Skip to main content
Category: Effects
Type: A belief, attitude, or behavior that moves opposite the position a persuasive message advocated. A result that does not move is a miss, not this effect
Origin: Xinyan Zhao and Edward L. Fink, Communication Monographs 88(3), 3 July 2021, pages 286–305, state the definition used here. The counted case is Robert Hornik, Lela Jacobsohn, Robert Orwin, Andrea Piesse, and Graham Kalton, American Journal of Public Health 98(12), December 2008, pages 2229–2236
Also known as: negative persuasion, Abelson and Miller’s 1967 name for an insult that made the hearer’s original stand more extreme. It is not a second title
Quick Answer — The Boomerang Effect is a later belief, attitude, or behavior that sits opposite the position a message asked for. Zhao and Fink state that definition in 2021 and separate it from a failure to persuade. In the 2008 evaluation of a national antidrug campaign, one pooled lagged association was γ = −0.07: more recalled exposure went with a lower share of prior nonusers who said they definitely would not use marijuana. Most tests in that paper found no effect. The authors call the unfavorable pattern irregular and the pro-marijuana claim tentative.

​
What is the Boomerang Effect?

The Boomerang Effect is movement of a message recipient’s belief, attitude, or behavior in the direction contrary to the position the message advocated.
A boomerang effect involves the movement of a message recipient’s belief, attitude, or behavior in a direction contrary to the advocated position in the message.
That sentence is the opening of Zhao and Fink’s abstract in 2021. The everyday picture is a thrown stick that curves back toward the hand that threw it. The measurement is not the stick. It is the later stand, set beside the stand the message asked for. A speech people ignore has not yet done this. The second stand has to point the other way. Psychological Reactance is a push back when a freedom feels threatened. That push can be one path. It is not this page’s name. Backfire Effect is a correction that leaves a false belief more convincing for some people. A correction is not the message this page starts from.

​
The Boomerang Effect in 3 Depths

  • Beginner: Write the ask on one line and the later stand on the next. Opposite directions are the Boomerang Effect. Silence, a shrug, or a flat score is a miss.
  • Practitioner: Name the row before you quote a gamma. In 2008 the starred pooled intention row was γ = −0.07 for general exposure. The same table’s pooled initiation row did not clear .05.
  • Advanced: The authors would not call even an antidrug success definite unless the trend, the same-round association, and the next-round association agreed. Their pro-marijuana reading failed that bar. They call it tentative.

​
Origin

Xinyan Zhao and Edward L. Fink, “Proattitudinal versus counterattitudinal messages: Message discrepancy, reactance, and the boomerang effect,” Communication Monographs 88(3), 3 July 2021, pages 286–305, give the definition above. The journal record lists an early online date of 2 September 2020. The DOI is 10.1080/03637751.2020.1813317. Zhao’s byline is the Hussman School of Journalism and Media at the University of North Carolina at Chapel Hill. The abstract says a between-subjects factorial study, N = 458, found a belief boomerang after a counterattitudinal message and persuasion after a proattitudinal message. It says a reactance process accounted for the path from discrepancy to that belief boomerang. The PDF was not re-read. No cell mean, no F, and no issue topic from that article is on this page. The abstract also says the outcome is distinct from a failure to persuade. Muzafer Sherif, Daniel Taub, and Carl I. Hovland, Journal of Experimental Psychology 55(2), 1958, pages 150–155, already call the social pattern “the frequently mentioned ‘boomerang effect.’” Their introduction says that under some conditions a new stand moves the person toward it, and under others the person rejects the proposal and sits more firmly in the original position. The publisher record is 10.1037/h0048784. The study that follows is a judgment of simple stimuli with anchors. Those stimulus results are not on this page. Carl I. Hovland, Irving L. Janis, and Harold H. Kelley, Communication and Persuasion (Yale University Press, 1953), is the volume later reviews often cite for the name. That book was not re-read. No sentence from 1953 is on this page. Some citations spell the third author Kelly. The book spells Kelley. Carl I. Hovland, O. J. Harvey, and Muzafer Sherif, Journal of Abnormal and Social Psychology 55(2), September 1957, pages 244–252, DOI 10.1037/h0048480, is the paper those authors use for a controversial issue. The PsycInfo abstract says a communicator unknown to the participant took a stand on prohibition of alcohol, in a dry state where the issue was live. It says distance from the person’s own stand, together with latitudes of acceptance and rejection, might predict reactions and susceptibility to change. The PDF was not re-read. No mean from 1957 is on this page. Sherif and Hovland, Social Judgment (Yale University Press, 1961), volume 4 of Yale Studies in Attitude and Communication, was not re-read. No table from that book is on this page. Robert P. Abelson and James C. Miller, “Negative persuasion via personal insult,” Journal of Experimental Social Psychology 3(4), October 1967, pages 321–333, DOI 10.1016/0022-1031(67)90001-7, tested an insult during an attempt to persuade. The abstract says the field setting was a park-bench talk arranged by a public-opinion interviewer, with a confederate as the partner. In the critical variation the confederate insulted the person. Other variations persuaded without an insult, or gave arguments without trying to persuade. Attitude change supported a more extreme original stand after the insult. Whether the insulter stayed or left, and whether a hostile crowd was present, did not produce significant differences. The abstract says the mechanism remained unexplained, though a social-equity account had some plausibility. The PDF was not re-read. No sample size and no quoted insult from that article is on this page.

