Mild depressive syndromes are highly prevalent among primary-care patients. Evidence-based treatment recommendations need to be derived directly from this diagnostically heterogeneous group. The primary aim was to assess the efficacy of sertraline and cognitive-behavioural group therapy for treatment of depressed primary-care patients, the secondary aim was to evaluate if receiving treatment according to free choice is associated with a better outcome than randomization to a particular treatment. We conducted a randomized, placebo-controlled, single-centre, 10-wk trial with five arms: sertraline (flexible dosages up to 200 mg/d) (n = 83); placebo (n = 83); manual-guided cognitive-behavioural group therapy (one individual session and nine group sessions per 90 min) (n = 61); guided self-help group (control condition, n = 59); and treatment with sertraline or cognitive-behavioural group therapy according to patients' choice (n = 82). From 1099 consecutively screened adult patients, 368 formed the intent-to-treat population with milder forms of depression. Primary outcome was a global efficacy measure combining z-converted Hamilton Depression Rating Scale and clinician-rated Inventory for Depressive Symptomatology scores. Sertraline was superior to placebo (p = 0.03). Outcome for guided self-help groups was worse compared to cognitive-behavioural group therapy (p = 0.002) and compared to all other treatment arms including pill placebo (secondary analyses). Outcome in the patients' choice arm was similar to that in the sertraline and cognitive-behavioural group therapy. Overall, sertraline is efficacious in primary-care patients with milder forms of depression. The superiority of cognitive-behavioural group therapy over guided self-help groups might partly be explained by 'nocebo' effects of the latter.