Sampling distributions for ten tests for comparing population variances in a two group design were generated for several combinations of equal and unequal sample sizes, population means, and group variances when distributional forms differed. The ten procedures included: (1) O'Brien's (OB); (2) O'Brien's with adjusted degrees of freedom; (3) Brown-Forsythe (BF); (4) Welch-James test on O'Brien's r transform (WJOB); (5) unaligned Klotz; (6) median-aligned Klotz; (7) mean-aligned Klotz; (8) unaligned Siegel-Tukey; (9) median-aligned Siegel-Tukey; and (10) mean-aligned Siegel-Tukey. Type 1 error rates and power estimates were compared using simulated data. The present results in conjuction with previous research indicates that no one procedure for comparing variances is uniformly the best strategy. With equal-sized samples a strategy that will yield appropriate Type 1 errors and reasonably powerful tests is to use OB unless there is evidence of heavy tails. In that case BF can be used to increase power while retaining appropriate Type 1 error rates. With unequal sample sizes and when the evidence suggests a direct relationship between the sample sizes and population variances the strategy might be modified to use WJCB in place of OB. (PN)