Content uploaded by Shu Imaizumi
Author content
All content in this area was uploaded by Shu Imaizumi on Feb 16, 2018
Content may be subject to copyright.
Available via license: CC BY 4.0
Content may be subject to copyright.
Imaizumiand Tanno BMC Res Notes (2018) 11:128
https://doi.org/10.1186/s13104-018-3245-5
RESEARCH NOTE
Rasch analysis oftheTrypophobia
Questionnaire
Shu Imaizumi1,2* and Yoshihiko Tanno1
Abstract
Objective: This study aimed to assess Rasch-based psychometric properties of the Trypophobia Questionnaire meas-
uring proneness to trypophobia, which refers to disgust and unpleasantness induced by the observation of clusters of
objects (e.g., lotus seed pods).
Results: Rasch analysis was performed on data from 582 healthy Japanese adults. The results suggested that Trypo-
phobia Questionnaire has a unidimensional structure with ordered response categories and sufficient person and
item reliabilities, and that it does not have differential item functioning across sexes and age groups, whereas the
targeting of the scale leaves room for improvements. When items that did not fit the Rasch model were removed, the
shortened version showed slightly improved psychometric properties. However, results were not conclusive in deter-
mining whether the full or shortened version is better for practical use. Further assessment and validation are needed.
Keywords: Psychology, Emotion, Disgust, Personality, Individual differences, Psychometric properties, Rasch
measurement model
© The Author(s) 2018. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License
(http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium,
provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license,
and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/
publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
Introduction
Trypophobia refers to disgust and/or unpleasantness
evoked by images depicting clusters of roughly circular
objects that are usually innocuous (e.g., lotus seed pods)
[1]. Trypophobic responses can also be somatic, such as
itch and nausea [2, 3], affecting both skin conductance [4]
and pupil constriction [5]. Trypophobic stimuli may be
associated with skin coloration of poisonous animals [1],
skin lesions [6, 7], and the observers’ skin disease history
[8], and consequently can trigger pathogen-avoidance
behaviors as trypophobic responses [3, 8, 9]. Moreover,
trypophobic images possess spatial-frequency character-
istics likely to induce visuoperceptual discomfort [1, 2,
10]. As these potential factors can vary between individu-
als, there is substantial inter-individual variability [2]; for
instance, 46 of 286 English adults exhibited aversion to a
trypophobic image while the others did not [1].
e Trypophobia Questionnaire (TQ) was developed
to assess the extent to which respondents can experience
subjective and somatic responses expressing disgust and/
or unpleasantness induced by trypophobic images [2],
and has been employed to elucidate mechanisms of try-
pophobia [4, 9–12]. Although some studies have con-
firmed a one-factor structure, reliability, and validity of
the TQ [2, 9, 11], there remains room for analyses of its
psychometric properties. is study investigated Rasch-
based psychometric properties of the TQ. Rasch analysis,
in contrast to classical test theory, computes the extent
to which the observed responses fit the responses pre-
dicted by the Rasch measurement model, and assesses
the scale’s unidimensionality and precision in measure-
ment [13, 14].
Main text
Methods
Participants
We recruited 584 Japanese adults via Lancers [15], a
crowdsourcing service. To sample the general population,
crowdsourcing was conducted with only the following
requirements: participants should be healthy, older than
18years, and native Japanese speakers. Sample size was
based on a guideline indicating that 500 participants are
required for precise and robust Rasch analysis [16]. One
Open Access
BMC Research Notes
*Correspondence: shuimaizumi@gmail.com
1 Graduate School of Arts and Sciences, The University of Tokyo, 3-8-1
Komaba, Meguro, Tokyo 153-8902, Japan
Full list of author information is available at the end of the article
Page 2 of 5
Imaizumiand Tanno BMC Res Notes (2018) 11:128
participant who did not complete the TQ and another
who provided invalid responses (i.e., extreme agree-
ment for all items including dummies) were excluded.
Data from 582 participants were analyzed [338 females,
age 19–81years, mean=39.59, standard deviation
(SD)=9.96].
ere was no sex or age bias in the sample: a Mann–
Whitney test showed no difference in age between sexes
(U=44,375, p=0.117, ρrb=0.076; ages in males and
females were not normally distributed, Shapiro–Wilk
Ws<0.98, ps<0.001). Participants younger than or
equal to the median age of 39years were comparable to
those older in terms of male/female ratio (χ2(1)=0.05,
p=0.833, φ=0.009).
