Journal of Autism and Developmental Disorders (2020) 50:12–29

https://doi.org/10.1007/s10803-019-04222-7

ORIGINAL PAPER

The Relation Between Autistic Traits, the Degree of Synaesthesia,

and Local/Global Visual Perception
Floor Burghoorn1 · Mark Dingemanse1,2,3 · Rob van Lier1 · Tessa M. van Leeuwen1

Published online: 14 September 2019

© The Author(s) 2019

Abstract
Synaesthesia is highly prevalent in autism spectrum disorder. We assessed the relation between the degree of autistic traits
(Autism Spectrum Quotient, AQ) and the degree of synaesthesia in a neurotypical population, and hypothesized both are
related to a local bias in visual perception. A positive correlation between total AQ scores and the degree of synaesthesia was
found, extending previous studies in clinical populations. Consistent with our hypothesis, AQ-attention to detail scores were
related to increased performance on an Embedded Figures Task and reduced susceptibility to visual illusions. We found no
relation between autistic traits and performance on a motion coherence task, and no relation between synaesthesia and local
visual perception. Possibly, this relation is reserved for supra-threshold synaesthetes.

Keywords Autism · Synaesthesia · Synesthesia · Neurotypical · Visual perception · Local/global

Introduction another neurodevelopmental condition. Synaesthesia, which

can be translated from Greek as ‘joined perception’, is char-
Until quite recently, autism spectrum disorder (ASD) was acterized by altered perceptual experiences: perceiving an
mostly known as a neurodevelopmental condition caus- inducing stimulus elicits an unusual concurrent sensation in
ing social dysfunctions (American Psychiatric Association the same or a different modality, e.g. letters automatically
2013). Even though sensory atypicalities in ASD had already evoke a colour. Whereas the general prevalence of synaes-
been acknowledged by Kanner (1943), the importance of thesia only lies around ~ 4% (Simner et al. 2006), approxi-
sensory and perceptual alterations in ASD has been empha- mately 20% of people with ASD have synaesthesia (Baron-
sized only recently (e.g. Simmons et al. 2009). This has led Cohen et al. 2013; Neufeld et al. 2013). No studies have yet
to the inclusion of sensory dysregulation, i.e. hypersensi- assessed the ASD prevalence in the synaesthesia population,
tivity or hyposensitivity to sensory input, as a diagnostic and hence the strength and full extent of the relationship are
criterion for ASD (American Psychiatric Association 2013). not yet known. However, the relatively high synaesthesia
A recent finding related to sensory atypicalities in autism prevalence in the ASD population and increased interest in
is a frequent co-occurrence of ASD with synaesthesia, sensory atypicalities in ASD have inspired studies inves-
tigating possible commonalities between synaesthesia and
Electronic supplementary material The online version of this autism.
article (https​://doi.org/10.1007/s1080​3-019-04222​-7) contains One commonality is that synaesthesia and ASD are both
supplementary material, which is available to authorized users. considered to be dimensional rather than categorical con-
structs, although for both conditions this issue remains con-
* Tessa M. van Leeuwen
[email protected] tinuously under debate. Over the last decades, an increasing
amount of evidence has been collected that suggests ASD
1
Donders Institute for Brain, Cognition and Behaviour, to fall on a continuum, ranging from a low to high degree
Radboud University, Montessorilaan 3, 6525 HR Nijmegen, of autistic traits (Baron-Cohen et al. 2001; De Groot and
The Netherlands
Van Strien 2017; Hoekstra et al. 2008). The quantitative
2
Max Planck Institute for Psycholinguistics, Nijmegen, nature of ASD has been termed the broader autism pheno-
The Netherlands
type (BAP). Similarly, several researchers have theorized
3
Centre for Language Studies, Radboud University, Nijmegen, synaesthesia to be of a dimensional nature, with quantitative
The Netherlands

13
Vol:.(1234567890)
Journal of Autism and Developmental Disorders (2020) 50:12–29 13

differences existing between individuals (Cohen Kadosh and synaesthetes outperformed controls on two tests requiring
Henik 2007). In line with this theory, studies have found attention to detail, extending the perceptual commonalities
evidence for a continuum of the strength of pitch-size (Bien between ASD and synaesthesia from the phenomenal (self-
et al. 2012) and color-vowel (Cuskley et al. 2019) associa- report) level to functioning at cognitive tasks.
tions in the non-synaesthetic population, suggesting that The goal of the present study is to shed more light on the
individual differences exist in how sensitive people are to cognitive functioning of individuals with ASD and/or syn-
commonalities in the combinations of sensory information aesthesia. Building on prior work, we investigate the rela-
in their environment. Other researchers, however, argue that tion between the degree of synaesthesia, autistic traits, and
there is a clear qualitative distinction between those who are performance on local/global visual perception tasks. Before
synaesthetes, and those who are not (Ward 2019). introducing the present research in more detail, we discuss
Since altered perception lies at the heart of synaesthesia, studies that have investigated local/global visual process-
it has been suggested that synaesthesia and ASD share per- ing separately in either ASD or synaesthesia. In addition to
ceptual and sensory characteristics (e.g. Ward et al. 2017). A elucidating perception in both conditions, these studies also
better understanding of similarities in perceptual functioning point to potential underlying neural mechanisms that could
in ASD and synaesthesia may lead to novel insights into per- drive a shared bias in local/global visual perception.
ceptual/sensory dysregulation in individuals with ASD. This
may also have practical implications by aiding individuals
with ASD (and their clinicians) to better understand and Evidence for a Local Bias from the ASD
cope with their perceptual/sensory symptoms. and Synaesthesia Research Fields
Several studies have directly compared sensory percep-
tion in synaesthesia and ASD. One aspect of sensory per- Various different tasks and paradigms have been used to
ception is sensory sensitivity, which is frequently atypical study local/global visual perception in ASD. First, individu-
in individuals with ASD, e.g. characterized by hyper- and/ als with ASD tend to perform superior to controls on the
or hyposensitivity to sounds or touch (for a review, see Embedded Figures Task (EFT), a task that requires partici-
Schauder and Bennetto 2016). More recently, similar pat- pants to detect a target shape embedded in a complex context
terns of increased and decreased sensory sensitivity (meas- (e.g. Pellicano et al. 2006; Brosnan et al. 2012). Focusing
ured with the Glasgow Sensory Questionnaire) were found on individual elements facilitates performance on this task.
in both ASD and synaesthesia (Ward et al. 2017, 2018; Second, a local bias is related to reduced susceptibility to
Van Leeuwen et al. 2019), revealing particularly strong visual illusions. To illustrate, the Ebbinghaus illusion con-
hyper- and hyposensitivity to auditory stimulation in both sists of one ‘core’ circle that is surrounded by other (smaller
conditions. or larger) contextual circles. Integrating the contextual cir-
Another aspect of perception that is of particular inter- cles with the core stimulus influences the perceived size of
est for the synaesthesia-ASD co-occurrence revolves around the latter. Focusing on the core circle (local bias) predicts a
local/global visual processing. Navon (1977) was the first reduced susceptibility to the illusion, which was indeed con-
to propose that whereas the majority of individuals have firmed for ASD (Happé 1996, Bölte et al. (2007). Chouinard
an inclination towards global processing (i.e. ‘global prec- et al. (2013) found AQ scores in neurotypicals to be nega-
edence’, extracting the general gist of an image), some indi- tively related to susceptibility to the Müller-Lyer illusion.
viduals are biased towards local processing. These individu- Not all recent studies were able to replicate these reports of
als might thus be prone to ‘seeing the trees before the forest’. reduced susceptibility in ASD, however (Chouinard et al.
Local/global visual processing is a widely studied topic 2016, Manning et al. 2017), calling for more studies to be
in ASD research, as many studies point to a bias towards performed in this area.
local processing in individuals with ASD (for a review, see Third, individuals with ASD showed impaired perfor-
Happé and Frith 2006). In 1994, Frith and Happé proposed mance compared to controls on a motion coherence task
the ‘weak central coherence theory’, which, in its most (MCT), which requires identification of the global motion
recent form, states that individuals with ASD have a ‘detail- direction of a group of moving dots. Individuals with ASD
focussed cognitive style’ (a local bias). were found to have an, on average, 10% higher motion
In line with the findings for autism, synaesthetes have coherence threshold than controls in this task, indicating
scored higher than controls on the Attention-to-detail sub- global processing deficits (for a review, see Dakin and
scale of the Autism Spectrum Quotient (AQ) (Mealor et al. Frith 2005). Jackson et al. (2013) found that this effect
2016; Van Leeuwen et al. 2019; Ward et al. 2017, 2018), was strongest when the ability to track single, local dots is
a self-report questionnaire on autistic traits. This finding limited and one is therefore forced to assess global move-
suggests a shared bias in local visual perception between ment. Finally, when individuals with ASD are presented
ASD and synaesthesia. Ward et al. (2018) also showed that with hierarchical stimuli (Navon 1977) they tend to attend

