Who is susceptible in three false memory tasks?

Memory 27(7): 962–984, DOI 10.1080/09658211.2019.1611862. Received 22 January 2019, accepted 16 April 2019. The file is the published version (Taylor & Francis), posted on E. F. Loftus's faculty site at UC Irvine; not open access. Nanyang Technological University and UC Irvine. Empirical, peer-reviewed. Provenance: papers/carry_on/nichols2019_false_memory_tasks.provenance.json.

What was read

All 1,459 lines of the pdftotext -layout text of the 24-page PDF: the article, Tables 1–8, the figure captions, the data statement and all references. Figures 1–2 (heard and performed judgments by condition) are images; their captions and the statistics reported in the text were read. Patihis, Frenda & Loftus (2018), the companion study cited here, has no openly accessible full text and was not read.

Question

If a person develops a false memory in one paradigm, are they likely to develop one in a different paradigm? And do personality, cognitive or other individual measures predict who is susceptible?

Method

• Sample. 373 undergraduates in California (277 women; mean age 20.8). Completed: misinformation 297, DRM 283, imagination inflation 244, individual-difference measures 367. Three lab sessions a week apart (Table 1).
• Misinformation. A 6 min 28 s silent video of an electrician stealing. A narrative later misstated four of eight critical items (for example Pepsi for Coke), counterbalanced. Test: 20 two-alternative forced-choice items, with source judgments (video, narrative, both, neither). A "robust" false memory is the misinformation item chosen and attributed to the video.
• DRM. 15 lists; a 120-item recognition test with confidence ratings. Items judged old were marked remember or know; a "robust" false memory is a lure given a remember judgment.
• Imagination inflation (after Goff & Roediger).
  • Session 1: 54 action statements, each performed, imagined or only heard.
  • Session 2: statements imagined 0, 1 or 5 times, including never-heard ones.
  • Session 3: tested whether each statement was heard in Session 1 and, if so, whether it was performed.
• Individual measures. Demographics; Big Five (BFI); cognitive ability (Over-claiming Questionnaire); Need for Cognition; Faith in Intuition; the Cognitive Reflection Test; imagery vividness (VVIQ); dissociation (DES); absorption (TAS); fantasy proneness (CEQ); religiosity; anomalous experiences (AEI).

Results

• Everyone was susceptible.
  • Misinformation: 8 of 297 endorsed no misinformation item. Accuracy was .77 on control items against .40 on misled items (d = 1.55), with equal confidence (4.25 vs 4.19). Robust false memories: .42 for misled items against .15 for controls.
  • DRM: critical lures were recognised at .63, studied items at .68, distractors at .30. Lures and studied items were "robustly" remembered at the same rate (.20 each). Internal consistency α = .67 for lures.
  • Imagination: only one other participant never judged an imagined, unheard statement as heard. Claims of having performed a statement never heard: 3.01% with no imagining, 9.16% after one imagining, 15.90% after five. Confidence did not vary with the number of imaginings.
• Little transfer between paradigms.
  • Every significant cross-paradigm correlation was positive but small: misinformation–DRM r = .14 and .17; misinformation–imagination r = .13–.15; DRM–imagination r = .13–.19.
  • Most were null. For example, 21 of the 24 misinformation–imagination correlations were not significant.
  • This fits earlier null or weak findings: Ost et al. 2013; Calvillo & Parong 2016; and the two earlier three-paradigm studies, Bernstein et al. 2018 and Patihis et al. 2018.
• Confidence transfers more than errors do. Confidence for DRM lures correlated r = .59 with confidence for imagined, unheard items, and r = .19 between misinformation and imagined items. The authors note the DRM and imagination tests shared a session and a yes/no format.
• Individual measures barely predict false memory.
  • The 19 predictors together explained no significant variance in misinformation false memories (p = .61), DRM lures (p = .09) or imagination false memories (p = .33).
  • Misinformation had no significant correlate at all; reported links with imagery, agreeableness, openness and dissociation did not replicate.
  • DRM: Cognitive Reflection Test intuitive answers r = .25; correct answers r = −.20.
  • Imagination: agreeableness r = −.14; over-claiming bias −.14 / −.18; dissociation .13 / .20; absorption .15; anomalous experiences .13.
  • The only predictor shared across paradigms was anomalous experiences (r = .13, for DRM robust lures and imagined "performed" judgments).
• Within one paradigm, susceptibility is stable (cited, not measured here): DRM false alarms correlated r = .76 over two weeks (Blair et al. 2002).
• Authors' conclusion. There is no false memory "trait": no type of person is especially prone or especially resistant.
  • Their practical warning: a low score on a suggestibility test, such as the Gudjonsson scale, does not show that someone is safe from suggestive techniques.
  • Their proposed mechanism (Patihis's dual encoding interference hypothesis): traits that strengthen encoding strengthen both true and false traces, so correlations with susceptibility come out null.

