Ghost ancestors: the people in our DNA who left no face behind

Science and Space Aug 21, 2026

A new genomic method finds evidence of ancient encounters written into living DNA — without a single fossil. A paper published in Science on July 30, 2026 offers something rarer than bones: biographies without bodies.

Every fossil is a body without a biography. We find the skull, date it, sketch the anatomy — and that is where the story ends. No name, no language, no account of where they went or who they met. The bones tell us a shape existed; they cannot tell us what happened afterward.

A paper published in Science on July 30, 2026 offers the inverse. Researchers using a new genomic method called TRACE found something rarer than bones: biographies without bodies. Segments scattered across the genomes of living people that record an ancient encounter — sex, children, lineages — but no skull to put in a museum case, no species name to attach, no face.

These are the ghost ancestors. And the fact that we can now detect them, precisely, without a single ancient bone, changes what we thought we knew about where modern humans came from.

The finding

The paper's headline claim is worth stating carefully, because the media coverage has already generated one significant misunderstanding.

TRACE identified two distinct archaic signals in modern human DNA that cannot be attributed to known Neanderthals or Denisovans. The first — the larger one — is a ghost lineage that appears to have mixed with the ancestors of all modern humans before the last major Out-of-Africa expansion. Its contribution to any living person today is estimated at 0.5–1% of the genome. Not enormous, but consistent and universal across the populations studied: present in people from Europe, East Asia, Africa, the Americas.

The second signal is older and stranger. It seems to have taken an indirect route: mixing first with Denisovans, then reaching some modern humans through subsequent Denisovan–sapiens contact. It shows up primarily in Oceanians, and its contribution is very small — around 0.002% of the genome.

What went wrong in most news coverage: several outlets added these two numbers to the familiar ~2% Neanderthal contribution and concluded that humans carry "2% of two unknown species." That arithmetic does not hold. The ~2% figure refers to total archaic ancestry — including the well-documented Neanderthal input — not to these two new signals alone. The ghost lineage at 0.5–1% is new; the super-archaic remainder is vanishingly small. Confusing these numbers makes the finding sound simpler than it is, and obscures what makes it genuinely important.

How TRACE works — and why it does not need ancient bones

The method is the story here. TRACE reconstructs what geneticists call an Ancestral Recombination Graph, or ARG.

The key to understanding this is recombination. Every time a human cell produces an egg or sperm, chromosomes shuffle. Segments that were together in your grandmother's genome get reshuffled in your father's, and reshuffled again in yours. Over many generations, any long stretch of DNA gets broken into smaller and smaller pieces, each carrying its own genealogical history — a slightly different version of the family tree, local to that segment.

A Homo heidelbergensis skull displayed at the Natural History Museum in London, viewed from the front against a dark background. — A Homo heidelbergensis skull at the Natural History Museum in London. Heidelbergensis — a contested label for a diverse group of Middle Pleistocene African hominins — is one candidate for the ghost lineage detected by TRACE. (image: Emőke Dénes / Wikimedia Commons, CC BY-SA 4.0)

An ARG makes all those local genealogies visible simultaneously. You can think of it as mapping not just who is related to whom, but where in the genome that relationship lives, and how deep it goes. When TRACE finds a segment with an unusually deep branch — one that diverges far earlier than the main human family tree — that is a candidate for introgressions from a population long separated from the main sapiens lineage.

What makes this powerful is what it does not require: ancient DNA. Most of what we know about Neanderthals and Denisovans comes from a handful of preserved samples, mostly from cold or cave environments in Europe and Siberia. Old DNA degrades; humid tropics preserve almost nothing; Africa — where most of human evolution happened — has yielded very little. TRACE sidesteps this problem entirely. It reads the living genome as an archive and looks for the deep signatures that ancient mixing would have left.

As a validation, the researchers ran TRACE on segments already known to be Neanderthal or Denisovan. It found them. That is not proof the method is infallible, but it is exactly the kind of sanity check that builds confidence in a new approach.

