Part A & B (Mal)

RESEARCH ARTICLE

Paranormal experiences, sensory-processing

sensitivity, and the priming of pareidolia

Jess M. WilliamsID*, Mark Blagrove

Department of Psychology, Swansea University, Swansea, Wales, United Kingdom

* [email protected]

Abstract

This investigation tested the effect of priming on pareidolia (the hearing of illusory words in

ambiguous stimuli). Participants (41 women, 20 men, mean age 29.95 years) were assigned

to primed (n = 30) or unprimed (n = 31) groups: the former were told the study was of ‘pur-

ported ghosts voices’, the latter ‘voices in noisy environments.’ Participants were assessed

for perception of human voices within recordings of purported electronic voice phenomena

(EVP), degraded human speech, normal human speech, and white noise. The primed

group had significantly higher perception of voices within EVPs than in degraded speech,

this difference was not found for unprimed participants. In contrast to the previous use of

this design, the primed group did not have higher perception of voices in EVPs and

degraded speech than did the unprimed group. The Aesthetic Sensitivity dimension of the

Highly Sensitive Person Scale (HSPS) was associated with detection of degraded stimuli,

but not with accuracy of stimulus identification. HSPS score was related to lifetime reporting

of anomalous and paranormal experiences. This study partially replicates a paranormal

priming effect and shows relationships between HSPS and detection of ambiguous stimuli

and anomalous and paranormal experiences.

Introduction

Although paranormal phenomena defy what is deemed physically possible [1], there remains

frequent reporting of such experiences within the general population [2, 3]. Attempts have

been made to explore reasons for this reporting through studies on human perception and per-

sonality. The present study addresses perceptual and personality correlates of one type of para-

normal phenomenon, that of electronic voice phenomena (EVP), defined as apparent ghost

voices attempting to communicate through typical audio recording equipment. Although

these ghost voices may be nothing more than random audio signals within white noise, inter-

preted by the perceiver as containing spoken content [4], they are commonly accepted as evi-

dence of the paranormal, and the vulnerability to perceive these as paranormal recordings has

been previously addressed by the literature.

The assumption behind the scientific approach to these perceptual illusions is that suscepti-

bility to the illusions is a result of evolutionary pressures for our ancestors in detecting threats

and subtle social cues [5]. In particular, EVP is an example of an illusion that results in the

PLOS ONE

PLOS ONE | https://doi.org/10.1371/journal.pone.0274595 September 14, 2022 1 / 19

a1111111111

a1111111111

a1111111111

a1111111111

a1111111111

OPEN ACCESS

Citation: Williams JM, Blagrove M (2022)

Paranormal experiences, sensory-processing

sensitivity, and the priming of pareidolia. PLoS

ONE 17(9): e0274595. https://doi.org/10.1371/

journal.pone.0274595

Editor: Fernando Blanco, University of Granada:

Universidad de Granada, SPAIN

Received: May 19, 2021

Accepted: September 1, 2022

Published: September 14, 2022

Copyright: © 2022 Williams, Blagrove. This is an

open access article distributed under the terms of

the Creative Commons Attribution License, which

permits unrestricted use, distribution, and

reproduction in any medium, provided the original

author and source are credited.

Data Availability Statement: All data files are

available from the Zenodo database (https://doi.

org/10.5281/zenodo.6368628).

Funding: The authors received no specific funding

for this work.

Competing interests: The authors have declared

that no competing interests exist.

perception of meaningful patterns in meaningless stimuli, referred to as pareidolia, occurring

in both visual and auditory forms, such as seeing faces in clouds or hearing phantom words in

white noise. Acting as a mechanism for specific paranormal events, a susceptibility to pareido-

lia may be associated with the holding of paranormal beliefs. For instance, Riekki et al [6] dem-

onstrate that paranormal believers are more likely to perceive faces in images regardless of

whether face-like cues are present. Believers also report hearing human voices in ambiguity,

whereas disbelievers report more sound-like signals such as sirens and swooshing [7]. This

may be explained by certain perceptual experiences being consistent with pre-existing para-

normal beliefs, leading to paranormal interpretations of such experiences [8, 9], closely related

to the effects of suggestibility.

Prior expectations and suggestion are also proposed to impact on perceptual experiences.

Halligan and Oakley highlight the effects of priming, such that prior knowledge or belief can

influence individuals’ thoughts, behaviour, and emotions, and exposure to paranormal sugges-

tion can result in the reporting of paranormal phenomena [10]. As an example, Wiseman and

colleagues [11] created a series of fake seances, and suggested to participants that an object was

moving, although no movement occurred. When asked later in the experiment, 31% of partici-

pants falsely recalled that the object had moved, revealing that the mere suggestion of paranor-

mal forces created false memories and paranormal experiences [12]. Interestingly, this false

reporting was significantly associated with paranormal belief. Similarly, Drinkwater et al [13]

found that 36% of their participants reported hearing voices in random noise after watching

videos informing them about EVP, demonstrating that the prior education and expectation of

EVP introduces misperceptions of such within pure noise.

Nees and Phillips [14] aimed to test whether a simple instructional prime could lead to a

perceptual shift, resulting in a higher reporting of voices within ambiguity. A total of twenty-

seven psychology students were recruited and assigned to either a primed or unprimed group.

The primed group were told it was a paranormal study, the unprimed group were left naïve to

the true nature of the experiment. All participants listened to a series of audio recordings, con-

taining examples of EVP, degraded human speech, normal human speech, and pure white

noise. The results revealed that primed participants detected more voices within ambiguous

speech conditions than did unprimed participants. However, there was no agreement across

participants of the spoken content of the EVP stimuli. Baker suggests that where sensory infor-

mation is limited, top-down processes attempt to fill these gaps, resulting in misperception,

and potential reporting of paranormal phenomena [15, 16]. A limitation of this study, how-

ever, was that the small and sceptical sample resulted in the inability to compare believers and

non-believers on their perceptions and susceptibility to priming.

Additionally, a temperament trait that has been associated with both perception and para-

normal belief/experiences is sensory-processing sensitivity (SPS), characterised by an

increased sensitivity to the environment and an enhanced, deeper processing of stimuli, sen-

sory information, and emotions. To describe individuals who experience high SPS, Aron [17]

devised the term Highly Sensitive Person (HSP), and the Highly Sensitive Person Scale (HSPS)

was developed to measure SPS by Aron and Aron [18]. This questionnaire measures aspects of

behaviour and perceptual experiences associated with being a HSP, such as being sensitive to

caffeine, becoming overwhelmed by large levels of sensory input, and being perceived as sensi-

tive or shy. Interestingly, SPS has been viewed as a negative trait within some literature, posing

detriments to the daily functioning of HSPs. For instance, in therapeutic environments, SPS is

confused for other disorders such as obsessive-compulsive disorder, anxiety, and depression

[19]. It has also been associated with negative personality traits such as neuroticism [20, 21], as

well as mental health vulnerabilities [22], poor stress management [23] and nightmare distress

[24, 25].

PLOS ONE Paranormal experiences and pareidolia

PLOS ONE | https://doi.org/10.1371/journal.pone.0274595 September 14, 2022 2 / 19

Consistent with the differential susceptibility framework, HSPs experience intense positive

reactions to positive stimuli, and negative reactions to negative stimuli. Thus, it can be

described as a for better or for worse trait. The previous literature has mainly focused on the for worse, associating SPS with negative emotionality and mental disorders, leading to the inflated

reporting of SPS resulting in negative reactivity to the environment. However, given the cor-

rect environment, HSPs are able to thrive, and function adequately and effectively, with Aron

[19] describing the benefits of having greater empathy and being more creative.

An additional characteristic of this trait is a potential perceptual advantage, due to a greater

depth of processing and awareness of external subtleties [26]. Jagiellowicz et al [27] tested par-

ticipants’ responses and brain activation to subtle changes of scenes in a change detection task.

Participants high in SPS took longer to respond to minor (versus major) scene changes,

thought to be a result of closer attention to the subtle details of the scene. There also seemed to

be higher levels of activation within the visual and attentional areas of the brain during such

changes, once again highlighting deeper processing of the images. In addition, Gerstenberg

[28] found that high SPS was associated with faster reaction times and fewer error rates on a

visual detection task, suggesting an increased processing of visual information, but also the

consequence of increased stress levels.

