In Japan, only a few rivers have been inspected for Cryptosporidium parvum contamination, and the methods used had low sensitivity. In 1998 and 1999, we used a method with higher sensitivity to examine all large rivers used as sources of water supply in one prefecture (which we divided into four areas) in western Japan for Cryptosporidium oocysts. One sample was collected at each of 156 sites along 18 rivers, and samples were tested for Cryptosporidium oocysts by immunomagnetic separation. Samples were classified as being obtained on an island with livestock and fishing industries, a densely populated urban area, a western region including farming villages, or a still more rural northern area with agriculture and fishing. Restriction fragment length polymorphism analysis was used for identification of the C. parvum found as the bovine or human type. C. parvum was detected in at least one sample from 13 of the 18 rivers and in 47% (74 of 156) of the samples. One-third to all of the samples from each area contained C. parvum oocysts. The number of C. parvum oocysts per 20 liters of river water varied in the same pattern as the number of cattle kept in the four kinds of areas (as determined by the Mantel extension test). Oocysts isolated were of the bovine type; the C. parvum detected in rivers probably came from cattle kept in that valley. As we had expected, when tested with a more sensitive method, river water in western Japan was found to be greatly contaminated with C. parvum oocysts, as reported in other countries.