A random sample of pregnant women in the cities of Odense and Aalborg, Denmark, was taken as a part of a quasi-experimental study. The time period of the study was from October 1984 to March 1985. Out of 175 randomly selected pregnant women, 146 filled out the questionnaire in the 36th week of their pregnancy. The questionnaire was concerned with eating, smoking, and drinking habits before and during their pregnancy. These 146 pregnant women participated in an interview 3 weeks later where some of the same questions on the questionnaire were asked by the interviewer. In spite of the rather short interval between the two data collections, there were some differences in the answers given at the individual level. This is especially true for eating habits and estimated average drinking habits during pregnancy. No tendency was observed for a more frequent recalling of "less socially acceptable norms" in the questionnaire method of obtaining information. Lack of knowledge of the "true" classification is probably the main reason for the lack of stability in the answers given. This leads to some unavoidable misclassification in studies on self-reporting life style habits among pregnant women.