Simple nitrogen (N) input/output balance calculations in agricultural systems are used to evaluate performance of nutrient management; however, they generally rely on extensive assumptions that do not consider leaching, denitrification, or annual depletion of soil N. We constructed a relatively complete N mass balance for the Big Ditch watershed, an extensively tile-drained agricultural watershed in east-central Illinois. We conducted direct measurements of a wide range of N pools and fluxes for a 2-yr period, including soil N mineralization, soybean N2 fixation, tile and river N loads, and ground water and in-stream denitrification. Fertilizer N inputs were from a survey of the watershed and yield data from county estimates that were combined with estimated protein contents to obtain grain N. By using maize fertilizer recovery and soybean N2 fixation to estimate total grain N derived from soil, we calculated the explicit change in soil N storage each year. Overall, fertilizer N and soybean N2 fixation dominated inputs, and total grain export dominated outputs. Precipitation during 2001 was below average (78 cm), whereas precipitation in 2002 exceeded the 30-yr average of 97 cm; monthly rainfall was above average in April, May, and June of 2002, which flooded fields and produced large tile and riverine N loads. In 2001, watershed inputs were greater than outputs, suggesting that carryover of N to the subsequent year may occur. In 2002, total inputs were less than outputs due to large leaching losses and likely substantial field denitrification. The explicit change in soil storage (67 kg N ha−1) offsets this balance shortfall. Although 2002 was climatically unusual, with current production trends of greater maize grain yields with less fertilizer N, soil N depletion is likely to occur in maize/soybean rotations, especially in years with above-average precipitation or extremely wet spring periods.