This meta analysis examines the effect of computer-assisted instruction (CAI) upon college student performance, addresses the impact of various study characteristics upon effects, and explores how media richness theory may predict CAI performance gains. Findings indicate that student performance gains are larger for CAI than traditional instruction (average r=.12, k=118, N=12,398). CAI benefits are greatest for social science disciplines, when the traditional instruction format is lecture/discussion, for undergraduate samples, in studies published after 1994, and for CAI delivered in multiple units. Analyses of media richness constructs indicate that CAI increases performance when delivered with an audio channel and is designed for a specific set of students. (Contains 2 tables.)