Thermal activity constraints play a major role in many aspects of ectotherm ecology, including vulnerability to climate change. Therefore, there is strong interest in developing general models of the temperature dependence of activity. Several models have been put forth (explicitly or implicitly) to describe such constraints; nonetheless, tests of the predictive abilities of these models are lacking. In addition, most models consider activity as a threshold trait instead of considering continuous changes in the vigor of activity among individuals. Using field data for a tropical lizard (Anolis cristatellus) and simulations parameterized by our observations, we determine how well various threshold and continuous-activity models match observed activity patterns. No models accurately predicted activity under all of the thermal conditions that we considered. In addition, simulations showed that the performance of threshold models decreased as temperatures increased, which is a troubling finding given the threat of global climate change. We also find that activity rates are more sensitive to temperature than are the physiological traits often used as a proxy for fitness. We present a model of thermal constraint on activity that integrates aspects of both the threshold model and the continuous-activity model, the general features of which are supported by activity data from other species. Overall, our results demonstrate that greater attention should be given to fine-scale patterns of thermal constraint on activity.