Contribution
It extends the authors’ 2023 two-level ALARM theory with a third phenomenon, reflexive self-consciousness, and argues that it is not a third kind of experience but general alertness carrying metacognitive, self-relational content. It sets out a function for each of the three (Table 1), and it argues that IIT, higher-order theories and GNWT all leave out the evolutionarily oldest form.
Key insight
Ask what consciousness did when it first evolved, and the answer is not integration, broadcast or monitoring but alarm: a felt signal that takes over the whole organism when a homeostatic variable leaves its range, because a reflex alone stops at withdrawal while a pain that persists keeps the injury protected and teaches avoidance in one trial. Later forms are layered on that, and they work only while the alarm is not overriding them.
Assumptions
- Phenomenal consciousness in wakefulness is the target; vegetative states and Block’s access/phenomenal distinction are set aside (Introduction).
- Phenomenality is independent of content: experience is explained by how a content is processed, not by what it is (Introduction; Vosgerau, Schlicht & Newen 2008).
- Consciousness is an adaptation, or part of one; epiphenomenalism is set aside without argument here (Introduction).
- Different realisers across taxa: birds and fish may be conscious with different neural mechanisms, so no single NCC cluster is expected (Introduction).
- Opt-out paradigms indicate metacognition (fn. 2).
- The pattern theory of self: no self as substance (fn. 4).
Key results
These are positions argued, with cited evidence.
- Basic arousal is subcortical (“Two framework observations”, a). Evidence cited: Merker’s hydranencephalic children, who seem to have affective experiences of comfort and distress; Bjursten et al.’s cats decorticated from infancy, which orient by vision and touch and solve a T-maze visual discrimination; Redinbaugh et al. (2020), where central lateral thalamus stimulation in anaesthetised macaques works as an on–off switch to a state with pain and perceptual responses but no capacity for attention-demanding tasks. NCCs for basic arousal lie in the upper brainstem and thalamus. Merker’s claim that this holds for all consciousness is rejected.
- Four advantages of felt over unfelt alarm (same section): pain keeps the hand away; it prompts care that supports healing; it generalises to new situations; it allows one-shot learning, which unconscious associative learning does not.
- General alertness (“Two framework observations”, b): attention-guided consciousness, the ability to select and switch between sensory signals; everyday human consciousness, and “clearly proven” in mice (citing Wimmer et al. 2015).
- Reflexive self-consciousness can be tested without language (“Reflexive self-consciousness”, a): mirror self-recognition at 18 months and in apes, elephants, dolphins and magpies; and opt-out metamemory tasks (Hampton 2001). The text describes the subjects of the latter as both “chimpanzees” and “monkeys”; Hampton’s 2001 paper, by its title, used rhesus monkeys.
- Reflexive self-consciousness is general alertness plus metacognitive self-relational content (“Reflexive self-consciousness”, b). Self-relational content comes with every sense of ownership, agency and perspectivity, and each develops from basic to complex forms. With basic arousal it is pre-reflexive self-consciousness (four- to nine-month-olds treat a live image of themselves differently from a mimicking other); with general alertness and metacognition it is reflexive (18-month mirror self-recognition, the 4-year explicit false-belief task). No third kind of experience is posited, on three grounds: parsimony, no phenomenological or neural evidence for one, and the neural overlap between attributing beliefs to oneself and to others.
- Critique of three theories (“The underestimation of the plurality of functions”). All three make neocortex essential (IIT posterior, GNWT prefrontal, Lau’s HOT prefrontal and parietal; citing the adversarial collaboration of Ferrante et al. 2025). IIT: on an a priori identity reading, integration is not a function of consciousness but what it is; its axioms are introspective and evolution plays no part in them; Φ is not constrained by cost; modular minds are a challenge. HOT: Rosenthal grants mental-state consciousness little or no function; Lau’s perceptual reality monitoring either needs no experience (blindsight, dual visual streams) or amounts to general alertness. GNWT: a theory of access, giving a synchronic function (Dehaene’s keep, compress, transfer) that the authors accept for general alertness, but with no phylogenetic account.