Procedures
Participants were directed to a survey website generated
by Qualtrics [17] via their own computers. ey reported
their sex and age, completed the TQ, and were paid 162
Japanese yen (approximately 1.4 US dollars).
e TQ is a one-factor structure questionnaire includ-
ing 17 items (Table1) to assess proneness to subjec-
tive and somatic responses induced by trypophobic
images and two dummy items irrelevant to the construct
of interest (“Want to laugh,” “Feel at peace”) [2]. We
employed the Japanese version, which has been validated
for use in an adult sample [18]. Participants observed two
trypophobic images (lotus seed pods, honeycombs) as
in the original [2], and rated their agreement with each
item on a 5-point scale ranging from 1 (“Not at all”) to
5 (“Extremely”). e scale score was the summed item
scores, excluding dummy items.
Data analysis
Descriptive statistics of the TQ except for dummy items
and its relationships with sex and age were analyzed
using JASP 0.8.5.1 [19]. Rasch analysis was performed
using Winsteps 4.0.1 [20]. As all items shared the same
polytomous response structure [21], the rating scale
model was employed [22]. e procedures of Rasch anal-
ysis were based on recent guidelines [21, 23].
Ordering of thresholds between five response catego-
ries of the TQ was assessed. resholds refer to points
at which two adjacent curves cross. Disordering implies
underused and/or indistinguishable categories.
Unidimensionality (i.e., to what extent the scale
assesses single construct) was assessed by principal
component analysis (PCA) of the residuals based on the
amount of raw variance explained by the measure and
the eigenvalue of unexplained variance in the first con-
trast (i.e., latent dimension). As in the previous studies,
we also reported factor analysis for descriptive purposes.
Infit and outfit mean-squares for each item were
the indices of the fit to Rasch unidimensional model.
Table 1 The Trypophobia Questionnaire andits Rasch-based psychometric properties
Dierential item functioning contrasts were reported in absolute values. Zstd, z-standardized statistic; DIF, dierential item functioning
Location
(standard
error)
Int mean-
square
(Zstd)
Outt mean-
square (Zstd) First
contrast
loading
Second
contrast
loading
DIF
contrast
forsex
DIF
contrast
forage
Feel freaked out − 0.50 (0.06) 1.07 (1.10) 1.08 (1.00) 0.55 − 0.17 0.14 0.00
Feel aversion, disgust or repulsion − 1.75 (0.05) 0.88 (− 2.10) 0.88 (− 1.80) 0.61 0.41 0.16 0.08
Feel uncomfortable or uneasy − 1.68 (0.05) 0.73 (− 4.80) 0.80 (− 3.10) 0.63 0.23 0.11 0.10
Feel like panicking or screaming 0.33 (0.07) 0.98 (− 0.20) 0.83 (− 1.70) 0.24 − 0.45 0.12 0.08
Feel anxious, full of dread or fearful − 0.52 (0.06) 0.87 (− 2.00) 0.84 (− 2.10) 0.46 − 0.31 0.10 0.00
Feel sick or nauseous 0.02 (0.06) 0.87 (− 1.90) 0.74 (− 3.10) − 0.13 − 0.24 0.00 0.06
Feel nervous (e.g., heart pounding, butter-
flies in stomach, sweating, stomachache,
etc.)