​
Key Points

Four readings keep a miss, a reversal, and a neighboring effect on separate lines.
1

Score the direction before the name

A message names a direction. The later measure can move toward it, stay flat, or move away. Only the third is the Boomerang Effect. Zhao and Fink’s abstract treats that third result as different from a failure to persuade. A class that still holds yesterday’s view after a lesson has not supplied the third result. Sleeper Effect is a later rise in agreement after a discounting cue. The cue there held the message down at first. It is not an ask that the later stand then opposes.
2

Keep the next round apart from the same round

Hornik and colleagues compared recalled exposure with marijuana outcomes in the same round and, separately, with outcomes at the next round. Same-round associations, after adjustment, gave little evidence of a link. The unfavorable pattern they discuss sits on the next-round rows. A dashboard that mixes today’s recall with today’s attitude has not read the row they marked.
3

Read γ = −0.07 as gamma, not as a use rate

Goodman and Kruskal’s gamma runs from −1 to 1. Zero is no association. In the pooled next-round table, among prior nonusers, the share who said they definitely would not use marijuana was 82.3 percent, 78.2 percent, and 78.4 percent across rising bins of general antidrug exposure. The gamma was −0.07, with a 95 percent interval from −0.13 to −0.01. The two higher bins sit below the lowest bin. 78.4 is not a further step down from 78.2. Do not subtract 82.3 from 78.2 and call the difference the effect.
4

Leave a warning label on its own page

Forbidden Fruit Effect is a higher wish-to-watch rating when a violent description adds advice. That rating is not a gamma on marijuana intentions. Abelson and Miller’s abstract supports a more extreme original stand after an insult. It does not print a tactic, and this page does not copy an insult. A far stand on prohibition is the 1957 topic. No 1957 percentage is printed here, so distance alone is not a guarantee of reversal.

​
Applications

Write the aimed direction, the measured direction, and the row before a reversal leaves the table it came from.

Put toward, flat, and away on three lines

Before a memo uses the Boomerang Effect, write what the message asked, what the next measure did, and which comparison was tested. Toward is persuasion. Flat is a miss. Away is the only line that matches this page. Zhao and Fink’s abstract says the distinction matters.

Separate the intention row from initiation

The starred pooled intention gamma is −0.07 for general exposure. The specific-ad intention gamma in that same table is −0.02, interval −0.07 to 0.03, and it is not marked. Pooled initiation is 10.7 percent, 11.2 percent, and 12.5 percent, γ = 0.07, interval −0.01 to 0.15, also not marked. Quote the row you mean.

Do not move the campaign onto a class

The counts come from in-home interviews of U.S. youths, ages 12.5 to 18 at the analyzed outcome, from 1999 to 2004. A lesson, a household rule, or a poster in another year was not that sample. A flat class discussion is not γ = −0.07.

Name the message before you pick a page

A threat to a freedom is the reactance page. A content warning on a program is the forbidden-fruit page. A correction of a false claim is the backfire page. A weak source whose conclusion looks stronger later is the sleeper page. The Boomerang Effect starts only when you can point to the ask and to a later move against it.

