Data obtained in a longitudinal study of caries incidence and caries-related factors were analysed with a view to producing a model for the prediction of caries. In direct correlations, caries incidence was significantly associated with bacterial, dietary and salivary variables; but when the data were examined by stepwise regression the strongest variables were the baseline caries score and misuse of sugar. Counts of Streptococcus mutans entered into the analysis but only as a relatively minor component. Similar significant relationships were seen with the determination of odds ratios. A caries activity test was formulated combining the following caries-associated variables: high counts of S. mutans, or lactobacilli, or the misuse of sugar, or frequent consumption of paediatric medicines. Regular use of fluoride tablets could compensate for paediatric medicine use or misuse of sugar. Such a caries activity test if it had been applied to the children at baseline would have had a positive predictive value of 0.76, a negative predictive value of 0.82, a sensitivity of 0.8 and a specificity of 0.78. Combining tests made the prediction of caries more accurate and in the population for which it was intended gave a reliable means of detecting those children most in need of enhanced caries prevention.