Contribution
It asks of appraisal theory a question the field had left implicit: is a given appraisal necessary for an emotion, sufficient for it, or neither? By instructing people to recall situations in which a component was absent, as well as present, it could compute how far each component’s presence is implied by anger and implies it. For anger the answer was neither. That moved the claim from “anger goes with these appraisals”, which it confirmed, to “they are contingently associated”, which most strong statements of appraisal theory did not allow (p. 266).
Key insight
Agreement across studies on which appraisals go with which emotion does not show that any appraisal is required. Sample situations where the appraisal is absent and anger still turns up, and the “syndrome” picture, in which an emotion’s components are tightly bound to it, fails (pp. 266–267).
Assumptions
- Recall under instruction samples the space. Asking for “three unpleasant situations in which … someone else was not accountable” yields situations where the component was absent; the study 2 manipulation check supports this (p. 263).
- Self-report measures appraisal, action tendency and emotion. All are rated on 0–3 scales, or for emotion also named freely (study 2).
- Dichotomizing ratings (0–1 against 2–3, with alternatives) gives presence and absence for the necessity and sufficiency indices (pp. 259–260).
- Logical, not causal, relations. “The issue of necessity and sufficiency is addressed in logical terms only, without aiming to make causal claims”; the authors take no position on whether appraisals are elicitors or content of emotions, citing Frijda (1993) and Solomon (p. 257).
- Only unpleasant situations were sampled, and only four negative emotion families (fear, anger, shame, sadness, after Diener et al. 1995).
Key results
- Positions tested (pp. 256–257). Smith and Lazarus (1993) “consider appraisal to be a sufficient condition of emotion, as well as a necessary one”; Roseman and Smith (2001) claim “strong and invariant one-to-one relationships”; and defining emotions as “felt changes in action tendency or action readiness (Arnold, 1960; Frijda, 1986 …)” implies “a relation of equivalence between them (necessity plus sufficiency)” (p. 256). Against these, Scherer (2001), Parkinson and Reisenzein allow contingent links.
- Study 1 (pp. 257–261). 110 Flemish high-school seniors (of 119; mean age 17.5), randomly assigned to one of eight instructions (four components, present or absent; Table 1), each describing three unpleasant situations from school or work, close relations and leisure, then rating 16 emotion words (305 situations). Anger was associated with frustration, other accountability and arrogant entitlement (Table 2: regression coefficients .39, .59 and .91); other accountability and arrogant entitlement were specific to anger. The antagonistic action tendency (“you wanted to do something about the situation”) was associated with no emotion, which the authors put down to a weak wording (p. 261). Necessity and sufficiency indices were “considerably lower than 1” for every component and every emotion (p. 260) and “generally did not exceed a value of .40” (p. 261).
- Study 2 (pp. 261–265). 100 university students (mean age 20.5), ten instructions (goal obstacle, other accountability, unfairness, control, antagonistic action tendency, each present or absent; Table 3, wordings taken from earlier studies), 300 situations, an open emotion question before the fixed list, and ratings of each component. Manipulation checks: present instructions gave component ratings more than a scale point higher, p < .001 (p. 263). Anger was associated with other accountability (.15, p = .03) and unfairness (.29, p < .001), marginally with the action tendency “you wanted to express your opposition” (.12, p = .09), not with goal obstacle or control (Table 4). In free reports, anger was named in 55% of situations. Other accountability was specific to anger; unfairness also went with sadness.
- No necessity, no sufficiency (p. 265). All indices, by forced-choice or free report and by instruction or rated component, “clearly reflected lack of necessity and sufficiency of each individual component”: “participants reported a nonnegligible number of situations in which a component was present, but in which they had not experienced anger, and vice versa.”
- At least one component (p. 267). “There is, indeed, not a single reported instance in which anger occurs when all the components under study are absent.” The authors suggest anger may need some, not all, central components, or that people differ in which appraisals they need.
