Most bird and mammal species that received hands-on conservation after falling below 50 animals in the wild later reached at least 100, according to a study published online on 4 November 2026 in Proceedings of the Royal Society B. Species left without that direct help almost never did. The comparison is observational. It does not prove that any one project caused a recovery, and it does not show that a small population is safe.

What was counted

Donal Smith, of Monash University, and colleagues at Monash and the Zoological Society of London reviewed every bird and mammal species known to have dropped below 50 wild individuals at any point since 1950. They started from 147 species and set aside six whose decline was too recent to judge, leaving 141: 96 birds and 45 mammals. Population figures came first from IUCN Red List assessments, then from other published accounts. The authors used broad size bands because old counts are uneven, and they took the low end of any range.

They split conservation into direct and indirect work. Direct work either tackled a limit on the population, such as predators, habitat or nest sites, or handled the animals themselves through captive breeding, translocation or release. Laws, park designations and publicity campaigns were coded as indirect. The paper says those steps may help, but they do not by themselves put animals back into a habitat.

Of the 141 species, 51 received direct conservation. Thirty-one of those 51, or 61%, later reached at least 100 individuals, and six, or 12%, went extinct. Of the 90 with no direct conservation, two, about 2%, reached 100, and 33, or 37%, went extinct. Among 37 species given work aimed at the outside pressures on them, the authors count only two extinctions. Across all 141 species, 39 are extinct, 69 are still under 100, and 33 have reached 100 or more. Of those 33, six are in the 100–249 band, 16 are in the 250–999 band, and 11 number 1,000 or more.

The recoveries were slow. Where the timing could be reconstructed, the median gap from the start of action to a population of at least 100 was 18 years, with a range of 1 to 57 years. Eight species came back from fewer than 20 wild animals. The Mauritius kestrel is the extreme the paper highlights: four wild birds in 1974, a later peak the authors put at 350–500 in the late 1990s, then a fall to under 250 by 2021. Seven further species had no wild animals left at their lowest point and persisted only in captivity. They now run from 340 black-footed ferrets to more than 6,000 European bison.

The Lord Howe woodhen is the paper's example of pairing threat control with captive breeding. An estimated 23 birds remained in the late 1970s. After pigs and cats were removed and birds were bred and released from 1980, the population was at least 221 by 2018. A later rodent eradication, with the woodhens held temporarily during baiting, was followed by a count of 1,638 birds in 2024.

A statistical model pointed the same way for recovery, not for a clean proof that conservation stops extinction. Combining pressure relief with hands-on population work gave the highest modelled chance of reaching 100: about 74% for a typical bird case and 71% for a typical mammal case, against about 6% and 5% with no recorded action. The estimated reduction in the odds of extinction was uncertain. In the main models, every credible interval for extinction crossed the line of no association. The links with recovery to at least 100, and to at least 250, stayed in place when the authors reran the analysis with different assumptions, including dropping a large share of the poorly known species that had been coded as unaided.

What the pattern does not settle

The authors say conservation was not assigned by lottery, so a high recovery rate among helped species could still reflect which species people chose to help. They looked for a simple version of that bias in Red List threat types and did not find one, and they report that action usually began while a species was still declining. They also say the toolbox is better for some problems than others. Rats on an island are a familiar target. Avian malaria in Hawaiian forest birds is not. A project aimed at an easier threat can make active conservation look more reliable than it would be if every hard case were taken on.

Reaching 100 animals is a doubling of the study's cutoff, not a secure population. Of the 33 species above that line, the paper's current Red List reading is one extinct in the wild, seven critically endangered, 11 endangered, eight vulnerable, three near threatened and three of least concern. Many still depend on management. The authors also say the work covers a few decades of counts, not the centuries that viability models use when they argue for thousands of animals, and that genetic damage from a bottleneck can show up later.

The study is birds and mammals only. Reptiles, amphibians, fishes, plants and invertebrates are outside it. Historical counts are patchy, and a cautious Red List entry could have placed a poorly known species under 50 when it was not, which would exaggerate the gap between helped and unhelped species. The recovery result held after the authors randomly removed up to half of those poorly known, unaided species. Extinction, in the main model, did not become a clear statistical effect.

Monash University's summary, published the next day, quotes Smith and co-author Rohan Clarke arguing against abandoning a species because the number left looks too small. The paper itself is narrower. It finds a strong association between direct conservation and a climb back over 100 animals, in a set of species that had already reached the edge, and it leaves causation, harder threats and the long term open. The work was supported by Research England. The authors declare no competing interests.