Reactance: Why a No Makes You Want It
Assertive advertising reliably irritates people and barely moves them, and the gap between those two findings is the useful part. The best pooled estimate in advertising says both halves at once.
Answer to the question
Why does being told no make you want it?
Being told no mainly makes you resist the message rather than want the thing. Pooled across 142 advertising samples in 2026, assertive wording produced a large rise in how threatened people felt, a moderate rise in reactance, and only a very small drop in persuasion. The annoyance is roughly nine times the size of the commercial damage.
Key takeaways
- Pooled across 142 advertising samples, assertive language raised felt threat to freedom by a large amount, raised reactance moderately, and lowered persuasion by a very small amount.
- The only moderator that survived that analysis was sector: non-profit campaigns drew stronger reactance than commercial ones.
- Warning labels drew people of all ages to violent television programmes in a study of 900 participants, while information labels did not.
- A preregistered meta-analysis of 52 but-you-are-free experiments and 19,528 people found a moderate effect overall and no effect at all in the seven studies rated at low risk of bias.
Telling people what to do annoys them reliably and persuades them hardly at all. Reactance, the push-back a person feels when a freedom looks like it is being taken away, is among the best-attested irritants in consumer research and among the feeblest commercial levers. The largest pooled estimate in advertising has the annoyance running large and the sales effect running very small, and the distance between those two findings is the story.
What Chu and colleagues pooled from 142 samples
A 2026 review in the Journal of Advertising, led by Tsz Hang Chu, gathered 142 samples of assertive advertising language and modelled the whole chain in one pass. Assertive language is the register of the instruction: buy now, stop wasting money, don't miss out.
The chain it supports runs one way, and each step is weaker than the one before it.
- The advert issues an instruction.
- The reader feels a freedom narrowing. This effect is large, and it is the most dependable thing in the literature.
- The feeling becomes state reactance, the mix of anger and silent counter-arguing that a threatened freedom produces. Moderate.
- Reactance lowers persuasion. Very small, and that is where the chain stops.
Pooled across those samples, assertive wording raised how threatened people felt by a large amount and moved whether they were persuaded by a very small one. The two sit on different measures and we are not going to divide one by the other. The point is the contrast in size. The net effect on persuasion was negative, which is the direction an advertiser fears, and small enough that an agency could run assertive copy for a year without seeing it in the numbers.
Assertive copy reliably annoys people. Whether it costs a sale is a much smaller question than the annoyance suggests.
One moderator survived. Campaigns in the non-profit sector drew stronger reactance than commercial ones, which is awkward, since those are the advertisers most confident that an instruction is justified. The authors are candid about what beat them: the spread between studies, they write, "could not be fully explained by the moderators we were able to code".
What Zemack-Rugar found in committed buyers
The reader who reacts worst to being told what to do is not the sceptic. It is the loyalist.
A 2017 paper in the Journal of Consumer Psychology, led by Yael Zemack-Rugar, argues that a committed relationship with a brand carries its own norms of compliance. A customer who already loves the brand feels more pressure to do as the advert says, and anticipates feeling guilty about ignoring it. The pressure raises reactance, which then eats the compliance the advert was asking for. Liking of the advert, liking of the brand and the money participants were willing to put towards it all fell. The authors' own summary is blunt: "Our results show the perils that assertive ads pose for marketers and their most valuable customers."
The Illusion of Choice, Richard Shotton's 2023 book on consumer bias and the seed for this piece, uses that study in its chapter on freedom of choice and credits it in the prose to its third author. The house rule is to credit the first author, which here means Zemack-Rugar.
A second moderator arrives with field data behind it, which is rare here. Ann Kronrod and colleagues, in the Journal of Marketing in 2012, ran three laboratory studies and a field experiment on Google search ads, across water saving, plastic recycling and pollution. Pushiness turned out to be conditional on how much the audience already cared. People "respond better to pushy requests in domains that they view as important", and want gentler appeals where they lack conviction to begin with. The field experiment counted clicks. It did not measure reactance, and no commercial field evidence here does.
What Brehm proposed and nobody could measure
Jack Brehm set out the theory in a monograph published in 1966, and it states in a sentence. Threaten or remove a specific freedom and you produce a motivational state aimed at getting it back, with the threatened option looking better than it looked before. We have not read that monograph, so nothing here is quoted from it.
