The temperature dependence of organic matter decomposition is a critically important determinant of any long-term changes of soil-carbon stocks in response to global warming. Because of practical experimental constraints, most knowledge of this temperature dependence is based on short-term studies. These studies generally show a strong temperature dependence of organic matter decomposition. At the same time, many modelling studies, especially global studies, or studies that investigate the effects of climate change, use longer time steps, such as annual. It is investigated here to what extent the use of short-term temperature dependencies are appropriate, or how they may need to be modified, for application over longer time steps. The work indicated that for global applications, it is critically important to explicitly consider seasonal temperature variations. Across the globe, observed annual mean temperature and the annual temperature range are negatively correlated. Inclusion of this correlation means that the strong short-term temperature dependence becomes much weaker when data are expressed as annual averages for the temperatures experienced across the globe. For short-term responses, the temperature dependence of organic matter decomposition is greater at low than high temperature and deviates strongly from an assumption of a constant Q10 across temperature. For annually averaged values, this pattern also weakens, and temperature dependencies change only slightly with temperature. Using short time steps for simulations leads to the expectation of more positive changes (sequestration) in soil carbon especially for cold regions of the globe than would be predicted for simulations at annual time steps without explicit consideration of seasonal temperature variations. These considerations help to reconcile some of the apparent differences in temperature dependencies obtained by different workers using different approaches.