Debate at a glance
- Summary: Two near-contemporaneous accounts posit a unified, momentary integration of body state, context, and memory as the substrate of subjective feeling — and place it in different lobes. Craig locates it at the anterior insula/frontal-operculum junction (the ‘global emotional moment’); LeDoux locates it in a prefrontal/parietal ‘cognitive workspace’ and gives the insula no role at all. The dispute bears directly on whether interoceptive cortex is the medium of feeling or merely one input to it.
- Status: open
Positions in the field
- A. D. (Bud) Craig: The anterior insular cortex is where awareness is instantiated. Salience across all conditions (homeostatic, environmental, hedonic, motivational, social, cognitive) is integrated along the posterior→anterior insular progression into a unified meta-representation of the sentient self ‘now’ — the global emotional moment. Feelings are the ‘computational common currency of awareness,’ grounded in homeostasis; a timebase-indexed sequence of these moments yields continuous subjective awareness. Von Economo neurons in AIC/ACC are the speculative substrate. (Craig (2009), Craig (2002))
- Joseph LeDoux: Feelings arise when a global organismic state (produced by survival circuit activation) is represented in a prefrontal/parietal cognitive workspace and there appraised and labeled. Six ingredients integrate: environmental stimuli, survival circuit activity, CNS arousal, body feedback, explicit memory, language. Body feedback is one input among six, not the medium. The insula plays no role in the account. Dorsolateral prefrontal cortex — largely absent in other mammals — is a key workspace component, which is what grounds his skepticism about animal feelings. (LeDoux (2012))
- Lisa Feldman Barrett / Kristen Lindquist: Closer to LeDoux on anatomy while opposed to him on categories. The anterior insula represents core affective feelings and interoceptive awareness (citing Craig approvingly), but conscious emotional experience additionally requires a conceptualization network (DMPFC, medial temporal lobe, retrosplenial/posterior cingulate) plus language (anterior temporal, VLPFC) and executive attention (DLPFC, VLPFC). So the insula is necessary-but-insufficient: an ingredient, with the categorizing work done cortically elsewhere. (Lindquist et al. (2012))
- Anil K. Seth / Karl Friston: Sides with the insula, but reverses its direction of operation. The AIC is a comparator in a predictive hierarchy where emotional content is constituted by interoceptive predictions and prediction errors, not by ascending integration. The agranular visceromotor areas (AIC, ACC, SGC, OFC) are cytoarchitecturally positioned as sources of descending predictions — so ‘the insula’ in Seth’s account is not doing Craig’s job even though it occupies Craig’s location. (Seth (2013), Seth & Friston (2016))
- Andrea Scarantino: Does not take a position on the lobe, but reports LeDoux’s own later view as insula-inclusive — citing LeDoux & Brown (2017) for locating higher-order representation in ‘general networks of cognition, which involve regions in the prefrontal cortex, posterior parietal cortex, and insular cortex that underlie attention, working memory, and metacognition.’ Endorses that account. Separately, he shares LeDoux’s higher-order structure (a feeling is a state plus a representation of it), which is what Craig denies: on Craig’s model the AIC’s integration is the awareness, with no second representational step. The structural divide may matter more than the anatomical one. (Scarantino (2018), cited via LeDoux & Brown (2017) A higher-order theory of emotional consciousness, PNAS)
Connections
- Related debates: inferential?, Do discrete emotions map onto distinct brain regions or networks?, Can we know whether other animals have conscious feelings?, Should emotion terms refer to feelings or to motive states?
- Related concepts: Global emotional moment (Craig’s model of awareness), Global organismic state, Insular cortex, Core affect, Interoceptive inference, Embodied selfhood, Higher-order theories of consciousness
Raised by the LeDoux (2012) ingest, which put a construct on the table that is structurally near-identical to Craig (2009)‘s and anatomically incompatible with it.
The near-collision
Set the two constructs side by side:
| Craig’s Global emotional moment (Craig’s model of awareness) | LeDoux’s Global organismic state | |
|---|---|---|
| What it integrates | salience across homeostatic, environmental, hedonic, motivational, social, cognitive conditions | stimulus/context, survival circuit activity, CNS arousal, body feedback, explicit memory, language |
| Unified? | yes — a single meta-representation of the sentient self at “now” | yes — a whole-organism state monopolizing brain resources |
| Momentary? | yes — sequenced by an endogenous timebase into a stream | yes — though prolonged by slow hormonal action |
| Is it the feeling? | yes — the moment is the awareness | no — it is raw material; a workspace must represent it |
| Where | anterior insula / frontal operculum junction | prefrontal / parietal cognitive workspace |
Both authors independently converged on “unified momentary multi-ingredient integration of body state, context, and memory” as the shape of the answer. The convergence is what makes the disagreement interesting rather than merely terminological: this is not two people describing different phenomena, it is two people assigning the same phenomenon to different cortex.
