Blog

In 2004, a graduate student walking the Khugenetsavkhlant badlands in the eastern Gobi Desert of Mongolia picked up the remains of an animal roughly the size of a chipmunk. Six or seven inches from nose to tail. A skull, and part of a hind limb. Not the kind of find that stops an expedition in its tracks, because the Gobi is generous with small Cretaceous mammals and stingy with complete ones.

The student was Andres Giallombardo. Twenty two years later, he is the lead author of the paper in Nature that describes that animal, names it Tamirkhan balcarceli, and uses it to take apart a reading of the mammal family tree that had stood since the 1980s. The person who carried it out of the desert is the person who finally explained it. That does not happen often, and it is worth sitting with for a moment, because the gap between those two events is really the story.

Here is what the animal did. For decades, a group of small Late Cretaceous mammals called zhelestids occupied an unusually important seat in the family tree. They had teeth built for grinding plants, and those teeth looked uncannily like the teeth of hoofed animals. Cows, deer, antelope, horses: that whole branch. So zhelestids were read as early relatives of that branch, which made them early members of the placental mammals, the group that includes almost every familiar mammal alive today and includes us.

That reading carried a very large consequence. If zhelestids were already placental mammals while dinosaurs were still walking around, then the great placental expansion did not wait politely for the asteroid. It began before it. That claim sat alongside molecular clock studies, which estimate divergence times from differences in living animals’ DNA and which have long pointed toward placental mammals originating well before the end of the Cretaceous. Fossil evidence and molecular evidence appeared to be shaking hands. Zhelestids were a significant part of why.

The new fossil says they were never placentals at all. They belong somewhere else: with the zalambdalestoids, a different lineage of Cretaceous mammals that sits outside the placental group. And the teeth that convinced everyone were not inherited from a shared ancestor. They were convergence, which is the word palaeontologists use for the same solution arrived at twice by unrelated animals.

Maureen O’Leary, a professor in the Department of Anatomical Sciences at Stony Brook University’s Renaissance School of Medicine and a research associate at the American Museum of Natural History, put the thesis of the whole paper in one plain sentence. Teeth remain among the most informative fossils available, she said, and this work shows that teeth cannot always tell us how the whole animal looked.

That is the sentence to hold onto, because it explains why this correction took twenty two years and a specific fossil rather than a better argument.

Convergence is the oldest trap in the business. Shape follows job. An animal that spends its life grinding tough plant matter needs ridged, hard wearing, high crowned teeth, and there are only so many ways to build those out of enamel. Two animals on opposite sides of a family tree, both chewing the same kind of food, will end up with teeth that look like cousins. Dolphins and extinct marine reptiles converged on the same body outline. Birds and bats converged on wings. Teeth do it constantly, and teeth are also the single most common mammal fossil, because enamel is the hardest tissue a body makes and it survives when everything else has gone.

So mammal palaeontology has a structural problem. The part of the animal that preserves best is also the part most likely to lie about ancestry. For most of the history of the field there was no way around it, because most Cretaceous mammals are known from a handful of teeth in a jaw fragment and nothing else. You classify what you have.

Zhelestids were a textbook case of exactly that. The family was named in the mid 1980s, from central Asian material, much of it from Uzbekistan. By the late 1990s, a formal analysis had placed them as ungulate-like mammals, close to the hoofed branch, and the idea entered the literature with real weight behind it. Nearly forty years of subsequent work referenced that placement. The great majority of zhelestid specimens were isolated teeth or scraps.

It is worth being accurate about what happened next, because the easy version of this story is wrong in a way that is unfair to several decades of careful people. The placental reading was not accepted without argument. Later researchers noticed features that sat badly with it, including bones in the pelvis associated with pouch-like reproduction in non placental mammals and various dental characters that looked archaic rather than modern, and some of them argued that zhelestids were eutherians outside the placental group. The debate existed. What it lacked was an animal complete enough to settle it.

That is the honest shape of the correction. Not forty years of unanimous error overturned by one heroic insight, but a long running uncertainty that nobody could close, closed at last by a better specimen.

It is worth knowing how a question like this actually gets decided, because the answer is not that somebody looked at the fossil and recognised it. The method is a phylogenetic analysis, and in practice it means building a very large table. Every relevant animal goes down one side. Every anatomical feature that varies between them goes across the top: the shape of a particular ridge on a molar, the presence or absence of a small opening in the skull, the way one ankle bone meets another. Hundreds of such characters, scored one by one against real specimens, which is why the work takes years and why it cannot be done from photographs.

Then the computer looks for the family tree that explains that table with the fewest evolutionary changes, or the one most probable under an explicit model. Characters will always conflict with each other, because convergence is real and so is loss. The point of the exercise is that the conflict becomes visible and countable instead of a matter of taste. When an animal has grinding teeth pointing one way and growing incisors, skull features and ankle construction all pointing the other way, the tree that needs the fewest coincidences is the one where the teeth are the coincidence.

That is why a partial skeleton beats a drawer of jaws. Teeth alone give you a handful of characters, all of them from the same functional system and all of them liable to the same bias. Add a skull and a hind limb and you add characters from systems that have nothing to do with diet, which is precisely what you need in order to catch diet driven convergence in the act.

Tamirkhan balcarceli is described as the most complete zhelestid found so far. That is a specific and modest claim, and it matters that it is stated that way, because it is the thing the fossil was good for. Once there is more of the animal than teeth, the teeth stop being the only witness. And the rest of this animal does not read as an early hoofed mammal at all.

It had long incisors that kept growing throughout its life, the way a rodent’s or a rabbit’s do. Several features of the skull point to zalambdalestoids. The hind limbs are long and slender, and the ankles have a distinctive construction that the authors describe as unmistakably zalambdalestoid. Long back legs and growing front teeth are not the kit of a small proto cow. They are the kit of something that hopped, or at least moved fast in short bursts, and gnawed.

Which is why the press cycle reached for the phrase Cretaceous rabbit, and why that phrase should be handled with care. The animal was not a rabbit and was not related to rabbits. It simply looked a bit like one, which is funny, given that the entire problem with its group was looking a bit like something it was not related to.

Pull zhelestids out of the placental stem and the downstream consequence is real but needs stating precisely, because the overstated version is everywhere and it is not what the paper claims. The finding does not prove that placental mammals arose after the extinction. It removes one of the fossil pillars that supported an earlier origin. The molecular clock estimates are untouched. What changes is that the physical evidence now points in a different direction from the molecular estimates, where before the two appeared to agree. That is a reopened argument, not a settled one, and reopened arguments are how the science actually works.

