Find Enlightenment

Churchland Debunked, Commonsense Psychology Vindicated

Will Wilkinson

In this paper I shall attempt to defend commonsense psychology (CP) [1] against Paul Churchland's attack on it. Others have undertaken the same task, but the arguments usually advanced against Churchland are often superficial, failing to address the fundamental epistemological positions driving his arguments. So I shall strive to argue against Churchland at this deeper level.

Churchland's attack on CP stands upon two tightly interwoven epistemological theses. First, Churchland believes that perception is deeply theory-laden. Second, Churchland holds that a concept's meaning is fixed by its place in a conceptual network. This network of concepts constitutes a theory. The contention that the concepts of this network determine what is perceived is the upshot of the claim that perception is penetrated by theory. I shall argue that both of these theses are false, and that if the second were true, Churchland's own position would likely be undermined. Additionally, I shall argue that Churchland is driven into an inconsistency if his positions on meaning and the social character of theoretical knowledge are both true. Lastly, I shall concur with J.D. Trout that the success of CP in the practice of science provides indirect confirmation for CP.

I. Theory and Perception

A thorough critique of Churchland's position would include a robust defense of theory-independent perception. That would be a big chore indeed. However, for present purposes, it should suffice to argue, with Fodor, that it is likely that at least some perceptual judgments are not infected by theoretical considerations owing to the insularity of perceptual modules and that it is likely that CP concepts are rooted, in part at least, to a similarly insular cognitive module.

Churchland argues with vigor for the proposition that what we take ourselves to see is a function of the concepts we hold.[2] This is to say that our conceptual scheme imposes itself upon our perceptual processes. Fodor, arguing against Churchland, stresses the insularity of our perceptual systems and argues that there is no evidence that our concepts penetrate the perceptual "modules."[3] Fodor also urges a distinction between the view that perception is, in some sense, a kind of problem solving mechanism and the view that perception is "comprehensively penetrated by background beliefs."[4] Fodor is willing to grant the former, but he denies the latter.

Fodor's idea is that although our perceptual systems are encapsulated, they operate inferentially in the sense that certain "hypotheses" about probable distal causes are written in to the systems (presumably by natural selection). When sensory systems register proximal stimuli, they are conveyed to the perceptual systems as premises over which to operate inferentially according to the principles of the embedded hypothesis. So perception is theoretical in a "sub-doxastic" sense, but this kind of "theory" is not open to significant revision.[5] Indeed, considering perceptual systems as embodying hypotheses and performing inferences is perhaps nothing more than a heuristically useful metaphorical figure.[6] One need not adopt Fodor's program and suppose there literally are hypotheses written in Mentalese lurking somewhere inside the head but outside the range of consciousness in order to make the point. We can simply consider perceptual systems as consisting of more or less biologically demarcated material arranged in a manner to support a particular circumscribed set of causal relations, like other organs of the body--like a lung or a kidney. And the general point holds: perception is facilitated by a functionally discrete set of causal mechanisms which operate separately from the mechanisms subserving semantic functions.

Fodor offers examples that purport to show that observation is sometimes free of the kind of theory Churchland is interested in. The well known Muller-Lyer Illusion tricks observers into judging that one of two lines is longer than the other when they are in fact equal in length. The illusion illustrates Fodor's point above. We are duped because the appearance of the figures in the illusion triggers a "theory" embedded in our visual system about distance. But, at the same time, being aware of the illusion as an illusion (and being aware of the theory behind why we succumb to the illusion) does not make the illusion go away. "Why," Fodor asks, "isn't perception penetrated by THAT piece of background theory?"[7] Fodor concludes:

To get from a cognitivist interpretation of perception to any epistemologically interesting version of the conclusion that observation is theory dependent, you need not only the premise that perception is problem solving, but also the premise that perceptual problem solving has access to ALL (or, anyhow, arbitrarily much) of the background information at the perceiver's disposal. Perceptual implasticities of the sorts we've just been noticing make it highly implausible, however, that this second premise is true.[8]

In order to defend CP on similar grounds, it is necessary to argue that CP concepts, like perceptual-level concepts, arise as part of the normal operation of an encapsulated cognitive mechanism, and thus, like perceptual concepts, are not subject to significant revision. The prospects of showing this to be the case seem strong. Recent studies on children's acquisition of the concept of belief and on the causes of autism point to the existence of a dedicated cognitive mechanism, an important function of which is to facilitate attributions of intentional states to others.[9]

In Wimmer and Perner's study[10] , children were shown a puppet show in which a puppet-child named Maxi puts his chocolate in a box and goes out to play. While he is out, Maxi's mother moves his chocolate from the box to the cupboard. When Maxi returns, where will he look for his chocolate? Five-year-olds say that Maxi will look in the box--a good prediction. After all, that's where Maxi left it. However, three to four-year-olds respond differently; they indicate, either verbally or by pointing, that Maxi will look in the cupboard.

