The prediction of caries risk has been of longstanding interest. Generally, few of the tests involving oral bacteria or their products have become accepted. Presently, the main focus is on counts of lactobacilli (L) and mutans streptococci (MS). Due to their positive numerical association with human caries and the linkage of this association to carbohydrate consumption, counts of L and MS may, potentially, serve not only as a caries risk predictor but also as an indicator of carbohydrate consumption, another caries-risk factor. The value of counts of L and MS as caries-risk predictors has been evaluated by means of studies providing data on test sensitivity, specificity, and predictive values. These and other studies indicate that their use for the prediction of caries risk of individuals is not possible but is more promising for that of the caries risk of groups (e.g., identification of high-caries-risk subjects); further, the prediction of low caries risk may be more reliable than that of high caries risk. The influence of test variables on the test results has been discussed. These include the level of caries increment, subject age, methods of caries evaluation, use of saliva or dental plaque as test sample, sampling frequency, type of bacterial growth medium, and the use of simplified methods rather than conventional laboratory procedures for microbial enumeration. An approach to optimize the use of microbiological caries-risk predictors in different populations as well as their use in conjunction with other caries-risk predictors has been discussed. The latter include the incipient caries lesion or past caries experience and salivary buffering capacity and flow rate. Due to the multifactorial nature of caries etiology, it is expected that multivariate approaches rather than the use of single parameters may improve caries risk prediction for individuals as well as groups of subjects.