The story is the stuff of advertising legend. In 1957 at a movie theater in New Jersey, messages too brief to be consciously detected were flashed on the screen during a movie: “DRINK COCA-COLA” and “EAT POPCORN.” Without knowing why they were doing it, people streamed to the concession stand and plunked down good money for overpriced movie snacks. The marketing researcher James M. Vicary reported a 37 percent increase in popcorn sales and an 18 percent increase in Coca-Cola sales (Karremans et al. 2006; Strahan et al. 2002). In the same year, Vance Packard published the bestselling Hidden Persuaders (1957) aimed at describing the subtle ways that marketers were influencing consumers’ choices, and in 1974, Wilson Bryan Key published Subliminal Seduction, filled with illustrations that he claimed demonstrated how advertisers were enticing consumers with sexual messages.
Despite the potential advantages for product marketing, many people found this all a bit disturbing, and the cultural climate in the United States didn’t always help matters. A year before Hidden Persuaders appeared, Invasion of the Body Snatchers was a box office hit, and in 1962, at the height of Cold War tensions, The Manchurian Candidate was released. Although it happens every day, people don’t like the prospect of being controlled by unseen forces. This worry seems to be one of the themes of current fears about artificial intelligence, as well as a possible reason many of my friends are afraid to have a smart speaker in their homes. “I don’t want Amazon listening to my private conversations all day long.” In response to worries about unfair practices, the United Kingdom and several other countries—but not the United States—have passed laws prohibiting subliminal advertising.
In a famous 1996 legal case, James Vance sued the British rock band Judas Priest, claiming that the band’s song “Better by You, Better Than Me” contained the subliminal phrase “Do It,” which caused him and a friend, Ray Belknap, to attempt suicide. Belknap died, but Vance survived with serious injuries. Judas Priest said they were unaware of any subliminal messages in their songs, but they failed to get the case dismissed on free speech grounds because the judge ruled that subliminal messages were an invasion of privacy that was not protected by the First Amendment. However, the judge did grant a summary judgment for the defendants because the plaintiffs could not prove that any subliminal messages were present in the songs.
Fortunately, many of the worst fears about subliminal messaging are unfounded. Vicary eventually admitted that his movie popcorn study was a complete hoax. It never happened. Furthermore, as we will see, in general, subliminal and other priming effects prove to be nonexistent or much more modest than previously believed (Karremans et al. 2006). For example, a study of subliminal training self-help audiotapes marketed to help people quit smoking, lose weight, or achieve higher self-esteem revealed that any benefit was caused by a placebo effect. Just giving people a tape and telling them it had subliminal messages on it was good enough, whether there were any subliminal messages or not (Greenwald et al. 1991).
What Are We Talking About?
For reasons I will speculate on later in this article, research into subliminal effects and the larger category of priming effects have been fraught with controversy from the beginning. The problem came to a head in 2012 when Nobel Prize–winning psychologist Daniel Kahneman wrote an open letter to priming researchers in which he claimed “your field is now the poster child for doubts about the integrity of psychological research” and suggested that a “train wreck is looming” for young researchers who are associated with this topic of study (Yong 2012). But before we get to the train wreck, let’s be clear about what priming effects are.
Figure 1. The relationship between the type of stimulus and the kind of effect it produces. Priming stimuli can be either subliminal (below the level of perception) or supraliminal (easily seen or heard), but their effects on behavior are unconscious. (Figure based on an article by Nick Kolenda.)
A typical priming effect is a fairly immediate response to something in the environment; however, unlike merely following directions or getting an idea to do something, the response to a priming stimulus is passive—not intended—and the person making it is unaware of its influence (Bargh 2016). The priming stimulus need not be subliminal. It can be quite obvious, but the individual is unaware of its effect. For example, in a classic but controversial experiment, John Bargh and colleagues gave college students scrambled sentences to unscramble (Bargh et al. 1996, Experiments 2a and 2b). The students were given groups of five words (e.g., “he it hides finds instantly”) and were asked to rearrange the words into grammatical sentences. For some participants, the scrambled-sentence task included words that suggested an elderly stereotype (e.g., “worried, Florida, old, grey,” etc.), and for other participants the scrambled words were neutral. Later, all the participants were surreptitiously timed as they walked down a hall leaving the laboratory, and Bargh and colleagues reported that the students exposed to words suggesting an elderly stereotype walked significantly slower leaving the laboratory. The participants had seen and understood all the words as they unscrambled them, but subsequent interviews determined that only one participant had any idea that the elderly words might be connected with their walking speed. For all the other subjects, the effect of the priming words had been completely unconscious.
