Abstract Soil leachate data from a long-term lysimeter (soil column) experiment were compared with the predictions of the MAGIC acidification model in an attempt to validate the latter. Acid production from nitrification was a necessary model input. In the surface horizon, the model gave a good prediction of base cation concentrations and their time trends, provided the weathering rate was set at about 250 mEq. m −2 year −1. However, the predicted pH and Al were far too high: this probably shows that the Al(OH) 3 control used in the model was inappropriate. At the base of the profile, pH and Al were modelled more satisfactorily but fluctuations in anion and base cation concentrations were out of phase from those predicted. This was because the model assumes instantaneous reaction throughout the soil profile whereas in reality downward transport takes a significant time. As the Al(OH) 3 solubility constant in the model was not temperature-dependent, MAGIC failed to predict the small annual fluctuation in Al concentration in the lower horizons. Because of the incorrect Al and H + modelling in the surface horizons, and possibly because vertical variation in selectivity coefficients was not accounted for, the model predicted a much larger decrease in soil base saturation than was actually observed. Soil processes in the MAGIC model appear to have been well chosen and modelled responses could all be observed in the experimental data. The model gave a good qualitative prediction of the experimental data and could be adjusted to give a reasonable quantitative prediction for many variables. In the surface horizon, however, and by extension in other soils with a dominating organic matter component, the Al(OH) 3 control used was inappropriate and could lead to misleading results.