The aim of this study was to ascertain accurately the prevalence of Tourette syndrome (TS) in a mainstream school population. All year 9 pupils (aged 13 to 14 years) in a mainstream secondary school were investigated using a twostage procedure. Standardized questionnaires were completed by parents, teachers, and pupils. CIass observations were also carried out to identify tics. Those pupils identified as having tics underwent a semistructured interview to determine whether they had TS according to DSM-III-R criteria. Data were available from'at least one source (parent, teacher, or self-reports) on 166 of the 167 pupils in the year. Five subjects were identified as having TS according to DSM-III-R criteria, yielding a prevalence estimate of 299 per 10 000 pupils in this age group. The results of this study suggest that TS in the community as a whole is more common and milder than those prevalence estimates and descriptions of the disorder based on TS encountered in secondary or tertiary health-care service settings. 1989, 1994, 1995).Most TS research has focused on describing the syndrome, its association with other psychopathologies, and o n attempting to investigate aetiological factors (Kobertson 1989, 1994). However, the populations studied have almost invariably been drawn from highly selected specialized tertiary refernl centres. Data derived from such studies will only be free from selection and attribution bhs if they are representative o f all cases of TS. This is unlikely for three main rcasons: first, complex and variable clinical filters exist, benvcen the individual with '15 in the community and specialist care, which need to be negotiated in order to access tcrtiary care; second, recognition and diagnosis ofTS is subject to variability; and third, referral may be dictated by associated bchavioural disturbance n t h e r than by 'IS itxlfwhich, in community samples. is often a mild disorder (Kurlan et id. 1986(Kurlan et id. . 1987 Kobertson and Gourdie 1990; McMahon et al. 1992).Thus, most study populations have been subject to substantial (and often unquantifiable) selection procedures which may limit the extent t o which descriptive and comorbidity data can be generalized. It is also likely that such methodologies will lead t o underestimates of the population prevalence o f the syndromc'. Prevalence estimates depcntl on accuntc data regarding the numerator (the number with thc disorder) and denominator (the size of the population from which the caws are dnwn). If the number of people identitied ;IS having ' I 3 by the rescarch methodology is only a fraction of the true number with ' l 3 then the population prevalence rate will be underestimated. In response to these difficulties, four studies have attempted to estimate the prevalence rate of -IS