It also means a great deal of published literature now needs reading with a pencil. Any analysis that included zhelestids as early placentals, any timeline calibrated partly on that assumption, any textbook paragraph about hoofed mammals having Cretaceous roots: all of it has to be revisited. Not thrown out. Revisited.

The author list makes that point better than any summary could. Michael Novacek, a curator in the Division of Paleontology at the American Museum of Natural History and one of the senior figures of exactly this field, is the last author on the paper. Eva Hoffman, an incoming postdoctoral researcher in O’Leary’s laboratory at the time of the announcement, is the second. Paul Velazco of Arcadia University and Shawn Zack of the University of Arizona round it out. A field correcting its own reading, with the generation that built the reading and the generation inheriting it on the same page of the same paper, is a healthier sight than a feud.

Then there is the name, which hides the nicest detail in the whole affair. Tamirkhan is the genus. The species epithet, balcarceli, honours a chief technician who worked on numerous expeditions and who is described as a co discoverer of the animal, according to an account published after the announcement. No press release in the cycle explains the name, which is a small shame, because technicians and preparators are the invisible infrastructure of palaeontology. They are the people who find things, carry things, stabilise crumbling bone with consolidant in a tent at the end of a long day, and then spend months at a bench under a microscope removing rock from a specimen one grain at a time. A fossil that reaches a published description has usually passed through more of their hands than anyone else’s. Having one of them written permanently into the animal’s scientific name is the field’s best way of saying so.

The twenty two years deserve a word too, because people outside the science tend to hear it as neglect. It is not. A specimen comes out of the ground in a jacket of plaster and rock, and what happens next is slow by nature. It has to be prepared, which can take a year on something this small and this fragile. It has to be scanned, usually with micro computed tomography, to see the parts still buried in matrix. It has to be compared, bone by bone, with every relevant specimen in collections scattered across several continents. Then it has to be put into a phylogenetic analysis alongside dozens of other taxa and hundreds of anatomical characters, and that analysis has to survive peer review by the exact specialists whose previous conclusions it is about to revise.

And all of that has to happen in the gaps between teaching, funding applications, other projects and the rest of a working life. Giallombardo was a graduate student in 2004 and the fossil’s description is the work of his career so far. Specimens waiting in museum drawers are not forgotten. They are queued. Natural history collections are, among other things, a vast reservoir of questions nobody has yet had the time or the tools to answer, and roughly every week somewhere in the world a drawer gets opened and something that has been sitting quietly for fifty years turns out to matter.

The place matters too. The Gobi is one of the few parts of the world where Cretaceous rock is exposed at the surface over huge distances, with little soil and almost no vegetation covering it, and where the sediments happen to be fine grained enough to preserve something as small as a chipmunk sized skull. That combination is rare. It is why the region has produced so much of what is known about Cretaceous mammals and small dinosaurs, and why museums have been sending expeditions into it for roughly a century, in collaboration with Mongolian institutions and researchers whose fieldwork makes the logistics possible at all.

Those expeditions are not glamorous. Weeks of walking slowly over gravel, eyes down, in heat and wind, looking for a fragment of bone a shade different from the rock around it. Most days produce nothing. The find rate is the reason a specimen like this one is treated as precious long before anyone knows what it will turn out to mean.

A note on caution, since the paper is new. One well preserved animal and one phylogenetic analysis is a strong argument, not a verdict. Mammal palaeontologists will test the reassignment against their own data, and the usual way these things resolve is that the broad outline holds while details get argued over for years. Nothing in the available reporting carries an independent specialist’s assessment yet, and anyone telling you the question is closed is getting ahead of the evidence. What is clear is the direction of travel: the group has been moved, and the reason it was moved is anatomy rather than interpretation.

The fossil itself is reported to be around 83 million years old, which puts the animal well inside the long stretch when mammals were small, mostly nocturnal, and sharing a landscape with dinosaurs that could not have cared less about them. Picture it properly and the scene is almost domestic. A chipmunk sized creature with ever growing front teeth and long back legs, moving through scrub in what is now a stony Mongolian badland, grinding up plants with teeth so well suited to the job that 80 million years later they would fool the people who found them.

That is the ending, and it is a better one than any correction notice. The feature everyone trusted most turned out to be the one doing the misleading. Not because anybody was careless. Because teeth are the loudest voice in the fossil record and the loudest voice is not always the most reliable one. The fix did not come from a cleverer reading of the same scraps. It came from a more complete animal, which had been waiting in a drawer at the American Museum of Natural History since a graduate student carried it out of the desert, and which needed only for someone to have the time, the tools and the twenty two years it takes to let an animal finish its sentence.

In a normal year, about twenty-five wild Atlantic salmon find their way back from the open ocean into the Narraguagus River in the far eastern corner of Maine. Not twenty-five thousand. Twenty-five. You could carry the entire returning run of a forty-mile river in the back of a pickup truck, and there would be room left over.

That number is the reason a small town called Cherryfield held a special meeting a few years ago and voted, with nobody against, to accept ownership of a dam from the United States federal government for the specific purpose of destroying it. It is the reason Congress later wrote the dam out of existence in a water resources bill. It is the reason roughly 9.1 million dollars of federal money has spent this year being turned into a staircase of boulder pools on the riverbed, built by a road contractor and designed by river engineers, so that a fish can climb a hill.

And it is the reason the people doing all of this will look you in the eye and tell you it might not be enough.

That last part is the unusual thing. Conservation projects do not normally come with their own warning label. Somebody cuts a ribbon, somebody says the word hope, and the press release ends. This one is different, because the biologists and managers attached to the Narraguagus have been remarkably unwilling to promise anybody a happy ending. They will tell you what they are building and why. They will also tell you that the thing most likely to kill these fish is happening a long way out to sea, where no amount of concrete and no number of boulders can reach it.

Start with the dam, because the dam is stranger than you would guess. It was not built for electricity. It was not built to hold drinking water or to float logs. The Army Corps of Engineers put it across the Narraguagus at Cherryfield in 1961 for one reason: ice. The river freezes, the ice breaks up, the broken ice piles into jams, and the jams push water into the town. The dam was there to catch the ice before it reached people’s houses. It did that job, more or less, for more than six decades.

Somebody in 1961 did at least think about the fish. The structure had a fish ladder built into one side of it. The trouble was that only some fish could read it. Alewives, small and abundant and determined, worked their way up it reasonably well. Adult salmon, according to the state fisheries officials who watched them for years, had a habit of gathering on the wrong side, the side with no ladder, swimming in circles against a wall and never finding the door that was twenty metres away. Shad, which are fussier again, mostly did not get through at all.

