On the Very Idea of a Science Forming Faculty
It has been speculated, by Chomsky and others, that our capacity for scientific understanding is not only enabled but also limited by a biologically endowed science forming faculty (SFF). I look at two sorts of consideration for the SFF thesis and find both wanting. Firstly, it has been claimed that a problem-mystery distinction militates for the SFF thesis. I suggest that the distinction can be coherently drawn for cases, but that the purported ‘evidence’ for even a fairly lose general demarcation of problems and mysteries is not best explained by a SFF. Secondly, I consider in detail a range of cognitive considerations for the SFF thesis and contend that it is at best moot whether science can be so construed as to make it feasible that it is a faculty competence.
I feel most deeply that the whole subject
is too profound for the human intellect.
A dog might as well speculate on the mind of Newton.
From a letter of Charles Darwin to Asa Gray
Noam Chomsky (e.g., 1975b, 1980, 1988, 2000a) conjectures that our capacity for science is due to a biologically endowed science forming faculty (SFF): what lies within the principles of the faculty are problems, what lies beyond them are mysteries. The brief of the sequel is to question the very idea of a SFF. It must be said that Chomsky’s conjecture is speculative. Even so, he takes the idea very seriously, and I shall pay the due respect by doing likewise. Moreover, the notion is employed, by McGinn (1991, 1993) in particular, to argue for the substantive claim that consciousness is mysterious (McGinn, in fact, appears to think that more or less everything philosophers think about is mysterious.) If my contentions are anywhere near correct, while consciousness (or free-will, or personal identity, or meaning, etc.) might well be mysterious, it will not be because there is a human SFF that fails to accommodate it.
2: Problems and Mysteries: A Preliminary Characterisation
Chomsky’s notion of a SFF is tied to that of a problem-mystery distinction. I shall describe a strong distinction; there is a weak one, relativised to the here and now, but that is one no-one should want to deny.
In a strong sense ‘problems’ covers questions we could answer, events we could explain or otherwise understand, properties whose constitution we could discern, and so on. It will be noted that problem has a modal aspect: problems are not necessarily things we shall solve, they are things we could solve. For example, Fermat’s last theorem remained a problem for over 300 years until Andrew Wiles’s positive proof. Now consider the closest possible world W just like the actual world save that Wiles (or a counterpart thereof) gives-up on his proof with no-one continuing his research, and that W-humanity meets its end without ever knowing whether or not ‘xn + yn = zn’ has integral solutions for n > 2. Is Fermat’s last theorem only a problem in W, i.e., could W-humanity find a proof? Yes; for, on our assumption that all else is equal, the mathematics is available in W for Wiles’s proof, even though no-one gets around to employing it. The point is this: problems are demarcated relative to our cognitive capacity or reach, where such a capacity is abstracted from the contingency of what we happen to do or are interested in; it is, though, constrained by the myriad of contingent factors that have contributed to the development of our brains and will, presumably, continue to do so. This last point bears emphasis: it is not that some domains are so simple, while others are so damned complex; the issue is to do with what our minds are constitutively able to represent and explain, independently of whether a given domain is simple or complex in an objective sense - whatever such a sense might be.
Mysteries also have a modal aspect: they are insoluble, inexplicable in principle. Unlike problems, which may contingently evade resolution, mysteries lie beyond our understanding. Before the discovery of DNA there was no known mechanism to instantiate the heritable traits upon which selection works. Even so, heritable traits were not mysterious before 1953, as the discovery and subsequent theory demonstrated, they were merely problematic. Dark matter might be mysterious; then again, it might smoothly be accommodated within current particle physics. It might be that no reformation of set theory we could formulate will tell us whether or not 2À0 = À1, in which case the continuum hypothesis would constitute a mystery (here I forego any intuitionistic scruples.) Alternatively, the negation, say, of the hypothesis might be unsatisfiable in a any model for some theory which supersedes ZF(+C). At the moment, as with dark matter, there is no way of knowing. Such is the way with mysteries: at least at the present state of play, we cannot tell if we are dealing with a deep problem or a mystery. Further, mysteries outstrip problems in the sense that, while any problem is formulatable qua soluble, mysteries are not so minimally accessible: the notion carries no implication that we can so much as frame the appropriate questions. After all, if the answers are constitutively beyond our ken, we should not expect, in every case, to be able to pose the right questions in the first place.
My aim will not be to refute this strong distinction; I have no argument that there are only problems, no mysteries. Indeed, I think it far from implausible that there are domains, perhaps as yet not thought of, which are such that our brains are ill-equipped to deal with them. Yet this thought does not provide us with the kind of distinction to which Chomsky and others appeal. In particular, while the thought allows us to hypothesise coherently cases of mystery, it does not allow us a demarcation of them. Chomsky (2000a, p. 83), for sure, does not think that the distinction can be drawn sharply, but nor is the distinction meant to be so lose that it does not reflect a real feature of our cognition. That is, Chomsky’s point is not merely that we are epistemically bounded; it is, rather, that such a boundary is endogenously determined to some specifiable degree. This thought is crucial. Pinker (1997, p. 558-65), who otherwise commends epistemic boundedness, thinks that the thesis is “almost perversely unprovable” (op cit, p. 562). I would go further: if the mysteriousness of mysteries is itself mysterious, then we shall never be in a position rationally to conclude that such and such is a mystery. By drawing an endogenous boundary, Chomsky may be understood as attempting to remove mystery from mysteries and so explain our epistemic boundedness. The determining endogenous factor is a human science forming faculty (SFF).
3: The Limits of Thought
Chomsky conjectures that the broad shape of human scientific accomplishment is a function of an innate SSF. As an initial characterisation, we can think of our putative SFF as analogous to the language faculty. Here is Chomsky (1975b, pp. 155-6; cf. Chomsky, 1968/72, pp. 90-3; 1971, p.49; 1988, pp. 156-9), making use of Peirce:
The fact that “admissible hypotheses” are available to [the SFF] accounts for its ability to construct rich and complex explanatory theories. But the same properties of mind that provide admissible hypotheses may well exclude other successful theories as unintelligible to humans… though these theories might be accessible to a differently organised intelligence.
Thus, where the language faculty realises a universal grammar (UG) which allows for the generation of the grammars humans may acquire, so the SFF realises a set of concepts and principles that allow for the formation of all the possible theories humans may understand. UG empirically defines the notion of a possible human language (a grammar or I-language), but it does not follow that UG determines every possible ‘language’. Aliens, if such there are, will, we may presume, possess a quite distinct UG (or something else entirely) that determines languages inaccessible to us. Our UG is not a general purpose device to construct languages, it is severely constrained by principles which allow only a finite amount of variation: a human language is one which can in principle be deduced from UG principles given the setting of a finite number of parametric values. Thus, some ‘languages’ are mysteries for us, i.e., those not determined by our UG. The same thought applies to SFF. SFF is not a general purpose device which can construct a true theory for any domain. The set of theories it determines is drawn from a fixed conceptual resource with a finite number of principles defined over it. Now consider the set of true theories of the universe and its furniture and the set of theories determined by SFF. The intersection of the two sets is the set of (true) theories accessible to humanity; what falls outside the intersection is inherently mysterious. The intersection there is, is a “chance product of human nature” (Chomsky, 2000a, p. 83). There are a number of quite slippery issues to do with how close we should understand this analogy to which I shall return at length in §4; for the moment the sketch above suffices.
In this section I shall look at some considerations that are understood to militate for a problem-mystery distinction independently of the notion of a SFF, but which may be taken to buttress the SFF thesis in that a SFF would provide a natural explanation of them. That is, the SFF thesis is supported to the extent that it provides the best explanation of an independently coherent problem-mystery distinction. In the following sections I shall look at considerations specific to the SFF thesis.
Chomsky (e.g., 1993, 2000a,) is fond of reminding us that we are organisms, put together by evolution (not necessarily, natural selection), we are not angels. We are not designed, by God or anything else, to know all there is to know. Independent of the SFF hypothesis, then, to claim jointly that the truth about reality is unconstrained by our cogitations and that every truth falls within our understanding is to attribute to ourselves strange powers unprecedented in the biota. It would seem, therefore, that once even a modest realism is accepted, mystery follows, lest we think ourselves angelic (cf. Fodor, 1983, §V).
