Differentiation of 13 positive emotions by appraisals

Study at a glance

• Authors: Eddie M. W. Tong
• Year: 2015
• Journal: Cognition and Emotion
• DOI: 10.1080/02699931.2014.922056
• N: 90
• Design: Within-participant recall study (Smith & Ellsworth 1985 method) across two cultures. 50 Singaporean and 40 American undergraduates each attended seven 1-hour sessions on separate days; in each, they recalled and wrote out a personal experience of one of 13 positive emotions, then rated it on 13 appraisal items (9-point scales, with a ‘not relevant’ opt-out per item). Analysis: discriminant analysis (classification accuracy), mixed ANOVAs and repeated-measure contrast analysis over the appraisal profiles, and a principal-component analysis reducing the 13 appraisals to 4 factors.

Key findings

• Appraisals classify 13 positive emotions well above chance — the wiki’s first primary evidence that the appraisal taxonomy has discriminative power. Discriminant analysis correctly classified 42.4% of 1170 emotion cases (leave-one-out 36.4%) against a chance rate of 7.7%, comparable to or higher than past studies classifying emotions in general. This is the appraisal-theory tradition (Cognitive appraisal (of emotion)) doing the thing it claims to do, held first-hand rather than through its critics for the first time in this wiki.
• 91 of 95 incongruent predicted emotion pairs differed as predicted, at a punishing correction (Bonferroni p = .00064). A positive emotion predicted to be higher (or lower) than another on a given appraisal was in fact significantly higher (lower) in the vast majority of cases. The four exceptions: compassion did not differ from joy on problems; contentment not from gratitude on agency-self; awe not from pride on agency-others; joy not from relief on effort. Support for the a-priori appraisal profiles (Table 1) is strong.
• But three emotions resist the appraisals — interest, joy and relief were classified at or barely above chance (leave-one-out: joy 11.86%, relief 12.00%, interest 10.20%). Tong’s own reading: ‘the appraisals are not apt at predicting all positive emotions,’ and he cannot say why. The taxonomy that works so well overall has a floor, and it falls exactly on some of the most canonical positive states.
• Contentment and joy could not be distinguished by any appraisal at all. They were assigned identical predicted profiles and showed no appraisal difference empirically. Tong offers two readings: a critical appraisal was omitted, or the two emotions ‘differ not qualitatively but quantitatively’ — same appraisals, different intensities. Either way, appraisal-based differentiation has a hard limit here, and it is worth setting beside the constructionist claim that discrete positive-emotion categories may not carry distinct signatures (see Bodily sensation maps (bodily fingerprints of emotion), Theory of constructed emotion). (The EMA follow-up Tong & Jia (2017) later recovers one differentiator — joy is tied to understandability where contentment and serenity are not — so the null softens but does not vanish; Tong still allows the quantitative-difference reading.)
• Pleasantness was the least-dispensable appraisal — the appraisal-theory echo of the wiki’s valence-primacy thread. Of 13 appraisals, pleasantness drew by far the fewest ‘not relevant’ responses (0.80% across emotions vs 22-23% for problems and goal attainment), followed by control-self and certainty. Tong reads this as ‘valence is viewed by laypersons (in agreement with experts) as fundamental to emotional experiences.’ But pleasantness alone is nowhere near sufficient — the other 12 appraisals are what do the differentiating. See the tension with Core affect below.
• Valence fails as the criterion for what makes an emotion ‘positive’. Challenge and compassion were rated low on pleasantness (challenge 3.90, compassion 3.67 on the 9-point scale) yet are treated as positive emotions. Tong’s conclusion canvasses and rejects every candidate criterion — valence, adaptation significance, approach motivation, mastery — and settles for none, ending on the deflationary suggestion that positive emotions may be ‘positive’ simply because they are judged desirable to experience. Directly relevant to Core affect and to what a valence-arousal primitive can and cannot carve.
• Four higher-order appraisal factors (59.9% of variance): Achievement, External Influence, Difficulty, Clarity — and Achievement dominates. PCA with varimax rotation. Achievement = goal attainment (.77), relevance (.74), agency-self (.67), control-self (.46) — high for pride, challenge, contentment, joy; low for compassion, awe, amusement. External Influence = control-others (.78), agency-others (.66), control- and agency-circumstances (.61) — high for compassion, hope, gratitude; lowest for pride. Difficulty = effort (.78), problems (.77), pleasantness (−.77) — high for challenge, compassion, hope, relief. Clarity = certainty (.66), predictability (.64). Achievement emerged as the most important factor, which Tong reads (via Lazarus et al. 1980) as the sense of accomplishment being central to differentiating positive emotions. External Influence and Difficulty reproduce factors from Smith & Ellsworth (1985).
• No cross-cultural differences in appraisal-emotion relationships. No main effect of culture (F(1,5) < .001, p = .99) and no significant culture interactions at either the omnibus or the individual-appraisal level; discriminant rates were comparable in Singapore (46.7%) and the USA (42.4%). Tong is careful: this echoes prior cross-cultural appraisal work (Scherer 1997) but on positive emotions for the first time — while cautioning that two English-speaking samples cannot support a universalist conclusion.
• Participants generally judged the traditional appraisals relevant (overall 10.59% ‘not relevant’), but unevenly. Challenge, contentment, hope, joy, pride, relief and romantic love drew few NRs (<10%); amusement, awe, compassion and serenity drew the most — and those are precisely the least-studied positive emotions in the appraisal literature, suggesting the standard appraisal set may be less apt for them.