Although Aron and Aron [18] intended for the HSPS to be a unidimensional measure of

SPS, there is conflicting research suggesting its multi-dimensionality. Smolewska, McCabe,

and Woody [29] proposed there were three subscales of the HSPS. The first, Ease of Excitation

(EOE), suggests a vulnerability to becoming mentally overwhelmed by external stimuli. Sec-

ondly, Aesthetic Sensitivity (AES) describes an “aesthetic awareness”, and thirdly, Low Sensory

Threshold (LST) refers to becoming unpleasantly aroused by external stimuli. Evans and Roth-

bart define AES as Orienting Sensitivity, relating to automatic attentional processing [30] and

interestingly, this subscale seems to represent a different type of sensitivity to EOE and LST.

For instance, Elst et al [31] found that EOE and LST amplified the associations between job

demands and emotional exhaustion in the workplace, and whereas these were considered per-

sonal vulnerability factors, AES acted as a personal resource, finding associations with both

increased helping behaviour and decreased emotional exhaustion. AES also positively corre-

lates with openness to experience and holds no associations with negative personality traits,

such as neuroticism [29].

This suggests that AES represents a positive element of high SPS, in that it results in deeper,

aesthetic processing of subtleties [32], whereas EOE and LST seem to result in overstimulation

in response to the environment and sensory experiences [33], signifying negative effects [30].

Thus, we will test whether the perceptual advantage of SPS, the ability to detect and process

subtleties within the environment, is associated with the positive AES element of SPS and/or

the negative facets of SPS (EOE/LST). Although the previous research claims that AES is asso-

ciated with these automatic sensory processes, there is no behavioural evidence that has tested

this.

Furthermore, many studies have addressed the relationship between reporting of paranor-

mal phenomena and other types of sensitivity, such as thin boundariness [34–36] and height-

ened environmental sensitivity [37–39]. There is limited research, however, which evidences

the association between the paranormal and SPS, although Irwin and colleagues [40] and Wil-

liams et al [41] (using the HSPS and the Survey of Anomalous Experiences [42]) found that

those high in SPS report a greater number of anomalous experiences with paranormal attribu-

tions. Further research is required to support this heightened reporting of paranormal experi-

ences in HSPs.

With these apparent relationships between SPS, paranormal experience and perception, a

paradox emerges in that SPS and paranormal experiences are differentially related to

PLOS ONE Paranormal experiences and pareidolia

PLOS ONE | https://doi.org/10.1371/journal.pone.0274595 September 14, 2022 3 / 19

perception although related to each other. For example, Williams et al [41] demonstrated that

higher reporting of paranormal experiences was positively associated with illusory perception

(pareidolia), that is the misperception of words within random noise [43, 44]. Paranormal

experiences were also related to SPS scores. However, SPS was not associated with mispercep-

tion, but instead to the accurate detection of degraded words. Therefore, this suggests that

both SPS and pareidolia are predictors of paranormal experiences, but are independent of one

another, possibly explaining the observed differences in relationships.

Aims and hypotheses

The methods of Nees and Phillips [14] were used for the current investigation due to the

appropriateness of the stimuli in testing perceptual accuracy and pareidolia. Participants were

split into two groups; one group received a simple instructional prime that informed them of a

paranormal nature to the study. The other group were not primed and instead told it was a

study about noisy recordings. The participants listened to the same four conditions of speech

recordings (34 recordings each of human speech, degraded human speech, white noise, and

EVP) and were instructed to state whether or not they heard a voice after each presentation.

Once the task was completed, they filled in the HSPS and Survey of Anomalous Experiences

(SAE) to measure SPS and paranormal experiences (respectively). The SAE provides partici-

pants a list of potential paranormal experiences which they could have encountered, and they

were to state whether they have had the experience and would attribute it as paranormal or

give it a natural explanation, or simply that they have not had the experience.

The aim of this study was to test the associations between paranormal experiences, SPS,

priming, and perception (including accurate perception and misperception). Firstly, although

the previous literature has demonstrated believers’ susceptibility to suggestion effects (e.g.,

French [12]), there is currently a lack of understanding in terms of the relationships between

paranormal and perceptual experiences. In particular, the research suggests that believers and

experiencers (although limited for the latter) show an enhanced susceptibility to perceptual

illusions, such as pareidolia, potentially providing one mechanism for the reporting of para-

normal encounters whereby ambiguous stimuli in the environment are interpreted as being

paranormal [6, 41]. It is therefore expected that those demonstrating an increased reporting of

paranormal phenomena will be more likely to detect voices within the ambiguous and eerier-

sounding EVP recordings, which can be defined as pareidolia [4]. Also, due to apparent

heightened proneness to priming in believers of the paranormal [11], it was also hypothesised

that participants with higher numbers of paranormal experiences will hear more ambiguous

voices within EVP when primed to expect paranormal recordings. There were no directional

hypotheses between the relationships with the other conditions of speech, although these were

assessed.

Secondly, due to the sparseness of literature focusing specifically on SPS and paranormal expe-

riences, these associations were examined. Thirdly, this study also tested the perceptual advantage

of high SPS. Williams et al [41] highlight that high SPS is associated with the detection of subtle,

degraded stimuli, and not pareidolia, supporting a lower filtering of incoming information, assist-

ing in processing true signals in the environment, and is a potential evolutionary advantage. Thus,

it is hypothesised that SPS (and possibly the facets of SPS, specifically AES) will correlate with

accurate detection of true, degraded stimuli and not with pareidolia. That is, the proportion of

voices detected within the degraded human speech recordings, which contain real voices that

have been audibly degraded to create difficulty with accurate speech perception.

Finally, the study aimed to replicate the results of Nees and Phillips [14]. The authors found

that participants told of the paranormal nature of the study were more susceptible to reporting

PLOS ONE Paranormal experiences and pareidolia

PLOS ONE | https://doi.org/10.1371/journal.pone.0274595 September 14, 2022 4 / 19

voices in all stimulus types, although in general, the perception of voices were similar between

the EVP and degraded human speech recordings (two types of ambiguous stimuli). They also

demonstrated that participants were unable to accurately identify the spoken content of the

EVP messages, as well as the lack of perceptual consistency amongst participants of these

recordings. In-line with this research, it was firstly expected that paranormal primed partici-

pants would be more susceptible to hearing voices in all stimulus types, especially the ambigu-

ous stimulus conditions of EVP and degraded human speech. It was also hypothesised, from

the findings of Nees and Phillips [14], that there will be no difference between the perceptions

of voices within the degraded speech and EVP conditions, due to the ambiguity of these sti-

muli. However, the degraded speech stimuli contain true signals, and thus it is possible that

this stimulus condition will demonstrate a higher percentage of detected voices compared with

the pareidolic stimuli of EVP, this was therefore explored. Also, due to difficulty with non-con-

textual speech perception, participants would also be unable to reliably identify purported

messages in EVP stimuli (due to the absence of human speech) as well as within the degraded

speech (due to sufficient degradation creating difficulty with speech perception). Furthermore,

it was expected that there would be a lack of agreement between participants’ perceptions of

the EVP spoken words and the perceptions of the original paranormal researchers. The cur-

rent study intended to extend these findings through the addition of the SAE and HSPS with a

larger sample, which is an advantage as the authors were limited due to their sample’s overall

scepticism of the paranormal. Additionally, Signal Detection Theory (SDT) was applied to

support relationships with perceptual experiences, providing the ability to differentiate

between true perception and response bias [45].

Materials and methods

Participants

62 participants were recruited through social media and the university psychology credit sys-

tem. Participants were awarded either university credits or a raffle ticket to win a prize. One

participant did not complete the computer task due to program failure, and their data were

excluded. There was a total sample size of N = 61, with 20 men and 41 women, ages 18 to 54

years (Mean = 26.95 years, SD = 9.86), and 90.2% were White. 13.1% of participants were edu-

cated to high school level (or below), 46.5% had completed college or sixth form (A-Level),

31.3% achieved Bachelor’s Level (undergraduate), and 8.2% achieved Master’s Level qualifica-

tions. Full written consent was provided by participants. The study was granted ethical

approval from Swansea University’s College of Human and Health Sciences Research Ethics

Committee.

