AIMS: To (1) establish the clinical usefulness of the 10-item Abbreviated Mental Test (AMT) and the 18-item Chinese Mini-Mental Status Examination (CMMSE) for detecting cognitive impairment associated with dementia in the elderly Chinese; (2) determine how the tests' optimal cut-off scores varied with the patients' educational level and age; and (3) evaluate which was the more accurate test. METHODS: 151 cognitively-healthy, community dwelling elderly Chinese subjects and 95 elderly Chinese outpatients with dementia were administered the AMT and CMMSE. Receiver-Operating Characteristic (ROC) analysis was used to determine the tests' optimal cut-off scores for each of the education-by-age subgroups and their areas-under-the-curve were compared non-parametrically to evaluate which test was more accurate. RESULTS: Both the AMT and CMMSE could identify cognitive impairment accurately, but higher cut-off values were necessary for the younger and more educated cohort, while lower values were adequate for the older and less educated subgroup. The AMT appeared to reach a ceiling effect in the more educated categories. The diagnostic accuracies of the two instruments were statistically equivalent; there was a trend, however, for the CMMSE to be performing better in the more educated subgroups. CONCLUSIONS: To maximise the diagnostic efficiency of these two clinically useful mental status tests, it is important to adjust their cut-off scores for the patients' education and age. Though no clear superiority of either instrument was established in this study, we recommend the AMT for patients with 0-6 years of education, whereas for those with greater levels of literacy, we think it better to administer the CMMSE.