This paper describes laboratory and field experiments which were conducted to study the dynamics of trichloroethylene (TCE) as it volatilized from contaminated groundwater and diffused in the presence of infiltrating water through the unsaturated soil zone to the land surface. The field experiments were conducted at the Picatinny Arsenal, which is part of the United States Geological Survey Toxic Substances Hydrology Program. In both laboratory and field settings the gas and water phase concentrations of TCE were not in equilibrium during infiltration. Gas-water mass transfer rate constants were calibrated to the experimental data using a model in which the water phase was treated as two phases: a mobile water phase and an immobile water phase. The mass transfer limitations of a volatile organic compound between the gas and liquid phases were described explicitly in the model. In the laboratory experiment the porous medium was nonsorbing, and water infiltration rates ranged from 0.076 to 0.28 cm h⁻¹. In the field experiment the water infiltration rate was 0.34 cm h⁻¹, and sorption onto the soil matrix was significant. The laboratory-calibrated gas-water mass transfer rate constant is 3.3×10⁻⁴ h⁻¹ for an infiltration rate of 0.076 cm h⁻¹ and 1.4×10⁻³ h⁻¹ for an infiltration rate of 0.28 cm h⁻¹. The overall mass transfer rate coefficients, incorporating the contribution of mass transfer between mobile and immobile water phases and the variation of interfacial area with moisture content, range from 3×10⁻⁴ h⁻¹ to 1×10⁻² h⁻¹. A power law model relates the gas-water mass transfer rate constant to the infiltration rate and the fraction of the water phase which is mobile. It was found that the results from the laboratory experiments could not be extrapolated to the field. In order to simulate the field experiment the very slow desorption of TCE from the soil matrix was incorporated into the mathematical model. When desorption from the soil matrix was added to the model, the calibrated gas-water mass transfer rate constant is 2 orders of magnitude lower than that predicted using the power law model developed for the nonsorbing laboratory soil.