Seven objective measures of language production--number of words, T-units, subordinate clauses, clauses of all types, words per clause, words per T-unit, and clauses per T-unit--were applied to 269 compositions written by seventh, ninth, and 11th graders to test assumptions about the relationships between quality of writing and students' age and maturity levels. Quality ratings of high (H), middle (M), and low (L) were given to each paper, using the evaluation procedures of the STEP test. The following relationships were tested: For total words, total T-units, words per T-unit, number of subordinate clauses, and total clauses, it was hypothesized that H=M, M>L, and H>L; for words per clause, H=M=L; and for clauses per T-unit, HL, and H>L. Results indicated that (1) distinctions between quality of papers were not signaled by differences in ratios of words to T-units, words to clauses, or clauses to T-units, (2) quality was distinguished by significant increases from low to middle to high in total words, total T-units, and total clauses, and (3) on the whole, low papers made use of significantly fewer subordinate clauses than did either middle or high. (Included are tables indicating the statistical measures used and a review of normative studies which previously examined developmental trends in composition quality. (JM)