The determination of predicted no-effect concentrations (PNECs) and sediment quality guidelines (SQGs) of toxic chemicals in marine sediment is extremely important in ecological risk assessment. However, current methods of deriving sediment PNECs or threshold effect levels (TELs) are primarily based on laboratory ecotoxicity bioassays that may not be ecologically and environmentally relevant. This study explores the possibility of utilizing field data of benthic communities and contaminant loadings concurrently measured in sediment samples collected from the Norwegian continental shelf to derive SQGs. This unique dataset contains abundance data for ca. 2200 benthic species measured at over 4200 sampling stations, along with co-occurring concentration data for >25 chemical species. Using barium, cadmium, and total polycyclic aromatic hydrocarbons (PAHs) as examples, this paper describes a novel approach that makes use of the above data set for constructing field-based species sensitivity distributions (f-SSDs). Field-based SQGs are then derived based on the f-SSDs and HCx values [hazardous concentration for x% of species or the (100 − x)% protection level] by the nonparametric bootstrap method. Our results for Cd and total PAHs indicate that there are some discrepancies between the SQGs currently in use in various countries and our field-data-derived SQGs. The field-data-derived criteria appear to be more environmentally relevant and realistic. Here, we suggest that the f-SSDs can be directly used as benchmarks for probabilistic risk assessment, while the field-data-derived SQGs can be used as site-specific guidelines or integrated into current SQGs. Copyright © 2005 American Chemical Society.