The objectives of this study were to examine the reliability and validity of a 26-point telephone version of the Cantonese Mini-mental State Examination (T-CMMSE) for a sample of 65 elderly patients, comprising 31 patients without dementia and 34 patients with dementia, in an acute regional hospital in Hong Kong, and to identify an optimal cut-off score to discriminate between those patients with dementia and those without. Participants were rated by using the face-to-face Mini-mental State Examination (MMSE) before inpatient discharge and the T-CMMSE after inpatient discharge, and were rated separately by two raters in two telephone follow-up sessions using the T-CMMSE. The results of the study indicated that the scale had excellent inter-and intra-rater reliabilities. There was substantial agreement between the two versions of the examination (kappa > 0.6-0.8 < or =) for orientation, registration, and recall items. An optimal cut-off score of < or = 16 was suggested for the T-CMMSE to discriminate between those with and without dementia. The T-CMMSE can be used in telephone follow-ups as an alternative to the conventional face-to-face version.