Pl. XVII

Part III · Intelligence in Action · Chapter 17

Numerical, Quantity, and Abstract-Concept Cognition in Cephalopods (with Cross-Modal and Mirror/Self Tests)

Cephalopod "higher-order" cognition splits into three uneven strands: solid evidence for approximate number/quantity representation and cognitive control (mostly in cuttlefish), suggestive but thin evidence for cross-modal integration and abstract concepts, and largely negative results for mirror self-recognition.

Numerical and quantity cognition. The strongest data come from cuttlefish. Yang & Chiao (2016, Proc. R. Soc. B) tested 54 juvenile Sepia pharaonis on prey-choice tasks and found reliable discrimination of 1v2, 2v3, 3v4, 4v5, and 1v5 shrimp—i.e., successful discrimination above a ratio of ≈1.25 (5/4). Reaction latency rose as numerosities converged, implying an analog-magnitude ("approximate number system") mechanism obeying Weber's law rather than exact counting or subitizing. Huang et al. (2019, Animal Cognition) extended this to "fractional" quantities, showing S. pharaonis discriminated 1v1.5, 1.5v2, and 2v2.5 but failed 2.5v3—again a ratio-dependent, Weber-compliant pattern hinting at a primitive proportion sense. Crucially, the same 2016 study demonstrated state-dependent valuation: hungry cuttlefish preferred one large shrimp, while satiated ones preferred two small shrimp, and prey number choice depended on prey quality (live vs. dead, large vs. small). This ties numerical assessment to internal state and value integration—an economically rational, not merely perceptual, computation. Dedicated octopus numerical work remains comparatively sparse; quantity-related competence is usually inferred from foraging optimization rather than controlled numerosity tasks, making the cuttlefish literature the field's numerical backbone.

Cognitive control / delay of gratification. Schnell, Boeckle, Rivera, Clayton & Hanlon (2021, Proc. R. Soc. B) adapted a Stanford-marshmallow-style delay-maintenance paradigm to Sepia officinalis. Cuttlefish forwent an immediately available, less-preferred prawn to wait up to 50–130 s for a preferred live grass shrimp, and—critically—individuals that tolerated longer delays also learned faster in a separate reversal-learning task. This is the first reported self-control/learning-performance link outside primates and some corvids, and a convergent-evolution centerpiece: a mollusc arriving at "willpower"-like inhibitory control along an independent lineage. Some observed cuttlefish rotated away from the tempting item, resembling attention-distraction strategies seen in children and apes.

Cross-modal integration. Octopuses maintain partly dissociable visual and chemotactile learning systems, yet Kawashima & Ikeda (2025, Zoological Science, 42:260–269) reported cross-modal object recognition in Callistoctopus aspilosomatis: animals that first learned an object by touch could then recognize it by vision alone. Notably, recognition was tactile-dominant—octopuses relied more on tactile representations, and vision-only encoding was less reliable—consistent with their "taste-by-touch" ecology mediated by cephalopod-specific chemotactile receptors (van Giesen et al.; Allard et al.). This addresses a long-standing question of whether the semi-autonomous arm nervous system and the central brain share unified object representations.

Abstract relational concepts. Evidence here is thin and contested. Octopus vulgaris performs conditional discrimination—modulating a response by context (e.g., attacking only when a contextual cue like bubbles is present)—and shows stimulus generalization, orientation/mirror-image discrimination, and reversal learning (reviewed by Schnell et al. 2021, Biological Reviews; Ponte et al. 2022). But rigorous same/different or identity-concept transfer tests of the kind passed by pigeons, corvids, and archerfish have not been convincingly demonstrated in cephalopods, so genuine abstract-relational cognition remains an open, actively debated question.

Mirror self-recognition / self-awareness. Results are largely negative or ambiguous. Maselli, Al-Soudy, Buglione et al. (2022, Frontiers in Physiology) ran a preliminary mark test on O. vulgaris: octopuses did not show interest in their reflection, and arm-directed exploration of a nail-polish mark occurred equally in sham-marked animals and without a mirror—implicating proprioception, not visual self-recognition. Comparative observations suggest squid react strongly and cuttlefish moderately to reflections (agonistic/exploratory), whereas octopuses react little. No cephalopod has passed a mark-test MSR, tempering strong self-awareness claims even as broader consciousness arguments (Godfrey-Smith; Birch et al.) continue.