Sample size, power calculation and sensitivity analyses. Nees and Phillips [14] found

two significant main effects (stimulus type: ηp 2 = .92; group: ηp

2 = .15) as well as a significant

interaction (ηp 2 = .15). The minimum effect size of ηp

2 = .15 (Cohen’s F = .42) inputted into

G�Power software [46] revealed a necessary sample of N = 50 to achieve adequate power (95%,

alpha = .05) for the between-subjects main effect and interaction. However, due to the poten-

tially inflated effect sizes found by the original authors, this might have been an underestima-

tion of the required sample. 61 participants were thus collected, and although the authors

aimed to recruit more, the restrictions posed during the COVID-19 pandemic and time

restraints did not allow for this. A sensitivity analysis found that a sample of N = 61 (95%

power, alpha = .05) was sensitive enough to detect an effect size of Cohen’s F = .19 (ηp 2 = .003)

for the within-subjects main effect and within-between interaction, and an effect size of

Cohen’s F = .37 (ηp 2 = .12) for the between-subjects main effect.

PLOS ONE Paranormal experiences and pareidolia

PLOS ONE | https://doi.org/10.1371/journal.pone.0274595 September 14, 2022 5 / 19

Materials

Stimuli. The stimuli were those used in Nees and Phillips [14]. There was a total of 136 sti-

muli arranged in four conditions (34 stimuli per condition). Each stimulus lasted approxi-

mately 600 milliseconds, including a 5-millisecond onset given by the original researchers. The

four conditions were EVP, human speech, degraded speech, and artificially produced white

noise. The EVP recordings were examples of purported ghost voices captured during real para-

normal investigations and can be defined as pareidolia [4]. Human speech recordings included

clearly spoken words or phrases, such as “Anything”, “Is there anybody”, and “Talk to us”. The

degraded speech stimuli were the same recordings used in the speech condition, degraded

within white noise, with the signal to noise ratio creating challenges for speech perception.

The speech and white noise stimuli acted as control conditions for the EVP and degraded

speech conditions. All audio files (except for the artificial white noise condition, created using

Audacity) were extracted from a series of videos titled Ghost Adventures—Guess that EVP on

the Travel Channel Website (see Nees & Phillips [14] for details of stimulus production).

Questionnaires. The Survey of Anomalous Experiences (SAE) [42] was used to measure

the extent to which participants report having previous paranormal experiences (Cronbach’s

alpha = .83). This questionnaire provides 20 anomalous experiences and participants respond

according to whether they have had the experience and would attribute it as paranormal, if

they have had the experience but believe there was an alternative explanation, or that they have

never had the experience. Questions 14 and 19 were reworded to create a broader anomalous

experience which could apply to more individuals, and the first response option for all items

was reworded to “I think it was. . .” instead of “It must have. . .”. For instance, “I have become aware of a scent in a room, yet there was nothing there that could have that smell”, gives the fol-

lowing response options: “Yes, and I think it was an instance of an apparition or ESP.” (Para-

normal Response), “Yes, but it was probably just an illusion or physiological anomaly.”

(Anomalous Response), or “No” (No Response). The total number of responses per option per

participant was calculated, resulting in three variables of SAE No Responses, Paranormal

Responses, and Anomalous Responses.

To measure SPS, the Highly Sensitive Person Scale [18] (HSPS; Cronbach’s alpha = .85) was

used, which asks about the sensory experiences of the individual. Participants responded to 27

questions with reference to a 7-point Likert scale (1 = Not at all; 7 = Extremely). Questions

include, “Are you easily overwhelmed by strong sensory input?” and “Do you startle easily?”. The

responses to each question were added together to give a total score, and a mean score was cal-

culated, with a higher score representing higher SPS. The mean scores for each of the subscales

were also calculated according to Smolewska et al.’s three subscales of AES (7 items; a = .68),

EOE (12 items; a = .75), and LST (6 items; a = .68) [29].

Procedure

The current design was a replication of the procedure used by Nees and Phillips [14], with the

addition of the SAE and HSPS. Firstly, participants were asked to read an information sheet

that contained all the relevant information regarding the study. The title given to all was

“Voice Detection in Sounds”, so that participants in the unprimed group remained naïve to the

paranormal aspect of the investigation. All provided informed written consent. Participants

sat in front of a MacBook Air (2017) and placed a pair of Sony MDR-ZX310APB headphones

over their ears.

Priming manipulation and computer task. The computer program randomly assigned

participants to one of two groups, either the paranormal primed group (n = 30) or the

unprimed group (n = 31). The only difference between these two groups was the first line of

PLOS ONE Paranormal experiences and pareidolia

PLOS ONE | https://doi.org/10.1371/journal.pone.0274595 September 14, 2022 6 / 19

the instructions. The primed group read the following, “This is an experimental study on the identification of electronic voice phenomenon–purported voices of ghosts in recordings from paranormal research.”, whereas the unprimed group were told, “This is an experimental study of the identification of voices in noisy environments.” The remainder of the instructions were

the same for both groups, that they would hear a variety of sounds in the headphones, and

after hearing each sound they should respond ‘Yes’ or ‘No’, depending on whether they heard

a voice in the recording. To move on from the instructions, participants pressed the spacebar,

and a proceed screen was displayed, asking participants to, “Please press the spacebar to begin a trial.”. This appeared before each of the 136 trials (stimuli) to ensure participants were pre-

pared for and not surprised by the sounds. After pressing, a blank screen appeared, and the

stimulus played after a 1000ms onset. Once the recording had finished, the program moved to

the question screen automatically, asking the participant if they heard a voice in the sound. If

they responded yes, they were immediately directed to an additional question screen, which

asked participants to type in a box a guess as to what they thought the voice had said, and to

not give responses such as “I don’t know” or “unsure”. If they responded no to hearing a voice,

they immediately moved to the next trial proceed screen. An unlimited amount of time was

given to answer, and the stimuli were presented in a different random order for each partici-

pant. The percentage of yes responses per condition was calculated by dividing the total num-

ber of yes responses in that condition by the number of trials (34 trials), as a measure of the

propensity to perceive human voices within the recordings. In addition, participants’ guesses

as to the spoken content of the recordings were extracted and used to test for spoken content

agreement.

Questionnaires. Once the computer task was completed, participants were asked to fill in

electronic versions of the SAE and HSPS on Qualtrics (Qualtrics, Provo, UT) using an Apple

iPad Air (2018). Data were also collected regarding participants’ age, gender, education level,

and ethnicity. Participants were fully debriefed at the end of the study, the purpose of the

investigation was explained, and their group assignment was revealed. The final question on

the questionnaire asked participants if they had guessed the true nature of the experiment, i.e.,

that there was a paranormal nature and/or that they had been randomly assigned to one of two

groups. This was specifically important to ensure the unprimed group remained naïve to this

throughout, so the priming manipulation was not diminished, and they remained unaware of

the paranormal aspect of the investigation. One example of how they could have been aware of

this during the experiment is if the participant had recognised the voices of the paranormal

investigators from the television show “Ghost Adventures”. Another reason was to control for

diffusion of the instructional manipulation that occurred, e.g., amongst class peers, a previous

limitation of the original study [14]. One participant did not give a response as there was an

error with the questionnaire software.

Statistical analyses

All analyses were conducted using IBM SPSS Statistics for Macintosh, Version 26.0 and Ver-

sion 28.0 (IBM Corp, Armonk, N.Y, USA), graphs were created using Python Programming

Language (version 3.8.8) and Microsoft Excel (version 16.57). The variables of interest mea-

sured perception (percentage of yes responses across the four stimulus conditions), paranor-

mal experiences (SAE Responses), and SPS (HSPS Score and scores on each subscale).

Bivariate correlations were conducted between these to test for associations. A multiple regres-

sion was also used to test the distinct relationships of SPS and paranormal experiences with

perception, in that SPS and pareidolia were expected to be independent predictors of the

reporting of paranormal experiences.

