Abstract The relationship between the age-specific mean number of DMF teeth and the age-specific mean number of DMF surfaces in the population has been studied using the original data of Knutson and epidemic/logic data from five Finnish materials, including data of 4718 subjects examined. Two models of regression between the variables were evaluated statistically by making a comparison between a linear and a curvilinear model of regression. For Knutson's original data, both models were suitable, apart from one data set with a relatively high caries prevalence, for which the association between the variables could be better expressed by the linear model. Concerning the recent Finnish materials, the difference in accuracy between the models seemed to be related to the range of the measurements; the larger the range of the measured caries prevalences, including low and high prevalence rates, the more preponderant the curvilinear model. For very low caries populations as well, the curvilinear model, a correction exponential curve, seemed to fit caries data somewhat better than the straight line.