One of my main aims in this book is to explain how mental phenomena—consciousness, intentionality, mental causation, and all of the other features of our mental life—fit into the rest of the universe. How, for example, does consciousness exist in a universe that consists entirely of physical particles in fields of force? How can mental states function causally in such a universe? So far, most of our investigation has been about conscious mental phenomena. In this chapter we will begin a serious exploration of the nature and mode of existence of unconscious mental states.
Let us begin by asking, naïvely, Do unconscious mental states really exist? How can there be a state that is literally mental and at the same time totally unconscious? Such states would lack qualitativeness and subjectivity and would not be part of the unified field of consciousness. So in what sense, if any, would they be mental states? And if such things do exist, how can they function causally as mental states while they are unconscious? We have become so used to talking about the unconscious, so comfortable with the idea that there are unconscious mental states in addition to conscious mental states, that we have forgotten just how puzzling the notion of the unconscious really is. For Descartes, the answer to the question, Do unconscious mental states exist? is obvious. The idea of an unconscious mental state is a self contradiction. Mind is defined for Descartes as res cogitans (thinking being) and “thinking” for Descartes is just another name for consciousness. So the idea of an unconscious mental state would be the idea of an unconscious consciousness, a plain self-contradiction. For a long time the Cartesian idea that there is a necessary connection between the mental and consciousness was extremely influential. It is only in the past century or so that the idea and importance of unconscious mental states has come to be generally accepted. Freud is usually given most of the credit for this acceptance, but his ideas were certainly anticipated by Nietzsche and by several literary figures, of whom Dostoyevsky is probably the most important.
So what exactly is an unconscious mental state, such as an unconscious belief or desire? I think many people, including some extremely sophisticated authors such as Freud, have the following rather simplistic picture. An unconscious mental state is exactly like a conscious mental state only minus the consciousness. The problem with this picture is that it is very hard to make any sense of it. To see this try it out: think to yourself consciously, “George Washington was the first president of the United States.” Now do exactly the same thing, only unconsciously. Subtract the consciousness. I have no idea what it would be like to do that, or what the instruction is supposed to mean. Yet the notion of the unconscious seems to be one we cannot do without, so we had better try to explain it.
My strategy in this chapter, as in earlier chapters, will be to begin with simple and unproblematic cases and then build the more difficult and puzzling cases on top of them. Let us start with some unproblematic cases of attributions of mental states to people where the attribution is not of a state that is conscious then and there. To take an obvious sort of case, it can be truly said of me, even when I am sound asleep, that I believe that George Washington was the first president of the United States. Now what fact corresponds to this claim? What fact about me makes it true that I have this belief even when I am not conscious? Notice furthermore that we can even say of a person who is wide awake, and who happens to be thinking about something else entirely, that he believes that George Washington was the first president of the United States. So again, what fact corresponds to these claims? Notice that neither of these is a puzzling or controversial attribution of unconsciousness. Descartes himself could have agreed to the truth of either of these claims. In both cases the fact that corresponds to the claims is that there is in him a structure that is capable of producing the state in a conscious form. If when the man is awake you ask him, for example, who was the first president, he is capable of giving the correct answer because he is capable of producing the conscious thought in question. Notice that in this case, we have identified a structure not in virtue of its intrinsic structural features but in virtue of what it is capable of causing. This sort of attribution is very common in all sorts of real-life unproblematic cases. We say of a substance in a bottle that it is a cleanser or bleach or poison without identifying the chemical structure any further. We just identify it by what it does, not by what structure enables it to do it; and I am suggesting that when we say the man has the unconscious belief that George Washington was the first president of the United States we are identifying a structure in him, not in virtue of its intrinsic neurobiological features, but in virtue of what it does, in virtue of the conscious state that it is capable of causing.
In these cases we have identified one type of unconscious mental state, an unproblematic type that Freud described as “preconscious.”
