The view that, because mental states and properties are items posited by a protoscientific theory (called folk psychology),  the science of the future is likely to conclude that entities such as beliefs, desires, and sensations do not exist. The alternate most often offered is physicalist and the position is thus often called 'eliminative materialism'.

Like its predecessor, the mind-brain identity theory, eliminativism claims that it is an empirical fact, rather than a conceptual necessity, that mental states are identical with brain states, and that this fact is justified only by scientific evidence.  But most historians of science now believe that scientific progress usually does not establish identities between the entities described by old theories and new ones.  The eliminativists argue that there is thus no reason to assume that such identities will be found when science develops a detailed alternative to the folk psychological view of mental states.
Most criticisms of eliminativism center around the claim that folk psychology is somehow different in kind from the theories that the eliminativists are advocating as replacements. Sometimes the argument is made that mental entities are different in kind from theories, because they are directly given to us, or are part of practical activities rather than theoretical discourse. It is also argued that folk psychology must remain autonomous with respect to physical or neuroscientific theories. Some justify claims of autonomy on the assertion that psychology cannot be reduced to neuroscience.  Others say that folk psychology is irreducibly normative, and because science is only descriptive, reducing folk psychology to a scientific theory would be attempting to derive an ought from an is.
Other criticisms are that the principles and scientific evidence used to justify eliminativism imply a pragmatist theory of truth, which does not support the traditional view of truth and reference that eliminativism presupposes.  Some pragmatist views of truth would require us to say that the posits of folk psychology must be true in some sense because they are so useful.  Also, some recent theories of reference claim that even if a theory is false, we cannot infer from this that what it describes does not exist, or that its terms do not refer.
Origins of the Debate
How Identity Theory gave rise to Eliminative Materialism
The identity theory of mind was first suggested  by E. G. Boring (1933), although it was made popular in philosophy by U.T. Place (1956) further developed by Place in dialogue with D.M. Armstrong (1965), J.C.C. Smart (1959) and others. Identity theorists claimed that mental states are really brain states. However, they did not claim that this was a conceptual truth. Identity statements between minds and brains are contingent identities, which we accept as true only  because scientific research tells us they true. However, it was far from clear that it was possible for science to prove that brain states were identical to mental states. Scientific research does discover causal connections between brain states and mental states, but connections are not identities.  As Shaffer put it "For one property to be reducible to another, they must be different; something cannot be reducible to itself" (p. 120 in Borst 1970). Feyerabend (1963) also pointed out that if the connection between material facts and mental facts is an identity, it has to be expressed as a biconditional. This biconditional " not only implies , as it is intended to imply, that mental events have physical features; it also seems to imply (if read from the right to the left) that some physical events, viz. central{brain} processes, have non-physical features" (p. 172 in Rosenthal 1971).
Furthermore, any claim that the mind-brain identity is a scientific fact has to take into account how science actually operates.  When we consider the most recent developments in the history and philosophy of science, eliminative materialism emerges as an alternative to the mind-brain identity. Eliminative materialism appeared to dissolve the problems with establishing identities between mental states and brain states , but it also created a whole set of new problems.
Eliminative Materialism and Philosophy of Science
In philosophy of science, arguments by Wilfrid Sellars and W.V.O. Quine prompted many philosophers to reject the idea that the properties of our mental states are directly given to us. These arguments criticize what Sellars called "the Myth of the Given" (Sellars 1963) and what Quine called a "Dogma of Empiricism". (Quine 1961)  Quine's arguments show that science does not rest on a foundation of immediately given experience, but is theoretical "all the way down", even to our experience of ordinary physical objects.  Sellars takes the implications of this even further, saying that even our experience of our own inner states is theory-based.  This means that if scientific evidence is what justifies the claim that mental states are brain states, it does so by establishing a relationship between mental concepts and neurological ones.  This relationship is not significantly different from what occurs when one scientific theory advances beyond another one, which means that the history of science becomes an essential discipline for philosophers of mind.
When philosophers of science began to take history of science seriously, it seemed natural at first to assume that a new theory always had some sort of identity relation with its predecessor. This relationship was  supposedly maintained by what Ernst Nagel called "Bridge Laws", (Nagel 1961) which established just the sort of connections between old and new theories that the identity theorists needed for brain states and mental states.  But thanks to historical work done by Kuhn (1962) Feyerabend (1962) and Laudan (1977,1981) it became clear that progress in science does not usually result in bridge laws. Most scientific progress comes from what Kuhn called 'Normal Science', which rarely, if ever, introduces new terms into scientific discourse, and thus does not need bridge laws. Scientific discourse usually adopts new concepts to refer to newly posited or discovered entities during what Kuhn called scientific revolutions. The thing that makes the introduction of these new concepts revolutionary, rather than evolutionary, is that there aren't anything like bridge laws to relate new theories to old ones.   Instead the old theory is often eliminated, and replaced with a better theory that rejects or ignores the ontological assumptions of the old theory.  For example, there are no identity relationships between the alchemical essences and the chemical elements, because we now claim that there are no alchemical essences.
Rorty (1965)  and Feyerabend (1963a and 1963b) thus concluded that if scientific progress was the model for the relationship between brain states and mental states, then there is no need to establish identities between the two. Once we have a sufficiently sophisticated neuroscience, we may be able to simply say that there are no mental states. This effectively disposes of the problems raised by Shaffer and Feyerabend mentioned above. The differences between identity and causal correlation were no longer of significance, because we were now talking about only one entity--the brain state-- the mental state having been consigned to the ontological trash heap.
