The clinical validity of diagnostic criteria for sleep orofacial motor activity--more specifically, bruxism--has never been tested. Polysomnographic recordings from 18 bruxers and 18 asymptomatic subjects, selected according to American Sleep Disorders Association criteria, were analyzed (1) to discriminate sleep bruxism from other orofacial motor activities and (2) to calculate sensitivity, specificity, and predictive values of research criteria. Clinical observations and reports revealed that all 18 bruxers reported frequent tooth-grinding during sleep. Tooth wear was noted in 16 out of 18 bruxers and jaw discomfort reported by six of them. These findings were present in none of the controls. The analysis of polysomnographic data showed that the asymptomatic subjects presented a mean of 1.7 +/- 0.3 bruxism episodes per hour of sleep (sustained or repetitive bursting activity in jaw closer muscles), while bruxers had a significantly higher level of activity: 5.4 +/- 0.6. Controls exhibited 4.6 +/- 0.3 bruxism bursts per episode and 6.2 (from 0 to 23) bruxism bursts per hour of sleep, whereas bruxers showed, respectively, 7.0 +/- 0.7 and 36.1 (5.8 to 108). Bruxism-like episodes with at least two grinding sounds were noted in 14 of the 18 bruxers and in one control. The two groups exhibited no difference in any of the sleep parameters. Based on the present findings, the following polysomnographic diagnostic cut-off criteria are suggested: (1) more than 4 bruxism episodes per hour, (2) more than 6 bruxism bursts per episode and/or 25 bruxism bursts per hour of sleep, and (3) at least 2 episodes with grinding sounds. When the polysomnographic bruxism-related variables were combined under logistic regression, the clinical diagnosis was correctly predicted in 81.3% of the controls and 83.3% of the bruxers. The validity of these clinical research criteria needs now to be challenged in a larger population, over time, and in subjects presenting various levels of severity of sleep bruxism.