Contribution
It turned pluralistic ignorance about drinking from a finding into an intervention and tested it in the field. Entering students who learned of the gap between private and attributed comfort reported drinking less four to six months later. It also asked how the intervention worked. The answer was not that students came to think peers approved of less drinking. It was that their perception of peers stopped governing their drinking through fear of others’ judgement.
Key insight
A norm that everyone privately doubts can still bind, because each person thinks the others endorse it, and thinks they endorse it uniformly. Showing people the doubt need not change what they think the norm is. It can free them from it by showing that the norm lacks the consensus that gives it force. The more consequential misperception may be of how unanimous the group is, not of where its average lies (pp. 2173–2174).
Assumptions
- The drinking norm is injunctive. It concerns what students think is appropriate, not how much they think others drink (p. 2154).
- Norms draw their force from perceived universal support (Turner 1991; Asch): “the presence of even one alternative viewpoint in a group sharply reduces the power of the group norm” (p. 2154).
- Fear of negative evaluation measures susceptibility to normative influence. The short-form FNE scale (Leary 1983), α = .90, was given after the discussion, since giving it before might have primed evaluative concerns. Condition did not affect it (fn. 9).
- Self-reported drinking is valid. The authors cite sensible relations with known correlates and no relation between reported drinking and FNE (fn. 6).
- The individual-oriented discussion is a neutral comparison. There was no untreated control, on the university’s ethical objection (fn. 3). The authors grant that the same pattern would follow if the individual-oriented discussion “heightened concern with drinking” (fn. 13).
Key results
- Sample (pp. 2155–2157). 452 first-year students in RA groups from four of Princeton’s five residential colleges attended September discussions (235 peer-oriented, 217 individual-oriented). Assignment was by RA group, “systematically” balanced across nights and colleges. 143 completed the follow-up 4 to 6 months later: 66 peer-oriented (35 men, 31 women) and 77 individual-oriented (44 men, 33 women). Return did not vary with condition, gender, ethnicity, comfort or FNE. Data collection ended when a student-newspaper article described the project (p. 2161).
- Manipulation (pp. 2160–2161, Appendix). A 7-minute video of party and drinking-game scenes, then a 20-minute peer-led discussion. The peer-oriented groups were told the self–other discrepancy finding and discussed why misperceptions arise. The individual-oriented groups discussed responsible personal decisions.
- Drinking (pp. 2161–2162, Table 1). Drinks per week (past week and a typical week, averaged; r = .70) were 3.10 (SD 4.15) in the peer-oriented condition against 5.05 (SD 6.70) in the individual-oriented condition. By gender: women 2.29 against 4.02, men 3.81 against 5.81. Condition F(1, 137) = 3.80, p = .05. Abstention did not differ (21.2% against 19.5%), so the difference is among drinkers.
- Pluralistic ignorance (pp. 2163–2165, Table 2). On 11-point scales, students rated their own comfort below the average student’s in September (6.98 against 7.76) and closer at follow-up (7.06 against 7.31): target × time F(1, 137) = 8.73. No interaction with condition (all F < 1). So the intervention did not change the norm students attributed to peers, and that cannot explain the drinking difference.
- The men’s convergence, and its direction (pp. 2164–2165, 2171). Time × gender F(1, 137) = 4.60. Men’s means went from 7.15 (self) and 7.87 (average student) to 7.05 and 7.12. The self–other gap closed (time × target for men, F(1, 137) = 7.32), almost wholly through their estimate of the average student. Women kept the gap: 6.76 and 7.63, then 7.08 and 7.54. The authors: “whereas Prentice and Miller’s second-year students reduced pluralistic ignorance by shifting their own comfort in the direction of the average student’s comfort, the first-year students in this study tended to do the reverse” (p. 2165).
