Three main techniques have been used to review research in the area of ability grouping: narrative review, meta-analysis, and best-evidence synthesis. Narrative review is the "traditional" method in which the reviewer surveys and comments in detail upon individual studies in the literature. While narrative review permits a great deal of evaluative commentary on the studies it includes, reviewers have always struggled with the difficulty of comparing studies with different results and different standards of measurement. Meta-analysis and best-evidence synthesis, the methods used in the two sets of reviews that form the focus of this article, were developed in order to make the results more replicable and quantifiable than the narrative technique permits.
The meta-analytic technique (used by James Kulik and Chen-Lin Kulik) requires the reviewer to locate studies of an issue through objective and replicable searches, code the studies for salient features, and describe study outcomes on a common scale. Kulik and Kulik itemized additional qualifications for the use of a study in their meta-analysis. In order for a study to be included in their review, the results had to be reported in quantitative form; the results had to be available from a conventionally instructed control group as well as from the one receiving the experimental treatment; the control group had to be similar to the experimental group in aptitude; and very important, the studies had to take place in actual classrooms, not labs. The best-evidence synthesis technique (used by Robert Slavin) is a combination of meta-analysis and narrative review. It has many characteristics in common with meta-analysis, including the computation of effect size* and the clear specification of inclusion criteria. There are, however, several crucial differences. One important difference is that studies were included whose effect size could not be computed. Such studies are characterized in the data analyses as positive, negative, or zero rather than excluded. In addition, individual studies and methodological and substantive issues are disclosed in the detail typical of narrative reviews. Finally, the Slavin review included studies that used calculations that Kulik and Kulik considered mathematically inappropriate for their meta-analytic techniques.
*Effect size is computed as the difference between the mean scores of experimental and control groups, divided by the standard deviation of the control group. It provides a common scale that standardizes the various measurements used in different studies (Kulik and Kulik 1989 and Slavin 1989, 1990)