Limits

• Undergraduates; laboratory materials; a single study.
• Test formats differ between paradigms: two-alternative forced choice for misinformation, yes/no for the others. The misinformation test had no novel foil, which may inflate the effect.
• There is no robust (memory, not belief) measure for imagination inflation, and no open-ended reports.
• Reliability of the misinformation and imagination measures could not be computed because of counterbalancing. Low reliability would attenuate the correlations, though the authors argue they would still have been detectable.
• Many correlations were tested, so some significant ones will be type I errors (acknowledged).
• Inconsistencies found:
  • The BFI is described as rated on a 5-point scale, yet Table 5 gives Big Five maxima of 6.00.
  • The imagination interaction is reported with exactly the same statistics as the main effect of imagination (F(1.87, 453.48) = 948.53, ηp² = .80), which looks like a copy error.
  • Session 2 has 44 action statements in the Materials and 144 in the Procedure.

What it means for Kurisutina

• Question 3: a person's way of misremembering is not one parameter. The previous summary (schacter2011_adaptive_distortion.md) proposed fitting a replica's memory transformation to its person's own forgetting and distortion. This study says how that must be done.

  • Susceptibility is stable within a kind of distortion (DRM retest .76).
  • It barely carries across kinds (r ≤ .19).
  • Traits do not predict it: Big Five, cognitive style and imagery are null or small.

  So the slot cannot hold one "suggestibility" or "false memory" setting inferred from a questionnaire. Each kind of distortion that matters for carrying on has to be measured in the person, in the form it will take: misinformation from others, gist and context intrusions, and inflation from imagining. This is the same pattern as Eckstein et al. for learning parameters: local, not portable.

• Confidence style may be the transferable part. Confidence ratings correlated across paradigms (up to r = .59), though format and session are confounded. A person-level metamemory style (how sure someone feels about what they "remember") is a candidate for the slot worth testing. It also bears on how a replica should express certainty.

• The questionnaire route fails here too. Peng et al. found 500 generic psychometric items carried little outcome-specific information. Nichols & Loftus find 19 trait and ability measures do not predict false memory. Both point to targeted, behaviour-based elicitation in the domain where the replica will act.

• Ethics and elicitation (brief 6.3, 10.2). Everyone is susceptible, and a low suggestibility score is no licence. Kurisutina should not screen participants as "safe" for suggestive or imagination-based elicitation. The design rules (no imagination prompts about unconfirmed events, the perturbation audit) apply to every participant.

• A test design consequence. In a question-3 test, the person's distortion profile has to be measured in more than one paradigm, and the replica scored against each separately. A replica that matches its person on misinformation but not on gist intrusions is not "forgetting like them"; the measures do not substitute for each other.

Cross-references

• summaries/carry_on/schacter2011_adaptive_distortion.md: why these distortions exist (adaptive processes).
• summaries/carry_on/eckstein2022_context.md: individual parameters that transfer only between similar tasks.
• summaries/carry_on/peng2025_funhouse.md: generic psychometrics carry little outcome-specific information.
• summaries/memory/roediger1995_false_memories.md (DRM), summaries/memory/lindsay1990_source_suggestions.md and summaries/memory/johnson1993_source_monitoring.md (misinformation, source monitoring).