The 1981 restatement that carried the idea into modern psychology is usually cited as Brehm and Brehm with Jack first. Sharon Brehm is its first author. That book also sets out the problem which held the field up for two decades: the state cannot be measured directly, only inferred from what somebody does next.
Mona Clee and Robert Wicklund moved the idea into shopping in 1980, casting promotional pressure, product unavailability and government regulation as events that threaten a shopper's freedom. It is a conceptual review with no data in it, and the origin of the consumer framing rather than evidence for it.
What Dillard and Shen made countable
James Price Dillard and Lijiang Shen broke the measurement deadlock in 2005 with two parallel studies, one pressing students to floss and one pressing them to drink less, just under two hundred participants in each. They specified four rival accounts of what the state is made of and tested all four. The winner was unglamorous. A single composite of anger and negative thoughts about the message carried the whole effect of a threatened freedom on attitude and intention.
Brian Quick and Michael Stephenson repeated that approach in 2007 across seven condom television advertisements, and it held in every one. Stephen Rains then pooled 20 studies and 4,942 participants in 2013 and found the best-fitting account was reactance as one underlying state with anger and counter-arguing as its two visible signs. Those papers settle what the thing is. None of them says how large its effects are, and the distinction gets lost constantly.
Two meta-analyses published in 2025 are where the sizes live. Zixi Li and Jingyuan Shi assembled 28 articles and 33 studies published between 2005 and 2024. Strongly freedom-threatening language raised anger, negative thoughts and measured reactance, each by a small to moderate amount, and anger and negative thoughts were in turn associated with less persuasion, again small to moderate.
The same analysis tested whether framing a message as a gain rather than a loss changed any of that. There was no detectable difference on anger, on negative thoughts or on reactance. Not a small difference, and not a demonstrated zero either. Nothing the pooling could detect.
The second, led by Rong Ma, covered 35 health studies and 10,658 people and landed in the same place with a frank caveat. The findings "generally align with prior theorising, with significant heterogeneity across studies", and the field still needs "additional work to unravel underlying mechanisms and define the scope of boundary conditions".
What a warning label does to desire
Resisting a message and wanting the restricted thing are two separate claims, and almost everything written on the subject runs them together. The pooled evidence above is about resistance. Desire has a thinner, older and more interesting file.
Brad Bushman asked 900 people, across five age bands from children of nine to adults over twenty-one, to read descriptions of violent and non-violent television programmes carrying either a warning label, an information label, or nothing, and say how much they wanted to watch. "Warning labels drew people of all ages to violent TV programs. Information labels did not draw viewers to violent programs."
The gap between those two labels is the mechanism. A warning tells you what to do and an information label tells you what is in the programme, so one threatens a freedom and the other does not. That is as close to a clean test as this corner of the literature gets, and it is worth holding in mind when reading what a warning label on a cigarette packet does.
Nobody in that study watched anything: the outcome was a stated wish to watch a programme they had read about. Three earlier experiments by the same researcher, published in 1996, measured the same sort of thing.
The oldest case is more fragile again. Stephen Worchel and colleagues, in 1975, had people rate cookies that were plentiful, scarce, or newly scarce because other people wanted them. Scarce cookies were rated better, and cookies made scarce by demand were rated best of all. The authors read their own results as an extension of commodity theory, and offered reactance only as a hypothesised intervening process for some of them.
What the chapter says about culture
Chapter 12 of The Illusion of Choice is a good short account of the theory for a commercial reader, and its advice, that mandates breed resistance so the wording should be softened, points the right way. Two things in it do not survive a check.
The first is clerical. The chapter names Brehm and the year 1966 on page 139 as the origin of the theory, and the chapter's reference list carries no entry for it. Seven entries cover roughly a dozen named claims, and the one that is missing is the monograph the argument rests on. A reader who wants to check the foundation cannot get there from the page.
The second is substantive. The chapter uses Eva Jonas's 2009 work on culture to the effect that individualistic cultures produce more reactance than collectivist ones. The review by Christina Steindl and colleagues, with Jonas herself as second author going back over that study, puts it differently: people with an independent sense of self "are more affected by threats to their individual, personal freedom", while people with an interdependent one are more affected by threats that reach their whole group. That is a finding about which freedom is threatened, not about how much of it a culture generates.
