word “law” did not. Even more abstract is a concept like the validity of an argument. You can easily grasp the validity of an argument like: All men are mortal, and Socrates is a man; therefore, Socrates is mortal. The conclusion plainly follows of necessity from the premises. But you cannot imagine validity in the sense of forming a mental picture of it. Validity is simply not the sort of feature that has a certain shape or color, sound or taste, texture or smell.

Concepts, then, simply cannot be identified with mental images of any kind. Nor, for similar reasons, can the propositions or complete thoughts that we build out of concepts be identified with images or sets of images. Take, for example, a thought the content of which is the proposition that snow is white. You might at first suppose that to have that thought is nothing more than to call before your mind the sentence “Snow is white.” But on further reflection that clearly cannot be correct. Someone who speaks German but no English can nevertheless entertain exactly the same proposition and in that sense have exactly the same thought. He would express that proposition in a sentence like “Schnee ist Weiss,” but the proposition itself, and its constituent concepts, are the same as those you express in English. 113 

It might seem that propositions could nevertheless be identified with images of things other than words and sentences. Take a thought the content of which is the proposition that the cat is on the mat. Might that not be identified with a mental picture of a cat sitting on a mat? But on closer consideration, this clearly won’t work either. 114 Mental images, like all pictorial representations, are inevitably ambiguous or indeterminate in their content. For example, suppose what you picture mentally is a gray cat sitting on a brown mat. Exactly which proposition is supposed to be identified with this mental picture? Is it the proposition that the cat is on the mat? Or the proposition that an animal is on the mat? Or the proposition that a gray cat is on a mat? Or the proposition that a cat is on a brown mat? Or the proposition that Tabby is on the mat (assuming that that is the cat’s name)? Or one of any number of other propositions that might be taken to describe the state of affairs pictured? There is nothing in the image itself that can tell you. An image is always at best ambiguous between alternative possible propositional contents. But a thought can be unambiguous in its propositional content – in which case a thought cannot be identified with any pictorial mental image. 115 

Then there are propositions the content of which is not even prima facie plausibly captured by any pictorial image. For example, consider propositions about a particular point in time, such as the proposition that it will rain at noon. You might call to mind a mental picture of rain falling during the daytime. But what in the image could capture the idea that it is raining at noon, exactly, as opposed to 11 am or 1 pm? What about cases where it is raining at noon, but there is no sunlight at all (for example, because the rain is occurring during the winter months in Alaska)? You might suppose that you could capture the time by adding to the mental picture an image of a clock that reads “12 pm.” But what is there in this new image that can determine that it corresponds to the (future tense) proposition that it will rain at noon, as opposed to the (present tense) proposition that it is raining at noon or the (past tense) proposition that it rained at noon – or indeed, as opposed to the proposition that it is raining at 11 am and there is a broken clock nearby that reads “12 am”? How would we represent a situation where it is raining at noon and there is no clock present? 

Consider also negations, which are propositions about what is not the case. For example, how could the proposition that it will not rain at noon correspond to a mental picture? By imagining, say, an ordinary contemporary street scene from which rain is absent? But why would that correspond to the proposition that it will not rain at noon as opposed, say, to the proposition that there are no extraterrestrials in the city, or the proposition that there are no dinosaurs in the city, or to any other of a countless number of propositions about what is not true of an ordinary street scene? 

Or consider the difference between conditionals (which are if-then propositions), conjunctions (which are both-and propositions), and disjunctions (which are either-or propositions). How would the proposition that if it is raining, then there will be bad traffic be conveyed in a mental picture? By imagining a scene in which it is raining next to a scene in which there is bad traffic? But why would this correspond to the proposition that if it is raining, then there will be bad traffic, rather than the proposition that it is raining and there is bad traffic, or the proposition that either it will rain or there will be bad traffic? Here too, nothing in the mental images themselves can determine which of several alternative propositional contents is the one being conveyed. 

Now, some philosophers have denied that mental images are best thought of on the model of pictures. Against this “pictorialist” conception of imagery, they have proposed the “descriptionalist” thesis that a mental image is best thought of on the model of a linguistic description of what is represented by the image. 116 Would imagism be more plausible given this alternative account of the nature of imagery? It would not be. For what exactly does a linguistic description of the sort in question amount to? If it amounts to a mental image of a spoken or written sentence, then the account will have exactly the same problems we’ve already considered. On the other hand, if we think of the linguistic description on the model of a proposition that might be expressed by a sentence, then we are no longer really talking about a theory that reduces thoughts to mental images, but rather one that goes in the opposite direction – that is to say, one that assimilates mental images to thoughts. Perhaps we even have a theory that implicitly denies that there are mental images, and instead posits only thoughts. 117 Whatever one thinks of such a view, it hardly amounts to a defense of imagism. 

It is important to emphasize that none of these critical remarks about imagism are meant to deny the familiar fact that when we entertain concepts and propositions, we do tend to call appropriate images to mind. For example, when you have the thought that the cat is on the mat, you may well “hear” in your mind the sentence “The cat is on the mat,” or you may visualize a cat on a mat. When you entertain the concept of being a triangle, you may well form a mental picture of a triangle. And so on. Indeed, we will see later on that such imagery is a necessary precondition of the normal functioning of the human intellect. The point, though, is that forming images is nevertheless distinct from entertaining concepts and propositions, even if it is associated with it. It is at most a necessary condition for thinking, and cannot be a sufficient condition. 118 

It is no less important to emphasize that the rejection of imagism by no means entails a commitment to the doctrine of innate ideas, which holds that at least some concepts are built into the human mind before anything enters into it via the senses. To be sure, in early modern rationalism, this doctrine came to be associated with the rejection of imagism. The assumption was that if concepts are not images, then they cannot have gotten into the mind via the sensory experiences of which images are faint copies. Meanwhile, the early modern empiricists, who rejected innate ideas and held that there can be no concepts in the mind without antecedent sensory experiences, concluded that concepts must be nothing more than images or copies of such experiences. The assumption was that one had to take either the rationalist option of rejecting imagism and adopting innate ideas, or the empiricist option of rejecting innate ideas and embracing imagism. 

But this is a false choice. A third option would be the AristotelianThomistic position that predated the modern rationalist/empiricist dichotomy. It holds that there can be no concepts in the mind without antecedent sensory experience and the imagery to which it gives rise, but also that the intellect can by abstraction draw concepts out of images, where the abstracted concepts are distinct from any image. Hence it rejects both innate ideas and imagism, both rationalism and empiricism in its modern, post-Lockean form. 119 Contemporary neo-empiricist philosophers like Jesse Prinz, who agree that imagism is false, nevertheless argue that some modified form of post-Lockean empiricism must be correct, on the grounds that rationalist accounts of concepts face serious philosophical and empirical difficulties. 120 What they fail to see is that the Aristotelian-Thomistic position provides another alternative. Indeed, Aristotelian-Thomistic philosophers would argue that the insuperable problems facing both imagism and innate ideas imply that there must be such an alternative. 121 We will have reason to return to these issues presently and in later chapters. Again, the point to emphasize for the moment is simply that rejecting imagism does not entail denying that sensation and imagination play crucial roles in the operation of the human intellect. 

What are concepts? 

To reject imagism is to say something about what concepts are not. Can we say more about what they are? We can, and we can start by noting what contemporary neo-empiricists and neo-rationalists alike agree on concerning their nature. 122 

First of all, a concept is a mental representation of the most fundamental sort. Take the concept of being a triangle. It represents triangles, just as the word “triangle” does. In the case of both the word and the concept, we therefore have what contemporary philosophers call intentionality. This is a technical term which must not be confused with what we have in mind when, in everyday contexts, we use words like “intentional” or “intentionally.” In particular, it has nothing to do with acting on purpose. Rather, for something to have intentionality in the technical philosophical sense is for it to aim at, be directed toward, or be about something (typically, though not always, something distinct from itself). For example, the word “triangle” aims at, is directed toward, or about geometrical figures of a certain kind, and the word “cat” aims at, is directed toward, or about animals of a certain kind. By contrast, the nonsensical sequence of letters “shmat” is not aimed at, directed toward, or about anything. As these examples indicate, it is the possession of intentionality that makes words meaningful, whereas it is the lack of intentionality that makes a sequence of letters like “shmat” meaningless. 

Now, though words like “triangle” and “cat” possess intentionality, they do not have it in a built-in or inherent way, but only as a matter of convention. Considered just by itself, the sequence of letters “cat” has no more inherent meaning than the sequence “shmat” has. Had the history of the English language gone differently, we might have used “shmat” to refer to animals of the feline type and “cat” as an example of a meaningless string of letters. Mental representations like concepts are unlike linguistic representations in this respect. Unlike the word “triangle,” the concept of being a triangle does have its intentionality in an inherent or built-in way. To aim at, be directed toward, or be about triangles is simply part of what it is to be the concept of being a triangle. There is just no such thing as a concept of being a triangle that doesn’t have such intentionality. To borrow John Searle’s way of putting the distinction, whereas linguistic representations like words have only “derived intentionality,” mental representations like concepts have “intrinsic intentionality.”123 

Keep in mind that the concept of being a triangle is not to be identified with any mental image, including a mental image of the way the English word “triangle” looks when written or the way it sounds when spoken. So, when I say that the concept of being a triangle has intrinsic intentionality rather than merely derived intentionality, I am not saying that the image you form when you see the word “triangle” in your mind’s eye, or the sound you bring to mind when you think about someone uttering it, has intrinsic intentionality. The intentionality of such mental imagery is as derived as that of a written or spoken word. Again, someone who speaks a language other than English and has never seen or heard that particular English word can nevertheless have the concept of being a triangle. He would call to mind different images than an English speaker would when contemplating triangles, but the concept would be identical to the one that an English speaker possesses. (If you’re trying to visualize concepts and finding it difficult, then you need to stop trying. A concept is not the sort of that can be visualized. If you are visualizing something, then what you’ve got is an image – and again, concepts are not images.) 

It is important to emphasize that to characterize a concept as a “mental representation” does not commit one to everything that has been associated with that phrase in the history of philosophy. In particular, it does not commit one to the view that concepts or any other mental representations are themselves the primary objects of our thoughts. When you think about triangles, it is indeed triangles that you are thinking about, rather than the concept of being a triangle. The concept is (usually) not itself the object of your thought, but rather that by which you grasp the object of your thought (just as eyeglasses are usually not themselves the object you are looking at, but rather that by which you perceive the object you are looking at). To characterize concepts as mental representations entails only that they are that by which the intellect represents things. It does not entail the rightly maligned early modern rationalist and empiricist model of the mind as directly aware only of its own ideas as they flit across the stage of an inner theatre. 124 

A second area of common ground between different theories of concepts is that concepts are that by which we categorize or classify things. Triangles are set apart from squares insofar as the former fall under the concept of being a triangle and the latter do not; cats differ from dogs insofar as they fall under the concept of being a cat and dogs don’t; and so on. (The class of things that fall under a concept is referred to by philosophers as the extension of the concept.) As with intentionality, this is an inherent feature of concepts. If something is the concept of being a cat, then it simply could not be the case that cats did not fall under it, or that non-cats (dogs, say) did. (That is not to say that there couldn’t, as a matter of contingent fact, be a situation where nothing fell under the concept. For example, if cats went extinct, then the extension of the concept of being a cat would be empty. The point is rather that even in that case, there could still in principle be something that fell under it, and if something did so, it would have to be a cat.) 

Note that I am not here addressing the question of how we know whether something falls under a certain concept. Nor am I addressing the question of whether all, or even any, of the classes represented by our concepts are objectively real or merely constructs of the mind. Those are very important questions, but they are beside the present point, which is simply that to be a concept is to be the sort of thing under which a class of things falls (however we know that something falls under it, and whether or not the class in question is something we invent rather than discover). 

A third widely acknowledged feature of concepts is what philosophers refer to as their compositionality. They can be combined into more complex concepts and, as noted already, into propositions. 125 For instance, the concept of being a triangle and the concept of being equilateral can be combined to yield the concept of being an equilateral triangle. The concept of being a man and the concept of being mortal can (together with what is called the universal quantifier) be combined to form the proposition that all men are mortal. (In fact, appeal to the role they play in propositions is another way in which contemporary philosophers often explain what it is to be a concept. A concept can be thought of as a constituent of a proposition, in something like the way a brick can be understood as a constituent of a wall.) 

As Jerry Fodor and Zenon Pylyshyn have emphasized, the compositionality of concepts accounts for two important features of thought, namely its productivity and systematicity. 126 Thought is productive or fertile in the sense that, out of a finite number of primitive concepts and rules of combination, we can construct an infinite number of compound concepts and propositions. To consider some trivial examples, we can combine concepts in such a way as to yield the proposition that there are two cats on the mat, or the proposition that there are three cats on the mat, or the proposition that there are four cats on the mat, and so on ad infinitum. Using connectives like “and” and “or,” we can construct compound propositions out of elementary ones. For example, from the proposition that the cat is on the mat and the proposition that the dog is on the log, we can form the compound proposition that the cat is on the mat and the dog is on the log. We could add another “and” and another proposition to yield the more complex proposition that the cat is on the mat and the dog is on the log and the mouse is in the house. And so on ad infinitum. Thought is systematic in the sense that the capacity to entertain one proposition typically goes hand in hand with the capacity to entertain others. For instance, I can entertain the proposition that the dog is on the log, but also the proposition that the log is on the dog. I simply put the same concepts together using the same rules of combination, but in a way that yields a different compound. 

A fourth attribute of concepts widely acknowledged by theorists of different stripes is their public character. One and the same concept can be entertained by different people, and by the same person at different times. For example, when you entertain the concept of being a triangle and I entertain the concept of being a triangle, it is one and the same concept that we are entertaining. By the same token, when each of us entertains the proposition that triangles have three sides, it is one and the same proposition we are entertaining. It is not that you are entertaining your own private concept or proposition and I am entertaining mine. If concepts were private, we could not so much as communicate with one another. We not only could not agree with one another, we could not disagree either. Unless you and I had the same thing in mind when we entertained the concept of being a triangle, we could neither agree that all triangles have three sides nor disagree about whether Feser overuses the triangle example. 

Now, it might seem that people often do not have the same concepts, given the very different beliefs and levels of knowledge they have about things. For example, a chemist has a much richer understanding of what it is to be water than a layman does, and libertarians and socialists have very different ideas about what justice requires. Can we really say, then, that the chemist and the layman have one and the same concept of water, or that libertarians and socialists have one and the same concept of justice? Yes, we can. Philosophers sometimes draw a distinction between concepts and conceptions. 127 A libertarian and a socialist might agree that a just society is one in which each person has what he is entitled to. But whereas the libertarian might say that people are entitled only to what they work for or are freely given, the socialist might say that they are entitled to an equal share of society’s wealth. We might in that case say that they have the same concept of justice, but different conceptions of what justice requires. A layman and a chemist might agree that water is a clear, tasteless, odorless substance that is liquid at room temperature, freezes when very cold, is the stuff found in lakes and rivers, and so on. But whereas that is more or less all that the layman has to say about the nature of water, the chemist will be able to go on to explain that water is a composition of hydrogen and oxygen, to say exactly how it will behave under various conditions and why, and so forth. We might say that the layman and the chemist have the same concept of water, but the chemist has a much more detailed conception of what water is. There are different ways in which this distinction between concepts and conceptions might be spelled out, but for present purposes it will suffice simply to note that the differences in the beliefs and levels of knowledge people have about a class of things do not by themselves entail that they are operating with different concepts. 128 

So much for what the different sides in the debate over concepts tend to agree about. Let us turn now to an issue they do not agree about. Apart from the question of whether any concepts are innate, the main point of contention between contemporary neo-rationalists and neo-empiricists is whether concepts are mental representations of a kind distinct from the representations associated with the different sensory modalities. The neorationalist view is that they are of a distinct kind – that concepts are “amodal” representations, representations in a “central code” unique to thought and distinct from the diverse ways information is coded in the different sensory modalities. The neo-empiricist view is that there are no such amodal representations, so that a concept is a representation in a modally-specific code. Hence, for the neo-rationalist, the concept of being a cat is a mental representation that is amodal in the sense of being distinct from any visual, auditory, tactile, gustatory, or olfactory perceptual representation of a cat. But for the neo-empiricist, the concept of being a cat is not amodal, but is instead encoded in one or more perceptual representations – for example, in a visual representation of how a cat looks or an auditory representation of the sound it makes. 129 

To be sure, and as noted already, neo-empiricists agree that concepts cannot be identified with mental images. Hence they do not regard the perceptual representations that encode concepts as representations of an imagistic kind, specifically. They would characterize the relevant representations in some alternative way. For example, a visual perceptual representation of a cat might be identified with a neural mechanism in the visual processing center of the brain that is causally related in the right sort of way to some cat in the external environment. An auditory perceptual representation of the cat would be identified with a mechanism in the auditory center of the brain that has the right sort of causal relation to the cat. And so forth. Distinct perceptual representations of various cats can then get linked together and stored in what Prinz calls a “long-term memory network” in the brain. 130 Activity in this network might in turn later generate a new perceptual representation that functions as a “simulation” of the sort of representation that might be triggered in the presence of a cat. 131 Insofar as this simulation can serve as a proxy for the long-term memory network causally correlated with cats, Prinz labels it a “proxytype.”132 The concept of being a cat can, on his version of neo-empiricism, be identified with such a proxytype. But the same simulation might serve as a different proxytype, and thus a different concept, depending on which long-term memory network triggers it. For example, a simulation of a visual perceptual representation of a cat that is generated by the long-term memory network causally correlated with cats in general will count as the concept of being a cat. But the same simulation would, if generated instead by the long-term memory network causally correlated with Persian cats, specifically, count as the concept of being a Persian cat. 

This account of concepts is less crude than the imagism of early modern empiricism, but it is nevertheless in the same spirit. Like imagism, neoempiricism essentially reduces the intellect to the activity of what the Aristotelian-Thomistic tradition calls the “internal senses.” It can capture at best the capacities of the most sophisticated sentient but non-rational forms of life, but not what is distinctive about concepts and the cognitive activities that concepts make possible, such as the entertaining of propositions and reasoning through arguments. In fact, despite its eschewal of imagism, neoempiricism is open to objections that are analogous to those that it acknowledges are fatal to imagism. 

