What is a species?

Interbreeding, ecological role, evolutionary history, morphology and genetic clustering can each pick out different boundaries. A single group of populations can be one species under one criterion and several under another.

This is not an academic problem. How species are counted changes how many count as endangered, and how biodiversity is measured. Groups that have been surveyed both ways come out 48% more numerous under a phylogenetic concept, each species with a smaller range and a smaller population.

A 1995 proposal took the opposite approach to defining a species by what keeps populations apart: it defined them by what is observed, as groups that remain genotypically distinct where they meet. Gene flow then acts on a species rather than defining one, and two clusters can stay separate under a good deal of it where selection against the intermediates is strong.

That handles hybridising species, which the textbook definition cannot, and its author argued it extends to predominantly clonal organisms, giving prokaryotes and eukaryotes one definition.

It is one of the more than twenty definitions in use, and it shows that the disagreement is about where boundaries fall, not only about words.

A 2007 paper argued that all the current definitions share one concept, that a species is a separately evolving lineage, and differ on which further property a lineage must have acquired to count. It proposed treating those properties as evidence that a lineage has separated, rather than as part of what a species is.

On that view, reproductive isolation, distinct ecology and genetic divergence are not rival definitions but different lines of evidence, which appear at different times as a lineage splits. Disagreements between them are informative rather than fatal: they locate a case in the middle of a split.

Many species exchange genes occasionally. Brown bears and polar bears mate in the wild and produce cubs, ducks and birds of paradise hybridise, and bear genomes show gene flow among several other species too, not only between close relatives. Our own lineage is in the same position: Neandertals shared more variants with present-day people in Europe and Asia than with people in sub-Saharan Africa.

A definition built on complete isolation cannot accommodate that. What is not agreed is how much gene flow a boundary can survive while remaining a boundary.

Bacteria and archaea exchange genes across enormous distances in the family tree, between prokaryotes and eukaryotes as well as among themselves, which raises the question of whether they form discrete species at all or a continuum. Working definitions fall back on genome similarity. A comparison of seventy genomes in 2005 found that about 94% average nucleotide identity matched the DNA reassociation standard the older definition had used.

The threshold started as a match to that older convention. A 2018 analysis found that it seemed to mark something: across eight billion pairs of prokaryotic genomes, 99.8% sat above 95% identity within a named species or below 83% between species, with very little between. A 2021 critique argued that the gap comes from sequencing many near-identical genomes of favoured species, and that it thins when each species is represented by its most different genomes. The original authors replied that populations sampled straight from the environment, without culturing, show a similar gap. Whether that gap is the kind of boundary a species concept needs is a separate question from whether it is there.