A second type of unconscious mental state is more problematic. It often happens that an agent has mental states that function causally in her behavior, where she is totally unaware of the functioning of the mental state and may even sincerely deny it. Some of these cases are of the sort that Freud described as repression. But more generally, we can characterize these, again using the Freudian vocabulary, as the dynamic unconscious. These are cases where the unconscious mental state functions causally, even when unconscious. A Freudian style example is the case of Dora, who develops a cough because of her unconscious sexual desire for Herr K.1 Freudian examples are often problematic and much of his clinical work, I think, was scientifically inadequate. But let us take some cases where there really is little doubt about the scientific accuracy of the description. We considered in chapter 8 an example of hypnosis where the agent clearly acts out of a motive of which he is unaware and would presumably deny if he were challenged. In the hypnosis case the man has the desire to obey the order, “Open the window when you hear the word ‘Germany’” even though he is unaware that he has been given any such order and is unaware of any desire to carry out the order. These second types of cases we will call, following Freud, cases of repressed unconscious mental states.
A third type of unconscious mental state is also very commonly discussed in the cognitive science literature. These are the cases where the agent not only cannot bring the mental state to consciousness in fact, but could not bring it to consciousness even in principle, because it is not the sort of thing that can form the content of a conscious intentional state. So, for example, in cognitive science it is commonly said that a child learns a language by “unconsciously” applying many computational rules of universal grammar, or that the child is able to perceive visually by performing “unconscious” computational operations over the input that comes into the child’s retina. In both of these kinds of cases, both in the acquisition of language and the forming of perceptions, the computational rules are not the kinds of things that could ever be consciously thought. Ultimately they reduce entirely to massive sequences of zeros and ones, and whatever the child can do when he or she thinks, he or she cannot think in zeros and ones, and indeed the zeros and ones are just a manner of speaking. The zeros and ones exist in the mind of the observer and form a manner of description of what is going on unconsciously in the child’s mind. Let us call these cases, where the agent operates with rules that are not only unconscious in fact, but not even the sort of thing that could be conscious, the “deep unconscious.”
In addition to these three types, there is a fourth form of neurobiological phenomenon that is not conscious. There are all sorts of things going on in the brain, many of which function crucially in controlling our mental lives but that are not cases of mental phenomena at all. So, for example, the secretion of serotonin at the synaptic cleft is simply not a mental phenomenon. Seretonin is important for several kinds of mental phenomena, and indeed some important drugs, such as Prozac, are used specifically to influence serotonin, but there is no mental reality to the behavior of serotonin as such. Let us call these sorts of cases the “nonconscious.” There are other examples of the nonconscious that are more problematic. So, for example, when I am totally unconscious, the medulla will still control my breathing. This is why I do not die when I am unconscious or in a sound sleep. But there is no mental reality to the events in the medulla that keep me breathing even when unconscious. I am not unconsciously following the rule “Keep breathing”; rather, the medulla is just functioning in a nonmental fashion, in the same way that the stomach functions in a nonmental fashion when I am digesting food.
To summarize then, we have identified four types of unconscious phenomena: the preconscious, the repressed unconscious, the deep unconscious, and the nonconscious. The first and the fourth seem to me to be unproblematic. What about the second and the third? In the sections that follow, I will argue that the way to understand the repressed cases is on the model of the first, the preconscious; and the way to understand the third, the deep unconscious cases, is on the model of the fourth, the nonconscious cases.
I turn now to the cases of repression. Our question is this, How can a repressed mental state exist and function as a mental state when it is completely unconscious? Well, we already saw the answer to that in the case of the preconscious. To ascribe a mental state to a person at a time when the state is unconscious is just to ascribe to that person a structure, the details of which may be completely unknown, that is capable of producing that state in a conscious form. There is really no difficulty in saying of someone who is asleep that he believes that George Washington was the first president and there is no difficulty in attributing all sorts of beliefs to a conscious person even though he is not thinking about those beliefs at the time of the attribution. Now this method, it seems to me, works just as well for the second class of cases, the repression cases. If I say that Sam acts out of a repressed hostility to his brother or that Wolfgang acts out of an unconscious desire to fulfill the command he was given during the hypnosis, in both cases I am attributing a neurobiological structure capable of causing a mental state in a conscious form.
But now that leaves us with what seems to be the most difficult problem. How can these unconscious states, when unconscious, succeed in causing actual human behavior? How do we account for the “dynamic unconscious”? It seems to me that when we attribute these unconscious mental states to an agent, we are attributing neurobiological features capable of causing consciousness. Not only are they capable of causing conscious states, but they are capable of causing conscious, or indeed unconscious, behavior. But the question is how can the state function causally as a mental state at a time when it is nothing but an unconscious neurobiological structure? The way to answer that question, as we have done with earlier difficult questions, is to work up to it by taking the simple and more obvious cases first.