Consider first the question of what gives a proxytype the specific conceptual content it is supposed to have – for example, what makes it the concept of being a cat, as opposed to the concept of being a Persian cat. Prinz’s answer, again, is that it has to do with which “long-term memory network” in the brain causes it. But this is no answer at all. For one thing, a proxytype will be causally related to all sorts of other mechanisms in the brain as well. So why is it the long-term memory network that determines the content of the proxytype, rather than one of these other mechanisms? For another thing, even if we concede that it is the proxytype’s causal relation to the long-term memory network that is relevant, this just kicks the problem back a stage. For what, exactly, gives a particular long-term memory network the precise content that it has? Suppose such a memory network is causally linked with cats. It will also be causally linked with all sorts of other things. For example, it will be causally linked with the ancestors of cats, with the light and sound waves that mediate our perceptual awareness of cats, and so on. So what makes it a memory network related to cats, as opposed to a memory network related to ancestors of cats, or one related to light and sound waves? 133

The problem, then, is that proxytypes and the sets of causal relations they enter into are, considered by themselves, systematically ambiguous or indeterminate in their content. There is nothing in them that can suffice to determine that they represent being a cat, say, as opposed to being a Persian cat. But concepts can have an umambiguous or determinate content. Hence concepts cannot be identified with proxytypes. Now, the problem of the ambiguity or indeterminacy of the content of imagism was among the problems facing imagism. Hence neo-empiricism is in this respect no improvement over imagism after all. 

Like imagism, neo-empiricism also has difficulty accounting for concepts that are highly abstract – in particular, those that are so abstract that they do not plausibly correlate with any perceptual representation any more than they correlate with any image. Earlier we considered examples like the concept of law, the concept of the validity of an argument, the notions of the negation and disjunction of propositions, and so on. None of these corresponds to a concrete physical object that might be seen, heard, tasted, touched, or smelled, and thus there can be no perceptual representations of them even of the kind the neo-empiricist speaks of, any more than there can be images of them. Prinz suggests that the neo-empiricist could appeal instead to perceptual representations of words, such as written and spoken instances of the English words “law,” “validity,” and so forth. 134 But this is as hopeless a strategy as identifying concepts with images of such words. For the meanings of these words are entirely conventional. “Law,” for example, has the meaning it does only because it is conventionally used to express the concept of law. Hence we already have to have some grasp of that concept independently of and prior to the introduction of the word. There is also the problem, noted earlier, that those who don’t speak English and have never encountered the word “law” can nevertheless have the very same concept of law that English speakers have. 

Then there is the problem that neo-empiricism cannot account for the way concepts enter into propositions and arguments. Consider the following inference: 

All tigers are cats. 

Tony is a tiger. 

Therefore, Tony is a cat. 

This inference is clearly valid. Now, its validity presupposes that each of the terms expresses the same concept from step to step. Otherwise we would have a fallacy of equivocation. For example, if “cat” conveyed the concept of being a cat in the first premise, but the concept of being a Persian cat in the conclusion, then the conclusion would not follow. According to Prinz’s theory of proxytypes, though, the same simulation of a perceptual representation can count as a different concept depending on which long-term memory network triggers it. For example, if causally linked to one such network, it will count as the concept of being a cat, and if causally linked to another, it will count as the concept of being a Persian cat. So what would guarantee that the same proxytype, and thus the same concept, is being deployed in both the first premise and the conclusion of the inference above, thereby preserving validity? 

Prinz’s answer would be that the relevant perceptual representations associated with the first premise and with the conclusion of the inference are triggered by the same long-term memory network. But now we have a problem. For causal relations of the kind in question are contingent. They could have been other than they are. But logical relations between concepts are necessary. If all tigers are cats and Tony is a tiger, then it cannot possibly fail to be the case that Tony is a cat. Proxytypes simply don’t bear the sorts of relations to one another that concepts bear to one another, and thus the latter cannot be identified with the former. The most the neoempiricist can identify are contingent causal relations which we might for some purposes treat as if they had the logical form a valid inference has. But what we need for an account of concepts is something that does in fact have that form. 135 

Nor does the empirical evidence neo-empiricists appeal to in defense of their position show what they think it does. Studies indicate that when asked to carry out conceptual tasks (such as listing a thing’s features), subjects will form simulations of perceptual representations of the things the concepts are concepts of. 136 But the most this shows is that the entertaining of concepts is associated with simulations of perceptual representations. It does not show that the entertaining of concepts is reducible to simulations of perceptual representations, any more than the fact that we tend to form mental images when we entertain concepts shows that concepts are reducible to mental images. With simulations of perceptual representations as with images, even if they were necessary for the entertaining of concepts, it would not follow that they were sufficient. 

There is also empirical evidence that at least some cognitive tasks are not tied to simulations of perceptual representations. For example, there seems to be no difference in the performance of subjects asked to estimate how many members there are in a set, whether the set involves visible objects or sounds. Nor is there difference in performance when they are asked to add their estimations of the number of things in two sets of objects known via the same sensory modality, or to add their estimations of the numbers of things in two sets known via different modalities. Hence such cognitive tasks do not seem tied to any particular sensory modality, but rather to involve some amodal form of representation. 137 

In short, the empirical arguments for neo-empiricism are weak at best and non sequiturs at worst, and the conceptual problems with it are as grave as those facing imagism. 138 To that extent, in the debate between contemporary neo-empiricists and neo-rationalists, the latter have the better of the argument. Neo-rationalists are correct to regard concepts as mental representations of an amodal kind, in the sense of being different from perceptual representations of any sort (whether images, proxytypes, or whatever). Now, one way in which the notion of an amodal mental representation might be developed is in terms a distinction drawn by the Thomist philosopher John Poinsot, also known as John of St. Thomas (1589- 1644). Poinsot distinguished between instrumental signs and formal signs. 139 Both sorts, being signs, signify or represent things. Accordingly, they have what I referred to earlier as intentionality. Now, an instrumental sign is a sign that is also something other than a sign. For example, a written or spoken instance of the word “cat” is a sign or representation of cats. But it is also something else, such as a set of ink splotches or sound waves. That is why its intentionality is of what Searle calls the “derived” kind. A formal sign, by contrast, is a sign that is nothing but a sign. Its entire nature is to represent or be about something, and there is nothing to it over and above its doing so. Its intentionality is therefore of what Searle calls the “intrinsic” kind. Its specific intentional content or meaning is built into it precisely insofar as it exhausts what it is. Now, concepts (and also the complete thoughts we entertain, which are built up out of concepts) are in Poinsot’s view signs of this type, formal rather than instrumental signs. There is nothing more to them than their representational content, which makes them not only unlike words in a language like English, but also unlike images and proxytypes, which have some nature over and above their representational content (such as resemblance in the case of images, and causal relations in the case of proxytypes). 

Thinking of concepts as formal signs gives us a way to see what is wrong with an objection raised by Prinz, who suggests that the view that concepts are amodal representations faces the same problems that its empiricist rivals do.140 If something could be a mental image of a cat or a simulation of a perceptual representation of a cat, and yet nevertheless fail to represent cats in general (as the concept of being a cat does), then why couldn’t an amodal mental representation fail to do so too? But the problem with this objection is that it implicitly assumes that amodal mental representations too would be instrumental signs, signs that have some nature over and above their representational content – so that, like images and proxytypes, they might in principle exist even if they did not have the representational content they do. But that is precisely what Poinsot’s proposal denies. What sets concepts qua formal signs apart from words, mental images, proxytypes, etc. is that there is no gap between their nature as signs and their representational content, precisely because, again, there is nothing more to their nature than their representational content. They are, to borrow an apt phrase from P.M.S. Hacker, “all message and no medium.”141 

The key argument for the existence of formal signs is that without them there could not be any signs at all. Instrumental signs, such as words in a natural language like English, derive their intentionality from other signs, as when a new word is defined in terms of existing words. Some of these other signs too might be instrumental, as when the existing words in question are defined in terms of yet other existing words. But not all of the signs from which these instrumental signs derive their meaning can be instrumental, on pain of vicious infinite regress or circularity. There must be signs which simply have their intentional content in an intrinsic or built-in way, or there would be nothing that signs with merely derived intentionality could derive it from. But to be a sign with such intrinsic intentionality is to be something whose entire nature just is its intentional content, something which does not have some distinct nature to which intentional content must be added. It is to be a formal sign. Now, concepts are precisely the sources of the derived intentional content possessed by merely instrumental signs like words in a natural language. They are what in fact do the job that only formal signs can do. Hence concepts are formal signs. 142 

Now, the contemporary neo-rationalist thesis that concepts are amodal mental representations is also often associated with a linguistic conception of these representations. The idea is that the right way to model concepts and propositions is on words and sentences, as in Fodor’s influential “Language of Thought Hypothesis.”143 Does our discussion so far commit us to this? Certainly concepts are like words in that they represent without bearing any sort of physical resemblance to what is represented, and insofar as they exhibit a kind of semantic compositionality which, as Fodor has rightly emphasized, makes possible a productivity and systematicity in the case of thought that parallels the productivity and systematicity of natural languages. Because concepts and propositions are in such respects analogous to linguistic phenomena, language provides at least a useful heuristic model for thought. 

However, this by no means entails a commitment to everything associated with the notion of a Language of Thought or to other themes of contemporary neo-rationalism. 144 It does not entail supposing that any concepts are innate. Nor does it entail agreement with Fodor’s physicalist supposition that mental representations are material symbols encoded in the brain. For that reason, it does not entail supposing that thinking is to be identified with the processing of symbols according to the rules of a computer program. In fact, all of these suppositions are false, as I will argue later. But it suffices for present purposes simply to note that what has been said so far leaves such questions open. 145 

Thought and language 

This naturally leads us to the question of the relationship between the intellect and language. Is the possession of language a necessary condition for having the capacities to grasp concepts, entertain propositions, and reason from one proposition to another? Might the possession of these intellectual capacities even be entirely reducible to the possession of a capacity for linguistic behavior? 

The latter proposal is famously associated with behaviorism. It is widely understood to be a further variation on the modern empiricist project of giving a reductionist account of the intellect. Instead of identifying the possession of concepts and thoughts with the having of mental images or simulations of perceptual representations, the behaviorist essentially identifies them with the capacity for certain kinds of linguistic utterance and other forms of behavior. For example, to have the concept of being a cat or the thought that the cat is on the mat would, on the behaviorist analysis, amount to the disposition to sort cats from dogs and other animals, to utter or assent to the sentence “The cat is on the mat” under certain circumstances, and so forth. Behaviorism is widely rejected today, but it is useful for the sake of completeness at least briefly to summarize the main problems with it. B. F. Skinner’s book Verbal Behavior is perhaps the best-known attempt to provide a behaviorist account of thought and language, precisely because it was the target of a famously devastating critique by linguist Noam Chomsky. 146 The objections raised by Chomsky and his followers against Skinner are applicable to behaviorism in general. 

Perhaps the most fundamental problem with behaviorism is that it has proved impossible to replace attributions of thoughts and other mental states with nothing more than descriptions of behavior and dispositions for behavior. Consider, for example, the suggestion that to believe that the cat is on the mat is to have a disposition to utter or assent to the sentence “The cat is on the mat.” A person will have such a disposition only under certain conditions, such as his having a desire to utter or assent to the sentence (which he might not have if, say, he wants to hide the presence of the cat from someone). So, analyzing one propositional attitude (the belief that the cat is on the mat) in terms of a behavioral disposition leads us to posit a further propositional attitude (the desire to utter or assent to a certain sentence). For the behaviorist reduction to be carried out completely will thus require an analysis of that further propositional attitude in terms of a behavioral disposition. But it turns out that such further attempted reductions only ever seem to raise the same problem over again. The behaviorist reduces a given propositional attitude to the having of a disposition for a certain kind of behavior, where the having of that disposition cannot be made sense without reference to a further propositional attitude, which the behaviorist attempts to reduce to a further disposition, which itself cannot be made sense of without reference to a propositional attitude, and so on ad infinitum. 

As Chomsky pointed out, the problem afflicts even the application of the behaviorist’s own fundamental theoretical notions. Skinner and other behaviorist psychologists hoped to explain the acquisition of language in terms of the responses of the language user to stimuli, and in particular in terms of those responses that are reinforced by rewards and punishments. The idea was to model even the most sophisticated linguistic behavior on the same basic mechanism by which behaviorists attempted to mold the behavior of rats and pigeons in the laboratory. A rat might be trained to press a certain bar (a response) in reaction to hearing a bell (a stimulus) by way of rewarding such behavior with a food pellet (a reinforcement). Once such conditioning is in place, further behavioral dispositions can be established by way of “response chaining,” in which a stimulus associated with a certain reward itself comes to reinforce a further response. For example, because of its association with the food pellet, the rat’s hearing of the bell might come to reinforce some further response like doing a dance upon seeing a light. 147 Mastery of language can, on the behaviorist analysis, be explained in terms of the operation of similar mechanisms.

But what counts as a stimulus or a response in the case of distinctively human behavior? To borrow an example from Chomsky, suppose someone makes a remark in the presence of a painting. 148 What exactly is the stimulus? We have no way of saying without first knowing the content of the remark. If the person says “It clashes with the wallpaper,” we might identify the stimulus as the color of the painting; if he says “It’s tilted,” we might identify the stimulus with the frame; if he says “It’s beautiful,” “Hideous,” or “I thought you liked abstract work,” we might identify yet some other feature as the stimulus (though which, exactly, is not so clear). But this means that we can’t characterize the stimulus in a way that is independent of the response of the person. Moreover, the characterization will presuppose facts about the mental states of the person – about how the painting looks to the person, what features he is concentrating his attention on, and so forth. This is even more obvious when the stimulus is claimed to be some complex phenomenon remote from the person whose utterance is the response, as when (to borrow another example from Chomsky) a confusing international situation triggers the comment “This is war.”149

 What counts as the verbal response is no easier to determine in a way that is independent of the mental states of the speaker. Suppose someone utters “The cat is on the mat” in the presence of a cat. Why is that to be regarded as the response to the stimulus, as opposed to just the words “on the,” or as opposed to some longer string of sounds made in the course of uttering the sentence (a cough, followed by a stammered “Um, the cat is on the mat,” followed by a grunt)? It is hard to see how to answer that question without making reference to the content of the speaker’s thought that the cat is on the mat, the convention we have of using words like those that make up “The cat is on the mat” to convey meaning (whereas there are no such conventions in the case of coughs and grunts), and so on. 

Then there is the fact that even the behaviorist tends to understand what counts as a reinforcement of a piece of behavior in terms of how we perceive and conceptualize things. Chomsky notes, for example, that Skinner characterizes what will reinforce a response in an artist or writer in terms of what the artist or writer hopes for or imagines happening in the future. 150 Again, the behaviorist’s own basic conceptual tools for re-describing human beings in a way that makes no reference to what is going on in their minds (notions like stimulus, response, and reinforcement) end up presupposing facts about what is going on in their minds.

There are other serious problems. For example, we exhibit many responses that are entirely novel and thus have never been reinforced in the past. For example, someone who has never been mugged or otherwise in physical danger before will turn his wallet over to a mugger if refusing to do so might cost him his life. 151 It is easy to understand this if we think of the victim as playing out the scenario in his mind, but not if, like the behaviorist, we confine our explanation to appeals to stimulus and response patterns that have been set up in the past. 

Then there are the details of distinctively linguistic behavior. Even response chaining cannot account for how we are able to put together sentences, as opposed to mere strings of related words (such as “up, down” or “north, south, east, west”). 152 Neither can it account for our mastery of the difference between disjunctives and conditionals. For instance, consider the two sentences “Either the cat is on the mat, or the dog is on the log” and “If the cat is on the mat, then the dog is on the log.” Building up a verbal response word-by-word via stimulus-response conditioning will not guarantee that a sentence that begins with “either” will continue with “or” as opposed to “then” after the word “mat” is reached, because the mechanism will be sensitive only to that immediately preceding word, and not to what was emitted by the speaker several words previously. 153 Especially influential is the “poverty of the stimulus” objection to behaviorist accounts of language, which notes that children are able to produce an indefinite number of grammatical sentences with a very high degree of accuracy (notwithstanding the occasional cute mistakes they are prone to make) on the basis of a relatively small number of examples provided by adults, some of which are themselves grammatically deficient. 154 

Reducing thought to language, then, is out of the question. We cannot make sense of linguistic behavior unless we take the intellect to be something real and distinct from that behavior even if its thoughts are expressed through it. Might we posit an even more radical separation of thought and language? Is thought not only distinct from language, but something that might exist entirely in the absence of language? 

Some arguments for this conclusion are certainly weak. For example, it sometimes pointed out that human infants and adults with certain cognitive impairments lack the ability to use language but nevertheless appear to possess some concepts. 155 The trouble with this argument is that human beings in their mature and healthy state do have the ability to use language, and even immature or impaired human beings are surrounded by fellow human beings who do in fact use it. So, for all the argument shows, infants and the impaired adults in question would not possess concepts if this larger linguistic context were not present. 

What about non-human animals which lack language altogether but nevertheless appear able to recognize and sort things? Doesn’t the presence of this ability show that they possess concepts? No, because such recognitional abilities are neither sufficient nor even necessary for the possession of concepts. 156 To grasp a concept, it is not enough to be able to put things into different groups (even a coin sorting machine can do that). One must have a grasp of the logical connections between concepts, as evinced in the capacity to make correct inferences, and that is typically manifested linguistically. (Whether any non-human animals possess language is a question I’ll address in a later chapter.) One can also possess a concept without knowing how to recognize or sort things that fall under it. To borrow some examples from Hacker, one can have the concept of being old without knowing how to identify pre-Cambrian rocks, and one can have the concept of being fake without knowing how to identity a fake painting. 157 

As followers of Ludwig Wittgenstein (1889-1951) emphasize, beings with the capacity for language are our paradigms for beings who possess the capacity to think (in the sense of grasping concepts, being able to entertain propositions, and being able to reason logically from one proposition to another). 158 It is precisely in the ability to use a word or phrase correctly that we typically manifest our possession of a concept, and certain intellectual activities (keeping records, codifying laws, sophisticated mathematics, etc.) would not be possible without written language. 