​
Case Study

Between 1998 and 2004 the U.S. Congress appropriated nearly $1 billion for the National Youth Anti-Drug Media Campaign. The White House Office of National Drug Control Policy supervised the campaign. The National Institute on Drug Abuse supervised the evaluation, which Westat carried out with the Annenberg School for Communication at the University of Pennsylvania. Robert Hornik and Lela Jacobsohn were at Annenberg. Robert Orwin, Andrea Piesse, and Graham Kalton were at Westat in Rockville, Maryland. The paper is American Journal of Public Health 98(12), December 2008, pages 2229–2236, DOI 10.2105/AJPH.2007.125849. Accepted 18 February 2008. The PMC text was read. The campaign bought advertising from September 1999 through June 2004 and aimed it at youths aged 9 to 18, their parents, and other adults. The youth goals were to reject illegal drugs, to avoid starting, especially marijuana and inhalants, and to have occasional users stop. The plan expected a youth to see 2.5 targeted ads a week. Sixty-four percent of the purchased gross rating points were on television and radio. The National Survey of Parents and Youth interviewed youths at home in four rounds. The counts were 8,117, 6,516, 5,854, and 5,126. The first-round response rate was 65 percent, and 86 percent to 93 percent of youths still eligible were interviewed later. Questions played through headphones. The outcome analyses use ages 12.5 to 18. The question was whether youths who recalled more antidrug ads held more antidrug intentions, norms, and beliefs, and whether fewer of them started using marijuana. Exposure was recalled, not assigned. General exposure asked how often the youth had seen or heard antidrug ads. Specific exposure played recent campaign television and radio ads and asked whether the youth had seen or heard them. More than 150 possible confounders were considered. Adjustment used propensity scores. The association is Goodman and Kruskal’s gamma. A same-round gamma cannot say which way the influence ran. A next-round gamma asks whether earlier exposure lines up with a later outcome after earlier confounders, including the earlier outcome, are held aside. Youths reported a high dose of messages. Ninety-four percent reported one or more general antidrug messages a month, with a median of about two to three ads a week. Fifty-four percent recalled at least weekly exposure to specific campaign television ads that had aired recently. Across the campaign, 15 percent, 31 percent, 38 percent, and 16 percent recalled under one, one to under four, four to under twelve, and twelve or more of those television ads a month. From 2000 to 2004 the prevalence of marijuana use in the 12.5-to-18 band did not change. Beliefs and attitudes moved a small way in the antidrug direction. Intentions, norms, and self-efficacy did not show a matching gain. Some year-to-year shifts were significant, including an antidrug shift in intentions from 2002 to 2004. The authors say a trend by itself cannot award the campaign success or failure. Same-round associations, after adjustment, showed little evidence that more exposure went with more antidrug cognitions among nonusers. In subgroup cuts by age, gender, race or ethnicity, and risk, 3 of 80 gammas were significant. The authors say those three may easily be chance. The next-round table pools the pairs of rounds. It is limited to youths who had not used marijuana and who were aged 12.5 to 18 at the earlier round. For general exposure, the adjusted shares saying they definitely would not use marijuana were 82.3 percent under four exposures a month, 78.2 percent at four to eleven, and 78.4 percent at twelve or more. Gamma was −0.07 (−0.13, −0.01), marked significant below .05. The specific-ad intention row was 84.0, 78.8, 77.6, and 78.5 percent, γ = −0.02 (−0.07, 0.03), not marked. The specific-ad social-norms index ran 107.1, 91.3, 83.3, and 76.9, γ = −0.05 (−0.08, −0.02), marked. That index had been scaled so an earlier group of nonusers had a mean of 100. Those points are not percentages. Pooled initiation was 10.7, 11.2, and 12.5 percent for the three general-exposure bins, γ = 0.07 (−0.01, 0.15), not marked. Two of ten pooled associations were significant, both in the pro-marijuana direction. Six of the other eight pointed that way without clearing .05. Read by round pair, Table 4 has no significant antidrug lagged association, and each pair has at least one significant pro-marijuana association. General exposure from round 1 to round 2 has γ = −0.16 (−0.27, −0.06) for the “definitely will not” item and γ = −0.06 (−0.12, −0.01) for norms. General exposure from round 3 to round 4 has γ = 0.11 (0.00, 0.22) for initiation. The paper calls that initiation link barely significant. The abstract’s −0.07 and −0.05 match the pooled rows. They are not the −0.16. Among never-users aged 12.5 to 18, 9 percent of those who said “definitely will not” reported use 12 to 18 months later, against 39 percent of those who gave any other answer. That pair is the item’s link to later use. It is not an effect of the ads. The discussion says an unfavorable influence of exposure is sometimes called a boomerang effect. The authors offer two ideas and call both speculative. One is reactance. They say analyses reported elsewhere did not support it. Those reports were not re-read. The other is that the ads carried a meta-message that marijuana use was common. They say more exposure lined up with that belief, and the belief lined up with later initiation. They mark those associations “data not shown.” No coefficient from that sentence is on this page. Their closing standard is the one they set at the start: they would claim an antidrug effect only if trends, same-round associations, and next-round associations all moved that way. Against that bar, a pro-marijuana claim has tentative, not definitive, support. The pattern was missing from the same-round tests, significant for only some lagged outcomes and periods, and tied to initiation only from round 3 to round 4. The lesson is narrow. A large antidrug buy can show a next-round association opposite the ask on some rows and no association on most others. The Boomerang Effect here is that opposite association, not the appropriation, and not the unchanged prevalence. Palmgreen and colleagues reported antidrug effects in two medium-sized cities for one six-month stretch of related advertising. That report was not re-read. No city percentage is on this page.

