Current knowledge of the prevalence of psychiatric disorders in the general population is basically derived from highly standardized and fully structured interviews such as the Diagnostic Interview Schedule (DIS) and the Composite International Diagnostic Interview (CIDI), which were administered by psychiatric lay interviewers. The feasibility of the most recent German version of the CIDI, the Munich-CIDI (M-CIDI) was analysed with respect to unit and item non-response. We also analysed the effects of editing and interviewer training, constituting quality assurance procedures that are inherent with the CIDI. A random sample of 4075 adults in a northern German region were interviewed by 56 psychiatric lay interviewers, using the M-CIDI, enabling an assessment to be made of the most frequent DSM-IV diagnoses over the whole lifespan. Editing was carried out by psychologists and was combined with weekly interviewer training and individualized feedback. Results show a negligible rate of interview-induced unit non-response (0.1% only partially completed interviews) and item-non-response due to refusal, except for illicit drug use (3%). Considerable item non-response due to ‘don't know’ answers occurred in time-related questions. Errors detected by editing led to data correction in 37% of all interviews. Alcohol, nicotine, and somatoform disorder sections were mainly affected on the level of items and diagnostic decision. Occurrence of errors decreased with increasing interviewer experience and training in CIDI administration, but was not associated with participants' characteristics such as age and educational level. Complex time-related questions, allocation of physicians' diagnoses, and calculation of alcohol consumption index were most frequently subject to editing. We conclude that the application of editing may significantly influence the result of psychiatric epidemiological studies for some diagnostic categories. Findings are discussed on the background of future improvements of the CIDI. Copyright © 2000 Whurr Publishers Ltd.