PLOS ONE Paranormal experiences and pareidolia

PLOS ONE | https://doi.org/10.1371/journal.pone.0274595 September 14, 2022 7 / 19

Signal detection analysis. To investigate and support associations between perception

and SPS, Signal Detection Theory (SDT) measures were used to distinguish true perceptual

sensitivity from a tendency to respond with bias. Responding “Yes” in the degraded speech

condition were considered “Hits” and responding “Yes” in the noise condition were consid-

ered “False Alarms”. Scores of zero were replaced with 0:5

34 = 0.014 and d’ (sensitivity) and β

(response bias) were calculated per participant, according to the method detailed by Stanislaw

and Todorov [47].

The analyses of Nees & Phillips [14] were replicated. A 2 (priming group) x 4 (stimulus con-

dition) mixed analysis of variance (ANOVA) was conducted, with the proportion of yes

responses as the dependent variable. A 2 (priming group) x 2 (stimulus condition) ANOVA

was also conducted using only the detection of voices within the EVP and degraded speech

stimulus conditions, due to the non-normality of the speech and noise conditions (Nees &

Phillips [14]; footnote 2, p29). Paranormal Responses and SAE No Responses were included as

covariates to control for participants’ existing susceptibility to experience paranormal phe-

nomenon as well as suggestibility. The current study involves priming, stimulus detection, and

paranormal experiences, and thus all variables are present within the ANOVAs.

It is important to note the potential confound that is introduced in that the recordings of

human speech were the basis also of the stimuli in the degraded human speech condition. Thus,

it is possible that an advantage due to priming is present for degraded human speech stimuli

that are presented after the speech equivalent. The task order effect was analysed for n = 40 par-

ticipants (specific order of presentation data were unavailable at time of analysis for the remain-

ing participants), for whom there was a mean of 17.55 (SD = 2.54) degraded stimuli presented

before the speech equivalent and 16.45 (SD = 2.54) presented after, there was no significant dif-

ference between these, t(40) = 1.37, p = .18. This did result in a priming effect, the mean per-

centage of voices detected was 50.24 (SD = 21.70) for degraded words presented after the non-

degraded words, and 44.23 (SD = 22.72) when presented before, t(40) = -2.151, p = .038. As the

order of presentation of spoken word and degraded word stimuli was balanced by randomisa-

tion, order of presentation was not included as a factor in the ANOVA and regression analyses.

Qualitative data (participants’ guesses) were collected and analysed for any agreement

amongst participants and a one-way ANOVA was conducted to test the difference between the

maximum agreement in each stimulus condition (34 trials each), with each trial as a case. Par-

ticipants’ agreement with the original interpretations of the EVP stimuli (according to the sub-

titled videos in which the EVP originated) were also compared to test for spoken content

agreement; the number of times in which participants agreed with these interpretations was

calculated.

Results

Density distributions of HSPS score and the SAE responses are displayed in Fig 1. The assump-

tion of normality was violated for the percentage of yes responses in the speech and noise con-

ditions, SAE Paranormal Responses, as well as for EOE (Shapiro-Wilk, p< .05). The mean

HSPS score was 3.99 (SD = 0.78), suggesting the sample was less sensitive than in Aron and

Aron [18] but is comparable to other studies on SPS (e.g., [41, 48, 49]). Independent t-tests

revealed no significant differences between males and females for any of the variables (all ps>

.05) and age was significantly correlated with the number of SAE No Responses (r(61) = .267,

p = .04) and Anomalous Responses (r(61) = -.343, p = .007), as well as with HSPS Score (r(61)

= -.39, p = .002).

57% of participants reported having at least one prior paranormal experience. The paranor-

mal experience with the most Paranormal Responses was item 5, which states “On at least one

PLOS ONE Paranormal experiences and pareidolia

PLOS ONE | https://doi.org/10.1371/journal.pone.0274595 September 14, 2022 8 / 19

occasion, I’ve had the impression of a figure nearby, yet nobody could possibly have been there”.

20 participants (32.8%) believed they had seen an apparition or ghost. On the other hand, the

least common paranormal occurrence was item 16 “In a life-threatening situation I have had the impression that my disembodied ’self’ was moving along a tunnel toward a light.” Only one

individual had experienced this and attributed it as paranormal, the remaining 60 participants

(98.4%) responded “No”.

All correlations statistics are displayed in Table 1. Relationships between the SAE responses

and voice detection were non-significant (p> .05) across all stimulus conditions, and this

remained the case when correlations were conducted separately for the primed and unprimed

groups (S1 and S2 Tables). HSPS score was positively associated with the number of Paranor-

mal Responses, rs(61) = .298, p = .02, as well as negatively related to the number of SAE No

Responses, r(61) = -.301, p = .02. This suggests that higher SPS is associated with an increased

reporting of paranormal experiences. The subscale AES also had significant associations with

SAE No Responses (r(61) = -.44, p< .001) and Paranormal Responses (rs(61) = .28, p = .03).

Furthermore, the only significant correlation found for SPS and perception was that between

Fig 1. Density distribution of mean HSPS scores, total Paranormal Responses, Anomalous Responses, and SAE No Responses. For HSPS score, lines

represent cut-off points for SPS groups as according to Lionetti et al [48]. The first line is the cut-off between the low SPS group (29.5% of the sample) and

medium group (HSPS score = 3.63) and the second line for medium and high SPS group (31.1% of the sample) (HSPS score = 4.41). Density distributions for

the subscales of the HSPS can be found in the S1 Fig.

https://doi.org/10.1371/journal.pone.0274595.g001

PLOS ONE Paranormal experiences and pareidolia

PLOS ONE | https://doi.org/10.1371/journal.pone.0274595 September 14, 2022 9 / 19

AES and the percentage of degraded words detected, r(61) = .30, p = .02, a higher AES score

was related to higher perception of words within this stimulus condition.

To test the paradox between SPS, paranormal experiences, and perception, a multiple

regression was calculated to predict Paranormal Responses based on two perceptual measures

(yes responses in the EVP (pareidolia) condition and degraded human speech condition) as

well as HSPS score and its three subscales. A significant regression equation was found, F (6,54) = 4.594, p< .001. The model explained 33.8% of the variance, and all variables signifi-

cantly contributed to the model (p< .01), except for the percentage of yes responses in the

EVP and degraded speech conditions (p> .05).

To further explore the relationship with AES and perception, SDT measures were calculated

to test sensitivity (d’) to perceiving voices where voices were present (hits, detection of

degraded human speech) versus when they were not (false alarms, detection of voices in the

noise condition), as well as response bias (β) (i.e., the inclination for participants to respond

with more “Yes” responses or “No” responses). AES was significantly associated with percep-

tual sensitivity (d’), r(61) = .263, p = .04, but not to response bias (β), r(61) = -.026, p = .842

(Fig 2). Correlations between SDT measures and the remaining subscales and SAE Responses

were non-significant (p> .05) (Table 1), and there were no significant differences between

primed and unprimed participants for perceptual sensitivity (t(61) = 0.23, p = .82) or response

bias (t(61) = -0.36, p = .72).

Finally, participants’ guesses were extracted for each trial (34 trials) within the degraded

speech condition, and the number of correct guesses per participant was calculated. Overall,

the number of correct guesses was low across the sample, ranging from 0 to 6, with a mean of

1.43 (SD = 1.50). This suggests that accurate speech recognition was difficult for participants.

Due to the non-normality of the number of correctly identified degraded words, a Spearman’s

Table 1. Correlations between the perception, paranormal experiences, and sensory-processing sensitivity.

