Back to Blog

Behavioural science had a reckoning. Starting around 2011, psychologists began re-running celebrated experiments under stricter conditions: larger samples, pre-registered predictions, and multiple independent labs running the identical protocol. A striking number of textbook findings shrank to nothing. This is the replication crisis, and it matters to marketers because the conference-stage version of behavioural science rarely mentions it. The bias sounds clever, it gets repeated, and nobody checks whether it survived.

Every study below is named, with its replication result and a link. We have split them into what broke and what held, because the useful skill is not knowing more biases. It is knowing which ones you can actually bet a budget on.

What quietly died

Willpower fatigue (ego depletion)

The idea: self-control is a finite battery, so a person who resists one temptation has less willpower left for the next. Marketers turned this into "reduce decision fatigue," the justification for stripping steps out of a checkout so the buyer does not run out of mental gas before paying.

The original 1998 experiments were influential for two decades. Then in 2016 a registered replication ran the test across 23 laboratories with 2,141 participants under one agreed protocol. The effect came out at essentially zero. The original author's rebuttal, titled "Misguided effort," did not produce a procedure that reliably reproduces it.

What to keep: reducing checkout friction still works, because every extra step is a chance to abandon. Just do not sell it internally as "preserving willpower." The friction argument stands on its own. The battery metaphor does not.

Source: Hagger, M. S., et al. (2016). A multilab preregistered replication of the ego-depletion effect. Perspectives on Psychological Science, 11(4), 546 to 573.

Subconscious priming

The idea: exposing people to a subtle cue changes their behaviour without their awareness. The famous demonstration had participants exposed to words about old age, who then supposedly walked more slowly out of the lab. It became the poster child for "prime the buyer and they will act."

In 2012, a team re-ran it with automated infrared timing instead of a person holding a stopwatch. The effect vanished. In a second experiment they showed that when the experimenters expected slower walking, they recorded slower walking. The original result was most likely the measurer's expectation, not the participant's behaviour.

What to keep: context and framing genuinely affect choices, that part is well supported elsewhere. But the strong claim, that a hidden cue reliably steers behaviour people cannot detect, did not survive a clean measurement. Be sceptical of any tactic that depends on the buyer not noticing.

Source: Doyen, S., Klein, O., Pichon, C. L., and Cleeremans, A. (2012). Behavioral priming: It's all in the mind, but whose mind? PLoS ONE, 7(1), e29081.

Power posing

The idea, from a 2010 study and one of the most-viewed talks of all time: holding an expansive posture for two minutes raises testosterone, lowers cortisol, and makes you act bolder. It got repackaged endlessly as pre-pitch confidence advice.

A 2015 replication with more than double the original sample found that people did report feeling more powerful, but there was no change in hormones and no change in actual risk-taking. In 2016 the original lead author stated plainly, "I do not believe that power pose effects are real." The feeling was real. The physiology and the behaviour were not.

What to keep: this one is a clean lesson in reading a study. A real subjective effect (feeling more confident) got oversold as a physiological and behavioural one that the data never supported. When a claim jumps from "people said they felt X" to "X changes your body and your decisions," check whether the second part was actually measured.

Source: Ranehill, E., et al. (2015). Assessing the robustness of power posing. Psychological Science, 26(5), 653 to 656.

What held up

Anchoring

The idea: the first number people see shapes every judgment that follows, even when the number is arbitrary. Show a high price first and the next price feels reasonable by comparison.

Unlike the effects above, anchoring was put through the same large multi-lab replication process and passed cleanly. In the Many Labs project it reproduced consistently across samples and settings, with one of the larger effect sizes in the whole set. It is among the most reliable findings in the field.

What to do with it: this is the evidence behind a struck-through original price, a premium tier placed next to the plan you actually want sold, and leading with your highest package. The first number frames the rest. Use it deliberately.

Source: Klein, R. A., et al. (2014). Investigating variation in replicability: A "many labs" replication project. Social Psychology, 45(3), 142 to 152.

Defaults

The idea: whatever option is pre-selected is the option most people end up with, because changing it takes effort and implies a decision. The cleanest evidence comes from organ donation, where the only difference between neighbouring countries is whether you are a donor by default.

Same culture, same border, near-opposite outcomes, driven entirely by which box was pre-ticked. Across countries, opt-out systems averaged far higher registered consent than opt-in ones. This finding has held up because it has been tested repeatedly in the real world, not just in a lab.

