Lisa Feldman Barrett (2017), Social Cognitive and Affective Neuroscience 12(1):1–23, online 19 October 2016 — DOI-10.1093/scan/nsw154
Key takeaways
- Emotion is not a set of dedicated circuits triggered by stimuli. The brain predicts. An instance of emotion is a categorization of predicted bodily and worldly sensation by an emotion concept learned from past experience. Emotion concepts are therefore populations of variable instances, not types with a shared physical fingerprint.
- Affect and emotion come apart. Affect, valence and arousal, is how the brain’s allostatic predictions about the body are felt, and the paper calls it a basic feature of consciousness “not unique to instances of emotion”. Emotion categories are species-specific and perceiver-dependent.
- The paper is a rival to appraisal theory, not a refinement of it. In its view the meaning of a sensory event “does not trigger action, but results from it”, so theories in which a response follows from a stimulus “evaluated for its meaning” are doubtful. Descriptive appraisals survive, as products of categorization.
Standing in the record
Filed on 2026-10-03 at the owner’s request, in the batch on emotion theory. It
is the record’s statement of psychological constructionism. Lazarus’s Emotion
and Adaptation (LIT-550) is the appraisal theory the paper names as its
example of a “classical” causal appraisal account, and Frijda’s action-readiness
theory, read here through Frijda, Ridderinkhof and Rietveld (LIT-540), is
the other side of the same dispute. The record holds this as rivals
(ADR-022), not compared_against or corrects. This paper argues from brain
architecture that appraisal theory has the causal order backwards: meaning
follows categorization rather than triggering the emotion. It does not test
that claim against appraisal theory, and the record has not read Lazarus.
It joins the record’s interoception and predictive-processing cluster, and places emotion inside it.
- Seth (LIT-135). Seth bases consciousness on interoceptive predictive processing that serves allostasis. This paper bases emotion on the same machinery, and cites Seth (2013), Seth et al. (2012) and Seth and Friston (2016) for it. The two agree that interoception is the core of the brain’s model. They differ in what they hang on it: Seth uses it to tie consciousness to being alive, Barrett to dissolve emotion categories as natural kinds.
- Friston (LIT-526). The paper cites Friston (2010) for active inference. It does not use the formal free-energy or Markov-blanket apparatus of LIT-526. Its “active inference” is the predictive-coding architecture, anchored in Barbas’s cortical anatomy. LIT-526 derives apparent inference from blanket statistics in any ergodic system. This paper takes prediction as an empirical claim about laminar cortex.
- Damasio and Carvalho (LIT-388). Cited here for the ascending interoceptive pathway, which runs through the nucleus of the solitary tract, parabrachial nucleus and periaqueductal grey to granular insula. The two papers agree that feeling is the experience of interoceptive, homeostatic (allostatic) state. They disagree about emotions. LIT-388 treats fear, anger and the rest as innate “action programmes” distinct from feelings. That is a form of what this paper calls the classical view, and the paper reads such programmes as pattern generators plus a perceiver’s categorization.
No instruction for machine-learning practice, and no anthology topic holds it.