The influence of heterogeneity in aquifer hydraulic conductivity (K) on contaminant plume patterns and the required thickness and length of permeable reactive barriers (PRBs) used for in situ remediation is evaluated using stochastic modeling. The results provide a quantitative means for evaluating the effects of (1) the level of aquifer heterogeneity as reflected by the standard deviation of the logarithm of K, σlnK, (2) the aquifer correlation structure anisotropy represented by the ratio of correlation lengths, λx/λy, and (3) DPRB representing the distance from the contaminant source zone to the PRB. In terms of PRB thickness, a probabilistic factor of safety related to uncertainty in influent groundwater seepage velocities (FS1,90) at the location of the PRB is quantified. In terms of PRB length, a probabilistic factor of safety related to uncertainty in the length of a PRB required to capture the contaminant plume, defined as the capture length ratio (CLR), is quantified. The mean and standard deviation of FS1,90 significantly increase as σlnK increases from 0.2 to 1.6, and slightly increase as λx/λy increases from 1.0 to 3.0 and DPRB increases from 15 to 45 m. The values for the factor of safety for PRB thickness versus σlnK compare favorably with previously published values based on a different methodology. The mean and standard deviation of CLR increase with increasing σlnK and with increasing DPRB, and decrease slightly with increasing λx/λy. Finally, the ranges in CLR are correlated with strongly divergent and strongly convergent plume patterns.