The Decision You Think Is Yours
A card in a hotel bathroom and a door hanger in California are both instruments of social influence, and the people they move rank them last among the things that moved them. That gap has been measured on metered households, which is a smaller claim than hidden motives and a far better evidenced one.
Answer to the question
Can you tell when social influence changed your mind?
Often you cannot. In Jessica Nolan's 2008 field experiment on 371 San Marcos households across five conditions, those given a card reporting that 77 per cent of neighbours used fans instead of air conditioning used measurably less electricity than households given any other reason, and rated that same card the least motivating of the messages tested. The behaviour and the explanation disagreed, on the same people, in the same summer.
Key takeaways
- The households given a card reporting that 77 per cent of neighbours used fans cut electricity use most and rated that card least motivating, in a field arm of 371 households across five conditions.
- Asch's line task replicated in 2023 with 210 participants at an error rate of 33 per cent, falling to 25 per cent when answers were paid for.
- Across all 126 trials run by two large nudge units, take-up rose 1.4 percentage points against 8.7 percentage points in published studies.
- Five studies in 2025 found people often described their own choice process more accurately than a sample of decision scientists predicted.
The question is usually put as whether other people influence what you buy, and the answer to that is boring. They do. The question worth asking is narrower and much harder: when social influence has moved you, in a room with nobody else in it, can you tell afterwards that it did?
That narrowing is deliberate, because four other pieces already hold the neighbouring ground. [Group Decisions at the Checkout](https://thewhyweproject.com/articles/group-decisions-at-the-checkout) is about two or more people deciding together, and here nobody else is present. [Everyone Thinks Everyone Else Wants It](https://thewhyweproject.com/articles/everyone-thinks-everyone-else-wants-it) is about misreading what others privately want, where this is about misreading yourself. [How Much Recommendations Really Sell](https://thewhyweproject.com/articles/how-much-recommendations-really-sell) measures the lift a recommendation produces, where this asks whether the person noticed it. And [You Do Not Know Why You Bought It](https://thewhyweproject.com/articles/you-do-not-know-why-you-bought-it) asks whether anyone can report their own mental processes at all, which is a sixty-year argument; the question here is smaller and has field experiments behind it.
A decision taken alone
A card in a hotel bathroom. A door hanger on a house in California. A line on an electricity bill telling you how your usage compares with your neighbours'. None of these involves a conversation, a crowd or a salesperson, and all of them are instruments of social influence, built on the assumption that telling you what other people do will change what you do.
They work, modestly, and the sizes are in the next few sections.
What makes the case interesting is not the effect. It is that the same studies which measure the effect also, occasionally, ask the people it worked on what moved them. When both measurements exist on the same households, you can compare the reason given against the reason that moved them, and you do not need any theory about hidden motives to do it.
The survivor of the replication decade
Start with whether the oldest finding here still holds, because our own [The Psychology You Learned Didn't Replicate](https://thewhyweproject.com/articles/the-psychology-you-learned-didnt-replicate) teaches readers to distrust mid-century social psychology by default. On conformity, that distrust is misplaced.
The paradigm is Solomon Asch's, from 1956, and it is as plain as experiments get. A participant says aloud which of three comparison lines matches a standard line, a judgement with an obvious correct answer, while sitting with confederates who give the same wrong answer and answer before them.
Axel Franzén and Sebastian Mäder repeated it in 2023 with 210 participants, five confederates to each naive subject. The error rate was 33 per cent, which is close to what Asch reported. In a condition where participants were paid for correct answers it fell to 25 per cent, and on political statements rather than lines it rose to 38 per cent. Of the personality traits they tested, only openness bore any convincing relation to susceptibility.
Paying people to be right reduced the error. It did not remove it.
A systematic review by Carla Capuano and Peggy Chekroun in 2024 gathered 48 articles on conformity published since 2004 and concluded that recent replications and meta-analyses underscore the effect to the point where its existence need not keep being assessed. The review also reports a 2005 meta-analysis by Rod Bond covering 125 Asch-type studies, which separated influence that works through what people approve of from influence that works through information about the world, and found the group-size relation differs between them. A 1996 meta-analysis by Rod Bond and Peter Smith covering 133 studies did the same for culture.
So the mechanism is real and well evidenced. What the trade gets wrong is its size.
