The typical critically endangered bird has about 5 times the research of a comparable common one, but most of what a literature search turns up for any bird is not research

Search the open research index OpenAlex for a bird's scientific name and, across the 9,113 species on this page, only 31% of what comes back for 2015 to 2024 is research. 59% are datasets, nearly all of them GBIF occurrence downloads, records created automatically each time someone downloads sightings from a biodiversity database, and 9% are Red List assessments and Birds of the World encyclopedia entries. Counting research alone, and comparing like with like (birds of the same family, size, range, time since they were named and rated attractiveness), a typical bird of Least Concern, the Red List's lowest-risk category, had 2.0 research works in those ten years and a Critically Endangered one 10.1: about 5 times as many. 30% of the birds had no research work at all. Being endangered goes with being studied, which none of the six published studies of research effort read in full for this page found (three found no link, two the opposite); which way the arrow runs is not settled here, because a bird cannot be listed as endangered without evidence, and evidence is research.

This data does not update. Counts fetched from OpenAlex on 24 October 2026; Red List 2026-1.

Works OpenAlex returns for a bird's scientific name in a title or abstract, 2015 to 2024, for the typical bird of each Red List category: grey, everything; blue, research record types only. Both from the same specification, with family, body mass, range, years since description and rated attractiveness held equal. Typical values do not add up, so the blue bar is drawn beside the grey one, not inside it; the 31% above is of all records pooled, which the famous species dominate.

What a search for a bird finds

Three kinds of record name a bird without being research about it. GBIF, the Global Biodiversity Information Facility, gives every download of its occurrence data a citable identifier and lists the species in its description, so each time a researcher, a student or a consultant pulls the sightings for a region, every bird in them collects a record: 59% of everything counted here. The Red List publishes each assessment of a species under its own identifier, and every reassessment gets a new one; the Cornell Lab's Birds of the World does the same for its encyclopedia accounts. Together those are 9%. 31% is journal articles, reviews, preprints, theses, books and reports, and that is the count this page uses; the rest, 1.4%, is peer-review reports, errata and other odds and ends.

For the typical bird, threatened ones carry more of every kind. With the same terms held equal, a Critically Endangered bird has 2.7 times the GBIF downloads of a Least Concern one and 1.5 times the Red List and Birds of the World entries: a listing writes some of its own record, as reassessments pile up. But the paperwork is not where the gap comes from. Counting every record, the Critically Endangered bird has 2.7 times a Least Concern one; counting only research, 5.0 times. Taking the paperwork out makes the gap larger, because the downloads that make up most of every bird's count rise less steeply with threat (2.7 times) than research does.

Averages tell a different story, and it is worth knowing why. A Critically Endangered bird averages 13 research works and a Least Concern one 11, hardly a difference, because about 140 well-known Least Concern birds with a hundred research works or more each, the red junglefowl (the chicken's wild ancestor), the rock dove, the mallard, the house sparrow, have hundreds or thousands each and pull the Least Concern average up. The typical bird, the obscure majority, is where threat and research go together.

Threat and research

Research works per bird species by Red List category: as counted, adjusted, and adjusted with public attention held fixed as well. "Typical" is the back-transformed mean of log(1 + research works), closer to a median than to a mean for counts this skewed. The adjusted series ask what each bird would have if it were in each category in turn, everything else about it unchanged. Whiskers are 95% intervals from resampling whole bird families.

As counted, a Least Concern bird has 2.3 research works and a Critically Endangered one 5.1. Adjusted, the figures are 2.0 and 10.1, with a 95% interval of 7.4 to 13.3 for the Critically Endangered species, and the adjusted values rise with every step: Near Threatened 3.1, Vulnerable 4.2, Endangered 6.0. (As counted, Near Threatened sits just below Least Concern.) The adjustment widens the gap rather than closing it, because threatened birds tend to be narrowly ranged and recently described, the kinds of bird that are hard to study, and holding that equal removes a handicap.

