Why You Defend What You Just Bought
The trade sells a simple rule: make the customer work for the choice and the customer will defend the purchase. The design that produced the evidence for it turned out to be unable to measure the thing it was pointed at.
Answer to the question
Is post-purchase rationalisation real?
No, and the consumer evidence runs the other way. Alexander Chernev and colleagues pooled 99 observations and 7,202 people in 2015 and found that higher decision difficulty predicts lower satisfaction, more regret and more switching. Post-purchase rationalisation itself is real but small, about half the size the older literature reported, resting on four studies of which two are two-tailed nulls.
Key takeaways
- The corrected meta-analysis of choice-induced preference change pooled only four studies and found an effect about half the size of the uncorrected literature's pooled estimate, with two of the four significant only on a one-tailed test.
- A 2024 multilab replication of induced compliance across 39 laboratories, 19 countries and 4,898 participants found no attitude difference between high-choice and low-choice conditions.
- In consumer research, 99 observations covering 7,202 people show higher decision difficulty predicting lower satisfaction, more regret and higher switching likelihood.
- The only field test of effort justification, in two Dutch studies of 202 and 46 people, found that a harsher initiation lowered liking for the group rather than raising it.
You spent three weeks on the mattress. Two shops, four spreadsheets, a long argument about springs. Ten minutes after the card clears, you are already building the case: the warranty was better, the firmness was right, the other one would have been too soft. Nobody asked. That reflex has a name, post-purchase rationalisation, and the marketing trade has built a rule on top of it. Make the customer work for the choice and the customer will defend the purchase. Add a configurator, add a comparison step, make them earn it, and you buy loyalty on the way out of the checkout.
Take that advice seriously for a moment, because the people selling it are not making it up. There is a seventy-year literature behind it, and the question is what the advice costs when you follow it all the way.
The Promise of a Hard Choice
The promise is specific. A difficult decision is supposed to leave the buyer with a small psychological debt, and post-purchase rationalisation is supposed to settle it by inflating the thing chosen and deflating the thing refused. The harder the call, the bigger the inflation, so a brand that makes you deliberate is buying an advocate.
It is worth saying where the idea does not come from. Daniel Kahneman's Thinking, Fast and Slow is the book usually waved at this claim, and it contains no chapter and no sustained passage on cognitive dissonance. What chapter 32 contains is the emotional bookkeeping that keeps a person tied to a decision already taken: mental accounting, escalation of commitment, regret as a punishment we administer to ourselves, and the sunk-cost fallacy, which is the machinery behind finishing a series you stopped enjoying and a separate argument from this one.
The nearest thing in the book to post-purchase rationalisation is borrowed from Daniel Gilbert and points the other way. People anticipate more regret than they end up feeling, Kahneman writes, because they underestimate how well their own psychological defences will work. That is a claim about defences being stronger than expected, not about effort buying loyalty. The loyalty version belongs to the trade press, and it has an older and more interesting ancestor.
Brehm's Spreading of Alternatives
In 1956 Jack Brehm published the experiment that started it. He had people rate a set of objects, choose between two of them, then rate everything again. After choosing, the chosen item rose in their estimation and the rejected one fell. He called the gap spreading, and he stated three derivations in his own words, the second of which is the exact ancestor of the modern sales pitch: the pressure to reduce dissonance is greater the more closely the two alternatives approach equal desirability. A close call produces a bigger spread.
Brehm credited the design to Leon Festinger's then-unpublished theory, and Festinger's own 1959 study with J. Merrill Carlsmith became the other founding paradigm, in which a person paid a trivial sum to say something they do not believe shifts towards believing it.
Three years after Brehm, Elliot Aronson and Judson Mills tested the effort version directly, asking whether a severe initiation makes people like a group more. Their framing is the most honest sentence in the whole file, and they wrote it in 1959. Is this common observation valid when tested under controlled conditions, and if it is, how can it be accounted for? They also raised the obvious alternative themselves. A club with a brutal initiation is joined only by people who badly want in, so the members would be expected to think highly of it whatever the initiation did. Their experiment exists to strip that confound out.
