How many plants should you save seed from?
Grow & Cook, AI-assisted writing ยท
How many plants should you save seed from? Fewer than most gardeners think is safe. Saving from the three best plants each year feels careful, yet after ten years it keeps only about 16% of the variety's genetic variety; twenty plants keep about 78% [1]. The difference is arithmetic that plant breeders have used for a century [2].
A seed jar, year after year
Picture the variety as a jar of seeds in twelve colours, one for each version of some gene the variety carries. Each year you grow plants from the jar and save seed from a few. Those few plants carry only some of the colours, two copies each, and next year's jar holds only what they carried. Nobody chooses which colours drop out; it is the luck of which plants you picked. Biologists call it genetic drift [2].
Each year of saving from N plants keeps, on average, a share of 1 − 1/(2N) of the variety's genetic variety [2]. With three plants that is five sixths a year, which sounds fine until it compounds: (5/6) to the tenth power is about 0.16 [1].
| Plants saved from each year | Variety kept after 10 years | After 20 years |
|---|---|---|
| 3 | 16% | 3% [1] |
| 5 | 35% | 12% [1] |
| 10 | 60% | 36% [1] |
| 20 | 78% | 60% [1] |
| 50 | 90% | 82% [1] |
The rare one you never noticed
Variety is not abstract. A version of a gene carried in one seed-gene in twenty might be the one that copes with a wet summer, or with a mildew your garden has not had yet. Saving from N plants passes on 2N copies of each gene, so the chance that a version with frequency q makes it into next year's jar is 1 − (1 − q)^(2N) [2]. From three plants that is about 26%; from twenty, about 87% [1]. Keeping it with 90% certainty takes 23 plants [1].
Squash is not tomato
The arithmetic above assumes plants cross freely, as squash, corn and the cabbage family largely do. Tomatoes and beans mostly pollinate themselves, so each plant is close to a fixed line and fewer plants keep the variety true to type [3]. The same arithmetic still decides how many of the variety's lines survive, so saving from more than one or two plants still pays. Saving from only the single best plant each year narrows a variety fastest of all: one plant carries at most two versions of each gene, so selecting hard for one trait drops much of what came along with the others.
Check yourself
You go from saving 5 plants to 10. Roughly how much more variety is left after ten years? About 60% instead of 35%: doubling the plants nearly doubles what survives [1].
Why can saving from only the very best plant shrink a variety faster than saving at random? Because one plant carries at most two versions of each gene, so everything else it happens to carry narrows along with the trait you picked for [2].
The squash is an illustration. This seed-shelf entry for growandcook.com was written with AI and checked by hand; every figure was computed from the formulas shown, and the box above recomputes them for your garden.