To determine the principal differences in syntactic structure between spoken and written American English, a corpus of the spoken (800 sentences) and written (280 sentences) utterances of six graduate students was described in terms of a transformational generative grammar. These utterances were used as a basis for a two-part grammar: (1) a source grammar describing all the utterances from the corpus of written English as well as those sentences from the corpus of spoken English that were non-deviant from the source grammar, and (2) an extension grammar specifying 50 rules designed to account for structures peculiar to spoken English. Extension rules signalled the lack of a formal correspondent in written English for certain types of spoken English structures. This comparative grammar showed that, with one exception, all structures peculiar to spoken English can be described by transformational extensions and that the greater variability of spoken English in surface structure made the comparative system generally more economical than an alternate system of analysis. Results indicated that the ability to discriminate between spoken and written styles is a linguistic competence which can be accounted for by a system of rules analogous to and isomorphic with the rules of the transformational component of a generative grammar. (Author/JB)