An electrical laryngeal analog was used to generate stimuli which varied in the magnitude of frequency differences between successive cycles. For each of two median fundamental frequencies, a 100 and a 200 cps (cycles per second) condition, stimuli were generated to have frequency variations on successive cycles of + 10 cps, 8 cps, 6 cps, 4 cps, 2 cps and 1 cps. 535 listeners judged the stimuli on the basis of which sounded the most rough. The results show that even very slight frequency variations, as little as [plus or minus] 1 cps around a median fundamental frequency of 100 cps, sounded rough. The data also indicate that the magnitude of the roughness judgement was directly related to the frequency differences between successive cycles. The preceding statement held true for a given median frequency value. Thus, for the two median frequency conditions used in this study, a 100 cps and a 200 cps median frequency, the degree of perceived roughness was directly related to the frequency differences between successive cycles for a single condition; the 200 cycle condition, however, was always judged to be less rough when the same frequency variations were programmed. Thus, it would be anticipated that if one were given two voices, a male and a female, with equal frequency variations, the male would be judged to have the most rough voice.