The extent to which populations will successfully adapt to continued warming temperatures will be a crucial factor in determining future health burdens. Previous health impact assessments of future temperature-related mortality burdens mostly disregard adaptation or make simplistic assumptions. We apply a novel evidence-based approach to model adaptation that takes into account the fact that adaptation potential is likely to vary at different temperatures. Temporal changes in age-specific mortality risk associated with low and high temperatures were characterised for Scotland between 1974 and 2018 using temperature-specific RR ratios to reflect past changes in adaptive capacity. Three scenarios of future adaption were constructed consistent with the SSPs. These adaptation projections were combined with climate and population projections to estimate the mortality burdens attributable to high (above the 90th percentile of the historical temperature distribution) and low (below the 10th percentile) temperatures up to 2080 under five RCP-SSP scenarios. A decomposition analysis was conducted to attribute the change in the mortality burden into adaptation, climate and population. In 1980-2000, the heat burden (21 deaths/year) was smaller than the colder burden (312 deaths/year). In the 2060-2080 period, the heat burden was projected to be the highest under RCP8.5-SSP5 (1285 deaths/year), and the cold burden was the highest under RCP4.5-SSP4 (320 deaths/year). The net burden was lowest under RCP2.6-SSP1 and highest under RCP8.5-SSP5. Improvements in adaptation was the largest factor reducing the cold burden under RCP2.6-SSP1 whilst temperature increase was the biggest factor contributing to the high heat burdens under RCP8.5-SSP5. Ambient heat will become a more important health determinant than cold in Scotland under all climate change and socio-economic scenarios. Adaptive capacity will not fully counter projected increases in heat deaths, underscoring the need for more ambitious climate mitigation measures for Scotland and elsewhere.