In previous work on computer speaker verification [Doddington, J. Acoust. Soc. Amer.49, 139(A) (1971)], low error rates were obtained when nonlinear time transformation was used for temporal registration of sample and reference utterances. In that study, utterances were represented by their pitch, amplitude, and formant‐frequency profiles. The present study considers how well one might do only with measures which can be obtained in real time. The procedure differs from that of the previous study in the following respects: (a) utterances are represented only by pitch and amplitude profiles, (b) these functions are smoothed by low‐pass filtering, and (c) a greater number of dissimilarity measures are computed. Because formants are not extracted, the verification is accomplished about 100 times faster than in the previous study. Performance was measured with the same speech material used in the earlier study. Included are utterances from eight “true” speakers and 32 “casual impostors.” A casual impostor is one who makes no attempt to mimic the true speaker. The acceptance‐rejection criterion was adjusteda posteriorifor equal rates of false acceptance and false rejection. The resulting error rate, averaged over speakers, was 1.0%.