This paper is designed to report on a theory of language acquisition embodied in a computational model, which is rather unique among theories of language acquisition in the degree to which it assumes an interplay between surface syntax and considerations of meaning. Since the theory is explicit as to where this interplay takes place, it will be possible to discuss the necessity of both semantic and surface structure in concrete terms. The theory and its computational model are both called PST -the Problem Solving Theory. The creation and improvement of PST has been taking place, with some interruptions, since 1967. The latest version of the computational model is called Instance 3.