Knowledge of long-term phosphorus (P) behavior is essential to improve soil structure, nutrient supply potential and environmental sustainability of cropping systems. A 35-year (1984-2019) long-term experiment under winter wheat-broomcorn millet and pea (W + M/P) was undertaken to investigate soil aggregation and aggregate associated P fractions using Hedley sequential fractionation method in the 60 cm soil profile of Northwest China. Treatments included (1) unfertilized control (CK), P fertilizer (P), N combined with P (NP), and NP plus farmyard manure (NPM). A settling tube apparatus was used to fractionate the soils physically into four sizes (> 250-, 125-250-, 63-125-, and < 63 mu m) aggregates. Results indicated long-term P fertilization regimes had no striking effects on promoting the size distribution of aggregates except for > 250 mu m in the 0-60 soil depths, where NPM treatment significantly increased the mass proportion by 50.7% at 0-20 cm, 78.8% at 20-40 cm, and 36.1% at 40-60 cm depths over CK, respectively. Compared with CK, long-term P fertilization markedly raised aggregate associated P fraction at 0-60 cm soil depths. In particular, the average labile P in all size aggregates increased by 93%, 63%, and 170%, moderately labile P by 117%, 118%, and 127%, and non-labile P by 324%, 268%, and 345%, respectively, under the P, NP, and NPM treatment that contributed higher in large (> 250 mu m) size aggregate than others. Our results suggested that long-term P fertilization, especially mineral P combined with farmyard manure better sustains soil structural stability and promotes soil P transformation and availability in soil aggregates in the soil profile, which would have environmental consequences. Therefore, P management needs to modify based on monitoring soil nutrients and P fractions over time.