If we look at the rest of the animal kingdom, we find cognitive closure. A favoured example is that of the maze-solving abilities of rats (e.g., Chomsky, 1991b, p. 41; 1993, p. 45). Over a large range of mazes (e.g., radial ones) rats perform at levels equal to or greater than humans, but some mazes prove intractable. For instance, a prime maze is one whose solution depends on the subject making a certain decision (left or right) at each prime choice. Rats’ poor performance with such mazes is naturally explained by their lack of number theory. Of course, it does not follow that the average person would perform much better than a rat, yet the average person has the concepts which would enable her easily to solve the maze. Even if one lacked the explicit notion of a prime, one could still work-out the maze by ‘discovering’ the concept. This a rat cannot do.
The point of the analogy is that just as a rat will scurry around the prime maze, fated by its cognitive short-comings never to find the solution, so humans scurry around with their problems, fated in some instances to remain in ignorance, constitutively lacking the concepts which would provide the correct solution. We do not, for sure, appear to ourselves to be rat-like, but we well might from the perspective of “a differently organised intelligence”. To think otherwise would effectively be to hold that humans have no cognitive closure. This appears to be a supernatural property, but to accept cognitive closure, it seems, is to accept a form of the SFF thesis: only a certain range of concepts are “admissible” to us; we lack the capacity to frame other concepts necessary for the understanding of certain domains.
The rat analogy is certainly striking; it has the desired humbling effect. Neither the analogy nor the surrounding argument, however, oblige us to seek a SFF explanation of our apparent cognitive closure.
Notwithstanding the potentially significant differences between rats and ourselves (e.g., language, culture, technology, etc.), the analogy certainly lends force to the thought that, for any species, there will be insoluble ‘problems’. So much, however, does not lead us to the SFF thesis as the natural explanation, and nor, therefore, to a definite sense of mystery for humans. Of course, we are just another species, but such modesty obliges us to concede no more than our lack of omniscience. The analogy may well convince us that there will be some problems or other that we are just not fit to solve, but this gives no support to an endogenous demarcation between problems and mysteries. Simply put: the rat analogy militates for our epistemic boundedness, but it does not tell in favour of the SFF thesis. Chomsky and McGinn appear to conflate the two ideas, but they are quite distinct. The SFF thesis would certainly count as an explanation of our cognitive limits, but we can be cognitively limited without a SFF; more to the point, a SFF appears to be the ‘best explanation’ simply because it is read into the supposedly independent explanandum.
The source of this illicit conflation, I think, is the contrast between cognitive closure and supernaturalism. Chomsky assumes that if there is no cognitive closure that allows for a demarcation (to some degree of precision) of problems and mysteries, an identification of mysteries as mysteries, then humanity is potentially omniscient (especially see Chomsky, 1988, pp. 158-9). By modus tollens, he arrives at the desired result. This inference is never questioned, perhaps because there are those, after Peirce, who have thought that evolution has equipped us with a sure way to the truth (cf., Dennett, 1995, chp.13.) Also, Chomsky does at times appear to identify the two notions: the SFF thesis is a mere tag which dignifies our ignorance of the biological basis of our epistemic boundedness. But if this is all that is intended, then it is unduly presumptive to speak of a faculty, still more so, a faculty for science. However Chomsky intends to gloss ‘SFF’, which will be investigated below, to reject epistemic immodesty is not to commend an endogenous demarcation. Chomsky’s inference harbours an exhaustive disjunction - endogenously determined closure or omniscience - we should not accept. Lack of such closure does not entail omniscience or anything remotely supernatural.
For the purpose of questioning the entailment, let us assume that we do not have fixed conceptual resources. There is no SFF; instead, it is genuinely indeterminate what we may understand. This may be so if a completed neuroscience and cognitive psychology would not provide us with a list, as it were, of domains we may understand; rather, we find that, cognitively speaking, the brain contains some relatively autonomous components that follow a ontogenetic pattern as default, while others are much more inter-modal and differentiated. Our completed theories do not tell us if we can know what dark matter really is or whether the Continuum Hypothesis is true or false. Indeed, we cannot even tell what range of concepts, bound or unbound, the brain can support: each brain, it turns out, is different in significant respects. For my present purposes I need not suggest that this scenario is true. My argument only requires that it is consistent with what we presently know about the brain. The scenario certainly deserves such modest credit, especially given that our concern is with fine grained notions of individual concepts and hypotheses.
Now the above view is not one of cognitive closure in the relevant sense; equally, it patently does not imply omniscience or any other supernatural property. One’s possessing a SFF is not a necessary condition for one not being a god. There may simply be no answer to what we may or may not understand, at least none from a complete science of the mind/brain. Human nature may leave undetermined the limit of our cognitive reach. A corollary is that if we still want to ask, ‘In principle, what can humans understand?’, then we should acknowledge that the question is no longer to be construed as straightforwardly empirical. Rather, we are asking something like, ‘If humans with their current cognitive make-up were to carry on indefinitely, what domains would resist explanation?’ With the secure foundation of fixed conceptual resources removed, it is very difficult to begin to assess this question, for any answer will be sensitive to a myriad of factors: not only our cognitive structure, but also many exogenous factors: the kind of traditions that develop, the kind of stuff there in fact is in the universe, the technology we develop, the kind of assistance, if any, we receive from alien life-forms, and maybe just sheer luck. This is not to say that there are no mysteries, only problems; the appropriate conclusion is that the distinction between the two cannot be empirically grounded; rather than being an issue in cognitive science, it is a piece of futurology, interesting enough to speculate on, but not something to be greatly exercised about. The dialectical moral is: a rejection of strong closure is not conceptually or empirically concomitant with our deification; far from the problem-mystery distinction being an independent notion the SFF thesis naturally explains, it is the thesis that motivates the supposed cognitive division. This conclusion should not be surprising: one can hardly expect to arrive at specific theses about human cognition from broad inchoate observations.
I shall shortly look at some arguments which seek to support the SFF thesis directly; before doing so, I shall look at another consideration - failure - of a more general nature.
Unlike McGinn (1991) on consciousness, Chomsky does not affect to know whether this or that domain is mysterious, even so he appeals to potential mysteries such as linguistic creativity. No-one is yet in a position to say that consciousness or linguistic creativity are definitely not mysteries; that would require coherent theories of the phenomena, something none of us possess. I do not think, though, that our historical failure to explain these phenomena or any others intimates that there is a SFF that lacks the appropriate conceptual resources. I shall argue for this negative thesis by suggesting that a history of failure may be properly explained in more modest terms.
Chomsky, of course, does not take the failure of previous accounts of creativity to demonstrate mystery; ditto for McGinn (1991) and Nagel (1986, 1995) vis-à-vis consciousness. Nevertheless, an inductive comfort is felt in past failings; they are taken to be “suggestive” or indicative that the time has come to give-up. We should, however, not be moved to pass from failure to mystery.
Patently, no amount of failure allows us to infer mystery. But in what way, then, is failure suggestive? Charitably, the history of science is one of equal proportion of failure and success; and where there is success, failure always threatens as research programs wax and wane and data accumulate. If we are to be moved by simple failure, we might as well declare the universe and all that’s in it a complete mystery. After all, science is not in the proof business. For failure to intimate mystery, the lack of success must be peculiar.
A mark of potential mystery to which some have appealed is that we, as it were, ‘stare blankly’ at a problem, nothing is forthcoming. This characterisation, however, is hardly descriptive, it is a judgement on the efforts made or, worse, an assessment of the authors’ own efforts. No problem induces blank stares, whether literally or metaphorically, in everyone. Consciousness certainly does not as the groaning book shelves and increasing number of ‘centres’ and conferences testify. Of course, one is free to think that such output does not amount to much more than a blank stare, but one would thereby be offering a slanted evaluation, not a neutral criterion of mystery. Alternatively, failure might take the form of an absence of science: a domain is identified, but neither a methodology nor predictive/explanatory theories are produced. Failure on such a scale would certainly intimate that something is grievously amiss, but, again, we are far from a suggestion of mystery.
