1. We asked whether an increase in food supply in the field would increase the ability of fish populations to withstand climate warming, as predicted by certain bioenergetic models and aquarium experiments. 2. We subsidised the in situ food supply of wild juvenile steelhead (Oncorhynchus mykiss) in a small stream near the species' southern limit. High-quality food (10% of fish biomass per day) was added to the drift in eight in-stream enclosures along a naturally-occurring thermal gradient. 3. The temperatures during the experiment were well below the upper thermal limit for the species (means of enclosures ranged from 15.1 to 16.5°C). Food supplements had no discernible effect on survival, but raised mean (± SD) specific growth rate substantially, from 0.038 ± 0.135 in controls to 2.28 ± 0.51 in feeding treatments. Food supplements doubled the variation in growth among fish. 4. The mean and variance of water temperature were correlated across the enclosures, and were therefore transformed into principal component scores T 1 (which expressed the stream-wide correlation pattern) and T 2 (which expressed local departures from the pattern). Even though T 1 accounted for 96% of the variation in temperature mean and variance, it was not a significant predictor of fish growth. T 2 was a significant predictor of growth. The predicted time to double body mass in an enclosure with a large T 2 score (cool-variable) was half that in an enclosure with a low T 2 score (warm-stable). 5. Contrary to expectation, temperature effects were neutral, at least with respect to the main axis of variation among enclosures (cool-stable versus warm-variable). Along the orthogonal axis (cool-variable versus warm-stable), the effect was opposite from expectations, probably because of temperature variation. Subtle patterns of temperature heterogeneity in streams can be important to potential growth of O. mykiss.