Although change is a central goal of oral health care interventions, little attention has yet been paid to change in self-perceptions of oral health status. This is an important omission, given the current emphasis on assessing health outcomes. This paper reports the results of a study which examined changes over a period of three years in self-perceived oral health among 611 community-dwelling Canadians aged 50 years and over. Change in self-perceptions was measured by means of a global transition item and change scores derived from repeat administrations of four subjective oral health status indicators. Overall, 20.5% reported that their oral health had deteriorated over the three-year observation period, 68.5% that it had remained the same, and 10.5% that it had improved. There was a significant association between these global change categories and change scores for the four subjective indicators. Because of the small number of edentulous subjects, the analysis of baseline characteristics predicting change was confined to dentate subjects. Bivariate and logistic regression analyses were used to compare the two groups reporting change with those whose oral health status remained stable over the observation period. The results suggest that, when compared with this reference group, those who deteriorated and those who improved were similar in some respects but distinct in others. Those who improved appeared to have specific oral conditions at baseline causing pain. Those who deteriorated had poor oral health in general and came from more disadvantaged backgrounds. However, the explanatory power of logistic regression models predicting change in self-perceived oral health was poor when judged in terms of model sensitivities. This was to be expected, given that the models did not include variables documenting the incidence of disease, receipt of dental care, or changes in social and personal circumstances over the observation period.