Published in 1993: "Nonconceptual Content and the Elimination of Misconceived Composites!", Mind and Language, volume 8, number 2, Summer 1993, (pp. 234-252)
Nonconceptual Content and the Elimination of Misconceived Composites!
(1) How good are the arguments against eliminativism?
Much of the critical discussion of eliminativism has been ineffective because it has been unclear as to its target. Powerful arguments ('self-defeating!', 'scientistic!') have been made against claims that eliminativism does not make, and only weak arguments ("put up or shut up") against claims that are made by eliminativism. In section (1) I show how its critics have tacitly assumed that the only kind of representational content is propositional or conceptual content. They have therefore begged the question against eliminativism's argument to substitute embodied representation for propositional meaning and for the propositional attitudes. I then outline a form of eliminativism which is not only not self-defeating — it is probably correct.
What does eliminativism propose to eliminate? There are three quite different alternatives:
(1) Folk psychological or commonsense psychological practice
(2) Content or meaning or belief or rationality ...
(3) The purported referents of Propositional Attitude theories of psychology
According to the first alternative, eliminativism proposes to eliminate the practice by which people interpret each others behaviour, and generally find our way in the human world. 'Folk-psychology' in this sense is not a theory but a highly sophisticated practice by which we locate ourselves and others in the community-based activities of communication, responsibility, duty and trust. To propose to eliminate such a practice cannot be false (it has the status of a political proposal not an empirical judgement), but it would be grotesque. However, I know of no sustained attempt to defend such a construal of eliminativism. (Were there to exist an adherent of such a doctrine, I would rather refer them to a psychiatrist than a philosopher).
The second alternative (to eliminate content, etc.) is not so much grotesque as impossible (short of armageddon). Meaning, content, belief and rationality are in the world and likely to remain so. As a philosophical proposal, eliminativism involves a claim about the eliminability of certain theoretical entities. But meaning and content are part of the ontology of the world, and can be neither reduced nor eliminated nor vindicated. The elimination of meaning and content would be like the elimination of witches. Witches are alive and well and on the streets of London (Luhrmann 1989); their elimination would land one in jail (or—in America—result in one's own elimination). Rather, it is a witch-theory-of-schizophrenia etc. which is abandoned in favour of a medical theory.
What it is proposed to eliminate is the third alternative: the purported referents of a particular theory of psychological practice. To claim that meanings are propositions and that beliefs are propositional attitudes is to make contentious theoretical claims about the nature of meaning and belief. Eliminativism—the real eliminativism—eliminates meanings-as-propositions and beliefs-as-propositional-attitudes by rejecting the propositional theory of meaning and the propositional attitude theory of belief. In the sense of propositional attitude theory, there are no beliefs and meanings. These theoretical referents are eliminated in favour of the entities of other theories of meaning and belief: theories of nonconceptual meaning and content, as carried by neuroscientific or computational or activity-theoretical representational vehicles.
Most of the arguments marshalled against eliminativism do not carry any weight against this third construal of eliminativism. Let us consider briefly the arguments reviewed in the target article: (1) that folk-psychology is not a theory vulnerable to being falsified by science, (2) that there are no alternatives to propositional attitude theory, (3) that eliminativism is self-defeating, (4) that eliminativism cannot explain "rational, stimulus-independent behaviour" and (5) that the consequences of eliminativism are epistemically too extensive and too disturbing to be accommodated. My purpose here is not to consider all the details of these arguments, but to point out how they all depend on a simple failure to understand what it is that eliminativism proposes to eliminate and what it is that eliminativism proposes to substitute.
The first argument is that eliminativism rests on the assumption that "commonsense propositional attitude psychology is a protoscientific empirical theory that purports to explain the nature of mental phenomena" and this assumption is false. But whether eliminativism rests on this assumption depends on what is meant by the phrase "commonsense propositional attitude psychology" [Hannan, p.3]. Sometimes it appears that what is meant by the target article is a way of structuring our everyday practice: "commonsense propositional attitude psychology exists for practical purposes of facilitating human interaction" [Hannan, p.9]. It would consist of ways of behaving, words that we utter to each other and to ourselves, and some rules-of-thumb, analogues in the psychological realm of folk-physics 'judgments' of the form 'you can't push over houses'. As Hannan points out, this is no theory, and its elimination would be immoral. (Of course!) But the problematic assumption employs the phrase "propositional attitude psychology", and this refers to something quite different from commonsense practices and rules of thumb. It refers to a philosophical and psychological theory of what beliefs and meanings are. I will say more about this below, but the theory includes the claims that beliefs, desires and intentions are attitudes to propositions, so that a sentence ascribing a belief, etc. to a person can be analysed as a believing, etc. relation between a person and a proposition, and that propositional attitude states play a causal role in the production of behaviour. Propositions are either Fregean Thoughts, Russellian propositions, syntactic forms or else other abstract objects identified by semantic theory in the tradition that runs from Frege, Russell and Carnap to Montague, Davidson and situation semantics. Eliminativism does assume that "Propositional Attitude Psychology" refers to a cluster of theories that model our everyday practice of finding our way in the human world of communication, belief and rationality. Eliminativism is the claim that this cluster of theories provide such poor models that its theoretical entities should be eliminated.
