Interoception Wiki

← RADlab

Lisa Feldman Barrett

11 min read

Researcher at a glance

  • Affiliation: Northeastern University; psychology / affective science
  • Orientation: Psychological constructionism; conceptual act theory / theory of constructed emotion; core affect

The principal contemporary voice against basic-emotion/natural-kinds assumptions, and thus against strong readings of Autonomic specificity of emotion. Cited by Friedman (2010) as the sharp counterpoint to the specificity tradition. See Basic emotions and natural kinds.

The constructive half, finally sourced

For eight ingests this page could only describe what Barrett denies. Barrett (2017) is the positive theory, and it changes how the rest of the page should be read: the natural-kinds critique is not the programme, it is a consequence of the programme. See Theory of constructed emotion.

The derivation runs from a claim about what brains are for, not about what emotions are:

A brain did not evolve for rationality, happiness or accurate perception.

It evolved to run a body — Allostasis. Regulating a system requires modelling it; modelling the world in a disembodied way “would be metabolically reckless”; so the model is built from the body’s perspective and necessarily includes the internal milieu. Interoception is therefore at the core of the brain’s internal model, and arises from allostasis. Affect is what the budget feels like, and — the line the wiki should hold onto — valence and arousal are “not unique to instances of emotion.”

Emotion arrives seventh in a seven-step derivation, as a special case of a general-purpose process. That is why there are no emotion-specific circuits to find: not because emotions aren’t real (they are as real as money), but because there is one mechanism, doing what it always does, with emotion concepts loaded. See Situated conceptualization (concepts as predictions), Degeneracy.

Two things this reframes on this page:

The framework run outward into social cognition (Theriault et al. 2021)

As senior author on Theriault, Young & Barrett (2021), Barrett extends the same machinery into the social domain — the wiki’s first sight of the theory doing work beyond emotion. The derivation is recognisably hers: a brain is for Allostasis; prediction error is metabolically costly (Metabolic cost of prediction error); so an unpredictable environment is a cost the organism avoids prospectively. The new step is that other people are the largest unpredictable part of a human’s environment, so conforming to their expectations is an allostatic strategy, and its felt form — the Sense of should — is Core affect pointed at social prediction error. Guilt, on this reading, is a constructed-emotion category unified by the sense of should (see that page). It shows the theory generalising as its authors claim it should: one mechanism, different concepts loaded — here, social rather than emotional ones. Lead author Theriault carries the framework’s novel pieces; Barrett’s role is as the architect of the underlying account.

As co-author rather than architect (Quigley et al. 2021)

A small entry, kept because it shows the framework in a role the wiki has not seen it in. On Quigley et al. (2021) Barrett is the fourth of five authors, supplying the affect-and-emotion section of a functional survey led by a psychophysiologist (Quigley) and a self-researcher (Tsakiris). Her material is restated compactly and without argument — Allostasis as the ground, interoception at the core of the internal model, affect as the budget’s readout, and the load-bearing line that interoception “is not specific to emotional events or emotional experiences.” She is cited alongside Seth (2013) and Seth & Friston (2016) as one of the “recent theoretical perspectives,” which is the wiki’s confirmation of the re-filing recorded above: she is on that shelf now, not opposite it.

What is worth recording is the candour of the section she is in. It states that the strongest evidence for interoception’s role in affect is neuroanatomical overlap, that the rest is temporal covariation, and that “this reliance on correlational evidence is a major limitation.” Barrett’s own writing (2017) does not concede this; the theory is presented there as derived rather than as awaiting evidence. Co-authorship is not endorsement of every sentence, and the concession is more plausibly Quigley’s — but it is the closest the wiki has to the constructionist camp marking its own evidential position on the body-to-feeling link. See Core affect.

Also worth noting for Are appraisals causes of emotion, or descriptions of it? and Cognitive appraisal (of emotion): the review reports Schachter & Singer-style physiological manipulation as difficult to replicate. Barrett’s account descends from the two-factor tradition’s structure (bodily state plus a categorizing step) while rejecting its physiology; a paper she co-authors saying the founding manipulation does not hold is a datum about the tradition’s empirical base, not about her theory.

The natural-kinds critique

Barrett (2005, 2006) argues that treating emotions as discrete entities independent of human perception “reifies” them; consistent emotion-specific somatic patterns may not exist outside subjective/conceptual experience. Emotions are constructed from core affect (a valence × arousal state) plus categorization.

Extending the critique into neuroimaging

Lindquist, Wager, Kober, Bliss-Moreau & Barrett (2012) operationalizes the natural-kinds critique as a testable neuroimaging hypothesis: does any brain region show both consistent and specific activation for a discrete emotion category? Meta-analyzing ~90+ published studies, the answer is no — amygdala, insula, ACC, and OFC each show functional selectivity for particular categories (e.g., insula and disgust) without functional specificity (the same regions are also active for other categories). The positive alternative — a domain-general “neural reference space” comprising core affect, conceptualization, language, and executive-attention networks — is this wiki’s most detailed empirical case for constructionism over the anatomical locationism implicit in strong readings of the basic-emotions tradition. See Do discrete emotions map onto distinct brain regions or networks?.

