The interpretation of thermal-gradient data depends on the behavioral drives reported or assumed, and on the underlying behavioral models explaining how such drives operate. The best-known example is positive thermotaxis, a thermoregulatory behavioral drive frequently linked to a dual set-point model of thermoregulation around a target range. This behavioral drive is often assumed as dominant among 'ectotherms', including amphibians. However, we argue that, because amphibians are extremely diverse, they may exhibit alternative behavioral drives in thermal gradients, and tackle this idea from two perspectives. First, we provide a historical review of original definitions and proposed limits for inference. Second, although caveats apply, we propose that a cross-study analysis of data of temperature settings of gradients and the temperatures selected by amphibians would corroborate alternative behavioral drives, including negative thermotaxis. Therefore, we analyzed published data focusing on such relationships and show that gradient temperature settings influence the temperatures selected by amphibians, with further effects of phylogeny and ontogeny. We conclude that thermal gradient experiments are outstanding tools to investigate behavioral drives, but no given drive can be assumed a priori unless additional information about thermoregulation is available. Based on the historical debate, we propose using selected temperatures and preferred temperatures as different concepts, the former merely operational and the second explicitly linked to positive thermotaxis (and thus compatible with dual set-point thermoregulation). Under this view, thermal preferences would stand for a hypothesis of a behavioral drive (positive thermotaxis) requiring formal testing. These considerations impact the scope for inference based on thermal gradient experiments, particularly ecological modeling and emerging disease.