Salmonid fishes are known to inhabit streams with ambient summer temperatures approaching or exceeding lethal limits. Under these conditions, localized areas of cool water facilitate the persistence of coldwater fishes, but these may be altered in regulated rivers. In this study we examined the variation in brown trout behavioral thermoregulation within three streams of the Hudson River drainage, the Cedar, Indian and Hudson Rivers, the latter two of which are impacted by recreational discharge events from an upstream dam that supports a summer whitewater rafting industry. We were particularly interested in evaluating the potential of thermal refugia dilution by these flow releases. Based on both laboratory derived tolerance values and field-based thresholds that incorporated metrics of temperature magnitude, duration, and fluctuation, all three rivers were thermally marginal for brown trout during the summers of 2005 and 2006. Behavioral thermoregulation was observed by adult brown trout in our study in all river reaches, albeit infrequently in the Indian River. We found that brown trout in the Cedar River (38%) were more often observed with body temperatures cooler than ambient river temperature than those in either the Hudson (29%) or Indian Rivers (4%). Fewer than 50% of stocked fish persisted over a 67 day period in all three of our study reaches. Persistence of stocked brown trout in the Cedar River in 2006 was greater than in either the Hudson or Indian Rivers in both 2005 and 2006. While recreational discharge events did not alter the mean or maximum daily temperature in either the mainstem Indian or Hudson River reaches, the patches of relatively cool water near tributary confluences were diluted by release events. Both daily temperature maxima and ranges increased significantly at these locations in concert with recreational flow releases. Recreational flow releases were not an important factor accounting for the thermal behavior of brown trout during any time period in the reference Cedar River reach (without dam releases); however, behavioral thermoregulation was reduced during flow releases in both the Indian and Hudson River reaches. In the absence of recreational discharge events, the most important factors affecting behavioral thermoregulation were whether a fish was located near a tributary confluence and the ambient river temperature. Brown trout were consistently cooler relative to ambient river temperature when located near tributaries during times when river temperature was within the upper critical range for brown trout. Behavioral thermoregulation increased as river thermal conditions became more stressful. Our results suggest that accessible thermal refuge areas are important resources that provide brown trout a haven from lethal summer temperature conditions in thermally marginal streams, such as these study reaches in the Upper Hudson River drainage. When low flow conditions correspond with peak summer temperatures, these refuge areas are likely most important and most vulnerable to altered flow regimes. Our results showed that pulsed discharge events altered both the thermal characteristics of refuge areas at tributary confluences and behavioral thermoregulation by stocked brown trout. Although poor survival of these trout in the affected reaches may be due to severe summer temperatures regardless of recreational releases, the observed reduction in behavioral thermoregulation suggests that pulsed discharge events may impair the ability of coldwater fish to survive in regulated systems.