BACKGROUND It is well recognized that sleep is severely disturbed in patients in intensive care units (ICU) and that this can compromise their rehabilitation potential. However, it is still difficult to objectively assess sleep quantity and quality and the determinants of sleep disturbance remain unclear. The aim of this study was therefore to evaluate carefully the impact of ICU sound intensity levels and their sources on ICU patients' sleep over a 24-h period. METHODS Sleep and sound levels were recorded in 11 ICU intubated patients who met the criteria. Sleep was recorded using a miniaturized multi-channel ambulatory recording device. Sound intensity levels and their sources were recorded with the Nox-T3 monitor. A 30-s epoch-by-epoch analysis of sleep stages and sound data was carried out. Multinomial and binomial logistic regressions were used to associate sleep stages, wakefulness and sleep-wake transitions with sound levels and their sources. RESULTS The subjects slept a median of 502.2 [283.2-718.9] min per 24 h; 356.9 [188.6-590.9] min at night (22.00-08.00) and 168.5 [142.5-243.3] during daytime (8 am-10 pm). Median sound intensity level reached 70.2 [65.1-80.3] dBC at night. Sound thresholds leading to disturbed sleep were 63 dBC during the day and 59 dBC during the night. With levels above 77 dBC, the incidence of arousals (OR 3.9, 95% CI 3.0-5.0) and sleep-to-wake transitions (OR 7.6, 95% CI 4.1-14) increased. The most disturbing noises sources were monitor alarms (OR 4.5, 95% CI 3.5-5.6) and ventilator alarms (OR 4.2, 95% CI 2.9-6.1). CONCLUSIONS We have shown, in a small group of 11 non-severe ICU patients, that sound level intensity, a major disturbance factor of sleep continuity, should be strictly controlled on a 24-h profile.