My distinction between fitting and guiding is, you see, the obvious and flat-footed one. Fitting is a matter of true description; guiding is a matter of cause and effect. Behaviour fits a rule whenever it conforms to it; whenever the rule truly describes the behaviour. But the behaviour is not guided by the rule unless the behaver knows the rule and can state it. This behaver observes the rule ( Quine: Methodological Reflections on Current Linguistic Theory).

Quine begins his critique of Chomskian linguistics by distinguishing between two different types of rules: Fitting and Guiding. He claims that Chomsky uses a third intermediate type of rule; this is a type of rule which Quine claims is an implicitly guiding rule. Quine claims that Chomsky thinks that ordinary speakers of English are guided by rules even though these rules cannot be stated by the English speaker. According to this Chomskian picture, we can have two extensionally equivalent grammars, each of which fits the behaviour of the child, neither of which explicitly guides the child, and only one of which is true of the child. Quine claims that if this intermediate way of following a rule is to be made sense of, we need some type of evidence which will help us decide which grammar the child is implicitly following. He claims that a person’s disposition to behave in determinate ways, in determinate circumstances, is the way to make sense of which grammar the person is following. However, obviously these dispositions must go beyond well-formedness if we are to use them to explain the distinction, because extensionally equivalent strings are indistinguishable from each other in terms of behaviour. He speculates that such dispositions which may be relevant are those such as the disposition to make certain transformations and not others; or certain inferences and not others. Quine further notes that he has no problem with dispositions; and points out that a body has a disposition to obey the law of falling bodies, while a child has a disposition to obey any and all extensionally equivalent grammars. However, he can make no sense of Chomsky’s intermediate notion of rule following. He alludes to the ironic fact that while Chomsky seems to have no difficulty with the obscure notion of implicit guidance by rules, he has serious difficulties with the humdrum notion of disposition.

Chomsky replied to this criticism by first questioning the analogy of a child following the rules of grammar with a body obeying the law of falling bodies. He argued that this is a singularly misleading analogy because the rules of English grammar do not determine what a speaker will do in a given context in the same way that the law of falling bodies will determine, if a person jumps from a building, he will hit the ground at a specified time. According to Chomsky, what the rules of English grammar tell us is that English speakers will understand and analyse certain sentences in certain ways, and not in other ways. In other words, the linguist in Chomsky’s sense is trying to discover regularities in a person’s linguistic competence; the linguist is not after a theory of performance. This reply of Chomsky’s has two different strands to it: first, his distinction between competence and performance; second his claim that we can have no scientific theory of performance. His argument that by using idealisations, such as abstracting from memory and performance, we can gain some understanding of subjects like phonology etc. has something to recommend it. Idealisation plays a vital role in any science, and the appropriateness of an idealisation is to be judged by the success of the science which uses it. Given that generative linguistics has had some success, we can tentatively accept his idealisation. However, Chomsky’s confident assertion that we cannot have a science of human performance/behaviour has little to recommend it. Quine’s claim that people are obeying extensionally equivalent rules if their performance conforms to those rules is a perfectly legitimate claim. And Chomsky has offered us no reason to think otherwise.

Chomsky adds that even if we leave aside the purported disanalogy between the law of falling bodies and linguistic competence, Quine’s formulation would still fail because he is guilty of treating physics and linguistics inconsistently. What Quine should have said if he wanted to remain consistent was:

English speakers obey any and all of the extensionally equivalent English grammars in just the sense in which bodies obey the law of falling bodies or the laws of some other system of physics that is extensionally equivalent (Chomsky: Reflections on Language p. 188)

Chomsky claims that when put in these terms, we see that linguistics is no more in need of a methodological cure than physics is. Chomsky’s argument here is merely repeating the criticism which he made against Quine’s Indeterminacy of translation argument. He is claiming that Quine is guilty of treating underdetermination as fatal in linguistics but as harmless in physics.