0.05 (0.06) 0.92 (− 1.10) 0.87 (− 1.40) − 0.09 − 0.21 0.17 0.06
Feel like going crazy 0.00 (0.06) 0.82 (− 2.60) 0.79 (− 2.40) − 0.14 − 0.27 0.00 0.22
Have an urge to destroy the holes 0.97 (0.08) 2.16 (9.90) 2.32 (7.10) − 0.06 − 0.22 0.51 0.33
Feel itchiness 0.24 (0.06) 1.56 (6.60) 1.40 (3.60) − 0.16 0.18 0.22 0.20
Feel skin crawl − 0.90 (0.05) 0.86 (− 2.30) 0.82 (− 2.70) − 0.20 0.66 0.17 0.09
Have goosebumps − 0.83 (0.05) 0.96 (− 0.70) 0.93 (− 0.90) − 0.40 0.62 0.08 0.08
Feel like crying 1.96 (0.10) 1.57 (4.50) 0.89 (− 0.50) − 0.37 − 0.37 0.13 0.19
Vomit 1.40 (0.08) 1.08 (0.90) 0.77 (− 1.40) − 0.41 − 0.24 0.37 0.07
Get chills − 0.26 (0.06) 1.14 (2.10) 1.01 (0.10) − 0.43 0.35 0.05 0.10
Have trouble breathing 0.63 (0.07) 0.97 (− 0.40) 0.86 (− 1.20) − 0.27 − 0.37 0.07 0.21
Shiver 0.83 (0.07) 0.98 (− 0.20) 0.72 (− 2.40) − 0.47 − 0.16 0.22 0.11
Page 3 of 5
Imaizumiand Tanno BMC Res Notes (2018) 11:128
Infit mean-square is based on the Chi square statistic
weighted using model variance and sensitive to inliers.
Outfit mean-square is based on the conventional Chi
square statistic and sensitive to outliers. Because mean-
squares indicate the amount of distortion of the meas-
urement system and their expected values are close to
1.00, values less than 1.00 indicate overfit and those
greater than 1.00 indicate underfit to the model. We
also reported infit and outfit z-standardized statis-
tics (i.e., standardized t-statistics with infinite degrees
of freedom), which indicate statistical significance of
mean-squares.
Rasch measure, which was computed for each person
and item and expressed in logits, indicates the location
on the unidimensional latent variable. We assessed tar-
geting, which is the difference between mean person and
item location measures and indicates how well item diffi-
culties match individuals’ abilities. Differential item func-
tioning (DIF) can be assessed by subgroup differentials of
the Rasch item measure, indicating whether a subgroup
of a sample scores on an item different from another sub-
group. We assessed DIF across two subgroups: sex (male
versus female) and age (younger or equal to versus older
than the median of 39years).
Person and item reliabilities (i.e., reproducibility) based
on the Rasch model were analyzed. High person or item
reliability indicates high probability that persons or
items with high estimated measures indeed show higher
measures than do persons or items with low estimated
measures. Specifically, person reliability reflects repro-
ducibility of person ordering that can be expected if the
same sample responded to another set of items measur-
ing the same construct, and item reliability reflects repro-
ducibility of items’ hierarchy and/or given item scores if
another sample responded to the same items [24]. We
also reported internal consistency (i.e., Cronbach’s alpha)
for descriptive purposes.
Results anddiscussion
Descriptive statistics
e mean TQ score was 32.02 [SD=13.71; range 17–85;
skewness (standard error)=1.10 (0.10); kurtosis=0.70
(0.20)]. TQ scores in total, male, and female samples were
not normally distributed (Ws<0.90, ps<0.001). While we
found no sex difference in TQ score (Meanmale=31.44,
SDmale=12.66, Meanfemale=32.44, SDfemale=14.44,
U=40,596, p=0.749, ρrb=−0.016), age negatively but
weakly correlated with TQ score (ρ=−0.227, p<0.001).
ese (null) effects of sex and age on TQ were consistent
with previous studies [9, 18]. As it is outside the scope
of this study, relief from trypophobia with age should be
investigated by future research.
Rasch analysis
We confirmed continuous ordering of thresholds of the
five response categories (Fig.1); their average measures
were −2.96, −1.47, −0.49, 0.40, and 1.37, respectively
[21]. is suggested that all response categories were dis-
tinguished and evenly used by participants.
Rasch-based PCA showed that the measures explained
64.1% of the raw variance, which was above the crite-
rion of 50.0% for unidimensionality of the scale [23]. e
eigenvalues of the unexplained variance in the first and
second PCA contrasts (i.e., latent dimensions) were 2.38
and 2.10, respectively, which exceeded the cutoff of 2.00
[23]. ese suggested that the TQ possesses unidimen-
sionality, but also that there may be other latent dimen-
sions in the residuals. We thus examined the presence of
multidimensionality in terms of the correlation between
item clusters within each contrast [21]. e items were
separated into three item clusters based on each of the
first and second contrast loadings (Table1). e corre-
lation between item clusters was reported as disattenu-
ated Pearson correlation coefficient, which removed the
standard error of measurement for each item cluster.
