Emotional Expressions Reconsidered: Challenges to Inferring Emotion From Human Facial Movements

Emotional Expressions Reconsidered: Challenges to Inferring Emotion From Human Facial Movements Lisa Feldman Barrett Ralph Adolphs Stacy Marsella Aleix M. Martı́nez Seth D. Pollak

Tag layer of the DHSS Meta-Tagging Project. The article text is reproduced from the staged copy named in sourceDesc and is not altered; the tags are held separately in standOff so that overlapping and nested spans are preserved.

Emotional Expressions Reconsidered: Challenges to Inferring Emotion From Human Facial Movements 2019 Psychological Science in the Public Interest 10.1177/1529100619832930 barrett2019-emotional-expressions-reconsidered barrett2019-emotional-expressions-reconsidered.txt, the project's staged copy bf3968ace26d0f11694407d3546229fa39f4f4be cc-by-sa psychology

the paper's own words

the tag layer; a sentence written by the tagger about the paper, always anchored to a quotation but not itself a quotation

0 spans located by searching the stored text for the quote's exact words, allowing any whitespace between them the inference from facial movements to emotional states, and whether the scientific evidence licenses the way that inference is being used in law, hiring, security, education and commercial software Quotation withheld because this record is not cleared for redistribution. whether the existing research on emotional expression and emotion perception is strong enough to justify how it is increasingly being used by those who consume it Quotation withheld because this record is not cleared for redistribution. a systematic review of production and perception studies against four named criteria, and a prescription for naming: call a facial configuration what was measured until the evidence licenses calling it an expression Quotation withheld because this record is not cleared for redistribution. critique Quotation withheld because this record is not cleared for redistribution. a measurement routinely given the name of the construct it is taken to reveal, so that the name carries an inference the data do not support Quotation withheld because this record is not cleared for redistribution. systematic-review systematic-review Quotation withheld because this record is not cleared for redistribution. offering a participant six emotion words and one face forces a choice among them, so agreement measures how the options were built rather than what the face conveys; remove the array and let people supply their own word and the rates fall response-set constraint manufacturing agreement Quotation withheld because this record is not cleared for redistribution. where happiness is the only pleasant category in the array, every distractor differs from the target in valence, so a participant can score perfectly by discriminating pleasant from unpleasant without perceiving any emotion category at all valence perception passing as emotion perception Quotation withheld because this record is not cleared for redistribution. calling a measured facial movement an 'emotional display' imports the conclusion into the description, so a later reader inherits an inference nobody tested; the same happens with 'recognizing' rather than 'inferring' and 'accuracy' rather than 'agreement' naming carrying the inference Quotation withheld because this record is not cleared for redistribution. a result reliable enough to publish can still be low in absolute terms, so a literature of individually valid findings accumulates into a consensus that no single finding supports above-chance publication threshold Quotation withheld because this record is not cleared for redistribution. the review covers six emotion categories, not the twenty-plus currently studied, because the evidence base for the other categories is far smaller - the conclusions are extended to them by argument rather than by review Quotation withheld because this record is not cleared for redistribution. evidence from small-scale remote cultures on expression PRODUCTION is inconclusive rather than negative: there are no systematic controlled observations of how people in these contexts spontaneously move their faces Quotation withheld because this record is not cleared for redistribution. validity - the fourth criterion - is not evaluated by the review either, because the underlying studies almost never address it Quotation withheld because this record is not cleared for redistribution. the authors' own earlier papers used exactly the design they criticise, presenting posed configurations without context; the critique includes its own authors Quotation withheld because this record is not cleared for redistribution. reliability, specificity, generalizability and validity - the criteria of psychological measurement - carried into an audit of what a whole literature's vocabulary is entitled to assert Quotation withheld because this record is not cleared for redistribution. reverse inference, and the conditional probability that formalises it, carried from neuroimaging methodology into the reading of faces Quotation withheld because this record is not cleared for redistribution. four criteria fixed in advance - reliability, specificity, generalizability, validity - so that the review's verdict is not a matter of the reviewers' impression eliminates: a review verdict that is the reviewers' impression of a literature they already disagreed with · outcome: partial Quotation withheld because this record is not cleared for redistribution. evidence sought across four independent populations - developed-nation adults, small-scale remote cultures, infants and children, and congenitally blind people - so that no single literature's method could carry the conclusion eliminates: a conclusion about facial expression drawn from one population - Western adults - and stated about people · outcome: held Quotation withheld because this record is not cleared for redistribution. results partitioned by the METHOD used to collect them, to test whether support for the common view is a property of the phenomenon or of the task eliminates: support for the common view