Background Sleep deprivation and delirium are major problems in the ICU. We aimed to assess the sleep quality by polysomnography (PSG) in relation to delirium in mechanically ventilated non-sedated ICU patients. Methods Interpretation of 24-h PSG and clinical sleep assessment in 14 patients. Delirium assessment was done using the confusion assessment method for the intensive care unit (CAM-ICU). Results Of four patients who were delirium free, only one had identifiable sleep on PSG. Sleep was disrupted with loss of circadian rhythm, and diminished REM sleep. In the remaining three patients the PSGs were atypical, meaning that no sleep signs were found, and sleep could not be quantified from the PSGs. Clinical total sleep time (ClinTST) ranged from 2.0-13.1 h in patients without delirium. Six patients with delirium all had atypical PSGs, so sleep could not be quantified. Short periods of REM sleep were found. ClinTST was median 8.5 h (range 0.4-13.8 h). EEG reactivity and wakefulness was found in all but one PSG. Four patients were CAM-ICU "unassessable" (unresponsive to voice). PSGs were atypical without reactivity or wakefulness, even though clinical wakefulness was documented. ClinTST was median 18.3 h (range 3.7-19.8 h). Paroxystic EEG activity was found in this subgroup. Conclusions The objective signs of sleep were absent in all but one PSG, so even though patients were not sedated, sleep could not be quantified. Even in patients without delirium, sleep could only be quantified in one of four patients. Paroxystic activity is frequent in unsedated patients, unresponsive to voice, but the implication is unknown.