Limitations

• The recall paradigm cannot separate lived experience from conceptual knowledge — which is the Are appraisals causes of emotion, or descriptions of it? problem, and Tong raises it himself. Participants recalled an emotion and then rated appraisals of the memory, so the appraisal-emotion correspondence could reflect ‘actual experience or culturalised conceptual knowledge’ about which appraisals go with which emotion word. Cross-cultural consistency argues against schematic knowledge being the sole driver, but Tong concedes ‘there is nothing in the current data to resolve these concerns conclusively.’ This is the crucial caveat for the wiki: the study demonstrates appraisal-emotion correspondence, which is exactly what both the causal and the descriptive readings of appraisal predict. It does not, and cannot, show that appraisals produce the emotions. (The memory-reconstruction half of this worry is later addressed by the same author’s experience-sampling study, Tong & Jia (2017), which recovers the same profiles from in-the-moment ratings — but the correspondence/causation gap remains, since EMA is also non-experimental.)
• Correlational and non-manipulated throughout. No appraisal was manipulated; nobody was assigned to appraise an event a given way. The design shows that recalled emotions carry distinct appraisal profiles, not that varying an appraisal changes the emotion.
• Single-item appraisal measures. Each of the 13 appraisals was one 9-point item (Appendix), which Tong notes ‘could imply some psychometric limitation’ — no internal-consistency estimate is possible.
• Within-participant design may inflate the differences. Rating all 13 emotions invites contrast effects; Tong argues the day-separated sessions and item randomisation mitigate this, but the differentiation the study reports is partly a within-person comparison and may overstate how separable these emotions are in the wild.
• Undergraduate samples, two cultures, both largely English-speaking. Singapore is ‘largely collectivistic’ but English-speaking and Westernised; the null culture effect cannot be read as evidence of universality. Ethnic composition within each sample was too uneven to analyse.
• Appraisal processes are partly non-conscious, so the ‘not relevant’ responses are of uncertain validity. Tong notes participants ‘might have misreported their NR responses’ — not all appraisals are available to awareness or remembered — so the relevance data measure reported relevance, not whether an appraisal occurred.
• No bodily or interoceptive measurement of any kind. This is a cognitive-appraisal study end to end; its relevance to the interoception wiki is thematic (Week 9 of the syllabus; the appraisal side of the constructionist debate) and structural (it is the primary appraisal-theory source the wiki lacked), not evidential about the body.

Connections

• Concepts: Cognitive appraisal (of emotion), Core affect, Basic emotions and natural kinds, Theory of constructed emotion
• Debates: Are appraisals causes of emotion, or descriptions of it?, Do discrete emotions map onto distinct brain regions or networks?
• Related studies: Tong & Jia (2017), Barrett (2017), Ferré et al. (2024), Lindquist et al. (2012)

The first primary appraisal-theory source in this wiki. Every prior page that touched Cognitive appraisal (of emotion) held the tradition through its critics — Friedman called Schachter–Singer’s appraisal model “exquisitely bad,” Dror treated its cognitive element as a historical retrofit, Barrett declared causal appraisal “highly doubtful.” The Are appraisals causes of emotion, or descriptions of it? debate closed on the honest admission that “the wiki has heard the case against causal appraisal four times and the case for it never,” naming this very file as the paper that might improve the balance. Here it is, read first-hand: a working appraisal theorist demonstrating that appraisals do a great deal of discriminative work.

The study asks a narrow, tractable question — how well can appraisals tell positive emotions apart, and how do the emotions differ on them? — and answers it cleanly. It does not answer the question the wiki most wanted answered, and the gap between the two is the most important thing to carry away.

What the study establishes

Ninety undergraduates recalled personal experiences of 13 positive emotions (amusement, awe, challenge, compassion, contentment, gratitude, hope, interest, joy, pride, relief, romantic love, serenity) across seven sessions, rating each on 13 appraisals (pleasantness, relevance, problems, goal attainment, three agency appraisals, three control appraisals, certainty, predictability, effort). Three results matter to the wiki:

1. The appraisals discriminate. 42.4% correct classification against 7.7% chance; 91 of 95 predicted pairwise differences confirmed at Bonferroni p = .00064. The taxonomy that Barrett grants is a good description of “how a situation is represented” is, at minimum, a good enough description to sort 13 fine-grained positive states apart. That is a real datum the constructionist critique has to accommodate, and the wiki now holds it directly rather than by report.

