It has often been assumed that all subjects with schizophrenia will eventually be admitted to hospital and therefore little bias is introduced by restricting research to hospitalized subjects. Using the Lothian Psychiatric Case Register, 66 subjects were identified who had been diagnosed in Edinburgh as suffering from schizophrenia between 1978 and 1989 but had no history of hospital admission by December 1991. This represented an adjusted average of 6.7% of the estimated annual rate of first diagnosis of schizophrenia: the proportion of such patients did not change over the period. Using a case-control design, the index cases were compared with a control group of schizophrenic patients who had been admitted to hospital within 3 months of diagnosis. At the time of diagnosis, the cases were generally less disturbed with lower levels of violent behaviour and less evidence of neglect or hallucinations. They had a longer duration of illness prior to diagnosis, were more often diagnosed by a consultant and unemployed. In a follow-up study of the index cases, 59 (89%) were traced, of whom 6 (10%) were deceased. The outcome of the illness was heterogeneous although the course was often chronic. The general practitioner provided most of the care they received. The small proportion of such patients suggests that their exclusion from most published series does not seriously bias our picture of the natural history of schizophrenia. Moreover, as there was no increase in the proportion over the period, first admission rates for schizophrenia in Scotland are a reasonable approximation to incidence rates.