By Greg Ransom

Greg Ransom -- Dept. of Philosophy, UC-Riverside. 

"We are trying to explain relations between a subject [human beings] and a physical world that we take ourselves to know something about .. What we know about [the] individuation [of physical, biological or psychological categories] is derived from reflecting on explanations and descriptions of going cognitive practices .. One cannot assume without serious discussion that psychology is correctly characterized as a science (only) of behavior." -- Tyler Burge, 1986.

".. common-sense mentalistic and intentional notions need no foundation in physicalism. Their legitimacy is assured, not by justification in non-intentional terms, but by their indispensable contribution to our cognitive enterprise." -- Lynne Rudder Baker, 1987.

"What [the rejection of conceptual realism] requires us to do is give up the picture of Nature as having its very own language which it is waiting for us to discover and use .. We make up uses of words .. and none of them is merely copied off the world itself. Yet for all that some of our sentences are true, and .. the truth of 'I had cereal for breakfast this morning' does depend on what happened this morning". -- Hilary Putnam, 1992

In this paper I both defend and criticize Tyler Burge's account of theoretical psychology as an explanatory science, and then I go on to support Gerald Edelman's pioneering work in psychology as both a scientific and conceptual breakthrough which avoids many of the difficulties which are implicit in Burge's own account of the science of theoretical psychology.

I first defend Burge in his claim that theoretical psychology seeks to explain the capacity of the human cognitive system to successfully represent the external world. I argue that Burge is correct when he claims that theoretical psychology is a legitimate explanatory enterprise which individuates cognitive states externally or non-individualistically using the categories of the physical sciences.

I then criticize Burge for his implied suggestion that the objective categories which are given to us by the physical sciences -- and are believed by us to exist in a world external to ourselves -- can be understood as independently existing 'information' which is delivered from the external environment to the human mind like a fancy hat in a ribboned box. I argue that Burge's notion of 'information' as categories which are given a priori in the environment and then conveyed to the human organism through a process of transferal or 'instruction' is both scientifically and philosophically dubious. It is scientifically dubious because of the limited informational capacity of the human genetic code and because of the massive variability displayed throughout development by the human neural wiring system. And it is philosophically dubious because it must assume the existence of an interpretive homunculus which acts as the 'mind' within the brain which is able to interpret the 'information' that has been transferred from the environment to the human organism.

In fact, even if the theory of information transfer through instruction could overcome these seemingly intractable problems, the theory would still be condemned for its explanatory inadequacy as a general mechanism of human categorization. The theory would be condemned for at least two reasons: (1) for its ongoing inability to account for context-sensitive categorization in an ambiguous and multifaceted world; and (2) for its ongoing inability to explain the development or emergence of a new category and its reliable projection from a limited sample into an array of novel situations.

Finally, I introduce Gerald Edelman's theory of neural group selection, a global theory within theoretical psychology which presents a neurologically plausible mechanism for the development of adaptively appropriate human categories. I argue that Edelman's theory explains how human categories could develop during the life of an individual without the need for information transfer through 'instruction' by the environment in much the same fashion as natural selection and clonal selection can explain the development of phenotypes and antibody responses without the need for informational instruction from the environment through prior design or through antigen templating. I go on to explain how Edelman's theory of neural group selection avoids the problem of the homunculus, solves the problem of context-sensitive categorization, and allows for the development and projection of a category from a limited sample into an array of novel situations.

Karl Popper suggests, rightly I think, that a central task in philosophy as in science is to correctly identify the nature of the problem. Let me begin then with a simple account of the field of theoretical psychology. Theoretical psychology is conceived of by most neuroscientists as a scientific approach to the problem of mind. What is the problem of mind? The problem of mind begins with the classical picture of the scientific observer as a psychologically transparent vacuity -- an empty slate. The procedure of the physical sciences as understood by philosophers since the time of Galileo is one which effectively removes the human mind from nature and thus from our scientific world view. Yet if we wish to account for the existence of human knowledge of a world which exists independently of ourselves the notion of a Galilean observer will not do.

Any proposed account of the physical sciences would be incomplete, of course, without the inclusion of an adequate account of the scientific observer. The very idea of an external world begins, for example, with the recognition that our purposive activities can influence the future course of events. We immediately distinguish between the external world and the domain of our own mind when we notice that external events do not regularly differ in their effects upon each other in the same way that they differ in their effects upon ourselves.

