Chlorinated hydrocarbon concentrations in eggs of fish-eating birds from contaminated environments such as the Great Lakes of North America tend to be highly intercorrelated, making it difficult to elucidate mechanisms causing reproductive impairment, and to ascribe cause to specific chemicals. An information- theoretic approach was used on data from 197 salvaged bald eagle (Haliaeetus leucocephalus) eggs (159 clutches) that failed to hatch in Michigan and Ohio, USA (1986–2000). Contaminant levels declined over time while eggshell thickness increased, and by 2000 was at pre-1946 levels. The number of occupied territories and productivity increased during 1981 to 2004. For both the entire dataset and a subset of nests along the Great Lakes shoreline, polychlorinated biphenyls (ΣPCBs, fresh wet wt) were generally included in the most parsimonious models (lowest-Akaike's information criterion [AICs]) describing productivity, with significant declines in productivity observed above 26 µg/g ΣPCBs (fresh wet wt). Of 73 eggs with a visible embryo, eight (11%) were abnormal, including three with skewed bills, but they were not associated with known teratogens, including ΣPCBs. Eggs with visible embryos had greater concentrations of all measured contaminants than eggs without visible embryos; the most parsimonious models describing the presence of visible embryos incorporated dieldrin equivalents and dichlorodiphenyldichloroethylene (DDE). There were significant negative correlations between eggshell thickness and all contaminants, with ΣPCBs included in the most parsimonious models. There were, however, no relationships between productivity and eggshell thickness or Ratcliffe's index. The ΣPCBs and DDE were negatively associated with nest success of bald eagles in the Great Lakes watersheds, but the mechanism does not appear to be via shell quality effects, at least at current contaminant levels, while it is not clear what other mechanisms were involved.