Noam Chomsky's interview on language with Ali G. Proof that Chomsky can suffer fools... well, if not gladly, at least gently.
The founder of modern linguistics, Noam Chomsky, was famous for decades for his dismissal of interest in the evolution of language. In recent years he has moderated his position and in a lecture recently made available on line (here, registration required) he outlines his scenario for how language evolved. It is about as different from the account being developed on this blog as a theory can be, making it of keen interest because it forces me to ask whether I have gone hopelessly astray and should change course quite sharply.
For one thing, while this blog sees the function of speech as piloting joint attention, and while many other thinkers about language origins focus on language as a means of communication, Chomsky is different. He says, “language evolved, and is designed, primarily as an instrument of thought” [italics mine; p. 22]. Communication, for Chomsky, is secondary and, therefore, there is to be no searching for social evolutionary pressures—neither in terms of group selection nor altruistic kin/gene selection.
A rewiring of the brain, “presumably the effect of some small mutation” occurred in an individual “not a group.” This individual was endowed with “complex thoughts, [superior] planning, [superior] interpretation, and so on.” Over generations this gene spread and dominated “a small breeding group.” Only then would there be a reason for “externalization” of the language capacity, i.e., only then would speech be possible or useful.
… so the [language] capacity would be linked as a secondary process to the sensorimotor system for externalization and interaction, including communication. [p. 23]
The speech that emerged was not a pidgin or protolanguage, but true language. The reason for Chomsky’s no-protolanguage position is technical, but important enough to be described here. Chomsky long ago abandoned the ideas about “deep structure” that first made him famous. He now sees the “core principle of language” as “unbounded Merge.”
Many visitors to this blog are already familiar with the mail-merge function of word processors. In that system users create lists of items (addresses, etc.) and basic letters. They then “merge” the two parts, inserting the items into the letters forms, producing a pile of unique and apt letters.
The Merge that Chomsky refers to is similarly primitive, but more abstract and capable of using its discrete elements to produce infinite variety. Merge’s central feature is that “it takes objects already constructed, and constructs from them a new object” [p. 20]. In mail merge, these already constructed objects can be lists of addresses and letter forms. In language, Merge those standard objects are words and the rules for categorizing and arranging them.
An important distinction between mail merge and language Merge is that Merge is unbounded. Mail merge can generate
Dear Joe, Thanks. Blair … Dear Pete, Thanks. Blair … Dear Jack, Thanks. Blair …
but although the mail merge output can continue forever, the form is limited. Language, however, can produce
Thanks … Thanks, Joe … Thanks, dear Joe … Thanks for the gift, dear Joe … Thanks for the gift, dear Joe, and for making me so happy …
Chomsky’s point is that without true language you cannot think deeply enough to gain a survival advantage and you cannot have true language without unbounded Merge. Protolanguage might use some limited form of Merge, but that gives a person no survival advantage. “Transition from [protolanguage],” Chomsky says, “to unbounded Merge is no easier" than from silence. Protolanguage, in this view, is just so much hemming and hawing by the theorizer.
My favorite passage in Chomsky’s essay comes just after he has described language’s profoundly asocial origins and speech’s late bloom:
It is not easy to imagine an account of human evolution that does not assume at least this much. And empirical evidence is needed for any additional assumptions about the evolution of language. [p. 23]
That confident, I-dare-you-to-knock-this-chip-off-my-shoulder tone has never left Chomsky. He still sounds like Euclid justifying the axioms of plane geometry. Evidence, I don’t need no evidence for these assumptions, but you do if you want to challenge them.
Particularly characteristic of Chomsky’s dare is the phrase this much, taking in one sweep the whole package of mutation, then Merge, then breeding community, and then speech. But this package all grows from the single assumption that thought is the primary function of language. If that claim falls, the whole package loses its legs.
Pure thought, which Chomsky says came before any “secondary” linking to a “sensory modality,” is not contaminated by the sound or words, the image of objects, or some other kind of perceptual sensation. Yet Merge assembles pre-existing objects. What pre-existing objects does he mean? Chomsky answers that Merge provides
An internal generative system that constructs thoughts of arbitrary richness and complexity [by] exploiting conceptual resources that are already available or may develop with the availability of structural expressions. [p. 22]
This passage sounds similar to the position emerging on this blog, that language began by taking advantage of a very old system of perceptions and the ability to direct attention. However, Chomsky is much too clear a thinker to confuse sensory organs with “conceptual resources.” The words and idiosyncratic rules that children learn as they acquire a particular language are secondary byproducts of the “mapping involved in externalization” [p. 25] and not a part of the “single internal language” [p. 24] common to us all. Chomsky spells this point directly by distinguishing between two “interfaces”:
- The systems of thought that use linguistic expression for reasoning, interpretation, organizing action, and other mental acts.
- The sensorimotor systems that externalize expressions in production and construct them from sensory data in perception. [p. 14]
These two interfaces, thought and speech, are “asymmetric” (i.e., of unequal size and power) and thought evolved first.
The thought that Chomsky refers to is symbol processing. Its individual symbols only acquire meaning when placed into a syntactic string (a sentence). We can see this acquisition of meaning simply enough if I show you an abstract symbol by itself: m. It could mean anything. If I put it into a string of symbols—F=ma—you can get its meaning.
I’m going to pass by the question of where our symbols came from originally (Chomsky makes no effort to account for them) and report why Chomsky thinks his account must be true. Suppose I say
John is too angry to eat.
I can take that to mean John won’t eat anything because he is angry, but it could also be understood to mean I won’t eat John because he is too angry. This last reading may seem ridiculous, but it is exactly how we would be likely to interpret the sentence John is too angry to invite. I won’t invite John because he is so angry. As Chomsky says in one of his charmingly gotcha sentences, “The surface form in themselves tell us little about the interpretation.” [p. 16]
The syntactical explanation for the different interpretations is that the verb to eat requires an object, e.g., eat an apple. If no object is presented, we assume a general one (won’t eat anything). The verb to invite requires an indirect object, e.g., invite to a party. If that is missing, we assume a general one (won’t invite to anything). The meaning of the sentence came from “the generative procedures that yield the expressions, but cannot be detected in the physical signal.” He goes on:
For that reason it seemed then [fifty years ago]—and still seems—that the language acquired must have the basic properties of an internalized explanatory theory. [p. 17]
That’s why language must be primarily an internal process and that any external features must be secondary: the surface structure of a sentence does not include enough information to make its meaning clear. To be understood we must have access to the generative mechanisms producing the sentence. A speaker generates and knows what it means because of the generative process. The listener reverse engineers the sentences, understanding them by discovering the rules that generated them.
In the late 1950s Chomsky argument carried the day because his opponent gave no place to any internal processes, either perceptual or conceptual. Behaviorism described only reflexive responses to unambiguous stimuli. The replacement school, cognitive psychology, gives us an internal symbol processor, but still has no room for sensation-based knowledge (perception). If you take as your axioms that animals are computers and language becomes meaningful by organizing symbols, it is hard to escape Chomsky’s logic. The only reason I can resist is that I believe animals think perceptually rather than conceptually, and that meaning comes from piloting attention rather than following syntactical rules. What do you believe?