CHOMSKY AND QUINE:THE ANALYTIC AND THE SYNTHETIC: PART 1
SECTION 1: THE OBJECT OF ATTACK
Despite the fact that Chomsky and Quine are both methodological naturalists, they nonetheless disagree on a number of issues. In this blog I will consider one such disagreement: their differing views on the status of the analytic/synthetic distinction. I will show that this dispute between them is an entirely naturalistic dispute. Traditionally, philosophers who used analyticity in their theorising did so in the service of a non-naturalistic philosophy. Such thinkers used analyticity as an explanation of non-naturalistic notions such as a-priori knowledge and our grasp of metaphysical necessity. One of my main reasons for discussing the dispute between Chomsky and Quine on analyticity is to illustrate a disagreement between both thinkers which, despite superficial appearances is entirely naturalistic. I will evaluate the evidence put forth by both thinkers and show that Chomsky has not offered sufficient evidence to justify his claim that analytic connections exist in natural language.
Quine’s attack on the analytic/synthetic distinction in his Two Dogmas of Empiricism is probably one of the most famous papers in the history of analytic philosophy. However, while the paper has been extremely influential, it has long been noted that the object of Quine’s attack was unclear even to Quine himself. In Two Dogmas Quine seemed to be attacking three different sources (1) Truth in virtue of meaning; (2) Necessary Truths; and (3) A Priori Knowledge. This fact was understandable because at the time Quine wrote Two Dogmas most philosophers believed that an analytic truth was a truth which was both an a priori truth and a necessary truth. Logical Positivists had argued that the only sensible way to account for supposedly a priori truths or necessary truths without postulating some mysterious faculty of intuition was to do so in terms of analyticity. So it is understandable that Quine would have believed that by attacking analyticity he was also attacking the a-priori and necessity.
Post Saul Kripke few philosophers would conflate analyticity, necessity, and the a-priori. It is worth bearing in mind the above distinctions when evaluating Quine’s arguments because doing so will help us to key in on what his arguments are directed towards, and on whether his arguments are effective. The a-priori/a-posteriori distinction is an epistemological distinction. The necessary/contingent distinction is a metaphysical distinction. The analytic/synthetic distinction is between purely semantic truths and truths which rely on semantic facts and extra-linguistic information. In the 1950’s, Quine, like most philosophers, believed that these distinctions lined up. He furthermore believed that, since analyticity was the only way of explicating the a-priori and necessity, then by attacking analyticity he was attacking the other distinctions. Quine was not merely concerned with attacking analyticity in so far as it was used as an explanation of the a-priori and necessity, he was also concerned with attacking the very notion of a purely semantic truth. He believed that truth by virtue of the meaning of the terms was a senseless notion. I will now discuss his actual arguments against analyticity in Two Dogma’s focusing firstly on his arguments against the semantical notion of analyticity.
SECTION 2: TWO DOGMAS OF EMPIRICISM: A SUMMARY
When Logical Positivists spoke of analytic truths they included truths which we would group as necessary truths, a priori truths, and purely semantic truths. Under the banner of analytic truths one would find philosophically important notions such as truths of logic, truths of geometry and arithmetic, as well as trivial semantic truths. So, for example, one would find complex mathematical statements grouped under the same category as a trivial statement such as ‘All bachelors are unmarried men’. Quine who was working within this logical positivist tradition grouped trivial analytic statements together with more philosophically important analytic truths.
In the first four sections of ‘‘Two Dogmas’’ Quine was critiquing the very notion of an analytic truth. Quine argued that we could make sense of what he called Logical Analyticity, an example of which is (1) ‘All bachelors are bachelors’. However he stated that he could make no sense of analytic truths which results from substituting synonyms for synonyms, for example (2) ‘All bachelors are unmarried men’. We will call analytic truths derived by substituting synonyms; Meaning Analyticity. Quine argued against Meaning Analyticity because he could find no cogent way of explicating what it amounted to. Any attempt to define meaning analyticity relied on using notions which themselves could only be made sense of if we helped ourselves to the prior notion of meaning analyticity. However, since analyticity was the very notion which we were trying to define, if the other terms which we use to define it themselves rely on the notion of analyticity to be made sense of, then our explanation will be circular. Quine’s objection to the notion of analyticity is usually called his ‘circle of terms’ argument. We cannot define analyticity without terms such as ‘‘synonymy’’, ‘‘necessarily’’, ‘‘meaning’’ etc. and these terms can only themselves be defined in terms of the ill-understood notion of analyticity. So Quine argued that, since we cannot define analyticity in a non-circular manner and the notion makes no sense behaviourally, then we have no justification to argue that analytic sentences, as traditionally conceived, exist.