So the dam aged into a problem with two halves. It still had an ice job to do, and the town still wanted that job done. It also sat across the throat of one of the last rivers in the United States where wild Atlantic salmon still exist at all. Removing it outright would have handed Cherryfield its ice jams back. Leaving it was a slow sentence on the fish.

What eventually came out of a five-year Army Corps study was a third answer: take the dam away and replace it with something that is not really a dam. The term the project uses is a nature-like fishway, which sounds like marketing until you see the drawings. It is a long, gradual channel of stepped pools made of large boulders, falling just a few centimetres from one pool to the next, with a water control structure and timber ice piers worked into it. The gentleness is the whole point. Spread the same drop in elevation over a much longer distance and a salmon can swim up it the way it would swim up a natural run of rapids, resting in each pool on the way. The piers and the structure still slow ice down before it reaches the town. The town keeps its protection. The fish get a route that does not require them to find a single hidden slot in a wall.

Getting from that idea to a construction site took more than ten years and an unusual chain of permissions. The residents of Cherryfield had to agree to take the dam off the federal government’s hands, knowing that what they were voting for was its demolition. They did, unanimously, at a special meeting. Corey Smith, a Cherryfield selectman in his thirties, has been plain about the result: nobody voted against it. Congress then de-authorised the structure, which is the legal step that lets a federal flood control project stop being a federal flood control project. NOAA’s Restoration Center put up about 9.1 million dollars through its fish passage grant programme, funded under the federal infrastructure law, with the money going to the Downeast Salmon Federation. The Nature Conservancy, Maine Coast Heritage Trust and the Atlantic Salmon Federation came in alongside the town. Inter-Fluve drew it. Sargent Corporation built it.

Note how many names that is. Four conservation organisations, three layers of government, two companies and a town meeting. No single one of them gets to call this theirs, and the project has been careful not to let any of them try.

Somewhere in the middle of all those logos is Ellie Mason, who is twenty-nine and whose job title is habitat restoration project manager at the Downeast Salmon Federation. In practice her job this year has been to stand on a river bank in a hard hat and be the person in the room who thinks like a fish. Construction sites generate a hundred small decisions a week about where a machine goes, how deep something sits, what gets moved and when, and most of those decisions have a consequence for an animal that is not there to argue. She is the argument.

Mason grew up on this river. And the sentence she uses about that is the one that should stay with you, because she says it without any drama at all: the thing about growing up here is that there are no salmon. She is not describing a loss she witnessed. She is describing an absence she inherited. By the time she was old enough to notice a river, the fish had already been functionally gone for a generation. She is building a door for an animal she has never seen use it.

She is not the only person on the project carrying something like that. Catherine Morse, who manages habitat restoration for the Atlantic Salmon Federation, grew up with a family story instead: her great-grandparents used to buy Cherryfield salmon for special occasions, back when buying a local salmon for a birthday was an ordinary thing to do. The recreational fishery closed in the late 1990s. Atlantic salmon were listed as endangered in 1999. Three generations of one family sit inside that detail, and only one of them had the fish.

It helps to know how far back the damage goes, because it is not a recent story and nobody should pretend it is. In the late 1800s, a single one-mile stretch of the Narraguagus near the head of tide in Cherryfield held one of the largest concentrations of water-powered mills in the whole state, according to a local historical account published in the 1970s. Dams, mills, log driving and two centuries of fishing did the original work. Before all of it, hundreds of thousands of Atlantic salmon are thought to have come back into Maine’s rivers in a year. The figure today, for the entire state, is around a thousand.

Which brings us to the arithmetic, and the arithmetic is where this stops being a nice story about a town and a dam.

A study published last spring in Frontiers in Ecology and Evolution looked at the Narraguagus specifically, using smolt trapping, adult trapping and spawning nest counts gathered over more than two decades. The authors are NOAA and Maine state fisheries scientists, among them Ernie Atkinson of the Maine Department of Marine Resources and James Hawkes of NOAA, both of whom also appear in the local reporting on the fishway.

Their numbers are brutal and worth reading slowly. Natural spawning plus conservation stocking on this river produces an average of about 1,700 smolts a year, the young fish that head out to sea. That is well under the production target. Of the naturally reared smolts, a little under one percent came back as adults after two winters at sea. Of the hatchery-raised smolts, the figure was five hundredths of one percent. Read those two side by side: a fish that hatched in the river was roughly twenty times more likely to survive the ocean and return than a fish raised in a tank. The average number of naturally originating adults coming back each year, across the whole study, was twenty-five.

So the twenty-five is real, it is wild, and it is the same twenty-five whether you count it at the trap or estimate it from survival rates. The paper’s conclusion is not about rivers at all. It is that low survival at sea is the primary obstacle to recovery.

Hawkes has put the problem in one line that functions as the honest version of every habitat press release ever written: all this restoration work is fantastic, but you need fish. Hatcheries have kept the population from blinking out, and they have not grown it. They are life support, not a cure.

Set that against what recovery officially requires. The federal recovery plan for this population of Atlantic salmon says that to move them from endangered to merely threatened, you need total annual returns of at least 1,500 wild-origin adults. To take them off the list altogether, you need at least two thousand wild adults returning to each of the three recovery units, six thousand across the whole population, sustaining themselves. The plan’s own estimate of how long that would take is seventy-five years, which it works out as roughly fifteen salmon generations. Its cost estimates run to tens of millions of dollars a year, and the reporting on it has put the total bill above 800 million dollars.

Twenty-five against fifteen hundred. Fifteen generations. Nobody now working on the Narraguagus will be alive to see whether it worked.

And the fishway cannot touch the reason it is failing. Valerie Ouellet, who runs research and environment for the Atlantic Salmon Federation, has made the genetic point: a population needs enough individuals to carry real diversity, including whatever heat tolerance exists in the gene pool, and heavy dependence on hatchery fish works against exactly that. Morse has made the blunter one: if warming water pushes the salmon’s food out of the places the salmon go to feed, no human intervention fixes that. Atlantic salmon start to suffer above about twenty degrees Celsius, sixty-eight Fahrenheit. The Narraguagus passes that in summer, regularly, and the warm stretch of the year is getting longer. The same study found the young fish are now leaving for the sea about three and a half days earlier per decade, which is a river telling you the calendar has moved.

All of which makes the quieter, older half of this project the part worth respecting most. Long before anyone touched the dam, a group called Project SHARE went looking for the smaller blockages, the road culverts that are too high or too narrow or too fast for a fish to pass. On this one river they identified sixty-one of them and set about replacing them. Christopher Federico, who runs the organisation, has also overseen a programme with the least glamorous name in conservation, chop and drop, which means felling large trees into the river on purpose so the channel narrows, scours deeper and runs cooler. Two decades of culverts and deliberately dropped trees. No ribbon, no press conference.

