# Observational Learning & Cognition Controversies in Octopuses

> Part III: Intelligence in Action · Chapter 5 of 17 — The Octopus Mind
> Canonical: https://octopuscognition.org/sections/observational-learning-cognition-controversies-in-octopuses/

## In brief

The single most cited claim in cephalopod social cognition is also its most disputed. In Fiorito & Scotto (1992, Science 256:545–547), naïve Octopus vulgaris "observers" watched trained demonstrators repeatedly attack one of two balls (red vs. white) in a simultaneous visual discrimination.

The single most cited claim in cephalopod social cognition is also its most disputed. In **Fiorito & Scotto (1992, *Science* 256:545–547)**, naïve *Octopus vulgaris* "observers" watched trained demonstrators repeatedly attack one of two balls (red vs. white) in a simultaneous visual discrimination. After merely four demonstrations, observers, tested alone, chose the demonstrator's target on their first trials and thereafter, and — strikingly — reached criterion *faster* than the demonstrators had during operant conditioning (which required ~16–21 rewarded/punished trials). The authors framed this as the first demonstration of observational learning in any invertebrate, implying a shortcut to knowledge that bypassed trial-and-error. It became a cornerstone of the "octopus is smart" narrative.

The backlash was immediate and substantive. **Biederman & Davey (1993, *Science* 259:1627–1628, "Social learning in invertebrates")** argued the design could not distinguish true imitation/observational learning from simpler, non-cognitive mechanisms: **stimulus (local) enhancement**, where the demonstrator's activity merely draws attention to a location or object; **response priming**; or exploitation of a pre-existing perceptual **bias toward red**. They noted that if octopuses innately prefer or are more reactive to red, apparent "copying" of red demonstrators is trivial. Fiorito & Scotto replied (same 1993 issue) that copying was obtained for *both* red and white targets, held stable across five days, and that a color bias alone cannot explain white-copying — but the exchange never fully resolved whether attention-directing (enhancement) versus genuine associative "learning what the demonstrator learned" was at work. This distinction — imitation vs. emulation vs. stimulus enhancement vs. local enhancement — remains the central interpretive fault line.

Notably, the antagonists then collaborated. **Fiorito, Biederman, Davey & Gherardi (1998, *Animal Cognition* 1:107–112)** tested whether *preexposure* to elements of the classic "jar-opening"/discrimination context would facilitate later problem solving — a way to probe latent/contingent learning and the priming account. Octopuses **failed to benefit** from familiarity with the training context or task elements, a result that sat awkwardly with strong observational-learning claims and underscored how sensitive these effects are to procedure. This is a genuinely unusual episode in comparative cognition: critics and original authors co-authoring a partly deflationary follow-up.

Replication and extension to other cephalopods produced mixed, cautious results. **Huang & Chiao (2013, *Animal Cognition* 16:481–490, "Can cuttlefish learn by observing others?")** tested *Sepia pharaonis* in a threat–place association: only a *subset* of observers acquired the association, not a clean group effect, and the authors were careful to frame it as, at best, weak observational conditioning. More positively, **Sampaio et al. (2021, *Animal Cognition* 24:23–32)** reported that *Sepia officinalis* hatchlings (neurally immature, ≤5 days old) inhibited predatory strikes after watching demonstrators fail — with *more* observers than demonstrators reaching criterion — interpreting it as **emulation/affordance learning** rather than imitation. Even here, the mechanism is framed conservatively.

The deeper puzzle is theoretical. Social learning is generally expected to evolve under social living, yet octopuses are famously **asocial, short-lived, semelparous, with no parental care and embryos dispersing after hatching** (reviewed in **Schnell, Amodio, Boeckle & Clayton, 2021, *Biological Reviews* 96:162–178**). If octopuses genuinely learn socially, either the trait is a by-product of general associative machinery repurposed in the lab, or our assumptions about the social-intelligence link need revision. Schnell et al. and others stress that behavioral flexibility is routinely over-read as "cognition."

This feeds a broader methodological reckoning. **Amodio et al. (2019, "Octopus intelligence: the importance of being agnostic," *Animal Sentience*; and related commentary on Mather)** argue the field should adopt an explicitly **agnostic, mechanism-first** stance: small sample sizes (often n < 10), lack of pre-registration, weak controls for non-associative explanations (sensitization, neophilia, arousal), difficulty of blind scoring, and publication bias toward "clever octopus" stories all threaten replicability — mirroring the wider replication crisis in animal cognition. The upshot: octopus observational learning is a landmark *claim* whose strong cognitive interpretation is not securely established. What is real is remarkable behavioral plasticity; whether it constitutes *true social learning* (let alone imitation) is still, three decades on, unresolved.

**Striking / counterintuitive:**
- Observers reportedly learned FASTER than the demonstrators who had undergone full operant conditioning — a striking, much-quoted claim from the 1992 paper.
- The original author (Fiorito) and his sharpest critics (Biederman & Davey) later co-authored a 1998 follow-up that failed to find a preexposure benefit — a rare adversarial collaboration in comparative cognition.
- Octopuses are asocial, semelparous, and provide no parental care, so a genuine social-learning capacity would be evolutionarily paradoxical.
- Even neurally immature cuttlefish hatchlings (≤5 days) appear to socially modulate predatory behavior, with more observers than demonstrators reaching criterion.

