10
alone, this would seem to be a rule of English. So one would expect the innately uninformed
child to infer that the third sentence can be made into a question using the same method. But this
is not how real flesh-and-blood children operate. Unlike the imaginary white-paper child, real
children already know that movement is sensitive to parts of speech. They do not make the sort
of mistake just described (Stromswold, 1999).
Discerning parts of speech requires an extra effort after determining which sounds in the
sentence are words. One might expect the child to arrive at the simpler hypothesis first, namely
that one forms a question just by moving items according to their original positions. Or, if one
ignores the problem of how the child recognizes parts of speech, one might expect the child to
follow the rule move the first verb to the beginning of the sentence. But even this rule would
produce deviant utterances, e.g. in transforming “The man who took the cash left town” into a
question. This means that children, in acquiring competence in moving sentential elements, are
not inferring to the simplest explanation of the data but are gravitating toward a remarkably
complex explanation. This strongly suggests not only that there is innate knowledge but that
there is innate knowledge specialized for syntax. Movement, in all natural languages in which
movement occurs, is sensitive to parts of speech. In no natural language are the rules for
movement linear, i.e. they are never sensitive merely to the position of elements in the original
sequence of sounds.
In discussing the incomplete and misleading nature of much of the child’s linguistic
experience, Chomsky does not quantify how much is “degenerate.” This may raise the question
of whether he has shown that enough of it is degenerate to make a good case for nativism. But,
for Chomsky, even if only a small percentage of the data are degenerate, this shows that the child
cannot be following some general-purpose hypothesis-forming procedure. One can see the point
by considering someone who does use a general-purpose hypothesis-forming procedure, namely