The reproduction of white-tailed sea eagles was monitored in 1964–1999 in 3 differently contaminated sub-populations: Baltic Sea coast (Bp), inland central Sweden (Ip) and Lapland (Lp). 249 dead eggs from 205 clutches were obtained for analyses of DDE and PCBs and for eggshell measurements. A desiccation index (D_i) value was calculated for each egg as a measure of water loss through the shell. In the highly contaminated Bp, p,p´-DDE concentrations in the eggs decreased continuously and 5-fold during the study period and PCB concentrations decreased 3-fold from the mid 1980s. The PCB pattern changed slightly over time towards more high-chlorinated congeners but the relative toxicity of the PCB mixture, expressed as 2,3,7,8-tetrachlorodibenzo-p-dioxin equivalents (TEQ), remained constant and TEQ can be assumed to have decreased in a similar way as PCB over time. Productivity (P), shell thickness (S_t), shell index (S_i) and D_i increased over time in the Bp but no change in D_i or productivity occurred in the Lp, where residue concentrations were 5–8 times lower. P of the Bp was not correlated to S_t or S_i but was negatively correlated to D_i, DDE and PCB. An S-shaped dose-response relationship was indicated between P and DDE. After 1988, when the PCB/DDE ratio was considerably higher than previously, PCB but not DDE concentrations were significantly higher in eggs with dead embryos as compared to undeveloped eggs, implying lethal concentrations of PCB, and a LOEL of 320 pg g⁻¹ TEQ is suggested for embryo mortality. In a subset of 21 eggs, representing productive and unproductive females, analyzed for a selection of coplanar PCB congeners, tris(4-chlorophenyl) methanol and bis(4-chlorophenyl) sulphone, there was no evidence for a correlation between P and any of these compounds. A reduction in residue concentrations in old females did not lead to increased P or improved D_i-values, indicating a remaining effect from a previous, higher exposure to contaminants. The inability to reproduce included a high rate of undeveloped eggs, indicating effects at a prezygotic stage. P showed the strongest correlation with D_i, and D_i was most strongly correlated to DDE. Thus, the remaining effect of previous exposure resulted in a stronger correlation to the symptom (D_i) rather than to the suggested causative agent (DDE). LOEL values for depressed P were estimated at 120 µg g⁻¹ DDE and 500 µg g⁻¹ PCB (lipid basis). It is concluded that the major reason for depressed P during the study period was DDE, but that effects also from PCB were largely concealed by the effects from DDE.