Sleep research has made extensive use of self-report measures relying on a response format that requires respondents to provide single, specific numerical estimates. The cognitive processes involved in storing and retrieving sleep-related information allow only approximate numeric estimates of sleep behavior. Based on a fuzzy set model of survey responses, a response format is proposed to better capture the inherent vagueness of quantitative estimates of sleep behavior. Ninety-three adults (mean age 29.3 years) participated in two interviews, 1 month apart, consisting of questions about health-related behaviors. Questions were asked in both traditional point estimate and fuzzy response formats. A subset of questions was repeated at the end of each interview to examine test-retest reliability. Subjects filled out daily diaries each morning during the month between interviews. Quantitative estimates of usual sleep behavior were found to be highly reliable. Point estimates differed significantly from fuzzy boundary estimates. Differences between lower and upper boundary estimates indicated substantial ranges in estimates of usual sleep: total sleep time (6-8.2 h), sleep latency (26-44 min), bed time (11:25 PM-12:56 AM) and waking time (7:24-8:39 AM). Mean diary-recorded bed and waking times fell between the lower and upper boundary estimates in more than 53% of cases, and between the lower and upper extreme estimates in more than 92% of cases. Fuzzy response formats provide an opportunity to report more completely what can be recalled about sleep behavior. They also provide a useful framework for better understand the meaning of traditional point estimate formats.