This review describes a meta-analysis of findings from 50 controlled evaluations of intelligent computer tutoring systems. The median effect of intelligent tutoring in the 50 evaluations was to raise test scores 0.66 standard deviations over conventional levels, or from the 50th to the 75th percentile. However, the amount of improvement found in an evaluation depended to a great extent on whether improvement was measured on locally developed or standardized tests, suggesting that alignment of test and instructional objectives is a critical determinant of evaluation results. The review also describes findings from two groups of evaluations that did not meet all of the selection requirements for the meta-analysis: six evaluations with nonconventional control groups and four with flawed implementations of intelligent tutoring systems. Intelligent tutoring effects in these evaluations were small, suggesting that evaluation results are also affected by the nature of control treatments and the adequacy of program implementations.