History of the Idea of Eliminative Materialism
Feyerabend, Rorty and the Churchlands
Feyerabend had nothing more to say on this subject after writing the two articles cited above.  Rorty, however, continued to defend his position against critical articles ( such as Cornman 1968a and 1968b, Lycan and Pappas 1972, and Bernstein 1968). Rorty had originally called his position the disappearance theory of mind, but he later adopted the term eliminative materialism, which had first been used by Cornman in his 1968a. Besides responding to many of Cornman's and Bernstein's objections in Rorty 1970, he also devoted all of Chapter II of Rorty 1979 to a more detailed version of the thought experiment in Rorty 1965.
Lycan and Pappas 1972 point out that if Rorty claimed that it is only possible in principle to replace mental talk with neurological talk, this would mean that mental states are as real as tables and chairs, because in principle all talk about tables and chairs could be replaced by talk about atoms and molecules. This would make Rorty's brand of eliminative materialism only trivially true. Lycan and Pappas claim that Rorty was somewhat equivocal in his 1965 and 1970 papers about whether the elimination of mental language was actually desirable, or only possible in principle. Paul and Patricia Churchland, however, developed a more aggressive form of eliminative materialism. They claimed that the elimination of talk about mental states was not only possible and desirable, but a fully viable goal for a neuroscientific research program.
It is important, however, not to confuse the Churchland's eliminative materialism with the more extreme and widely publicized position of B.F. Skinner.  Because the Churchland's work is also discussed by non-philosophers in the cognitive science community, it is quite common for people trained in psychology to react negatively to eliminative materialism because they are reminded of passages like this from B.F. Skinner's "Beyond Freedom and Dignity."
Aristotle argued that a falling body accelerated because it grew more jubilant as it found itself nearer home. All of this was eventually abandoned, but . .  .  Almost everyone who is concerned with human affairs. . . continues to talk about human behavior in this prescientific way. ( Skinner 1971. pp. 8-9)
The implication of this and other passages in Skinner was that any talk that presupposed something like folk psychology had already been proven to be false, and that talk about mental states could be successfully replaced right now by Skinner's own behaviorist psychology.  These philosophical presuppositions of Skinnerian behaviorism were a serious impediment to many kinds of psychological research, and have now justly been relegated to the history of psychology. It is natural for psychologists to assume that the Churchlands are making the same claims for neuroscience that Skinner once made for behaviorism, and to resist eliminative materialism for fear that it would compel a similar narrowness of focus.  The Churchlands recognize, however, that as long as there are important nuances in folk psychology that the neuroscientific view cannot capture, folk psychology is not ready to be eliminated. As Paul Churchland puts it "If and when the change ever happens, it will be because we are all gleefully pulled rather than grudgingly pushed. We will be pulled, if at all, by the manifold, personal, social, and practical advantages of the new framework: by the clarity it represents, by the freedoms it makes possible, by the cruelties it diverts, and by the deeper interactions it affords." (Churchland 1998)
Misunderstanding is also encouraged by the fact that the Churchlands alternately emphasize two different aspects of their message, which can sometimes create an illusion that they believe that folk psychology is already on the way out. On one hand, they admit that our current neurological language has a long way to go before it could eliminate mental language, and that it is purely an empirical question whether folk psychology will actually be eliminated.  But they also insist that folk psychology does have some very serious problems, and that current developments in neuroscience appear to be on the way to solving those problems. This is why they see eliminative materialism not as a mere logical possibility of interest to no one but philosophers, but as a genuine empirical possibility that deserves serious scientific attention.  Many of the Churchlands' critics actually accept that the elimination of folk psychology is possible, and attack only the claim that elimination is probable. (such as Horgan and Woodward 1985, Horgan and Graham 1990). The Churchlands have less at stake with respect to these kinds of criticisms, and have in fact softened their position in response to some of them. But this should not be taken as an abandonment of the core principles of eliminative materialism.
Stich and the Case against Belief
Stephen Stich (1983),  claims that logic-based cognitive science  provides evidence against the validity of folk psychology, because it is based on what Stich calls a syntactic theory of mind. (STM). This means that it must focus exclusively on processes that are assumed to take place entirely within the brain, and ignore how those processes relate to the outside world.  The folk concept of belief, however, is necessarily semantic, not just syntactic. The belief that Paris is the capitol of France is what it is not only because of its relationship to other sentences, but also because of its relationship to Paris and France. Consequently, there is no place for the folk concept of belief in the ontology of cognitive science.
Stich 1983 devotes several pages to the Churchlands' eliminativism, stating their numerous  agreements and relatively few disagreements with him, and Patricia Churchland returns the compliment in P.S. Churchland 1986. Stich has also argued that Dennett should be an eliminativist about beliefs, given the numerous criticisms they both have of the concept. (Stich 1980, 1981). But, as Susan Haack points out, the Churchland and Stich arguments to some degree work against each other. One of Churchland's arguments for eliminative materialism is that we have good reason to believe that all sciences will be ultimately reducible to the physical sciences.  Stich's  syntactic theory of mind, in contrast, is justified by the need to keep psychology autonomous (and thus not reducible to physics) (Haack 1993 pp. 159-160) Haack's argument against Stich is strengthened by Fodor's many arguments that logic-based cognitive science is a refinement of the principles that underlie folk psychology and thus would only refine and not eliminate folk psychology.  [1]  Also,  Fodor 1994 uses something like Stich's argument to cut in the opposite direction, arguing that cognitive science is on the wrong track because it is incapable of explaining reference and other semantic properties of language. The fact that some scientists and philosophers now claim that cognition cannot be understood without making reference to an organism's environment also shows that Stich's syntactic theory of mind is perhaps not even the best, let alone the only, option. (Clark 1997, Bechtel and Abrahamsen 1993).
Ramsey, Stich, and Garon 1990 brought Stich's arguments somewhat more in line with the Churchlands by demonstrating that at least one connectionist model of the mind was also incompatible with the folk psychological concept of belief.  But Stich eventually became a critic of eliminativism, for reasons that we will discuss later on.
