The New Science of the Mind Page 17
Abstracting from the details, a very definite picture of visual perception emerges. First of all, perception involves information processing: the transformation of information-bearing structures-condition (1) of the criterion. The retinal image is transformed into the raw primal sketch. Further processing operations then transform this into the full primal sketch, and so on. The result of these transformations is the making available of information, to subsequent processing operations, that was previously unavailable-condition (2). Thus, in the transformation of the retinal image into the raw primal sketch, new information becomes available for subsequent processing-information that was not present in the retinal image. And in the transformation of the raw primal sketch into the full primal sketch, further information becomes available, information that was not available in the raw sketch. The culmination of the perceptual process is the 21/21) sketch. This sketch carries information that is available for further processing operations-the postperceptual operations that result in the formation of 3D object representations (which are, in turn, available to play a further role in belief formation, etc.). The general picture is clear: at each stage in the operation, it is possible to identify a new structure, one that carries novel information that is available to subsequent processing operations. Marr's theory, thus, provides a graphic illustration of condition (2).
At each stage in the operation, we find a new representational item. The retinal image, though informationally impoverished, does carry some information about the environment. The goal of visual processing is to successively transform this into an item sufficiently rich in informational content to provide the basis of visual perception and postperceptual judgments. Each stage of the process, therefore, culminates in an item that carries more information about the environment than its predecessor. Once we leave the retinal image behind, each successive item is normative in at least the following minimal sense: given that the item is instantiated with the properties it has, the world is supposed to be a certain way. With the retinal image, there is no distinction between the way the world is and the way it is supposed to be: the retinal image is caused by whatever causes it. However, the raw primal sketch contains new informationinformation contributed by the first stage of perceptual processing. This, in effect, is the brain's "guess" about the way the world would have to be in order to have produced the retinal image being processed. As such, given the "guess," the world is supposed to be a certain way; and if it is not, the "guess" was mistaken. At each successive stage of processing, therefore, we find an item that carries information and makes normative claims on the world. Such items are basic representational states. Moreover, the content they carry as representational states does not derive from the content of representational states that lie outside the particular processing stream. That is, whereas the content of the 21/2D sketch derives from that of the full primal sketch, which, in turn, derives from that of the raw primal sketch, the content involved in the successive transformations that constitute this processing straw does not derive from the content of representational states that lie outside this stream. This is an innocuous sense of "derived" that cannot be what is at issue when objectors to the extended mind insist on cognitive processes containing nonderived content (see the earlier discussion of the distinction between derived and nonderived content). The content embodied in this particular processing stream is, therefore, nonderived content in the sense required by condition (3).
5 Extending Perception
Conditions (1)-(3) of the proposed criterion of the cognitive can, therefore, be painlessly extracted from an examination of a paradigmatic internalist approach to cognitive-scientific practice. Therefore, these conditions can hardly be accused of being devised for the purposes of advocating the theses of the embodied mind and extended mind. Nevertheless, (1)-(3) can also be extracted, in an equally effortless and straightforward manner, from extended models of cognition.
Let us begin with the extended model of perception I defended in The Body in Mind (Rowlands 1999). The basis of this model was a creative reinterpretation of the work of Gibson. Whether you think the resulting account remains sufficiently Gibsonian to merit the epithet (for what it's worth, I do) does not matter for our purposes.
To recap once again: according to Gibson (1966, 1979), the environment is filled with rays of light traveling between the surfaces of objects. At any point, light will converge from all directions. Therefore, at each physical point in the environment, there exists a set of solid visual angles composed of differential light intensities and wavelengths. We can, therefore, imagine an observer as a point surrounded by a sphere that is divided into tiny solid angles. The intensity of light and the mixture of wavelengths vary from one solid angle to another. This spatial pattern of light is the optic array. Light carries information because the structure of the optic array is nomically determined by the nature and position of the surfaces from which it has been reflected.
The optic array is an external information-bearing structure. It is external in the quite obvious sense that it exists outside the skin of a perceiving organism and is in no way dependent on this organism for its existence. By acting on the optic array, and thus transforming it, the perceiving organism is able to make available to itself information that was, prior to this action, present-at least conditionally-but not immediately available. When an observer moves, the entire optic array is transformed, and such transformations contain information about the layout, shapes, and orientations of objects in the world. The transformation of the array makes available to the organism information that was, prior to the action, present in only a conditional or dispositional form.
Specifically, by effecting transformations in the ambient optic arrayby transforming one array into another systematically related arrayperceiving organisms can identify and appropriate what Gibson calls the invariant information contained in the array. This is information contained not in any one static array as such, but in, and only in, the transformation of one optic array into another. In the absence of such transformations, invariant information is present, but only in conditional form: conditionally upon certain types of transformation being systematically related to certain changes in sensory input.