The advance of the physical sciences comes with the development of a system for classifying external events which provides a more accurate description of the regularities found in nature than is provided by the system of classification which is provided by our own senses. Yet even with the sophisticated development of physical theory the question still remains why external events appear to us as they do. The advance of the physical sciences simply exposes the existence a mental order which itself becomes a scientific problem and the legitimate focus of scientific inquiry. The problem of mind is therefore the proper subject for a theoretical psychology which aims to provide a neurological explanation for the categories of human perception. It is, in fact, hard to image what an adequate or complete account of human knowledge would look like which did not include such an account.

Of course, the traditional 'Galilean' stance which posits a psychologically transparent scientific observer has been taken by many philosophers to be a complete and adequate account of human knowledge. But the Galilean stance is in fact insufficient as an account of human knowledge because any successful account of our knowledge of the physical sciences must include a physical account of psychological events. This follows from the fact that any adequate and complete account of the physical sciences must posit the existence of a scientific observer. Yet as Gerald Edelman explains, when we attempt to account for the scientific observer we may remove our own mind from this account but we may not remove his mind. In fact, it is only by means of the interaction between the purposive activities and the sensory experiences of the scientific observer that the constructions of the physical sciences may be tested. As a result, any adequate theory of knowledge which includes the physical sciences will present us with the problem which is addressed by theoretical psychology.

As I have suggested, the problem of theoretical psychology is precisely the opposite of the problem confronted by the physical sciences. The problem has been well cast by Friedrich Hayek. According to Hayek, the problem which is addressed by the physical sciences arises because objects which appear alike to us do not always behave in the same manner towards other objects and, similarly, objects which appear to us to be quite different often behave in a similar manner. By contrast, theoretical psychology confronts the problem which arises because events as they are classified by the physical sciences produce in their effect on our senses a quite different order of events. As Hayek explains, (closely paraphrasing his own words), what we routinely take to be the physical order of the external world in our everyday affairs is in fact an order of sensory qualities which we experience whenever we learn anything about the external world but which has no place the physical account of the external world which has been developed by modern science. The mystery which the domain of theoretical psychology presents for the philosopher is generated by the seeming paradox of a scientific discipline which attempts to provide a physical account for an order of sensory experiences which find no place in our scientific picture of the external world. Yet the existence of such an order of sensory qualities is exactly the problem which theoretical psychology must address.

My argument thus far has been that the philosophy of mind is properly concerned with the epistemological legitimacy and metaphysical coherence of what I have called theoretical psychology. But what is theoretical psychology? What are its presuppositions and what does it explain? In what way if any is theoretical psychology tied to the understanding and explanation of human behavior?

Theoretical psychology begins with the premise that we can understand and explain human action in a manner we cannot do with physical phenomena. It assumes that we can understand and explain human action directly from the model of our own behavior -- which is something we cannot do when we seek to understand and explain physical phenomena. This understanding is direct and nonlinguistic although it may be shaped by the indirect and often metaphorical techniques which are used to linguistically characterize such understanding. The development of language is in fact built upon this prior capacity for human understanding.

In the past, philosophers often felt that the fact that our ordinary explanatory understanding of human behavior cannot be directly confirmed in the same manner that physical explanations were thought to be confirmed was good enough reason to de-legitimatize this form of explanation. Philosophers now recognize that no form of explanation is justified in the manner once demanded. In fact, although there are no objective tests for successful human understanding in the sense which is in fact found within the physical sciences, the coherence and stability of human social and linguistic practices lends overwhelming support to the legitimacy of the direct explanatory understanding of human behavior. The legitimacy of the thesis of direct explanatory understanding is also supported by our common selective history -- a fact which was unknown to or little understood by earlier philosophers.

What does it mean to have a direct explanatory understanding of the behavior of another human beings? It means that we are justified in assuming that other human beings treat objects in the world as alike or unalike in the same manner that we ourselves do. It means that we are justified when we suppose that other human beings recognize a distinct contrast between the differences of events in their effects upon each other and the differences which events in the world have in their effects upon us. We might call this distinctive order of classification the order of human understanding -- or perhaps the order of the human mind.

It is no surprise then and scientifically quite untroubling that researchers in theoretical psychology take it for granted that human beings share a common psychological order of classification which differs from the order of classification which is given to us by the physical sciences. As I have argued, theoretical psychology is on very strong grounds when it assumes that the categories of human understanding -- what we are want call the human 'mind' -- are an ordered domain we know something about because we are human beings. Not only do we -- as human beings -- share a common genetic make-up which provides the basic foundations for a common neurological organization -- we also share a common social nexus which remains unfalsified by any sort of systematic failure to spontaneously intercoordinate our everyday activities with others -- an intercoordination which is dependent upon an unreflective anticipation of the well-ordered behavior of others.

