A self-administered questionnaire using the mark-sheet method (MSM), in which responses of subjects are computer processed directly through an optical scanning device, has recently been utilized in epidemiologic surveys. Compared to the data coding process for a conventional questionnaire, in which a keypuncher enters the responses manually into a computer (manual method; MM), optical scanning requires less time and cost. Accuracy of the MSM for use in the general population in Japan, however, remains uncertain. Therefore the response rates, frequencies of missing values, validity and reproducibility of the answers in self-administered questionnaires were compared between the MSM and MM. Subjects were 463 residents aged 40-69 years living in 6 local districts of a rural town in northeastern Japan. They were randomly allocated, by district basis, to the MSM group (n = 242) or the MM group (n = 221). The questionnaire was delivered and collected at the subject's home by volunteers. Two weeks after collecting the original questionnaire, the same type of questionnaire was again distributed to half of the responders randomly chosen to investigate reproducibility. The overall response rate did not differ in MSM and MM (96.7% vs 98.2%, p = 0.312). Among questions with a multiple-choice type of answer, proportions of missing values were not different for most of the items, but it was lower in MSM for all of the 33 food frequency items. Reproducibilities of food frequency items measured by Spearman's rank correlation did not differ substantially in two groups.(ABSTRACT TRUNCATED AT 250 WORDS)