The number at the centre of all this is 99.2 percent. That is the probability a biogeographic model now assigns to Europe as the place the first bats appeared, somewhere between roughly 65 and 60 million years ago, in a study that assembled 103 bat genomes covering every one of the 21 living bat families, folded in 44 fossil species, and took 137 researchers in 64 countries to produce. It went online in Nature in late September. The senior author is Emma Teeling, professor of zoology at University College Dublin, and the reason the result lands differently for her than for almost anybody else is that she has been arguing a version of it since she was a doctoral student whom most of the field politely disagreed with.
Bats are the hardest large group of mammals to trace backwards, and the reason is the thing that makes them bats. Powered flight demands lightness. Lightness means bones that are thin and partly hollow, and thin hollow bones are almost never what survives 50 million years in sediment. The result is a fossil record with holes in it that are not small gaps but structural absences. On the researchers’ own accounting, close to half of each bat family’s early history, 48.7 percent on average, is represented by no known fossil at all.
Worse, the fossils that do exist arrive too late to be useful for the one question everyone wants answered. The oldest bat fossils anybody has ever found are roughly 52 to 56 million years old, and they are already finished bats. They fly properly. The inner ear anatomy says they already echolocate. There is no known fossil of a gliding proto bat, no halfway creature with a stretched hand and no sonar. The transition that produced the second most species rich group of mammals on the planet, more than 1,400 living species, around a fifth of all mammal diversity, is simply missing from the rock.
Which is why, in the 1990s, the standard account of bats was built on anatomy and on reasonable inference rather than on evidence anybody could point at. That account split the order into two: megabats, the large fruit eating and nectar feeding bats of the Old World tropics, which navigate largely by sight and smell, and microbats, the smaller insect hunters that find prey by shouting and listening. It was tidy. It matched how the animals look and how they behave. It also implied something obvious about echolocation, namely that sonar belonged to the microbats, had evolved in that lineage, and was the thing that defined it.
Teeling grew up in Dublin, studied zoology at University College Dublin, did a masters in animal behaviour at Edinburgh that included fieldwork with swift foxes in Canada, and then went to Queen’s University Belfast and the University of California at Riverside for a doctorate. Her thesis, finished at the start of the 2000s, was on a molecular perspective on bat systematics, which in plain terms meant testing the anatomical family tree against gene sequences. The sequences did not agree with the anatomy.
What the molecular data kept saying was that the microbats were not a single natural group. Some of the echolocating bats, the horseshoe bats and their relatives in particular, turned out to be closer kin to the big fruit bats than to the other echolocators. That arrangement is now known as Yinpterochiroptera, and it broke the old division in half. Teeling published the microbat paraphyly result in the early 2000s, and in 2005 she and five colleagues, including the mammal geneticist William Murphy, published a molecular phylogeny of bats in Science under a title that is worth reading twice given what happened afterwards: a molecular phylogeny for bats illuminates biogeography and the fossil record.
That paper carried an uncomfortable implication, and she said so. If horseshoe bats sit on the fruit bat side of the tree, then either laryngeal echolocation evolved twice independently, in two lineages, by two separate routes to the same extraordinary capability, or it evolved once in the common ancestor of all modern bats and was subsequently lost by the fruit bats. Teeling argued for the second. Evolved once, lost once. It was the less fashionable reading, because losing sonar sounds like throwing away a superpower, and because the convergence story had the comfort of being the default.
There is a reason the loss argument is more plausible than it first sounds. Echolocation is expensive. It requires a larynx modified to produce calls at intensities that would be painful to stand next to if they were pitched lower, an inner ear tuned to catch the returning echo of an insect wing, and a brain able to do the arithmetic on the delay in real time while the animal is flying. All of that is worth paying for if your food is a moth that is actively trying not to be caught. It is worth considerably less if your food is a fig. Fruit does not evade. Large bats feeding on fruit and nectar have good eyes, excellent noses and no particular need to shout, and biology does not maintain costly machinery for which there is no longer a job. Framed that way, the fruit bats did not throw away a superpower. They stopped funding one they had no use for.
For two decades that stayed an argument rather than a settled fact, and arguments in phylogenetics are usually settled by data volume. Which is roughly the premise of Bat1K. Teeling co founded the consortium with Sonja Vernes, now a professor at the University of St Andrews who studies the genetics of vocal communication, with the stated goal of sequencing the genomes of all living bat species. It is not a modest project. The consortium announced itself properly in 2020, when six high quality bat genomes made the cover of Nature and produced the findings on bat immunity and longevity that have since turned bats into a model system for research on ageing. Teeling’s own separate line of work asks how a seven gram bat can live past forty, which it should not be able to do on any mammalian metabolic logic, and appears to involve unusually well managed telomeres.
The paper that has just appeared is the family tree Bat1K was built to produce. Of the 103 genomes, 42 are newly generated chromosome level assemblies, which means the sequence is resolved into whole chromosomes rather than scattered fragments, and 61 were already in hand. Every living family is represented. The team also reconstructed what the ancestral bat chromosome set probably looked like, reporting 26 ancestral chromosomes, which is the sort of result that only becomes possible once you have assemblies good enough to track where whole chunks of genome moved.