SAE No Responses Anomalous Responses Paranormal Responses HSPS Score EOE LST AES

Percentage of Words Detected:

Degraded Human Speech -.214 .055 .117a .116 -.122a .118 .300a

Human Speech -.009a .140a -.021a .062a .236a .131a -.137a

EVP -.123 .093 -.001a -.031 -.109a .045 .135

Noise -.061a .012a .107a .000a -.030a -.001a .072a

SDT Measures: Perceptual Sensitivity (d’) -.216 .052 .109a .095 -.138a .121 .263b

Response Bias (β) .045 .087 -.087a -.073 -.144a -.007 -.026

Paranormal Experiences: SAE No Responses - -.349c -.583ad - - - -

Anomalous Responses - - -.427ad - - - -

Sensory-Processing Sensitivity:

HSPS Score -.301b .058 .298ab - - - -

EOE -.107a .048a .108a .824ad - - -

LST -.169 .114 .158a .837d .603ad - -

AES -.439c .114 .279ab .733d .445ad .452d -

All DFs = 61. aSpearman’s Rho correlations conducted due to non-normality. bp< .05 cp< .01 dp< .001

https://doi.org/10.1371/journal.pone.0274595.t001

PLOS ONE Paranormal experiences and pareidolia

PLOS ONE | https://doi.org/10.1371/journal.pone.0274595 September 14, 2022 10 / 19

rho was conducted and revealed a non-significant correlation between AES and the number of

correctly identified degraded words (rs(61) = .07, p = .61), as well as with HSPS score (rs(61) =

.11, p = .39).

Replicating the analysis of Nees and Phillips [14], a 2 (instructional priming group) x 4

(stimulus condition) mixed ANOVA was conducted with percentage of yes responses as the

dependent variable: due to the design of the current study, Paranormal Responses and SAE No

Responses were added as covariates. The assumption of sphericity was violated, and thus

Greenhouse-Geisser correction was used. A significant main effect of stimulus condition on

yes responses was revealed, F(1.78, 101.52) = 29.42, p< .001, ηp 2 = 0.34. Pairwise comparisons,

with an adjusted significance threshold using Bonferroni correction (a = .008), were used to

test the differences between the four speech conditions. There were significant differences

between all four conditions of stimulus (p< .001), except between the degraded speech and

EVP conditions (p> .05) (Fig 3, for 95% Confidence Intervals see S3 Table). On the other

hand, there was a non-significant main effect of priming group, F(1, 57) = 0.04, p = .84, ηp 2 =

.001. There were non-significant differences between the groups for the detection of voices in

degraded speech (t(59) = -.72, p = .47, 95% CI [-14.686, 6.691]), EVP (t(59) = 1.13, p = .26, CI

[-4.116, 14.799]), speech (t(46.64) = -1.05, p = .30, 95% CI [-0.580, 0.182]), and noise (t(33.28)

= -1.10, p = .14, 95% CI [-1.873, 0.557]). There was also a non-significant interaction between

group and stimulus type, F(1.78, 101.52) = 1.43, p = .24, ηp 2 = .025. The same results held with-

out inclusion of the covariates (as conducted in Nees and Phillips [14]).

Nees and Phillips [14] noted the violations of the assumption of normality for both the

noise and speech conditions (see footnote 2, p.29), which is seen to be the case in the current

study. With these conditions omitted, there was no change to the original authors’ findings of

two significant main effects and a significant interaction. However, removing these and con-

ducting a 2 (group) x 2 (stimulus condition) ANOVA with the current data (with Paranormal

Responses and SAE No Responses as covariates) reveals a significant interaction between

group and stimulus condition, F(1,57) = 4.08, p = .048, ηp 2 = 0.07), although the main effects

were non-significant: stimulus condition, F(1,57) = 0.16, p = .69; group, F(1,57) = 0.09, p = .77.

For the primed group, a paired samples t-test revealed a significant difference between these

two stimulus conditions, t(29) = -2.47, p = .02, 95% CI [-15.057, -1.414]. The primed partici-

pants reported hearing a higher percentage of voices in the EVP stimuli compared with the

degraded speech stimuli. However, for the unprimed group, there was a non-significant

Fig 2. The relationship of AES with Signal Detection Theory measures of perceptual sensitivity and response bias. Scatterplots

with confidence intervals demonstrating the relationship between AES and sensitivity (d’) and the lack of association between AES

and response bias (β). Larger d’ values indicate greater tendency to differentiate signals from noise whereas larger β values

correspond to a bias towards responding “No” (no voices present) [47].

https://doi.org/10.1371/journal.pone.0274595.g002

PLOS ONE Paranormal experiences and pareidolia

PLOS ONE | https://doi.org/10.1371/journal.pone.0274595 September 14, 2022 11 / 19

difference between these conditions, t(30) = 0.38, p = .71, 95% CI [-5.080, 7.972]. This suggests

that participants informed of the paranormal aspect of the study reported hearing more voices

in the EVP than the degraded speech condition (Fig 3).

Qualitative data (participants’ word guesses) were collected and analysed for the agreement

amongst participants as to the spoken content of the stimuli within the four conditions. A total

of 3,954 responses included scoreable guesses as to what the voices within the recordings had

said. Scoreable guesses did not include words such as ‘I don’t know’, ‘Not sure’, ‘Man’s voice’,

or ‘Interference’, and following the previous researchers, partial matches or rhyming words

were not counted as in agreement, but homophones (e.g., ‘here’, ‘hear’) and spelling mistakes

were. There was no trial in the noise condition in which more than one participant agreed

upon the spoken content, thus, with the three remaining stimulus conditions, each trial (stim-

ulus) with more than one response was analysed and the maximum percentage agreement

across participants was calculated. This was done by identifying the word with which the most

agreed and dividing the number of participants who agreed with the word by the total number

of guesses for that trial.

Analysing differences between the percentage agreements in each stimulus condition (EVP,

speech, and degraded speech), a one-way ANOVA was conducted. Stimulus condition was the

independent variable, with each trial as a case (102 cases in total), and the mean percentage

agreement as the dependent variable. The ANOVA revealed a significant difference between

the stimulus conditions, F(2, 99) = 391.16, p< .001, ηp 2 = .89. Bonferroni post-hoc compari-

sons were used to explore the differences between these, with an adjusted alpha value of a =

.017. There were differences between the degraded speech and speech conditions (p< .001),

and between speech and EVP (p< .001). However, there were no significant differences

between the EVP and degraded speech conditions, the two types of ambiguous stimuli (p> .05). Descriptive statistics are displayed in Table 2.

Fig 3. The percentage of words heard in each stimulus condition for the primed and unprimed groups. The mean percentage of yes

responses in the EVP, speech, degraded speech, and noise stimulus conditions. Error bars represent 95% confidence intervals.

https://doi.org/10.1371/journal.pone.0274595.g003

PLOS ONE Paranormal experiences and pareidolia

PLOS ONE | https://doi.org/10.1371/journal.pone.0274595 September 14, 2022 12 / 19

Lastly, participants’ guesses of the spoken content of the EVP stimuli were compared with

the interpretations of the original paranormal researchers (Ghost Adventures). The original

videos in which the EVP were extracted were subtitled with interpretations of the spoken con-

tent. As above, homophones and spelling mistakes were scored as being in agreement, but par-

tial matches were not. 34 trials of EVP were included, with 984 scorable guesses: only 9

responses (0.91%) agreed with the paranormal researchers’ interpretations.

Discussion

With large numbers of paranormal experiences reported in the general population [3], under-

standing these and their impact on thoughts, behaviour, and experiences is essential. The cur-

rent study aimed to replicate the findings of Nees and Phillips [14], as well as extend this to test

the associations of detection of ambiguous stimuli, priming, and paranormal experiences with

sensory-processing sensitivity (SPS). The original findings were partially replicated, and sup-

port was found for the associations between paranormal experiences and SPS and SPS and

accurate perception.

Firstly, there were no associations found between the reporting of paranormal experiences

and perception, revealing no indication of a susceptibility to pareidolia in those who have had

previous paranormal experiences. This contradicts the previous research which suggests that

paranormal believers and potentially experiencers are more susceptible to misperceiving pat-

terns in ambiguity [6, 7]. Furthermore, no support was found for experiencers of the paranor-

mal being more susceptible to priming, although previous research has demonstrated that

believers and experiencers are more likely to interpret anomalous phenomenon as paranormal,

specifically when this is suggested to them [9, 11, 50].

Two explanations are provided for the lack of findings. First, it is possible that the percep-

tion of voices within EVP (pareidolia) is not related to paranormal experiences as measured by

the SAE. Thus, alternative measures of paranormal experiences may provide different results.

