Soil and sediment samples from oil gas (OG) and coal gas (CG) manufactured gas plant (MGP) sites were selected to represent a range of PAH concentrations (150-40,000 mg/kg) and sample matrix compositions. Samples varied from vegetated soils to lampblack soot and had carbon contents from 3 to 87 wt %. SFE desorption (120 min) and water/XAD2 desorption (120 days) curves were determined and fit with a simple two-site model to determine the rapid-released fraction (F) for PAHs ranging from naphthalene to benzo[ghi]perylene. F values varied greatly among the samples, from ca. 10% to >90% for the two- and three-ring PAHs and from <1% to ca. 50% for the five- and six-ring PAHs. Release rates did not correlate with sample matrix characteristics including PAH concentrations, elemental composition (C, H, N, S), or "hard" and "softs" organic carbon, indicating that PAH release cannot easily be estimated on the basis of sample matrix composition. Fvalues for CG site samples obtained with SFE and water desorption agreed well (linear correlation coefficient, r2 = 0.87, slope = 0.93), but SFE yielded higher F values for the OG samples. These behaviors were attributed to the stronger ability of carbon dioxide than water to desorb PAHs from the highly aromatic (hard) carbon of the OG matrixes, while carbon dioxide and water showed similar abilities to desorb PAHs from the more polar (soft) carbon of the CG samples. The combined SFE and water desorption approaches should improve the understanding of PAH sequestration and release from contaminated soils and sediments and provide the basis for subsequent studies using the same samples to compare PAH release with PAH availability to earthworms.