Prior to Darwin, it is fair to say that while there were theories of evolution (witness Lamarck and Geoffrey), they did not provide sound mechanical explanations of the origins and inheritance of traits which lead to species diversity and similarity. Indeed, the very idea of species evolution was tendentious; perhaps the then dominant view in biology was the neoclassical one that dismissed the very idea that one species may ‘change’ into another. A reasonable person might well have declared, and many did, that life was a mystery, the province of divine ordinance. With the re-discovery of genes and the discovery of DNA, Darwin’s theory is now the background for modern biology. Such has been this success that the very idea of an elan vital is now as egregious as that of a res cogitans. This transformation from not even a recognition of evolution to advances favourably comparable to those of post-Galileo physics took just over a hundred years. Thus, there is precedent for ‘blank stares’ to metamorphose quickly into paradigmatic science. It is always too soon, it seems, to gainsay intellectual advance.
More tendentiously, the human sciences, in contrast to the physical sciences, exhibit a failure to progress and in many cases predictive or explanatory hypotheses are not even sought. An assessment of the human sciences by the present criterion might lead one to think mystery endemic in the human domain. Would this be a reasonable conclusion?
Well, is Homo more complex, mysterious even, than DNA, quantum mechanics, analysis, relativity theory, etc.? We are encouraged to think so merely on the basis of the lack of scientific success. We have, however, no clear, neutral sense of what conceptual complexity amounts to, still less a domain-independent metric of it. The relevant variables for any interesting social problem might be too astronomical to control for, but this would not constitute a mystery in the present sense. Friendly aliens might lend us their super computers. Consider: a four-colour-like theorem might be unprovable in the absence of computer assistance, but it would not therefore be mysterious. Perhaps the problem with Homo is more mundane.
Chomsky (1979, p. 57) himself has likened the methodology of sociology to butterfly collection: lots of interesting data, if one likes that kind of thing, but nothing approaching explanation. Chomsky (1968/72, pp. 24-6) also suggests that a deep problem with the scientific investigation of that we are most familiar with is that we think we already know the facts, and so waste our efforts trying to systematise and explain what are in reality chimera. Chomsky’s assault on behaviourism is a paradigm of the required process of defamiliarisation. Perhaps we are awaiting a similar development vis-à-vis consciousness. In short, a more modest judgement is that there is nothing inherently mysterious about Homo, the problem is that we continuously adopt the wrong approach. It is an interesting historical question why this should be so, but there need be no portent of mystery. Indeed, what is probably most inimical to the progress of the human sciences is the unfortunate, though perhaps unavoidable, usurpation of method by political agendas, both left and right. Again, this is something Chomsky has taught us.
I should not suggest that the retrieval of a criterion of mystery from a history of failure is impossible, but I fail to see how it could be reasonably made in the face of historical precedent and more modest explanations. Let us, however, move to Chomsky’s particular considerations for the existence of a human SFF rather than mysteries in general.
4: Faculties and Science
Let us assume that cognition is not served by a general purpose device; instead, the mind is divided into a number of dedicated devices that support a range of competencies and capacities. Think of the nomenclature ‘faculty’ as (intentionally) picking out such devices in terms of the domain-specific principles and concepts particular to them. This somewhat fuzzy characterisation is for a reason.
Fodor (1983, 2000) reads Chomsky’s notion of a faculty epistemologically rather than functionally, i.e., a Chomskyan faculty is a body of information a subject knows, not an architectural component. I think Fodor is right in as much as Chomsky’s theories are not processing stories, as many assume. One could, therefore, accept that h-theory, binding theory, et al. are innate while holding that the mind is a general purpose device. Fodor, however, is not quite right. Chomsky uses ‘faculty’ (and ‘organ’ and ‘module’) ambiguously (somewhat like his use of ‘grammar’): it sometimes denotes a body of information, that which we cognize, at other times it denotes the cognitive mechanism that supports that information and interfaces with performance systems (information cannot interface with anything.) This is not sloppiness: for Chomsky, there is no question about the ‘psychological reality’ of a grammar apart from its explanatory worth; if it proves so worthy, then the grammar is an accurate, albeit abstract, intentional description of a yet unknown physical mechanism. Chomsky is assuming (contra the generalist) that there are dedicated devices (brain areas) without speculating upon their operational specification; and so the devices do not individuate the faculties (see Chomsky, 1988, pp. 7-8). Consequently, a ‘Chomskyan’ faculty is a loser notion (qua intentional) than Fodor’s (1983) modules (effectively, input systems). In particular, faculties need not be automatic, inaccessible, or encapsulated. They do, though, follow a biologically fixed maturation process: given a uniform initial state and experiential input, they determine a final state that supports a mature competence. This final state may also be uniform, as it appears to be with, say, the theory of mind faculty; or it might be non-uniform, as it clearly is with the language faculty, i.e., different input determines different I-languages. As indicated, the bare idea of modules or faculties admits great variation (more of which below), let us though stay with the sketch at hand.
Now if we conjecture that the human mind has a faculty architecture, some diagnostics have to be in play so that we may identify the faculty based competencies, for not all competencies are so supported. For example, line dancing, car repair, origami, chicken sexing, etc. are all competencies, but we are not moved to posit, say, a line dancing faculty. But we do posit faculties for language, theory of mind et al. This difference provides an angle on the diagnostics appropriate for a faculty competence.
Faculties are fixed as part of our biological endowment; the principles therein specified are thus innate, unlearnt. This gives us some ready diagnostics. First, a candidate faculty based competence must be uniform across the species within intelligible bounds of difference; it cannot be a culturally specific capacity. In short, the competence must be a trait of the species. Second, the competence must follow a fairly strict ontogenic course; for since the blueprint of the development of the competence is genetically coded for, the competence should be invariant across a wide variety of experiences. Explicit teaching, for example, should not make a significant difference to the speed of the development or the final competence arrived at. Third, the competence and its development should, to some degree, be invariant over various pathologies, injuries and differences in intelligence. A faculty F is a device dedicated to a specific domain; disturbance to another faculty, therefore, should not necessarily lead to disturbance to F. Pace Fodor (1983, 2000) and his modules, it is perfectly coherent to view some faculties as enjoying proprietary interfaces with one another, while others may work in isolation (Collins, 2000). Fourthly, the competence should reach normal maturity in the face of a poverty of stimulus. This diagnostic is essentially just another way of saying that the competence acquired is underdetermined by the data available to the child. After all, if a competence were determined by some learning regime or a certain set of stimuli (no matter how complex), it would be redundant to claim that it is supported by a faculty, for the competence could apparently be acquired independent of any prior principles or concepts specific to the competence’s domain (here I exclude general principles of, say, association, if such there be).
It should be transparent that the diagnostics delineated do not fit line dancing, etc. The diagnostics do fit linguistic competence and face recognition, and a good although still highly controversial case can be made for them fitting theory of mind. Let us hypothesise, then, that these diagnostics are indeed criterial of a faculty competence. Do the diagnostics identify our scientific endeavours as faculty based? This is a difficult question because Chomsky is, I think, somewhat unclear on how the notion of a SFF is to be understood.
The problem is this: we can give ‘science’ a strict construal under which it primarily covers our paradigms of successful scientific theories. Under this reading, a SFF is a kind of theory selector, determining those domains in which we can achieve some success. Alternatively, we may construe science in a liberal way as covering any thinking (practise) that is guided by certain meta-principles (supra-empirical virtues). So read, a SFF is simply the seat, as it were, of the set of principles which enter into our construction and evaluation of theories. Now while a case can be made, I think, for the view that some such principles are innate and uniform throughout the species, it also seems obvious that such principles are domain general, not specific, and that they support neither the problem-mystery distinction nor the associated model favoured by Chomsky whereby our SFF is supposed to determine a subset of the set of true theories. If, faced with such difficulties, we revert to the first, strict construal of science, so that the putative SFF meets these demands, then we lose the positive readings on the diagnostics. Either way, therefore, the SFF thesis appears to be in some disrepair.
