The role of temperature on biological activities and the correspondent exponential relationship with temperature has been known for over a century. However, lacking to date is knowledge relating to (a) the recovery of ectotherms subjected to extreme temperatures in the wild, and (b) the effects repeated extreme temperatures have on the temperatures that induce behavioural thermoregulation (aggregations). We examined these questions by testing the hypothesis that thermal thresholds which initiate aggregations in juvenile Atlantic salmon (AS) (Salmo salar) are not static, but are temporally dynamic across a summer and follow a hysteresis loop. To test our hypothesis, we deployed custom-made underwater camera (UWC) systems in known AS thermal refuges to observe the timing of aggregation events in a natural system and used these data to develop and test models that predict the temperatures that induce thermal aggregations. Consistent with our hypothesis our UWC observations revealed a range of aggregation onset temperatures (AOT) ranging from 24.2°C to 27.1°C, thus confirming our hypothesis that AOTs are dynamic across summer. Our models suggest it take ~ 11days of non-thermally taxing temperatures for the AOT to rebound in the study river. Conversely, we found that as the frequency of events increased, the AOT declined, from 27.1°C to 24.2°C. Integrating both model components led to more robust model performance. Further, when these models were tested against an independent data set from the same river, the results remained robust. Our findings illustrate the complexity underlying behavioural thermoregulation in AS-a complexity that most likely extends to other salmonids. The frequency of extreme heat events is predicted to increase, and this has the capacity to decrease AOT thresholds in AS, ultimately reducing their resilience to extreme temperature events.