In a dimly lit laboratory tank at Dartmouth College, an octopus sits motionless inside a small open-topped chamber. A few seconds after a crab image flickers into view on a screen behind it—visible only as a mirror reflection in front—the animal makes a decisive turn, climbs over the chamber wall, and navigates directly to the spot where the actual crab is being projected. It cannot see the crab directly. It has never seen it. But it knows, somehow, where to go.
That behavior—deliberate, spatially informed, and repeatable—is at the heart of a study published June 3, 2026, in Current Biology by researchers at Dartmouth. The paper reports that three California two-spot octopuses (Octopus bimaculoides) learned to use mirrors to locate a hidden food reward, making correct navigational choices roughly 73 percent of the time. It is the first demonstration of mirror-mediated spatial localization in any invertebrate, and it adds a new chapter to the decades-long scientific inquiry into the cognitive lives of cephalopods.
“Our findings are the first to demonstrate that invertebrates can use mirrors to understand their environment to find prey. It’s a skill that previously has only been documented in vertebrates, such as in some mammals and some birds,” said Mary Kieseler, the lead author of the study.
A Mirror in the Tank
The experiment, designed by Kieseler and senior author Peter Tse, a professor of psychological and brain sciences at Dartmouth, proceeded in three stages. First, the octopuses were habituated to the presence of a mirror in their experimental tank—allowed to approach it, examine it, and come to understand it as a reflective surface rather than a passageway or a conspecific. This last point was not trivial: prior research has established that cuttlefish and octopuses often react to their reflections as if confronting another animal, displaying agonistic signaling behaviors toward mirror images.
Once habituated, the animals entered a learning phase. A live crab was hidden from direct view and made visible only through a mirror. To retrieve the crab, an octopus had to suppress its impulse to lunge at the reflection and instead make a 90-degree turn to reach the actual jar. Learning to do this consistently took between 10 and 12 trials per animal—a finding the researchers note suggests that predatory behavior in octopuses is not purely reflexive but is subject to executive control.1
The final testing phase introduced a virtual crab—a white silhouette projected onto a translucent screen—visible only as a reflection in a mirror positioned at the center of the tank. The octopus, placed in a small start chamber, had to exit, make a 180-degree reversal, and navigate to whichever side of the tank the projection matched. A real crab was dropped from above as reward for a correct choice. No punishment was given for incorrect choices.
“Just as new drivers learn to use a rearview mirror to track other vehicles, octopuses can also learn how to use a mirror to infer where things are in the world,” said Tse.
More Than Association?
The most striking behavioral observation came not from the overall success rate but from a particular class of choices: in 59 percent of correct trials, octopuses climbed over the side walls of the start chamber rather than exiting through the front and swimming around. This behavior moved them away from the salient mirror reflection and toward a visually occluded location that was spatially aligned with where the crab image was being projected. The researchers argue this suggests something beyond simple stimulus-response association.2
In associative learning, an animal learns that visible stimulus X predicts reward at location Y. But climbing the chamber wall requires the animal to first move away from the only visible cue—the reflection—toward a location it cannot yet see, implying some internal spatial model of the tank. The octopuses also succeeded on the first testing trial despite the experimental layout being substantially different from the training setup, which would be unexpected if they had learned only to repeat specific movements.
The authors are careful not to overclaim. They acknowledge that the study was not designed to definitively distinguish between associative and spatial-reasoning mechanisms, and they call for future research using novel mirror configurations, generalization tests, and larger sample sizes. But the pattern of behavior, they write, is more consistent with an internal spatial representation—a mental map—than with simple reinforcement learning alone.