This seems to indicate that three to four-year-olds don't have the ability to accurately attribute beliefs. These children are not able to tell what others are in a position to know, and attribute belief to others egocentrically, on the basis of their own knowledge. On theory theories of CP, like Churchland's, this must mean that they either lack a theory about belief, or grasp it inadequately. As the studies show, at around four to five children vastly increase their capacity to make allowances for what others aren't in a position to know. How is this accomplished? Is this new ability due to the acquisition of additional theoretical knowledge that the child comes to employ? That is, do four to five-year-olds internalize a large number of of laws and generalizations in which the term 'belief' occurs, and then, on the basis of this internalized theoretical knowledge proceed to accurately attribute beliefs to others? This seems unlikely for a number of reasons. For one thing, this seems to require unrealistic intellectual sophistication on the part of five-year-olds. And there is an alternative explanation that is more plausible.

Robert Gordon and Alvin Goldman argue that the capacity for belief attribution (and commonsense psychological capacities in general) flows from a capacity for what they call "offline simulation."[11] Gordon points out that we have an astonishing degree of predictive success regarding our own behavior. For instance, just a moment ago, I predicted, accurately, that I would remain seated and type this very sentence. Gordon remarks that, "We have in this department a success rate that surely would be the envy of any behavioral or neurobiological science."[12]

We also have a related capacity to predict our own behavior in hypothetical cases.[13] If I wonder how I would react if my computer were to suddenly crash, all I need do is imagine that my computer his in fact suddenly crashed and observe how I react in the pretend situation (I'm mad and I cuss!). Since this simulation does not effect how I am actually behaving at the moment, it is "offline." Now, this kind of hypothetical simulation can be extended to predict other's intentional states and behavior as well. Expert chess players report that they often visualize the board from the other side, taking the opposing pieces for their own, decide what they would do, and thereby predict their opponent's move. The hypothesis of offline simulation offers a plausible explanation for the success of four to five-year-olds in predicting where Maxi will look. By this age they have developed a capacity to shift perspective and suppose hypothetically that they are in the position of another. They predict that Maxi will look in the box because that is what they would do if they were in Maxi's position.

That there is considerable consistency in the age at which children can first pass the belief attribution test provides a reason to suspect that there is a functionally specific cognitive mechanism that normally develops around this age. The hypothesis that there is some such mechanism would be strengthened if there were evidence that the specific functions of that mechanism could be impaired while leaving other capacities intact. And there is evidence to this effect.

Baron-Cohen, Leslie, and Frith administered a Wimmer-Penner type test to autistic children between the ages of six and sixteen.[14] Almost all of these children gave the wrong answer--the three-year-old answer. This result is not due to the fact that many autistic children are also mentally retarded. The autistic children tested had I.Q. levels in the average to borderline range, while Down's Syndrome children, with significantly lower I.Q. levels, performed very well on the belief attribution test. So the failure of the autistic children points to a highly specific cognitive deficiency. Since autistic children are well known for their inability for pretend play there is reason to believe that what they lack is a capacity for offline simulation.

Churchland often argues against CP by comparing it to some bygone scientific doctrine, such as the theory of phlogiston or caloric, and then proceeding to tell a story about how this well-entrenched, but ultimately misguided, theory eventually fell from favor. However, there is a significant disanalogy between CP and early scientific theories and Churchland fails to take this into account.

Scientific theories, like the phlogiston theory, arise in a particular period in history and have a sharply delimited influence, whereas the core concepts of CP seem to span all of known human history and to have been held in every human population. Chuchland himself points out that "The [folk psychology] of the Greeks is essentially the [folk psychology] we use today, and we are negligibly better at explaining human behavior in its terms than was Sophocles."[15]

Churchland takes this as a sign of a very bad theory. But one could just as well see in this a sign that CP is not so much a developed theory as it is a natural outgrowth of basic cognitive mechanisms that human beings have had for a long time. Since there is evidence that the acquisition of at least some CP concepts are related to the normal function of a dedicated cognitive module, there is reason to believe that CP is a more or less permanent part of human cognitive architecture and is thus not subject to elimination.[16] Elimination would require brain surgery, not a just paradigm shift.