The size of the multiple depends on what is being compared, and the page gives each. On the typical bird it is 5.0 times (95% interval 3.7 to 6.8, resampling whole families), and on journal articles and reviews alone about the same, 5.0 times. On the average count the gap is smaller and much less certain. A count model, which compares averages and so is dominated by the few hundred birds with large literatures, gives 2.3 times, with an interval from 1.0 to 5.1: only just distinguishable from no difference, and in that model Vulnerable birds sit below Near Threatened ones. So the finding is about the typical bird, the obscure majority: a critically endangered one of those has several times the research of a comparable Least Concern one, while the famous few dominate the averages whatever their status.

Some of the gap travels with public attention. Holding English Wikipedia views equal as well takes the multiple from 5.0 to 3.0. Which causes which is not knowable from one snapshot; a bird that is studied more may get a better article and more readers, as easily as the reverse.

What else goes with being studied

What the full model's fit depends on: the fall in R² when each term is removed and the model refitted, with 95% bootstrap intervals over families. The full model adds Wikipedia views to the adjusted model above; R² rises from 0.48 to 0.58.

How often people look a bird up on Wikipedia carries the most of the fit (0.101), and which family it belongs to nearly as much (0.083): some families are studied as model organisms, game or pests. Red List category is fifth of seven (0.013), behind years since description and range size. Rated attractiveness, scored from photographs by several thousand people, adds 0.000: beauty, the quality this page set out to test, explains almost nothing once the rest is known, with a coefficient indistinguishable from zero (p 0.57).

Why the published record disagrees

Ten species-level models of research effort across mammals, birds, reptiles, turtles and plants have been published since 2014, eight of them with a threat-status term in a model. Of the 6 whose full text could be read for this page, 3 find threat status null, 2 find it negative (one of them, the one other study of all birds, from 2014, a comparison with nothing else held equal) and one finds that being assessed at all, rather than being threatened, goes with more study. Those studies counted papers in Web of Science, Zoological Record or Scopus, mostly on a single class. This page uses a different index, a later window and one class, and any of the three could produce a different sign. A single result against a literature is more likely to be wrong than the literature, so what this page reports is what one instrument measured on one class over one decade.

"Assessed, positive" means the paper found that species on the Red List at all are more studied than unassessed ones, with no further effect of the category; the authors note the circularity, since an assessment needs data to exist first. "Unadjusted" marks a comparison with nothing else held equal. On British birds, the international Red List category was null while the national action-plan listing, with money and a monitoring duty behind it, predicted more study. Rows read from the abstract alone carry the abstract's own result; unread rows are left unread rather than guessed.

Across whole classes the picture is the familiar one. Amphibians are 41% threatened, the highest best estimate the Red List gives for any vertebrate class, and hold 25% of the threatened vertebrate species, but get 4% of the research works that name their class: 6 per threatened species, against 89 for birds.

The five vertebrate classes: the share of all threatened vertebrate species each holds, against the share of research works of 2015-2024 whose title or abstract names the class. Threatened counts from Red List 2026-1, Table 1a¹⁰; works from OpenAlex, biology and environment topics, research record types only. The Red List has assessed 78 to 100% of each vertebrate class and far less of most other groups, which is why no other group is drawn.

Method

Birds are the one class where every term has a free, species-level source. AVONET gives body mass¹¹, range size and family for 11,009 species. The iratebirds project¹² gives a rated attractiveness score for 9,386 of them. The GBIF backbone gives the year¹³ each species was described (23 from the Catalogue of Life where GBIF had none). English Wikipedia gives a public-attention measure¹⁴ for the 10,645 species with an article. OpenAlex gives the number of papers¹⁵ and other works of 2015 to 2024 whose title or abstract names the species, split by record type; research means the types article, review, preprint, letter, book, book chapter, dissertation, conference paper and report, a list fixed before any model was fitted on the split. The Red List gives the category. Joined, keeping the species with every term and a category that ranks risk, that is 9,113 birds.

The response is log(1 + research works). The adjusted model is ordinary least squares with a fixed effect for every family, log body mass, log range size, years since description to 2024, rated attractiveness, and the Red List category with Least Concern as the reference; the full model adds log(1 + mean monthly Wikipedia views). Standard errors are clustered by family. Category means are predictions, not coefficients: every species is predicted under each category with its other terms as they are, and the predictions are averaged and back-transformed. The count model is a Poisson regression on the same terms. Intervals on the means are percentile bootstrap over families, 300 resamples, seed 20260905. Family fixed effects stand in for a phylogeny, which is assumed; the published bird model used a tree¹⁶.

