Managing nitrogen (N) inputs to sustain high yields while minimizing losses to adjacent environments remains among the foremost aims in agroecosystems. We studied the N balance in a study established in 1911 at Lethbridge, AB, Canada. The experiment includes three cropping systems — continuous wheat (W), fallow–wheat–wheat (FWW), and fallow–wheat (FW) — with a factorial of two N rates (0 and 45 kg N ha^-1) and two phosphorus (P) rates (0 and 20 kg P ha^-1) superimposed beginning in 1967. In unfertilized subplots, grain yields generally increased for the first eight decades, but then declined, perhaps partly because of growing N deficiency. Yield response to N increased over time, especially under continuous cropping and when co-applied with P. Soil N concentration in the surface 15 cm declined in the first few decades, and then approached an apparent steady state. Application of N increased soil N, roughly in proportion to the amount of residue returned. For the first half-century (1911–1967), N removal was approximately equivalent to the loss of soil N in the surface 15 cm. Since then, however, when the soil organic N was near steady state, removals of N in grain exceeded N inputs by approximately 20–30 kg N ha^-1 yr^-1, suggesting an input from outside sources, perhaps partly from atmospheric NH₃. This study demonstrated the importance of long-term experiments in evaluating the N balance of cropping systems, and indicated the potential significance of non-fertilizer N inputs from outside sources in such ecosystems.