Genetic algorithm optimization of pump-and-treat ground water remediation is used to explore the extent of bias introduced into remediation designs and costs by sorption assumptions. Remediation problems with equilibrium sorption and two-site kinetic sorption with different mass transfer rates are addressed. Under the time-scales and flow conditions of contamination, all of these sorption assumptions created similar initial plumes. Thus, calibration of sorption parameters to the initial plume may be insufficient information to differentiate between sorption rates. Two optimization formulations were considered for each sorption assumption. In one formulation, the final water quality goal is with respect to the aqueous concentration; in the other, the goal is with respect to the total concentration (aqueous and sorbed), While remediation designs could be found with the aqueous concentration formulation. the total mass of contaminant remaining within the aquifer increases with decreasing mass transfer rates, creating the potential for concentration rebound after the pumping period ends. The designs found with the total concentration formulation avoid this difficulty, although it may: not always be possible to find feasible remediation schemes within the specified remediation period, If an optimal design based on the assumption of equilibrium sorption is applied to an aquifer with mass transfer limitations, the water quality goals may not be met within the specified remediation period. The length of time required to meet the water quality goals with this suboptimal design may increase exponentially as the mass transfer rate decreases, although aquifer heterogeneity reduces the scale of this effect.