The manipulation of the optic array-the transformation of one optic array into another-is the transformation of one information-bearing structure into another, and thus satisfies condition (1) of the criterion of cognition. The result of this transformation is the making available, either to the organism or to subsequent processing operations, of informationinvariant information-that was previously unavailable, and thus satisfies condition (2).
Now we come to the creative reinterpretation of Gibson's theory employed in the extended model of perception developed in The Body in Mind. The reinterpretation was based on distinguishing what Gibson said about his theory, or is widely thought to have said about his theory, from what his theory actually requires. Gibson's account of perception is widely thought to be hostile to the postulation of representations and to the concept of representation more generally. Whether or not this widespread perception is correct (I think it is not), the extended model of perception developed in The Body in Mind was based on the claim that Gibson's account did not require any general hostility to representations (even if Gibson, arguably, thought that it did). Rather, what Gibson's position actually requires is best expressed in terms of two cognate ideas, one epistemological, the other metaphysical.
The epistemological idea is that one cannot understand the internal information-processing task facing an organism unless one understands the extent to which the organism is capable of manipulating, exploiting, and transforming relevant information-bearing structures in its environment. The metaphysical claim is that visual perception does not start at the retina, but consists, partly, in operations that transform ambient information-bearing structures. It is these claims, I argued, that delineate the content of Gibson's work and underwrite its significance, and not any hostility to representation
per se (Rowlands 1995, 1999).
The extended model of perception developed in The Body in Mind was, therefore, a Gibsonian account purged of any gratuitous hostility toward representations. On this model, the transformation of one optic array into another consists in the manipulation of information-bearing structures. This can be supplemented, when circumstances dictate, with the manipulation of intracranial information-bearing structures. The culmination of processes of this sort is the production, in the subject, of a representational state: the visual recognition that the world is such-and-such (condition 3). And this recognition has been achieved by way of processes-the manipulation and transformation of information-bearing structures-that straddle the boundary of the skin.
The processes described in the extended model of perception developed in The Body in Mind, therefore, satisfy conditions (1)-(3) of the criterion or mark of the cognitive. The arguments of The Body in Mind, in other words, operate on the basis of a criterion of the cognitive that was never made sufficiently explicit in that work, but was (arguably) there in implicit form. It is the satisfaction of these conditions that, pending satisfaction of condition (4), explain why these processes of manipulation and transformation of external structures should be regarded as cognitive and not as merely extraneous causal accompaniments to the supposedly real, internal processes of cognition. The argument is, emphatically, not that they count as cognitive because they are closely causally coupled with internal cognitive processes. And the argument is not that they count as cognitive because they provide a milieu in which cognitive processes are embedded. Rather, they count as cognitive because, pending satisfaction of condition (4), they conform to the mark or criterion of the cognitive. To the extent that the objections to the idea of extended cognitive processes all turn, in one way or another, on the mark of the cognitive objection, the model of extended perception I described in The Body in Mind is, pending satisfaction of condition (4), immune to those objections.
6 Extending Cognition
Consider, now, the extended account of memory I developed in The Body in Mind. This account was structured around four principles, two of which are pertinent to our current concerns. These are:
(A) An organism can process information relevant to memory task T through the manipulation of physical structures in its environment (Rowlands 1999, 122).
(B) In certain circumstances, acting on, or manipulating, external structures is a form of information processing (ibid., 123)
In suitably fleshing out these principles we find, I shall argue, an implicit commitment to the criterion of the cognitive advanced above.
Consider our reliance, highlighted by Luria and Vygotsky (1930/1922), on external information-storage structures in the constitution of memory. The Peruvian kvinu officer who employs a system of knots to store information employs his biological memory in a very different way from that of the envoy of a culture in which external forms of information storage have not been invented (Rowlands 1999, 134-136). The latter must rely on biological memory to retain information, and must do so afresh for each item of information he needs to retain. But the kvinu officer need deploy his biological memory only in the remembering of the "code" that allows him to tap into the information contained in each knot. Once he does this, a potentially unlimited amount of information becomes available to him through his abilities to manipulate and exploit such external structures.
The knot is an information-bearing structure that exists outside the skin of the subject who deploys it: it is, in this sense, external to the subject. Deployment of knots can take various forms: one can tie them, one can modify them, one can read them, and one can use the information they contain as an aid in the construction of further knots. These are all ways in which the knots can be manipulated or exploited. When one ties a knot, for example, or modifies a knot that one has already tied (in order, for example, to register some change in pertinent information), then one is, I argued, manipulating or transforming an information-bearing structure. Thus, the deployment of knots by the kvinu officer satisfies condition (1) of the criterion of the cognitive.
