This paper summarizes three companion computer-assisted instruction (CAI) case studies in a series for the last several years. These studies were designed to investigate a various of impacts of an Earth Science Computer-Assisted Tutorial (ESCAT), developed by the research team, on students' Earth science achievement and their attitudes toward Earth Science in senior high schools. The pretest-posttest control-group experimental design was adopted by all the three studies. Quantitative data were collected on students' pre- and post-treatment achievement and attitudes toward Earth Science measures. The multivariate analysis of covariance (MANCOVA) and analysis of covariance (ANCOVA) revealed that (a) the ESCAT has promise; (b) the "problem-solving" design of ESCAT is generally better than the "nonproblem-solving" design of the ESCAT; and (c) the "teacher-directed" ESCAT seems to produce statistically greater student gains than the "student-controlled" ESCAT. These findings suggested that instruction, such as the ESCAT should be more broadly developed and widely employed in the secondary Earth Science classrooms.