Quantitative structure-toxicity relationship (QSTR) plays an important role in toxicity prediction. With the modified method, the quantum chemistry parameters of 57 benzoic acid compounds were calculated with modified molecular connectivity index (MCI) using Visual Basic Program Software, and the QSTR of benzoic acid compounds in mice via oral LD50 (acute toxicity) was studied. A model was built to more accurately predict the toxicity of benzoic acid compounds in mice via oral LD50: 39 benzoic acid compounds were used as a training dataset for building the regression model and 18 others as a forecasting dataset to test the prediction ability of the model using SAS 9.0 Program Software. The model is LogLD(50) = 1.2399 x (0)J(A) + 2.6911 x (1)J(A) -0.4445 x J(B) (R-2 = 0.9860), where (0)J(A) is zero order connectivity index, (1)J(A) is the first order connectivity index and JB = (0)J(A) x (1)J(A) is the cross factor. The model was shown to have a good forecasting ability.