Quine’s criticism was that Chomsky was using a third type of rule which was obscure because it went beyond fitting the behaviour of the person, but was not consciously guiding the behaviour of the person. Chomsky notes that people are rarely, if ever, guided by rules in the sense of being able to state the rule they are following. Furthermore, he argues that linguists can go well beyond rules that merely fit the behaviour of people. He says that if we accept the same realistic attitude towards linguistics that we do with physics, then we can say that people are obeying rules which are really encoded in their brains, but of which they are usually not consciously aware. So we can use various different pieces of evidence to decide between extensionally equivalent grammars.

To make his criticism of Chomsky’s conception of rules explicit, Quine discussed a toy example of two extensionally equivalent grammars which a person could obey. He asks us to imagine a string (abc) which can be divided up as extensionally equivalent grammars (ab) (c) versus (a) (bc). He notes that from a behavioural point of view, we could say that a person is following either of the grammars and that if we ask the native which is the correct grammar he will not be able to tell us. In this situation, he claims that Chomsky’s view that the person can be said to be implicitly guided by rule (ab) (c) as opposed to (a) (bc) is senseless.

Chomsky argues that that there is no mystery in deciding whether the child is implicitly guided by rule system (a) as opposed to (b). In, this situation, he argues all we have is a problem which can be solved by the ordinary methods of the natural sciences. Quine had argued that a natural solution to deciding whether (ab) (c) or (a) (bc) was the rule which implicitly guided the people, would be to ask them which rule they followed. He notes, of course, that the people will not be able to tell us. So he argues that since the natives cannot decide which rule they are implicitly following, and both rules are compatible with the dispositions of the ordinary language user, then the notion of one of the rules implicitly guiding the child is senseless. Chomsky agrees that asking the native will not get us far in deciding between rule (a) and rule (b), and he suggests a different way of dealing with the problem. He cites various different types of evidence which could theoretically bear on the problem. He discusses how the linguist could have evidence which suggests that intonation patterns are determined by structure. This evidence could be derived by studying our own language as well as other languages. He argues that such evidence might bear on the choice between the two proposed grammars. So we might discover that the rules needed for other cases give the correct intonation if we take the constituents to be those of (a) instead of (b). Whether this type of evidence really occurs or not is irrelevant. Chomsky’s point is that Quine is incorrect to assume that no evidence on the topic is attainable beyond evidence gained by questioning the natives.

Chomsky cites a quote from Quine’s paper ‘‘Linguistics and Philosophy’’ where Quine claims that there is much innate apparatus which will need to be discovered to tell us how the child gets over the great hump that lies beyond ostension. Quine further notes that if Chomsky’s anti-empiricism says nothing more than that conditioning is not sufficient to explain language learning, then it is of a piece with his indeterminacy of translation argument. Chomsky notes that if Quine really believes that there is further innate apparatus waiting to be discovered by science, then this casts doubt on his claims in both his ‘‘Methodological Reflections’’ paper and in Word and Object. Chomsky asks us to consider the sentence: ‘ABC’. Quine had claimed that we have no evidence which can help us determine whether a subject is implicitly following the rule (ab) (c) or (a) (bc). However, according to Chomsky, given that Quine has no problem with innate mechanisms of any sort, then surely it is possible we will discover an innate mechanism in our species which determines that we follow (ab) (c) as opposed to (a) (bc) or vice versa. So in this respect, he claims, Quine (1969) holds doctrines which run counter to the doctrines he accepts in (1960) and (1970).

Quine would deny this charge of Chomsky’s. What Quine has said is that he has no problem with innate apparatus of any sort as long as such apparatus can be made sense of behaviourally. When he is discussing the notion of rule following, he is claiming that if two grammars both equally fit the behaviour of the subjects, and each subject is not consciously guided by these rules, then to claim that the person is implicitly guided by the rule is senseless. The postulation of innate apparatus which ensures that the child follows the rule (ab) (c) as opposed to rule (a) (bc) is pointless unless we have some behavioural evidence to justify such a postulation. The type of evidence which Chomsky claims could be useful to the linguist in deciding whether the subject is following rule A or B, i.e. perceived structure in intonation, is behavioural evidence so it does not bear on the type of considerations which Quine is concerned with.

