Albert Newen and Carlos Montemayor (2025), Philosophical Transactions of the Royal Society B 380(1939):20240314, theme issue “Evolutionary functions of consciousness” — DOI-10.1098/rstb.2024.0314
Key takeaways
- Two levels of experience, three functions. Basic arousal (level 1) is signal-driven, uses short-term and procedural memory, and has one function, alarming the bodily agent: (i) triggering survival behaviour, (ii) caring for the body, (iii) initiating generalised, even one-shot, learning. General alertness (level 2) adds selective, goal-directed attention and the interaction of short- and long-term memory, and its function is advanced learning and decision-making. Reflexive self-consciousness is the same level-2 experience with metacognitive, self-relational content, and its function is future-directed long-term planning that takes one’s own or another’s mindset into account (Table 1).
- Basic arousal does not need cortex. The evidence offered is all cited: hydranencephalic children (Merker), cats decorticated in infancy that still orient and solve a T-maze discrimination, and macaques under anaesthesia that wake to central lateral thalamus stimulation and fall back when it stops, responding to pain but unable to do tasks that need focused attention (Redinbaugh et al. 2020). More complex consciousness needs cortex, modulated by the thalamus. Merker’s stronger claim, that consciousness never depends on cortex, is rejected.
- The leading theories cover only general alertness. IIT (posterior cortex), GNWT (prefrontal cortex) and Lau’s higher-order theory (prefrontal and parietal monitoring) all make neocortex essential. None constrains itself by evolutionary function, and so none accounts for basic arousal. The authors claim to stay “neutral” on which of the three wins.
Standing in the record
Filed on 2026-10-03 at the owner’s request, with Maldarelli & Güntürkün (LIT-650) from the same theme issue, for the record’s holdings on the evolution and function of consciousness. No anthology topic holds it, and it carries no instruction for machine-learning practice.
It rivals Veit (LIT-105) on the original function of consciousness. Both are evolutionary, affect-first accounts: the first consciousness was felt bodily value, and it was an adaptation. They exclude each other on what that first consciousness did. For Veit’s pathological complexity thesis, as read in NOTE-187, hedonic valence originated as a common currency for trade-offs in action selection among animals with many degrees of freedom; motivational trade-offs are his criterion for attributing consciousness. For ALARM, basic arousal gives the alarm signal “100% weight” and runs “rather rigid behavioural programmes without choices”; choice first appears with general alertness (“The various functions of phenomenal consciousness”). An affect whose first job is to arbitrate between options, and one whose first job is to remove the option, cannot both be the original function. The two also split on learning: ALARM accepts that basic arousal brings a new quality of learning, close to Ginsburg and Jablonka’s unlimited associative learning (fn. 1), while Veit holds that unlimited associative learning explains later features of consciousness, not its origin. Neither cites the other.
The rest of the record it touches:
- Damasio & Carvalho (LIT-388) and THEORY-045. The closest support. They place the first integrated body maps, and feeling, in the upper brainstem, not the insula, as basic arousal requires. ALARM’s alarm is a special case of THEORY-045’s claim that felt affect is the experience of the body’s regulatory state, valenced by what it means for staying alive: the case where a variable leaves its tolerance range. ALARM sides against prediction-only accounts in one sentence: pain alarms “in ways that mere predictions concerning statistical probabilities of likelihood cannot capture” (“Two framework observations”, a). That bears on THEORY-045’s open question, prediction (Barrett, LIT-542) versus map (Damasio), but it is asserted, not argued. Damasio & Damasio (LIT-387, unread) state the same affect-first origin.
- Maldarelli & Güntürkün (LIT-650), in the same issue, argue that GNWT’s requirements are met by the bird’s nuclear, non-cortical pallium. That contradicts ALARM’s reading of GNWT as needing prefrontal cortex: GNWT may need a hub-and-module connectome rather than neocortex. The two papers also differ on the mirror test, which ALARM takes as a test of reflexive self-consciousness that magpies pass, and which Maldarelli and Güntürkün doubt, noting a failed magpie replication.
- IIT 3.0 (LIT-445) and Barrett et al. (LIT-185). ALARM’s objection is that IIT’s axioms are introspective and unconstrained by evolution or cost, and that modular minds sit badly with a measure of integration. This is close in spirit to Veit’s complaint, in NOTE-187, that IIT rests on unrevisable axioms. It is weaker than Barrett et al.’s, that Φ is ill-defined for real systems.
- Gallagher (LIT-453) splits the minimal self into a sense of ownership and a sense of agency. ALARM uses those two, adds a sense of perspectivity, and grades each from a pre-reflexive form (with basic arousal) to a reflexive one (with general alertness and metacognition). THEORY-064, from Metzinger (LIT-568), says that what is experienced as a self is the content of a self-model the system cannot recognise as a model. ALARM’s reflexive self-consciousness is a self-model the system does represent as its own, built on the pattern theory of self (fn. 4). The two are not rivals: they describe different levels.
- Chalmers (LIT-406). ALARM sets epiphenomenalism aside in its Introduction and argues only about what consciousness does. It is an answer to easy problems in Chalmers’s sense, the functions of consciousness, by design.
self is tagged for the section on reflexive self-consciousness, a third of the paper, on reflexive and pre-reflexive
self-consciousness. emotion-and-affect is tagged for basic arousal, which
the paper identifies with Denton’s primordial emotions.