​
Boundaries and Failure Modes

The Boomerang Effect requires a measured move against the ask. Hornik and colleagues’ own same-round tests mostly did not find one. Three significant gammas inside 80 subgroup tests are the kind of count they say may be chance. A national prevalence that did not change from 2000 to 2004 is not a hidden reversal. In 2000, more than half of youths already reported recent antidrug messages from other places: 66 percent in school, 52 percent in talks with friends about bad consequences, 54 percent in two or more talks with parents, and 54 percent in weekly coverage that was not advertising. Two ads of 15 to 30 seconds are up to about one minute a week. The authors say that small addition makes a null easier to imagine. It does not turn the null into a boomerang. A second boundary is the authors’ own caution. The unfavorable pattern was lagged, irregular, and not a randomized assignment of ads. Recall can move with the outcome as well as the other way. The reactance account of these ads was not supported in the analyses they cite, and those analyses are not reprinted here. The meta-message account has no coefficient on this page. Unintended Consequences is a result that was not the aim, including a side effect that does not reverse the ask. The Chinese page for that law lists 回力镖效应 among its aliases. That alias is not this gamma. Cobra Effect is a reward that pays people to produce the thing you meant to reduce. A bounty is not a persuasive sentence. The common misuse is to treat a harder slogan, or the 1967 insult, as a tool for flipping a person. Abelson and Miller reported a more extreme stand after an insult. They did not publish a script, and this page has no cell from their PDF. Exporting −0.07, 82.3 percent, or $1 billion onto a classroom, a household, a meeting, or another year changes the study.

​
Common Misconceptions

Three swaps turn a careful evaluation into a slogan. Each one drops a comparison the papers actually printed.
Reality: Reactance is the tension after a freedom feels threatened. The Boomerang Effect is the later move against the ask. Zhao and Fink’s abstract says a reactance process accounted for discrepancy in their factorial study. Hornik and colleagues say a reactance reading of the antidrug ads was not supported in analyses reported elsewhere. One path in one experiment is not the name of every opposite row.
Reality: Failure to persuade leaves the stand where it was. The Boomerang Effect moves it the other way. In 2008 most analyses showed no effect. Same-round tests were largely null. Only two of ten pooled next-round associations cleared .05. A flat row is not γ = −0.07.
Reality: Gamma is not a percentage of youths and not a share of the appropriation. The −0.07 is the pooled association between general exposure and the next-round “definitely will not” item among prior nonusers. Prevalence of use in the age band did not change from 2000 to 2004. The initiation gamma that is marked is a different number, 0.11, from round 3 to round 4, with an interval that starts at 0.00.
These pages meet the Boomerang Effect at the edge of a message. None of them is the gamma on the next-round intention row.

Psychological Reactance

A motive to restore a freedom that feels threatened. It can sit on the path to an opposite move. It is not the 2008 gamma, and the campaign authors did not find support for it in analyses reported elsewhere.

Forbidden Fruit Effect

A higher wish to watch a violent program when the description adds a warning. The wish mark is not a marijuana-intention gamma.

Backfire Effect

A correction that can leave a false belief more convincing for some people. The starting message is a correction, not the antidrug ask in the 2008 tables.

Sleeper Effect

Agreement with a discounted message rises later. The discount held the message down at first. The later rise is not a move against the message’s ask.

Unintended Consequences

A result that was not the aim. The Chinese alias line on that page includes 回力镖效应. A side effect that does not reverse the ask is not this page.

Cobra Effect

A reward that makes the punished thing multiply. The mechanism is the payoff, not the direction of a persuasive sentence.

​
One-Line Takeaway

When a message misses, write the ask and the next measure on two lines. Use the Boomerang Effect only when the second line points the other way, and name the row that moved.