LeDoux’s half-endorsement

LeDoux (2012) is the wiki’s most useful stress-test of how far the natural-kinds critique carries, because he accepts the destructive half and rejects the constructive half — from inside the animal-circuit tradition Barrett’s critique targets.

  • Agrees: “I concur with her conclusion that the foundation of support for the idea that basic emotions, as conventionally conceived, have dedicated neural circuits is weak.”
  • Disagrees on the inference: this shows the categories are wrong, not that innate circuits are absent. Enter Survival circuits.
  • Disagrees on the method: Barrett’s argument from similar fMRI activation across emotions to shared mechanism “does not have the resolution necessary” — area-level similarity can conceal microcircuit-level difference (Lin et al. 2011). See Amygdala.

Worth registering that they are closer than the polemic suggests. Both require a categorization/labeling step; both give language real work; both treat a pre-categorical bodily state as raw material (her core affect, his global organismic state); and LeDoux names core affect as the most plausible account of whatever animals do have. The live disagreements are (1) whether hard-wired function-specific circuitry survives the critique, and (2) what imaging can license.

The challenge she has not answered here

Volynets et al. (2020) is the strongest empirical case in this wiki against her position, and the exchange has not happened: she is not cited in it, and no reply from her appears in the wiki’s sources. Recorded as an open gap rather than a defeat.

The finding is exactly what her account is usually taken to deny — 13 emotions with distinct, discrete bodily topographies, stable across 101 countries and both sexes, with emotion category swamping culture, language, education, BMI and sex in a single model. Nummenmaa’s group reads it as vindicating Ekman.

Two things make it less damaging than it looks, and both are available from within the source:

  • It is phenomenology, not physiology. The maps are self-reported felt topography, and the authors concede they “are thus subjective feeling maps” that say nothing about emotion-specific bodily states. Her critique targets the latter. See Bodily sensation maps (bodily fingerprints of emotion), Autonomic specificity of emotion.
  • The cue is a word. Subjects were shown emotion terms, in English, and asked where they would feel them. If emotion categories are constructed from core affect plus conceptualization, and the concepts are widely shared, then cross-culturally concordant maps are a prediction of her theory, not a refutation of it. The paper’s only defence is an inherited citation (Nummenmaa et al. 2014: word-cued maps ≈ induction-cued maps).

So the constructionist has a good answer available — and owes it, because the pattern is large and robust rather than the weak, contested effects the natural-kinds critique was built to dismiss. See Are emotions culturally universal?.

Update: the answer is now sourced, three years early

Barrett (2017), ingested after this section was written, contains the second bullet’s argument in her own words — not aimed at Volynets (which postdates it), but in terms that transfer exactly:

The categories of social reality, such as flowers and weeds, or emotion categories, are conceptual because functions are imposed on physically disparate instances by virtue of collective agreement.

Shared concepts, shared maps. If emotion categories are held together by collective agreement rather than by physical similarity, then word-cued maps concordant across 101 countries measure the reach of the agreement, not the existence of an essence. The wiki’s inference was right and can stop being an inference.

What this does not do is close Are emotions culturally universal?. Barrett still has not engaged Nummenmaa’s data, and the gap this section records is a gap in the exchange, not in the theory’s resources. Two live questions survive:

  • Why would collective agreement be this uniform? Concordance at rs > 0.82 across 101 countries needs an account of how the agreement got distributed that widely. A constructionist can give one (shared bodies, shared situations, translated instruments, global media), but must give one — “it’s conceptual” names the mechanism without measuring it.
  • Her own statistical objection cuts here too. Volynets’s maps are group averages over thousands of subjects. By Barrett’s own 3.14-person-family argument, the averaged anger map need not describe any individual. That objection is available to her and would be more damaging than the concepts reply — and neither she nor Nummenmaa’s group has made it.

Contested appropriation of James

Both camps claim James. Barrett cites his acknowledgement of individual variation and his warning against treating “anger” as an “entitative” thing. Friedman/Ellsworth reply that James, read in full context, held within-emotion resemblance “trumps” individual variation and that his causal model presupposes basic emotions “that have a distinct bodily expression.” This exegetical dispute is part of Do basic emotions have distinct autonomic signatures? — see William James.

Graph View

  • Studies
  • Concepts
  • Methods
  • Debates
  • Gaps
  • Researchers, other
  • Studies
  • Concepts
  • Methods
  • Debates
  • Gaps
  • Researchers, other

Backlinks