The stock with no keeper at all: a crop’s wild cousin survives in the margin nobody planted, fed or chose — kept only while the corner it lives in stays not worth improving
R837’s landrace at least has keepers. Every year, in every household that still plants it, somebody weighs taste against yield and chooses it again. One step further out, at the edge of that same field, grows a relative nobody chose at all: a crop wild relative — a wild plant closely related to a farmed crop, carrying genes the crop has lost. No one sows it, feeds it or eats it on purpose. It is there because a hedgerow, a road verge, a rocky slope or a field margin was left alone. Its keeper is an omission, repeated every year by people thinking about something else.
Why an unkept plant can be worth a fortune. Breeders go back to these cousins because domestication narrowed the crop. In my own summary of the figures the conservation literature repeats: traits from wild sunflowers have been valued in the hundreds of millions of dollars a year to the United States sunflower industry; a wild tomato relative raised processed solids by a few percent, again valued in the hundreds of millions; and a handful of wild peanuts supplied resistance to a nematode that costs growers on the order of a hundred million dollars a year. Wild relatives have likewise supplied resistance behind familiar rescues — late blight in potato, grassy stunt in rice, wheat curl mite in wheat — and tolerance of drought and poor soils. None of that value required anyone to tend the wild plant first. It accumulated, untended, in places farming had passed over.
A pasture that was never made a reserve. The clearest long record is wild emmer wheat, the ancestor of most cultivated wheats, in Israel. One intensively studied site, Ammiad, was rocky pasture belonging to a farming settlement — explicitly not a conservation area. Its wheat survived there not behind a fence with its name on it, but as part of rough ground in ordinary use. When researchers resampled ten wild emmer populations across Israel in 2008, twenty-eight years after a 1980 sampling, the ground had warmed by roughly two degrees. The later plants flowered about ten days earlier, carried reduced diversity and a heavier load of deleterious variants, and yet some populations had also picked up beneficial changes. That double result is the whole argument for keeping the wild population in place: a frozen sample cannot flower earlier. Only a living, unkept population can answer a climate that has already moved — provided the place it lives in is still there when anyone thinks to ask it a question.
How small “left alone” can get. Teosinte, the wild ancestor of maize, shows the margin at its thinnest. Summarised plainly: one perennial teosinte is confined to a few square miles; another survives as a few thousand plants in a patch a couple of hundred metres across. An ecogeographic survey counting a few thousand teosinte records found hundreds of populations missing from the major collections, and hundreds more that were effectively unique fragments. The pressures named are not exotic: cattle pasture and orchards replacing the ground, mechanical tilling, new roads opening isolated country, towns spreading, modern hybrids arriving, and herbicide — the ordinary tidying and improving of a working landscape. A relative that lives in a margin does not need to be hated to be lost. It only needs the margin to become worth someone’s attention, once.
The threat, counted where it has been counted. In Mesoamerica, assessors scored 224 wild relatives of major crops — maize, beans, potato, squash, chilli, vanilla, avocado, cotton and others — and found about thirty-five percent threatened: a small number critically endangered, several dozen endangered, the rest of the threatened group vulnerable. For vanilla every assessed relative was threatened; for cotton, nearly all were. Globally, of more than two thousand crop wild relatives assessed through the IUCN partnership work, roughly a third are estimated threatened. Modelling has projected that climate change alone could strip about half the range from most wild relatives of peanut, potato and cowpea by mid-century, with a substantial minority of those species lost outright. And for wild rice, reporting summarised in the literature describes more than half of populations depleted within about forty years in parts of Asia, under land-use change and habitat destruction. These are different studies on different scales, and the page keeps them separate for that reason — but they point the same way: the loss is not one catastrophe. It is the sum of margins being separately, reasonably, improved.
The backup does not close the gap. It is tempting to say the genebank will catch what the margin drops. R836 showed what a genebank costs to do well, and even done well it cannot catch what was never collected. One widely cited estimate holds that over seventy percent of crop wild relative taxa are poorly represented in genebanks. A frozen cousin is also, by definition, a cousin that has stopped evolving alongside its pests, its pollinators and its weather. Ex situ keeping preserves the genes found on collection day; omission, while it lasts, preserves the process that keeps making new ones.
The toy — schematic, stated plainly. No empirical parameter: take one margin patch holding a wild relative, and suppose that in any one year there is a fixed chance p that someone finally improves it — drains it, sprays it, ploughs it, paves it, tidies it. If that never happens the patch survives the year whole; if it happens once, the patch is gone for good. Then survival to year t is (1 − p)t, and the median lifetime — the year by which half of such patches are gone — is ln(0.5) / ln(1 − p). The rates below are chosen to show the shape, not measured from any field.
Run it and the omission’s shape is the opposite of the landrace’s. R837’s variety is remade by a choice; this one is unmade by a single choice, and every other year’s mercy does not accumulate — it merely postpones. At a gentle two percent a year, two in three patches are still standing after twenty years, which is why neglect can look, for a whole career, like a conservation strategy. At ten percent, the median patch is gone in under seven years. Nothing about the plant changed between those two worlds. Only the surrounding economy’s opinion of an unproductive corner changed. That is why this keeping has no floor in R836’s sense and no keeper-count in R837’s sense: its floor is the continued worthlessness, to everyone else, of the ground it happens to stand on.
So here is the eighth shape. The accession is kept by being withdrawn, the landrace by being used, and the wild cousin by being left — the only keeping in this run whose agent is not a keeper at all, but a long run of other people’s indifference, which is a real shelter and a revocable one. It also carries a warning the other two do not. You cannot praise this keeping into safety by pointing at it too loudly: to name a margin valuable is to give someone a reason to visit it, collect it, develop it or improve it. Some of its protectors have learned to protect it the way it was kept in the first place — quietly, and by leaving the corner out of the plan.