The second possible explanation concerns the definition of EVP as pareidolia, in that present-

ing experimentally compiled EVP to measure the tendency to perceive patternicity may be

erroneous. The initial perception of voices within white noise can be defined as pareidolia,

however, spontaneous pattern perception is eliminated once sounds (voices) have been identi-

fied and re-presented, the original perceiver having already attributed these signals as mean-

ingful. In this case, the original perceiver of the EVP stimuli may have experienced pareidolia

upon listening. Presenting to others as meaningful and voice-like encourages the perception of

voices in these noises, and thus eradicates the randomness and spontaneity of the original pat-

tern perception. Importantly, this poses implications on the use of these stimuli to objectively

measure illusory perception. Future research should aim to investigate auditory pareidolia

with random signals or noise that have not been attributed as meaningful prior to the experi-

ment, enabling the ability to test spontaneous pattern perception. One suggestion, with the

adequate resources, is allowing participants to record their own EVP sessions [16] and identify

Table 2. Descriptive statistics for the maximum agreement of the spoken word content per condition.

Mean SD Minimum Maximum

Human Speech 89.03 17.78 31.15 100.00

Degraded Speech 10.69 10.31 0 40.00

EVP 11.80 10.17 0 43.33

The means, standard deviations (SD), minimum, and maximum values for the maximum spoken content agreement in each stimulus condition. Noise was not included

as there was no trial on which more than one participant guessed the same word.

https://doi.org/10.1371/journal.pone.0274595.t002

PLOS ONE Paranormal experiences and pareidolia

PLOS ONE | https://doi.org/10.1371/journal.pone.0274595 September 14, 2022 13 / 19

patterns within their recordings. However, it would be important here to differentiate misper-

ception due to a natural tendency to identify patterns in randomness and misperception that is

enhanced due to a priming effect, i.e., the suggestion that participants will capture meaningful

recordings [13].

Secondly, the number of Paranormal Responses provided was significantly related to HSPS

Score, suggesting that highly sensitives report a larger frequency of prior paranormal experi-

ences. This is supportive of previous findings by Irwin et al [40] and Williams et al [41] and

extends the research focusing on alternative types of sensitivity [34, 37–39]. A potential media-

tor of the relationship between SPS and paranormal experiences could be openness to experi-

ence, a trait that has been previously associated with both SPS [18, 32] and reporting

paranormal experiences [51]. Open individuals tend to report a higher number of these inci-

dents due to their inherent flexibility and experience-seeking behaviours [51], and thus, if

HSPs are more open, this may leave them vulnerable to finding themselves in situations

whereby typical, perceptual experiences can be interpreted as paranormal. This is simply spec-

ulation, and future research should fully explore this mediation effect further.

Although there was a lack of support for HSPS score being related to the detection of true

degraded stimuli, it was found that a higher score on the subscale AES was related to a higher

proportion of voices detected within the degraded human speech condition. This was also

demonstrated using SDT measures, which distinguish perceptual sensitivity from response

bias, in that AES scores were associated with heightened ability to distinguish signals from

noise and not with the tendency to respond with bias. These findings partially support the per-

ceptual advantage of SPS and the multi-dimensionality of the HSPS, such that AES might rep-

resent an aspect of sensitivity that differs from the other subscales, EOE and LST [32].

AES is defined as an awareness of aesthetic subtleties, leading to more positive emotional,

sensory, and perceptual experiences [29, 30]. AES also has associations with positive elements

of personality and behaviours, such as openness to experience and helping behaviours in the

workplace [31, 32]. Therefore, these findings suggest that it is the AES aspect of SPS that cap-

tures the depth of processing described by previous research [18, 19, 27, 28, 41]. This finding

can prove useful in highlighting that high SPS is not a deficit and can result in beneficial behav-

iours, depending on the environment. Demonstrating such advantages can aid in the under-

standing of SPS, such as in therapeutic environments, which support and promote the

flourishing of HSPs in their everyday lives [19].

Furthermore, there were no associations between SPS (or the subscales) and the number of

correctly identified words in the degraded speech condition. This could imply that SPS (AES)

may be associated with stimulus fluency (detecting presence of voices) but not accuracy in

identifying the content of such stimuli (spoken words). Nevertheless, the number of correctly

identified words was very low across the entire sample (ranging from zero to six), suggesting

that the signal to noise ratio simply created too much ambiguity and degradation to allow for

accurate speech identification, regardless of SPS. Such a high level of degradation was the aim

of the original researchers [14] upon creation of the stimuli. Thus, it would be interesting to

explore degradation levels at which accurate stimulus detection and identification can occur,

and the associations of this with SPS, so as to test thresholds for accurate stimulus detection

and identification.

Finally, Nees and Phillips [14] demonstrated that a simple instructional prime resulted in a

perceptual shift; participants that were told it was a study containing paranormal recordings

heard more voices within the ambiguous stimulus conditions. Contrary to this, and with a

larger sample of participants, the prime in the current study did not introduce differences

between the primed and unprimed participants in this measure, both groups identified a simi-

lar proportion of voices across the stimulus types. However, an alternative priming effect was

PLOS ONE Paranormal experiences and pareidolia

PLOS ONE | https://doi.org/10.1371/journal.pone.0274595 September 14, 2022 14 / 19

revealed: primed participants detected significantly more voices within the EVP condition

compared with the degraded human speech, whereas this difference did not occur for the

unprimed group. These findings imply that the EVP stimuli produced a unique response in

those expecting to hear “ghost voices”, perhaps due to the spooky, ominous qualities that these

EVP recordings possess compared with the artificially produced degraded speech. Where

there is no existing expectation of hearing ghostly voices, both types of ambiguous recordings

are interpreted similarly. This emphasises the importance of prior expectations on paranormal

perception, as well as the notion that expectations could lead to false conclusions in real-world

situations [12, 52, 53].

As found previously, within primed and unprimed groups there was a lack of agreement

between participants on the spoken content of the EVP and degraded speech stimuli. Further-

more, there was little to no agreement on the apparent spoken content of EVP items between

the participants in the current study and the original Ghost Adventures researchers’ interpre-

tations of the voices. Thus, although there seems to be acoustical differences in the way parei-

dolic and truly degraded stimuli are perceived, there are contextual similarities in individuals’

attempts to identify the spoken content of such recordings, evidencing the difficulty of non-

contextual speech perception, and suggesting a role of top-down processes for ‘filling the gaps’

where sensory information is limited [15, 16].

Limitations

Although the current study aimed to capture a more representative sample of participants

than in Nees and Phillips [14], the majority (90%) of the participants were White and resided

in Wales, United Kingdom. With notable differences between ethnicities in paranormal

experiencing [54], the generalisability is questionable. Another important limitation to note is

the sample size, which was sensitive enough only to detect an effect size of ηp 2 = .12 for the

between-subjects effect of priming group. Thus, if the true effect size was smaller than this, the

difference between groups would not be detectable.

In addition, the instructional prime was very subtle (only one sentence) and could have

been easily missed by participants. If this occurred, then the prime was ineffective from the

beginning and could explain the lack of difference between the perceptions of primed and

unprimed participants. This is supported by the finding that only seven (24%) primed partici-

pants “guessed” the true nature of the experiment when asked at the end, the true nature being

that there was a paranormal aspect to the investigation. However, it is plausible that partici-

pants misunderstood what this question was truly asking, for example, they may have thought

the question was directed towards the perceptual and personality testing, for which they would

have also been naïve from the outset. Thus, a greater prime may have elicited a stronger effect

on participants’ perceptions. On the other hand, a subtle priming effect was found for the

primed group only, such that they detected more EVP voices than degraded speech. Perhaps

participants had unconsciously processed the paranormal aspects of the instructions which

resulted in distorted perceptions of voices within random noise.

As mentioned previously, there was also a confound introduced by the nature of the stimuli

used in the current paradigm. Specifically, the degraded human speech recordings were the

same as those used in the human speech condition. This means that participants may have had

a detection and identification advantage on trials where the human speech equivalent was

heard prior to the degraded speech recording, leading to inflated scores in the degraded speech

condition only. This has implications for the comparisons conducted with the other condi-

tions, including the EVP condition (another type of ambiguous stimuli). Also, this “prime”

may not have been consistent across participants in the sample, and specific individual

PLOS ONE Paranormal experiences and pareidolia

PLOS ONE | https://doi.org/10.1371/journal.pone.0274595 September 14, 2022 15 / 19

differences (e.g., being a HSP) could determine variation in the effect of the confound across

the sample. Therefore, generalisability issues arise, as the perceptual bias of presenting the

human speech prior to the degraded speech may affect some individuals more than others.

