The estimation capacities of two optimization methodologies, response surface methodology (RSM) and artificial neural network (ANN) were evaluated for prediction of biosorptive remediation of As(III) and As(V) species in batch as well as column mode. The independent parameters (viz. pH, initial arsenic concentration, temperature, and biomass dose in the case of batch mode and bed height, flow rate, and initial arsenic concentration in the case of column mode) were fed as input to the central composite design (CCD) of RSM and the ANN techniques, and the output was the uptake capacity of the sorbent. The CCD was used to evaluate the simple and combined effects of the independent parameters and to derive a second-order regression equation for predicting optimization of the process. The sets of input-output patterns were also used to train the multilayer feed-forward networks employing the backpropagation algorithm with MATLAB. The application of the RSM and ANN techniques to the available experimental data showed that ANN outperforms RSM indicating the superiority of a properly trained ANN over RSM in capturing the nonlinear behavior of the system and the simultaneous prediction of the output. © 2011 American Chemical Society.