According to condition (4) of the criterion, if a process is to qualify as cognitive it must belong to, or be owned by, a subject. There are no subjectless cognitive processes. For our purposes, the notion of a subject can, I think, be understood quite broadly. For example, I do not wish to rule out the possibility that the subject in question might be a group rather than an individual.' Nevertheless, even when understood liberally, the appeal to a subject might be thought problematic. In particular, it might be thought to make the criterion circular. After all, it is not any sort of subject that can own a cognitive process: the subject in question must be a cognitive subject. But a cognitive subject, it seems, is a subject of cognitive processes. Therefore, the criterion presupposes, rather than explains, an understanding of what it is for a process to be cognitive.' However, this circularity is only apparent. What this shows, in fact, is that the criterion is recursive rather than circular. All I mean by "cognitive subject" is any organism that satisfies conditions (1)-(4) of the criterion. More precisely, S is a cognitive subject if it owns (i.e., is engaged in) informationprocessing operations (i.e., the manipulation and/or exploitation of information-bearing structures), where these operations have the proper function of making available, either to the subject or to subsequent processing operations instantiated by the subject, information that was hitherto unavailable, and where this proper function is realized by way of the production in the subject of a representational state that possesses nonderived content. That is: a "subject" in the sense required by condition (4) is an individual that owns processes that satisfy conditions (1)-(3). This is all that I mean by the term "subject."
The real problem, I think, is not explaining what a cognitive subject is but rather explaining the sense in which such a subject can own (or instantiate) its cognitive processes; that is, the problem is one of explaining the sense in which a cognitive process can legitimately belong to a subject.' I am going to spend much of this chapter convincing you that this is a problem-a deep and difficult problem. More than that, I shall try to convince you that it is a problem for any account of cognitive processeswhether orthodox internalist or some other version of the amalgamated mind. The problem of ownership is an equal opportunity problem that does not discriminate between Cartesians and anti-Cartesians. Providing a solution to this problem is the task of the reminder of this book. A useful way to begin examination of the problem is by way of a problem that was, in the previous two chapters, largely sidestepped in a blur of promissory notes: the problem of cognitive bloat.
The problem of cognitive bloat is usually thought of as one that arises in connection with cognitive states rather than processes. Specifically, the charge of cognitive bloat typically arises in connection with Clark and Chalmers's claim that the entries in Otto's notebook, when appropriately combined with various neural processes undergone by Otto, constitute a subset of his beliefs. If we are willing to allow that the entries in Otto's book can qualify as beliefs, why stop there? Given that Otto makes regular and reliable use of the entries in his telephone directory, for example, why not say that these entries also number among his beliefs? Or given that Otto makes regular and reliable use of the Internet, why not claim that his beliefs include the contents of the World Wide Web?
Given that the version of the extended mind I am defending in this book is formulated in terms of processes, and given that the usual problem of bloat is formulated in terms of states, it might be thought that the version of the extended mind I am defending is immune to the problem of bloat. I would, after all, deny that the entries in Otto's notebook are beliefs. However, the problem is not so easily sidestepped. There is a version of the problem of bloat that applies to cognitive processes rather than states. In this opening section, I am going to identify this version of the problem.
To return to an earlier example, suppose I am using a telescope.' The telescope is, let us suppose, a reflector, and therefore works by transforming one mirror image into another. Mirror images are information-bearing structures-their properties are systematically determined by way of a mapping function that is itself determined by the specific properties of the mirror and by the properties of the visual environment. Therefore, the operation of the telescope is based on the transformation of informationbearing structures, and so it satisfies condition (1). The processes occurring inside the telescope are, in combination with other processes, of the sort normally capable of yielding a representational state. This is true even when the content of this state is nonderived. Thus, in combination with other processes-ones occurring inside me-the processes can yield a representational state-for example, my visual perception of Saturn's rings. This is a representational state with nonderived content. Therefore, the processes occurring inside the telescope satisfy condition (3). And the proper function of the processes occurring inside the telescope is making information available, both to me and to subsequent processing operations within me (for example, processes of inference), of information that was previously unavailable (for example, the current orientation relative to Earth of Saturn's rings), and thus satisfy condition (2). Therefore, the processes occurring inside the telescope satisfy conditions (1)-(3) of our criterion of cognition. If these conditions were sufficient for cognition, then the intratelescopic processes would have to be classified as cognitive.
Relevantly similar examples can be easily generated. How can we rule out, for example, processes occurring inside my calculator, or my computer, from counting as cognitive ones? They (1) seem to involve the transformation of information-bearing structures. They are also (3) the sort of process that, when combined with other processes, can produce representational states. Thus, when combined with operations occurring inside my brain, the processes occurring inside the calculator can produce a representational state in me-for example, when I read off the result of their operations from the screen. And (2) the proper function of these processes is, it seems, to make available to me information-information that I might subsequently employ in further processing operations-where this information was not available prior to the operations of the calculator or computer. Therefore, the processes occurring inside the calculator and computer seem to satisfy conditions (1)-(3), and, without further constraints, would therefore count as cognitive.