II. Meaning Holism

Churchland's position on the theory-ladenness of perception is intimately tied to his theory of concepts, which he derives from Sellars. On this theory, the meaning of a concept is fixed by its relations (inferential, evidential, etc.,) to other concepts in the system. Churchland often speaks as if the entire conceptual network constitutes a theory. Since everything within the network is interconnected and the meaning of any element in the network is determined by these interconnections, it looks like the fundamental unit of meaning is not the individual concept, but the entire system, that is, the whole theory. But this conception of meaning has some problems. Fodor notes that it is difficult to see how two theories could possibly dispute the truth of some proposition, since the proposition would literally mean something different in each theory.[17] Thus there could not be genuine disagreement. People with different theories would simply talk past each other. But, of course, there are genuine disagreements.

Additionally, on Churchland's kind of semantic holism, it seems that two theories that differ anywhere must differ everywhere.[18] And if this is the case, it poses severe problems for Churchland's case against CP. Churchland's idea is that a completed neuroscience will provide a conceptual framework from within which CP concepts will no longer be necessary. His basis for this idea is the remarkable explanatory power of current neuroscience. However, if Churchland is right about his theory of meaning, one consequence of this view is that within the current, pre-elimination conceptual scheme, the meaning of neuroscientific concepts are in part fixed by their relations to CP concepts. But then this means that if CP concepts are eliminated, then the meaning of neuroscientific concepts will change. Given such a change in meaning, the explanatory power of these concepts is called into question.

The point can be given an even stronger formulation. Since, on the holist view, the reference of a term is fixed by its relations in the network, if any term in the network fails referentially, then every term fails referentially.[19] If CP concepts are, as Churchland argues, pieces of fiction, referring to nothing, then any conceptual system that has CP concepts fails as a whole to refer. Since neuroscientific concepts currently reside side by side with CP concepts, the referential failure of CP concepts would spell the referential ruin of neuroscience. And not only neuroscience! As Devitt observes, "Given such truisms as 'If "cat" does not refer then there are no cats,' this loss of reference threatens loss of world."[20] This is, to say the least, drastic.

But let us take the weaker stance in which neuroscientific concepts do not undergo total failure (along with the rest of the conceptual system), but merely undergo meaning alteration post-elimination. In this case, it is sensible to ask whether these concepts will remain as explanatorily powerful as they were prior to CP elimination, or whether we will be unable to make sense of them absent their connections to the concepts of CP. The answer is not obvious if one accepts semantic holism. So it is incumbent upon Churchland to show that the coherence of neuroscientific theory in no way rests upon relations to CP concepts.

Churchland has admitted recently that "many 'pragmatic paradoxes' do indeed attend the eliminativist's current position."[21] This seems to be one of them. One possible way out is to argue that we do not possess a single monolithic conceptual scheme, but rather a plurality of more or less related schemes -- local wholes, partitioned according to domain of inquiry and explanatory purpose. Harold Brown, who shares with Churchland a Sellarsian conceptual role theory, argues as follows.

[A] language is not a single formal structure. Rather, we can view an individual as wielding a number of different languages that are essentially independent of each other, although each of these languages may share some common terms with other languages. A change in role of a term in one of these languages will have a ripple effect on the meanings of other terms in the language, but the effect will be relatively isolated. If I have learned both classical physics and biology, and my physical concept of time undergoes a transformation, my original biological concept of time can remain intact.[22]

This move from global holism to local holism is attractive. However, if Churchland were to accept it, it would give much of his case away to the CP theorist. Brown correctly identifies that different theories have different purposes. Biology and physics are different kinds of sciences which attempt to explain different kinds of phenomena. Since the concept of time serves one explanatory purpose in classical physics and another in biology, one need not crowd out the other. Now, the CP theorist can argue, along similar lines, that CP concepts serve a different explanatory purpose than neuroscientific concepts, and thus it would make no sense to eliminate one theory in favor of another. Just as biology is not physics, psychology is not neuroscience.

Indeeed, perhaps a more fundamental issue concerns the goal of having theories at all. That is to say, why do we want beliefs that match well with the world? Churchland does not address this, but it is clear what his answer is, at least in Scientific Realism. The purport of our theories is to mirror the ultimate structure of the world. By why should we be interested in this? It is obvious that most of us aren't. Will we all be happier and better satisfied with life if we each possess comprehensive knowledge of the Final Theory of Everything (and if this knowledge is reflected in our perceptual judgments)? Well, it's doubtful. Some individuals are deeply gratified by the possesion of scientific knowledge and all of us benefit enormously from the work of those who possess it. But the scientific minority is, after all, a minority. Since the goal of the scientist is unique to his station, there are surely principles of epistemic procedure uniquely suited to his endeavor. But they may not be equally well suited to other endeavors. Given her ends in the courtroom, the lawyer is best off thinking like a lawyer, not a physicist. And even the neuroscientist, qua lover, is apt to fail to make the right nerves fire if he continues to think of about himself and his lover too much in terms of nerve-firings.

