1.In an extensive field study, we tested the hypothesis that warming during a sensitive period alters trophic interactions in pelagic food webs of dimictic lakes: Matching of predation on the key herbivore Daphnia by vertebrate (Perca fluviatilis) and invertebrate (Leptodora kindtii) predators would destabilise Daphnia populations during summer. To predict food-web-mediated effects of climate warming, we relied on phase-specific warming trends (during winter, early thermal stratification, summer stagnation) of an 11-year (19992009) period to study instantaneous or time-delayed ecological responses. 2.Warming during the period of early thermal stratification (here corresponding to May) rather than during winter or summer was found to generate complex, time-delayed cascading effects on Daphnia via timing and rates of predation. In contrast to our hypothesis, warming exceeding a critical mean temperature (14 degrees C) during May increased the stability of the Daphnia population during summer. 3.The termination of predation on Daphnia by older perch and L.kindtii was advanced even stronger (14days per degree Celsius warming during May) than the start of predation. Warming during May thus shortened the period of synchronised predation on Daphnia by older perch, young-of-the-year perch and L.kindtii (match index) by up to 4weeks, thereby decreasing total rates of predation on Daphnia during July by up to 40%. 4.The timing of the top-down control of Leptodora dynamics by perch proved to be a key process for population stability of Daphnia during summer compared to the less important direct cascading effect of fish on Daphnia and temperature effects on Leptodora growth and consumption or on Daphnia reproduction. 5.We conclude that predicting the consequences of climate warming for plankton dynamics requires a mechanistic understanding of lake-specific temperature-driven changes in trophic cascades, especially when an invertebrate predator is dynamically coupled both to the key herbivore and to their predators.