Abstract —The aim of this study was to evaluate the predictive power of the frequency of initial caries lesions in selecting persons at high risk for caries. The subjects (n= 124) were 11–13 yr old at the beginning of the follow‐up. Caries was registered initially and after 5 yr. Evaluation of the predictions was based on cross‐tabulations of the subjects according to their predicted caries risk and the actual 5‐yr caries increment. Logistic regression analyses were used to produce the individual risk scores. In addition to the caries scores, age of the subjects, was included in the prediction function. In bivariate analyses the total initial caries score at baseline correlated better with caries increment than FS and DS scores did (r = 0.40, 0.13, and 0.31, respectively). When past caries experience was used as predictor, 69% of the children were classified correctly into the low and high risk groups. Surprisingly, only two more subjects (2%) were classified correctly when initial caries score was included in the function. These results suggest that inclusion of initial caries score adds little to the predictive power of the conventional DS and