Researchers who work with octopuses learn a first rule early: never approach the tank from above. To the animal, you are a shadow, and a shadow reads as a predator. That rule was on a lot of minds this June, when a team at Dartmouth published the first evidence that any invertebrate can use a mirror to find prey it cannot see.

On June 3, 2026, Mary Kieseler and her supervisor Peter Tse at Dartmouth reported in Current Biology that three California two-spot octopuses, Octopus bimaculoides, had learned to use a mirror to locate a hidden food reward, picking the correct side roughly 73% of the time [1][2]. The behavior, mirror-guided spatial localization, had previously been documented only in certain mammals and a handful of birds.

"We don't enter the world knowing how to use a mirror but learn how to use a mirror. Octopuses can also learn how to use a mirror to infer where things are in the world."

That is how Tse, the senior author, put it [1]. The octopuses learned it too.

The Mirror Test, But Not the Famous One

The phrase "mirror test" carries baggage. In animal cognition, it usually refers to the Gallup self-recognition test, in which a researcher places a mark on an animal's body and watches whether the animal uses the mirror to investigate the mark. Octopuses do not pass that test. They treat their reflection as another animal, sometimes posturing, sometimes trying to grapple with it [4]. The Dartmouth experiment is more specific. The animals were not asked whether they recognized themselves. They were asked whether they could use a reflection as a guide to where a real stimulus was hidden.

The team stripped the search to vision. Octopuses taste by touch, so a live crab would have leaked odor cues; they projected a virtual crab instead [1][4]. Over ten to twelve trials per animal, the octopuses learned to suppress the impulse to attack the reflection and travel to the real location. In some trials, an animal climbed up and over the start box rather than swimming around it, suggesting flexible problem-solving rather than a fixed path [1][3].

A Mind That Lives Mostly in Its Arms

What makes the result stranger is what the octopus is doing this with. An adult has roughly 500 million neurons, and only about a third sit in any central brain. The other two-thirds, around 320 million, are distributed through the eight arms, with roughly 40 million per arm [12]. Classic experiments from Binyamin Hochner's lab at Hebrew University have shown that a severed octopus arm can still reach, grasp, and respond to stimuli. The arms do not just receive instructions, they make their own decisions [11].

"Nine brains" is the popular shorthand, and it is approximate but points to something real. The central brain, with roughly 180 million neurons, sits in a doughnut of tissue between the eyes and houses the vertical lobe, a learning and memory structure that evolved independently from the vertebrate hippocampus but appears to do something analogous [11]. Locally, the arms decide how to grip, how to probe, when to let go.

The rest of the body is no less foreign. An octopus has three hearts: two branchial hearts pushing blood through the gills, and one systemic heart serving the body. The systemic heart stops beating when the animal swims, one of the reasons octopuses prefer to crawl [13]. The blood is blue, because the oxygen carrier is hemocyanin, a copper-based molecule that works better than our iron-based hemoglobin in cold, low-oxygen water [13]. Nine brains, three hearts, blue blood, and arms that think locally.

Sleep That Looks Like Dreaming

You may have noticed the color changes in a sleeping octopus. In 2023, Hochner's team described the phenomenon in Nature as a two-stage sleep architecture with a striking vertebrate parallel [5]. The animal settles into long quiet stretches with closed eyes and pale skin. Roughly once an hour, it slips into a brief active phase: the skin ripples through patterns, the eyes twitch, breathing grows more variable. In a typical 24-hour period at 22°C, the animals experience around ten active bouts of about 75 seconds each, separated by twelve quiet bouts of roughly 50 minutes [5].

The authors stop short of saying the animals are dreaming. The signatures resemble REM sleep, but resemblance is not identity. Chromatophores, the same pigment cells the animal uses for camouflage, drive the color cycling. Active bouts are short and fragmented, lacking the full postural atonia of mammalian REM. Two-stage sleep has now been documented in mammals, birds, and cephalopods, lineages separated by more than half a billion years of independent evolution [5][6]. The most parsimonious reading is that complex cognition comes with a particular kind of off-line processing, whatever its subjective content turns out to be.

Tools, Problems, and the Question of Maps

The Dartmouth octopuses are not the first cephalopods to surprise us. In 2009, Julian Finn and colleagues described veined octopuses, Amphioctopus marginatus, collecting coconut shell halves in Indonesian waters, carrying them across the seafloor, and assembling them into portable shelter when threatened [8][9]. The behavior qualifies as tool use because the shells are transported for future use, not immediate purpose.

Add the puzzle-solving of captive octopuses, cuttlefish signaling, and a picture starts to form. The octopus is not just clever; it is built on a body plan so unlike our own that the usual comparisons mislead. As Tse put it, hunters benefit from an internal map of their territory, and octopuses may have something like that too [1]. Kieseler framed the broader implication as convergent evolution [1][3].

A Distant Inheritance

The last common ancestor of humans and octopuses sits deep in evolutionary time. Dartmouth's team dates that ancestor, a small worm-like creature, to between 350 and 500 million years ago [1]. Closer to 550 million years is where the 2023 Nature sleep paper places the broader vertebrate-cephalopod split [5]. A Hebrew University press release frames the divergence at roughly 700 million years [11]. None is exactly wrong; they refer to different nodes on the same deep tree. The safe shorthand is that the last common ancestor lived more than half a billion years ago, and the two lineages have been building minds independently ever since.

This is what makes the cephalopod story important beyond cephalopods. Peter Godfrey-Smith's 2016 book Other Minds argues that complex, body-supported intelligence has evolved independently in arthropods, cephalopods, and vertebrates, three times, on three different body plans [10]. The mirror test from Dartmouth, the coconut shells from Indonesia, the REM-like sleep from the 2023 paper, the puzzle-solving in labs: all of it can be read as data points in that larger comparative project. When a remote lineage arrives at a sophisticated cognitive capacity, the simplest explanation is that the problem of being a minded, behaving animal is solved, by physics, in roughly the same way.

The octopus at Dartmouth does not know any of this. With its nine-brained, arm-thinking, copper-blooded, dream-cycling architecture, it probably cannot frame a question about itself. The next time a mirror is angled in its tank, it does what it has learned: it goes around.