When the coefficient approaches 1.000, a pair of item
clusters measures the same construct [21]. We indeed
found that the coefficients were very high: 0.928–1.000
for the first contrast and 0.988–1.000 for the second.
is suggests that the item clusters defined by two latent
dimensions (i.e., contrast) indeed measure the same
construct, supporting the unidimensionality of the TQ.
Factor analysis also confirmed its one-factor structure,
consistent with previous studies [2, 11, 18] (see Addi-
tional file1).
Most items were well fitted to the unidimensional
model; infit and outfit mean-squares were between 0.73
Fig. 1 Category probability curves for the Trypophobia
Questionnaire. The curves show ordered thresholds between five
response categories
Page 4 of 5
Imaizumiand Tanno BMC Res Notes (2018) 11:128
and 1.14, within a criterion range (i.e., 0.70–1.30), except
for the items “Have an urge to destroy the holes,” “Feel
itchiness,” and “Feel like crying,” whose infit and/or out-
fit mean-squares were above 1.30 (Table1). ese can be
interpreted as low-quality fit to the scale [23]. erefore,
these three items might be candidates to be removed
from the TQ (see next section).
e Rasch person and item location measures are dis-
played in a Wright map (Fig.2). e targeting index of
2.13 [i.e., item measure: mean (SD)=0.00 (0.97); per-
son measure=−2.13 (1.89)] exceeded the cutoff of
2.00, suggesting that the TQ has a low level of match-
ing between item difficulty and person ability [23]. is
might be because a minority of individuals experience
trypophobia, given that a previous study reported that
16.1% of adults exhibited aversion to a trypophobic image
[1]. All items showed insignificant DIFs for sex (i.e., dif-
ferentials of item measures less than 0.37, not exceeding a
cutoff of 0.50; Table1), except for the item “Have an urge
to destroy the holes,” which showed a DIF of 0.51 and
was unfit for the model (see above). Moreover, all items
showed an insignificant DIF for age, at less than 0.33.
While zero-order correlations suggested a weak correla-
tion between TQ and age in the present and previous [18]
studies, the present Rasch analysis suggested that the TQ
indeed possesses unproblematic DIF and remains stable
regardless of sex and age [23].
e person reliability of 0.86 and item reliability of 0.99
were above the criteria for sufficiency of 0.80 and 0.90,
respectively [21], suggesting that the TQ has sufficient
reproducibility of respondent classification and item hier-
archy. Furthermore, the Cronbach’s alpha of 0.95 was suf-
ficiently high and comparable to the previous studies [2, 9,
11, 18], demonstrating good internal consistency of the TQ.
In sum, the TQ possesses a unidimensional structure
with ordered response categories measuring a single con-
struct (i.e., proneness to trypophobia) and has sufficient
reproducibility, although the targeting leaves room for
improvement. Nevertheless, three items did not fit well to
the unidimensional structure. A shortened version with-
out these items might improve psychometric properties.
Follow-up withoutunt items
We performed follow-up Rasch analysis without the
three unfit items (see Additional file1 for details). Infit
and outfit mean-squares for the 14-item version of TQ
were within the criterion range (infit: 0.78–1.29; outfit:
0.81–1.22), demonstrating a better fit to the model, as
expected. Response categories ordered well again. PCA
revealed that 67.1% of the raw variance was explained
by measures, but the eigenvalue of the first contrast
was 2.33, exceeding the cutoff. Nevertheless, item clus-
ters defined by the first contrast loading were highly
correlated (i.e., disattenuated correlation coefficients of
0.849–1.000), suggesting unidimensionality comparable
to the full version of the TQ. Targeting of 2.11 indicated
low quality, comparable to the full version. e DIFs
were inconsequential, as all differentials of item meas-
ures across sex and age subgroups were less than 0.46.
e person reliability of 0.87 and item reliability of 0.99
were sufficiently high, comparable to the full version.