The Red List column came from the Red List API, version 2026-1, matched by scientific name to the names AVONET uses; 10,669 of 11,009 matched. Species with no English Wikipedia article are given zero views, which keeps the obscure species in. 17 species otherwise in the data are left out for want of a category the model can use: most because their name finds no match in the Red List, and a few because they are Data Deficient, Extinct or Extinct in the Wild. The species query is pinned to OpenAlex's title-and-abstract filter, because its short search parameter changed meaning in February 2026¹⁷: it now means the full text, which returned 3.4 times as many works for the lion; a test asserts the query the fetcher builds.

Limits

Association, not cause, and here that is not a formality. A bird cannot be listed Critically Endangered without evidence: the higher categories need population, trend and range data, which is research, and Least Concern needs much less. So the arrow may run from research to listing, and the gradient on this page is what rising evidence requirements would produce on their own. The category that means nobody has enough data points the same way: the 34 Data Deficient birds have a median of 0 research works, against 1 for Least Concern, in a wider frame of 10,669 birds with a category and a count. That is too few species for a coefficient and is given none.

Record types are OpenAlex's, not this page's. A report or a thesis is counted as research and a short note the same as a monograph; a paper that names a bird only by its common name is missed, and a few scientific names are also ordinary words (Indicator indicator, the greater honeyguide, collects every paper that writes "indicator indicator"). A spot check of the record sources behind 40 sampled species found 96% of the datasets were GBIF downloads and 100% of the reference entries Red List assessments or Birds of the World accounts.

The model needs a rated attractiveness score, which 1,472 of the 10,585 otherwise-complete birds do not have. Fitting the same terms on both samples, without attractiveness, gives a Critically Endangered term of +1.11 on the wider sample against +1.30 on the modelled one (on the model's log scale, where the headline is +1.30), with overlapping intervals: the selection moves the size a little and not the direction. The attention measure is the weakest input: 364 species have no English Wikipedia article and are given zero, and 273 species share 129 articles, each credited with the whole of an article's traffic. Refitting only on species with an article roughly halves the Critically Endangered term with views held equal, from +0.86 to +0.43 on the same scale: how much of the gap survives holding attention equal depends on how the birds without an article are treated.

The species model stops at birds because birds are the one class with a rated-attractiveness dataset, and the class figure stops at vertebrates because the Red List has assessed few species outside them. Mammals, with a public h-index dataset of their own¹⁸, are the obvious next class and are not built.

The code

The fetch scripts, one for each source and each resumable; the build, which joins, fits and bootstraps; the figure script; and the page renderer. Every input whose license allows it travels: AVONET and the attractiveness scores under CC BY 4.0, the Wikipedia and OpenAlex counts under CC0, the Red List summary table as a cited aggregate. The Red List column does not: the IUCN Red List's terms (version 3.1) bar redistributing Red List data "in their original format" (section 4), while placing "no restrictions on use of the IUCN Red List Categories associated with each named taxonomic entity" (section 3), and the archive takes the cautious reading. fetch_iucn.py gets the column with a free personal token, and manifest.json carries the SHA-256 of the 11,185 species and categories this page was built from, 1f1d4b6f1ff2c81d, so build_beauty.py --verify-categories can tell a reader whether their fetch matches. One shipped file, data/lc_matched.csv, lists 1,000 species that are all Least Concern, and so does reveal that category for them; whether it keeps shipping is an open decision.

OpenAlex has metered its API since February 2026. The record-type counts cost $10.67 to fetch for 10,669 species; a reader who trusts the shipped counts pays nothing, and one who re-fetches them needs a key and 11 days of a free account's allowance, or about $11 of credit.