that is a property of the TASK used to collect it rather than of the phenomenon · outcome: failed Quotation withheld because this record is not cleared for redistribution. the authors turning the same critique on their own published work rather than only on other laboratories' eliminates: a critique that applies to every laboratory except the one making it · outcome: held Quotation withheld because this record is not cleared for redistribution. testing the common view against specific alternative hypotheses about context, rather than only against chance eliminates: specific alternative accounts of the expression-emotion link, in particular context-based ones - testing against the null alone rules out chance and nothing else · outcome (stated by the tagger, no quotation): absent-stated-about-other Quotation withheld because this record is not cleared for redistribution. reverse inference reverse inference Quotation withheld because this record is not cleared for redistribution. reliability reliability Quotation withheld because this record is not cleared for redistribution. specificity specificity Quotation withheld because this record is not cleared for redistribution. generalizability generalizability Quotation withheld because this record is not cleared for redistribution. validity validity Quotation withheld because this record is not cleared for redistribution. emotional expression emotional expression Quotation withheld because this record is not cleared for redistribution. facial configuration facial configuration Quotation withheld because this record is not cleared for redistribution. emotional expressions emotional expressions Quotation withheld because this record is not cleared for redistribution. facial configurations facial configurations Quotation withheld because this record is not cleared for redistribution. accuracy accuracy Quotation withheld because this record is not cleared for redistribution. Consistency Consistency Quotation withheld because this record is not cleared for redistribution. Common view Common view Quotation withheld because this record is not cleared for redistribution. stereotype stereotype Quotation withheld because this record is not cleared for redistribution. prototype prototype Quotation withheld because this record is not cleared for redistribution. Choice-from-array tasks Choice-from-array tasks Quotation withheld because this record is not cleared for redistribution. free labeling free labeling Quotation withheld because this record is not cleared for redistribution. context context Quotation withheld because this record is not cleared for redistribution. Ecological validity Ecological validity Quotation withheld because this record is not cleared for redistribution. Conditional probability Conditional probability Quotation withheld because this record is not cleared for redistribution. classical view classical view Quotation withheld because this record is not cleared for redistribution. Facial movement Facial movement Quotation withheld because this record is not cleared for redistribution. Facial expression Facial expression Quotation withheld because this record is not cleared for redistribution. Perceiver-dependent Perceiver-dependent Quotation withheld because this record is not cleared for redistribution. Perceiver-independent Perceiver-independent Quotation withheld because this record is not cleared for redistribution. Forward inference Forward inference Quotation withheld because this record is not cleared for redistribution. Percent agreement Percent agreement Quotation withheld because this record is not cleared for redistribution. percent accuracy percent accuracy Quotation withheld because this record is not cleared for redistribution. Recognition Recognition Quotation withheld because this record is not cleared for redistribution. perception perception Quotation withheld because this record is not cleared for redistribution. Discrimination Discrimination Quotation withheld because this record is not cleared for redistribution. identification identification Quotation withheld because this record is not cleared for redistribution. Emotional granularity Emotional granularity Quotation withheld because this record is not cleared for redistribution. "emotional expression" - AS LICENSED: a facial configuration that reliably and specifically communicates an emotional state. The paper's glossary adds that many so-called emotional expressions are errors of reverse inference Quotation withheld because this record is not cleared for redistribution. "emotional expression" - AS USED IN THE LITERATURE: any facial configuration labelled with an emotion category, applied without checking reliability, specificity, generalizability or validity Quotation withheld because this record is not cleared for redistribution. "accuracy" - AS DEFINED: correspondence to an intended performance that must be perceiver-independent and not subject to the whims of the experimenter Quotation withheld because this record is not cleared for redistribution. "accuracy" - AS IT SHOULD BE RENAMED in this literature: agreement or consensus, because what these studies measure is whether people label a face the same way, not whether the label is correct Quotation withheld because this record is not cleared for redistribution. 