A systematic sleep questionnaire has been devised for assessing the previous night's sleep of a subject. It has been designed for repeated use. It is completed by the subject and is framed with the needs of the hospital patient in mind. In the present study it was given to 93 subjects in four different groups: 16 surgical inpatients, 21 medical inpatients, 32 psychiatric inpatients (in a general hospital unit), and 24 normal volunteers. Test retest reliability correlations have been derived using a nonparametric correlation coefficient (Kendall's tau). Each of the items achieved statistically significant reliability (p less than 0.0001) in all four groups, with the value for tau on the total sample varying from 0.70 to 0.96. The St. Mary's (or SMH) Sleep Questionnaire is put forward as an instrument that is a systematic inquiry into the subject's experience of sleep and that is composed of items of demonstrable reliability.