Dennett and the Intentional Stance
Daniel Dennett agrees with many of the key points of the eliminative materialists. He admits that he sees the future of the sciences of the mind "very much as Churchland does, but with some shifts in emphasis" (Dennett 1987 p.235). These shifts arise partly because, in Dennett 1975 and 1989, he is addressing at least part of his message to psychologists who were trained in Skinnerian behaviorism, and are now interested in Cognitive Science.
Talking about the mind, for many people is rather like talking about sex: slightly embarrassing, undignified, maybe even disreputable. . . .Those in other disciplines who are newly eager, or at any rate reluctantly willing, to indulge in various mentalistic sorts of talk find that philosophers, who have never been shy about talking about the mind, have a lot to tell them about how to do it. (Dennett 1987 p. 1)
Dennett thus assumes that much of his audience will already be inclined to accept the eliminativist dismissal of folk psychology. Consequently, even though Dennett accepts, and even frequently defends, the eliminativist position, he also emphasizes how useful folk psychology is, not only in daily life, but in scientific disciplines like Artificial Intelligence. Because philosophers have not been shy about talking about the mind, eliminativism was a radical position in the philosophy community. But for behaviorist psychologists, it was a stagnating orthodoxy that needed to be transcended.
Dennett's term for folk psychology  is "The Intentional Stance". He says that when we take a certain "stance" towards other human beings, we can think of them as if they had beliefs, desires, and the ability to make inferences from those beliefs and desires. Taking this stance makes it much easier to predict the behavior of humans than if we assume that people are only mechanisms governed by the laws of physics.   We can, for example, use the intentional stance to predict that Sally will go to the refrigerator and get milk, if we know that she wants milk and believes that there is milk in the refrigerator. This stance also works, with noticeably less effectiveness, for animals, and even for artifacts such as chess-playing computers. If we were omnipotent, we could predict that Sally would go to the refrigerator using nothing but the laws of physics, but because we are not, it makes sense to continue to use the concepts of the intentional stance/folk psychology. Despite it's usefulness, however, the intentional stance must be seen as only a stance --i.e. a way of looking at things which is strictly speaking, or in some sense, false or incomplete. This is because it breaks down on numerous occasions. If Sally has been has been poisoned or is suffering from a brain tumor, we cannot explain her behavior by referring to her beliefs and desires; we  need chemistry and/or neuroscience. But most of the time, the intentional stance is by far the best conceptual system we have for predicting human behavior. Dennett usually argues that it would be for all practical purposes  impossible to do scientific research or daily social interactions without it, at least at this point in time.
Dennett spends a fair amount of time discussing both the limitations of the intentional stance (Dennett 1978 chapters 3, 11;  1987 chapter 5) and its usefulness (Dennett 1978 chapter 1; 1987 chapters 2,3, 4, 7). Consequently, when it comes to claims about the ontological status of the entities described by the intentional stance, Dennett's position is more ambivalent than that of Stich or the Churchlands. Dennett expresses his ontology of the mental in analogies (beliefs are like voices, or centers of gravity) that get interpreted in conflicting ways by different commentators. In Dennett 1991a, he describes his position as "semi-realism" which grants  "quasi-existence" to mental entities. (p. 1 ) But ultimately he responds to Haugeland's detailed exegesis of that paper (in Dahlbom 1993) by saying "I am shy about drawing ultimate conclusions about Reality, Truth. . .and the other grand topics of metaphysics and epistemology". (p, 204).  Dennett apparently prefers an equivocal position on these ontological questions to a position that he feels would force him to either downplay or ignore the strengths and weaknesses of folk psychology.
In Dennett's later writings,  however, he does seem to be more willing to think in terms that de-emphasize the intentional stance. He is overtly eliminativist about the concept of a unified self  in Dennett 1991b, (Although he softens this eliminativism with an analogy in chapter 1 of Dennett 1991b, and concludes that book by saying that he is only replacing one metaphor with another.) And he seems to imply in Dennett 1995 that "the universal acid" of Darwinian natural selection in some sense falsifies or eliminates the concept of intentionality. Perhaps the only real difference between Dennett and the other eliminative materialists is which possibilities they each consider to be the most probable. Dennett is betting that the ultimate relationship between neuroscience and the intentional stance will be closer to a reduction than an elimination.(at least in his earlier writings.) The Churchlands and (the early) Stich are betting on an outright elimination. But both sides agree that neither can be certain that the other is wrong.
Analysis of the Eliminative Argument
The basic form of the Eliminative argument is something like this  [2] : 
  1. Folk psychology is not significantly different from obsolete scientific theories like Alchemy, Phlogiston etc.
  2. Alchemy, Phlogiston, etc. are false, do not apply to reality, and the entities they describe do not exist.
  3. Therefore, It is possible that folk psychology is false, does not apply to reality, and the entities it describes don't exist.
Because eliminative materialism is a speculation about the future, based on analogy, this argument is not a straight forward logical entailment. It does, however, give a good starting point for describing the history of the criticisms of eliminative materialism. Most criticisms have attacked the first premise, by trying to prove that in some sense folk psychology is significantly different from other theories, or not a theory at all. This is an understandable approach, because the eliminativist's first premise is contrary to common sense, and until Quine and Sellars, was contrary to most philosophical thinking as well. The second premise, in contrast, seems so self-evident that any argument that denied it would seem to be a reductio ad absurdum.  There are, however, arguments raised against the second premise in Rockwell (1995) which claim that Churchland's own pragmatism is inconsistent with the idea that alchemy and phlogiston are unconditionally false theories, and by Stich (1996, 1998) who argued (against his own earlier eliminativism) that there is no reason to assume that a theory does not refer to anything just because it is false.

