This paper isolates and examines separately three distinct sources of error in essay scores: lack of agreement between raters; inconsistencies in performance within mode of discourse, and inconsistencies in performance between modes of discourse. Essay prompts in the Iowa Tests of Basic Skills (ITBS) Writing Supplement were designed to assess students' skills in five modes of discourse: the narrative, the explanation, the description, the informative report, and the persuasive essay. A subsample of 46,000 students participated in the spring, 1985, standardization of the ITBS. One quarter of these 46,000 were assigned to the standardization of the ITBS Writing Supplement, and this report considers data from 10 of 20 subsamples from this quarter. Estimates of reader reliability range from .83 to .95, those for within-mode score reliability range from .37 to .69, and those for between-mode score reliability are generally in the .30s. This study reports several indicators of test validity: (1) measures of prompt appropriateness and scoring protocol usefulness; (2) evidence that the essays collected measure both small increments of growth and significant increases in writing proficiency; and (3) evidence for the discriminant validity of the essays. (Author/PN)