A fossil from the Gobi overturns what the teeth said about an entire mammal group

Zhelestids, a group of small Cretaceous mammals known almost entirely from their teeth, have been misfiled for nearly four decades. A skeleton from Mongolia’s eastern Gobi Desert, named Tamirkhan balcarceli, shows the animals were not early relatives of placental mammals at all. They were members of the zalambdalestoids, a lineage of slender, insect-eating mammals, and their rounded, plant-grinding molars were a case of evolution arriving at the same shape by a different road.

The animal was about the size of a chipmunk, six to seven inches from head to tail.

A graduate student’s find and the teeth that misled

Andres Giallombardo found the specimen in 2004 as a graduate student on an American Museum of Natural History expedition to the Gobi. He is now the lead author of the formal description, which appeared in Nature as “Cretaceous zhelestid mammals are zalambdalestoids”. The co-authors include Eva Hoffman of Yale University, Maureen O’Leary and Paúl Velazco, Shawn Zack, and Michael Novacek.

Zhelestids had been known since the 1980s from jaws and isolated teeth, mostly from Central Asia. Their molars were low-crowned and rounded, the sort of dental kit associated with an animal that eats plants. Several paleontologists read that as a sign the group sat near the base of the placental mammals, the branch that today includes everything from shrews to whales. A herbivorous placental radiation in the Late Cretaceous, alongside the dinosaurs, was a tidy and influential story.

Hind legs, ankles and incisors that say zalambdalestoid

The Nature abstract describes the new specimen as combining zhelestid-like molars with a skull and a cursorially adapted partial hind limb. Those leg bones are the problem for the old reading. According to the release as republished by ScienceDaily, the hind legs are long and slender with distinctive ankles, traits described as “unmistakably zalambdalestoid.” The lower incisors are specialized, and the hindlimb bones are fused, both features the authors treat as diagnostic of the zalambdalestoid group.

Zalambdalestoids were small runners and leapers that fed on insects, a lifestyle that suits long legs and ever-growing front teeth and poorly suits a browsing animal.

Convergence in the molars

If the molars did not signal placental ancestry, what produced them? The museum’s account gives the answer: the rounded teeth associated with zhelestids were probably an example of convergent evolution, in which unrelated lineages independently arrive at similar traits because they face similar demands. A zalambdalestoid that began eating more plant matter would face the same mechanical problem as any other plant eater, and similar crushing surfaces are a likely solution. Textbook cases include the placental sabre-toothed cats and the marsupial Thylacosmilus, which grew near-identical stabbing canines from separate ancestries. In each case the shared feature reflects a shared job, not shared descent, and only the rest of the skeleton reveals which explanation applies. That is the position the zhelestid molars occupied until the Gobi skeleton supplied the missing evidence of the legs, ankles and incisors.

Maureen O’Leary of Stony Brook University, a co-author of the study, framed the lesson in the museum’s release. “While teeth remain among the most informative fossils available,” she said, the work shows that teeth cannot always reveal what the whole animal looked like. Teeth survive far more often than skeletons, and mammal classification in the Mesozoic leans on them heavily, so a dental trait that proves convergent is a warning about a lot of other placements.

MorphoBank scoring of the Tamirkhan skeleton

The authors did not rest on a visual impression. According to the museum, the team scored the specimen in MorphoBank, an open database in which researchers code hundreds of anatomical characters across many mammals, and let the data place the animal on the tree. The Nature abstract reports that the result resolves a puzzle that had stood for 38 years: the molars said one thing, the skeleton another, and only a fossil that preserved both could referee.

Giallombardo drew the practical conclusion in the museum’s account. “This discovery illustrates why fieldwork remains indispensable to understanding life’s history,” he said. The skeleton is the holotype, the single reference individual to which the species name is permanently attached, and Nature’s page records it as a complete individual from the Upper Cretaceous of Mongolia. It was collected two decades before it rewrote anything; the description is also indexed in PubMed and carries the DOI 10.1038/s41586-026-10861-5.

Collected in 2004 and described in Nature only in 2026, the Tamirkhan skeleton settled a 38-year puzzle for one group, yet it gives no verdict on how many other fossil groups defined by teeth alone sit in the wrong place. Placental origins in the Cretaceous are still argued largely from dental evidence, and each such case needs a skeleton of its own.

This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.

More from Morning Overview