Objections against the First Premise
Eliminative Materialism is Self-Contradictory
Many people claim that folk psychology is different from scientific theories because it is impossible to think without it. The simplest version of this argument is that when eliminativists say that they believe that there are no such things as beliefs, they are contradicting themselves, and therefore their arguments are invalid. Despite the fact that this argument has been answered by both Churchlands (P.M. Churchland 1989 p. 22;  P.S. Churchland 1986 pp. 397-398) and in Ramsey 1990, it continues to reappear. (Rudder-Baker 1987; Haack 1993 pp. 179-180; Putnam 1987 p.16; Miller 1992 ).  Patricia Churchland's answer is as follows:
What the eliminativist is fumbling to say is that folk psychology is seriously inadequate as a theory. Now  within the confines of that very theoretical framework we are bound to describe the eliminativist as believing there are no beliefs; however, this is not because folk psychology is bound to be true,  but only because we are confined within the framework the eliminativist wishes to criticize and no alternative framework is available. . . It would be foolish to suppose folk psychology must be true because at this stage of science to criticize it implies using it. All this shows is that folk psychology is the only theory available now.  (pp. 397-398)
She goes on to add that a similar argument could be made in defense of vitalism, if one argued that anyone who denied the existence of vital spirit must be asserting that she herself is dead.
A closely related argument is that the empirical research that the Churchlands claim supports eliminativism actually refutes it. Herman Philipse (1998), for example, reminds us that the network theory of meaning defended by Quine and Sellars was the primary justification for eliminativism, and he claims that eliminativism necessarily accepts Sellars' claim that all awareness is a linguistic affair. Philipse claims that because contemporary neuroscience shows that thought is fundamentally non-linguistic, it contradicts the network theory of meaning and therefore contradicts eliminativism.
This would be a valid criticism of Rorty, who repeatedly quotes and paraphrases the Sellars slogan mentioned above as a defense of eliminativism (in Rorty 1970, among other places). But  Paul Churchland has been aware of these possibilities from the beginning. As he points out in Churchland 1998, the last chapter of Churchland 1979 is devoted entirely to the possibility that all thought is fundamentally non-linguistic. In his first formulations of eliminativism, Churchland did claim that we should see folk psychology as a set of law-like sentences from which the folk make logical inferences (Chapter one Churchland 1989). But, in chapter one of Churchland and Churchland 1998, he also gives reasons why his current connectionist arguments show that "The claim that FP {folk psychology} is a corrigible theory need not be hobbled by its initial logical positivist dress." (p.15). The fact that he first had to compare scientific and folk theories by thinking of them in linguistic terms is really no more surprising than the fact that he has to say he believes there could be no such thing as beliefs.  At the time Churchland wrote those words, the best description we had of folk psychology, or any other theory, was to call it a set of sentences from which logical inferences could be made. The fact that we are now groping towards a better description doesn't invalidate the general point that Churchland was trying to make: whatever we use to explain the workings of scientific knowledge is equally applicable to folk psychology.
As for the point that the network theory requires accepting a linguistic theory of knowledge, the holistic nature of neural networks makes them every bit as capable of  supporting eliminativism as were linguistic networks. The characteristic of the network theory of meaning that made eliminativism unavoidable was not the fact that the networks were made of words. It was the fact that the network theory eliminated the idea of perception as an immediately given foundation for thought and science.  However, this idea of perception as direct awareness of the given is still so deeply ingrained that many philosophers still attempt to use it as an argument against eliminativism, as if Quine's and Sellars' arguments against the given had never been written.
Theory vs. the Experienced World

Phillipse is not the only one to insist that "surely a change of theory will not transform the phenomenal content of perceptual and inner awareness." (p. 887)  Bernstein made almost the exact same objection against Rorty (Bernstein 1968 p.218 in Rosenthal 1971). John Haldane claims that 'theory. . .may filter down to the observational level. However, such penetration does not eliminate pretheoretical observation." (p. 269 in Christensen/Turner 1993).  And Searle (1992 ) asserts that "Beliefs and desires, unlike phlogiston and caloric fluid, were not postulated as some special theory, they are actually experienced as part of our mental life"(p.60).  The Churchlands have no answer to these objections, because they believe that Quine and Sellars have already answered them. ( see Churchland and Churchland 1998 p. 26) Until someone shows why Sellars and Quine are wrong about the points that support eliminativism, there is no reason for eliminativists to formulate separate answers of their own. The only critic who makes these sorts of points who cannot be dismissed with a Quine or Sellars footnote is Susan Haack, whose "foundherentism" is a genuinely new alternative to both the network theory and classical empiricism. Interested readers will find a clear presentation of this view in Haack 1993.
However, tangled in amongst these empiricist arguments there is another set of arguments that requires a different answer. Bernstein (1968) has a quote from Merleau-Ponty saying that "the whole universe of science is built upon the world as directly experienced. . .we must begin by reawakening the basic experience of the world of which science is the second order expression." (p. 200-201 in Rosenthal 1971) The claim that there is a unified basic experience of the world is completely untouched by the Sellarsian arguments against immediately given and independent sense data.  Sellars' critique was against the idea that what was given in immediate experience was propositional, i.e. that one could make inferences based on bits of information which were presented to us in single observations. The arguments against sense data in Sellars 1963, which were the basis of the holism that lead to eliminativism, do not dispose of the possibility that there exists something like a lebenswelt, or world of lived experience, distinct from the world of scientific knowledge.