Although psychometric properties of the shortened
TQ improved slightly, the full and shortened versions
had comparable qualities according to the criteria [21,
23]. To determine whether the TQ should be formally
shortened, further studies should compare the validity
Fig. 2 Wright’s person-item map of the Trypophobia Questionnaire.
Person locations are on the left column, item locations are on the
right. Each period represents one to four participants, and each
hash represents five. The item names on the right column were
abbreviated for brevity. M, mean; S, standard deviation from the
mean; T, two standard deviations from the mean
Page 5 of 5
Imaizumiand Tanno BMC Res Notes (2018) 11:128
of the full and shortened versions by examining behav-
iors [2, 18] and other psychological constructs (e.g.,
anxiety [2, 11], disgust sensitivity [3, 9]).
Conclusions
Rasch analysis suggested that the TQ has a unidimen-
sional structure with ordered response categories and
sufficient person and item reproducibility, although the
targeting leaves room for improvement. Although incon-
clusive, a revised TQ without unfit items might improve
its psychometric properties, but further comparative
studies and validations are required.
Limitations
Rasch-based psychometric properties of the TQ were
shown using its Japanese version and online sampling.
To generalize our findings, future studies should replicate
the results using the English version and paper-and-pen-
cil sampling.
Abbreviations
DIF: differential item functioning; PCA: principal component analysis; SD:
standard deviation; TQ: Trypophobia Questionnaire.
Authors’ contributions
SI and YT conceived the study. SI performed the survey, analyzed the data, and
drafted the manuscript. YT provided critical revisions. Both authors read and
approved the final manuscript.
Author details
1 Graduate School of Arts and Sciences, The University of Tokyo, 3-8-1 Komaba,
Meguro, Tokyo 153-8902, Japan. 2 Japan Society for the Promotion of Science,
5-3-1 Kojimachi, Chiyoda, Tokyo 102-0083, Japan.
Acknowledgements
Not applicable.
Competing interests
The authors declare that they have no competing interests.
Availability of data and materials
The dataset supporting the conclusions of this article is included within
Additional file 2.
Consent for publication
Not applicable.
Additional les
Additional le1. Supplemental results for the full and short versions of
the Trypophobia Questionnaire. This file describes factor analysis on the
full version of the Trypophobia Questionnaire, and follow-up analyses on
the shortened 14-item version of the scale, including descriptive statistics,
Rasch analysis, and factor analysis.
Additional le2. Raw data from the survey. Dataset includes demo-
graphic data and responses to the Trypophobia Questionnaire from 582
healthy native Japanese speakers who were recruited via online.
Ethics approval and consent to participate
This study was approved by the ethical committee of the Graduate School of
Arts and Sciences, The University of Tokyo (Approval Number: 468). All partici-
pants provided electronic informed consent prior to participation.
Funding
This study was supported by Grant-in-Aids for JSPS Research Fellow (16J00411)
and Young Scientists (B) (17K12701) from the Japan Society for the Promotion of
Science, who had no involvements in the study conceptualization, data collec-
tion and analysis, or manuscript preparation.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published
maps and institutional affiliations.
Received: 20 January 2018 Accepted: 10 February 2018
References
1. Cole GG, Wilkins AJ. Fear of holes. Psychol Sci . 2013;24:1980–5.
2. Le AT, Cole GG, Wilkins AJ. Assessment of trypophobia and an analysis of its
visual precipitation. Q J Exp Psychol (Hove). 2015;68:2304–22.
3. Kupfer TR, Le ATD. Disgusting clusters: trypophobia as an overgen-
eralised disease avoidance response. Cogn Emot. 2017. https ://doi.
org/10.1080/02699 931.2017.13457 21.
4. Pipitone RN, Gallegos B, Walters D. Physiological responses to trypophobic
images and further scale validity of the Trypophobia Questionnaire. Pers
Individ Dif. 2017;108:66–8.
5. Ayzenberg V, Hickey MR, Lourenco SF. Pupillometry reveals the physiological
underpinnings of the aversion to holes. PeerJ. 2018;6:e4185.
6. Skaggs W. Fear of holes. Sci Am Mind. 2014;25:12.
7. Furuno M, Imaizumi S, Maeda K, Hibino H, Koyama S. The influence of
background objects on unpleasantness evoked by lotus-seed-pods-on-the-
living-body images (“Hasu-colla”). Int J Affect Eng. 2017;16:221–30.
