Biological effects of climate change are expected to vary geographically, with a strong signature of latitude. For ectothermic animals, there is systematic latitudinal variation in the relationship between climate and thermal performance curves, which describe the relationship between temperature and an organism's fitness. Here we ask whether these documented latitudinal patterns can be generalized to predict arthropod responses to warming across mid and high temperate latitudes, for taxa whose thermal physiology has not been measured. To address this question, we used a novel natural experiment consisting of a series of urban warming gradients at different latitudes. Specifically, we sampled arthropods from a single common street tree species across temperature gradients in 4 US cities, located from 35.8 to 42.4° latitude. We captured 6746 arthropods in 34 families from 111 sites that varied in summer average temperature by 1.7 to 3.4 °C within each city. Arthropod responses to warming within each city were characterized as Poisson regression coefficients describing change in abundance per °C for each family. Family responses in the two mid latitude cities were heterogeneous, including significantly negative and positive effects, while those in high latitude cities varied no more than expected by chance within each city. We expected high-latitude taxa to increase in abundance with warming, and they did so in 1 of the 2 high latitude cities; in the other, Queens (New York City), most taxa declined with warming, perhaps due to habitat loss that was correlated with warming in this city. With the exception of Queens, patterns of family responses to warming were consistent with predictions based on known latitudinal patterns in arthropod physiology relative to regional climate. Heterogeneous responses in mid latitudes may be ecologically disruptive if interacting taxa respond oppositely to warming. This article is protected by copyright. All rights reserved.