When Paul Churchland speaks of 'folk-psychology' he is talking about the body of theory that governs "the 'Propositional Attitudes', as Russell called them" [Churchland 1981, p.70]. The propositional attitudes are said to be analogous to the 'numerical attitudes' in the physical sciences: the analogue of the quantitative relations amongst the numbers referred to in a statement of numerical attitude is the logical relations of entailment, equivalence, etc. amongst the propositions referred to in a statement of propositional attitude. In other words by 'folk psychology' he means an account which models psychology on logic. It is unfortunate that many people have been misled by the use of the phrases 'folk-psychology' and 'commonsense psychology' into supposing that the elimination of logical models of psychology entails the elimination of our ordinary practice of finding our way in the cognitive and social worlds. There are excuses: Churchland says, for example, that "beliefs and desires are of a piece with phlogiston, caloric and the alchemical essences" [1989, p.125], but a reading of the context always makes clear that what he means by 'a belief and a desire' is a believing attitude to a proposition and a desiring attitude to a proposition.
Against this it can be argued that the propositional attitude theory is a recent development going back no further than Frege, whereas Churchland says things like "The folk psychology of the [ancient] Greeks is essentially the folk psychology we use today" [1981, p.74]. But here I think the idea is that although the ancient Greeks didn't themselves formulate propositional attitude theory, it makes sense for us, informed as we are by that theory, to interpret their inscriptions about human behavior as manifesting a loose and tacit theory of the mind that, if made articulate and rigorous, would resemble our theory of the propositional attitudes. I don't know whether this is correct about the ancient Greeks, but then I have no wish here to go further into the past in my endeavours of historical hermeneutics than the inscriptions of the Churchlands. The point about them, it seems to me, is that they are struck by the insensitivity of most philosophical talk in the present and in the past to the ways in which meaning and belief are embodied. What is most characteristic of propositional attitude theory is that meaning and belief are analysed in terms of abstract propositions whose structure is wholly independent of the structure of embodiment. It is a reaction to this anti-naturalism which drives the claims of eliminativism. So the core claim of Churchlandish eliminativism is that we should reject those theories of behaviour (and eliminate their posits) which employ notions of content which are explanatorily independent of embodiment, and that therefore we should reject propositional attitude theory. If it is propositional attitude theory of our folk-practice which is eliminativism's target, rather than a loose collection of practical folk-platitudes, then the argument that 'folk-psychology' is not vulnerable to being shown false by 'natural philosophy' has no force.
While the first argument mistakes what is to be eliminated, the other four arguments mistake what theory they are to be eliminated in favour of. The view amongst eliminativism's critics seems to be either that there is no theory of meaning and belief to substitute for the propositional attitude theory, or else that the eliminativist aims to reject all theories of meaning and belief (ie., all theories of representational content). Both of these views are false. It is no part of the Churchlands' program to engage in the dubiously coherent exercise of rejecting all possible theories of representational content. Their neuroscientific and computational accounts of mental activity are explicitly and intentionally shot through with representational notions (eg., P.S. Churchland and T. Sejnowski 1992 and P.M. Churchland 1989, ch. 5 and chs. 9-14). Consider, for example, how Paul Churchland responds to the 'self-defeating' charge in (Churchland, 1984, p.48): "the hole in this argument is the premise concerning the conditions necessary for a statement to be meaningful. It begs the question. If eliminative materialism is true, then meaningfulness must have some different source". That is, meaningfulness is preserved because it has a source other than in the propositional attitudes. Hence worries about whether eliminativism is self-defeating, whether it can provide explanatory generalizations that cover stimulus independent behaviour and whether it entails the impossibility of our 'cognitive lives' are at best premature. We need to look carefully at the alternative 'source' of meaning, and thereby determine what norms and generalizations it can support. It is quite wrong to suggest independently of such an analysis, as Hannan reports Rudder Baker as suggesting, that taking eliminativism seriously involves abandoning the idea that there is a difference between lying and telling the truth, between an intentional act and an accident, even the idea that there are persons. These consequences would follow if eliminativism abandons representational content, but it does not, so we cannot yet conclude that they do.
How can these misunderstandings occur so often? One answer is that we have allowed ourselves to assume that the only theory of representational content is the conceptual theory of propositional content. After all, where are the alternatives? Thus Hannan repeatedly slides between content and propositional-attitude-content: for example, "In order for rational, stimulus-independent behavior to occur, the meanings of stimuli must be internally represented. And, if the meanings of stimuli are internally represented, then there are internal states appropriately characterised as having propositional content" [p.17]; also, "we are constrained by the very concept of cognition ... to recognize the existence of causally-relevant, representational states of persons (propositional attitudes)" [p.19, my emphasis]; and "As things stand, propositional attitude concepts appear to be not only essential to our self-conception, but clearly exemplified. According to the only conceptual scheme we have, people obviously possess contentful mental states and act rationally" [p.21]. Given this assumption, it would be right to conclude that eliminativism entails the elimination of meaning and belief. (If there is only one theory of content, which is rejected and its referents eliminated, then there would be nothing left for meaning or belief to be.) The assumption also accounts for the typical tacit slide between structured practice and theory which is indicated by the hybrid phrase "commonsense propositional attitude psychology", which we encountered in the first objection to eliminativism. When a theory presents itself as the only theory of a subject matter, one has to work much harder to keep in mind the ontological difference between the theory and the practice of which it is a theory. The very success of the formal tradition in logic has blinded us to the fact that this tradition explores only a tiny sub-space of the whole territory available to the theory of representational content. My moral, then, is that any good contemporary discussion of eliminativism should discuss non-propositional and non-conceptualist conceptions of representational content in order to discover which of the different theories of representation give the best account of human psychological practice and whether propositional attitude theory is explanatorily autonomous from the level of nonconceptual content, or is reduced to it, or is eliminated in favour of it. I shall try to follow my own moral.
