Attachment of E. coli to soil particles affects bacterial transport in overland flow and through soil. The objective of this research was to investigate the existence of strain-dependent variations in attachment of manure-borne E. coli to soil particles of different sizes using rep PCR techniques. The Tyler clay loam soil was fractionated to obtain particles of coarse sand, medium sand, fine sand, silt, and clay sizes. The aged manure was cultured with the goal of producing a uniform inoculum for attachment studies without selecting for any particular E. coli strains. Serial dilutions of the fecal coliform suspension were mixed with one gram of soil in 10 mL of water to obtain fecal coliform concentrations of 10² and 10³ CFU mL ^-1. Fifty E. coli isolates for each soil fraction supernatant and 50 isolates representing the E. coli suspension without soil were utilized for the extraction of DNA to be used for polymerase chain reaction (PCR) using ERIC primers. The DNA fingerprint analysis was done with pseudo-gel images from electrophoregrams using BioNumerics. The DNA concentrations in the supernatants were on average about 90 ng L^-1 and did not exhibit a significant dependence on soil particle size. The cluster analysis led to identification of five clusters with the similarity level within them higher than 80% and 17 clusters with the similarity level within them higher than 90%. The chi-square test was applied to test the hypothesis that the strain distribution among the clusters does not depend on adsorbent. This hypothesis could be rejected at 0.0001 probability level. The preferential attachment of different strains to particles of different size classes may be attributable to differences in particle surface properties. It remains to be seen whether the attachment to mineral surfaces for pathogenic E. coli strains may be different from that for non-pathogenic strains.