The Mini-Mental State Examination (MMSE) is a commonly used instrument for assessing mental impairment. Previous proposals for its underlying structure have focused on scores obtained from a single administration of the test. Because the MMSE is widely used in longitudinal studies, we examined the pattern of relations among the rates of chance of the items. Data were obtained from 63 subjects for 1.5 years or more. The relations among the rates of change of the MMSE items were described by a five-factor solution that accounted for 75% of the variance and comprised factors pertaining to orientation and concentration, obeying commands, learning and repetition, language, and recall. This was in contrast to the structure of the scores obtained from a single administration of the MMSE, which was best described by a two-factor solution. In order to provide a clinical validation, factor scores derived from the MMSE factors were used to predict scores on the Memory and Behavior Problems Checklist and the Brief Cognitive Rating Scale.