The towel sign and its size
The most-cited descriptive-norm field experiment is Noah Goldstein, Robert Cialdini and Vladas Griskevicius's hotel towel study of 2008, and its effect sizes are the part nobody prints. In the first experiment, a sign reporting that most guests reuse their towels produced reuse in 44.1 per cent of rooms against 35.1 per cent for the standard environmental appeal. In the second, pooled across 1,595 instances of potential reuse, the standard environmental appeal reached 37.2 per cent and the descriptive-norm conditions combined beat it by about six rooms in a hundred. The provincial version, naming guests in this room, reached 49.3 per cent against 42.8 per cent for the other norm wordings.
Converted into something a hotel manager could use, the first experiment moved about nine rooms in a hundred and the second about six.
P. Wesley Schultz and colleagues found the complication in 2007. A message reporting average neighbourhood electricity use pulled heavy users down and pushed light users up, the second being the boomerang effect, and adding a line conveying approval or disapproval removed it. A descriptive norm is not a lever that only moves one way.
Hunt Allcott's 2011 study is the scaled version. Home Energy Report letters comparing a household's electricity use with its neighbours' went to randomised groups across 600,000 treatment and control households in the United States, and consumption fell by 2.0 per cent, which he calculates as equivalent to a short-run price rise of 11 to 20 per cent. The heaviest-using tenth of households cut 6.3 per cent. The lightest tenth cut 0.3 per cent.
The saving was concentrated among the heaviest users.
Two replications and a discount
Gerd Bohner and Lisa Schlüter revisited the towel study in 2014 across two German hotels and 928 guests, with messages stating that 75 per cent of guests had reused their towels, varied by how close the reference group was and how recent the figure. Both the descriptive-norm messages and the standard environmental appeal raised reuse above a no-message baseline. The descriptive-norm message did not beat the standard appeal, and the proximity effects were inconsistent between their two studies.
So putting a sign up replicated. The thing the field is famous for did not.
Then the discount that applies to every figure above. Stefano DellaVigna and Elizabeth Linos assembled all 126 randomised trials run by two large United States nudge units, covering 23 million people, and compared them with nudge trials published in academic journals. The published trials show a take-up effect of 8.7 percentage points. The nudge-unit trials show 1.4 percentage points, still statistically significant, and roughly a quarter of the published effect once each is read as a share of its own control average, which is a different comparison from dividing the two percentage-point figures. Their model attributes roughly 70 per cent of the gap to selective publication, amplified by low statistical power.
Three things follow, and they apply to any norm figure a deck shows you:
- Ask whether the trial was published because it worked, or run because an agency had a letter to send.
- Expect the saving to sit mostly with heavy users, because the heaviest-using households supplied much the largest cut and the lightest-using barely moved.
- Keep the direction and discard the magnitude, because the direction has replicated and the magnitude has not.
None of that touches the question this article is actually about, which is whether social influence announces itself to the person it moves. A small effect can still be an undetected one.
The gap, measured in kilowatt hours
Jessica Nolan led the study that settles it, published in 2008 with the same group behind the towel sign. It did two things at once, which is why it is worth more than any of the above.
In the first part, a survey arm, 810 Californians were questioned by telephone about energy conservation. Their beliefs about what their neighbours actually did predicted their own behaviour better than any other belief measured, and those same survey respondents rated such norms the least important factor in their conservation decisions.
The second part was a separate field experiment on different people, in San Marcos, California, starting from 981 households and analysing 371 of them across five conditions. Each household got a door hanger urging the use of fans instead of air conditioning, with one of four reasons printed on it, or a control. One card appealed to saving money, up to 54 dollars a month. One offered the prevention of up to 262 pounds of greenhouse gases a month. One appealed to social responsibility. The fourth said that in a recent survey, researchers at Cal State San Marcos had found 77 per cent of San Marcos residents often use fans instead of air conditioning.
Households given the neighbours' figure used measurably less electricity in the short term than the other conditions combined, 12.97 against 14.17 kilowatt hours a day.
And in the follow-up interview, the households given that card rated it the least motivating of all the messages tested, significantly below the environmental card and the social-responsibility card.
> The condition that changed behaviour most was rated least motivating by the very people whose behaviour it changed.
The condition that changed behaviour most was rated least motivating by the very people whose behaviour it changed.
The authors' reading is that people fall back on their prior theories of what ought to motivate someone to save energy. Their conclusion is that although people may not believe the behaviour of others should motivate them, it does. Nothing in that design requires anybody to be concealing anything from themselves. It requires two measurements inside the one field arm, a meter and a rating scale, which is what makes it the strongest evidence in this piece.
The asymmetry and its mechanism
Why the rating comes out that way has its own literature. Emily Pronin, Jonah Berger and Sarah Molouki ran five studies in 2007 in which people judged others as more conforming than themselves, across consumer purchases and political views. The comparison was not restricted to an average peer either: they rated themselves less susceptible than named peers in specific instances of conformity.