That is the context the nine million dollars sits in. It is not a rescue. It is the last and most expensive link in a chain of unglamorous work, and the people who built the chain are the ones saying out loud that the ocean may undo it.

As of now, the fishway is not finished. That matters, and it is worth saying clearly, because plenty of coverage has already slid into the past tense. The old dam has indeed come down. In-water construction ran through the summer and into the autumn, and the schedule has it wrapping up around the turn of the season, with the renovated park due to reopen before the end of the year and replanting carrying on into next summer. In the final weeks of in-water work the crews were setting the last of the big boulders into the top step pools, pouring concrete foundations for the ice piers and building the timber structures above them, with the first controlled release of water down the new stepped channel, the removal of the sandbags at the downstream end and the connection of the new channel to the old riverbed all still ahead of them. The fishway is nearly built. It is not built.

Nor will anybody know quickly whether it works. You cannot count twenty-five salmon and learn anything; the number is too small to tell signal from noise. So the monitoring is being done with a different fish and a cleverer method. Shad are even more sensitive to bad passage than salmon are, which makes them a good canary. Researchers are taking water samples above and below the site and sending them to be tested for traces of shad DNA. The baseline, taken in the spring before the dam came out, showed what you would expect from a barrier: more shad DNA below it than above it. They will sample again next summer, and what they want to see is the concentrations evening out, which would mean fish are getting past. A helicopter with an infrared camera has also been mapping the river’s temperature, looking for the cold tributaries where fish can shelter when the main stem gets too warm.

That is the honest shape of it. A town gave up a federal flood defence and took on the liability of knocking it down. Nine million dollars became a boulder staircase. The measure of success for the first year is whether some shad DNA shows up in a water sample. And the fish everyone is actually doing it for number about two dozen, with the main threat to them sitting somewhere in the North Atlantic, outside anybody’s permit.

You could read all that as futility. The people on the river bank do not, and their reasoning is practical rather than sentimental. Marine survival is the thing they cannot control, so the obvious response is to be ruthless about the things they can. Make the freshwater side as good as it is possible to make it. Open every mile of it. Cool it where cooling is possible. Then, if conditions at sea ever improve, even slightly, the fish come home to a river that is ready rather than a river that is still broken. If the ocean never improves, the river was still worth fixing, and the shad and the alewives and the sea-run trout get the benefit either way.

Which is why the right place to end this is not with a ribbon and not with a fish. It is with a twenty-nine-year-old standing in a construction hard hat beside a river in her home town, doing the day-to-day work of arguing on behalf of an animal that was already gone before she was old enough to miss it, on a project whose own timetable for success is longer than her working life, and which the scientists attached to it will not promise her is going to work. She is building the door anyway. Somebody has to be the one who puts it there before anything arrives, and that person never gets to see the arrival.

The usual version of a new species story goes like this. Somebody walks for days into a forest almost nobody has been into, lifts a log, and finds an animal science has never seen.

This is not that. The animal at the centre of this story is a stick insect roughly the length of your hand, living in coastal scrub and suburban gardens on the east side of Australia. People have photographed it on fence palings and lemon trees. Insect enthusiasts have kept it and bred it. It sits in museum drawers with a label on it. It has been logged in public nature-recording databases by ordinary people who went out into their own backyards with a phone. The earliest record of it that anybody can now point to dates to 1976.

Nobody had given it a name. For about fifty years, every single one of those photographs, specimens and records was filed under the wrong species.

It has one now. It is Anchiale robusta, and the work that gave it a name was led by Dr Braxton Jones, a postdoctoral researcher in molecular ecology and evolution at the University of Sydney’s School of Life and Environmental Sciences. The paper, a full revision of the Australian members of the genus Anchiale, went up in the journal Austral Entomology in the second half of September, and it did considerably more than name one backyard insect. Before the revision, Australia was thought to have two species in the genus. After it, there are five. Two of those five are new to science. One is a species that had been described long ago, written off, and has now been reinstated.

The reason this is a better story than a jungle discovery is precisely that nothing was hidden.

Think about what fifty years of misfiling actually means in practice. It means that every time somebody uploaded a photograph of this insect to a citizen-science platform, the identification that came back attached it to a different animal. It means that the hobbyists who were rearing it and raising its eggs believed they were rearing something else. It means that when a conservation assessor or a state environment department went looking at what lived in a particular patch of coastal vegetation, this species did not appear on the list, because as far as any list was concerned it did not exist. It had a stand-in. The stand-in absorbed all its data.

The specific error has a traceable origin, which is unusual and rather satisfying. According to a detailed account of the paper, the 1976 record of what is now Anchiale robusta was an egg case, and that egg case was identified as belonging to a species then called Ctenomorphodes briareus, which was itself later shuffled into the genus Anchiale in 2007 and became Anchiale briareus. So the mistake did not start with somebody misreading a whole insect. It started with somebody looking at an egg, reasonably, and reaching for the closest thing they knew. Everything downstream inherited it. Fifty years of records built on top of a single plausible call about a small brown object.

What finally broke it open was not one technique but five braided together, and the braid is the interesting part.

The first strand was DNA. Sequencing lets you ask whether two populations that look alike have actually been exchanging genes, and the answer for a stick insect is frequently no. Stick insects are champions of convergent appearance. Their entire evolutionary strategy is to look like a twig, and a twig does not offer much scope for variation. Two lineages that separated millions of years ago can end up looking almost identical because both of them are solving the same problem, which is how not to look like food. Genetics cuts straight past the twig problem.

The second strand was museum collections. Specimens in drawers carry locality labels and dates, sometimes going back more than a century, and they are often the only physical evidence of what lived where before a place was cleared or built on. A revision like this one lives or dies on whether somebody is willing to go through the drawers.

The third strand was field survey, which means somebody going out at night with a torch and looking at foliage, because that is still how you find a phasmid.

The fourth strand was citizen science. Thousands of ordinary people photographing insects in their gardens and uploading the results produces something a research team could not generate on any budget: a dense, dated, geographically spread picture of where a common animal actually is. The limitation of that data is that it is only as good as the names attached to it. The strength is that the photographs are the photographs regardless. Once you know what you are looking at, every one of those records becomes a usable data point retroactively. That is a genuinely new feature of modern taxonomy, and it means the hobbyists and backyard naturalists in this story are not colour. They are part of the evidence base.

The fifth strand was captive populations. The insect being kept and bred by enthusiasts gave the team access to something the wild rarely offers on demand: the full life cycle, the eggs, the nymphs, the adults, in a known lineage, observable for as long as you need. For an animal defined partly by features of its eggs, that matters enormously.