13
14 Journal of Autism and Developmental Disorders (2020) 50:12–29

to the local rather than the global level of these stimuli global perceptual characteristics are caused by different
(Koldewyn et al. 2013; Muth et al. 2014). neural mechanisms in synaesthesia and ASD, respectively.
A meta-analysis by Cribb et al. (2016) showed that the
results pointing towards a local bias also hold for individu- Purpose of the Current Research
als who possess a high degree of autistic traits (i.e. neu-
rotypicals). This supports the idea of autism as a dimen- The present research has three goals. First, we examine the
sional (as opposed to categorical) construct, with autistic relation between the degree of autistic traits and synaesthe-
traits distributed across a spectrum. sia in neurotypicals, attempting to extend on prior studies
However, for both clinical and subclinical samples, that showed synaesthesia to be more common in individuals
results regarding a local bias in ASD are not fully con- with diagnosed ASD. Second, we seek to reduce the incon-
clusive, with several studies failing to support it (e.g. sistency of evidence regarding the weak central coherence
Chouinard et al. 2016; Manning et al. 2017 for the visual theory by studying the relationship between local/global
illusions) and a large heterogeneity appearing in meta- processing and the degree of autistic traits. Finally, the rela-
analyses (Cribb et al. 2016; Muth et al. 2014; Van der Hal- tionship between synaesthesia and local/global processing
len et al. 2015). These discrepancies call for more studies abilities is examined in more detail.
on the relation between ASD and local bias. Two features of the current study distinguish it from pre-
In synaesthesia only a few studies have investigated vious research in this field. First, the study is performed
local/global perception, but these do suggest that synaes- in a neurotypical population (i.e. in a population of indi-
thetes exhibit a local bias similar to individuals with ASD. viduals not classified as having synaesthesia or ASD), using
First, Ward et al. (2018) found synaesthetes to outperform continuous measures of the degree of synaesthesia and of
controls on the EFT. Synaesthetes were also more accurate autistic traits and thereby treating ASD and synaesthesia as
than controls on a Change Blindness Test, which requires dimensional constructs (Baron-Cohen et al. 2001; Cohen
detecting small changes in a visual environment. Second, Kadosh and Henik 2007; Cuskley et al. 2019; De Groot and
Janik McErlean et al. (2016) found that although synaes- Van Strien 2017). Second, the current study is the first to
thetes outperformed non-synaesthetes in facial identity specifically assess local/global perception in relation to both
tasks that required featural (local) discriminations of autistic traits and synaesthesia in the same study population.
facial features, they did not on tasks that require configu-
ral (global) face processing. Finally, Banissy et al. (2013) Study Approach and Hypotheses
found grapheme-colour synaesthetes to have increased
motion coherence thresholds in the MCT, suggesting a The degree of autistic traits was measured by the Autism
possible global processing deficit. So far, however, only Spectrum Quotient (AQ; Baron-Cohen et al. 2001). The
these three studies have addressed local/global processing degree of grapheme-colour synaesthesia was measured by
in synaesthesia; no studies have addressed the suscepti- an extensive grapheme-colour synaesthesia consistency
bility to visual illusions in synaesthetes. More research test (Eagleman et al. 2007). Given our focus on local/global
is needed to replicate these findings, and to extend their visual perception, the Attention to detail-subscale of the AQ
results to other local/global processing tasks. (AQ-detail) was of particular interest. We hypothesized that
In addition to commonalities in performance on local/ synaesthesia scores and AQ-detail scores would correlate
global perceptual tasks, studies into ASD and synaesthesia positively, given previous findings (Mealor et al. 2016; Van
have found neural similarities related to visual perception, Leeuwen et al. 2019; Ward et al. 2017, 2018). In addition,
such as enhanced sensitivity of the parvocellular visual we expected synaesthesia scores to be positively correlated
pathway (sensitive to fine detail and high contrast) (see with AQ-total scores. This latter correlation was investigated
Brown and Crewther 2017; Sutherland and Crewther 2010; to gain insight into the relation between the overall AQ-
Jackson et al. 2013 for ASD, and Barnett et al. 2008; Van construct (formed of several subscales) and synaesthesia
Leeuwen et al. 2013 for synaesthesia), as well as increased scores. A second reason for including AQ-total in our analy-
local and decreased global cortical connectivity (see Just ses is that although previous research has found the relation
et al. 2012 for ASD, and Hänggi et al. 2011, for synaes- between synaesthesia and ASD to be strongest for the Atten-
thesia). These studies provide suggestions for the neu- tion to detail-subscale, synaesthetes have also been found to
ral mechanisms that might underlie a shared local bias. obtain elevated scores on the remaining four subscales (Van
However, it should be acknowledged that a direct causal Leeuwen et al. 2019; Ward et al. 2017), especially when
relation between the shared neural mechanisms and per- these were added together into an ‘AQ-other’ score (Ward
formance on perceptual task has not been established so et al. 2018).
far. Hence, it is possible that the shared sensory and local/ Three experiments were devised to measure local/global
visual perception. Experiment 1 was a motion coherence

13
Journal of Autism and Developmental Disorders (2020) 50:12–29 15

task, requiring participants to detect the global motion direc- between the degree of synaesthesia/autistic traits and visual
tion of a group of moving dots (MCT; Newsome and Paré perception, such as general increased/decreased percep-
1988). This task can be generally performed using one of tual functioning, or an increased/decreased motivation for
two strategies. With a ‘global strategy’, one focuses on the accuracy. We predicted that only in the context condition,
global movement of all dots together. With a ‘local strategy’, a higher degree of synaesthesia and autistic traits (AQ-total
however, one picks a single dot and tracks this to identify its and AQ-detail) would be related to a smaller discrepancy
direction. For the present study, we created two task condi- between the to-be-adjusted stimulus and the reference stimu-
tions, one with a limited (60 ms) and one with an unlimited lus, indicating a decreased susceptibility to visual illusions.
(600 ms) dot lifetime. In the limited dot lifetime condition, In the context-free trials, we expected no relation between
participants are forced to use the global strategy, as the dot the degree of synaesthesia/autistic traits and performance.
lifetime is too short to track single dots. Because of the
hypothesized local bias, we expect a higher degree of autistic
traits and synaesthesia to be related to impaired performance Methods
in this condition. In contrast, in the unlimited dot lifetime
condition, participants can use a local strategy. In line with Participants
results from Jackson et al. (2013), we expect higher AQ/
synaesthesia scores to be related to increased performance At the time of study design, research on local/global visual
in this condition. That is, we hypothesize that because of perception in synaesthesia was minimally available, and
their local bias, participants with higher AQ/synaesthesia a large heterogeneity existed in reported effect sizes in
scores are better at tracking single dots than participants ASD (Cribb et al. 2016; Muth et al. 2014; Van der Hal-
with lower AQ/synaesthesia scores. It should be noted that len et al. 2015). Therefore, it was difficult to determine an
a local strategy does not guarantee accuracy; that is, because expected effect size for a priori sample size calculations.
only a subset of the dots are moving in the same direction, Using a medium effect size (r = .40 for correlations and f =
there is a chance of selecting and tracking the ‘wrong’ dot. .25 for repeated measures ANOVAs; Cohen 1988) in these
However, assuming that people with both low and high AQ/ calculations (G*Power; Faul et al. 2007), we included 39
synaesthesia scores use the local strategy in the unlimited participants to aim for a power of 0.80. These participants
dot lifetime condition, this chance of picking the wrong dot (of which N = 26 university students) took part in the study
is the same for everyone. after providing informed written consent. Three participants
In the second experiment, an Embedded Figures Task were excluded due to neurological/psychiatric disorders,
(EFT; Witkin et al. 1971) was used to assess local visual leaving 36 participants for analyses (10 males, mean age
perception. A small target shape needed to be identified 25.47, SD = 9.07). Participants were recruited via an online
within a complex background figure, requiring focus on study participation website and flyers around the university
local elements while ignoring the background context. It campus. Participation was compensated with 12.50 euros or
was expected that both an increased degree of autistic traits 1.5 course credits. Informed consent was obtained from all
(AQ-detail and AQ-total) and an increased degree of synaes- individual participants included in the study. The study was
thesia would be related to superior performance (decreased approved by the Ethics Committee of the Faculty of Social
reaction times and/or a lower error percentage) on this task. Sciences (ECSW) at the Radboud University Nijmegen.
In Experiment 3, we assessed the susceptibility to visual
illusions using a method-of-adjustment task devised by Man- Degree of Autistic Traits
ning et al. (2017). Participants adjusted the size of a stimulus
(either the Ebbinghaus or Müller-Lyer illusion, Fig. 3) to The degree of autistic traits was measured by the Dutch ver-
make it match a reference stimulus. This method is sensi- sion of the Autism Spectrum Quotient (AQ-NL; Hoekstra
tive to the extent to which the illusion is being perceived et al. 2008), a self-report questionnaire (Baron-Cohen et al.
and therefore gives a graded indication of the susceptibility 2001) providing a continuous measure of autistic traits in
to the illusion, contrary to a same/different judgment task adults with normal intelligence. It consists of 50 statements
(Manning et al. 2017). In addition to the main condition concerning personal habits and preferences that can be
containing illusory stimuli, participants completed a control agreed or disagreed with on a 4-point Likert scale (definitely
condition in which they had to adjust context-free stimuli agree, slightly agree, slightly disagree, and definitely disa-
(e.g. simple circles). It is assumed that only in the illusory gree): e.g. item 13 states ‘I would rather go to a library than
context condition, a local bias would aid performance, as to a party’. Subscores are available for five subscales: Social
this facilitates separating the illusory context from the ‘core’ skills, Communication, Imagination, Attention to detail, and
of the stimulus. Using both types of conditions allowed us Attention switching. The total AQ score is the sum of the
to control for potentially confounding factors of the relation subscores; higher scores indicate a higher degree of autistic