I once had a fractured wrist. This injury caused me a fair amount of pain during the day and the pain increased if I was not careful about the movements I made with my arm. I noticed an interesting thing in sleep. I would be totally sound asleep, so that I felt no pain whatever, and yet, my body would move during the night in such a way as to protect the injury. How should we describe such a case? Should we say that during sleep I had an unconscious pain and my unconscious pain caused me to behave in such a way as to avoid aggravating the pain? Or should we say, on the other hand, that while I was sound asleep I had no pain whatever, but rather the underlying neurobiology that was capable of causing the pain in a conscious form acted on me causally in such a way as to prevent any pain stimulation? It seems to me that the facts are the same in both cases. We do not normally talk about unconscious pains; but we easily could, and cases like this would give us the motivation for talking about unconscious pains. Notice in this case the neurobiology is capable of causing the pain in a conscious form, even though when I am sound asleep I do not consciously feel any pain. But, and this is the crucial point for this part of the discussion, the neurobiology that is capable of causing the pain in a conscious form is also capable of causing behavior appropriate to avoiding the pain even at a time when I do not feel the pain. Now that seems to me exactly right for describing the cases of the dynamic repressed unconscious. The agent is not conscious of any motivation when the dynamic unconscious is active. Nonetheless, there is a neurobiological structure that is capable both of causing the motivation to occur as part of his conscious thoughts and capable of causing behavior appropriate to having that motivation. The only disanalogy between this case and the pain case is that the agent may have extra reasons for not wanting to admit the motivation to himself. But—and this is the answer that I am proposing to the question—the mode of existence, the ontology, of the unconscious motivation, when unconscious, is that of a neurobiological structure capable of causing the motivation in a conscious form and capable of causing behavior that is appropriate to having that motivation. This, incidentally, is why the Freudians were so anxious to bring the unconscious to consciousness. As long as it remains unconscious, it is not in our control. We cannot reflect on it, or appraise it, or evaluate it, or subject it to rationality, in a way we normally can with motivations that exist as part of our conscious rational thought processes in the gap.
So far, then, in this chapter I have suggested that there are completely unproblematic cases of the unconscious, the cases we called the preconscious. These are the cases that even someone like Descartes could accept. But I have also, more controversially, maintained that these provide the appropriate model for considering the repressed cases, the cases where the “dynamic unconscious” is in operation. I am suggesting that the same sort of neurobiological processes that can cause a conscious state can also cause behavior appropriate to having that conscious state. So we have assimilated the first two types of cases of unconsciousness to what we already know about the brain and how it works, as well as what we know about our conscious mental life. No metaphysical mystery remains about the notion of the unconscious, at least for these sorts of cases.
But now let us turn to our third class of cases, the deep unconscious, and here the thesis I want to maintain can be stated quite simply. There are no such cases. There is no such thing as a deep unconscious mental state. There are nonconscious neurobiological processes that we can describe as if they were intentional, and there are neurobiological processes capable of producing states in the conscious form; but to the extent that the mental state is not even the kind of thing that could become the content of a conscious state, it is not a genuine mental state. We have been discussing these cases as if the neurobiology were intentional, as if it were mental, as if it were following rules; but that is not the case. The thesis I am putting forward is that we understand an unconscious mental state only as a state that, though not conscious then and there, is capable of becoming conscious; and when we attribute such a state to an agent, we are describing a brain mechanism, not in terms of its neural biological properties, but in terms of its capacity to cause conscious states and behavior. I call this view the “Connection Principle,” because it claims that our notion of the unconscious is logically connected to the notion of consciousness. An unconscious mental state must be the kind of thing that could be a conscious mental state.2
What is the argument for this apparently startling conclusion? We saw in our explanation of intentionality (chapter 6) that all intentional phenomena have aspectual shapes. But in the case of the deep unconscious there is no aspectual shape. There is no form of the intentional states that determines one intentional content rather than another. The argument that I am making here is that we should assimilate the third type of unconscious, the deep unconscious, to the fourth kind, the nonconscious, because the deep unconscious cases do not have the essential feature of intentionalistic phenomena, the aspectual shape of the intentional state that enables it to function in mental causation and therefore to justify the mentalistic forms of causal explanations. There are no deep unconscious mental states. There are, rather, neurobiological features that behave as if they had intentionality.
