Eutrophication, a major problem in many fresh and brackish waters, is largely caused by non-point source pollution by phosphorus (P) from agricultural soils. In Sweden, agriculture is estimated to contribute about 40% of the total anthropogenic net load of P to the surrounding coastal areas. Several P leaching studies have stressed the importance of the subsoil for P leaching but few attempts have been made to quantify the effect of the subsoil being a source or sink of P. This study with two clay and two sandy soils investigated the influence of soil P content and sorption characteristics in topsoil and subsoil on P leaching measured during 30 months in undisturbed soil columns collected with (1-m long) and without (0.8-m long) the topsoil. Total P losses during the period varied between 1.0 and 14.4 kg ha^-1 in lysimeters with topsoil and between 1.0 and 12.3 kg ha^-1 in lysimeters without. In all soils, leaching of dissolved reactive P was smaller from lysimeters without topsoil than from lysimeters with. In contrast, leaching of particulate P was higher from subsoil lysimeters than from lysimeters with topsoil and subsoil in the sandy soils and one clay soil, which indicates presence of less stable soil structure in the subsoil. Soils with high sorption capacity (PSI) and low P saturation in the subsoil (<10% of PSI) had small P leaching (<2.5 kg ha^-1), despite high extractable P (Olsen-P) in the topsoil. One of the soils, with low sorption capacity, high P saturation (>35% of PSI) and high Olsen-P in the profile, had large P leaching (>12 kg ha^-1), in lysimeters both with and without topsoil. These preliminary results indicate that subsoils could act as either source or sink for P leaching and that topsoil studies are insufficient for assessing P leaching in many soils.