United Chrome Products, Inc. (Corvallis, OR), discharged liquid chrome-plating wastes into an on-site dry well from 1957 to 1977. Samples were collected from ground and surface waters during February and December 1984 and evaluated for toxicity potential using 96-h Selenastrum capricornutum bioassays. Bioassays of the February samples were used to calculated predicted EC50 values for different chromium concentrations, which were then compared with the chromium concentrations measured in the December samples and with the results of S. capricornutum bioassays on these samples. An excellent correlation was obtained between the predicted EC50 values and the actual bioassay response values for ground water and drainage ditch samples. However, no correlation was obtained with the off-site surface water samples. These results could not be explained by changes in chromium valence, and they demonstrate that, even in simple systems, toxicity cannot readily be predicted based solely on chemical analysis.