The traditional-Cartesian-way of thinking about cognitive processes draws a firm distinction between cognitive processes proper and the framework or scaffolding within which these processes occur. This scaffolding can causally affect, and so facilitate, the cognitive processes proper. But, according to Cartesian cognitive science, we must be careful not to confuse real cognition with extraneous causal accompaniments (Adams and Aizawa 2001). Notice that this point has, in effect, been developed in terms of the language of embedding. Thus, endorsing the thesis of the embedded mind can actually provide a way of attacking non-Cartesian cognitive scienceat least in the way I have developed this idea.
The correct response to this is not to deny the distinction between cognition and scaffolding. Such a move might be tempting, but only, I suspect, because it has, as Bertrand Russell once put, all the advantages of theft over honest toil. We need to reiterate the distinction between causation and constitution and cannot fall into the trap of supposing that whatever causally contributes to a cognitive process is part of that cognitive process (cf. Wheeler 2008). To say that the distinction between cognition and scaffolding has been drawn in the wrong place (or has, in general, not been drawn in the right place) is not to deny that there is a distinction between cognition and scaffolding. Without this distinction, nonCartesian cognitive science becomes true by stipulation.
If non-Cartesian cognitive science is to provide a genuine (i.e., interesting and informative) alternative to its Cartesian counterpart, then it must be based on a stronger claim than the mere environmental embedding of cognitive processes. The required claim, I am going to argue, is that processes occurring in the environment-that is, outside the braincan, in part, literally constitute cognitive processes. That is, things existing or occurring outside of the brain can be partial constituents of cognitive processes. So, the question that should be used to demarcate non-Cartesian cognitive science from the Cartesian alternative is this: can states, processes, and structures that exist outside the brain form part-a genuinely cognitive part-of cognitive processes? Cartesian cognitive science answers no: cognitive processes are exclusively realized by brain processes. To the extent that cognitive processes rely on processes occurring outside the brain, these form part of the scaffolding or framework within which these processes are situated, but do not literally form part of cognition.
Non-Cartesian cognitive science, at least as I shall develop this idea, answers yes: some cognitive processes-not all by any means, but somecan have, as literal constituents, processes occurring outside the brain. "Outside the brain," of course, does not necessarily mean "outside the body." Here, the new science bifurcates. The embodied strand of the science emphasizes the way on which cognitive processes can, in part, be constituted by processes and structures occurring outside the brain but inside the body of the cognitive organism. The extended strand, however, argues that cognitive processes can, in part, be made up of processes and structures that occur outside the body in the wider environment of the cognitive organism.
At stake, of course, is the Cartesian vision of mental phenomena as ones located inside the heads of thinking subjects. This vision has under written both our commonsense conception of the mind, and the scientific attempts to study the mind that have been erected on the foundation of common sense. The Cartesian vision stands or falls on this question: can processes occurring outside of the brain form genuinely cognitive parts of cognitive processes? If they can, and if cognitive processes are mental processes, then the Cartesian vision of the mental must be abandoned. The aim of this book is to show that the Cartesian vision must in fact be abandoned.
1 Cartesian Cognitive Science at Work: Marr's Theory of Vision
The primary aim of this chapter is to show that non-Cartesian cognitive science is not a peculiar eruption of philosophy or philosophers of idiosyncratic bent. Rather, at least in its most recent incarnation, it is motivated by empirical work in a variety of cognate disciplines. This chapter is, therefore, aimed at the uninitiated. It comprises a surveynecessarily brief and for that reason unsatisfactory-of this empirical work. Those already familiar with this work can safely skip to chapter 3.
In the previous chapter I used the expression "non-Cartesian science of the mind"-a science built on a non-Cartesian vision of mental phenomena. At present this expression is largely aspirational rather than descriptive. It applies to something that does not yet exist. However, the expression is not entirely empty. It has meaning to the extent that it picks out, or denotes, a collection of theories of, or approaches to, the mental phenomena that, though often quite different, cluster around a recognizable and recurring theme: mental operations are not confined to processes occurring in the brains of cognizing organisms. Instead, they can be made up, in part, of processes that extend into the organism's body and even into the organism's world.