The primary aim of the first four sections was to criticise the notion of meaning analyticity. In these sections Quine was concerned with attacking the notion of purely semantic truth. In the final two sections Quine is attacking what he takes to be the dogma of reductionism. He begins by attacking the verification theory of meaning. He notes that, according to the verification theory of meaning, the meaning of a statement is the method of confirming or infirming it (1951,37). He then notes that an analytic statement is that limiting case which is confirmed no matter what (ibid., 37). Quine claims that it is possible to argue for a conception of analyticity using the verification theory of meaning. One could say that two statements are analytic if they have the same conditions of verification.
Quine rejects the definition analyticity in terms of conditions of verification because of deficiencies which he sees in the verification theory of meaning. He famously argued that the unit of empirical significance is the whole of science (ibid., 42). He pointed out that all empiricist attempts at reduction had failed. Furthermore, he noted that such failures were failures in principle; all such attempts relied on undefined terms to construct their reduction. It was because of this failure that Quine argued for his famous web of belief. When cashing out the nature of the web of belief Quine claimed that: ‘our statements about the external world face the tribunal of experience not individually but as a corporate body’ (ibid., 46).
An important consequence of this holism is its effect on the nature of falsification of our scientific theories. Since our total theory of the world is an interconnected web of belief, and this web faces the tribunal of experience as a corporate body then a falsification of a prediction made by the theory will reverberate throughout the theory. However, Quine notes that if we are faced with a supposed falsification of a prediction, how we modify the theory which made the false prediction is to be dictated by pragmatic considerations. So, for example, since mathematical and logical principles are so deeply connected to every area of our scientific theories they will be the last thing that we will modify if our theory makes a false prediction. Nonetheless there is no reason in principle that we cannot revise mathematical or logical laws if it seems the best way to accommodate recalcitrant experience.
Laws of mathematics have traditionally been viewed as being special in that they could not be explained empirically. Kant claimed that mathematical laws were a priori synthetic truths, because, while they were a priori true, they were not analytic truths. An analytic truth, was for Kant, a statement where there was nothing in the predicate concept that was not already contained in the subject concept. However, Kant noted that in the case of mathematical truths we could not explicate them in terms of analyticity. So, for example, if we consider ‘2+2=?’, according to Kant, no analysis of preceding concepts will give us the answer ‘4’. Kant claimed that if we want to discover an answer in maths we have to construct it in intuition, merely analysing concepts will get us nowhere.
Logical Positivists argued, contrary to Kant, that mathematical truths could be reduced to logical truths. They recognised two types of truth, analytic truth and synthetic truth. Mathematical truths, like truths of logic, were viewed by them as analytic truths.
Quine’s web of belief argument, with its emphasis on the fact that mathematical and logical truths are revisable in principle but in practice are rarely ever revised, is pertinent to the considerations of Kant and the Logical Positivists. To explain the special status of mathematics Kant had to postulate a mysterious kind of knowledge (a priori synthetic knowledge). Logical Positivists tried to explain mathematics in terms of logic which was in turn explained by linguistic convention. Quine demonstrated that the doctrine of truth by convention was dubious because in order to consistently apply any logic we would need to presuppose a logic prior to such an application (1934). While Kant’s postulation of a priori synthetic knowledge was never made sense of in a non-mysterious way, Quine’s web of belief picture explained why mathematical and logical truths seemed to have a special status in our overall theory of the world. Furthermore, it managed to avoid the difficulties which plagued the pictures of the Positivists and Kant.
Quine’s web of belief argument shows that we do not need to appeal to analytic truths in the sense of truth in virtue of meaning to explain the apparent specialness of mathematics and logic. It furthermore shows that we do not need to appeal to a-priori synthetic knowledge to explain what is apparently special regarding mathematics. Quine’s picture of mathematics and logic being deeply embedded into our total theory of the world explains their apparent specialness adequately. Furthermore, by admitting that basic laws of mathematics and logic can be revised, he is making sense of the history of science in a way that people like Kant do not.