8. Yamada Y, Sasaki K. Involuntary protection against dermatosis: a preliminary
observation on trypophobia. BMC Res Notes. 2017;10:658.
9. Imaizumi S, Furuno M, Hibino H, Koyama S. Trypophobia is predicted by
disgust sensitivity, empathic traits, and visual discomfort. SpringerPlus.
2016;5:1449.
10. Sasaki K, Yamada Y, Kuroki D, Miura K. Trypophobic discomfort is spatial-
frequency dependent. Adv Cogn Psychol. 2017;13:224–31.
11. Chaya K, Xue Y, Uto Y, Yao Q, Yamada Y. Fear of eyes: triadic relation
among social anxiety, trypophobia, and discomfort for eye cluster. PeerJ.
2016;4:e1942.
12. van Strien JW, van der Peijl MK. Enhanced early posterior negativity in
response to trypophobic st imuli. Psychophysiolo gy. 2015;52: S90.
13. Andrich D. Rating scales and Rasch measurement. Expert Rev Pharmaco-
econ Outcomes Res. 2011;11:571–85.
14. Hobart JC, Cano SJ, Zajicek JP, Thompson AJ. Rating scales as outcome
measures for clinical trials in neurology: problems, solutions, and recom-
mendations. Lancet Neurol. 2007;6:1094–105.
15. Lancers Inc. Lancers. 2016. https ://www.lance rs.jp. Accessed 20 Jan 2018.
16. Linacre JM. Sample size and item calibration stability. Rasch Measure Trans.
1994;7:328.
17. Qualtrics LLC. Qualtrics. 2018. https ://www.qualt rics.com. Accessed 20 Jan
2018.
18. Imaizumi S, Furuno M, Hibino H, Koyama S. Development of the Japanese
version of Trypophobia Questionnaire. Jpn J Pers. 2016;25:17 1–3.
19. JASP Team: JASP (Version 0.8.5.1) [Computer software]. 2017. https ://jasp-
stats .org. Accessed 20 Jan 2018.
20. Linacre JM. Winsteps® Rasch measurement computer program. 2017.
http://www.winst eps.com. Accessed 20 Jan 2018.
21. Linacre JM. A user’s guide to WINSTEPS® MINISTEP: Rasch-model computer
programs. Beaverton: Winsteps.com; 2017.
22. Andrich D. A rating formulation for ordered response categories. Psycho-
metrika. 1978;43:561–73.
23. Khadka J, McAlinden C, Pesudovs K. Quality assessment of ophthal-
mic questionnaires: review and recommendations. Optom Vis Sci.
2013;90:720–44.
24. Wright BD, Masters GN. Rating scale analysis. Chicago: MESA Press; 1982.
1
Additional file 1: Supplemental results for the full and short versions of the Trypophobia Questionnaire
Full version
To replica te previous studies [1, 2], factor analysis using the maximum-likelihood method with promax rotation
was performed on the 17 items of the Japanese version [1] of the Trypophobia Questionnaire (TQ) [3] using
SPSS 24.0 (IBM, Armonk, New York). Data from 582 adults were analyzed. As a result, the Kaiser–Meyer–
Olkin index of sampling validity was 0.954 and Bartlett’s test for sphericity was significant (χ2(136) = 7399.78,
p < 0.001), suggesting that this analysis was valid. As expected, the first factor had an eigenvalue of 9.62 and
explained 56.56% of the variance, while the second factor had an eigenvalue of 1.26 and explained 7.41% of
the variance. Given the large difference between eigenvalues of the two factors, we interpreted the TQ as a one-
factor structure scale, consistent with previous studies [1-3]. Factor loadings ranged from 0.418 to 0.834 (Table
S1), again implying that all items loaded on a single factor, thought to be a proneness to trypophobia.
Tabl e S1. One-factor structure of the Trypophobia Questionnaire.