So, even if CP is a theory, Churchland would still need to provide adequate criteria for theory evaluation and choice for the domain in which CP operates.[23] But he does not do this. He generalizes criteria for scientific theory choice to every domain, as if it were patently clear that these criteria will provide us with theories capable of doing the jobs we want done in every domain. But this is not patently clear. Churchland cannot simply assume that the criteria are the same across the board. He owes us an argument.

III. The CP-ladenness of Theory

Churchland stresses that theories are by no means idle things. They are integral to the actions of many kinds of people. But he also stresses that learning a theory isn't so much swallowing whole an inert body of facts as it is becoming inducted into the practices of an epistemic community and acquiring certain kinds of skills.[24] For the sake of argument, I propose to grant this point to Churchland. However, once this is granted, it should be well noted that the sort of socialization that attends Churchland's kind of integrated theory acquisition is, as a matter of fact, replete with CP enticements and methods of enforcement.

For example, standards of theory-guided practice in the sciences are enforced by hovering, tacit (or not so tacit) threats of ostracism or outright rejection from the community--tactics that appeal to the ongoing desires of the participants to become or remain members of the community. Participants are urged to believe that there is a special kind of honor and prestige attached to membership in the community, and that failure to conform to the standards of "good science" (by fudging data, or whatever) is immoral and dishonorable, and that one ought to feel shame when one breeches the code of the community. Now, the maintenence of these kinds of desires and beliefs in the members of epistemic communities seem to be so integral to the acquisition and sustenance of the theories embodied in the communities' practices that it is hard to mark a sharp divide between the CP attitudes and the theories. These attitudes are essential to the maintence of the community and the shared paractices and skills that by and large define it. When considered in this light, it is hard to understand CP as a theory in competition with theories in the hard sciences; rather, CP seems to be a necessary adjunct to theoretical practice. The success of theoretical practice is due in no small part to CP's ability to promote the goals and to police the standards of the scientific community. Consequently, these theoretical successes should be understood as a kind of indirect confirmation of CP's claims.

Call this the "argument from the CP-ladenness of theory". We can state the argument more clearly this way. If a theory is, as Churchland claims, more than a set of propositions, but essentially involves the internalization of skills, then a social process is required to inculcate these skills and a community is required to maintain them. However, it is an empirical fact that the social process of skill inculcation and the resulting epistemic community are both largely constituted by CP principles. Now, as we saw above, on Churchland's theory of concepts, a concept's meaning is determined by the inferences it licenses. So, if Churchland is right about concepts and if he is right the social character of theories, then if someone proposes a theory, we ought to be able to infer that they propose a certain social structure. And since the concept of a social structure includes principles and thus concepts of CP, someone who proposes a theory proposes the employment and, a fortiori, the acceptance of CP concepts. Thus, if someone proposes a theory that will eliminate CP concepts, she both proposes that CP concepts be accepted and eliminated, and this is inconsistent.

Notice that the argument from the CP-ladenness of theory is quite different from another charge of internal self-inconsistency frequently brought against the eliminativist. The argument goes like this: EM claims that the concept of belief ought to be eliminated, but also urges that we believe this to be true, and thus asks us to believe that there is no such thing as a belief, which no one can consistently do. Though I think that this objection is really better than some suppose, I agree with Patricia Churchland that we should be careful not to confuse a theory's truth with its unique availability, and that the apparent inconsistency may arise simply because we do not yet possess a new conceptual framework in which to criticize our current one.[25] The objection from the CP-ladenness of theory is different because it takes seriously several of the theses central to Churchland's case and purports to show that the joint acceptance of these theses leads to a inconsistency.

The argument from the CP-ladeness of theory resembles a complementary argument for CP advanced by J. D. Trout.[26] Trout argues that the use of CP concepts by scientists in the process of research puts CP under "theoretical stress." Let me paraphrase the main line of Trout's argument as I understand it. Science is cumulative and cooperative. It is impossible for a single scientist to check all of her colleagues' work. She must take much of it for granted (this is, by the way, why scientists so ferociously attack those who violate community standards). That is, an implicit hypothesis upon which her own work rests is the hypothesis that other scientists (the ones whose research she uses) desire to seek out the truth, are faithful to the standards of the scientifc community, have the relevant body of beliefs necessary to enable sound research, etc. Her attributes these CP states to other scientists is a basic assumption of her own work. When she tests her own hypothesis, these implicit CP hypotheses come under theoretical stress. Confirmation of the main hypothesis, provides indirect confirmation of the implied hypotheses under theoretical stress. So the success of CP in facilitating scientific research under conditions of theoretical stress demonstrates CP's reliablity and serves as indirect confirmation of CP's claims.

