A literature search shows that robust techniques are rarely used in the animal sciences. Traditionally, normal theory tests are used to determine whether there are significant differences among group means. However, normal theory tests are optimal only if the distribution of error terms is normal and in practice, non-normal distributions are more prevalent. Therefore, in this study, it is aimed to present robust tests based on modified maximum likelihood (MML) estimators for testing the equality of the group means. A real data set about the effect of zeolite and initial weight on feedlot performance of Brown Swiss cattle is the subject of the study. Results showed that the test statistics presented in this study are more powerful than the traditional normal-theory tests. They are also more robust against departures from assumed distribution and outliers.