This study was designed to reveal some of the facets of writing proficiency that different kinds of timed writing exercises might reflect. The approach taken was to examine how performance on these exercises correlates with performance on several nontest indicators of writing skill: self-assessments of writing skills; self-reports of writing-related accomplishments and relevant undergraduate grades; and independently graded samples of writing. The writing exercises investigated in this study were of several kinds: composing a relatively traditional expository essay, writing somewhat briefer responses involving either explaining the meaning of a quotation or presenting ideas in a specified format, and revising a poorly written composition. A geographically diverse sample of 470 undergraduate students, all of whom had taken the GRE General Test, were recruited to participate in the study. Subgroups of the study sample tried various combinations of the writing tasks. The exercises appeared to have differential utility for a GRE writing test. A traditional issue- based expository essay appeared to be the most promising candidate. An exercise involving the revision of a composition also appeared to have potential. Two other experimental tasks, however, did not, from their relationships with the writing indicators, seem as defensible.