R. C. Blair (1991) developed tables of critical values for the generalized "t" and generalized rank sum tests that do not suffer inflation of Type I error. This study evaluated the critical values generated by Blair for situations in which sample size varies more than a maximum of a factor of two. The accuracy of the values was explored through a Monte Carlo study in which random samples were generated from populations with known characteristics. The variance in each population, the difference in population means, and the sample sizes were manipulated, with 5,000 samples of each size, from 20 size combinations, generated for each condition (equal size and three ratios of unequal sizes). Results suggest that the critical values proposed by Blair maintain the Type I error rate for the generalized t-test and generalized rank sum tests, even with sample ratios as large as 1:8, and with population variance ratios as large as 1:9. Only under small ample conditions with a sample size ratio of 1:8 and with a conservative alpha level of 0.01 did the generalized t-test not control the Type I error rate using these critical values. (Contains four tables and six references.) (SLD)