Download source · 1,699 KB All code

Sources

1. Brooke, Bielby, Nambiar & Carbone, PLoS ONE 9(4):e93195, 2014, doi:10.1371/journal.pone.0093195.286 carnivores; IUCN status dropped from the model beside body mass, range and diet breadth at the second simplification step (likelihood-ratio p = 0.848, Table S5).
2. McKenzie & Robertson, PLoS ONE 10(7):e0131004, 2015, doi:10.1371/journal.pone.0131004.225 British breeding birds; Red List status null (univariate p = 0.34; multivariable t = 0.45, p = 0.65), national action-plan status positive in the multivariable model (t = 3.14).
3. Fleming & Bateman, Mammal Review 46(4):241-254, 2016, doi:10.1111/mam.12066.331 Australian mammals; the threat-term result is in the paywalled full text and was not read for this page.
4. dos Santos et al., Animal Conservation 23(6):679-688, 2020, doi:10.1111/acv.12586.4,108 terrestrial mammals, hurdle model; the abstract reports threat status as weakly associated with research presence and volume. Read from the abstract; the full text is open access but was not reached.
5. Tam, Lagisz, Cornwell & Nakagawa, GigaScience 11:giac074, 2022, doi:10.1093/gigascience/giac074.5,497 mammals; IUCN status linear -16.4 [-19.3, -13.6] with a quadratic term; Google Trends a strong positive predictor.
6. Mammola et al., eLife 12:RP88251, 2023, doi:10.7554/eLife.88251.3,019 species across 29 phyla; being on the Red List raises interest whether the species is endangered or of Least Concern.
7. Guedes, Moura & Diniz-Filho, Ecography 2023:e06491, 2023, doi:10.1111/ecog.06491.10,531 reptiles; body size, description year and proximity to institutions dominate. Full text open access (CC BY), not reached past the publisher's bot wall.
8. Adamo et al., Nature Plants 7:574-578, 2021, doi:10.1038/s41477-021-00912-2.113 Alpine plants; color, conspicuousness and range predict research attention; the authors' published code has no threat-status term. Paywalled; not read for this page.
9. Ducatez & DeVore, PLoS ONE 21:e0347198, 2026, doi:10.1371/journal.pone.0347198.370 turtles; threatened vs non-threatened -0.10, t = -0.46, p = 0.64 (the extinction-risk ordinal +0.10, p = 0.14); phylogeny 66% of the variance; assessed species more studied than unassessed.
10. IUCN, The IUCN Red List of Threatened Species, version 2026-1, summary statistics Table 1a, 2026, https://www.iucnredlist.org.Described, evaluated and threatened species by major group, and the category per bird species through the Red List API.
11. Tobias et al., Ecology Letters 25(3):581-597, 2022, doi:10.1111/ele.13898.Body mass, range size, family and order for 11,009 bird species on the HBW-BirdLife v5 taxonomy; CC BY 4.0.
12. Santangeli et al., npj Biodiversity 2:20, 2023, doi:10.1038/s44185-023-00026-2.Rated attractiveness per species and sex from over 400,000 photograph ratings by more than 6,000 people; the deposit is CC BY 4.0.
13. GBIF Secretariat, GBIF Backbone Taxonomy, checklist dataset, doi:10.15468/39omei.The full scientific name with authorship, from which the year of description is read.
14. Mittermeier et al., Conservation Biology 35(2):412-423, 2021, doi:10.1111/cobi.13702.Wikipedia pageviews as a measure of public interest in species, the method this page borrows.
15. Priem, Piwowar & Orr, arXiv 2205.01833, 2022.The open index of scholarly works whose title-and-abstract search gives the paper counts; data CC0.
16. Ducatez & Lefebvre, PLoS ONE 9(2):e89955, 2014, doi:10.1371/journal.pone.0089955.10,064 birds; phylogeny takes 74% of the variance in research effort; Least Concern species have twice the papers of threatened ones in a bivariate comparison, not a model term.
17. OpenAlex, “New features and usage-based pricing”, 24 February 2026, https://blog.openalex.org/openalex-api-new-features-and-usage-based-pricing/.The date the bare search parameter was redefined as full-text search and the API became metered, at $0.0001 for a filtered list call.
18. Tam et al., GigaDB dataset 102237, 2022, doi:10.5524/102237.Per-species h-index, publication count, Red List category and Google Trends for 7,521 mammals; CC0.