Their account of the mechanism is the useful part. When judging their own susceptibility to social influence, people weigh what they can introspect; when judging someone else's, they weigh what that person did. Two different kinds of evidence, one standard applied to yourself and another to everyone else, and the verdicts come apart.
W. Phillips Davison named the broader pattern in 1983 as the third-person effect, after New Yorkers told him a political attack message had little effect on their own vote and more on the average New Yorker's. He supplies the name; Emily Pronin and Jessica Nolan supply the measurements.
The case for self-knowledge
This is where the article has to be careful, because the opposite case is strong and recent.
Adam Morris, with Ryan Carlson, Hedy Kober and Molly Crockett, published five studies in 2025 in which participants made choices and then reported how they believed they had made them, while computational modelling recovered the process their choices actually revealed. Accuracy varied substantially between people. But participants were often highly accurate, more accurate than a sample of decision scientists predicted they would be, and more accurate than informed third-party observers given the same information. Their conclusion is that these results challenge notions that we are strangers to ourselves.
That is not a small finding, and the honest limit is in their own abstract: the domain is value-based choice between options with several attributes, not social influence from a card on a door.
Ben Newell and David Shanks made the broader case in 2014, reviewing the evidence for unconscious influences on decision making and concluding there is little convincing evidence of them. They name three reasons the field believed otherwise: poor procedures for checking what participants were aware of, a failure to test ordinary explanations of landmark results, and a readiness to accept conclusions that match intuition. The open commentary published with them pushes back, with Ian McLaren and colleagues agreeing on the main point while warning that readers may conclude unconscious influence is impossible, which it is not.
Both arguments survive contact with the Nolan result, because that result does not claim an unconscious process at all. It claims a mismatch between a rating and a meter.
The unfalsifiable frame and its cost
The idea that we hide our motives from ourselves reaches most readers through Kevin Simler and Robin Hanson's The Elephant in the Brain, which argues that much of our behaviour serves motives we do not admit, including to ourselves. Their chapter on consumption claims that we read other people's purchases well and our own badly. No study is cited where that claim is made, and the examples offered are, in their own words, contrived.
The book's apparatus is the strongest in our library, at roughly 44 per cent peer-reviewed across about 480 bibliography entries, and its notes cite evidence against its own claims, which no other book on our shelf does. Our own hand count across their chapters puts about 253 claims as cited findings and about 485 as the authors' readings or conjectures, which is our count rather than theirs and is approximate at the margins.
The structural problem is not the ratio. A theory whose content is that people conceal a motive from themselves converts every denial into confirmation, so agreement and disagreement both arrive as support.
Nothing a reader could say would count against it.
They present that reach as a strength. Having argued that laughter signals that something is not serious, they note at page 137 that the theory also explains the cases where we do not laugh, and then the case where laughter returns. At page 68 they allow that neither party to a signal need be conscious of it, which, as our audit records, means no introspective report can refute the claim. To their credit, page 306 concedes that these explanations are more compelling about the species than about any individual, where they are often hollow and unpersuasive. And page 106 says it outright: if they seem to have selectively chosen their sources and evidence, then it is probably because they have.
### Where this is weakest
Five limits worth stating plainly. Nolan and colleagues ran two arms on two different samples, a telephone survey of 810 Californians and a field experiment on San Marcos households, so the two measurements this piece turns on are the meter and the rating inside the field arm, and the survey's own least-important ranking is separate evidence pointing the same way rather than part of the same comparison.
The Nolan field experiment analysed 371 households in one Californian town in one summer, and its behavioural result is a short-term one. Every field figure in this piece carries the DellaVigna and Linos discount, so treat the directions as sound and the magnitudes as inflated. The strongest counter-case, from Adam Morris and colleagues, tested a different kind of decision from the one studied here, and its own authors report wide variation between people. And nothing here shows that anyone is hiding a motive; a gap between a rating and a meter is compatible with concealment and does not establish it.
Here is our verdict, and it is narrower than the one the popular version offers. We think the honest claim is underdetection rather than self-deception: social influence can change what a person does while being ranked last among the things that person believes moved them, and that has been measured on metered households rather than argued from a frame. The hidden-motive account is the more interesting story and the less testable one, and we would rather have the smaller claim that could have come out flat and did not.
Which leaves you with a practical dilemma rather than a conclusion. The next time you decide something alone and can give a clean account of why, you have two sources of evidence about yourself and they may disagree. Trust the account, or read the meter.