Sellars was arguably not in complete agreement with this more holistic concept of experience, advocated by writers such as Dewey and Merleau-Ponty. Rorty's interpretation of Sellars does not permit this kind of distinctness, for he claims in his reply to Bernstein that the very idea of a non-linguistic awareness is what Sellars is saying we should give up (Rorty 1970 p. 229). But there are interpretations of Sellars which claim he is advocating something like this kind of non-linguistic awareness for what he calls the Manifest Image.  And although Paul Churchland never uses the Continental terminology, he believes strongly that there is a distinction between knowledge and experience. He argues at some length in Churchland 1979 that there is an important difference between knowing a theory and experiencing the world in its terms. (pp. 25-45) For him, the thing that is most exciting about eliminativism is that it holds forth the possibility of experiencing the world in scientific terms, which is very different from merely knowing the facts of science. Churchland does not believe that the experienced world is autonomous with respect to the world described by science. But he did see experiencing as distinct from knowing, even though the two are locked in a dialectical embrace that makes them constantly reshape each other. (And because Merleau-Ponty frequently cites scientific experiments in his phenomenology of lived experience, he probably agreed with Churchland on this.) This is the main significance of what Churchland calls "the Plasticity of Perception". Because our theories shape our perceptions, we should not be satisfied with merely knowing the truth of science. We must experience the world in scientific terms as well. 
There are also versions of this argument that come out of the Oxbridge ordinary language tradition. Wilkes 1991 tries to drive a wedge between the abstract concepts of science and the engaged experience of everyday life when she says that common sense psychology (CSP) "has countless other things to do besides description and explanation. . . {such as} joking, jeering, exhorting, discouraging, blaming, praising. . .  This means that not only is CSP not a theory; it could never become one, and nobody should wish it to" (p. 176 in Christensen and Turner 1993). It is also possible to interpret the above Searle quote in this manner, given Searle's repeated arguments that  "the background" of experienced life is what makes linguistic knowledge possible. (Searle 1992 Chapter 8). Paul Churchland replies to this objection, not by resisting, but by accepting it and going one step further.  (Churchland and Churchland 1998, chapter 3.) Scientists, like everyone else, have experiences that are shaped by their practices. A scientist not only knows things that other people don't know, he can do things that other people can't do, and perceive things that other people can't perceive. In order to be a practicing scientist one must do many things besides describe and explain: one must know how to operate lab equipment, perform diagnoses etc. and these skills require the ability to perceive subtle differences in colors and shapes e.g.  that appear on the screens of laboratory measuring devices.  Churchland believes that knowledge, experience, and practice always condition each other, and that some of the practices and perceptions of science could trickle down and transform the practices and perceptions of ordinary experience. He is not claiming that theoretical knowledge can ever replace practice and experience.
The Normative vs. Descriptive Argument
In Epistemology Naturalized, (Quine 1969), Quine argued that because there could be no such thing as a first philosophy which provided a foundation for all other inquiry, epistemology and the natural sciences could not be separated from each other. In many ways, the Churchlands' Neurophilosophy  is a concrete application of the project described in that essay, for it is based on the assumption that studying the brain is an essential part of figuring out how the mind knows things. The Churchlands' eliminativism derives from the belief that as we continue our scientific study of the mind/brain, we may come up with an epistemology that eliminates our current ones, which are based on Folk Psychology.
Jaegwon Kim argues (in Kim 1993, chapter 12) that no physical science could ever replace epistemology, because the physical sciences are merely descriptive and epistemology is necessarily normative. He claims that even the concept of belief (which is essential to epistemology and all other aspects of folk psychology) is necessarily normative. This is because we can make no sense of the idea of an organism having beliefs unless we presuppose that those beliefs are part of a network of rationally interrelated concepts. Because the concept of rationality is normative, the concept of belief also presupposes normative categories. Kim acknowledges that we must have purely descriptive criteria for recognizing when something that occurs in the world conforms to a norm. But this does not mean that the norm is itself reducible to such a description. Kim accuses  Rorty of making precisely this mistake in his defense of eliminativism in Rorty 1979. Kim claims that the relationship of Supervenience, which is the main topic of Kim 1993, is the only concept that can explain the relationship between the normative and the descriptive. McCauley 1988 and 1992 applies similar arguments to the Churchlands' eliminativism, claiming that "attempts to completely eliminate the normative will be either forever incomplete or inimical to the progress of science" (McCauley 1988 P.14).
Paul Churchland acknowledges that "it may be true that normative discourse cannot be replaced without remainder by descriptive discourse"  but claims that the most important point is that "it is only the autonomy of epistemology that must be denied". (Churchland 1989 p. 196 )   In other words, epistemology should be informed by science, but this need not imply that science can replace it. Churchland's claim that "normative issues are never independent of factual matters" (ibid.) is at the very least compatible with Kim's supervenience relationship between the normative and the descriptive. Churchland, however, sees norms as being guided by prototypes of the sort that Kuhn calls paradigms. To see a human activity as guided by prototypes is not to describe it purely factually. In a prototype-based value system, certain activities, practices and/or artifacts are considered to be exemplary, and all others are judged by how close they come to both fulfilling and extending the ideal exemplified by those prototypes. But the prototypes are real pieces of exemplary work or behavior in the world, they are not abstract Platonic ideals. Consequently, new scientific work will always influence those ideals, and it is in this sense that Churchland's epistemology is both normative and naturalized. Old norms may be eliminated in favor of new norms, but this does not mean that normativity itself will be eliminated. The fact that Paul Churchland now frequently writes about ethics  (for example in Churchland 1995) shows that he now wants to contribute to normative discourse, rather than eliminate it.
The Autonomy Arguments
There are several arguments against eliminativism which claim that folk psychology is a conceptual system with some kind of autonomy with respect to the physical sciences. These arguments often concede that folk psychology may be reduced or even eliminated by scientific progress. But they claim that scientific psychology, not physics, will replace folk psychology, and psychology will remain in some sense autonomous from the physical sciences.
Horgan and Woodward 1985 accept that it is unlikely that the categories of folk psychology will be identified with anything in the categories of neuroscience, even if the latter eventually "provide a marvelous account of the nature and behavior of Homo Sapiens" (p. 149 in Christensen and Turner 1993). But they claim that this would not make folk psychology false, if we accept the anamolous monism of Donald Davidson. Horgan and Woodward claim that Davidson's position is naturalistic and materialist without being reductive or eliminative, and if he is right, psychology would not be reducible to any physical science. Jaegwon Kim, however,  argues at some length that Davidson's non-reductive materialism is inconsistent, and that we can be either materialist or non-reductive but not both. ( Kim 1993 chapter 11). If Kim is right about this, then Davidson provides no aid and comfort to those who want an autonomous scientific psychology.
