Interoception Wiki

← RADlab

Positive Emotion, Appraisal, and the Role of Appraisal Overlap in Positive Emotion Co-Occurrence

11 min read

Study at a glance

  • Authors: Eddie M. W. Tong, Lile Jia
  • Year: 2017
  • Journal: Emotion
  • DOI: 10.1037/emo0000203
  • N: 395
  • Design: Ecological Momentary Assessment (experience sampling). 395 National University of Singapore undergraduates carried Palm handhelds that signalled at ~half-hour intervals over two weekdays (up to 36 observations each), yielding 11,141 responses (21.36% nonresponse). At each beep participants picked the event that had most engaged their attention in the last 15 min and rated 12 positive emotions (amusement, awe, challenge, contentment, gratitude, hope, interest, joy, love, pride, relief, serenity) and 13 single-item appraisals (7-point scales, with a per-item ‘not applicable’ opt-out). Analysis: hierarchical linear modelling for within-person emotion–appraisal associations (156 models); Euclidean-distance appraisal-overlap indices between emotion pairs (Kuppens et al. 2004; Pe & Kuppens 2012); paired-sample t tests on Fisher-normalised overlap and co-occurrence indices; false-discovery-rate correction throughout (Benjamini & Yekutieli 2001).

Key findings

  • The appraisal taxonomy replicates in daily life, not just in recall — which is what the wiki most wanted from this file. 61 of 75 a-priori predictions were supported in the zero-order HLM analyses (55 of 75 in the more conservative simultaneous regressions), cross-culturally consistent with Western findings. This is Tong (2015)‘s recall result carried into experience sampling, addressing that study’s headline limitation — that recall ‘invites concerns over memory biases and schematic responses.’ The appraisal profiles are not a memory artefact; they show up when appraisals and emotions are rated ‘online’ in real events.
  • Appraisal overlap predicts positive-emotion co-occurrence — the paper’s central new claim. Positive emotions co-occur substantially (pairwise r = .15–.56). Using a hypothesis-driven split of each emotion pair’s shared appraisals into predicted overlapping appraisals (POA) versus nonpredicted (NOA), the paper shows (Postulate 1) that emotion pairs overlap more strongly on POA than NOA, t(63) = 3.02, p = .004, η² = .13; and (Postulate 2) that co-occurrence correlates more strongly with POA overlap than NOA overlap, t(62) = 3.62, p < .001, η² = .17. Emotions co-occur to the degree they share the theoretically relevant appraisals — see Appraisal overlap.
  • Positive emotions do not co-occur only because they are all pleasant — but pleasantness is overwhelmingly the strongest single tie. Dropping pleasantness, POA overlap still beats NOA overlap (Postulate 3, t(44) = 6.28, p < .001, η² = .47) and still predicts co-occurrence better than NOA (Postulate 4, t(44) = 2.41, p = .012, η² = .12). So nonpleasantness appraisals carry real co-occurrence structure. Yet pleasantness alone overlaps far harder than every other appraisal (dPleasantness < dNOA, η² = .94; < dPOA-minus-pleasantness, η² = .92) — ‘the central unifying denominator underlying positive emotions.’ The same dual verdict as Tong (2015): valence is the floor and not the building. Directly relevant to Core affect.
  • Contentment, joy and serenity are near-indistinguishable — but understandability finally splits joy off, refining the Tong 2015 ‘contentment ≡ joy’ result. The three co-occur highly and share nearly identical appraisal profiles, except that joy is tied to understandability (joyful people think they grasp the situation) while contentment and serenity are not. Tong (2015) had reported contentment and joy as separable by no appraisal at all; the EMA data locate one differentiator. Tong calls it ‘a novel finding’ and still concedes more work is needed — the emotions may ‘differ not qualitatively but quantitatively.’
  • Joy misbehaves in the overlap analysis and the authors say they don’t know why. For joy, only 3 of 11 POA-vs-NOA comparisons went the predicted way (‘the reason for these results is unknown’); awe and contentment supported only 6 of 11. Many predictions for relief also failed. The aggregate postulate tests hold, but the emotion-by-emotion picture is uneven, and the paper is candid about it.
  • Several appraisal surprises, reported honestly rather than smoothed over. Hope was positively related to control-self (predicted negative) and to predictability (predicted negative) — read post hoc via Snyder’s hope theory. Gratitude was positively related to agency-self (‘it is unclear why’). Agency-self drew a 50.71% ‘not applicable’ rate, which the authors read as a collectivistic reluctance to credit the self for good events — while control-self drew only 11.46%, so the inhibition is about credit, not control.
  • Challenge is the one positive emotion negatively related to pleasantness — so valence cannot even sort the set into the right bin. Challenge tracks problems, effort and consideration, not pleasantness, and does not co-occur with the highly pleasant emotions (amusement, contentment, joy, serenity). The authors use this to reopen the question Tong (2015) left unresolved: what makes an emotion ‘positive’ if not hedonic value? They float eudaimonic value (growth, social benefit) as a criterion beyond pleasantness, and settle nothing — the same open end as the earlier paper.

