Three California two-spot octopuses learned to use a mirror to find a live crab hidden behind them, choosing the correct side of their tank about 73% of the time. Published in Current Biology on June 3, 2026, the Dartmouth College experiment is the first documented case of any invertebrate using a reflected image to locate food outside its direct line of sight. The result places a mollusk alongside dogs and other vertebrates in a small club of animals that can extract spatial information from a mirror, a skill that requires translating a flat reflection into real-world direction.
Why mirror-guided foraging in a mollusk changes the cognitive map
Until now, mirror-guided food localization had been confirmed only in vertebrates. A Monash University study on dogs showed that canines could use a mirror to find hidden food, but the behavior was assumed to depend on the kind of centralized brain architecture vertebrates share. Octopuses have roughly 500 million neurons, two-thirds of which sit in their arms rather than in a central brain. Showing that this distributed nervous system can still convert a mirror image into a correct left-or-right movement decision forces a rethink of what neural hardware the task actually requires.
One plausible explanation draws on earlier work showing that octopuses already integrate vision with arm position. A 2011 study in marine neurobiology demonstrated that Octopus vulgaris uses visual information to determine the location of its arm, confirming that the animals possess a feedback loop between what they see and where their limbs are in space. If mirror-guided navigation recruits these same sensorimotor circuits, the new behavior may not demand an entirely separate cognitive module. Instead, it could represent an extension of circuits the animals already use to coordinate visually guided reaching. Individual differences in arm proprioception accuracy, as measured in that 2011 task, might predict how quickly each animal learns the mirror trick, though no study has tested that link directly.
The work also broadens the comparative framework for animal cognition. Most mirror experiments have been run in primates, birds, and mammals, often with a focus on self-recognition. By showing that an invertebrate can use a reflection for a practical, goal-directed task, the Dartmouth team adds a new data point to debates about whether complex cognitive skills require a vertebrate-style brain. It also underscores how much can be learned from species whose nervous systems are organized in radically different ways from our own.
Three octopuses, a crab in a jar, and a 73% success rate
The Dartmouth team worked with three Octopus bimaculoides, a small species native to the California coast. Each animal was placed in a tank with a mirror positioned so that a reward, a live crab sealed inside a jar, was visible only through the reflection. The crab sat behind and to one side of the octopus, completely outside its normal field of view. After an acclimation period in which the animals grew comfortable with the mirror, training trials began. According to Dartmouth’s institutional release, the octopuses traveled to the correct side about 73% of the time in the mirror-only condition.
That figure matters because it sits well above the 50% baseline that random guessing would produce, yet it also falls short of the near-perfect accuracy vertebrates sometimes achieve on similar tasks. The gap could reflect genuine cognitive limits, or it could stem from the small sample of three animals and the inherent variability of working with octopuses in captivity. Full trial-by-trial performance data and the statistical analyses behind the 73% figure are available only in the peer-reviewed paper itself, not in the press materials.
The choice to use a virtual crab, visible solely through the mirror, was deliberate. It ruled out the possibility that the octopuses were simply following chemical cues or water currents to the food. Because the crab was sealed in a jar, odor plumes were minimized. And because the jar sat behind the animal, only the mirror could reveal its location. That design isolates vision as the information channel and the mirror as the mediating tool.
The experiments also included control conditions in which the mirror was removed or repositioned. In those cases, the animals’ performance dropped toward chance, strengthening the case that they were not memorizing a fixed route in the tank but actively using the reflection to infer where to go. Still, with only three individuals, it remains possible that idiosyncratic strategies or personalities affected the results.
What separates mirror use from mirror self-recognition
The Dartmouth finding is distinct from the more famous, and more contentious, question of whether octopuses recognize themselves in a mirror. A recent assessment of self-recognition in Octopus vulgaris explored whether the animals changed their behavior when confronted with their own reflection, a standard test associated with self-awareness in great apes, elephants, and certain birds. That study did not produce clear evidence of self-recognition, and the question remains open.
Mirror-guided navigation, by contrast, does not require the animal to understand that the reflection is itself. It requires only that the animal extract directional information from the image: the crab is on the left in the mirror, so the crab must be on the right behind me. This is a spatial inference task, not a self-concept task. Conflating the two would overstate what the Dartmouth experiment shows. What it does show is that an invertebrate can perform a mental rotation of reflected spatial information and act on it, which is itself a significant cognitive operation.
For researchers, the separation between these abilities is useful. It suggests that mirror-based tests can be decomposed into component skills: perceiving and tracking objects in the reflection, mapping those perceptions onto real-world coordinates, and, potentially, recognizing one’s own body as an object among others. Octopuses now appear to clear at least the first two hurdles.
Open questions and what to watch next
Several gaps remain in the evidence. The experiment involved only three animals, a sample size that limits the strength of any statistical claim. Whether the 73% success rate holds across a larger and more diverse group of octopuses is unknown. Future work could test additional individuals, including other cephalopod species, to see whether mirror-guided foraging is widespread or an outlier.
Researchers will also want to know how flexible the behavior is. Can octopuses use a mirror to find different kinds of objects, or to avoid an unpleasant stimulus rather than approach a reward? Can they adapt when the mirror is rotated or when the reflected scene changes in more complex ways? Systematically varying the geometry of the tank and the position of the mirror would help reveal whether the animals are building a generalizable rule about reflections or relying on simpler learned associations.
Neuroscientists, meanwhile, are likely to focus on mechanisms. Because octopus arms contain their own dense neural networks, one possibility is that each arm independently integrates visual and proprioceptive information, with the central brain coordinating only high-level goals. Recording neural activity in the brain and arms while an octopus solves a mirror task could show where spatial information from the reflection is first transformed into a movement plan.
The study also highlights the value of publishing detailed methods and data in venues such as Current Biology, where other labs can scrutinize the design and attempt replications. As more groups test cephalopods with mirrors, patterns may emerge about which environmental conditions, training regimes, and individual traits support successful mirror use.
For now, the Dartmouth results stand as a provocative demonstration that a creature with a decentralized nervous system can use a mirror as a tool for navigation. That single finding does not settle debates about consciousness in octopuses, nor does it prove they possess a human-like understanding of reflections. It does, however, push the boundaries of what kinds of minds can manipulate abstract spatial information-and suggests that even animals whose brains are spread into their limbs may share more with us, cognitively speaking, than their alien bodies first imply.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
