Two grammars (traditional and transformational) and two methods of teaching (direct and indirect) were measured to determine their effects upon the writing performance of students of low socioeconomic backgrounds. While the direct teaching of grammar was concerned with the correction of specific errors, the indirect teaching was concerned with a generalized grammar with no specific errors emphasized. The subjects, 160 ninth grade students from Memphis, Tennessee, were required to write four compositions--the better of the first two serving as a pretest, the better of the second two, as a posttest. These were then analyzed and rated according to (1) the number and proportions of well-formed sentences, (2) the number of transformations in the well-formed sentences, and (3) average structural complexity scores for the well-formed sentences. Findings indicated no significant differences in the proportion of well-formed sentences or in structural complexity scores. Neither was there any significant interaction among the type of grammar, method of teaching, and the before and after scores. However, there was a significant correlation between IQ and proportions of well-formed sentences. (Author/HS)