Common questions
Can you tell when social influence changed your mind?
Often not. Jessica Nolan's 2008 field experiment analysed 371 San Marcos households across five conditions. The households given a card reporting that 77 per cent of neighbours used fans used measurably less electricity than those given other reasons, and then rated that same card the least motivating of the messages tested. The behaviour and the explanation disagreed, on the same households, in the same summer.
How large are social influence nudges in practice?
Smaller than published studies suggest. Stefano DellaVigna and Elizabeth Linos examined all 126 trials run by two large nudge units, covering 23 million people, and found a take-up effect of 1.4 percentage points against 8.7 in the academic literature, attributing about 70 per cent of the gap to selective publication. Hunt Allcott's 600,000-household energy study found a 2.0 per cent reduction overall.
Do people know why they made a choice?
Sometimes, and better than experts expect. Adam Morris and colleagues reported five studies in 2025 where people described how they had reached a choice and were often highly accurate, beating both decision scientists' predictions and informed observers. Accuracy varied widely, and the choices were between options with several attributes rather than decisions under social influence.
Is conformity still a real finding?
Yes, and it replicates. Axel Franzén and Sebastian Mäder repeated Asch's line task in 2023 with 210 participants and found an error rate of 33 per cent, falling to 25 per cent when answers were paid for and rising to 38 per cent on political statements. A 2024 systematic review of 48 articles concluded that recent replications and meta-analyses underscore the effect to the point where its existence need not keep being assessed.
Sources
Asch, S. E. (1956). 'Studies of independence and conformity: I. A minority of one against a unanimous majority.' Psychological Monographs: General and Applied, 70(9), 1-70. doi:10.1037/h0093718. [Citation and sole authorship confirmed at the OpenAlex item record, Swarthmore College. ABSTRACT NOT READ: none is deposited at Crossref or OpenAlex, the monograph runs seventy pages and no open copy was located in this pass]: the origin, and the article must describe the paradigm from the peer-reviewed secondary sources it did read rather than from the monograph. The design as the 2024 PRISMA review describes it: participants verbally identify which of three comparison lines matches a standard line, a task characterised by its simplicity and lack of ambiguity, seated alongside confederates who give unanimous incorrect answers on certain trials, with the naive participant answering last. Nothing is attributed to this paper directly.
Bond, R. (2005). 'Group Size and Conformity.' Group Processes & Intergroup Relations, 8(4), 331-354. doi:10.1177/1368430205056464. [Abstract read; full text not retrieved. The effect size below is taken from the 2024 systematic review's report of this paper and NOT from the paper itself, and must be attributed that way]: the most recent meta-analysis of the paradigm, covering 125 Asch-type conformity studies. The 2024 PRISMA review reports it as finding conformity a robust behaviour exhibiting a weighted average effect size of 0.89. Bond's own abstract frames the paper as a challenge to the assumption of one fixed group-size function in formal social influence models, arguing that normative influence should dominate with public face-to-face responses and informational influence with private responses and indirect communication with the majority, and finding that the size-conformity relationship differs accordingly and by whether the study used the Asch or the Crutchfield paradigm. Sole author. CAUTION: the vault holds People notes for both Rod Bond and Charles Bond; this is Rod Bond.
Franzen, A., & Mader, S. (2023). 'The power of social influence: A replication and extension of the Asch experiment.' PLOS ONE, 18(11), e0294325. doi:10.1371/journal.pone.0294325. [Abstract read in full at journals.plos.org; open access]: the modern direct replication and the single most useful source in the conformity section. Four stated goals: a replication of the original Asch experiment with five confederates and one naive subject in each group, N = 210; a randomised trial incentivising the line decisions with money; an extension to political statements; and a test of whether intelligence, self-esteem, need for social approval and the Big Five predict susceptibility. Results in the authors' own words: an error rate of 33 per cent for the standard length-of-line experiment, which replicates the original findings by Asch; in the incentivised condition the error rate decreases to 25 per cent; for political opinions a conformity rate of 38 per cent; and besides openness, none of the investigated personality traits are convincingly related to susceptibility to group pressure. The money result is the interesting half for this article: paying people to be right lowers the error rate without removing the influence. Two authors, both at the Institute of Sociology, University of Bern, Franzen first and corresponding, confirmed at the OpenAlex item record.
