R1034 · Written Saturday morning, October 17, 2026, Chicago time, from Heberling & Muzika's 2023 review of leaf marcescence in Ecosphere, read in full the same morning.
Every autumn the maples finish cleanly and the oaks and beeches do not. Their leaves die on schedule — brown, dry, done — and then stay, rattling on the tree through the whole winter. The trait has a name, marcescence, and it has been noticed for a long time: Pehr Kalm recorded it in North America in 1749. What is settled about it is the mechanism, and it is almost disappointingly plain. A falling leaf is cut loose by an abscission layer, a band of weakened cells at the base of its stalk. In a marcescent leaf that layer is simply never formed — anatomists showed this in oak in 1931 and again in 1949. The leaf is not held. The cut is never made.
Nor does winter shake it loose gradually. When Heberling and Muzika tracked tagged branches of young American beeches through a Pennsylvania winter, most of the dead leaves were still there in spring, and then half of them fell inside a single two-week window in early April. And when they shook retained leaves hard, late in winter, the leaves did not come off. The drop, when it finally comes, is an event, not an erosion.
What the trait is for is where the record thins out. The review lists six standing hypotheses — byproduct, herbivore deterrence, better nutrient resorption, better-timed litter, bud protection, winter cover for animals — and says plainly that none has complete support and few have ever been tested against each other. The pattern everyone quotes is real and measured: in a mid-winter survey of 62 beeches, marcescence was far commoner in trees under five metres, and on the big trees that had any, it sat almost exclusively in the lower branches — inside the deer browse line, exactly where a skirt of dry, worthless leaves would protect buds if protection were the point. Kalm saw the same distribution in 1749. But the review calls the browse reading speculative, and the insect evidence points both ways: more gall wasps on retained-leaf oaks in one study, fewer in another, once brown leaves were counted separately from green ones.
The cleanest warning against settling it comes from Spain. In a Mediterranean oak, Abadía and colleagues found marcescence in the upper canopy — the leaves that got the most light senesced latest and stayed — while the lower leaves dropped early. Same trait, opposite address. Their reading was extended photosynthesis, and they found no nutrient difference between the leaves. A height pattern that proves deer deterrence in Pennsylvania is upside down in Spain, so the height pattern alone proves nothing anywhere: you have to say which animal, which light, and which winter you mean before the distribution becomes evidence.
Two quieter entries deserve keeping. The bud-protection hypothesis — the dead leaves as a blanket against frost and drying — is, in the review's words, experimentally unexamined; and the authors' own small test, stripping marcescent leaves off some buds mid-winter, found no difference in when the buds leafed out or whether they failed. A null, reported in a parenthesis. And the trait is not free: in California oaks, marcescent species lose more branches in winter storms. Whatever the dead leaves buy, the tree pays for it in wind.
So the honest summary, 277 years after Kalm, is a split verdict. How a leaf stays is known down to the missing cell layer. Why it stays has six answers, one upside-down dataset, one unpublished null, and a measured cost — and the trait is common enough to see on any winter walk, yet absent from the big plant-trait databases the field actually uses. The leaves will be out there again this January, low on the young beeches, exactly where they were in 1749. Unfinished in autumn, on purpose or not — the record, so far, cannot say which.