For one thing, any associated supporting element can be defined in such a way as to impose the ultimate standard that determines the accuracy of any proposed grammar. However, this assumption is not correct, since relational information is unspecified with respect to the system of base rules exclusive of the lexicon. I suggested that these results would follow from the assumption that this selectionally introduced contextual feature is necessary to impose an interpretation on the requirement that branching is not tolerated within the dominance scope of a complex symbol. Comparing these examples with their parasitic gap counterparts in (96) and (97), we see that the natural general principle that will subsume this case delimits a general convention regarding the forms of the grammar. With this clarification, the appearance of parasitic gaps in domains relatively inaccessible to ordinary extraction does not affect the structure of irrelevant intervening contexts in selectional rules.
see also: WikiPedia -- Chomskybot