The experience-sampling companion to Tong (2015), and the file the Are appraisals causes of emotion, or descriptions of it? debate page named as still pending — “one further appraisal paper remains in raw/… on present evidence it is unlikely to supply the missing manipulation either.” That prediction is confirmed. What the paper does supply is the thing the 2015 recall study could not: the appraisal profiles of positive emotions, captured in daily life as events happen, rather than reconstructed from memory.

(Filed as 2017 per the Emotion volume — Vol. 17, No. 1, 40–54 — though published Online First 11 July 2016; the same journal-volume convention the wiki used for Barrett (2017). Received Oct 2014, accepted May 2016. Second author Lile Jia, also NUS, is not given his own page — Tong is the appraisal specialist and lead; see Eddie M. W. Tong.)

What is new here relative to Tong (2015)

Three things, in order of importance to the wiki.

1. The taxonomy survives the move out of the memory. Tong (2015)‘s single largest flagged limitation was that its recall-and-rate paradigm ‘invites concerns over memory biases and schematic responses’ — participants rated appraisals of a remembered emotion, so the correspondence could be ‘culturalised conceptual knowledge’ rather than lived experience. This study answers that specific worry with method: appraisals and emotions are rated within 15 minutes of a real event, sampled quasi-randomly across two ordinary days. 61 of 75 predictions replicate. The appraisal–emotion correspondence is not an artefact of reconstructing memories.

Be precise about how much this buys. It removes the memory-reconstruction component of the schematic-knowledge concern. It does not remove the deeper concern — that even an in-the-moment rating of “how much joy did you feel / how pleasant was this event” could apply the same learned associations. And it supplies no manipulation, so it says nothing about whether the appraisal produced the emotion. See the debate section below.

2. Appraisal overlap explains why positive emotions co-occur. This is the paper’s title contribution and its genuinely novel construct — Appraisal overlap. Positive emotions are notoriously entangled (they correlate far more than negative emotions do); the standard explanation is shared valence. Tong & Jia show, with an a-priori hypothesis-driven design, that two positive emotions co-occur to the extent they share their theoretically predicted appraisals — and that this holds even after pleasantness is stripped out, so co-occurrence is not merely a valence effect. See Appraisal overlap for the POA/NOA machinery and the four postulates.

3. A differentiator for contentment and joy. Tong (2015) reported that contentment and joy could be separated by no appraisal at all — the taxonomy’s clearest failure. Here the EMA data recover one: understandability. Joyful people rate the situation as understood; contented and serene people do not necessarily. It is a single thread, and Tong still allows that these emotions may differ only in intensity — but it is a real refinement of the earlier null, arrived at with a better method.

Why it still does not close the causes-or-descriptions debate

The debate page predicted this file would not supply the missing manipulation, and it does not. The paper is explicit — more explicit than Tong (2015):

Because appraisals can influence emotions and emotions can also shape appraisals, and also because the current study employed a nonexperimental paradigm, this article does not take a stand on the causal directionality between appraisals and emotions.

That sentence concedes the whole causal question and even grants bidirectionality (citing evidence that emotions bias subsequent appraisals). What the study demonstrates is real-time correspondence between appraisal and emotion — which Barrett’s descriptive reading (categorization produces both at once) predicts exactly as well as the causal reading (appraisals generate the emotion). EMA improves the quality of the correspondence evidence — it is now naturalistic, not remembered — without touching the mechanism dispute. The wiki’s position on causation is unchanged; its confidence that the appraisal taxonomy is empirically real, not a recall artefact, goes up. See Are appraisals causes of emotion, or descriptions of it?.

The valence thread

The paper lands squarely on the wiki’s recurring valence-primacy finding, and lands on both sides of it, as Tong (2015) did.

For valence’s primacy: pleasantness is the appraisal on which positive emotions overlap most massively — dPleasantness is smaller than the overlap on all nonpredicted appraisals (η² = .94) and even smaller than the overlap on all other predicted appraisals combined (η² = .92). Those are enormous effects. Pleasantness is ‘the central unifying denominator underlying positive emotions.’

Against valence’s sufficiency: with pleasantness removed, the remaining predicted appraisals still carry co-occurrence structure (Postulates 3 and 4). Positive emotions are not merely a bag of pleasant states; they share specific nonpleasant appraisals, and those predict which ones travel together. And challenge — negatively related to pleasantness yet counted a positive emotion — shows valence cannot even assign the right sign.

This is the appraisal-theory image of the split Core affect documents from the interoceptive side: valence is where affect is grounded (the least-dispensable dimension) and not what individuates the emotions built on it. Held together with Tong (2015), the wiki now has two first-hand appraisal-theory sources delivering the identical dual message about valence, from recall and from experience sampling.

What the wiki gains, precisely

A second primary appraisal-theory source; the naturalistic replication that retires Tong (2015)‘s memory-bias caveat; a new construct (Appraisal overlap) linking appraisal structure to emotion co-occurrence; a refinement of the contentment≡joy null; and one more independent affirmation that pleasantness is the strongest but not the only organiser of positive emotion. What it does not gain — again — is any evidence that appraisals cause emotions, or any measurement of the body. See Eddie M. W. Tong for the author and the tradition.

Graph View

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

Backlinks