However, as Putnam pointed out in his paper ‘‘The Analytic and the Synthetic’’, while Quine’s argument for the web of belief is persuasive as an explanation of non-trivial analytic sentences, it is far from persuasive when it comes to trivial analytic sentences. While it is sensible to explain away truths of mathematics conceived as analytic truths by appeal to them being deeply embedded in our web of belief, it is not sensible to claim that a trivial statement such as ‘All bachelors are unmarried men’ is deeply embedded into our web of belief. I will explore this weakness with Quine’s explanation of analyticity later in Section 4. Firstly I will explicate Chomsky’s difficulty with Quine’s claim that there is no sensible distinction to be drawn between analytic and synthetic sentences.
SECTION 3: CHOMSKY’S DEFENCE OF THE DISTINCTION
It appears, then, that one of the central conclusions of modern philosophy is rather dubious: namely, the contention- often held to have been established by the work of Quine and others- that one can make no principled distinction between questions of fact and questions of meaning, that it is a matter of more or less deeply held belief. (2000b, 63)
As a linguist, Chomsky was primarily with whether an analytic/synthetic distinction was, as a matter of empirical fact, a distinction that ordinary speakers of natural language recognised. He argued contra Quine that analytic connections do exist in natural language, and that it is the job of the cognitive scientist to study such connections. As in the case of syntax, we begin with people’s intuitions; in the case of analytic connections, we need to test whether people can reliably distinguish sentences which are analytic from sentences which are synthetic. So the question arises: would an ordinary speaker of English be able to distinguish between an analytic sentence like ‘All Bachelors are Unmarried men’ from a synthetic sentence like ‘All Bachelors like Seinfeld’? For Chomsky’s claim that analytic connections are a fact of natural language to be shown to be correct, experiments need to be conducted. However, experimental research on the analytic/synthetic distinction has not exactly been overwhelming. There have been some experiments on the distinction, for example: Apostel, L., W. Mays, A. Morf, and J. Piaget. Les liaisons analytiques et synthetiques dans les comportements du sujet. Arne Naess. Interpretation and Preciseness, as well as Katz and Fodor’s paper, ‘‘The Structure of a Semantic Theory’’. These studies do imply that people have a gradualism of statements that they find intuitively synonymous. And this view seems to support Chomsky’s claim that people do have intuitive conceptions of what sentences in natural language are analytic and what ones are synthetic.
Chomsky argues that since the experiments show that people intuitively distinguish between analytic and synthetic sentences, then we need to explain this fact. His explanation centres on poverty of stimulus considerations. He argues from the speed that children acquire words (about 12 words a day at peak periods), and the incredible complexity of the words acquired, to the claim that the children must be merely labelling concepts they are born with. From his argument for innate concepts, he draws the following conclusion:
This would appear to indicate that the concepts are already available, with much or all of their intricacy and structure predetermined, and that the child’s task is to assign labels to concepts, as might be done with limited evidence given sufficiently rich innate structure. And that these conceptual structures appear to yield semantic connections of a kind that will, in particular, induce an analytic-synthetic distinction, as a matter of empirical fact. (2000b, 62)
On the face of it, this argument simply does not work. Chomsky’s poverty of stimulus argument for concepts relies on the assumption that when children acquire words this ability indicates a grasp of a complex concept which the word means. So, for example, it implies that when a child acquires a word such as ‘Mama’, the child understands the same complex concept which adults do when they explicate what the concept MAMA means. Chomsky has never provided any evidence to support this wildly implausible claim. If we take away the unproven claim that the words children are learning at peak periods of language acquisition express full-blown adult concepts, then the acquiring of 12 words a day is less impressive than Chomsky claims. If a child using a word such as ‘mama’ had only a simple understanding of what it meant, then the fact that the child was learning such words so fast would not seem so miraculous. And there would therefore be no need to postulate innate concepts to explain the speed of word acquisition.
