Awe or horror: differentiating two emotional responses to schema incongruence

ABSTRACT

Experiences that contradict one's core concepts (e.g. of the world, people, the self) elicit intense emotions. Such schema incongruence can elicit awe, wherein experiences that are too vast to understand with existing cognitive schemata cause one to feel that schemata should be updated [i.e. a “need for cognitive accommodation” (NFA); Keltner & Haidt, 2003. Approaching awe, a moral, spiritual, and aesthetic emotion. Cognition and Emotion, 17(3), 297–314]. However, other emotional responses to schema incongruence, such as horror, have not been investigated. The current studies compared awe and horror to investigate if they are distinct emotional responses to schema incongruence. Study 1 observed significant differences between awe and horror in cognitive appraisals (e.g. certainty, legitimacy), indicating several areas of dissimilarity. Study 2 found evidence that awe and horror are both responses to schema incongruence, as schema incongruence and NFA were salient in awe and horror, but not a contrast emotion. However, awe and horror were elicited by different types of schema incongruence: awe by spiritual vastness, horror by extremity. Awe-eliciting experiences also appeared to be easier to assimilate than horrifying experiences, as NFA and uncertainty were significantly lower in awe than in horror. Differences in the functions of horror and awe are also discussed.

KEYWORDS:

• Awe
• horror
• emotional appraisals
• cognitive schema

Disclosure statement

No potential conflict of interest was reported by the authors.

ORCID

Pamela Marie Taylor http://orcid.org/0000-0002-7528-1738

Notes

1 Previous psychological research on horror has been limited to “art horror” (e.g. psychological responses to cinematic fiction), which is different from “natural horror” towards real-life events (Carroll, 1990).

2 However, trauma and horror are not equivalent. Horror is one potential etiological factor in PTSD (Dalgleish & Power, 2004), but most horrifying events are not traumatic.

3 Because we use the terms “awe” and “horror” to denote three different things (study conditions, Likert emotion ratings, emotions), we use the following conventions to distinguish between them: study conditions are capitalised (Awe, Horror), emotion ratings are italicised and lower-case (awe, horror) and emotions (in discussion, not the data) are unitalicised and lower-case (awe, horror).

4 Data was collected before publication of Gordon et al. (2017), so their findings on “certainty” appraisals did not inform our hypotheses.

5 Because these strict exclusion criteria caused many participants to be excluded, we separately analyzed data from all participants who missed ≤ 2 attention check questions (N = 212). All results were consistent with those reported here, with no differences in p-values.

6 Prior to data collection, sample sizes were determined using Cohen (1992)'s rule of thumb for medium effect sizes for .05 alpha and doubling it to obviate inattentiveness of online participants.

7 Note that in Study 1, horror was defined as “vast” for participants, which is inconsistent with Study 2's results. When Study 1 was conducted, we assumed schema incongruence in both awe and horror was “vastness”. Study 2 tested this assumption, but found no support for it.

8 Study 1 was part of a larger study that included variables not reported here.

9 Four items were split into two items each: Anticipated Effort-effort (“mental effort”, “physical effort”), Situation-Control (“fate/destiny was in control”, “chance/coincidence was in control”), Situation-Responsibility (“fate/destiny was responsible”, “chance/coincidence was responsible”) and Legitimacy-cheated (“you were cheated”, “someone else was cheated”).

10 Similar to Gordon et al. (2017), the reliability for Personal Agency was low, so separate t-tests were conducted on “personal control” and “personal responsibility”. For both individual and composite variables, ratings in Awe were significantly higher than in Horror (ts ≤ 3.15, ps = .002).

11 We separately analyzed data from all participants (N = 181). All results were entirely consistent with those reported, including EFA item loadings and ANOVA results (both omnibus and pairwise).

12 Sample size was based on Everitt (1975), stipulating a N:p ratio of 10.

13 To discriminate elicitation by a person from elicitation by a person's actions, “another person” was defined as elicitation by a person's traits or the fact that such a person exists; elicitation by a person's actions were coded as “events”.

14 “Information” included hearing/reading something second-hand or discovering first-hand evidence (rather than directly witnessing).

15 Due to a database error, twelve horror scenarios were not coded.

16 The two-coder alpha for the “event” category was 0.59, so a third coder independently coded all scenarios for this theme; the three-coder alpha was .69.

17 KMO = .835, Bartlett's χ²(91) = 2338, p < .001. Eigenvalues and parallel analysis in SPSS (O’Connor, 2000) indicated two factors. Four items were removed: social significance and personal significance (failed to load), power (double-loaded) and physical size/number (communality < .20). The two factors showed good discriminant validity, r = .17 (i.e. shared 2.9% of variation).

18 KMO = .877 Bartlett's χ²(45) = 2235, p < .001. Eigenvalues and parallel analysis indicated two factors. Three items were removed: absorbed (communality < .20), shocked (cross-loaded), easy to understand (cross-loaded). Amazement cross-loaded, but was retained because the loading difference was almost .40. The two factors were negatively correlated (−.49), sharing 24% of variance.