Desorption of cadmium (Cd) from sand was studied by both batch and flow-through methods. Batch experiments were conducted at three pH values (5.5, 6.0 and 6.5). In each case, the amount of Cd desorbed was low compared with the quantity of Cd adsorbed previously. Desorption of Cd in the batch experiments can be described adequately by a Freundlich isotherm. The Freundlich isotherm coefficient, K- f, increased with pH. Hysteresis between the sorption/desorption isotherms was observed in all batch experiments. Flow-through experiments in soil columns were conducted for the same three pH values, with the results used to determine transport and sorption/desorption parameters. Again, the desorption isotherms bore little resemblance to the corresponding adsorption isotherms. The experimental breakthrough curves were well fitted by a nonequilibiium desorption model, however the time scale of the desorption process was much larger than measured in batch experiments. This model was therefore rejected as lacking realism. A simple linear retardation (including hysteresis) model that utilises different isotherms was found to simulate column breakthrough curves well. The Freundlich isotherm coefficients, K-f, in all batch and flow-through desorption experiments were different to values evaluated from the corresponding adsorption experiments. However, in contrast to adsorption, desorption in flow-through experiments was not noticeably affected by changes in pH. The effect of pore-water velocity on desorption was also studied at pH 6.0. No trend was established between flow velocity and the desorption coefficient.