What the Replication Crisis Means for Marketing Psychology

Many of the behavioral science findings that marketers rely on have failed to replicate. It is time to update the playbook.

Anika van der Berg · February 4, 2025

In 2015, the Open Science Collaboration published the results of an ambitious project: they had attempted to replicate 100 psychology studies published in three leading journals. Only 36% of the replications produced statistically significant results in the same direction as the originals. The average effect size in the replications was roughly half the size of the original studies.¹

This was not an isolated finding. In the years since, large-scale replication projects have confirmed the pattern across subfields of psychology. The Many Labs projects, the Social Sciences Replication Project, and numerous individual replication attempts have collectively established that a substantial proportion of published psychology findings—perhaps a majority—are either false positives or dramatically overstated.

For academic psychologists, this has prompted painful but ultimately productive methodological reforms: pre-registration, larger sample sizes, open data, and a shift toward valuing replication. For marketers, however, the implications have been largely ignored. Marketing textbooks, conference talks, and strategy decks continue to cite findings that have been seriously challenged or outright debunked. The playbook has not been updated.

This essay examines several of the most popular marketing psychology concepts that have been affected by the replication crisis, and what marketers should believe now.

Power Poses: The Confidence That Wasn't

In 2010, Carney, Cuddy, and Yap published a study (n=42) claiming that adopting expansive "power poses" for two minutes increased testosterone, decreased cortisol, and increased willingness to take risks. The finding was enormously popular. Amy Cuddy's subsequent TED talk has been viewed over 70 million times. Marketing consultants incorporated power posing into presentation coaching. Brand strategists discussed "powerful" body language in advertising imagery.

The original study did not replicate. Ranehill et al. (2015, n=200) found no significant effects of power posing on hormones or risk-taking behavior. In 2016, Dana Carney, the first author of the original paper, published a public statement disavowing the finding: "I do not believe that 'ichpower pose' effects are real." A comprehensive meta-analysis by Gronau et al. (2017) confirmed that the hormonal effects were not supported by the evidence.

A caveat is warranted here. There is some evidence that adopting expansive postures may produce small, transient effects on self-reported feelings of confidence. Cuddy and colleagues have published follow-up work supporting this more modest claim. But the dramatic narrative—that two minutes of standing in a particular way produces measurable physiological changes that affect behavior—has not survived empirical scrutiny.

For marketers, the lesson is not that body language is irrelevant to advertising or presentation. It is that the specific, quantified claims about power posing that were being cited ("increases testosterone by 20%") were based on a single, small, unreplicated study. The confidence with which these numbers were deployed in marketing contexts was never warranted.

Ego Depletion: The Willpower Tank That May Not Exist

The ego depletion model, pioneered by Baumeister, Bratslavsky, Muraven, and Tice (1998, n=67), proposed that self-control draws on a limited resource that is depleted with use—like a muscle that fatigues. The concept was rapidly adopted by marketers: it suggested that consumers who had already made many decisions would have depleted willpower and would be more susceptible to impulse purchases, upsells, and low-effort default options.²

Conversion optimization strategies were built on this idea. Present the upsell after the consumer has already navigated a complex configuration process (their willpower is depleted, so they will accept the default). Schedule email promotions for late evening (decision fatigue makes consumers more impulsive). Reduce decision points early in the funnel to "preserve" willpower for the purchase decision.

The empirical foundation has weakened substantially. A pre-registered, multi-lab replication attempt (Hagger et al., 2016) involving 23 laboratories and 2,141 participants found no significant ego depletion effect (d = 0.04, 95% CI [-0.07, 0.15]).³ The overall effect was essentially zero, with the confidence interval comfortably including null.

Subsequent meta-analyses have yielded mixed results, with some finding small effects and others finding none, and strong evidence of publication bias in the original literature. The current consensus among researchers is that if ego depletion exists at all, it is substantially smaller than originally reported and may depend on beliefs about willpower rather than on an actual resource depletion mechanism (Job, Dweck, & Walton, 2010).

This does not mean that decision fatigue is entirely fictitious. There is decent evidence that people simplify their decision strategies as they make more decisions in sequence—but this may reflect changing motivation rather than resource depletion. The distinction matters for marketing strategy. If decision fatigue is about motivation rather than depletion, then the solution is not to "deplete" consumers into compliance but to maintain their engagement and motivation throughout the decision process.

