Psychiatric disorders in general and schizophrenia in particular often begin with a prodromal phase progressing into more specific syndromes long before the first treatment contact. More longitudinal studies on course and outcome of schizophrenia begin with first hospital admission not taking into account the preceding early course. The reason for this is the unsurmountable difficulties in observing and assessing real onsets with unspecific symptomatology directly and collecting a sufficient number of 'precontact' cases of rare diseases in the general population. There are in principle two practical ways of assessing time and type of onset and the early course of schizophrenia: the application of a screening procedure for establishing a prospective approach, e.g. the study of individuals at high risk for schizophrenia, and the retrospective assessment of the preceding course carried out at first treatment contact. Methodologically both strategies share the same problems, especially recall bias. In a comprehensive discussion of methodological issues ways are described to reduce and even control recall deficits by using appropriate techniques (standardised assessment with the IRAOS, the 'Interview for the Retrospective Assessment of the Onset of Schizophrenia'; time grids; interval assessment, comparison of independent sources of information etc.). A representative sample of 232 patients with a first episode of schizophrenia were selected from the 276 first admissions of the ABC schizophrenia study taken from a German population of 1.5 million. Results are represented for the crucial questions: when, how and with what symptoms does schizophrenia begin and how does the disorder develop up to the first admission? For example, the gender-specific mean age at different points in time during the development of the disease is presented, and the sequence and cumulation of symptoms in the early course examined. Additionally, the consecutive fulfillment of DSM-III criteria sets for schizophrenic and schizophreniform disorders are presented along with the frequency of different types of onset in schizophrenia.