Robert McCauley has argued in several articles (McCauley 1986, 1987, 1996) that the Churchlands have mistakenly conflated intralevel reductions with interlevel ones. Intralevel cases are concerned with successive theories at the same level of analysis; for example when chemistry replaced alchemy.  McCauley acknowledges that elimination frequently occurs during these intralevel reductions. But he claims that it is far less common for elimination to occur with interlevel reductions e.g. reductive relationships between a science operating at lower level (such as physics) and one operating at a higher level (such as psychology).
In interlevel cases corrections can arise because the upper level theory is insufficiently fine grained to handle certain problems. By contrast, in intralevel cases corrections always arise because the earlier theory is wrong--by a little in evolutionary cases, by a lot in revolutionary ones (McCauley 1996 p.31)
For this and other reasons, McCauley claims that there is little historical evidence to support the conjecture that psychology will be eliminated by neuroscience. McCauley also claims that Patricia Churchland implicitly accepts this in Churchland 1986 when she says that psychology and neuroscience co-evolve i.e. that each is constantly shaped by the need to account for the discoveries of the other. McCauley quotes Churchland and Sejnowski as saying "the co-evolutionary advice regarding methodological efficiency is 'let many flowers bloom'" (McCauley 1996 p.33),  McCauley sees this maxim as consistent with a position he calls "explanatory pluralism", which he claims is in some sense inconsistent with the Churchland's earlier eliminativism.
Paul Churchland offers some historical examples of his own to counter McCauley's claim that eliminations do not occur in interlevel reductions (McCauley 1996 pp. 222-225). And he says that even if a more abstract level is needed to account for psychological phenomena, we need not assume that folk psychology is the best framework to account for that abstract level. "Legitimating the office need not legitimate the current office holder." (ibid. p. 225) Nevertheless, this concession implies that Churchland now acknowledges that a unified science is not essential to his position, which is a significant departure from eliminative materialism as he first defended it. In Churchland 1989 chapter 1, many of his arguments for Eliminative Materialism were based on the claim that folk psychology could not be integrated into the unified structure which is modern physical science.
Fodor 1975 and Putnam 1975 advocated somewhat different versions of  the multiple realizability  argument for the autonomy of psychology. (also called the functionalist  argument). Fodor points out that psychological and sociological categories have too many different physical ways that they could be instantiated for them to be reduced to single physical predicates. For example, a monetary exchange could be instantiated physically by handing over a dollar bill, or by writing a check, or by using a string of wampum, and it would obviously be only an improbable coincidence if any of these actions had anything in common physically. (Fodor 1975 p. 15) Because physics only studies what things are made of, psychology and the other sciences of the mental must necessary be autonomous with respect to physics. The underlying assumption of the social, psychological, and biological sciences is that the things they study are constituted by how they function, not by what they are made out of.
At one point Putnam was willing to consider the possibility that human minds were functionally equivalent to Turing machines, which meant that computer science was the functionally autonomous science of mind that could never be reduced or eliminated by any of the physical sciences. Although he specifically rejected this position in Putnam 1975 chapter 14, he still believed at that time that it was necessarily true that "[m]entality was a real and autonomous feature of our world" (ibid. p.291). In Putnam 1988, however, he renounced functionalism altogether, because he claimed that the relationship between the mental and the computational was every bit as problematic as the relationship between the mental and physical. Fodor, however was more cautious in Fodor 1975, claiming only that "The world could turn out to be such that every kind corresponds to a physical kind . . . It's just that as things stand, it seems very unlikely . . ." (Fodor 1975 p.15).
Paul Churchland agrees that even the most radical eliminative materialist must endorse functionalism "construed broadly as the thesis that the essence of our psychological states resides in the abstract causal roles they play in a complex economy of internal states mediating environmental inputs and behavioral outputs" (1989 p.23).  But he argues that multiple realizability with respect to physics has nothing to do with whether a theory is true or not. He points out that alchemy describes the behavior of macroscopic substances in a way that could be multipley realized by a variety of submicroscopic theories. (ibid. pp. 12-14) Consequently, he claims we could conclude from the functionalist arguments that alchemy is not a false theory, which is obviously absurd. The terms used by alchemists do not refer to anything, because there are no alchemical essences.
Few people will question the intuition that the terms used by many obsolete sciences describe things that do not exist. We all agree that there is no such thing as phlogiston, or caloric, or the aether. Eliminativism, however, invalidates many of the assumptions about truth and language that support that intuition. It is thus possible to use the eliminativist's radical new assumptions to undercut the second premise of the eliminativist argument i.e. the assumptions that obsolete sciences are 1) false, 2) do not apply to reality, and 3) the entities they describe do not exist. The pragmatist argument described below questions all three of the claims in the second premise, without distinguishing among them. The reference argument accepts the first claim (that obsolete theories are false) but denies that this must imply the other two claims.
Objections to the Second Premise
The Pragmatist Argument
The philosophy of science that made eliminativism possible had other implications, some of which created problems for many of the presuppositions of eliminativism. Laudan 1981 argued that scientific realism, as it was commonly defined, could not be defended once we admitted that new scientific theories eliminated old ones. And although Paul Churchland continued to refer to himself as a scientific realist, he clearly accepted Laudan's interpretation of the scientific historical record.
