A forest rodent uses its lips and teeth to check whether a cache will keep
Picture an agouti in a Panamanian forest, holding a peanut in its paws. It turns the nut over, presses its front teeth firmly against the shell without breaking it, turns it again, and presses again. Then it either eats the peanut on the spot or runs off and buries it.
This careful quality check is part of agouti scatter hoarding, the habit of burying single seeds across a territory to retrieve later. Researchers from the University of Pennsylvania and the Smithsonian Tropical Research Institute argue that those few seconds of fiddling are a test of whether a cache will keep, and that the animal is judging something other than what you might expect.
Eat Now or Save for Later?
Agoutis are large, daytime, fruit-and-seed-eating rodents. They bury single seeds in shallow holes, about 4 cm deep on average, across their territory and return for them later. Seeds they never retrieve can grow into trees, which makes agoutis important seed spreaders. Each burial is a bet that the food will still be good, and that no one will steal it, by the time it’s needed.
Earlier studies suggested that scatter-hoarding rodents decide by a seed’s perceived internal value, often inferred from its weight. This study tested that idea in the wild at the Smithsonian’s station in Gamboa, Panama, over two field seasons (spring 2025 and winter 2025-26). Cameras filmed the animals, who came and went freely and were recognized by scars, notched ears, and fur markings.
Across 251 recorded encounters with fruit, agoutis treated foods very differently. They never buried apples, buried bananas 2% of the time and oranges 5%, grapes 12%, and peanuts 38%. Soft, juicy fruit gets eaten, while hard, dry, shelled food gets saved. Peanuts aren’t native to Panama, but the authors say they are similar in size to many nuts agoutis find there, so the team focused on them.
Agouti Scatter Hoarding Step by Step: How They Bury Nuts
Scatter-hoarding is a long, structured routine, and the researchers filmed every stage:
- Transport. The agouti carries the nut in its mouth, far from where it found it, to a chosen spot.
- Dig. On suitable soil, it digs a small pit with its forelimbs.
- Place. It drops the nut into the hole and presses it down firmly.
- Cover. Using forward strokes of its forelegs, it pushes soil back over the nut.
- Finish with a leaf. It often finds a leaf or other debris and lays it over the spot, which probably helps hide the cache.
Once an Agouti Starts Hoarding, It Keeps Hoarding
The researchers gave each agouti peanuts one after another, so they could record every eat-or-bury decision in order. The 21 agoutis didn’t gradually become hoarders. Each ate the first few, buried one, and from then on mostly buried.
The team compared four statistical models, scoring each with AIC, which rewards fit while penalizing complexity (lower is better). A model with two hidden internal states, eating and hoarding, scored 281. A single abrupt change scored 461, a steady trend 468, and a constant chance of burying 591. In the best-fitting model, the hoarding state never switched back to eating within the experiments, though animals could still eat an occasional peanut while in hoarding mode.
The switch also changes behavior. Before the first burial, agoutis rotated peanuts in about 5% of cases. Afterward, about 48%. In the field data, peanuts that ended up buried were far more likely to have been rotated than peanuts that were eaten (76% versus 17%). For grapes there was no such difference, since rotation was common either way (73% versus 80%).
The Empty Peanut Trick
To separate what’s inside a nut from its shell, the researchers opened peanuts along the seam, removed the kernels, and glued the shells shut. Empty peanuts weighed 0.9 g against 2.5 g for full ones. Sixteen agoutis already in hoarding mode buried both kinds at statistically indistinguishable rates and inspected them equally.
Then the team damaged shells with a 1 cm hole, a 2 cm hole, or crushing. Burying dropped significantly with damage, even though the nuts were just as nutritious. A smaller test with seeds from a native palm, Attalea butyracea, showed a similar pattern, though it involved only three animals.
Why a Cracked Shell Is a Bad Investment
The researchers buried peanuts 4 cm deep, protected by an iron grid, and dug samples up over a week. Peanuts with a 1 cm hole lost dry mass about nine times faster than intact ones, and crushed peanuts about eight times faster. Damaged peanuts also molded faster and softened more.
Damaged nuts were also easier to steal. When an intact and a crushed peanut were buried 20 cm apart in front of camera traps, the crushed one was dug up first in 77% of 26 pairs. Most thieves were agoutis (95%), and the rest were coatis. The authors suspect a more conspicuous smell. A damaged nut will spoil and get stolen, so eating it now is the better move.
