A two-year field experiment was conducted to investigate ammonia volatilization from Gleyed paddy soil in the Taihu Lake region.The experiment was designed to have four N levels (0,180,255,330 kg hm^-2) and four P₂O₅ levels (0,30,90,180 kg hm^-2),forming six NP combinations (CK-N₀P₀,LN-N₁₈₀P₉₀,OPT-N₂₅₅P₉₀,LP-N₂₅₅P₃₀,HP-N₂₅₅Psub>180 and HN-N₃₃₀P₉₀),Ammonia volatilization was measured with the enclosure method.Results show that ammonia emission occurred within the first 6 days after fertilizer application,and it was less after the basal dressing and the first top-dressing than after the second top-dressing.N loss through ammonia emission accounted for 0.4%～12% of the total nitrogen applied for the basal and the first top dressings.However for the second top-dressing the percentage was 5.8%,9.7%,25.6%,15.6%,11.6% during the 2002 rice season,and 27.4%,26.2%,30.0%,35.1%,27.6% during the 2003 rice season for Treatments LN,OPT,LP,HP and HN,respectively.Ammonia loss was low when fertilization was followed by rainy days or was done right at the elongating or booting of the rice crop,Positive correlation was observed between NH₃ loss and NH₄⁺-N concentration in the surface water in all the three fertilization periods.When the rate of phosphorus fertilizer remained the same,NH₃ loss increased with N fertilizer rates,whereas when at the rate of N fertilizer stayed unchanged,NH₃ loss was higher in treatments higher and lower in P rate treatments than in treatments optimum in P rate,suggesting that ammonia loss would be greater when NP combination is unbalanced.P₂O₅ is recommended to be no more than 90 kg hm^-2 for rice.