'common view' / 'classical view' - THE FRAMEWORK THIS PAPER IS AGAINST HAS TWO NAMES, AND THE SECOND APPEARS EXACTLY ONCE: 'a set of beliefs were refer to as the common view (also called the classical view)'. A definitions row already glosses 'Common view'; what was missing is that the literature this paper argues with calls it something else. A reader searching the corpus for 'classical view' - which is what the opposing literature calls itself - would not have reached this record. Quotation withheld because this record is not cleared for redistribution. Facial movement - an objective, perceiver-independent configuration description that makes no emotional inference Quotation withheld because this record is not cleared for redistribution. Facial expression - a perceiver-dependent inference that a facial configuration expresses an internal state Quotation withheld because this record is not cleared for redistribution. Perceiver-dependent - interpretation depends on human judgment and may be consistent without being accurate or valid Quotation withheld because this record is not cleared for redistribution. Perceiver-independent - observation does not depend on human judgment, though independence comes in degrees Quotation withheld because this record is not cleared for redistribution. Forward inference - inferring an effect from a known cause, such as frowning conditional on anger Quotation withheld because this record is not cleared for redistribution. Percent agreement - inter-rater consistency, explicitly not percent accuracy Quotation withheld because this record is not cleared for redistribution. Percent accuracy - not interchangeable with agreement because agreement is more perceiver-dependent Quotation withheld because this record is not cleared for redistribution. Recognition - acknowledging an independently confirmed existent, contrasted with interpretive perception Quotation withheld because this record is not cleared for redistribution. Perception - inference and interpretation rather than independently confirmed recognition Quotation withheld because this record is not cleared for redistribution. Discrimination - deciding that stimuli differ without identifying what they are or recognizing what they mean Quotation withheld because this record is not cleared for redistribution. Identification - deciding what a stimulus is, not merely that it differs from another Quotation withheld because this record is not cleared for redistribution. Emotional granularity - experiencing or perceiving emotion through many differentiated categories Quotation withheld because this record is not cleared for redistribution. critiques → the common view that each of six emotion categories is reliably and specifically expressed by, and recognisable from, its own configuration of facial movements Quotation withheld because this record is not cleared for redistribution. critiques → automated emotion-recognition technology that infers emotional states from facial movements alone Quotation withheld because this record is not cleared for redistribution. qualifies → choice-from-array emotion-perception findings, which supply the strongest support for the common view and cannot produce disconfirming evidence Quotation withheld because this record is not cleared for redistribution. reinterprets → the canonical six facial configurations - as Western stereotypes rather than as prototypes or universal displays Quotation withheld because this record is not cleared for redistribution. emotional expressions —distinct-from→ facial configurations emotional expressions facial configurations differs in: whether reliability, specificity, generalizability and validity have been demonstrated · the relation itself is separately evidenced · strength of the relation: 0.95 · differs in (stated by the tagger, no quotation): whether reliability, specificity, generalizability and validity have been demonstrated · the relation stated: the second is what was measured; the first adds an inference the evidence must license Quotation withheld because this record is not cleared for redistribution. accuracy —distinct-from→ Consistency accuracy Consistency differs in: whether there is a perceiver-independent correct answer to be right about · the relation itself is separately evidenced · strength of the relation: 0.95 · differs in (stated by the tagger, no quotation): whether there is a perceiver-independent correct answer to be right about · the relation stated: a group can consistently believe something that is wrong Quotation withheld because this record is not cleared for redistribution. stereotype —distinct-from→ prototype stereotype prototype differs in: whether the configuration is actually the most frequent instance of the category · the relation itself is separately evidenced · strength of the relation: 0.95 · differs in (stated by the tagger, no quotation): whether the configuration is actually the most frequent instance of the category · the relation stated: a prototype is the typical instance; a stereotype is a belief over-applied Quotation withheld because this record is not cleared for redistribution. Choice-from-array tasks —contrasted-with→ free labeling Choice-from-array tasks free labeling differs in: whether the participant may supply a word the experimenter did not offer · the relation itself is separately evidenced · strength of the relation: 0.9 · differs in (stated by the tagger, no quotation): whether the participant may supply a word the experimenter did not offer · the relation stated: the constrained task inflates agreement and cannot produce disconfirming evidence Audit 2026-09-28: DOUBTFUL - read against its full span and kept, but the span describes the choice-from-array method without naming it. Quotation withheld because this record is not cleared for redistribution. common view —same-as→ classical view common view classical view the relation itself is separately evidenced · strength of the relation: 1 · note: A bare parenthetical synonym, no hedge. THE AXIS IS THE AUTHOR'S OWN VS THE OPPOSING LITERATURE'S: Barrett coins 'common view' for the position she is attacking, and notes that it already has a name. lists discipline, language, profession and venue as the axes this type carries; this is a fifth - THE NAME A POSITION'S CRITICS GIVE IT AGAINST THE NAME IT GIVES ITSELF. · the relation stated: the paper gives the second name parenthetically as an alternative for the first Quotation withheld because this record is not cleared for redistribution. facial configuration —distinct-from→ Facial expression facial configuration Facial expression the relation itself is separately evidenced · strength of the relation: 1 · note: The glossary explicitly says the objective configuration is not synonymous with the inference-laden expression. Quotation withheld because this record is not cleared for redistribution. Perceiver-independent —distinct-from→ Perceiver-dependent Perceiver-independent Perceiver-dependent