It is widely accepted that a congener-specific analysis of polychlorinated biphenyls (PCBs), rather than traditional Aroclor equivalent total PCB analysis, is required for risk assessment. This is based on the fact that environmental processes alter the original distribution of PCB congeners in Aroclors and that toxicity varies considerably among the congeners with dioxin-like PCBs (dl-PCBs) generally being among the most toxic. Using the largest known dl-PCB fish dataset, here we present a likely composition of dl-PCBs in fish. In contrast to common perception, we found that the dl-PCB composition is relatively constant (within approximately a factor of 2) regardless of fish species and total PCB level. The abundance of dl-PCBs expressed as a percentage of total PCB (25-75 quartile range) in fish is generally in the order of PCB-118 (3.0-6.2%) > PCB-105 (1.1-2.4%) > PCB-156 (0.39-0.75%) > PCB-167 (0.20-0.43%) > PCB-123 (0.11-0.26%) > PCB-157 (0.09-0.19%) = PCB-114 (0.08-0.18%) > PCB-189 (0.045-0.094%) > PCB-77 (0.018-0.093%) > PCB-126 (0.015-0.036%) > PCB-81 (0.002-0.007%) = PCB-169 (0.001-0.006%). The most toxic dl-PCB congeners PCB-126 and -169 contribute on average only 0.027 and 0.004% of total PCB, respectively. The statistically significant relationships presented between individual di-PCB and total-PCB concentrations can be used as a practical tool to estimate dl-PCBs for risk assessment purposes. A comparison of the dl-PCB pattern presented here with other studies suggests that this dl-PCB composition is applicable to fish from North America and perhaps from other geographical regions throughout the world.