The New Science of the Mind

by Mark Rowlands

  Functionalism takes a similar view of the nature of mental properties. That is, such properties are defined by what they do-by their functional role. What is it that mental phenomena do? Fundamentally, they relate to each other, to perception, and to behavior in various complex, but in principle analyzable, ways. Take a belief, for example, the belief that it is raining. This is a belief that is typically caused by perception of certain environmental conditions, rain being the most obvious. Of course, perception of other environmental conditions might also produce the belief; for example, someone, unknown to you, using a hose outside your window. But rain is the most typical cause of the belief. The belief can, in turn, go on to produce certain sorts of behavior. Because of your belief you might, for example, carry an umbrella with you when you leave the house. However, the belief has these sorts of ramifications for your behavior not in isolation but only in combination with other mental states. You will carry the umbrella because you believe it is raining, but only if you also want to stay dry. Your wanting-your desire-to stay dry is necessary for your behavior too. And this belief-desire combination will produce your behavior only if you also believe that the umbrella will keep you dry; only if you believe that it is not too windy to use an umbrella; indeed, only if you believe that what you have picked up is an umbrella; and so on. What emerges is a complex network of mental states, perception, and behavior. According to functionalism, each mental state is defined by its place in this network: by the relations in which it stands to perception, to other mental states, and to behavior. To specify the place of a mental state in this network is, according to functionalism, to define that state. Of course, any such definition would be grotesquely long. Indeed, if you spent your whole life attempting to give a functional definition of even one mental state, you might well not have time to finish the task. But, its practicalities aside, the strength of functionalism consists in giving us a general vision of what mental phenomena are. The vision is of mental phenomena forming a vast causal system-a system of interrelated causal connections-where each mental property is individuated by way of its place in this system: by way of its causal connections to other mental states, to perceptual stimuli, and to behavioral responses. (See also box 1.1.)

  The most important assumption underlying the thesis of the extended mind seems to be functionalism about mental states and processes (Clark 2008a,b). Indeed, the functionalism presupposed by the thesis of the extended mind is of a peculiarly liberal sort. As we saw earlier, Clark and Chalmers's use of Otto in developing their case for the extended mind makes clear their functionalist commitments. What is decisive in determining whether or not the entry in Otto's notebook qualifies as a belief is the way it interacts with perception, behavior, and other mental states. Thus, when combined with his desire to see the exhibition, the entry in his notebook causes Otto to set off in the direction of 53rd Street. In general, Clark and Chalmers argue that the entries in Otto's notebook have a functional role in Otto's psychology that is sufficiently similar to the functional role of belief in Inga's psychology that the notebook entries should be counted as among Otto's beliefs. Though there are differences between the entries in Otto's notebook and more standard cases of belief, these differences are shallow ones-insufficiently significant to disqualify the entries from counting as beliefs.

  This commitment to functionalism, however, is not peculiar to Clark and Chalmers's development of the extended mind. My case for the extended mind developed in The Body in Mind was based on the idea that cognitive processes could be defined in terms of what they do and the way they do it. Roughly, the idea was that cognitive processes are ones whose function is to enable organisms to accomplish cognitive tasks (e.g., perceiving the world, remembering perceived information, reasoning on the base of remembered information), and they do this by way of the manipulation, transformation, and exploitation of information-bearing structures. As a matter of contingent fact, I argued, some of the information-bearing structures involved were ones external to the body of the cognizing organism, and therefore, the manipulative processes involved were similarly external. So, to this extent, my case for the extended mind was also a functionalist one.

  Indeed, not only are most arguments for the extended mind functionalist ones, there is a way of understanding functionalism according to which the thesis of the extended mind emerges as a straightforward, almost trivial, consequence.' Functionalism, it is widely supposed, is based on a principled indifference to the details of the physical structures that realize mental processes. What is crucial to a mental state or process is its functional role, not its physical realization. This indifference to physical realization is principled in this sense: the structure or mechanism in question has to be capable of realizing the functional role that defines a mental state or process; but this is all that is required of it. As long as it is capable of doing that, nothing else about the mechanism matters. In other words, the structure or mechanism is only indirectly relevant; it is relevant insofar as it realizes or underwrites a functional role, and only insofar as it does this. For the functionalist, if it walks like a duck, and talks like a duck, then it is a duck. How it manages to walk and talk like a duck is not directly relevant.

