The diffusion of 2,2,4-trimethylpentan (TMP) and 2,2,5-trimethylhexane (TMH) vapors out of residually contaminated sandy soil from the U.S. Environmental Protection Agency (EPA) field research site at Traverse City, Michigan, was measured and modeled. The headspace of an intact core sleeve sample was swept with nitrogen gas to simulate the diffusive release of hydrocarbon vapors from residual aviation gasoline in and immediately above the capillary fringe to a soil-venting air flow in the unsaturated zone. The resulting steady-state profile was modeled using existing diffusivity and air porosity estimates in a balance of diffusive flux and a first order source term. The source strength, which was calibrated with the observed flux of 2,2,4-TMP leaving the sleeve, varied with the residual gasoline remaining in the core, but was independent of the headspace sweep flow rate. The finding suggested that lower soil-venting air flow rates were in principle as effective as higher air flow rates in venting LNAPL vapors from contaminated soils.