Bryant and Smith [(2001). Refining the architecture of aggression: A measurement model for the Buss-Perry Aggression Questionnaire. Journal of Research in Personality, 35, 138-167] refined the 29-item Aggression Questionnaire and obtained a 12-item short form of the AQ that has yet to be validated with samples outside of North America and England. Confirmatory factor analysis (CFA) with two Asian adolescent samples revealed that while the hierarchical AQ model had an acceptable fit to the data, the multidimensional AQ model demonstrated a superior fit. The unidimensional AQ model was shown to be an unacceptable measurement model in both samples. Results obtained replicated previous findings using the original 29-item measure and supported the AQ's multidimensionality. Multigroup CFA analyses revealed partial measurement invariance for the 12-item fourfactor AQ model for both samples with respect to gender. Supplementary t-tests revealed a consistent gender difference on physical aggression in the male direction for both samples and a gender difference on verbal aggression in the male direction only for one sample. No other gender differences were found. These results are consistent with recent meta-analytic research.