Four shallow pits in a broken bird leg do not sound like much of a fossil record. In this case, they preserve one of the rarest kinds of evidence paleontologists can find: physical traces of one formidable predator feeding on another.

The leg belonged to a large terror bird that lived in what is now Colombia during the Middle Miocene, more than 10 million years ago. A 2025 Biology Letters paper led by Andres Link concluded that the tooth marks fit a large caimanine, most plausibly a juvenile or subadult Purussaurus neivensis measuring roughly 4.6 to 4.8 meters long.

That is the firm part of the story. The fossil cannot tell us whether the caiman attacked a living terror bird or found its body near the water and scavenged it. This is one study, not settled consensus, and the distinction between a kill and a meal matters.

It is tempting to assemble prehistoric South America’s saber-toothed mammals, terror birds, giant snakes and crocodilians into one enormous cast. The fossil record is less tidy. These groups ranged across different environments and many millions of years, and not every famous animal met every other one.

La Venta, the Colombian fossil site at the center of this story, was extraordinary without any help from that kind of compression. A major review of the La Venta biome records more than 150 vertebrate taxa, including almost 100 mammals. Turtles, fishes, dolphins, manatees, large running birds and at least eight crocodylian taxa occupied a landscape shaped by rivers, channels and wetlands.

The land predators were unfamiliar. Terror birds were flightless, fast-running carnivores with deep beaks. Sparassodonts were predatory metatherian mammals, distant relatives of living marsupials rather than cats, even when some evolved saber-like teeth. Terrestrial sebecid crocodyliforms added another body plan to the mix.

One 2023 ecological analysis found unusually low diversity and abundance among La Venta’s terrestrial predators and argued that large reptiles may have occupied some of the macro-predatory space. That is a model of the community, not a head count taken 13 million years ago, but it helps explain why the waterways deserve as much attention as the land.

The specimen is cataloged as MT-0200. It consists of the lower end of a left tibiotarsus, a bird’s fused lower-leg bone, from a large member of the phorusrhacid family. The fossil had previously been described in a 2024 paper on a giant terror bird from La Venta, although a partial leg cannot supply a complete portrait of the animal.

On the back of the bone are four pits arranged in an oblique row. Their parallel spacing indicates that several teeth pressed into the bone during one movement of a jaw. The smallest mark measures about 5 by 4.1 millimeters. The three larger pits approach 10 millimeters across, and the deepest reaches roughly 3.4 millimeters into the cortical bone.

Several marks also have grooves running along their bottoms and small indentations at their ends. These details matter because they preserve the shape of the tooth tips and their cutting edges, not merely the violence of the bite.

None of the four pits shows healing.

That tells us the bird did not survive long enough for its bone to repair itself. It does not tell us whether the bite happened seconds before death, contributed to the death, or occurred after the animal was already dead.

Link and his colleagues did not simply match the holes to the largest available predator. They scanned the fossil in three dimensions, isolated the marked strip of bone, created a digital negative of the pits and compared it with teeth from La Venta’s known crocodyliforms.

The location of the damage provided one clue. Carnivorous mammals often gnaw the ends of long bones to reach spongy tissue and marrow, leaving chewed joint surfaces. Crocodylians tend to grip, pull apart and swallow sections of a carcass, which can leave serial marks along the shaft while the joint surface remains comparatively intact. MT-0200 fits the second pattern.

Tooth shape narrowed the field further. La Venta’s narrow-snouted crocodylians had slender, pointed teeth that should have left more conical pits. A terrestrial sebecid carried compressed, serrated teeth more likely to score bone. Several smaller caimans had broadly suitable tooth shapes, but their teeth were not large enough.

The spacing between the three larger marks was about 27 millimeters. When the researchers scaled a modern black caiman’s jaw to that spacing, they arrived at a total length of about 4.6 to 4.8 meters. Among the animals actually known from La Venta, the best match was a not-yet-fully-grown P. neivensis, the site’s largest crocodyliform.

It remains an attribution, not an identification tag. No Purussaurus tooth was found embedded in the bone, and the paper uses careful language such as “possibly” for good reason.

The most honest sentence in the 2025 paper is also the one most easily lost in retelling: the authors could not discard or confirm scavenging.

The absence of healing means this was not a wound the terror bird carried for weeks. But an unhealed mark can be made during an attack, around the moment of death or on a fresh carcass. Fossil bone rarely preserves the sequence well enough to separate those possibilities.

The geological setting offers a plausible scene, not proof. The layer containing the bird has been interpreted as a place where water channels once ran. If the bird was alive, it may have approached a channel to drink or hunt and entered the caiman’s effective range. If it was dead, its carcass may have washed into the water or simply lain close enough for a caiman to retrieve.

In either scenario, a large aquatic predator consumed part of a large terrestrial predator. Only the active-predation scenario allows us to say the terror bird itself became prey. The fossil keeps that possibility open, but it does not settle it.

“Apex predator” can sound like a permanent rank. Ecology is more local than that. A terror bird equipped to run down or strike animals on firm ground carried those advantages to the water’s edge, but it also entered a place where concealment, buoyancy and an ambusher’s jaws favored the caiman.

Modern ecosystems offer the same basic warning about fixed rankings. Large crocodilians take terrestrial animals at drinking sites, while jaguars sometimes kill caimans. Which predator has the advantage depends on body size, position, surprise and whether the encounter begins on land or in water.

The terror-bird bone is also not the only fossil hint that P. neivensis dealt with large land animals. In 2020, Rodolfo Salas-Gismondi and colleagues reported 46 bite marks on the shin of a ground sloth from the proto-Amazonian Pebas wetlands. Those marks matched a young or subadult P. neivensis around four meters long. The authors favored predation in that case, while acknowledging how difficult it is to distinguish hunting from scavenging by marks alone.

I think the two fossils are more informative together than either is alone. They do not establish a menu or tell us how often such attacks happened. They do show that the interface between wetland and dry ground repeatedly brought large caimans into feeding contact with substantial terrestrial animals.

There is one more trap in this story. Popular descriptions often take the largest estimate for Purussaurus brasiliensis, a later and larger species, then apply it to the whole genus. That can turn every Purussaurus into a ten-meter animal whether the fossil supports it or not.

P. neivensis was a distinct Middle Miocene species, and its size estimates depend on the specimen, the skull dimension used and the equations chosen to scale an incomplete animal. A 2022 reassessment of Miocene caimanine body size shows why those methodological choices matter.

The terror-bird study makes a narrower claim. The jaw that marked MT-0200 belonged to a caimanine about 4.6 to 4.8 meters long, and P. neivensis is the best candidate available. That animal was enormous beside most living caimans, yet probably juvenile or subadult. We do not need to enlarge it further to make the encounter interesting.

The fossil does not give us a slow-motion attack between prehistoric champions. It gives us something quieter and more defensible: evidence that the boundary between land and water made dominance conditional.

An animal could command one habitat and become food in the next.