This is the problem of cognitive bloat, applied to cognitive processes rather than states. At root, the problem is a reflection of the idea that the concept of cognition comprises three separable aspects: why, how, and, crucially, who? The why and how aspects of cognition concern how cognitive processes do what they are supposed to do and why they do it. The why of cognition-its function, broadly understood-is to make previously unavailable information available, either to subsequent processing operations or to the organism itself. They do this-the how of cognition-by producing, perhaps in combination with other processes of the same general sort, representational states. And the production of such a state is the result of information-processing operations-the manipulation and transformation of information-bearing structures. The first three conditions of the criterion of the cognitive capture these why and how aspects of cognition. However, in addition to the why and the how aspects of cognition, there is also an irreducible who aspect. Whatever else is true of cognitive processes, whatever the specific details of their form and function, such processes always belong to someone or something. There are no subjectless cognitive processes: they always have an owner-an owner that is the subject of states and processes that satisfy conditions (1)-(3) of the criterion. Condition (4)-the ownership condition-attempts to capture this idea. And if we fail to adequately capture this idea, we are immediately faced with the problem of bloat as it applies to cognitive processes: the processes in my telescope and in my computer count as cognitive.
There are, it seems, two ways in which the thesis of the extended mind might try and deal with this problem. The first is to embrace its conclusion: the processes occurring in my telescope and my computer are, intuitions notwithstanding, cognitive. Th
e second is to show that the thesis of the extended mind can avoid this conclusion: the thesis is not committed to the claim that these processes are cognitive. There is, however, also a more subtle option. The two outlined responses are not, in fact, incompatible. They are not incompatible because the concept of a cognitive process covers two quite different kinds of process. On the one hand, there are personal-level cognitive processes. On the other, there are subpersonal cognitive processes. This distinction is captured in condition (2) of the criterion. This condition talks disjunctively of cognitive processes that have the function either of making information available to a subject or of making information available to subsequent processing operations, where in both cases such information would have been unavailable without the completion of the relevant cognitive processes. Processes that have the function of making information available to the subject of these processes are personal-level cognitive processes. This is true whether or not they also make information available to subsequent processing operations. Processes that have the function of making information available only to subsequent processing operations are subpersonal cognitive processes. The advertised subtle option, therefore, is based on the idea that we might be able to embrace the problem of bloat with respect to subpersonal-level processes, but avoid it with respect to personal-level processes. The possibility of personal-level bloat is far more objectionable than that of subpersonal bloat. Furthermore, embracing the relatively anodyne claim of subpersonal bloat, I shall argue, makes it possible to reject the far more objectionable personal-level version. This is the strategy that I shall pursue in this chapter.
In more detail, I am going to argue for four claims. First, the problem of ownership is just as much of a problem for the Cartesian internalist as it is for the defender of the extended mind. Focusing unduly on the problem of cognitive bloat does have the drawback of masking this fact. However, I shall try and show that the idea that the Cartesian internalist has an unproblematic grip on the conditions under which a cognitive process is owned by a subject, whereas the defender of the amalgamated mind does not, is an idea that cannot be sustained.
Second, I shall argue that the processes occurring in the telescope, computer, or calculator are, when the appropriate conditions are met, subpersonal cognitive processes. The "appropriate" conditions are those identified in the criterion of cognition. However, the processes occurring in the telescope, computer, or calculator are not personal-level cognitive processes. That is, I am going to embrace the problem of bloat for subpersonal cognitive processes, but reject it for personal-level processes. This is not to say that processes occurring in the telescope, computer, or calculator could never be personal-level cognitive processes. But the conditions under which this could happen would be very unusual. Typically, and almost always, intratelescopic processes and their ilk are not personal-level cognitive processes.
Third, the distinction between personal and subpersonal cognitive processes forces us to qualify the ownership condition. There are two quite different ways in which a cognitive process might be owned by a subject. Personal-level ownership of a cognitive process is quite different from subpersonal ownership. That is, the conditions that must be met for a subpersonal cognitive process to be owned are quite different from the conditions that must be met for a personal-level cognitive process to be owned. Typically, the ownership of subpersonal cognitive processes is understood in terms of a certain sort of causal integration. Subpersonal cognitive processes belong to an individual when they are appropriately integrated into the overall cognitive life of that individual. And when that subject is the possessor of personal-level cognitive processes, this means that the subpersonal processes are appropriately integrated when they make an appropriate contribution to the personal-level cognitive processes undergone by the subject. Spelling out the precise nature of the integration required for a subpersonal process to make an "appropriate" contribution to the unfolding of personal-level processes is a fiendishly difficult technical problem. I shall assemble some general arguments for thinking that the ownership of subpersonal processes is best understood in terms of the idea of causal integration. But, I shall not attempt the tricky technical task of providing a criterion that specifies which forms of integration are sufficient for ownership. Instead, my concern is with a more basic problem. This brings us to the fourth, and ultimately most important, task of this chapter.
