The magnitude and timing of the response of the soil carbon reservoir to changes in land use or climate is a large source of uncertainty in global carbon cycle models. One method of assessing soil carbon dynamics, based on modeling the observed increase of 14C in organic matter pools during the 30 years since atmospheric weapons testing ended, is described in this paper. Differences in the inventory and residence time of carbon are observed in organic matter from soils representing tropical (Amazon Basin, Brazil) and temperate (western slope of the Sierra Nevada mountains, California) forest ecosystems. The majority of the organic carbon in the upper 22 cm of the tropical soil (7.1 kgC m-2) has residence times of 10 years or less, with a minor component of very refractory carbon. The estimated annual flux of carbon into and out of the soil organic matter in this horizon of the mineral soil, based on modeling of the 14C data, is between 1.9 and 5.5 kgC m-2 yr-1. In contrast, organic matter in the temperate soil over a similar depth interval (0--23 cm; 5.2 kgC m-2), is made up of approximately equal amounts of carbon with residence times of 10, 100, and 1000 years. The estimated annual flux of carbon into and out of this soil is 0.22 to 0.45 kgC m-2 yr-1. Rapid turnover of organic matter with density 1.6--2.0 g cm-3 removed 14C-enriched components from the temperate soil but had no effect on the 14C content of the residue in 0--22 cm layer of the tropical soil.The results presented here show that carbon cycle models which treat soil carbon dynamics as a single reservoir with a turnover rate based on radiocarbon measurements of bulk soil organic matter underestimate the annual fluxes of organic matter through the soil organic matter pool, particularly in tropical regions.