Let us first look at the strict construal, which is, I think, highly implausible; it does, though, have its interest. Science, we might say, is paradigmatically represented by the theories found in text books of say Newtonian mechanics, statistical thermodynamics, general relativity, Bohr’s atom theory, etc., where generalisations are sought that are explanatory and predictive of future cases on the basis of postulated unobservables that unify otherwise disparate phenomena. Under this construal, a SFF may be understood as a device that gives us access to those concepts and principles required for fecund generalisations over the domains in question (e.g., atoms or the structure of space) while other domains remain closed to us, for our SFF simply lacks the appropriate conceptual resources. When Chomsky and McGinn claim that our SFF determines but a subset of the set of possible scientific theories, and that we may determine what domains are mysterious to us, it is difficult not to have such a construal in mind. My initial analogy between the SFF as and the language faculty (and the attendant quotation from Chomsky) followed such a line, for it is one that makes perfect sense of the problem-mystery distinction. This conception, however, appears to tell us that the kind of science typical of the West for the past 400 or so years is as cognitively determined as language is. If this is the conception, then it is surely mistaken.
Science, as exemplified by, say, general relativity theory, is a fairly recent product of Western culture, there is no evidence whatsoever for its being a species trait. Nor, of course, does it have an ontogeny: normal human maturation does not produce scientists; it requires a great amount of explicit instruction for one to grasp the theories characteristic of the last few centuries. Equally, being a scientist, so to speak, is not invariant under differences or changes of intelligence or cognitive capacity: no-one expects scientific competence to be selectively spared or impaired and there is, of course, no evidence for any such pathological profile. The reason for this is that grasping and working with a theory appears to require a battery of competencies and capacities: distinct kinds of reasoning (e.g., deductive and analogical), good long and short term memory, mathematical and linguistic knowledge, experimental design, etc. So much I take to be indisputable. Finally, the poverty of stimulus diagnostic does not apply either (I shall separately look at this diagnostic in §5 with reference to the more plausible liberal construal.) We learn scientific theories, we do not acquire them from partial and degraded data. Indeed, to acquire a theory we typically need to be inundated with stimulus (lectures, text books, conversations, experiments, etc.), and even then we consistently make all kinds of errors. Perhaps, then, Chomsky has something different in mind.
Consider: “The basic elements of rational inquiry may have some of the properties of such cognitive systems as the language faculty, though the ways they are employed are surely quite different: scientific knowledge does not grow in the mind of someone placed in an environment” (Chomsky, 1980, p. 140). Quite! Notwithstanding the apparent support the strict construal receives from Chomsky, here he seems to be advancing what I earlier called the liberal construal under which science is simply a kind of thinking marked by our predisposition to judge according to certain principles. By this reading, our SFF, in some sense, leads us one way rather than another through the space of theories, but it does not code for any such route, we can not, as it were, read-off the theory of natural selection, say, from the neonatal brain. Where we are at a given period will be a function of a background of past theories, especially the successful ones, but this history and future progress is shaped or canaled by the kind of answers our SFF permits according to its principles. This is the model Chomsky appears to present in his 1988, chp. 5. Chomsky (2000a, pp. 82-3; also 1980, p. 140) offers empirical test, elegance, and criteria of intelligibility, as potential candidates for such principles; we may add simplicity, exhibition of causal structure (‘Mill’s methods’, perhaps) and other such meta-empirical notions.
As I indicated above, while this latter construal does not suffer from the same impairments as its restrictive counterpart, it does have its own problems. The first thing to note about the proposed principles is that they are not domain specific: simplicity, elegance, testability, etc. are applicable to any field. Thus, if a SFF has no domain specificity, then it really makes no difference whether one says that the history of science has been shaped by our SFF or, vacuously, by our thought. Put only slightly otherwise, a SFF would simply be for thinking as such. Indeed, Fodor (1983, 2000) takes cognition which is governed by such global principles to be precisely that kind of thinking which is not domain specific, in contrast to the cognition that is served by modules with their proprietary databases. Of course, the operative notion of a domain is somewhat vague; still, one reason why language and face recognition, say, appear to be faculty competencies is that their domains are so idiosyncratic: our proficiency in the domains calls for specific information about verb structure, vertical symmetry of eyes, mouth, et al. In contrast, the principles under consideration appear to have no idiosyncratic domain. From finding one’s way home or finding a lost sock to arranging a wedding or building a kennel for the dog requires principles of reason and testing, even if, perhaps, only in the imagination. Any rational belief fixation requires some constraining, otherwise, we would be afflicted with the frame problem, which, in point of fact, we never are. Consider, specifically, the notion of causal structure, which appears to have a strong innate basis (Sperber, et al., 1995): we like theories to give us causal mechanisms; such is why, inter alia, Einstein gives us a better theory of gravitation than Newton and why no-one but cranks take morphic resonance to be a serious hypothesis. Equally, however, we impose causal structure on everything we come across: if we cannot discern a causal pattern, we tend to retire in bemusement, but not always. Sometimes we forego causal structure (see below).
It might be that we can delineate science-specific notions of simplicity, elegance, etc., but we cannot assume that there are such principles without begging the question at issue, for such an assumption amounts to a presumptive specification of a SFF. Moreover, we have no independent good reason to think that there are any science-specific abductive principles. As it stands, therefore, ‘SFF’ is a misnomer; for why speak of a faculty for science when precious little is excluded? The whole point of faculty theorising is to divide and conquer, to isolate specific competencies and attempt to see what kinds of peculiar information and principles best explain the observed proficiency. There is no theoretical gain in hypothesising a faculty which appears to serve (more or less) the whole of thought.
How, then, are we to understand the faculty-ness of the SFF? Perhaps we are employing ‘faculty’ too precisely; Chomsky (1975b, pp.155-6; 1988, pp. 156-9) does explicitly draw the analogy between language and science, but it is clearly not meant to be a tight one. The issue here is the extent to which a SFF can be domain neutral (unlike language) without becoming indistinguishable from a vacuous notion of general intelligence or rationality. Let us look at some potentially pertinent proposals.
On Fodor’s (1983, 2000) view, the mind/brain broadly divides into some components that are domain specific (for Fodor, these are modules that serve input cognition: vision, olfaction, parsing, etc.) and others that are domain general (or perhaps just one), these serving central cognition, i.e., rational belief fixation. Although it is rarely, if ever, noted, the theory theory view (e.g., Gopnik and Wellman, 1994, and Gopnik and Meltzoff, 1997) is close to Fodor’s position to the extent that both resist the view that central, rational thought is modular, i.e., rational belief fixation is not served by an ensemble of dedicated, domain specific, encapsulated components; still less do Fodor and the theory theorists commend a module dedicated to science. Instead, both posit innate domain general principles of theory formation and confirmation which, we may presume, have shaped the history of science. So, is this not a tale of a SFF? Fodor himself is non-committal as to the organisation of central cognition: organisation there surely is, but without new concepts, we haven’t a clue how to account for it (for Fodor, 1983, p. 107, “the more global [i.e., less modular] a cognitive process is, the less anybody understands it.”.) Even so, the principles of belief fixation captured by our putative principles are, for Fodor, innate and can be selectively impaired (e.g., it follows from Fodor’s account that rational belief fixation could be impaired while linguistic competence remains intact.) Where Fodor demurs on principle, the theory theory view can be understood as offering a story about belief fixation (see Gopnik and Meltzoff, 1997, pp. 63-7). Peripheral modules output the data which forms the evidence for theory construction on the basis of an initial innate structure, guided by meta-empirical principles. A proper assessment of Fodor’s account and the theory theory view is beyond my present scope; fortunately though, at least for my dialectic, neither approach is flush with the SFF notion Chomsky appears to favour.
Chomsky clearly does not view faculties as theories that are developed according to domain general principles; on the contrary, he eschews the very idea of anything approaching a general intelligence in favour of common-sense faculties, a mathematics faculty, a musical one, etc. Chomsky (1988, p. 47-8) avers: “in any domain…, specific… capacities enter into the acquisition and use of belief and knowledge”; general mechanisms, “if they exist” enjoy, at best, a “doubtful” role. Chomsky (e.g., 1980, p. 135) certainly rejects the application of domain general theorising with respect to linguistic development, which, for Chomsky, has nothing whatsoever to do with finding simple and elegant hypotheses or analogical reasoning or any hypothesis testing at all.