A powerful defense of the view that the cognitive abilities most distinctive of human beings cannot exist in animals without language has been developed by José Luis Bermúdez – powerful, and also ironic insofar as Bermúdez defends this conclusion in the context of a larger argument for attributing thoughts of other kinds to animals that lack language. 159 Consider first what Bermúdez characterizes as second-order cognition, or thinking about thought processes themselves, as we do when we recognize a defect in some plan or argument, try to come up with ways to remedy it, reflect on the logical connections between steps in an argument we find compelling, and so forth. Here the objects of our thoughts are not things outside the mind, but, again, thoughts themselves, and so in order to reflect on them we need some way to represent them to ourselves. Now, unconscious mental representations, like those posited in Fodor’s Language of Thought hypothesis, cannot be the form this representation takes, precisely because they are unconscious and the second-order cognitive processes in question are conscious. Mental images cannot do the job either, since images do not have inferential relationships to one another the way propositions do.160 The only sorts of representations that seem suited to represent thoughts and their inferential relations are linguistic representations. Language is, after all, compositional, productive, and systematic in a way that, as we noted earlier, thought is. Words combine into sentences and sentences into arguments in a manner that parallels the way in which concepts combine into propositions and propositions into chains of inference. This gives to language a fitness for representing thought that other media of representation lack. We are able to engage in second-order cognition precisely insofar as we can linguistically represent thoughts and their logical relationships. 

As Bermúdez argues, what is true of second-order cognition is true also of other distinctively human cognitive activities. For example, we are capable of higher-order desire, or the desire to have or to be free of certain desires. 161 This requires us to be able to represent to ourselves what having a certain desire involves and what its logical implications are for circumstances we have not experienced. We are capable of attributing beliefs, desires, and other propositional attitudes to other people (or as the jargon would have it, we possess a “theory of mind”). That requires entertaining propositions we may neither regard as true ourselves nor want to be true, but which we think others have the attitudes of belief or desire towards. We are able to entertain compound propositions of the kind discussed earlier, such as conjunctions, disjunctions, and conditionals. That too requires us to be able to entertain propositions in the abstract, and consider how their truth or falsity determines the truth or falsity of the more complex propositions they make up. We are able to consider the modal status of a proposition (that is to say, whether it is true or false of necessity or merely possibly), and we can grasp tense (that is to say, whether a proposition is true, was true, or will be true). This also requires us to grasp the proposition in the abstract and determine whether or not to attach a certain modal status or tense to it. All of these cognitive abilities require us to be able to represent to ourselves not particular things in the world (as we do in perception and imagination), but thoughts or propositions about things. Again, sentences provide a vehicle for doing so that no other medium of representation can. 

Now, Bermúdez does not deny that animals without language are in some sense capable of having beliefs and of changing them in the light of the evidence. He gives the example of a rat which comes to realize that a food reward will appear whether or not it presses a lever, and adjusts its future behavior accordingly. But he characterizes such belief modification as “direct” rather than “reflective.”162 It such a case, a perceptual encounter with the world produces an immediate alteration in how the world is represented. It does not involve the entertaining of propositions and considering their logical or evidential relations to one another. Nonlinguistic animals obviously can also have desires, but only what Bermúdez calls “goal-desires” rather than “situation-desires.”163 The object of the former sort of desire is some thing in the world (such as water, or prey), whereas the object of the latter sort of desire is that some proposition be true or situation obtain (for example, that it be true that one always has enough to eat and drink). Non-linguistic animals can also possess a rudimentary sort of “theory of mind.” Take the distinction between “simple seeing” and “epistemic seeing.”164 Characterizing the former entails identifying some object or situation that the perceiver is aware of, as when we say that the dog sees the food in front of it or sees that there is a man in the vicinity. But to characterize the former requires goes beyond this and attributing propositional content to the state of the perceiver that the perceiver might go on to relate to other propositions. For example, we might say of a human being not only that he sees that a man is present but that he thereby grasps perceptually the truth of the proposition that a man is present and can therefore go on to infer that, since all men are mortal, a mortal thing is present. Non-linguistic animals are capable of simple seeing and of being aware that other animals are engaged in simple seeing. An ape might not only see a banana but see that another ape sees the banana. But that is not the same thing as epistemic seeing or the attribution to other animals of epistemic seeing. Epistemic seeing and the attribution to others of epistemic seeing, like reflective belief modification and situation-desires, requires taking thoughts or propositions themselves as the object of one’s thoughts, and that requires linguistic representation. 

As Bermúdez notes, we can also distinguish between believing a general proposition, and having a disposition to form a certain kind of belief about particular things of a general type. 165 An example of the former would be believing that all tigers are dangerous. An example of the latter would be having a disposition, when in the presence of tigers, to form the belief that this thing is dangerous. To be capable of the latter does not entail being capable of the former. While non-linguistic animals are capable of something analogous to the latter, they are, argues Bermúdez, not capable of the former. The reason is that grasping a universal generalization requires representing a proposition in thought and breaking it down into its components, as when in modern formal logic we analyze the proposition that all tigers are dangerous as having the structure: For any x, if x is a tiger then x is dangerous. Again, to do this sort of thing requires representing the proposition in a sentence, and thus linguistically. 

Much more could be said, but the key point to emphasize is that words like “thought,” “belief,” and the like are not univocal, so that we need to disambiguate when addressing the question whether thinking requires language. If by “thinking” we mean the capacity for what Bermúdez calls direct belief modification, goal-desires, simple seeing, dispositions to form beliefs about things of a general type, and so forth, then non-linguistic animals are capable of thinking. But all of this falls within the range of capacities made possible by what Aristotelians traditionally call the “internal senses.” It doesn’t follow that non-linguistic animals are capable of what Bermúdez calls second-order cognition, higher-order desires, the attribution to others of propositional attitudes, reflective belief modification, a grasp of compound, modal, tensed, and general propositions, situationdesires and epistemic seeing, and so forth. These capacities are characteristic of human beings as animals with distinctively intellectual rather than merely sentient powers. If that is the sort of thing we mean by “thinking,” then there is good reason to conclude that non-linguistic animals are not capable of thinking. 

The word “language” is also not univocal and must be disambiguated before we can fruitfully address the question whether non-linguistic animals can think. Karl Popper drew a distinction between four main functions of language. 166 The first is the expressive function, which involves the outward expression of an inner state. Here language operates in a way comparable to the sound an engine makes when it is revved up, or an animal’s cry when in pain. The second is the signaling function, which adds to the expressive function the generation of a reaction in others. Popper compares it to the danger signals an animal might send out in order to alert other animals, and to the way a traffic light signals the possible presence of cars even when there are none about. The third is the descriptive function, which involves the expression of a proposition, something that can be either true or false. The paradigm here would be the utterance of a declarative sentence, such as “Roses are red,” “Two and two make four,” or “There is a predator in the area.” Notice that the latter example differs from an animal’s cry of warning in having a conceptual structure. A bird’s squawk might cause another bird to feel fear and take flight. What it does not do is convey an abstract concept like eagle, predator, or danger, and thus it does not convey the sort of propositional content that presupposes such concepts. The fourth is the argumentative function, which involves the expression of an inference from one or more propositions to another in a manner than can be said to be either valid or invalid, as when we reason from the premises that all men are mortal and Socrates is a man to the conclusion that Socrates is mortal. 

While the most primitive of animals may be incapable of behavior that performs even the first two of these functions, there are also obviously animals that can perform them, and in that sense they can be said to have language. But to be capable of performing the expressive and signaling functions of language does not entail being capable of the descriptive and argumentative functions. Those functions require linguistic representations that mirror what we have called the compositionality, productivity, and systematicity of thought, such as the words by which we express concepts and the sentences by which we express propositions. Whether animals lacking that particular sort of language, can be said to think in the sense of entertaining concepts, propositions, and logical inferences is the most interesting and important way of interpreting the question of whether nonlinguistic animals can think. As we have seen, there is good reason to conclude that, when the question is interpreted that way, the answer must be in the negative. 167 (Note that the question whether any non-human animals possess language of that sort, and thus can think in that sense, is a separate question. Again, I will address that issue in a later chapter.) 

These results are perfectly compatible with the irreducibility of thought to linguistic behavior. Certainly nothing in the critique of behaviorism summarized above entails that the capacity for concepts, thought, and reasoning could exist in the absence of language. I have noted already that even though concepts and propositions cannot be reduced to mental imagery, it could still be true that human beings require imagery as an aid to entertaining concepts and propositions. Similarly, even though thought cannot be reduced to the capacity for language, it could still be true that the capacity for language is a precondition for human thought. 

On the other hand, as I will argue in later chapters, it is certainly possible for there to be intellects (albeit ones very different in their operation from our own) entirely divorced from matter, and thus without language of the kind with which we are familiar (embodied in spoken utterances and written marks, and thus presupposing bodily organs and material media). For the moment we can simply note that we needn’t settle this particular issue for the purposes of this chapter, which requires showing only that the intellect is not reducible to the capacity for language any more than it is reducible to the capacity for mental imagery or the simulation of perceptual representations. 

Eliminativism 

Much more radical than the view that concepts, thoughts, etc. can be reduced to mental imagery, perceptual representations, or linguistic behavior is the thesis that there simply are no such things – that we should not bother trying to reduce them to something else, but rather ought partially or entirely to eliminate such notions from our account of human nature. 

The standard argument for this sort of view appeals to the premise that one or more of the notions in question cannot plausibly be accommodated by science – say, by neuroscience, 168 or by the best computational models of the mind developed by cognitive scientists, 169 or by experimental psychology. 170 The idea is that if concepts, propositional attitudes, intentionality, or some other aspect of the mind are not to be found in what Wilfrid Sellars famously called the “scientific image of man,”171 then they must not really be there in man himself. 

This kind of argument rests on a number of assumptions. First of all, it presupposes, at least implicitly, that science commits us to a “mechanistic” conception of the natural world which attributes to matter only those properties susceptible of mathematical description (size, spatial location, velocity, etc.), and which takes the world to be devoid of what Aristotelians call final causality or teleology (i. e. directedness toward an end). Second, it presupposes scientism, according to which we should not regard as real anything but what is captured in the picture of the world provided by science, interpreted in broadly mechanistic terms. Third, it assumes that one or more of the key mental notions we have been discussing (concepts, the propositional attitudes, intentionality, etc.) cannot be fitted into this scientific picture of the world (again, interpreted mechanistically).

This last assumption is controversial even among those committed to the first two assumptions. Many contemporary materialist philosophers hold that the mind can be explained in terms compatible with a mechanistic conception of nature. However, as I will argue in later chapters, they are wrong, and the eliminativist’s third assumption is correct. If you exclude teleology or directedness to an end from your picture of nature, there is no way you are going to fit notions such as intentionality (the directedness of thought toward an object) into it. 

But the specific purposes of the present chapter don’t presuppose this rejection of materialism, and in any event, the third assumption will support eliminativism only given the first two assumptions. Those assumptions, however, are essentially simply taken for granted by eliminativists rather than argued for rigorously. Moreover, the assumptions are false. The mechanistic conception of nature has not been proved true by science, and there is more to the world than can be captured by science interpreted in mechanistic terms. Hence, even if concepts, propositional attitudes, intentionality, etc. cannot be explain in terms of mechanistic science, that gives no support at all to eliminativism. 

Naturally, to defend these claims requires a treatment of deep issues in metaphysics and the philosophy of science that cannot be addressed thoroughly in a book on human nature, specifically. But I have defended these claims in depth elsewhere, 172 and we will have reason to address some of the relevant issues in later chapters. For the purposes of the present chapter, it will suffice to note some more specific problems with eliminativist arguments – and to show that eliminativism simply cannot, in any event, be true. 

To begin with a less radical form of eliminativism, consider Edouard Machery’s view that psychology ought to abandon the notion of a concept. 173 This suggestion is not quite as extreme as it might seem. Machery notes that psychologists have developed several theories of what a concept is. Some take a concept to be a prototype, in the sense of a representation of a set of properties that a standard example of a member of some class will tend to have. Some take a concept to be an exemplar, in the sense of a representation of some remembered instance of a category. Some take a concept to be a theory, in the sense of a body of knowledge about members of a class. These different approaches each have their strengths, and Machery allows that there really are prototypes, exemplars, and theories. But he suggests that psychological research does not give us reason to think that there really is some one kind of thing, concepts, of which the prototype view, the exemplar view, and the theory view are all alternative accounts. Rather, psychology should just affirm that there are prototypes, exemplars, and theories, and leave it at that, giving up the assumption that the notion of a concept is needed to tie these three kinds of thing together. 

Machery’s argument simply takes it for granted that what matters is what contemporary psychologists happen to need or find useful when developing their theories. But this assumption is problematic even apart from the usual general difficulties with scientism. As Machery himself emphasizes, contemporary psychologists and philosophers are often talking past one another when discussing concepts. The former are typically not even addressing the issues that concern the latter (where the sorts of issues that concern philosophers are the ones we’ve been addressing in this chapter). Why, then, conclude from the state of play in psychology that the notion of a concept should be dispensed with? Why not conclude instead that psychologists ought to address the issues that philosophers are concerned about, in which case they may find that the notion of a concept is indispensable? Or why not conclude that psychology and philosophy are complementary approaches to the subject, so that even if the former does not need the notion of a concept, the notion is still indispensable when dealing with the issues addressed by the latter? 

Nor is it plausible in the first place that psychology really can do without the notion of a concept. As Georges Rey points out, it seems impossible even to characterize the variety of phenomena that psychology must account for without deploying the notion. 174 Rey observes: 

On the face of it, concepts are the stuff of which psychological claims and explanations are made. Generalizations and explanations of, e.g., cognitive development, fallacies in reasoning, vision and language understanding (to take some of the more successful areas of recent psychology) – all these presuppose concepts as shared constituents of the propositional attitudes the explanations concern. It’s not clear how even to describe the phenomenon of the Müller-Lyer illusion unless we can presume that people share a concept of longer than, or the gambler’s fallacy, without them sharing [the concept of] more likely. Concepts seem to be natural kinds at least to the extent that they are the kinds of entity over which psychology generalizes.

Moreover, people’s prototypes, exemplars, and theories differ, yet they are still able to engage in fruitful debate with one another. That implies that there are concepts over and above these that they share and that provide a common reference point for agreement and disagreement. Indeed, Machery and his critics themselves could hardly disagree if there were not some shared conceptual content by reference to which they could understand what their dispute is about in the first place. In these ways, looking to psychology for grounds to deny the reality of concepts seems inevitably self-defeating. 

The self-defeating nature of eliminativism is even more obvious, indeed notorious, when its more extreme versions are considered. These would be views which claim that there are no such things as beliefs, desires, or other propositional attitudes, and in general nothing that possesses the intentionality or representational content definitive of concepts and thoughts – where this view is, again, put forward on the grounds that propositional attitudes and intentionality cannot be reduced to properties of the kind recognized by neuroscience or other natural sciences. 

That there is something fishy about this line of reasoning should be obvious from the fact that it is never pushed through consistently. As Stephen Stich (himself a former eliminativist) and Stephen Laurence emphasize, there are all sorts of notions for which we have no good reductionist analysis (they give examples like couch, car, war, famine, ownership, mating, and death) but which few would propose we eliminate from our picture of reality. 175 But the fact is that we simply could not possibly coherently eliminate from that picture everything which cannot be given a reductionist analysis. In particular, the very attempt to eliminate intentionality and related notions will always implicitly presuppose them. 

The simplistic or “pop” way of making this point is to say that eliminativists claim to believe that there are no beliefs, which is a performative self-contradiction. As eliminativists rightly point out, by itself this is not a very impressive objection, because the eliminativist can always avoid using locutions like “I believe that…” and make his point in other terms instead. However, the real question is whether the eliminativist can spell out his position in a way that entirely avoids terms that explicitly or implicitly presuppose intentionality. That cannot be done. Even if you get rid of intentionality in one area it will, like the proverbial whack-a-mole, always rear its head somewhere else. 176 

Hence, consider the eliminativist’s central claims – that intentionality is illusory, that descriptions of human beings as possessing intentionality are false, that it is a mistake to try to reduce rather than eliminate it, etc. All of these notions are as suffused with intentionality as any the eliminativist wants to overthrow. They presuppose the meaning of a thought or of a statement that has failed to represent things accurately, or a purpose that one has failed to achieve or that one should not have been aiming to achieve in the first place. Yet we are told by the eliminativist that there are no purposes, meanings, representations, aims, etc. of any sort whatsoever. So, how can there be illusions, falsehoods, and mistakes? 

For that matter, how can there be truth or correctness, including the truth and correctness the eliminativist would ascribe to science? For these concepts too presuppose the meaning of a thought or statement that has represented things accurately, or the realization of a purpose. Thus, “Water is composed of hydrogen and oxygen” is true, while “Water is composed of silicon” is false, and the reason has to do with the meanings we associate with these sentences. Had the sentences in question had different meanings, the truth values would not necessarily have been the same. By contrast, “Trghfhhe bgghajdfsa adsa” is neither true nor false, because it has no meaning at all. Yet if eliminativism is right, “Water is composed of hydrogen and oxygen” is as devoid of meaning as “Trghfhhe bgghajdfsa adsa” is – in which case it is also as devoid of a truth value as the latter is. Moreover, if eliminativism is right, every statement in the writings of elminativists themselves, and every statement in every book of science, is as devoid of meaning as “Trghfhhe bgghajdfsa adsa” is, and thus just as devoid of any truth value. But then, in what sense do either science or eliminative materialist philosophy give us the truth about things? 

