What it is
The smallest population that has a high probability of persisting for the long term in the face of random shocks.
The concept came from Mark Shaffer's 1981 paper, which proposed 50/500 as a rough rule. 50 breeding adults to avoid inbreeding depression in the short term. 500 to maintain enough genetic variation to adapt over generations. Modern updates push the long-term number to 1,000 or 5,000 for vertebrates.
Below those thresholds, a population can look healthy in a single survey and still be sliding toward local extinction. The decline is genetic before it is demographic.
Why it works
Four threats compound below the threshold.
Inbreeding depression. Close-relative mating exposes recessive harmful alleles. Florida panthers crashed to 20 individuals in the 1990s with kinked tails, undescended testicles, and heart defects. Outcrossing with Texas pumas in 1995 rescued the population almost overnight.
Genetic drift. In small populations, random sampling loses rare alleles each generation. Once gone, they cannot come back without immigration. The toolkit for future adaptation shrinks.
Demographic stochasticity. Random variation in births and deaths can sink a small population by chance. With 30 individuals and a balanced sex ratio, a run of male-heavy litters across three years can leave too few females to recover.
Environmental catastrophe. One fire, one flood, one disease outbreak takes out a small isolated population. The Tasmanian devil facial tumour disease swept through low-diversity populations precisely because they could not mount varied immune responses.
A connected population of 5,000 across a wide range absorbs all four. An isolated 50 cannot.
Real numbers
Cheetahs. Roughly 7,100 left in fragmented populations. They went through a genetic bottleneck around 10,000 years ago and now show almost no genetic variation between individuals. Skin grafts between unrelated cheetahs are not rejected, the way grafts between identical twins behave in humans. Fertility problems and high cub mortality follow.
Wollemi pine. Discovered in 1994 in a single canyon in Australia with under 100 mature trees. Genetic analysis shows zero variation between individuals. Survival rests entirely on the ex-situ propagation programme that distributed clones worldwide.
Iberian lynx. Down to 94 individuals in 2002, split between two isolated populations. Captive breeding plus corridor restoration has pushed numbers above 2,000 by 2024, but inbreeding signals remain.
Florida panther. The 1995 Texas puma intervention saved the lineage. Hybrid offspring showed triple the survival rate of inbred kittens. Texas blood still flows in every Florida cat today.
Design for it
You do not save a species by protecting one reserve. You save it by stitching populations together so genes flow.
Corridors over patches. A 50 ha woodland connected to a 200 ha block by a hedgerow corridor functions as a 250 ha population. Disconnect them and you have two doomed populations. See wildlife corridors.
Stepping stones. When a continuous corridor is impossible, plant a chain of patches 1 to 5 km apart. Most temperate forest birds and mammals will cross gaps of that scale through farmland and gardens. Beyond 10 km the gene flow drops to near zero.
Translocation. When natural movement is blocked (highways, ocean, dam), move animals deliberately. Two to four introductions per generation are usually enough to maintain genetic health. The European bison programme in Bialowieza Forest has done this since 1929, taking 12 captive founders to over 7,000 free-ranging animals across Europe.
Plant populations. The same logic applies to trees and herbs. Collect seed from at least 50 unrelated mother plants across the source population. Plant the seedlings mixed, not segregated by mother. Single-source plantings carry hidden inbreeding load that shows up generations later. See seed banking.
When it goes wrong
Reserve too small. Even with perfect habitat, a 100 ha patch will lose its top predators and large herbivores. They need range, mates, and gene flow. Plan reserves at the landscape scale or accept that you are building a habitat museum, not a functioning ecosystem.
One population, one event. California condors crashed to 22 birds in 1987 partly because all were in one area facing the same lead-poisoning threat. Distributed populations buffer single-event risks. Build two or three sub-populations from the start.
Connectivity exists on paper, not on the ground. A "corridor" with a six-lane highway through it is a barrier. Test connectivity with camera traps, GPS collars, or genetic samples before you trust the map.
Founder effect ignored. Reintroductions from a handful of captive animals carry only a slice of the original gene pool. Source widely. Mix lineages. See seed banking and native species.