13
16 Journal of Autism and Developmental Disorders (2020) 50:12–29

traits. For the AQ-NL, total AQ scores lie between 50 and whether or not they experienced any form of synaesthesia
200; each subscore has a potential minimum of 10 and a in daily life (see Supplementary Material). Participants who
maximum of 50. Since we only predicted that AQ-total and scored below 1.43 on the test and indicated experiencing
AQ-detail would relate to a higher degree of synaesthesia synaesthesia in daily life were subsequently excluded from
and a bias towards local visual perception we focussed on all analyses (N = 3). One other participant obtained a score
these two scores. within the synaesthetic range (1.27), but because this indi-
The AQ-NL has been validated in a Dutch population by vidual did not report any synaesthetic experiences they were
Hoekstra et al. (2008), evaluating its criterion validity, inter- retained in the sample.
nal consistency reliability (α = .81) and test–retest reliability Additionally, 4 of the 33 remaining participants failed
(α = .78) to be satisfactory. Since none of the 36 participants to complete the synaesthesia test correctly (i.e. they did not
had been diagnosed with ASD, and the AQ cannot be used to pick any associating colours, or chose black or white for all
diagnose people but only indicates a person’s position on the items) and no synaesthesia score could be determined. These
autism spectrum scale, no participants were excluded based 4 participants were excluded from analyses concerned with
on their scores. Participants completed the AQ in LimeSur- the degree of synaesthesia. Therefore, all analyses related to
vey (https​://www.limes​urvey​.org/), which was run on the the degree of synaesthesia were performed with N = 29. The
laboratory computer. 4 participants who failed to complete the synaesthesia test
did complete the AQ, which resulted in a sample of N = 33
Degree of Synaesthesia participants for the analyses concerning autistic traits and
visual perception. For these reasons, it was decided that for
The extent to which participants experienced grapheme- each experiment, two separate analyses would be run: one
colour synaesthesia was assessed with an online grapheme- with the AQ scores (in which N = 33), and one with the syn-
colour synaesthesia test similar to the Synesthesia Battery aesthesia scores (N = 29).
(Eagleman et al. 2007). The test was developed for a ‘Groot
Nationaal Onderzoek’ (Large National Survey), a crowd- General Procedure
sourcing initiative in the Netherlands (http://gno.mpi.nl/
test/synes​thesi​etest​; Van Leeuwen and Dingemanse 2016). The experiment consisted of one 75-min laboratory session,
Participants were serially presented with the numbers 0–9 in which all tests and questionnaires were administered.
and letters A–Z in random order, and were asked to indicate After providing informed consent, participants received
on a colour spectrum with which colour they associated this instructions on the general procedure. The complete experi-
number or letter. All items were presented three times. Dif- ment was presented on a 24″ BenQ screen with display reso-
ferences in RGB value between the three instances of each lution set to 1920 × 1080, controlled by a Dell laboratory
grapheme were used to compute a difference score that indi- computer running Windows 7. Distance to the computer
cates the consistency of the associated colour experience. screen was 50 cm. Participants first performed the motion
Difference score values below 1.43 are considered to signal coherence task, followed by the Embedded Figures Task
consistent synaesthetic color experiences, a cut-off that was and the visual illusions task. After these three tasks were
established in a test of specificity and sensitivity (Rothen performed, the synaesthesia test and AQ questionnaire were
et al. 2013) of the widely used Synaesthesia Battery (Eagle- completed immediately after each other. At the end of the
man et al. 2007), which is generally accepted as a classifi- experiment, participants were debriefed on the research’s
cation tool for synaesthesia. Lower scores thus indicate a purpose and hypothesis and thanked for their participation.
higher degree of synaesthesia.
Since this study investigated visual perception in a popu- Experiment 1: Motion Coherence Task (MCT)
lation of neurotypical individuals, we excluded participants
classified as synaesthetes from further analyses. This pre- Apparatus
vented the potential association between synaesthesia scores,
AQ scores and visual perception to be potentially inflated by Participants performed an online motion coherence task on
a few influential scores at the higher end of the continuum the laboratory computer. This test was developed in HTML
(caused by synaesthetes), while the majority of the scores 5 (HTML/CSS/JavaScript), accessible via a URL specifying
might be at a lower level. The cut-off for synaesthesia was the task parameters, and run in Google Chrome.
at 1.43, in line with Rothen et al. (2013). To ensure that
achievement of this score was due to the actual conscious
experience of synaesthesia (rather than, for instance, a mere
high memory performance), all participants completed
a short post-test questionnaire on which they indicated

13
Journal of Autism and Developmental Disorders (2020) 50:12–29 17

Table 1  Descriptive statistics

Measure M SD Range

Synaesthesia/AQ
Synaesthesia 2.26 0.44 1.27–3.15
AQ-total 107.39 11.67 89–136
AQ-detail 25.15 3.83 18–34
MCTa
60 ms 0.28 0.18 0.07–0.77
600 ms 0.08 0.03 0.03–0.17
EFTb
% errors (total) 12.54 6.74 0–26.56
DL 1 3.22 4.97 0–18.75
DL 2 3.41 5.87 0–25.00
DL 3 14.02 12.00 0–43.75
DL 4 29.73 15.15 0–56.25
RT (s) (total) 3.69 2.99 0.80–18.16
DL 1 1.88 0.63 0.80–18.16
DL 2 2.50 1.18 0.93–6.14
DL 3 4.04 2.58 1.08–12.21
Fig. 1  Schematic overview of the MCT. A subset of the presented DL 4 6.30 3.92 0.98–18.16
dots moved in a coherent direction; this direction had to be identified Visual ­illusionsc
by the participant
Ebb-NC 1.72 0.52 0.70–2.89
Ebb-C 5.38 1.92 2.20–10.45
Stimuli ML-NC 4.94 1.18 2.10–8.25
ML-C 21.04 10.78 8.15–48.46
The visual display consisted of 200 white dots (diameter a
Motion coherence task: 60 ms = limited dot lifetime condition,
0.15°) that moved with a speed of 6°/s across a 11.7 by 600 ms = unlimited dot lifetime condition
11.7 cm grey square background (see Fig. 1). A subset b
Embedded Figures Task: DL = difficulty level; RT = response times
of the dots moved in a coherent direction, either right- (s)
c
wards (0°), downwards (90°), leftwards (180°) or upwards Visual illusions task: Ebb-NC = Ebbinghaus illusion-no context;
(270°). Two different dot lifetimes conditions were cre- Ebb-C = Ebbinghaus illusion-context; ML-NC = Müller-Lyer illusion-
no context; ML-C = Müller-Lyer illusion-context
ated. In the limited dot lifetime condition individual dots
could be tracked to a limited extent (60 ms) (see Simmons
et al. 2009, Table 1, for an overview of MCT parameters three runs of 60 staircase trials each were completed. Each
in ASD studies). In the unlimited dot lifetime condition, participant thus completed both conditions (2 × 180 trials).
individual dots could be tracked during the full duration The order of dot lifetime conditions was counterbalanced
of the trial (600 ms). across participants.

Data‑Analysis
Procedure
For both dot lifetime conditions (60 ms and 600 ms) sepa-
At the start of each run the motion coherence level was set rately, the motion coherence thresholds at the end of each
to 0.5 (i.e. 50% of dots moving coherently). Participants run (trial 60, 120 and 180) were averaged and taken as an
used the arrow keys on the keyboard to indicate the motion indication of performance level. After 60 trials, the stair-
direction during or directly after each trial, after which case procedure has typically led to stable motion coher-
the next trial began immediately. After a correct trial the ence thresholds.
coherence level for the next trial was reduced logarith- Two repeated measures ANCOVAs (one with the
mically by dividing the current coherence level by ­100.1 AQ and one with the synaesthesia scores) were used to
(power bel function), while after an incorrect answer the analyse the main and interaction effects of dot lifetime
coherence level was multiplied by 1­ 00.1, making the task (60/600 ms), AQ-total and AQ-detail, and synaesthesia
easier. The minimum motion coherence value was close to consistency scores on the motion coherence threshold (the
0, the maximum was 1.0. For each dot lifetime condition, dependent variable). For the interactions to be analysed,