So many past theories, rightly judged excellent in their time, have since proved to be false. And their current successors, though even better founded, seem  but the next step in a probably endless and not obviously convergent journey. (Churchland 1989 p.140)
It is expected that existing conceptual frameworks will eventually be replaced by new and better ones, and those in turn  by frameworks better still, for who will be so brash as to assert that the feeble conceptual achievements of our adolescent species comprise an exhaustive account of anything at all? (ibid.  p.52)
However, if scientific truth is this mutable, folk psychology's "falsehood" is no longer as serious as it was from the traditional scientific realist perspective.  Rockwell (1995) argues that the eliminativist view of scientific progress requires us to believe not only that all our past scientific theories are false, but also that our present and future theories are falsifiable. Unless we are willing to accept some sort of quasi-Hegelian view that each scientific theory is true "for its time", we must conclude from this that when a theory is falsified, we are actually discovering that it was really false all along.  Thus, if each theory is falsified by its successor, and every theory is succeeded by some other theory, therefore all scientific theories that ever existed or will exist are false.  To some degree Feyerabend did succumb to these skeptical implications of eliminative materialism, which is why he ended up writing books with titles like Farewell to Reason, and Against Method.  Churchland, however, gradually shifted towards a pragmatic pluralism which saved him from skepticism, but which no longer supported eliminativism very effectively.
In their later writings, the Churchlands described reduction (which validates some aspects of the old theory) and elimination (which falsifies the old theory) as being endpoints on a continuum rather than an either/or choice. (a position that was later developed in greatest detail in Bickle 1998).  Paul Churchland's analysis of connectionist neuroscientific research also supported this view.  In Churchland 1989, he argued that, while this research showed that the learning process in a neural net involved reducing errors by reconfiguring weight spaces "nothing guarantees that there exists a possible configuration of weights that would reduce the error messages to zero" ( p. 194). Thus the difference between a bad theory and a good one was a difference in degree, which was not accurately captured by the traditional either/or distinction of true/false. He also pointed out that there could be different, even contradictory, theories of equal epistemic virtue, "all of them equally low in error, all of them carving up the world in quite different ways . " (ibid.). (This position is developed at greater length in chapters 15 and 17 of Churchland and Churchland 1998). Although these changes in his thinking resulted from the exegesis of scientific research, they seriously undermined Churchland's earlier self-described scientific realism (Churchland 1979) and transformed it into a "pluralistic form of pragmatism" (Churchland 1989 p. 194)

However, Churchland's later pragmatic pluralism requires him to give folk psychology  essentially the same status that Dennett gives to the intentional stance: a theory with serious shortcomings, which nevertheless has genuine epistemic merit. The comparison to alchemy does not reduce folk psychology to a false theory, for if Churchland denies all epistemic status to alchemy because we now have better theories, we would have to deny epistemic status to our current theories because they will eventually be replaced by better ones. And this would leave us with the universal skepticism described earlier. We could save ourselves from this skepticism if we were willing to accept the possibility that truth and falsity need not be binary, and that alchemy and folk psychology (as well as special relativity theory and whatever theory replaces it) all possess degrees of epistemic virtue. From this pragmatist perspective, none of these theories would be true in the classical sense, but that does not imply they are all false, anymore than the fact that no one possesses vital spirit implies that everyone is dead.
This pragmatist perspective also saves us from an objection to eliminativism that dates back to Lycan and Pappas' 1972 criticism of Rorty, and which continues to resurface: the fact that eliminativist scientific realism requires us to claim that there are no such things as tables, chairs and the other  functionally defined terms of common sense. Searle, in his  1992, dismisses eliminative materialism by pointing out that if  the entities posited by folk psychology don't exist because they cannot be reduced smoothly to the entities of science, we must also deny ontological status to split level ranch houses, tennis rackets, golf clubs, etc. (p.47). Similar arguments are also raised by Andrew Cling in his 1991, and by Putnam in his 1992. However, these arguments usually rest on a clash of intuitions that leads nowhere. The assumption is that obviously alchemical essences do not exist, and also obviously tables and chairs exist, and therefore the two must be different. But all the arguments that attempt to show why they must be different have been effectively answered by the Churchlands. The pragmatist, unlike the scientific realist, can dissolve this problem by saying that alchemy does possess some epistemic merit , and the only reason that we no longer refer to the alchemical essences is that we now possess a better theory which has more epistemic merit. Paul Churchland, however, has so far refused to take this step. Instead, in his 1992, he attempts to maintain a distinction between "(legitimately functional) tables and chairs" and (the avowedly non-existent) phlogiston, caloric, and the four principles of medieval alchemy ". This, however, does not sit well with the thought experiment in chapter 1 of his 1989,  which is designed to show that the categories of alchemy are functional, and therefore that functionality is no proof of  legitimacy. It also seems unlikely that a notion from vulgar discourse like "chair" is going to be more ontologically rigorous than a concept like "phlogiston" which, in its time, was the best that the greatest living scientific minds could produce.
Those who cannot accept the idea that obsolete theories like Alchemy have truth value might prefer another alternative proposed by Stephen Stich: Obsolete theories are false, but that need not imply that the terms in those theories don't refer to anything.
Eliminativism and the Problem of Reference
In Stich 1996, Stich turns from being eliminativism's dearest friend to its severest critic. His criticisms focused on certain assumptions about truth and reference that neither side of the eliminativist debate have questioned. Given that the central claim of eliminativism is that folk psychology might be false, it is obviously crucial to determine what we mean by truth and falsity. The true-false dichotomy is usually assumed to be tightly bound up with questions of reference and existence, so that the expression "theory x is false" was assumed to inevitably imply both "the entities described by theory x do not exist" and "the terms in theory x do not refer to anything". Stich argues that these are three distinct claims, and that an argument is needed to prove why they must go together. Because there is no such argument, it is possible to acknowledge that folk psychology is wrong about most things and in serious need of replacement, but not infer from this that the entities folk psychology describes are non-existent, or that its terms do no refer to anything.