Might a SFF be understood as an abstraction from, or construction out of, a collection of domain-specific faculties or theories? A SFF, by such a suggestion, would be constituted from our endowed folk understanding of biology, physics, etc. This suggestion might be finessed by appeal to Carey and Spelke (1994). They rightly acknowledge that explicit science is quite different from our developmental theorising (if such is what we do) precisely because it is not restricted to, nor directly constrained by, the domain-specific ‘core’ knowledge which may reasonably be thought of as innate. But they also picture science proper as enabled (and so limited?) by analogical mappings across the innate domains which constitutes ‘constructed’ knowledge. Well, what makes Carey and Spelke’s hypothesis, and perhaps other such bootstrapping models, provisionally sound is that the notion of analogical mappings allows scientific thinking to come under the purview of developmental cognitive science without egregiously restricting science to what the child naturally arrives at. The notion, however, is also so lose as to be of scant help in constructing a workable notion of a SFF. After all, anything is analogous to anything else. I cannot imagine what the evidence would look like for the claim that the admissible theories are restricted by innate analogical possibilities. For example, we tend to analogise on the basis of our most complex machines (water pumps, clocks, computers, etc.), but none of this is innately specified in the relevant respect. Indeed, if we consider the extent to which explicit theorising in any given domain has departed from our intuitive outlook, then it seems that the restriction our ‘natural’ view places on science is approximately zero. The data there are to support the existence of ‘science’ faculties indicates that they are of a distinctively Aristotelian stripe, an outlook long since rejected in every area of our understanding of the natural world (see, e.g., Keil, 1989).
It bears noting that if Chomsky were commending the SFF thesis as a story about the principles of central cognition, then he would be quite inconsistent. Chomsky’s language faculty is not a Fodorian peripheral module; rather, it is a system that is for thinking (if anything), it is not a mere parser. Chomsky, then, cannot consistently hold to the claim that central cognition (the seat of thinking) is governed by domain general principles. There is, however, a more fundamental problem with the idea of a SFF answering to either the Fodor or theory theory view.
Chomsky’s discussions of the feasibility of a SFF are always presented in the context of the problem-mystery distinction. The two ideas appear to be mutually supporting: the SFF hypothesis offers a cognitive explanation of the distinction; without it we would have no independent ground to say of any domain that it is mysterious. Concomitantly, the supposed intuitive coherency of the distinction (as discussed in §3) gives credence to the view that science is not the play of a general, unbounded intelligence. Neither the Fodor nor theory-theory view support such a position on the problem-mystery distinction.
Fodor (1983, §V) certainly thinks that there are mysteries; he reasons that since cognition is innately structured, there are endogenously determined limits on the kind of hypotheses we can entertain. Crucially, however, this has little to do with Fodor’s particular architecture of modules and central systems: any view that gives cognition a fixed architecture of information control and access is bound to admit the possibility that the world might throw up a problem which cannot be answered by our minds. This reasoning, though, is in line with my happy concession in §3. There I suggested that human thinking, like that of any other organism, is most certainly epistemically bounded. But this is just to admit that we are not potentially omniscient; it is not to concede that the structure of cognition demarcates between theories, determines a subset of true theories. Fodor’s account, then, does not give us the kind of cognitive explanation Chomsky expects. Indeed, Fodor agrees with Chomsky that if cognition is thoroughly modular, then we shall have a clear demarcation of problems and mysteries, but Fodor’s (1983, 2000) key claim is precisely to deny the antecedent here: that the mind is not massively modular is what makes it mysterious!
The theory theory view also fails to give Chomsky the support he wants for the problem-mystery distinction. The point here is straightforward and is independent of the details of the theory theory approach. The general principles at issue are comparative ones: they help us decide between theories or hypotheses, they do not produce the theories for us. It makes very little sense, for example, to say that we only select simple theories. We favour the simpler theory, ceteris paribus. The rider accommodates the fact that we happily neglect the dictates of a given principle, if so doing gives us a greater all round fit with the other principles: as regards causal structure, the move from Cartesian to Newtonian mechanics is an example; so is, perhaps, the development of quantum mechanics last century. Being comparative, these principles cannot preclude certain hypotheses or theories from consideration, they can only advice us against them once they are on the table, as it were. Consider, for example, one of Chomsky’s (2000a, p. 85) speculative examples: dark matter. There are a number of proposals on the market, involving, variously, the size and shape of the universe, the presence of super-massive (hence, super dense) black holes, hitherto undetected elementary particles, etc. Now let us hypothesise that the nature of dark matter is mysterious for us. In what way could this be a function of our meta-empirical principles? It is certainly not to the point to say that the true theory is too complex. It might be too complex, but this would have nothing to do with our favouring simplicity; after all, if the correct theory is too complex to entertain, then we cannot get around to judging its simplicity relative to another theory. What if the truth of the matter did not admit to a causal explanation? Again, this would not necessarily portend mystery. A variety of indeterminacy hypotheses have been happily entertained and accepted in the absence of a settled interpretation (read ‘mechanism’), notoriously, the collapsing wave packet in quantum theory. We like causal mechanisms, but we can and do forego them; mutatis mutandis, I submit, for the other principles. The principles help us decide between the theories on the table, but they will not reject all the theories, still less provide a licence for us to say that no theory will do. Again, therefore, the theory theory approach, even if it were otherwise acceptable to Chomsky, does not provide for a cognitive explanation of the problem-mystery distinction.
The position we have arrived is that while the strict construal of science is patently inadequate to satisfy any of the faculty diagnostics, it clearly does make sense of the problem-mystery distinction in the way Chomsky sets it up. On the other hand, while the views we have just been considering do offer the kind of principles relevant to theory construction and assessment that may well be innate (if not quite faculty-like), they do not offer a cognitive ground for the theory demarcation Chomsky wants. Perhaps, as it seems, Chomsky wants it both ways. It must be kept in mind that Chomsky’s SFF is an unabashed speculation and, for all we presently know, there might be innate principles that are rich and specific enough to determine the set of theories we may access. As it stands, I think the proposal falls between the two stools of the strict and liberal construals.
In the next and final section I shall look at Chomsky’s specific argument for a SFF based upon the poverty of stimulus diagnostic. This separate treatment is apposite, for Chomsky, at least in one place, appears to understand our fourth diagnostic as the crucial one.
5: Induction and Convergence
For Chomsky, the most telling indicator that a competence is faculty based is its satisfaction of our fourth diagnostic. We have already seen, though, that science is not a competence in the way that language is, say: our acceptance of the theory of geodesic planetary orbits is patently not underdetermined by data in the same way that our acceptance of the principles of binding theory is. In what respect, then, does the fourth diagnostic militate for a SFF? Chomsky (1975b, p.24-5; cf. 1971, pp.49-50) has contended that without a SFF “it would be impossible for scientists to converge in their judgement on particular explanatory theories that go far beyond the evidence at hand,… while at the same time rejecting much evidence as irrelevant or beside the point, for the moment at least” [my emphasis]. The nature of the inference here appears to be that if our convergence is to be possible (not a miracle), we need extra-empirical principles to weed out all but a few of the contrary theories that we could otherwise find to comport with the data; a SFF is simply the seat of such principles. I must say that I find this argument to be very weak indeed; before demonstrating why, however, a word of caution is in order.
Chomsky’s early views were certainly motivated by Quine and, especially, Goodman’s work on induction: Chomsky agreed with many others that unconstrained induction is untenable, whether as a model of learning or the norms of science, but he also rejected any empiricist band-aids to cover the problem (see Chomsky, 1975a, Introduction). Chomsky, to my knowledge, however, offers the above direct argument to a SFF only in the two places cited above (see below for a qualification); his more recent discussions simply associate the SFF thesis with the coherency of the problem-mystery distinction (e.g., Chomsky, 1991b, p. 41; 1993, pp.44-5; 2000a, pp.82-3). Perhaps, then, Chomsky’s considered view is that there is no inference from the inductive practise of science to the existence of a SFF (cf., Chomsky, 1980, pp.139-40). Whatever the case may be, the argument is worth considering, for it purports to offer precisely the backing the SFF thesis requires.
