Accelerated decomposition of the soil carbon (C) reservoir in response to global warming could transform mineral soils from a net sink of atmospheric CO₂ to a net source. We examined the apparent sensitivity of old soil C decomposition in forests to temperature using field-based estimates of soil C decomposition across gradients in mean annual temperature (MAT) and laboratory-based estimates of soil C decomposition for soils sampled along a wide MAT gradient in North America. We found that the sensitivity of soil carbon decomposition rates in the field responded to changes in MAT with a Q10 of ~ 1.2. Using boot-strap statistical approach, we show that this Q10 estimate has a confidence interval of 1.0 to 1.4, ruling out a statistical explanation for the apparent lack of temperature sensitivity of soil C examined across our MAT gradient. Using a combination of long-term and short-term incubations, we also report that for old carbon (that C remaining after 300 days of incubation), apparent decomposition responses to temperature (Q10) were 20% to 60% lower depending on temperature than the apparent response of new carbon (newly incubated soils).  Taken together, our results suggest that decomposition of older soil C in upland temperate and tropical forests is unlikely to respond more strongly to rising global temperatures than labile C.