Full version
14-item version
Factor loading
Communality
Factor loading
Communality
Feel freaked out
0.732
0.548
0.734
0.548
Feel aversion, disgust or repulsion
0.827
0.843
0.837
0.847
Feel uncomfortable or uneasy
0.834
0.814
0.843
0.816
Feel like panicking or screaming
0.755
0.582
0.755
0.586
Feel anxious, full of dread or fearful
0.787
0.624
0.789
0.625
Feel sick or nauseous
0.808
0.685
0.806
0.696
Feel nervous (e.g., heart pounding,
butterflies in stomach, sweating,
stomachache, etc.)
0.787
0.639
0.785
0.646
Feel like going crazy
0.795
0.659
0.794
0.669
Have an urge to destroy the holes
0.418
0.200
–
–
Feel itchiness
0.634
0.413
–
–
Feel skin crawl
0.822
0.678
0.821
0.674
Have goosebumps
0.795
0.633
0.792
0.631
Feel like crying
0.493
0.384
–
–
Vomit
0.641
0.558
0.629
0.546
Get chills
0.747
0.576
0.742
0.575
Have trouble breathing
0.724
0.626
0.713
0.609
Shiver
0.720
0.649
0.711
0.642
Shortened 14-item version
For descriptive purposes, in the same manner as the full version, we analyzed responses from the same 582
2
adults to the 14-item version of the TQ, excluding unfit items “Have an urge to destroy the holes,” “Feel
itchiness,” and “Feel like crying”). For details about interpreting the results, see the Main Text.
The mean TQ score was 27.65 [standard deviation (SD) = 12.26; range 14–70; skewness (standard error) = 1.05
(0.10); kurtosis = 0.50 (0.20)]. The TQ scores in the total, male, and female samples were not normally
distributed (Ws < 0.91, ps < 0.001). While there was no sex difference in TQ score (Meanmale = 27.12, SDmale =
11.32, Meanfemale = 28.02, SDfemale = 12.90, U = 40665, p = 0.775, ρrb = -0.014), age weakly correlated with TQ
score (ρ = -0.221, p < 0.001). These (null) effects of sex and age were consistent with the full version.
Ordering of thresholds of the five category probability curves (Figure S1) and average measures for the five
categories (i.e., 1 to 5) of -3.13, -1.58, -0.47, 0.59, and 1.88, respectively, suggested that all response categories
were distinguished and evenly used, consistent with the full version.
Figure S1. Category probability curves for the 14-item version of the Trypophobia Questionnaire.
As expected, 14 items were well fitted to the unidimensional model; infit and outfit mean-squares ranged from
0.78 to 1.29 (Table S2), within the criterion range of 0.70–1.30. Similar to the full version, principal component
analysis showed that the measures explained 67.1% of the raw variance, and the eigenvalues of the unexplained
variance in the first and second contrasts were 2.33 and 1.99, respectively. These suggested that the TQ has
unidimensionality, but also other latent dimensions in the residuals. To examine potential multidimensionality
in the scale, we analyzed the disattenuated Pearson correlation between three item clusters based on each of the
first and second contrast loadings (Table S2). The coefficients were very high: 0.849–1.000 for the first contrast
and 0.980–1.000 for the second. This suggests that the item clusters defined by two latent dimensions measure
the same construct, suggesting unidimensionality. For descriptive purposes, factor analysis—the same as in the
full version—was performed on the short version. Kaiser–Meyer–Olkin index of 0.950 and significance in
Bartlett’s test (χ2(91) = 6658.18, p < 0.001) supported the validity of factor analysis. The first factor had an
eigenvalue of 8.30 and explained 59.26% of variance, while the second had an eigenvalue of 0.81 and explained
5.81% of the variance. Factor loadings for all items were above 0.629 (Table S1), again implying that all items
0.0
0.2
0.4
0.6
0.8
1.0
-10 -8 -6 -4 -2 0 2 4 6 8 10
Category probability
Measure relative to item difficulty
1: Not at all 2: Slightly 3: Moderately
4: Considerably 5: Extremely
3
loaded on a single factor. The short version also had a unidimensional one-factor structure.
Tab le S 2. The 14-item version of the Trypophobia Questionnaire and its psychometric properties. Differential
item functioning (DIF) contrasts are reported in absolute values. Zstd: z-standardized statistic.