The result of this manipulation or transformation of knots is the making available to the cognitive subject of information that was previously unavailable. Indeed, not only does the knot do this: it is its proper function to do this. We are, of course, dealing with a case of remembering, rather than the transformation, of novel information by a distinct individual. So, the relevant scenario would look something like this. The person who ties the knot would otherwise have forgotten the information that the knot contains, owing to, let us suppose, other demands on his biological memory resources. But when he picks up the knot again-the next day, for example-the information it contains is once again available to him. In this case, the knot has the proper function of making available to the subject information that would have been, without the tying of the knot, unavailable. It thus satisfies condition (2) of the proposed criterion of the cognitive.
The way in which this process of manipulating or exploiting knots makes information available to the subject is in the form of the production in that subject of a representational state: a perception of the knot and subsequent doxastic representations of the informational content contained therein. It is, of course, no part of the thesis of the amalgamated mind to claim that processes entirely external to (i.e., outside the skin of) a cognizing subject can count as cognitive. According to this thesis, cognitive processes are entirely internal or are coupled wholes composed of operations occurring both inside and outside the subject's skin. That is, according to the amalgamated mind, cognitive processes always contain a noneliminable internal element. According to most versions of the amalgamated mind, it is here and only here-in the internal component-that we find representational states that possess nonderived content (but see Rowlands 2006 for an alternative). Manipulation of the knot, an external information-bearing structure, thus makes information available to the subject by way of its production in that subject of representational states that possess nonderived content. Therefore, this manipulation of an external information-bearing structure satisfies condition (3) of the proposed criterion of the cognitive. Note, since this will prove important shortly, that there is nothing in the thesis of the amalgamated mind that requires us to claim that the knots, or any other form of external informationbearing structure, possess non-derived content. The nonderived content need be possessed only by the internal states produced in a subject through its deployment of these sorts of external information-bearing structures.5
Therefore, the kinds of manipulation, exploitation, and transformation of external information-bearing structures described for the process of remembering satisfy conditions (1)-(3) of the criterion of the cognitive in precisely the same way as the manipulation of internal information-bearing structures described by Marr for the process of perceiving. Therefore, if (1)-(3) do indeed partly delineate what it is for a process to qualify as cognitive, then with respect to these conditions at least, external operations of the sort invoked by the thesis of the extended mind seem to qualify as cognitive in the same way that classical internal operations qualify as cognitive.
7 The Objections Revisited
This, then, is the strategy that should be, and arguably largely has been, employed by the defenders of the amalgamated mind (understood as the combination of extended mind and embodied mind). Cognitive processes are hybrid ones that straddle both neural processes and wider bodily and environmental processes. Pending satisfaction of condition (4), these hybrid processes satisfy the criterion of cognition advanced above, and therefore count as cognitive. They qualify as cognitive because they (1) involve manipulation or exploitation of information-bearing structures, where this manipulation or exploitation (2) has the function of making information available to a subject or subsequent processing operations, which it does so by way of (3) the production in that subject of a representational state. Because, pending satisfaction of condition (4), they satisfy this crit
erion, these processes of manipulating or exploiting extracranial structures qualify as cognitive processes.
Their satisfaction of this criterion also makes it clear why there are no purely extracranial cognitive processes. We can divide the thesis of the amalgamated mind into two claims. First, a process of manipulating, transforming, or exploiting an external structure never counts as cognitive unless it is combined with an appropriate internal (i.e., neural) process. Second, once combined in this way, the external process can be regarded as cognitive to the same extent, and for the same reasons as the internal one. The criterion explains both of these claims. The external process on its own never counts as cognitive because cognition, I have assumed of the sake of argument, always involves representational states with nonderived content. This, it is commonly supposed, can only be found on the inside.' Cognitive processes always involve the production in a subject of a representational state-for this is how they satisfy condition (3). Therefore, conditions (1)-(3) of the criterion collectively entail that all cognitive processes are purely intracranial or a combination of intra- and extracranial processes. There are, therefore, no purely extracranial cognitive processes. Second, once combined with an appropriate internal process, the external process counts as cognitive because, pending condition (4), it satisfies the criterion of cognition: it consists of (1) manipulation or transformation of information-bearing structures, where this has the proper function of (2) making available to the subject or subsequent processing operations of information that was, prior to this manipulation or transformation, present but not available, which it does by way of (3) the production in the subject of a representational state.