Knowledge of the gaseous diffusion coefficient is necessary to properly model gas movement in porous media. In this study, a nonreactive tracer (freon-12) and a reactive tracer (hexafluorobenzene) were used to evaluate the relationship between the ratio of the gas diffusion coefficient in soil (D g s ) to that in free air (D g a ) and the volumetric air content (a) with a two-chamber diffusion system. The measured D g s /D g a values in our experiments (i.e., the relative gas diffusion coefficients) showed a similar relationship to air content as data obtained from various studies reported in the literature. Our data and values taken from other studies were used to test the validity of various models. When compared with data obtained from experiments with disturbed soils, the Penman model D g s /D g a = 0.66 a was found to overestimate the measured relative diffusion coefficient, whereas the Millington-Quirk model D g s /D g a = a 10/3 /φ 2 , where φ is the soil porosity, underestimated it. Another Millington-Quirk relationship D g s /D g a = a 2 /φ 2/3 , which has been largely overlooked in the literature, was found to provide significantly better agreement with measured relative diffusion coefficients in various disturbed soils of different texture. On the contrary, no universal relationship was found when data from both disturbed and undisturbed soil experiments were evaluated. The Troeh model D g s /D g a = [(a - u)/(1 - u)] v had the flexibility to fit all of the experimental data when both of the model parameters were varied simultaneously ; however, no obvious correlation was found between soil properties and the parameters. The limitations of any universal form for gas tortuosity model in natural soils were analyzed with percolation theory.