Most previous studies investigating the factor structure of psychosis have focussed on chronic samples. First episode samples with longitudinal follow up are few. To investigate the stability and validity of symptom factors, a sample of 257 patients with DSM IV nonaffective psychoses were assessed using the PANSS during the acute first episode and at 3- and 18-month follow up. Exploratory factor analysis of the changes in PANSS item scores over time gave a five-factor solution. This was consistent with the solutions to factor analyses at the initial assessment and each of the follow-ups. However, there was progression over follow-up. Confirmatory factor analysis demonstrated that symptom ratings at 18-month follow-up fitted the models from existing research, in relatively chronic samples, better than the ratings at the initial assessment. A psychomotor poverty factor showed most stability over time and a positive symptom factor most change. Factors showed different associations with demographic and external variables, further supporting their validity.