It also helps to know what a name actually costs to produce, because the phrase “described a new species” makes it sound like a declaration. It is closer to a legal filing. To name a species you have to designate a holotype, a single physical specimen that from that moment forward IS the definition of the name. If anybody ever disputes what Anchiale mabiensis means, the answer is not an argument, it is a particular dead insect in a particular collection with a particular number on it. You have to publish a description detailed enough that somebody who has never seen your animal can tell it apart from its relatives. You have to work out whether any of the names already sitting in the literature, some of them from the 1800s, might already apply to your animal, because if one does, that older name wins and yours is invalid. That last step is why revisions resurrect species. Somebody goes back through the old names, finds one that was lumped in with another animal by a nineteenth-century author working from a damaged specimen, and establishes that it was a real thing after all.

The muddle in this particular genus had help from geography. Anchiale is not purely Australian. It ranges through New Guinea and the islands to the north, and genera with that kind of spread tend to accumulate a messy history, because specimens arrived in European museums over a century and a half from scattered collectors who could not compare notes. Australia’s share of the genus ended up treated as two species largely because nobody had sat down with all of the material and all of the tools at once. Phasmids as a group are also chronically under-described relative to how many of them there are, partly because they are nocturnal, partly because they are hard to find by design, and partly because the people who know them best have often been enthusiasts rather than salaried specialists.

Which brings us to the other new species, and to the strangest detail in the entire paper.

Anchiale mabiensis lives in Queensland’s Mabi rainforest, on the Atherton Tablelands in the far north of the country. Mabi is not a large place. It is a distinctive type of complex vine forest that once covered a decent spread of the tablelands and has been reduced to scattered fragments, with something in the order of 1,050 hectares left as of a 2021 assessment. That is a critically endangered ecological community by any measure. A new large insect described from it is a conservation story wearing a taxonomy story’s clothes.

The females are substantial animals. One measured out at 16.3 centimetres when stretched, which is to say roughly the length of a toothbrush, and that is not even close to the Australian ceiling; Ctenomorpha gargantua, from the same part of the country, can reach around 60 centimetres and sits among the longest insects known anywhere.

But the thing that makes entomologists sit forward is the egg.

Stick insect eggs are already one of the more remarkable objects in the insect world. They do not look like eggs. They look like seeds, often very convincingly, down to surface texture and the little attachment structures that some plants use to persuade ants to carry their seeds underground. Some stick insects drop their eggs onto the forest floor and the ants do exactly that, burying them in a nest where they are safe from fire and from most things that eat eggs. The egg also has a hatch in the top, called an operculum, a genuine lid that the emerging nymph pushes open from inside. All of that is standard phasmid equipment and has been described for a very long time.

What Anchiale mabiensis has is a structure in that lid that has not been recorded in any stick insect, or in their closest relatives, before. The published description refers to inner rings on the lid of the egg. The team’s framing is that it is unlike anything previously documented in the group.

Nobody is claiming yet to know what it does. That is the honest state of it, and it is worth sitting with rather than rushing past, because it is a fair description of how this kind of science actually proceeds. You find a structure. You describe it accurately. You note that nothing else has it. Then somebody, possibly years later and possibly not you, works out what it is for. There are plausible directions. The lid is the point of mechanical failure on a sealed container, so a reinforced or differently constructed lid might be about surviving burial, or about controlling water loss in a forest that has been chopped into hot, dry-edged fragments, or about timing: holding the nymph in until conditions outside are right to come out. Those are guesses and the paper does not make them. What the paper establishes is that the structure is real and that it is unique as far as current knowledge goes.

It is also sitting inside roughly a thousand hectares of remnant forest.

The eggs of the backyard species are described too, which gives a sense of how fine the distinctions in this work can be. Anchiale robusta lays a darker brown egg, glossy when it is freshly laid, which becomes bumpy and matte as it ages. These are the characters a revision turns on. Not dramatic differences in a large animal, but the surface finish of an object a few millimetres across, read correctly and checked against DNA.

The author list is worth a moment, because it is not the list you would expect from a university molecular lab. Alongside Jones, the paper draws in Paul D. Brock, a stick insect specialist with decades of work on the group and a reputation built substantially outside the conventional academic track; Ross M. Coupland, who took habitat photographs in the Queensland rainforest for the study; Angus J. Emmott, a Queensland grazier and naturalist associated with James Cook University, who photographed the female holotype of the new rainforest species; and Nathan Lo, who runs the molecular side at Sydney. A land manager, an independent specialist, a field photographer and a genomics lab, on one paper. Australian natural history has always worked like this to a degree that surprises people from elsewhere, and the serious amateurs are not a quaint tradition. They are load-bearing.

The funding tells its own small story: an Australian Biological Resources Study National Taxonomy Research Grant, plus a CSIRO postgraduate scholarship. There is a standing complaint in biology that descriptive taxonomy has been quietly starved for a generation, treated as a finished nineteenth-century chore while money went to work that looks more modern. A paper that takes a well-studied genus in a well-studied, heavily populated part of a wealthy country and finds that the true number of species is more than double what everybody thought is a fairly direct argument about what that starvation costs.

Because the practical consequence is not academic. An animal without a valid name cannot be assessed for extinction risk. It cannot be listed. It cannot be written into a protection measure or a development condition or a recovery plan, because legal instruments operate on names. For fifty years, Anchiale robusta had no name of its own, which means that for fifty years it was invisible to every mechanism a country has for noticing that something is in trouble. It happens to be common and coastal and probably fine. The new rainforest species is neither common nor widespread, and it was in exactly the same position until September.

Jones’s own framing of why the work matters is unfussy and lands harder for it: that this is about protecting threatened ecosystems and getting Australia’s biodiversity documented before species disappear without ever having been discovered at all.

Which is the part that should actually unsettle anybody who reads this far. The headline version of the story is the insect in the garden, the funny fifty-year clerical error, the animal people had in a tank and misnamed. That is the hook, and it is a good one. But the insect in the garden was never the one at risk. It was photographed thousands of times precisely because it is everywhere, and all it lost was its name.

The one with the lid on its egg, the structure nobody has seen anywhere else in the group, lives in a forest that has been cut down to fragments. It got its name in September, at the age of however many million years, with about a thousand hectares of the world left that it is known to live in. Fifty years of looking straight at a common insect and not seeing it is a story about human attention. The other one is a story about timing, and we very nearly did not make it.

In a back room at the Aquarium of the Pacific in Long Beach, California, somebody spreads a paste of spirulina, fish food and powdered calcium onto small plates and lowers them into tanks of tadpoles. The plates are there so the tadpoles can graze the way they would graze algae off a wet rock in a mountain stream. It happens four or five times a day. Every so often the water flow in the tanks gets turned up, not because the tadpoles need it but because the frogs they are about to become will need strong legs and will need to have felt moving water before they meet it for real.

