Predictive models for the solute diffusion coefficient, DS, dependency on volumetric soil-water content, &thetas;, are often applied in simulations of solute transport and fate in natural, undisturbed soils. However, all available DS(&thetas;) models have been developed from measurements on sieved, repacked soil. In this study, DSfor chloride was measured in both repacked and undisturbed loamy sands at different soil-water contents. The measurements on undisturbed soil were carried out using a modified half-cell method, where the source half-cell is a sieved and repacked soil core and the other half-cell is an undisturbed soil core. Thus, the problems of (i) incomplete contact area at the interface between undisturbed half-cells and (ii) potentially different diffusion properties in undisturbed half-cells can be avoided. The modified half-cell method requires that the diffusion coefficient in sieved, repacked soil is determined separately and that the experimental data is analyzed with a numerical solution to the diffusion equation. No significant difference in chloride DS(&thetas;) between undisturbed and sieved, repacked soil was observed for a Danish (Foulum) loamy sand and a Japanese (Hiroshima) loamy sand. A recently presented soil type dependent DS(&thetas;) model, derived from repacked soil data, shows it to be applicable also for predicting solute diffusion coefficients in natural, undisturbed soils.