Another explanation for why eliminativism has often been misinterpreted has to do with an incorrect assimilation of two quite different components in the work of the Churchlands: the argument for the elimination of beliefs and meanings as propositional attitudes, and the suggestions for how our ordinary psychological practice may be modified in the light of discoveries, metaphors and artefacts employed by the neurosciences, by the cognitive sciences and by our increasing philosophical understanding of the nature of representation. The argument for the elimination of beliefs-as-propositional attitudes has been confused by critics of eliminativism with the claims about the modifiability of our ordinary practices and ways-of-thinking to yield the outlandish claim that somehow our ordinary practice and understanding should be eliminated. Consider just one example of how badly wrong this gets the position of the Churchlands. In (1979, pp. 30-36) Paul Churchland describes a modification of our ordinary practice of observation of the night sky given our understanding of the Copernican theory of the arrangement and motions of the solar system. After providing a sort of manual by which the reader can attempt to be "more at home in our solar system", he writes that "what I do advocate at the social level is that we do what we can to assist the concepts used in common sense to evolve towards whatever counterparts they may have in the wider conception that science has provided ... presumably this is precisely what common sense has always been doing — evolving in pursuit of the ever-advancing front of new and successful theory — so I am advocating only that we assist a process that has been underway for many millennia" [his emphasis]. This overstates his case, because the modifications should be conceived of more symmetrically: commonsense practice in the light of theory, but also theory in the light of our evolving commonsense practice. However, even in the overstatement, it is emphatically clear that what is envisaged is not elimination but modification of practice.
The insight behind the claim that our ordinary practice and understanding is modifiable by science is that our folk thought is not and should not be insulated from our more institutional (scientific and philosophical) thought. One sometimes detects an almost fundamentalist disposition amongst the critics of eliminativism to wall-off our 'folk psychology' in a protected explanatory space where it cannot be 'threatened' by science. However stagnant folk-psychology may, or may not, have been in the the past, it would surely become so in the future if eliminativism's critics were to have their way. Worse: it would become a dangerous model for how a realm of knowledge and practice can gain an illegitimate authority. The authority of "walled-off folk-psychology" would derive from boundaries established between its truth and the truth which belongs to the rest of culture on the far side of the boundary. This contrasts badly with democratic authority that derives internally to a practice because that practice can sustain itself in open communication with the rest of intellectual culture (instead of deriving externally from the boundary walls themselves). In order to maintain open communicative exchange in our culture we must temper the pressures towards explanatory pluralism with the pressures towards unity (explanatory exchange), as well as tempering the pressures towards unity with the pressures towards pluralism. Elsewhere, I consider some arguments against what I take to be an excessive pluralism.
One of the great virtues of eliminativism has been the encouragement which it has provided to the search for naturalistic theories of meaning and psychological practice that are radically unlike "propositional theory". It is false to claim, as the target article does in several places, that "we have absolutely no idea what it would be like to describe human action, and human life as subjectively experienced, without propositional attitude concepts". The nonpropositional, nonconceptual theories of representation provide just such alternative conceptions. There are two components to nonpropositional theories of representation: theories of representational vehicles which are typically highly unlike sentential vehicles, and theories of representational contents which are unlike propositional contents. Thus, the Churchlands and Paul Smolensky  amongst others have explored accounts that appeal to representational vehicles such as connectionist vectors and vector-spaces, gradient descent through weight / error space and partitions of activation-vector space. And Gareth Evans , Christopher Peacocke [1989, 1992a & 1992b] and Adrian Cussins [1990 and 1992b] amongst others have explored accounts of nonconceptual contents which are experiential modes of presentation whose structure is dependent on how they are embodied in animals and embedded in the physical and social environment. Nonconceptual contents and multi-dimensional phase-spaces fit well together to provide a multitude of interesting ways to modify and extend our ordinary understanding of "human life as subjectively experienced". If eliminativism is ultimately to withstand the self-defeating charges and the charges of 'calamitous consequences', then it must combine theories of non-sentential vehicles with theories of nonconceptual contents to reveal new ways in which for animals like us, in environments like ours, a world is presented to subjects of experience.
(2) Of Conflations and Misconceived Composites
In section (3) I will provide a thumbnail sketch of my favoured theory of nonconceptual content in order to show that conceptualist propositional contents should be eliminated in favour of a quite different conception of our conceptual capacities. But first I want to consider briefly in section (2) what is involved in the posits of one theory being eliminated in favour of those of another rather than being reduced to them. This is a large topic in the philosophy of mind and the philosophy of science, so I shall not attempt to do justice to it here. But I do want to sketch what I hope will be an intriguing suggestion of which more can be made elsewhere.
What distinguishes elimination from reduction? What is the difference between the relation that Phlogiston theory bears to Lavoisier's theory and the relation that our understanding of temperature in gases bears to the theory of mean molecular kinetic energy in gases? The accounts of the distinction between elimination and reduction in the literature are disappointing. Consider, for example, what Ramsey, Stich and Garon (1990) say:
There is, in the philosophy of science literature, nothing that even comes close to a plausible and fully general account of when theory change sustains an eliminativist conclusion and when it does not. In the absence of a principled way of deciding when ontological elimination is in order, the best we can do is to look at the posits of the old theory — the ones that are at risk of elimination — and ask whether there is anything in the new theory that they might be identified with or reduced to. If the posits of the new theory strike us as deeply and fundamentally different from those of the old theory, in the way that molecular motion seems deeply and fundamentally different from the "exquisitely elastic" fluid posited by caloric theory, then it will be plausible to conclude that the theory change has been a radical one, and that an eliminativist conclusion is in order. But since there is no easy measure of how "deeply and fundamentally different" a pair of posits are, the conclusion we reach is bound to be a judgment call.
