Excessive daytime sleepiness (EDS) is an important symptom that needs to be quantified, but there is confusion over the best way to do this. Three of the most commonly used tests: the multiple sleep latency test (MSLT), the maintenance of wakefulness test (MWT) and the Epworth sleepiness scale (ESS) give results that are significantly correlated in a statistical sense, but are not closely related. The purpose of this investigation was to help clarify this problem. Previously published data from several investigations were used to calculate the reference range of normal values for each test, defined by the mean+/-2 SD or by the 2.5 and 97.5 percentiles. The 'rule of thumb' that many people rely on to interpret MSLT results is shown here to be misleading. Previously published results from each test were also available for narcoleptic patients who were drug-free at the time and who by definition had EDS. This enabled the sensitivity and specificity of the three tests to be compared for the first time, in their ability to distinguish the EDS of narcolepsy from the daytime sleepiness of normal subjects. The receiver operator characteristic curves clearly showed that the ESS is the most discriminating test, the MWT is next best and the MSLT the least discriminating test of daytime sleepiness. The MSLT can no longer be considered the gold standard for such tests.