  The thesis of the extended mind is, in effect, based on a further consequence of functionalism-less familiar but ultimately no less obvious. For the functionalist, not only are the details of the physical structures and mechanisms only indirectly relevant, so too is their location. For functionalism, the only thing that is directly relevant to whether or not something qualifies as a mental state or process of a certain sort is whether it plays the required functional role. It doesn't matter what mechanism realizes or accomplishes this role-as long as it does so. And, crucially, it doesn't matter where this mechanism is located when it realizes or accomplishes this role-as long as it does, in fact, realize this role. According to func tionalism, when consistently understood, the location of the mechanisms that realize functional roles is no more relevant than the physical details of those mechanisms. Both are only indirectly relevant. What is always crucial, according to functionalism, is whether or not a given functional role is realized, not how and not where this role is realized. How something manages to walk and talk like a duck does not matter for the functionalist; neither does it matter where it walks and talks like a duck.

  This understanding of functionalism is sometimes referred to as liberal, where this is opposed to chauvinistic. There are many different explanatory levels at which functional description of a process might be useful. Therefore, for some versions of functionalism, it does, in fact, matter how something might walk and talk like a duck. This more chauvinistic conception of functionalism is presupposed by Rupert (2004), in an influential critique of the thesis of the extended mind. This commitment is clear in the way he prosecutes his differences argument. Rupert argues that there are significant differences between internal cognitive processes and the sorts of environmental processes invoked and claimed to be cognitive by the extended mind. These differences are significant enough to preclude the latter being regarded as part of the same psychological kind as the former.

  For example, in attacking the extended conception of memory I developed in The Body in Mind, Rupert (2004) argues that there are significant differences between the fine-grained profile of internal memory operations and external memory stores of the sort invoked in the case of Otto. For example, internal (neural) memory operations seem subject to what is known as the generation effect. The ability of a subject to remember the second term of a paired associate upon being presented with the first is augmented if the subject is allowed to generate meaningful connections between each associate. Rupert argues that this generation effect will fail to occur in at least some extended memory systems. For example, it will fail to occur in a notebook-based system where there are connection sentences between paired associates but where, crucially, these sentences have been entered into the notebook by the experimenter rather than the subject. Rupert doe
s accept that the effect might occur if the subject both creates the meaningful association and records this in the notebook-and this, of course, is far closer to Clark and Chalmers's case of Otto. However, he argues that this is simply a contingent feature of the system. Such differences, Rupert claims, undermine any attempt to regard internal and extended memory systems as forming part of a single explanatory kind.

  I think there are several points that can be raised against Rupert's argument. Most obviously, suppose someone's internal, biological memory failed, for whatever reason, to exhibit the generation effect. Every other aspect of his memory was functioning normally. If you ask him facts about the world, he is, if he was aware of those facts, able to reply correctly. If you ask him to describe events from his childhood, he is able to do so. In short, his memory is what we would regard as a perfectly normal one, except for one thing: it fails to exhibit the generation effect. Would we really want to say, in such a case, that he does not really remember? This would be implausible. The generation effect is a peripheral feature of human remembering whose absence, by itself, is nowhere near decisive enough to disqualify a process from counting as one of remembering (Wheeler 2008).

  However, my purpose here is not to evaluate Rupert's argument but to identify its guiding presupposition. And this seems to be a more chauvinistic form of functionalism than the thesis of the extended mind can afford to accept. Rupert's argument is based on the idea that a coarse-grained functional profile is not, by itself, decisive in determining the identity conditions of a psychological kind. Fine-grained functional details, of which the generation effect would be one example, are also crucial determinants of these identify conditions.

  The latter claim, however, is not one that the thesis of the extended mind can accept. That thesis presupposes a liberal form of functionalism, according to which gross functional role is crucial. Rupert's objections presuppose a far more chauvinistic conception of functionalism, according to which fine-grained functional details are also crucial. To this extent, Rupert's objections are question-begging (Wheeler 2008). Matters would be different if we possessed some independent reason for preferring the chauvinistic version of functionalism presupposed by Rupert. Partly in anticipation of this sort of response, Rupert argues that any psychological kind general enough to subsume both internal and extended memory would be so lacking in detail as to be explanatorily useless. However, this seems merely to beg further questions. If it would, in fact, be implausible to deny that a person who exhibited every feature normally associated with memory-recall except the generation effect was able to remember, then the generation effect does seem to be an accidental rather than a defining feature of memory-which is, of course, precisely what the liberal conception of functionalism entails. And if this is correct, we have no reason for withdrawing the attribution of remembering to the extended system.

