The purpose of this work was to examine the effect of porous-media heterogeneity, nonuniform distribution of dense nonaqueous phase liquid (DNAPL), and sampling method on the performance of the partitioning tracer method for measuring DNAPL saturation in water-saturated subsurface systems. Experiments were conducted in an intermediate-scale flow cell that contained two discrete zones of trichloroethene (TCE) at residual saturation. One zone (zone 2) consisted of ∼10% saturation formed in the same sand as used for the flow-cell matrix. The other zone (zone 1) consisted of ∼10% saturation in a finer sand emplaced within the coarser matrix. Aqueous samples were collected using depth-specific sampling, using vertically integrated sampling, and at the extraction well. A dual-energy γ radiation system was used to measure TCE saturation before and after the tracer experiment, allowing the measurements obtained from the tracer experiment to be compared to a previously tested method. The saturations estimated using the data collected at point-sampling ports located downgradient of zones 1 and 2 were approximately 7% and 50% of the true values, respectively. The saturations estimated using the data obtained from the vertically integrated ports were 0% and 6% of the true values, respectively. Finally, the saturation estimated using the extraction-well data was 30% of the cell-wide averaged value. These results indicate that the presence of porous-media heterogeneity and a variable distribution of DNAPL saturation can lead to reduced accuracy of the partitioning tracer test. The reduced performance can be improved, in part, by using depth-specific sampling.