Abstract: Using a long-term fertilization experiment set up in 1978 in Pingliang, Gansu Province, changes in soil phosphorus (P) and P fertilizer efficiency (PFE) were determined and used to guide rational application of P fertilizer in dry farmlands. Six fertilizer treatments-no fertilizer (CK), chemical nitrogen (N), chemical nitrogen and phosphorus (NP), straw plus chemical NP (SNP), farmyard manure (M) and farmyard manure plus chemical NP (MNP)-were used in the cropping system of 4-year continues winter wheat followed by 2-year spring corn. After 38 years, the total soil phosphorus and Olsen-P contents in CK and N treatments decreased, and phosphorus activation coefficient dropped. Total soil phosphorus contents under NP, SNP, M and MNP treatments increased respectively by 22.8%, 14.0%, 38.6% and 56.1%. Then Olsen-P contents increased respectively by 99.1%, 48.4%, 206.4% and 375.6%, while P activation coefficients were 1.7, 1.3, 2.2 and 3.1 times as that in 1978. Total soil P contents in N and SNP treatments were negatively correlated with the number of experimental years with annual decline rates of 1.9 mg·kg^-1 and 2.6 mg·kg^-1, respectively. Total soil P contents under NP, M and MNP treatments were positively correlated with the number of experimental years, with annual increase rates of 1.2 mg·kg^-1, 1.9 mg·kg^-1 and 2.8 mg·kg^-1, respectively. Total soil P contents under CK remained basically unchanged with the number of experimental years. Soil available P contents under CK and N treatments decreased with increasing number of experimental years, and the annual decline rates were respectively 0.03 mg·kg^-1 and 0.09 mg·kg^-1. Soil available P contents under NP, SNP, M and MNP treatments were positively correlated with the number of experimental years, with annual increase rates of respectively 0.29 mg·kg^-1, 0.24 mg·kg^-1, 0.46 mg·kg^-1 and 0.89 mg·kg^-1. Crop yield was positively correlated with soil available P (wheat R²=0.116 9, n=132; corn R²=0.332 4, n=54). P using rate (PUR), P recovery rate (PRR) and P agronomic efficiency (PAE) of corn were greater than those of wheat for all 4 P treatments (NP, SNP, M and MNP), but P physiological efficiency (PPE) of wheat was greater than that of corn. The ranked order of above four indexes of P fertilizer efficiency (PFE) for different treatments was SNP > NP > MNP > M. The four indexes of PFE of maize were positively correlated with the number of experimental years. Compared with MNP treatment, P input with M decreased 14.2%. Wheat and corn PFE decreased by 14.3%-69.5% and 0.8%-75.5%, respectively. In summary, combined application of organic and inorganic fertilizers was optimal choice for improving soil fertility, increasing crop yield and enhancing resource use efficiency in Heilu soil of the Loess Plateau dryland.