Social Priming: The Subtle Influence That Was Too Good to Be True

Social priming—the idea that subtle environmental cues unconsciously influence complex social behavior—was one of the most exciting areas of psychology in the 2000s. The seminal study by Bargh, Chen, and Burrows (1996, n=34) reported that participants who unscrambled sentences containing words associated with the elderly subsequently walked more slowly down a hallway. The finding spawned hundreds of follow-up studies and a cottage industry of marketing applications.

Marketers embraced priming enthusiastically. Use certain colors on your website to prime purchasing behavior. Play French music in the wine aisle to increase French wine sales. Prime consumers with luxury cues to increase willingness to pay. The book Nudge by Thaler and Sunstein (2008) popularized many of these ideas, though the authors were more careful than their popularizers.

The replication record for social priming has been poor. Doyen, Klein, Pichon, and Cleeremans (2012, n=120) failed to replicate Bargh's walking speed finding when using automated rather than experimenter-timed measurement, and found evidence that the original effect may have been driven by experimenter expectancy effects. A large-scale replication of another classic priming study—the "professor priming" effect on trivia performance (Dijksterhuis & van Knippenberg, 1998)—failed across 40 laboratories with a total sample of 4,493 participants (O'Donnell et al., 2018).

Important caveats: not all priming effects have failed to replicate. Basic cognitive priming (e.g., faster recognition of the word "nurse" after seeing "doctor") is robust and well-established. And some environmental effects on consumer behavior have adequate support—the relationship between music tempo and dining speed, for example, has been replicated with reasonable consistency. But the broader claim that subtle, unconscious environmental cues can reliably steer complex purchasing behavior remains inadequately supported.

What Still Holds Up

It would be a mistake to conclude from the replication crisis that all of marketing psychology is suspect. Many foundational findings have replicated well and remain reliable guides for practice.

Loss aversion has been challenged by some researchers (Gal & Rucker, 2018), but the preponderance of evidence still supports the basic finding that losses are weighted more heavily than equivalent gains in most decision contexts. The effect may be smaller and more context-dependent than originally stated, but it is real.

Anchoring effects—the tendency for initial numbers to influence subsequent numerical judgments—replicate reliably and with large effect sizes across many contexts (Furnham & Boo, 2011). This remains one of the most useful and robust findings in behavioral science.

Choice overload has a complicated replication history. Iyengar and Lepper's famous jam study (2000, n=249) has been difficult to replicate exactly. However, meta-analyses by Chernev, Bockenholt, and Goodman (2015) suggest that choice overload is a real phenomenon, but it is moderated by several factors: it is more likely when options are difficult to compare, when the decision maker lacks clear preferences, and when the choice set lacks a dominant option. The blanket claim that "more options always reduce conversion" is false. The nuanced claim that "too many similar options can paralyze certain decision makers" is supported.

Default effects—the tendency for people to stick with pre-selected options—replicate consistently and powerfully. Johnson and Goldstein's (2003) classic comparison of organ donation rates between opt-in and opt-out countries, and numerous subsequent studies of default options in retirement savings, insurance selection, and software settings, confirm that defaults exert large and reliable effects on behavior.

Social proof in its basic form (people look to others' behavior to guide their own, especially under uncertainty) is well-established and replicates. Cialdini's work on towel reuse in hotels, and subsequent studies of social norms messaging in energy conservation, tax compliance, and health behavior, hold up well.

How to Update the Marketing Playbook

The practical question for marketers is how to respond to this state of affairs. Complete skepticism toward all behavioral science is unwarranted—many findings are robust and practically useful. Complete credulity toward the pre-crisis literature is also unwarranted—too many popular findings have not survived scrutiny.

A reasonable middle ground involves three principles. First, check the replication record before building strategy on a finding. Resources like the Curate Science database, the Replication Index, and the "replications" section of Google Scholar make this feasible, though not always easy. Second, weight replications with large samples more heavily than originals with small samples. A study with n=42 that is contradicted by a replication with n=2,000 should be treated as disconfirmed, regardless of the prestige of the original authors. Third, be wary of effects that seem too clean, too large, or too universal. The replication crisis has revealed that the largest and most dramatic effects in the original literature were often the most inflated.