The presumptions of theoretical psychology only become problematic in the context of philosophical commitments which are shared by most modern philosophers of language and a good many cognitive and neural scientists but which are themselves highly controversial. Although these philosophical commitments have a long intellectual history it is only in the contemporary period that the incompatibility which exists between these commitments and the very possibility of a theoretical psychology has been clearly exposed -- if not well understood.

The problem which I wish to focus upon is what Edelman refers to as the problem of 'information'. There are currently two principle views of scientific psychology -- both of which take for granted the problematic notion of 'information' which is somehow given in the environment and directly transferred to the human mind like a package in the mail. The first -- and dominant -- view of scientific psychology -- popularized by Jerry Fodor -- attempts to interpret psychology as a 'science' of behavior which formulates lawlike generalizations and cites true causal relations in the intensional language of content. This view seeks to 'vindicate' belief/desire psychology as a legitimate explanatory enterprise by showing how intentional states may be discharged in the philosophically respectably extensional language of the physical sciences.

The second view -- popularized by Tyler Burge -- sees scientific psychology as an attempt to account for the neural processes and cognitive structures which are responsible for human cognitive success or failure. This view seeks to characterize the actual explanatory enterprise of theoretical psychology as it is conducted by working cognitive and neural scientists. On this account theoretical psychologists are attempting to model the neural processes and cognitive structures which are responsible for our ability --or inability -- to successfully classify our external and social environment.

This second view disputes the explanatory adequacy of the view of psychology which characterizes it as a science of human behavior. Burge, for example, charges that such a view mistakes the evidence of psychology for its subject matter. The subject matter of psychology, according to Burge, is concerned with the sort of molar abilities which are treated by contemporary theories of vision, memory, learning, linguistic understanding, belief formation, and categorization. Bernard Kobes also criticizes the Fodor tradition for its impoverished conception of the explanatory aims of psychology. According to Kobes, for example, the study of psychological categories which are structured by prototypes does not fit the dominant conception of psychology as a causal science of bodily motions.

On the second view of scientific psychology the semantic content of the explanatory kinds of theoretical psychology -- e.g. the everyday categories of human perception -- are type identified non-individualistically for explanatory purposes by the external environment in which human beings find themselves -- as that environment is understood by scientifically sophisticated and linguistically competent working psychologists. On this view, in other words, the correct application of the everyday categories of human perception -- e.g. color categories, middle size object categories, spatial categories, etc. -- is fixed externally or non-individualistically either by the objective categories which are hypothesized by the theories developed within the physical sciences or by the conventional categories of natural language which are developed and enforced by competent or authoritative speakers. The categories of human perception, the categories of the physical sciences, and the categories of social environment are each ordered domains with which the psychologist is assumed to be intimately familiar. Indeed, the theoretical psychologist sees it precisely as his task to establish the connections which exist between these domains. As Burge explains, "We are trying to explain relations between a subject and a physical [or social] world that we take ourselves to know something about."

Burge uses a number of thought experiments to establish that representational states are ascribed to the states of human brain so that these states may be evaluated for their practical or semantic success -- i.e. for their truth. Burge shows how for purposes of semantic evaluation we may image two quite different scenarios in which a single individual's brain states are correlated with two quite different external environments even while that person's internal physical history remains unchanged. The thought experiment is founded upon the well confirmed premise that two different causes my have the same effect on another structure -- for example the human brain. According to Burge, in this scenario an external observer is warranted in saying for explanatory purposes that the individual's brain states are true of or are 'about' two different things or causal environments.

