In 2008, a team of Dutch researchers published a study in the journal Science that seemed to demonstrate a remarkable form of unconscious influence. Participants who completed a word-scramble task containing words such as “pure,” “clean,” and “washed” later expressed a stronger preference for a specific brand of laundry detergent than participants who unscrambled neutral words. The finding was neat, intuitive, and quickly became a textbook example of social priming—the idea that subtle cues in the environment can shape our behaviour without our awareness.

The 2008 Paper That Changed Laundry Detergent Choices

The study, led by Ap Dijksterhuis of Radboud University, was part of a wave of research on unconscious priming that had gained momentum since the late 1990s. The experimental design was straightforward: participants were asked to unscramble sets of words. For one group, the words were related to cleanliness; for the other group, they were neutral. Afterwards, participants were shown two hypothetical detergent brands and asked which they would prefer. The primed group showed a clear shift toward the detergent whose packaging and name evoked purity.

The effect size appeared large, and the authors argued that the priming had activated a “cleanliness” concept that then influenced a seemingly unrelated consumer choice. The paper was cited rapidly—within a decade it had accumulated over 500 citations in psychology, marketing, and behavioural economics journals. It became a go-to example for popular science writers and TED talkers who wanted to illustrate the power of the unconscious mind.

Part of the appeal was the story it told: we are not rational decision-makers; our preferences can be nudged by something as trivial as a few words on a page. The study aligned with a broader cultural fascination with unconscious processes, from Malcolm Gladwell’s “Blink” to the rise of neuromarketing. For researchers, it offered a clean, replicable paradigm for studying how abstract concepts influence concrete choices.

To appreciate the study’s impact, consider how it was used in marketing textbooks and consumer behaviour courses. It was presented as evidence that subtle environmental cues—even those unrelated to the product—could shift brand preferences. A few marketing firms even experimented with in-store priming, such as playing scent-based cues or displaying cleanliness-related imagery near detergent aisles. The study seemed to offer a low-cost, high-impact tool for influencing consumer behaviour, though these applied efforts rarely published their results, making it difficult to assess their success.

Why the Original Result Looked So Convincing

The study’s design was clever in its simplicity. The word-scramble task was a well-established priming method that had been used in dozens of studies. The cleanliness words were chosen to be subtle—no direct mention of detergent or laundry. The dependent measure—a hypothetical choice between two detergent brands—was similarly indirect. Participants never suspected the link between the task and the choice.

The effect also matched a plausible psychological mechanism: activating the concept of cleanliness might make a product associated with purity more attractive. This narrative coherence made the finding feel intuitively correct. In the late 2000s, social priming was a booming field, with studies reporting that priming participants with words related to old age made them walk more slowly, or that priming with money made them more selfish. The detergent study fit right in.

Moreover, the original sample was small—roughly 40 participants per condition—which is typical for laboratory studies of that era. With such samples, even modest effects can appear statistically significant if the true effect is moderate or if random noise happens to align in the predicted direction. The study’s p-value was below the conventional threshold of 0.05, and the authors reported that the effect was robust to various analytical choices.

But there were early warning signs that went largely unnoticed. The effect size reported was unusually large for a social priming study—Cohen’s d around 0.8, which is considered a large effect. In a field where many priming effects are small to medium (d around 0.2–0.4), such a large effect should have prompted scepticism. Moreover, the confidence interval around that effect was wide due to the small sample, meaning the true effect could have been much smaller or even zero. At the time, however, the field lacked the tools and norms to flag such concerns. Journals rarely required power analyses or confidence intervals, and replication was not a priority.

The Replication Attempt That Unraveled the Finding

In 2014, the Many Labs 2 project—a large-scale replication effort involving dozens of labs across multiple countries—included the detergent priming study among its targets. The replication used the exact same materials and procedure as the original, but with a much larger sample: roughly 1,500 participants in total, giving it high statistical power to detect even a small effect.

The result was stark. The effect did not replicate. The difference between the primed and neutral groups was near zero, and the confidence intervals excluded the original effect size. The replication was published in Social Psychology as part of a special issue on replicability. For many researchers, it was a sobering moment: a well-known, highly cited finding had evaporated under scrutiny.

The failure to replicate did not necessarily mean the original finding was fraudulent. But it raised serious questions about the robustness of social priming effects. The Many Labs 2 team noted that the original study had a small sample, and that the effect might have been a false positive—a statistical fluke that happened to reach significance. Alternatively, the effect might be real but highly context-dependent, varying with subtle differences in procedure or participant population.

Notably, the replication team also conducted a mini meta-analysis combining the original and replication data. The overall effect was essentially zero, with a very narrow confidence interval. This analysis suggested that if a real effect existed, it was likely too small to be practically meaningful—perhaps equivalent to a shift in preference of less than 1% on a typical rating scale. For a finding that had inspired marketing applications and theoretical debates, this was a humbling conclusion.

Procedural Choices That Made the Difference

The original study was conducted using paper questionnaires in a laboratory setting. The replication was administered online, via computer. Could that shift explain the failure? Some researchers argued that the online environment might dilute the priming effect because participants are less engaged or more distracted. Others countered that if an effect is so fragile that it cannot survive a change in medium, it is not a robust phenomenon worth building theories on.

Another difference was sample size. The original had roughly 40 participants per cell; the replication had hundreds. Small samples are more vulnerable to both false positives and inflated effect sizes—a phenomenon known as the “winner’s curse.” When a small study happens to produce a significant result, the estimated effect size is often much larger than the true effect. The replication, with its larger sample, provided a more precise estimate, and that estimate was essentially zero.

