Using a piezoelectric transducer, wrist activity was recorded simultaneously with electroencephalogram (EEG), electro-oculogram (EOG), and submental electromyogram (EMG) to obtain 102 recordings--39 from hospital patients and 63 from nonpatients. On a minute-to-minute basis, wrist activity alone was used to estimate Sleep Time. Blind independent scoring of the EEG-EOG-EMG records was also done to distinguish Sleep and Wake phases. Results from the two Sleep/Wake estimations agreed for 94.5% of the minutes ((96.3% among nonpatients). Correlations between the two methods were determined for Total Sleep Period (r = 0.90). Total Sleep Time (r = 0.89), Wake After Sleep Onset (r = 0.70), and the number of Midsleep Awakenings (r = 0.25). Correlation coefficients were higher when the 39 patients were excluded from the computations. On the average, the actigraphic method overestimated Sleep Time by 15 min. Continuous wrist activity recordings provide simple and inexpensive, but rather accurate, estimates of sleep duration.