Hong Kong has an assessment for learning policy and a cultural context that emphasizes examinations. In addition to associating student grading with improvement, important improvement-oriented conceptions have been identified among Hong Kong teachers and which were not fully instantiated in the original Teachers’ Conceptions of Assessment (TCoA) inventory. An expanded Chinese-TCoA inventory was administered to in Chinese in both Hong Kong and southern China. The intended 6 construct structure was not supported. Exploratory factor analysis (MLE, oblimin rotation) identified 7 factors, which were further reduced to 3 major inter-correlated constructs (i.e., improvement, accountability, and irrelevance). Improvement and accountability were strongly inter-correlated (r=.80), while improvement had a weak negative correlation with irrelevance and accountability had a weak positive correlation with irrelevance. The model had 7 factors based on 33 items and acceptable fit (χ2=3856.97, df=426; χ2/df=9.05, p<.001; CFI=.85; RMSEA=.065; SRMR=.065; gamma=.90). Two group nested invariance testing (i.e., Hong Kong vs. southern China) showed that the model was statistically equivalent, except for item residuals (k=62; χ2=4612.89, df=891; χ2/df=5.18, p=.02; CFI=.83; RMSEA=.047; SRMR=.065; gamma=.94). Differences in mean scores between the two groups showed that Hong Kong teachers agreed more with the ideas of assessments helping student learning, being accurate, and being examinations; whereas southern China teachers agreed more that assessment was irrelevant. This study contributes to our understanding of how assessment is understood by teachers working within Chinese contexts.[Copyright © 2010 The Authors.]