The New Science of the Mind

by Mark Rowlands

  If the arguments developed in the preceding sections are correct, the hybrid processes invoked by the amalgamated mind can satisfy at least the first three conditions of the criterion of the cognitive. To this extent, they qualify as cognitive processes-pending satisfaction of condition (4). However, I have also argued that the other objections to these theses collapse into the mark of the cognitive objection. Thus, there is no reason for thinking that the wider bodily and environmental processes provide merely an external causal scaffolding or milieu within which the real, internal processes of cognition can do their work. The hybrid amalgamations of processes together satisfy conditions (1)-(3) of the mark of the cognitive and, therefore, pending satisfaction of condition (4), qualify as cognitive in their own right.

  At this point, any further objections to the thesis of the extended mind tend to coalesce around the role of representation. Condition (3) claims that the information made available to a cognitive subject is done so by way of the production in that subject of a representational state. This representational state, I have assumed, must possess nonderived content. This renders the criterion-and the thesis of the amalgamated mind predicated upon it-immune to some further objections associated with Adams and Aizawa (2001) and Fodor (2009).

  A perceived reliance on derived representations forms the basis of one of one of Fodor's objections to Clark and Chalmers's version of the extended mind:

  [I]f something literally and unmetaphorically has content, then either it is mental (part of the mind) or the content is "derived" from something that is mental. "Underived" content (to borrow John Searle's term) is the mark of the mental; underived content is what minds and only minds have.

  Items external to the brain, Fodor seems to think, do not possess nonderived intentionality-their content is determined by convention rather than nature. Thus, the sentences in Otto's notebook are not mental items because, while being about something (e.g., the location of the Museum of Modern Art), this aboutness or intentionality is merely derived from the (inner) mental states of Otto, or anyone else who uses the book.

  Here, Fodor is reiterating a point made by Adams and Aizawa: "If you have a process that involves no intrinsic content, then the [nonderived content] condition rules that the process is noncognitive" (2010, 70). This is a version of their mark of the cognitive objection. As we have seen, Clark and Chalmers have been widely interpreted as arguing that the sentences in Otto's notebook constitute a subset of his beliefsand these sentences are, of course, examples of external visuographic storage structures that possess only derived content in the sense introduced earlier. For reasons discussed earlier, I reject Clark and Chalmers's claims that the sentences in Otto's book are to be numbered among his beliefs. However, my own development of an extended account of remembering also accords a central role to external visuographic information-bearing structures-of which language is the most familiar and important-in the constitution of memory, and so I might also be thought vulnerable to Adams and Aizawa's objection. However, in fact, when the thesis of the extended mind is properly understood, neither I nor Clark and Chalmers are vulnerable to this objection. So, in the argument to follow, I shall assume that Clark and Chalmers are correct in claiming that the entries in the notebook are among Otto's beliefs. This is an assumption because both Fodor's and Adams and Aizawa's objections are framed in terms of this claim; I make the assumption purely for the sake of the argument to follow and, in particular, to show why these objections do not work.

  According to Adams and Aizawa, if a process is to count as cognitive it must involve states that possess nonderived content (2001, 2010). I agree, or at least have made this assumption in advancing my criterion of the cognitive. However, what is puzzling is that they think this provides an objection to the thesis of the extended (or, for that matter, the amalgamated) mind. As I have been at pains to emphasize, the thesis of the extended mind does not claim that extracranial processes-processes of environmental manipulation, exploitation, and transformation, for example-can, by themselves, be cognitive. It is not that a cognitive process might ever consist in processes entirely and exclusively outside the skin, or even outside the skull, of a cognizing organism. On the contrary, the thesis of the extended mind claims that cognitive processes are either purely internal or are hybrid processes straddling both internal and external components. They are never purely external. The thesis of the extended mind is a claim about the character of the external or extended processes when, and only when, they are appropriately combined with the relevant sorts of internal processes: the claim is that these external or extended processes are genuinely cognitive components of the overall cognitive process-a process that straddles both internal and external elements. The external processes are not merely noncognitive accompaniments that facilitate the "real" process of cognition that occurs inside the head or in which the "real" process of cognition is causally embedded.

  Therefore, the thesis of the extended mind is not only compatible with but actually insists on the claim that the external processes involved in cognition are dependent on internal cognitive processes for their status as cognitive: the former could not count as cognitive without the latter. However, the extended mind thesis is then defined by the claim that once these external processes are appropriately combined with internal cognitive process of the requisite sort, the label "cognitive" can be as legitimately applied to the overall process-the hybrid combination of internal and external processes-as it can to the internal process itself. And this book cashes out this claim in terms of the idea that both the overall process and its external component satisfy the mark or criterion of the cognitive.

