Nonlinear complex-resistivity (NLCR) cross-hole imaging of the vadose zone was performed at the A-014 Outfall at the Savannah River Site, Aiken, SC. The purpose of this experiment was to field-test the ability of this method to detect dense nonaqueous phase liquids (DNAPLs), specifically tetrachloroethene (PCE), known to contaminate the area. Five vertical electrode arrays (VEAs) were installed with ∼15-ft (3 m) separations in and around the suspected source zone to depths of 72 ft (22 m), and measurements were carried out at seven nearest-neighbor panels. Amplitude and phase data were edited for quality and then inverted to form three-dimensional (3D) images. The comparatively small magnitude of the nonlinear resistivity Hilbert distortion allowed approximate linearized imaging of the 3D distribution of this quantity as well. Laboratory analysis of nearby soil contaminated in situ indicated that the NLCR response to the PCE-clay reaction is maximized near 50 mHz, leading to the development of a metric involving the phase and resistivity Hilbert distortion to infer the 3D distribution of PCE. Variations in PCE content were independently detailed at three drilling locations within the NLCR survey area using direct penetration-based soil-collection tools. Approximately 400 soil samples were collected and analyzed for chlorinated solvent mass composition at 1-ft (0.3-m) vertical intervals to compare with the NLCR-predicted distribution of DNAPL. The optimum performance for 1,000 mg/kg PCE was ∼80% detection (true positives) with ∼30% false alarms (false positives) at an effective resolution of 4 ft (1.2 m), or ∼1/4 of the interwell separation. When smoothed to 12-ft (3.7 m) resolution (comparable to well spacing), detection was 100% with just 12% false alarms. NLCR successfully predicted the general distribution of PCE at parts-per-thousand soil-mass fractions, specifically widespread near-surface contamination and a zone of discontinuous stringers and pods below the source.