The method is not without limits. The reconstruction depends on demographic assumptions, estimates of mutation rates, and the composition of the modern sample. In Africa in particular, deep population structure — long-separated groups that later merged within Homo sapiens — can produce signals that resemble archaic introgressions without actually being one. The honest scientific framing, which the paper itself maintains, is that these are powerful genomic signals consistent with archaic mixing. Not an excavated person. Not a confirmed species.

Who could they be?

This is where the paleoanthropology gets speculative, and deliberately so. The researchers have not named a species. They cannot. There is no fossil, no ancient genome, no anatomical description to compare against.

For the first ghost lineage — the one that merged with all modern human ancestors before the Out-of-Africa expansion — the chronologically plausible candidates are Middle Pleistocene hominins in Africa. Homo heidelbergensis is often mentioned, but that name is genuinely contested; it functions more as a label for a diverse group of fossils that do not fit neatly elsewhere than as a coherent species with a clear identity.

For the super-archaic signal, Homo erectus fits the timing: a lineage that split from the main hominin trunk around 1.8 million years ago, spread across Eurasia, and was still present in parts of the world when Denisovans were active. Whether any erectus population actually interbred with Denisovans remains speculative.

What the paper does well is reframe the question. Instead of saying "we think it might be X," it specifies, very precisely, what kind of fossil or ancient biomolecule would confirm or refute the hypothesis. Genomics defines the shape of the absence. Paleoanthropology now knows what to look for.

The tree was always a network

The deeper shift in this research is conceptual, and the paper's authors say so directly. The human family tree is not a tree. It is a network.

The older model — comfortable, teachable, clean — depicted human origins as a series of replacements. One group gives way to another. Moderns arrive; archaics disappear. The picture was always an approximation, but it held because the evidence for mixing was thin. Neanderthal DNA in modern Eurasians was the first serious crack. Denisovans widened it. Now TRACE finds signals that predate both, mixed into ancestors before those ancestors became modern humans.

The model that fits the data is messier and more interesting: populations diverging for hundreds of thousands of years, moving across continents, occasionally meeting again, producing children who survived. "Extinction" in this context does not mean what it seems to mean. A population can stop existing as a distinct group while its DNA continues — dispersed into other lineages, unrecognizable without the right tools, but present.

Composite image of Homo erectus stone tools including handaxes and bifaces from the Eastern Desert, Sudan, arranged in a vertical panel. — Stone tools associated with Homo erectus, including handaxes and bifaces from the Eastern Desert, Sudan. Such tools mark the presence of archaic hominins across Africa and Eurasia — populations whose genetic contribution to modern humans is only now becoming traceable. (image: Bioanthropologist1 / Wikimedia Commons, CC BY-SA 4.0)

That is what these ghost ancestors are. People who existed. Who made children. Whose children made children who eventually made us. Their group did not survive, their fossils have not been found, their language and culture left no trace anyone has yet identified. But some of what they were is in us, scattered in fragments across every person whose genome the study examined.

What this changes

The practical implications for paleoanthropology are significant. There is now a genomic target: researchers know, approximately, how old the divergence was, how much DNA the mixing introduced, and which modern populations carry the strongest signal. That information can guide fossil hunting, dating strategies, and the prioritization of sites where ancient biomolecules might survive.

For everyone else, the finding is a small recalibration of what "human origin" means. The common metaphor — that we are descended from a single ancestral population, a founding group who spread and diversified — is not wrong, but it is incomplete. The founding population itself already carried the traces of older encounters. And those encounters go back further than TRACE can currently reach.

Genomics cannot yet fully separate deep population structure from archaic introgressions in some African populations. Better ancient DNA from African sites, if it becomes recoverable, would sharpen the picture considerably. The current estimates will likely be revised. The existence of the signals seems unlikely to be revised away.

The stronger point, and the one the paper ends on, is about the shape of human diversity: not a tree with clean forks, but something more like a braided river — channels joining and separating, some running dry, others carrying water to the sea. The ghost ancestors ran dry as a group. Their water is still here.