The direction also reverses when the restriction lands on somebody else. In the same 2015 review, participants from collectivist backgrounds felt more reactance on behalf of a restricted stranger, especially a member of their own group, than participants from individualist backgrounds did, while the individualists reacted harder when restricted themselves. The chapter's version inverts half of that. It is a fault in the reading, and the correction sits in a review the researcher it reports co-wrote.
What this evidence cannot settle
Nothing above licenses a prediction. There is still no pooled estimate of reactance outside advertising and persuasion, where almost all of the pooling has been done. It explains why an appeal backfired. It cannot tell you by how much the next one will.
The warning-label studies measured stated wishes rather than behaviour. The advertising pooling could not test product type, calls to action or message length at all, because too few of the underlying studies varied them or published their materials. The two 2025 meta-analyses are health and persuasion evidence, and the commercial base under all of this is thin: one advertising meta-analysis, one marketing field experiment on green messaging, one consumer-psychology paper on brand commitment.
The best-studied way of defusing reactance is the "but you are free" line tacked onto a request. A meta-analysis led by Adrien Fillon in 2023, preregistered, meaning its analysis plan was filed before the results were seen so it could not be chosen afterwards, pooled 52 experiments and 19,528 people and found a moderate overall effect. Only seven of those experiments were rated at low risk of bias, and pooling those seven on their own found nothing at all. The technique belongs with the other findings that did not survive replication until somebody runs it properly.
Two worlds are on sale here at the same price. In one, reactance is a real cost centre, and every mandate softened and every "but you are free" bolted on buys back persuasion you can bank. In the other, it is a dependable annoyance with almost no price attached, and the budget spent policing the tone of the copy would have done more work inside the offer. The advertising evidence sits much closer to the second. Nobody is asking which one they bought.
Common questions
Why does being told no make you want it?
Sometimes, and not in the way the phrase suggests. What the evidence shows reliably is resistance to the message: an instruction raises how threatened people feel and lowers how persuasive they find it. Wanting the restricted thing more is a separate claim, supported mainly by warning-label studies in which 900 people said how much they wanted to watch labelled violent television.
How much does reactance cost an advertiser?
Very little, on the best available estimate. A 2026 meta-analysis of 142 samples led by Tsz Hang Chu found that assertive advertising language produced a large rise in felt threat to freedom, a moderate rise in the state itself, and a very small negative effect on persuasion. The annoyance is large and the commercial damage is very small, on measures that are not directly comparable.
Does "but you are free" actually work?
Nobody yet knows. A 2023 preregistered meta-analysis led by Adrien Fillon pooled 52 experiments and 19,528 people and found a moderate effect overall, but the seven experiments rated at low risk of bias showed no effect at all. Treat it as an untested tactic rather than a reliable one.
Do individualistic cultures show more reactance?
No, and this is a common misreading. The finding from Eva Jonas's work, as her own 2015 review with Christina Steindl states it, is about which freedom is threatened: independent selves react to personal restriction, interdependent selves to restriction of their group. Where the restriction was imposed on somebody else, collectivist participants reacted more strongly, not less.
Sources
Brehm, J. W. (1966). A Theory of Psychological Reactance. New York: Academic Press. No DOI; monograph predates DOI registration. [Catalogue record read, Open Library, work OL9393463W; the monograph itself was not read and is not quoted]
Brehm, S. S., & Brehm, J. W. (1981). Psychological Reactance: A Theory of Freedom and Control. Academic Press. Chapter DOIs under prefix doi:10.1016/b978-0-12-129840-1. [Catalogue record read, Open Library, work OL2024889W; first author is Sharon S. Brehm]