How Agoutis Inspect Nuts: A Search Pattern That Beats Standard Search Methods in Simulation
The inspection is a loop of rotate, clamp, rotate, clamp, with some sniffing and licking. A clamp means biting down firmly on the shell without breaking it. Both the upper and lower teeth leave marks, which the team mapped by letting agoutis bite clay peanuts. The authors describe the lips and incisors as working like a fingertip, a “somatosensory fovea” in the authors’ words, likely supported by an expanded touch-processing region of the brain’s cortex.
By scoring 282 videos and tracking symbols marked on the shells of 30 peanuts, the team found that clamps are not random:
- Clamps concentrate on the centers of the two lobes.
- Moves run mostly along the nut’s length or around it, and diagonal moves are avoided.
- After a lengthwise move, another lengthwise move follows 65% of the time. After a move around the nut, the animal goes back to lengthwise 74% of the time. A typical sequence is two steps along, one around.
- Among 93 tracked lengthwise moves, those starting at a tip turned back toward the nut’s waist 90% of the time, which keeps the search on the shell.
Does the Pattern Actually Work?
The researchers ran 10,000 computer simulations of each search strategy on a 3D model of a peanut shell, with a hole at the waist and a budget of about eight clamps. The agouti-derived pattern found the hole more often than the other strategies at every hole size. In a logistic regression, the odds of success were about 0.93 times the agouti pattern’s for a self-avoiding walk and 0.92 times for a correlated random walk. The Lévy walk did worst, at 0.61 times, possibly because its long jumps overshoot a small target. A further test showed the pattern fits the object: when the shape was reversed from long to wide, a reversed search pattern did better.
One Touch of the Teeth Decides
In a follow-up experiment, 11 agoutis were offered peanuts with holes from 0.3 to 2.5 cm wide. The bigger the hole, the less often they buried the nut. Of all the behaviors the researchers measured (sniffing, licking, rotating, clamping, time spent), only one reliably predicted that an agouti would eat the nut: its mouth touching the hole. This happened in about 12% of clamps.
The teeth turned out to matter most. In 83% of detections the lower incisors touched the hole, and the upper teeth accounted for the other 17%.
The agoutis also changed how they searched just before finding the hole. Two moves around the nut in a row became twice as likely (from 26% to 54%), and 40% of detections were preceded by one lengthwise move followed by two around the nut. The last moves turned the nut about 90 degrees. The authors propose a two-stage sequence, which they have not directly confirmed:
- The lips, probably the upper lip, first feel the edge of the hole.
- The agouti then makes one long turn around the nut, swinging the lower incisors over the opening. Touching the hole with the teeth confirms the flaw and triggers the decision to eat.
To explain how one touch can settle the decision, the authors fit a drift diffusion model. Picture an internal meter that starts at zero. Each clamp on a sound shell nudges it a little toward “bury.” If it reaches +0.31, the animal buries the nut. If it falls to −0.53, the animal eats it. In the fitted model, touching a hole with the teeth knocks the meter down by 0.52, which the authors describe as about 61% of the distance between the two thresholds and equivalent to roughly five ordinary clamps pushing the other way. In short, mild reassurance builds up slowly, but one clear sign of damage can decide the matter at once.
What to Keep in Mind
The samples are modest: 11 to 21 animals per experiment. The animals were free-ranging but habituated to people, and the main tests used non-native peanuts. Hunger, competitors, and perceived risk all vary in the wild, though the authors note the results followed shell integrity consistently anyway. The lip-then-tooth sequence is a proposal based on behavior, not a direct measurement. The drift diffusion model is a fit to the data, not a direct measurement of the animal’s brain.
Bigger Picture
The study suggests a wild rodent plans ahead by judging how well food will keep rather than how rich it is. It combines a clean switch between modes, a touch-driven search tuned to an object’s geometry, and a decision that one clear signal can settle. The authors also see a parallel to human and primate eyes, whose scanning patterns adapt to the shape of the scene being searched.
Source
Study: Oro-manual assessment of food preservability in a wild rodent
Authors: Lucas Dumargne, Esther Tibi, Margaret Williams, Dumas Galvez, Camille Testard, Juan I Sanguinetti-Scheck (2026)
Read the full paper: https://www.biorxiv.org/content/10.64898/2026.09.23.752726v2
Leave a Reply