Kurisutina
Library/Carrying on

Kritzler, Rakhshani, Terwiel, Fassbender, Donnellan, Lucas & Luhmann (2021/2022). How are common major life events perceived? Exploring differences between and variability of different typical event profiles and raters

Published in European Journal of Personality (2022), DOI 10.1177/08902070221076586, which is closed access. Read as the PsyArXiv preprint, version 1 (May 2021), before peer review (DOI 10.31234/osf.io/fncz3, CC BY 4.0). Numbers and wording may differ in the published version. Ruhr University Bochum, Michigan State, Siegen. Provenance: papers/carry_on/kritzler2022_event_perception_profiles.provenance.json.

What was read

All 1,836 lines of pdftotext -layout output (46 pages): abstract, all sections, footnotes, data statement, references, Tables 1–3 and the captions of Figures 1–5. The figures, which hold the profiles and the individual lines, were not inspected. Not read: the online supplement with per-event SDs and correlates, the Shiny app, the preregistration, and the OSF data and scripts (osf.io/eyu4p).

Questions

  1. What is the typical profile of perceived characteristics for common major life events?
  2. How much do perceptions vary within the same event?
  3. Do outsiders' ("ex situ") ratings match those of people who lived the event ("in situ")?

Method

  • Instrument. The Event Characteristics Questionnaire (ECQ), nine dimensions on 1–5: challenge, change in world views, emotional significance, external control, extraordinariness, impact, predictability, social status change, valence.
  • Ten events: becoming a parent, breakup, death of a close family member, graduation, marriage, new friendship, new job, new relationship, serious illness or injury, vacation.
  • Samples. Six samples, 2,210 people and 3,555 ratings, mean age 22.9, 72% students.
    • In situ (Samples 1–4): 778 people, 1,001 ratings. Each freely named an event from the last three months, which was coded into the ten types (κ = .84). Earlier ECQ versions were used. n per event ranges from 23 (marriage) to 214 (vacation).
    • Individual ex situ (Sample 5): 981 North American students, 1,201 ratings of how they would experience two randomly assigned events they had never had. Breakup, new relationship and vacation were not included.
    • Generalized ex situ (Sample 6): 451 people, 1,353 ratings of how young adults typically experience three randomly assigned events.
    • The ex situ samples used adapted ECQ short forms.

Results

  • Profiles. Each event has a distinct average profile.

    • Breakup: high emotional significance, challenge and impact; negative.
    • Vacation: positive and predictable; low challenge and impact.
    • New friendship and new relationship are nearly identical.

  • Common core. All ten events rate high on emotional significance (means of the event means 4.27) and impact (3.86), and low on external control (2.45) and extraordinariness (2.48). This is partly a consequence of the definition and the instructions.

  • Dimensions that separate events. Valence (SD of event means 1.42, range 1.28–4.75), challenge (0.79) and predictability (0.66).

  • Within-event variability.

    • By event, aggregated over dimensions: SD 0.70 for new friendship up to 0.91 for serious illness.
    • By dimension, aggregated over events: valence varies least (SD 0.56), predictability most (1.15).
    • By event and dimension: the valence of a death in the family and of a new friendship (SD 0.39 each) against the valence of a breakup (SD 0.90). Some people rated a breakup "very negative", others "moderately positive".

  • My rough calculation, not the paper's. Comparing the SD of event means with the pooled within-event SD:

    • valence is about 87% between event types;
    • predictability is about 25% between event types.

    The figures are unweighted over unequal n and only indicative.

  • Correlates (age, weeks since the event): near zero overall (r = .04 and .02), with scattered event-specific correlations.

    • Older participants rated a breakup and a new friendship as more extraordinary.
    • With time since the event, a new friendship was rated more positive and more impactful, and a serious illness more challenging and emotionally significant.

  • Outsiders against those who lived it.

    • Individual and generalized ex situ ratings did not differ overall (b = 0.02, p = .26). Imagining "how would I feel" gave the same average as "how do people feel".
    • A multilevel model found zero variance due to ex situ raters.
    • Ex situ ratings were slightly higher overall (b = 0.13). They overstate change in world views (+0.83), social status change (+0.39), challenge (+0.28) and external control (+0.18), and understate extraordinariness (−0.66), valence (−0.21, i.e. less positive) and emotional significance (−0.12). All are Bonferroni-significant.
    • Predictability matched (0.00). The authors call most differences small (η²G ≤ .039) and the profiles "very similar".

Limits

  • Cross-sectional; young, mostly student, WEIRD samples; self-selected, memorable events; post-hoc coding of free text into broad types, which by the authors' account inflates within-type variance.
  • Person and instance cannot be separated (the authors say so). Within-event variance mixes differences between people with differences between instances of the event (two breakups are not the same breakup) and measurement error.
  • In situ and ex situ samples differ in country, age, instructions and ECQ version (earlier item sets against adapted short forms). The in/ex situ contrasts confound rater type with instrument and sample. The paper notes the differing items but does not treat this as a limitation of the comparison.
  • Judgements of mine:
    • Calling a +0.83 difference on a 1–5 scale "small" is generous when within-event SDs are about 0.6–1.2.
    • The individual-versus-generalized comparison is between different samples at the aggregate level. It does not show whether a person's own forecast predicts their own later experience.

What it means for Kurisutina

  • Question 1: where individual appraisal can matter.
    • Valence is mostly fixed by the event type (rough share 87%), except for ambiguous events such as a breakup. Predictability is mostly individual or instance-specific.
    • Haehner et al. (2022) found valence the only dimension predicting the direction of trait change.
    • Together: for unambiguous events (a death in the family) a category-level model will get the appraisal nearly right, and there is little room for "reacting like her". Individuality in appraisal is largest for ambiguous events: breakups, and by extension job changes, moves, separations.
    • Tests of whether a replica reacts like its person should oversample such events, or they will measure mostly the population profile.
  • A replica imagining an event is an ex situ rater. Humans who have not lived an event overstate its challenge, world-view change and status effects, and understate how unusual it feels.
    • A replica asked about an event it has only been told about may show the same forecasting biases, or the base model's version of them. The person's in situ ratings are the target.
    • That people's "how would I feel" matched "how do people feel" on average warns that even humans default to the stereotype when imagining. An own-slot replica must beat the generalized profile, not just match the mean.
  • Concrete baseline for the ECQ test proposed in the haehner2022 summary.
    • The typical in situ and ex situ profiles here are a ready population baseline, the counterpart of the population tables in the GSS pilot.
    • Score a replica's appraisal of a person's new event against three references: the person's own rating, the typical profile, and the empty-slot replica.
    • The OSF data (osf.io/eyu4p) would supply the baseline; its licence was not checked. Proposal only.

Cross-references

  • summaries/carry_on/haehner2022_event_perception.md: perceived valence and trait change; the ECQ test proposal.
  • summaries/carry_on/luhmann2014_its_about_time.md: minor events, non-events.
  • summaries/carry_on/schacter2011_adaptive_distortion.md: imagining and remembering share machinery (ex situ forecasts).
  • docs/research/gss_pilot_design.md: population tables as the baseline to beat.

This summary is our record of the paper, written after reading the full text and published as written; links into our own repository have been removed.