the relation itself is separately evidenced · strength of the relation: 1 · note: The glossary contrasts the two categories through dependence on human judgment and a 'whereas' example. Quotation withheld because this record is not cleared for redistribution. Percent agreement —distinct-from→ percent accuracy Percent agreement percent accuracy the relation itself is separately evidenced · strength of the relation: 1 · note: The glossary states directly that percent agreement is not percent accuracy. Quotation withheld because this record is not cleared for redistribution. Discrimination —distinct-from→ identification Discrimination identification the relation itself is separately evidenced · strength of the relation: 1 · note: The glossary says discrimination is separate from identifying what stimuli are. Quotation withheld because this record is not cleared for redistribution. Discrimination —distinct-from→ Recognition Discrimination Recognition the relation itself is separately evidenced · strength of the relation: 1 · note: The glossary says discrimination is separate from recognizing what stimuli mean. Quotation withheld because this record is not cleared for redistribution. the six canonical facial configurations occur during their predicted emotions above chance, but weakly, and evidence for specificity is lacking in almost all research domains the six canonical facial configurations occur during their predicted emotions above chance, but weakly, and evidence for specificity is lacking in almost all research domains Quotation withheld because this record is not cleared for redistribution. there are many-to-many mappings between facial configurations and emotion categories - anger is expressed more broadly than a scowl, and scowls express more than anger there are many-to-many mappings between facial configurations and emotion categories - anger is expressed more broadly than a scowl, and scowls express more than anger Quotation withheld because this record is not cleared for redistribution. the strength of support for the common view varies systematically with the research method used, and the strongest support does not generalise the strength of support for the common view varies systematically with the research method used, and the strongest support does not generalise Quotation withheld because this record is not cleared for redistribution. Choice-from-array tasks severely limit the chance of observing disconfirming evidence, because they restrict the labels offered. Choice-from-array tasks severely limit the chance of observing disconfirming evidence, because they restrict the labels offered. Quotation withheld because this record is not cleared for redistribution. where happiness is the only pleasant option, apparent emotion perception may be valence perception where happiness is the only pleasant option, apparent emotion perception may be valence perception Quotation withheld because this record is not cleared for redistribution. emotion-reading technology detects facial movements, not emotional expressions emotion-reading technology detects facial movements, not emotional expressions Quotation withheld because this record is not cleared for redistribution. the literature confounds measurement with interpretation - calling movements expressions, inference recognition, and agreement accuracy the literature confounds measurement with interpretation - calling movements expressions, inference recognition, and agreement accuracy Quotation withheld because this record is not cleared for redistribution. a facial configuration, bodily movement or vocal expression that reliably and specifically communicates an emotional state - and many so-called emotional expressions are errors of reverse inference emotional expression Quotation withheld because this record is not cleared for redistribution. that certain emotion categories reliably cause specific patterns of facial muscle movements, and that specific configurations are diagnostic of certain emotion categories Common view Quotation withheld because this record is not cleared for redistribution. extent to which a participant's performance corresponds to the intended performance, which requires the correct answer to be perceiver-independent accuracy Quotation withheld because this record is not cleared for redistribution. an outcome that does not vary greatly across time, context or individuals - and not accuracy, since a group can consistently believe something wrong Consistency Quotation withheld because this record is not cleared for redistribution. the probability that X occurs given that Y has occurred; p(frown given sadness) is the form a face-reading claim takes Conditional probability Quotation withheld because this record is not cleared for redistribution. the extent to which findings generalise to real-life settings, and the extent to which a protocol captures valid aspects of the real world Ecological validity Quotation withheld because this record is not cleared for redistribution. assuming an observed facial movement reveals an emotional state you cannot access directly reverse inference Quotation withheld because this record is not cleared for redistribution. the correlation range the paper calls weak reliability - a person sometimes scowls in anger, but not most of the time r = .20 to .39 Quotation withheld because this record is not cleared for redistribution. reliability reliability Quotation withheld because this record is not cleared for redistribution. specificity specificity Quotation withheld because this record is not cleared for redistribution. generalizability generalizability Quotation withheld because this record is not cleared for redistribution. null hypothesis null hypothesis Quotation withheld because this record is not cleared for redistribution. context context Quotation withheld because this record is not cleared for redistribution.

The full text of this article is not redistributable, so it is not included here. The evidence sentences are withheld with it; the tag layer and its grounding status remain. The article itself is at https://doi.org/10.1177/1529100619832930.