Biotic responses to climate change will vary among taxa and across latitudes, elevational gradients, and degrees of insularity. However, due to factors such as phenotypic plasticity, ecotypic variation, and evolved tolerance to thermal stress, it remains poorly understood whether losses should be greatest in populations experiencing the greatest climatic change or living in places where the prevailing climate is closest to the edge of the species' bioclimatic envelope (e.g., at the hottest, driest sites). Research on American pikas (Ochotona princeps) in montane areas of the Great Basin during 1994-1999 suggested that 20th-century population extirpations were predicted by a combination of biogeographic, anthropogenic, and especially climatic factors. Surveys during 2005-2007 documented additional extirpations and within-site shifts of pika distributions at remaining sites. To evaluate the evidence in support of alternative hypotheses involving effects of thermal stress on pikas, we placed temperature sensors at 156 locations within pika habitats in the vicinity of 25 sites with historical records of pikas in the Basin. We related these time series of sensor data to data on ambient temperature from weather stations within the Historical Climate Network. We then used these highly correlated relationships, combined with long-term data from the same weather stations, to hindcast temperatures within pika habitats from 1945 through 2006. To explain patterns of loss, we posited three alternative classes of direct thermal stress: (1) acute cold stress (number of days below a threshold temperature); (2) acute heat stress (number of days above a threshold, temperature); and. (3) chronic heat stress (average summer temperature). Climate change was defined as change in our thermal metrics between two 31-y.r periods: 1945-1975 and 1976-2006. We found that patterns of persistence were well predicted by metrics of climate. Our best models suggest some effects of climate change; however, recent and long-term metrics of chronic heat stress and acute cold stress, neither previously recognized as sources of stress for pikas, were some of the best predictors of pika persistence. Results illustrate that extremely rapid distributional shifts can be explained by climatic influences and have implications for conservation topics such as reintroductions and early-warning indicators. ?? 2010 by the Ecological society of America.