We describe a new PCR-based method for distinguishing human and cow fecal contamination in coastal waters without culturing indicator organisms, and we show that the method can be used to track bacterial marker sequences in complex environments. We identified two human-specific genetic markers and five cow-specific genetic markers in fecal samples by amplifying 16S ribosomal DNA (rDNA) fragments from members of the genus Bifidobacterium and the Bacteroides-Prevotella group and performing length heterogeneity PCR and terminal restriction fragment length polymorphism analyses. Host-specific patterns suggested that there are species composition differences in the Bifidobacterium and Bacteroides-Prevotella populations of human and cow feces. The patterns were highly reproducible among different hosts belonging to the same species. Additionally, all host-specific genetic markers were detected in water samples collected from areas frequently contaminated with fecal pollution. Ease of detection and longer survival in water made Bacteroides-Prevotella indicators better than Bifidobacterium indicators. Fecal 16S rDNA sequences corresponding to our Bacteroides-Prevotella markers comprised closely related gene clusters, none of which exactly matched previously published Bacteroides or Prevotella sequences. Our method detected host-specific markers in water at pollutant concentrations of 2.8 x 10(-5) to 2.8 x 10(-7) g (dry weight) of feces/liter and 6.8 x 10(-7) g (dry weight) of sewage/liter. Although our aim was to identify nonpoint sources of fecal contamination, the method described here should be widely applicable for monitoring spatial and temporal fluctuations in specific bacterial groups in natural environments.