Working on the assumption that listeners would be more convinced and persuaded by a speaker with whom they could identify, a study asked black and white college students to evaluate a speaker with a black or white dialect, introducing them to an innovation in mathematics. Half of the black subjects were randomly assigned to listen to a black speaker; the other half were assigned to listen to a white speaker. The same procedure was followed for the white subjects. It was hypothesized that the speaker with a dialect similar to the recipient would be rated more credible than the dissimilar counterpart, and would produce more attitude change and information recall. The analysis revealed that none of the three hypotheses was supported. However, when the data were validated by selecting only those subjects who had correctly indicated ethnic similarity/dissimilarity with the speaker, it was found that the similar speaker was thought to be significantly more credible than a dissimilar speaker. The two hypotheses regarding attitude change and recall remained unsupported. (HOD)