Variables that best explain variance in student attrition for 168 freshmen students from a college of agriculture within a major midwestern university were studied in 1979. Variables were selected from a theoretical causal model of student attrition (Bean, 1981), a review of the literature (Bean, 1978), and five empirical investigations conducted by Bean. Nine main variables were: parents' educational attainment, performance in high school, absenteeism, the practical value of one's education, confidence in being a successful student, intent to leave, university grades, informal contact with faculty, and course attractiveness. Responses to a Likert-scale type attitude measure and information about the students were analyzed using stepwise multiple regression analysis. Intent to leave was the strongest indicator of attrition. The nine variables and the interaction of five of these with grade point average were found to account for 62 percent of the variance in attrition. Students with different levels of academic ability withdrew from the institution for systematically different reasons. The effects of the variables (particularly practical value, absenteeism, confidence, faculty contacts, and courses) differed depending upon academic ability. Academic performance is not only an important predictor of student attrition, but interacts strongly with other variables affecting decisions to leave school. The importance of studying small, homogeneous groups in order to accurately describe a selected group (e.g., agricultural college students) is noted. A bibliography is appended. (SW)