Hidden God Concepts in CAPTCHAs Offer a Cleaner Way to Prime Religious Thought

Every few days, millions of people prove their humanity by deciphering distorted text, clicking traffic lights, or assembling jigsaw tiles. Those fleeting moments of verification have now been recruited for a very different purpose: quietly activating religious concepts in the human mind. In a study published in Behavior Research Methods, a team of psychologists led by Fırat Şeker and Ensar Acem of Kadir Has University, working with colleagues including Ozan Isler of the University of Queensland and Onurcan Yilmaz, introduces what they call allusive priming, a technique that embeds religious themes inside CAPTCHA-like tasks. The approach is designed to solve one of the most stubborn methodological problems in the psychology of religion: how to put an idea into someone’s head without that person realizing that this is exactly what the experiment is trying to do.

Religious priming has a long and contested history. Since the influential 2007 finding by Ara Norenzayan and Azim Shariff that subtly reminding people of God increased their generosity in an anonymous economic game, researchers have used scrambled sentences containing religious words, subliminal images, and explicit passages about divine judgment to test whether religious cognition shapes moral behavior. A 2016 meta-analysis by Shariff and colleagues suggested the effects were real but modest, while later preregistered replications, such as the 2015 effort by Gomes and McCullough, often failed to confirm the original results. The field has been left with a methodological dilemma that the new study directly confronts.

That dilemma concerns the two broad families of priming techniques. Implicit primes, such as word fragments or subliminal flashes, are meant to bypass awareness, but critics argue they often lack construct validity: researchers cannot be sure the prime actually activates the intended concept rather than something adjacent or nothing at all. Explicit primes, such as a paragraph asking participants to reflect on God, are transparent and demonstrably effective, but they invite demand characteristics, the tendency of participants to guess the hypothesis and behave accordingly, along with social desirability biases. Because religious people tend to score higher on socially desirable responding, as meta-analytic work by Sedikides and Gebauer has shown, explicit religious manipulations are particularly vulnerable to contamination. The validation crisis in psychology, documented by researchers such as Schimmack and by Flake and colleagues, has made it clear that manipulations must be validated, not merely assumed to work.

The allusive priming technique was built to thread the needle between these extremes. Instead of hiding religious content below awareness or announcing it openly, the researchers embedded it in plain sight within tasks that participants perform routinely online. Following a preliminary calibration study with 104 participants to select the most effective materials, the team constructed CAPTCHA-style challenges in which religious themes were woven into the stimuli themselves. The primes targeted four distinct conceptions of God: punitive, rewarding, forgiving, and neutral. Participants might, for example, complete word puzzles or select images whose content subtly evoked divine punishment or divine reward, all while the task appeared to be an ordinary verification step rather than a psychological manipulation.

The construction of the materials was itself a hybrid of human judgment and artificial intelligence. The researchers used large language models to generate candidate word lists for each category, including 60 punitive, 60 rewarding, and 60 neutral religious words, and then employed ChatGPT in structured dialogues to classify whether individual words carried religious meanings on their own. Neutral filler words unrelated to any theme were generated separately, with explicit instructions to avoid conjunctions. This AI-assisted pipeline allowed the team to assemble a large pool of candidate stimuli quickly, which the calibration study then filtered down to the items that best conveyed the intended concepts without drawing attention to themselves. The authors note that images representing the forgiving aspect of God were harder to depict clearly than punitive or rewarding imagery, which is why a wider range of alternatives was considered for that condition.

The main test was a preregistered online experiment with 2,770 participants from the United States, a sample large enough to detect even small effects with confidence. Participants were randomly assigned to one of two allusive priming conditions or to a standard explicit prime, and then completed two downstream measures. The first was a Word Completion Task, in which fragmented words could plausibly be completed as either religious or non-religious terms; the number of religious completions served as an index of how accessible religious thoughts had become. The second was a dictator game, a classic economic paradigm in which participants receive an endowment and decide how much, if anything, to transfer to an anonymous stranger, providing a behavioral measure of generosity that has been the standard outcome in religious priming research since the field’s earliest experiments.

The results were encouraging on the methodological front and sobering on the theoretical one. One of the two allusive primes significantly increased religious word completions compared with the control, demonstrating that the technique genuinely activated religious cognition. Critically, participants who received the allusive primes were less aware of the manipulation’s purpose than those who received the explicit prime, which is precisely the pattern one would hope for if the goal is to reduce demand effects. In other words, the CAPTCHA-based approach managed to be both effective and covert, a combination that traditional implicit and explicit methods have struggled to achieve simultaneously. The authors argue that this positions allusive priming as a promising alternative that offers greater flexibility and ecological validity for priming a wide range of psychological constructs, not only religious ones.

The generosity result, however, told a different story. Despite the successful activation of religious thoughts, religious priming did not significantly influence how much money participants transferred in the dictator game. This null finding adds to a growing body of evidence questioning whether religious primes reliably change prosocial behavior at all. The theoretical backdrop here is the supernatural monitoring hypothesis, developed by scholars such as Dominic Johnson and Jesse Bering, which holds that belief in watchful, punishing gods promotes cooperation because it deters cheating even in anonymous settings. Supporting work includes the 2006 finding by Bateson, Nettle, and Roberts that images of watching eyes increase real-world cooperation, and studies showing that children who believe in an invisible watcher cheat less. Yet a 2024 meta-analysis of nearly 60 years of research by Kelly, Kramer, and Shariff found that religiosity predicts prosociality most clearly when measured by self-report, raising the possibility that the link reflects how religious people present themselves rather than how they behave when no one, human or divine, is watching.

The new study’s design also speaks to broader concerns about the credibility of psychological experiments. The work was preregistered, its data, code, and materials are openly available on the Open Science Framework, and it received ethical approval from Kadir Has University with informed consent from all participants. Funding came from the Templeton Religion Trust. By validating the manipulation itself, checking that the primes actually raised the salience of religious thought before drawing conclusions about downstream behavior, the researchers modeled the kind of construct validation that Chester and Lasko and others have urged social psychologists to adopt. The separation between manipulation check and outcome measure also allowed the team to interpret the null effect on generosity with unusual clarity: the prime worked, but the behavior did not follow, which is a very different conclusion from one in which a failed manipulation masks a failed theory.

What makes the technique genuinely novel is its ecological grounding. CAPTCHAs are among the most familiar interactive elements on the web, and their format, brief puzzles requiring careful attention to distorted or fragmented stimuli, turns out to be an ideal vehicle for delivering content that participants process but do not scrutinize. The authors suggest that allusive priming could be adapted to activate many constructs beyond religion, from security concerns to environmental attitudes, wherever researchers need a manipulation that is both potent and inconspicuous. Whether the activation of religious concepts can ever reliably translate into changed behavior remains an open question, and the dictator game null result suggests caution. But the study demonstrates that the tools of everyday digital life can be repurposed for rigorous experimental science, and that the path forward for priming research may lie not in more powerful primes, but in smarter, subtler, and better-validated ones.

Subject of Research: A CAPTCHA-based allusive priming technique for activating religious concepts in psychological experiments

Article Title: Divine verification: Introducing a CAPTCHA-based religious priming technique

Article References: Divine verification: Introducing a CAPTCHA-based religious priming technique. (n.d.). https://doi.org/10.3758/s13428-026-03147-1

Image Credits: AI Generated

DOI: 10.3758/s13428-026-03147-1

Keywords: religious priming, CAPTCHA, allusive priming, demand characteristics, construct validity, dictator game, prosociality, supernatural punishment, Behavior Research Methods, preregistered experiment, cognitive science of religion, word completion task