Purpose of review: This article will evaluate the rationale and feasibility of detecting psychosis and schizophrenia earlier than is currently the case. Recent findings: Schizophrenia incidence may vary more than has been believed previously. Early detection studies fall into two groups. Firstly, operational criteria now exist for prodromal or at risk mental states which predict transition to psychosis of 20-40% over 1 year. The first randomized trials of antipsychotic drug and psychological interventions aimed at reducing this transition rate have shown promising results. Secondly, duration of untreated psychosis in the first episode seems genuinely to be associated with clinical outcome but how much of the association is truly causal remains disputed. Summary: Shortening duration of untreated psychosis by early detection of full psychosis appears feasible but its benefits are not yet unequivocally demonstrated. © 2005 Lippincott Williams & Wilkins.