It would be both inaccurate and premature to describe this cluster of theories as a new science. It would be inaccurate because some of the theories that make this up are not new; shortly, we shall look at some work conducted within this general framework that is nearly a century old. It would be premature because, despite their antipathy toward the central tenet of traditional cognitive science, the theories involved are often disparate; some might even prove to be incompatible. Traditional cognitive science is a systematic and comprehensive framework for understanding cognitive processes, a framework whose foundational principles are explicit and well understood. The aspiring new science is, as yet, not even on nodding terms with this sort of systematic self-awareness that constitutes a mature science. The new science, as yet, has no clear conceptual foundation. This chapter is going to focus on identifying some of the accounts of mental processes that make up the cluster of theories that are candidates for the new science. The rest of the book will be concerned with supplying the relevant foundation for this science. It would, however, be useful to begin with an examination of a classic-indeed paradigmatic-exemplar of Cartesian cognitive science. This will allow us to identify more precisely exactly what it is that the aspiring new science brings to the table. And, as far as Cartesian cognitive science goes, you just do not get more paradigmatic than David Marr's (1982) theory of vision. Indeed, so canonical is Marr's account, it not only provides an example of traditional cognitive science at work, it also played an important role in shaping how cognitive scientists came to understand their enterprise. The example is, therefore, both illustrative and partly constitutive of traditional cognitive science at its best. That is why I am going to use it to provide a foil to much of the discussion to follow-both in this and subsequent chapters.
From the perspective of common sense, visual perception seems to be an obvious example of a process that takes place both in the world and in the brain. For visual perception to occur, light has to first strike the retina, and this is a worldly occurrence rather than an intracranial one. Admittedly, for visual perception to occur, the brain must then process the information it receives. But, one might think, visual perception straddles both worldly and intracranial processes. However, traditional cognitive science relegates the first half of the process to sensation as opposed to perception. The study of perception is, therefore, the study of the processes occurring in the brain whereby it transforms visual input or sensation into visual perception.
According to David Marr's (1982) theory of vision, visual perception begins with stimulation of the retina (sensation) and culminates with the construction of a visual representation of the world (visual perception proper). To understand vision is to understand the processes in betweenthe processes whereby stimulation of the retina is progressively transformed into genuine visual representation. If we think of retinal stimulation as input, and the production of a visual representation of the world as output, then to understand vision is to understand the intervening steps whereby input is transformed into output.
Stimulation of the retina takes the form of electromagnetic energy distributed over an array of retinal locations. T
his pattern of stimulation is often called the retinal image. The retinal image, however, falls far short of a visual representation. Images are flat, static, and notoriously ambiguous. Vision, on the other hand, yields perception of a three-dimensional, mobile, and interpreted world. It seems, therefore, that visual perception must involve processes that embellish and augment the information contained in the retinal image: the image must be processed in some way. To understand vision, according to Marr, is to understand this processing.
The retinal image consists in varying intensities of light distributed over an array of locations on the retina. The variation in light intensity is a result of the way in which light is reflected by structures in the observer's environment. The goal of what Marr calls early visual processing is to create from the retinal image a representation of the visible structures of observed objects, where visible structures are ones that can be observed directly without, as Marr sees it, any need for postperceptual cognitive processes. The shapes of surfaces, and their orientations to and distances from the observer are examples of visible structures in this sense. In contrast, recognition of an object as being of a certain type-table, dog, car, and so onwould involve inferential processes that fall outside the scope of early visual processing.
The first stage in early visual processing consists in the construction of what Marr calls the primal sketch. The primal sketch represents the intensity changes present in the retinal image and makes some of the more global image structures present. Construction of the primal sketch consists of two stages. First, there is the construction of the raw primal sketch. The raw primal sketch is a representational structure instantiated in the brain: a representation of the pattern of light distributed over the retina in which information about the edges and textures of objects is made explicit. To make this information explicit, the brain registers the information contained in the retinal image, and then applies certain transformational rules to it.
These transformational rules are, in essence, something like a process of inference-although this does, of course, undermine Marr's distinction between direct perception and perception based on inference. Alternatively, we can think of them as something like guesses that the brain makes. The brain registers the information contained in the retinal image, and then attempts to work out from this the sorts of edges and textures and object would have to possess in order to produce a retinal image like that. Consider, for example, the transformational rule known as good continuation. The brain detects in the retinal image the presence of a line, which it takes to be indicative of an object's edge. The line stops at a certain point, but shortly after this, another line is found that continues the course of the original line. The rule of good continuation allows the brain to infer that the two lines belong to the same edge. In other words, the brain interprets the gap in the line to indicate the occlusion of a single object rather than the presence of two distinct objects. This "guess" is, of course, fallible. But it is based on a rule of good continuation that is hardwired in the brain. And it became hardwired in the brain only because, all things being equal, it was correct more often than the alternative inference to two distinct objects. So, by way of transformational rules of this sort, the brain essentially embellishes the information contained in the retinal image.
