Paranormal beliefs and cognitive function

The association between cognitive functioning and paranormal beliefs has been researched over several decades. Such functions include memory, attention, language, and executive function (the umbrella term used to describe set-shifting ability, inhibitory control, and working memory updating; for a full description of executive function, see Miyake et al.’s work [37]).

As important for cognitive function is an individual’s belief system. Religious and spiritual beliefs have been associated with slower cognitive decline in older adults [38, 39] but have also been shown to have an inverse relationship with memory performance [40] and intelligence [41, 42]. Similarly, so-called “epistemically unwarranted beliefs” [19], which includes belief in conspiracy theories, has been linked with lower educational attainment and reduced analytical thinking [43, 44]. Conspiracist beliefs are similarly associated with increased illusory pattern perception [45, 46], decreased need for cognition and cognitive reflection [47–49], biases against confirmatory and disconfirmatory evidence [50], and hindsight bias (for discussions on this topic see [51–53]).

The last published review to examine the relationships between paranormal beliefs and various aspects of cognition was conducted by Irwin in 1993 [53]. That non-systematic narrative review of 43 studies is now almost 30 years old and may have introduced bias by “…citing null results only when these form a substantial proportion of the available data on a given relationship” (p.6). At the time of his review, Irwin [53] concluded that, owing to the variable findings, support for the cognitive deficits hypothesis remained uncertain.

Research has grown considerably since Irwin’s [53] review and an updated and systematic review is timely. The current review has two key aims: first, to provide the first assessment of study quality [54] in this area and second, to systematically review and summarise key associations between paranormal beliefs and a range of cognitive functions.

Search strategy

A systematic literature review was chosen for this area owing to its strength as a method to synthesise relevant evidence from large bodies of research [56, 57]. Our searches included both peer-reviewed articles published in scholarly journals and “grey literature” (concerning unpublished works such as doctoral theses).

We searched the electronic databases Scopus, ScienceDirect, SpringerLink, and OpenGrey from inception to May 2021. Our search terms were: (1) “paranormal belief” AND cogni*, (2) “paranormal belief” AND thinking, and (3) “paranormal belief” AND (memory OR “executive function”). For databases that did not permit wildcard Boolean operators (ScienceDirect), one of the above search terms was amended and entered as: “paranormal belief” AND (cognition OR cognitive), to best replicate the effect of the Boolean operator. Following exclusion of duplicate articles across databases, titles and abstracts were assessed to identify studies relevant to the review. Full-text assessment of eligible studies was performed to determine final inclusion. Full-text copies were unavailable for five studies, which were subsequently sought for retrieval. Finally, we hand-searched reference lists for each included article to identify any additional relevant articles. The PRISMA flow diagram presented in Fig 1 illustrates the full screening and selection process. The PRISMA checklist for abstracts is presented in S1 Appendix, and the full PRISMA checklist is presented in S2 Appendix.

Inclusion/Exclusion criteria

Studies were eligible for inclusion if they were: published in the English language, conducted with a healthy adult sample (age 18 or over) and presented original data involving both a measure of paranormal belief and a measure of cognitive function. As cognitive functions have been shown to peak at different ages (for a detailed discussion on this topic, see [58]), we excluded samples that included children and adolescents under the age of 18 as some cognitive functions are still developing in these younger individuals.

Data extraction

We used a detailed data extraction form to collate the following information from included studies: sample sizes and demographic details (including sex, age and education), the measures of self-rated paranormal belief, the aspect of cognition assessed, the tests of cognitive functions used, and findings relating to the relationship between paranormal beliefs and cognitive function. We categorised eligible outcome measures broadly to include both global cognitive function and domain-specific cognitive functions. Any measure of cognitive function was eligible for inclusion (e.g., neuropsychological tests, self-report measures). Results for both paranormal beliefs and cognitive functioning could be reported as an overall test score that provides a composite measure, subscale scores that provide domain-specific measures, or a combination of the two. When multiple cognitive outcomes were investigated, we included all measures. To assess the strength of the relationships between paranormal beliefs and various cognitive functions, we calculated the number of positive, negative, or null findings reported by each study included in the review. Measures of paranormal belief were examined to determine the extent to which established questionnaires have been used.

In line with our preregistered protocol, we synthesised evidence narratively. Meta-analyses could not be undertaken because of the heterogeneity of study designs and outcome measures. We did, however, develop summary tables that include information relating to: sample size, gender composition, mean sample age, cognitive domain, outcome measure, and key findings. Given the range of outcome measures, we attempted to categorise the included studies by common cognitive domains. As the review took an explorative approach, and did not specify domains of interest, categorisation took place after full-text evaluation of included studies.

Assessment of study quality and risk of bias

The preregistration for this review specified using a bespoke series of questions to assess study quality, but we subsequently decided to use a more well-established and validated measure of study quality in the Appraisal tool for Cross-Sectional Studies (AXIS) tool [60]. Of the 20 AXIS items, seven assess reporting quality (items: 1, 4, 10, 11, 12, 16 and 18), seven relate to study design (items: 2, 3, 5, 8, 17, 19 and 20), and six to possible biases (items: 6, 7, 9, 13, 14 and 15). Two authors (DG and CED) independently rated each study, and these two sets of ratings had almost-perfect agreement (93%) with Kappa = .84.

Following previous research [61], we classified AXIS quality scores according to the number of “Yes” responses for the 20 items for each study—poor quality for scores <50%, fair quality for scores between 50 to 69%, good quality for scores of 70% to 79%, strong quality for scores of 80% and higher. Three in four studies were rated as either ‘strong’ (26/71: 37%) or ‘good’ (27/71: 39%). By contrast, 17/71 (24%) were rated as ‘fair’ and only 1/71 (1%) was rated as ‘poor’. The mean quality rating score across all 71 studies was in the ‘good’ range; however individual AXIS items are not weighted and so this total score provides a general, but limited, classification that should be interpreted with some caution. The number of papers meeting each AXIS criterion (‘Yes’) is presented in Table 1. The number of