Pl. II

Part I · The Architecture of an Alien Mind · Chapter 2

Embodied Cognition and Autonomous Arm Control in Octopuses

The octopus is the canonical animal model for embodied cognition. Of an estimated ≈500 million neurons, roughly two-thirds reside outside the central brain—about 350 million distributed along the eight arms in axial nerve cords and ganglia—which motivates the popular framing of a body that partly "thinks" for itself. The foundational demonstration came from Sumbre, Gutfreund, Fiorito, Flash, and Hochner (2001, Science), who showed that the stereotyped bend-propagation reach of Octopus vulgaris is a peripheral motor program. Arm extensions evoked mechanically or electrically in arms whose connection to the brain had been surgically severed reproduced the kinematics of voluntary reaches almost exactly: a bend forms near the base and propagates distally at a characteristic velocity profile. The brain need only issue a "go" command and specify direction; the arm's own circuitry computes the rest. This drastically simplifies control of an appendage with effectively infinite degrees of freedom.

Sumbre et al. extended this to purposive movement. In Nature (2005) and Current Biology (2006), they analyzed the arm-to-mouth "fetching" motion and found the octopus transiently converts its soft, hyperredundant arm into a quasi-articulated limb with three dynamic joints—a vertebrate-like, jointed strategy. Strikingly, the medial joint forms where two waves of muscle activation, propagating toward each other from opposite ends, collide; one wave is triggered by the central motor command, the other by sucker sensory input contacting the object. They argued that a kinematically constrained, joint-controlled limb is the optimal solution for precise point-to-point movement, and that octopuses and humans "evolved similar strategies" despite ≈500 million years of divergence—a case of convergent motor-control logic.

The tidy "autonomous arms" story has since been qualified. Classical work (Wells, 1970s) held that octopuses lack proprioception—they reportedly cannot learn tasks requiring knowledge of their own arm position by touch alone, and their motor system was thought to sacrifice body-awareness for flexibility. Gutnick, Byrne, Hochner, and Kuba (2011, Current Biology) challenged this with an elegant transparent-maze reaching task: an octopus had to guide a single arm out of the water (losing chemotactile guidance) along a maze to a food reward, relying on vision to direct the arm. Six of seven animals learned within 61–211 trials; when the transparent maze was swapped for an opaque one, performance collapsed to naïve levels—showing the octopus can visually track and steer one of its own arms, a form of goal-directed complex movement not previously demonstrated. This is often summarized as the octopus using visual information to determine "the location of its arm."

Gutnick, Zullo, Hochner, and Kuba (2020, Current Biology 30:4322–4327) closed the loop on the proprioception debate. Using a two-choice single-arm Y-maze where the correct path could only be sensed by the arm inside the maze (not by the eyes), they showed 5 of 6 octopuses learned the operant task—the central brain must therefore receive and use non-visual, peripheral (proprioceptive and tactile) information to make the decision, since "the learning takes place centrally but the information is detected only by the arm." Gutnick's reframing became widely quoted: rather than "an octopus with nine brains," it is better described as "one brain and eight very clever arms." This is a decisive move away from the folklore of fully independent arms toward bidirectional central–peripheral integration.

Coordination among arms is also less autonomous—and less rhythmic—than expected. Levy, Flash, and Hochner (2015, Current Biology) provided the first kinematic analysis of crawling and found it uses no central-pattern-generator rhythm: Fourier analysis revealed no periodicity in arm recruitment. The octopus simply elongates one or more arms to push the body the opposite way ("push right, go left"), and—exploiting radial symmetry—decouples crawling direction from body orientation with no preferred leading arm, a control scheme unlike any bilaterally symmetric animal. Anatomically, Kuuspalu et al. (2022, Current Biology) described multiple inter-arm nerve pathways—an interbrachial commissure linking each arm to its neighbors plus a connecting ring, and crossing oral/aboral intramuscular nerve cords—offering peripheral routes for inter-arm signaling (possibly proprioceptive) that could coordinate arms without routing through the brain.

What remains genuinely unknown: whether the brain accesses a continuous body-schema/map of arm posture or only sparse task-relevant signals; the actual information carried by the inter-arm commissures; how the ≈350 million peripheral neurons implement the bend-propagation and wave-collision computations; and whether "the arm knows where it is" in any experiential sense (Godfrey-Smith's Other Minds and "Where is it like to be an octopus?" press the philosophical version). The consensus is now a hierarchical, embodied division of labor—not brain-in-charge, not arms-fully-autonomous, but a shifting delegation contingent on task.

