Abstract. The relative importance of nitrogen inputs from atmospheric deposition and biological fixation is reviewed in a number of diverse, non-agricultural terrestrial ecosystems. Bulk precipitation inputs of N (I-12kgN ha-' yr-') are the same order of magnitude as, or frequently larger than, the usual range of inputs from nonsymbiotic fixation (< 1 -5 kg N ha-' yr-I), especially in areas influenced by industrial activity. Bulk precipitation measurements may underestimate total atmospheric deposition by 3040% because they generally do not include all forms of wet and dry deposition. Symbiotic fixation generally ranges from z lo-160 kg N ha-' yr-' in ecosystems where N-fixing species are present during early successional stages, and may exceed the range under unusual conditions.Rates of both symbiotic and nonsymbiotic fixation appear to be greater during early successional stages of forest development, where they have major impacts on nitrogen dynamics and ecosystem productivity. Fates and impacts of these nitrogen inputs are important considerations that are inadequately understood. These input processes are highly variable in space and time, and few sites have adequate comparative information on both nitrogen deposition and fixation.The existing information on N inputs is deficient in several areas: -more intensive studies of total atmospheric deposition, especially of dry deposition, are needed over a wide range of ecosystems; -additional studies of symbiotic fixation are needed that carefully quantify variation over space and time, examine more factors regulating fixation, and focus upon the availability of N and its effects upon productivity and other nutrient cycling processes; -process-level studies of associative N-fixation should be conducted over a range of ecosystems to determine the universal importance of rhizosphere fixation; -further examination of the role of free-living fixation in wood decomposition and soil organic matter genesis is needed, with attention upon spatial and temporal variation; and -investigations of long-term biogeochemical impacts of these inputs must be integrated with process-level studies using modem modelling techniques.