Although autumn leaf colours of deciduous trees have been shown to protect against photo-oxidative damage, they are sometimes seen as signals to pests and predators. Here I modify the coevolution hypothesis of autumn leaf colours. I suggest that much of the within-population variation in autumn leaf colours can be explained by differences in the allocation of resources to sexual reproduction. According to the novel hypothesis, reproductively active woody plants produce early and intense autumn leaf colours in order to protect seeds and other reproductive tissues from pests that lay eggs in the autumn. If many seeds mature at times of leaf senescence or during the next summer, a woody plant will reallocate plenty of nitrogen to seeds. If sucking insects reproduce on such hosts, their flightless offspring will suffer poor-quality food after the ripening of seeds. Before this, however, insects will probably concentrate around the ripening seeds to forage on nitrogen-rich veins. This will decline the quality and quantity of developing seeds. If, on the other hand, insects are able to recognize reproductively active plants while laying eggs in the autumn, both the insects and the plants benefit. The flightless offspring of insects feeds on plants that supply sufficiently nitrogen for longer than reproducing plants do, while these optimise their reproduction by avoiding pests, which also contributes to the abundance of specialist pests. Hence, I suppose that while physiological factors are the origin of autumnal colour changes of deciduous leaves, the visible cue utilized by insects has evolved several times to an honest signal that reveals the unsuitability of the potential host in the near future. The reproductive insurance hypothesis may help us to understand why bright autumn leaf colours are rare among herbaceous plants, and why plants at high altitudes and latitudes are often brightly coloured in autumn.