Measurements of elevated concentrations of polychlorinated biphenyls (PCBs) in fish and in streambed sediments of the Millers River Basin, Massachusetts and New Hampshire, have been reported without evidence of the PCB source. In 1999, an investigation was initiated to determine the source(s) of the elevated PCB concentrations observed in fish and to establish the extent of fish exposure to PCBs along the entire main stems of the Millers River and one of its tributaries, the Otter River. Passive samplers deployed for 2-week intervals in the water-column at 3 1 stations, during summer and fall 1999, were used to assess PCB concentrations in the Millers River Basin. The samplers concentrate PCBs, which diffuse from the water column through a polyethylene membrane to hexane (0.200 liters) contained inside the samplers. Only dissolved PCBs (likely equivalent to the bioavailable fraction) are subject to diffusion through the membrane. The summed concentrations of all targeted PCB congeners (summed PCB) retrieved from the samplers ranged from 1 to 8,000 nanograms per hexane sample. Concentration and congener-pattern comparisons indicated that the historical release of PCBs in the Millers River Basin likely occurred on the Otter River at the upstream margin of Baldwinville, Mass. Elevated water-column concentrations measured in a wetland reach on the Otter River downstream from Baldwinville were compatible with a conceptual model for a present-day (1999) source in streambed sediments, to which the PCBs partitioned after their original introduction into the Otter River and from which PCBs are released to the water now that the original discharge has ceased or greatly decreased. Two four-fold decreases in summed PCB concentrations in the Millers River, by comparison with the highest concentration on the Otter River, likely were caused by (1) dilution with water from the relatively uncontaminated upstream Millers River and (2) volatilization of PCBs from the Millers River in steep-gradient reaches. A relatively constant concentration of summed PCBs in the reach of the Millers River from river mile 20 to river mile 10 was likely a consequence of a balance between decreased volatilization rates in that relatively low-gradient reach and resupply of PCBs to the water column from contaminated streambed sediments. A second high-gradient reach from river mile 10 to the confluence of the Millers River with the Connecticut River also was associated with a decrease in concentration of water-column summed PCBs. Volatilization as a loss mechanism was supported by evidence in the form of slight changes of the congener pattern in the reaches where decreases occurred. Exposure of fish food webs to concentrations of dissolved PCBs exceeded the U.S. Environmental Protection Agency's water-quality criterion for PCBs throughout most of the Millers River and Otter River main stems. Because the apparent source of PCBs discharged was upstream on the Otter River, a large number of river miles downstream (more than 30 mi) had summer water-column PCB concentrations that would likely lead to high concentrations of PCBs in fish.