Reactance

Evidence

Brehm set out the theory in 1966, in a book that is not available here; it is therefore reported from his own summary of 2006. When a person sees their freedom of action threatened, a motivational state arises aimed at restoring that freedom. In practice this means the threatened option becomes more attractive and the pressing source loses credibility.

Brehm himself held that reactance could not be measured. “We cannot measure reactance directly”, Brehm and Brehm write in 1981; it has the status of a hypothetical intervening variable. Dillard and Shen (2005) disputed that and tested four models against each other: reactance as pure cognition, as pure feeling, as two separate routes, and as an alloy of both. Two parallel studies with 196 and 200 students in Wisconsin, on flossing and on binge drinking. The alloy fitted best (flossing: AGFI .94 against .87, RMSEA .006 against .110), and Rains (2013) confirmed it meta-analytically across 20 experiments with 4,942 participants — while calling the support for it tentative.

The copy Dillard and Shen worked with is crude. The pressing version read, among other things, “as any sensible person can see, there is really no choice when it comes to flossing: you simply have to do it” and “only a fool would possibly argue with it”. The restrained version said “you may want to try it today”. No shop writes like the first.

And the absolute values temper the expectation considerably. On a five-point scale the pressing version reached 2.98 against 2.31 on perceived threat to freedom — barely past the midpoint. Anger, on a scale from 0 to 4, came to 0.90 against 0.45 (flossing) and 1.44 against 0.70 (alcohol). The differences are significant; the amount of anger is not: most participants were barely annoyed.

Three figures from it temper the expectation. A threat to freedom correlates with anger at r = 0.23 and with rejecting thoughts at r = 0.21. The r measures how closely two quantities go together; 0 means not at all, and values around 0.2 count as small. Anger and rejecting thoughts correlate with each other at only r = 0.31, although in this model they are supposed to indicate the same thing. And the direct link between a threat to freedom and attitude was not significant (r = 0.06, confidence interval −0.03 to 0.14). Only the indirect route through reactance is demonstrable.

Brehm himself summarised the theory forty years on (Miron & Brehm, 2006), and that paper also carries the experiment from the 1966 book. Eighth-graders rated several films; some were told they would get to choose, others that they would be assigned one. Then the film each had ranked second dropped out. Those who lost one of three options found the lost film markedly more attractive than those who lost one of six — and only among those who had expected a choice of their own. In the same paragraph the authors record that later work has produced both confirming and disconfirming evidence on this point.

Brehm later qualified the theory himself. Reactance does not simply rise with pressure, he argued, but depends on the difficulty of restoring the freedom: little where it is easy, most where it is moderately hard, none where it is impossible — resignation then takes its place (Miron & Brehm, 2006).

That point, however, is not established. Steindl and colleagues (2015) list it explicitly as an open question: what becomes of reactance when a freedom cannot be restored is unresolved, and more research is needed on the role of perceived difficulty. Only the direction has support so far, and it comes from elsewhere: Laurin, Kay and Fitzsimons (2012) found that people rationalise a restriction when it is unavoidable and respond with reactance when it is not. Both are stronger when the matter concerns them personally. This finding is reported here at second hand, through Steindl and colleagues.

And reactance does not always win out. Miron and Brehm list counterforces that suppress it: a powerful sender, an expected further dealing with that sender, group pressure, and anticipated regret — people who first consider what resistance will cost them show less of it. Their conclusion: where the counterforces exceed the urge to restore, people relinquish the freedom.

Steindl and colleagues (2015) carry the research on from Miron and Brehm; Rosenberg and Siegel (2018) sort five decades into five waves. Three findings from those reviews are directly usable for shop copy.

First: the sender counts. The same patronising message produced more perceived threat and more reactance the more human it appeared — plain text, text with a picture of a robot figure, text with a short film clip of the same figure (Roubroeks et al., 2011). For chat assistants on shop pages that is the most relevant line in this literature.

Second: loss framing creates threat, but not the same mechanism. “When you do not use sun protection you will pay costs” produced more perceived threat to freedom than the gain version — and there the threat correlated with anger, not with rejecting thoughts (Cho & Sands, 2011). So the combined model does not hold equally everywhere.