This process of inference continues in the construction of the full primal sketch. Here the brain applies various grouping principles-proximity, similarity, common fate, closure, and so on. Again, the underlying idea is the same. Suppose you have a mishmash of regions, edges, and textures-and this, according to Marr, is essentially what the raw primal sketch is. The next task you will have to accomplish is working out which regions, edges, and textures go together. That is, does this edge belong with this region or that one? Does this texture go together with this edge or that one? The function of the grouping principles is to answer these sorts of questions. Again, these principles are akin to a form of inference whereby the brain embellishes the information contained in the raw primal sketch with its best guess as to what this must mean for the full primal sketch. The result of the application of the various grouping principles is the identification of defined larger structures, boundaries, regions, and so on. This new representational structure is the full primal sketch.
The brain then further embellishes the information contained in this new representational structure with further forms of inference. Here principles concerning depth, motion, shading, and so on are brought into play. The result is the construction of a new representational structure: the 21/2D sketch. This is the culmination of early visual processing and describes the layout of structures in the world from the perspective of the perceiver.
A further, and equally essential, aspect of vision is recognition of objects. Marr argued that in order to recognize the object to which a particular shape corresponds, a third representational layer of processing is required, a layer that centers on the object rather than the observer. This level of processing yields what Marr calls 3D object representations. Again, construction of this new representational structure is akin to a process of inference. Here, the brain utilizes a stored set of object descriptions, and works out which description is most likely to be satisfied by the information contained in the 21/2D sketch.
The language of "inference" and "guessing" is, of course, metaphorical. The brain does not infer in the way we do when we solve a logical puzzle. And it does not guess in anything like the way we do when we guess who it was that just knocked on our door. Nevertheless, Marr's account of what the brain does when we visually perceive the world employs ideas that are similar enough to these familiar concepts to make this language illuminating. Abstracting from the details, Marr's account is built around two related concepts: representations and rules. And these are inseparable.
The retinal image is not a representation (although there has been much confusion on this score). The retinal image can be neither true nor false; neither correct nor incorrect. And it is simply caused by whatever in fact caused it. The key to any representation is that it makes a normative claim on the world. What this means is that a representation makes a claim about the way the world is, and if the representation is instantiated or activated then the world should be that way. If the world is not that way, then the representation is false or incorrect. The retinal image is not the sort of thing that can be false or incorrect: it simply is what it is. Even the raw primal sketch, however, is different. For example, in applying its good continuation guess, and constructing the primal sketch in one way rather than another, the brain is, in effect, making a claim about the way the world is. If the brain applies the good continuation rule, then the observer's world should contain a single occluded object rather than two distinct objects. If this world in fact turns out to contain two distinct objects, the representation is incorrect; and it is incorrect because the brain's "guess," although generally reliable, was in this case erroneous.
So, the primal sketch (in either form) is a representation, whereas the retinal image is not. But what makes the primal sketch into a representation is precisely the fact that it has been constructed via the application of transformational rules (such as good continuation). It is because these rules have been applied that the primal sketch makes a claim about the way the world is-as opposed to the retinal image that is simply what it is. And, therefore, it is because these rules have been applied that, if the world should turn out otherwise, we can legitimately talk of the primal sketch being incorrect or inaccurate. It is the application of these rules that makes the primal sketch the sort of thing that can make a claim on the world, and therefore be either true or false. It is the application of transformational rules that makes the primal sketch a normative item. But being normative is a necessary condition of an item's being a representation. Therefore, ultimately, it is the application of transformational rules to a structure that, at least in part, makes that structure into a representation. In Marr's theory of vision, and in Cartesian cognitive science in general,
the concepts of a representation and (transformational) rule are indissolubly entwined.
The Cartesian idea that the mind is located somewhere inside the head was, therefore, underwritten in Cartesian cognitive science by way of two claims:
1. Mental representations are structures instantiated in the brains of cognizing animals-structures that make claims on the world.
2. Cognitive processes consist in the application of transformational rules to mental representations.
The internality of cognitive processes, therefore, follows from the internality of mental representations. Cognitive processes are rule-governed operations performed by the brain on mental representations. Both the representations and the operations by which they are transformed are internal to the brain.
2 Check This Out!
A perfect place to begin exploration of the cluster of theories that are candidates for the new, non-Cartesian, science of the mind is Kevin O'Regan's Web site (http://nivea.psycho.univ-paris5.fr/). O'Regan is a perceptual psychologist, working at the CNRS in Paris, who made his name studying a phenomenon known as change blindness; and his Web site provides some very good examples of this phenomenon. There you will find photographs of various scenes: Notre Dame Cathedral is in one of them; a fighter plane landing on an aircraft carrier is another; people canoeing on a lake against a backdrop of mountains; and so on. The pictures are going to changesignificantly. I won't tell you precisely how, as that would spoil things. But the changes are significant enough that you would easily notice them under normal circumstances. However, in a change blindness experiment, circumstances are very far from normal.
The New Science of the Mind Page 4