13
18 Journal of Autism and Developmental Disorders (2020) 50:12–29

Procedure

On each trial, participants clicked—as fast and as accurate

as possible- on the complex figure in which they thought the
target shape was hidden. Feedback was given on each trial:
after a correct answer a green border appeared around the
correct figure, and an arrow appeared allowing the partici-
pants to proceed to the next trial. After an incorrect answer
a red border appeared around the selected figure, and par-
ticipants had to try again until succeeding. To complete this
session, all 64 trials had to be answered correctly.
Fig. 2  EFT visual display. The target shape is presented in the upper
part of the display. This target shape is also hidden in one of the Data Analysis
three complex figures at the bottom, in this case the right figure (at
66% continued lines, difficulty level 3) Adapted version of Fig. 2 of Error percentages and response times (RTs) were recorded
De-Wit et al. (2017) under a Creative Commons CC-BY 4.0 license
(http://creat​iveco​mmons​.org/licen​ses/by/4.0/) for each participant. The overall error percentage was cal-
culated as the percentage of the 64 trials that was answered
incorrectly on the first presentation (i.e. number of errors/64
× 100%). RTs on incorrect trials were removed. RTs that
the AQ-total, AQ-detail and synaesthesia consistency
were 1.5 interquartile ranges (IQRs) above the third quartile
scores were standardized and used as covariates in the
(75th percentile) for each difficulty level were denoted as
analysis (as recommended by Ellis 2016).
outliers (Ellis 2014), and were removed from the analysis
[4.42 trials (6.91%) on average]. The remaining response
Experiment 2: Embedded Figures Task (EFT)
times were averaged per participant per difficulty level. The
average RTs and error percentages were subsequently used
Apparatus
as dependent variables in two doubly multivariate repeated
measures ANCOVAs, with difficulty (1–4) as within-
Participants completed an online Embedded Figures Task, of
subjects factor and AQ-detail and AQ-total (in the first
which the stimuli and procedure were developed at the Uni-
ANCOVA) and synaesthesia consistency scores (in the sec-
versity of Leuven (de-Wit et al. 2017). The experiment was
ond ANCOVA) as covariates. Difficulty was incorporated in
programmed in HTML 5 (HTML/CSS/JavaScript), acces-
the analyses for exploratory purposes; to investigate whether
sible via a URL specifying the task parameters, and run on
the hypothesized main effects of AQ-detail, AQ-total and
Google Chrome.
synaesthesia scores are dependent on the level of difficulty.
The covariates were standardized to analyse the interaction
Stimuli
between the covariates and the within-subjects factor (Ellis
2016).
Each stimulus display consisted of a target shape at the
top and three complex figures at the bottom of the display
(Fig. 2). The general background was grey; the target shape
Experiment 3: Visual Illusions Task
and complex figures consisted of grey lines presented on
Apparatus
white squares. The target shape was hidden in one of the
three (left, middle or right) complex figures. The stimulus
The experiment was run in Matlab Psychtoolbox (Matlab
set consisted of 16 target shapes with their corresponding
R2015b 64-bit, Psychtoolbox 3.0.11, MathWorks). The
complex figures and each target shape was presented four
method-of-adjustment task used was developed by Manning
times, at four different levels of difficulty. The difficulty
et al. (2017) for studying the susceptibility to visual illusions
manipulation was achieved by varying the number of target
in autistic children. The task was adapted for the current
shape lines that continued into the background context of
research (e.g. number of trials, procedure).
the complex figure. (0%, 34%, 64%, and 100% on average;
for details see de-Wit et al. 2017). Thus, 16 trials of each
Stimuli
difficulty level were presented in the experiment resulting in
a total of 64 randomly presented experimental trials. Nine
The experiment consisted of two different tasks: an Ebbing-
practice stimuli were used to familiarise the participants
haus task (Fig. 3a, b) and a Müller-Lyer task (Fig. 3c, d). In
with the task.
both tasks, the visual display consisted of two side-by-side

13
Journal of Autism and Developmental Disorders (2020) 50:12–29 19

a b

c d

Fig. 3  Visual illusion task visual display with stimulus examples. other line. The to-be-adjusted stimuli could occur on either side of the
Both the Ebbinghaus and Müller-Lyer stimuli were presented with- display. In this figure, all to-be-adjusted stimuli are identical in size
out (a, c) or with (b, d) surrounding context. In the Ebbinghaus task to the reference stimulus. Adjusted from Figs. 1 and 2 from Manning
(a, b), participants matched the size of the to-be-adjusted inner (b) et al. (2017) permitted under a Creative Commons CC-BY 4.0 license
or single (a) circles to the other circle. In the Müller-Lyer task, the (http://creat​iveco​mmons​.org/licen​ses/by/4.0/)
to-be-adjusted inner (d) or single (c) lines had to be matched to the

white stimuli (RGB level 250): either two Ebbinghaus stim- Performance was defined as the physical discrepancy in pix-
uli or two Müller-Lyer stimuli, on a grey (RGB level 70) els between the to-be-adjusted and reference stimulus, with a
background. One of the two stimuli was the reference stimu- smaller discrepancy indicating higher performance and less
lus, the other was the to-be-adjusted stimulus. In the Ebb- susceptibility to the illusions.
inghaus task, the central circle of the reference stimulus had Each participant performed four task sessions: two ses-
a fixed diameter of 1.25°. This central circle was either sur- sions of the Ebbinghaus task and two sessions of the Müller-
rounded by eight small context circles (diameter .42°) that Lyer task, in counterbalanced order. Each session consisted
had a distance of 1.25° from the centre stimulus, or by four of one practice trial and 20 experimental trials. In the prac-
large context circles (diameter 1.67°) that had a distance of tice trials participants adjusted the size of a yellow star to
2.08° from the centre stimulus. The starting diameter of the make it match a reference star to familiarise themselves with
central circle of the to-be-adjusted stimulus randomly varied the task. Of the 20 experimental trials per session 10 trials
between .68° and 1.82°. In the Müller-Lyer task, the refer- were context-free and 10 trials had context. The context- and
ence stimulus consisted of a horizontal line (length 3°) with context-free trials were blocked, and across the two sessions
inward or outward context arrows at the end (at an angle of per task, it was counterbalanced whether participants started
45° compared to the horizontal line). The starting length of with 10 context-trials or 10 context-free trials. Within each
the to-be-adjusted stimulus randomly varied between 2.43° 10 trials trial order was randomised. In the context-free tri-
and 3.86°. The location of the reference and to-be-adjusted als two context-free circles or lines were presented; in the
stimulus (left/right) was varied across trials. For both the context-trials, two circles or lines with context were pre-
Ebbinghaus and Müller-Lyer tasks, control (context-free) sented. In half of the trials per session, small circles (in the
stimuli were created (Fig. 3a, c), which were identical to Ebbinghaus task) or inward arrows (in the Müller-Layer
the experimental stimuli except that they had no contextual task) appeared on the left side of the screen and large circles
circles or arrows. (Ebbinghaus) or outward arrows (Müller-Lyer) appeared on
the right side of the screen; in the other half, it was reversed.
Procedure
Data Analysis
At the start of each trial a small green rectangle (1000 ms)
cued participants which of the two subsequently presented Preprocessing and data-analysis procedures were identical
stimuli should be adjusted. Participants adjusted the size of for the Ebbinghaus and Müller-Lyer task. First, performance
the to-be-adjusted stimulus (using the up and down arrow on each trial was calculated by comparing the size of the to-
keys) so that its size matched the reference stimulus and be-adjusted stimulus with the size of the reference stimulus.
pressed the space bar to proceed to the next trial. There Performance scores that were 1.5 interquartile ranges (IQRs)
was no time limit and reaction times were not measured. above the third quartile (75th percentile) were denoted as

13
20 Journal of Autism and Developmental Disorders (2020) 50:12–29

outliers (Ellis 2014), and were removed from the analysis

(2.79 trials on average, i.e. 3.49%). Next, for each type of 130

illusion separately, performance scores on all trials were

averaged into one context-free score and one context score.
This resulted in four measures per participant: Ebbinghaus- 120
no context, Ebbinghaus-context, Müller-Lyer-no context,
Müller-Lyer-context. These measures were used as depend-

AQ−total score
ent variables in a repeated measures ANCOVA, with context
(present/absent) and type of illusion (Ebbinghaus/Müller- 110

Lyer) as within-subjects factors and AQ-total and AQ-detail

(in the first analysis) and synaesthesia consistency scores
(in the second analysis) as covariates. For the interactions 100
between AQ-total by context, AQ-detail by context and
synaesthesia by context to be analysed, standardized values
(z-scores) of AQ-total, AQ-detail and synaesthesia scores
were used as covariates in the analysis (Ellis 2016). 90

1.5 2.0 2.5 3.0

Synaesthesia consistency score
Results
Fig. 4  Correlation between the degree of autistic traits (the AQ-total
Due to the separation of the analyses concerning the degree score) and the degree of synaesthesia. Pearson’s r = − .38. Note: lower
synaesthesia consistency scores indicate a higher degree of synaes-
of autistic traits and synaesthesia (see methods section), the thesia
general task performance analyses (i.e. the effects of dot
lifetime, type of illusion, difficulty) were performed twice.
There were no differences between these results in terms of (2013). For this reason, and because analysing AQ-detail
significant/non-significant effects; the only difference was and AQ-total separately would increase the probability of a
that test statistics tended to be slightly smaller in the syn- Type-I error, it was decided to retain both variables in the
aesthesia analyses due to the smaller sample size. To avoid analyses.
unnecessary repetition, it was decided to only discuss the Consistent with the participants’ self-reports, no AQ
general task performance results as they were found in the scores of 145 and above—the ASD cut-off score used by
AQ analyses. The descriptive statistics of all variables can Hoekstra et al. (2008)—were obtained in our sample. Syn-
be found in Table 1. All figures were created using the pack- aesthesia scores for all but one participant (see Methods)
age ggplot2 (Wickham 2016) in R (R Core Team 2018) and were within the non-synaesthetic range (M = 2.26), above
Rstudio (R Studio Team 2016). the cut-off for synaesthesia (score of 1.43).

