By combining adjunctions and certain deformations, the appearance of parasitic gaps in domains relatively inaccessible to ordinary extraction is not quite equivalent to the requirement that branching is not tolerated within the dominance scope of a complex symbol. To provide a constituent structure for T(Z,K), this analysis of a formative as a pair of sets of features can be defined in such a way as to impose a stipulation to place the constructions into these various categories. If the position of the trace in (99c) were only relatively inaccessible to movement, a case of semigrammaticalness of a different sort delimits the strong generative capacity of the theory. Suppose, for instance, that a subset of English sentences interesting on quite independent grounds cannot be arbitrary in an important distinction in language use. A consequence of the approach just outlined is that the fundamental error of regarding functional notions as categorial is unspecified with respect to irrelevant intervening contexts in selectional rules.
see also: WikiPedia -- Chomskybot