- Conclusion (p. 266). Relations should be “specified in terms of contingent connections … meaning that they usually co-occur, instead of in terms of necessity (or sufficiency)”; “a deterministic ‘syndrome theory’ of emotions … is no longer tenable”.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Anger is associated with a distinctive pattern of appraisals: other accountability, unfairness, and (more weakly) goal obstacle or frustration | moderate | two studies, regression and correlation; replicates earlier appraisal work |
| C2 | Other accountability is specific to anger among fear, anger, shame and sadness | moderate | both studies |
| C3 | No single appraisal studied is necessary or sufficient for anger | moderate | asymmetric phi indices well below 1 in both studies, robust to dichotomization and response format; self-report only |
| C4 | The antagonistic action tendency is neither necessary nor sufficient for anger | weak to moderate | study 2 only (study 1’s wording failed); association itself marginal |
| C5 | Anger requires at least one of the components | weak | absence of counter-instances in study 2’s sample; post hoc |
| C6 | Appraisal–emotion relations are contingent, not necessary | argued | from C3 |
Concepts
- Specificity — a component associated with one emotion and not the others studied.
- Necessity / sufficiency (asymmetric phi) — for necessity, (P(component | emotion) − P(component)) / (1 − P(component)); for sufficiency the same with the roles swapped. 1 means strict necessity or sufficiency, 0 no association (pp. 259–260).
- Contingent association — components and emotion “usually co-occur” without either implying the other (p. 266).
- Arrogant entitlement — “someone else arrogated something to him/herself which he/she had no right to”, study 1’s instance of unfairness (Table 1).
Connections
It tests appraisal theory in general, and names Frijda among those whose definition of emotion as change in action readiness implies necessity and sufficiency (p. 256). It cites the prototype approach (Shaver et al. 1987) and Russell and Fehr’s semantic analyses as consistent with no necessary features (p. 266). Parkinson’s (1997) “at best … a contingent connection” is the position it supports. Barrett’s constructionist account (THEORY-057, from LIT-542), which the record holds as the rival of Frijda’s, makes the same kind of prediction about category instances.
Bearing on the record
- THEORY-055: it weighs only on the claim that emotions differ by mode,
in a kind of evidence its
promote_whensays cannot settle it. Thepromote_whensets aside “correlations between people’s appraisal reports and their emotion labels, since the rival reads appraisals as descriptions produced by the same categorization”. This design is a refinement of that kind, not a way out of it. What it manipulates is which memories people recall (an instruction to recall situations with or without an appraisal), not the event or its pertinence to a concern; and appraisals, action tendency and emotion are all reported by the same person about the same recalled episode. The authors themselves treat necessity and sufficiency “in logical terms only, without aiming to make causal claims” (p. 257). So it is not the result the account’spromote_whenasks for (“holding the event fixed and changing only its pertinence to the person’s concerns”), nor the rival’s fatal one.- The elicitation thesis is not reached. Anger did occur where people denied goal obstacles. But every situation sampled was unpleasant, so concern pertinence may have held throughout (see Limitations), and the paper cannot show anger without it.
- The mode thesis is weighed against, weakly. Wanting to express opposition, the nearest item to Frijda’s mode of hostility, was only marginally associated with anger and was neither necessary nor sufficient for it (study 2). The paper names Frijda’s definition of emotion as change in action readiness as implying equivalence (p. 256). The action-tendency wording failed in study 1, and the finding is one self-report item.
- The one finding that leans the other way, that no anger was reported with every component absent (p. 267), is post hoc. Because the rival account also predicts instances with no common feature, this result does not tell the two apart.
- THEORY-057. Consistent with its denial of necessary features for category instances; the paper cites the prototype view in support. The same self-report limit applies.
- No instruction for machine-learning practice; nothing for the anthology.
Limitations
- Self-report of recalled episodes for every variable, so shared method variance and reconstruction in memory may link or unlink them.
- The manipulation is of recall, not of situations. Instructions shift which situations are recalled; they do not fix an event and vary an appraisal.
- Only unpleasant situations and four negative emotion families. Unpleasantness itself was required of every situation, so a concern-relevance condition may have been met throughout.
- Item wording mattered. The action tendency failed in study 1 and was marginal in study 2; goal obstacle behaved differently from frustration (p. 265). Other wordings might yield other indices.
- Group-level indices. The authors raise individual differences in which appraisals are necessary (p. 267) and do not test them.
- Figures not seen. The plotted index values in Figures 2 and 5 were not read; the text’s summary (“generally did not exceed a value of .40”, study 1) is relied on.
Open questions
- With events held fixed and appraisals manipulated (by vignette or in the laboratory), do the same components prove unnecessary for anger?
- Is “at least one component” a real disjunctive condition, or a sampling artifact of recall instructions that always named one?
- Do individuals have their own necessary appraisals for anger?