No doubt it is a 'judgment call', but it would be nice to gain a better theoretical grip on what it is that underlies such a decision, what it is that would make the judgment true rather than false. (One might wonder why DNA is not 'deeply and fundamentally different' from the Mendelian conception of a gene, or why wave-motion is not 'deeply and fundamentally different' from sound.) Often it is said that theories whose posits are eliminated are 'very badly false'; more so than theories whose posits are reduced. But this gets things back to front: we say that a theory is very badly false because it exhibits those characteristics that, in relation to another theory, lead us to eliminate its posits. We have no independent grip—at least in the interesting cases of theories which exhibit good predictive success—on 'degree of falsity'. What, then, are these characteristics?
I want to distinguish relations between an early theory and a successor theory where some of the concepts of the early theory are conflations of concepts of the successor theory from relations between two theories where some of the concepts of the early theory are misconceived composites of concepts of the successor theory and / or elsewhere. My hypothesis is that where there are only conflations, the early theory may be reduced to the successor theory. But where there are misconceived composites, the early theory is eliminated in favour of the successor theory.
A conflated concept is a semantically unstructured joining together of two or more concepts which are from the same explanatory domain (where which concepts belong to the same explanatory domain is determined from the point of view of the successor theory) and which subserve the same or linked functional interests. The use of conflated concepts often leads to imprecision in science but can be a useful feature of discourse where more structural precision would be inappropriate. One example is the everyday notion of weight which is a conflation of the notions of mass and force. Mass and force both belong to the same explanatory domain from the point of view of mechanics, and subserve common functional interests in equations of the form, F=m*a. Thus, a conflation of mass and force yields imprecision, but not incoherence. I use the ordinary term 'conflation' here because whether it bears a pejorative sense depends on which community is using it: for philosophers conflation is bad news, but for historians and lexicographers it is an aspect of their skill. Thus whether the conflation of weight is a matter for reprimand depends on whether it is employed within a community of physicists or a community of weight-watchers. We speak of the reduction of weight to force and mass because the more basic terms reveal useful structure which is invisible to a weight perspective. From the perspective of physical theory we may criticize the concept of weight as imprecise, whilst recognizing that that degree of vagueness may be appropriate, even useful, in the context of concerns which are remote from those of physics. (Compare our use of indexicals like 'around here'). But we don't criticize it as misconceived or illegitimate, a joining together of aspects of concepts which do not subserve the same explanatory purposes.
Contrast misconceived composites where the early theory employs concepts which are a combination of components of other concepts which come from distinct explanatory domains, or which subserve quite different functional interests, and are therefore illegitimate from the perspective of the successor theory. In these cases the composite conceptions lead us to form arguments (or engage in practices) which are conceptually confused (from the successor perspective), not merely imprecise or insufficiently revealing of microstructure. I will mention a number of plausible examples of misconceived composites, although no doubt a more detailed analysis of any one of the examples might lead to a different diagnosis of how the concept fails. Consider first the phrenologist's conception of physiology (or mental organs). I shall follow the account in Shapin (1979), according to which the formal codified corpus of phrenology owed its origins to the work of Franz Joseph Gall (1758 - 1828) and his associate Johann Gaspar Spurzheim (1776 - 1832). The brain was held to be a "congeries of organs, topographically distinct, each of which subserved a distinct mental function". There were from 27 to 33 mental organs such as organs of 'adhesiveness' (clinginess to surrounding objects, such as one's house); of 'hope' and of 'philoprogenitiveness' (the tendency to procreate and nurture children). The size of the cerebral organ was a measure of the power of its functioning so that a small organ of 'tune' might render its possessor insensitive to melody. The contours of the skull followed the variations in the sizes of the underlying cerebral organs so that one could make a mental diagnosis on the basis of visible bumps and depressions in the skull. Shapin describes the 'physiological' relation between structure and function in phrenology as follows:
Three sources for the organology were repeatedly mentioned by Gall and Spurzheim. Gall himself liked to emphasize how he observed clear correlations between the behaviour of his school-mates and their external cranial contours. Later on, he and Spurzheim displayed busts and portraits of individuals celebrated for certain traits and abilities, pointing out their possession of the appropriate 'bumps'. Thus engravings of Chaucer ('ideality' large) and Locke ('ideality' small) were crucial visual confirmation of the organology. The second alleged line of research derived from cranial comparative anatomy. The skull shapes of various animals were displayed as evidence of their accurate reflection of the beasts' 'known' psychic and behavioural attributes. Thus, the 'bump' for amativeness is appropriately large in rabbits; that for 'cautiousness' is well-reflected in the crania of birds. And, finally, 'knowledge' of racial and sexual traits was also brought to bear in the construction of an organology. 'Veneration' was found to be large amongst the superstitious and credulous Negroid races; 'amativeness' was well-known to be small in women. This sort of evidence, which correlated structure and function and made each illuminate the other, the phrenologists liked to term 'physiological', and it was by physiological, and not by anatomical, investigations that the enumeration and mapping of organs was achieved. The skull did not require to be opened up and the brain displayed in order for the system to be established. [Shapin, 1979, pp.142-3]
The phrenologist's concept of physiology is a misconceived composite of notions linking brain anatomy with a mix of psychological, social and evaluative notions such as hope, veneration, ideality, cautiousness, adhesiveness and amativeness. This concept of physiology is such that its acceptance would render reasonable the diagnostic use of the three kinds of evidence described in the quotation. But from the perspective of a successor neuropsychology the diagnostic use of these three kinds of evidence is not reasonable: the two sets of concepts employed in the phrenologist's notion of physiology neither serve linked explanatory goals nor share a common functionality. In this example, they do not, as they are intended, stand to each other as structure and function, for the successor theory treats them as belonging to distinct explanatory domains. That degree of veneration or degree of caution should be reflected in brain organ size in the way required by the phrenologists borders on incoherence from a successor perspective, and the arguments that were made about the social (eg. educational) consequences of phrenology are radically misconceived. The failure here is not merely a failure to be precise or revealing of structure, and hence the phrenologist's conception of physiology is an appropriate target for eliminativism.
