Matching synoptically collected chemical and laboratory bioassay data (n = 1,068) were compiled from analyses of surficial sediment samples collected during 1990 to 1993 to evaluate the predictive ability of sediment quality guidelines (SQGs), specifically, effects range—low (ERL), effects range—median (ERM), threshold effects level (TEL), and probable effects level (PEL) values. Data were acquired from surveys of sediment quality performed in estuaries along the Atlantic, Pacific, and Gulf of Mexico coasts. Samples were classified as either nontoxic (p > 0.05 relative to controls), marginally toxic (p < 0.05 only), or highly toxic (p < 0.05 and response greater than minimum significant difference relative to controls). This analysis indicated that, when not exceeded, the ERLs and TELs were highly predictive of nontoxicity. The percentages of samples that were highly toxic generally increased with increasing numbers of guidelines (particularly the ERMs and PELs) that were exceeded. Also, the incidence of toxicity increased with increases in concentrations of mixtures of chemicals normalized to (divided by) the SQGs. The ERMs and PELs indicated high predictive ability in samples in which many substances exceeded these concentrations. Suggestions are provided on the uses of these estimates of the predictive ability of sediment guidelines.