Capuano, C., & Chekroun, P. (2024). 'A Systematic Review of Research on Conformity.' International Review of Social Psychology. doi:10.5334/irsp.874. [Full text read, from the open journal]: a PRISMA systematic review of conformity research conducted since 2004, identifying 48 relevant articles from a literature search run between January and April 2023. The abstract's position is that recent studies confirm the prevalence of conformity across diverse contexts, echoing Asch, while the field needs a unified understanding of influencing factors including age, gender and culture, with contextual variables central, and that digital technology has expanded the research possibilities through computer-mediated communication and virtual reality. The Introduction carries the numbers this article needs and all are the review's report of other papers: Bond 2005 on 125 studies at a weighted average effect size of 0.89; Goodmon and colleagues 2020 finding 82.67 per cent of their participants conformed to the majority at least once; Usto and colleagues 2019 replicating Asch 1956 at a conformity rate of 59.2 per cent against Asch's 75 per cent; and Franzen and Mader's 33 per cent. The review's verdict in its own words: recent replications and meta-analyses on conformity underscore the robustness of this effect, obviating the necessity to continually assess its existence as it persistently manifests. The same paragraph cites Pronin and colleagues 2007 for the observation that individuals occasionally portray themselves as less conformist than their peers, which is the bridge to the self-report half of this article. Two authors, Capuano first at Universite Paris Nanterre, Chekroun second.
Bond, R., & Smith, P. B. (1996). 'Culture and conformity: A meta-analysis of studies using Asch's (1952b, 1956) line judgment task.' Psychological Bulletin, 119(1), 111-137. doi:10.1037/0033-2909.119.1.111. [Citation and author order confirmed at the Crossref item record: Bond first, Smith second. ABSTRACT NOT READ. No abstract is deposited at Crossref; api.semanticscholar.org returned the record with the abstract field explicitly elided by the publisher; no open copy was located]: secondary summaries agree on 133 studies and on higher conformity in collectivist than in individualist cultures, and several report a decline in effect sizes between the 1950s and the 1990s, but every reachable source for the decline in this pass was a study-aid or exam page rather than the paper or a peer-reviewed secondary. THE TIME-TREND QUESTION IS THEREFORE NOT VERIFIED. The article may cite this paper for the existence of a culture-and-conformity meta-analysis of 133 studies and must not assert the decline. What it can say instead, read at first hand and better evidence anyway: the 2024 PRISMA review and the 2023 direct replication both find conformity rates close to Asch's, which is positive evidence against a large decline. CAUTION: the vault's People note for Peter Smith is the correct Peter B. Smith.
Goldstein, N. J., Cialdini, R. B., & Griskevicius, V. (2008). 'A Room with a Viewpoint: Using Social Norms to Motivate Environmental Conservation in Hotels.' Journal of Consumer Research, 35(3), 472-482. doi:10.1086/586910. [Full text read, from the open copy at the Carlson School of Management]: the hotel towel study, and the article's job is to print the effect sizes the decks never print. Experiment 1 used 190 rooms in a midsized, midpriced hotel in the Southwest that was part of a national hotel chain, with guests unaware they were participants, and two signs tested. The descriptive-norm sign produced a towel reuse rate of 44.1 per cent against 35.1 per cent for the environmental-protection condition, chi-square(1, N = 433) = 3.72, p = .05, phi = .09, and the authors note the normative sign is one which they have never observed employed by any hotel. Experiment 2 set the descriptive-norm conditions combined against the standard environmental appeal at 37.2 per cent, chi-square(1, N = 1,595) = 4.94, p = .03, phi = .06. The provincial-norm finding, about guests in this room, gave 49.3 per cent against 42.8 per cent for the other three descriptive-norm conditions combined, chi-square(1, N = 1,318) = 4.18, p < .05, phi = .06. Authors' stated scope: theoretical and practical implications for managing pro-environmental efforts. The paper also addresses a discrepancy against towel-hanger suppliers' data, noting its signs informed guests that the majority recycled at least one towel at some point during their stay. The effect sizes are phi = .09 and phi = .06 and the article should state that in words, because the sign that launched a thousand nudge decks moved about nine rooms in a hundred and then about six. Author order confirmed at OpenAlex: Goldstein first.
