Total sleep time is important in investigations of obstructive sleep apnoea, since the diagnosis is usually based on the average number of apnoeas per hour of sleep. Sleep estimates instead of exact EEG-recorded total sleep time is often used in the clinical setting. However, an overestimated sleep time would underestimate the degree of the disease and vice versa. The purpose of this study was to investigate the accuracy of subjective sleep time and time-in-bed as sleep estimates. One hundred patients undergoing diagnostic polysomnography for suspected obstructive sleep apnoea were asked to estimate their sleep time in a questionnaire. Seventy-five patients were diagnosed as suffering from obstructive sleep apnoea syndrome. The mean difference between self-scored and EEG-recorded total sleep time was 4 +/- 74 min. However, 30% scored with a difference greater than 1 h. The intra-class correlation coefficient was fair (0.58, CI: 0.43-0.70). Fifty-three patients overestimated their sleep time and 47 patients underestimated it. All but four patients underestimated their number of awakenings (P<0.001). The mean difference between time-in-bed and EEG-recorded total sleep time was 110 +/- 63 min. This difference was significantly larger than the difference between subjective sleep time and EEG-recorded total sleep time (P<0.001). The intra-class correlation coefficient was poor (0.38, CI: 0.20-0.54). Mean AHI was 27 +/- 27 using subjective sleep time and did not change significantly compared with the mean AHI of 25 +/- 21 based on EEG-recorded total sleep time. Mean AHI decreased significantly to 20 +/- 17 (P<0.001) when time-in-bed was used. In conclusion, 'time-in-bed' time is a poor predictor of total sleep time and should not be used when calculating the apnoea-hypopnoea index. Subjective sleep time is better as an approximation, but the individual differences are large.