The animals being fed are southern mountain yellow-legged frogs, and there are fewer than two hundred of them left in the wild.

That figure is the one to hold onto, because everything else about this story pulls against it. Erin Lundy, the aquarium’s conservation initiatives manager, runs the facility where these frogs are grown. She and her team passed a milestone this year: roughly a thousand frogs raised in Long Beach and let go into the mountains above Los Angeles since 2020. One report puts the running total at 1,032. It is a real number and a good one. It is also not the same thing as a thousand more frogs living in the San Gabriel Mountains, and the people who do this work are careful about the difference in a way that the headlines around them often are not.

To understand why a fish tank in a coastal city has become critical infrastructure for a mountain amphibian, you have to go back through three separate things that went wrong, none of which had anything to do with the others.

The first was trout. Starting in the late 1800s, non-native trout were put into high mountain lakes and streams across California so that people could fish them. Those waters had been fishless, and the frogs that bred in them had spent a very long evolutionary history without anything in the water that wanted to eat their young. Tadpoles of this species stay in the water for a year or more before they transform, which is a long time to be small and slow in a stream that now contains a predator built for exactly that. Across large parts of the historic range, the frogs simply stopped recruiting. Adults hung on in places for a while, and then they did not.

The second was a fungus. Batrachochytrium dendrobatidis, usually shortened to chytrid, is a waterborne pathogen that infects amphibian skin. Amphibians breathe and regulate water and salts through their skin, so an infection of the skin is not a surface problem for them. Chytrid has moved through amphibian populations on most continents over the past several decades and is associated with declines and extinctions on a scale that has few parallels in modern wildlife disease. The southern mountain yellow-legged frog turned out to be badly susceptible. Tali Hammond, a recovery ecology researcher with San Diego Zoo Wildlife Alliance, has described chytrid as still one of the biggest obstacles to this species recovering, especially in Southern California. It has not been solved. It has been worked around.

The third was fire. In late 2020 the Bobcat Fire burned through the San Gabriel Mountains and took something in the order of 95 percent of the habitat the species had left. By then the population had already been squeezed into a handful of isolated stream reaches. A fire of that size across that small a footprint is not a setback in the usual sense. Biologists pulled around 130 frogs out ahead of the flames, which tells you how thin the margin had become: the emergency response to a wildfire was to physically carry a significant fraction of a species out of the mountains by hand.

Put those three together and you get the arithmetic that the nurseries exist to answer. One count in 2001 put the wild adult population at 79 animals. By 2021 the estimate was under two hundred. Across the species’ historic range, something like 92 percent of the ground it used to occupy no longer has it. What survives sits in roughly ten isolated locations spread across three mountain ranges, the San Gabriel, the San Bernardino and the San Jacinto. Ten separated pockets is a dangerous shape for a population, because a single bad year in any one of them has nowhere to recover from.

So the work became head-starting. The idea is simple and the execution is not. Eggs and tadpoles are brought into captivity, grown past the stage at which a trout or a drought or a disease would most likely kill them, and released as four-limbed juveniles with a fighting chance. In the wild, a tadpole might spend one to three years in the water before transforming. In a tank with controlled temperature, steady food and no predators, that same animal gets through the bottleneck faster and in better condition. At the Long Beach aquarium, somewhere between 60 and 65 percent of the tadpoles make it to the point of having four legs, no tail and a release date. Lundy has described that rate as high. For an amphibian starting at the egg, it is.

The reason the rooms are built like clean rooms is the fungus. A nursery concentrates a lot of highly susceptible animals of a single species in a small volume of water. If chytrid gets in, it does not take a few individuals, it takes the batch. So the newer facilities are what the people running them call bio-secure, meaning controlled water, controlled equipment, dedicated tools per tank, staff who do not carry anything between rooms, and a hard separation from anything that has been in a stream. The Los Angeles Zoo opened a bio-secure amphibian head-starting facility and it is the reason the zoo’s numbers changed this year. It is an unglamorous building and it is the single most consequential piece of kit in the programme.

Which brings up the thing that makes the public accounting of all this genuinely confusing, and it is worth being plain about rather than tidying away.

Over two days late in the summer, biologists released more than 650 southern mountain yellow-legged frogs into streams in the San Gabriel Mountains. The Los Angeles Zoo described it as 653 frogs. Press coverage out of Long Beach described it as more than 650, of which nearly 400 were raised at the Aquarium of the Pacific. Both are correct and they are the same release. The zoo contributed 227 animals from its new bio-secure facility, the aquarium contributed 394, and the rest came from elsewhere in the partnership. Nobody released 653 frogs and then separately released another 400.

The same thing happens at the next scale up. San Diego Zoo Wildlife Alliance reported releasing more than 3,000 frogs and tadpoles this year, into the San Bernardino and San Jacinto mountains, and has produced over 10,000 animals since 2006. The Los Angeles Zoo has bred and released close to 7,000 since its programme began in 2007. Those totals overlap in partners and in release sites but they are not additive in any simple way, because one institution is counting animals it raised, another is counting animals it released, and a third is counting animals that passed through its building on the way to somewhere else. A reader can handle the fact that the zoo counts releases and the aquarium counts animals raised. What a reader cannot handle is a single invented number that is wrong.

There is also a naming trap in this story that has tripped up more than one outlet. There are two mountain yellow-legged frogs. Rana sierrae lives in the Sierra Nevada. Rana muscosa, the southern animal, is the one in the mountains behind Los Angeles and San Diego. At least one account of this year’s releases placed them in the Sierra Nevada. They were not in the Sierra Nevada. They were a short drive from a metropolitan area of several million people, which is one of the quieter oddities of the whole thing: the last ground of an endangered mountain frog is about forty minutes from freeway traffic.

The partnership doing this is larger than any one logo on it. Alongside the Long Beach aquarium, the Los Angeles Zoo and San Diego Zoo Wildlife Alliance, the recovery programme pulls in Omaha’s Henry Doorly Zoo and Aquarium, the United States Geological Survey, the Forest Service, the Fish and Wildlife Service, the California Department of Fish and Wildlife, the Santa Ana Zoo, Big Bear Alpine Zoo and Birch Aquarium at Scripps. Byron Wusstig, curator of ectotherms at the Los Angeles Zoo, has made the case for the controlled-environment approach in straightforward terms: raising sensitive amphibians in a specialised, temperature-controlled setting markedly improves the odds they survive. The zoo also sent 112 tadpoles on to the Santa Ana Zoo, which is how a programme like this spreads its risk. If one facility has a bad year, the species does not.