A Theorem and Its Critics
For fifty years the free-choice design carried the evidence for post-purchase rationalisation, and in 2010 M. Keith Chen and Jane Risen pointed out that it cannot. Their argument is uncomfortable and simple. If a person's ratings are an imperfect measure of what they actually prefer, and their choice is guided even partly by what they actually prefer, then the design will show spreading whether or not any preference changed. The choice reveals information the ratings missed, and the second round of ratings picks it up.
They proved a theorem identifying the conditions for this, then showed experimentally that spreading appeared where choice could not have caused it. Their own statement of scope is maximal: the problem applies to the basic paradigm and to every variant examining moderators and mediators of spreading.
The dissent deserves its best form. In 2015 Carlos Alós-Ferrer and Fei Shi argued that the theorem is mathematically incorrect except in a particular case, presenting a model that satisfies every assumption Chen and Risen make while implying that spreading need not be positive absent any real preference change. Their conclusion is carefully hedged: the paradigm is flawed, and because reasonable models of behaviour need not predict consistent positive spreading, previous results remain informative.
The mathematical dispute is live. The empirical half is not. Three separate studies have since produced significant spreading in the paradigm with no change in true preference, so whatever the proof is worth, the instrument demonstrably reports an effect that is not there.
The Four Corrected Studies
Here is the part that decides the argument, and it arrived in 2013. Keise Izuma and Kou Murayama reviewed the field, ran simulations, and meta-analysed only the studies that had properly addressed the flaw. Three designs qualify. A blind choice, where the chooser cannot see what they are choosing. A rate-rate-choose design, where the artefactual spread is measured and subtracted. An implicit choice design, where preference change is measured for items never directly compared.
| Design | What it was taken to show | Where it stands now |
|---|---|---|
| Free choice, 1956 onwards | Choosing makes you like what you chose | Cannot separate preference change from pre-existing preference |
| Induced compliance, 1959 onwards | Acting against your view shifts your view | Choice manipulation produced no effect across 39 labs |
| Blind choice, 2010 | Choice alone shifts preference | Works as a manipulation, effect small and fragile |
| Rate-rate-choose, 2010 | Spreading survives the correction | Works, and the pooled effect is about half the old one |
Four studies met the bar. Pooled, the effect is small and reliably above zero, and roughly half the size of the uncorrected literature's moderate estimate. Izuma and Murayama put it plainly: choice-induced preference change exists, and past studies substantially overestimated it because of the artefact Chen and Risen identified. Post-purchase rationalisation survives, at half price.
The effect is real, roughly half the size the textbooks reported, and resting on four studies.
Two further facts belong with that number. Two of the four surviving studies, Tali Sharot's blind-choice experiment and Izuma's own brain-imaging study, reach significance only on a one-tailed test, which means that by the conventional two-tailed standard they are nulls. And Louisa Egan, with Laurie Santos and Paul Bloom, corrected their own 2007 children-and-monkeys study with a blind-choice version in 2010, which is to their considerable credit, and Izuma and Murayama still judged the findings not sufficiently compelling because the post-choice preference was itself measured by further choices.
### Thirty-Nine Labs and a Null
The largest test of dissonance came in 2024, and it was not a free-choice study. David Vaidis led a multilab replication of the induced-compliance paradigm across 39 laboratories in 19 countries and 14 languages, with 4,898 participants. Participants wrote a counterattitudinal essay under high choice, the same essay under low choice, or a neutral essay.
The primary analysis failed to support the core hypothesis. No significant difference in attitude appeared between writing the essay under high choice and writing it under low choice, and the authors report that pattern holding up against data exclusions, lab variability and different ways of measuring attitude. A difference did appear between the counterattitudinal and the neutral essay, so something happened. The choice manipulation, which is the whole mechanism, did nothing detectable in nearly five thousand people.
That is a null about induced compliance and not about free choice, and the two paradigms must be kept apart. Conflating them overstates the damage. It is still the biggest single piece of evidence anyone has ever gathered on dissonance, and the thing it failed to find is the ingredient the sales pitch sells.