Logic is also suffused with intentionality, insofar as inferences aim at truth and insofar as the logical relationships between statements presuppose that they have certain specific meanings. “Socrates is mortal” follows from “All men are mortal” and “Socrates is a man” only because of the meanings we associate with these sets of symbols. If we associated different meanings with them, the one would not necessarily follow from the others. If each was as meaningless as “Trghfhhe bgghajdfsa adsa” is, then there would be no logical relationships between them at all – no such thing as the one set of symbols being entailed by, or rationally justified by the others. But again, if eliminativism is right, then every sentence, including every sentence in every work of eliminativist philosophy and every sentence in every book of science, is as meaningless as “Trghfhhe bgghajdfsa adsa” is. In that case there are no logical relations between any of the sentences in any of these writings, and thus no valid arguments (or indeed any arguments at all) to be found in them. So in what sense do either science or the assertions made by eliminativist philosophers constitute rational defenses of the claims they put forward? 

Notions like “theory,” “evidence,” “observation,” and the like are as suffused with intentionality as the notions of truth and logic are. Hence if there is no such thing as intentionality, then there is also no such thing as a scientific theory, as evidence for a scientific theory, as an observation which might confirm or disconfirm a theory, etc. Eliminativism makes of all statements and all arguments – scientific statements and arguments no less than metaphysical ones, and indeed every assertion of or argument for eliminativism itself – a meaningless string of ink marks or noises, no more true or false, rational or irrational than “Trghfhhe bgghajdfsa adsa” is. As M.R. Bennett and P.M.S. Hacker put it, the eliminativist “saws off the branch on which he is seated,” undermining the very possibility of science in the name of science. 177 

The eliminativist owes us an explanation, then, of how he can so much as state his position in a coherent way in the absence of all these notions his position requires him to jettison. The stock eliminativist move at this point is to claim that future neuroscience will provide new categories to replace these old ones, at which point the view can be stated in a more consistent way. But that is like someone asserting that 2 +2 = 23 and then, when asked what exactly this claim can mean given what it is to add, what it is for numbers to be equal, etc., responding that he can’t really say but that future mathematicians will come up with a way of making sense of it. Until we have such an explanation, we don’t even know so much as what the claim is that we are being asked to consider, much less whether it is correct. The same can be said for eliminativism. Until we are given a coherent way of formulating the thesis, we don’t really have anything that amounts to a thesis, much less one we have reason to take seriously.

Hilary Putnam reports that the eliminativist Paul Churchland once acknowledged in conversation that he needs a “successor concept” to the notion of truth, and that he doesn’t know what it will be. 178 This is doubly problematic. For one thing, if an eliminativist like Churchland is admitting both that he cannot claim that eliminativism is true (since a consistent eliminativist has to regard the notion of truth as illusory as that of intentionality) and that he has nothing to put in place of truth, then it is not clear exactly what he is trying to say about eliminativism or to convince us to say about it. (It cannot be “Eliminativism is true,” but at best something like “Eliminativism is _____,” with no explanation of how to fill in the blank. Until we have such an explanation, what is it that are we supposed to do with this utterance?) For another thing, the notion of a concept is as suffused with intentionality as the notions of truth, meaning, etc. are. So, if Churchland is consistent, he cannot say that he needs a “successor concept” for the notion of truth, but rather a “successor _____,” where we now have a second blank to fill in with a term that does not entail the existence of intentionality, and where we are once again at a loss as to what that term might be. 

The upshot of our discussion in this chapter, then, is that the intellect, like the self whose intellect it is, is real and irreducible. Whether it is material or immaterial in nature is a question to which we will return. But if it should turn out that no materialist account of our intellectual powers can succeed, it will follow that the problem is in that case with materialism and not with the commonsense supposition that we really do have concepts, thoughts, and rationality.

4. The Will 

What is the will?

 As we saw in chapter 1, according to Boethius’s classic definition, a person is an individual substance of a rational nature. The subsequent two chapters defended the commonsense view that we are indeed persons in this sense. I argued in chapter 2 that it cannot coherently be denied that the self is a substance that persists over time. I argued in chapter 3 that it cannot coherently be denied that this substance has the rational or intellectual powers of grasping concepts, putting them together into propositions, and reasoning logically from one proposition to another. I also claimed in chapter 1 that free will is no less real and irreducible than the self and the intellect. I defend that claim in this chapter. But before we can understand why and in what sense the will is free, we need first to understand what the will is. 

The short answer is that the will is the power or capacity of a rational substance to pursue what the intellect judges to be good. (Not necessarily morally good, but at least good in the broad sense of being suitable or desirable in some way.) It is also traditionally known as the “rational appetite.” It is an appetite insofar as it involves an inclination or tendency toward some object or end. It is a rational appetite insofar as the object is pursued as something conceptualized in a certain way, and thus as something about which we might reason and come to know true propositions. This distinguishes the will from the appetites that non-human animals possess, which involve inclinations toward non-conceptualized objects of sensory perception; and from the sorts of inclinations or tendencies that a plant or an inanimate substance might exhibit, which are not even conscious, let alone conceptualized. A dog can see water and desire what it sees, but it cannot desire it as water, because it does not have the concept of water or any other concept. A tree can be inclined to sink its roots in the direction of the water below ground, but it cannot see or otherwise perceive the water or consciously desire it, much less conceptualize it. Water itself has an inclination or tendency to move toward the lowest point it can reach, but when it does so it is not exercising some organic capacity (as a tree is when it sinks roots into the ground), much less perceiving, desiring, or conceptualizing anything. A human being, by contrast, can not only perceive and desire water, but can know that it is water, and desire it precisely as something conceptualized in that specific way. To be able to do that sort of thing is just what it is to have a will. This is, in any event, the account of the will famously associated with Aquinas, and the aim of this chapter is to establish the reality and freedom of the will in this sense of “will.”179 

Again, that is the short answer to the question about what the will is. A longer answer will have to address various further questions raised by the short answer, such as: What is a substance? What is a power? What is it for a power to be inclined or tend toward some object or end? Now, these are issues of general metaphysical import, a thorough treatment of which is beyond the scope of a book devoted to human nature, specifically. But for present purposes, a brief overview will suffice (and in any event, I have provided a more thorough treatment of these topics elsewhere). 180 

In fact we have already discussed the notion of a substance in chapter 2, but let’s say a little more. One way to understand what a substance is is to contrast it with what it is not. A substance is not an attribute, nor an aggregate, nor an artifact. For example, redness is an attribute rather than a substance, insofar as it exists only in other things – in an apple, say, or in a Stop sign – whereas a substance does not exist in another thing in the same sense. It has a kind of independence that an attribute lacks. A pile of stones is an aggregate rather than a substance, insofar as it is really nothing over and above the sum of the stones that make it up, each of which would be exactly as it is whether or not it was part of the pile. A substance has a kind of unity that an aggregate lacks, insofar as it is more than the sum of its parts, each of which is what it is only relative to the whole substance of which it is a part. A watch is an artifact rather than a substance, insofar as its characteristic end or purpose of telling time is observer-relative, something imposed on it from outside by the designers and users of the watch. A true substance is directed toward the ends that are characteristic of it in an intrinsic way, of its very nature rather than merely by virtue of the purposes of some artificer. 

The technical Aristotelian-Thomistic way of capturing the difference between a true physical substance and a mere aggregate or an artifact is to say that a physical substance has qua substance a substantial form, whereas an aggregate or an artifact has qua aggregate or artifact a merely accidental form. But let’s pause to say something about the notion of form in general before explaining the distinction between substantial and accidental form in particular. The form of a thing is, to a first approximation, its essence or nature, as contrasted with the matter that takes on that form, essence, or nature. For example, there is the form of being a sphere, which is distinct from the different kinds of matter (plastic, rubber, stone, wood, etc.) that might take on that form. In Aristotelian-Thomistic philosophy, to identity the form and matter of a thing is to identify two of its four causes, namely its formal cause and its material cause. The remaining two are its efficient cause, which is whatever brought it into existence, and its final cause, which is the end or set of ends toward which it tends. In the case of a ball, for example, we might say that its formal cause is its spherical shape, its material cause is the rubber it is made out of, its efficient cause is the manufacturing process in some factory that brought it about, and its final cause is to roll down incline surfaces, bounce when hitting a solid object, and so on. 

Now, a substantial form is a form that makes something a substance of a certain kind (such as a stone or a dog), and thereby gives it its fundamental character. An accidental form (such as the shape of some stone or the color of a certain dog) is a secondary feature that merely modifies a substance that would still be a substance of the same kind even without the feature. A mark of a thing’s having a substantial form (and thus of its being a true substance rather than a mere modification of a substance) is the presence in it of irreducible properties and causal powers. Any example is bound to be controversial, but suppose for the sake of argument that water is a true substance, and thus has qua water a substantial rather than merely accidental form. The idea would be that the properties and causal powers of water (being liquid at room temperature, acting as a solvent, and so on) are irreducible to the properties and causal powers of hydrogen, oxygen, or any mere aggregate of hydrogen and oxygen. By contrast, the properties and causal powers of a random pile of ice cubes made out of frozen water are reducible to the aggregate of the properties and powers of the cubes that make up the pile, and the properties and causal powers of an igloo constructed out of blocks of ice are reducible to the aggregate of the properties and powers of the blocks. Hence to be an aggregate like a pile of ice cubes or an artifact like an igloo is to have a merely accidental rather than substantial form. As these examples indicate, things with merely accidental forms presuppose the existence of things having substantial forms. You can’t have random piles of ice cubes and igloos unless you first have frozen water. Hence substances are metaphysically more fundamental than aggregates and artifacts. 

I’ve brought up powers again, so let’s turn to those. A power is a capacity to act or operate in a certain way, such as acting on another thing so as to bring about a change in it. For example, frozen water or ice has the power to cool down the surrounding air. Now, a power is a kind of attribute, and like other attributes it exists only in a substance rather than in a freestanding way. The power to cool down the surrounding air is a power of the ice. Strictly speaking, it is not the power to cool that cools down the air; it is the ice that cools down the air, by virtue of having the power to do so. 

The French playwright Molière (1622-73) famously mocked this sort of talk as tautological or uninformative, but it is not. Consider his example of opium, to which we can attribute a dormitive power, or power to cause sleep. To attribute such a power to opium is not a tautology, as it would be a tautology to say that opium causes sleep because it causes sleep. When we attribute such a power to opium, we are saying that there is something about the opium itself that produces sleep, that it is not something merely in the circumstances under which it is ingested that does so (even if the circumstances play some role). We are also saying that it is not merely something in a particular sample of opium that causes sleep, but something in opium as such that does so. Of course, this is not the most informative claim in the world. We would also like to know exactly what it is about opium that gives it this effect, and a chemist might tell us that. But the claim really does give us at least some information about opium. 

Now, as I have indicated, the powers of a physical substance are grounded in its substantial form, and different kinds of substance will have different powers, as water and opium obviously do. Powers themselves differ insofar as they aim at or are directed toward different outcomes, as being cooled and being made to sleep are different outcomes. It is through its powers that a substance manifests its final causality, also known as its teleology (as we saw in chapter 3). Indeed, a causal power is a kind of link or middle man between a substance’s substantial form and its teleology. As Aquinas says, “some inclination follows every form,”181 and “every agent acts for an end.”182 That is to say, an agent (a substance considered as an efficient cause) will, by virtue of its distinctive substantial form, be inclined toward, and act to bring about, some particular kind of end or set of ends. But a substance’s substantial form aims it toward those ends through its causal powers, as it were. 

Early modern philosophers eschewed the notions of formal and final causality, and, especially after David Hume, they tended to throw out the notion of causal powers as well. For post-Humean philosophy of causation, causes and effects are conceived of as inherently “loose and separate” in the sense that, in principle, any effect or none might follow upon any cause. This is precisely what we should expect if no cause has a substantial form or teleology by virtue of which it naturally aims at the production of some particular effect or range of effects. Hence, if a certain kind of effect B does in fact happen always to follow from a certain cause A, that must, for the post-Humean philosopher, be attributed to some “law of nature” extrinsic to A and B, rather than to some power intrinsic to A. It is something outside of A (the law in question) rather than something internal to it (a causal power) that accounts for what it does. This conceptual shift could hardly fail to have repercussions for the modern understanding of how the will is related to the agent whose will it is and to the actions that follow upon its operation (an understanding very different from that of Aquinas). 

Before we get to that, one last general point must be made about causal powers, which is that whether and how they operate depends crucially on circumstances. We must distinguish between a power and its manifestation on any particular occasion. A sample of opium has the power to cause sleep even if it is never ingested and thus never in fact causes sleep. Even if it is ingested, it may not cause sleep if this outcome is blocked by some other causal factor – say, by your having taken a handful of amphetamines before taking the opium. The manifestation of a causal power may depend on its operating in tandem with other causal powers, and also on the absence of the operation of yet other powers. 

As contemporary theorists of causal powers often point out, such facts have dramatic implications for our understanding of laws of nature. It is often assumed that laws of nature are fundamental to physical reality, and that at least the paradigmatic examples of laws of nature operate in a strict or exceptionless way. But in fact, physical phenomena will exhibit the regularities described by a law of nature only given that certain causal powers are acting in tandem or failing to do so, and in some cases this condition never actually holds. For example, Newton’s first law tells us that an object in motion, if it is not acted upon, has a tendency to remain in motion forever. But objects never actually do remain in motion forever, because all of them always are in fact acted upon by other objects. Hence the regularity described by the law never really holds. As this indicates, causal powers are actually more fundamental to physical reality than laws of nature are. A law is really just a description of how things will go if substances manifest their causal powers in such-and-such a way. As philosopher of science Nancy Cartwright famously put it, the laws of physics “lie.”183 They do not describe physical reality as it really is, but only some idealization of physical reality. 184 

As we will see, this has implications for the traditional debate over determinism and free will. Determinism holds that what happens in the physical world at any given moment is necessitated by what was happening at earlier moments. This presupposes a picture of the world on which the laws of nature are metaphysically fundamental, and causal processes are secondary, operating in a way that conforms strictly to the laws. But for the Aristotelian-Thomistic picture of the world that I am describing, this has things backwards. It is substances and their powers that are fundamental, and laws are secondary, describing the patterns that will result if certain combinations of powers manifest or fail to manifest. Hence determinism simply gets nature fundamentally wrong. Note that it is not only human beings that it gets wrong, but also non-human animals, plants, and inanimate phenomena. Since much of the modern debate about free will presupposes this deterministic picture of nature, that debate simply gets off on the wrong foot. 185 

We’ll come back to that. For the moment, note that while the behavior of non-human animals is not deterministic, neither is it random or arbitrary. It is perfectly intelligible given their causal powers together with the powers of the things they interact with. Yet non-human animals do not have wills, much less free will. So, being non-determined but also non-random is not sufficient for having a will, free or otherwise. What more is needed? Aquinas’s answer is that the will is a power of the kind of thing whose activity arises from within it in a certain way – specifically, by way of having knowledge of the kind of which a rational substance alone is capable. 186 To act voluntarily is to know what you are doing in the way that human beings do. 

But it will be useful, before we say more about what this involves, to say something about intermediate cases – intermediate, that is to say, between inanimate things like opium and water on the one hand and human beings on the other. A living thing is distinguished from non-living things in being selfperfecting. 187 That is to say, in addition to having powers by which it affects other things, it also has powers to complete and maintain itself, as a plant does when it grows and takes in water and nutrients. In this thin sense, any living thing, including even a plant, can be said to be a kind of self-mover, and thus to act by virtue of something within it. (The technical way of putting this is that it exhibits immanent causality as well as transeunt causality.) Still, a plant, like water or opium, does not know what it is doing, and thus cannot be said to act voluntarily or from will. 

Now, though the behavior of non-human animals is not quite voluntary either, animals are closer than plants are to possessing the crucial characteristic that makes for will, in a way that is instructive. Suppose I yank a plant from the ground, or tie one of its branches back so that the leaves are kept permanently in the shade. Of course, no one would regard these movements of the plant as voluntary. The plant’s natural tendencies would be to sink its roots into the ground, and to grow toward the light. Being contrary to its natural tendencies, the movements I’ve caused in it are violent, and violent movements are ipso facto involuntary. But even when the plant does do what is natural to it, we wouldn’t say that it acts voluntarily. It sinks its roots into the ground and grows toward the light, but again, we would say that even then it doesn’t know what it is doing. That is why we wouldn’t blame a plant for failing to sink its roots into the ground or grow toward the light. We would judge such a failure to be the result of the plant’s being afflicted by some sort of damage or disease, rather than anything analogous to choice. We wouldn’t say that the plant should know better, because it doesn’t know anything at all, and you have to know something before you can be held responsible.

 Again, for Aquinas, voluntary behavior is behavior that arises from within the agent in a certain specific way. The plant’s roots being yanked from the ground is a movement that does not arise from within the plant in any sense, since, in my example, I am the one who makes them move that way. The plant’s roots sinking into the ground is a movement that does arise from within the plant in a sense, since it results from the plant’s natural selfperfective powers qua plant. However, though the power and inclination are in the plant, the end toward which the power is inclined is in no sense present within the plant, nor even represented within it. An end’s being represented within an agent in some way is, Aquinas holds, a necessary condition for the agent’s behavior arising from within it in the relevant sense. 

Now, in a non-human animal, the ends toward which its powers are inclined are represented within it. For an animal can perceive a thing toward which it is naturally inclined, such as the food it eats. Even before it actually eats it, the food is in the animal qua perceived. Of course, an animal can misperceive things, as when a greyhound in a race chases a fake rabbit. But that reinforces the point that it is something in a sense internal to the animal, namely the end as perceived (or misperceived), that moves it to act. A plant, by contrast, doesn’t even rise to the level of misperception, let alone perception. 

So, unlike a plant, an animal has a kind of knowledge of the end it is pursuing. However, we still wouldn’t say that an animal knows what it is doing, and thus wouldn’t hold it morally responsible for what it does, any more than we would hold a plant morally responsible. For while the animal knows the end that it pursues, it doesn’t know that it knows it. It doesn’t know the end as an end, for it does not have the concept of an end, nor indeed any concepts at all. Hence it also lacks the concept of being a means to an end, and thus doesn’t know that its action is a means to the end it pursues, much less that other means might be possible. This lack of concepts entails a certain inflexibility in the animal’s behavior. Like a plant, an animal cannot do otherwise than whatever it in fact does. If the animal is hungry and there are no countervailing circumstances present (such as a predator’s being in the vicinity of the food, and frightening the animal away), the animal’s perception of the food will result in it trying to eat it. Because of this lack of concepts and consequent inflexibility in behavior, Aquinas says that a nonhuman animal acts voluntarily only in an imperfect sense. 