Birch Aquarium joined in 2024 and released 223 frogs this year, 74 of them at a creek near Idyllwild in the San Jacinto Mountains and 149 at a second site a few weeks later, bringing its own running total past 440. Sean Bruce, the assistant curator of fishes and invertebrates there, has talked about the particular satisfaction of watching an animal you raised from a tadpole make the last jump out of the container. It is the sort of thing that sounds like press-release warmth until you have seen how many of these animals do not get that far.

There is a detail in the Birch Aquarium story that says more about what these programmes are actually attempting than any release total does. Before the San Jacinto work, the aquarium’s first reintroduction went to Bluff Lake, in the San Bernardino Mountains, inside a wildlife preserve managed by The Wildlands Conservancy. The species had last been recorded at Bluff Lake in 1951. Tim Krantz, the conservancy’s conservation director, was part of putting them back. That is not a population being topped up. That is an animal being returned to water it had been absent from for seventy years, and it is the shape the whole effort is trying to achieve at ten sites at once.

What that costs, day to day, is worth saying out loud, because the romance of releasing frogs into a mountain stream obscures how much of this work is janitorial. Bio-security means dedicated nets and siphons per tank, water that is treated and monitored rather than plumbed in, staff who change what they are wearing between rooms, and an absolute rule against bringing anything back from a stream into the building. It means somebody is checking temperatures and flow rates on a weekend. It means the plates of spirulina paste go in four or five times a day, every day, for the year or more that a tadpole takes to become a frog, and that if the schedule slips the batch notices. The reason the newer facilities are built this way is not fastidiousness. It is that a nursery is the single best place a fungus could ever hope to find: hundreds of susceptible animals of one species, sharing water, in a warm room.

And at that first San Jacinto site, the field team found something more useful than a feeling. They saw several adult frogs from earlier San Diego Zoo Wildlife Alliance releases, alive, grown, in the creek. They managed to tag one for future tracking.

That is the real news inside all of this, and it is smaller and harder than the big release totals. Released animals are surviving to adulthood in the wild. What none of this year’s reporting establishes is the next step, which is whether those surviving adults are breeding and producing young that themselves survive. Release is not recruitment. A programme can put thousands of animals into streams every year for a decade and still be running a very expensive conveyor belt rather than a recovery, if the wild population never starts replacing itself. The federal recovery plan for the species, published in 2018, sets a long-term target of 25 self-sustaining populations. There are currently about ten pockets, propped up.

Nobody involved pretends otherwise. The strategies being layered on top of the releases are an admission of exactly this problem: genetic management so the captive stock does not become a bottleneck of its own, antifungal treatment of animals before and after release, and deliberately scouting drier release sites where chytrid is less able to persist. That last one is a quiet concession. It means choosing habitat not because it is the best habitat for a frog, but because it is the worst habitat for the thing killing frogs.

Which is why the thousandth frog out of Long Beach is a hook and not an ending. It marks the point at which one aquarium has demonstrated it can reliably manufacture juvenile frogs. It does not mark the point at which the mountains can keep them.

But there is one moment from a release this year that is worth carrying out of the story, and it has nothing to do with totals. Lundy was at a stream in the San Gabriels watching the containers open. One frog, raised entirely indoors on a diet spread on a plate by a human being, jumped out, landed on a rock, and ate a housefly.

It had never seen a housefly. Nobody taught it. Whatever it is that makes a frog a frog had come through the tanks and the plates and the adjusted water flow completely intact, and it switched on in the first second of contact with the actual world. The nursery had not produced a pet. It had produced a frog, and the frog knew what to do.

Whether the mountain lets it keep doing that is the part still being decided.

The animal weighed something in the region of half a kilogram, about as much as a large orange, and once the anaesthetic took hold the team had fifty minutes. Not fifty minutes to do the injection. Fifty minutes to do everything, because that was the outer edge of what a monkey that small could safely be kept asleep for, and because the vaccine itself was on a separate clock: once diluted, it stops being any use after three hours. Two timers running at once, in a tent in coastal rainforest, on an island off the southern coast of Brazil, over an animal from a species that numbers fewer than four hundred individuals and has not one single representative in captivity anywhere on earth.

That last detail is the one to hold onto, because it changes the character of everything else. Most endangered species work carries a quiet safety net underneath it. If a field procedure goes wrong, if a population crashes, if a disease sweeps through, there is a captive group in a zoo or a breeding centre somewhere that can supply animals for a reintroduction. The black-faced lion tamarin has no such group. There is no insurance copy. Every one of them is out in the forest, and every decision taken over a sedated one is a decision that cannot be unwound.

In July, a team from SPVS, the Brazilian non-profit whose name translates roughly as the Society for Wildlife Research and Environmental Education, caught three of them on Superagui Island in the state of Parana. Two males and one female. Veterinarian Paula Mangini, who works for the organisation, gave each of them a yellow fever vaccine adapted from the formulation given to humans. According to the reporting on it, this was the first time the species had ever been vaccinated against anything.

It is worth being precise about what the first actually was, because this is the sort of claim that drifts. Lion tamarins have been vaccinated against yellow fever before. Golden lion tamarins, a different and better known species from the forests near Rio de Janeiro, were given doses from late 2020 onward in what was itself described at the time as an unprecedented programme. What happened on Superagui was a first for Leontopithecus caissara specifically, the black-faced or caicara lion tamarin, and not a first for the genus or for wild primates generally. The distinction matters because the golden lion tamarin work is exactly what made the Superagui work possible. Somebody had already shown it could be done.

Now the numbers, which deserve a moment of care because three different figures circulate and they are not three attempts at the same measurement. A survey in 2011 produced a count of 392 individuals in the wild. The IUCN, working from that same survey, estimates roughly 250 mature adults, which is a smaller number because it deliberately excludes juveniles and counts only the animals capable of breeding. SPVS, on its own account of the programme, says around 400 have been recorded in the wild. Those are a population census, a breeding-adult estimate, and an organisational working figure respectively. Averaging them would produce a number that means nothing. The honest summary is that the total is somewhere under four hundred, the breeding core is probably a good deal smaller, and the most recent proper survey is now well over a decade old.

The species itself was only described scientifically in 1990, by Vanessa Persson and Maria Lucia Lorini. That is recent enough to be slightly disorienting. A reasonably large, strikingly coloured primate with a black face and a mane of golden fur went formally unrecognised until a human generation ago, and it is plausible that it will be gone again within another one. It lives in a narrow strip of coastal Atlantic Forest straddling the border between Parana and Sao Paulo, in a habitat called restinga, a sandy low forest that grows on old coastal deposits. Superagui National Park holds the largest known population. There is a second cluster around Ariri and the Lagamar de Cananeia State Park further north.