While there is some evidence that children from as young as four months of age demonstrate some conceptual understanding of concepts such as OBJECTS, CAUSATION and AGENT, this evidence does not support Chomsky’s more radical claims about concepts. I will discuss the evidence for children’s conceptual abilities in chapter three; here I will merely note that it is at best suggestive. The evidence does show that children have some concepts prior to learning language. However it does not conclusively demonstrate that these concepts are innate, nor that children possess most of their concepts prior to learning language. Chomsky is making much more radical claims than that some abstract concepts such as object, or causality are innate. In a personal communication with me he made the following claims:
Also, there is good evidence that innateness of concepts goes far beyond the philosophically interesting examples that you mention. It’s hard to imagine how else people could acquire the meanings of the simplest words in the language — “river,” “tree,” “person,”…. They have rich properties for which there is no empirical evidence for the child. (Chomsky: Personal Communication)
Furthermore in his New Horizons in the Study of Language and Mind, Chomsky made even more radical claims about the innate conceptual abilities of Children. Chomsky considered an argument from Putnam (1988) that evolution could not have equipped us with an innate set of concepts including ones like Carburettor and Bureaucrat. Putnam argued that evolution could not have equipped us with these concepts because in order to do so evolution would need to be able to anticipate all future contingencies. Chomsky replied to this by noting that Putnam’s argument is incorrect because we do not need to assume that evolution anticipated all possible contingencies, just the particular ones under question.
Chomsky claimed that a similar argument to Putnam’s was used in immunology and that recent work by Niels Kaj Jerne challenges this idea:
Notice that a very similar argument had long been accepted in immunology: namely, the number of antigens is so immense, including every artificially synthesized substances that had never existed in the world, that it was considered absurd to suppose that evolution had provided an ‘innate stock of antibodies’, rather formation of antibodies must be a kind of ‘learning process’. But this assumption might well be false. Niels Kaj Jerne won the Nobel Prize for his work challenging this idea… (ibid, 65)
This analogy proves nothing, it is not in doubt that it is possible that people are born with an innate stock of concepts like CARBURETTOR and TREE. What is in doubt is whether there is any evidence that supports the claim that such concepts are innate. Chomsky has provided none whatsoever here. Chomsky vaguely gestures in the direction of poverty of stimulus considerations to support his conjecture about innate concepts:
Furthermore, there is good reason to suppose that the argument is at least in substantial measure correct even for such words as Carburettor and Bureaucrat, which, in fact, pose the familiar problem of poverty of stimulus if we attend carefully to the enormous gap between what we know and the evidence on the basis of which it is known. (2000b, 65)
Here Chomsky is using a poverty of stimulus argument similar to the argument he uses in the area of syntax. I will discuss the poverty of stimulus argument in the area of syntax in chapters two and four. The poverty of stimulus argument in syntax has the merit of being precise and testable. Chomsky has provided no such explicit model of a poverty of stimulus argument for innate concepts. So we have no reason to believe Chomsky’s radical claims about innate concepts until more evidence is provided.
It is of course possible that Chomsky is correct that children are born with innate concepts such as: CARBURETTOR, TREE, BUREAUCRAT, RIVER, etc.; however an incredible amount of evidence is needed to support such an incredible claim. Chomsky vaguely points to poverty of stimulus considerations. However, he provides no details of where the supposed gap between the knowledge children display of the concept CARBURETTOR and the data the child receives when learning the concept lies. So until he spells out his poverty of stimulus argument in more detail his claims on these matters should be viewed with extreme scepticism.
More importantly, even if Chomsky did provide evidence that such concepts are innate, it would not follow that they will automatically yield analytic connections. Jerry Fodor has long agreed with Chomsky that most of our concepts are innate; however he does not think that analytic connections exist in natural language. It is worth briefly considering Fodor’s views on innateness and analyticity because doing so will help us better understand Chomsky’s views.
Fodor agrees with Chomsky that most of our concepts are innate. However he does not use the same arguments as Chomsky to reach this conclusion. While Chomsky’s poverty of stimulus argument points to a gap between what we know and the inability of the data of experience to account for our knowledge, Fodor argues that because of certain facts about the structure of concepts, all known theories about how concepts are learned are incorrect in principle. He claims that there are two different ways of accounting for our knowledge of concepts. The first one is to say that most of our concepts are definitions which are defined in terms of primitive concepts. The primitive concepts are either defined as sensory primitives such as RED, SQUARE etc., or as abstract concepts such as CAUSATION, AGENCY, and EVENT etc. What the primitive concepts are will depend on the nature of the theory being expounded. So for example, the empiricist philosopher David Hume argues that the primitive concepts are sensory experiences. While the linguist Stephen Pinker argues that the primitive concepts are abstract concepts such as CAUSATION, AGENCY, and OBJECT which we are born with knowledge of. Fodor thinks that Pinker, Hume and all definitional theorists are wrong because concepts cannot be explicated in terms of definitions.
