Biology

Tiny, long-armed dinosaur leads to rethink of dinosaur miniaturization

Biology, dinosaurs, evolution, paleontology, Science / Tim Belzer / March 8, 2026

Small size seems to have come before a change in diet for a tiny dinosaur lineage.

Alvarezsaurids were mostly small-bodied theropods that paleontologists originally misinterpreted as early flightless birds, only to later recognize them as an ant-eating lineage of non-avian dinosaurs. For years, we suspected that Alvarezsaurids underwent a rare process of evolutionary miniaturization directly coupled to a diet of social insects like ants and termites. It was a tidy hypothesis: They got smaller to become more efficient at catching ants.

Now, a recently discovered fossil of one of the smallest alvarezsaurids ever found suggests that the evolution of miniature dinosaurs likely wasn’t as neat and linear as we thought. This new species, called Alnashetri cerropoliciensis, probably did not feed on ants at all. “It was a pursuit predator actively hunting insects and small mammals,” said Peter Makovicky, a paleontologist at the University of Minnesota.

The oddball

Alverezsaurids, found mostly in the Late Cretaceous rocks of Asia and South America, had short forelimbs tipped with a single oversized thumb claw built for digging. They also had minute teeth and sensory adaptations akin to those in modern nocturnal birds—everything necessary to work on termite mounds. “The explanation of their small body size has been tied to this specialization,” Makovicky explained.

The dinosaur he and his colleagues found, however, did not look like a specialized ant-eater.

The fossil of Alnashetri cerropoliciensis was unearthed from the Candeleros Formation at the Cerro Policía locality in Argentina’s Río Negro Province and is estimated to have lived roughly 90 million years ago. It currently stands as the most complete and smallest Alvarezsaurid skeleton found in South America.

While missing its skull roof, parts of its right arm, its lower right leg, and much of its tail, the skeleton preserves plenty of its crucial anatomy. Its bone tissue reveals that the alvarezsaurid was a subadult, likely approaching sexual maturity, as indicated by the presence of what appears to be medullary bone, a temporary tissue associated with egg-laying in modern birds. Despite being nearly fully grown, this dinosaur is estimated to have weighed a mere 700 grams.

The real surprise, though, came when researchers realized that Alnashetri wasn’t a highly specialized, late-stage Alvarezsauroid. Instead, despite living in the Late Cretaceous, it occupied an early-branching position among earlier, basal members of the clade.

This combination of tiny size and early-branching status fundamentally breaks our previous model of how these animals evolved. If the miniaturization of Alvarezsauroids was strictly tied to their lifestyle as stubby-armed insect-eaters, an early-diverging species like Alnashetri should have some transitional features on a steady, clade-wide march toward that extreme endpoint. But it didn’t look that way.

“It’s a very long-limbed animal, so it was probably fairly fast. My best analogy would be something like a roadrunner from the American West,” Makovicky said.

Arms and teeth

Late Alvarezsaurids had tiny, robust forelimbs that were less than half the length of their femurs. Alnashetri, though, sported comparatively long forelimbs that were 61 percent of the length of its entire hindlimb. While it had three-fingered hands with a robust first digit, a hallmark of its group, it still retained slender second and third digits, unlike its later cousins.

Other features that challenge the established evolutionary model of miniature dinosaurs are Alnashetri’s jaws and teeth. Its dentition features non-serrated teeth set into sockets, but importantly, these teeth are not extremely small, as they were in the late Alvarezsaurids like Shuvuuia or Jaculinykus. “This decoupled the evolution of small body size from anatomical specializations,” Makovicky explained.

The team concluded that extreme miniaturization in Alvarezsaurids did not necessarily co-evolve with either the evolution of smaller arms more suitable for digging or small teeth built for crushing ants and/or termites. Instead of a clade-wide trend where the entire lineage steadily shrank over time, a new evolutionary model that includes Alnashetri suggests that Alvarezsaurid body mass fluctuated repeatedly. Alnashetri, it turns out, achieved its 700-gram frame independently from the other, highly specialized alvarezsaurid species.

But Alnashetri didn’t just upend the understanding of how Alvarezsaurids evolved their tiny bodies. It also redrew the map of where they lived.

Museum tour

Before Makovicky’s study, it was a mystery why Alvarezsaurids were found almost exclusively in the late Cretaceous rocks of Asia and South America. The previous leading hypothesis suggested that the group must have dispersed back and forth between these two landmasses relatively late in the game. But placing Alnashetri, a remarkably basal member, into their evolutionary tree created a massive ghost lineage. The phylogenetic analysis linked geographically close South American species to much older, geologically distant Asian taxa like Bannykus and Xiyunykus, implying that the group must have diverged way back in the Jurassic period.

To explain this chronological and geographic gap, Makovicky and his colleagues started digging through historical museum collections to see if early Alvarezsaurids had been hiding there under different names. It turned out they had.

The team successfully reidentified a small, fragmentary theropod from the Upper Jurassic Morrison Formation in North America, as well as a Lower Cretaceous taxon from the Isle of Wight in Europe. These were early, diverging Alvarezsaurids, and they possessed distinct features such as specialized ball-and-socket joints in the neck vertebrae that are unique in the Alvarezsaurid clade. These museum reidentifications entirely changed the biogeographical story.

If Alvarezsaurids were roaming North America and Europe in the Jurassic and Early Cretaceous, they weren’t just performing a late-stage migration between Asia and South America. Instead, the new model proposed by Makovicky and his team reconstructs a widespread Pangaean distribution. Early Alvarezsaurids were likely present across the globe before the supercontinent Pangaea fully fractured.

The Late Cretaceous distributions we see in the fossil record today would therefore be the result of populations slowly becoming isolated as the continents drifted apart, combined with regional extinctions that wiped them out in places like North America and Europe. The populations in Asia and South America represent surviving pockets.

Still, Makovicky’s work produced far more questions than answers. If at least some Alvarezsaurids did not evolve their miniature bodies as an adaptation to eating ants, what made them so small?

Messy evolution

“We sort of falsified this nice narrative where Alvarezsaurid body size change was driven by ecology, but unfortunately, we don’t have anything good to replace it,” Makovicky acknowledged.

The classic story of Alvarezsaurids—a lineage steadily shrinking in lockstep as it committed to a life of termite-hunting, finally migrating across the Late Cretaceous globe—was neat and logical, but it’s apparently gone now. “That’s science. Sometimes you can falsify a hypothesis without necessarily finding a better one to support,” Makovicky added. But his team is already busy looking for evidence documenting the new, more complex and messier version of Alvarezsaurid evolutionary history. “We have a couple of angles we’re pursuing,” he said.

