Previous studies have demonstrated a high miscarriage risk among women in laboratories and women exposed to anaesthetic gases. Many of these studies have methodological shortcomings. The aim of the investigation was to study the pregnancy outcome, mainly the miscarriage rate in these occupational groups, and to study such potential sources of error as response bias and inaccurate diagnosis. A questionnaire study was performed among women who had worked at a Swedish hospital as well as women employed in laboratories at the University of Gothenburg (U.G.). Reported pregnancies were accepted only if they could be verified in hospital records or if they were anamnestically reported in medical records. Pregnancies among non-respondents were also sought in medical records. All women suffering a miscarriage after exposure to anaesthetic gases reported their miscarriage while one-third of all miscarriages occurring to women who were not exposed during pregnancy were not reported in the questionnaire. Twelve per cent of the miscarriages among laboratory employees at U.G. could not be verified or found anamnestically in medical records. Work with solvents during pregnancy was more frequently reported for these miscarriages as compared to those which were found in medical records. No significantly increased miscarriage risk after exposure to solvents in laboratory work or to anaesthetic gases could be demonstrated after correction for these errors. It is concluded that use of questionnaires as the only source of information in studies of pregnancy outcome could give biased results and should thus be avoided.