The other theory of concept acquisition thinks that our concepts are basically prototypes which are learned statistically. Fodor argues against this view because he thinks that the prototype theory cannot account for the fact of compositional concepts. I will not here consider Fodor’s arguments against prototype theory because they are not relevant to the concerns of this thesis. Furthermore I will not consider the fact that Fodor has lately retracted his views on radical nativism because again this is not relevant to the overall argument of the thesis. I am only concerned with explicating why Fodor does not follow Chomsky in arguing that analytic connections follow from the fact of innate concepts. Fodor’s reason for not believing that analyticity follows from innate concepts is that he does not think that innate concepts have internal structure. His argument that concepts do not have internal structure is derived from his belief that concepts are not definitions. Fodor essentially uses three different arguments against the claim that concepts are definitions:
(1) Lack of definitions: In general we have not found any definitions for concepts. There are millions of concepts and very few definitions (maybe fifty or so). (1998, 46)
(2)Developmental argument: All concepts cannot be definitions. Some of the concepts must be primitive. The concepts which are definitions must be defined partly in terms of the primitives. This being so we must assume that there is a developmental stage when children only know primitive concepts. However we have no evidence to support this claim. Therefore, concepts cannot be definitions. (ibid, 47)
(3) Production argument: If concepts were definitions, then sentences which contained complex concepts would be harder to produce than those which contained simpler concepts. The reason is that a complex concept would take longer to call to mind the definition of when speaking the sentence. However the experimental evidence indicates that the production of sentences involving complex concepts takes no longer than the production of sentences which contain less complex concepts. (ibid, 49).
Fodor believes that his arguments against definitions are decisive. He claims that without definitions, concepts have no internal structure, and furthermore if concepts have no internal structure, then there are no analytic sentences. Kant famously defined an analytic sentence as a sentence in which the predicate is entirely contained within the subject. However obviously if a concept has no internal structure, then nothing can be contained within it. For this reason, Fodor thinks that there are no analytic sentences. He notes that one of the main arguments for definitions is that they explained people’s felt intuitions of analyticity. So this gives us two options: (1) accept that people’s intuitions of analyticity are a reason to believe in conceptual connectedness; (2) explain away the intuition. Since Fodor has offered reasons to deny that there are complex concepts, he opts for number (2), to explain away the intuitions of analyticity.
EXPLAINING AWAY ANALYTICITY INTUITIONS:
As we saw above Quine explained away our intuitions of analyticity in terms of his web of belief picture. This worked for supposed non-trivial analytic sentences like F=MA. However it was less plausible as an explanation for apparently more trivial analytic sentences like ‘‘All bachelors are Unmarried men’’. We cannot plausibly explain away cases like ‘‘All Bachelors are Unmarried men’’ by saying that they are deeply embedded into our total web of belief.
As we have already seen, Putnam claimed that Quine’s argument worked as a criticism of the a priori but not as a criticism of apparently trivial analyticity. He claimed that a sentence such as ‘‘All Bachelors are Unmarried men’’ is an analytic sentence, and its analyticity derives from the fact that it is a one criterion concept. The concept BACHELOR is not according to Putnam connected to any other concepts in our web of belief other than to the concepts UNMARRIED and MEN. Putnam argues that sentences which people intuitively find analytic but which are not deeply connected to our web of belief, can be explained by his notion of one-criterion concepts. Fodor disagrees with this explanation because he feels that it is a circular argument. In order to define which concepts are learned by one criterion, you need to know that it is analytic already. So your explanation of analyticity presupposes the existence of the very thing we are doubting. So we are back to Quine’s circle of terms again.