Bohner, G., & Schluter, L. E. (2014). 'A Room with a Viewpoint Revisited: Descriptive Norms and Hotel Guests' Towel Reuse Behavior.' PLOS ONE, 9(8), e104086. doi:10.1371/journal.pone.0104086. [Abstract read in full at journals.plos.org; open access]: the replication, and the counterweight the article has to carry. Two hotels in Germany, Study 1 N = 724 and Study 2 N = 204, so 928 guests. Descriptive-norm messages stating that 75 per cent of guests had reused their towels, or a standard message appealing to environmental concerns, were placed in guests' bathrooms, with the norm messages varying by reference-group proximity, hotel guests against guests in this room, and by temporal proximity, currently against two years previously. Towel reuse was unobtrusively recorded. The result in the authors' words: reuse rates were high overall and both standard and descriptive norm messages increased reuse rates compared to a no-message baseline; however, descriptive norm messages were not more effective than the standard message, and effects of proximity were inconsistent across studies. Their discussion addresses cultural and conceptual issues in comparing their findings with the original. So putting a sign up at all replicated; the thing the field is famous for, the descriptive norm beating the standard appeal, did not. Bohner first.
Schultz, P. W., Nolan, J. M., Cialdini, R. B., Goldstein, N. J., & Griskevicius, V. (2007). 'The Constructive, Destructive, and Reconstructive Power of Social Norms.' Psychological Science, 18(5), 429-434. doi:10.1111/j.1467-9280.2007.01917.x. [Abstract read; full text not retrieved in this pass, and the 290-household figure that circulates for this study could not be confirmed and must not be used]: a field experiment on household energy conservation in which a descriptive normative message detailing average neighbourhood usage produced either desirable energy savings or the undesirable boomerang effect, depending on whether the household was already a below-average or above-average consumer, and adding an injunctive message conveying social approval or disapproval eliminated the boomerang. The boomerang is the article's honest complication: the same message that pulls heavy users down pulls light users up. Author order confirmed at OpenAlex: Schultz first at California State University San Marcos, then Nolan, Cialdini, Goldstein and Griskevicius. CAUTION: the vault holds a People note for Wolfram Schultz, who is a different person from P. Wesley Schultz.
Allcott, H. (2011). 'Social norms and energy conservation.' Journal of Public Economics, 95(9-10), 1082-1095. doi:10.1016/j.jpubeco.2011.03.003. [Abstract read, including the heterogeneity result from the RePEc and IDEAS listings of the published abstract; full text not retrieved]: the scaled-up version, and the number the article needs for the gap between a laboratory result and a national programme. A series of programmes in which the company OPOWER sent Home Energy Report letters to residential utility customers comparing their electricity use with their neighbours', using data from randomised natural field experiments at 600,000 treatment and control households across the United States. The result is a 2.0 per cent reduction in consumption, which the author states is equivalent in effect to a short-run electricity price increase of 11 to 20 per cent, with cost-effectiveness comparing favourably to traditional conservation programmes. The heterogeneity carried in the published abstract: the highest-consumption decile cut usage by 6.3 per cent while the lowest decile cut it by only 0.3 per cent. Sole author. An earlier 2009 working paper by the same author used about 80,000 Minnesota households and reported 1.9 to 2.0 per cent; the 600,000 figure belongs to the published version and the two must not be mixed. The vault already holds a People note for Hunt Allcott.
DellaVigna, S., & Linos, E. (2022). 'RCTs to Scale: Comprehensive Evidence from Two Nudge Units.' Econometrica, 90(1), 81-116. doi:10.3982/ECTA18709. [Abstract read, from the Econometric Society listing of the published abstract; full text not retrieved]: the single most important source in this file for the article's honesty, and it belongs in the piece at full strength. The authors assembled 126 randomised controlled trials covering 23 million individuals, being all trials run by two large United States nudge units, and compared them with nudge trials published in academic journals drawn from two earlier meta-analyses. The academic-journal trials show a take-up effect of 8.7 percentage points, about a third above the control mean; the nudge-unit trials show a smaller but still statistically significant effect of 1.4 percentage points, an increase of roughly 8 per cent over control. A meta-analysis model attributes about 70 per cent of the gap to selective publication in the academic sample, amplified by low statistical power, with differences in nudge characteristics accounting for most of the remainder. Version note, recorded because it is checkable: the NBER working paper w27594 gives 33.5 per cent and 8.1 per cent where the Econometrica abstract gives 33.4 per cent and 8.0 per cent, and the working paper goes further than the published version in saying publication bias can account for the full difference. Cite the Econometrica figures. DellaVigna first.
