For decades, scientists have expected that the shedding of leaves from temperate trees will get later and later under ongoing climate change. And early observations supported that idea, as warming caused leaves to stay on the trees later over recent decades, driving increased growing season length that could help to slow the rate of climate change. However, a large-scale study of European trees now suggests that this trend is beginning to change, and in fact, tree leaves may start to fall earlier as the productivity of those trees increases. The results build on growing evidence that plant growth is limited by the ability of tree tissues to use and store carbon. While changes in the growing-season lengths of temperate trees greatly affect global carbon balance, future growing-season trajectories remain highly uncertain because the environmental drivers of autumn leaf senescence are poorly understood. Autumn leaf-shedding at the end of the growing season in temperate regions is an adaptation to stressors, such as freezing temperatures. A common related assumption is that alleviating some of these stressors - as a warmer climate could - would allow leaves to persist longer to fix more atmospheric carbon by photosynthesis. However, the role of photosynthesis in governing the timing of leaf senescence has not been widely tested in trees. To do this, Deborah Zani and colleagues used long-term observations from dominant Central European tree species from 1948 to 2015, and experiments designed to modify carbon uptake by trees, to evaluate related impacts on senescence. Collectively, their data show that increased growing-season productivity in spring and summer due to elevated carbon dioxide, temperature, or light levels can lead to earlier - not later - leaf senescence. This is likely because roots and wood cease to use or store leaf-captured carbon at a point, making leaves costly to keep. The authors used their observations to build a model to improve autumn senescence prediction under a business-as-usual climate scenario. It forecasts the possibility of slight advances, not delays, in autumn leaf-dropping dates over the rest of the century. The results "substantially lower our expectations of the extent to which longer growing seasons will increase seasonal carbon uptake in forests," they write, though the universality of this pattern in other forest types remains unknown. They note an important next avenue of research is implementing such growing-season length constraints in Earth system and vegetation models, which currently do not consider these dynamics when predicting seasonal carbon dioxide uptake of plants. A related Perspective discusses the results in more detail.