Fodor offers a different account of one-criterion concepts which he thinks is less problematic than Putnam’s is. Fodor agrees with Putnam that one-criterion concepts exist. He furthermore thinks that he can use these one-criterion concepts to explain people’s intuitions of analyticity. However, Fodor argues that since he is not using one-criterion concepts to explicate analyticity, then his story is not circular in the way that Putnam’s is. Here is Fodor’s story in a nutshell:
Suppose you think the only epistemic route from the concept C to the property it expresses depends on drawing inferences that involve the Concept C*.Then you will find it intuitively plausible that the relation between C and C* is conceptual; specifically, that you can’t have C unless you have C*. And the more you think that it is counterfactual supporting that the only epistemic route from C to the property it expresses depends on drawing inferences that involve the concept C*, the stronger your intuition that C and C* are conceptually connected will be. (Ibid, 83)
Fodor argues that people’s intuitions of analyticity are really just epistemic intuitions which are confused for semantic connections. So for example I can only know that someone is a bachelor by knowing that he is an unmarried man. According to Fodor, I know that all bachelors are unmarried men because of epistemic access to the facts in the world, not because of the internal structure of concepts. He holds that people’s intuitions of analyticity are illusions which result from conflating epistemic properties with semantic properties.
So we can see from Fodor’s argument that analyticity does not follow from innateness. If the innate concepts are unstructured, then they will not automatically yield analytic connections. Chomsky assumes that the children’s innate concepts are structured and as a result will automatically yield analytic connections. Furthermore unlike Fodor and Quine, he takes it for granted that people’s intuitions of analytic connections are good evidence that such connections exist.
In a personal communication I asked Chomsky why he believed that analyticity followed from innateness:
You’re quite right that it isn’t obvious, but I think it is correct. It’s an empirical issue, in principle, but a hard one to investigate — and like other empirical questions, certainty is unattainable, just a high degree of plausibility. Personally, I think that serious inquiry would reveal that “tomorrow is two days after yesterday” or “my uncle is male” or “nightmares are dreams” and much else would withstand the harshest tests, and I don’t think that Quine or Fodor have suggested any reason to doubt these expectations. (Chomsky: Personal Communication)
Chomsky’s reply is instructive because he simply claims that certain constructions are obviously analytic and that any future tests would confirm this belief. He does not answer the question about why analyticity follows from innateness. He merely claims that it is in principle an empirical question. Presumably his reasoning is that people do recognise a distinction between analytic and synthetic sentences, and he believes that people are born with a massive amount of innate concepts. From this he believes it follows that the structure of these innate concepts are our best explanation for our intuitions of a distinction between analytic and synthetic sentences. However this vague argument relies on an unproven claim about innate concepts, and a further unproven claim about the structure of these purported innate concepts. So Chomsky thus far provides no real explanation for people’s felt intuitions of a distinction between analytic and synthetic sentences.
Chomsky appeals to the fact that certain sentences are obviously analytic, presumably he is here claiming that such sentences are intuitively felt to be analytic. However obviously the fact that people have intuitions of analyticity does not tell us whether Fodor’s story is correct or if Chomsky’s is.
Since Chomsky has offered no compelling evidence to support his claim that humans are born with innate concepts, a debate between him and Fodor on whether innate concepts have internal structure is a nonstarter. Chomsky has offered no evidence for innate concepts, so he cannot support his views on analyticity by appealing to innate internally structured concepts. Fodor has offered some arguments for innate concepts, however they offer little support to Chomsky’s position.
It is of course open to Chomsky to use Fodor’s arguments for innate concepts; however a vital part of Fodor’s argument, is that concepts are not definitions, so they have no internal structure. So if Chomsky did use Fodor’s arguments for innate concepts, he would be left in a situation where he could claim that concepts are innate but have no internal structure, so do not yield analytic connections. Either approach leads to the conclusion that analytic sentences do not exist. So Chomsky has offered us no real reason to believe that we can derive analyticity from innate concepts.
We have shown that analyticity does not follow from innate concepts. It is worth noting that the converse is also true. If one is denying that innate concepts exist one is not thereby denying that analytic connections exist in natural language. Empiricists such as Ayer, Carnap and Hume who are not committed to the view that concepts are innate, nonetheless argue that analytic connections exist in natural language. Ultimately, then, Chomsky’s arguments for innate concepts are irrelevant to facts about analyticity unless he can provide evidence that such innate concepts are structured in such a way as to yield analytic connections. Thus far Chomsky has provided no evidence about the structure of these supposedly innate concepts, so we therefore have no reason to believe that innate concepts lead to analytic connections in natural language.