Clee, M. A., & Wicklund, R. A. (1980). 'Consumer Behavior and Psychological Reactance.' Journal of Consumer Research, 6(4), 389 onwards. doi:10.1086/208782. [Abstract read, OpenAlex record; conceptual review with no data]
Worchel, S., Lee, J. W., & Adewole, A. (1975). 'Effects of supply and demand on ratings of object value.' Journal of Personality and Social Psychology, 32(5), 906-914. doi:10.1037/0022-3514.32.5.906. [Abstract read, OpenAlex record]
Bushman, B. J., & Stack, A. D. (1996). 'Forbidden fruit versus tainted fruit: Effects of warning labels on attraction to television violence.' Journal of Experimental Psychology: Applied, 2(3), 207-226. doi:10.1037/1076-898x.2.3.207. [Abstract read, OpenAlex record]
Bushman, B. J. (2006). 'Effects of Warning and Information Labels on Attraction to Television Violence in Viewers of Different Ages.' Journal of Applied Social Psychology, 36(9), 2073-2078. doi:10.1111/j.0021-9029.2006.00094.x. [Abstract read, Crossref deposited abstract; N = 900 across five age groups]
Dillard, J. P., & Shen, L. (2005). 'On the Nature of Reactance and its Role in Persuasive Health Communication.' Communication Monographs, 72(2), 144-168. doi:10.1080/03637750500111815. [Abstract read, publisher record; two studies, N = 196 and N = 200]
Quick, B. L., & Stephenson, M. T. (2007). 'Further Evidence That Psychological Reactance Can Be Modeled as a Combination of Anger and Negative Cognitions.' Communication Research, 34(3), 255-276. doi:10.1177/0093650207300427. [Abstract read, publisher record; sample size not stated in the abstract]
Rains, S. A. (2013). 'The Nature of Psychological Reactance Revisited: a Meta-Analytic Review.' Human Communication Research, 39(1), 47-73. doi:10.1111/j.1468-2958.2012.01443.x. [Abstract read, publisher record; K = 20 studies, N = 4,942]
Steindl, C., Jonas, E., Sittenthaler, S., Traut-Mattausch, E., & Greenberg, J. (2015). 'Understanding Psychological Reactance: New Developments and Findings.' Zeitschrift fuer Psychologie, 223(4), 205-214. doi:10.1027/2151-2604/a000222. PMC4675534. [Full text read, PubMed Central, open access]
Jonas, E., Graupmann, V., Niesta Kayser, D., Zanna, M. P., Traut-Mattausch, E., & Frey, D. (2009). 'Culture, self, and the emergence of reactance: Is there a "universal" freedom?' Journal of Experimental Social Psychology, 45(5), 1068-1080. doi:10.1016/j.jesp.2009.06.005. [Bibliographic record verified at Crossref and OpenAlex; abstract not read, so this study is reported here only as summarised by Steindl et al. 2015]
Zemack-Rugar, Y., Moore, S. G., & Fitzsimons, G. J. (2017). 'Just do it! Why committed consumers react negatively to assertive ads.' Journal of Consumer Psychology, 27(3), 287-301. doi:10.1016/j.jcps.2017.01.002. [Abstract read, Crossref deposited abstract]
Kronrod, A., Grinstein, A., & Wathieu, L. (2012). 'Go Green! Should Environmental Messages be So Assertive?' Journal of Marketing, 76(1), 95-102. doi:10.1509/jm.10.0416. [Abstract read, Crossref deposited abstract; three laboratory studies and one Google AdWords field experiment]
Li, Z., & Shi, J. (2025). 'Message effects on psychological reactance: meta-analyses.' Human Communication Research, 52(1), 38-52. doi:10.1093/hcr/hqaf016. [Abstract read, Crossref deposited abstract; 28 articles, 33 studies, 146 effect sizes, 2005 to 2024]
Ma, R., Ma, Z., Kalny, C. S., & Walter, N. (2025). 'Words that trigger: a meta-analysis of threatening language, reactance, and persuasion in health.' Journal of Communication, 75(5), 348-357. doi:10.1093/joc/jqaf004. [Abstract read, Crossref deposited abstract; k = 35, combined N = 10,658]
Chu, T. H., Liu, D., & Gan, H. (2026). 'Should Ads Be Assertive? A Meta-Analytic Review of the Effects of Assertive Language on Perceived Freedom Threat, State Reactance, and Persuasion Outcomes.' Journal of Advertising, advance article, 1-20. doi:10.1080/00913367.2026.2725644. [Full text read, Taylor and Francis, open access; 142 samples, d = 0.72 freedom threat, d = 0.39 state reactance, d = minus 0.08 persuasion]
Fillon, A., Souchet, L., Pascual, A., & Girandola, F. (2023). 'Effectiveness of the "But-you-are-free" technique.' Meta-Psychology, 7. doi:10.15626/mp.2020.2640. [Abstract read, Crossref deposited abstract; 52 experiments, N = 19,528, overall g = 0.44, g = 0.11 in the seven low-risk-of-bias studies]
Shotton, R. (2023). The Illusion of Choice: 16 1/2 Psychological Biases That Influence What We Buy. Harriman House. Chapter 12, Freedom of Choice, pages 139 to 148. [Read in full; vault reliability audit at Books/The Illusion of Choice/The Illusion of Choice.md]