Based on artificial neural network (ANN) and extracted characters of dead fish in polluted piscatorial waters, an identifiable model of toxicant type bringing on dead fish is presented in the paper. There are eighteen nerve cells in input-layer, nine nerve cells in middle-layer and one nerve cell in output-layer of the artificial neural network. Network parameters were optimized. Results show that any pollution to deaths can judge death causes by simple addition and multiplication operations using trained B - P artificial neural network, so this method has high applicability and practicability.