The Degree of Autistic Traits and Synaesthesia AQ‑Synaesthesia Relation

For descriptives, see Table 1. The AQ-total and AQ-detail The degree of synaesthesia and AQ-total correlated signifi-
scores were significantly correlated, r(31)= .50, p = .006, cantly, r(27)= − .38, p = .044, 95% CI [− 0.63, − 0.05], see
95% CI [0.12, 0.77]. This is not surprising, since AQ-detail Fig. 4 (note the correlation is negative because lower syn-
forms a subscore of the AQ-total. Using two related vari- aesthesia consistency scores indicate a higher degree of syn-
ables in the same analysis (when analysing the perception aesthesia). This correlation indicates that a higher degree of
experiments) could lead to collinearity issues. This could synaesthesia is related to a higher degree of overall autistic
reduce power, increase standard errors and potentially cause traits. There was no significant correlation between synaes-
unstable estimates in the ANCOVAs (Baguley 2012; Dor- thesia and AQ-detail scores, r(27)= − .23 p = .23, 95% CI
mann et al. 2013; Yoo et al. 2014). To examine this, we [− 0.57, 0.09].
computed the variance inflation factor (VIF), a widely used
measure of collinearity. The VIF was 1.28, amply below Experiment 1: Motion Coherence Task (MCT)
the critical value of 10 that is recommended by Dormann
et al. (2013) and Yoo et al. (2014). In addition, we computed General Task Performance
the tolerance, condition index (CI) and variance decomposi-
tion proportions (VD); none of these measure of collinear- The repeated measures ANCOVA showed a significant main
ity exceeded the critical value described by Dormann et al. effect of dot lifetime, F(1,30) = 43.57, p < .001, η2 = .59. The

13
Journal of Autism and Developmental Disorders (2020) 50:12–29 21

mean motion coherence threshold was higher in the lim-

ited (60 ms) dot lifetime condition than in the unlimited
(600 ms) dot lifetime condition (Table 1). This reflects a
general increase in performance with increasing dot lifetime. 4000

EFT response time (ms)

Relation Between AQ Scores and Performance

There were no significant relations between AQ-total 3500

and AQ-detail and the overall motion coherence thresh-

old (independent of dot lifetime, F(1,30) = 0.01, p = .934,
and F(1,30) = 0.22, p = .640, respectively). Further-
more, no significant dot lifetime by AQ-total interaction, 3000

F(1,30) = 0.13, p = .726, nor dot lifetime by AQ-detail inter-

action, F(1,30) = 0.60, p = .446, was found. Thus, the rela-
tion between AQ-total/AQ-detail and the motion coherence
threshold was not moderated by dot lifetime. 2500

20 25 30
Relation Between Synaesthesia Scores and Performance AQ−detail score

No significant relation between synaesthesia consistency Fig. 5  Correlation between the AQ-Attention to detail subscores and
scores and mean overall motion coherence thresholds was response times on the Embedded Figures Task. Pearson’s r = − .47,
indicating faster response times with higher AQ-Detail
found, F(1,27) = 1.34, p = .256. Also, no significant synaes-
thesia by dot lifetime interaction was found, F(1,27) = 1.06,
p = .313, indicating that the relation between synaesthesia Relation Between AQ Scores and Performance
consistency scores and mean motion coherence threshold
did not differ for the two dot lifetime conditions. On a multivariate level, AQ-detail had a significant main
effect on overall EFT performance, F(2,29) = 4.53, p = .019,
Experiment 2: Embedded Figures Task (EFT) η2 = .24; univariate tests showed this to be significant for
response times, F(1,30) = 8.71, p = .006, η2 = .23, not for
General Task Performance error percentages, F(1,30) = .001, p = .980, η2 < .001. A
post hoc correlational analysis showed a negative relation
Difficulty (levels 1–4) had a significant main effect on over- between AQ-detail score and response times, r(31)= − .49,
all EFT performance, multivariate F(6,25) = 69.48, p < .001, p = .004, 95% CI [− 0.71, − 0.13], indicating that higher
η2 = .94. Univariate contrasts show that the differences in AQ-detail scores were associated with faster performance
response times between level 1 and 2, F(6,30) = 79.21, (Fig. 5).
p < .001, η 2 = .73, 2 and 3, F(1,30) = 168.54, p < .001, Next, on a multivariate level, AQ-total was not related
η2 = .85, and 3 and 4, F(1,30) = 54.69, p < .001, η2 = .65, to overall EFT performance, F(2,29) = 0.54, p = .589. Also,
were all significant. Response times gradually increased there were no significant multivariate interaction effects
as a function of difficulty level (Table 1). Error percent- between AQ-detail and difficulty, F(6,25) = 1.17, p = .172,
ages did not differ between level 1 and 2, F(1,30) = 0.02, or between AQ-total and difficulty, F(2,29) = 0.75, p = .618.
p = .881, whereas the differences between level 2 and Therefore, the univariate tests of these analyses (which
3, F(1,30) = 27.79, p < .001, η 2 = .48, and 3 and 4, separated response times and error percentages) were not
F(1,30) = 36.93, p < .001, η2 = .55, were significant. Simi- considered.
larly as for response times, higher difficulty levels led to
higher error percentages (Table 1). The results are line with Relation Between Synaesthesia Scores and Performance
those reported in the original paper introducing this task
(de-Wit et al. 2017). Synaesthesia consistency scores did not have a signifi-
Error percentages and response times did not correlate cant main effect on overall EFT performance, multivariate
significantly, r(31)=− .16, p = .371, 95% CI [− 0.47, 0.08], F(2,26) = 0.45, p = .641. Also, no significant interaction
so we have no indications for a speed-accuracy trade-off. between synaesthesia consistency scores and difficulty was
found, multivariate F(6,22) = 0.16, p = .958. Hence, univari-
ate tests, separated on response time and percentages, were
not considered.

13
22 Journal of Autism and Developmental Disorders (2020) 50:12–29

Müller−Lyer task Ebbinghaus task

50
No Context
10.0
Context
Discrepancy between stimuli (pixels)

Discrepancy between stimuli (pixels)

40

7.5

30

(*)
5.0
20

10 2.5

0
20 25 30 20 25 30
AQ−detail score AQ−detail score

Fig. 6  Significant interaction between context and AQ-detail on task was found that relates a higher AQ-detail score to higher task perfor-
performance in the visual illusions task, separated by the type of illu- mance. Results from the Ebbinghaus task (right panel) demonstrate
sion. A lower discrepancy between the to-be-adjusted and reference no relation between AQ-detail and task performance in context or
stimulus indicates higher task performance. Results from the Müller- context-free trials
Lyer task (left panel) indicate that only in the context trails, a trend

Experiment 3: Visual Illusions Task Relation Between AQ Scores and Performance

General Task Performance No main effect of AQ-total on the mean overall task per-
formance (independent of the type of illusion and context)
The repeated measures ANCOVA demonstrated a sig- was found, F(1,30) = 0.79, p = .381. AQ-detail did have
nificant main effect of type of illusion, F(1,30) = 121.03, a marginal main effect on the overall task performance,
p < .001, η2= .80. The mean discrepancy between the to- F(1,30) = 3.87, p = .058, η 2= .11 (medium effect size).
be-adjusted and reference stimulus was lower in the Ebb- Post-hoc correlational analysis on overall task perfor-
inghaus task (M = 3.55, SD =1.02) than in the Müller-Lyer mance revealed a trend indicating that a higher AQ-detail
task (M = 12.99, SD = 5.28), reflecting overall higher perfor- score was accompanied by a lower discrepancy between
mance in the Ebbinghaus than Müller-Lyer task. the to-be-adjusted and reference stimuli (i.e. a higher
Also, a significant main effect of context was found, performance; r(31) = − .302, p = .088, 95% CI [− 0.55,
F(1,30) = 91.11, p < .001, η2= .75, with the mean discrep- − 0.01]).
ancy between the two stimuli being higher in the contex- A significant context by AQ-detail interaction effect with
tual (M = 13.21, SD = 5.87) compared to the context-free medium effect size, F(1,30) = 4.21, p = .049, η2= .12, was
(M = 3.33, SD = 0.73) trials, confirming the effectiveness found (Fig. 6). This is relevant because we hypothesized
of the contextual stimuli as visual illusions. Furthermore, a that AQ scores would only be related to performance in
significant type of illusion by context interaction effect was the context trials. Additionally, a marginal 3-way interac-
found, F(1,33) = 60.70, p < .001, η2= .65. Post-hoc paired tion between type of illusion, context and AQ-detail with
sampled t-tests revealed that the effect of context was sig- medium effect size was present, F(1,30) = 3.66, p = .065,
nificant in both the Ebbinghaus, t(32) = − 10,79, p = < .001, η2= .13. Examination of the regression coefficients revealed
and Müller-Lyer, t(32) = − 8.30, p < .001 illusion. However, that in the Müller-Lyer task (Fig. 6a), there was a marginal
the effect of context was higher in the Müller-Lyer (mean relation between AQ-detail and performance only in the
difference = 16.09) than in the Ebbinghaus (mean differ- contextual trials (b = − 4.29, SE = 2.21, t = − 1.94, p = .062,
ence = 3.66) task. η2= .11), not in the context-free trials (b = 0.27, SE = 0.24,

13
Journal of Autism and Developmental Disorders (2020) 50:12–29 23

Müller−Lyer task Ebbinghaus task

No Context
Discrepancy between stimuli (pixels)

Discrepancy between stimuli (pixels)

