AIMS: Although the prevalence of alcohol problems amongst detainees is suspected to be high, often only more flagrant problems are detected, restricting the possibilities for intervention in alcohol misuse and reducing opportunities for preventive efforts. This study examined the re-test reliability of AUDIT (the Alcohol Use Disorder Identification Test) in screening prisoners METHODS: AUDIT was administered for the first time on the day of entry to prison and again about 15 days later. The results were analysed according to two AUDIT thresholds: a score of 8 or higher and 12 or higher. RESULTS: Of 75 consecutive entrants tested, 47 male prisoners completed the study. At the first administration, 19.1% of these 47 men met criteria for a probable alcohol problem but this percentage rose to 59.6% on the second occasion (P = 0.0001). The proportion of subjects with a score 12 or higher (probably dependent) was 10.6% the first time versus 42.6% the second time (P = 0.0001). In the 19 who scored positive at the second administration only, changes in answers to the 10 items were coherent with a total score growing from 3.0 to 18.1 (P = 0.0001). No prisoner had a lower AUDIT score on the second administration. As alcohol problems are not routinely considered during the medical and biological examination at entry, no confirmation of the AUDIT results could be obtained, although those obtained at the second administration fitted well with the prevalence rates in previous reports. CONCLUSIONS: AUDIT, for the purpose of giving a prevalence estimate or to enter appropriate prisoners into more detailed assessment or interventions, should not be conducted immediately at entry, but some weeks later.