What is wrong with just saying of the processes in the brain that they are unconscious intentional states occurring right then and there as unconscious intentional states? Why do we have to go through this elaborate dispositional analysis where we say that the attribution of unconscious intentionality is like describing something as poison or bleach? The answer is that the neurobiology as such has no aspectual shape. We can see this if we consider examples. Imagine a man who wants to drink water. Now he may have a desire for water but not a desire for H2O, simply because he does not know that water is H2O. But the external behavior will be exactly the same in the two cases: the case of desiring water and the case of desiring H2O. In each case he will seek to drink the same sort of stuff. But the two desires are different. How is this difference to be captured at the level of the neurophysiology? The neurophysiology, described in terms of synaptic strength and action potentials, knows nothing of aspectual shape. Yet we do want to be able to say that the man who has an unconscious desire for water is in a different intentional state from the man who has an unconscious desire for H2O, even though the manifestation of this desire in the form of behavior would be exactly the same in the two cases. The answer that I am proposing, and indeed the only answer that I can think of that would make any sense at all, is that we are describing the neurobiological structure in terms of its capacity to cause conscious thoughts and conscious behavior. For the person who does not know that water is H2O, the neurobiology that corresponds to the desire “I want water” is different from the neurobiology that corresponds to the desire “I want H2O.” But all the same, at the level of the neurobiology, these different aspectual shapes do not exist as aspectual shapes, but, for example, as differences in neuronal structure. So we can give a legitimate sense to the notion of the unconscious, provided we describe it in terms of the causal capacities of the brain to cause consciousness.
But this has an interesting consequence. It means that we have no notion of the unconscious except in terms of the conscious. Something that is not even the sort of thing that could be brought to consciousness cannot be an intentional state because it cannot have aspectual shape. For that reason there are no deep unconscious mental states. There are neurobiological structures capable of causing conscious states and capable of causing behavior appropriate to those mental states, and these cover both the preconscious and repressed unconscious states, and there are neurobiological structures capable of causing behavior that is as if it were intentionally motivated, but where the sort of motivation could not be a conscious intentional content and therefore has no psychological reality.
I have given a dispositional analysis of unconscious mental states. An unconscious mental state, when unconscious, consists in a capacity of the brain to produce that state in a conscious form and to produce behavior appropriate to that state. But this result has an unexpected consequence for our earlier analysis of intentionality. I made a distinction between the network of intentional states and the background of capacities that enables these states to function. But what are the elements of the network when they are unconscious? What, for example, is the status of my belief that George Washington was the first president when I am sound asleep? On the dispositional analysis I just gave, it consists of a brain capacity. But then, the background also consists of such capacities. So it turns out that the network of intentionality, when unconscious, is a subclass of background capacities; it is the special capacity to produce certain forms of conscious thoughts and behavior.
The topic of the unconscious differs from most of the other topics we have discussed in this book in that it is not immediately experienced but rather is something we have found it necessary to postulate for some other purpose. Why is it so important to us? Why does it matter to us that we give an account of the unconscious, when the unconscious is by definition not even experienced?
The answer is that the unconscious has come to figure hugely in our explanation of human behavior. It is because we want to explain our behavior that we postulate the unconscious at all. I have heard philosophers claim that the reason we say that people have beliefs and desires is so that we can explain their behavior. Frankly, I think that is about as unintelligent as saying the reason we say that people have feet is so that we can explain their walking behavior. No, the reason we say they have feet is because they have feet, and the reason we say they have beliefs and desires is because they have beliefs and desires. But the postulation of the unconscious really is part of an explanatory need. The reason we say that people have unconscious motivations is that we have found no other way to explain some forms of their behavior. The postulation of unconscious mental states, unlike the “postulation” of feet, or beliefs and desires, really is done for an extraneous purpose: the explanation of human behavior. That is why we had a special problem about the ontology of the unconscious and that is why it is worth going through the effort to get an account of the unconscious that is consistent with our overall conception of the physical world and the role of the mental in that world.
