AIM The study aims to examine whether an extended Theory of Planned Behaviour (TPB) model predicts the frequency of past and intended cannabis use in young adults with psychosis. METHODS A cross-sectional correlational design was used with 32 participants (aged 18-36 years old) within 3 years (mean = 7 months) of their first psychotic episode. TPB variables and a measure of social influence (descriptive norms) were measured. RESULTS The TPB accounted for a large amount of the variance in both the frequency of past cannabis use (25%) and in the intended frequency of future cannabis use (36%). The addition of descriptive norms (beliefs about the frequency of others' cannabis use) accounted for a significant additional amount of the variance in past use (19%). Adding the 'friends' component of descriptive norms only led to a model that accounted for 52% of the variance in past cannabis use. CONCLUSIONS Descriptive social norms should be considered when trying to understand cannabis use in early psychosis. Interventions should take account of the individual's attitude towards cannabis, their perceived control over its use and its social context.