The suspected sources of confounding in current meta-analytic studies of computer based instruction (CBI) are uncontrolled effects of instructional method and/or the John Henry Effect (i.e., compensatory rivalry). To determine which confounding is most plausible, a random 30% sample of the 128 studies which formed the original Kulik meta-analyses for primary, secondary, and college levels was selected for a descriptive review. Evidence from reviewer coding of study features strongly suggests that the achievement gains found in CBI studies are actually due to the uncontrolled but robust instructional methods employed in these treatments. Cautions are offered for those who wish to implement CBI in schools and for research design, and it is suggested that the achievement gains attributed to the computer mode of delivery are probably due instead to the instructional methods employed in their software, and these methods could probably be delivered by a teacher. A list of references and appendices with data tables complete the document. (Author/JB)