Previous studies of climate change impacts on stream fish distributions commonly project the potential patterns of habitat loss and fragmentation due to elevated stream temperatures at a broad spatial scale (e.g. across regions or an entire species range). However, these studies may overlook potential heterogeneity in climate change vulnerability within local stream networks. We examined fine-scale stream temperature patterns in two headwater brook trout Salvelinus fontinalis stream networks (7.7 and 4.4 km) in Connecticut, USA, by placing a combined total of 36 pairs of stream and air temperature loggers that were approximately 300 m apart from each other. Data were collected hourly from March to October 2010. The summer of 2010 was hot (the second hottest on record) and had well below average precipitation, but stream temperature was comparable with those of previous 2 years because streamflow was dominated by groundwater during base-flow conditions. Nonlinear regression models revealed stream temperature variation within local stream networks, particularly during warmest hours of the day (i.e. late afternoon to evening) during summer. Thermal variability was primarily observed between stream segments, versus within a stream segment (i.e. from confluence to confluence). Several cold tributaries were identified in which stream temperature was much less responsive to air temperature. Our findings suggested that regional models of stream temperature would not fully capture thermal variation at the local scale and may misrepresent thermal resilience of stream networks. Groundwater appeared to play a major role in creating the fine-scale spatial thermal variation, and characterizing this thermal variation is needed for assessing climate change impacts on headwater species accurately. Copyright © 2013 John Wiley & Sons, Ltd.