Background: Self-help for social phobia has not received controlled empirical evaluation. Aims: To evaluate the efficacy of pure self-help through written materials for severe social phobia and self-help augmented by five group sessions with a therapist. These conditions were compared with a waiting-list control and standard, therapist-led group therapy. Method: Participants with severe generalised social phobia (n=224) were randomised to one of four conditions. Assessment included diagnoses, symptoms and life interference at pretreatment, 12 weeks and at 24 weeks. Results: A larger percentage of patients no longer had a diagnosis of social phobia at post-intervention in the pure self-help group than in the waiting-list group, although this percentage decreased slightly over the next 3 months. Symptoms of social anxiety and life interference did not differ significantly between these groups. Augmented self-help was better than waiting list on all measures and did not differ significantly from group treatment. Conclusions: Self-help augmented by therapist assistance shows promise as a less resource-intensive method for the management of social phobia. Pure self-help shows limited efficacy for this disorder. Declaration of interest: R.M.R. authored the book used for bibliotherapy and receives royalties.