Fourth, I am going to begin the task of explaining what it is for a subject to own his, her, or its personal-level cognitive processes. Beginning the task, in this chapter, involves canvassing and examining an initially plausible account of ownership of personal-level cognitive processes. Despite its initial plausibility, however, this account will ultimately turn out to be merely derivative. Its derivative character, however, is important; and it points us in the direction of the fundamental roots of ownership of personal-level cognitive processes.
Developing this account is, in effect, the subject of the rest of this book. And I have high hopes for it. I shall argue that when properly developed, the theses of the embodied and extended mind-and so also the amalgamated mind, understood as the conjunction of the two-emerge smoothly and elegantly from this account, as obvious, indeed entirely quotidian, consequences.
2 Ownership: Integration versus Containment
If we are to properly understand cognition, we must understand not only its why and its how, but also its who dimension. There are no subjectless cognitive processes. If this is correct, then the ownership condition is one that must be satisfied by intracranial as well as extended cognitive processes. That satisfying the condition appears less problematic for intracranial processes stems from tacit adoption of an untenable model of ownership: the assumption that ownership of a cognitive process by a subject can be explained in terms of its spatial containment within the boundaries of that subject. Roughly, the idea is that a cognitive process P belongs to subject S just in case P occurs inside S. This criterion of ownership is, I shall argue, indefensible. Indeed, I doubt that spatial containment is a plausible criterion of ownership for any of the primary bodily processes we undergo; a fortiori it is not true of cognitive processes.
To see this, consider, as an example of a noncognitive biological process, digestion.' It may seem obvious that what makes a digestive process mine, and not anyone else's, is the fact that it occurs inside of me and not anyone else. This claim, however, should be resisted: spatial containment is only a fallible guide to the ownership of digestive processes. For a digestive process to be mine, it is neither necessary nor sufficient that it occur inside my body. Consider, first, the issue of necessity. Imagine a case whereby one's digestive processes become externalized. Suppose, for example, one cannot produce enough of the relevant enzymes in one's digestive tract. The solution, drastic and implausible, but nonetheless a solution, is to reroute one's tract into an external device where the relevant enzymes are added, before routing the tract back into one's body where it finishes its work in the usual way. The most natural way of understanding this scenario is, I think, as a case where my digestive processes pass outside my body and receive the required external aid. The processes do not stop being mine just because they are, for a time, located outside my body. We shall examine the intuitions underlying this claim momentarily.
If this is correct, then the specific character of the external device is largely irrelevant-as long as it permits this proper function to be realized. Suppose, for example, that the external device were the body of someone else. That is, suppose my digestive tract were temporarily rerouted through the tract of someone else. We will suppose that the food contents of each tract are (somehow!) kept separate, but that the coupling allows the other person's digestive enzymes to pass over into my tract, thus aiding in the digestion of food that I ingested. Afterward, my tract passes back into my body where it culminates in the usual manner. This would appear to be a case where my digestive processes are spatially located inside someone else's body, and indeed ma
ke use of someone else's digestive enzymes. We can make the same points about sufficiency. If, for example, the position were reversed, and someone's digestive tract rerouted through mine, then a process of digestion would, temporarily, be taking place inside my body: but this digestive process would not be mine.
Underlying both the lack of necessity and sufficiency are two related intuitions. First, a digestive process is defined by way of its proper function: breaking down ingested food and releasing energy for subsequent respiratory processes. Second, a digestive process is mine if it fulfills this proper function with respect to me: it is my food that is being broken down, and energy is being released for my respiratory processes. Digestive processes are defined by way of the proper functions, and a digestive process is mine if it fulfills its function with respect to me. If this is correct, then the connection between ownership and location of a digestive process is merely contingent. The ownership of digestive processes seems to be determined not by spatial containment but by a kind of integration. For a digestive process to be mine, it is necessary and sufficient for it to be integrated into my other biological processes in the right way. It is integrated into my ingestive processes to the extent that it consists in the breaking down of food that I have taken in. And it is integrated into my other respiratory processes to the extent that it releases energy that enables these processes to continue. The criterion of appropriate integration is determined by proper function: a digestive process is appropriately integrated into my other biological processes when it is fulfilling its proper function with respect to those processes.
This, I think, provides the right model for understanding ownership of subpersonal cognitive processes. The failure of spatial containment as a criterion of ownership of subpersonal cognitive processes follows from a general commitment to functionalism-the idea that a cognitive process is defined by what it does.' Just as details of physical implementation are not decisive in determining the identity conditions of psychological kinds, neither, and for the same reason, are details of physical location decisive in determining ownership of instances of those kinds. Once you accept functionalism, you cannot cling to a conception of cognitive ownership based on spatial containment. Where the process is-that is relevant only to the extent that it impinges on what the process does (and, crucially, with respect to whom it does it).
The New Science of the Mind Page 19