We present a likelihood-based statistical method for examining the pattern or rate of evolution of reproductive isolation. The method uses large empirical datasets to estimate, for a given clade, the average duration of two phases in the divergence of populations. The first phase is a lag phase and refers to the period during which lineages diverge but no detectable reproductive isolation evolves. The second is an accumulation phase, referring to the period during which the magnitude of reproductive isolation between diverging lineages increases. The pattern of evolution is inferred from the relative durations of these two phases. Results of analyses of postzygotic isolation data indicate significant differences among taxa in the pattern of evolution of postzygotic isolation that are consistent with predictions based on genetic differences among these groups. We also examine whether the evolution of postzygotic isolation is best explained by either of two models for the rate of accumulation: a linear model or a quadratic function as may be suggested by recent studies. Our analysis indicates that the appropriateness of either model varies among taxa.