Respiration, which is the second most important carbon flux in ecosystems following gross primary productivity, is typically represented in biogeochemical models by simple temperature dependence equations. These equations were established in the 19th century and have been modified very little since then. Recent applications of these equations to data on soil respiration have produced highly variable apparent temperature sensitivities. This paper searches for reasons for this variability, ranging from biochemical reactions to ecosystem-scale substrate supply. For a simple membrane-bound enzymatic system that follows Michaelis–Menten kinetics, the temperature sensitivities of maximum enzyme activity (Vmax) and the half-saturation constant that reflects the affinity of the enzyme for the substrate (Km) can cancel each other to produce no net temperature dependence of the enzyme. Alternatively, when diffusion of substrates covaries with temperature, then the combined temperature sensitivity can be higher than that of each individual process. We also present examples to show that soluble carbon substrate supply is likely to be important at scales ranging from transport across membranes, diffusion through soil water films, allocation to aboveground and belowground plant tissues, phenological patterns of carbon allocation and growth, and intersite differences in productivity. Robust models of soil respiration will require that the direct effects of substrate supply, temperature, and desiccation stress be separated from the indirect effects of temperature and soil water content on substrate diffusion and availability. We speculate that apparent Q10 values of respiration that are significantly above about 2.5 probably indicate that some unidentified process of substrate supply is confounded with observed temperature variation.