**Open questions:**
- Can the 1992 observational-learning result be replicated with modern controls (pre-registration, blind scoring, adequate n) that rule out stimulus enhancement and arousal?
- Is any cephalopod 'social learning' genuine imitation, or is it always reducible to emulation, local/stimulus enhancement, or general associative priming?
- Why would robust social-learning machinery evolve in solitary, short-lived octopuses — is it a by-product of general-purpose learning rather than an adaptation?
- How much of the octopus-cognition literature would survive replication given small samples and publication bias toward 'clever' results?
- Do octopuses show any latent/contingent learning, or does the failure of preexposure paradigms (Fiorito et al. 1998) indicate a real limit?

*Key researchers/labs: Graziano Fiorito (Stazione Zoologica Anton Dohrn, Naples), Pietro Scotto, Gerald B. Biederman & Vaughan A. Davey (critics), Chuan-Chin Chiao (National Tsing Hua University, cuttlefish), Piero Amodio (agnostic/comparative-cognition critique), Alexandra K. Schnell & Nicola S. Clayton (University of Cambridge), Jennifer Mather (University of Lethbridge), Eduardo Sampaio & Rui Rosa (University of Lisbon).*

### Key papers
- **Graziano Fiorito & Pietro Scotto (1992).** *Observational Learning in Octopus vulgaris.* Science 256(5056):545–547 — Founding (and most contested) claim: naïve octopuses copied trained demonstrators' color choice after ~4 demonstrations, faster than operant learning.
- **Gerald B. Biederman & Vaughan A. Davey (1993).** *Social Learning in Invertebrates.* Science 259(5101):1627–1628 — Core critique: results explainable by stimulus/local enhancement, response priming, or an innate red bias rather than true imitation.
- **Graziano Fiorito, Gerald B. Biederman, Vaughan A. Davey & Francesca Gherardi (1998).** *The role of stimulus preexposure in problem solving by Octopus vulgaris.* Animal Cognition 1(2):107–112 — Critics and original author co-author a partly deflationary follow-up: octopuses did NOT benefit from context/task preexposure, complicating strong learning claims.
- **Kuan-Ling Huang & Chuan-Chin Chiao (2013).** *Can cuttlefish learn by observing others?.* Animal Cognition 16(3):481–490 — Extension to Sepia pharaonis: only a subset of observers formed the threat-place association — weak, partial evidence for observational conditioning.
- **Alexandra K. Schnell, Piero Amodio, Markus Boeckle & Nicola S. Clayton (2021).** *How intelligent is a cephalopod? Lessons from comparative cognition.* Biological Reviews 96(1):162–178 — Frames the asocial paradox: social learning is theoretically unexpected in solitary, short-lived, non-parental octopuses; cautions against over-reading flexibility as cognition.
- **Eduardo Sampaio, Catarina S. Ramos, Bruna L. M. Bernardino, et al. (Rui Rosa) (2021).** *Neurally underdeveloped cuttlefish newborns exhibit social learning.* Animal Cognition 24(1):23–32 — Sepia officinalis hatchlings inhibited predatory strikes after watching demonstrators; interpreted as emulation/affordance learning, not imitation.
- **Piero Amodio et al. (2019).** *Octopus intelligence: the importance of being agnostic (commentary on Mather).* Animal Sentience — Calls for a mechanism-first, agnostic stance; behavioral flexibility is not proof of complex cognition, and evidence is under-controlled.

## Resolved source links

- [Observational Learning in Octopus vulgaris.](https://doi.org/10.1126/science.256.5056.545) — DOI 10.1126/science.256.5056.545
- [Social Learning in Invertebrates.](https://doi.org/10.1126/science.259.5101.1627) — DOI 10.1126/science.259.5101.1627
- [The role of stimulus preexposure in problem solving by Octopus vulgaris.](https://doi.org/10.1007/s100710050015) — DOI 10.1007/s100710050015
- [Can cuttlefish learn by observing others?.](https://doi.org/10.1007/s10071-012-0573-z) — DOI 10.1007/s10071-012-0573-z
- [How intelligent is a cephalopod? Lessons from comparative cognition.](https://doi.org/10.1111/brv.12651) — DOI 10.1111/brv.12651
- [Neurally underdeveloped cuttlefish newborns exhibit social learning.](https://doi.org/10.1007/s10071-020-01411-1) — DOI 10.1007/s10071-020-01411-1
- [Octopus intelligence: the importance of being agnostic (commentary on Mather).](https://search.crossref.org/?q=Octopus%20intelligence%3A%20the%20importance%20of%20being%20agnostic%20(commentary%20on%20Mather).)

## Related trails

- [Intelligence in Action](https://octopuscognition.org/trails/intelligence-in-action/index.md): Which celebrated octopus abilities survive close methodological scrutiny?