Now, a rational substance is a substance with an intellect, and to have an intellect is precisely to have the capacity to conceptualize what one knows. A human being qua rational animal not only knows the food that he tries to eat, but knows it as food of a certain kind (an apple, say) and as something that would be good for him to eat given his need for nutrition. He also knows that biting into it straightaway would be one possible means of eating it, and that this has certain advantages and disadvantages compared to other means such as baking it into a pie, turning it into applesauce, or squeezing the juice out of it and drinking it. What all of this makes possible, as Aquinas emphasizes, is deliberation. Since a rational being possess universal concepts, he can identify particular ends and means as instances of general kinds. 188 He can make comparisons between these different possible ends and different possible means to the ends. 189 He can arrive at judgments about which ends and means are best, and also make judgments about those judgments themselves. 190 Hence, when he decides to eat the apple, he knows what he is doing in a way that opium, or a plant, or a non-human animal does not know what it is doing. 

We have, then, a hierarchy of degrees to which the source of a thing’s behavior is to be found within it. Inanimate things like frozen water and opium have causal powers which are directed toward certain ends (for example, cooling down the surrounding air and causing sleep, respectively), but these ends are entirely external to the inanimate things themselves. Merely vegetative forms of life have powers directed toward ends internal to them (such as growth), but without any sort of knowledge of these ends. Given their inherent directedness toward an end, the powers of these animate and inanimate non-sentient substances can be thought of as appetites of a sort, but as merely natural appetites (to use the Aristotelian-Thomistic jargon) insofar as they reflect a thing’s natural tendencies rather than any sort of knowledge. Since a non-human animal perceptually represents the ends toward which it is directed, it does have a kind of knowledge (albeit an imperfect kind). Its powers are classified as sensory appetites. Human beings as rational animals have a more perfect knowledge of the ends toward which they are directed, insofar as they can conceptualize them, make judgments about them, and draw inferences about them. They know these ends as ends, deliberate about whether they are worthy of pursuit and about the best means to pursue them, and decide how to act on the basis of this deliberation. The activity of rational beings thus arises from within them in the most perfect possible way. They possess a kind of causal power which we can classify as rational appetite. 

To have a will is, for Aquinas, precisely to have rational appetite, to have in the most perfect way possible the source of one’s activity within one. To be a rational substance entails having a will, because every kind of substance has its own distinctive causal powers, and a will just is the kind of appetite or power characteristic of a rational substance. Because rational substances have, in this way, the source of their activity within them, they possess a kind of self-determination that nothing else has. 191 What we rational substances do is up to us, whereas what non-rational things do is not up to them. This is, as we will see, crucial to understanding the nature of the will’s freedom. 

The thesis that the will is to be understood as a power of a rational substance is by no means merely some idiosyncratic or arbitrary claim of Aquinas’s. It is implicit in the very concept of choice and in the phenomenology of our experience of choosing. As John Searle notes, we cannot make sense of choice except as the activity of an irreducible self with reason and agency that persists over time. 192 For choice is the outcome of deliberation, which is a rational process extended over time, and it presupposes a continuing subject who persists from the beginning of that process through to the end of it. Choice is also active, a bringing of something about rather than passively experiencing it as occurring. The subject of this activity cannot be merely a bundle of events or properties of the kind posited by Hume, because there must be something that ties together the various cognitive acts that enter into the deliberative process, and makes of them a single chain of reasoning that results in choice. Of course, the skeptic might insist that all of this is illusory. But the point for the moment is that the notion of will is necessarily tied to the notions of a persisting self with rationality, so that we cannot coherently affirm the reality of the will without affirming the reality of a rational self. I have, in any event, already argued in chapters 2 and 3 for the reality of such a persisting self and for its rationality. We will return presently to the question of why we should affirm that it does indeed possess will. 193 

Aquinas holds that the intellect is prior to the will, a thesis sometimes called intellectualism (where voluntarism is the rival view that the will is prior to the intellect). What has been said should make it clear what this means and why he holds it. For the will is a power of a substance, and a power is ontologically less fundamental than the substance of which it is a power. And the kind of substance of which it is a power is a rational or intellectual substance. Hence, the will is posterior to the intellect in just the way that the power to cause sleep is posterior to opium, or the power to sink roots into the ground is posterior to a plant. To say that the will is prior to the intellect would be like saying that a plant’s power to sink roots is prior to its vegetative nature. It would get things the wrong way around, putting the volitional cart before the intellective horse. Moreover, a thing cannot be willed unless it is known, and it is the intellect that knows. An appetite is an appetite for some object, and if the object is not known intellectually, then the appetite would be a natural appetite or a sensory appetite rather than a rational appetite, and thus would not be a will. By contrast, if, per impossibile, an intellectual substance lacked a will and therefore were not inclined toward or away from what it knows, it would still know what it knows. 

To be sure, there is a sense in which the will can move the intellect, as even Aquinas allows. But when properly understood, this does not in any way conflict with what has been said so far. For the will cannot move the intellect, any more than it can do anything else, unless the intellect first judges such movement to be good and worthy of pursuit. For example, suppose we say that the will moves the intellect to contemplate some topic. That is precisely because the intellect judges such contemplation to be worth pursuing, and the will simply follows its lead. When we speak of the will doing this or that, it is easy to fall into the trap of thinking of the will as if it had a mind of its own, which somehow competes with the intellect. In fact the only mind the will has is the intellect itself. When the will moves the intellect, it is really the intellect moving the intellect by means of the will. Or, better still, it is the rational substance moving itself, by means of the same power by which it moves other things, namely the will. To think that the will somehow moves the intellect on its own is like thinking that, when I pull a rug out from under myself and fall down, it is the rug that has knocked me down. In fact I knocked myself down by means of pulling the rug, and in the same way, I move my intellect myself when I will either to contemplate or not contemplate some topic. 

Here we must avoid a fallacy of hypostasization, of thinking of the will as if it were a substance in its own right as opposed to being a power of a substance. To speak of the will doing this or that is like speaking of the power of opium causing sleep. Properly understood, it is really just a shorthand for speaking of what a substance is doing. It is the opium that causes sleep by virtue of its dormitive power, and it is a rational substance as a whole to which the will’s activity is properly attributed. It is not really my will that does anything. I do things, by virtue of having a will. Reductionist theories of the self can facilitate the fallacious tendency to hypostasize the will. If we think of the person, not as a unity but rather as an aggregate of parts, then the intellect, the will, our various bodily parts, and the other aspects of human nature come to seem like quasi-substances in their own right, interacting like billiard balls knocking into one another. We end up with a model of action on which the intellect directing the will to pick up an apple is conceptualized on analogy with the cue ball knocking the eight ball into a third billiard ball. One then can start wondering whether it might not really be the eight ball that is knocking into the cue ball, and be led thereby into voluntarism. But this model is a travesty. Strictly speaking, there is just one agent here, the human being himself, who picks up the apple by virtue of knowing it and willing to acquire what it knows. 

We must also avoid confusing the will with the emotions that are often associated with its exercise. 194 When you will to take a bite of an apple, this may be associated with a strong desire for the apple, and when you will to run away from a snake, this may be associated with fear. But the act of will is distinct from the desire and the fear. The will is related to the intellect in a way that is analogous to the relationship between emotions and the sensations and mental imagery that can trigger them. The sight of a snake can generate fear, and forming a mental image of an apple can generate desire. In the same way, when the intellect conceptualizes something as an apple or a snake, the will might be drawn toward it in the first case or repelled by it in the second. But just as it would be a mistake to identify the concept of an apple or a snake with the mental images associated with these things, so too would it be a mistake to identify the will with the desire or the fear. Desire and fear, like sensation and imagination, can exist in creatures without intellects. Will cannot. 

Volitions and teleology 

I have been elaborating upon what it means to hold, as Aquinas does, that the will is the rational appetite, the power or capacity of a rational substance to pursue what the intellect judges to be good. But is will, in this sense, real? I have already argued, in chapter 2, that we are indeed persisting substances, and, in chapter 3, that we do indeed have intellect or rationality. 195 Does that not suffice to establish the reality of the will too, given that it is essentially just what a rational substance exercises when it acts? For example, isn’t it obvious that you have just now willed either to read on or to put this book down? 

Some philosophers think not. What I have been saying implies that the will is a capacity over and above our capacities to believe something, to desire that it be the case, or to take any of the other propositional attitudes referred to in chapter 3. I have also said that the will has a teleological character, that its objects serve as the final causes of the actions that result from willing. An exercise of the power of will – an act of willing something – is commonly referred to in the philosophical literature as a volition. Some philosophers deny that there are any such things as volitions, and hold that action can be entirely explained instead in terms of propositional attitudes like belief and desire. Some philosophers also deny that action is inherently teleological, and hold that it can instead be analyzed entirely in terms of efficient causation (unsurprisingly given that, as I noted above, modern philosophy has tended to eschew the notion of teleology or final cause). In order to see that the will as I have characterized it is real, then, we need to see why these philosophers are mistaken. 

One objection raised against the notion of a volition is that in order for a volition to be the cause of some piece of behavior, it would have to be something we could characterize independently of that behavior, since causes and their effects are, in general, distinct things. For example, heat from the sun might cause a certain ice cube to melt, but we can describe the sun completely independently of its relation to this particular effect. Now, in the case of volitions, it seems we cannot characterize them apart from the behavior of which they are said to be the causes. For example, I cannot describe the specific volition that is my willing to raise my arm without making reference to my behavior of raising it. Thus, the objection holds, there is a tight conceptual connection between volitions and the behaviors they purportedly cause of a kind that does not hold in the case of things genuinely related as cause and effect. 196 

The trouble with this objection is that while we can concede that a cause and its effect must be distinct entities, it doesn’t follow that we must in every case be able to characterize the former independently of the latter. For while this is possible in some cases (as in the example of the sun and the melting of the ice cube), there may be special features of other cases that make it impossible. In the case of volitions, they have (as beliefs and desires do) a kind of intentionality or directedness toward an object, and specifying a particular volition’s intentional content is, naturally, going to part of how we identify it. Thus do we characterize the volition we have taken as our example as willing to raise my arm, specifically. But it doesn’t follow that the volition is not a distinct entity from the behavior of raising my arm (just as a belief or desire is a distinct entity from what it represents). I might, after all, will to raise my arm without my arm actually being raised, if damage to the relevant nerves prevents this (just as I might believe that it is raining or desire that there be rain even if it is not in fact raining). In that case, the volition is clearly a distinct entity from the behavior toward which it points, and which in the normal case it causes. 197 

An especially influential set of objections to the notion of a volition was raised by Gilbert Ryle (1900-1976). 198 He notes that volitions are said to explain what makes a piece of behavior voluntary. For example, if my arm goes up as a result of a muscular spasm, we would say that such a motion was involuntary. But when I deliberately raise my arm, the claim is that this counts as a voluntary action because it was caused by a volition. Now, Ryle raises a dilemma for this account. Are volitions themselves voluntary? Either way we answer, the notion of a volition seems problematic. For if we say that they are not voluntary, then how can they explain what makes a piece of behavior voluntary? (After all, the neural processes that cause a muscular spasm are not voluntary, and they hardly make the spasm itself voluntary.) But if we say that volitions are voluntary, then it seems we are faced with a vicious regress. For if what makes a piece of behavior voluntary is that it was caused by a volition, wouldn’t it follow that what makes that volition itself voluntary is a further volition? Wouldn’t that further volition itself be made voluntary by yet another volition, and so on ad infinitum? 

But the dilemma is bogus, for in fact there is no problem for the notion of a volition either way we answer. For suppose we do say that volitions are themselves voluntary. It doesn’t follow that they are, or need to be, voluntary in the same sense in which the behavior they explain is voluntary. We can say that the sidewalk is wet, and that water is wet, but they are obviously not wet in the same way. The sidewalk is wet because it is covered with water, but water is wet, not because it is covered with water, but rather because it is that by virtue of which other things are wet. Similarly, both the raising of my arm and my act of willing to raise my arm can be said to be voluntary, but in different senses. The raising of my arm is voluntary because it was caused by the volition or act of will, and the volition is voluntary because it is that in virtue of which the behavior it causes is voluntary. 199 

Suppose instead that we say that a volition is not itself voluntary. It doesn’t follow that it would not be capable of explaining how the behavior it causes is voluntary. That would follow if the sense in which a volition is involuntary is the same as the sense in which a muscle spasm or a reflex is involuntary, but volitions are not involuntary in that sense. Volitions do, after all, have features that spasms and reflexes don’t have, such as being conceptually related to the cognitive processes that give rise to them, and this gives volitions and the behavior they cause a rationality that spasms and reflexes lack. 200 It is precisely this rationality that accounts for the difference between voluntary action and mere spasms and reflexes. Even if volitions are not themselves voluntary, they can still intelligibly be said to be that by virtue of which behavior is voluntary, and thereby play an explanatory role. By way of analogy, consider that the moon is illuminated, and that because it is it can illuminate other things, such as a tree you see outside your window when the moon is full. On a moonless night, let us imagine, the tree is not illuminated. Now, the sun is not illuminated either. But obviously, the sense in which the sun is unilluminated is not the same as the sense in which the tree is unilluminated on a moonless night, and neither is the sun unilluminated in a sense that would prevent it from illuminating other things. Rather, the sun is that by which things like the moon are illuminated. It is not less than illuminated, but rather more than being merely illuminated. Volitions are like that. If they are not themselves voluntary, that is not because they are less than voluntary, but rather because they are more than merely voluntary, being that by virtue of which other things count as voluntary. 

Ryle also objects that the notion of a volition is the product of philosophical theorizing rather than something derived from experience. The average person is familiar with thoughts, feelings of boredom, headaches, and the like, but not volitions. But as E. J. Lowe points out, we are in fact sometimes aware of volitions, especially when they are inefficacious (as when injury prevents us from doing what we will to do). 201 Moreover, the reason we are usually not aware of them is not because they are rare, but precisely because they are so common. As Lowe says, they are like one’s nose, which is rarely noticed in one’s field of vision precisely because it is always there. 202 

Ryle claims that it is often difficult to say exactly how many volitions are involved in a particular case, such as when reciting a poem backwards. One problem with this objection, as Lowe notes, is that it would prove too much. It is no less difficult to say how many utterances are involved in reciting a poem backwards, but Ryle and other critics of the notion of volition wouldn’t deny that utterances are real. 203 Another problem noted by Lowe is that Ryle’s question is not as unanswerable as he supposes. How many volitions are involved in a case like the one he mentions may be clearer once we further specify the circumstances. Reciting the poem backwards would presumably involve fewer volitions for someone wellrehearsed than it would for someone who is attempting this for the first time, and who may thus have to stop several times to try to remember the next line, and then will to continue the recitation.

Critics of the notion of volition allege, in any event, that it is simply not needed in order to account for action. It is often claimed, for example, that we can explain an action instead in terms of what the agent believes together with some “pro-attitude” he takes, such us desiring or wishing. Hence, we can explain why Bob went to the liquor store by noting that he desired some gin, believed that he could get it there, and that this conjunction of belief and desire caused him to act. 204 But the trouble with such an account, as Lowe observes, is that belief and desire are not by themselves sufficient to generate action. 205 Obviously, I can believe that gin is available at the liquor store without that prompting me to go there. But even if I desire some gin, that won’t necessarily generate the action either. I may, after all, judge that the desire would be a bad one to indulge (if, say, I am trying to overcome a drinking habit). Nor would it help to appeal to a mental event like forming the intention to go to the liquor store, since, like a desire, an intention need not result in action either immediately or at all. (Even if I sincerely intend to go to the liquor store, I may do so only long after forming the intention, or may forget about it and not do it at all.) What we need to posit is something with an inherent tendency immediately to generate action (which desires and intentions lack) and that is precisely what a volition or act of will is. 206 This is not to deny that beliefs, desires, and intentions play a role in generating volitions, which then go on to trigger behavior. But we shouldn’t conclude from this that the volition is somehow unnecessary, and that belief, desire, and intention can trigger behavior without this middle man. As Lowe notes, this would be like saying that since a spark might cause an explosion which in turn causes a building to collapse, we might as well suppose that a spark could directly cause the collapse without the middle step of the explosion. 207 

As Searle notes, in rational action there is always a “gap” between the having of a certain belief and desire on the one hand, and the carrying out of the action on the other. 208 I may desire some gin and believe that I can get it at the liquor store, but the action of going to the liquor store will count as rational precisely only if the having of that desire and belief does not causally necessitate it – that is to say, only if I might nevertheless refrain from the action despite having that desire and belief. For if the mere having of the desire and belief does suffice to guarantee that I will carry out the action, then we would have a case in which I am acting out of addiction, obsession, or the like, and thus in a paradigmatically irrational way. Of course, sometimes we do act in such ways. But not in the normal case. As Searle concludes, the thesis that actions are caused by beliefs and “proattitudes” alone is true not of actions in general and certainly not of rational actions, but only of irrational and non-rational actions.

The thesis that action is not irreducibly teleological is also commonly spelled out in terms of the idea that behavior can be explained in terms of beliefs and “pro-attitudes.” For example, that Bob went to the liquor store can, on this view, be explained in terms of his desire for gin and his belief that gin is available at the liquor store acting in tandem as efficient causes of his behavior. 209 No reference to final cause is needed. Even if Lowe’s point is conceded that reference need also be made to a volition or act of will that comes between the belief and desire on the one hand and the behavior on the other, the opponent of teleology might insist on interpreting this volition as simply one further efficient cause in the chain. 

