The multiple sleep latency test (MSLT) is commonly used as an objective measure of sleepiness. We retrospectively correlated MSLT scores from 147 patients with other information relating to sleepiness, namely demographic information, data from nocturnal polysomnograms (PSGs), and subjective assessments. The only variable that showed a valid and statistically significant correlation with the MSLT score was sleep latency on the PSG. The results were largely similar within subgroups focusing on (1) individuals with the ability to fall asleep on every nap, (2) subjects with abnormally short MSLT scores, (3) nap attempts that were successful, and (4) patients with particular diagnoses. We conclude that the mean sleep latency on the MSLT, in a clinical population, does not correlate well with a number of variables expected to influence sleepiness. While the validated utility of the MSLT in separating patients from normals or in identifying narcolepsy is not disputed, the accuracy of the MSLT as a measure of sleepiness appears to be limited.