But if we require the notion of unconscious mental states in order to explain human behavior, then we need a prior conception of human behavior and its explanation before we can know how to apply the concept of the unconscious. I have, in a preliminary fashion at least, given an account of the structure of human action in chapter 6, Intentionality. That chapter has certain implications for the explanation of human actions, and I now want to spell some of those out.
The key notion for the explanation of a human action is the notion of a reason. We saw in our discussion of mental causation that the content of the explanation has to match the content in the mind of the agent whose behavior is being explained. This is a point of stunning importance for such disciplines as history and the social sciences. It is disguised from us by the enormous complexity of actual explanations. So we say, for example, that the rise in American interest rates caused a rise in the value of the dollar. And on the surface that looks very simple, like saying that the rise in temperature caused a rise in pressure. But in fact the explanation in terms of interest rates is immensely complicated. To spell it all out, we would have to explain how the perception of higher interest rates in the United States led investors to desire to invest in American securities so that they could make a higher level of profit because of the higher interest rates and how that desire in turn led to a desire to buy more dollars with which to make these investments. So when I say that the intentional content in the explanation has to match the intentional content in the minds of the agents whose behavior is being explained, I do not mean there is any simple one to one matching in the actual surface of the explanation.
What then is a reason for an action? That looks like a very simple question but the answer to it is immensely complicated, and to spell it out in detail would take us beyond the scope of this book. I have in fact written a book about it (Rationality in Action, MIT Press, 2000), so you can look up the details. Let me just say the following: if you ask yourself how you explain your own behavior, for example, why did you vote for the candidate you did in fact vote for in the last election, you will find that your answers fall into two categories. Either you will give some sort of motivation, for example, “I wanted lower taxes,” or you will give some fact that you believed is related to the motivation, for example, “I believed the Republicans would lower taxes.” Taken together, this complex forms what I call a “total reason.” Reasons are always propositional in form and something is a reason only if it is part of a total reason. The key point for the discussion of the unconscious is this. There are some forms of human behavior that make sense only if we postulate a reason for action of which the agent himself is unconscious.
A special subcategory of reasons for action are rules governing human behavior, and a special form of intentional causation occurs in rule-governed behavior. The agent does what he does at least in part because he is following a rule. But what does it mean to follow a rule?
The explanatory power of the postulation of unconscious mental processes largely depends on the assumption that these processes are cases of unconscious rule following. The idea is that our intelligent behavior is explained by a lot of unconscious mental processes that consist in our following rules of which we are not aware and could not become aware. But if we are going to understand the notion of unconscious rule following then we have to understand the notion of rule following in the first place; and that would seem to require that we understand conscious rule following. What is it exactly that one does when one performs an action by way of following a rule? The answer to that question is by no means obvious. In order to explore it, we will have to specify some of the features of rule following. The first distinction we need to make, and it is crucial for everything that follows, is that between rule-governed behavior and rule-described behavior. Rule-governed, or rule-guided, behavior is such that the agent who is following the rule is causally influenced in his behavior by the rule. The rule functions causally in producing the very behavior that constitutes following it. So, for example, if I follow the rule “Drive on the right-hand side of the road” then the content of that rule must function causally in producing my behavior. This is not to say that the behavior is entirely determined by the rule. No one goes out driving just for the sake of following that rule, but all the same the content of the rule must function causally or it is not the case that one is following the rule. In this respect rule-following behavior differs from rule-described behavior. So, the ball rolling down the inclined plane can be described by the rules of Newtonian mechanics, but it does not follow that the ball is in any sense following those rules. The behavior of the ball rolling down the inclined plane is rule described but is not rule following.
What, then, are some of the features of rule-following behavior? Let us list them.
1. As we just stated, the content of the rule must function causally in producing the behavior.
2. Because of feature 1, rules have the logical properties that are common to volitional intentional states and directive speech acts. This is why the analogy is often made between following a rule and obeying an order. Specifically, the conditions of satisfaction of the rule have the world-to-rule direction of fit. The behavior must change so as to match the content of the rule. The rule also has the causal self-referentiality that we saw earlier was characteristic of prior intentions and intentions-in-action. The rule is followed only if the rule itself causes the behavior that constitutes following it.
