Background We consider the evidence for the proposition that the early phase of psychosis (including the period of untreated psychosis) is a critical period' in which (a) long-term outcome is predictable, and (b) biological, psychological and psychosocial influences are developing and show maximum plasticity.Method First-episode prospective studies, predictors of outcome and the genesis of patients' key appraisals of their psychosis are reviewed.Results The data support the notion of the ‘plateau effect’, first coined by Tom McGlashan, which suggested that where deterioration occurs, it does so aggressively in the first 2-3 years; and that critical psychosocial influences, including family and psychological reactions to psychosis and psychiatric services, develop during this period.Conclusions The early phase of psychosis presents important opportunities for secondary prevention. We outline a prototype of intervention appropriate to the critical period. The data challenge the widely held assumption that first-episode psychosis is a benign illness posing little risk.