A Chomskian could argue that if Quine would accept evidence such as structure intonation as a way of distinguishing between extensionally equivalent grammars, then he would be committed to accepting the Chomskian intermediate notion of a rule which implicitly guides the subject. Quine would of course rightly deny this. He would claim that, if we assume for the sake of argument that Chomsky’s made-up example of structure in intonation pattern is correct, then it would follow that the two extensionally equivalent grammars do not fit the totality of linguistic behaviour of the relevant subjects. He could argue that his flat-footed distinction between the two different types of rules can accommodate the type of example which Chomsky put forth. Quine could further claim that Chomsky has not fully explained his notion of the third intermediate notion of a rule.

This debate about the notion of a rule is a real point of contention between both thinkers. Chomsky has tried to explicate his intermediate notion of a rule as something in the brain which unconsciously guides the language user. This conception of a rule does not fit in neatly with Quine’s two conceptions of a rule; the question is whether Chomsky’s third notion captures something real which is missed by the two Quinean notions.

For simplicity’s sake, let us consider a particular rule of English, the subject-auxiliary rule which states that when forming a question, one must move the main auxiliary to the front of the sentence. According to Chomsky, the subject auxiliary rule is a rule of English (and Spanish) though not of all languages; so it is a rule which requires some experience in order for acquisition of it to take place. Children are only capable of learning this rule because they are genetically programmed to follow certain linguistic universal principles such as the following one: All languages are structure dependent.

On the Chomskian conception, the child is guided by certain rules which he does not know and cannot state (unless he has knowledge of linguistic theory). The question we now need to ask is why Quine would have difficulties with these rules, and can such rules be accommodated within his flat-footed conception of rules? Quine’s emphasis on reinforcement and induction means that he would more than likely feel that the postulation of such innate rules is unnecessary. Nonetheless let us assume that Chomsky is correct that poverty of stimulus considerations dictate that rules such as structure dependence are known innately and that parametric variations of these universal rules result in particular languages being spoken. The question which needs to be asked at this point is whether we can make sense of the notion of ‘rule’ which Chomsky is postulating here?

When Quine wrote his ‘‘MRCLT’’ in the early 1970s Chomsky was operating with a rule-based conception of language. From about 1980 onwards, with the inception of his principles and parameters approach, Chomsky’s conception of language changed. Though the change from a rule-based approach to a principles and parameters approach was a significant empirical advance, it does not have any effect on the criticisms which Quine brings to bear in his ‘‘MRCLT’’. So let us consider a question such as the following one: ‘Is the man who is happy over thirty?’ which is derived from the statement ‘The man who is happy is over thirty’. When forming this question, Chomsky claims the child begins with the statement and unconsciously applies the rule: move the main auxiliary to the front of the sentence. According to Chomsky, none of this is done consciously; rather, what happens is that a mechanism in the brain which is genetically programmed will interpret the data of experience and construct a model. The mechanism will determine what the rules of the language are, using the universal principles of the language and having the parameters set through experience. When these parameters are set, we can say that the child is unconsciously following various different rules of English, such as the subject/auxiliary inversion.

Suppose we assume (falsely) that Quine would accept Chomsky’s poverty of stimulus argument and agree that rules such as structure dependence are innate and that these rules determine that ordinary humans will derive rules such as auxiliary inversion when placed in certain linguistic environments. The question we now need to answer is how would Quine characterise these facts? The key point here is that Quine would focus on performance facts; he would ask whether people are disposed to form questions in ways consistent with these rules. Answering these questions would involve studying various different corpuses and seeing how people actually form questions when talking with others, or when writing various different texts. It would also involve constructing various different controlled experiments to see whether the manner in which questions are formed varies in the circumstances of speaking/writing. What Quine would object to is the reliance of Chomsky et al. on the intuitions of others as to what is or is not an acceptable way of forming questions. Quine would claim that the salient aspect of our studies should be the behaviour of the subject, not the intuitions of acceptability or unacceptability by the subjects of various constructions.