3. It follows from 1 and 2 that every rule must have an intentional content that determines a certain aspectual shape. So you might have extensionally equivalent rules that were not at all equivalent in the conditions under which they are followed. The rule, for example, “Drive on the right-hand side of the road” in my car would give the same result as “Drive in such a way that the steering wheel is near the center line of the road, and the passenger seat is near the curb.” Given the structure of American cars this rule will produce exactly the same result as the initial rule, but the two rules, though extensionally equivalent, are not the same rule because they have different aspectual shapes.
4. Rule following is typically voluntary. In order that the rule should be able to guide behavior it has to be the kind of thing that the agent can follow voluntarily. The gap, in short, is present in rule-governed behavior. This is why, for example, the “rules” according to which I digest carbohydrates are not cases of rule following, but cases of rule-described behavior. This is because it is not up to me. In short, it is a feature of rule following that the rule can either be followed or broken. But where the rule cannot be broken, it cannot be followed either.
5. Rules, like any other intentional contents, are always subject to different interpretations. It is always possible to offer another interpretation of the rule. So, for example, most rules of human behavior are what are sometimes called “other-things-being-equal” or ceteris paribus rules. And this is because the rule is subject to interpretations. So, for example, I do indeed follow the rule “Drive on the right-hand side of the road,” but when I follow this rule I do not simply stop when confronted with an obstruction blocking the right hand side of the road; I swerve around it onto the left side of the road. I interpret the rule in such a way as to allow me to do things that are not specified in the content of the rule.
This feature of rule following, that it is always subject to different interpretations, has led to a certain form of skepticism. On one interpretation of Wittgenstein’s famous private-language argument, Wittgenstein is arguing that any behavior at all can be made consistent with a rule so long as we allow ourselves liberty of interpreting the rule.3 And his answer to that, according to some interpretations, is to say that our following of the rule is a social practice and that society makes it possible to achieve agreement about what constitutes following the rule. For this reason, Wittgenstein is supposed to have shown that a “private language” would be impossible because there would be no public check on the interpretations of the rule.
6. Human conscious rule following goes on in real time. When I follow the rule, “Drive on the right-hand side of the road,” the rule functions causally in my real psychological time to determine conditions of satisfaction. As far as this ordinary sense of rule following is concerned, it is impossible that there should be, for example, thousands of computational rules that I follow more or less instantaneously in a way that a commercial digital computer does. Rule following takes a certain amount of time and it goes on in real time.
These are the paradigmatic features of conscious rule following. But when we postulate unconscious rule following (and such postulations are all too common), how many of these features can we keep? If we are talking literally about rule following, these are the features we need to preserve. If talk about unconscious rule following is to be taken literally, then such rule following has to have these features: the rule functions causally with the world-to-rule direction of fit and at the rule-to-world direction of causation. The rules have to have an aspectual shape, be followed voluntarily, be followed in a way that is subject to different interpretations; and they have to be followed in real time. Some postulations of unconscious rule following, such as rule following in the performance of speech acts, meet these conditions. But many postulations of unconscious rule following, as in the cognitive science accounts of visual perception and language acquisition, do not meet these conditions.
The conclusion of this chapter is somewhat depressing. The notion of the unconscious is one of the most confused and ill-thought-out conceptions of modern intellectual life. Yet it seems we cannot get on without it. What we need to do, then, is to try to develop a coherent notion of the unconscious, which we can fit into what we know about the rest of reality, including what we know about how the brain works. The result is the Connection Principle. Most of the people who work in these fields object to my account of the Connection Principle, but I have not seen them present any alternative coherent conception of the unconscious. The upshot is that we can continue to use the notion of the unconscious legitimately, but we have to recognize that we are using it as a dispositional notion. To say of an agent that he has such-and-such an unconscious intentional state, and that that state is functioning actively in causing his behavior, is to say that he has a brain state that is capable of causing that state in a conscious form, even though in a particular instance it may be incapable of causing it in a conscious form because of brain damage, repression, etc. I am not entirely satisfied with this conclusion, but I cannot think of an alternative conclusion that is superior to it.