The first involves taking a closer look at Alnashetri’s anatomy using CT scans. The goal here is to treat Alnashetri as a starting point to understand the stepwise evolution of its ant-eating, specialized cousins. Most of this meticulous scanning is currently happening in Argentina. The second angle, though, seems way more thrilling. “By pure luck, we found another Alvarezsaur in the same general area,” Makovicky said.

The other Alvarezsaur is bigger than Alnashetri and has slightly shorter forelimbs. “It’s still being prepared, but I think it will sort of give us the next chapter in the story of how Alvarezsaurids evolved,” Makovicky explained. “It’s probably a few years out in the making.”

Makovicky’s work on Alnashetri is published in Nature: https://doi.org/10.1038/s41586-026-10194-3

Jacek Krywko is a freelance science and technology writer who covers space exploration, artificial intelligence research, computer science, and all sorts of engineering wizardry.

Tiny, long-armed dinosaur leads to rethink of dinosaur miniaturization Read More »

A unicorn-like Spinosaurus found in the Sahara

Biology, dinosaurs, evolution, paleontology, Science, spinosaurus / Tim Belzer / March 7, 2026

A unique head spike and fish-eating jaws help make sense of these dinosaurs.

The Spinosaurus is a sail-backed, crocodile-snouted dinosaur that Hollywood depicted as a giant terrestrial predator capable of taking down a T. rex in Jurassic Park 3. Then they changed their mind and made it a fully aquatic diver in Jurassic World Rebirth—a rendering that was more in line with the latest paleontological knowledge.

But now, deep in the Sahara Desert, a team of researchers led by Paul C. Sereno, a paleontologist at the University of Chicago, discovered new Spinosaurus fossils suggesting both scientists and filmmakers might have got it all wrong again. The Spinosaurus most likely wasn’t an aquatic diver because, apparently, it couldn’t dive.

Bones in the sand

While the T. rex-beating version of the Spinosaurus was considered unlikely due to its relatively fragile skull, the newer depiction as an aquatic diver made more sense in light of paleontological evidence. Until now, all remains of these predators were pulled from coastal deposits near ancient seas and oceans. That geographic distribution was consistent with the aquatic lifestyle interpretation. If a creature lived on the coast, maybe it swam out to sea like a prehistoric seal, only crawling out to the beaches to rest just as it was depicted in Jurassic World Rebirth.

But the Spinosaurus found by Sereno and his colleagues lived in a completely different neighborhood. The fossils were discovered in the central Sahara of Niger, in what was a terrestrial area called Jenguebi. “When you want to find something really, truly new, you have to go where few have been or maybe nobody has been,” Sereno says. “In the case of Jenguebi, I don’t think it’s seen a paleontologist before.” His team managed to find the site, led by local Tuareg guides after driving for over a day and half through the desert. “We had a team of nearly 100, including paleontologists, filmmakers, guides, and 64 armed guards. You feel like you’re in an Indiana Jones movie,” Sereno recalls. But the effort paid off.

Back in the Cenomanian stage of the Late Cretaceous, the Jenguebi was an inland basin laced with rivers—a riparian habitat situated between 500 and 1,000 kilometers away from the nearest marine shoreline. In these riverbank sediments, Sereno and his team unearthed multiple specimens of the new Spinosaurus species they called S. mirabilis. The skeletons were buried right alongside massive, long-necked dinosaurs, including various species of titanosaurian and rebbachisaurid sauropods. To Sereno, the proximity of these bones left no doubt that the animals they belonged to lived and died together in the same inland freshwater environment. And this inland existence drives a pretty big nail in the coffin of the aquatic diver idea.

Prehistoric heron

The researchers point out that all large-bodied secondarily aquatic tetrapods like whales, mosasaurs, or plesiosaurs, are marine. Finding a giant Spinosaurus thriving in an inland river system strongly supports the idea that it was a semiaquatic, shoreline ambush predator that would wade into shallow waters like a giant crane or heron. But there were other hints that the Spinosaurus was not a diver.

“When you calculate this animal’s lung volume and the air that was permanently in its bones, you’ll find out it was buoyant,” Sereno explains. The permanent air sacks in the bones, an anatomical feature shared by many modern birds, most likely kept the Spinosaurus afloat even when it exhaled all the air out of its lungs. “Birds that dive get rid of those air sacks—penguins got rid of them,” Sereno says. “It’s a balloon you can’t fight against.” He added that even its limbs were far too long to be effectively used as paddles.

This wading lifestyle, the team argues in the paper, was not something unique to the S. mirabilis but extended to other Spinosaurus species as well—the skeletal features of the newly discovered S. mirabilis were found fundamentally similar to its shoreline cousins like S. aegyptiacus on which the Jurassic World Rebirth vision was largely based. Sereno argues it’s highly unlikely that one was a wading river monster while the other was a deep-diving pursuit predator with limited land mobility.

But there was one thing that made S. mirabilis different from S. aegyptiacus. The word “mirabilis” in the newly discovered Spinosaurus’ name translates to “astonishing” in Latin. What Sereno’s team found so astonishing was the prominent crest atop the animal’s head, one of the largest we’ve ever discovered.

The scimitar crown

Instead of the bumpy, fluted ridge seen on S. aegyptiacus, S. mirabilis sported a blade-shaped, scimitar-like bony crest that arched upward and backward from its snout, reaching an apex high over its eyes. This structure was composed of solid bone, unlike the highly porous, pneumatic casques found on some modern birds. However, the bone itself was etched with fine longitudinal striations and deep grooves, indicating that the bony core was just the foundation.

The newly discovered skull, along with a model of what its spike might have looked like on a living animal. Credit: UChicago Fossil Lab

In a living S. mirabilis, this crest would have been enveloped and substantially extended by a keratinous sheath, much like the vibrant growth developed by modern helmeted guinea fowls. If scaled up to a fully mature adult, the bony core alone would measure around 40 centimeters in length; with its keratinous sheath, it could have easily exceeded half a meter. For Sereno, the purpose of this “astonishing” scimitar crown was similar to crests worn today by cranes and herons. “It was asymmetrical. It varied between individuals. So, I think it was solely for display,” Sereno explains.

His team hypothesizes that visual signaling was the primary function of both the cranial crests and the massive trunk and tail sails that define spinosaurids. In the crowded shoreline and riverbank habitats, a towering, brightly colored crest or sail would be an excellent way to broadcast your size, maturity, and genetic fitness to rivals and potential mates without having to engage in a costly physical brawl.

