Background: We aimed to examine the association between illicit substance use and age at onset in psychotic disorders in an Australian cohort. Methods: Retrospectively acquired information on substance use during the year prior to illness onset was collected from 1642 participants enrolled in the Australian National 2010 Survey of High Impact Psychosis study (SHIP), with an ICD-10 diagnosis of schizophrenia spectrum or affective psychosis. Latent class analysis was performed according to illicit substance use, using age as an active covariate; identified classes were subsequently validated. Cox regression was used to examine the independent contribution of the identified substance use classes and several confounding variables to the prediction of age at onset of psychosis. Results: Three classes according to substance use were identified: non-users (n= 803), cannabis predominant users (n= 582), and polysubstance users (n= 257). For participants with schizophrenia spectrum disorders, cannabis predominant users had a higher hazard of earlier age at onset than for non-users (adjusted HR = 1.38, 95% CI = 1.2-1.6); polysubstance users had an even higher hazard (adjusted HR = 1.95, 95% CI = 1.5-2.4). In contrast, for participants with affective psychosis, cannabis predominant users (adjusted HR = 1.10, 95% CI = 0.8-1.4) and polysubstance users (adjusted HR = 0.87, 95% CI = 0.6-1.3) did not have a higher hazard of earlier age at onset compared with non-users. Conclusions: Illicit substance use in the 12. months prior to psychosis onset has a differential effect on age at onset in schizophrenia spectrum and affective psychotic disorders. Our findings are compatible with the notion that illicit drugs bring forward age at onset in schizophrenia spectrum disorders but not affective psychotic disorders. © 2014.