Discrimination of temporal patterns has been suggested as a relevant process in speech recognition by subjects with normal hearing [Sorkin, J. Acoust. Soc. Am. 87, 1695–1701 (1990)]. This paper investigates whether performance of Nucleus multichannel cochlear implant subjects on a temporal pattern discrimination task is an efficient and valid psychophysical measure of speech recognition ability. Stimuli consisted of temporal sequences defined by twelve 35-ms tones and eleven randomly generated temporal gaps separating the tones. A fixed-level same/different paradigm was used to measure the discriminability of these sequences as a function of their average correlation across a block of trials. On each trial, the ‘‘standard’’ sequence was generated randomly by drawing gap durations from a Gaussian distribution. The gaps of the comparison sequence were generated in a similar fashion with a specified average correlation with the gaps of the first sequence. Performance of implanted and normal hearing subjects decreased monotonically with increasing average sequence correlation. However, performance across implanted subjects ranged from that observed for acoustically stimulated subjects with audiometrically normal hearing to levels near chance. Comparing these data with measures of speech recognition in the same subjects, we have found that performance on standard speech recognition tests correlates with ability to discriminate among such random temporal patterns.