- Functions (“The various functions of phenomenal consciousness”; Table 1). Basic arousal gives the extreme signal “100% weight” over slow homeostatic updating, with rigid programmes and no choice. General alertness organises costs by choice: holding a hot plate rather than dropping it at a greater cost. Reflexive self-consciousness weighs one’s reputation: holding the plate to be seen bearing the pain. A “managerial analogy” orders them: each later manager works only while the older one does not override it.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | A basic form of phenomenal consciousness can be realised without cortex, by upper brainstem and thalamus | moderate: three cited lines of evidence, none new here; whether decorticate behaviour or thalamically aroused macaques are conscious is the point at issue | hydranencephaly, decorticate cats, Redinbaugh et al. 2020 |
| C2 | The first function of consciousness is alarm: survival behaviour, bodily care and generalised learning, with no choice | weak to moderate: an evolutionary argument from the advantages of felt pain over reflex; no comparative or phylogenetic data | ”Two framework observations”, a; Table 1 |
| C3 | Reflexive self-consciousness is general alertness with metacognitive self-relational content, not a new kind of experience | weak: parsimony plus an absence of contrary evidence plus neural overlap of self- and other-belief attribution | ”Reflexive self-consciousness”, b |
| C4 | IIT, HOT and GNWT each make neocortex essential and omit basic arousal | moderate for the descriptive part (cited); the inference that they are therefore incomplete depends on C1 | ”The underestimation of the plurality of functions” |
| C5 | Each of the three phenomena has a distinct dominant function, and later ones operate only when not overridden by earlier ones | weak: argued by examples and analogy | Table 1, Conclusion |
Concepts
- basic arousal: level-1 phenomenal consciousness; sensitivity to pain, pleasure and other basic affect without focused attention; identified with Denton’s primordial emotions; realisable subcortically.
- general alertness: level-2 phenomenal consciousness; attention-guided, selective and goal-directed; cortical with thalamic modulation.
- reflexive (self-)consciousness: general alertness whose content is meta-representation of an agent’s mindset, one’s own or another’s.
- mindset: the cluster of a person’s mental states and processes (emotions, desires, beliefs).
- self-relational content: information about the agent produced alongside information about the world, in the senses of ownership, agency and perspectivity.
- pre-reflexive self-consciousness: self-relational content experienced with basic arousal only; self–other differentiation without an explicit self-representation (fn. 3).
Connections
- Veit (LIT-105, NOTE-187). The rival on the original function of consciousness, declared on the LIT. Both are affect-first and adaptationist. Veit’s first consciousness is valence as a currency for trade-offs; ALARM’s is an alarm that removes the trade-off.
- Damasio & Carvalho (LIT-388). Brainstem origin of feeling: the neuroanatomy basic arousal needs. Not cited (the paper cites Damasio 1999).
- Barrett (LIT-542) and THEORY-045. ALARM’s one sentence against prediction-only accounts of pain bears on THEORY-045’s open question, but only as an assertion.
- Maldarelli & Güntürkün (LIT-650). Same theme issue. Their bird evidence fits ALARM’s methodological commitment to different realisers, but it contradicts ALARM’s reading of GNWT as needing prefrontal cortex.
- Gallagher (LIT-453) for ownership and agency; Metzinger (LIT-568) and THEORY-064 for a self-model account at a different level.
- IIT 3.0 (LIT-445), Barrett et al. (LIT-185). The record’s IIT holdings, against which ALARM’s critique is an outside, functional one.
Bearing on the record
- Candidate co-source, with Damasio & Carvalho (LIT-388), for a THEORY that a basic, affective form of consciousness is realised subcortically and precedes cortical consciousness. It should not be the only source, since its evidence is cited rather than new.
- It would also enter a THEORY on the first function of consciousness, set against Veit’s.
- No ML instruction. Nothing here belongs in the anthology.
Limitations
- No new data; the subcortical case rests on three cited findings.
- What separates basic arousal from unconscious alarm is argued from felt pain’s advantages, which presuppose that it is felt. No marker is offered that would tell a conscious alarm from an unconscious one in an animal.
- The critique treats GNWT as essentially prefrontal-cortical; a GNWT reading as connectome architecture (as in Maldarelli & Güntürkün) escapes it.
- The metamemory example is described inconsistently (“chimpanzees” and “monkeys”).
Open questions
- What finding would separate basic arousal from unconscious homeostatic alarm in a non-verbal animal?
- Does ALARM predict that one-shot avoidance learning is absent where basic arousal is absent, and is that testable in decorticate or thalamically silenced animals?
- If GNWT can be met without cortex, does ALARM’s objection to it survive?