We examined 129 patients recruited from two sleep clinics to study the sleep apnea syndrome (SAS), defined by the apnea-hypopnea index (AHI) > or = 10. Information was registered from a self-administered questionnaire, basal physical measurements, and polysomnography. In 68 subjects recorded for two consecutive nights, a high correlation was found between first- and second-night AHIs (r = 0.89). Habitual loud snoring and breathing arrests during sleep were associated with AHI > or = 10. A model including these two variables, sex, age, and body mass index was created in order to predict AHI > or = 10 and with which it was possible to successfully classify almost three of four patients. Among subjective sleep questionnaire items, only daytime sleepiness was related to drops of transcutaneous oxygen tension. These discrepancies in the observed relationship between sleep parameters and subjective sleep items reduce the questionnaire value in epidemiologic settings where it aimed to detect SAS, as defined solely by the AHI value.