Ten Scottish soils were cropped in a glasshouse with three sowings of perennial ryegrass (Lolium perenne, cv. Dutch Barenza) without added magnesium. After 25 cuts (185 weeks) growth was poor in some soils, but soil Mg was not limiting growth in others. Uptake of Mg and the Mg concentration in the ryegrass correlated well over the entire cropping period with the equilibrium Mg-concentration ratio Mg/(Ca, Mg) = CR0, determined from the Mg quantity/intensity isotherms measured before cropping in 001 m chloride solutions. A CR0 value of 016 was required in the early stages of cropping to give 02 % Mg in grass dry matter. Seven soils released 12-37 mg non-exchangeable Mg/kg, between one-fifth and one-third of the total Mg uptake. These reserves probably came from vermiculite minerals, but the release was too slow to maintain Mg concentrations of 02 % in the ryegrass.