Our work assessed the accuracy of the original zygosity classification in the Vietnam Era Twin (VET) Registry using new information from DNA markers on a subset of participants. We then constructed an updated zygosity classification algorithm. The VET Registry includes 7,375 male–male twin pairs who served in the military during the Vietnam era. During the mid-1980s 4,774 twin pairs completed a zygosity questionnaire of 20 items. Additionally, military record information, including blood group, was available. Items from the zygosity questionnaire and blood group were used in the original zygosity classification. Between 1990–2009 DNA was obtained from 612 twin pairs and concordance between co-twins was used to classify zygosity. Next logistic regression was used to construct predicted probabilities of zygosity using items from the zygosity questionnaire with this subsample. All twins were reclassified according to the new zygosity prediction model and compared with the original zygosity assignment. The original and new predicted probabilities of zygosity were highly correlated (r = 0.962) and concordance for the classification of zygosity was similarly high (kappa = 0.936). Errors in the original zygosity assignment were primarily due to monozygotic twins that were misclassified as dizygotic based on military record blood group data. Removing the military record blood group data markedly improved the accuracy of the original classification. Zygosity assignment based on a zygosity questionnaire was highly predictive of DNA-based zygosity. Augmentation of such a zygosity classification from administrative data, military records, or other records, should be done with caution.