Independent of his unproven claims about innate concepts yielding analytic connections, Chomsky’s arguments for analytic connections in natural language focus on three primary facts. One is that ordinary language speakers can readily tell whether a sentence of natural language is true by virtue of the meaning of the words involved and extra linguistic fact, or whether it is made true entirely by facts of meaning. His second reason is that he has never seen a convincing counter example to clear cases of analytic sentences, such as ‘Tomorrow is two days after Yesterday’. And his third reason is that he thinks that particular sentences of a relational structure such as the sentence ‘‘If John killed Bill then Bill is Dead’’ are more difficult to find counter examples to than the simple examples focused on in the philosophical tradition such as ‘‘All Bachelors are Unmarried men’’.
Chomsky’s first claim that people intuitively recognise a distinction between statements which are synonymous and those which are not has been tested experimentally through the using of questionnaires. The small literature does seem to indicate that people have an intuitive sense of a distinction which loosely corresponds to what is traditionally called an analytic/synthetic distinction. However such studies involve showing people different sentences which are categorised into two different groups. The fact that people categorise in a way consistent with an analytic/synthetic distinction proves nothing. Such categorisation may reflect an epistemic categorisation rather than a semantic one. The various different studies do not distinguish between the Quine/Fodor story and the Chomsky story. The Quinean model would predict that people would have an intuitive sense of a distinction between truths like ‘‘2+2=4’’ and ‘‘Most dogs are less than six feet tall’’. The difference between Quine and Chomsky on the question of Analyticity is whether; a person’s intuitive sense of Analyticity can be explained best in terms of an innate system of concepts, or in terms of deeply embedded beliefs in a person’s holistic theory of the world. The weak point of Quine’s views on analyticity is that his web of belief story works well for epistemologically significant concepts such as ‘‘2+2=4’’ but not for ‘‘All Bachelors are Unmarried men’’ which cannot be reasonably claimed to be deeply embedded into our total theory of the world. However Putnam’s one criterion concepts as modified by Fodor works perfectly to capture trivial cases of analyticity. Fodor has shown that such cases can be viewed as epistemic knowledge confused with conceptual knowledge. The fact that people have intuitions of analytic connections in natural language is not that significant, such intuitions could be epistemic knowledge confused with conceptual truths. They can be explained away as intuitions of truths deeply embedded in to our web of belief, or one-criterion concepts.
Chomsky’s second reason is not very convincing, as philosophers have shown over the last fifty years that some statements which seem immune to revision are in principle revisable; see for example Putnam on the sentence ‘‘All Cats are Animals’’. Now Chomsky could reply to this that while some sentences which are analytic have been shown to be in principle revisable, the paradigm statements have not been shown to be revisable. If he were to argue so, in order to make his claim a testable one, he would need a list of paradigm cases of analytic sentences. He would then need to show how such sentences could withstand the probing of the most strenuous of tests. Chomsky has never done anything like this. His remarks on analytic sentences, remain just that, remarks; and remarks which are unproven.
