Whether the thermal sensitivity of an organism's traits follows the simple Boltzmann-Arrhenius model remains a contentious issue that centers around consideration of its operational temperature range and whether the sensitivity corresponds to one or a few underlying rate-limiting enzymes. Resolving this issue is crucial, because mechanistic models for temperature dependence of traits are required to predict the biological effects of climate change. Here, by combining theory with data on 1,085 thermal responses from a wide range of traits and organisms, we show that substantial variation in thermal sensitivity (activation energy) estimates can arise simply because of variation in the range of measured temperatures. Furthermore, when thermal responses deviate systematically from the Boltzmann-Arrhenius model, variation in measured temperature ranges across studies can bias estimated activation energy distributions toward higher mean, median, variance, and skewness. Remarkably, this bias alone can yield activation energies that encompass the range expected from biochemical reactions (from ~0.2 to 1.2 eV), making it difficult to establish whether a single activation energy appropriately captures thermal sensitivity. We provide guidelines and a simple equation for partially correcting for such artifacts. Our results have important implications for understanding the mechanistic basis of thermal responses of biological traits and for accurately modeling effects of variation in thermal sensitivity on responses of individuals, populations, and ecological communities to changing climatic temperatures.