Some more examples: A familiar idea is that an epistemologically robust sense of reality takes our knowledge and interaction with middle-sized dry material goods as being paradigmatic. Our notion of, and access to, reality should, according to this idea, be modelled on our notion of and access to material objects like chairs and trees. But sometimes in philosophy one comes across a notion of the physical which is a misconceived composite formed from this notion of the material and also from the notion of being characterized in the physical sciences. Two quite different conceptions of the physical, which come from distinct explanatory domains, are joined together in a way which licenses misconceived claims by concealing the argumentative transitions on which such claims would have to be based. Thus the misconceived composite notion of the physical allows us to argue as follows: An epistemologically robust sense of reality takes the physical as paradigm, and the physical is explicated in the physical sciences, therefore an epistemologically robust sense of reality takes the physical sciences as paradigm. Such a misconceived philosophical notion of the physical is ripe for elimination.
Or consider mermaids: Nowadays, mermaids are mostly fictional, mythical creatures. Given our contemporary conception of mermaid, a voyage of exploration to discover marine animals that are mermaids is not merely unlikely to succeed, for it involves a conceptual misunderstanding. But 200 years or so ago the mermaid concept was not like this. At these times the concept was a misconceived composite of mythical and scientific elements, as judged and described from our successor perspective. Stories of a fish-tailed siren "provide an eloquent expression of latent eagerness to believe in one of the most romantic myths created. Born from the 'Mother and love of men, the sea' and embodying a sexual fantasy of universal appeal, the mermaid is clearly too glamorous a creature to be understood by the laws of science alone. Forever alluring, yet never to be possessed by a living man, her reality is embedded deep in the collective unconscious, the magnetic focus of a ceaseless inchoate longing to dare the wild Unknown" [Phillpotts, 1980]. For someone employing the misconceived conception, it makes good sense that a marine voyager might come across a sea mammal which is a mermaid in the sense of the mythical animal "embedded deep in the collective unconscious". How shocking that would be! And people have 'identified' surfacing sea cows (dugongs and manatees) and basking seals as mermaids.
Or, to use an idea of Brian Smith's, recent work in non-linear dynamics and chaos theory has shown the need to pull apart two components in the familiar philosophical notion of determinism: the idea of being patterned or non-random and the idea of being predictable. Chaos theory reveals that philosophical notion to be a misconceived composite because it shows how the ideas of non-randomness and of predictability subserve quite different functional interests. It is therefore misconceived to use the concept of determinism to license the inference from the non-randomness of the world to the predictability of the world. (It might be that the world itself is the most efficient 'predictive' algorithm for physical effects such as the weather.) One could also argue from a feminist successor perspective that the notion of Man is a misconceived composite formed from the idea of universal humankind and the idea of certain masculine qualities. It licenses illegitimate inferences to the form or nature of humanity. And again the proper response is eliminativist: in that sense of Man, there is no Mankind.
(3) The Nonconceptual Construction of Concepts
I argue that the successor theory of representational content shows that the propositional or conceptualist notion of concept is a misconceived composite of three notions which have quite distinct functional purposes, and hence it conceals illegitimate inferential transitions. Conceptualist concept should therefore be eliminated. Given that the conceptualist nature of content is a fundamental assumption of propositional attitude psychology and given the claim in section (1) that the eliminative materialism of the Churchlands is targeted at propositional attitude theory, it follows that the argument of this section for the elimination of conceptualist content is an argument for the more general eliminativist conclusion.
Content is a mode of presentation of the world in experience or in thought, and is therefore normatively assessable. Perception (or memory, imagination, 'thinking', ...) typically carries both a conceptual and a nonconceptual content. The conceptual content is the content which is canonically specified in terms of the realm of reference of the content, and the nonconceptual content is the content which is canonically specified in terms of the realm of embodiment of the content. The realm of reference of the content is that domain of objects, properties and states of affairs with respect to which the content is evaluated as being correct or incorrect, true or false, accurate or inaccurate, satisfied or not. The realm of embodiment of the content is that range of skills, abilities, dispositions or activities in virtue of which the animal-in-its-environment grasps and manifests the content. Conceptualist theory of content includes the claim that realm of reference specifications of content are explanatorily basic and can be understood independently of realm of embodiment specifications of content. Nonconceptualist theory includes the denial of this claim. A concept is an objective content and is therefore either assessable as true or false, or is a constituent in a content which is assessable as true or false. Conceptualist and nonconceptualist theories give different accounts of what concepts are. Since conceptualist theory takes realm of reference specifications to be explanatorily independent of realm of embodiment specifications it has the consequence that what it is for a content to be true or correct can be understood and specified independently of the nature of the embodiment of the content. This consequence is a crucial component of propositional attitude theory, and provides much of the motivation for the eliminativist's rejection of propositional attitude theory.
