The Arctic is warming rapidly due to climate change, causing landscape changes such as permafrost thaw that could shift the regional carbon balance from a net sink to a net source. Land surface models are necessary to predict the future carbon balance, however, they have large uncertainties in simulating permafrost carbon dynamics. Observations from experiments are a valuable tool to validate and refine permafrost representations within models, particularly though the use of functional relationships, which can inform for example how the response of carbon release with temperature will behave under warmer conditions. Here, we use the Tundra flux database consisting of 52 pan-Arctic experimental warming sites over a 19-year period from 2003-2021. Using these observations, we evaluate the response of ecosystem respiration (Reco) with temperature, and how the shape and magnitude of these responses is altered through experimental warming and changing environmental conditions. In doing this analysis, we also evaluate the best practices for the creation of robust functional relationships, including the impact that amount and temporal resolution of observations have on the temperature sensitivity of Reco. Across the Arctic, control sites indicate a peak in Reco (4.36 gC m^-2 d^-1) when soil temperatures reach ~20 °C, at which Reco then begins to decline. Reco responses from warmed plots using open top chambers exhibit a similar shape of response, but tend to show an increased magnitude of Reco relative to control plots, with a peak around 6.1 gC m^-2 d^-1. Individual sites also display a range in the shape and magnitude of Reco in response to temperature, likely due to site-level differences in soil moisture and vegetation presence. Overall, our findings provide evidence of an increased sensitivity in magnitude of Reco with temperature as climate warms. These results can be applied as benchmarks to test whether simulated responses of Reco with temperature from land surface models are capturing the shape and magnitude as seen by the observed responses.