During the 1990s, the use of meta-analytic methods in educational research has been widespread, and few aspects of education have escaped the meta-analytic revolution. The acceptance has not been complete, however, and several threats to validity remain. Prominent among these are the "normality" problem and the "independence" problem (whether multiple effect sizes from a single study should be analyzed independently). As a result, resampling methods have been proposed when it is assumed that distributions are nonnormal and multiple effect sizes are independent. However, resampling methods for nonnormal dependent multiple effect sizes have not been found. This paper discusses methods for resampling meta-analysis with dependent multiple effect sizes. First, the literature regarding the use of resampling for a univariate meta-analysis is reviewed. Then, a review of the independence problem (i.e., multiple effect sizes) is provided. Finally, resampling methods for countering the problems of nonnormality and nonindependence for the multivariate meta-analytic case are described. Educational researchers involved with meta-analysis are likely to find multiple effect sizes to be of issue. In most cases, if not all, multivariate methods will be preferred over univariate. Resampling methods can improve multivariate meta-analytic applications. (Contains 15 references.) (SLD)