Since meta-analysis was described in 1976 (Glass) as the application of familiar experimental methods to the integration of available research, at least five coherent approaches to meta-analysis have appeared in common use. These approaches can be divided into two broad groups. In the first group (including procedures by Robert Rosenthal, Larry Hedges, and Frank Schmidt and John Hunter), meta-analysis is used to approximate data-pooling. This type of meta-analysis attempts to answer the same questions as primary research, only larger samples are used by combining information from many studies. The alternate view (shown in the procedures of Gene Glass and James Kulik) is that meta-analysis is a form of literature review. As a literature review, meta-analysis is not meant to test a hypothesis but to summarize features and outcomes of a body of research. The differences in approaches to meta-analysis indicate that the procedure is still in a period of invention and change. It is important that editors, consumers, and critics of meta-analysis know about these differences so that they can make more informed evaluations of meta-analytic findings. (Author/PN)