A fragment constant method for prediction of toxicity (LC50) to rainbow trout was developed based on the experimental LC50 values of 258 chemicals obtained from the literature. The dataset was randomly divided into a training set and a validation set for purposes of model development and validation. The final model was established using all of the experimental LC50 values by pooling the two sets together. The coefficient of the determination for the final model was 0.9495 with a mean residual of 0.42 log-units. The model's robustness was tested using jackknife tests.