Pl. IV

Part III · Intelligence in Action · Chapter 4

Problem Solving & Tool Use

Octopuses are the textbook invertebrate problem-solvers, and the evidence spans controlled manipulation tasks, wild object use, and famous escapes. The most cognitively resonant finding is defensive tool use in the veined (coconut) octopus, Amphioctopus marginatus (Finn, Tregenza & Norman, 2009, Current Biology). During more than 500 diver-hours on soft-sediment bottoms at ≈18 m off northern Sulawesi and Bali, the team observed more than 20 individuals excavating buried coconut shell halves (and discarded clam shells), cleaning them with water jets, and carrying them—often stacked one inside another—across open seafloor to assemble later into a spherical shelter. Transport used an ungainly "stilt-walking" gait, in which the octopus rigidly extends its arms around the load and tiptoes on the arm tips, a less efficient form of locomotion than normal crawling. Finn's argument for tool use rests on deferred benefit and apparent foresight: the animal incurs a present locomotor cost (and exposure) to gain a shelter usable only at a future, unspecified time, rather than solving an immediate problem.

That interpretation is genuinely debated, and the disagreement is definitional. James Wood favors a conservative definition—a tool is a detached object used to act on and change the environment or solve an immediate problem—and quipped that a carried shelter is more like a house than a tool ("My house isn't a tool for me—it's my house"). Jennifer Mather also declined to call the coconut a tool on the grounds that the octopus does not modify the shell or use it to alter other objects, yet she does credit octopuses with tool use elsewhere. Finn himself cautioned that the object use of animals "likely forms a continuum" and that associative learning cannot be fully excluded. Notably, the shell qualifies as a borderline "tool" only because it is a portable, secondarily deployed manufactured/found object—unlike a hermit crab's permanently occupied shell, which is used continuously and never set down.

Related wild behaviors sharpen the debate. Mather (1994, J. Zool. London—"'Home' choice and modification by juvenile Octopus vulgaris… specialized intelligence and tool use?") described octopuses clearing dens with siphon water jets and stacking rocks, shells, and even bottles to barricade den entrances. More strikingly, Godfrey-Smith, Scheel and colleagues (2022, PLOS ONE, "In the line of fire") documented gloomy octopuses (Octopus tetricus) at Jervis Bay throwing silt, shells and algae by holding material in the arms and jetting it with the siphon: 102 throws in 2015 footage (55 shell, 35 silt, 11 algae), with silt more common in social contexts and ≈17 throws hitting other octopuses. Throws using atypical arm positions and higher vigor were more likely to strike conspecifics, and targets sometimes ducked—suggesting some throws are directed at others. The authors were careful to call this projectile use, not tool use, since the siphon (not the arms) propels the material; it remains one of few non-human examples of possible targeted throwing.

In the laboratory, the jar-opening paradigm is the classic problem-solving assay. Octopus vulgaris readily removes a plastic plug from a transparent jar containing a live crab, with unsuccessful attempts declining over trials—consistent with trial-and-error or stimulus–response learning rather than insight (studies from the Fiorito/Naples tradition; e.g., work on preexposure in Animal Cognition). Anderson and Mather's Seattle Aquarium work popularized giant Pacific octopuses (e.g., "Billye") opening childproof medication bottles, initially taking ≈55 minutes and improving to an average of ≈5 minutes with practice—without instruction. Octopuses also unscrew jar lids, including from the inside. Richter, Hochner & Kuba (2016, PLOS ONE, "Pull or Push?") gave seven O. vulgaris an L-shaped container that had to be extracted through a tight Perspex hole across five escalating levels (transparent then opaque partitions, reversed and randomized orientations). All seven solved every level, showed faster acquisition on later levels, and used individualized strategies, evidencing behavioral flexibility rather than a single fixed routine. The famous 1992 Science claim of observational learning (Fiorito & Scotto) remains influential but has faced replication and interpretive skepticism.

Finally, escape behavior—most famously "Inky," who in 2016 slipped from a National Aquarium of New Zealand tank, crossed the floor, and descended a ≈50 m drainpipe to the sea—dramatizes octopus opportunism, though such anecdotes are natural-history observations, not controlled cognition. What remains unknown is whether any of these behaviors reflect genuine planning/mental representation versus flexible learning, and where octopus object use sits on the tool-use continuum.

