Mast-seeding plants often produce high seed crops the year after a warm spring or summer, but the warm-temperature model has inconsistent predictive ability. Here, we show for 26 long-term data sets from five plant families that the temperature difference between the two previous summers (ΔT) better predicts seed crops. This discovery explains how masting species tailor their flowering patterns to sites across altitudinal temperature gradients; predicts that masting will be unaffected by increasing mean temperatures under climate change; improves prediction of impacts on seed consumers; demonstrates that strongly masting species are hypersensitive to climate; explains the rarity of consecutive high-seed years without invoking resource constraints; and generates hypotheses about physiological mechanisms in plants and insect seed predators. For plants, ΔT has many attributes of an ideal cue. This temperature-difference model clarifies our understanding of mast seeding under environmental change, and could also be applied to other cues, such as rainfall.