Perception of major life events and personality trait change

European Journal of Personality, advance online publication 2022, DOI 10.1177/08902070221107973. Read as the PsyArXiv preprint, version 3 (May 2022), which states that it was accepted on 24 May 2022 and is "not the copy of record" (DOI 10.31234/osf.io/kxz2u, CC BY 4.0). Ruhr University Bochum, Michigan State, Siegen. Provenance: papers/carry_on/haehner2022_event_perception.provenance.json.

What was read

All 2,174 lines of pdftotext -layout output (49 pages): title page, abstract, all sections, footnotes, references, Tables 1–11 and the captions of Figures 1–2. The figure drawings were not inspected. Not read: the supplement (Tables S1–S19: preregistration deviations, event codings, invariance, model equations, all other simple slopes, other traits for RQ6, effect sizes in other studies), the preregistrations, and the OSF data and scripts.

Question

Does how a person perceives a major life event, rather than the event's category, predict change in their Big Five traits?

Method

• Sample. What's Next? Study (Germany, 2018–19). 857 registered; 433 analysed, namely those with T1 and T2, quality checks passed, and an event dated between T1 and T2. 76% women, mean age 21.8. N at T3–T5: 343, 299, 293.
• Occasions. Five online surveys at 0, 12, 24, 36 and 48 weeks.
• Event. At T2, each person freely named the most important major life event since T1 and rated it on the Event Characteristics Questionnaire (ECQ; Luhmann et al. 2021): 37 items on nine dimensions.
  • The dimensions are valence, impact, predictability, challenge, emotional significance, change in world views, social status change, external control and extraordinariness.
  • Events were also coded into 35 fine and 7 broad categories (e.g. starting college, new friendship, relocation, new job).
• Traits. BFI-2-XS, 3 items per trait, α .45–.69.
• Analysis.
  • Latent change scores for T1→T2, T3, T4 and T5 (3, 6, 9 and 12 months), in T1 SD units.
  • Multilevel models of change (with sign, or absolute value for "amount") on interval, one ECQ dimension and their interactions. An omnibus model comparison is required before slopes are interpreted.
  • α = .05 for the preregistered RQ1–3; α = .01 for the exploratory RQ4–5.
  • RQ6 was added at review and not preregistered: trait scores over time × ECQ dimension × broad category.

Results

• Mean change (Table 3). Neuroticism fell by 0.08, 0.16, 0.17 and 0.24 T1-SD at 3, 6, 9 and 12 months. Other traits moved by 0.07 or less.
• RQ1, valence.
  • The omnibus test improved fit for agreeableness (χ²(4) = 12.49, p = .014) and neuroticism (10.80, p = .029). Extraversion, openness and conscientiousness were not significant, so H1 (valence raises extraversion) failed.
  • Agreeableness: +1 SD valence meant 0.020 SD less decline at 3 months. The slope was 0.002 at 6 months, 0.010 at 9 and 0.020 at 12.
  • Neuroticism: the slopes run −0.036, −0.009, +0.021, +0.008 from 3 to 12 months. The sign flips; no simple slope differs from zero, so H2 (valence lowers neuroticism) failed.
• RQ2–3, impact.
  • Perceived impact did not predict the amount of change for any trait (all p > .085).
  • Impact × valence predicted nothing (all p > .091).
  • This is the social-investment hypothesis (role change drives trait change), and it failed.
• RQ4. No other ECQ dimension predicted the direction of change at α = .01 (0 of 35).
• RQ5, amount of change. 4 of 35 tests reach α = .01:
  • world views → agreeableness and external control → agreeableness, both −0.02 at 12 months;
  • predictability → neuroticism (−0.054 at 9 months);
  • predictability → openness (+0.016 at 3 months).
• RQ6 (neuroticism only; other traits in the supplement). Time × category × social status change (p = .004) and time × category × valence (p = .002).
  • Example: an educational event perceived as status-threatening (+1 SD) goes with −0.31 SD neuroticism over a year. A health event perceived the same way goes with +0.47 SD.
• Overall. The largest per-SD effect is about 0.05 SD of trait change. The authors call this about as large as checklist-based event effects (e.g. Denissen et al. 2019). It is about half the size of ECQ effects on subjective well-being (Luhmann et al. 2021). Every association depended on the interval.

Limits

• No no-event group. Every analysed person had an event, so the neuroticism decline (young-adult maturation) cannot be separated from event effects. Only differences between perceptions are identified.
• Perception was rated at T2, alongside T2 personality. State or trait change could colour the rating. The paper does not discuss this reverse path.
• Measures and sample. Three-item trait scales with low α; young, educated, 76% female German sample; the most-important, mostly positive and normative events; attrition to 293 by T5.
• Many tests. 5 omnibus tests at .05 (2 significant), 70 exploratory at .01 (4 significant), plus RQ6. The preregistered hypotheses all failed as stated.
• Inconsistencies found:
  • The text credits "the significant cross-level interactions" for the agreeableness pattern. In Table 5 only Time_6M × valence is significant (p = .040); 9M and 12M are not.
  • Figure 2A's caption says positive valence gave "smaller decreases" in neuroticism over 12 months. The 12-month simple slope (+0.008) is not different from zero.
  • The author note gives the analysis preregistration as osf.io/783px. The method gives osf.io/u5m7v. Blinding placeholders ("[INSTITUTION BLINDED FOR PEER REVIEW]") remain in the accepted version.
  • Arithmetic checked: the Table 4 simple slopes equal the Table 5 main effects plus interactions, and the Table 3 d values equal the Table 5 intercepts plus time effects (e.g. neuroticism 12 months −0.079 − 0.165 = −0.244 against −0.24).

What it means for Kurisutina

• Question 1: the same event is not the same experience.
  • The paper's premise, citing Kandler & Ostendorf (2016), is that people rate e.g. separation across the whole valence range. Perceived valence, not the event, carried what little signal there was.
  • For a replica this means reacting like Alice starts with appraising a new event as Alice would. A slot that maps an event category to a reaction skips the step where individual differences first enter.
• A cheap, direct test of "reacts like her".
  • The ECQ gives nine rated dimensions per event.
  • With a real participant, collect their ECQ ratings of new events. Have the replica, given the participant's slot, rate the same events. Compare with the empty-slot and other-slot ratings.
  • This measures appraisal agreement without waiting months for trait change, and it has a population baseline built in. Proposal only, not started.
• Big Five scores are the wrong outcome for single events. Effects are about 0.05 SD per SD of perception, below the noise of short scales. Subjective well-being responds about twice as strongly. A replica's self-rated traits should barely move after one event. Its well-being and affect answers are where carrying on will show. That matches Anusic, Kettlewell and Luhmann 2012.
• Timing again. Signs flip between 3 and 12 months (Luhmann et al. 2014). Any test of a replica's change after an event needs fixed, declared intervals.
• GSS pilot and LISS. Neither dataset records how the person perceived the event. That is one unmeasured source of individual difference that no slot can recover from these records. It lowers the ceiling for the own-slot advantage in both.

Cross-references

• summaries/carry_on/luhmann2014_its_about_time.md: time-interval dependence, non-events.
• summaries/carry_on/luhmann2012_adaptation.md: event effects on subjective well-being.
• summaries/carry_on/yap2012_personality_moderation.md, summaries/carry_on/anusic2014_hilda_personality.md: traits as moderators (the reverse direction).
• summaries/carry_on/sliwinski2009_stress_bursts.md: reactivity as a partly stable personal slope.
• docs/research/liss_q1_design.md, docs/research/gss_pilot_design.md.