In another study, a liquor store featuring displays of both German and French wines played French music and German music on the store sound system on alternate days. The researchers found that wine sales fluctuated with the music being played (e.g., more French wine sold on days when French music was being played) and that, in a post-purchase questionnaire, customers were unaware of the influence of the music on their behavior (North et al. 1999).
When they occur, priming effects are presumed to be caused by unconscious associations. The knowledge and memories that we carry around in our heads are not randomly filed in there. Things that we have encountered together in the past are held together in memory: old people and Florida, hotdogs and mustard, French music and French wine. Stimulating the memory of a particular sound, object, or idea has a spreading effect on a network of related associations, and these associations unconsciously affect our behavior.
A Slow-Motion Train Wreck
In 2005, Stanford University professor John Ioannidis published a paper with the provocative title, “Why Most Published Research Findings Are False.” He identified a bias among journals in favor of publishing successful experiments and evidence of widespread statistical manipulation to produce the desired outcome of a successful experiment. In medicine, psychology, and other related fields, there was a growing sense that many studies that had appeared in textbooks for decades might not be reliable. The controversy later became known as the “Reproducibility Crisis,” and the problem in psychology was confirmed when a large-scale attempt to reproduce the results of 100 studies previously published in three prominent psychology journals found that, although 97 percent of the original studies had reported statistically significant results, only 68 percent of the replicated studies were significant (Open Science Collaboration 2015). Not a great track record.
Part of the problem was publication bias. Journals in many fields have a strong bias in favor of publishing positive results—studies in which the hypotheses being tested are shown to be valid. Daniele Fanelli of the University of Edinburgh conducted a survey of 2,434 published articles in twenty different science disciplines ranging from the traditionally “hard” sciences (e.g., space science) to the “soft” sciences (e.g., psychology/psychiatry). As Figure 2 shows, she found that the highest level of positive findings were in articles published in psychology/psychiatry (Fanelli 2010). Based on her sample, studies in psychology, psychiatry, economics, and business were approximately five times as likely to report significant results than space science. This kind of publication bias creates great pressure among researchers to come up with statistically significant findings, and in an earlier column I confessed to my own past as a “p-hacker,” manipulating the results of psychology studies to eke out a statistically significant result, traditionally indicated by a probability, p, of less than five percent. At the time, long before the Reproducibility Crisis, we were taught p-hacking was standard operating procedure.
Figure 2. A graph from Fanelli (2010) showing the percentage of published papers from twenty different fields of science that reported support for the tested hypothesis.
It was in this atmosphere of reexamination and skepticism about research in psychology that things started to come apart for priming research. In 2012, Stéphane Doyen of the Université Libre de Bruxelles and colleagues attempted to recreate the results of Bargh’s elderly stereotype priming study and were unable to reproduce the original results (Doyen et al. 2012). Doyen and colleagues used twice the number of participants as Bargh and introduced a number of controls. For example, Bargh and his collaborators timed the students walking down the hall with a stop watch, introducing the possibility of experimenter error. In contrast, Doyen and colleagues automatically timed their participants’ speed of walking using infrared sensors at each end of the hallway. In a second experiment, Doyen and colleagues were able to duplicate the slow-walking effect but only after manipulating the experimenter’s expectations about the outcome. Although in Bargh’s original experiment the people who timed the walking were unaware of whether or not the participants had received the elderly stereotype primes, it was unclear what the experimenters administering the scrambled sentences knew about the expected results. As a result, there was a possibility of unconscious experimenter bias affecting the behavior of the student participants.
The Doyen et al. (2012) article provoked a blistering response from John Bargh on his Psychology Today blog. (The post has since been removed but is preserved on the Open Science Framework website). But this was not the only blow to priming research. A few months before the Doyen et al. (2012) study appeared, a committee at Tilburg University in the Netherlands, released an interim report on psychologist Diederik Stapel, a rising star in the priming field. Tipped off by several young researchers who had worked in his laboratory or had knowledge of his research, the committee found evidence of a shocking level of research fraud. Stapel ultimately admitted to fabricating data on a total of fifty-eight different publications. The website Retraction Watch, which monitors cases of research papers being retracted by their authors or publishers due to fraud or error, reported on Stapel’s most recent retraction in 2015. At that time, Stapel was listed fourth on Retraction Watch’s “leader board” for researchers with the most known retracted papers. He has since dropped to seventh.
