At many contaminated sites, nonaqueous phase liquids (NAPLs) persist in the vadose zone for long periods of time. This occurs because the permeability of the NAPL becomes negligible at some saturation and downward movement ceases, resulting in residual NAPL. To obtain data that can be used to study the development of a residual NAPL saturation and to test corresponding models, a detailed transient experiment was conducted in a 170 cm long by 90 cm high by 5.5 cm wide flow cell. Fluid saturation measurements were obtained with a dual-energy γ radiation system. The experimental conditions reflected those at the Hanford Site in Washington State, where an estimated 363 to 580 m3 of carbon tetrachloride (CCl4) was disposed to the subsurface. A key subsurface feature at the Hanford Site is a sloped Plio-Pleistocene caliche layer, which was reproduced in the experiment as a sloped lens in a medium-grained, uniform, sand matrix. The caliche contains considerable amounts of CaCO3 and may have fluid wettability properties other than strongly water wet. A total of 800 mL of CCl4 was injected into the experimental domain at a rate of 0.5 mL min−1 from a small source area located at the surface. After apparent static conditions were obtained with respect to CCl4 redistribution, saturation measurements indicated that all of the dense nonaqueous phase liquids (DNAPL) that had initially moved into the caliche remained in this layer. Water was subsequently applied to the surface at a constant rate over the full length of the caliche layer to study CCl4 displacement as a result of changing water saturations. Water saturation in the caliche layer rose to as high as 0.91 during water infiltration. Results show that 25% of the DNAPL present in the caliche migrated from this layer as a consequence of water infiltration, while 75% remained in the caliche layer. The experimental results could not be reproduced with numerical multifluid flow simulations based on common constitutive theory. This indicates that improvements in constitutive theory may be needed to accurately model air−DNAPL−water flow behavior.