Context
7.5
30

5.0
20

10 2.5

1.5 2.0 2.5 3.0 1.5 2.0 2.5 3.0

Synaesthesia consistency score Synaesthesia consistency score

Fig. 7  Interaction between context and synaesthesia consistency of synaesthesia. In the Müller-Lyer task (left panel), a non-significant
scores on task performance in the visual illusions task, separated by trend was found suggesting a positive relation between the degree of
the type of illusion. Note that a lower discrepancy between to-be- synaesthesia and task performance only in the context trials. No such
adjusted and reference stimulus indicates higher task performance, trend was found in the Ebbinghaus task (right panel)
and a lower synaesthesia consistency score indicates a higher degree

t =1.13, p = .269). In the Ebbinghaus task (Fig. 6b), the rela- explored this similarity with the AQ results and examined
tion between AQ-detail and the discrepancy between stimuli the regression coefficients. It appeared that (analogous to the
was not significant in the contextual (b = − 0.57, SE = 0.40, AQ results) a higher degree of synaesthesia was marginally
t = − 1.43, p = .163) nor context-free (b = − 0.07, SE = 0.11, related to a lower discrepancy between the to-be-adjusted
t = − 0.64, p = .526) trials. and reference stimulus (i.e. higher task performance) only
No interaction effects between AQ-total and context, in the context trials of the Müller-Lyer illusion (in which
F(1,30) = 1.69, p = .204, or other 2-way or 3-way interac- b = 3.20, SE = 1.71, t = 1.88, p = .071, η2= .12). We note that
tions between the within-subjects and covariates were found. this effect is not significant and cannot confirm a relation
between synaesthesia scores and performance; all we can
Relation Between Synaesthesia Scores and Performance conclude is that the relation is in the same direction as for
the AQ scores. No other 2- or 3-way interactions involving
No significant main effect of synaesthesia consistency scores the synaesthesia scores were obtained.
on mean overall task performance was found, although a
trend was present, F(1,27) = 3.19, p = .085, η2= .11. There
was a marginal interaction between context and synaesthe- Discussion
sia scores with medium effect size, F(1,27) = 3.82, p = .061,
η2= .12, while the 3-way interaction of context, synaesthesia We investigated the relation between the degree of autis-
score and type of illusion was not significant, F(1,27) = 3.13, tic traits (as measured with the Autism Quotient) and the
p = .088, η2= .10. Figure 7 shows the pattern of performance degree of grapheme-colour synaesthesia (as measured with
in relation to the synaesthesia scores is similar to the relation a consistency test) in neurotypicals, and whether this rela-
between AQ-detail and task performance (Fig. 6). Because tion is accompanied by a shared bias towards local (detail-
the effect size of the context by synaesthesia score interac- focussed) visual perception. In line with our first hypothesis,
tion was as large as the effect size of the interaction effect a positive relation exists between the degree of autistic traits
for AQ-detail, and we hypothesized the relation between (AQ-total scores) and the degree of synaesthesia. In addi-
synaesthesia scores and performance would be similar to tion, and supporting our second hypothesis, a relation was
the relationship between AQ-scores and performance, we found between the AQ-attention to detail subscores and a

13
24 Journal of Autism and Developmental Disorders (2020) 50:12–29

bias towards local visual perception, as indicated by per- Brosnan et al. (2012), however, also found fewer errors in
formance on the Embedded Figures Task (EFT) and (to a individuals with ASD, which we could not confirm in our
lesser extent) the visual illusions task. Performance on the sample. We furthermore found a relation between AQ-detail
motion coherence task (MCT) was not related to AQ scores. scores and a reduced susceptibility to visual illusions, as
Finally, no relation between the degree of synaesthesia and indicated by a significant AQ-detail by context interaction.
visual perception was found (with the exception of a non- This points towards an increased ability to focus on relevant,
significant trend in the visual illusions task resembling the local visual aspects and ignore contextual distractions. This
results obtained for autism). This contradicts our hypothesis is consistent with several previous studies (e.g. Bölte et al.
that synaesthesia would also be related to a tendency towards 2007; Happé 1996). When breaking down this AQ-detail
local visual perception. Therefore, two of our three main by context interaction for each visual illusion separately, we
hypotheses were supported. found a trend only in the Müller-Lyer (and not in the Ebb-
inghaus) illusion. This result is in accordance with Choui-
The Degree of Synaesthesia and Autistic Traits nard et al. (2013), who also reported a relation between AQ
scores and susceptibility to the Müller-Lyer illusion, and
Our finding of a relation between autistic traits and the acknowledge that there is no consensus on why this effect is
degree of synaesthesia in neurotypicals confirms research only present in this particular type of illusion. One possible
that previously established this link in clinical (i.e. supra- explanation could be that the Müller-Lyer illusion can be
threshold) populations (Baron-Cohen et al. 2013; Neufeld classified as a ‘within-object contextual illusion’, meaning
et al. 2013). As far as we know this study is the first to report that the contextual elements (the surrounding arrows) are
this relation in neurotypicals, supporting the idea of ASD physically attached to the local elements (the lines)—some-
and synaesthesia as dimensional constructs. The correlation thing that is not the case in the Ebbinghaus-illusion, which
between the degree of autistic traits and synaesthesia scores can be classified as a ‘between-object contextual illusion’
raises the question whether the ASD and synaesthesia con- (Ben-Shalom and Ganel 2012). Therefore, it might be more
tinuums share certain characteristics. Based on the existing difficult for most people to not integrate the context when
literature on visual perception in (clinical) ASD and syn- perceiving the Müller-Lyer illusion, and having the ability
aesthesia, we hypothesized a local bias in visual perception to focus on the local elements (i.e. a local bias) could result
might be shared. However, we found the synaesthesia scores in greater performance benefits on such a within-object illu-
to be related only to the AQ-total scores, not to the AQ-detail sion than in between-object contextual illusions. It should be
scores. The lack of a relation between the degree of synaes- noted that in Chouinard et al. (2013), the susceptibility to the
thesia and this subscale (assessing self-reported attention Müller-Lyer illusion did not relate to AQ-Attention-to-detail
to detail) is in line with the results of our experiments on (but only to total AQ scores).
visual perception and its relation with synaesthesia scores, The degree of autistic traits and performance on the
as explained below. motion coherence task (MCT) did not relate, and there
was no moderation by dot lifetime. It is not clear whether
The Degree of Autistic Traits and Visual Perception enhanced local perception in ASD is indeed always
accompanied by impaired global perception: although
Our results show a relation between the degree of autistic researchers traditionally supported this idea (Frith and
traits (AQ-detail scores, not AQ-total) and a local bias in vis- Happé 1994), recent studies suggest that individuals with
ual perception on an embedded figures task and an illusion ASD merely have a reduced preference to process global
task. Our findings support prior studies that demonstrated a information, not a reduced ability (Happé and Frith 2006;
local bias in individuals with ASD, and are therefore in line Koldewyn et al. 2013; Van der Hallen et al. 2015). This
with the detail-focussed cognitive style as proposed by the possibly explains why autistic traits did not relate to per-
weak central coherence theory (Happé and Frith 2006). The formance in the limited dot lifetime condition (60 ms) in
results are also in line with the findings from Cribb et al. which participants were forced to process global motion.
(2016), showing that this tendency towards detail-focused In the unlimited dot lifetime (600 ms) condition, which
perception exists along the autistic (sub- and supra-clinical) allowed the tracking of single dots, we anticipated a posi-
spectrum. tive relation between AQ-scores and performance. One
The strongest evidence for the local bias was found in the potential reason for the lack of this relation might be that
EFT, as a higher degree of autistic traits was related to faster the assumption underlying our hypothesis was not cor-
response times on this task. This is consistent with studies rect. That is, we assumed that in the unlimited dot lifetime
performed in clinical populations, which also found faster condition, participants with both low and high AQ-scores
response times in individuals diagnosed with ASD com- would use a ‘local strategy’ (i.e. tracking a single dot to
pared to controls (Brosnan et al. 2012, Pellicano et al. 2006). identify its motion direction). However, it is also possible

