There has been a long-standing dispute about the efficacy of computer assisted instruction (CAI) with regard to the interpretation of effect size estimates in reviews using techniques of meta-analysis. It has been claimed that the data used to calculate these estimates come from studies which are methodologically flawed. The aim of this study was to provide an updated meta-analysis on the learning effect of (CAI) over a broad range of study features with particular attention focused on the effectiveness debate. Using standard procedures, the results and estimates were similar to previous reviews and showed a learning benefit for CAI. The mean effect size for CAI was (.24) for the years 1987-1992, with more recent studies showing an average of (.33). Although moderate, these estimates tended to raise the average student from at least the 50th and 60th percentile. However, studies which controlled for teacher and materials, and were of longer duration, and studies using pencil and paper equivalents of CAI showed no learning advantage over traditional forms of instruction. It is suggested that what accounts for the typical learning advantage of CAI in this meta-analysis and others is the better quality instruction provided by CAI materials. These materials seem versatile enough to be used effectively over a broad range of subjects and educational settings. While the materials did not seem to improve substantially over the past two decades as reflected by effect sizes, these estimates did not include the newer multimedia technology. It is concluded that educational approaches should be judged by a number of criteria including achievement gains and when this is done CAI may far surpass other forms of instruction. *This research was supported by a grant from Telecom, NZ. Special thanks are extended to Bill Flinn for provision of the combinational formulae used for the statistical analysis.