Despite the importance of anhydrous ammonia (AA) and urea as nitrogen (N) fertilizer sources in the United States, there have been few direct comparisons of their effects on soil nitrous oxide (N 2O) and nitric oxide (NO) emissions. We compared N oxide emissions, yields, and N fertilizer recovery efficiency (NFRE) in a corn (Zea mays L.) production system that used three different fertilizer practices: urea that was broadcast and incorporated (BU) and AA that was injected at a conventional depth (0.20 m) (AAc) and at a shallower depth (0.10 m) (AAs). Averaged over 2 yr in an irrigated loamy sand in Minnesota, growing season N 2O emissions increased in the order BU < AAc < AAs. In contrast, NO emissions were greater with BU than with AAc or AAs. Emissions of N 2O ranged from 0.5 to 1.4 kg N ha -1 (50-140 g N Mg -1 grain), while NO emissions ranged from 0.2 to 0.7 kg N ha -1 (20-70 g N Mg -1 grain). Emissions of total N oxides (NO + N 2O) increased in the order AAc < BU < AAs. Despite having the greatest emissions of N 2O and total N oxides, the AAs treatment had greater NFRE compared with the AAc treatment. These results provide additional evidence that AA emits more N 2O, but less NO, than broadcast urea and show that practices to reduce N 2O emissions do not always improve N use efficiency.