Biogeochemical monitoring for 45 years at the Hubbard Brook Experimental Forest in New Hampshire has revealed multiple surprises, seeming contradictions, and unresolved questions in the long-term record of ecosystem nitrogen dynamics. From 1965 to 1977, more N was accumulating in living biomass than was deposited from the atmosphere; the "missing" N source was attributed to biological fixation. Since 1992, biomass accumulation has been negligible or even negative, and streamwater export of dissolved inorganic N has decreased from ~4 to ~1 kg of N ha(-1) year(-1), despite chronically elevated atmospheric N deposition (~7 kg of N ha(-1) year(-1)) and predictions of N saturation. Here we show that the ecosystem has shifted to a net N sink, either storing or denitrifying ~8 kg of N ha(-1) year(-1). Repeated sampling over 25 years shows that the forest floor is not detectably accumulating N, but the C:N ratio is increasing. Mineral soil N has decreased nonsignificantly in recent decades, but the variability of these measurements prevents detection of a change of <700 kg of N ha(-1). Whether the excess N is accumulating in the ecosystem or lost through denitrification will be difficult to determine, but the distinction has important implications for the local ecosystem and global climate.