  The thesis of the extended mind, therefore, is predicated on a liberal conception of functionalism that sees gross functional profile as decisive in determining the type-identity conditions of psychological kinds. Any objection to the thesis that presupposes a more chauvinistic form of functionalism is, therefore, question-begging. However, charges of questionbegging can cut both ways. It is one thing to argue that, in the case of memory, the generation effect is a peripheral aspect of memory that is not decisive in deciding whether or not a process counts as one of remembering; it is quite another to argue that there are general reasons for preferring liberal over chauvinistic forms of functionalism. This is so for two reasons. First, perhaps there are other aspects of memory-different from the generation effect-that are more important in qualifying a process as one of remembering; these processes might not be captured within the broadbrush way of identifying functional kinds adopted by liberal approaches. Second, perhaps a liberal conception of functionalism is simply more appropriate to some kinds of cognitive process, whereas a chauvinistic form is more appropriate to others. If, so the dispute between liberal and chauvinistic forms of functionalism would have to be adjudicated on a case-by-case basis. The question, then, is: how do we so adjudicate? On what basis do we decide between the claims of liberal and chauvinistic functionalism? On what criteria do we draw to make such adjudications?

  There is, however, a more pressing problem. If the extended mind does indeed rest on a form of liberal functionalism, then, in the eyes of many, this makes it perilously close to being incompatible with the embodied mind. For it is commonly thought, at least by its protagonists, that the latter requires rejection of any liberal form of functionalism (Clark 2008a). Thus, in the previous chapter, we saw how Shapiro understands his embodied mind thesis as being opposed to what he calls ST, the separability thesis-the thesis that minds make no essential demands on bodies and that therefore a humanlike mind could very well exist in a nonhumanlike body. Shapiro understands ST as, in essence, an expression of a liberal form of functionalism. As we have seen, for liberal functionalism, details of the physical implementation of a cognitive process are of only derivative importance: what is crucial is whether or not the relevant functional role is realized. Therefore, if a nonhumanlike body were to realize the functional roles definitive of humanlike cognitive processes, that, and that alone, would be decisive in determining whether the nonhumanlike body possessed a humanlike mind.

  However, although Shapiro's embodied account is incompatible with liberal functionalism it is perfectly compatible with a more chauvinistic form. To see why, recall Shapiro's analogy of the submarine instruction manual. What this analogy in fact undermines is not a functionalist account of cognitive processes itself, but a particular-and peculiarly liberal-conception of what this account must look like. A more sophisticated form of functionalism, it might be thought, might accept that functional roles cannot be understood in isolation from the bodily structures that they include and depend on. The relevant submarine instruction manual would, then, look something like this. It contains a series of instructions on how to pilot the submarine, but these make noneliminable reference to features of that submarine. So, an instruction for increasing depth might look something like this: "Pull out the red-colored knob in the column on the far left. Keep this knob out until the blue bulb in the second column from the left starts flashing. Then push it back in." You can neither fully understand nor successfully employ this manual unless you are already in the process of piloting the submarine, or at least physically placed to do so. But, nevertheless, the instruction manual can be regarded as a functionalist program of a certain sort. The manual-the program-remains, and does so independently of the existential fortunes of the submarine; it is just that without the submarine, the program can be neither employed nor properly understood.

  The form of functionalism described here is compatible with Shapiro's body centrism, but is not sufficiently abstract to be compatible with the standard arguments for the extended mind. In a nutshell, Shapiro's body centrism can be rendered compatible with a form of functionalism, but only if this form is sufficiently chauvinistic to include explicit reference to bodily structures and mechanisms. And, as we have seen, the thesis of the extended mind presupposes a liberal form of functionalism. Without this, it is vulnerable to the sorts of objections raised by Rupert (2004). Of course, if there could be a more general manual, sufficiently abstract to subsume instructions for submarines and other forms of vehicle, then the rapprochement of embodied mind with extended mind would be back on the table. However, the thesis of the embodied mind is, in effect, that there is no such general-purpose manual.

  9 Desiderata for Further Development of the Amalgamated Mind

  This chapter has identified two clear imperatives for the further development of the amalgamated mind. First of all, there is the imperative that centers on the issue of functionalism. This is crucial for the amalgamated mind thesis in two different but related ways. First of all, the extended mind's reliance on a liberal form of functionalism can leave it vulnerable to charges of question-begging. Second, and more worryingly, th
e reliance of the thesis on liberal functionalism seems to make it incompatible with the thesis of the embodied mind, a thesis that can accept functionalism only in its more chauvinistic forms. The first imperative facing the amalgamated mind, therefore, is to resolve these problems.

  In the chapters to come, I propose to satisfy this imperative by, insofar as this is possible, taking functionalism out of the equation. Thus, in chapters 7 and 8, I am going to develop an argument for the theses of the embodied mind and the extended mind-and hence for the amalgamated mind-that does not rely on functionalism. This is not to say, of course, that the argument I shall develop is necessarily incompatible with functionalism. However, it does not rely on functionalist assumptions, whether liberal or chauvinistic. Indeed, it provides us with the means to adjudicate, in particular cases, the competing claims of liberal and chauvinistic forms of functionalism. Moreover, since the argument is one that supports both the thesis of the embodied mind and the extended mind, and does so for precisely the same reasons, the issue of the compatibility of the two theses is resolved.