However, the next question to ask is what Quine’s response would be if the behaviour of the subjects and the acceptability tests lined up perfectly. So Quine would have the following facts to account for[1]: poverty of stimulus considerations, subject’s intuitions of the acceptability or unacceptability of certain constructions, and subject’s behaviour in certain determinate circumstances. In this circumstance, Quine would claim that the person’s behaviour fits with any and all extensionally equivalent grammars which capture the behaviour of the subject. He would say that we are justified in claiming that the person’s behaviour conforms to a particular rule system (and other extensionally equivalent systems). He would have to object to any postulation of innate apparatus because there is no justification for postulating one rule system over another one. He would accept any innate apparatus if it was determined by behavioural facts; however, no behavioural facts will help us decide between attributing rule system (A) over rule system (B), if they are extensionally equivalent. So on this Quinean picture, the rules which we claim the behaviour of the child fits do so in the same way the behaviour of physical objects fits certain rules of physics. So rules discovered in this behaviourist manner would easily conform to Quine’s flat-footed conception of rules fitting the behaviour of the subject. There would be no need for the Quinean to postulate an intermediate type of rule which implicitly guides the subject.

There is, however a difficulty with this Quinean conception of the behaviour of the child fitting certain rules. The difficulty stems from normative issues. The Chomskian picture conforms to our pre-theoretical intuitions about language in one clear sense. It seems obvious that when I construct a new sentence, the sentence will be grammatical or ungrammatical according to the rules of the language. So, for example, most people would believe that if I construct a new sentence there will be a fact of the matter as to whether the sentence is grammatical or not. However, if we are deriving our rules of language by studying how people actually perform, then at best, all we can speak of is the probability of an utterance with a certain syntactic structure occurring or not. We will have no warrant to claim that the sentence is grammatical or ungrammatical. From a behavioural point of view, we can say that the sentence is atypical but not that it is incorrect. This seems like a serious difficulty with Quine’s conception of the nature of rules. He could reply to this concern by claiming that we could test whether the sentence was grammatical by the lights of the linguistic community by asking its members. However, this reply does not solve our problem; rather, it reintroduces the problem in a different guise. If we ask people whether this or that construction is acceptable, we are testing their intuitions in the same manner that Chomsky recommends. If people’s intuitions all agree that a certain construction is ungrammatical, but performance data indicates that the construction is used and accepted in ordinary speech, then we have arrived at an impasse. Furthermore, if people’s intuitions of acceptability disagree with their actual performance, then Quine would argue that we would have to give preference to performance data over people’s intuitions of acceptability. The problem with this is that if people’s intuitions cannot be used to help us decide between a correct or an incorrect linguistic utterance, then we seem to have no way of doing so. All we have are linguistic regularities, and constructions which are irregular from that point of view.

To clarify the above difficulty let us consider a computer. From a performance point of view, the computer exhibits certain regularities. So, for example, if I press the caps lock key when typing in Microsoft Word it results in the words I type afterwards being capitalised. When I press the caps lock key again, the computer no longer types in capital letters. Now imagine if one day I press the caps lock key and the following symbol appears on my screen *. Imagine I continued pressing the key for a while and the symbol * kept appearing. You can be quite sure that if this happened, nobody would think to themselves ‘Strange, the probability of my letters being capitalised when I press the caps lock button has just been reduced’. It is a safe bet that anybody who noticed that the caps lock key being pressed resulted in * being typed would assume that the computer was broken. This assumption would be based on the fact that we know that the computer was designed for a particular purpose which it is no longer achieving efficiently. So we would assume that some part of the computer was broken and set about getting it repaired.