Chomsky’s concern, of course, is with the cognitive basis of our rationality, not with the clarification of the concepts of validity or justification. That humans, especially in their scientific mode, are concerned to cleave to rational norms, tells us a simple truth about human thought: there is a slack between our receipt of data and the convergent beliefs we arrive at. Such a slack, however, amounts to no more than the fact that we are not blank slates. Crucially, the possibility remains open that the innate equipment that enters into our ability to take up the slack is not the set of the methodological principles that govern self-conscious scientific investigation and even if it is, it does not follow, as made clear above, that the equipment makes up a faculty. All that does follow is that the scientist needs some innate equipment to be so much as a thinker as opposed to a S-R device. Let us see this in some detail.
An underdetermination thesis UT says, for some set of methodological cannons C, that data-set D does not confirm or corroborate a theory T on the basis of C at the expense of all other contrary theories. Now if we do in fact converge on T, then, relative to UT, any potential addition to C would override the underdetermination problem, i.e., UT would not show that there is no rational justified choice to be had. It might be that every theory is underdetermined by every possible D given any C, but I know of no argument that attempts to show that this is so. To assess any given UT then, we need to ask whether the associated C is reasonable or realistic. If the answer is ‘No’, then we do not have a sound underdetermination claim; if the answer is ‘Yes’, then we do. So, Chomsky is perfectly correct in thinking that some C is required, but the requirement is not based on a need to overcome underdetermination; C is needed so that our theories may rationally confront the data in the first place, whether or not the confrontation leads to underdetermination. Where, then, does the claim come from that the scientists’ C must be innate? For all that has been said, it is still open to think of C as the product of our thought, rather than being our thought as such, as it were. It looks as if the claim that scientists’ C is the innate content of a faculty floats free of any underdetermination thesis.
Consider Hume on induction. Hume demonstrated the deductive underdetermination of theories (hypotheses); that is, for any D, if theory T is confirmed by entailing D, then there are contrary theories that are equally confirmed. This notion of underdetermination, however, amounts to the now trivial claim that scientific inference is non-demonstrative or, so as not to exclude the Popperian, deduction cannot amount to justification. As regards Chomsky’s inference to a SFF, if this is the only underdetermination a scientist must face, then all that follows about convergence is that it cannot be explained on the assumption that the scientist’s cannons are wholly deductive. For example, theories T1, T2,…Tn might all entail D, but if T1 were the only theory consistent with an associated favoured theory Tn+1, then it would be rational to choose T1. Otherwise put, Chomsky’s argument can be spiked by so little as an appeal to a single cannon of reason (e.g., choose the theory that least disrupts your other commitments) that goes beyond the entailment of D; there is not a whiff here of an inference to an innate SFF. I do not mean to suggest, of course, that any such cannon(s) would in fact uniquely single out a theory in any epistemic scenario; my point is only that deductive underdetermination leaves such an option open.
As indicated above, a notion of underdetermination closer to what Chomsky has in mind is the one due to Goodman’s (1954/83) ‘new riddle of induction’. Familiar details aside, Goodman shows that if hypothesis H (e.g., ‘All emeralds are green’) is confirmed by its observed instances (i.e., green emeralds), then a contrary hypothesis H* (‘All emeralds are grue’) is also confirmed by the same instances, where ‘grue’ means ‘either green if first examined before 2050AD or blue if not first examined before 2050AD’. The riddle is that we unerringly take the ‘green’ hypothesis to be confirmed (to some degree) by the observation of emeralds, even though the grue hypothesis is confirmed to the same degree by the emeralds. If we did not so converge, we would have no shared sense of laws, natural kinds, explanation, prediction, etc., but the basis for the convergence appears to be prior to the actual framing and corroboration of hypotheses. Do we need the SFF thesis to explain this?
To keep things finite, let us assume that Goodman’s claim is well-founded. What Goodman shows is that enumerative induction is not sufficient to give a hypothesis a unique degree of confirmation. So, if we wish to explain convergence on, say, ‘green’ rather than ‘grue’ we have no greater license from the riddle than to add to C (= enumerative induction). This, of course, is precisely what Goodman did, for good or ill, with his historical notion of entrenchment: roughly, we converge on ‘green’ because, unlike ‘grue’, it has been successfully projected in the past. Alternatively, Harman (1994) proposes a practical principle of simplicity to rule out ‘grue’. Now, these and many other ‘solutions’ do not necessarily speak to the ‘genetic problem’ of how humans do in fact converge (Chomsky, 1971, p.6). But the same point holds for the riddle itself: it does not so much as indicate the shape of a cognitive solution, still less necessitate one along faculty lines. At best, the riddle shows that enumerative induction is inadequate as either a model of justification or, indeed, cognition. This is an instance of the general moral: underdetermination arguments are negative; they work against a given set of methodological cannons, they do not establish the identity of the cannons which in general are required for rational convergence.
I have not insisted that we should read Chomsky’s direct argument from convergence to a SFF as being a gloss on Goodman’s riddle. For sure, Chomsky (1975a, pp. 33-4; 1971, pp.6-8) does appeal with perfect legitimacy to the riddle as a central plank in the argument against a general empiricist model of leaning, but no direct association with a SFF is made. Perhaps, though, Chomsky, at least in (1975b) where the direct argument is made, is conflating the SFF thesis with Fodor’s position on concept learning. The proceedings of the 1975 Royaumont conference (Piattelli-Palmarini, 1980) suggest this. Fodor and Chomsky there argued that the moral of Goodman’s riddle is that any induction is “logically impossible” without an “a priori [innate] ordering of hypotheses [or predicates]”; Fodor took this to be “so self-evident that it is superfluous to discuss it” (Piattelli-Palmarini, 1980, pp. 259-61). Now I do not so much think that this line is false, as woefully underspecified. What is “self-evident”, let us grant, is that blank-slate induction is impossible. Goodman (1954/83, p. 82) himself claimed this in arguing that a mere habituated fixation on a regularity a la Hume can not establish which predicates are the projectible ones, because green and grue instances are equally regular by definition. The point is well taken but it hardly leads us to credit the scientist with a SFF, even one which consists of a predicate metric alone. The inference is blocked because the scientist is patently not otherwise a blank slate. Such, indeed, is Goodman’s point: scientists’ choices are a function of history. As regards the developing child, if one is running a hypothesis confirmation model of learning, as Chomsky (1965) was with language and Fodor (1975) was considering with concepts, then one faces the problem of explaining how the child fixates on a given hypothesis when the data does not determine such a choice. As with underdetermination generally, the answer is to be found in a methodological organon C. In the learning case, both Chomsky (1965) and Fodor (1975) take C to consist, inter alia, of a simplicity metric defined over an innate set of grammars and predicates respectively. For present purposes I have no problem with such proposals, although both Chomsky (1981) and Fodor (1981) rejected them in favour of triggering models (such is perhaps why Chomsky ceases to appeal to Goodman after the development of the P&P approach.) My point is simply that the scientist faces a quite different underdetermination problem from the child: the child, on the assumption that she is a hypothesis confirmer, needs some C prior to the data; this is enough to undermine blank-slate empiricism in favour of an indeed self-evident nativism. The issue for the scientist, on the other hand, is what, if anything, can go into C to enable one to arrive rationally at some hypothesis, but there is no argument here to say that whatever C comprises must be what the child has; a fortiori there is no argument which says that the scientist’s C must be innate. I should say that if Chomsky and Fodor were guilty of this conflation of the child with the scientist at Royaumont, the confusion did not last. There is some dispute about this.
Putnam has persistently attributed to the ‘Chomskyan’ (a nomenclature apparently co-extensive with ‘Fodorian’) an ambition to find an inductive algorithm which would explain our scientific reasoning, an ambition that is spiked by Goodman’s riddle. In point of fact, though, even in his (1975) Fodor was not arguing for a theory of inductive concept learning; his claim was that if that is one’s model, then one is de facto committed to an innate inductive logic (his major point, of course, was that such a model does not amount to concept learning - there is no such thing - but to belief fixation.) Fodor’s (1981) triggering proposal is by definition non-rational: a trigger is caused, it isn’t warranted. Since then, Fodor (1983, 1987, 2000) has claimed that we have not a hope of a prayer of discovering any general inductive algorithm, because our computational theory of mind does not account for non-modular processes, viz. abductive ones. Chomsky (1980, p. 140; cf. Piattelli-Palmarini, 1980, p. 320) agrees with Fodor (see quotation in §4.)
