Muir, Madill & Brown (2022/2023). Reflective rumination mediates the effects of neuroticism upon the fading affect bias in autobiographical memory

Self and Identity 22(1): 102–128 (2023; online 14 February 2022), DOI 10.1080/15298868.2022.2041080. The published version was not read. Read as the authors' accepted manuscript from White Rose Research Online (48 pages; "© 2022 Informa UK Limited … uploaded in accordance with the publisher's self-archiving policy"; all rights reserved). Provenance: papers/carry_on/muir2022_neuroticism_fab.provenance.json.

What was read

All 2,242 lines of pdftotext -layout output: repository cover, abstract, three studies, general discussion, footnotes i–vii, references and the captions of Figures 1–4. The figures (bar charts) were not inspected.

Question

Is the fading affect bias (FAB) smaller in more neurotic people? Is that explained by how often, and why, they privately rehearse unpleasant memories?

Method

• Three correlational studies with one design. Each person recalls three pleasant and three unpleasant events from the last 12 months (not the last week), and rates how intense the emotion was then and is now in one session. Fading = then − now.

  • Neuroticism: Eysenck's EPQR-S, a 12-item yes/no scale.
  • For each event: how often it was privately rehearsed (one item, 1–7).

• Samples and additions:

  Study	Sample	Year	Events analysed	Additions
  1	174 students and staff	2010–11	972	Unipolar intensity scale
  2	168 students, 91% female	2014	959	Bipolar scale; six reasons for rehearsal (Ritchie et al. 2006)
  3	141 UK adults (Prolific), aged 18–76	2019	804	Event-level brooding (adapted BSRI) and reflection (adapted RRS)

• Analysis. PROCESS models with a person dummy and event age:

  • moderation: valence × neuroticism on fading;
  • mediation: neuroticism → rehearsal or rumination → fading, run separately for pleasant and unpleasant events.
  • Only significant results are reported "for clarity and space".

Results

• A FAB in every study. Unpleasant against pleasant fading:

  Study	Unpleasant	Pleasant
  1	1.57	0.68
  2	0.95	0.42
  3	0.71	0.31

• Neuroticism barely moderates it.

  • The valence × neuroticism interaction is significant only in Study 2 (p = .009). Studies 1 and 3 give p = .60 and .37.
  • Pooled over studies (footnote vi): p = .003, R² change = .003. More neuroticism goes with less fading of negative affect (b = −.04), and no change for positive affect.

• Rehearsal is the consistent correlate. In all three studies, more private rehearsal of an event goes with less fading of its affect:

  • unpleasant events: b −.08 to −.29;
  • pleasant events: b −.11 to −.13.

  Neuroticism predicts more rehearsal of unpleasant events (b .05–.07), and the indirect effect is significant in all three studies.

• Which rehearsal. Of six reasons for rehearsal, only "to reflect on the meaning of the event or to better understand it" mediated, in Studies 2 and 3.

  • In Study 3, event-level reflective rumination mediated (b = .08 on neuroticism, −.11 on fading).
  • Brooding did not predict fading.

• Interpretation, left open by the authors. Reflection might be maladaptive in neurotic people, or a slower route to eventual relief. No outcome data can decide between them.

Limits

• Correlational, with retrospective ratings of "then" and "now" in one sitting.
• Rehearsal is measured with single self-report items.
• Mostly female samples. Only significant results are reported, with no correction for testing six rehearsal types.
• Inconsistencies found:

  • The Study 3 discussion says "we did not find any relationship between neuroticism and frequency of brooding". Footnote v reports neuroticism predicting brooding over unpleasant events, b = .13, t(393) = 5.00, p < .001.

  • Several confidence intervals exclude their own estimates:

    • Study 1, pleasant indirect effect: b = .006, CI [−.01, −.008];
    • footnote vi, the pooled main effect: b = −.006, CI [−.02, −.008], reported as n.s.;
    • Study 2 reflection: b = −.02, CI [.04, .01].

    In Study 2, several slopes carry opposite signs in b and t (e.g. b = −.14, t = 5.56).

  • Study 2 reports its descriptive "S.D." as .05 and .04 for fading scores that range over several points. These are probably standard errors.

  • Pooled sample: "n = 483" participants and "n = 2746" events.

    • 483 = 174 + 168 + 141 includes the 12 Study 1 participants who gave no neuroticism score.
    • 2,746 = 972 + 970 + 804 uses Study 2's count before its 11 valence-flipped events were removed (959).

  • Study 3's main effect of neuroticism is reported with t(2746), the pooled df.

  • Some p-values are too large for their t: t(393) = 3.71 given as p = .002, and t(393) = −3.84 as p = .001. Both are .00024 and .00014 (my computation).

• Checked:
  • Event counts: 522 + 522 = 1,044 in Study 1; 490 + 480 = 970 in Study 2; 412 + 396 = 808 in Study 3.
  • The pleasant and unpleasant df in the pooled footnote (1,385 + 1,357 = 2,742) match its F(1, 2742).

What it means for Kurisutina

• Q3: how often a memory is revisited governs how much of its feeling it keeps.
  • Across three samples, events a person privately rehearsed more kept more of their emotional intensity, for unpleasant and pleasant events alike.
  • For a replica, retrieval is rehearsal. Fading of affect should depend on how often the replica has retrieved that memory and why, not only on elapsed time.
  • The rehearsal count is a natural field in the memory store (proposal).
• A trait alone is a weak parameter for this. Neuroticism explains almost none of the variance in the FAB (R² change .003 pooled). What differs between people is their rehearsal pattern: which unpleasant events they return to, and why.
  • The slot should hold the person's rehearsal habits, e.g. elicited per event ("how often have you thought about this, and why"). A Big Five score would not do the job.
• Reflection is not automatically healing. In more neurotic people, rehearsing to understand an unpleasant event went with keeping its negative feeling.
  • A replica design that has the model "reflect" on stored negative episodes (generative-agents style) would therefore not necessarily soften them.
  • Whether it should depends on the person. This complements Speer et al., where positive meaning-finding softened memories.
• Design caution for Q3 baselines. All FAB evidence read so far rests on one-session retrospective ratings, except Ritchie's Sample 10.
  • Human reference values for a replica's affect fading need a prospective design, rate then re-rate, from the person themself.

Cross-references

• summaries/carry_on/ritchie2015_fading_affect_pancultural.md: the FAB across ten samples.
• summaries/carry_on/speer2021_positive_meaning_memory.md: positive reappraisal at recall updates memories.
• summaries/carry_on/zhong2023_memorybank.md: recall-strengthened forgetting (content, not affect).
• summaries/carry_on/park2023_generative_agents.md: reflection as a memory operation in LLM agents.