Good and poor explanatory essays of 40 college freshmen were analyzed for 18 cohesive ties and chains to determine the appropriateness of the cohesion system for teaching and evaluating writing. The questions that were specifically addressed were, (1) How do writers use the cohesive resources of the language? and (2) How is cohesion related to teachers' perceptions of writing quality? The analysis revealed that the density of ties and length of chains increased disproportionately to the length of essays. A review of individual specimen essays suggested that greater variety and maturity of lexical choice characterized the good essays. Poor essays had frequent pseudochains--long strings of common high-frequency words bearing very little semantic import. Most good and poor essays had a dominant chain connecting several paragraphs. The findings suggest that the cohesion system lacks content and domain selection validity to be appropriate as an evaluation scheme. The system could be used in instruction by a teacher at the point of responding and suggesting revisions, but not as the central emphasis of instruction. (HOD)