What to do with it, honestly: defaults are powerful, which is exactly why they need to be used with care. Pre-selecting the plan that genuinely fits most customers is good design. Pre-ticking a box that adds cost the buyer did not ask for is a dark pattern, and increasingly an illegal one. The power is real. So is the responsibility.

Source: Johnson, E. J., and Goldstein, D. (2003). Do defaults save lives? Science, 302(5649), 1338 to 1339.

Social proof

The idea: people copy what similar people are doing. In two hotel field experiments, a sign saying "the majority of guests reuse their towels" got more guests to reuse than the standard "please help the environment" appeal. The strongest version was the most specific one: "the majority of guests in this room reused their towels."

The lesson that survives is not "add a testimonial." It is that specific, relevant, true social proof beats generic virtue. "Join thousands of happy customers" is weak. "127 Calgary contractors switched to this last quarter" is the same principle done right, because it is concrete and it matches the reader's situation.

One honest caveat: the exact towel study has been re-run with mixed results, so treat the specific numbers with care. The broader principle, that concrete descriptive norms move behaviour, is supported by a large body of field work. It is sturdier than most of what is on this page, but it is not magic, and vague social proof does little.

Source: Goldstein, N. J., Cialdini, R. B., and Griskevicius, V. (2008). A room with a viewpoint: Using social norms to motivate environmental conservation in hotels. Journal of Consumer Research, 35(3), 472 to 482.

How to read behavioural science without getting burned

Three habits separate the marketers who use this well from the ones who quote a dead study at a strategy meeting.

First, ask whether the effect was replicated, not just published. A single striking result from one lab in one year is a hypothesis, not a tactic. The findings that held up here all went through large, pre-registered, multi-lab tests. The ones that died did not survive that same test.

Second, watch for the jump from feeling to behaviour. Power posing measured a real feeling and got sold as a change in hormones and decisions. A lot of behavioural claims quietly slide from "people said they preferred X" to "X makes people buy," and the second half is often where the evidence runs out.

Third, test it on your own numbers before you commit. Even a robust effect like anchoring shows up at different sizes in different contexts. Behavioural science tells you where to look and which direction to expect. Your own A/B test tells you whether it is true for your buyers, your price, and your category. Use the research to form the hypothesis. Use your data to settle it.

Frequently Asked Questions

Is ego depletion or "decision fatigue" real?

The strong version did not hold up. A 2016 registered replication ran the ego-depletion test across 23 laboratories with 2,141 participants under one pre-agreed protocol and found essentially no effect (Hagger et al., Perspectives on Psychological Science). Reducing checkout steps still helps conversion because each step is a chance to abandon, but the "willpower runs out like a battery" explanation is not supported by the replication evidence.

Does subconscious priming actually change buyer behaviour?

The famous demonstration did not replicate. When Doyen and colleagues (2012, PLoS ONE) re-ran the elderly-priming study with automated timing instead of a human stopwatch, the effect disappeared, and they showed the original result tracked the experimenters' expectations. Context and framing affect choices, but the strong claim that a hidden cue reliably steers behaviour people cannot detect is not well supported.

Does power posing work?

Partly, and not the part that was famous. A 2015 replication with double the original sample (Ranehill et al., Psychological Science) found people did feel more powerful after an expansive pose, but there was no change in testosterone, cortisol, or risk-taking. In 2016 the original lead author stated she no longer believes the power-pose effects are real. The subjective feeling replicated. The hormonal and behavioural claims did not.

Which behavioural effects are safe to build marketing on?

The ones that passed large, pre-registered, multi-lab or repeated real-world tests. Anchoring reproduced strongly in the Many Labs project (Klein et al., 2014). Defaults are supported by cross-country organ-donation data (Johnson and Goldstein, 2003, Science). Concrete descriptive social norms are supported by a broad body of field experiments (Goldstein, Cialdini, and Griskevicius, 2008, Journal of Consumer Research). These are the safer bets, though you should still validate them on your own numbers.

How do defaults change behaviour so much?

Because changing a pre-set option takes effort and feels like an active decision, so most people stay with whatever is already selected. In organ donation, opt-out countries such as Austria show registered consent near 99% while opt-in countries such as Germany sit around 12%, despite similar cultures (Johnson and Goldstein, 2003). Used honestly, that means pre-selecting the plan most customers actually want. Used to sneak in unwanted charges, it is a dark pattern and often illegal.

Related Reading

• Five marketing myths peer-reviewed research says are wrong, the companion piece on beliefs the evidence contradicts.
• Performance vs brand marketing, why the split is a false choice and how to divide budget by stage.