Meanwhile the broader reexamination of published psychology studies chugged on, and additional priming studies were among the experiments that fell from glory. For example, in a very large-scale replication effort called Many Labs 2, rigorous replications were attempted for twenty-eight classic and contemporary published studies, and the authors found that only fifteen of the replications (54 percent) produced significant results similar to the original studies. Among the studies that could not be reproduced were three priming experiments. For example, one experiment used a scrambled sentence task similar to Bargh’s 1996 study. People in one group were primed with words suggesting “heat” and in another with words suggesting “cold” (Zaval et al. 2014). Subsequently, people in the heat condition expressed significantly greater belief in and concern about global warming than those in the cold priming condition. Unfortunately, a more rigorous replication of the study as part of the Many Labs 2 project came up empty: no significant differences between the heat- and cold-primed groups.
Daniel Kahneman was moved to write his 2012 open letter about the priming “train wreck” because he had been personally burned by it. Although he was not actually a priming researcher, in his 2011 bestselling book Thinking, Fast and Slow, he had included a section called “The Marvels of Priming,” in which, among other examples, he described Bargh’s elderly stereotype study at some length. Unfortunately, soon after the book appeared, things began to unravel, and it looked like priming research might follow the ignominious path of subliminal messages: lots of excitement followed by a burst balloon. A decade has passed since Kahneman wrote the open letter, but it appears he is still down on the field of priming. In a 2022 online lecture for Edge.org, he said, “And behavioral priming research is effectively dead. Although the researchers never conceded, everyone now knows that it’s not a wise move for a graduate student to bet their job prospects on a priming study.”
Part of what bothered Kahneman was the remarkably defensive response of priming researchers, particularly John Bargh, as exemplified by Bargh’s Psychology Today blog post. In Kahneman’s view, priming researchers had shown a disappointing reluctance to attempt to reproduce their previous results, in sharp contrast to Kahneman’s research philosophy. Like most academic psychologists, Kahneman has had his share of disagreements with other researchers, but unlike almost all other academic psychologists, Kahneman has sought to engage with his rivals in an effort to find to sensible resolution. The subject of his Edge.org talk was “adversarial collaboration,” a strategy in which researchers who are at odds with each other agree to collaborate on the design of experiments that they both acknowledge have the potential to resolve the issues at hand. Kahneman has used this approach on a couple of occasions, but he admits that it has not always worked. It is difficult to change people’s minds. But the impulse to try to solve the problem by cooperatively seeking out more information seems far superior to entrenched positions and academic warfare. In my view, Kahneman’s gentle manner and his “make love not war” approach to scientific conflicts are two of the many reasons he is held in such great esteem.
How This Happened
I think I know how this happened, at least in broad strokes. Part of the explanation is reflected in Kahneman’s phrase “The Marvels of Priming.” These studies are a bit mind-blowing and, as a result, guaranteed to grab attention. Indeed, some of what made control by subliminal messages so scary to the general public is precisely what makes priming studies so attractive for researchers. First, the control is unconscious. We tend to think of ourselves as being in control of our actions. The prospect of being controlled by someone else without our knowledge or consent is quite unsettling. On the other hand, psychology researchers are likely to be thrilled about the prospect of nudging us toward French or German wine without our awareness.
The second factor, in my view, is the outsized power of priming effects. Typically, a very subtle stimulus—a few words in a scrambled sentence task or background music in a wine store—produces a much larger effect in the unsuspecting participants. This kind of power is intoxicating and bound to get attention. In a recent article, Jeffrey Sherman and Andrew Rivers described the most controversial studies as “eye-catching and counter-intuitive,” as well as “the kind of sexy research that popular science writers love to describe” (Sherman and Rivers 2021, 6). The prospect of your research creating a publicity buzz would also be a powerful incentive in favor of positive results.
These factors don’t explain the defensiveness of priming researchers, except perhaps that psychologists might be reluctant to cast aside studies that bring them lots of attention. But for whatever reason, some primacy researchers did respond defensively, and this did not serve to quell the controversy.
Is Priming Dead?