Well, if Chomsky and Fodor do not run child and scientist together, we appear to be left with no argument from underdetermination to a SFF as the only possible explanation of convergence. Notwithstanding Chomsky’s choice of modality for his direct argument, let us relax his inference to a best explanation one. The thesis now becomes quite tempting, for it seems that its only competitor is a sociological account, which threatens to make scientific convergence no more rational than our allegiances to football teams. The choice between cognition or society, however, does not exhaust the options.
Let us agree that individual scientists are rational, they hold to their theories because they judge them to be true or at least well corroborated. So much, of course, does not explain convergence, but can we not say that agreement arises due to individuals being trained within a research program and their theoretical energies being spent therein? Otherwise put, scientists employ their cognition to a highly restricted space whose shape is due to exogenous factors; convergence is explained by ‘society’ determining the options from which cognition chooses. Well, this appears quite inadequate. Chomsky’s SFF thesis is offered as an explanation of convergence, an appeal to research programs seems to be simply a re-description of the explanandum; we want to know why there are such programs in the first place. Such a riposte is apt for those who take the sociological to be primitive, but that is not what I am suggesting.
Since scientists are rational creatures, as we all are, then endogenous factors clearly have a role to play, but the fact of convergence does not lead us to any particular thesis about those factors. It is perfectly intelligible to appeal to a central, global rationality along the lines of Fodor (1983, 2000) or perhaps some ensemble of faculties. We must be careful, though, not to run child and scientist together again. There may be deep similarities as the theory theory teaches, but while it is true, I think, that children’s commonsensical convergence is principally due to cognitive factors, the same explanation does not work for scientists. It only takes one person to produce a hypothesis that answers certain questions or makes novel predications that the extant competitors fail to do. This relative success means that the research gets taught, it has possible technological side-effects, it attracts funding, it is popularised to a general audience,… it snowballs. In short, the norm is for scientists to converge on given extant hypotheses; the norm is not for them, as it is for children, individually and creatively to converge, although sometimes it happens. Again, there is nothing irrational in following a good idea. Individual scientists are not drones to the program, they simply tend to be unconcerned in their day to day activities with formulating novel theories, for those they have work very well and there is much interesting testing and tweaking to be done. After all, if there is nothing obviously wrong with one’s theory - it withstands attempted falsifications and continues to explain novel data - then one would be irrational to forsake it.
If this model is anywhere near correct, then individual scientists within communities, both through their education and in their maturity, do not face the array of possible theories with data in hand and plum for just a few out of the indefinite options. Were this the case, a SFF would indeed be required, unless we thought science a miraculous affair. But scientists are always somewhere, occupying a theoretical position. As such, the choices they make are of the form ‘I would rather be here than there’, or vice versa. They are not nowhere, child-like, deciding where they wish to be, but faced with underdetermination where ever they turn. A scientist will stay where he is, and he is rational to do so, because his theory does enough good work. If the work dries up, or some other theory does more work better, or (indeed!) is simpler, or more elegant, or integrates better with some other accepted theory, or if its mathematics is more user friendly,… then, ceteris paribus, he will be rational to vacate his theory. I fail to see, then, the SFF thesis as the best explanation of convergence: the rationality of science is not exhibited by convergence in the face of underdetermination, it is manifest in the subtle interplay of factors which determines when a scientist should move or when he should just stay put.
6: Concluding Remarks
I have not sought to refute the SFF thesis either empirically or conceptually. At our current state of knowledge, the former route is unavailable and to follow the latter one would simply be to misunderstand the issue. Moreover, I have ran with the speculation, offering a number of substantiating proposals and batting off some ill-founded ripostes. Charitably construed then, my claim is just that, on reflection, it is unclear what the SFF thesis amounts to, more work needs to be done before we can seriously treat it as a hypothesis. I wish, though, for my conclusion to be slightly less modest: the deep problem with the thesis is its presupposition of a universal individualism, the idea that the endogenous factors which shape any of our practises are specifiable independently of, and have primacy over, the exogenous ones. Such thinking is what drives the problem-mystery distinction and it is this that gives so much sustenance to the SFF thesis, so much so that Chomsky nigh conflates the two. I have argued in respect to both problems and mysteries and scientific convergence that a two way street approach is at least as viable for something as broad and amorphous as science. Individualism is a sound assumption, I think, when we are dealing with a competence that has the look of a faculty, but science is not such a competence. Still, who knows… Would that we were all capable of the Chomskyan speculations that have turned out to be true.
 Chomsky (e.g., 1988, p. 158) has since dropped the appeal to Peircean abduction (see below).
 Chomsky (1965, p. 56) does contend that grammars which are not generated from UG may still be acquirable through our more general problem solving capacity. Even so, there is no guarantee that any, still less all, ‘alien’ languages will so succumb.
 Ironically, while Chomsky understands our biological nature virtually to guarantee our epistemic boundedness, he also speculates that UG is a perfect solution to the engineering problem of fitting language to the legibility conditions imposed by the other systems of the mind; but perfection is not a property found elsewhere in nature. The moral should perhaps be: Don’t infer properties of particular organisms from general claims about the biota; one always has to look at the particular. I shall not, though, press the moral, for the difference between the two cases is what is important. See Chomsky (1995) for the perfection speculation at work; for more informal discussions, see Chomsky (2000a, chp.1; 2000b).
 Such a view is perhaps close to that of Dennett (1995), Churchland (1989), and Clark (1996). More generally, an empiricist theory of cognition would tend towards indeterminacy about what can be known, whereas a rationalist or Kantian one would tend towards determinacy. My point, though, is orthogonal to this traditional divide. Cognitive design space is vast and we know very little about the area humans occupy, as such, any inference from non-omniscience to architecturally specifiable epistemic limits is quite unsafe.
 Chomsky (1986, 1991a) dubs the question of creativity Descartes’s problem: How are we able to use language for the free expression of our thoughts? Ironically, precisely because we have some very good ideas about the structure of language, a better case can be made, I think, for the mystery of language use than can be for consciousness. Thanks to the great advances made in linguistics, we have a quite precise working notion of linguistic competence, and much corroborating data. Against such a background we might reasonably hope for at least a working explanation of creativity. Yet this we do not have.
 McGinn (1991, p. 7): “longstanding historical failure is suggestive, but scarcely conclusive”. Nagel (1995, p. 97): “The various attempts to carry out this apparently impossible task [i.e., explaining consciousness] and the arguments to show that they have failed, make up the history of the philosophy of mind during the past fifty years”. Pinker (1997, p. 562): “the species’ best minds have flung themselves at the puzzles for millennia but have made no progress”.
 The now standard view is that, before Darwin, a dogmatic and degenerating Aristotelian essentialism prevailed (see, e.g., Mayr, 1982; for a dissenting voice, see Depew and Weber, 1995). What is certainly true is that after Darwin (ultimately, the New Synthesis) evolution is not a phenomenon to be seriously disputed and, furthermore, natural selection is recognised as the principal mechanism of change, if not the whole story.
 Fodor (2000, p.62-3) does entertain what might be called distributional encapsulation, where modules enjoy (architecturally constrained) access to other modules’ databases, but such organisation is clearly exceptional for Fodor.
 See Segal (1996) for, to my mind, a sound defence of the modularity of theory of mind.
 The closeness is not even noted by Gopnik and Meltzoff (1997), who run Fodor together with the central module crowd in evolutionary psychology (Ibid, p. 58), while also suggesting the theory theory view can tell a story about Fodor’s central system! They also fallaciously conflate Fodor’s parser module with Chomsky’s language faculty (Ibid, Chp. 2, passim) (see below, especially n.15).
 One striking correspondence is that Gopnik and Meltzoff (1997, pp. 26-7/53) claim that the innate principles (prediction, explanation, etc.) which govern theory formation also explain scientific convergence. This is one of Chomsky’s key claims (see §5).