Still, when it came down to it, S. mirabilis, weighing in at well over 7 tons, totally could brawl. “The Spinosaurus was enormous. I think it could have eaten anything it wanted even though its mainstay was fish,” Sereno says.

Crocodile jaw

The showpiece on its forehead aside, the S. mirabilis was a highly specialized killing machine. Its snout featured a low profile with parallel dorsal and ventral margins, terminating in a mushroom-shaped expansion at the tip. The upper and lower jaws allowed the teeth to interdigitate perfectly—there was a notable diastema, a gap in the upper row of teeth, that neatly accommodated the large teeth of the lower jaw. The S. mirabilis jaw structure appears similar to that of modern long-snouted crocodiles, optimized for snatching and snaring aquatic prey with a rapid, trap-like closure. Surprisingly, S. mirabilis showed greater spacing between the teeth in the posterior half of its snout compared to S. aegyptiacus despite being otherwise nearly identical.

Analysis of the animals’ overall body proportions led Sereno and his colleagues to suspect these dinosaurs resided in the functional middle ground between semiaquatic waders like herons and aquatic divers like darters, placing them in an ecological niche entirely separate from all other predatory theropods. Based on Sereno’s paper, the evolutionary history of the spinosaurids started in the Jurassic, when their ancestors first evolved that distinctive, elongate, fish-snaring skull before splitting into two main lineages: baryonychines and spinosaurines.

Then, during the Early Cretaceous, the spinosaurines enjoyed a golden age, diversifying across the margins of the Tethys Sea, a late Paleozoic ocean situated between the continents of Gondwana and Laurasia, to become the dominant predators in their respective ecosystems. What most likely brought an end to their reign was climate change.

The end of the line

The final chapter in the Spinosaurus history played out just before the Late Cretaceous, as the Atlantic Ocean was opening up. This is when spinosaurines, limited geographically to what today is Northern Africa and South America, pushed their biological limits, attaining their maximum body sizes as highly specialized shallow-water ambush hunters. This specialization, though, probably led to their extinction.

Around 95 million years ago, at the end of the Cenomanian stage, the world started to shift. An abrupt rise in global sea levels driven by climate changes drowned the low-lying continental basins and created the Trans-Saharan seaway. The complex, shallow river systems and coastal swamps that supported giant wading spinosaurines vanished beneath the waves. “We don’t see spinosaurid fossil records beyond this period,” Sereno explains. The spinosaurid lineage, unable to dive and adapt to more aquatic lifestyles, was brought to an end.

But we still don’t know much about its beginning. “This is going to be the subject of our next paper—where did the Spinosaurus come from?” Sereno says.

Sereno’s paper on the S. mirabilis is published in Science: https://doi.org/10.1126/science.adx5486

Jacek Krywko is a freelance science and technology writer who covers space exploration, artificial intelligence research, computer science, and all sorts of engineering wizardry.

A unicorn-like Spinosaurus found in the Sahara Read More »

Why are vertebrate eyes so different from those of other animals?

Biology, evolution, eyes, Science / Kris Guyer / March 7, 2026

“We think that in this early deuterostome, the median eye contained both ciliary and rhabdomeric cells,” Kafetzis explains. As a result, both cellular lineages were incorporated into a single, ancient, cyclopean eye, which later evolved into the vertebrate eyes.

The vertebrate third eye

A trace of this transformation may still survive in the pineal complex at the base of the brain—often referred to as a vertebrate “third eye.” Scientists have long recognized striking similarities between the retina and the pineal organ, leading many to suspect that the two evolved from a single ancestral structure, with the pineal representing a more rudimentary version.

Kafetzis and his colleagues see it differently.

Many researchers suspect that one class of neurons—the bipolar cells—is unique to the retina and represents a key evolutionary innovation of the vertebrate eye. Bipolar cells connect rods and cones to ganglion cells (hence the name “bipolar”). “We think that these bipolar-like cells already exist in the pineal,” says Kafetzis. “It’s just that they don’t look like the typical bipolar—they don’t have a cell before and a cell after.”

For this reason, Kafetzis and his colleagues argue that bipolar neurons are not a de novo evolutionary invention but instead have a chimeric origin, blending features of both rhabdomeric and ciliary cells and bridging the two photoreceptor lineages.

Though grounded in existing ideas and data, the new proposal offers a potentially far-reaching synthesis. Several aspects still require firmer evidence. The idea that the ancestral chordate adopted a burrowing lifestyle remains debated, and the claim that early bilaterians already possessed paired lateral eyes is still speculative.

The authors acknowledge that their model now needs testing. In the paper, they lay out several ways to do so—from molecular comparisons of pineal and retinal cells to developmental studies and broader sampling of eye development across other deuterostome species.

“We want to put forward some literature-based and inspired hypotheses that are testable, and now we can go out and test them,” concludes Kafetzis.

Cell, 2026. DOI: 10.1016/j.cell.2025.12.056

Federica Sgorbissa is a science journalist; she writes about neuroscience and cognitive science for Italian and international outlets.

Why are vertebrate eyes so different from those of other animals? Read More »

Large genome model: Open source AI trained on trillions of bases

AI, Biology, DNA, genes, Genetics, Genomics, Science, training / Kris Guyer / March 4, 2026

System can identify genes, regulatory sequences, splice sites, and more.

Late in 2025, we covered the development of an AI system called Evo that was trained on massive numbers of bacterial genomes. So many that, when prompted with sequences from a cluster of related genes, it could correctly identify the next one or suggest a completely novel protein.

That system worked because bacteria tend to cluster related genes together—something that’s not true in organisms with complex cells, which tend to have equally complex genome structures. Given that, our coverage noted, “It’s not clear that this approach will work with more complex genomes.”

Apparently, the team behind Evo viewed that as a challenge, because today it is describing Evo 2, an open source AI that has been trained on genomes from all three domains of life (bacteria, archaea, and eukaryotes). After training on trillions of base pairs of DNA, Evo 2 developed internal representations of key features in even complex genomes like ours, including things like regulatory DNA and splice sites, which can be challenging for humans to spot.

Genome features

Bacterial genomes are organized along relatively straightforward principles. Any genes that encode proteins or RNAs are contiguous, with no interruptions in the coding sequence. Genes that perform related functions, like metabolizing a sugar or producing an amino acid, tend to be clustered together, allowing them to be controlled by a single, compact regulatory system. It’s all straightforward and efficient.

Eukaryotes are not like that. The coding sections of genes are interrupted by introns, which don’t encode for anything. They’re regulated by a sequence that can be scattered across hundreds of thousands of base pairs. The sequences that define the edges of introns or the binding sites of regulatory proteins are all weakly defined—while they have a few bases that are absolutely required, there are a lot of bases that just have an above-average tendency to have a specific base (something like “45 percent of the time it’s a T”). Surrounding all of this in most eukaryotic genomes is a huge amount of DNA that has been termed junk: inactive viruses, terminally damaged genes, and so on.

