Quantifying compliance and acceptance through public and private social conformity

Consciousness and Cognition 65: 359–367, doi 10.1016/j.concog.2018.08.009; open access (CC BY 4.0), PMC6204883. Read by researcher R5a, 5 October 2026. Provenance: papers/change/sowden2018_public_private_conformity.provenance.json.

What was read

• Read in full: the Europe PMC XML, converted with tools/pmc2txt.py (self-check passed): highlights, abstract, introduction, methods, results, discussion, conclusions, acknowledgements, references.
• Not read: Figures 1–4 as images (Figure 4's values are in the text). No supplement.

Definitions used

• Compliance: conforming in public but not in private. It is attributed to normative influence (the group's power to accept or reject).
• Acceptance: the group's view is internalised, so it shows in private too. It is attributed to informational influence (others as evidence about the world, strongest under ambiguity).

What they did

• Participants. 22 women recruited at King's College London (mean age 21.2, SD 2.8); one guessed the deception and was excluded, leaving 21. Three female confederates. The participant always answered last.
• Task. Judge the grey level of a square on a 1–99 scale. 10 practice and 153 test trials, about 45 minutes.
• Trial types, within each person:
  • silent (everyone types; 18 trials);
  • private (the participant types after the confederates answer aloud; 27);
  • public (everyone answers aloud; 27);
  • reduced majority (one confederate types, so the participant answers aloud after only two spoken answers).
• Confederates. Wrong on two thirds of trials: ±15 from the truth plus 0–3 jitter. Otherwise within 3 of the truth.
• Measure. Discrepancy from the truth, signed toward the confederates.
  • Private effect (incongruent minus congruent) = informational influence.
  • Public effect minus private effect = normative influence.

Main results (verified)

• Baseline accuracy: mean absolute error on silent trials 1.42 (SEM .16).
• Private: congruent 0.57 (SEM .34), incongruent 2.10 (.36); t(20) = 3.10, p = .006, d = 0.96. Informational effect 1.53 (SEM .50).
• Public: congruent −0.26 (.34), incongruent 4.07 (.41); t(20) = 7.79, d = 2.51. Public effect 4.33 (SEM .56).
• Condition × congruency: F(1, 20) = 12.66, p = .002, ηp² = .39.
• Normative increment (public − private): 2.80 (SEM .79), p = .002. It is not significantly larger than the informational part (t(20) = 1.06, p = .303).
• Majority size:
  • incongruent public discrepancy 4.07 with three confederates, 3.21 with two (t(20) = 2.72, d = .51);
  • normative increment 2.80 against 1.74 (t(20) = 2.29, p = .033).
• My computation, against the 15-point push:
  • private shift ≈ 10%;
  • public ≈ 29%;
  • the private part is ≈ 35% of the public shift.
• Literature as cited (originals not read):
  • Asch 1956, Experiment 4: 43% conformity in public against 12.5% in private, and 1% errors without group pressure.
  • Abrams et al. 1990: 58% public against 33% private.
  • Deutsch and Gerard 1955: a 16-point drop with privacy.
  • In Asch, "only a quarter of participants remained completely independent".
  • "Off-line" paradigms (fake group feedback on a computer) give smaller conformity than live groups (Bond and Smith 1996).

Limits

• Small and narrow sample: n = 21, women aged 19–30, one site. The authors call the effect sizes inexact and the study task development.
• "Private" means an immediate private answer, typed seconds after hearing the group. There is no delayed retest, so acceptance here is not lasting change.
• A perceptual task with a correct answer. Opinions and moral judgements may differ.
• Inconsistencies found:
  • Reduced-majority congruency effect. The text gives 3.27, but its own means give 3.21 − (−0.59) = 3.80.
  • Degrees of freedom. The private effect is tested at t(21) = 3.10 in one place and t(20) = 3.10 in another, with 21 participants.
  • Two contrasts, one statistic. The same t(20) = 2.29, p = .033 is printed for both the public-effect and the normative contrast (three against two confederates), with different d (.49 and .32). The contrasts are algebraically identical, because the private term cancels; the d values differ in their denominators.

What it means for Kurisutina (inference)

• This is the design a replica probe should copy: the same push, in the same session, answered in public and in private, within one person.
  • About a third of the immediate shift survived privacy here.
  • Asch Experiment 4 (as cited) gives 12.5/43 ≈ 0.29, and Abrams (as cited) 33/58 ≈ 0.57 (my ratios).
• For B7:
  • Expression channel: carries the normative part, which should end when the audience does.
  • Belief channel: carries only the informational part.
  • Majority size: raises the normative part (three against two), so it belongs to the population rule.
• No duration data. This paper says nothing on how long the private part lasts.