- Mechanism (pp. 2165–2169, Table 3, Figure 1). A simultaneous regression of drinking on gender, own comfort, average-student comfort, condition, FNE and their interactions explained 39% of the variance. Own comfort β = .60; average-student comfort (relative to own) β = −.25; condition β = −.17; FNE × average-student comfort β = .27; the three-way FNE × condition × average-student comfort β = −.25, t(129) = −1.98, p = .05. In the individual-oriented condition, low-FNE students drank less the more comfortable they took the average student to be, relative to themselves. High-FNE students “uniformly drank at an above-average level” (p. 2169). In the peer-oriented condition both groups showed the low-FNE pattern. The result held among drinkers only, with less power (fn. 11).
- The 1993 finding, as this paper states it (p. 2153). “Prentice and Miller (1993) … found evidence for attitudinal conformity among male students, who modified their private attitudes over time in the direction of the position they mistakenly assumed to be held by the average student.” In a separate study both sexes showed alienation from the university when they believed their attitudes deviant.
- Two explanations of the reversal, both labelled speculative (p. 2172). First-years may be less certain of peer opinion, with estimates inflated by orientation stories and accounts of pre-coeducation Princeton. Or campus sentiment may really have shifted after years of university campaigns. Either way, “Men resolved the discrepancy, and women retained it”, and the men’s resolution is suspected to be “only temporary” (p. 2172).
- Uniformity, not location (pp. 2173–2174). Students in the two conditions may have rated the average student alike but represented the distribution differently. Through discussion the peer-oriented students learned “that different people feel differently”, which dispels “the illusion of universality (Allport, 1924) that gives peer opinion its prescriptive force”. So “their more consequential error may be the underestimation of the variability of the group”. Students feel pressure “because they believe that they are outliers”.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Exposing pluralistic ignorance to entering students reduces their drinking months later, relative to a decision-making discussion | weak to moderate | one field study, p = .05; self-report; 143 of 452 followed up; no untreated control |
| C2 | The reduction is among drinkers, not more abstention | moderate | equal abstention rates (21.2% against 19.5%) |
| C3 | The intervention did not change the comfort attributed to the average student | moderate | no condition effects on comfort ratings (all F < 1) |
| C4 | Exposure reduced the norm’s prescriptive strength: sensitivity to evaluation stopped moderating the link between perceived peer comfort and drinking | weak | one three-way interaction at p = .05, FNE measured after the manipulation, and the alternative in fn. 13 not excluded |
| C5 | First-year men reduced pluralistic ignorance by revising their estimate of others, not their own attitude; women did not reduce it | moderate | gender × time and simple effects; Table 2 |
| C6 | Prentice and Miller (1993) found men shifting their private attitudes toward the misperceived norm | as reported | restated (pp. 2153, 2165, 2171–2172), not reanalysed |
| C7 | The more consequential misperception is of the group’s variability, the illusion of universality | speculation | offered to explain C3 with C4; not measured |
| C8 | Behavioural norms depend for their power on perceived private support, and revealing that support as illusory reduces their power without revising what they prescribe | weak; generalization | C3 and C4, from one study |
Concepts
- Pluralistic ignorance — students “assume that their own private attitudes are more conservative than are those of other students, even though their public behavior is identical” (p. 2152, after Miller and McFarland 1991). It involves two errors: misplacing the group’s central tendency and underestimating its variability (pp. 2173–2174).
- Illusion of universality — overestimating how uniform others’ support for a norm is (Allport 1924; pp. 2152, 2173).
- Injunctive against descriptive norm — what people approve of against what they do (Cialdini, Kallgren & Reno 1991). The misperception studied is of the injunctive norm (p. 2154).
- Level against prescriptive strength of a norm — what a norm prescribes against how much it compels. These are the two routes the paper tests (p. 2154).
Connections
It applies Prentice and Miller’s 1993 programme, and Miller and McFarland’s account of pluralistic ignorance, to an intervention. Prentice and Miller 1993 (LIT-1370), the source of the self–other measure used here, is unread in this record. Crandall’s model of binge eating in sororities supplies the social-influence frame: behaviour is a function of social pressure and one’s vulnerability to it (p. 2170). Perkins and Berkowitz’s social-norms work is its neighbour. List’s falsely attributed common attitudes (NOTE-554) cite Prentice and Miller as their example, and this paper is the record’s first read account of that example.