 Gopnik and Wellman (1994) and Gopnik and Meltzoff (1997) follow Karmiloff-Smith (1992) in assuming that modularity (the faculty approach) has trouble explaining development. The thought is wholly confused. Of course modules develop, the point is that the crucial determinant of development is the normal maturation of the brain under normal stimulus conditions. Chomsky makes this point every time he puts pen to paper. It is the theory theory approach, I should say, that is in trouble (see n.14).
 Pinker (1997, Chp.5) appears to share this view insofar as he explicitly rejects a faculty for science, but (Chp.8) argues that we are epistemically bounded due to our core endowed knowledge. Pinker’s reasoning here, however, unlike that of Chomsky, is that natural selection has constrained what we might understand to being less than the whole truth.
 There is a serious question as to whether the theory theory approach is distinctive enough to give us the required domain general principles. The problem is this. The child is understood to fixate on certain evidentially constrained theories; the child’s brain is built to arrive at them. This is evidenced by the fact that the theories are more or less uniform across the species. But thesis and evidence are here unstable: to account for the uniformity, the theory theorist must enrich the principles the child employs, but the more rich they become, the less evidence the child requires and the less theory-like the ‘theories’ become. It is therefore moot, I think, whether the theory theory view offers a coherent alternative to the faculty approach. In short, the child just ain’t like the scientist, or vice versa. (see Leslie, 2000 for other deep worries.)
 For Chomsky’s explicit rejection of the Fodor view, at least regarding language, see Chomsky (1986, p.14, n.10; 1991a, pp.19-21; 2000a, pp.117-8).
 E.g., Fodor (1983, p.123) thinks that even a wild-eyed generalist position such as Hume’s is still endogenously restricted because such ‘minds’ can only access information derived from perceptual input.
 For the association of Chomsky with inductive logic, with references to Goodman’s ‘refutation’ thereof, see Putnam (1981, pp. 125-6; 1983, p. viii; 1988, pp. 82-3; 1992, pp. 14-6).
 I am in the debt of two anonymous referees for insightful constructive criticism of an earlier draft.
Carey, S. and Spelke, 1994, “Domain-specific Knowledge and Conceptual Change.”
In L. Hirschfeld and S. Gelman (eds.), Mapping the Mind: Domain Specificity
in Cognition and Culture (pp. 169-200). Cambridge: Cambridge University Press.
Chomsky, N., 1965, Aspects of the Theory of Syntax. Cambridge, MA: MIT Press.
Chomsky, N., 1968/72, Language and Mind (extended edition). New York: Harcourt
Chomsky, N., 1971, Problems of Knowledge and Freedom: The Russell Lectures.
New York: Pantheon.
Chomsky, N., 1975a, The Logical Structure of Linguistic Theory. Chicago: Chicago
Chomsky, N., 1975b, Reflections on Language. New York: Pantheon.
Chomsky, N., 1979, Language and Responsibility: Based on Conversations with
Mitsou Ronat. New York: Pantheon.
Chomsky, N., 1980, Rules and Representations. New York: Columbia University
Chomsky, N., 1981, Lectures on Government and Binding. Dordrecht: Foris
Chomsky, N., 1986, Knowledge of Language: Its Nature, Origin, and Use. New
Chomsky, N., 1988, Language and Problems of Knowledge: The Managua Lectures.
Cambridge, MA: MIT Press.
Chomsky, N., 1991a, “Linguistics and Adjacent Fields: A personal View.” In A.
Kasher (ed.), The Chomskyan Turn (pp.3-25). Oxford: Blackwell.
Chomsky, N., 1991b, “Linguistics and Cognitive Science: Problems and Mysteries.”
In A. Kasher (ed.), The Chomskyan Turn (pp. 26-53). Oxford: Blackwell.
Chomsky, N., 1993, Language and Thought. London: Moyer Bell
Chomsky, N., 1995, The Minimalist Program. Cambridge, MA: MIT Press.
Chomsky, N, 2000a, New Horizons in the Study of language and Mind. Cambridge:
Cambridge University Press.
Chomsky, N., 2000b, The Architecture of Language. New Delhi: Oxford University
Churchland, P., 1989, A Neurocomputational Perspective: The Nature of Mind and
the Structure of Science. Cambridge, MA: MIT Press.
Clark, A., 1996, Being There: Putting Brain, Body and World Together Again.
Cambridge, MA: MIT Press.
Collins, J., 2000, “Logical Form, Theory of Mind, and Eliminativism.” Philosophical
Psychology, 13, 465-490.
Dennett, D., 1995, Darwin’s Dangerous Idea: Evolution and the Meanings of Life.
Depew, D. and Weber, B., 1995, Darwinism Evolving: Systems Dynamics and
Natural Selection. Cambridge, MA: MIT Press.
Fodor, J., 1975, The Language of Thought. Cambridge, MA: Harvard University
Fodor, J., 1981, “The Present Status of the Innateness Controversy.” In
Representations: Philosophical Essays on the Foundations of Cognitive Science
(pp. 257-316). Cambridge, MA: MIT press.
Fodor, J., 1983, The Modularity of Mind. Cambridge, MA: MIT Press.
Fodor, J., 1987, “Modules, Frames, Fridgeons, Sleeping Dogs, and the Music of the
Spheres.” In J. Garfield (ed.), Modularity in Knowledge Representation
and Natural-Language Understanding (pp. 25-36). Cambridge, MA: MIT Press.
Fodor, J., 2000, The Mind Doesn’t Work That Way: The Scope and Limits of
Computational Psychology. Cambridge, MA: MIT Press.
Goodman, N., 1954/83, Fact, Fiction, and Forecast (fourth edition). Cambridge,
MA: Harvard University Press.
Gopnik, A. and Meltzoff, A., 1997, Words, Thoughts, and Things. Cambridge,
MA: MIT Press.
Gopnik, A. and Wellman, H., 1994, “The Theory Theory.” In L. Hirscheld and S.
Gelman, (eds.), Mapping the Mind: Domain Specificity in Cognition and
Culture (pp. 257-93). Cambridge: Cambridge University Press.
Harman, G., 1994, “Simplicity as a Practical Criterion for Deciding what
Hypotheses to Take Seriously.” In D. Stalker (ed.), Grue! The New Riddle of
Induction (pp. 153-171). La Salle: Open Court.
Karmiloff-Smith, A., 1992, Beyond Modularity: A Developmental Perspective on
Cognitive Science. Cambridge, MA: MIT Press.
Keil, F., 1989, Concepts, Kinds, and Cognitive Development. Cambridge, MA:
Leslie, A., 2000, “How to Acquire a Representational Theory of Mind.” In D. Sperber
(ed.), Metarepresentations: A Multidisciplinary Perspective. Oxford: Oxford
Mayr, E., 1982, The Growth of Biological Thought: Diversity, Evolution, and
Inheritance. Cambridge, MA: Harvard University Press.
McGinn, C., 1991, The Problem of Consciousness. Oxford: Basil Blackwell.
McGinn, C., 1993, Problems in Philosophy: The Limits of Inquiry. Oxford:
Nagel, T., 1986, The View from Nowhere. Oxford: Oxford University Press.
Nagel, T., 1995, Other Minds. Oxford: Oxford University Press.
Piattelli-Palmarini, M. (ed.), 1980, Language and Learning: The Debate between
Jean Piaget and Noam Chomsky. Cambridge, MA: Harvard University Press.
Pinker, S., 1997, How the Mind Works. London: Penguin.
Putnam, H., 1981, Reason, Truth and History. Cambridge, MA: Harvard University
Putnam, H., 1983, “Forward.” In Goodman (1954/83): pp. vii-xvi.
Putnam, H., 1988, Representation and Reality. Cambridge. MA: MIT Press.
Putnam, H., 1992, Renewing Philosophy. Cambridge, MA: Harvard University Press.
Segal, G., 1996, “The Modularity of Theory of Mind.” In P. Carruthers and P. Smith
(eds.), Theories of Theories of Mind (pp. 141-57). Cambridge: Cambridge
Sperber, D., Premack, J., and Premack, A. (eds.), 1995, Causal Cognition: A
Multidisciplinary Debate. Oxford: Clarendon Press.