Furthermore, this confound could only be analysed for n = 40 participants due to a file error,

which meant the original (trial) presentation order of n = 21 participants was lost and could

not be restored. Future research adopting this paradigm should take this confound into con-

sideration when determining the presentation order of stimulus conditions, perhaps fully test-

ing the impact it has on perceptual ability and individual differences. Alternatively, the

confound could be avoided by using different words for the speech and degraded speech

conditions.

Conclusion

Perceptual illusions and prior experience and belief can influence humans to erroneously mis-

interpret stimuli and situations, posing real world implications, such as errors in decision-

making, pseudoscientific beliefs, and susceptibility to misinformation about global events [55,

56]. From the current findings, it is shown that increased reporting of paranormal experiences

is not associated with the detection of voices within examples of EVP, and these individuals

are no more susceptible to priming effects. Also, SPS and paranormal experiences are posi-

tively correlated, and the subscale of AES is related to increased perceptual sensitivity to detect-

ing truly degraded stimuli, supported by SDT. We also conclude that the original priming

effect found in Nees and Phillips [14] does not fully replicate, however, a simple instructional

prime resulted in the detection of a larger proportion of ghost voices than of true, degraded

human voices.

Supporting information

S1 Fig. Density distributions for the subscales of the HSPS.

(TIF)

S1 Table. Correlations between perception and paranormal experiences for the primed

group only. The Pearson’s (r) and Spearman’s Rho (rs) correlations between the percentage of

yes responses in each stimulus condition and SAE responses for primed participants.

(PDF)

S2 Table. Correlations between perception and paranormal experienced for the unprimed

group only. The Pearson’s (r) and Spearman’s Rho (rs) correlations between the percentage of

yes responses in each stimulus condition and SAE responses for unprimed participants.

(PDF)

S3 Table. Mean (M) differences between the stimulus conditions and 95% Confidence

Intervals (CI [Upper, Lower]).

(PDF)

Author Contributions

Conceptualization: Jess M. Williams, Mark Blagrove.

Data curation: Jess M. Williams.

Formal analysis: Jess M. Williams.

Funding acquisition: Mark Blagrove.

PLOS ONE Paranormal experiences and pareidolia

PLOS ONE | https://doi.org/10.1371/journal.pone.0274595 September 14, 2022 16 / 19

Investigation: Jess M. Williams.

Methodology: Jess M. Williams.

Project administration: Jess M. Williams.

Resources: Jess M. Williams.

Software: Jess M. Williams.

Supervision: Mark Blagrove.

Validation: Mark Blagrove.

Visualization: Jess M. Williams.

Writing – original draft: Jess M. Williams, Mark Blagrove.

Writing – review & editing: Jess M. Williams, Mark Blagrove.

References 1. American Parapsychological Association. Paranormal. Retrieved from: https://parapsych.org/articles/

53/335/paranormal.aspx [Accessed 5 August 2020].

2. MacDonald WL. The popularity of paranormal experiences in the United States. J. AM. Cult. 1994; 17

(3). Retrieved from https://search.proquest.com/docview/200667091?accountid=14680

3. Castro M, Burrows R, Wooffitt R. The paranormal is (still) normal: the sociological implications of a sur-

vey of paranormal experiences in Great Britain. Sociol. Res. Online. 2014; 19(3)

4. Banks J. Rorschach audio: ghost voices and perpetual creativity. Leonardo Music J. 2001; 11: 77–83

5. Palmer CJ, Clifford CWH. Face pareidolia recruits mechanisms for detecting human social attention.

Psychol. Sci. 2020; 31(8): 1001–1012 https://doi.org/10.1177/0956797620924814 PMID: 32697673

6. Riekki T, Lindeman M, Aleneff M, Halme A, Nuortimo A. Paranormal and religious believers are more

prone to illusory face perception than skeptics and non-believers. Appl. Cogn. Psychol. 2013; 27: 150–

155

7. Simmonds-Moore C. Exploring the perceptual biases associated with believing and disbelieving in para-

normal phenomena. Conscious. Cogn. 2014; 28: 30–46 https://doi.org/10.1016/j.concog.2014.06.004

PMID: 25036936

8. Houran J, Lange R. Hauntings and poltergeist-like episodes as a confluence of conventional phenom-

ena: a general hypothesis. Percept. Mot. Ski. 1996; 83: 1307–1316 https://doi.org/10.2466/pms.1996.

83.3f.1307 PMID: 9017745

9. Rogers P. Paranormal believers’ proneness to probabilistic reasoning biases: a review of the empirical

literature. In Galbraith N, editors, Aberrant Beliefs and Thinking: Current Issues in Thinking and Rea-

soning. 1st ed. London: Psychology Press; 2014.

10. Halligan PW, Oakley DA. Hypnosis and beyond: exploring the broader domain of suggestion. Psychol.

Conscious. 2014; 1(2): 105–122, https://doi.org/10.1037/cns0000019

11. Wiseman R, Greening E, Smith M. Belief in the paranormal and suggestion in the séance room. Br. J.

Psychol. 2003; 94(3): 285–297, https://doi.org/10.1348/000712603767876235 PMID: 14511544

12. French C. Fantastic memories: The relevance of research into eyewitness testimony and false memo-

ries for reports of anomalous experiences. J. Conscious. Stud. 2003; 10(6): 153–174

13. Drinkwater K, Denovan A, Dagnall N, Parker A. Predictors of hearing electronic voice phenomena in

random noise: schizotypy, fantasy proneness, and paranormal beliefs. J. Parapsychol. 2020; 84(1):

96–113

14. Nees M, Phillips C. Auditory pareidolia: effects of contextual priming on perceptions of purportedly para-

normal and ambiguous stimuli. Appl. Cogn. Psychol. 2015; 29(1). https://doi.org/10.1002/acp.3068

15. Baker MJ. Linguistic comprehension of electronic voice phenomena: an experiment in auditory percep-

tion accuracy. Accessed From: https://www.academia.edu/11547676/Linguistic_Comprehension_of_

Electronic_Voice_Phenomena_An_Experiment_in_Auditory_Perception_Accuracy

16. Buckner VJE, Buckner RA. Talking to the dead, listening to yourself. Skeptic, 2012; 17(2): 44–49

17. Aron EN. Counseling the highly sensitive person. Couns. Dev. 1996; 28: 1–7.

PLOS ONE Paranormal experiences and pareidolia

PLOS ONE | https://doi.org/10.1371/journal.pone.0274595 September 14, 2022 17 / 19

18. Aron EN, Aron A. Sensory-processing sensitivity and its relation to introversion and emotionality. J.

Pers. Soc. Psychol. 1997; 73, 345–368. https://doi.org/10.1037/0022-3514.73.2.345 PMID: 9248053

19. Aron EN. Psychotherapy and the highly sensitive person. 1st Ed. East Sussex, England: Taylor &

Francis Group; 2010.

20. Ahadi B, Basharpoor S. Relationship between sensory processing sensitivity, personality dimensions

and mental health. J. Appl. Sci. 2010; 10(7): 570–574

21. Lionetti F, Pastore M, Moscardino U, Nocentini A, Pluess K, Pluess M. Sensory processing sensitivity

and its association with personality traits and affect: A meta-analysis. J. Res. Pers., 2019; 81, 138–152

22. Hofmann SG, Bitran S. Sensory-processing sensitivity in social anxiety disorders: relationship to harm

avoidance and diagnostic subtypes. J. Anxiety. Disord. 2007; 21(7): 944–954

23. Brindle K, Moulding R, Bakker K, Nedeljkovic M. Is the relationship between sensory-processing sensi-

tivity and negative affect mediated by emotional regulation? Aust. J. Psychol. 2015; 67, 214–221

24. Carr M, Nielsen T. A novel susceptibility framework for the study of nightmares: evidence for trait sen-

sory processing sensitivity. Clin. Psychol. Rev. 2017; 58, 86–96. https://doi.org/10.1016/j.cpr.2017.10.