One aspect of this motivating naturalism is the idea that representational theory should not begin with the notions of truth and reference in place. We can begin only with brain processes and environmental, historical and social activities. I have developed an account of representational theory according to which these processes and activities can be understood as representational without having to presuppose some prior notions of truth and reference. The naturalist's task is to explain what truth and reference (the world!) is by means of an understanding of embodied animal activity; to explain a conceptual scheme as an adaptive ecological system of animal and environment. To presuppose the conceptual resources of late twentieth century, western intellectual human culture (folk and scientific) in explaining the nature of cognition and rationality would be to assume that our very parochial, contingent animal ecology is the center of the explanatory universe, and would thus be to give up on naturalism's rejection of the privilege of 'our' perspective. That is, for our theory of objectivity in representation to presuppose the conceptual resources by means of which the world is given to us would be to give up on the explanatory ambitions of naturalism. But, equally, to confine oneself to non-representational notions in the description of activity is to render impossible the naturalist's ambitions. The solution is to provide an account of activity in terms of nonconceptual representational content: a non-referential mode of experiential access. By making the primary mode of content be the experiential accessibility of embodied skills and ways of negotiating the environment, the nature of our embodiment is no longer invisible (as it is for the conceptualist) because located at the point of origin of our vision. If we can explain how conceptual contents are constructed from nonconceptual contents, then we can explain how our conceptual scheme is an out-growth of our mode of embodiment.
Truth and reference are properties of concepts: objective contents. Beginning, then, without truth and reference in place in our ecology, is beginning without objectivity in place. Objectivity is the gap between mind and world, between representation and represented in virtue of which it is possible to be wrong or in error, and therefore in virtue of which it is possible to be right. Thus naturalism requires that the explanatory base for our account of activity is representation for which there is (as yet) no gap or divide between mind and world: a field of activity which is partially both mind and world because it is not initially either yet develops into both. Or, better, a field of environs - mental activity. The character of any cognitive phenomenon may be described as the negotiation or structuring of this field of environs - mental activity. We may plot the course of this structuring by means of two variables: what I call "pd ratio" and "stabilization". PD ratio is a measure of how well the field of activity supports way-finding through the field. In a jungle or a prairie, way-finding is supported by trails through the territory. Trails are a function of both the landscape itself and also of the passage of creatures through it. So we can use the idea of trails through environs - mental activity ("cognitive trails") to mark the idea of a structuring of the field of activity which is not thought of as being due to independent mind and world components: a stored 'map' within the mind of the creatures, which allows them to navigate through a territory of mind-independent features, some of which are marked on the map. Rather, the navigational ability can't be assigned to either mind or world (or some combination of mind and world components) because it consists in the network of trails themselves. PD ratio, then, is a measure of how extensive the network of cognitive trails is with respect to the whole extent of the field of environs - mental activity. If the network covers most of the field then the PD ratio will be close to 1, and if the network is fragmentary then the PD ratio will be close to 0.
If a network of cognitive trails is unstabilized, then the intersections of trails must be constantly negotiated and renegotiated. This is like a troop of baboons each member of which must repeatedly renegotiate their kinship, friendship and sexual relationships by means of the changing spatial relations amongst all the members of the troop, because these relationships are not marked by names, labels, clothing or other visible, stable signs. It involves an immense amount of hard, real-time experiential work. Compare negotiating a business deal in a board room where all the principal banks, suppliers, customers, shareholders and management is represented, and where none of the parties can count on prior established relationships of friendship, trust or obligation. Stabilization of the network of cognitive trails provides a way of putting a 'black box' around some part of the network of relations so that part of the network stays put whilst one works on renegotiating some other part. Signs, institutions, the relative stability of the physical world, reliable artefacts and public language (our technology) do this work of stabilization for us. Landmarks in the network of trails are partially stabilized regions by means of which navigation around the field of activity is enhanced. (Likewise, whole networks can be stabilized, as in 'relativity theory' or 'F=m*a'). Thus the structuring of environs - mental activity is a function of both pd ratio and degree of stabilization.
The course of a cognitive phenomenon (a dynamic, representational activity) may be plotted on a graph whose axes are the pd ratio of the cognitive trails and the degree of stabilization of the cognitive trails. Let us suppose that an activity starts out with low pd ratio and low stabilization. As the field starts to become structured — the creatures start to find their way around a landscape (as the theorist would say) — pd ratio will increase. A network of cognitive trails is temporarily established, and this provides for the possibility of stabilization. Both stabilization and pd ratio continue to increase, until the work concentrates almost entirely on the stabilization of trails that are in place. However, once a network of trails is tightly stabilized it becomes less flexible, and as the nature of the field of activity changes over time, pd ratio will start to decrease as stabilization increases. Further improvement in way-finding will then require that a stabilized region of cognitive trails be destabilized for a period of time in order to allow pd ratio to increase again. In other words, virtuous representational activity is the effective trade-off of the relative merits and demerits of pd ratio and stabilization. Virtuous activity may itself be represented as a figure, a shape, in the two dimensional space of the pd ratio / stabilization graph. It is not hard to see that the virtuous form of representational activity has the shape of a spiral:
What, in terms of this diagram, is a concept? There are three main contenders: (a) the highly stabilized cusps at each layer of the spiral, (b) a point in the graph which has the maximum value on both dimensions (or else, that point on the graph which is such that the line from the origin to it is longer than all other lines from the origin to other points on the graph), and (c) the virtuous shape of the graph itself. Each of (a), (b) and (c) has quite different functionality, yet each has a claim to be the representation of a concept. Often what we folk-call a concept functions like a stabilized cusp. A word or a phrase or a sentence has a meaning which is linguistically highly stabilized: appropriate use of or response to the linguistic item does not require that we go inside the nonconceptual content of the meaning and renegotiate its form. We say that the word expresses a concept, and we allow our analysis of meaning to stop with atomic concepts. But if the cusps of each layer have a claim on conceptuality, then surely so does that point of the graph which maximizes not only stabilization but pd ratio as well? Yet we also want our notion of conceptuality to capture all that is virtuous in meaningful representational activity, and we know that maintaining a highly stabilized condition for too long may be counterproductive. Fundamentalist ideologies are just such over-stabilized cusps, where a very high price is paid in inflexibility for the virtues of maintaining old, familiar forms. Perhaps, then, the virtuous shape of the whole spiral should be treated as the representation of a concept?
