The Octopus and the Question of Where Intelligence Lives

Part 1 of a three-part series on distributed biological intelligence.

Octopuses have become something of a celebrity in popular neuroscience, and it's easy to see why. They open jars, they recognize individual humans, they squeeze their soft bodies through implausibly small holes, and they do all this with a nervous system that looks nothing like ours. The detail that gets repeated most often is that an octopus has roughly 500 million neurons, but only a small fraction of them are in the head. Most are spread out through the eight arms.

That number is real. The central brain is estimated at around 40 to 50 million neurons. The paired optic lobes add something like 130 million on top of that. And the eight arms together house around 350 million — somewhere in the neighborhood of two-thirds of the entire nervous system. Connecting all of it is a surprisingly thin wire: by classic estimates, only about 30,000 nerve fibers run between the central brain and the arm cords.

If you're used to the idea that intelligence happens in a brain, full stop, this picture is genuinely disorienting. Where is the octopus actually thinking?

The honest answer is that we don't fully know. But we know enough to say that the octopus arm is doing real work. Each arm contains a thick axial nerve cord running its full length, four smaller intramuscular nerve cords alongside it, and a small ganglion at the base of every single sucker. There can be hundreds of suckers per arm. Recent imaging has sharpened the picture considerably. Olson, Schulz and Ragsdale published a striking 2025 paper in Nature Communications showing that the axial nerve cord isn't a uniform tube but is organized into segments, with cell bodies arranged in repeating columns and a topographic map of the suckers built right into the wiring. A 3D molecular atlas published the year before by Winters-Bostwick and colleagues in Current Biology added another layer of detail, identifying multiple distinct neurochemical cell types whose distribution differs from arm base to arm tip.

So here is the part of the story that's solid: the octopus has outsourced an enormous amount of work to its arms. The brain doesn't have to micromanage every sucker. The arm has its own circuitry for grabbing, exploring, and propagating the wave of muscle contractions that produces the species' signature reaching motion.

So far, so cool. Now we need to talk about where popular accounts have run ahead of the evidence.

You've probably read that severed octopus arms can keep behaving for hours after detachment — that they grasp objects, withdraw from pain, and even, in some tellings, learn or remember. The first two are true. A landmark 2001 paper by Sumbre and colleagues showed that an isolated arm, given a small electrical pulse to its nerve cord, will produce the species-typical reaching motion with kinematics nearly identical to the real thing. Other work has shown that detached arms will curl away from a pinch and grip a finger that touches them. This is real, and it's strange and beautiful.

But the arm isn't thinking. What you're seeing are reflex arcs and stored motor programs running on whatever cellular fuel is left in the tissue. The arm is executing — it isn't deciding, and it isn't learning. When researchers actually test learning carefully, the central brain turns out to be required. A 2020 study by Tamar Gutnick and colleagues trained octopuses on a maze that required them to insert a single arm. Once an animal had learned the task with one arm, it could perform it correctly using arms that had never been used in training. That's a clean signal that the learning happened centrally and was then made available to all the arms — not that each arm learned independently. The accompanying commentary in Current Biology was bluntly titled "Octopus Arms Need Brains to Learn Their Way."

It's a subtle point, but worth getting right, because the next two essays in this series ask whether something analogous is happening inside the human body. The story I want to tell is genuinely interesting only if we're honest about what's been demonstrated and what's been wished for.

So let me try to summarize the octopus the way I think the evidence warrants.

The octopus is a real example of distributed neural computation. Most of its nervous system is not in its head. Its arms contain genuine local circuitry that handles fast, automatic, body-shape-aware movement without waiting for central permission. The brain issues high-level intentions — explore here, grab that — and the arms figure out how. The arm is not a puppet. It's also not a separate mind. It's something in between, and we don't yet have a clean word for it.

That "something in between" is the thing I want to look at next, because the human body has its own version of it. We don't have eight arms full of neurons. But we do have something else: a long, mostly-sensory nerve that wanders from the brainstem down into the chest and abdomen, sampling the heart, the lungs, the gut, and the immune system. It's called the vagus, and the more closely you look at it, the less it resembles a wire.

Next: Your Vagus Nerve Is Not Just a Wire.

5 comments