2. The taxonomy has a floor. Interest, joy and relief classify near chance; contentment and joy cannot be distinguished by any appraisal. The instrument that separates amusement from awe cannot separate contentment from joy — which either means an appraisal is missing or means these emotions differ only in degree. For a wiki that has spent several ingests on whether discrete emotions have distinct signatures (see Bodily sensation maps (bodily fingerprints of emotion), Lindquist et al. (2012), Theory of constructed emotion), an appraisal theorist reporting that some canonical positive emotions have no distinct appraisal fingerprint is worth noticing. It is the same shape of finding as the constructionists’, arrived at from inside the appraisal tradition.

3. The structure reduces to four factors, and “accomplishment” leads. Achievement, External Influence, Difficulty, Clarity. External Influence and Difficulty reproduce Smith & Ellsworth (1985); the novel claim is that Achievement (goal attainment + relevance + agency-self + control-self) is the most important axis differentiating positive emotions.

Why this does not close the causes-or-descriptions debate

The temptation is to file this as “the case for appraisal theory, finally.” Resist it, for a reason Tong states before the wiki has to:

The same recall method used in past studies was employed here… [it] carries the concern of whether the results reflect actual experience or culturalised conceptual knowledge.

The study measures the correspondence between recalled emotions and reported appraisals. Correspondence is predicted equally by the causal reading (appraisals generate the emotion, so of course they track it) and the descriptive reading Barrett endorses (Are appraisals causes of emotion, or descriptions of it?: categorization produces both the emotion and the appraisal-describable content, so of course they track each other). Nothing in a recall-and-rate design distinguishes the two. No appraisal was manipulated; no emotion was caused; the arrow of production is not observed anywhere in the data.

So what this paper supplies is not “the case for causal appraisal” — it is the primary source the appraisal tradition was missing entirely, demonstrating that the taxonomy is empirically rich and cross-culturally stable. That upgrades the debate from “one side is unrepresented” to “both sides are represented, and the disagreement is about the mechanism behind a correspondence both sides grant.” That is a genuine improvement in the wiki’s position, and a smaller one than the debate’s old framing might lead a reader to expect. Cross-cultural consistency does cut against pure schematic knowledge — but “not purely schematic” is a long way from “causal.”

The tension with core affect

The study presses on Core affect from the appraisal side, and the pressure runs both ways.

Against a valence-arousal primitive: 13 positive emotions, all sharing positive valence and varying arousal, are nonetheless separated by their appraisal profiles at well above chance. If feeling were just valence plus arousal, positive emotions should collapse toward a corner of the affective circumplex and be hard to tell apart — the very problem Tong opens with (positive emotions correlate highly). His answer is that appraisal structure differentiates what the affective dimensions cannot. Challenge and compassion make the point sharpest: both are negatively valenced yet count as positive emotions, so valence does not even sort the set into the right bin.

For valence’s primacy anyway: pleasantness was the single least-dispensable appraisal (0.80% “not relevant”), which is the appraisal-theory image of the wiki’s recurring finding that valence is the most fundamental affective dimension. The wiki has three interoceptive/lexical legs for a valence–arousal dissociation (Pollatos et al. (2005), Dunn et al. (2010), Ferré et al. (2024)); Tong adds, from a fourth direction, that valence is where laypeople and experts agree emotion is grounded — while insisting it is a floor, not the whole building.

The reconciliation the wiki should hold: valence/arousal may be the primitive substrate (Core affect) and under-determine which discrete emotion is experienced, with appraisal structure (or, in Barrett’s telling, conceptualization) supplying the rest. Tong and Barrett agree the affective dimensions are insufficient to individuate emotions; they disagree about what does the individuating — a learned appraisal system versus ad hoc conceptualization. That is the same fault line as Do discrete emotions map onto distinct brain regions or networks?, drawn over positive emotions.

What the wiki gains, precisely

Not a refutation of Barrett, and not proof that appraisals cause anything. What it gains is (a) a first-hand demonstration that the appraisal taxonomy carries real discriminative information about positive emotions, cross-culturally; (b) an appraisal theorist independently reporting the limits of that taxonomy (interest/joy/relief; contentment ≡ joy), which the constructionist reading can absorb; and (c) a clean statement, from inside appraisal theory, that valence cannot define “positive emotion” — a result Core affect and the whole positive-emotion literature (including the wiki’s wellbeing applications) have to reckon with. See Eddie M. W. Tong for the author and the tradition he works in.