Location
(standard
error)
Infit mean-
square
(Zstd)
Outfit
mean-
square
(Zstd)
First
contrast
loading
Second
contrast
loading
DIF
contrast
for sex
DIF
contrast
for age
Feel freaked out
-0.31 (0.06)
1.19 (2.80)
1.22 (2.70)
-0.55
-0.05
0.18
0.00
Feel aversion, disgust or
repulsion
-1.72 (0.06)
0.92 (-1.30)
0.90 (-1.50)
-0.51
0.54
0.18
0.08
Feel uncomfortable or
uneasy
-1.64 (0.06)
0.78 (-3.90)
0.84 (-2.50)
-0.55
0.37
0.12
0.10
Feel like panicking or
screaming
0.63 (0.07)
1.08 (1.10)
0.95 (-0.40)
-0.30
-0.41
0.11
0.10
Feel anxious, full of dread
or fearful
-0.34 (0.06)
0.96 (-0.60)
0.93 (-0.90)
-0.50
-0.22
0.13
0.00
Feel sick or nauseous
0.28 (0.07)
0.96 (-0.50)
0.81 (-2.20)
0.12
-0.38
0.00
0.06
Feel nervous (e.g., heart
pounding, butterflies in
stomach, sweating,
stomachache, etc.)
0.32 (0.07)
1.01 (0.20)
0.96 (-0.40)
0.07
-0.31
0.18
0.07
Feel like going crazy
0.26 (0.07)
0.90 (-1.40)
0.86 (-1.50)
0.13
-0.36
0.06
0.25
Feel skin crawl
-0.76 (0.06)
0.95 (-0.80)
0.90 (-1.50)
0.39
0.59
0.19
0.09
Have goosebumps
-0.69 (0.06)
1.07 (1.00)
1.01 (0.20)
0.59
0.48
0.08
0.09
Vomit
1.84 (0.09)
1.23 (2.30)
0.91 (-0.50)
0.35
-0.34
0.46
0.12
Get chills
-0.04 (0.06)
1.29 (4.00)
1.16 (1.80)
0.51
0.22
0.07
0.11
Have trouble breathing
0.97 (0.07)
1.12 (1.50)
0.99 (0.00)
0.18
-0.42
0.10
0.23
Shiver
1.20 (0.08)
1.10 (1.20)
0.81 (-1.50)
0.43
-0.24
0.27
0.10
Rasch person and item location measures are displayed in a person-item map (Figure S2). The targeting of 2.11
[mean item measure (SD) = 0.00 (0.98); mean person measure = -2.11 (2.09)] exceeded the cutoff of 2.00,
suggesting that the short version had a low quality of matching between item difficulty and person ability,
comparable to the full version. All items showed inconsequential levels of differential item functioning for sex
and age (i.e., differentials of item measures less than 0.46 below the cutoff of 0.50; Table S2). The person
reliability of 0.87 and item reliability of 0.99 were above the criteria for sufficiency of 0.80 and 0.90,
respectively. The classical internal consistency index Cronbach’s alpha was 0.95 and sufficiently high. The short
version of the TQ showed sufficient reliability, comparable to the full version
4
Figure S2. Person-item map of the 14-item version of the Trypophobia Questionnaire. Each period represents
one to four participants, and each hash represents five. The item names on the right column are abbreviated for
brevity. M: mean; S: standard deviation from the mean; T: two standard deviations from the mean.
Supplemental references
1. Imaizumi S, Furuno M, Hibino H, Koyama S. Development of the Japanese version of Trypophobia
Questionnaire. Jpn J Pers. 2016;25:171-3.
2. Chaya K, Xue Y, Uto Y, Yao Q, Yamada Y. Fear of eyes: triadic relation among social anxiety, trypophobia,
and discomfort for eye cluster. PeerJ. 2016;4:e1942.
3. Le AT, Cole GG, Wilkins AJ. Assessment of trypophobia and an analysis of its visual precipitation. Q J
Exp Psychol (Hove). 2015;68:2304-22.
Measure Person - Map - Item
More Rare
5
4
3
2
Vomit
Shiver
1 #Have trouble breathing
Feel like panicking
Feel sick; Feel nervous; Feel like going crazy
0# Get chills
Feel freaked out; Feel anxious
Have goosebumps
Feel skin crawl
-1 #
Feel aversion; Feel uncomfortable
-2
-3
#
-4
-5
Less Frequent