  Adams and Aizawa object to the thesis of the extended mind on the grounds that Otto's notebook entries possess merely derived content. But, when the thesis of the extended mind is properly understood, this is irrelevant. According to this thesis, the external component of the overall hybrid process would not on its own count as cognitive. And the overall process certainly does contain components that possess nonderived content. Thus, although the sentences in Otto's book do not-we might accept-possess nonderived content, Otto's perceptions of these sentences certainly do (Menary 2006, 2007). So too does his belief that sentence s communicates content c. That is: the sentences in Otto's book may not possess nonderived content, but his perceptual and doxastic apprehensions of them certainly do. This point can also be explained in terms of our criterion of the cognitive. According to condition (3) of the criterion, a cognitive process must make information available to an organism or subsequent processing operations by way of the production, in the cognizing organism, of a representational state. When Otto forms a perceptual or doxastic representation of the entry in his notebook, this condition is satisfied, and it is satisfied irrespective of the status of the notebook entries themselves. Whether they are representations that possess only derived content is completely irrelevant to the thesis of the extended mind.

  Otto's perceptual representation-an internal, nonderived, state of Otto-is part of the overall hybrid cognitive process: the overall process in the context of which, and only in the context of which, Otto's notebook entries can qualify as beliefs (if we believe the widespread interpretation of Clark and Chalmers's version of the extended mind). However, since this hybrid process clearly does involve states with nonderived content, Adams and Aizawa's requirement that cognition involve nonderived content would seem to be satisfied. Accordingly, it is puzzling why they think they have provided an objection to the thesis of the extended mind.

  Matters are even clearer on my process-oriented version of the extended mind that eschews the identification of cognitive states and notebook entries. The process of manipulating an external information-bearing structure does not count as cognitive on its own, but only when it is suitably combined with a relevant internal (i.e., neural) process. However, once it is combined in this way, the external process of manipulation satisfies the mark of the cognitive and hence counts as a genuinely cognitive
part of the overall (amalgamated) cognitive process. Since any cognitive process always contains a noneliminable internal component, it will also, for that reason, be made up of states that possess nonderived content. No case has been made for the extended mind to answer.

  Of course, matters would be substantially different if Adams and Aizawa could show that every part of a cognitive process must involve a state that possesses nonderived content-or, conversely, that the possession of derived content by a state automatically precluded that state from qualifying as cognitive. However, this is an extremely implausible claim, and one that Adams and Aizawa explicitly, and correctly, deny:

  Although we have good reasons to believe in the existence of intrinsic content, we have no good reasons to think that cognitive states must consist entirely of intrinsic representations or that cognitive states must be, in their entirety, content bearing. This is why we said that "it is unclear to what extent each cognitive state of each cognitive process must involve non-derived content." (Adams and Aizawa 2010, 69)

  They are, of course, entirely correct in this denial. Most cognitive processing does not involve states with nonderived representational content, for the simple reason that most cognitive processing does not involve representational states at all. Rather, with respect to the cognitive operations of a subject, we have compelling reasons to accept a general picture of small islands of representational states in a large sea of nonrepresentational processing. This is equally true whether you are an externalist or internalist.

  Thus, to return to our paradigmatically internalist model of visual perception, the transformational operations postulated in Marr's theory of vision do not represent anything. The intermediate states that these operations produce might be regarded as representational; but the processes that produce them are not. For example, the application of principles such as common fate and good continuation (among others) to the raw primal sketch produces the full primal sketch. However, what does the application, to the raw primal sketch, of common fate or of good continuation represent? What content does it possess? The answers are, respectively: nothing and none. The raw primal sketch possesses content (of a sort), and when the operative principles are applied to it, the full primal sketch can possess content-again of a sort. But operations according to which the one is transformed into the other do not represent anything at all. (This fact is often obscured by the misguided assimilation of cognitive transformations to inference rules. They may be describable in terms of inference rules, but they are not themselves forms of inference.) Cognitive states are representational; but the processes whereby one cognitive state is transformed into another are not. Nonetheless, it would be implausible to deny them the status of cognitive: if Marr's transformational operations do not count as cognitive, it is difficult to see what does. And their status as cognitive is, of course, preserved in the criterion of the cognitive proposed here. The processes are cognitive because, when they are fulfilling their proper function, and when they are combined with other processes of the requisite sort, they are capable of yielding representational states, thus making information available to the subject or subsequent processing operations that was previously unavailable. If our criterion of the cognitive is correct, a cognitive process must be the sort of thing that, in isolation or in combination with other processes, can be normally capable of yielding a state with nonderived content. But it is simply false to claim that anything that is to count as cognitive must possess nonderived content.

