A general model is described which can be used to represent the four common types of meta-analysis: (1) estimation of effect size by combining study outcomes; (2) estimation of effect size by contrasting study outcomes; (3) estimation of statistical significance by combining study outcomes; and (4) estimation of statistical significance by contrasting study outcomes. The model explicitly allows for nonindependence among study outcomes, providing exact statistical solutions when the nonindependence can be estimated. Some consequences of ignoring nonindependence are described. When combining outcomes, failure to adjust for nonindependence will result in a Type I error. This will also happen if two studies are compared across correlated units. When two different outcomes within studies are compared, failure to adjust for nonindependence will inflate the Type II error rate. It is concluded that the effect of nonindependence on the outcome of a meta-analysis depends on the type of question being asked. (GDC)