Burge uses David Marr's theory of vision to illustrate his claim that working psychologists individuate the representational content of human brain states by reference to the environment in which the human organism finds itself -- as that environment is characterized in the objective terms of contemporary physical science. According to Burge, "Marr's claim that the structure of the real world figures in determining the nature of the representations that are attributed in the theory is tantamount to the chief point about representation or reference that generates our thought experiment .. ". Burge characterizes Marr's theory of vision as one which attempts to characterize the mechanisms which account for our success in acquiring 'information' about the world. For example, according to Marr's theory the activation of a set of retinal mechanisms which are sensitive to discontinuities of light -- i. e. a set of 'zero crossing segments' -- can be identified as the mechanisms which represent a single kind of physical phenomena -- i.e. a single edge in the external environment as characterized by the physical sciences -- because the activation of such a set is typically correlated with a single physical edge in the external environment. Of course, it remains true that in some situations such a set may fail to be activated by a single edge in the external environment and in other situations a set may be activated by events in an external environment which cannot be characterized as a single physical edge -- although the activation of a set of 'zero crossing segments' is in fact typically caused by a single physical edge in the eternal environment,

The representational content of these sets of retinal mechanisms is individuated by the current external environment not simply because that environment is the physical environment in which man currently finds himself and in which these mechanism are typically activated -- but more importantly because it is the environment to which these mechanisms are thought to be adapted. It should be understood, of course, that if man had found himself in an alternative external environment the activation of these same sets of retinal mechanism might have been typically correlated with an environmental cause which consist of something other than a single physical edge. If this external environment was the one to which he had been adapted, then such would be the representational content of his retinal mechanisms. We can, that is, imagine the representational content of these sets of retinal mechanisms as being something quite different without any concurrent change in the internal physical constitution of the human organism. Still, in our external environment the activation of sets of zero crossing segments represent or fail to represent physical edges -- because these states are individuated for evaluative purposes by the external environment in which we finds ourselves -- as that environment is understood by the physical sciences.

Unfortunately, Marr-style theories are severely inadequate if they are meant to provide a plausible and complete account of human perceptual categorization. Although Burge's use of Marr's theory of vision is an effective example for illustrating the way in which practicing theoretical psychologists individuate human brain states by reference to the external environment, as a general model which hopes to account for the success or failure of human perceptual categorization the computational form of Marr's theory is severely deficient. The central difficultly with Marr-style theories of perceptual categorization is the assumption that the determination of the representational content of human brain states can understood as a simple algorithmic process which directly transfers structure from the environment to the organism.

The theory of direct computational transfer of structure -- i.e. 'instruction' -- from the environment to the organism is deficient for at least three principle reasons. First, what we know about developmental biology makes it highly unlikely that the human sensory system exhibits the sort of precise and prespecified point-to-point wiring which is demanded by a theory of direct computational transfer. Second, the instruction model can only account for the representational content of those human brain states which can be understood to be directly correlated with the general features of the external environment to which our species had been directly adapted in the past. It is, however, almost impossible to directly correlate any particular feature of the external environment to which we have been adapted in the past with particular cognitive structures which are clearly adapted for any specific purpose -- and without such a strict correlation between adaptive environment and adaptive response we have no clear criteria for attributing particular representational content to particular brain states.

Finally, and most significantly, the aspect of human cognitive capacity which we most wish to explain and understand is our ability to categorize novel objects and events in a perceptual environment which is new to the human organism from an adaptive point of view. Unfortunately, instructional models of human cognitive capacity have been singularly unsuccessful when it comes providing a plausible account of adaptively novel categorization which is useful for explaining representational success or failure.

The need to account for the categorization of adaptively novel objects or events generates two central problems for the instructive model. If the instructive model attempts to account for representational content of a brain state by reference to the causes of such a state it is either faced with the problem of disambiguating the representational content of any particular brain state from its complex causal history or it is faced with the problem of allowing for misrepresentation or cognitive error. The first problem -- the qua-problem -- is the problem of providing a principled reason for identifying the precise physical element in a complex causal environment which is to be identified as a brain state's representionally specifying cause in the external environment. If the brain state was caused by a complex chain of objects and events why should any particular one of those objects or events be selected out as the representational content of the brain state?

The second problem -- the disjunction problem -- is the problem of providing principled criteria for specifying the precise range of physical characteristics which a particular brain state represents in a manner which allows for cognitive error or misrepresentation. If a disjunctive range of different objects and events can all cause the same brain state, why should any particular class of those objects and events be selected out as the brain states representational content? Without access to an adaptive story to draw these distinctions the instructionist model cannot answer these questions.

In fact, it is exactly our common cognitive capacity to respond in any number of identifiably appropriate ways to an infinite number of different examples of an adaptively novel category after only a few initial confrontations with a small number of its members which is the primary source of problems for the instructionist model of categorization. The problem is to explain how human beings are capable of recognizably appropriate behavior towards objects and events which are encountered in novel contexts after only a few encounters with those objects or events in entirely different contexts.

