Field experiments were conducted on three typical farmland soils (loess soil, fluvo-aquic soil, and cinnamon fluvo-aquic soil) in Northern China to study the grain yield, phosphorus agronomic efficiency (PAE), and phosphorus use efficiency (PUE) of wheat under effects of long-term fertilizations. Seven treatments were installed, i.e., non-fertilization (CK), nitrogen fertilization (N), nitrogen-potassium fertilization (NK), nitrogen-phosphorus fertilization (NP), nitrogen-phosphorus-potassium fertilization (NPK), NPK plus straw returning (NPKS), and NPK plus manure application (NPKM). The averaged wheat grain yields under long-term P fertilizations (treatments NP, NPK, NPKS, and NPKM) ranged from 2914 kg x hm(-2) to 6219 kg x hm(-2), being 200%-400% higher than those under no P fertilizations (treatments CK, N, and NK), and no significant differences were observed between the P fertilizations. In the early years of the experiment, the PAE in treatment NPK on the loess soil, fluvo-aquic soil, and cinnamon fluvo-aquic soil was 17.0 kg x kg(-1), 20.3 kg x kg(-1), and 13.3 kg x kg(-1), and the PUE was 15.3%, 31.2%, and 23.8%, respectively. After 15-year fertilization, the PAE and PUE in treatment NPK increased annually by 3.9 kg x kg(-1) and 1.3% on loess soil, 2.5 kg x kg(-1) and 0.9% on fluvo-aquic soil, and 2.8 kg x kg(-1) and 1.0% on cinnamon fluvo-aquic soil, respectively. There were no significant differences in the PAE and PUE among the P treatments for the same soils. In Northern China, long-term P fertilization could increase the wheat grain yield and PUE significantly, and the mean annual increase of PAE and PUE in treatment NPKM was higher on loess soil than on fluvo-aquic soil and cinnamon fluvo-aquic soil.