The effects of varying amounts of computer assisted instruction (CAI) in mathematics, on the academic performance of 446 fifth and sixth graders scoring below norm on the California Test of Basic Skills (CTBS) were investigated. Independent variables were: (1) beginning-of-year CTBS mathematics pretest score; (2) teacher's verbal ability; (3) years of teacher experience; (4) teacher degree level; (5) student self-expectation; (6) number of CAI sessions that a student had in the year; and (7) intelligence test score. By regression analysis, end-of-year CTBS posttest scores were compared with these independent variables using linear, Cobb-Douglas (homogeneous and nonhomogeneous), and transcendental logarithmic model specifications of student achievement. CAI had a significant and positive impact on achievement in most cases. Typically, 100 5-10 minute CAI sessions--given daily to weekly--can, during the course of a school year, raise a disadvantaged student's grade placement in mathematics by perhaps .3 years over what the achievement gain would have been otherwise. Furthermore, at the present cost of $25 to $75 per student for provision of 100 sessions per year, CAI is substantially less expensive than most alternatives for compensatory education. Advantages and limitations of the Cobb-Douglas and transcendental-logarithmic specifications are also discussed. (Author/CP)