So, is priming research really dead? In one sense it is, and in another it isn’t. In Kahneman’s sense, priming research is probably dead—at least for the time being. The field is still so fraught with controversy that few young researchers, not yet safely ensconced in tenured positions, are likely to risk conducting priming research only to have their work looked at with skepticism or—worse yet—shown to be unreliable someday in the future. For a while, at least, most priming research will be limited to brave and well-established researchers. Also, in no small part due to the influence of Kahneman’s letter and the widespread publicity surrounding the problems in the field, priming research in general has a bit of a dark cloud hanging over it.
In a second sense, the reports of the death of priming have been greatly exaggerated. Many studies have stood the test of replication, and the general consensus is that many priming effects are real and likely to influence us in unknown ways every day (Bargh 2016; Strack and Schwarz 2021). Even subliminal priming has risen from the ashes of Vicary’s “EAT POPCORN” and “DRINK COCA-COLA” disaster. Recent studies suggest that subliminal stimuli, below the level of conscious awareness, can influence behavior—but in a far more modest fashion than first imagined. For example, Johan Karremans and colleagues (2006) found that subliminal priming of a particular brand (e.g., Lipton’s Iced Tea) could influence the brand of drink chosen, but only in participants who were already thirsty. Thus, consumer research suggests that subliminal messages can’t make you do things you aren’t already motivated to do (Bargh 2016).
Furthermore, there is a growing understanding that many of the studies that fell to the reproducibility crisis were “under-powered” studies with relatively small sample sizes and weaker experimental designs (Sherman and Rivers 2021). The more stringent standards that have emerged in recent years in response to these issues should lead to more reliable results in the future. In particular, the relatively new movement called “open science” has promoted more collaborative and transparent methods in science, and many scientific journals have adopted open science methods as standards or incentives for acceptance of submitted articles. This movement may not solve all the research problems in the priming field, but it can only help. The clouds have not yet lifted, but perhaps before long priming research will be reestablished as a solid and dependable field.
References
Bargh, John A. 2016. Awareness of the prime versus awareness of its influence: Implications for the real-world scope of unconscious higher mental processes. Current Opinion in Psychology 12: 49–52.
Bargh, John A., Mark Chen, and Lara Burrows. 1996. Automaticity of social behavior: Direct effects of trait construct and stereotype activation on action. Journal of Personality and Social Psychology 71(2): 230–44.
Doyen, Stéphane, Olivier Klein, Cora Lise Pichon, et al. 2012. Behavioral priming: It’s all in the mind, but whose mind? PLoS ONE 7(1): 1–7.
Fanelli, Daniele. 2010. ‘Positive’ results increase down the hierarchy of the sciences. PLoS ONE 5(4): 1–10.
Greenwald, Anthony G., Eric T. Spangenberg, Anthony R. Pratkanis, et al. 1991. Double-blind tests of subliminal self-help audiotapes. Psychological Science 2(2): 119–122.
Kahneman, Daniel. 2011. Thinking, Fast and Slow. New York, NY: Farrar, Straus, Giroux.
Karremans, Johan C., Wolfgang Stroebe, and Jasper Claus. 2006. Beyond Vicary’s fantasies: The impact of subliminal priming and brand choice. Journal of Experimental Social Psychology 42(6): 792–98.
North, Adrian C., David J. Hargreaves, and Jennifer McKendrick. 1999. The influence of in-store music on wine selections. Journal of Applied Psychology 84(2): 271–276.
Open Science Collaboration, Brian A. Nosek, Alexander A. Aarts, et al. 2015. Estimating the reproducibility of psychological science. Science 349(6251): 1–69.
Sherman, Jeffrey W., and Andrew M. Rivers. 2021. There’s nothing social about social priming: Derailing the ‘Train Wreck.’ Psychological Inquiry 32(1): 1–11.
Strahan, Erin J., Steven J. Spencer, and Mark P. Zanna. 2002. Subliminal priming and persuasion: Striking while the iron is hot. Journal of Experimental Social Psychology 38(6): 556–68.
Strack, Fritz, and Norbert Schwarz. 2021. What’s on your mind? Psychological Inquiry 32(1): 35–37.
Yong, E. 2012. Replication studies: Bad copy. Nature 485: 298–300.
Zaval, Lisa, Elizabeth A. Keenan, Eric J. Johnson, et al. 2014. How warm days increase belief in global warming. Nature Climate Change 4(2): 143–147.