There can be little doubt that Burge looks upon Marr's theory of vision as a instructionistic model of perceptual categorization in which structure is directly transferred from the environment to the organism. According to Burge, "The primary aim of referring to contingent physical facts and properties is to enable the theory to explain the visual systems' reliable acquisition of information about the physical world: to explain the success or veridicality of various types of visual representations." (underlying added) Burge also remarks that , ".. the theory is set up to explain the reliability of a great variety of processes and sub-processes for acquiring information .. ". (underlying added)

The problem with Burge's account of representational content is not his claim that brain states are individuated for evaluative purposes externally by reference to the physical environment (as that environment is characterized by the physical sciences). The real difficulty with Burge's account is that it takes for granted the deeply problematic notion of 'information' -- a notion which tacitly presupposes the functioning of an instructive mechanism which is able to link brain states directly to physical categories. As we have already seen, however, no instructive mechanism can be viewed as tying brain states directly to physical categories in the external world unless we are also warranted in presupposing the existence of a selective story which picks out the adaptively relevant aspects of the external environment with which brain states are strictly correlated .

What Burge has not explained is exactly how an instructive mechanism can be imagined to link 'information' in the external environment directly to a particular brain state when that 'information' is an adaptively novel category of objects or events which is not only new to the particular individual but which is also new to the human species from the perspective of biological adaptation -- e.g. alcohol, beer, carburetor, Doberman pinscher, etc. In fact, then, although the idea of 'information' transfer provides Burge with a hypothetical 'mechanism' which can be imagined to be responsible for representational success or failure -- it also commits him to the existence of a mysterious process which magically links brain states to objects and events in the external environments as if there were a preexisting categorical structure in the world which was tied by a preestablished harmony to particular brain states.

Burge's notion of 'information' assumes that the representational content of a brain state can be tied by direct causal instruction to the external environment although nothing in our evolutionary history can be cited as establishing that connection. The implicit assumption contained within the notion of 'information' seems to be that the external world somehow comes in nice little packages with category labels on them which can be delivered directly to the human brain with intrinsic semantic significance like a message from Western Union. But as we have seen, no instructive mechanism which is not magical has been able to account for how adaptively novel categories of object and events in the external environment which are not universally linked to particular brain states can be identified as the representational content of a particular brain state in a manner which allows for misrepresentation. The problem for the instructive model is to explain how an adaptively novel category of objects or events may be considered to be linked to particular brain states on the basis of only a very few exposures to the presumed members of the new category and then linked again and again for evaluative purposes to those same brain states in complex and never before experienced contexts.

Gerald Edelman has repeatedly criticized the sort of linguistically inspired 'top-down' accounts of representational content which implicitly assume that 'information' preexists independently in the external environment and is then somehow innately associated with or directly communicated to human brain states through a physical process of 'instruction'. Edelman argues that the notion of 'information' and the concomitant hypothesis of 'instruction' must be rejected for at least two reasons: 1) the nervous system of individual human beings show an enormous variability in the connective structure of their neuronal maps which is underdetermined by their genetic code; and 2) human beings are capable of categorization and association in widely variant environments full of novelty.

According to Edelman, a process of 'instruction' can be likened to the process in which specific structural aspects of a key have been directly transferred to a lock during its construction. When the key is removed from the lock after the lock has been internally configured various characteristics of the external structure of the key remain imprinted within the lock. Those aspects of this external structure which allow for the repeated or interchangeable use of the same or similar keys for opening the lock can -- from the perspective of the locksmith -- be understood as 'information' about the key which is contained within the lock.

Edelman's example illustrates the fact that in what are recognized as paradigm cases of the process of 'instruction' those particular aspects of structure which are to count as 'information' must be defined in advance by a human agent with his own plans and projects. We recognize that the template left within the lock contains 'information' about the key for the locksmith because aspects of that template serve his conscious purposes. Edelman's point is that the hypothesis of 'instruction' assumes that only selected aspects of the structure which is passed from one item to another are to count as 'information' -- and that these aspects are defined a priori. As a result, the classic difficulty for 'information' theorists is to identify the 'agent' in the brain or in nature which performs this function. Edelman's argument is that theories of 'instruction' are inevitably left with the problem of homunculus -- i.e. who or what decides what counts as 'information'? -- exactly for the reason that those theories start with the problematic assumption that 'information' preexists in the external world in the form of immutable categories of objects and events which are established prior to our encounter with the world and which are then directly transferred to the human brain through a process of instruction.

