We compared the usefulness of t-test and parametric and rank-sum tests in the statistical analysis of significant differences in the so-called "decision tree" for the quantitative data obtained from toxicity studies. The Dunnett's multiple comparison test had lower analytic power than the t-test when one of the groups showed a marked difference in variance. The Dunnett's test was less efficient with the increase in the number of groups. If one group showed a decrease in the number of animals, this test was less efficient than parametric tests, because the rank-sum tests should be chosen. The rank-sum test is required occasionally to attach the asterisks of significant difference to the mean +/- SD even in showing the same mean values. The nonparametric Dunnett's test could not be used for analysis of significant differences when the mean value for the control and treated groups showed big differences. The nonparametric Dunnett's and parametric Scheffé tests were not as efficient as the other parametric tests probably because of the vague evaluation or overlooking the effect of the test substance.