Chomsky has also claimed that philosophers erroneously focus on analytic sentences with to simple a structure; he claims that sentences of a relational structure provide much more evidence for analytic sentences. To test Chomsky’s claims we will need to consider some examples of sentences of a relational structure and test whether they support his claims over Quine’s. He has used two examples over and over again, they are: (1) If John killed Bill, then Bill is Dead. (2) If John persuaded Bill to go to college, then John caused Bill to intend to go to college. Here Chomsky discusses these sentences:
Notice again that we appear to have connections of meaning in such cases as these; we have a rather clear distinction between truths of meaning and truths of fact. Thus, if John persuaded Bill to go to college, then Bill at some point decided or intended to go to college and did so without duress; otherwise, John did not persuade Bill to go to College. Similarly if John killed Bill, then Bill is dead (though John may or may not be depending on the facts). These are truths of meaning not of fact. (2000b, 62)
Sentences1 and 2 are examples of sentences with a relational structure. The use of the predicates ‘killed’ and ‘persuade’ in the antecedents will have direct bearing on what words are allowed in the consequent. Take, for example, Sentence 1 above- Let us consider the sentence as an argument schema: If X killed Y then Y is….?. What Chomsky seems to want to say here is that the only way to make the argument schema true is to fill ‘dead’ into the blank or something synonymous with ‘dead’. Of course such an interpretation is patently false as can be seen by the following; we can insert ‘is not going to play football next week’, ‘is not going to run a marathon next week’ etc. and the truth value will remain invariant. So it could be claimed that the relational structure does not in any obvious sense indicate that, from the fact of John killing Bill, we have to infer that Bill is dead. However, upon closer inspection it becomes clear that anything we fill into the blank must be consistent with the fact that Y is dead. So, for example, if one were to offer the following to fill in the blank ‘is going to sing his favourite song tomorrow’, one would be constructing a false argument. And the reason is that singing a song is not consistent with being dead. The two attempts to fill in the blank above which result in true arguments work because obviously not doing something, is largely consistent with being dead. From this we can conclude that any blank we fill into the consequent must at least be consistent with being dead, however obviously it does not have to be synonymous with being dead. So if we consider ‘X killed Y so Y is -’, only something consistent with being dead can be slotted into the consequent and the truth value remain invariant. So any conditional which has an antecedent such as ‘Y being killed, so Y is…’ must have a consequent which is consistent with Y being dead or the conditional will be false.
If one puts ‘is dead’ into the consequent one gets an analytically true statement. Likewise, if one puts ‘is not playing tennis’ one also gets an analytic sentence. This construal of the argument leaves us in a situation where we cannot equate the analyticity with synonymy. Obviously the statement that ‘X killed Y’ is not synonymous with the statement that ‘Y is not playing tennis’, despite the fact that Sentence 1‘X killed Y so Y is not playing tennis’ is an analytic statement. Chomsky could argue that Sentence 1 is analytic because of the internal structure of the complex concept KILLED. He could argue that the complex structure of such concepts which we are born with will automatically tell us what inferences can be derived from it purely based on the meaning of the concept. So our language faculty determines what arguments can be filled into the schema ‘X killed Y so Y is-’ in order to yield analytic connections. Here we have an argument which seems to show how we can go from an innate concept such as KILLED to construct analytic sentences based on the internal structure of the concept. However, such an argument again relies on assumptions about innate concepts which have not been proven correct. So if Chomsky wants to construct such an argument he needs to find evidence to support his belief in innate concepts such as KILLED. Independent of the argument from innate concepts all that is left is the claim is that some analytic statements are difficult to refute, but of course this claim is largely consistent with Quine’s claims about analytic statements being deeply embedded into our overall theory of the world.
So Chomsky’s claim that sentences of a relational nature provide stronger kinds of evidence for analytic connections in natural language than the simpler cases that philosophers typically talk about has been shown to be erroneous. His assumption that this is so relies heavily on his belief in innate concepts. The claim that sentences of a relational structure seem to offer cases of analytic connections can be explained in terms of how embedded notions like KILL and INTENTION etc. are to our overall theory of the world.
The above discussion shows that Chomsky has offered no compelling arguments which demonstrate that Quine’s critique of an analytic/synthetic distinction is incorrect. It is possible in principle that Chomsky’s positive views on analyticity are correct; however, we do not as of yet have any empirical evidence to support Chomsky’s claims on this matter. In the next blog I will try to explicate what type of analyticity Chomsky is committed to if we assume he is correct that analytic sentences exist in natural language as a result of certain constraints which are imposed by the concepts which we are born with. In the next section I will assume that Chomsky is correct on this point, and consider how much analytic connections being encoded in the brain of a subject meets the criterion of analyticity as traditionally conceived.
 John Stuart Mill was an exception to this rule.
 I will not here evaluate whether Fodors arguments for innate concepts work because it has no relevence to the argument of the thesis.
 For a detailed discussion of issues connected to the above discussion see Fodor 1970 ‘Three Reasons for Not Deriving Kill from Cause to die’, and Pietroski’s ‘Small Verbs, Complex Events, Analyticity without Synonomy’. The debate between Pietroski and Fodor would take us too far a field from the concerns of this thesis but IS worth reading in its own right.