Now in natural language such an approach is possible as well. Take, for example, people with severe schizophrenia, or with some form of aphasia. People with schizophrenia sometimes speak with what is known as word-salad. Such word-salad sentences sometimes exhibit syntactic, semantic and pragmatic deviance. People with some forms of aphasia are sometimes incapable of forming sentences into syntactic units. Analysing sentences from schizophrenics and aphasics is no trivial matter, and understanding the way such sentences go wrong is a flourishing field of psycholinguistics and neurolinguistics. Any theory which claimed that the speech of aphasics and schizophrenics was just statistically unlikely rather than incorrect seems to be seriously deficient. Furthermore, as opposed to the case of the computer, we cannot say in this case that such people are in error because they are behaving contrary to their designers’ intentions. Who designed English that we would be sinning against if we speak ungrammatically? Chomsky would answer that people speaking in such a deviant manner are breaking the implicit rules of universal grammar and this explains our judgements that such people are speaking ungrammatically[2]. Quine, who rejects appeal to implicit rules which govern what sentences we accept as grammatical, will need to tell a different story of how such sentences are viewed as deviant.

In the rough and ready world of ordinary discourse, Quine recognises that intentional idioms are indispensable. It is only when we are trying to limn the true and ultimate scheme of reality that such idioms have no place. It is also in the rough and ready world that people’s behaviour (including their verbal behaviour) is viewed as deviant. It is from this rough and ready vantage point that people are judged to be suffering from schizophrenia and aphasia[3]. Quine would have no problem with using this pragmatic idiom in daily discourse while using the more precise discourse of neurology and behavioural science when trying to limn the ultimate nature of reality. So in this sense the schizophrenic and aphasic objection does not in any way affect Quine’s argument.

To the objection that Quine’s flat-footed conception of rule following cannot handle normative notions like correct and incorrect grammar the same reply as above will suffice. First, Quine can point out that in ordinary discourse, when applying the dramatic idiom of intentionality, we can judge that certain statements are unclear, badly structured, pragmatically deviant, etc. However when we are limning the true and ultimate structure of reality, we can do no better than say that people are following whatever extensionally equivalent grammars can be constructed to systematise their utterances. To people such as Chomsky who claim that people follow one true grammar as opposed to any and all extensionally equivalent grammars, Quine would reply that we have in-principle no behavioural evidence which can decide between them. And without such evidence, all we are justified in positing is Quine’s flat-footed conception of rule following.

As we have already seen, Chomsky would greatly object to this characterisation because he does not think that syntactic rules can be studied in terms of performance. The important thing to note is that at this point we do not know whether a theory of syntax, semantics, etc. is possible in terms of performance. We do not know whether people’s grammatical intuitions match up with their actual linguistic behaviour, though from the empirical data to date they clearly don’t. A further difficulty stems from the fact that Chomsky does not typically offer statistical analysis of people’s grammatical intuitions, and when studies have been done[4] the result has revealed much greater variety in people’s intuitions than Chomsky would admit. So Chomsky’s confident assertions aside, we do not know whether it is possible to construct a theory of performance yet, and discovering whether this is possible or not will require empirical research, not rhetoric. This bears on the debate between Chomsky and Quine on the nature of rules. If a science of behaviour is tractable then Quine’s conception of the behaviour of various subjects fitting syntactic rules will be the most accurate way of conceiving the facts. However, if it is shown that Chomsky is correct that a science of human behaviour is impossible, then Chomsky’s conception of a third intermediate type of rule will be the correct picture.

[1] In reality, Quine does not have this collection of facts to account for because as, Chomsky’s poverty of stimulus argument is dubious, and performance data does not match up closely with competence data.

[2] Obviously, there is more to schizophrenic word-salad than ungrammatical sentences; however ,I am just focusing on ungrammaticalness because it is directly relevant to the debate between Chomsky and Quine on this particular point.

[3] The rough and ready vantage point includes the judgements of psychiatric workers, who help themselves to intentional idioms and physical idioms as is useful for their purposes.

[4] For some research on the statistical analysis of people’s grammitical intuitions see Lappin and Clark Linguistic Nativism and Poverty of Stimulus.