That complexity has made eukaryotic genomes more difficult to interpret. And, while a lot of specialized tools have been developed to identify things like splice sites, they’re all sufficiently error-prone that it becomes a problem when you’re analyzing something as large as a 3 billion-base-long genome. We can learn a lot more by making evolutionary comparisons and looking for sequences that have been conserved, but there are limits to that, and we’re often as interested in the differences between species.

These sorts of statistical probabilities, however, are well-suited to neural networks, which are great at recognizing subtle patterns that can be impossible to pick out by eye. But you’d need absolutely massive amounts of data and computing time to process it and pick out some of these subtle features.

We now have the raw genome data that the process needs. Putting together a system to feed it into an effective AI training program, however, remained a challenge. That’s the challenge the team behind Evo took on.

Training a large genome model

The foundation of the Evo 2 system is a convolutional neural network called StripedHyena 2. The training took place in two stages. The initial stage focused on teaching the system to identify important genome features by feeding it sequences rich in them in chunks about 8,000 bases long. After that, there was a second stage in which sequences were fed a million bases at a time to provide the system the opportunity to identify large-scale genome features.

The researchers trained two versions of their system using a dataset called OpenGenome2, which contains 8.8 trillion bases from all three domains of life, as well as viruses that infect bacteria. They did not include viruses that attack eukaryotes, given that they were concerned that the system could be misused to create threats to humans. Two versions were trained: one that had 7 billion parameters tuned using 2.4 trillion bases, and the full version with 40 billion parameters trained on the full open genome dataset.

The logic behind the training is pretty simple: if something’s important enough to have been evolutionarily conserved across a lot of species, it will show up in multiple contexts, and the system should see it repeatedly during training. “By learning the likelihood of sequences across vast evolutionary datasets, biological sequence models capture conserved sequence patterns that often reflect functional importance,” the researchers behind the work write. “These constraints allow the models to perform zero-shot prediction without any task-specific fine-tuning or supervision.”

That last aspect is important. We could, for example, tell it about what known splice sites look like, which might help it pick out additional ones. But that might make it harder for it to recognize any unusual splice sites that we haven’t recognized yet. Skipping the fine-tuning might also help it identify genome features that we’re not aware of at all at the moment, but which could become apparent through future research.

All of this has now been made available to the public. “We have made Evo 2 fully open, including model parameters, training code, inference code, and the OpenGenome2 dataset,” the paper announces.

The researchers also used a system that can identify internal features in neural networks to poke around inside of Evo 2 and figure out what things it had learned to recognize. They trained a separate neural network to recognize the firing patterns in Evo 2 and identify high-level features in it. It clearly recognized protein-coding regions and the boundaries of the introns that flanked them. It was also able to recognize some structural features of proteins within the coding regions (alpha helices and beta sheets), as well as mutations that disrupt their coding sequence. Even something like mobile genetic elements (which you can think of as DNA-level parasites) ended up with a feature within Evo 2.

What is this good for?

To test the system, the researchers started making single-base mutations and fed them into Evo 2 to see how it responded. Evo 2 could detect problems when the mutations affected the sites in DNA where transcription into RNA started, or the sites where translation of that RNA into protein started. It also recognized the severity of mutations. Those that would interrupt protein translation, such as the introduction of stop signals, were identified as more significant changes than those that left the translation intact.

It also recognized when sequences weren’t translated at all. Many key cellular functions are carried out directly by RNAs, and Evo 2 was able to recognize when mutations disrupted those, as well.

Impressively, the ability to recognize features in eukaryotic genomes occurred without the loss of its ability to recognize them in bacteria and archaea. In fact, the system seemed to be able to work out what species it was working in. A number of evolutionary groups use genetic codes with a different set of signals to stop the translation of proteins. Evo 2 was able to recognize when it was looking at a sequence from one of those species, and used the correct genetic code for them.

It was also good at recognizing features that tolerate a lot of variability, such as sites that signal where to splice RNAs to remove introns from the coding sequence of proteins. By some measures, it was better than software specialized for that task. The same was true when evaluating mutations in the BRCA2 gene, where many of the mutations are associated with cancer. Given additional training on known BRCA2 mutations, its performance improved further.

Overall, Evo 2 seems great for evaluating genomes and identifying key features. The researchers who built it suggest it could serve as a good automated tool for preliminary genome annotation.

But the striking thing about the early version of Evo was that, when prompted with a chunk of sequence that includes known bacterial genes, some of its responses included entirely new proteins with related functions. Now that it was trained on more complex eukaryotic genes, could it do the same?

We don’t entirely know. If given a bunch of DNA from yeast (a eukaryote), it would respond with a sequence that included functional RNAs, and gene-like sequences with regulatory information and splice sites. But the researchers didn’t test whether any of the proteins did anything in particular. And it’s difficult to see how they could even do that test. With bacterial genes, they could safely assume that the AI-generated gene should be doing something related to the nearby genes. But that’s generally not the case in eukaryotes, so it’s difficult to guess what functions they should even test for.

In a somewhat more informative test, the researchers asked Evo 2 to make some regulatory DNA that was active in one cell type and not another after giving it information about what sequences were active in both those cell types. The sequences that came out were then inserted into these cells and tested, but the results were pretty weak, with only 17 percent having activity that differed by a factor of two or more between the two cell types. That’s a major achievement, but it isn’t in the same realm as designing brand new proteins.

What’s next?

Overall, given that this has come out less than four months after the paper describing the original Evo, it’s not at all surprising that there wasn’t more work done to test what Evo 2 can do for designing biologically relevant DNA sequences. Biology experiments are hard and time-consuming, and it’s not always easy to judge in advance which ones will provide the most compelling information. So we’ll probably have to wait months to years to find out whether the community finds interesting things to do with Evo 2, and whether it’s good at solving any useful protein design problems.

There’s also the question of whether further training and specialization can create Evo 2 relatives that are especially good at specific tasks, such as evaluating genomes from cancer cells or annotating newly sequenced genomes. To an extent, it appears the research team wanted to get this out so that others could start exploring how it might be put to use; that’s consistent with the fact that all of the software was made available.

The big open question is whether this system has identified anything that we don’t know how to test for. Things like intron/exon boundaries and regulatory DNA have been subjected to decades of study so that we already knew how to look for them and can recognize when Evo 2 spots them. But we’ve discovered a steady stream of new features in the genome—CRISPR repeats, microRNAs, and more—over the past decades. It remains technically possible that there are features in the genome we’re not aware of yet, and Evo 2 has picked them out.

