This study examined the role of affect and conation in high school students science test performance, using a sample of 491 students. It provides a profile of partial corrections of standardized multiple-choice and constructed response test scores with affect and conation scores (after accounting for general ability and student background) at three distinct levels of generality: domain-specific, task-specific, and situation-specific. Results show differential patterns of correlations, varying with level of generality of affective and conative constructs, and with different aspects of science achievement represented by the dimensions of basic knowledge and reasoning, quantitative science reasoning, and spatial-mechanical reasoning. The discussion invokes several theoretical frameworks to interpret these results. The report concludes by stressing the need for empirical and theoretical integration in the study of noncognitive elements in academic task performance. (Contains 7 figures, 1 table, and 27 references.) (Author/SLD)