002 PMID: 29074230

25. Carr M, Matthews E, Williams J, Blagrove M. Testing the theory of Differential Susceptibility to night-

mares: The interaction of sensory processing sensitivity with the relationship of low mental wellbeing to

nightmare frequency and nightmare distress. J. Sleep. Res. 2020; 29, e13200. https://doi.org/10.1111/

jsr.13200 PMID: 32985043

26. Greven UC, Lionetti F, Booth C, Aron EN, Fox E, Schendan HE, et al. Sensory processing sensitivity in

the context of environment sensitivity: A critical review and development of research agenda. Neurosc.

Biobehav. Rev. 2019; 98: 287–305. https://doi.org/10.1016/j.neubiorev.2019.01.009 PMID: 30639671

27. Jagiellowicz K, Xu X, Aron A, Aron E, Cao G, Feng T, et al. The trait of sensory processing sensitivity

and neural responses to changes in visual scenes. Soc. Cogn. Affect. Neurosci. 2011; 6: 38–47. https://

doi.org/10.1093/scan/nsq001 PMID: 20203139

28. Gerstenberg FXR. Sensory-processing sensitivity predicts performance on a visual search task fol-

lowed by an increase in perceived stress. Pers. Individ. Differ, 2012; 53: 496–500. https://doi.org/10.

1016/j.paid.2012.04.019

29. Smolewska KA, McCabe SB, Woody EZ. A psychometric evaluation of the Highly Sensitive Person

Scale: the components of sensory-processing sensitivity and their relation to the BIS/BAS and “Big

Five”. Pers. Individ. Differ. 2006; 40(6): 1269–1279

30. Evans D, Rothbart MK. Temperamental sensitivity: two constructs or one? Pers. Individ. Differ. 2008:

44(1): 108–118

31. Elst TV, Sercu M, Van den Broech A, Van Hoof E, Baillien E, Godderis L. Who is more susceptible to

job stressors and resources? Sensory-processing sensitivity as a personal resource and vulnerability

factor. PLOS ONE, 2019; 14(11)

32. Grimen HL, Diseth A. Sensory processing sensitivity: factors of the highly sensitive person scale and

their relationship to personality and subjective health complaints. Percept. Mot. Ski. 2016; 123(3): 637–

653

33. Liss M, Mailloux J, Erchull MJ. The relationships between sensory processing sensitivity, alexithymia,

autism, depression, and anxiety. Pers. Individ. Differ. 2008; 45(3): 255–259

34. Spinelli SN, Reid HM, Norvilitis JM. Belief in and experience with the paranormal: relations between per-

sonality boundaries, executive functioning, gender role, and academic variables. Imagin. Cogn. Pers.

2002; 21(4): 333–346

35. Thalbourne MA, & Maltby J. Transliminality, thin boundaries, unusual experiences, and temporal lobe

lability. Pers. Individ. Differ. 2010; 44(7): 1617–1623. https://doi.org/10.1016/j.paid.2008.01.022

36. Rabeyron T, Watt C. (2010). Paranormal experiences, mental health and mental boundaries, and psi.

Pers. Individ. Differ. 2010; 48(4): 487–492. https://doi.org/10.1016/j.paid.2009.11.029

37. Jawer M. Environmental sensitivity: A neurobiological phenomenon? Sem. Integr. Med. 2005; 3(3):

104–109. https://doi.org/10.1016/j.sigm.2005.10.003

38. Jawer M. Environmental sensitivity: inquiry into a possible link with apparitional experience. J. Soc. Psy-

chical. Res. 2006; 70(1): 25–47

39. Massullo B. Environmental Sensitivity and Paranormal Experiences. [Master’s Thesis] University of

Edinburgh; 2008. Accessed From https://pdfs.semanticscholar.org/ef7c/

b6c6e661b9645b08b7c9bb587f1613f746fa.pdf

40. Irwin HJ, Schofield MB, Baker IS. Dissociative tendencies, sensory processing sensitivity and aberrant

salience as predictors of anomalous experiences and paranormal attributions. J. Soc. Psychical. Res.

2014; 78(917): 193–206

PLOS ONE Paranormal experiences and pareidolia

PLOS ONE | https://doi.org/10.1371/journal.pone.0274595 September 14, 2022 18 / 19

41. Williams JM, Carr M, Blagrove M. Sensory processing sensitivity: Associations with the detection of real

degraded stimuli, and reporting illusory stimuli and paranormal experiences. Pers. Individ. Differ.

2021;117, https://doi.org/10.1016/j.paid.2021.110807

42. Irwin HJ, Dagnall N, Drinkwater K. Parapsychological experience as anomalous experience plus para-

normal attribution: A questionnaire based on a new approach to measurement. J. Parapsychol. 2013;

77(1): 39–53

43. Deutsch D. Phantom Words and Other Curiosities. La Jolla: Philomel Records.

44. Deutsch D. Musical illusions and phantom words: How music and speech unlock mysteries of the brain.

Oxford, England: Oxford University Press; 2019

45. Lynn SK, Barret LF. “Utilizing” Signal Detection Theory. Psychol. Sci. 2014; 25(9):1663–1673 https://

doi.org/10.1177/0956797614541991 PMID: 25097061

46. Faul F, Erdfelder E, Lang A, Buchner A. G*Power 3: A flexible statistical power analysis program for the

social, behavioural, and biomedical sciences. Behav. Res. Meth. 2007; 39: 175–191

47. Stanislaw H, Todorov N. Calculation of signal detection theory measures. Behav. Res. Meth. 1999; 37:

137–149 https://doi.org/10.3758/bf03207704 PMID: 10495845

48. Lionetti F, Aron A, Aron EN, Burns GL, Jagiellowicz J, Pluess M. Dandelions, tulips and orchids: evi-

dence for the existence of low-sensitive, medium sensitive and high sensitive individuals. Transl. Psy-

chiatry. 2018;24 https://doi.org/10.1038/s41398-017-0090-6 PMID: 29353876

49. Acevedo BP, Aron EN, Aron A, Sangster MD, Collins N, Brown LL. The highly sensitive brain: an fMRI

study of sensory processing sensitivity and response to others’ emotions. Brain. Behav. 2018; 4(4):

580–594

50. Harte TM. Contextual mediation of perceptions during hauntings and poltergeist like experiences: a rep-

lication and extension. Percept. Mot. Ski. 2000; 91: 451–459 https://doi.org/10.2466/pms.2000.91.2.

451 PMID: 11065304

51. Perdue A. The relationship between the big five personality traits and paranormal belief (Master’s The-

sis): University of Central Florida, Florida. HIM 1990 2015. 1540

52. Pizzagalli D, Lehmann D, Brugger P. Lateralised direct and indirect semantic priming effects in subjects

with paranormal experiences and beliefs. Psychopathology, 2001; 34: 75–80

53. Lev-ari S, Dodsworth R, Meilke J, Peperkam S. The different roles of expectations in phonetic and lexi-

cal processing. Interspeech 2019. 2019: 2305–2309.

54. Irwin HJ. Belief in the paranormal: a review of the empirical literature. J. Soc. Psychical Res. 1993; 87

(1): 1–39

55. Otgaar H, Schell-Leugers JM, Howe ML, Vilar A, Houben STL, Merckelbach H. The link between sug-

gestibility, compliance, and false confessions: a review using experimental and field studies. Appli.

Cogn. Psychol. 2021; 35(2): 445–455. https://doi.org/10.1002/acp.3788

56. Agley J, Xiao Y. Misinformation about COVID-19: evidence for differential latent profiles and a strong

association with trust in science. BMC Public Health. 2021; 21(89). https://doi.org/10.1186/s12889-020-

10103-x

PLOS ONE Paranormal experiences and pareidolia

PLOS ONE | https://doi.org/10.1371/journal.pone.0274595 September 14, 2022 19 / 19

Copyright of PLoS ONE is the property of Public Library of Science and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.