The effect of long-term fertilization on soil-denitrifying communities was determined by measuring the abundance and diversity of the nitrite reductase genes nirK and nirS. Soil samples were collected from plots of a long-term fertilization experiment started in 1990, located in Taoyuan (110°72″ E, 28°52″ N), China. The treatments were no fertilizer (NF), urea (UR), balanced mineral fertilizers (BM), and BM combined with rice straw (BMR). The abundance, diversity, and composition of the soil-denitrifying bacteria were determined by using real-time quantitative PCR, terminal restriction fragment length polymorphism (T-RFLP), and cloning and sequencing of nirK and nirS genes. There was a pronounced difference in the community composition and diversity of nirK-containing denitrifiers responding to the long-term fertilization regimes; however, less variation was observed in communities of nirS-containing denitrifiers, indicating that denitrifiers possessing nirK were more sensitive to the fertilization practices than those with nirS. In contrast, fertilization regimes had similar effects on the copy numbers of nirK and nirS genes. The BMR treatment had the highest copy numbers of nirK and nirS, followed by the two mineral fertilization regimes (UR and BM), and the lowest was in the NF treatment. Of the measured soil parameters, the differences in the community composition of nirK and the abundance of nir denitrifiers were highly correlated with the soil carbon content. Therefore, long-term fertilization resulted in a strong impact on the community structure of nirK populations only, and total organic carbon was the dominant factor in relation to the variations of nir community sizes.