As an alternative which avoids the problem of the homunculus, Edelman suggests that basic human perceptual categories are generated from the 'bottom-up' through a process of selection rather than from the 'top down' through a process of instruction. Edelman points out that all selective processes assume the prior existence of structural variations whose causal origins are not directly influenced by a posteriori events but whose functional significance is later determined by a posteriori events. For example, the functional significance of particular gene mutations cannot be assessed until some natural selective condition acts upon the phenotype -- and the prior existence of such a gene mutation is causally independent of this later selective event. Alternatively, particular antibody variations are generated spontaneously by the immune system but lack a functional significance within that system until a cell already bearing that antibody has a response to a particular antigen. For example, a cell bearing an antibody which has never before responded to an foreign antigen may bind to a newly synthesized artificial organic compound -- i.e. an hapten -- producing a functionally significant immune response which has never been produced in the prior history of the species.

According to Edelman, as a theory of categorization, the selective hypothesis offers a biologically and neurologically plausible alternative to the hypothesis of 'instruction' which does not require that the significance of one neural structure should be determined in advance by some aspect of another structure -- whether this structure is in the brain or in the external world. Edelman's argument is that because there is no particular semantic order given to it a priori in the external world the human brain must form representational categories for itself from the 'bottom-up' in the same manner that functional kinds are formed in biology or immunology -- by a process of selection over independently generated structural variations. As Edelman explains, the hypothesis of selection avoids the homunculus problem by supposing that the environment acts indirectly and dynamically on the potential orderings already exhibited by neural structures rather than supposing that these structures respond directly to 'information' which already exists in the environment.

Edelman's leading idea is that the categories of human perception are not themselves present in the environment independently of human cognition but must be constructed by each individual according to what is adaptive for its species and its own particular circumstances. On this view the human brain is a creator of criteria leading to the production of information rather than a receiver of such information from the environment. Edelman's selective account of human categorization explains how information comes to exist in an unlabeled world without assuming a magical process of direct instruction or prearranged harmony. As Edelman puts it, "The world becomes 'labeled' as a consequence of behavior that leads to particular selective events within such neural structures in each animal."

We are now in a position to express the theoretical task of psychology in a form which avoids the problematic notion of information while remaining true to the claim that human brain states are individuated externally or non-individualistically so that they may be evaluated for their success in representing the world as it in understood either by the physical sciences or by the social community. We are in a position, in other words, to describe theoretical psychology like Burge as an attempt to, "illumine as specifically as possible the mechanisms that underly and make true our evaluations" without agreeing with him that the neurological mechanisms under consideration to can be correctly describes as "processes or subprocess for acquiring information."

The central task of theoretical psychology is to explain how human perceptual categorization could take place in the human brain. Neurobiologists working in the field of theoretical psychology seek to provide a biologically plausible and philosophically sound in principle explanation of the neural bases of perceptual categorization. Specifically, as Edelman suggests, theoretical psychology seeks to account for the organizational features of neural networks and their synapses which permit an organism to recognize larger numbers of different examples of a category after initial confrontation with just a few of instances of that category.

In this paper, I have suggested that the problem of how the human brain could be capable of perceptual categorization in contexts full of novelty can only be solved by positing the existence of a selective mechanism within the brain and cannot be solved by the hypothesis that the human brain is a simple mechanism of instruction. I have also suggested that a neurobiological theory which seeks to account for our ability to successfully categorize objects and events in contexts we have never before encountered must not only assume a common network of human practices but must also make reference to contingent physical facts and properties in the external environment. I have suggested, in other words, that scientific understanding of the problem of the successful perceptual categorization in a world of novelty can only be made in the context of a publicly shared network of appropriate classifications and with reference to the external world as characterized by the physical sciences. In this I follow Burge. However, unlike Burge, I have rejected the possibility that such a capability can be explained by an instructive process in which 'information' is transferred from the environment to the brain.

