Polycyclic aromatic hydrocarbon (PAH) bioavailability to Hyalella azteca was determined in 97 sediments from six former manufactured-gas plants and two aluminum smelter sites. Measurements of Soxhlet extractable, rapidly released based on mild supercritical fluid extraction, and pore water dissolved concentrations of 18 parent and 16 groups of alkyl PAHs (PAH34) were used to predict 28 daysurvival based on equilibrium partitioning and hydrocarbon narcosis models. Total PAH concentrations had little relationship to toxicity. Amphipods survived in sediments with PAH34 concentrations as high as 2990 microg/g, while sediments as low as 2.4 microg/g of PAH34 resulted in significant mortality. Equilibrium partitioning using either total extractable or rapidly released concentrations significantly improved predictions. However, pore water PAH34 concentrations were best for predicting amphipod survival and correctly classified toxic and nontoxic sediment samples with an overall model efficiency of 90%. Alkyl PAHs accounted for 80% of the toxicity, demonstrating that careful measurement of the 16 alkyl clusters in pore water is required. Regression analysis of the pore water PAH34 data from 97 field sediments against amphipod survival resulted in a mean 50% lethal residue value of 33 micromol/g of lipid, consistent with 32 micromol/g of lipid for fluoranthene determined by others in controlled laboratory conditions, thus demonstrating the applicability of EPA's hydrocarbon narcosis model when using pore water PAH34 concentrations.