Unlike phosphate, an index of sulphate retention of soils has not commonly been defined. However, sulphate sorption by soils is more sensitive than phosphate sorption to the amount of their colloidal components. Sulphate has less sorption affinity than some organic acids or phosphate, and is sorbed by hydrous Al/Fe such as allophane but not by Al–humus complexes. Therefore, we attempted to categorize soils as allophanic or non-allophanic Andosols by sorption experiments with sulphate. The method employed for sulphate retention (S-retention) was based on the common method for P-retention, using potassium sulphate solution (3.12 mmol SO42− l−1, equivalent to 100 mg S l−1) buffered by potassium acetate (pH 4.0) and shaking for 1 hour with a soil:solution ratio of 1:5. From the results from 172 soil samples of 30 forest soils in eastern Japan, a linear regression was obtained between S-retention and Alo + 1/2Feo (r = 0.780, P < 0.001). All aluandic (non-allophanic) soil samples (Sio < 0.6% and Alp:Alo > 0.5, in accordance with the WRB classification system) had poor S-retention (< 60%) and were clearly divided from silandic (allophanic) soil samples. Highly weathered Cambisols with small activity ratios of free iron oxides (Feo:Fed < 0.4) tended to have greater S-retention (> 30%) than less weathered Umbrisols and Cambisols with Feo:Fed > 0.4. In addition, these four soil types were separated as distinct domains on a scatter diagram of P-retention plotted against S-retention.