If Trout is correct, then his argument significantly strengthens the case for CP. I argued above that, given different epistemic goals, CP may be well justified on criteria of confirmation and theory choice different from those of science. If Trout is right, CP is well justified according to scientific criteria as well. Thus, it appears that CP is succesful in multiple domains (a domain being defined by a distinct set of epistemic goals). Rather than seek to jettison such multiply successful theories, reason implores us to embrace them.

I have argued that Chruchland's case against CP fails for a variety of reasons. His case for the theory-ladenness of perception has problems, as does his conceptual role theory meaning. Furthermore, his positions lead in multiple ways to paradoxes and inconsistencies. Finally, I attempted to show that CP plays an important regulative role in scientific practice and that the success of this practice redounds to CP as well as the theories it helps build and maintain. Thus, I believe there is ample reason to hang on tight to CP.


[1] I prefer "commonsense psychology" over the usual "folk psychology" because the latter seems to me rhetorically prejudicial. Back

[2] Paul M. Churchland, Scientific Realism and the Plasticity of Mind (Cambridge: Cambridge University Press, 1979), Chap. 2. Back

[3] Jerry Fodor, "Observation Reconsidered", Philosophy of Science 51 (1984), pp. 23-43. Back

[4] Ibid, p. 35. Back

[5] Ibid, p. 36. Back

[6] See John R. Searle, The Rediscovery of the Mind, (Cambridge, MA: MIT/Bradford Books, 1993). Back

[7] Fodor, p. 34. Italics are Fodor's. Back

[8] Ibid, p. 35. Back

[9] See H. Wimmer and J. Perner, "Beliefs about beliefs: Representation and constraining function of wrong beliefs in young children's understanding of deception," Cognition, 13 (1983), pp. 103-28 and S. Baron-Cohen, A. M. Leslie, and U. Frith, "Does the autistic child have a 'theory of mind'," Cognition, 21 (1985), pp. 37-46. Back

[10] Ibid. Back

[11] Robert M. Gordon, "Folk Psychology as Simulation," Mind and Language, Vol. 1 No. 2 (Summer 1996); Alvin I. Goldman, "Interpretation Psychologized," Mind and Language, Vol. 4, No. 3, (Autumn 1989). Back

[12] Gordon, p. 61. Back

[13] Ibid, pp. 61-62. Back

[14] S. Baron-Cohen, A. M. Leslie, and U. Frith, "Does the autistic child have a 'theory of mind'," Cognition, 21 (1985), pp. 37-46. Back

[15] Paul Churchland, "Eliminative Materialism and the Propositional Attitudes," Journal of Philosophy, 78, p. 78. Back

[16] On the "normal operation" of cognitive modules, see Ruth Millikan, Language, Thought, and Other Biological Categories, chaps. 1 and 2. Back

[17] Fodor, p. 27. Back

[18] Ibid. Back

[19] Michael Devitt, Coming to our Senses: A Naturalistic Program for Semantic Localism (Cambridge: Cambridge University Press, 1996), p. 21. See also David Papineau, "Representation and Explanation", Philosophy of Science 51, p.98. Back

[20] Ibid. Back

[21] Paul Churchland, "Evaluating our Self Conception", Mind and Language, Vol. 8, No. 2 (Summer 1993), pp. 211-222. Back

[22] Harold I. Brown, "Sellars, Concepts, and Conceptual Change", Synthese 68 (1986), pp. 275-307. Italics added. Back

[23] For a similar critique of eliminativism from the relativity of epistemic goals, and a good discussion of the relation of epistemic concepts to varying epistemic goals, see Paul Moser, Philsophy after Objectivity : Making Sense in Perspective (New York: Oxford University Press, 1993). Back

[24] Paul M. Churchland, A Neurocomputational Perspective: The Nature of Mind and the Structure of Science (Cambridge, MA: MIT Press, 1989), Chap. 9. Back

[25] Patricia Smith Churchland, Neurophilosophy (Cambridge, MA: MIT Press, 1986), p. 97. Back

[26] J. D. Trout, "Belief Attribution in Science: Folk Psychology Under Theoretical Stress", Synthese 87, pp. 379-400. Back