13
Journal of Autism and Developmental Disorders (2020) 50:12–29 25

that participants with low AQ-scores refrained from using of similarity to our previous results provides support for
the local strategy altogether, and used the global strategy the relative stability of our initial estimates. This confirms
regardless of dot lifetime. Here, people with both low and that the interdependence between AQ-total and AQ-detail
high AQ-scores might have obtained high performance, did not have a major influence on our results.
but using different strategies. In addition, only people who
used the local strategy (participants with high AQ-scores The Degree of Synaesthesia and Visual Perception
according to this line of reasoning) might sometimes have
suffered from selecting and tracking the wrong dot, an In contrast to the findings for the degree of autistic traits,
error that is inherent to the local strategy. This may have we did not find a relation between the degree of synaes-
prevented participants with high AQ-scores to outperform thesia and local visual perception in neurotypicals. This
participants with lower AQ-scores on the unlimited dot result, combined with the fact that we found no corre-
lifetime condition. lation between the synaesthesia consistency scores and
Alternatively, a failure to find the hypothesized posi- the AQ-detail scores, raises the possibility that the rela-
tive relation between AQ scores and performance in the tion between the degree of synaesthesia and the AQ-total
unlimited dot lifetime condition could have been due to a scores should be sought somewhere else than in a shared
ceiling effect. It is possible that the unlimited dot lifetime local bias. To investigate this possibility, we decided to
condition was too easy compared to the limited dot life- explore the correlation between the synaesthesia scores
time condition, resulting in all participants scoring sig- and the remaining AQ-subscores separately (Social skills,
nificantly better and not allowing for a differentiation to Communication, Imagination and Attention switching), as
be made based on AQ score. Similar results were obtained well as when they were added together into an AQ-other
by Manning et al. (2015), who used dot lifetimes of 83 subscale. However, no significant correlations were found,
versus 1000 ms and found no moderating effect of dot providing no support for this potential explanation.
lifetime. Jackson et al. (2013) used dot lifetimes of 80 and We found a non-significant trend for the relation
300 ms and did find an AQ by dot lifetime interaction, between the degree of synaesthesia and susceptibility to
supporting this explanation. Future studies are encouraged visual illusions that was very similar to what we found in
to systematically vary these dot lifetimes, to investigate in the AQ analyses. This tentatively points towards an alter-
what range (between 50 and 600 ms) the dot lifetime forms native explanation for our results that does not completely
a moderating effect on the relation between the degree of exclude a potential local bias in synaesthesia: a possible
autistic traits and performance on the MCT. bias towards local visual perception could be stronger in
Finally, one issue that should be discussed relates to supra-threshold synaesthetes. Support for this explana-
our earlier discussion about the interdependence of AQ- tion stems from recent studies reporting a local bias in
detail and AQ-total, and the use of both variables in the supra-threshold synaesthetes. Ward et al. (2018) and Van
same analyses. Although we have shown that the relation Leeuwen et al. (2019) found increased AQ-detail scores in
between these variables does not cause statistical colline- supra-threshold synaesthetes, which we did not find in our
arity issues, it should be acknowledged that AQ-detail and non-synaesthete sample. Van Leeuwen et al. (2019) report
AQ-total are, by definition, not independent (as AQ-detail decreased performance on the limited dot lifetime MCT
forms a subscore of AQ-total). Therefore, one might still in synaesthetes and both studies demonstrated a decreased
wonder what the relation between the degree of autistic error percentage on the EFT. Interestingly, however, they
traits and visual perception looks like when this depend- did not find decreased response times, which could poten-
ency between AQ-detail and AQ-total is excluded. In order tially be explained by differences in testing procedure
to examine this, we reran our analyses in all three visual (online versus in the laboratory).
perception experiments, replacing AQ-total by an AQ- To see whether our own data could provide some further
other score, a score composed of all AQ subscales except (speculative) support for the differences between sub- and
AQ-detail (similar to Ward et al. 2018). To be certain, we supra-threshold synaesthetes, we explored the scores of
assessed the degree of collinearity between AQ-detail and the three synaesthete participants who were excluded from
AQ-other, which was found not to be problematic (e.g. our neurotypical sample. Supporting previous research in
VIF = 1.03). The results of the analyses were very similar synaesthetes (Ward et al. 2017) and extending on the cor-
to our previous results. For the MCT and EFT, we found relation in neurotypicals, the AQ-total/synaesthesia cor-
no differences in results. For the visual illusions task, the relation was maintained and even became more prominent
only difference we found was that the multivariate AQ- when including these 3 participants, r(30)= − .48, p = .006,
detail by Context interaction became marginally signifi- 95% CI [− 0.70, − 0.19]. This suggests that the relation
cant [F(1,30) = 3.30, p = .079, η2= .10]. This shows that between the degree of synaesthesia and autistic traits
although the results are not completely similar, the degree we found in neurotypicals can be extended to a sample

13
26 Journal of Autism and Developmental Disorders (2020) 50:12–29

including supra-threshold synaesthetes and encourages alternative underlying traits. For instance, it could indicate
future studies to investigate the possible existence of a a high memory performance, a high attention to colours,
local bias along a synaesthesia continuum. or a high degree of conscientiousness. The latter possibil-
ity, however, would not account for the relation between the
Strengths and Limitations of the Present Study degree of synaesthesia and autistic traits that we found, as
a recent meta-analysis (Lodi-Smith et al. 2018) found con-
Our study assessed visual perception using three different scientiousness to be negatively correlation with ASD char-
experiments, allowing for a balanced assessment of a poten- acteristics. This limitation concerns the widely discussed
tial local bias in visual perception. Furthermore, we system- debate whether synaesthesia should be considered as a con-
atically investigated certain task properties (e.g. dot lifetime, tinuous traits or all-or-none phenomenon, as discussed in the
difficulty, type of illusion) that could potentially moderate introduction. Future research is encouraged to investigate
the relation between the degree of autistic traits/synaesthesia the value of synaesthesia tests as a measure of the degree of
and visual perception. This extends on previous reviews and synaesthesia in neurotypicals.
meta-analyses that discussed the possible moderating influ-
ence of such task properties (see King et al. 2017 on visual
illusions, and Simmons et al. 2009 on dot lifetime); possibly Conclusions
explaining inconsistencies in previous research as discussed
in the introduction. Third, we used an extensive grapheme- Our study has provided insight into the perceptual expe-
colour synaesthesia test. By complementing these test scores riences and abilities inherent to synaesthesia and ASD, as
with a self-report questionnaire on the possible experience well as into their relationship. We found a positive relation
of synaesthesia, we managed to exclude individuals with between the degree of synaesthesia and autistic traits, one
synaesthesia from our sample while still obtaining reliable of the first demonstrations of this relation in a neurotypi-
measures of grapheme-colour consistency. cal sample. Furthermore, we found a relation between the
Our study also has some limitations, which are important degree of autistic traits and a bias towards local visual per-
for future studies to consider. First, we only assessed the ception. We did not find evidence for a relation between
degree of synaesthesia for the grapheme-colour type. The the degree of synaesthesia and a bias towards local visual
relation between the degree of synaesthesia and the degree perception. One explanation for this last result might be
of autistic traits might also extend to other types of synaes- that this local bias is expressed stronger in supra-threshold
thesia: if so, then this relation could be ascribed to a more synaesthetes than in sub-threshold synaesthetes; this expla-
fundamental, underlying synaesthesia ‘trait’ (Rouw and nation is supported by other studies into synaesthetes and by
Scholte 2016; Rouw et al. 2011). Future studies could inves- a speculative exploration of our own data. Future studies are
tigate whether the relation between the degree of synaes- encouraged to study this alternative explanation, and/or to
thesia and local/global visual perception might be depend- extend our research to other types of synaesthesia.
ent on the specific synaesthesia type. For instance, using a Gaining a better understanding of the relationship
questionnaire to assess local/global bias, Mealor et al. (2016) between ASD and synaesthesia can lead to important
found evidence for a self-reported local bias in sequence- insights into perceptual alterations in ASD. This might also
space synaesthesia, but not in grapheme-colour synaesthesia. have practical implications; that is, as perceptual dysregula-
Future studies could use visual perception tasks to confirm tion has been acknowledged as an important clinical feature
these findings, and could extend this to other types of syn- of ASD, an increased understanding hereof might aid indi-
aesthesia. A second limitation of our study comes from the viduals with ASD, as well as their clinicians.
fact that, due to exclusion of several participants from our
original sample (N = 39), we performed our analyses with a Acknowledgments We thank Wilbert van Ham of the Technical Sup-
relatively small sample size (33 in the AQ analyses and 29 in port Group for online task implementation and Lee de-Wit for provid-
ing the Embedded Figures test stimuli. FB performed this research as
the synaesthesia analyses). Although we did find significant
part of the Honours Programme of Psychology at the Radboud Uni-
effects in several of our analyses, the small sample size made versity, Nijmegen.
it harder to reliably detect relatively subtle effects. Future
studies could seek out to validate our results more firmly Author Contributions TvL and FB conceived of the study, participated
using a larger sample of participants. in the design and coordination of the study, and participated in the
interpretation of the data. FB collected the data, performed the sta-
Finally, a potential limitation concerns the use of a syn-
tistical analyses, and drafted the manuscript. MD contributed to the
aesthesia test as a continuous measure of the degree of synaesthesia test. TvL, MD and RvL edited the manuscript. All authors
synaesthesia in healthy controls. Since our participants did read and approved the final manuscript.
not actually have any synaesthesia, it should be noted that a
high grapheme-colour consistency score might also reflect

