KurisutinaJournal of Personality and Social Psychology 83(6): 1298–1313, doi 10.1037/0022-3514.83.6.1298. Read by researcher R5b (change with reasons), 5 October 2026, from the publisher's PDF hosted on the second author's website. Provenance: papers/change/tormala2002_resistance_certainty.provenance.json.
What was read
- Read in full: all 16 pages, converted with
pdftotext -layout: abstract, theory, Experiments 1–4, the follow-up studies reported in the General Discussion and in footnote 5, Tables 1–2, figure captions, references. - Figure 3 (attitude change in Experiment 3) was rendered and inspected; its values below are read off the bars (about ±0.05). Figures 1 and 2 were not inspected as images; Figure 1's cell means are given in the text, Figure 2's coefficients too.
- Nothing supplementary exists.
What it is
- Claim: resisting persuasion is not a null event. People observe their own resistance and infer, like an outside observer, that their attitude must be valid, so their certainty rises. They correct for the situation: resisting an attack they believe is strong is diagnostic; resisting one they believe is weak is not ("there may be stronger arguments out there").
- Paradigm (all four experiments): Ohio State undergraduates read a counter-attitudinal message proposing senior comprehensive exams at their university (Petty and Cacioppo's arguments: two strong, two weak, mixed in one message). They were told to list counterarguments. The same message was framed as containing "only the strongest (weakest) of all the arguments … the best (worst) possible reasons". Controls read an irrelevant article. Attitudes: 1–9 semantic differentials (α .88–.96). Certainty: one 1–9 item ("How certain are you of your opinion…?"). Two blind judges rated counterargument number, quality and focus.
Main results (verified)
| Experiment | n | Manipulation | Attitudes | Certainty (1–9) |
|---|---|---|---|---|
| 1 | 56 | perceived strong, weak, control | equal: 5.03, 4.91, 4.82 (F = 0.16) | 6.16 against 4.33 and 5.10; strong > others, F(1,53) = 7.31 |
| 2 | 113 (72 in the 2 × 2, 41 control) | strength × false feedback "your attitude changed" (index 9) or "did not change" (index 2) | no change pre to post in any condition | told not changed: strong 6.78, weak 5.00, control 5.24; told changed: 5.66, 5.65, 5.37 (no difference). Interaction F(1,68) = 4.50 |
| 3 | 70 | as Exp. 1, then 15 min filler, then a second message with three new strong arguments, no counterargument instructions | first message resisted in all conditions | not measured (to avoid demand) |
| 4 | 112 | as Exp. 1, plus voting intention | equal: 4.81, 4.96, 4.68 | 5.39 against 4.78 and 4.83 (overall F p = .10; contrast F(1,109) = 4.63) |
- Mediation (Exp. 2, no-change group, n ≈ 33): perceived message strength predicted certainty (β = .34) and perceived success at resisting (β = .38); with both in the model, success β = .53 and strength β = .14 (n.s.); Sobel z = 1.95, p = .05.
- Later resistance (Exp. 3, Figure 3): after the second, strong message, attitudes moved by about +0.4 points after resisting a perceived-strong first message (4.93 to 5.33, n.s.), against about +1.1 after a perceived-weak one (4.73 to 5.82) and +1.3 in controls (4.88 to 6.21); interaction F(2,67) = 3.40, p < .04. Overall 4.85 to 5.79.
- Attitude–intention consistency (Exp. 4): attitude–voting-intention correlation r = .89 after resisting a perceived-strong message, against .72 (perceived weak) and .68 (control); z = 2.13 and 2.46. Adding the attitude × certainty interaction removed the attitude × condition interaction (β .18, n.s.), consistent with certainty carrying the effect.
- Not explained by counterarguments: across experiments the number, quality and focus of counterarguments did not differ between perceived strong and weak (e.g. Exp. 1: 2.94 against 3.44 counterarguments; quality 5.41 against 5.69).
- Generalizations reported in the discussion:
- a consumer study (48 students, an aspirin advertisement, no "best/worst" wording): certainty 6.61 after resisting the perceived-strong advertisement against 4.48 (weak) and 4.97 (control); attitudes unchanged;
- footnote 5 (57 students, messages actually strong or weak, unlabelled, counterargued): equal attitudes (5.18 and 4.73) but certainty 5.81 against 4.89, t(55) = 1.97, p = .05.
- Background the authors cite (not tested here): more certain attitudes changed less under persuasive arguments and were more stable over two weeks (Bassili 1996); certain attitudes predict behaviour better (Fazio and Zanna 1978).
Limits
- Small samples (n 18–41 per cell), undergraduates, one main topic; one-item certainty; certainty effects partly marginal (Exp. 4 omnibus p = .10; mediation p = .05).
- Resistance was instructed ("list your arguments against"), and the authors assume it was easy; they predict difficult resistance, very strong attacks, or "illegitimate" resistance ("I just ignored the arguments") could lower certainty, untested here.
- Time scale: 15 minutes at most between attacks. Persistence of the certainty gain is not measured.
- Motivation to think was high throughout (personally relevant topic); the authors expect the effect mainly when people think extensively.
What it means for Kurisutina (inference)
- An attempt at persuasion that fails still changes the person. A rule of change needs a second state per belief besides its value: certainty. It rises when the person resists what they took to be a strong attack, and then lowers the yield to the next attack (about a third of the control's move here) and ties behaviour more closely to the attitude.
- What raises certainty is the person's own reading of the episode (I resisted; the attack was strong), not the content of their counterarguments. In a replica this is a metacognitive write: an episode stored as "I held my view against a strong argument" should strengthen the belief, while "I held it against a weak one" should not.
- For M1: a "resisted strong argument" family has a checkable rule here only as a direction: value unchanged, certainty up, later yield down. Sizes come from a 1–9 scale in small samples.
- For the battery: the human pattern is a usable signature: unchanged attitude, higher certainty and lower later yield after resisting a strong attack, but not a weak one.