Stemmer Weighs in on Grammar

The story so far. Skinner advanced a theory (or a research program, or something) about how human language is simply "verbal behavior," a structure of stimulation and response subject to operant conditioning, like the movements of a rodent through a maze.

Chomsky objected, on grounds of method, vocabulary, and grammar.

In 1970, Kenneth MacCorquodale replied on Skinner's behalf, and seems to have done a better job on the issues of method and vocabulary than on the issue of grammar, which many had seen from the start as the heart of Chomsky's critique (and is certainly the heart of his own professional work as a linguist). Grammars consist of very complex structures, and in his view our ability as humans to acquire this structure can not be explained as the result of a process of generalization. For example: Children who already know the sentence "The man who is tall is in the room" easily learn that the way to turn that into a question is this: "Is the man who is tall in the room?"

If behavioralism is right, Chomsky thinks, children would at least sometimes make the mistake of asking, "Is the man who tall is in the room?" It is often the case that when turning a statement into a question, we transfer the occurrence of the word "is" to the start of the corresponding question. "The tall man is here" becomes "Is the tall man here" and so forth. Following that practice mechanically, simply generalizing it, could produce mistakes like the italicized sentence above. "Children make many mistakes in language learning," Chomsky says, but never that one.

So our story comes to 1990, when Nathan Stemmer wrote his own reply to Chomsky.

Stemmer replies that behavioralists don't have to expect children to treat the word "is" in isolation and generalize its moves in such a way. Children, rather, likely learn the active sentence pattern first "X is Y," and latter generalize it over time to learn substructures, so that "X" might include the word "is" internally.

Consider, then, "The man who is tall is in the room." The phrase "The man who is tall" is X in the old "X is Y" pattern. The second appearance of the word "is" then becomes the copula -- the one that connects X with the Y of "in the room." Generalization, then, Stemmer says, "does not simply transform certain word sequences into other word sequences but rather certain structures into other structures," and does so without importing a Chomskian machine-shop into the brain. Is [the man who is tall] in the room. Voila.

So (you might ask after all of this) what is my view? As a curious amateur, on-looker, and stand-up philosopher I have to say I think both sides are wrong. I do think that the Skinnerians are right to distrust the presumption of human uniqueness that runs through Chomsky's work. And I side with them (and the whole empirical tradition) against innate ideas, even grammatical ones. Stemmer's contention that we don't need innate ideas to understand how statements can be transformed into questions within a natural language seems to me sound. On the other hand, the Skinnerians' materialistic, reductive philosophy is repugnant on its own account.

William James used to refer to himself as squeezed between the "upper and the lower dogmatisms," between the positivists and the Hegelians of his day. It is always thus. Closer to our time, Skinner has revived the role of W.K. Clifford and Chomsky is writing like one of those "priggish Hegelians" on the top side of the two dogmatisms.

Next weekend I hope to return to these issues from another angle.