  Given Adams and Aizawa's apparent acquiescence on this point, therefore, it is mystifying why they think they have made any case for the thesis of the extended mind to answer. The sorts of hybrid processes posited by the thesis clearly always involve states that possess nonderived content. The criterion of cognition defended here, in effect, elevates this into a necessity. According to the criterion, the culmination of cognitive processing is the making available of information (to an organism or subsequent processing operations) by way of the production, in the cognizing subject, of a representational state. In the case of Otto, this requirement would be satisfied by way of his perceptual or doxastic representations of the contents of his notebook. When the thesis of the extended mind is properly understood-as entailing the possibility of hybrid, or amalgamated, rather than purely external cognition-then the nonderived content condition is one that it clearly meets. Therefore, the Adams and Aizawa objection has no substance.

  There is, perhaps, one option remaining for Adams and Aizawa. Consider the sorts of hybrid processes postulated by the thesis of the extended mind-processes that straddle both internal and external components. The crucial difference between components, they might argue, is that the external component never contains states with nonderived content. Nonderived content might always be involved in these hybrid processes, but it always attaches to the internal operations and never to the external ones. That is why the external processes are never more than merely accompaniments to cognition.

  There are two ways one might respond to this. One can challenge Adams and Aizawa's premise; or one can grant the premise and show that the conclusion still does not follow. I have, in effect, challenged the premise elsewhere (Rowlands 2006). Here, however, for the sake of argument, I propose to grant Adams and Aizawa's premise and argue that their desired conclusion still does not follow.

  Imagine an unlikely but certainly conceivable internalist proposal concerning the implementation of the functional roles that constitute cognition. This implementation is hybrid in a manner that parallels the extended mind's conception of (some) cognitive processes. That is, the neural processes involved turn out to be localized in two causally coupled but functionally and structurally isolable regions-brain region R (for representational) and region P (for processing). Region R provides the neural realizations of representational states. Region P supplies the neural basis of transformational operations performed on those states. The two regions are causally coupled in the sort of intimate manner required to underwrite their successful cooperation. Region R contains states with nonderived content; but region P does not, for the simple reason that it does not contain states with any sort of representational content.7

  If the details of the neural implementation turned out as described above, then it would be implausible to deny the transformational operations occurring in region P the status of cognitive, simply because this region contains no states with nonderived representational content. That is, would we want to claim that these transformational processes are merely extraneous causal accompaniments to the "real" process of cognition that occurred in the region R? That would be a strange and deeply implausible form of legislation. And it is, of course, a form of legislation precluded by the criterion of cognition defended here. According to this criterion, the processes are cognitive because, when they are fulfilling their proper function, and when they are combined with other processes of the requisite sort, they are capable of yielding representational states, thus making information available to the subject or subsequent processing operations that was previously unavailable. A cognitive process must be the sort of thing that can, in isolation or in combination with other processes, be normally capable of yielding a state with nonderived content. But it is simply false to claim that anything that is to count as cognitive must possess nonderived content. Some processes that are clearly cognitive never possess nonderived content-for the simple reason that they never possess content at all. It is the role in contributing to the production of states that possess nonderived content that is crucial in determining the cognitive status of a process: whether or not it actually is a state that possesses such content is irrelevant.

  8 Conclusion

  In this chapter, I identified a mark or criterion of the cognitive: a sufficient condition for a process to qualify as cognitive. I defended this criterion by showing that it is presupposed by, and makes sense of and systematizes, paradigmatically internalist models of cognition. I then argued that according to this criterion, hybrid
processes of the sort invoked by both the extended mind and the embodied mind qualify as cognitive. However, discussion so far is provisional and incomplete. I have argued only that the hybrid processes in question satisfy the first three conditions of the criterion; I now need to show that they satisfy the fourth.

  According to the fourth condition, anything that is to count as a cognitive process must belong to the subject of the representational state adverted to in condition (3). That is, the process must be owned by a cognitive subject. Accordingly, I shall call this the ownership condition. Whereas the first three conditions are relatively straightforward, the ownership condition is, I shall argue, deeply problematic. It is not, however, problematic in a way that would provide succor to the internalist opponent of the new science. On the contrary, it is problematic for both parties in equal measures. Satisfying the ownership condition will take us on an extensive exploration of the notion of consciousness, and the ways in which it can be both embodied and extended; that is the subject of the second half of this book. In the next chapter, however, I shall examine the ownership condition and explain why it is so problematic.

  1 Ownership and the Problem of Bloat

  The previous chapter presented and began the defense of a mark or criterion of the cognitive. The criterion is formulated in terms of processes: it provides conditions sufficient for a process to count as cognitive. A state can also qualify as cognitive, but only by way of the contribution it makes to the cognitive processes instantiated in an organism. Thus, the primary role for states in the criterion is as representational items. And the significance of this is that they are produced by cognitive processes and, as representational items, can function to make information available either to the organism or to subsequent processing operations.