The test-retest reliability of the Multiple Sleep Latency Test (MSLT) was evaluated in 14 healthy normal subjects. Each slept a single night in the laboratory (8 h time in bed) and received the MSLT the following day (1000, 1200, 1400, and 1600 h) on two occasions separated by 4-14 months. Mean sleep latency (four tests) was highly reliable from MSLT to MSLT (r = 0.97, p less than 0.001). The test-retest reliability did not change as a function of the interval of time between tests or as a function of the level of sleepiness (range = 4-20 min) within the population. However, as the number of tests comprising the MSLT was reduced below three, the reliability was reduced such that only 50% or less of the variance could be predicted.