Striking / counterintuitive:

• A severed, brain-disconnected octopus arm still produces a near-normal reaching movement when stimulated—the reach 'program' lives in the arm, not the brain (Sumbre et al. 2001).
• To fetch food, the soft arm temporarily builds a jointed, elbow-like structure with three dynamic joints, and the middle 'elbow' forms exactly where two muscle-activation waves collide (Sumbre et al. 2005/2006).
• Octopuses and humans converged on the same joint-level, quasi-articulated control strategy for point-to-point reaching despite ≈500 million years of separate evolution.
• Octopus crawling has no rhythm and no gait: Fourier analysis finds no periodicity, and the animal can crawl in any direction independent of which way its body faces, with no preferred 'lead' arm ('push right, go left').
• The long-standing textbook claim that octopuses lack proprioception was overturned in 2020—the central brain does read arm-position information, just not in the way vertebrates do.
• The catchy 'octopus has nine brains' is now considered misleading; the researcher who tested it reframes it as 'one brain and eight very clever arms.'

Open questions:

• Does the octopus central brain maintain a continuous body-schema/map of arm posture, or does it only access sparse, task-relevant peripheral signals on demand?
• What information actually travels through the interbrachial commissure and crossing intramuscular nerve cords—proprioceptive, motor, both, or something else?
• How do the ≈350 million peripheral arm neurons physically implement the bend-propagation and counter-propagating-wave computations at the circuit level?
• Where is the true boundary of delegation—which movements are fully peripheral, which require central command, and how does the split shift with task difficulty or learning?
• Is there any experiential or 'felt' dimension to arm-local sensing (the Godfrey-Smith 'where is it like to be an octopus' question), or is it purely reflexive computation?
• How much do findings from Octopus vulgaris generalize across cephalopod species (e.g., O. bimaculoides, cuttlefish, squid) with different ecologies and arm morphologies?

Key researchers/labs: Binyamin Hochner (Hebrew University of Jerusalem — Octopus Research Group), Tamar Flash (Weizmann Institute — motor control / computational), Germán Sumbre (ENS Paris; formerly Hochner lab), Tamar Gutnick (OIST / Hebrew University), Michael J. Kuba (OIST / Hebrew University), Graziano Fiorito (Stazione Zoologica Anton Dohrn, Naples), Letizia Zullo (Italian Institute of Technology), Guy Levy (Hebrew University), Melina Hale & Adam Kuuspalu (University of Chicago — arm neuroanatomy), Peter Godfrey-Smith (philosopher of mind, University of Sydney).

Key papers #

• Sumbre G, Gutfreund Y, Fiorito G, Flash T, Hochner B (2001). Control of Octopus Arm Extension by a Peripheral Motor Program. Science — Severed arms reproduce normal reaching kinematics; the bend-propagation motor program is embedded in arm circuitry, not the brain.
• Sumbre G, Fiorito G, Flash T, Hochner B (2005). Neurobiology: Motor control of flexible octopus arms. Nature — Arm-to-mouth fetching forms a vertebrate-like quasi-articulated limb with three dynamic joints from a soft appendage.
• Sumbre G, Fiorito G, Flash T, Hochner B (2006). Octopuses Use a Human-like Strategy to Control Precise Point-to-Point Arm Movements. Current Biology — Medial joint forms at the collision of two counter-propagating muscle-activation waves (central command + sucker input); convergent with human joint-level control.
• Gutnick T, Byrne RA, Hochner B, Kuba M (2011). Octopus vulgaris Uses Visual Information to Determine the Location of Its Arm. Current Biology 21(6):460-462 — Octopuses visually guide a single arm through a transparent maze to a goal; 6/7 learned; opaque maze abolishes performance.
• Gutnick T, Zullo L, Hochner B, Kuba MJ (2020). Use of Peripheral Sensory Information for Central Nervous Control of Arm Movement by Octopus vulgaris. Current Biology 30(21):4322-4327 — Y-maze solvable only via arm-sensed (non-visual) cues; 5/6 learned—proves the CNS uses peripheral proprioceptive/tactile input, revising the 'no proprioception' view.
• Levy G, Flash T, Hochner B (2015). Arm Coordination in Octopus Crawling Involves Unique Motor Control Strategies. Current Biology 25(9):1195-1200 — Crawling is non-rhythmic (no CPG), decouples direction from body orientation, no preferred arm—'push right, go left.'
• Kuuspalu A, Cody S, Hale ME (2022). Multiple nerve cords connect the arms of octopuses, providing alternative paths for inter-arm signaling. Current Biology 32(24):5415-5421 — Interbrachial commissure plus crossing intramuscular nerve cords give peripheral, brain-bypassing routes for inter-arm (likely proprioceptive) signaling.
• Hochner B (2012). An Embodied View of Octopus Neurobiology. Current Biology — Framework arguing octopus motor control offloads computation to body morphology and peripheral circuits—'embodied organization.'
• Godfrey-Smith P (2016). Other Minds: The Octopus, the Sea, and the Deep Origins of Consciousness. Farrar, Straus and Giroux — Popular/philosophical synthesis of distributed octopus cognition and the 'thinking with the body' question.