Polychlorinated biphenyl (PCB) congeners were analyzed in four deep, dated sediment cores from the Ashtabula River (OH, USA), for the purpose of identifying relevant PCB sources and congener patterns. The time span for three of the cores is from the mid 1960s to 1998, whereas the fourth has a time span of six years. The total PCB concentrations are in the range of 0.4 to 6.8 mug/g dry weight, with the highest concentrations observed in samples from the 1970s. A factor analysis (FA) model with nonnegative constraints was used to investigate the sources and patterns of PCBs. Additionally, a new model, based on a least squares method, was developed to identify possible patterns of anaerobic dechlorination of PCBs in the sediments, and to quantify the relevant dechlorination pathways. Both models were validated successfully either by artificially created data sets (FA model) or by using laboratory data from the literature (dechlorination model). The FA model revealed two significant sources. The first was identified as a slightly altered Aroclor 1248. The second did not resemble any Aroclor closely, but was very similar to the overall average congener profile of all samples. Simulation of anaerobic dechlorination on an Aroclor 1248 profile from the literature, according to dechlorination activities H/H', as defined in the literature, yielded a congener profile very similar to that of the second pattern. This indicates the likelihood of anaerobic dechlorination of PCBs in Ashtabula River sediments.