Striking / counterintuitive:

• The coconut octopus adopts a slower, clumsier 'stilt-walking' gait specifically to carry shells—paying an immediate locomotor and predation cost for a shelter it can only use later, the crux of the 'foresight' argument
• Giant Pacific octopuses learn to open human childproof medication bottles (a push-and-turn task) unaided, improving from ≈55 minutes to ≈5 minutes
• Gloomy octopuses throw silt and shells with siphon jets and sometimes appear to aim at other octopuses, who occasionally duck—one of very few non-human cases of possibly targeted throwing
• Octopuses can unscrew jar lids from the inside
• 'Inky' the octopus escaped a New Zealand aquarium by crossing the floor and squeezing down a ≈50 m drainpipe to the ocean

Open questions:

• Does coconut-shell carrying reflect genuine planning/mental representation of future need, or flexible associative learning? Finn explicitly could not rule out associative learning.
• Where should the definitional line for 'tool use' be drawn—immediate problem-solving, environmental modification, or deferred deployment—and is a set-aside, re-deployed shelter a tool?
• Is the debris-throwing in O. tetricus intentionally aggressive/social signaling, or an incidental byproduct of den-clearing that sometimes hits others?
• Do laboratory puzzle solutions (jars, L-container) involve any insight, or are they entirely trial-and-error/stimulus–response?
• How robust is the 1992 observational-learning result given later replication and interpretive concerns?

Key researchers/labs: Julian K. Finn (Museums Victoria) — cephalopod behavior, coconut octopus tool use, Mark D. Norman (Museums Victoria) — octopus taxonomy and behavior, Jennifer A. Mather (University of Lethbridge) — octopus cognition, den modification, play, Roland C. Anderson (Seattle Aquarium, dec.) — octopus enrichment, personality, jar-opening, Peter Godfrey-Smith (University of Sydney) — philosophy of mind, octopus behavior, David Scheel (Alaska Pacific University) — octopus social behavior, Jervis Bay fieldwork, Binyamin Hochner & Michael J. Kuba (Hebrew University of Jerusalem) — octopus learning and neuroscience, Graziano Fiorito (Stazione Zoologica Anton Dohrn, Naples) — octopus learning, observational learning, James B. Wood — cephalopod biology, tool-use skeptic.

Key papers #

• Julian K. Finn, Tom Tregenza, Mark D. Norman (2009). Defensive tool use in a coconut-carrying octopus. Current Biology 19(23):R1069–R1070 — First reported invertebrate tool use: veined octopus carries and assembles coconut shells as portable shelter with deferred benefit
• Peter Godfrey-Smith, David Scheel, Stephanie Chancellor, Stefan Linquist, Matthew Lawrence (2022). In the line of fire: Debris throwing by wild octopuses. PLOS ONE 17(11):e0276482 — 102 documented siphon-propelled throws in Octopus tetricus, some directed at conspecifics—projectile use, not tool use
• Jonas N. Richter, Binyamin Hochner, Michael J. Kuba (2016). Pull or Push? Octopuses Solve a Puzzle Problem. PLOS ONE 11(3):e0152048 — Seven O. vulgaris solved a five-level L-shaped container extraction task using individualized, flexible strategies
• Jennifer A. Mather (1994). 'Home' choice and modification by juvenile Octopus vulgaris (Mollusca: Cephalopoda): specialized intelligence and tool use?. Journal of Zoology (London) 233:359–368 — Documented den-building: water-jetting sand and barricading entrances with rocks, shells and debris
• Graziano Fiorito, Pietro Scotto (1992). Observational Learning in Octopus vulgaris. Science 256:545–547 — Influential but contested claim that octopuses learn a discrimination task by watching a trained demonstrator
• Roland C. Anderson, Jennifer A. Mather (popularized) (2010s). Jar/childproof-bottle opening in giant Pacific octopus (Seattle Aquarium). Seattle Aquarium / popular science reporting — Octopuses learn to open screw-top jars and childproof caps, improving speed with practice (e.g., 'Billye')