Bearing on the record
- CLAIM-415: partly, and partly against its one named example. The
claim names Prentice and Miller’s men as its one example of private
change under a misperceived consensus. Its defeat condition counts as
defeat any finding that such effects are “fully explained as conformity
of expression, with no change in how the person interprets or organizes
their own activity, which is what Vilone and Polizzi’s model and
Prentice and Miller’s women show at most”. This paper bears on that in
three ways.
- The 1993 result is restated, not checked. The paper says Prentice and Miller’s men “modified their private attitudes over time in the direction of the position they mistakenly assumed to be held by the average student” (p. 2153). That confirms what the claim attributes to the 1993 paper, at second hand and from one of its authors. It is not a reading of the 1993 data.
- Its own data do not replicate the men’s private shift. First-year men here closed the gap by moving their estimate of others toward themselves (7.87 to 7.12), not their own comfort toward others (7.15 to 7.05). Women kept the gap, as in 1993. The authors explain the difference by the first-years’ weaker knowledge of peers or by a change on campus, and expect the men’s resolution to be “only temporary”. So the record’s one cited case of private change is now a result that its own authors’ follow-up did not reproduce in a different cohort. That weakens it as the claim’s support.
- What it does show is conduct, not belief, moved by an attributed norm. Months of self-reported drinking, not what students said in public, tracked the comfort they attributed to peers. It did so in proportion to their fear of others’ judgement, and private attitudes and attributed norms were the same across conditions (C3, C4). That goes beyond “conformity of expression” in the narrow sense, since it is how much people drank. But it is compliance, not reinterpretation: the paper’s mechanism is fear of censure, not a person organizing their activity around the judgement. So on the claim’s defeat condition the paper sits between the two options, and the claim must say which side conduct under normative pressure falls on.
- Step 1’s “false unity” has an empirical counterpart here. The paper’s speculative conclusion is that a norm’s force comes from its perceived uniformity. It suggests the more consequential error is “the underestimation of the variability of the group” (p. 2174). That is close to the claim’s “imagined social audience … organized into a false unity”. In the paper it is a hypothesis to explain C3 with C4, not a measured result.
- THEORY-242 and LIT-939. Vilone and Polizzi’s silenced majority holds private beliefs fixed. This paper is a field case with private attitudes unchanged by the intervention (C3) and conduct changed. It is consistent with a misperceived public norm that governs conduct without converting belief.
- No instruction for machine-learning practice; nothing for the anthology.
Limitations
- No untreated control. Both groups had an intervention. The authors grant that the result is equally consistent with the individual-oriented discussion raising concern with drinking (fn. 13).
- Attrition and assignment. 143 of 452 were followed up, and follow-up stopped when a newspaper article appeared. Assignment was by RA group, so students were clustered in groups that shared a discussion. The analysis treats students as independent.
- Self-report, between subjects only. No baseline drinking was taken, since it would not have been a valid baseline for first-years (fn. 5). The effect rests on the between-condition comparison at follow-up.
- Thin significance. The main effect is p = .05, the mechanism’s three-way interaction p = .05, and the drinking distributions are skewed (SDs larger than means, interquartile ranges starting at 0).
- FNE measured after the manipulation. No condition effect on it was found (fn. 9), but its role as a pre-existing moderator is assumed.
- What the authors do not claim. They say exposing pluralistic ignorance is no “magic-bullet” (fn. 14), and both their explanations of the reversal from 1993 are labelled speculative.
Open questions
- Does the 1993 finding of private attitude change in second-year men replicate? This paper’s first-years moved the other way, and the paper cannot tell cohort from campus change.
- Is it the perceived variability of opinion that carries the effect? The paper did not measure it. A measure of perceived dispersion, before and after, would test C7.
- Does the effect survive against an untreated control, with drinking measured other than by self-report?