Anxiety Enhances Recall of Supernatural Agents

ABSTRACT

The motivational account of religious belief – that belief fulfills some psychological need – has been historically popular, and recent studies have identified a causal role for anxiety in particular. However, the cognitive mechanisms by which anxiety ultimately produces religious belief are unclear. In two studies, we show that anxiety intensifies a known cognitive bias to recall supernatural agents via preferential processing of the threatening characteristics of these agents. Across the two studies, participants exposed to an anxiety manipulation at encoding (but not at retrieval) exhibited a stronger recall bias for supernatural agents than controls, regardless of how anxiety was elicited and regardless of participants’ religiosity. The results suggest that people in anxious states are more likely to remember and accumulate representations of supernatural or “godlike” agents than people in non-anxious states, potentially biasing them toward religious belief in these agents. This work therefore lends support and detail to the motivational account, addresses the puzzle of why some malevolent gods attract believers, and, by illustrating the importance of anxiety in recall for supernatural agents, argues for the construction of cognitive-motivational models of religious belief.

Acknowledgments

The authors thank Samantha Smith for research assistance. They also thank Brittany Cardwell, Jonathan Jong, two anonymous reviewers, and researchers in the University of Otago’s Social Cognition Laboratory for their helpful feedback.

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplementary material

Supplemental data for this article can be accessed on the publisher’s website.

Notes

¹ Or 194 participants for each of the encoding and retrieval procedures (see below).

² No participants were added after analysis. We recruited more than the required number of participants in anticipation of exclusions.

³ See Dance (2015) for a discussion about the validity and reliability of internet studies and of Mechanical Turk. Although participants recruited online are more diverse in their age and geography than undergraduate or local participant pools, there may be greater potential for inattention, distraction, cheating, and familiarity with psychological research. To partially address these concerns in our studies, we employed quality controls (participants had a >95% past approval rate for completed tasks, were excluded if they took part in similar studies with our lab, and were asked to not discuss the survey with anyone), an attention check, a question about cheating, and time-based exclusion criteria (see below).

⁴ Despite the apparent positivity of the control condition, Pacheco-Unguetti et al. (2010) did not report a significant reduction from baseline anxiety levels.

⁵ Other ontological categories (objects, other living things, etc.) are less relevant to god beliefs.

⁶ Reported unpleasantness is a product of anxiety. A more extensive manipulation check, such as a state anxiety scale, was not used as the procedure is time dependent, and Pacheco-Unguetti et al. (2010) had confirmed the manipulation produces anxiety.

⁷ In total, for the encoding procedure, 15 participants were excluded from the anxiety condition and 21 from the control condition. For the retrieval procedure, 15 participants were excluded from the anxiety condition and 17 from the control condition.

⁸ No participants were added after analysis. We recruited more than the required number of participants in anticipation of exclusions.

⁹ Participants were recruited from Mechanical Turk and paid US$.70. All were native English-speaking U.S. nationals currently living in the U.S. Their average age was 36.3 years (SD = 12.3); 128 identified with a religion (of which 94% were Christian); 89 identified as “agnostic”, “atheist”, or “none”; and 4 identified as “other” (e.g., spiritual). There were 97 female, 122 male, and 2 identifying as “other”. A power analysis, using the effect size reported by Pacheco-Unguetti et al. (2010) for the Study 1 conditions, indicated that 180 participants would be required to obtain 80% power with α set to .05.

¹⁰ Between conditions (see Table 2), 10 participants were excluded from “social,” 16 from “impersonal harm,” 11 from “interpersonal harm,” 6 from “impersonal death,” 9 from “interpersonal death,” and 17 from “control.”

¹¹ Quantitatively, this method does not differ from the ANOVA procedure, which also tests differences between means.

¹² Comparing each of the anxiety conditions to control yielded effect sizes ranging from r = .10 to r = .17, with the latter equaling the effect size observed in Study 1.

¹³ Although ecological validity could be improved with use of a culturally “real” religious narrative, it should be clear that many gods and spirits are believed to possess the supernatural abilities in our recall materials, and that our narrative could be perceived as taking place in an archaic setting (as in many religious narratives). Nevertheless, the category of supernatural agents includes many cultural, fictional, and mythological creatures (e.g., Santa Claus, Superman, Dracula), which is why a supernatural recall bias cannot fully explain religious belief beyond an initial “inclination” or attraction toward such agents (Swan & Halberstadt, 2019), and why we suggest motivational processes and cultural learning heuristics are necessary for belief in particular agents.

Additional information

Funding

This research was supported by grants from the John Templeton Foundation [52257 and 60624] and The New Zealand Royal Society’s Marsden Fund [UOO1312].