But there is a well-known problem with this sort of analysis, having to do with cases involving “deviant causal chains.” To take a different example, consider an explanation like: Bob knocked over the glass of water for the purpose of distracting Fred. The phrase “for the purpose of” is teleological. But the opponent of teleology would suggest that the explanation can be reformulated as: Bob had the intention of distracting Fred and this caused him to knock over the glass of water, where this new description eliminates the teleological element. Bob’s intention is characterized as the efficient cause of his action, and this, it might seem, suffices to explain his behavior. But on closer inspection, we can see that this reformulation won’t work. For consider the case where Bob’s intention to knock over the glass makes him so nervous that his hand shakes uncontrollably, and knocks over the glass before he otherwise would have. 210 Then it is certainly true that Bob had the intention of distracting Fred and this caused him to knock over the glass of water, but it is not true that Bob knocked over the glass of water for the purpose of distracting Fred. For in this case, Bob knocked over the glass not for the purpose of distracting Fred (even though he did want to do that at some point), but rather because he lost control of his hand. So, the two descriptions are not equivalent after all. That is to say, there is content to the statement that Bob knocked over the glass of water for the purpose of distracting Fred that is not captured in the reformulation Bob had the intention of distracting Fred and this caused him to knock over the glass of water. The teleological language captures a crucial aspect of the situation that is not conveyed in the reformulation. 211 

To try to salvage the reformulation, one might add to it the idea that the intention in question causes the resulting action only via bodily motions that the agent has guidance of or control over, rather than by involuntary shaking and the like. 212 But the trouble with this is that “guidance” and “control” are themselves teleological notions – for guidance or control is always guidance or control towards an end or goal – so that the revised analysis will not truly have eliminated teleology at all. 213 

Alternatively, Alfred Mele suggests that the reformulation might stipulate that the intention is the direct cause of the behavior, whereas in the case in which Bob nervously knocks over the water prematurely, the intention causes the behavior only indirectly. 214 Now, even in the case where nervousness plays no role, there is presumably neural activity that comes between the intention and the behavior. So why wouldn’t that make the intention only an indirect cause even in that case? Mele’s response is to include all of that neural activity as part of the behavioral effect that the intention directly causes. But now another problem for the reformulation arises. For if, in this case, the neural activity that generates the knocking over of the glass is to be included in the behavioral effect that the intention directly produces, why couldn’t Bob’s nervousness in the other case also be included as part of the behavioral effect that the intention directly produces? And if it is, then Mele’s appeal to what the intention directly causes will fail to salvage the reformulation. Bob’s intention will be the direct cause of the effect in both cases, so that directness will not suffice to explain what makes the purposive behavior different from the case where Bob is nervous. 215 

Another option might be simply to stipulate, in the reformulation, that nervousness played no role. In other words, we can reformulate the statement that Bob knocked over the glass of water for the purpose of distracting Fred by saying instead that Bob had the intention of distracting Fred and this caused him to knock over the glass of water (but without nervousness playing a causal role). But it turns out that even this won’t work. For there are cases of purposive action in which nervousness does play a role. For instance, consider a weightlifter who is able to lift an extremely heavy barbell only because his intention to do so is associated with nervous excitement that provides an extra burst of energy. 216 Since that is possible, it is not correct to say that purposive action is distinguished by the absence of nervousness, in which case reference to such an absence cannot salvage the reformulation. 

G. F. Schueler notes another problem facing the thesis that any action can be explained in non-teleological terms by citing the belief and desire that served as efficient causes of the action. 217 The problem is that identifying the desire in question will itself inevitably bring back in an implicit appeal to teleology. Suppose I donate to a charity, you ask me why I did so, and I respond: “Because it’s a worthy cause.” You, as an observer of my action, might reasonably attribute to me the desire to support the cause and explain my action by reference to this desire. But from my first-person point of view, I may, nevertheless, not conceptualize my reason for acting in terms of such a desire. That is to say, when I consider whether I should make the donation, what occurs to me would in normal cases be a judgment like the judgement that this is a worthy cause, rather than the judgment that donating to it will satisfy a desire I have. The fact that I have such a desire, even when it is obvious that I do, would normally not enter into my deliberative process. What enters into the process are questions like “Is this a worthy cause?” rather than questions like “What sorts of desires do I have?” 218 

Moreover, the third-person description of me as having such a desire is parasitic on the first-person conceptualization of my action. There are, after all, several possible desires you could attribute to me in order to explain my action. For example, you could attribute my action to the desire to be seen by others as generous, or even (though this would be strange!) the desire to get rid of some excess cash. The reason you in fact attribute to me instead the desire to support the cause is that I have told you that the reason I made the donation is that the cause is worthy. In other words, it is that I have the end or goal of supporting the cause that justifies the attribution to me of a desire to support it. But that is a teleological description of my action. Hence the attribution of the desire presupposes, at least implicitly, such a teleological description. It hardly makes sense, then, to suggest that an explanation of my actions in terms of my beliefs and desires, considered just as efficient causes, can entirely replace a teleological description of my action. For such a description is implicit in the very attribution to me of a desire. 

The lesson of all this is, again, that there is an aspect to volitional behavior that cannot be captured in a description that makes reference only to efficient causality. As Scott Sehon observes after surveying the relevant contemporary philosophical literature: 

The [efficient] causal theory of action is still the dominant theory… But despite the hegemonic status of the causal theory, the problem of deviant causal chains has yet to be successfully addressed, and, thus, the alleged elimination of teleology has not been carried through. Moreover, it appears that there is a steady pattern in the failed attempts, consistent with what we would expect if there is an irreducibly teleological element to human agency. 219 

But might not the critic of teleology at this point simply dig in his heels and argue that if we can’t give a reductionist account of the teleology of human action, we should just dismiss it as an illusion? No, because this eliminativist move is no more coherent than is the eliminativist approach to concepts criticized in the previous chapter. Even to propose doing this is itself to posit an end or goal – the goal of solving the problem teleology poses for the reductionist – and thus to manifest teleology in the very act of denying it! As Sehon says of eliminativists who suggest that they are speaking of teleology only for the sake of formulating a reductio ad absurdum against it: “When we argue by reductio, we assume the truth of something in order to show that it leads to a contradiction and thus must be false after all. That is, to argue by reductio is to do something for a purpose.”220 Similarly, if the eliminativist says that he uses teleological language to describe his proposal only for the sake of argument, or only for the sake of communicating his position, he is manifesting teleology in the very act of denying that it is real. As Alfred North Whitehead (1861-1947) once observed, those who are “animated by the purpose of proving that they are purposeless constitute an interesting subject for study.”221 

What is freedom? 

On analysis, then, we simply cannot make coherent sense of the suggestions that there are no volitions or that action involves only efficient causality rather than teleology. Hence we must conclude that the will is real and that it really is directed towards the ends grasped by the intellect, as toward a final cause. Let’s turn now to the question of whether the will is free. Here too I need to begin by defining my terms, and once again the conception I will be defending is Aquinas’s. Modern discussions of free will often identify two conditions, one or both of which are taken to be necessary and sufficient for an agent’s acting freely: first, that the agent is the source of his own actions (sometimes called the “principle of ultimate authorship”222 ); and second, that the agent could have acted otherwise than he does (often called the “principle of alternate possibilities” 223 ). Aquinas too characterizes free choice in terms of these conditions. 224 But he doesn’t regard them as equally fundamental. As I have said, for Aquinas, to have a will is to be the source of one’s own activity in the fullest possible way, which entails acting from intellectual knowledge (as opposed to acting from natural appetite or from sensory knowledge). And that is also, in his view, what it is to have freedom of choice. The ability to do otherwise is a byproduct of this. 225 

We’ll come back to that, but let’s begin by noting that though I have been speaking both of free will and of free action, the former is the more fundamental notion and the latter is not meant to be understood independently of it. For there is a sense in which one might be said to act freely and yet not have free will, and a sense in which one might be unable to act freely and still have free will. 226 For example, a dog that slips the leash and runs off can be said to act freely insofar as it is doing what it wants to do, but since it lacks an intellect, it is not acting from free will. A human being might be bound and thus prevented from acting freely, but his will would nevertheless remain free insofar as he can still will to perform the acts he is prevented from carrying out. When I speak of “free action,” I mean, specifically, action that flows from free will, and thus action of which a rational substance is the ultimate author.

Note also that freedom as I understand it does not entail the will’s utter indifference to the ends toward which it might aim. Like every other substance, a rational substance is of its nature directed toward some end. Aquinas would say that that end is its happiness, which is the realization of what is good for it. Some might dispute this, but for present purposes we needn’t settle the question of what, specifically, is our natural end. We need note only that in the case of a rational substance, an end toward which it is directed can be known intellectually, and that the will or rational appetite just is that power by virtue of which such a substance can aim at an end qua grasped by the intellect. 227 A will that was not of its nature directed toward what the intellect takes to be good would be like an acorn that was not of its nature directed toward becoming an oak. The latter just wouldn’t really be an acorn, and the former just wouldn’t really be a will. 228 

We have here a kind of necessity that Aquinas calls natural necessity, since it has to do with what a thing cannot not do given its nature or essence. 229 This contrasts with what he calls the necessity of coercion, which has to do with a thing’s being forced to do what is contrary to its natural tendencies, as when I imagined tying a plant’s branches back so that its leaves were kept in the shade. Obviously, necessity of coercion is incompatible with freedom of choice, since it would involve forcing a rational substance to do something contrary to what its intellect takes to be good, and thus contrary to what it wills. But natural necessity is not contrary to freedom of choice. On the contrary, just as you might say that a plant is more free if I do not tie its branches back – that is to say, if I let it do what it “wants” to do, what it cannot help aiming to do given the kind of thing it is – so too a will is more free if it is not prevented from doing what it cannot help doing given the kind of thing it is, namely pursuing what the intellect takes to be good. A tree that did not aim at getting its branches toward the sun would not be a freer tree, but just a defective tree. And a will that did not aim at the good would not be a freer will, but just a defective will. 

Now, Aquinas distinguishes a third kind of necessity, which, perhaps somewhat confusingly, he calls necessity of the end. The reference here is not to our ultimate natural end, which, as I have said, Aquinas takes to be happiness, understood as the realization of what is in fact good for a rational substance. The reference is rather to what is perceived to be good, or to the means by which the good might be attained. Now, the intellect can be mistaken about what is in fact good, and it might note that several different means to attaining what it takes to be good are possible. In that sense, it is not necessitated toward the ends it pursues. Necessity of the end in the present sense would exist if the intellect had such a penetrating grasp of what is in fact good that it could not not perceive it to be good, or if there were no possible means of attaining the good other than the one the intellect takes to be a means. In this sort of situation, the will cannot will otherwise. Importantly, for Aquinas, this kind of necessity, like natural necessity, is not contrary to the will’s nature and not contrary to freedom. 

This is why I say that, for Aquinas, ability to do otherwise is a byproduct of acting from intellectual knowledge, and thus of freedom. But it is not itself the essence of freedom. The essence of freedom is acting from intellectual knowledge, and in cases where we are capable of misperceiving what is in fact good, or of entertaining alternative means of achieving the good, this yields alternative possible courses of action. The deep underlying reason for this is that knowing things intellectually entails bringing them under concepts, and there are different ways that the same things can be conceptualized. As Elizabeth Anscombe notes in a slightly different context, “a man may know that he is doing a thing under one description, and not under another,”230 and there are typically alternative possible descriptions under which we might know a thing or an action. John Haldane writes: 

[T]o think of an item is always to think of it via some conception… For any naturally individuated object or property there are indefinitely many non-equivalent ways of thinking about it. That is to say, the structure of the conceptual order, which is expressed in judgements and actions, is richer and more abstract than that of the natural order. 231

To borrow an example from Haldane, every triangle is necessarily a trilateral and vice versa. Hence, triangles and trilaterals are not different objects. Nevertheless, the concept of being a triangle is distinct from the concept of being a trilateral. Thus, one might think about a given object in terms of one of these concepts and yet not think about it in terms of the other. To borrow a famous set of phrases from Wilfrid Sellars (1912-1989), the intellect always brings things into “the logical space of reasons,” of which the conceptual order is a part, and that space is wider than the “space of causes” into which our behavior and the objects we interact with also fall. 232 The same one natural order of causally related physical objects can be carved up conceptually by the intellect in innumerably different possible ways. 

This reflects what contemporary philosophers call the thesis of the indeterminacy of meaning, which generalizes what we saw in chapter 3 about the impossibility of reducing concepts to either mental images or patterns of behavior. As we there noted, there are always alternative possible ascriptions of conceptual content to any mental imagery or behavioral pattern. Conceptual content outstrips anything that could be captured in a description of mental images or behavior. But it turns out that it also outstrips any description we could give, however thorough, of a thinker’s mental imagery and behavior together with the details of his neurophysiology and of the larger physical world with which the thinker causally interacts. No matter what this collection of facts turns out to contain, there are always alternative possible ways the thinker might conceptualize the world, consistent with this set of facts. The collection of physical facts is therefore indeterminate between different possible ascriptions of concepts, beliefs, etc. to a thinker – that is to say, it is not by itself sufficient to determine what a thinker is thinking. (We will have reason in a later chapter to revisit this thesis of the indeterminacy of meaning, and to see why philosophers of very different basic metaphysical commitments, materialist no less than dualist, have affirmed it. 233 ) 

Now, since the intellect can in principle conceptualize or describe the same one thing in different ways and thus, depending on the conceptualization or description, judge the very same thing either as a good end (or means), or judge instead that some other end (or means) is better, it follows that the agent could in principle either will or not will the same end (or means). Thus he has the ability to do otherwise, just as a consequence of having an intellect. Non-human animals and inanimate things lack this ability to do otherwise precisely because they do not bring things under concepts, and therefore don’t have alternative ways of conceptualizing things and the flexibility of action that that entails. 

This gives us a way to understand the role of circumstances in choice and action. 234 Recall that I said that whether and how the causal powers of a substance will manifest depends crucially on context, and in particular on which of the causal powers of other things are either manifesting or not manifesting. This, you might also recall, is why I said that it is a mistake to think of laws of nature as the fundamental level of physical reality. Laws of nature are instead a description of the way things will go if there obtains a (typically idealized) situation in which substances and their powers are operating in tandem in a certain specific way. Hence determinism, which presupposes that laws of nature are fundamental and that they describe the way things always actually go, gets nature wrong. However, this does not entail that there is no necessitation in nature. On the contrary, though the idealized situation described by a law may not in fact obtain, if it obtains, then things really will go the way the law says they will. That is why it is possible to discover at least idealized laws of nature. 

But though this is true of the order of inanimate things, and of plants and non-human animals, it is not true of human beings. For the context in which our causal powers operate does not include merely the other substances, together with their powers, with which we may interact. It also includes the way we actually conceptualize all of this other context, and, again, there are typically alternative possible ways in which we might conceptualize it. Whatever particular conceptualization we adopt is part of what governs how we act, but because there is nothing in the intellect as such that entails that one conceptualization rather than another will prevail, there is nothing that necessitates a particular outcome, the way that a certain outcome will be necessitated when non-human physical substances and their powers interact in just the right way. 

This is the reason for what Donald Davidson (1917-2003) famously called the “anomalism of the mental” – the fact that there are no strict laws correlating mental and physical phenomena, the way there are strict laws of physics and chemistry. 235 That the conceptual order outstrips the physical order rules out there being such laws. Now, one might argue, on the basis of this anomalism, that our intellectual capacities are immaterial. Indeed, I will argue for that conclusion in a later chapter. But Davidson himself did not draw that conclusion, and we need not insist upon it for present purposes. We need not appeal to mind-body dualism in order to defend free will, any more than we needed to do so in order to defend the reality and irreducibility of the self or the reality and irreducibility of the intellect. What has been said so far commits us only to rejecting a reductionist brand of materialism, even if we were to grant for the sake of argument that the mind might in some way be material. One alternative to materialist reductionism would be an account that relates mind to body in terms of the Aristotelian four-causal explanatory framework described earlier. For example, consider the action of my typing up this chapter. That my intellect conceptualizes what I am doing as an act of typing (as opposed, say, to the typing being an involuntary spasm) is the formal cause of the action; the end or goal of typing up the chapter, toward which my will is directed, is the final cause; and the relevant bodily processes are the material and efficient causes. I am a single substance of which these four irreducibly different factors are aspects. 236 

That is the case, anyway, if indeed I am a substance with bodily properties. But as I have said before, for the purposes of the first four chapters of this book, we can bracket off the question of whether the substance that I am is wholly material, wholly immaterial, or some compound of the material and immaterial. This is a question that naturally arises when we consider the relationship of the intellect and will to human physiology, and it is one I will address in later chapters. But it can largely be put to one side when considering the nature of free will per se. On the Thomistic conception that I am defending, the will is free insofar as it is oriented toward an end that the intellect conceptualizes as worth pursuing (which entails that volition arises from within the agent in the fullest possible way, as the “principle of ultimate authorship” requires); and there are typically alternative ways the intellect might conceptualize an end (which entails that the agent could have willed otherwise, as the “principle of alternate possibilities” requires). This is so whether the actions that result take the form of bodily movements (as when I type up this chapter) or exercises in thought (as when I lie back in bed and think through what I want to write before getting up to do it). How this relates to physiological processes is of secondary importance (albeit I will have reason to say a little more about that presently, and will say a lot more about it later in the book). The will could be free in the sense described whether we go on to spell out its relation to human physiology in a Cartesian manner, a materialist manner, or the Aristotelian hylemorphist manner I favor. 

Some contemporary philosophers have suggested that the key to understanding the nature of free will is the insight that action must be analyzed in teleological terms, and not merely in terms of efficient causality. 237 Since I have emphasized the irreducibly teleological character of action, I would agree with this as far as it goes. However, it is crucial to emphasize that teleology as such is merely a necessary condition for freedom, not a sufficient condition. As I have suggested, non-human animals, and even plants and inanimate things, can be said to be directed towards ends or goals, but they do not have wills, let alone free will. What makes the will free is that it is directed, specifically, toward an end as conceptualized by the intellect. 238 

Now, given the way I have characterized the nature of free will, its reality follows straightaway from what I have argued for so far. In particular, I have argued that we are substances of the kind that have intellects; that the actions of these substances are (at least in some cases) directed toward what the intellect judges to be worth pursuing (which satisfies the “principle of ultimate authorship”); 239 and that there are typically alternative ways in which the intellect could conceptualize the ends it considers and, consequently, alternative ends the will might pursue (which satisfies the “principle of alternate possibilities”). Since that is just what it is to have free will, it follows that we have free will. 

