In recent decades, many climate manipulation experiments have investigated biosphere responses to global change. These experiments typically examined effects of elevated atmospheric CO₂, warming or drought (driver variables) on ecosystem processes such as the carbon and water cycle (response variables). Because experiments are inevitably constrained in the number of driver variables tested simultaneously, as well as in time and space, a key question is how results are scaled up to predict net ecosystem responses. In this review, we argue that there might be a general trend for the magnitude of the responses to decline with higher-order interactions, longer time periods and larger spatial scales. This means that on average, both positive and negative global change impacts on the biosphere might be dampened more than previously assumed.