This study reports on the determination of zygosity in Chinese adult twins by simple questionnaire and physical features comparison. The subjects were 511 twin pairs from two cities and their town areas, consisting of 371 monozygotic (MZ) and 140 same-sex dizygotic (DZ) pairs, identified by ABO blood group and multiplex polymerase chain reaction of several polymorphic short tandem repeat markers. The twins themselves responded to 8 questionnaire items, 4 items on twin similarity, and 4 items on the frequency of mistaking one twin for another by parents, relatives, teachers and strangers when they were 6 to 13 years old. Research assistants responded to 20 items regarding twins' physical features at the moment of interview. A parsimonious model established using stepwise logistic regression analysis of the 28 items showed that the total accuracy of zygosity diagnosis was 90.1%. The accuracy was 89.2% when using only the items dealing with the confusion of twins and 85.4% using only similarity. In the questionnaire, 'facial appearance', 'mistaken by teachers' and 'mistaken by strangers' had stronger discriminating power between MZ and DZ twins. Two physical features--'eyelid' and 'middigital hair'--were informative to some extent. There was no statistically significant sex and area difference in the validity of such questionnaire and physical features comparison-based classification. In conclusion, questionnaire-based zygosity assessment in this Chinese adult twin sample could still be regarded as a valid and valuable classification method. Physical features comparison, however, could only provide limited information for zygosity determination.