Half a Billion Years of Separation
Cephalopods diverged from the vertebrate lineage more than 520 million years ago, making them among the most evolutionarily distant animals from humans with demonstrably sophisticated cognition. In the intervening half-billion years, their nervous systems developed along an entirely independent trajectory. Unlike vertebrates, whose centralized brains are housed in the skull, octopuses distribute roughly two-thirds of their neurons across their eight arms. Their evolution of complex cognition—spatial learning, problem-solving, camouflage control, tool use—represents what researchers describe as a convergent solution to the challenges of being a predator in a complex environment.3
Earlier work had already established that O. bimaculoides is capable of spatial learning in laboratory settings, and that octopuses can be trained to respond to virtual representations of crabs with natural predatory behavior. The Dartmouth study builds on that foundation by demonstrating a qualitatively new capacity: not just responding to visual stimuli, but integrating reflected visual information with a three-dimensional representation of surrounding space.4
“Octopuses are among the most evolutionarily distant animals from humans, as our last common ancestor was a worm that lived 350 to 500 million years ago. Given that such a remote organism has independently evolved the means to use a mirror as a tool to process spatial cognition suggests that the underlying cognitive processes might be subject to convergent evolution,” said Kieseler.
What Counts as Mirror Use?
Mirror-mediated localization—using a reflection to find an object outside the direct line of sight—has been documented across a range of vertebrate taxa, including multiple primate species, dogs, pigs, New Caledonian crows, and African grey parrots. It is considered by some researchers to be a cognitive precursor to self-recognition, the more famous ability tested by the Gallup mark test, because it requires linking a visible image to a real-world spatial position. Until now, no invertebrate had been shown to possess this capacity.5,6
A separate and still-unresolved question is whether octopuses might pass the mark test itself. A preliminary protocol for adapting the Gallup mirror self-recognition test for Octopus vulgaris was recently developed, and one study has found that O. vulgaris can discriminate between individual conspecifics for at least 24 hours—a cognitive feat that may provide scaffolding for the kind of self-other distinction that self-recognition requires. The Dartmouth study does not speak to self-recognition, but the authors note that their findings open that question for further investigation.7
Hunters with Mental Maps
Peter Tse frames the significance of the work in ecological terms. Octopuses, he notes, are ambush predators operating in complex, three-dimensional environments—coral reefs, rocky seafloors, tide pools—where prey can be obscured by terrain and threats can approach from any direction. In such environments, a hunter with an accurate internal map of its territory has a decisive advantage.
“Hunters are very effective when they have a mental map of their territory, so that they know where they are in relation to their environments,” Tse said. “Our work suggests that octopuses might also have internal maps, an internal representation of space.”
The efficiency data from the study support the idea that learning was genuine. Travel duration to the reward location decreased significantly across trials, with octopuses moving faster as experience accumulated. Path length also shortened across correct trials. The animals were not wandering to solutions—they were becoming more directed.
Whether that directionality reflects spatial mapping or sharpened associative prediction remains a productive open question. For now, the Dartmouth study stands as the first evidence that an invertebrate can parse a mirror reflection and use it to navigate toward something it cannot see—a finding that, whatever its ultimate mechanistic explanation, pushes the known boundaries of cephalopod cognition another step further from where anyone expected them to be.
Endnotes
1. Kieseler, M., et al. “Octopus bimaculoides can learn to utilize a mirror to localize a reward outside the line of sight.” Current Biology 36 (June 22, 2026). https://doi.org/10.1016/j.cub.2026.05.012
2. Ibid.
3. Moroz, L.L. “On the Independent Origins of Complex Brains and Neurons.” Brain, Behavior and Evolution 74 (2009): 177–190. https://doi.org/10.1159/000258665
4. Boal, J.G., et al. “Experimental evidence for spatial learning in octopuses (Octopus bimaculoides).” Journal of Comparative Psychology 114, no. 3 (2000): 246–252.
5. Anderson, J.R. “Mirror-mediated finding of hidden food by monkeys (Macaca tonkeana and M. fascicularis).” Journal of Comparative Psychology 100 (1986): 237–242.
6. Broom, D.M., H. Sena, and K.L. Moynihan. “Pigs learn what a mirror image represents and use it to obtain information.” Animal Behaviour 78 (2009): 1037–1041.
7. Amodio, P., and G. Fiorito. “A preliminary attempt to investigate mirror self-recognition in Octopus vulgaris.” Frontiers in Physiology 13 (2022): 951808.