We can agree, then, with Bernard Kobes that theoretical psychologically aims at, "explaining a subject's classificatory successes and failures, where success and failure are judged against public standards or correct application of the subject's concepts." Yet when Kobes goes on to contend that, "a subject's concept truly applies to all and only things that share the underlying nature of the sample, where the sample is picked out in virtue of causal interactions between concept users in one's community and real-world items," we can add that human beings have no God's eye view or direct 'informational' access to the sample in question other than that which is revealed in our own common practices and the well-bounded but disjunctive characterization of that sample which is provided by the natural sciences. So, although it is scientifically respectable to ascribe representational content to human brain states for explanatory purposes by refering to physically characterized objects and events, it is a mistake to interpret this representational content as an 'informational' object which can be pointed to as existing either in the world or in the brain. As Putnam explains, "Meanings are not objects in a museum, to which words somehow get attached." The claim of my paper, in other words, has been that a selective theory of the brain can help to explain how facts about our psychologies and the structure of the world can give rise to brain states which are semantically evaluable for their success in representing a world which is full of new objects and novel contexts.

Dec., 1992

Click here to return to Greg Ransom Home Page.

See Gerald Edelman, 1987, Neural Darwinism, New York: Basic Books, pp. 38-41.

This discussion follows Gerald Edelman, 1989, The Remembered Present, New York: Basic Books; Friedrich Hayek, 1952, The Sensory Order, Chicago: U. of Chicago Press; Friedrich Hayek, 1979, The Counter-Revolution of Science, Indianapolis: Liberty Press; and Ludwig von Mises, The Ultimate Foundation of Economic Science, Kansas City: Sheed Andrews and McMeel. See also Karl Popper, 1983, Realism and the Aim of Science Totowa: Rowman and Littlefield; Tyler Burge, 1986, "Individualism and Psychology", Phil. Rev. 95, pp. 3-45; and Bernard Kobes, 1989, "Semantics and Psychological Prototypes", P. Phil Q., 70, pp. 1-18.

Galileo himself, of course, accepted the existence of a mental order -- as do most physical scientists. See Hayek, 1982, p. 288.

Compare Popper, 1983, pp. 97-100; and Lynne Rudder Baker, 1987, Saving Belief, Princeton: Princeton U. Press, p. 171.

See Hayek, 1952, where he say, "The task of the physical sciences is to replace that classification of events which our senses perform but which proves inadequate to describe the regularities in these evens, by a classification which will put us in a better posistion to do so. The task of theoretical psychology is the converse one of explaining why these events, which on the basis of their relations to each other can be arranges in a certain (physical) order, manifest a different order in their effect on our senses."

Compare Hayek, 1952, p. 7, "The physical sciences, even in their ideal perfect development, give us only a partial explanation of the world as we know it through our sense and must always leave an unexplained residue .. even after we have learnt to distinguish events in the external world according to the different effects they have upon each other, and irrespective of whether they appear to us as alike or different, the question of what makes them appear alike or different to us still remains to be solves. .. once we have learned that the things in the external world show uniformity in their behavior toward each other only if we group them in a way different from that in which they appear to our senses, the question why the appear to us in that particular way, and especially why they appear in the same way to different people, becomes a genuine problem calling for an answer."

Even eliminitive materialists concede that we have the capacity to identify behavioral patterns from the example of our own case. See, for example, Paul Churchland, A Neurocomputational Perspective, Cambridge: MIT Press, pp. 118-119, where he says, ".. an enormous amount of one's appreciation of the internal states and overt behavior of other humans derives form one's ability to examine and to extrapolate from the facts of one's own case .. One learns from every example of humanity one encounters, and one encounters oneself on a systematic basis." Compare Stephen Stich, From Folk Psychology to Cognitive Science, Cambridge: MIT Press, p. 7, were he says, ".. in ascribing content to a belief we are, in effect, comparing the believer to ourselves." See also Friedrich Hayek, "The Facts of the Social Sciences", in Friedrich Hayek, 1948, Individualism and Economics Order, Chicago: U. of Chicago Press; and the later work of Ludwig Wittgenstein.

See Mark Johnson, 1987, The Body in Mind, Chicago: U. of Chicago Press.

The later Ludwig Wittgenstein can be read as a long illustration of just this point. For a revealing statement which broadly hints of this view which Wittgenstein never directly argues for see Ludwig Wittgenstein, 1980, p. 31, where he says, "The origin and the primitive form of the language game is a reaction; only from this can more complicated forms develop. Language .. is a refinement, 'in the beginning was the deed'. Compare also Willard Quine, Word and Object, Cambridge: MIT Press, p. 82-3; and Friedrich Hayek, 1967, "Rules, Perception and Intelligibility", pp. 43-65, in his Studies in Philosophy, Politics and Economics, Chicago: U. of Chicago Press.

See Quine, 1960; Daniel Dennett, 1969, Content and Consciousness, London: Routledge & Kegan Paul; Burge, 1986; Gerald Edelman, 1987; Baker, 1987.