The Consumer Evidence Against It
Nothing above is a purchase. Every study so far has measured post-purchase rationalisation over stickers, holiday brochures and student essays. The consumer question is whether a harder buying decision predicts a happier, stickier customer, and the best available evidence says the opposite.
Alexander Chernev and colleagues pooled 99 observations and 7,202 people in 2015. Higher decision task difficulty makes choice overload more likely, and the authors report that satisfaction or confidence, regret, choice deferral and switching likelihood are interchangeable measures of it. Translated into trade language, a harder decision predicts less satisfaction, more regret and more switching, across 7,202 people. That is also the literature behind what pooling the choice-overload experiments showed, which found the raw count of options close to inert.
The field test of effort justification goes the same way. Hendrik Lodewijkx and Jacques Syroit ran two longitudinal studies in real Dutch groups, 202 people and 46 people. Severity experienced early in the initiation correlated with frustration and loneliness in the first study and depressed mood in the second, those feelings persisted, and they lowered liking for the group. Only the authors' rival account survived, that companionship predicts attractiveness irrespective of severity.
The dose-response has failed developmentally too. Lauren Marsh and colleagues found in 2017 that the effect named after flat-pack furniture is not moderated by how much effort five and six year olds put in, and that the results are incompatible with existing theories of it. The adult version has a boundary condition worth remembering: in Michael Norton's studies labour led to love only where the task was finished, and we have been through that in full elsewhere.
The Limits and the Verdict
The record has real holes. The corrected-design evidence base appears to still be the four studies pooled in 2013, and we looked for a preregistered replication with a corrected design since then and found none. No sample size is verified for Brehm, for Festinger and Carlsmith, for Aronson and Mills, for Egan, for Sharot or for Izuma, so none is printed here.
The older pooled estimate the corrected figure is compared against comes to us through Izuma and Murayama, because we could not retrieve the 2011 meta-analysis itself. We also searched for a field study testing whether the difficulty of a real purchase predicts later satisfaction or loyalty, which is the question a shop actually has about post-purchase rationalisation, and what came back was cross-sectional survey work in journals we will not cite.
So, our position. Post-purchase rationalisation is real and it is about half the size the textbooks reported, resting on four studies of which two are two-tailed nulls, with a different dissonance paradigm flat across 4,898 people. The specific thing the trade sells, effort buying loyalty, is the part with the field evidence running against it. A hard decision is not a loyalty programme, and this is a different species of failure from the effects that did not survive re-testing, because the instrument was never able to measure the thing it was pointed at.
Across 7,202 people, a harder decision predicts less satisfaction, more regret and more switching.
Which leaves the question nobody has answered. Does a purchase that cost you something in deliberation feel better a year later, or only ten minutes later, when you are still talking yourself into it in the car park? Brehm measured minutes. So did almost everyone who followed him, which is why the correction took half a century to arrive and why the corrected number is half the size of the one the textbooks still carry, and it remains the easiest thing in this entire literature to put right, because it costs nothing but a second measurement taken before the choice rather than after it.
Common questions
Is post-purchase rationalisation real?
The consumer evidence says the opposite. Alexander Chernev and colleagues pooled 99 observations and 7,202 people in 2015 and found higher decision difficulty makes choice overload more likely, with lower satisfaction, more regret and more switching. The only field test of effort justification, in real Dutch groups, found that a harsher initiation lowered liking. Post-purchase rationalisation exists; effort buying loyalty is a separate claim with evidence against it.
How big is the effect once the design is fixed?
Smaller than advertised. Keise Izuma and Kou Murayama's 2013 meta-analysis of the four studies that corrected the design flaw found a small effect reliably above zero, roughly half the size of the older pooled estimate. Two of those four reach significance only on a one-tailed test, so by the usual two-tailed standard they are nulls. The honest summary is a real effect resting on thin evidence.
What was wrong with the original dissonance experiments?
The measurement, not the idea. In 2010 M. Keith Chen and Jane Risen showed that if ratings imperfectly measure preference and choices partly follow preference, the free-choice design produces spreading with no preference change at all. Three later studies demonstrated exactly that. Carlos Alós-Ferrer and Fei Shi dispute the mathematics while accepting the design is flawed.