It’s possible to imagine ways to use the tools described here to query Evo 2 and pick out new genome features. So I’m looking forward to seeing what might ultimately come out of that sort of work.

Nature, 2026. DOI: 10.1038/s41586-026-10176-5 (About DOIs).

John is Ars Technica’s science editor. He has a Bachelor of Arts in Biochemistry from Columbia University, and a Ph.D. in Molecular and Cell Biology from the University of California, Berkeley. When physically separated from his keyboard, he tends to seek out a bicycle, or a scenic location for communing with his hiking boots.

Large genome model: Open source AI trained on trillions of bases Read More »

Neanderthals seemed to have a thing for modern human women

Biology, evolution, Genomics, human evolution, Neanderthals, Science / Rejus Almole / February 26, 2026

By now, it’s firmly established that modern humans and their Neanderthal relatives met and mated as our ancestors expanded out of Africa, resulting in a substantial amount of Neanderthal DNA scattered throughout our genome. Less widely recognized is that some of the Neanderthal genomes we’ve seen have pieces of modern human DNA as well.

Not every modern human has the same set of Neanderthal DNA, however; different people will, by chance, have inherited different fragments. But there are also some areas, termed “Neanderthal deserts,” where none of the Neanderthal DNA seems to have persisted. Notably, the largest Neanderthal desert is the entire X chromosome, raising questions about whether this reflects the evolutionary fitness of genes there or mating preferences.

Now, three researchers at the University of Pennsylvania, Alexander Platt, Daniel N. Harris, and Sarah Tishkoff, have done the converse analysis: examining the X chromosomes of the handful of completed genomes we have. It turns out there’s also a strong bias toward modern human sequences there, as well, and the authors interpret that as selective mating, with Neanderthal males showing a strong preference for modern human females and their descendants.

What type of selection are we looking at?

Given how long modern humans and Neanderthals had been evolving as separate populations, some degree of genetic incompatibility is definitely possible. Lots of proteins interact in various ways, and the genes behind these interaction networks will evolve together—a change in one gene will often lead to compensatory changes in other genes in the network. Over time, those changes may mean re-introducing the original gene will actually disrupt the network, with a negative impact on fitness.

That means the introduction of some Neanderthal genes into the modern human genome (or vice versa) would be disruptive and make carriers of them less fit. So they’d be selected against and lost over the ensuing generations. Of course, some segments would likely be lost at random—the genome’s pretty big, and the modern human population was likely large and growing, allowing its DNA to dilute out the influence of other human populations. Figuring out which influence is dominant can be challenging.

Neanderthals seemed to have a thing for modern human women Read More »

Boozy chimps fail urine test, confirm hotly debated theory

animal behavior, animals, Biology, primates, Science / Paul Patrick / February 25, 2026

The urine of chimpanzees contains high levels of alcohol byproduct, most likely because the chimps regularly gorge themselves on fermented fruit, according to a new paper published in the journal Biology Letters. It’s the latest evidence in support of a hotly debated theory regarding the evolutionary origins of human fondness for alcohol.

As previously reported, in 2014, University of California, Berkeley (UCB) biologist Robert Dudley wrote a book called The Drunken Monkey: Why We Drink and Abuse Alcohol. His controversial “drunken monkey hypothesis” proposed that the human attraction to alcohol goes back about 18 million years, to the origin of the great apes, and that social communication and sharing food evolved to better identify the presence of fruit from a distance. At the time, skeptical scientists insisted that this was unlikely because chimpanzees and other primates just don’t eat fermented fruit or nectar.

But reports of primates doing just that have grown over the ensuing two decades. Earlier this year, we reported that researchers had caught wild chimpanzees on camera engaging in what appears to be sharing fermented African breadfruit with measurable alcoholic content. That observational data was the first evidence of the sharing of alcoholic foods among nonhuman great apes in the wild. The authors measured the alcohol content of the fruit with a handy portable breathalyzer and found almost all of the fallen fruit (90 percent) contained some ethanol, with the ripest containing the highest levels—the equivalent of 0.61 percent ABV (alcohol by volume).

And last September, Dudley co-authored a paper reporting the first measurements of the ethanol content of fruits favored by chimps in the Ivory Coast and Uganda, finding that chimps consume 14 grams of alcohol per day, the equivalent of a standard alcoholic drink in the US. After adjusting for the chimps’ lower body mass, the authors concluded the chimps are consuming nearly two drinks per day.

A thankless task

The next step was to sample the chimps’ urine to see if it contains any alcohol metabolites, as was found in a 2022 study on spider monkeys. This would further refine estimates for how much ethanol-laden fruit the chimps eat every day. That thankless task fell to Aleksey Maro, a UCB graduate student who spent last summer in Ngogo, sleeping in trees—protected from the constant streams by an umbrella—to collect urine samples. Sharifah Namaganda, a Ugandan graduate student at the University of Michigan, showed him how to make shallow bowls out of plastic bags hung on forked twigs for more efficient collection. He also collected samples from puddles of urine on the forest floor.

Boozy chimps fail urine test, confirm hotly debated theory Read More »

Dinosaur eggshells can reveal the age of other fossils

Biology, dinosaurs, evolution, paleontology, radioactive dating, Science / Kris Guyer / February 21, 2026

When dinosaur fossils surface at a site, it is often not possible to tell how many millions of years ago their bones were buried. While the different strata of sedimentary rock represent periods of geologic history frozen in time, accurately dating them or the fossils trapped within them has frequently proven to be frustrating.

Fossilized bones and teeth have been dated with some success before, but that success is inconsistent and depends on the specimens. Both fossilization and the process of sediment turning to rock can alter the bone in ways that interfere with accuracy. While uranium-lead dating is among the most widely used methods for dating materials, it is just an emerging technology when applied to directly dating fossils.

Dinosaur eggshells might have finally cracked a way to date surrounding rocks and fossils. Led by paleontologist Ryan Tucker of Stellenbosch University, a team of researchers has devised a method of dating eggshells that reveals how long ago they were covered in what was once sand, mud, or other sediments. That information will give the burial time of any other fossils embedded in the same layer of rock.

“If validated, this approach could greatly expand the range of continental sedimentary successions amenable to radioisotopic dating,” Tucker said in a study recently published in Nature Communications Earth & Environment.

