The second of four symposium papers brings together the results of four separate statistical analyses of findings on computer-based instruction (CBI) in 199 comparative studies, of which 32 were conducted in elementary schools, 42 in high schools, 101 in universities and colleges, and 24 in adult education settings. Each was a controlled, quantitative study which met predefined standards for methodological adequacy. The following results and study features were observed: (1) students generally learned more in classes in which they received CBI; (2) students learned their lessons with less instructional time in classes in which they received CBI; (3) students liked classes more when they received computer help; (4) students developed more positive attitudes toward computers when they received help from them in school; (5) computers had little effect in the areas of course withdrawal and attitude toward subject matter; (6) study results were consistently stronger in published studies than in those that were not published; (7) effects were greater when different teachers taught experimental and control groups; (8) effects tended to be greater in more recent and in short studies; and (9) effects were somewhat greater in studies reported in journal articles than in dissertations. It is suggested that editorial screening, experimental design flaws, and instructional quality may have contributed to this favorable picture of CBI. The four meta-analyses are listed as references. (MES)