The study of latitudinal gradients can yield important insights into adaptation to temperature stress. Two strategies are available: resistance by limiting damage, or tolerance by reducing the fitness consequences of damage. Here we studied latitudinal variation in resistance and tolerance to frost and heat and tested the prediction of a trade-off between the two strategies and their costliness. We raised plants of replicate maternal seed families from eight populations of North American Arabidopsis lyrata collected along a latitudinal gradient in climate chambers and exposed them repeatedly to either frost or heat stress, while a set of control plants grew under standard conditions. When control plants reached maximum rosette size, leaf samples were exposed to frost and heat stress, and electrolyte leakage (PEL) was measured and treated as an estimate of resistance. Difference in maximum rosette size between stressed and control plants was used as an estimate of tolerance. Northern populations were more frost resistant, and less heat resistant and less heat tolerant, but-unexpectedly-they were also less frost tolerant. Negative genetic correlations between resistance and tolerance to the same and different thermal stress were generally not significant, indicating only weak trade-offs. However, tolerance to frost was consistently accompanied by small size under control conditions, which may explain the non-adaptive latitudinal pattern for frost tolerance. Our results suggest that adaptation to frost and heat is not constrained by trade-offs between them. But the cost of frost tolerance in terms of plant size reduction may be important for the limits of species distributions and climate niches.