Neither cites the other. Craig 2009 and LeDoux 2012 are three years apart in the same literature.
What actually turns on it
Is interoceptive cortex the medium of feeling, or an input to it? This is the question the wiki’s whole Craig/Seth spine presupposes one answer to. On Craig’s account, feelings are made of interoception — the AIC’s re-representation of body state is subjective feeling, and Embodied selfhood follows. On LeDoux’s, body feedback is one of six ingredients, and a feeling is a workspace representation that could in principle be assembled with a weak bodily contribution.
Does the answer license the animal-feelings skepticism? LeDoux’s argument in Can we know whether other animals have conscious feelings? depends on his anatomy: DLPFC is lacking in most mammals, therefore feelings of the human kind are unlikely in them. If Craig is right that the AIC constitutes feeling, the argument loses its premise — other mammals have insular cortex. The two debates are not independent.
Which way does the arrow run? Note that Craig and Seth agree on the location while disagreeing about direction (ascending integration vs. descending prediction) — that is inferential?, a separate axis. A full position needs an answer on both: where, and which way.
Points of partial agreement
- Craig, LeDoux, and Barrett all reject a feeling as a simple read-out of a peripheral signal; all posit an integrative step.
- LeDoux and Barrett/Lindquist both require a categorization/labeling step and both give language real work — LeDoux notes >30 English words for gradations of fear and doubts specific emotions arise without words. Their disagreement is about categories, not about anatomy, where they are close. See Do discrete emotions map onto distinct brain regions or networks?.
- Lindquist et al. cite Craig approvingly for the insula’s interoceptive-awareness role while denying the insula suffices for emotional experience. So “insula is necessary but not sufficient” is a genuinely available middle position, and is arguably where the weight of this wiki’s sources sits.
A complication from the Scarantino ingest: is the insula back in?
Scarantino (2018) does not argue about lobes, but he reports something that bears directly on this page’s framing. Endorsing LeDoux & Brown (2017), he describes the higher-order representations required for consciousness as cortically implemented by “general networks of cognition, which involve regions in the prefrontal cortex, posterior parietal cortex, and insular cortex that underlie attention, working memory, and metacognition.”
The 2012 paper this debate is built on gives the insula no role. The 2017 co-authored paper, as reported, lists it. Three readings:
- A shift. LeDoux’s position moved, and this page’s “the insula plays no role in the account” is dated to 2012.
- No shift — different job. The insula enters as one workspace component supporting attention/working memory/metacognition, not as the medium of feeling. The Craig opposition survives intact, since the dispute was never whether the insula is involved but whether the AIC’s integration of body state is the feeling.
- Scarantino’s compression of LeDoux & Brown’s emphasis rather than their claim.
Reading (2) is most likely and changes little. Reading (1) would materially weaken Can we know whether other animals have conscious feelings?, whose force comes entirely from DLPFC being absent in most mammals — an insula-inclusive substrate is one other mammals have. LeDoux & Brown (2017) is not in raw/; this cannot be settled from the commentary. Flagged for a human as the highest-value missing source for this debate.
The deeper divide: one stage or two?
The Scarantino ingest surfaces something this page had not recorded, and it may matter more than the lobe disagreement.
Every position here except Craig’s is a two-stage account: a first-order state, plus something that represents, conceptualizes, or labels it. LeDoux (global organismic state → workspace representation), Barrett/Lindquist (core affect → conceptualization), Scarantino (activated basic emotion → higher-order representation), and behind all three, two-factor theory.
Craig’s global emotional moment is one-stage. The AIC’s integration is the awareness. There is no higher-order representation because there is nothing left for one to do.
So the question underneath “where” is: is a feeling what a certain integration is like, or what happens when a state gets represented again? Craig answers the first; both parties to the survival-circuits dispute answer the second. That is a divide about the structure of consciousness, not about anatomy — and it means Craig’s opponents on this page agree with each other more than the lobe framing suggests. See Higher-order theories of consciousness.
Status: open
No source in the wiki adjudicates this directly, and the two principals do not engage each other. LeDoux’s own concession is relevant to how open it is: “the brain mechanisms that underlie conscious emotional feelings in humans are still poorly understood,” and global organismic states are currently unmeasurable by his own account (no method gives whole-brain cellular resolution simultaneously with body physiology). Craig’s list of what his model does not explain is comparably long (how a feeling is constructed; the nature of the timebase; the salience metric).
Flagged for a human: this debate has no empirical discriminator recorded yet. A useful next step would be lesion evidence — whether insular damage abolishes feeling (Craig predicts yes; LeDoux predicts impairment of one ingredient) — which none of the current sources supply.