Third: a reason changes the course. Where a restriction came without a reason, heart rate rose immediately; where it was justified, it rose only after a delay. The reason does not prevent the reaction, it shifts it from reflex to reflection.

For marketing this is less a tactic than a side effect: reactance is why the same lever that works in moderation can flip at higher intensity. Scarcity cues, countdown timers, repeated reminder emails and very direct address are the typical triggers.

Limits

The evidence comes mostly from health communication. The 20 experiments in Rains concerned flossing, alcohol, exercise, sun protection, organ donation, soft drinks, phones in lecture halls and the temperature setting on a washing machine. Exactly one involved a commercial product, a camera phone. Whether a pricing page triggers the same reaction is unresolved.

Reactance is an explanation that easily explains everything. If a measure works, the pressure was well judged; if it does not, it was reactance. Without measuring the intermediate state that is not a test but a retrospective reading. Rains supplies his own evidence for that: he expected his model to fit better where the threat to freedom was stronger. It fit equally well however strong the pressure was.

Measurability is an open problem of the theory itself. Rodrigues and colleagues (2019) reviewed 382 works from five decades bibliometrically. Their conclusion states that it is practically impossible to establish whether reactance is actually occurring: a manipulation check such as “please indicate the extent to which you felt your freedom of choice was threatened by this advertising” they consider unusable. They quote an assessment from 1969 that the weak state of the evidence stems from the phenomenon being hard to measure experimentally. Brehm’s original record experiment, too, they report as having been contested repeatedly because the results were only marginally significant.

That work is not itself a study of the effect. Rodrigues and colleagues count citations and group them; about the size or existence of reactance they say nothing from their own measurement. Their statistical basis is thin as well — the sampling adequacy of their factor analysis was initially below the usual threshold and was reached only by dropping five works. It stands here solely for its methodological admission and the historical detail.

German-speaking countries respond differently from collectivist ones. People who see themselves as independent persons — in this research, Germany and Austria — react more strongly to threats against their own freedom; people who see themselves more through the group react more strongly when the group’s freedom is threatened or when someone else is affected (Jonas et al., 2009; Sittenthaler and colleagues). Findings from US samples therefore do not transfer straightforwardly, and for a German-speaking market the direction suggests more reactance to direct pressure rather than less.

Both reviews advise against measuring reactance as a personality trait. Rosenberg and Siegel (2018) work through the three common scales: of Merz (1983) they record that it lacked the psychometric quality to warrant continued use; of the Therapeutic Reactance Scale they quote that its use in its current form should be undertaken with caution; of the Hong scale, that without a major revision practitioners would be well advised to discontinue it. Miron and Brehm (2006) write that there is little to gain from conceiving of reactance as a personality trait.

One finding even reverses the expectation. Silvia (2006) found that a pushy message went down best with precisely those who scored high on the reactance scale — the people who should have disliked it most disliked it least. This finding is reported here at second hand, through Rosenberg and Siegel.

The newer state scales do not convince the authors either. Rosenberg and Siegel object that the items of the two more recent instruments capture the precondition of reactance — the perceived threat to freedom — rather than reactance itself. Their recommendation remains the combined model of anger and rejecting thoughts, which is exactly the model whose support Rains calls tentative.

Reactance is bound to the situation, not to a principle. Miron and Brehm stress that the theory deals not in general precepts but in specific behaviours within a particular context: what counts as a freedom in one setting does not in another. A sentence read as an overreach here may be of no consequence elsewhere.

Brehm’s 1966 book is no longer in the source list. Its central experiment is documented in Miron and Brehm (2006) — written up by Brehm himself — so the finding is supported without the book.

The more recent papers supply mechanisms and boundary conditions, not a larger effect.

Rosenberg and Siegel are cited here from the author manuscript in a university repository, not from the journal’s typeset version. Differences in content and in page numbering are possible.

The model replicates; the effect does not do so throughout. Rains (2013) confirmed Dillard and Shen’s model across 20 experiments. In that same meta-analysis, however, the direct link between a threat to freedom and attitude was not significant (r = 0.06). Only the indirect route through reactance is demonstrable — a mechanism, then, whose end point cannot be measured directly.