Phosphorus (P) fertilization must be conducted with a clear understanding of P forms and their distribution in the soil profile. Excessive application of P fertilizers could eventually cause P accumulation, especially in the topsoil. P loss from a nutrient-enriched agricultural field can lead to eutrophication of water bodies, which has been a recent global concern. Traditionally, the P obtained after filtering a soil solution by using a 0.45-μm microporous membrane is defined as "particulate" or "dissolved" P species. Compared to the immobile soil matrix, colloidal particles (size, typically, between 1 nm and 1 μm) have larger surfaces and stronger sorption capacities. Colloid-facilitated P transport in agricultural soils has received much attention in recent decades because the binding of P species to colloidal particles can largely enhance its mobility. Therefore, colloidal P is an important contributor of P according to soil particle size fractionation, and it plays a significant role in the distribution, transformation, and variation of P in the soil environment. However, to date, few studies have shown the changes in soil colloidal P under long-term P fertilization. This study was based on long-term field experiments conducted in Hangjiahu Plain. We measured the changes of soil colloidal P contents and analyzed its distribution characteristics in a paddy soil profile, under different fertilization managements. Four P fertilizer treatments were applied: no fertilizer control (CK), low P fertilizer (P1, 26 kg P /hm²), high P fertilizer (P2, 39 kg P /hm²), and manure treatment (M, 26 kg P /hm²). Soil samples were collected after oilseed rape and rice harvests, and the soil profiles were divided into four layers, 0-5, 5-30, 30-60, and 60-100 cm. Soil total P concentration, mass of soil colloidal particles, and colloidal and truly dissolved P in water extraction were determined. Soil colloidal and truly dissolved P were determined using extraction with water, centrifugation, and ultra-centrifugation at 300000r/min for 2 h to remove the colloids; P concentrations were determined using spectrophotometry based on binding with molybdenum and antimony. The mass of soil colloidal particles and colloidal P was calculated as the difference between non-ultracentrifuged and ultracentrifuged samples. Our results showed that colloidal P occupied at least 85% of the P in a soil colloidal solution and 0.1%-2% of the total P in the soil. P fertilization increased soil colloidal P concentration in all treatments, especially in the 0-5 cm layer in the M treatment, in which colloidal P was 8.0 mg/kg. Furthermore, soil colloidal P decreased with soil depth, except in the 0-5 cm and 5-30 cm layers after rice harvest. Colloidal P decreased significantly in these two layers after rice harvest, which might have resulted from the anoxic conditions during the flooded period of rice production. At 0-5 cm depth, soil colloidal P after rice harvest was 90% lower than that after oilseed rape harvest; the reduction was less for the manure treatment. In general, compared to inorganic P fertilizer treatments, manure treatment had a more significant influence on soil colloidal P content. Colloidal P increased from 5.9% to 18.3% in the 30-60 cm layer after rice harvest compared to that after oilseed rape harvest, mainly because of the vertical transport of colloidal P. Our results provide scientific evidence for the existing P forms within the soil profile and the transformation characteristics of P. Our study also offers guidance for evaluating the environmental risk of varying levels of colloidal P.