Source
- Barrett (2017), LIT-542, as read in NOTE-434: “What is a brain for?”, “How does a brain perform allostasis?” and the Glossary.
- Damasio & Carvalho (2013), LIT-388, as read in NOTE-330: the abstract, Box 1, the section on insular and brainstem lesions, and the concluding remarks.
What was actually shown
Neither paper reports an experiment that tests this claim. Barrett gives a synthesis of cited work, and Damasio and Carvalho give an opinion piece argued from lesion cases. The account is the record’s reading of where two independent lines agree.
What they agree on.
- What is felt is the body’s regulatory state. Barrett holds that interoception “is at the core of the brain’s internal model”, and that “interoceptive sensations are usually experienced as lower dimensional feelings of affect” (LIT-542, claim C1 in NOTE-434). Damasio and Carvalho open with “Feelings are mental experiences of body states” (LIT-388, claim C1 in NOTE-330), and say that a feeling is what an interoceptive map is like when it is experienced.
- The valence comes from the body’s state. In Damasio and Carvalho, body states “are necessarily valenced”, and feelings portray “the advantageous or disadvantageous nature of a physiologic situation”. Exteroceptive senses borrow their valence from the body. In Barrett, every concept “represent[s] changes in affect” because every prediction serves allostasis.
- Affect is not specific to emotion. Barrett calls valence and arousal “basic features of consciousness … that, importantly, are not unique to instances of emotion”. Damasio and Carvalho’s list of feelings runs from hunger, thirst and colon distension to compassion. Both extend affect beyond humans. Barrett says a fly “may feel affect”, and Damasio and Carvalho argue from the conserved brainstem that other species very likely feel (C6 in NOTE-330).
- Affect is not the emotion. Both papers separate the felt state from what an emotion is. For Damasio and Carvalho, emotions are action programmes that cause feelings and can run without them (C2). For Barrett, an emotion is a categorization of affect and of exteroceptive sensation by an emotion concept. This account is about the felt layer only, and so it takes no side in the rivalry over what an emotion is (THEORY-057 against THEORY-055). Frijda’s action readiness (LIT-540) is compatible with it as well.
What they disagree on.
- Prediction or map. For Barrett the felt state is a prediction. It is allostatic, it is issued before the sensory input arrives, and it is corrected by prediction error (C1–C3 in NOTE-434). Damasio and Carvalho never use prediction except to say that feelings help anticipate “future adverse or favourable conditions”. Their feeling is a topographic map of a body state that has already changed. The somatic marker paper (LIT-384, NOTE-332) sits between the two. Its “as if body loop” has somatosensory cortex take on a body-state pattern directly, with the body bypassed, and Damasio believes adults now use it more than the body loop. A state generated inside the brain and felt as the body is close to what Barrett calls a prediction. The paper asserts this, and does not show it (C6 in NOTE-332).
- Where it is generated. Barrett’s visceromotor predictions start in agranular limbic cortex, in the anterior cingulate and ventral anterior insula (C3 in NOTE-434). Damasio and Carvalho put the generation of feeling in the upper brainstem, and hold that the insula is not necessary (C4–C5 in NOTE-330). They rest this on patient B, whose insulae were destroyed by herpes simplex encephalitis and who still feels. Barrett cites a patient with the same aetiology (“Roger”, Feinstein et al. 2010) who still has emotions, and explains the case by degeneracy. Neither paper says whether the two are the same person. Either way, each account absorbs the case, so it does not decide between them.
The consciousness holdings
- Seth (LIT-135, NOTE-169). Seth grounds consciousness in interoceptive predictive processing that serves allostasis (C4), the “beast machine”. Barrett cites him for the base layer. His claim is about consciousness, and the reading holds no statement from him about affect as such. So he is a neighbour, not a source. He shares Barrett’s predictive version, and his claim is the larger one: that being alive is the ground of experience in general.
- Affect as the ground of consciousness. Damasio and Carvalho call the “experiential aspect of homeostatic neural mappings … the lowest level of the mind and consciousness” and end on “the advent of feelings was simultaneously the advent of the mind” (C10, an assertion). Damasio’s “Feelings Are the Source of Consciousness” (LIT-387) is filed unread. This account does not claim that consciousness is grounded in feeling. It claims only that, where affect is felt, what is felt is the body’s regulatory state.
- THEORY-023 (AI consciousness cannot currently be settled). This account does not move that deadlock, but it locates it for affect. On both readings a system with no body to regulate has no affect in this sense, which is the condition Roberts sets for affective states (LIT-207, NOTE-157). The open question is a system that predicts and regulates a body, simulated or built. Barrett’s version is stated as computation over predictions, and a computational functionalist would credit such a system with affect. Damasio and Carvalho tie feeling to living tissue, down to a speculation about unmyelinated interoceptive fibres (C7 in NOTE-330), and Seth requires life. That is Birch’s Janus problem in miniature: the same architecture counts as evidence for one side and against the other (LIT-111). Man and Damasio’s design proposal for homeostatic “feeling machines” (LIT-386) is filed unread.
- Friston (LIT-526, NOTE-421). Barrett’s “active inference” is the neural predictive-coding scheme, not LIT-526’s Markov-blanket theorem. Her need for an internal model rests on the good-regulator theorem instead (footnote 5). Nothing here depends on the claim that every blanketed ergodic system “appears” to infer. Seth’s route to life passes through a variational–thermodynamic equivalence that LIT-526 does not supply. LIT-526 is not evidence for this account.
What this does not say
- It does not say that affect is an emotion, or that emotions are body states. The body-state view of emotion is Damasio’s (LIT-384), and Barrett rejects it for emotions while keeping it for affect.
- It does not say that prediction is how affect is generated. That is Barrett’s version. The record holds a read account without it, so the prediction claim is the open half.
- The record holds a dissent. Minsky (LIT-432, NOTE-348) holds that “feelings” are attempts by parts of the mind “to describe large-scale aspects of mental conditions”. Their objects are brain cascades, the body is “one more set of resources”, and they need reflective self-models. If he is right, what is felt is the brain’s state and not the body’s, and this account is false. He argues by example and offers no data, which is why it stays Proposed and does not count as refuted. His account is now stated as THEORY-570, a rival of this one.
- “Is the experience of” is not an identity claim about phenomenal character. The account says what felt affect is about and what varies with it. It does not say why it is felt at all.
Note of 2026-10-14: a later, consistent restatement
Theriault et al. (LIT-996, read in NOTE-761), Box 1, restate from the
Barrett group’s earlier papers that “affective experience, or mood,
functions as a low-dimensional allostatic barometer”. That is consistent
with this account’s Barrett half and is a later statement of it, not
independent evidence: the claim is restated, not argued, and the paper
places allostatic generation in limbic cortex and brainstem nuclei together,
which neither settles nor sharpens the disagreement with Damasio recorded
above. It is not added to source:. The statement is unchanged.