What we have here is an example of that familiar phenomenon whereby a successor theory (in this case a successor theory of representational content) identifies a more basic level of description which reveals structure invisible to the earlier theory. Where conceptualist theory operates with just one notion of concept, nonconceptualist theory identfies three quite different notions, each of which answers to some aspect of the conceptualist notion of concept. So we must ask: is the conceptualist concept a conflation of the nonconceptualist notions, or a misconceived composite of them?
That it is a misconceived composite, and should therefore be eliminated, can be seen from the following: Within conceptualist theory, conceptuality—if measured at all—is always measured along one dimension, usually called "generality". Stabilization and pd ratio are thus treated as equivalent, so reduction in generality (conceptuality) is always equivalent to reduction in the contentfulness of the phenomenon. No gap opens between concept and content, so that it is hardly surprising that we so often find the slide referred to earlier between content and propositional attitude content. The very character of conceptualist theory impedes us in our search for alternative theories. But where the notion of concept is formed as a composite not only from maximum stabilization and maximum pd ratio, but also from the virtuous shape of diachronic representational activity, the consequences are not merely unfortunate but dangerous. The misconceived composite licenses the illegitimate transition from 'greater conceptuality is always to the representational good' (where conceptuality is read as virtuous form) to the conclusion that 'greater stabilization is always to the representational good'. But, as we saw, with stabilization we can easily have too much of a good thing: excessive stabilization is an ideology. And the maintenance of ideology is hegemonic and exclusionary.
Let us not transform 'folk-psychology' into an ideology!
Boghossian, P. 1990: "The Status of Content", The Philosophical Review, vol. XCIX, no.2 (April 1990), pp.157-184
Churchland, Pat 1986: Neurophilosophy: Toward a Unified Science of the Mind/Brain, Cambridge: M.I.T. Press
Churchland, Paul 1979: Scientific Realism and the Plasticity of Mind, Cambridge University Press
—— 1981: "Eliminative Materialism and The Propositional Attitudes", Journal of Philosophy 78, no. 2 (1981); also in Churchland (1989); and in Mind and Cognition: A Reader, (ed) Lycan, W., Blackwell, 1990
—— 1984: Matter and Consciousness, Cambridge: M.I.T. Press
—— 1989: A Neurocomputational Perspective: The Nature of Mind and the Structure of Science, Cambridge: M.I.T. Press, 1989
Churchland, P.S. and T. Sejnowski 1992: The Computational Brain, Cambridge: M.I.T. Press
Crane, T., 1992: "The Nonconceptual Content of Experience" in Crane (ed) The Contents of Experience: Essays on Perception. Cambridge: Cambridge University Press
Cussins, Adrian 1990: "The Connectionist Construction of Concepts" in M. Boden ed. The Philosophy of Artificial Intelligence, in the Oxford Readings in Philosophy Series, Oxford: Oxford University Press, pp.368-440
—— 1992a: "Limitations of Pluralism" in D. Charles and K. Lennon eds. Reduction, Explanation and Realism. Oxford: Clarendon Press, pp. 179-223
—— 1992b: "Content, Embodiment and Objectivity: the theory of Cognitive Trails", Mind, 101, October 1992, pp. 651-688
Evans, G. 1982: The Varieties of Reference, Oxford: Oxford University Press
Hannan, B. 1993: "Don't Stop Believing: The Case Against Eliminative Materialism," Mind and Language
Luhrmann, T.M. 1989: Persuasions of the Witch's Craft: Ritual Magic in Contemporary England, Harvard University Press
Peacocke, C 1989: "Perceptual Content" in "Themes from Kaplan", edited by J.Almog, J.Perry, H.Wettstein, pp.297-329, Oxford: Oxford University Press
——1992a: "Scenarios, Concepts and Perception" in Crane (1992) (ed) The Contents of Experience: Essays on Perception. Cambridge: Cambridge University Press
—— 1992b: A Study of Concepts, Cambridge: M.I.T. Press
Philpotts, Beatrice 1980: Mermaids, New York: Random House
Putnam, H. 1988: Representation and Reality, Cambridge: M.I.T. Press
Ramsey W; Stich S; and Garon J, 1988: "Connectionism, Eliminativism, and the Future of Folk Psychology" in Tomberlin (ed.) Philosophical Perspectives 4: Action Theory and Philosophy of Mind, Ridgeview
Rudder Baker, Lynne 1992: Review of "A Neurocomputational Perspective" by Paul M. Churchland, Philosophical Review, vol. 101, October, pp.906-908
Shapin, S 1979: "The Politics of Observation: Cerebral Anatomy and Social Interests in the Edinburgh Phrenology Disputes" in R. Wallis (ed) The Sociological Review Monography No. 27: On the Margins of Science: The Social Construction of Rejected Knowledge, University of Keele.
Smolensky, P. 1988: "On the Proper Treatment of Connectionism" Behavioral and Brain Sciences, 11:1-71, with continuing comentary from Behavioral and Brain Sciences, 1990 2:399-412
 I restrict myself to eliminativism in the general area of belief and meaning, and so don't consider eliminativism with respect to qualia, etc.
 It's not that I deny that there are philosophical positions with respect to which meaning and content are theoretical posits. For example, with respect to a debate over the Lockean view of the world as atoms-in-the-void, meaning is a theoretical posit. The debate over eliminativism with which I am concerned is not conducted at that level of philosophical granularity, but rather with respect to competing theories of meaning and belief. As I take the first question in the debate over eliminativism to be whether content is essentially embodied (that is, whether the theory of content is, or is not, explanatorily isolated from the theory of embodiment), I take content to be an ontological category with respect to the eliminativist debate. That is, the falsity of the atoms-in-the-void view is a presupposition of the debate. So while there is a debate over the elimination of content, that is not the debate of principle concern to the Churchlands, and both kinds of debate have been compromised by failing to clearly separate them: a failure that arises because of the tacit assumption of the uniqueness of the propositional theory of content.
