OBJECTIVES: Studies investigating suicidal behaviour in psychosis rarely focus on incidence cohorts of first-episode patients. This is important, because patients who refuse study participation have higher rates of comorbid substance use disorders and longer duration of untreated psychosis as well as worse course illness, variables potentially linked to higher prevalence of suicidal behaviour. The aims of the present study were therefore to examine the prevalence and predictors of suicide and suicide attempt before and during the first 18-24 months of treatment. METHOD: A retrospective file audit of 661 patients was carried out. RESULTS: Six patients (0.9%) died by suicide, 93 (14.3%) attempted suicide prior to entry, and 57 (8.7%) did so during treatment. Predictors of suicide attempt were: previous attempt (odds ratio (OR)=45.54, 95% confidence interval (CI)=9.46-219.15), sexual abuse (OR=8.46, 95%CI=1.88-38.03), comorbid polysubstance (OR=13.63, 95%CI=2.58-71.99), greater insight (OR=0.17, 95%CI=0.06-0.49), lower baseline Global Assessment of Functioning Scale and Scale of Occupational and Functional Assessment score (OR=0.96, 95%CI=0.62-0.91; OR=0.98, 95%CI=0.95-0.99), and longer time in treatment (OR=1.05, 95%CI=1.03-1.08). CONCLUSIONS: The prevalence of suicidal behaviour was high, indicating that suicidal behaviour in incidence populations is higher than in non-epidemiological cohorts of first-episode patients. The rate of repetition of suicide attempt among the sample, however, was lower than expected, suggesting that specialist services can play a role in reducing suicide risk.