Among 600 medical clinic attenders in Lahore, Pakistan, screened using the Bradford Somatic Inventory in Urdu, significant differences in symptom endorsement between 'organic' and 'mood-disordered' patients were found in 14 BSI items. Counting only those symptoms reported as present on more than 15 days during the previous month, significant differences in symptom endorsement were found in 21 BSI items. Another 60 patients attending the same clinics were recruited to evaluate the BSI as a screening questionnaire for mood disorders. Each patient was interviewed by a psychiatrist and a diagnosis made according to DSM III-R and ICD-9. Optimal cutting points for the 14-item, 21-item and 44-item BSI were determined by Relative Operating Characteristic (ROC) analysis. The areas under the fitted ROC curves (an index of discriminating efficiency) were similar for the three versions. Using a cut-off score of 13/14, the BSI-21 identified DSM III-R psychiatric disorders with an estimated sensitivity of 75% and specificity of 75%. Assuming a 30% prevalence in medical clinics, a positive value of around 60% can be expected when screening for psychiatric disorders.