Various claims have been made about the efficiency of different composition scoring schemes. This paper reports the results of a systematic analysis of the robustness of holistic scoring, analytical scoring, and an obective diagnostic test, Davidson's "Test of the Ability to Subordinate." Compositions from an advanced English as a Second Language composition course were assigned holistic scores by two independent raters. The following variables were quantified for each paper: analytical scores, ratings on grammar, fluency, relevance, vocabulary, and mechanics. In addition, data from the "Test of Ability to Subordinate" were also collected. The data were submitted to a multiple regression analysis, the holistic scores being the dependent variable and all other scores being the predictor variables. The analytical variable accounted for 89% of the variance in the holistic scores. The only other variables to approach significance in accounting for variance in the dependent variable were ratings on vocabulary and fluency. It was not expected that a proficiency metric and a diagnostic metric would yield the same results. A further study is called for employing two different evaluator pools in order to confirm or disconfirm the findings of the present study. (Author/AMH)