Dioxins/furans are ubiquitous environmental contaminants whose primary route of human exposure occurs via the consumption of fatty foods of animal origin. The US FDA conducted a market basket survey of dairy products and commercial fish and shellfish to obtain data on levels of 17 dioxin/furan congeners (2, 3, 7, 8-congeners) in the US. The dairy products sampled included various cheeses (American, cheddar, Swiss, cottage), ice cream, yogurt, butter, and milk. The finfish and shellfish (molluscs and crustacea) sampled are those marine species consumed in the greatest amounts and include canned tuna, shrimp, cod, blue crab, and oysters. Catfish was sampled because it is the dominant aquaculture species. Samples were collected in 1995/96 and analysis for 17 dioxin/furan congeners was performed by high-resolution gas chromatography following extraction and clean-up. Limits of detection (LOD) and quantitation (LOQ) for each congener in each food were reported. Point estimates of exposure were calculated using a 3-day (1-day diary plus 2-day recall) food consumption survey for eaters-only and for the general population (USDA/CSFII, 1989-92). Toxicity equivalency factors (TEFs) developed by the World Health Organization (1997) were used to derive overall dioxin/furan toxicity equivalents (TEQ) for each sample food. Mean estimates of TEQ exposure for each food were derived using five values for non-detects (ND = 0; ND = 1/2 LOD or LOQ, ND = LOD or LOQ) on both a total sample and eaters-only basis. Using zero and the LOD provide lower and upper bounds on the range of estimated exposure, respectively. The bounds on mean dioxin intakes (pg/person/day) calculated for consumers of specific foods were estimated as follows (using zero or LOD for non-detects): butter (0.5-11), cheese (1.6-3.2), ice cream (4-19), yogurt (0.8-28), catfish (148-150), fish (other than catfish) (0.03-9), crustacea (32-35), mollusks (16.1-16.6), and shrimp (0.09-4.5). Exposure estimates derived by the five ND-methods are strongly dependent on the LOD and LOQ and represent upper bound estimates of exposure. Uncertainty in the exposure estimates is reduced with refinements in the analytical method.