Bacterial source tracking (BST) has been used to identify the sources of fecal contamination in surface waters. Methods used for BST can be library based or nonlibrary based. For the library-based approach, it is important to assess the reliability of source assignments. In this study, we constructed an E. coli rep-PCR DNA fingerprint library and evaluated the reliability of the library for bacterial source tracking purposes. E. coli isolates were obtained from nine host sources, i.e., human, cattle, dog, horse, chicken, wild turkey, waterfowl, beaver and deer, in the Catoma Creek watershed of Alabama. BOX A1R primer was used to generate rep-PCR DNA fingerprints; data were analyzed using BioNumerics software. The decloned library consisted of 945 unique DNA fingerprints. Jackknife analysis showed that the average rate of correct classification was 71% for the library. Library accuracy was also tested using 51 challenge isolates from human, cattle, dog and waterfowl, which were not present in the known source library. ID Bootstrap analysis showed that 93%, 60%, 57%, and 33% of the isolates from human, cattle, dog and waterfowl, respectively, were correctly classified. Further assessment involved 20 E. coli isolates from animal species not present in the library. ID Bootstrap analysis indicates that 100% of cat, bird, and goat E. coli and 40% of hog E. coli were classified correctly. When conducting BST studies, it is important to understand the limitations of the methodology.