The counts of mutans streptococci (MS) and lactobacilli (LB) in saliva have been used for prediction of dental caries. The results have not justified putting the methods into clinical practice as the explanatory values of the bacteria for caries are low. There is therefore reason to question how representative the methods for sampling and estimation of those bacteria are. The aims of this study were to investigate if the variation in caries is better explained by the bacteria, estimated as means of numbers in double saliva samples, than by numbers in single samples, and to determine whether chewing on either the preferred or the non-preferred side influences the bacterial counts. Twenty-five healthy subjects were included in the study. Two saliva samples were collected with a 1-wk interval by chewing on a piece of sterile paraffin wax on the preferred side only. One month later this was repeated on the opposite side. In order to test if there were variations in the numbers of bacteria due to the laboratory methods, a paired cultivation of 20 samples was carried out. The results showed that the coefficients of variation (c.v.) for the differences between the individual double samples varied between 110 and 276% of the total mean of the double samples. The paired cultivation gave r values for MS of 0.93 and for LB of 0.99. Using double samples did not increase the explanatory values for caries. There were higher explanatory values for caries on the preferred side than on the non-preferred side, this being much more pronounced for lactobacilli than for mutans streptococci.