Did cognitive dissonance fail to replicate?
One paradigm did. David Vaidis led a 2024 replication of induced compliance across 39 labs, 19 countries and 4,898 participants, and found no attitude difference between writing a counterattitudinal essay under high choice and under low choice. That is induced compliance, not the free-choice paradigm, and the two are separate. The free-choice literature was invalidated as a measuring instrument instead.
Sources
Brehm, J. W. (1956). 'Postdecision changes in the desirability of alternatives.' The Journal of Abnormal and Social Psychology, 52(3), 384-389. doi:10.1037/h0041006. [Opening of the article read via OpenAlex reconstructed text; sole authorship verified at Crossref and OpenAlex. NOT VERIFIED: sample size and composition, with the article closed at APA and no deposited abstract, so no N is printed.]
Festinger, L., & Carlsmith, J. M. (1959). 'Cognitive consequences of forced compliance.' The Journal of Abnormal and Social Psychology, 58(2), 203-210. doi:10.1037/h0041593. [Published record read at Crossref, author order confirmed. NOT VERIFIED: sample size, payment conditions as published, and all figures; closed at APA with no deposited abstract and no open-access location.]
Aronson, E., & Mills, J. (1959). 'The effect of severity of initiation on liking for a group.' The Journal of Abnormal and Social Psychology, 59(2), 177-181. doi:10.1037/h0047195. [Opening of the article read via OpenAlex reconstructed text; author order confirmed. The authors state the self-selection alternative themselves. NOT VERIFIED: sample size, as the retrievable text stops before the method.]
Chen, M. K., & Risen, J. L. (2010). 'How choice affects and reflects preferences: Revisiting the free-choice paradigm.' Journal of Personality and Social Psychology, 99(4), 573-594. doi:10.1037/a0020217. [Abstract read in full; author order confirmed at Crossref and OpenAlex. NOT VERIFIED: number of experiments and their sample sizes; closed at APA with no open-access version located.]
Izuma, K., & Murayama, K. (2013). 'Choice-induced preference change in the free-choice paradigm: A critical methodological review.' Frontiers in Psychology, 4, 41. doi:10.3389/fpsyg.2013.00041. PMC3566335. [Full text read at PubMed Central, open access. Corrected-design meta-analysis k = 4, d = 0.26, 95% CI 0.10 to 0.42, Q(3) = 0.618, p = 0.89, against the uncorrected pooled estimate of d = 0.61, 95% CI 0.55 to 0.66 as reported from Kenworthy et al. 2011. Two of the four corrected studies significant only one-tailed.]
Vaidis, D. C., Sleegers, W. W. A., van Leeuwen, F., DeMarree, K. G., Saetrevik, B., Ross, R. M., et al. (2024). 'A multilab replication of the induced-compliance paradigm of cognitive dissonance.' Advances in Methods and Practices in Psychological Science, 7(1). doi:10.1177/25152459231213375. [Abstract read in full; author order confirmed at Crossref, first of 104 authors. 39 labs, 19 countries, 14 languages, N = 4,898. Primary analyses found no significant attitude difference between high-choice and low-choice counterattitudinal essays.]
Egan, L. C., Santos, L. R., & Bloom, P. (2007). 'The origins of cognitive dissonance: Evidence from children and monkeys.' Psychological Science, 18(11), 978-983. doi:10.1111/j.1467-9280.2007.02012.x. [Abstract read in full; author order confirmed. NOT VERIFIED: sample sizes for either species. A free-choice paradigm study and therefore subject to the 2010 critique in full.]
Egan, L. C., Bloom, P., & Santos, L. R. (2010). 'Choice-induced preferences in the absence of choice: Evidence from a blind two choice paradigm with young children and capuchin monkeys.' Journal of Experimental Social Psychology, 46(1), 204-207. doi:10.1016/j.jesp.2009.08.014. [Abstract NOT retrievable: OpenAlex returns null, closed at Elsevier, no open-access version located. Published record and author order verified at OpenAlex. The design and the standing objection to it are known to us through the full text of Izuma and Murayama 2013 and are cited that way.]
