The carbon balance of the world's terrestrial ecosystems is uncertain. Both top-down (atmospheric) and bottom-up (forest inventory and land-use change) approaches have been used to calculate the sign and magnitude of a net terrestrial flux. Different methods often include different processes, however, and comparisons can be misleading. Differences are not necessarily the result of uncertainties or errors, but often result from incomplete accounting inherent in some of the methods. Recent estimates are reviewed here. Overall, a northern mid-latitude carbon sink of approximately 2 Pg C yr−1 appears robust, although the mechanisms responsible are uncertain. Several lines of evidence point to environmentally enhanced rates of carbon accumulation. Other lines suggest that recovery from past disturbances is largely responsible for the sink. The tropics appear to be a small net source of carbon or nearly neutral, and the same uncertainties of mechanism exist. In addition, studies in the tropics do not permit an unequivocal choice between two alternatives: large emissions of carbon from deforestation offset by large sinks in undisturbed forests, or moderate emissions from land-use change with essentially no change in the carbon balance in undisturbed forests. Resolution of these uncertainties is most likely to result from spatially detailed historical reconstructions of land-use change and disturbance in selected northern mid-latitude regions where such data are available, and from systematic monitoring of changes in the area of tropical forests with satellite data of high spatial resolution collected over the last decades and into the future.