This study investigated liberal arts college students' preferences for various teaching methods and testing techniques in relation to their measured cognitive style. A total of 100 students completed the Gregorc Style Delineator and a questionnaire on instructional methods commonly used in college classes. The results indicated that the sequential-random dimension of Gregorc's model was the stronger predictor of learning style. High sequential scorers preferred teaching methods such as structured lecture and independent lab experiments, and tests comprised of problems with concrete answers. These students tended to be science majors rather than social science or humanities majors. High random scorers preferred group discussion and group projects, and assessment by projects or class discussion. These students tended to be humanities majors. It is concluded that Gregorc's cognitive styles appear to be related to the instructional preference of liberal arts college students. (Contains 10 references.) (Author/MDM)