Periods of extremely hot and cold weather can cause significant mortality and morbidity in both temperate and more extreme climates. In the UK, their occurrence prompts the issuing of number and colour coded warnings providing an assessment of the level of risk. These are designed to minimize health impact by prompting timely and appropriate mitigating actions by the public. Drawing upon the interdisciplinary notion of framing, I report on a study that identified a central role for language in mediating how these warnings are interpreted and evaluated. I use an innovative approach that combines the quantitative tools of corpus linguistics to identify the language used to represent warnings and the risks of extreme temperature in the mass media, with qualitative analysis of focus group discussions of typical texts. A comparison of both datasets indicates a multi-layered interactivity between the myriad ways in which language can give salience to aspects of a risk scenario and an interpreter's knowledge and perception of a threat, and that underlying such interactions, is the conceptualization of risk as scalar property.