Fifty-five pharmaceuticals were monitored at four rivers and inlets and/or outlets of three sewage treatment plants (STPs) in Yodo River watershed, Japan over 17 sampling events. Twenty-six quantified pharmaceuticals were classified by source and fate. The load per person (LPP) of nine pharmaceuticals, including six with observed mass balance in studied river stretch of <80%, was appreciably lower in river water (RW) than in the effluent (EF) of STPs (RW/EF <0.5), indicating that they were susceptible to in-stream attenuation in the study area, while the others were relatively conservative. The LPP of 12 pharmaceuticals in RW were within ±50% of that in EF. Because their mass loadings in rivers were correlated with human population in the catchment and most people use the sewer system, the major source of the 12 pharmaceuticals was considered to be STPs. The LPP of the three most labile pharmaceuticals in STPs (caffeine, theophylline, and acetaminophen) was >1.5 in RW/EF and <1.0 in RW/influent (IF) of STPs. Poorly treated sewage discharged from households without using the sewer system was considered to be influential source of the three pharmaceuticals. The LPP (RW/EF) of caffeine, a pharmaceutical contained in food and beverage, was considerably higher than that of the other two, and this is attributable to untreated gray water discharged at households using the night-soil treatment system. The LPP of two veterinary drugs (sulfamonomethoxine and lincomycin) were >1.5 (RW/EF) and >1.0 (RW/IF). Their mass loadings in rivers showed a positive correlation with swine population in the catchment, although sulfamonomethoxine is equally used in both cattle and swine farming. This was attributable to application of cattle excrement as manure, and lability of sulfamonomethoxine during composting processes. The major source of the two veterinary drugs was considered to be on-site treatment facilities of swine urine.