Terrestrial ecosystems currently offset one-quarter of anthropogenic carbon dioxide (CO2) emissions because of a slight imbalance between global terrestrial photosynthesis and respiration. Understanding what controls these two biological fluxes is therefore crucial to predicting climate change. Yet there is no way of directly measuring the photosynthesis or daytime respiration of a whole ecosystem of interacting organisms; instead, these fluxes are generally inferred from measurements of net ecosystem-atmosphere CO2 exchange (NEE), in a way that is based on assumed ecosystem-scale responses to the environment. The consequent view of temperate deciduous forests (an important CO2 sink) is that, first, ecosystem respiration is greater during the day than at night; and second, ecosystem photosynthetic light-use efficiency peaks after leaf expansion in spring and then declines, presumably because of leaf ageing or water stress. This view has underlain the development of terrestrial biosphere models used in climate prediction and of remote sensing indices of global biosphere productivity. Here, we use new isotopic instrumentation to determine ecosystem photosynthesis and daytime respiration in a temperate deciduous forest over a three-year period. We find that ecosystem respiration is lower during the day than at night—the first robust evidence of the inhibition of leaf respiration by light at the ecosystem scale. Because they do not capture this effect, standard approaches overestimate ecosystem photosynthesis and daytime respiration in the first half of the growing season at our site, and inaccurately portray ecosystem photosynthetic light-use efficiency. These findings revise our understanding of forest-atmosphere carbon exchange, and provide a basis for investigating how leaf-level physiological dynamics manifest at the canopy scale in other ecosystems.