Nolan, J. M., Schultz, P. W., Cialdini, R. B., Goldstein, N. J., & Griskevicius, V. (2008). 'Normative Social Influence is Underdetected.' Personality and Social Psychology Bulletin, 34(7), 913-923. doi:10.1177/0146167208316691. [FULL TEXT READ, from the open copy at the California State University San Marcos repository]: THIS IS THE ARTICLE AND EVERYTHING ELSE IS SCAFFOLDING. The abstract in the authors' own words: the first study surveyed 810 Californians about energy conservation and found that descriptive normative beliefs were more predictive of behaviour than were other relevant beliefs, even though respondents rated such norms as least important in their conservation decisions; Study 2, a field experiment, showed that normative social influence produced the greatest change in behaviour compared to information highlighting other reasons to conserve, even though respondents rated the normative information as least motivating; results show that normative messages can be a powerful lever of persuasion but that their influence is underdetected. Study 1 as the paper gives it: 810 Californians surveyed by Computer Assisted Telephone Interviewing, stratified by region across Northern, Bay Area, Central, Los Angeles and Southern. Study 2 as the paper gives it: a starting sample of 981 households in San Marcos, California, of which 509 participated in a post-intervention interview, 52 per cent, with 371 households included in the analysis, 73 per cent of those interviewed. Door-hanger messages urging the use of fans instead of air conditioning in four reason conditions plus a control, the descriptive-norm card reading in part that in a recent survey of households in your community researchers at Cal State San Marcos found that 77 per cent of San Marcos residents often use fans instead of air conditioning, with the self-interest card offering up to 54 dollars a month, the environmental card up to 262 pounds of greenhouse gases, and a fourth appealing to social responsibility. The behavioural result: participants in the descriptive-norm condition used significantly less energy in the short term than participants in the combined other conditions, adjusted mean 12.97 against 14.17, SE .44 and .20, N = 46 in the norm condition, F(1, 268) = 5.99, p < .05. AND THE RESULT THAT MAKES THIS THE SPINE, from the same paper on the same people: participants in the descriptive-norm condition reported that the messages were the least motivational, M = 1.76, SE = .10, significantly lower on pairwise comparison than participants in the environmental condition at M = 2.07, SE = .08, p < .05, and than the social-responsibility condition. The authors' own reading: these results suggest that people relied on their a priori theories of what should motivate someone to conserve energy, and although people may not believe that the behaviour of others should motivate them to conserve energy, it does. Author order at the published record, and the secondary literature gets this wrong: Jessica M. Nolan first, at the University of Arkansas, then P. Wesley Schultz, Robert B. Cialdini, Noah J. Goldstein then at the University of Chicago, and Vladas Griskevicius at the University of Minnesota. Funded by the Hewlett Foundation, with Study 1 forming part of a master's thesis at San Marcos. One secondary index renders the first author as J. P. Nolan; that is an error.
Pronin, E., Berger, J., & Molouki, S. (2007). 'Alone in a Crowd of Sheep: Asymmetric Perceptions of Conformity and Their Roots in an Introspection Illusion.' Journal of Personality and Social Psychology, 92(4), 585-595. doi:10.1037/0022-3514.92.4.585. [Abstract read in full]: the self-report half, and five studies of it. Across five studies people judged others to be more conforming than themselves, in areas ranging from consumer purchases to political views, rating themselves as less susceptible than their average peers on broad descriptions of social influence and less susceptible than specific peers in specific instances of conformity. The authors argue the effect is not simply social desirability and is rooted in how people weigh introspective against behavioural information: people give more weight to introspective evidence when judging their own susceptibility than when judging someone else's, which is the introspection illusion. Author order: Pronin first at Princeton, Berger second, Molouki third, given in full as Sarah Molouki at the PubPeer record. The vault already holds a People note for Emily Pronin. A 2020 honours thesis at Appalachian State reports a replication with a larger and more diverse sample finding the effect specific to self against others rather than self against a close friend; it is an unpublished thesis and must not be cited as a replication.
Davison, W. P. (1983). 'The Third-Person Effect in Communication.' Public Opinion Quarterly, 47(1), 1-15. doi:10.1086/268763. [Citation and sole authorship confirmed at the OpenAlex item record, which gives Columbia University, where Davison was Professor of Journalism and Sociology. ABSTRACT NOT READ: no deposited abstract was located and the article is behind Oxford's paywall. OpenAlex records the first page as 1 and carries no last page; the pagination 1 to 15 is the standard citation and is given as such]: the originating paper for the effect the seed book correctly credits to him. The claim, as the vault's digest of the seed book records it and as the secondary literature uniformly reports it: people deny being influenced by media while believing others are influenced, Davison's illustration being New Yorkers saying a gubernatorial attack message had little effect on their own vote but a greater effect on the average New Yorker. The article should credit Davison for naming and originating the effect and carry its evidence from Pronin and from Nolan, both of which were read at first hand.
