The main aim of this study was to develop and optimize a questionnaire-based zygosity determination method in Chinese adolescent twins. Participants were 471 pairs of same-sex twins (345 monozygotic, 126 dizygotic) with a mean age of 14.56 years (SD = 2.62). A second sample was recruited for cross-validation, including 382 pairs of same-sex twins (261 monozygotic, 121 dizygotic) with a mean age of 12.53 years (SD = 2.22). The questionnaire consisted of 12 questions dealing with co-twin similarity or frequency of confusion. Two means were put forward to improve the predictive accuracy of the questionnaire — adding parent-reports to the analysis, and using a 2-point rather than 3-point response format. DNA genotyping was performed on nine short tandem repeat loci, with an estimated zygosity classification accuracy very close to 100%. The validity of all questionnaires was assessed by being compared to the results of DNA analysis. Results of stepwise logistic regression analysis showed that the predictive accuracy of the 3-point self-reported questionnaire is 83.8%. Using parent-reports and 2-point scale led to 3.9% and 4.6% increase in predictive accuracy, respectively. When using the parent-reports and children's self-reports jointly, the predictive accuracy was enhanced to 90.6%. For the cross-validation, the equations and cut-offs derived from the first sample led to an acceptable accuracy (91.3%) in the second sample. In conclusion, the method we developed can be used in future studies among Chinese adolescent twins. Multiple-rater and 2-point response format were suggested for all twin studies for enhancing the predictive accuracy of questionnaires.