So why bother vaccinating against yellow fever at all, when the forest has not been emptied of tamarins by it?

Because of the howler monkeys. SPVS staff have described the reasoning plainly, and it is the kind of evidence that is impossible to unsee once you have looked at it. The team knows yellow fever has passed through Superagui Island in recent years, and the way it knows is that the howler monkeys disappeared. Howlers are exceptionally susceptible to the virus. They die fast and in numbers, which is why public health services across South America watch for howler die-offs as an early signal that yellow fever is circulating in an area. An absence of howlers where there used to be howlers is not a mystery. It is a reading on an instrument.

And the virus does not respect the boundaries between primate species. The outbreak that moved through southeastern Brazil in 2017 and 2018, concentrated mainly around Rio de Janeiro, killed howlers, marmosets, capuchins and golden lion tamarins. That outbreak is the direct reason the vaccine work began at all. Nobody has ever recorded a black-faced lion tamarin dying of yellow fever. Edson Montilha, a biologist with the Fundacao Florestal in Sao Paulo, has said so directly: there is no registered case. He frames the vaccination as an extra strategy held in reserve against the possibility of a new outbreak, rather than a response to one in progress.

That is a harder position to occupy than it sounds, and it is the real decision at the heart of this story. The team is not treating sick animals. It is taking a measurable, immediate risk with individual members of a species that has no margin, in order to reduce a risk that is serious and well evidenced but has not yet arrived. Catching a wild primate and putting it under general anaesthetic is not a trivial act. Animals die under anaesthesia. They die of capture stress. They are injured in handling. When there are only a few hundred of something and no captive reserve, losing one in the course of protecting it is not a statistical footnote, it is a meaningful fraction of what exists.

So the team built the procedure to be as short as it could possibly be made, and then filled it to the edges.

The vaccine is a live attenuated virus preparation, an adaptation of the human product made by Bio-Manguinhos, the immunobiological arm of Brazil’s Fiocruz foundation. It came through the national action plan for Atlantic Forest primate conservation, with researchers from Fiocruz and from ICMBio, the federal body that runs Brazil’s protected areas and issues the permissions for this kind of work. The three-hour window is a property of the product. Reconstituted yellow fever vaccine is notoriously perishable, which is why human vaccination campaigns are organised around getting an opened vial used up quickly. On an island with a field station set up in sandy coastal forest, that clock starts the moment the dilution is done and does not care how the trapping is going.

The fifty-minute clock is a property of the animal. A primate of five hundred to seven hundred grams has very little thermal mass and very little margin on blood pressure, and the longer it is held under, the more the arithmetic turns against it. So everything that needs doing while an animal is unconscious has to be sequenced, rehearsed and done once. Blood samples, drawn for baseline health screening and for the immunological work. Hair samples. Faecal samples. A microchip, so the individual can be identified for the rest of its life. A radio collar, which is the single most consequential thing on the list, because without telemetry the team has essentially no way of finding that particular animal again in dense coastal forest. The collars are how they locate groups to work with in the first place. One capture has to deliver what would otherwise be a year’s worth of separate visits, and it has to fit inside the time it takes to drink a cup of coffee slowly.

None of this was improvised. The chain runs back through a sequence of multi-institutional workshops held between 2021 and 2023, where SPVS, ICMBio, Fiocruz and Bio-Manguinhos worked through the health threats facing the species and what could be done about them. Then, at a planning meeting in April 2025, that process produced something more binding than a recommendation. SPVS wrote a standing technical protocol: from that point on, every black-faced lion tamarin captured in the course of field research must be vaccinated against yellow fever.

Read that again with the captive population in mind. The organisation did not decide to vaccinate three animals. It decided, in advance and on paper, that it would take the anaesthesia risk every single time it got its hands on one of these animals, indefinitely, on a species with no spares. A conservation decision made in a meeting room over a year before it became a sedated animal on a table.

Which brings us to the part that is easy to skip over and is actually the point. Three animals were vaccinated. Three, out of fewer than four hundred. Less than one per cent of the species. On its own, that does nothing whatsoever for the population’s odds against a yellow fever outbreak. If the virus came through Superagui next year, three immunised individuals would not save the island’s tamarins.

This was not a vaccination campaign. It was a test, and it should be read as one.

The confirmation of that came in late August, when the team went back out and recaptured the same three animals, the two males and the female, in order to draw blood again. The purpose was to find out whether the vaccine had actually done anything, which is to say whether the adapted human formulation produces a usable immune response in this species at this dose. That is the whole scientific content of the exercise. Three animals is a sample size suited to answering the question does this work and is it safe, and it is nowhere near a sample size suited to protecting anything. Elenise Sipinski, the technical coordinator of the black-faced lion tamarin conservation programme, has been straightforward about the stage they are at: one group of three done so far, with other groups to be captured and vaccinated.

Six more animals were scheduled to receive doses at the Lagamar de Cananeia State Park in Sao Paulo. Nine out of four hundred is still a pilot. It is simply a larger one.

What makes this more than a veterinary procedure report is that none of it happens in an empty landscape. The restinga forests of the Superagui and Ariri region are inhabited. Caicara communities, the coastal fishing and smallholding populations the tamarin shares its common name with, live in and around the tamarin’s entire range, and have for generations. The island’s geography has itself been rearranged by human decision: the Varadouro canal, cut through at the start of the 1950s, runs some six kilometres long and fifty metres wide and is the reason Superagui is an island rather than part of the mainland at all. Adrianne Pires, a biologist with SPVS, runs the organisation’s community relations work, which is not a soft add-on to the conservation programme. Monitoring in this region depends substantially on residents reporting sightings, and in a range this small, whether local people regard the animal as theirs to protect is a variable of the same order as disease.

There is something clarifying about a conservation story told at this scale. It is small enough to see all the way through. A virus leaves a legible trace by taking out the species most sensitive to it. A set of meetings turns that trace into a written rule. The rule turns into a vet with fifty minutes, a diluted vial with three hours on it and a monkey the size of a kitten. Then a drive back out six weeks later to find out whether any of it took.

The team does not yet know the answer in full, and the honest reporting on it does not pretend otherwise. What is already settled is the commitment. The protocol was adopted before anyone knew whether the vaccine would work in this species, and it does not have a stopping condition written into it. So long as the research continues and animals keep being caught, the needle goes in, over and over, on the only four hundred of these animals that exist, with nothing in a zoo anywhere to fall back on if a capture goes wrong. The three on Superagui were not a victory lap. They were the first three times anyone found out what that promise costs.