But more can be said. Aquinas offers an argument for the reality of freedom. 240 It begins like this: “Man has free choice. Otherwise, deliberations, exhortations, precepts, prohibitions, rewards, and punishments would make no sense.”241 Now, you might think that what he is doing here is really just giving an argument for why it would be horrible if free choice were an illusion, insofar as the basic presupposition of morality and related practices would in that case be undermined. And you may go on to point out that this doesn’t really show that free choice is real, but merely gives a motivation for wanting it to be real. 

But I don’t think that that is Aquinas’s argument at all. I would suggest that what he is saying is that it is simply a datum that our practices of deliberating, exhorting, and so forth do in fact make sense. If anyone claims to doubt this, we can point out that such doubt is, on analysis, incoherent – that one has actually to engage in deliberation and the like precisely in the act of trying to justify one’s denial of it. Now, if this datum makes sense only on the assumption that we have free choice, then, Aquinas concludes, we must have free choice. 242 Aquinas elaborates as follows: 

Now a man acts by judgment, since through his cognitive power he judges that something should be pursued or avoided. But the reason why he acts by free judgment and is able to go in alternative ways is that in the case of a particular action this judgment arises from a comparison made by reason and not from natural instinct. For with respect to contingent matters, reason has an openness with respect to opposites, as is clear from dialectical syllogisms and rhetorical persuasions. But particular actions are contingent matters, and so with respect to them the judgment of reason is related to different alternatives and is not determined to just one. Accordingly, by the very fact that he is rational, man must have free choice. 243 

Aquinas’s reference to reason’s “openness with respect to opposites” and to its being “related to different alternatives and… not determined to just one” is, I suggest, essentially an appeal to the way that the conceptual order outstrips the order of things conceptualized. And it is this outstripping that entails the possibility of willing otherwise. Because the intellect is not necessitated to conceptualizing things one way rather than another, the will is not necessitated to being moved to one end rather than another. But neither is choice random or arbitrary. For choice is always for a reason, even if the reason is not necessitated. And to act for a reason is precisely not to act randomly or arbitrarily. 244 

Searle presents an argument which is similar to Aquinas’s, and which elaborates on the idea that engaging in rational deliberation while denying the freedom of the will is ultimately incoherent. 245 He notes, first, that even when we have reasons for action that we take to be decisive, we cannot help but judge that it is up to us whether or not to go ahead and act on those reasons, and that the reasons by themselves don’t suffice to generate action. 246 We experience action as a matter of our making things happen, rather than of our merely observing that they happen. We experience a reason for action as something which we ourselves have to make effective, rather than being effective on its own. And we experience actions that simply occur automatically as a consequence of our having a reason for doing them (such as acting from a compulsion to seek a certain end) precisely as outside our control. Here Searle is describing what, as we noted earlier, he calls the “gap” between having a belief and desire on the one hand, and actually deciding and acting on the other, that exists whenever we take ourselves to act freely. 

Now, Searle argues that even if I judge that these features of the phenomenology of choosing are illusory, I nevertheless cannot possibly act except in a way that presupposes that they are not illusory. 247 For example, I cannot “sit back” and passively wait for my actions to happen rather than actively trying to make them happen. For one thing, even when, after “sitting back” in this way, some action finally does occur, it will be experienced precisely as something I myself did rather than something I merely observed occurring. For another thing, even deciding to carry out this policy of “sitting back” and waiting for action to occur is itself experienced as an active choice on my part. We cannot avoid having to choose, and we cannot experience the choices we make except as up to us, and thus free. 

Searle himself stops short of claiming that this argument establishes the reality of free will, but in my estimation he is wrong to do so. His view seems to be that, even if I cannot coherently deliberate and choose while at the same time taking myself to lack free will, it might for all that still be that this is just a matter of finding it psychologically irresistible to believe in something that is not in fact real. But it is no good merely to say that it might be unreal. The skeptic needs to provide for us a coherent scenario on which freedom is unreal even though we seem to experience it. And I submit that there is no such scenario. Searle says that, though at the psychological level there is nothing that necessitates what I will choose, it may at least in principle be the case that there is something at the neural level that does so, of which I am unaware. 248 But this cannot be right, and we have already seen the reason why. It has to do with the “anomalism of the mental,” the fact that there is nothing at the physical level of description that necessitates any particular description at the mental level (where, you will recall, this will be the case even if materialism is true). Hence, no matter what is going on at the neural level, it will not suffice to determine my choice. The anomalism of the mental entails that the “gap” described by Searle is metaphysical in nature rather than merely epistemic – that it is a gap in reality itself rather than in our knowledge of reality. (And again, this is so even if materialism is true.) So, it’s not just that we cannot help but believe that free will is real. It’s that there is no coherent scenario in which free will merely seems to be real but is not. 

I have said that acting from intellectual knowledge is the essence of the will’s freedom, and that the ability to do otherwise is a byproduct of this, rather than being itself the essence of freedom. I have also qualified the claim that there are alternative ways in which the intellect might conceptualize things by saying that this is typically the case. That leaves it open that there might be cases where the intellect is unable to conceive of something other than as a good to be pursued, or where it might know there to be only one possible means to attaining some good. In such cases, I have said, the will is subject to what Aquinas calls necessity of the end. Even in these cases, since the end the will is fixed on is an end grasped by the intellect, the agent is the source of his own activity in the fullest possible way. The agent thus remains free, on the Thomistic conception of freedom that I have been advocating. 249 

Does this entail that the ability to do otherwise drops out as irrelevant to free will? By no means. For we have to consider how an agent comes to be unable to conceive of some end except as good, and thus unable to will otherwise. Suppose someone could, initially, will either to perform some action or not to perform it, but through habitually choosing not to perform it molds his character in such a way that he eventually becomes unable to will it. Robert Kane describes the actions that result from such choices as “selfforming actions.” 250 And as he notes, it is natural to characterize the actions that flow from our having a certain kind of moral character as done “of our own free will,” even if we could not have done otherwise. For it is precisely because of earlier actions where we could have done otherwise that we have come to have such a character. This does not entail that we could at some point in the past have willed otherwise in absolutely every respect. As I noted earlier, for Aquinas, the will is as a matter of natural necessity oriented toward what the intellect (correctly or incorrectly) takes to be good. But to the extent that exactly what the intellect takes to be good can vary, so too can what the will chooses vary. 

Readers familiar with the modern philosophical debate over free will might wonder whether Aquinas’s position amounts to a compatibilist theory of freedom. 251 I would answer that it most certainly does not. For one thing, compatibilism holds that freedom is consistent with determinism, and as we have seen, from a Thomistic point of view, determinism is a wrongheaded conception of the natural world. So, reconciling freedom with determinism is simply not the sort of project that a thinker like Aquinas has in view. For another thing, precisely because of its determinism, compatibilism makes something outside the agent – namely the laws of nature together with the state of the universe prior to the agent’s existence – the source of his activity. That conflicts with what I have said is Aquinas’s account of the essence of free choice. Again, for Aquinas, the agent himself, and not the laws of nature or the state of the physical universe prior to the moment of choice, is the ultimate author of his actions. 252 

Is Aquinas’s view of free choice what philosophers call a “libertarian” view, then? 253 As Eleonore Stump notes, this depends on how one understands libertarianism. 254 If libertarian free will is taken to entail the ability to do otherwise without qualification, then Aquinas’s account is not libertarian. But Stump seems to think that a view according to which it is sufficient for freedom that an action originate in the intellect and will of the agent could reasonably be construed as a version of libertarianism. And on this construal, Aquinas’s account would be libertarian. Specifically, it seems closest to what is in contemporary philosophy called an “agent-causal” libertarian position. 255 Perhaps the issue is, at the end of the day, semantic. But I would point out that participants on all sides of the contemporary debate about free will, compatibilism, and libertarianism usually make philosophical assumptions that Aquinas and other Thomistic philosophers would reject. Hence it is not easy to situate Aquinas’s position within that debate, and trying to do so can threaten to obscure his position as much as it might illuminate it. In any event, the conception of free will that I have been defending in this chapter is Aquinas’s conception, specifically. Whether or not it corresponds to free will as others conceive of it is, for my purposes, neither here nor there.

Arguments against free will 

Let us turn now to arguments purporting to show that free will is an illusion. As Mark Balaguer notes, there have historically been three main arguments of this sort, which he labels the classical argument, the random-orpredetermined argument, and the scientific argument. 256 Let’s consider each of them in turn. We’ll then consider some further, less well-known but nevertheless important arguments against free will.

The classical argument 

The classical argument is the traditional argument from determinism. Determinism holds that the state of the physical universe at any particular moment, together with the laws of nature, necessitate what occurs at any later moment. The claim of the classical argument against free will is that given this determinist thesis, our choices and actions cannot be free. One influential response to this argument is the compatibilist position that properly understood, free will is compatible with determinism. Another is to appeal to the Cartesian dualist conception of the mind as an immaterial substance, and argue that free will is secured by virtue of this immaterial substance’s causal interference with the material world. But as I have indicated, I do not accept compatibilism, and I do not think that an appeal to dualism is necessary in order to uphold the reality of free will. 257 

The first thing to say in response to the classical argument is that it rests on a false premise. In particular, and as I have already suggested, determinism is simply mistaken about the nature of material substances and their causal relationships. As causal powers theorists Rani Lill Anjum and Stephen Mumford argue: 

Causes genuinely produce their effects. But this does not require that they necessitate them. Causing an effect is not the same as guaranteeing it. Causal dispositionalism offers an alternative account of causal production… Causes tend or dispose toward their effects with varying degrees of strength in different cases. They often succeed in producing these effects but, even when they do so, they did not through any necessitation. 258 

When physical substances get into relatively stable relationships so that their causal powers operate in tandem, the result is what Nancy Cartwright calls a “nomological machine,” a system whose behavior approximates the description given by a law of nature. 259 An example would be the solar system, whose constituents are related to one another in such a way that their causal powers, operating together, give rise to a pattern of behavior that approximates Kepler’s laws of planetary motion. On this account, laws of nature are not fundamental features of the natural world, and the world does not behave in the rigidly deterministic way the laws describe. Rather, a law of nature is an idealized description of how a system of physical objects will behave if the causal powers of those objects are activated in tandem. 

This account has many able advocates in contemporary philosophy, and is motivated independently of the debate over free will. Naturally, it requires a detailed exposition and defense. Given the treatment of issues in general metaphysics and philosophy of science that that would require, this is beyond the scope of a book on human nature, though I have provided such exposition and defense elsewhere. 260 The point to emphasize for present purposes is that determinism can hardly be taken for granted. It too requires defense, and in fact (I and other contemporary causal powers theorists maintain) is mistaken in its account of the natural world and its causal properties. Hence, insofar as the classical argument against free will takes determinism for granted, it begs the question at best and rests on a false premise at worst. 

Nor, even apart from these issues in philosophy of science, can determinism claim to have modern science on its side. As Balaguer points out, a major problem with the classical argument against free will is that while it seemed to have scientific support in the heyday of Newtonian physics, it no longer does so after the rise of quantum mechanics. 261 For quantum physics contains laws that are probabilistic rather than deterministic. To be sure, some would speculate that there might be deeper laws that have not yet been discovered but which would be consistent with determinism. But the point is that science itself does not currently support such a speculation. 

Others would propose that even if determinism does not hold at the quantum level, at the macroscopic level of the brain, quantum effects cancel out and the physical world operates more or less as determinism describes it. 262 One problem with this reply is that it too is open to challenge. Some have argued that there are physical mechanisms by which quantum effects can be amplified sufficiently to make a difference at the macroscopic level of the brain. 263 But a deeper problem is that the concession that at least the quantum level is probabilistic rather than deterministic effectively undermines the classical argument against free will. For the appeal is now no longer to the thesis of universal determinism, but rather to the purported features of some specific part of the natural world. That makes it a different argument from the classical argument against free will – specifically, a version of the scientific argument, which we’ll consider below. 

If the classical argument gets matter wrong, it also presupposes a mistaken conception of the mind’s relationship to it. Given the “anomalism of the mental” referred to earlier, the totality of physical facts at a given time would not determine the mental facts at any later time even if they determined the physical facts at every later time. Hence they would not determine what we think and will at any later time. Davidson thus took his thesis of the anomalism of the mental to secure free will. 264 As I have noted, Davidson did not take this anomalism to entail metaphysical dualism. For Davidson, the physical and the mental comprise two irreducibly different modes of description of human beings and their behavior, but they do not correspond to two irreducibly different realities. Rather, they are different ways of describing the same one reality, which Davidson took to be physical. His view is thus known as “anomalous monism.” It is a kind of “monism” insofar as it posits a single, physical reality to which both physical and mental predicates can be ascribed. It is “anomalous” insofar as it denies that there can be any laws of nature by which the totality of physical facts would entail the mental facts. 

Some readers are bound to wonder whether such a position can work. It might seem that either anomalous monism makes of mental descriptions mere ways of talking that capture nothing in objective reality or, if they do capture something real, that the view entails a kind of dualism in disguise. I sympathize with this suspicion, and will argue in a later chapter that the considerations that motivate Davidson’s anomalism do in fact entail the immateriality of the intellect and will. But as is evident from the fact that Davidson and his followers would not agree, this claim requires further argumentation. Commitment to the anomalism of the mental does not by itself entail dualism, but does support free will. Hence to appeal to it in rebutting the classical argument does not require insisting that the mind is immaterial. 

The random-or-predetermined argument 

The second of the main arguments against free will purports to show that we don’t have free will even if determinism is false. It takes the form of a dilemma. Either determinism is true, or it is not true. If it is true, then our choices are predetermined by forces outside our control, and free will is an illusion. But if determinism is not true, then our choices are uncaused. In that case, they amount to random occurrences, and a random occurrence cannot amount to a free choice. Hence, even if determinism is false, free will is an illusion. 265

As Anjum and Mumford point out, one problem with this sort of argument is that it rests on a false dilemma. 266 It assumes that if an effect is not causally predetermined, then it must be random. But as we have seen, there is an alternative way to understand causality, according to which causes tend or dispose toward their effects. Because they tend toward their effects, the effects are not random. It isn’t the case that just any old effect could have followed. But because causes merely tend or dispose rather than necessitate, effects are not predetermined either. 

Another problem is that the argument presupposes that the intelligibility of a choice is to be found only in identifying its efficient cause. Hence, in the scenario where a choice is not determined by some antecedent efficient cause, the argument claims that the choice must be random. But as I emphasized earlier, there is in human action an irreducible element of final causality or teleology, and this is crucial to its intelligibility. If an action is not predetermined but is nevertheless done for the sake of some particular end, then it is hardly random. That the agent decided to carry out the action with that end in view makes the action intelligible. Again, on the Thomistic account of free will that I am defending, freedom is a matter of our choices resulting from our rational deliberation about ends and means. And it makes no sense to regard as “random” a choice that reflects such deliberation. 

Balaguer makes the additional point that as long as I made the choice I did, the fact that nothing outside me caused me to make it is hardly inconsistent with free will; on the contrary, that is precisely what it is to have free will. Hence the random-or-predetermined argument’s interpretation of the “random” side of the purported dilemma that it poses is just confused. Properly understood, the scenario in which our choices are not determined by forces outside us is not inconsistent with free will but rather just is our having free will. 267 I think this is correct as far as it goes, insofar as it reflects the point that having free will is fundamentally about being the ultimate author of one’s actions. But it is important to emphasize that what it means to be the ultimate author of one’s actions is for those actions to flow from one’s rational deliberation about ends and means. This is crucial to understanding why an agent’s not being caused to choose as he does by anything outside him does not entail that the choice is arbitrary or unintelligible, as the accusation of randomness implies. 

For the same reason, it is also crucial that Anjum and Mumford’s point be supplemented with an emphasis on the fact that it is, specifically, rational deliberation that tends or disposes us toward the choices we make. It is this consideration, and not tending or disposing as such, that makes freedom possible, since inanimate objects and non-human living things also have tendencies or dispositions but lack free will. To be sure, Anjum and Mumford recognize that mere tending or dispositionality is not sufficient for free will, and that reference to deliberation and the normative considerations it takes account of is also essential. 268 However, I would add that the teleological element that I have emphasized is no less essential. I have allowed that there is something at least approximating causal necessitation in the case of non-rational substances, when the conditions are right for the triggering of their causal powers in tandem. But there is not even an approximation of such necessitation in the case of human beings. If we are thinking of rational deliberation entirely as a matter of the efficient causation of our actions, this might seem to entail a residue of randomness or unintelligibility. This disappears when we keep in mind that action is ultimately to be understood in terms of final causes. 

The scientific argument 

Of the three most influential arguments against free will, it is the scientific argument that has in the contemporary debate gotten the most attention from the general public. Recent years have seen the popularization of the claim that neuroscience and psychology, in particular, have shown free will to be an illusion. 269 

The first thing to note in response to such claims is that determinism turns out to be as much of a red herring in this context as it did in the case of the classical argument. Recall that, in response to the point that quantum mechanics has undermined the scientific case for determinism, it is often suggested that quantum effects cancel out at the macroscopic level, so that the brain might more or less operate in a deterministic way even if determinism does not hold at the quantum level. As Balaguer points out, the problem with this suggestion is that modern neuroscience, like quantum mechanics, is at least in part probabilistic rather than deterministic. 270 In particular, both a neuron’s release of a neurotransmitter and the neurotransmitter’s triggering of a neural firing operate in a probabilistic rather than deterministic way. 