See Edelman, 1987, pp. 12-13; 24-33; 38-45; Edelman and Reeke, 1988, "Real Brains and Artificial Intelligence, Daedalus, 117: 1, pp. 143-173; Edelman, 1989, pp. 40-43.

For the purposes of this paper I shall ignore the claims of the eliminative materialists that there can be a coherent scientific psychology which is bind to the existence of the common-sense categories of everyday human understanding -- but see footnote 13 below.

Eliminative materialists accept the view of scientific psychology as a science of behavior -- yet reject the 'radically false' intensional categories of belief/desire psychology in favor of non-intensional replacement categories in the purely extensional 'language' of neuroscience or functional psychology which are viewed as limning reality in exactly the same empirically adequate way that the categories of micro-physics are held to limn reality. But once we have abandoned not only the received view of theories and deductive-nomological model of explanation but also the notion of truth -- as Paul Churchland has done (see Churchland, 1989, p. 121, p. 149, pp. 153-154) -- then one wonders what is left of the claim that a scientific psychology which takes the common-sense categories of human understanding for granted should be conceived as a radically false and explanatorily degenerate theoretical research program in the nomological science of behavior. Putting these terms up for grabs only weakens the plausibility and coherence of Churchland's case against the epistemic acceptability of the common-sense categories of everyday human understanding -- just as it weakens the very motivation for such a program. See, for example, Hilary Putnam, 1992, "Truth, Activation Vectors and Possession Conditions for Concepts", in Phil. & Phen. R. 52:2, pp. 431-447; and Hilary Putnam, 1988, Representation and Reality, Cambridge: MIT Press, pp. 57-71. See also Baker, 1987, pp. 124-128, and pp. 134-148.

Burge, 1986, p. 12.

Kobes, pp. 12-13.

Burge, 1986, p. 12.

Burge, 1986, p. 27

Edelman, 1987, pp. 38-39

According to Edelman, "The context driven character of perceptual categorization and the capacity to generalize from a few learned examples together strongly challenges any .. information driven explanation of the data."

Burge, 1986, p. 32.

Burge, 1986, p. 29.

Compare Putnam, 1988.

Edelman's claim is very similar to Wittgenstein's point that the meaning of a word cannot be handed from one person to another by a single act of ostensive definition.

Edelman is director of the Neurosciences Institute at Rockefeller University.

See Gerald Edelman, 1982, "Through a Computer Darkly", Bul. of the Am. Ac. of Arts & Sci., 36:1, 20-49; Gerald Edelman and Leif Finkel, 1984, "Neuronal Group Selection in the Cerebral Cortex", in G. Edelman, W. Gall, and W. Cowan, eds. Dynamical Aspects of Neocortical Functions, New York: Wiley; Gerald Edelman, 1985, "Neural Darwinism" in M. Shafto, ed. How We Know, New York: Harper & Row; Edelman, 1987; Edelman and Reeke, 1988; Edelman, 1989.

See Edelman, 1987, pp. 38-39; Edelman, 1989, pp. 4041

Compare Putnam's claim that, "If we confine ourselves to .. 'structural' properties describable in physical science, then there is no scientifically describable property common to all red things..". (Putnam, 1988, p. 2.)

Edelman, 1987, p. 41.

Burge, 1986, p. 25.

Burge, 1986, p. 29.

Edelman, 1987, p. 4.

Compare Hayek, 1952.

Edelman, 1987, p. 25.

Massimo Piattelli-Palmarini suggests that the history of biology consists in large part of the transition from instructive theories to selective theories. See Massimo Piattelli-Palmarini, 1989, "Evolution, Selection and Cognition: From 'learning' to parameter setting in biology and in the study of language", in Cognition, 31, 1-44. See also Massimo Piattelli-Palmarini, 1986, "The Rise of Selection Theories: A Case Study and Some Lessons from Immunology", in W. Demopoulos and A. Marras (eds.) Language Learing and Concept Acquisition, Norwood: Ablex Pub., pp. 117-130.

Kobes, 1989, p. 13.

Ibid.

Compare Putnam, 1988, p. 2 and p. 109. The lack of a God's eye access to semantic categories also seems to be the point of many of Wittgenstein's examples -- including his well known discussion of 'slab' in the Philosophical Investigations.

Putnam, 1988, p. 119.

Compare Kobes, 1989. p. 13.