Nisbett, R. E., & Wilson, T. D. (1977). 'Telling more than we can know: Verbal reports on mental processes.' Psychological Review, 84(3), 231-259. doi:10.1037/0033-295X.84.3.231. [Abstract read; author order confirmed at the OpenAlex item record, Nisbett first and Wilson second, which is the ordering the seed book gets wrong on p. 99 and right on p. 102]: cited here only for positioning, because the vault's article You Do Not Know Why You Bought It owns this material. The claim at abstract level: there may be little or no direct introspective access to higher-order cognitive processes, and people attempting to report on their cognitive processes do so not on the basis of true introspection but on the basis of a priori implicit causal theories. The article's one sentence about it should be the boundary line: the general question about introspective access is unsettled and sixty years old, and the narrower question about detecting social influence on oneself has field experiments behind it.
Morris, A., Carlson, R. W., Kober, H., & Crockett, M. J. (2025). 'Introspective access to value-based multi-attribute choice processes.' Nature Communications, 16(1), 3733. doi:10.1038/s41467-025-59080-y. [Abstract read in full; open access]: the best counter-case in the literature and it is recent, and it has to be in the piece rather than in a footnote. The design in the authors' words: across five studies participants make choices and then report how they believe they made them, with computational modelling used to identify the process revealed in their choices and compared against those self-reports. The finding: while we observe substantial variation in accuracy, participants are often highly accurate about their choice process, more accurate than predicted by a sample of decision scientists and more accurate than informed third-party observers. The conclusion, unusually direct and worth quoting: these results challenge notions that we are strangers to ourselves. Note what the comparison is and say it in the piece: people were not merely better than chance but better than experts expected and better than observers given the same information. Data, code and materials on OSF at doi:10.17605/OSF.IO/TMQJU. The honest limit, stated in their own abstract: substantial variation in accuracy, and the domain is value-based multi-attribute choice rather than social influence. Author order confirmed at OpenAlex: Adam Morris first and corresponding at Princeton, Carlson at Chicago Booth, Kober at Berkeley, Crockett last and also corresponding at Princeton. The vault already holds a People note for Molly Crockett. CAUTION: the vault holds a People note for Michael Morris, who is a different person from Adam Morris.
Newell, B. R., & Shanks, D. R. (2014). 'Unconscious influences on decision making: A critical review.' Behavioral and Brain Sciences, 37(1), 1-19. doi:10.1017/S0140525X12003214. [Abstract read; the open copy at UCL Discovery was located but the full text was not read in this pass]: the overreading case, made at length in the journal designed for it, with open peer commentary attached. The review covers multiple-cue judgment, deliberation without attention and decisions under uncertainty, with shorter treatment of priming and of awareness in movement and perception. Its conclusion is that there is little convincing evidence of unconscious influences on decision making and that such influences should not therefore be assigned a prominent role in theories of decision making. Its diagnosis of how the field got there names three causes: inadequate procedures for assessing awareness, failure to test artifactual explanations of landmark results, and a readiness to accept conclusions that match intuitions. Its recommendation is that future work should use tasks that divert participants' attention away from the experimenter's hypothesis rather than highly reflective tasks. The honest qualification, from the open commentary: McLaren and colleagues agree the case against an intelligent cognitive unconscious holds but worry the paper could lead readers to think unconscious influences on decisions are impossible, and give reasons why that may not be so. Author order confirmed at OpenAlex: Newell first at the University of New South Wales, Shanks second at University College London. The vault already holds People notes for Ben Newell and David Shanks.
Smith, E. R., & Miller, F. D. (1978). 'Limits on perception of cognitive processes: A reply to Nisbett and Wilson.' Psychological Review, 85(4), 355-362. doi:10.1037/0033-295X.85.4.355. [Citation, pagination and author order confirmed at the OpenAlex item record: Eliot R. Smith first at the University of California Riverside, Frederick D. Miller second; several secondary reference lists give the second author as F. S. Miller and that is wrong. ABSTRACT NOT READ. No deposited abstract exists at Crossref; api.semanticscholar.org returned the record with the abstract field explicitly elided by the publisher; psycnet.apa.org returned an empty body; no open copy was found]: secondary summaries report the objections as inappropriate statistical tests, a nonfalsifiable statement of the argument, an incorrect definition of causality, missing essential definitions and other evidence running against the theory, but no peer-reviewed or primary source carrying those objections was reached in this pass. The article may name this paper as the contemporaneous reply and must not quote or paraphrase its arguments. Worth noting only if the abstract is later obtained: the vault's own finding about the seed book is The Unfalsifiable Frame, and the contemporaneous reply to the introspection paper is reported to have made the same objection about the same family of claim in 1978.