 Compare Pat Churchland (1986, pp.280-1): "By making theories the fundamental relata, much of the metaphysical bewilderment and dottiness concerning how entities or properties could be reduced simply vanished". Likewise for theoretical posits and elimination.
 The 'words that we utter' include 'belief' and 'desire' as phonemically or graphemically identified. Nothing follows about the role of the propositional attitude, or any other, theory of belief and desire. Our practice may be modified over the centuries so that there may come a time when we do not use the words 'belief' and 'desire'. Again, nothing follows about whether there are beliefs and desires, and what the theory of behaviour should look like for them, or for us.
 So the criticisms of Lynne Rudder Baker (1992, pp.906-908) that Churchland "conflates views on the nature of knowledge and views on the mechanisms that encode it" and "confuses what is thought about (here, theories) with neural mechanisms that enable us to think about them" beg the question. Propositional attitude theory puts a lot of weight on the distinction between the representational content and its embodiment in different kinds of representational vehicles, and it has done this in order to protect our aspirations to the universality of truth. What it is that we know (the objects of knowledge) has to be kept explanatorily isolated from the ways in which animals like us are able to know (the media of knowledge) if our science and the rest of our 'high status' knowledge is not to be polluted by the contingencies of our psychology and our evolutionary past. In rejecting propositional attitude psychology, eliminativism denies that much weight can be put on the distinction between what it is that we know and how we know it. The open question then becomes: how much of our aspiration to the universality of truth can be recovered within a naturalistic theory of representation?
 See also Churchland 1981, p.89: "one cannot simply assume that particular theory of meaning without begging the question at issue". Nor is this a position which combines 'irrealism about mental content' with 'realism about linguistic content', as suggested by Boghossian (1990), p.170. This is an especially odd position to ascribe to the Churchlands, given the characteristically Churchlandish move to deny that linguistic representation is basic. Rather, the view is that mental content and linguistic content have a common source explained in a successor theory which eliminates the posits of the propositional theory of meaning and the propositional attitude theory of belief.
 He is much closer to a symmetric view in the present volume where he writes of the "intimate connection of theory with practice" and the skill-based nature of theory.
 Cussins, 1992a
 Eliminativism of the kind I am defending requires both a theory of representational vehicles and a theory of representational contents. Appeal to theories of nonconceptual content strikes me as the proper response to criticisms such as those of Rudder Baker that "Churchland offers no naturalistic account of what makes a given activation vector represent a particular environmental feature" (Rudder Baker, 1992, p.907).
 By 'reduction' here I mean that explanatory relation or relations, whatever it (they) turn out to be, between theories or their posits that hold in the 'classic examples of reduction': thermodynamics, temperature, sound, genes, etc. It may be that this relation (or, one of these relations) is best analysed not in terms of the traditional account of reduction, but in terms of some notion of 'construction' (Cussins, 1992a). Here I just wish to note that we will need an account of the reduction / elimination distinction however weak or strong our account of the classic cases of reduction.
 The reading of 'judgment call' according to which the decision between eliminativism and reduction is just indeterminate cannot be correct exegesis of Ramsey, Stich and Garon because they go on to give arguments that elimination, and not reduction, is the correct response to the propositional attitudes.
 This account presupposes an explanation of what makes one theory a successor to another. I am not aiming to provide such an explanation here.
 In 1842 an advertisement for a show at "Mr. P.T. Barnum's American Museum and Gardens" appeared, announcing that "the manager being ever desirous to gratify his Numerous Patrons ... has in accordance with Universal Desire! effected an engagement for a short time longer with the Proprietor of the most wonderful curiosity ever discovery [sic], the Real Mermaid! Which was exhibited during the past week at Concert Hall, in Broadway, and which elicited the wonder and amazement of Hundreds of Naturalists and other Scientific Persons ...
 See footnote 5 and surrounding discussion. The attempt to isolate content from embodiment entails conceptualism becuase it prevents the possibility of a nonconceptual level of analysis: see below.
 In (Cussins, 1990) I attempted to provide a neutral characterization of nonconceptual content that would apply equally well to the accounts given by Evans, Peacocke, Dretske and myself. Tim Crane discusses this characterization in Crane (1992). But in this brief summary I describe things from within my own perspective.
 In (Cussins 1992b) where I offer an argued explication of the theory sketched here. Since I think that the debate over eliminativism really comes down to a debate between those who take the space of representational theory to be small and well explored and those who take it to be very large and poorly explored, I think there is some value in giving even just a little sense of what ways of talking about representation that belong to the latter camp might look like.
 Because those concepts which are thoughts are the primary bearers of truth, (and given that I am assuming a Fregean-style tight connection between meaning and truth) a theory of concepts must also be a theory of truth. So there are resources here to answer Putnam's question to Paul Churchland (Putnam 1988, p. 60). And also to resist Boghossian's assumption that "contents just are truth conditions" or that content "consists simply in the idea of a truth condition" (Boghossian 1990, pp.173-4).
 See Evans's 'generality constraint' in Evans (1982).
 I would like to thank Pat and Paul Churchland, Charis Cussins, Stefan Heck, Joe Ramsey, Brian Smith and Steve Yalowitz for discussion and lots of help with the examples of misconceived composites. And Pat and Paul for their love of life.