Computer-administered questionnaires have been little explored as a potentially effective and inexpensive alternative to pencil and paper screening tests. A self-administered computerised form of the revised Clinical Interview Schedule (CIS-R) was compared with the Composite International Diagnostic Interview (CIDI) in a two-phase study of 2032 Australian high school students (mean age 15.7 years) drawn from a stratified random sample of 44 schools in the state of Victoria, Australia. Prevalence, sensitivity and specificity were estimated using weighting to compensate for the two-phase sampling. Point prevalence estimates of depression using the CIS-R were 1.8% for males and 5.6% for females--an overall prevalence of 3.2%. Prevalence estimates for depression in the past 6 months using the CIDI were 5.2% for males and 16.9% for females--an overall estimate of 12.1%. The CIS-R had a positive predictive value (PPV) of 0.49 and negative predictive value (NPV) of 0.91 for CIDI depression in the past 6 months. Specificity was very high (0.97) but sensitivity low (0.18), indicating that a majority of those with a CIDI-defined depressive episode in the past 6 months were not recognised at a single screening using the CIS-R. Even so, the CIS-R has proved at least as good as any pencil and paper questionnaire in identifying cases for nested case-control studies of adolescent depression. Further exploration of strategies such as serial screening to enhance sensitivity is warranted.