Striking / counterintuitive:

• Cuttlefish reverse their numerical preference based on hunger: hungry animals pick one large prey, satiated ones pick two small—number choice is value-driven, not fixed.
• Longer delay-of-gratification tolerance predicts faster learning in cuttlefish, mirroring the human 'marshmallow test' correlation in a mollusc with no shared ancestry for such control.
• Some cuttlefish physically turn away from tempting food, resembling self-distraction strategies documented in children and apes.
• Octopus object recognition is tactile-DOMINANT: they trust touch over vision when forming representations of novel objects, the reverse of the human default.
• The 'self-directed' mark-touching octopuses do in mirror tests happens just as often with no mirror and in sham-marked animals—it's proprioception, not self-recognition.
• Numerical discrimination follows Weber's law (ratio-dependent), meaning cuttlefish use an analog approximate-number system rather than exact counting or subitizing.

Open questions:

• Do any cephalopods possess genuine abstract relational concepts (same/different, identity) that transfer to novel stimuli, as pigeons and archerfish do? No convincing demonstration exists yet.
• Is octopus numerical competence comparable to cuttlefish, or is the near-total reliance on cuttlefish data an artifact of species testability?
• Does the semi-autonomous arm nervous system share fully unified cross-modal representations with the central brain, or only partially?
• Why do squid, cuttlefish, and octopus differ so sharply in mirror-reflection responsiveness, and does any paradigm could reveal non-visual self-representation in octopuses?
• Can Weber-law quantity discrimination in cephalopods be dissociated from continuous non-numerical cues (surface area, density, movement)?
• How do these convergent 'cognitive control' abilities map onto cephalopod neuroanatomy (vertical lobe) relative to the mammalian prefrontal cortex?

Key researchers/labs: Chuan-Chin Chiao (National Tsing Hua University) — cuttlefish number sense, Alexandra K. Schnell (Cambridge / MBL) — cuttlefish self-control & comparative cognition, Nicola S. Clayton (University of Cambridge) — comparative cognition, delay of gratification, Roger T. Hanlon (Marine Biological Laboratory) — cephalopod behavior, Yuzuru Ikeda & Sumire Kawashima (University of the Ryukyus) — octopus cross-modal recognition, Valeria Maselli / Anna Di Cosmo lab (University of Naples Federico II) — octopus mirror/self tests, Piero Amodio & Graziano Fiorito (Stazione Zoologica Anton Dohrn) — cephalopod cognition & learning, Peter Godfrey-Smith — philosophy of cephalopod minds/consciousness.

Key papers #

• Yang, T.-I. & Chiao, C.-C. (2016). Number sense and state-dependent valuation in cuttlefish. Proceedings of the Royal Society B, 283(1837):20161379 — 54 juvenile Sepia pharaonis discriminated numerosities (1v2 up to 4v5, ratio >1.25) with Weber-law latency effects, and reversed prey-number preference by hunger state—linking analog number sense to economic valuation.
• Schnell, A.K., Boeckle, M., Rivera, M., Clayton, N.S. & Hanlon, R.T. (2021). Cuttlefish exert self-control in a delay of gratification task. Proceedings of the Royal Society B, 288:20203161 — Sepia officinalis waited up to 50–130 s for a preferred reward, and longer waiting predicted better learning—first self-control/intelligence link outside primates/corvids.
• Huang, Y.-H., Lin, H.-J., Lin, L.-Y. & Chiao, C.-C. (2019). Do cuttlefish have fraction number sense?. Animal Cognition — Sepia pharaonis discriminated fractional quantities (1v1.5, 1.5v2, 2v2.5) but failed 2.5v3, evidencing a Weber-law-governed primitive proportion sense.
• Kawashima, S. & Ikeda, Y. (2025). Cross-Modal Object Recognition and Reliability Between Visual and Tactile Senses in Octopus (Callistoctopus aspilosomatis). Zoological Science, 42(3):260–269 — Octopuses transferred object knowledge learned by touch to vision, with tactile information dominant—demonstrating unified cross-modal representations.
• Maselli, V., Al-Soudy, A.-S., Buglione, M., et al. (2022). A preliminary attempt to investigate mirror self-recognition in Octopus vulgaris. Frontiers in Physiology, 13:951808 — Octopuses failed the mark test; mark-directed arm behavior occurred without a mirror and in sham controls, attributing responses to proprioception, not visual self-recognition.
• Schnell, A.K., Amodio, P., Boeckle, M. & Clayton, N.S. (2021). How intelligent is a cephalopod? Lessons from comparative cognition. Biological Reviews, 96(1):162–178 — Authoritative synthesis situating cephalopod numerical, conditional-discrimination, and self-control abilities within convergent-evolution and comparative-cognition frameworks.