In chapter 1 of Stich 1996, he drags out an old argument but adds a new twist that gives it genuine bite. The Churchlands have frequently admitted that the theory which replaces folk psychology could be fundamentally non-linguistic; in fact they even emphasize this in their later writings. Stich, however, points out that truth and falsity are properties of language, and if folk psychology is constituted by tacit connectionist structures which are fundamentally non-linguistic, this means that it cannot be either true or false. In other words, if the Churchland's beloved connectionist theories can be developed into a complete and accurate account of the skills and abilities that we call folk psychology, then folk psychology is neither true (as many of the Churchlands' critics have claimed) nor (possibly) false (as the Churchlands themselves have claimed.)
Stich also claims eliminativism is inevitable only if we accept a descriptive or conceptual role semantics, which determines the reference of a term by the place it occupies (or fails to occupy) in a network of concepts. This has certainly been Paul Churchland's view of semantics (see Churchland 1979 pp. 56-58.), but it is not the only possible semantics. Stich claims that if we accepted a causal-historical semantic theory, such as those proposed by Putnam and Kripke, (Kripke 1972, Putnam 1975),  even very seriously mistaken theories would still refer to their subject matters. For example, Norse mythology claims that the stars are holes in the skull of a dead giant, which is about as mistaken as one could possibly be about anything. Yet when the Vikings spoke about the stars, their talk was clearly about the stars, it did not refer to nothing at all. The reason for this is that the stars themselves had some sort of causal relationship to the Viking's discourse. The light from the stars themselves caused the Vikings to think and speak about them, and thus the Vikings' speech and thought were about the stars.
The obvious objection to causal theories of reference is that surely not all causal relationships between language users and objects are reference relations, and it is not easy to figure out which causal relations are reference relations and which are not. If dust causes me to sneeze, my sneeze is not "about" the dust. If my dry mouth causes me to ask for water, my question does not refer to my mouth, it refers to the water. So how can we tell the difference between causal-reference relations and other kinds of causal relations? Stich acknowledges that these problems have been around for a while, and says this is largely because the defenders of the causal theory have not considered the possibility that there is not a single reference relation at all. Stich claims that the so-called reference relation is probably several different relations related only by a loose family resemblance, and in many cases what constitutes a reference-fixing causal relation is probably decided by politics and consensus.
Stich develops this position in greater detail in Bishop and Stich 1998, arguing that confusing questions of ontology with questions of reference causes problems in other areas of philosophy of science as well. Schouten and De Jong 1998 also extrapolate from Stich's position with regard to eliminativism and reference. They argued that reference fixation is not arbitrary, as Stich seemed to be implying, but acknowledge that it is "local, partial and context dependent" (p.1) In Schouten and De Jong's view, our scientific practices are constantly creating and destroying reference relations, all of which are provisional, and none of which are as rigid as the first causal-historical reference theories claimed. Consequently reference is maintained even when theories are radically falsified. The eliminativist claim that folk psychology might be false loses much of its bite when the concept of falsity is redefined so as to permit a false concept to refer to the world.
Eliminativism Then and Now
Both the pragmatist and the reference objections to eliminativism can be avoided by simply redescribing eliminativism as "Whatever happened to obsolete theories like alchemy and phlogiston could also happen to folk psychology. "  If we bypass those questions in this way, however, we are no longer committed to the claim that folk psychology could be false, as we ordinarily understand that term.  Because this claim is the most common one-sentence description of eliminativism, it may seem to some that this is an admission of defeat.  But the essence of eliminativism is the willingness to question our ordinary understanding, which includes concepts like truth and falsity. Churchland himself said that future epistemologies might cause us to eliminate even the concept of truth itself. (Churchland 1989 p. 150) Perhaps the arguments in Quine and Sellars, along with the Churchlands' analysis of connectionism, have already led us close to this point. To claim that alchemy has some truth value, or that it is possible for a false theory to refer, are certainly major changes in the traditional concept of truth, which arguably edge that concept towards the eliminativist end of the reduction-elimination spectrum. But the debates about eliminativism reveal inconsistencies in our folk psychological concept of truth that appear to have no other way of being resolved.
Furthermore, the philosophical implications of the neuroscience which support the Churchlands' current position are very different from the more purely theoretical considerations that originally inspired their eliminativism. When Paul Churchland wrote "Eliminative Materialism and the Propositional Attitudes" (chapter one of Churchland 1989.), one of the main arguments in that paper was that folk psychology was a likely candidate for elimination because it was different from physics in several crucial ways. As their neurologically-inspired alternative has actually developed however, it has lead to a pluralism which has changed the significance of their original claim that "active coherence with the rest of what we presume to know is a central measure of credibility to any theory." (Churchland and Churchland 1998 p.6.) The need for scientific theories to eliminate folk theories is no longer as compelling when scientific progress is produced by a dynamic and coevolutionary process, rather than a monolithic bottom-up reduction to physics. Active coherence is still to be strived for, but it is more likely to be a regulative ideal than an obtainable goal.
This should be seen, however, as a natural part of the growth of our understanding of epistemological issues . There was a time when epistemology was widely considered to be an a priori  enterprise, and the eliminativist claim that epistemologists needed to understand the brain was considered preposterous by most professional philosophers.  The ontologically cautious downsizing of eliminativism given above still preserves the eliminativist's rejection of a priori  epistemology, and on this point the eliminativists have been decisively vindicated. Thanks to the questions raised by the eliminativist debate, philosophers and scientists are now working together to analyze the philosophical implications of neuroscientific data, which has given rise to a whole new multi-disciplinary enterprise(see the articles on Connectionism and Cognitive Science in this encyclopedia.) This research has produced a variety of theories, some of which will surely be contradicted by future research. But naturalized epistemology, unlike it's a priori  ancestor, can take falsification of its past in stride.
Teed Rockwell