Indeed, even to speak of probabilities here is potentially misleading if it is taken to imply that neural causes always make their effects highly probable. A historically influential conception of scientific explanation known as the “covering-law model” claims that the way science explains a phenomenon is to show that it is entailed by, or at least made probable by, a description of the conditions holding antecedent to the phenomenon together with some law of nature. But as philosopher of neuroscience Carl Craver notes, this model simply does not fit the way explanations in neuroscience actually work. 271 Rather, such explanations typically involve identifying the underlying mechanisms by which some higher-level phenomenon is brought about, where the phenomenon often not only does not follow with necessity from the operation of the mechanism, but (given all the factors that need to be in place together with possible interfering circumstances) may even occur with well under fifty percent probability. 

Hence contemporary neuroscience gives no more support to the deterministic premise of the classical argument than contemporary physics does. If opponents of free will are going to find support in neuroscience, they have to look elsewhere. However, some of them claim to find it in the work of neuroscientist Benjamin Libet (1916-2007). 272 In Libet’s famous experiments, subjects were asked to flex a wrist whenever they felt like doing so, and then to report on when they had become consciously aware of the urge to flex it. Their brains were wired so that the activity in the motor cortex responsible for causing their wrists to flex could be detected. While an average of 200 milliseconds passed between the conscious sense of willing and the flexing of the wrist, the activity in the motor cortex would begin an average of over 500 milliseconds before the flexing. Hence the conscious urge to flex, it is suggested, seems not to be the cause of the neural activity which initiates the flexing, but rather to follow that neural activity. 

Now, Libet himself qualified his conclusions, allowing that though we don’t initiate movements in the way we think we do, we can at least either inhibit or accede to them once initiated. But according to philosopher Alex Rosenberg, the work done by Libet and others “shows conclusively that the conscious decisions to do things never cause the actions we introspectively think they do” and “defenders of free will have been twisting themselves into knots” trying to show otherwise. 273 Similarly, biologist Jerry Coyne assures us that: 

“Decisions” made like that aren’t conscious ones. And if our choices are unconscious, with some determined well before the moment we think we’ve made them, then we don’t have free will in any meaningful sense. 274 

However, as several critics have pointed out, this line of argument commits a number of fallacies. 275 The first problem is that Libet didn’t show that the kind of neural activity he measured is invariably followed by flexing. Given his experimental setup, only cases where the activity was actually followed by flexing were detected. He didn’t check for cases where the neural activity occurred but was not followed by flexing. So, we have no evidence that that kind of neural activity is sufficient for the flexing. For all Libet showed, it may be that the neural activity in question leads to flexing (or doesn’t) depending on whether it is conjoined with a conscious free choice to flex. 276 

A second problem is that the sorts of actions Libet studied are highly idiosyncratic. The experimental setup required subjects to wait passively until they were struck by an urge to flex. But many of our actions don’t work like that, especially those we attribute to free choice. Instead, they involve active deliberation, the weighing of considerations for and against different possible courses of action. It’s hardly surprising that conscious deliberation has little influence on what we do in an experimental situation in which deliberation has been explicitly excluded. And it’s a fallacy of hasty generalization to extend conclusions derived from these artificial situations to all human action, including cases which do involve active deliberation. 277 

Third, even if the neural activity Libet identified had invariably been followed by a flexing of the wrist, that still wouldn’t show that the flexing wasn’t a product of free choice. For why should we assume that a choice is not free if it registers in consciousness a few hundred milliseconds after it is made? 278 Think of making a cup of coffee. You don’t explicitly think: “I will now proceed to move my hand toward the kettle; now I will pick it up; now I will pour hot water through the coffee grounds; now I will put the kettle down; now I will pick up a spoon.” You simply do it. You may, after the fact, bring to consciousness the various steps you just carried out; or you may not. We take the action to be free either way. After all, you are not having a muscle spasm, or sleepwalking, or hypnotized, or under duress, or in any other way in circumstances of the sort we would normally regard as incompatible with acting of your own free will. The notion that a free action essentially involves a series of fully conscious episodes of willing, each followed by a discrete bodily movement, is a straw man. 279 

It is also simply wrongheaded to think of voluntary actions as prompted by feelings and urges. (Recall the point made earlier that a volition should not be confused with the feelings or emotions associated with it.) As M. R. Bennett and P.M.S. Hacker point out, feeling an urge to sneeze does not make a sneeze voluntary. 280 Since Libet is willing to allow that we might at least inhibit actions initiated by unconscious neural processes, even if we don’t initiate them ourselves, Bennett and Hacker observe that: 

Strikingly, Libet’s theory would in effect assimilate all human voluntary action to the status of inhibited sneezes or sneezes which one did not choose to inhibit. For, in his view, all human movements are initiated by the brain before any awareness of a desire to move, and all that is left for voluntary control is the inhibiting or permitting of the movement that is already under way. 281 

As Bennett and Hacker go on to emphasize, being moved by an urge – such as an urge to sneeze, or to vomit, or to cough – is in fact the opposite of a voluntary action. Once again, Libet’s model of voluntary action is simply a straw man, so that his experiments have dubious relevance to the question of free will. 

A fourth problem is that Libet and those who draw sensationalistic conclusions from his work fail to consider alternative interpretations of the neural activity in question. Perhaps it correlates, not with the intention to flex, but rather with preparing to flex without necessarily intending to do so, or with imagining or thinking about flexing. Or perhaps it correlates with a general intention to flex as opposed to a proximal intention to do so.282 Think again of the coffee example. Suppose when you got up in the morning, you decided you wanted to make some coffee. You could be said to have formed a general intention to do so. But suppose also that you don’t actually make it until several minutes later, after using the bathroom, getting dressed, and going to the porch to get the newspaper. Only then did you decide it was time to go to the kitchen and actually make the coffee. At that point you formed a proximal intention to make the coffee. Similarly, the participants in Libet’s experiments could be said to form both a general intention that they will flex their wrists once they have a certain feeling, and then a proximal intention once the feeling actually arises. Nothing in Libet’s experiment tells us that the neural activity he cites correlates with the one kind of intention rather than the other, even if we were to concede (as we should not) that there is any reason to correlate it with an intention in the first place. 

As Raymond Tallis points out, the nature of the intentions involved even in this simple action of flexing the wrist is actually more complex than this last point indicates. 283 There is a sense in which the intention to perform the action could be said to have been formed many minutes before the subject flexed his wrist, when he had the experimental setup explained to him; or hours before, when he left the house to come take part in the experiment; or even days or weeks before, when he first agreed to participate. A long and complex series of psychological and physiological events played a role in what happened when the wrist was actually flexed. So why fixate on one particular bit of neural activity taken in isolation as the cause of the action? After all, neural activity and bodily movements do not by themselves entail action, free or otherwise. The spasmodic twitch of a muscle involves both neural activity and bodily movement, but it is not an action. 

So, the precise significance that a bit of neural activity or a bodily movement has for a given action cannot be read off from the physiological facts alone. It is only within the larger psychological context that we can make sense of it. For it is only the person as a whole, and not some subpersonal part of him such as an isolated bit of neural activity, who can properly be said to intend and to act. And so it is only the person as a whole, and not the neural activity, who can be said to be the cause of his actions. In pretending otherwise, Libet and those who appeal to his research in order to cast doubt on free will are simply presupposing a reductionist account of human nature – an account which defenders of free will would reject. Hence they cannot claim that Libet’s research supports reductionism and undermines free will without begging the question. 284 

Another influential argument against free will appeals to the work of neuroscientist J. D. Haynes and his colleagues, but it essentially commits the same fallacies as Libet’s argument. 285 To be sure, on the surface, Haynes’s results seem more impressive than Libet’s. Subjects in the Haynes study were asked to decide whether to press one of two buttons while their brain activity was measured using fMRI. It was found that activity in two brain regions predicted which button would be pressed up to ten seconds before the conscious decision was reported. The problem for free will seems obvious: If a person’s choice can be predicted from brain activity ten seconds before the choice is made, how can the choice be free? 

But there is much less to this argument than meets the eye. For one thing, the accuracy of the predictions in question is in fact only about 60 percent (where 50 percent would be pure chance). That is consistent with an interpretation of the relevant brain activity as corresponding, not to an unconscious decision to press a certain button, but rather to a mere unconscious bias or inclination toward pressing it, an inclination the subjects can and often do go on to resist. 286 Indeed, there is independent evidence that the brain regions in question are associated, not with willing, but with planning. 287 If the subjects in question are merely planning to press a certain button up to ten seconds before they do, it is hardly surprising if they often then go on to press it – but also, around forty percent of the time, change their minds. Certainly there is nothing in that that is incompatible with free will. 

As with Libet’s argument, the argument from Haynes’s study simply makes tendentious philosophical assumptions about how to interpret the brain activity in question. And as with Libet’s experimental setup, the action studied by Haynes is highly idiosyncratic. It involved choosing one action rather than another for no particular reason. Again, the actions we typically regard as free are not like that, but instead involve conscious deliberation. Hence it would be a fallacy to generalize from cases like the ones studied by Haynes to all action. 

Other influential scientific arguments against free will appeal to psychological research that shows that unconscious factors, social pressure, and the like can influence the choices we make. But these arguments are even feebler than the ones considered so far. 288 For example, consider psychologist Stanley Milgram’s famous obedience experiments, in which participants were instructed to administer what they falsely supposed were genuine electric shocks to people who gave incorrect answers to questions put to them. Many participants reluctantly obeyed these commands even when they seemed to be causing severe pain. Some have argued that such data casts doubt on the reality of free will. But as Mele says, “it’s [hard] to see exactly what the argument is supposed to be.”289 Is the claim that Milgram’s experimental setup made it inevitable that participants would obey? That can’t be it, because not every participant obeyed the commands. Is the idea merely that situations exist in which people find it difficult to disobey authority figures? If so, what defender of free will ever denied that? 

Social psychologist Daniel Wegner studied unusual cases in which subjects perform actions they are not conscious of. 290 He suggests that these cases show that free will is an illusion. But here too the argument is a non sequitur. That some of our behavior is not caused by conscious choices simply does not entail that none of our behavior is ever caused by such choices. 291 And it is especially fallacious to draw such a sweeping conclusion from idiosyncratic examples. Moreover, social scientific research on what are called “implementation intentions” shows that of two groups assigned to do a certain task, the one instructed consciously to decide how to carry it out is more likely actually to do it than the group that is not so instructed. This indicates that conscious intentions do in fact influence what we do.292 Of course, that is in any case just common sense. But the point is that, even if we are looking at the issue simply from the perspective of social psychology, the evidence does not point in the direction of Wegner’s extreme conclusion. 

Other arguments 

Let’s turn, finally, to some arguments that are less well-known to the general public, but which have gotten attention in academic philosophy. I have suggested that the Thomistic conception of free will that I have been defending is similar to what in contemporary philosophy is called “agentcausal libertarianism.” Against this view, Derk Pereboom has raised what has come to be known as “the luck objection.”293 The state of the physical universe at any particular time together with the laws of nature, whether conceived of in deterministic or probabilistic terms, would lead us to expect agents to behave in the future in certain specific ways. Meanwhile, agentcausal libertarianism claims that the behavior of these agents is the result of their free choices, where these choices have no causal antecedents external to the agent. Together these claims seem to imply that the behavior that results from these choices happens to correspond to what we would expect it to be from the laws of nature. But this entails “wild coincidences,” according to Pereboom, and “the proposal that agent-caused free choices do not diverge from what the statistical laws predict for the physical components of our actions would be so sharply opposed to what we would expect as to make it incredible.”294 Now, Pereboom acknowledges that agent-causal libertarians could deny the assumption that free choices will in fact generally correspond to what the laws of nature would lead us to expect. 295 And in my opinion, we should indeed deny it. But the trouble with this move, says Pereboom, is that “we have no evidence” that there really are any actions that diverge from what the laws of nature would lead us to expect. 296 

The problem with Pereboom’s claim that “we have no evidence” of this is not that it is false, but that it is true only in a trivial and uninteresting sense. For in fact the laws of nature as science knows them today give us no evidence one way or the other for any claim about how, specifically, any of us will act. Pereboom makes it sound as if we actually know of physical laws that entail or make probable what we will do at any given point in the future, and as if the question is what the agent-causal libertarian can say in the face of these laws. But of course, we know of no such laws. Reductionists assume that such laws must exist, but the assumption is based on a general metaphysical theory, not on any actual scientific findings. Hence the “evidence” no more favors Pereboom’s position than it undermines the agent-causal position. Talk of the “evidence” is thus a red herring. The dispute between Pereboom and the agent-causal libertarian is not going to be settled by existing empirical evidence, but by philosophical argumentation. 

Now, I would say that what a sound philosophical analysis of the situation really shows is the following. Laws of nature are not fundamental to physical reality. Rather, the natural world is, at the fundamental level, made up of substances of various kinds with their distinctive causal powers. When certain kinds of substance get into proximity to one another in a stable way, the result is what Cartwright calls a “nomological machine,” whose behavior approximates laws of nature. That is true, anyway, of relatively simple collections of substances, such as a solar system. With much more complex systems, there are no known laws, though in some cases that might change with further scientific investigation. But with human beings, not only are there no known laws, there can be no laws. The reason is that human beings are rational beings, and the “anomalism of the mental” makes it impossible in principle to subsume rational deliberation and its outcomes under laws of nature. Hence human beings and collections of human beings can never be “nomological machines,” so that there cannot be laws of psychology the way there are laws of planetary motion, laws of chemistry, and the like. 

If this is correct, then there can be no question of agent-causal libertarianism predicting that human actions will be contrary to what the laws of nature would lead us to expect them to be, because there not only are no laws of the relevant sort, but never could be such laws in the first place. Pereboom may disagree with this analysis, but the point is that his objection merely assumes it is wrong, and does nothing to show that it is. Hence he simply begs the question. 

Galen Strawson has developed a variation on the random-orpredetermined argument, holding that free will is impossible whether or not determinism is true. 297 In particular, he argues that the kind of selfdetermination that is essential to free will is in principle impossible. For what such self-determination requires is not just that the agent choose to act the way he does for certain reasons, but also that he is responsible for having those reasons in the first place. That requires him to have chosen those reasons, and in order for him rationally to have done so, there must be yet further reasons in light of which the reasons were chosen. But those reasons too will have to have been chosen, in light of yet further reasons, and so on ad infinitum. So, self-determination requires the making of an infinite series of choices, which is impossible. Hence the free will that depends on selfdetermination is also impossible. 

But the defender of free will need not accept Strawson’s assumption that freedom requires that the agent has chosen the reasons for which he acts. 298 Certainly the defender of Aquinas’s conception of free will need not accept it. As I have said, on Aquinas’s conception, freedom of the will is consistent with the natural necessity by which the will is directed toward what the intellect takes to be good. We don’t decide that the will will be directed toward what the intellect takes to be good; it just is so directed, of its very nature. Free will as Aquinas understands it is also consistent with necessity of the end, which obtains when the intellect cannot fail to conceive of some end as good or where there is only one possible means by which the end it takes to be good can be realized. In such cases, the agent cannot choose reasons other than those that are present to the intellect. There may also be cases where the agent simply judges some end to be worth pursuing for its own sake, rather than for the sake of anything else, even if some other judgement was possible. For example, you might on some occasion find it worthwhile for its own sake to tap your fingers on a certain glass. 299 Now, whether it is the will’s general directedness toward what the intellect perceives to be good, or the intellect’s inability to conceive of some specific end except as good or of some means as an avoidable way of achieving a good, or the intellect’s simply finding some good as worth doing for its own sake, we have in each case reasons for action that are simply given, as it were, rather than chosen. The agent is nevertheless free. For one thing, the reasons, even if they are in one of these ways just given, are the agent’s reasons, so that he is still the source of any action that they give rise to (as the “principle of ultimate authorship” requires). For another thing, it is still up to the agent whether or not to act on those reasons (as the “principle of alternate possibilities” requires). 

Now, this implies (to borrow some terminology from Stefaan Cuypers) a distinction between the object of the will and the exertion of the will. 300 When I say that some reasons are simply given rather than chosen, I am saying that the object toward which the will is directed can simply be given rather than chosen, consistent with the will’s freedom. There is a termination to any regress of final causes for the sake of which the will acts. I may pursue one end for the sake of another, and that for the sake of yet another, but at some point I simply have certain ends which are not chosen but in some sense are just given. Hence the regress posited by Strawson is not after all infinite, but this is consistent with the self-determination required for free will, both because the reasons in question are still my reasons and because it is up to me whether I act on them. 

But it might now seem that another vicious regress looms. For what about the exertion of the will? What makes it the case that I do in fact act or refrain from acting on the reasons in question? Doesn’t a choice to act require a cause, and if the act is to be a free and thus self-determined one, doesn’t that entail that the cause must be some further choice on my part? If that is the case, doesn’t that further choice itself require yet another choice as a cause, and so on ad infinitum? The answer is that this does not follow. Just as the will’s ultimate ends in the order of final causes are simply given rather than pursued for the sake of yet further ends, so too its ultimate acts in the order of efficient causes can bring about effects (such as bodily movements) without themselves being caused. The will is in this sense an uncaused cause or unmoved mover. 301 

This does not entail that its operation is random, either in the sense of being an exception to the principle of causality or in the sense of being unintelligible or devoid of explanation. It is not an exception to the principle of causality, because events require causes insofar as they contain some passive potentiality that requires actualization. For example, a statue requires a cause because the material out of which it is made initially only potentially has the shape of the statue. Something has to actualize that potential in order for the statue to exist. The exertion of the will, by contrast, is entirely active rather than passive. It is a bringing about that is not itself brought about by anything else. Yet it is not unintelligible or devoid of explanation, because it is done for a reason – in particular, for the sake of the end grasped by the intellect – and something done for a reason is the opposite of unintelligible. Hence the defender of agent-causal free will is not stuck with the dilemma of having to posit either “an infinite regress or a nonrational flip-flop.”302 

This completes the long answer to the question “What is mind?”, the short answer to which I gave in chapter 1. That short answer, you will recall, is that to have a mind is to be a self or a person, and that this entails being a substance which by nature possesses an intellect and a will. We have, over the course of the subsequent chapters, now seen in much greater detail what this means, and why it cannot coherently be denied that we have minds in this sense. We turn now to consider the nature of the body.