Although regional and global models of nitrogen (N) cycling typically focus on nitrate, dissolved organic nitrogen (DON) is the dominant form of nitrogen export from many watersheds and thus the dominant form of dissolved N in many streams. Our understanding of the processes controlling DON export from temperate forests is poor. In pristine systems, where biological N limitation is common, N contained in recalcitrant organic matter (OM) can dominate watershed N losses. This recalcitrant OM often has moderately constrained carbon:nitrogen (C:N) molar ratios (approximately 25-55) and therefore, greater DON losses should be observed in sites where there is greater total dissolved organic carbon (DOC) loss. In regions where anthropogenic N pollution is high, it has been suggested that increased inorganic N availability can reduce biological demand for organic N and therefore increase watershed DON losses. This would result in a positive correlation between inorganic and organic N concentrations across sites with varying N availability. In four repeated synoptic surveys of stream water chemistry from forested watersheds along an N loading gradient in the southern Appalachians, we found surprisingly little correlation between DON and DOC concentrations. Further, we found that DON concentrations were always significantly correlated with watershed N loading and stream water [NO3-] but that the direction of this relationship was negative in three of the four surveys. The C:N molar ratio of dissolved organic matter (DOM) in streams draining watersheds with high N deposition was very high relative to other freshwaters. This finding, together with results from bioavailability assays in which we directly manipulated C and N availabilities, suggests that heterotrophic demand for labile C can increase as a result of dissolved inorganic N (DIN) loading, and that heterotrophs can preferentially remove N-rich molecules from DOM. These results are inconsistent with the two prevailing hypotheses that dominate interpretations of watershed DON loss. Therefore, we propose a new hypothesis, the indirect carbon control hypothesis, which recognizes that heterotrophic demand for N-rich DOM can keep stream water DON concentrations low when N is not limiting and heterotrophic demand for labile C is high.