Depressions in general practice tend to be mild and non-endogenous. The value of antidepressants and indications for their use in such depressions are not well documented. One hundred and forty one general practice depressives were treated from four to six weeks with amitriptyline in median dose 125 mgs daily or placebo in a randomised controlled trial. Significant drug placebo differences were apparent by two weeks and were strong by six weeks. Detailed examination of symptom ratings showed that the effect was a true antidepressant one, on core symptoms of depression. Interactions were examined between drug-placebo differences and a range of predictor characteristics. There were no significant interactions for demographic characteristics, history of illness or endogenous depression in terms of symptom pattern, absence of life stress or non-neurotic personality. Drug was superior to placebo in all subgroups. Only for initial severity of illness were significant interactions found. Amitriptyline was superior to placebo in probable and definite major depressions on the Research Diagnostic Criteria but not in minor depressions. It was also superior to placebo in patients with initial Hamilton Scale scores of more than 12, but not 6-12. Overall these findings indicate that tricyclic antidepressant is of clear therapeutic benefit in a spectrum of milder depressions except for the most mild of these.