How long have we been talking?

The shape of the vocal tract, inferred from the base of the skull. The hyoid bone, and where it sat in the vocal tract. The size of the canal carrying nerves to the tongue and of the one carrying nerves to the muscles of breathing, which speech controls finely.

Two of those do less work than they look. The position of the larynx cannot be read reliably off the base of the skull, and the tongue canal overlaps in size with chimpanzees.

And genetics: variants in genes involved in speech and language development, whose distribution across populations and across archaic humans can be dated.

A 2013 review assembled the case that Neanderthals had the relevant anatomy, including a hyoid bone essentially like ours, the widened spinal canal the review ties to fine control of breathing, and the same variant of a gene implicated in speech that modern humans carry.

The canal argument leans on one Homo erectus skeleton, from Nariokotome, whose canal is narrow. A 2015 study found only three of its vertebrae constricted, which suggests a spinal disorder, placed the Homo erectus vertebrae from Dmanisi within the modern human range, and contested the idea that vertebrae say anything about breathing control or language.

They also hafted tools with pitch, used pigment, buried their dead and looked after the injured. Those behaviours do not prove language. The review argues they would not have been possible without it, and its authors call reading language off the archaeological record a controversial endeavour.

If Neanderthals had language, the capacity predates the split between their lineage and ours, which puts its origin at least half a million years ago rather than in the last hundred thousand.

Other animals communicate by sound, and a few learn their calls. A 2002 paper by Hauser, Chomsky and Fitch proposes that the only uniquely human core of language is open-ended combination: a finite set of elements producing an unlimited range of expressions. Critics reply that this leaves out much of grammar that is not recursive, such as sound systems and word structure.

A gradual account has to describe what the intermediate stages were for, since half a syntax is not obviously useful. A sudden account has to explain how such a thing appears at once. Neither has a settled answer.

A gene once thought to hold a key to the rise of modern language is a regulator of other genes, active in the brain, the lung and the gut, and involved in the fine motor control that speech needs. Neanderthals carry the same variant.

That is a useful case of a finding surviving while the story attached to it collapses. The 2013 review pictures the biology of language as the slow adaptation of many factors, not a mutation in one or a few genes.

The 2002 paper that put a gene at the centre of this reported that the human version of FOXP2 differs from the chimpanzee version at two amino acid positions, and estimated that those changes spread through the human population within roughly the last two hundred thousand years.

The gene was tied to speech through one family that carries a mutation in it, and through an unrelated person in whom it is disrupted. Those with the mutation have severe difficulties with articulation, with language and grammar, and with fine control of the mouth and face.

The dating did not survive either. A 2018 re-analysis using a far larger and more diverse set of human genomes found no signal of a recent sweep at the locus, which removes the part of the 2002 result that put the change inside the window when anatomically modern humans appear.

It is a real finding about a real gene. What it is not is a switch for language, and the same variant turning up in Neanderthals means it cannot mark the arrival of language in our lineage alone.