This goes way back

Vertebrates have been laying calcified eggs for hundreds of millions of years (although the first dinosaur eggs had soft shells). What makes fossil eggshells so useful for figuring out the age of other fossils is the unique microstructure of calcium carbonate found in them. The way its crystals are arranged capture a record of diagenetic changes, or physical and chemical changes that occurred during fossilization. These can include water damage, along with breaks and fissures caused by being compacted between layers of sediment. This makes it easier to screen for these signs when trying to determine how old they are.

Dinosaur eggshells can reveal the age of other fossils Read More »

From chickens to humans, animals think “bouba” sounds round

Behavioral science, Biology, evolution, language, neurobiology, perception, Science / Rejus Almole / February 19, 2026

Does “bouba” sound round to you? How about “maluma”? Neither are real words, but we’ve known for decades that people who hear them tend to associate them with round objects. There have been plenty of ideas put forward about why that would be the case, and most of them have turned out to be wrong. Now, in perhaps the weirdest bit of evidence to date, researchers have found that even newly hatched chickens seem to associate “bouba” with round shapes.

The initial finding dates all the way back to 1947, when someone discovered that people associated some word-like sounds with rounded shapes, and others with spiky ones. In the years since, that association got formalized as the bouba/kiki effect, received a fair bit of experimental attention, and ended up with an extensive Wikipedia entry.

One of the initial ideas to explain it was similarity to actual words (either phonetically or via the characters used to spell them), but then studies with speakers of different languages and alphabets showed that it is likely a general human tendency. The association also showed up in infants as young as 4 months old, well before they master speaking or spelling. Attempts to find the bouba/kiki effects in other primates, however, came up empty. That led to some speculation that it might be evidence of a strictly human processing ability that underlies our capacity to learn sophisticated languages.

A team of Italian researchers—Maria Loconsole, Silvia Benavides-Varela, and Lucia Regolin—now have evidence that that isn’t true either. They decided to look for the bouba/kiki effect well beyond primates, instead turning to newly hatched chickens, only one or three days old. That may sound a bit odd, but chickens have a key advantage beyond ready availability: unlike a 4-month-old human, newly hatched chicks are fully mobile and able to interact with the world.

From chickens to humans, animals think “bouba” sounds round Read More »

Rare gifted word-learner dogs like to share their toys

animal cognition, animals, Biology, gifted word learner dogs, Science / Mike M. / February 19, 2026

This time around, the group recruited 10 GWL dogs and 21 non-GWL dogs, all border collies, since this is the most common breed to fall into the GWL category. They compiled a list of eight toys: two labeled, two unlabeled, and four that were new to each dog.

What’s their motivation?

There was a two-week period during which owners familiarized the dogs with the toys once a day for at least 10 minutes. Each toy was presented separately. For the labeled toys, owners moved the toy while crouched on the floor, repeatedly naming the toy (“Look at the [toy name]! Here is the [toy name]”). They did not name the unlabeled toys. Owners devoted an equal amount of time to all the toys. Novel toys were excluded from the familiarization phase.

After that period, each dog participated in two 90-second trials. The dogs were provided free access to the toys (washed with soap to control for odor cues). In the first trial, owners entered first and placed the labeled and unlabeled toys, plus two of the novel ones, on the floor and stood at a distance, passive and ignoring the dogs as the latter explored the toys. After a five-minute break, the test was repeated with the other two novel toys. All tests were recorded remotely and the footage subsequently analyzed.

Human babies are known to pay more attention to named objects, and the authors thought the GWL dogs would show a similar response, but that’s not what happened. All the dogs, whether they were GWL dogs or not, strongly preferred the new toys, and there were no significant differences between the two groups of dogs in terms of how much time they spent playing with labeled versus unlabeled. So just hearing toy names does not automatically increase a dog’s attention.

Rare gifted word-learner dogs like to share their toys Read More »

Tiny, 45 base long RNA can make copies of itself

biochemistry, Biology, evolution, origin of life, ribozymes, RNA, Science / Tim Belzer / February 13, 2026

Self-copying RNAs may have been a key stop along the pathway to life.

By base pairing with themselves, RNAs can form complex structures with enzymatic activity. Credit: Laguna Design

There are plenty of unanswered questions about the origin of life on Earth. But the research community has largely reached consensus that one of the key steps was the emergence of an RNA molecule that could replicate itself. RNA, like its more famous relative DNA, can carry genetic information. But it can also fold up into three-dimensional structures that act as catalysts. These two features have led to the suggestion that early life was protein-free, with RNA handling both heredity and catalyzing a simple metabolism.

For this to work, one of the reactions that the early RNAs would need to catalyze is the copying of RNA molecules, without which any sort of heritability would be impossible. While we’ve found a number of catalytic RNAs that can copy other molecules, none have been able to perform a key reaction: making a copy of themselves. Now, however, a team has found an incredibly short piece of RNA—just 45 bases long—that can make a copy of itself.

Finding an RNA polymerase

We have identified a large number of catalytic RNAs (generically called ribozymes, for RNA-based enzymes), and some of them can catalyze reactions involving other RNAs. A handful of these are ligases, which link together two RNA molecules. In some cases, they need these molecules to be held together by a third RNA molecule that base pairs with both of them. We’ve only identified a few that can act as polymerases, which add RNA bases to a growing molecule, one at a time, with each new addition base pairing with a template molecule.

Black on white image showing 3 different enzymatic activities. One links any two nucleic acid strands, the other only links base paired strands, and the third links one base at a time. — Some ligases can link two nucleic acid strands (left), while others can link the strands only if they’re held together by base pairing with a template (center). A polymerase can be thought of as a template-dependent ligase that adds one base at a time. The newly discovered ribozyme sits somewhere between a template-directed ligase and a polymerase. Credit: John Timmer

Obviously, there is some functional overlap between them, as you can think of a polymerase as ligating on one base at a time. And in fact, at the ribozyme level, there’s some real-world overlap, as some ribozymes that were first identified as ligases were converted into polymerases by selecting for this new function.

While this is fascinating, there are a few problems with these known examples of polymerase ribozymes. One is that they’re long. So long, in fact, that they’re beyond the length of the sort of molecules that we’ve observed forming spontaneously from a mix of individual RNA bases. This length also means they’re largely incapable of making copies of themselves—the reactions are slow and inefficient enough that they simply stop before copying the entire molecule.

Another factor related to their length is that they tend to form very complex structures, with many different areas of the molecule base-paired to one another. That leaves very little of the molecule in a single-stranded form, which is needed to make a copy.

Based on past successes, a French-UK team decided to start a search for a polymerase by looking for a ligase. And they limited that search in an important way: They only tested short molecules. They started with pools of RNA molecules, each with a different random sequence, ranging from 40 to 80 bases. Overall, they estimated that they made a population of 10¹³ molecules out of the total possible population of 10²⁴ sequences of this type.

