In such a phrase-structure grammar," anywhere you have a grammatical sentence with one phrase of a certain type," says Peters, "you can just substitute for it any other phrase of the same type, and the result will be a grammatical sentence."
Known as a "context-free phrase-structure grammar," it consists of a finite list of phrase types and a finite set of rules for piecing together sentences.
Chomsky's theory "overcame some apparent shortcomings of phrase-structure grammar," says Peters, "but it opened the door to too much potential variation in languages." It's like a physicist inventing a theory with seven quarks instead of three, he says.
In the August LINGUISTICS AND PHILOSOPHY, two papers show for the first time that there are human languages that cannot possibly be described by a context-free phrase-structure grammar. Stuart M.
Similarly, Chomsky contended that while context-free phrase-structure grammars may succeed in distinguishing the grammatical sentences of some languages, they do so only with unnecessarily complex rules.
In particular, several researchers began to look again at phrase-structure grammars to see if they could come up with a theory that was mathematically neater and more tightly constrained than transformational theory.
These conclusions encouraged Pullum and others to examine context-free phrase-structure grammars more closely.
In general, says Pullum, context-free phrase-structure grammars can't define languages in which a "string" of arbitrary length in one part of a sentence must have an exact duplicate elsewhere.
But context-free phrase-structure grammars are hard to give up.