Sharot, T., Velasquez, C. M., & Dolan, R. J. (2010). 'Do decisions shape preference? Evidence from blind choice.' Psychological Science, 21(9), 1231-1235. doi:10.1177/0956797610379235. [Abstract read in full; author order confirmed. NOT VERIFIED: sample size. Per Izuma and Murayama 2013, the effect is significant only on a one-tailed test.]
Izuma, K., Matsumoto, M., Murayama, K., Samejima, K., Sadato, N., & Matsumoto, K. (2010). 'Neural correlates of cognitive dissonance and choice-induced preference change.' Proceedings of the National Academy of Sciences, 107(51), 22014-22019. doi:10.1073/pnas.1011879108. [Abstract read in full; author order confirmed. Rate-rate-choose control with self-report and striatum activity as preference measures. NOT VERIFIED: sample size. Per Izuma and Murayama 2013, the effect is significant only on a one-tailed test.]
Alos-Ferrer, C., & Shi, F. (2015). 'Choice-induced preference change and the free-choice paradigm: A clarification.' Judgment and Decision Making, 10(1), 34-49. doi:10.1017/s1930297500003168. [Abstract read in full; author order confirmed. Argues the 2010 theorem is mathematically incorrect except in a particular case, while accepting that the paradigm is flawed and concluding that previous experimental results remain informative.]
Chernev, A., Bockenholt, U., & Goodman, J. (2015). 'Choice overload: A conceptual review and meta-analysis.' Journal of Consumer Psychology, 25(2), 333-358. doi:10.1016/j.jcps.2014.08.002. [Abstract read in full. 99 observations, N = 7,202. Higher decision task difficulty facilitates choice overload; satisfaction or confidence, regret, choice deferral and switching likelihood reported as interchangeable measures of it.]
Lodewijkx, H. F. M., & Syroit, J. E. M. M. (1997). 'Severity of initiation revisited: Does severity of initiation increase attractiveness in real groups?' European Journal of Social Psychology, 27(3), 275-300. doi:10.1002/(sici)1099-0992(199705)27:3<275::aid-ejsp822>3.0.co;2-s. [Abstract read in full; author order confirmed. Two longitudinal field studies, N = 202 and N = 46. Severity correlated with frustration, loneliness and depressed mood and lowered liking for the group; only the competing affiliation-attraction account was supported.]
Marsh, L. E., Kanngiesser, P., & Hood, B. (2017). 'When and how does labour lead to love? The ontogeny and mechanisms of the IKEA effect.' Cognition, 170, 245-253. doi:10.1016/j.cognition.2017.10.012. [Abstract read in full; open-access PDF at ScienceDirect. The effect emerges at age 5 and is not moderated by amount of effort invested in 5-to-6-year-olds; the authors report the results incompatible with existing theories of the effect.]
Norton, M. I., Mochon, D., & Ariely, D. (2012). 'The IKEA effect: When labor leads to love.' Journal of Consumer Psychology, 22(3), 453-460. doi:10.1016/j.jcps.2011.08.002. [Abstract read in full; author order confirmed at Crossref. Labour led to love only where the task was completed; the effect dissipated when creations were destroyed or left unfinished. NOT VERIFIED: sample sizes.]
Kahneman, D. (2011/2013). Thinking, Fast and Slow. Farrar, Straus and Giroux, New York. ISBN 978-0-374-53355-7. [Full vault read-through, pp. 3-482 including both appendices and the endnotes. Ch. 32, Keeping Score, pp. 342-352. The book contains no chapter and no sustained passage on cognitive dissonance: the complete thirteen-file page-level reading record was searched for 'dissonance', 'post-purchase', 'justification' and 'rationalisation', and the word dissonance appears nowhere in the body-text record.]
Kenworthy, J. B., et al. (2011), the meta-analysis giving the uncorrected pooled estimate. NOT VERIFIED and not cited as a reference: the paper is not retrievable after two OpenAlex query formulations and one WebSearch naming the presumed authors and journal. The figure is used only as reported inside the full text of Izuma and Murayama 2013, whose acknowledgements thank Kenworthy for sharing the meta-analysis data.