For one thing, a case of semigrammaticalness of a different sort does not affect the structure of the ultimate standard that determines the accuracy of any proposed grammar. Nevertheless, the theory of syntactic features developed earlier is to be regarded as an abstract underlying order. Clearly, the fundamental error of regarding functional notions as categorial may remedy and, at the same time, eliminate an important distinction in language use. Notice, incidentally, that relational information cannot be arbitrary in nondistinctness in the sense of distinctive feature theory. Thus the systematic use of complex symbols is not subject to a descriptive fact.
see also: WikiPedia -- Chomskybot