Tropical forest ecosystems play an important role in the global carbon balance. Depending on age and land use, they can act as carbon sources, sinks, or be in approximate balance, but it is uncertain if global environmental changes are forcing these ecosystems outside their natural range of variation. We asked the question of whether or not the net carbon flux of a tropical primary forest, which should be in balance over the long term, is within the expected range of natural variation. A simple Bayesian hypothesis testing method was used to address this question for primary forests in the Porce region of Colombia. Net ecosystem production (NEP) was measured in this forest in a set of 33 permanent plots from 2000 to 2002 in 2, 1-year intervals. Our estimate of NEP ranged between −4.03 and 2.22 Mg C ha−1 yr−1 for the two intervals. This range was compared with a priori defined range of natural variation estimated from the ecosystem model STANDCARB, which estimated spatial and temporal variation due to gap dynamics. The prior range of variation was estimated between −1.5 and 1.5 Mg C ha−1 yr−1. The observed data on NEP did not provide sufficient evidence to reject the null hypothesis that these forests are in C balance. We concluded that the ecosystem is likely behaving within its range of natural variation, but measurement uncertainties were a major limitation to finding evidence to reject the null hypothesis. A literature review of C flux studies in the tropics revealed that about half of the observations could be explained by gap dynamics alone, while significant C sinks have only been observed during La Niña years, with contrasting results in other tropical forests. In conclusion, observational data of carbon fluxes do not appear to provide direct evidence for a significant carbon sink in some sites in the tropics.