It is worth pausing on what chromosome level means in practice, because it is the difference between the old molecular trees and this one. A cheap genome assembly gives you the sequence in thousands of disconnected pieces, like a shredded book you can read sentence by sentence without knowing the chapter order. A chromosome level assembly gives you the chapters. That matters for a family tree, because some of the most informative evolutionary evidence is not in the letters at all but in the order, in which blocks of genes sit next to which other blocks, and whether a whole region has been inverted or moved wholesale to a different chromosome. Rearrangements of that kind are rare, hard to reverse, and therefore unusually reliable markers of shared ancestry. You cannot see any of them in a shredded assembly.
The method for getting a date out of this is worth following, because it is where most of the honest uncertainty lives. Genomes on their own give you a shape, not a timescale. They tell you the branching order and roughly how much sequence change accumulated along each branch, but sequence change does not come with a clock face attached. To convert it into years you need fossils: specimens whose position on the tree you can defend from anatomy and whose age you can defend from the rock they came out of. Each one pins a minimum age to a node. The model then stretches and compresses the rest of the tree around those pins. That is what the 44 fossil species are for. They are not illustrations. They are the calibration.
Then a separate model asks where. It takes the dated tree, takes the known geographic range of every living and fossil lineage on it, and works backwards to the distribution that best explains the whole pattern with the fewest improbable moves. The answer that came out, with that 99.2 percent figure attached, was Europe, in the late Paleocene. Previous estimates had tended to put bat origins in Asia, Africa or North America, and later, around 50 million years ago.
Sixty five million is a specific kind of interesting, because the asteroid that ended the age of the large dinosaurs struck about 66 million years ago. On this reading the first bats appear almost immediately afterwards, in the emptied world, among the mammal lineages that expanded into the space.
Here is where precision matters more than excitement. Europe is an inference, not an observation. Nobody has found a bat fossil in Europe or anywhere else that is 65 million years old. The oldest is still the 52 to 56 million year old material, which leaves a gap of roughly a decade of millions of years between the modelled origin and the first physical bat. The 99.2 percent is a confidence figure produced inside a model, given the tree, the calibrations and the geography that went into it. It is not the probability that the conclusion is true in the way a reader might naturally hear it. Change the fossil calibrations and the number moves.
There is a sampling problem sitting underneath it too, and it is the obvious objection. Europe has the best early bat fossils in the world largely because Europe has the right rocks. The Messel oil shale in Germany, a drowned volcanic crater lake, preserves whole articulated bats with their wing membranes and sometimes their last meal, at a level of completeness almost nowhere else matches. One of the circulating images from this study is a Messel specimen of Palaeochiropteryx tupaiodon, about 47 to 48 million years old, photographed by Matthew Jones, who is one of the paper’s authors. A region that preserves bats unusually well will also appear to have unusually many early bats. Any serious biogeographic model has to try to correct for that, and no correction is perfect.
What the genomes plus fossils do support strongly, and what is arguably the real intellectual payoff, is the second finding. A French fossil named Vielasia, around 50 million years old, with ear anatomy indicating sophisticated echolocation already in place, sits on the oldest branch of the bat tree in this analysis. If the earliest surviving lineage was already echolocating, then echolocation and powered flight were both present before the modern bat groups began to diversify. Not acquired repeatedly afterwards. Present at the start.
That is the 2005 position, now resting on 103 whole genomes instead of a handful of genes, and it collapses an argument that ran for a generation about which came first. The careful version of the claim is narrower than some of the coverage: the analysis supports a single early origin of both traits in the common ancestor of modern bats rather than multiple independent inventions, and it pushes the origin of echolocation earlier than many had assumed. It does not show flight and sonar appearing in the same animal in the same instant, because the fossil that would show that does not exist.
The other thing worth saying about this paper is what it cost in logistics rather than in computation. One hundred and thirty seven researchers in 64 countries is not a figure about prestige. It is a figure about tissue. Every one of those 103 genomes started as a sample from a specific animal in a specific place, collected under that country’s permits, exported under the rules governing genetic resources, which since the Nagoya Protocol are considerably more involved than mailing a tube. The samples came from field researchers working over decades in places from New Zealand to Madagascar. Laurel Yohe, an assistant professor of bioinformatics at the University of North Carolina at Charlotte and one of the paper’s authors, is among the people who collected material that went into the database, and now works on how a gene implicated in vocal learning varies across bats with different echolocation styles. Daniel Whitby, whose photographs of British horseshoe bats have carried much of this story’s coverage, is also an author. The first author is Ariadna Morales, an evolutionary genomicist who has moved between Frankfurt, Montpellier and New York and who also runs the unglamorous infrastructure projects that make this kind of work repeatable, a database of bat occurrence records and a project digitising the scattered historical literature on bats.
The honest state of play, then, is this. The shape of the bat family tree is now about as well resolved as the shape of any mammalian order, every living family has a reference genome behind its position, and two long running disputes have moved decisively. Echolocation almost certainly evolved once, early, and the fruit bats lost it, which is what a graduate student in Belfast and Riverside concluded from far less data and was not widely believed about. The place of origin is Europe on the strongest available model, with a specific probability attached and a specific weakness underneath it.
And the thing that would turn the European origin from a very good inference into a fact has not been found. It would be a bat, or something on its way to being one, in a European deposit older than 56 million years, with the wrong proportions and perhaps no sonar at all. The tree now says where to look and roughly how deep. Somebody will go and look, which is the most anyone can ask of a model.