Are appraisals causes of emotion, or descriptions of it?
Opened by Barrett (2017), the wiki’s first source to attack appraisal theory directly rather than absorb it. The cognitive-appraisal page had been recording four readings of appraisal, all of which kept it in some form. This debate exists because a fifth source says the causal version has to go.
Caveat the wiki should keep visible: the causal theorists named in Barrett’s attack — Lazarus, Scherer, Roseman, and Clore & Ortony on the descriptive side — are still represented only through Barrett’s citations, not held first-hand in raw/. What the wiki now holds first-hand from the appraisal tradition is two Tong studies — Tong (2015) (recall) and Tong & Jia (2017) (experience sampling) — a working appraisal theorist, but both studies demonstrate appraisal-emotion correspondence without arguing (or being able to test) the causal claim. So the page still records an attack and the defences available in other sources rather than a symmetrical exchange on causation — but it is no longer true that the tradition appears here purely as a target. Scherer in particular has a documented history of objecting that Barrett’s camp misdescribes his position — see locationist-vs-constructionist-brain-emotion — and cannot be assumed to accept the characterisation here.
The argument
Barrett’s objection is not empirical. It is about ordering, and it is a one-step deduction from predictive coding:
As detailed in Figure 5, meaning does not trigger action, but results from it. This makes classical appraisal theories highly doubtful, because they assume that a response derives from a stimulus that is evaluated for its meaning.
In her architecture, visceromotor predictions come first; efferent copies cascade to motor and sensory cortices; the sensory consequences are anticipated before the input arrives. “For a given event, perception follows (and is dependent on) action, not the other way around.” A theory that runs stimulus → evaluation → response has the arrow pointing the wrong way, because there is no moment at which an unevaluated stimulus arrives and waits to be appraised.
Note the scope: she says this calls into question all classical theories of emotion, “even those that explain emotion as iterative stimulus → response sequences.” Iterating a wrong ordering does not fix it.
What survives
The concession is real and easy to miss, and it is what makes this a debate rather than a demolition:
Appraisals as descriptions of the world, however, are produced by categorization with concepts.
So the appraisal literature’s taxonomy — novelty, goal-conduciveness, agency, coping potential — is not being called false. Those may be perfectly good descriptions of how a situation is represented. What Barrett denies is that they name a mechanism. On her account the categorization produces both the emotion and the appraisal-describable content, in one act; the appraisal is a readout, not a cause.
This is Clore & Ortony’s position, and Barrett cites them approvingly (and cites Si et al., 2010, for appraisal being produced by categorization). Worth noticing that she is not opposing “appraisal theory” as a body of work — she is siding with one wing of it against another.
Why this is the wiki’s Week 9 question
The cognitive-appraisal page’s “through-line” section had already identified the hinge: how much of emotion is bodily signal and how much is interpretation of it? Four answers were recorded. Barrett supplies a fifth that reframes the question rather than answering it:
| answer | |
|---|---|
| appraisal theory | interpretation dominates (arousal is undifferentiated) |
| autonomic specificity | the signal is already differentiated |
| interoceptive-inference (Seth) | the distinction dissolves — two directions of one predictive hierarchy |
| LeDoux | both, in series, in different tissue |
| Barrett | the question presupposes an ordering that doesn’t exist; there is no unevaluated signal to interpret |
Seth and Barrett look close here and are not identical. Seth keeps appraisal by generalizing it — interpretation is inference, so appraisal theory was right in outline and imprecise in vocabulary. Barrett demotes it — inference is not appraisal-under-another-name, because appraisal claims to explain how a response is generated and inference denies there was a response-generation step. Seth’s move is friendly; Barrett’s is not.
The strongest reply available
Not made in any source the wiki holds, so recorded as an argument rather than a position, and flagged as the wiki’s own reasoning:
Barrett’s objection assumes appraisal theorists are committed to a temporal claim — that evaluation happens after the stimulus and before the response, as events in sequence. Scherer’s component process model explicitly treats emotion as an emergent process requiring a dynamic computational architecture, with recursive and parallel checks; it is not obvious it requires the serial ordering Barrett attacks. And Scherer has already objected once, on the record, that constructionists mischaracterize his theory as locationist when its appraisal checks were always domain-general. The same objection may transfer: appraisal-as-continuous-recursive-evaluation may be describing the same loop Barrett calls prediction and error, in another vocabulary.
If that is right, the disagreement is smaller than Barrett’s framing implies and largely terminological — which would be the second time in this wiki that a constructionist critique lands on a version of appraisal theory that its author does not hold. The wiki cannot check this without Scherer in raw/.
Status: open
Open, and the wiki is still poorly positioned to close it — but for a sharper reason than before. The wiki now holds two primary appraisal-theory sources, Tong (2015) and Tong & Jia (2017), and it turns out the primary sources do not settle the debate, because their method cannot.
Tong shows the appraisal taxonomy is empirically rich: 13 appraisals sort 13 positive emotions well above chance, cross-culturally, with the predicted profiles overwhelmingly confirmed — first from recall (2015), then replicated in experience sampling (2017). What he does not show — cannot show, using either recall-and-rate or naturalistic observation — is that appraisals produce the emotions. A correspondence between emotion and reported appraisal is exactly what Barrett’s descriptive reading predicts (categorization produces both the emotion and the appraisal-describable content) and exactly what the causal reading predicts (appraisals generate the emotion, so of course they track it). The data are silent between them, and Tong says so — in 2015 the correspondence “could reflect actual experience or culturalised conceptual knowledge,” and in 2017, flatly, the non-experimental design “does not take a stand on the causal directionality between appraisals and emotions.”
The one thing the 2017 paper does change: it was named on this page as the last pending appraisal file, with the prediction it would “not supply the missing manipulation either.” It does not. But it does retire the debate’s standing worry that the 2015 result might be a memory artefact — the profiles now replicate in ratings made within minutes of real events, so the taxonomy is not merely a reconstruction of what people believe goes with what. That strengthens the descriptive standing of appraisal theory without moving the causal question a step.
So the debate’s shape is settled into its stable form: the appraisal tradition is represented first-hand and its taxonomy is vindicated as description, in recall and in daily life — but the causal question turns on a manipulation (vary an appraisal, watch the emotion change) that no source in raw/ performs. Both remaining Week 9 appraisal files are now ingested; the missing manipulation is not among them.
The honest summary now: the wiki has heard the case against causal appraisal four times, and has now heard the appraisal tradition speak for itself twice — but on differentiation and co-occurrence, not on causation, which remains untested from either side.