A history of theories of transformational generative grammar is presented, and four theories are characterized as: Chomsky's initial theory, Chomsky's standard theory, the semantically-based theory, and finally Chomsky's nonstandard theory. The main issue appears to be the question of what determines sentence meaning. Specifically, it is stated that Chomsky's nonstandard theory is in opposition to McCawley's nonstandard theory and to Chomsky's standard theory, the major revision in Chomsky's nonstandard theory being the hypothesis that sentence meaning depends not only on the deep structure but also on the surface structure. Chomsky's nonstandard theory is examined in order to answer two questions: whether it admits a theory whereby it is possible to state whether or not transformations change meaning; and whether it admits that deep structures, but not intermediate structures, determine certain meanings. It is concluded that Chomsky's nonstandard theory is not clear on these two questions. A nonstandard theory is then proposed which differs from Chomsky's in that it attempts to explain how semantic representations of sentences are determined and how Pi and Pn participate in their determination. It is concluded that the greater accuracy of one theory or the other can only be determined after further descriptive study within the framework of both theories. (AM)