Present addresses of 12,752 like-sexed twin pairs born in the period 1915-1960 were identified. A questionnaire, concerning the similarity of pair members, was sent to all individuals. Responses were obtained from 83.7% of the subjects. The zygosity of 207 pairs was established by examination of genetic markers. By using discriminant analysis on the responses from this subgroup, functions were obtained for prediction of zygosity from questionnaire data. It was estimated that 2.4% of the pairs would be misclassified if the questionnaire responses from both pair members were used, and 3.9% if only the response from one of the twins was used. Accordingly, zygosity could be predicted with satisfactory reliability also for twin pairs where only one of the twins had responded. The predicted percentage of monozygotic (MZ) pairs among pairs where one or both twins had responded, was 39.4 (4,402/11,175). The percentage of MZ pairs was significantly lower (34.5) in death-discordant pairs than in pairs in which both twins were alive (39.6). The zygosity questionnaire data are sufficient to adequately score twin pairs for zygosity in the great majority of cases.