This study was built on the hypothesis that linguistic cues distinguish the speech of the poor and the ethnically different. Eight samples of connected interview discourse were taken from a two-by-two-by-two matrix of sex, race and social status. The samples were played to groups of elementary, secondary and college students who rated each voice for the highest job each speaker could hold. Listeners agreed strongly about the relative quality of speech demanded by each occupation, but they did not agree with one another in their ratings of the voices or with themselves when they were retested. This study does not support the viability of dialect as a reliable cue in social perception. (Author)