The files of a narrowly defined group of schizophrenic patients, admitted to hospital for the first time in their lives in 1963, and who were representative of cases from a large un-selected population, were examined for Schneider's first-rank symptoms. Forty-three such patients were followed-up in 1977. Using stepwise regression, it was found that Schneider's first-rank symptoms relate to outcome such that they can ‘explain’ from 17 per cent to 26 per cent of the variance on four outcome measures. The presence of some first-rank symptoms correlated positively with good outcome, others negatively, whilst some correlated differently with different outcome measures. Thus, assessment of only the presence, absence or number of first-rank symptoms is unlikely to relate to outcome.