Contextual cues also differed. The original study was run in the Netherlands, with Dutch participants. The replication involved participants from multiple countries. Cultural associations with cleanliness or detergent brands might vary. But the replication used the same brand names and wording, so cultural variation seems an unlikely full explanation. More plausibly, the original finding was simply a false positive.

There is also the issue of demand characteristics. In the original study, participants completed the word-scramble and then immediately made a choice between two detergents. The connection, though indirect, might have been more obvious in a face-to-face lab setting, where participants feel observed. Online, participants might have been more distracted or less motivated, reducing any subtle priming effect. However, if the effect is so sensitive to such factors, it raises the question of whether it is a genuine psychological phenomenon or an artefact of a particular experimental setup.

The Broader Replication Crisis in Social Priming

The detergent study was not an isolated case. Around the same time, several other high-profile social priming findings failed to replicate. John Bargh’s classic 1996 study on elderly priming—where participants walked more slowly after being exposed to words related to old age—proved difficult to reproduce. The power-posing effect, which claimed that adopting expansive postures increases feelings of power and risk-taking, also came under fire after replication attempts yielded mixed results.

These failures triggered what became known as the replication crisis in psychology. Methodologists pointed to questionable research practices—such as selectively reporting only significant results, stopping data collection once significance was reached, or running multiple analyses and reporting only the most favourable—as likely contributors. The crisis led to widespread calls for reform: pre-registration of study designs, larger sample sizes, and direct replications before accepting findings as established.

In social priming specifically, several meta-analyses suggested that the average effect size across studies was much smaller than initially reported, and that publication bias was severe. Studies with null results were rarely published, creating a distorted literature. The detergent study became a textbook example of how a single, small-sample study could mislead the field for years.

One striking example of the problem came from a meta-analysis of the elderly-walking priming effect. Researchers collected all known replications and found that the overall effect was small and heavily influenced by publication bias. When they corrected for bias, the effect disappeared. Similarly, a multi-lab replication of the money-priming effect—where participants exposed to money-related words behave more selfishly—found no consistent effect across labs. These cases reinforced the lesson that small, underpowered studies can produce dramatic but unreliable results.

What This Case Teaches About Experimental Design

The rise and fall of the detergent priming study offers concrete lessons for experimental design. First, small samples inflate the false-positive rate. Even if a study is perfectly conducted, a sample of 40 per condition has limited power to detect a small effect, and any significant result is likely to be an overestimate. Larger samples provide more stable estimates.

Second, context dependency is real but can be overused as a post-hoc explanation. If an effect cannot be reproduced under slightly different conditions, it may not be worth building theories on. Researchers should specify the boundary conditions of their effects a priori, rather than invoking context after a failed replication.

Third, pre-registration—publicly documenting the study design, sample size, and analysis plan before data collection—can prevent many questionable practices. If the original study had been pre-registered, the authors would have committed to a specific analysis, reducing the temptation to try different approaches until one worked. The replication crisis has spurred many journals to adopt registered reports, where peer review occurs before results are known.

Fourth, the importance of direct replications cannot be overstated. The detergent study was cited hundreds of times before anyone attempted a large-scale replication. Had the field invested earlier in replication efforts, the fragility of the finding would have been exposed sooner. Some funding agencies now require replication plans for large grants, and journals such as Psychological Science have introduced badges for pre-registration and data sharing.

Fifth, researchers should be wary of narrative plausibility. A finding that tells a compelling story—like cleanliness words making you prefer a pure detergent—can be persuasive, but narrative coherence is not evidence. The history of psychology is filled with intuitively appealing findings that later failed to replicate, from ego depletion to the facial feedback hypothesis. The best defence is methodological rigour, not storytelling.

Lessons for Behavioral Science and Beyond

The detergent priming case has influenced how funding bodies and journals approach evidence. Some funding agencies now require replication plans for large grants. Journals such as Psychological Science have introduced badges for pre-registration and data sharing. The number of registered reports has grown steadily since the mid-2010s.

For practitioners—marketers, policymakers, educators—the lesson is to demand robust evidence before acting on a priming effect. A single study, especially one with a small sample, should not drive a campaign or a policy. Meta-analyses and multi-lab replications provide a more reliable basis for decisions. For example, if a marketing team is considering using priming cues in an advertisement, they should look for converging evidence from multiple well-powered studies, ideally conducted in realistic settings.

None of this means that priming effects do not exist. Some priming phenomena, such as semantic priming in cognitive psychology, are well-established. But the kind of complex, behaviourally consequential social priming that the detergent study claimed remains controversial. The effect, if it exists, appears to be small, fragile, and highly dependent on subtle features of the experimental context.

The story of the detergent priming study is not a story of fraud or incompetence. It is a story of how the scientific process, when it includes systematic replication, can correct itself. The original authors, to their credit, participated in the replication effort and accepted the outcome. The field moved forward, adopting more rigorous standards. And researchers now have a clearer sense of what it takes to produce reliable knowledge about the unconscious mind.

Ultimately, the case underscores a fundamental tension in science: the desire for exciting, novel findings versus the need for cautious, cumulative evidence. The replication crisis has not destroyed social priming; it has refined it. Researchers now ask harder questions, use larger samples, and share their data more openly. The detergent study, once a star in the priming firmament, now serves as a reminder that even the most elegant experiment can be a mirage. The path to reliable knowledge is slow, iterative, and often humbling—but it is the only path worth taking.