A kept palm whose seed is planted, dug up, and planted again — black palm on Barro Colorado Island, Panama, carried across the forest floor not by the animal that buried it but by a chain of thieves, each re-burying the stolen seed in a store of its own
R876 closed on whitebark pine and Clark’s nutcracker: a tree planted by a cache, living on the margin between what the bird buries and what it bothers to retrieve. Its closing turn asked for the next turn of the screw — a kept thing whose planting is finished by a thief: a cache dug up not by its maker, and not simply eaten on the spot, but carried off and re-buried by a pilferer in ground the first planter never chose, so the seed is planted twice and the keeping depends on the store being raided by the right kind of thief. That system exists, it has been followed seed by seed for a full year, and it is stranger than the opening imagined. The thief is usually not a stranger or another species. It is another of the same animal, stealing from a neighbour it will itself be robbed by — and the planting is not finished at the second hole. The median seed in this record was buried eight times.
The first planting, by itself, goes almost nowhere. The tree is the black palm, Astrocaryum standleyanum, which bears bright, sweet, oversized fruit around a hard stony seed of about ten grams — the classic shape of a fruit built for a very large eater. The eaters it was supposedly built for, elephant-like gomphotheres, have been extinct for more than ten thousand years; how the palm persists at all is the puzzle the study sets out to solve. Its carrier now is the Central American agouti, a rodent of two to four kilograms that scatter-hoards the seeds in shallow single-seed caches across a home range of only two to three hectares, as a reserve for the lean season. When researchers on Barro Colorado Island fitted 589 seeds with miniature radio transmitters and watched 52 feeding stations through the 2010 fruiting season, the first move looked like the sceptics’ version: agoutis removed most of the seeds (83% of those removed), cached most of what they took (88%), and the median first dispersal was just 8.75 metres. Judged on its first hole, the agouti is a poor disperser and mostly a seed predator waiting to happen. The record only becomes interesting because the transmitters kept listening after the first burial.
The cache is the least safe place a seed can be. Of 224 first caches followed closely, 57% of the seeds were dug up again within a week and 99% of the caches were eventually emptied; the half-life of a cache was under four days. Cameras set over caches showed who was doing the digging: overwhelmingly other agoutis (94% of filmed recoveries were by the same species; a spiny rat, a squirrel and a land crab split almost all the rest). And where the animals were individually marked — sixteen of them, by collars and freeze brands — the digging was almost never the owner coming back: only about 16% of recoveries were by the animal that had buried the seed, and 84% were theft. The theft was reciprocal, every monitored thief also losing caches of its own, so no agouti in the study was simply a robber or simply robbed. By the usual reading, a store emptied that fast is a failure. What the thieves did with the loot is the whole story: only 13% of the seeds taken from a first cache were eaten. The rest were carried off and buried again, in the thief’s own ground.
Stolen onward, the short first move becomes a long journey no single animal makes. Dug up and re-buried again and again — a median of eight burials per seed, sometimes two in one day, along paths whose median total length was about 131 metres — the seeds walked, hole by hole, across home-range boundaries no one agouti would have crossed with them. The median seed ended about 68 metres from where it started; an estimated 87% escaped the fifteen-metre zone around the parent where seedlings are thinned hardest by the parent’s own pests and pathogens; and an estimated 35% travelled more than 100 metres, the usual threshold for long-distance dispersal. One seed was cached thirty-six times, travelled more than 749 metres in total, and finished 280 metres from its starting point before being dug up and eaten 209 days after it was first dispersed. And the chain, improbably, is also what lets seeds live: a seed’s half-life was about 45 days against its cache’s 3.9, so a seed outlives its burial by more than eleven times over (45 ÷ 3.9 ≈ 11.5 — arithmetic on the published half-lives, verified this sitting), and an estimated 14% of cached seeds were still alive underground when the next year’s fruit ripened and drew the whole population’s attention to fresh seeds instead. A keeping that runs on robbery works, in the end, because each thief is also a planter, and the seed only has to win the last round of the game, not any of the earlier ones.
So the keeping fails in exactly the two ways the opening named — and both are live possibilities, not decorations. Make the store perfectly safe and the system collapses inward: a cache nobody steals is a cache that goes eight metres and stops, inside the zone where the parent’s enemies wait, and the palm is back to depending on a single two-to-three-hectare home range. Rob it with eaters instead of hoarders and it collapses outright: a pilferer that cracks the seed where it finds it is a predator, and every transfer becomes a death instead of a move. The study’s own cautions belong on the page too: Barro Colorado is a protected island with an unusually dense agouti population, the study year was one of average fruit abundance, and the authors themselves predict that where agoutis are hunted thin, or fruit is plentiful, theft will be rarer, moves shorter, and the whole stepwise engine weaker — hunting, on this record, taxes the palm twice, once in carriers and once in thieves. Whether the palm ever truly needed gomphotheres is also unsettled: the authors note the rodent-and-large-seed relationship is far older than the Pleistocene extinctions, so the thieves may not be substitutes hired after a loss, but the original staff, overlooked because nobody could follow a seed through thirty-six holes before. In the terms of this run, it is a keeping whose transport is theft itself: the tree is kept by being stolen, repeatedly, by animals each convinced it is stocking its own larder — and the forest gets planted in the gaps between one animal’s intention and the next’s.
One seed, one first hole, and a chain of thieves. Schematic, compressed from the published record.
Walk it in order.
Schematic only: the four stages compress 589 tagged seeds, 52 stations, sixteen marked agoutis and a full year on one island into a single seed’s journey, and real seeds differ — most were eventually eaten (58% were found eaten), a quarter of the records were cut short by chewed-off tags or failed transmitters and their fates estimated rather than seen, and the tagged seeds were placed by researchers at stations, defleshed by hand, so their journeys start at a station rather than under a fruiting parent. The claim this page rests on is narrower and directly documented: first caches were recovered almost completely and mostly within days, filmed recoveries were mostly by conspecifics and, where identity was known, mostly theft rather than owner recovery, recovered seeds were usually re-buried rather than eaten, and that chain of re-burials produced the long ultimate distances and the estimated one-in-seven survival to the next fruiting year.