These random molecules were fed a collection of three-base-long RNAs, each linked to a chemical tag. The idea was that if a molecule is capable of ligating one of these short RNA fragments to itself, it could be pulled out using the tag. The mixtures were then placed in a salty mixture of water and ice, as this can promote reactions involving RNAs.

After 11 rounds of reactions and tag-based purification, the researchers ended up with three different RNA molecules that could each ligate three-base-long RNAs to existing molecules. Each of these molecules was subjected to mutagenesis and further rounds of selection. This ultimately left the researchers with a single, 51-base-long molecule that could add clusters of three bases to a growing RNA strand, depending on their ability to base-pair with an RNA template. They called this “polymerase QT-51,” with QT standing for “quite tiny.” They later found that they could shorten this to QT-45 without losing significant enzyme activity.

Checking its function

The basic characterization of QT-45 showed that it has some very impressive properties for a molecule that, by nucleic acid standards, is indeed quite tiny. While it was selected for linking collections of molecules that were three bases long, it could also link longer RNAs, work on shorter two-base molecules, or even add a single base at a time, though this was less efficient. While it worked slowly, the molecule’s active half-life was well over 100 days, so it had plenty of time to get things done before it degraded.

It also didn’t need to interact with any specific RNA sequences to work, suggesting it had a general affinity for RNA molecules. As a result, it wasn’t especially picky about the sequences it could copy.

As you might expect from such a small molecule, QT-45 didn’t tolerate changes to its own sequence very well—nearly the entire molecule was important in one way or another. Tests that involved changing every single individual base one at a time showed that almost all the changes reduced the ribozyme’s activity. There were, however, a handful of changes that improved things, suggesting that further selection could potentially yield additional improvements. And the impact of mutations near the center of the sequence was far more severe, suggesting that region is critical for QT-45’s enzymatic activity.

The team then started testing its ability to synthesize copies of other RNA molecules when given a mixture of all possible three-base sequences. One of the tests included a large stretch in which one end of the sequence base-paired with the other. To copy that, those base pairs need to somehow be pried apart. But QT-45 was able to make a copy, meaning it synthesized a strand that was able to base pair with the original.

It was also able to make a copy of a template strand that would base pair with a small ribozyme. That copying produced an active ribozyme.

But the key finding was that it could synthesize a sequence that base-pairs with itself, and then synthesize itself by copying that sequence. This was horribly inefficient and took months, but it happened.

Throughout these experiments, the fidelity averaged about 95 percent, meaning that, in copying itself, it would make an average of two to three errors. While this means a fair number of its copies wouldn’t be functional, it also means the raw materials for an evolutionary selection for improved function—random mutations—would be present.

What this means

It’s worth taking a moment to consider the use of three-base RNA fragments by this enzyme. On the surface, this may seem a bit like cheating, since current RNA polymerases add sequence one base at a time. But in reality, any chemical environment that could spontaneously assemble an RNA molecule 45 bases long will produce many fragments shorter than that. So in many ways, this might be a more realistic model of the conditions in which life emerged.

The authors note that these shorter fragments may be essential for QT-45’s activity. The short ribozyme probably doesn’t have the ability to enzymatically pry base-paired strands of RNA apart to copy them. But in a mixture of lots of small fragments, there’s likely to be an equilibrium, with some base-paired sequences spontaneously popping open and temporarily base pairing with a shorter fragment. Working with these base-paired fragments is probably essential to the ribozyme’s overall activity.

Right now, QT-45 isn’t an impressive enzyme. But the researchers point out that it has only been through 18 rounds of selection, which isn’t much. The most efficient ribozyme polymerases we have at present have been worked on by multiple labs for years. I expect QT-45 to receive similar attention and improve significantly over time.

Also notable is that the team came up with three different ligases in a test of just a small subset of the possible total RNA population of this size. If that frequency holds, there are on the order of 10¹¹ ligating ribozymes among the sequences of this size. Which raises the possibility that we could find far more if we do an exhaustive search. That suggests the first self-copying RNA might not be as improbable as it seems at first.

Science, 2026. DOI: 10.1126/science.adt2760 (About DOIs).

Tiny, 45 base long RNA can make copies of itself Read More »

Bringing the “functionally extinct” American chestnut back from the dead

American chestnut, Biology, extinction, genetic engineering, Genetics, Genomics, Invasive species, pathogens, plants, Science, trees / Tim Belzer / February 12, 2026

Wiped out in its native range by invasive pathogens, the trees may make a comeback.

Very few people alive today have seen the Appalachian forests as they existed a century ago. Even as state and national parks preserved ever more of the ecosystem, fungal pathogens from Asia nearly wiped out one of the dominant species of these forests, the American chestnut, killing an estimated 3 billion trees. While new saplings continue to sprout from the stumps of the former trees, the fungus persists, killing them before they can seed a new generation.

But thanks in part to trees planted in areas where the two fungi don’t grow well, the American chestnut isn’t extinct. And efforts to revive it in its native range have continued, despite the long generation times needed to breed resistant trees. In Thursday’s issue of Science, researchers describe their efforts to apply modern genomic techniques and exhaustive testing to identify the best route to restoring chestnuts to their native range.

Multiple paths to restoration

While the American chestnut is functionally extinct—it’s no longer a participant in the ecosystems it once dominated—it’s most certainly not extinct. Two Asian fungi that have killed it off in its native range; one causes chestnut blight, while a less common pathogen causes a root rot disease. Both prefer warmer, humid environments and persist there because they can grow asymptomatically on distantly related trees, such as oaks. Still, chestnuts planted outside the species’ original range—primarily in drier areas of western North America—have continued to thrive.

There is also a virus that attacks the chestnut blight fungus, allowing a few trees to survive in areas where that virus is common. Finally, a handful of trees have grown to maturity in the American chestnut’s original range. These trees, which the paper refers to as LSACs (large surviving American chestnuts), suggest that there might have been some low level of natural resistance within the now-vanished population.

Those trees are central to one of the efforts to restore the American chestnut. If enough of them have distinct means of resisting the fungi, interbreeding them might produce a strain that not only survives the fungi but can also thrive in the Appalachians.

A related approach took advantage of the fact that the American chestnut can produce fertile hybrids with the Chinese chestnut, which had co-evolved with the introduced fungi and were thus resistant to lethal infections. The hope was that continued back-breeding of these hybrids with American chestnuts would result in trees that were very similar to American chestnuts yet retained the fungal resistance of their Asian cousins.

