The objective of this study was to determine if dithionite- and oxalate-extractable Fe and Al can be used to predict the P sorption capacity of Thai soils. Forty-five samples from diverse soil types were taken from surface and subsurface horizons of soils on sandstone, shale/limestone, granite, and basalt. The samples were analysed for P sorption, dithionite- and oxalate-extractable Fe and Al (Fe_d, Fe_o, Al_d, Al_o), specific surface area (SSA), and other soil properties. Generally P sorption data for these soils were slightly better fitted by the Langmuir equation than the Freundlich equation. The Langmuir P sorption maximum ranged from 35 to 1111 μg/g with a median value of 370 μg/g soil. Soils developed on basalt had higher values of P sorption maximum (x_m) (range 400–1111 μg/g, median 597 μg/g) than soils on other parent materials. Fe_d concentrations in soils (4–74 g/kg) were much higher than Fe_o concentrations (0.2–13.8 g/kg) with values of Fe_o/Fe_d ranging from 0.01 to 0.28 (median 0.09), indicating that most of the free iron oxides were crystalline. Amounts of Al_d and Al_o were about equal with median values of 1.6 and 1.0 g/kg, respectively. About 80% of the samples had SSA values <40 m²/g. Both the P sorption maximum and Freundlich k were linearly related to SSA (R² = 0.77, 0.74), Al_d (R² = 0.78, 0.79), Al_o (R² = 0.64, 0.74), Fe_d (R² = 0.48, 0.41), Fe_o (R² = 0.43, 0.72), and clay content (R² = 0.48, 0.36). Stepwise regression indicated that 81% of the variability in P sorption by these soils could be explained by a combination of dithionite and oxalate Fe and Al, however, Al_d alone is almost as effective in predicting the P sorption capacity of Thai soils.