Soil carbon is returned to the atmosphere through the process of soil respiration, which represents one of the largest fluxes in the terrestrial C cycle. The effects of climate change on the components of soil respiration can affect the sink or source capacity of ecosystems for atmospheric carbon, but no current techniques can unambiguously separate soil respiration into its components. Long-term free air CO2 enrichment (FACE) experiments provide a unique opportunity to study soil C dynamics because the CO2 used for fumigation has a distinct isotopic signature and serves as a continuous label at the ecosystem level. We used the 13C tracer at the Duke Forest FACE site to follow the disappearance of C fixed before fumigation began in 1996 (pretreatment C) from soil CO2 and soil-respired CO2, as an index of belowground C dynamics during the first 8 years of the experiment. The decay of pretreatment C as detected in the isotopic composition of soil-respired CO2 and soil CO2 at 15, 30, 70, and 200 cm soil depth was best described by a model having one to three exponential pools within the soil system. The majority of soil-respired CO2 (71%) originated in soil C pools with a turnover time of about 35 days. About 55%, 50%, and 68% of soil CO2 at 15, 30, and 70 cm, respectively, originated in soil pools with turnover times of less than 1 year. The rest of soil CO2 and soil-respired CO2 originated in soil pools that turn over at decadal time scales. Our results suggest that a large fraction of the C returned to the atmosphere through soil respiration results from dynamic soil C pools that cannot be easily detected in traditionally defined soil organic matter standing stocks. Fast oxidation of labile C substrates may prevent increases in soil C accumulation in forests exposed to elevated [CO2] and may consequently result in shorter ecosystem C residence times.