NOTE-488: Conscious birds

Contribution

A compact review from the Bochum avian-cognition group. It sets the two avian electrophysiology studies of sensory consciousness, the avian connectome and the new cortex-like-circuit anatomy beside the requirements of three theories of consciousness, and it argues that bird mirror studies show the mark test’s limits rather than birds’ lack of self-awareness. It adds no data.

Key insight

The question “could a bird be conscious under theory T?” splits into an anatomical requirement and a physiological one, and the avian brain answers them separately. The anatomy that matters for GNWT is the connectome’s shape, not cortical layering, and birds have it in a nuclear pallium. The physiology GNWT predicts, report-tracking activity in the prefrontal analogue, has been recorded. IIT’s markers have simply not been looked for.

Assumptions

• Consciousness is not uniquely human, and theories of it must account for both its neural basis and its behavioural consequences, “because evolution can only operate by selecting an animal’s behavioural output” (§1).
• NCL is the functional analogue of mammalian prefrontal cortex (§2, citing Güntürkün 2005; Herold et al. 2011).
• Report tracks experience: a neuron that follows the animal’s “seen” report rather than the stimulus is a correlate of subjective experience (§2).
• PCI is “a good substitute” for Φ (§2).
• Mirror self-recognition and its ecological variants measure self-awareness, graded rather than all-or-nothing (§4).

Key results

• Bistable perception (§2). Pigeons shown apparent-motion stimuli report alternating horizontal and vertical motion (Vetter et al. 2000).
• Crow NCL (§2). In a delayed detection task at threshold, NCL single-unit activity followed the “seen” report regardless of stimulus presence, and was silent for “not seen” (Nieder et al. 2020). Separate NCL populations coded reported presence and reported absence when absence was trained as a reportable category (Wagener & Nieder 2024). The authors say the data “could be explained by any of these theories”. The absence-coding result could “expand or challenge” GNWT, since absence too produced ignition-like activity.
• Avian anatomy (§3). 3D polarised-light imaging and tracing show layers, columns and repeated canonical circuits in avian sensory pallium (Stacho et al. 2020). NCL lacks them and is nuclear. The pigeon telencephalic connectome has hubs (NCL, arcopallium dorsale and intermedium, area parahippocampalis, area corticoidea dorsolateralis) that are densely interlinked and tied to local modules (Fig. 1B, from Shanahan et al. 2013). Recurrent connectivity is found throughout the visual pallium and NCL.
• Theory scorecard (§3 summary; §5). GNWT: anatomical requirements “largely met”, with NCL activation correlating with awareness. RPT: anatomical requirements met, physiological ones unproven without switching recurrence off. IIT: requirements “not met (yet)”; no PCI measured and no grid-like lattice found, although sensory pallium’s cortex-like modularity is the precondition for one.
• Self-awareness (§4). Magpies (Prior et al. 2008, but a 2020 replication failed) and Indian house crows pass the mark test; many corvids fail it, though crows and ravens show self-directed behaviour and contingency checking. Pigeons discriminate live from delayed self-video (1–7 s delays) and show less stress-related behaviour to their mirror image than to a food competitor. Roosters give alarm calls to a visible conspecific but not to their mirror image, even with a hidden conspecific behind the mirror (Hillemacher et al. 2023; Fig. 2).

Claims

id	claim	strength	support
C1	An avian brain region tracks reported perception rather than stimulus, a neural correlate of sensory consciousness	moderate: two single-unit studies from one lab, in crows; report-based, so access and phenomenal consciousness are not separated	§2, refs. 17, 19
C2	Conscious processing does not need laminar cortex, since the avian prefrontal analogue is nuclear	moderate as an inference, conditional on C1 being a correlate of consciousness rather than of report	§3
C3	Birds meet GNWT’s anatomical requirements	moderate: a hub-and-module connectome is shown (Shanahan et al. 2013), but GNWT’s “ignition” has not been shown in birds, nor the contrast between conscious and unconscious trials beyond NCL	§3, Fig. 1
C4	Birds meet RPT’s anatomical requirements; its physiological ones are unproven	moderate for anatomy; the authors say recurrence was never switched off experimentally	§3
C5	IIT’s requirements are not (yet) met in birds	weak as stated: “not met” means “not measured”	§3, §5
C6	The mark test lacks ecological validity and gives false negatives in birds	moderate: pigeons and roosters discriminate self-image from conspecific under ecologically relevant conditions	§4, refs. 18, 83, 85
C7	Widespread consciousness across distant taxa makes it likely that consciousness confers evolutionarily successful abilities	weak: an inference stated in one sentence in the Conclusion	§5

Concepts

• NCL (nidopallium caudolaterale): the associative avian pallial region treated as the functional analogue of prefrontal cortex; nuclear in organisation.
• sensory awareness vs self-awareness: the paper’s two headings for evidence of consciousness in birds.
• “seeing” vs “knowing” (RPT): recurrence confined to sensory areas gives a basic, non-reportable experience; recurrence extending to prefrontal areas gives reportable experience with working-memory access.
• mirror-audience test: a rooster’s alarm calling toward a mirror image versus a conspecific, used as an ecological self-recognition test.

Connections

• Butlin et al. (LIT-056). The same theories, RPT and global workspace, turned into indicators for AI there and into requirements for a bird here.
• IIT 3.0 (LIT-445) and Barrett et al. (LIT-185). Barrett et al. argue that PCI is a proxy that captures at most two of IIT’s axioms. That weakens the paper’s plan to test IIT in birds by measuring PCI.
• Newen & Montemayor (LIT-626). Same theme issue. They hold that GNWT needs prefrontal cortex; this paper argues that a nuclear prefrontal analogue suffices. They cite the magpie mark-test result as a pass, which this paper treats as contested.
• Veit (LIT-105). Both are gradualist and both infer that consciousness is an adaptation from its spread across taxa.
• THEORY-023. Evidence of the comparative kind its promote_when names. The NCL result counts against cortical lamination as the biological property that matters.

Bearing on the record

• Candidate source for a THEORY that conscious report in vertebrates does not require laminar cortex. It would need the primary crow studies read first, and should state that report-based paradigms do not separate access from phenomenal consciousness.
• self is justified by §4, which is a third of the paper and is about self-recognition.
• No ML instruction. Nothing here belongs in the anthology.

Limitations

• A review: the evidence is second-hand, and one lab’s crow studies carry the sensory-awareness case.
• Report-based: it cannot separate phenomenal from access consciousness, a limit the paper does not discuss.
• IIT is judged unmet only because it is untested.
• Binocular rivalry and flash suppression, milestone paradigms in humans, have not been run in birds, as the authors note.

Open questions

• Is activity in primary and associative visual areas the same on seen and unseen trials, with NCL added only on seen ones? That is the paper’s statement of what GNWT predicts (§3), and it is untested.
• What does avian PCI look like across wake, NREM sleep and anaesthesia?
• Does the avian sensory pallium have a grid-like lattice of cell-type columns?