Different methods reported for assessing mutans streptococci (MS) make the direct comparisons of results across studies difficult. To quantitate the variations of MS estimates attributable to differences in method, stimulated and unstimulated saliva samples and oral swab samples were compared with pooled dental-plaque samples. Detection of MS in stimulated saliva samples was in excellent agreement with the presence of MS in pooled plaque samples. MS detection in unstimulated saliva samples, however, was significantly discordant with that in either pooled plaque or oral swab samples. When caries status was used as the criterion of validity of MS estimates, stimulated saliva samples demonstrated a sensitivity (94%) similar to that of pooled plaque samples, but exhibited a lower specificity (11%) than that of pooled plaque samples (17%). As a result, the measure of association between MS and caries was biased (towards null) when MS status was based on stimulated saliva samples. Interestingly though, in enumerating MS, stimulated saliva samples yielded significantly higher levels of MS (about 1.5 log10 increase) with a lower variability compared to unstimulated saliva samples. The use of different culture media for detection of MS gave different results as well. MS detection was poor (κ =0.31) and MS levels were lower (p =0.0001) when samples were grown on glucose-sucrose-potassium tellurite-bacitracin agar compared to mitis-salivarius-bacitracin agar. Together with the relative ease of sampling and processing, these findings collectively justify the use of plaque samples for the qualitative assessment of MS and stimulated saliva samples for the quantitative assessment of MS, while providing a basis for adjustment of estimates when comparing results across studies. © 1995.