Geotechnical centrifuge testing is used to examine the preferential (fingered) flow of a nonaqueous phase liquid (NAPL) in a uniform dry sand. The results of nine experiments, containing a total of 87 observations of NAPL finger behavior, are analyzed. The observed finger tip velocities range from 0.01 to 0.3 cm/s, while the observed finger widths range from 0.3 to 3.6 cm. From the experimental data it is concluded that, asymptotically, the NAPL fingers are not fully saturated. For comparison, the behavior of water fingers is examined using the same experimental setup. In contrast to the NAPL fingers, and in agreement with other work reported in the literature, the water fingers are found to be fully saturated. In addition, it is confirmed that the water finger properties can be well predicted from known porous medium and fluid properties. A scaling analysis is presented that allows the NAPL finger properties to be inferred from models developed to describe water finger properties. The analysis predicts NAPL finger velocities to within 15% and NAPL finger, widths to within 50% if both finger types are assumed saturated. By adjusting the analysis to account for the fact that the NAPL fingers are not fully saturated, NAPL finger widths can be predicted within to 10%, and NAPL finger velocities to within 30%.