Both efforts suffered from the same problem that faces any biologist working on trees: They are slow-growing and can take years to reach a size at which they produce seeds. The situation was further complicated by the fact that the American chestnut can’t pollinate itself, so you need at least two trees before any breeding is possible.

Concerned about what this might mean for the potential reintroduction of the chestnut into the Appalachians, a third project turned to biotechnology. Research had identified oxalic acid as a key factor in the blight’s virulence. Wheat naturally produces an enzyme that degrades oxalic acid, and researchers inserted the gene that encodes that enzyme into the American chestnut genome, creating a genetically modified tree that can potentially disarm the fungus’ attack.

Without understanding the nature of resistance or the effectiveness of the transgenic gene, there’s no way to know which method would be most effective. So researchers from the American Chestnut Foundation assembled a massive collaboration to examine all these options and determine what would be needed to reintroduce blight-resistant chestnuts into the wild.

Tracking resistance

The scale of the effort is immense. All told, the team infected over 4,000 individual trees with the blight fungus and tracked their growth in Appalachian nurseries for an average of over 14 years. The trees were scored for resistance on a zero-to-100 scale based on the damage caused by the infection. This data was combined with some serious lab work; the team produced the highest-quality chestnut genomes yet (of both American and Chinese species) and gathered biochemical data on how the trees respond to infection.

It quickly became apparent that there were significant differences in the growth rates of some of the resistant trees. When planted at sites where viruses kept the blight in check, the Chinese chestnuts grew more slowly than native trees, while hybrids grew at an intermediate rate. That could make a big difference, as rapid growth may have enabled the chestnut to reach its former dominance of the canopy.

Somewhat surprisingly, this slow growth turned out to be a problem for the genetically modified American chestnuts as well. By chance, the wheat gene ended up being inserted into a gene known to be important for the growth of other plants. It seems to be important in the chestnut as well; plants with two copies of the inserted genes survived at 16 percent of their expected rate, and those with a single copy grew 22 percent slower than unmodified trees.

That said, there was a lot of variability among the genetically modified trees, with 4 percent of the tested trees showing both high blight resistance and growth comparable to that of unmodified American chestnuts. It will be important to determine whether this collection of traits remains consistent in ensuing generations.

In a bit of good news, the progeny from surviving American chestnuts grew like American chestnuts. In less good news, among 143 of these trees, only seven had resistance levels of above 50 on the team’s 100-point scale. It’s possible that interbreeding these trees could further boost resistance, but it also poses the risk of creating a population that’s too inbred to thrive after reintroduction.

Root causes

The research team decided to use their testing to investigate the genetic basis of resistance. There’s a very practical reason for this: If resistance is mediated by just a handful of genes that each have large impacts, it should be possible to continue breeding resistant strains back to regular American chestnuts and selecting for resistance. But if there are many factors with relatively small impacts, it will require directed interbreeding of hybrids to maximize both resistance and DNA originating from the American chestnut.

The team completed the highest-quality chestnut genomes for both the American and Chinese species, identifying about 25,000 to 30,000 genes in the different assemblies. They then used this information for two types of genetic analysis: quantitative trait locus identification and genome-wide association. Both approaches aim to identify regions of the genome associated with specific properties and estimate their impact.

The work suggested that resistance arises from a relatively large number of sites, each with relatively minor effects. For example, the sites in the genome identified by quantitative trait analysis typically boosted resistance by about 10 points on the researchers’ 100-point scale. In the genome-wide analysis, 17 individual genetic differences were associated with about a quarter of the heritable resistance traits. All of this suggests that, for the hybrids (and likely for the weaker blight resistance found in surviving American chestnuts), directed breeding among surviving trees will be needed.

For the root rot fungus, in contrast, it looks like there are a limited number of important alleles with a large impact.

The researchers also took an alternative approach to identify resistance factors, comparing 100 chemicals produced by resistant and susceptible strains. Among the 41 chemicals detected at higher levels in the Chinese chestnut, the researchers found a metabolite, lupeol, that completely suppressed the growth of the fungal pathogen. Another, erythrodiol, limited its growth. If we can identify the genes involved in producing those chemicals, we could use that knowledge to guide directed breeding programs—or even engage in gene editing to increase their production.

The team’s current plan is to use genomic predictions to select hybrid seedlings for planting in test orchards, aiming to identify plants with high growth and resistance. From there, the process can be repeated. But even after the exhaustive exploration of resistance traits, the researchers seem to believe that all three approaches—selecting resistant American chestnuts, breeding hybrids derived from Chinese chestnuts, and directed genetic modification—can help bring the American chestnut back.

The researchers warn, though, that as environmental disturbances and invasive species continue to push some key species to the brink of extinction, we need to get better at this kind of species rescue operation.

Science, 2026. DOI: 10.1126/science.adw3225 (About DOIs).

Bringing the “functionally extinct” American chestnut back from the dead Read More »

Watch Kanzi the bonobo pretend to have a tea party

animal behavior, animal cognition, Biology, bonobos, primate studies, Science / Kris Guyer / February 5, 2026

Such studies have nonetheless been met with skepticism when it comes to interpreting the behavior as evidence of animals’ ability to engage in make-believe. For instance, it’s possible the animals are responding to behavioral cues, like the direction of a gaze, to solve such tasks.

“Kanzi, let’s play a game!”

Enter Kanzi, a 43-year-old bonobo who lives at the Ape Initiative and is capable of responding to verbal prompts, either by pointing or using a lexigram of more than 300 symbols. There had also been anecdotal observations of Kanzi engaging in pretense. Krupenye et al. conducted three distinct experiments with Kanzi, each involving an 18-trial session.

In the first experiment, a scientist would offer a verbal prompt: “Kanzi, let’s play a game! Let’s find the juice!” They would then place two empty transparent cups on a table and pretend to fill them from an empty transparent pitcher, with another verbal prompt (“Kanzi, look!”). The scientist would pretend to pour the “juice” in one of the cups back into the pitcher, placing the pitcher under the table. Then they asked, “Kanzi, where’s the juice?” and recorded which cup the bonobo pointed to first.

“If Kanzi could only track reality (that both cups were empty), he should have chosen at chance between the two options, whereas if his choices were guided by stimulus enhancement, he should have selected the incorrect cup that had been ‘emptied’ above chance,” the authors wrote. “In contrast, if Kanzi could represent the pretend juice, he should have chosen above chance the cup containing the ‘imaginary’ juice, the empty cup that had not been ‘poured’ back into the pitcher. That is exactly what Kanzi did.” Kanzi selected the correct cup 34 out of 50 times (68 percent).

Watch Kanzi the bonobo pretend to have a tea party Read More »