13
Journal of Autism and Developmental Disorders (2020) 50:12–29 27

Funding This study was supported by a Nederlandse Organisatie voor Chouinard, P. A., Noulty, W. A., Sperandio, I., & Landry, O. (2013).
Wetenschappelijk Onderzoek (NWO) Veni grant to TvL [451.14.025] Global processing during the Müller-Lyer illusion is distinctively
and by an NWO Groot Nationaal Onderzoek grant to MD and TvL. affected by the degree of autistic traits in the typical popula-
tion. Experimental Brain Research, 230, 219–231. https​://doi.
Data Availability All data associated with this paper can be retrieved org/10.1007/s0022​1-013-3646-6.
from: http://hdl.handl​e.net/11633​/aacbl​xq5 Chouinard, P. A., Unwin, K. L., Landry, O., & Sperandio, I. (2016).
Susceptibility to optical illusions varies as a function of the
autism-spectrum quotient but not in ways predicted by local-
Open Access This article is distributed under the terms of the Crea-
global biases. Journal of Autism and Developmental Disorders,
tive Commons Attribution 4.0 International License (http://creat​iveco​
46, 2224–2239. https​://doi.org/10.1007/s1080​3-016-2753-1.
mmons​.org/licen​ses/by/4.0/), which permits unrestricted use, distribu-
Cohen, J. (1988). Statistical power analysis for the behavioral sciences.
tion, and reproduction in any medium, provided you give appropriate
Hillsdale, NJ: Lawrence Erlbaum Associates.
credit to the original author(s) and the source, provide a link to the
Cribb, S. J., Olaithe, M., Di Lorenzo, R., Dunlop, P. D., & Maybery,
Creative Commons license, and indicate if changes were made.
M. T. (2016). Embedded figures test performance in the broader
autism phenotype: A meta-analysis. Journal of Autism and Devel-
opmental Disorders, 46, 2924–2939. https​://doi.org/10.1007/
References s1080​3-016-2832-3.
Cuskley, C., Dingemanse, M., Kirby, S., & Van Leeuwen, T. M. (2019).
Cross-modal associations and synesthesia: Categorical percep-
American Psychiatric Association. (2013). Diagnostic and statistical
tion and structure in vowel–color mappings in a large online
manual of mental disorders (5th ed.). Washington, DC: American
sample. Behavior Research Methods, 51, 1651–1675. https​://doi.
Psychiatric Association.
org/10.3758/s1342​8-019-01203​-7.
Baguley, T. (2012). Serious stats: A guide to advanced statistics for
Dakin, S., & Frith, U. (2005). Vagaries of visual perception in
the behavioral sciences. Basingstoke, United Kingdom: Palgrave
autism. Neuron, 48, 497–507. https​://doi.org/10.1016/j.neuro​
Macmillan.
n.2005.10.018.
Banissy, M. J., Tester, V., Muggleton, N. G., & Ward, J. (2013). Synes-
De Groot, K., & Van Strien, J. W. (2017). Evidence for a broad autism
thesia for color is linked to improved color perception but reduced
phenotype. Advances in Neurodevelopmental Disorders, 1, 129–
motion perception. Psychological Science, 24, 2390–2397. https​
140. https​://doi.org/10.1007/s4125​2-017-0021-9.
://doi.org/10.1177/09567​97613​49242​4.
De-Wit, L., Huygelier, H., Van der Hallen, R., Chamberlain, R., &
Barnett, K. J., Foxe, J. J., Molholm, S., Kelly, S., Shalgi, S., Mitch-
Wagemans, J. (2017). Developing the Leuven Embedded Figures
ell, K. J., et al. (2008). Differences in early sensory-perceptual
Test (L-EFT): Testing the stimulus features that influence embed-
processing in synaesthesia: A visual evoked potential study.
ding. PeerJ. https​://doi.org/10.7717/peerj​.2862.
NeuroImage, 43, 605–613. https​://doi.org/10.1016/j.neuro​image​
Dormann, C. F., Elith, J., Bacher, S., Buchmann, C., Carl, G., Carré,
.2008.07.028.
G., et al. (2013). Collinearity: A review of methods to deal with it
Baron-Cohen, S., Johnson, D., Asher, J., Wheelwright, S., Fisher,
and a simulation study evaluating their performance. Ecography.
S. E., Gregersen, P. K., et al. (2013). Is synaesthesia more
https​://doi.org/10.1111/j.1600-0587.2012.07348​.x.
common in autism? Molecular Autism, 4, 1–6. https​: //doi.
Eagleman, D. M., Kagan, A. D., Nelson, S. S., Sagaram, D., & Sarma,
org/10.1186/2040-2392-4-40.
A. K. (2007). A standardized test battery for the study of synes-
Baron-Cohen, S., Wheelwright, S., Skinner, R., Martin, J., Clubley,
thesia. Journal of Neuroscience Methods, 159, 139–245. https​://
E. (2001). The autism-spectrum quotient (AQ): Evidence from
doi.org/10.1016/j.jneum​eth.2006.07.012.
Asperger syndrome/high-functioning autism, males and females,
Ellis, J. L. (2014). Statistiek voor de psychologie, deel 1: Verdeling
scientists and mathematicians. Journal of Autism and Develop-
van een variabele, samenhang tussen twee variabelen. Den Haag:
mental Disorders, 31, 5–17. Retrieved from http://www.sprin​ger.
Boom Lemma uitgevers.
com/psych​ology​/child​+%26+schoo​l+psych​ology​/journ​al/10803​
Ellis, J. L. (2016). Statistiek voor de psychologie, deel 4: GLM en
Ben-Shalom, A., & Ganel, T. (2012). Object representations in visual
non-parametrische toetsen. Amsterdam: Boom Lemma uitgevers.
memory: Evidence from visual illusions. Journal of Vision, 12,
Faul, F., Erdfelder, E., Lang, A. G., & Buchner, A. (2007). G*Power
1–11. https​://doi.org/10.1167/12.7.15.
3: A flexible statistical power analysis program for the social,
Bien, N., Ten Oever, S., Goebel, R., & Sack, A. T. (2012). The sound of
behavioral, and biomedical sciences. Behavior Research Methods,
size: Crossmodal binding in pitch-size synaesthesia: A combined
39(2), 175–191.
TMS, EEG and psychophysics study. NeuroImage, 59, 663–672.
Frith, U., & Happé, F. (1994). Autism: Beyond ‘theory of mind’. Cog-
https​://doi.org/10.1016/j.neuro​image​.2011.06.095.
nition, 50, 115–132. https:​ //doi.org/10.1016/0010-0277(94)90024​
Bölte, S., Holtmann, M., Poustka, F., Scheurich, A., & Schmidt, L.
-8.
(2007). Gestalt perception and local-global processing in high-
Hänggi, J., Wotruba, D., & Jäncke, L. (2011). Globally altered struc-
functioning autism. Journal of Autism and Developmental Disor-
tural brain network topology in grapheme-color synaesthesia. The
ders, 37, 1493–1503. https:​ //doi.org/10.1007/s10803​ -006-0231-x.
Journal of Neuroscience, 31, 5816–5828. https:​ //doi.org/10.1523/
Brosnan, J., Gwilliam, L. R., & Walker, I. (2012). Brief report: The
JNEUR​OSCI.0964-10.2011.
relationship between visual acuity, the embedded figures test and
Happé, F. G. E. (1996). Studying weak central coherence at low lev-
systemizing in autism spectrum disorders. Journal of Autism
els: Children with autism do not succumb to visual illusions. A
and Developmental Disorders, 42, 2491–2497. https ​ : //doi.
research note. Journal of Child Psychology and Psychiatry, 37,
org/10.1007/s1080​3-012-1505-0.
873–877. https​://doi.org/10.1111/j.1469-7610.1996.tb014​83.x.
Brown, A. C., & Crewther, D. P. (2017). Autistic children show a sur-
Happé, F., & Frith, U. (2006). The weak coherence account: Detail-
prising relationship between global visual perception, non-verbal
focused cognitive style in autism spectrum disorders. Journal
intelligence and visual parvocellular function, not seen in typically
of Autism and Developmental Disorders, 36, 5–25. https​://doi.
developing children. Frontiers in Human Neuroscience. https​://
org/10.1007/ss108​03-005-0039-0.
doi.org/10.3389/fnhum​.2017.00239​.

13
28 Journal of Autism and Developmental Disorders (2020) 50:12–29

Hoekstra, R. A., Bartels, M., Cath, D. C., & Boomsma, D. I. (2008). mind and executive control. Development and Psychopathology, 18,
Factor structure, reliability and criterion validity of the Autism- 77–98. https​://doi.org/10.1017/S0954​57940​60600​56.
Spectrum Quotient (AQ): A study in Dutch population and patient R Core Team. (2018). R: A language and environment for statistical com-
groups. Journal of Autism and Developmental Disorders, 38, puting (Version 3.5.1) [Computer software]. Retrieved from https​://
1555–1566. https​://doi.org/10.1007/s1080​3-008-0538-x. www.R-proje​ct.org/
Jackson, B. L., Blackwood, E. M., Blum, J., & Crewther, D. P. (2013). R Studio Team. (2016). RStudio: Integrated development environment
Magno- and parvocellular contrast responses in varying degrees for R (Version 1.1.456) [Computer software]. Retrieved from http://
of autistic trait. PLoS ONE, 8, e66797. https​://doi.org/10.1371/ www.rstud​io.com/
journ​al.pone.00667​97. Rothen, N., Seth, A. K., Witzel, C., & Ward, J. (2013). Diagnosing syn-
Janik McErlean, A. B., Susilo, T., Rezlescu, C., Bray, A., & Banissy, aesthesia with online colour pickers: Maximising sensitivity and
M. J. (2016). Social perception in synaesthesia for colour. Cogni- specificity. Journal of Neuroscience, 215, 156–160. https​://doi.
tive Neuropsychology, 33, 378–387. https​://doi.org/10.1080/02643​ org/10.1016/j.jneum​eth.2013.02.009.
294.2016.12618​20. Rouw, R., & Scholte, H. S. (2016). Personality and cognitive profiles of
Just, M. A., Keller, T. A., Malave, V. L., Kana, R. K., & Varma, S. (2012). a general synesthetic trait. Neuropsychologia, 88, 35–48. https​://doi.
Autism as a neural systems disorder: A theory of frontal-posterior org/10.1016/j.neuro​psych​ologi​a.2016.01.006.
underconnectivity. Neuroscience and Biobehavioral Reviews, 36, Rouw, R., Scholte, H. S., & Colizoli, O. (2011). Brain areas involved in
1292–1313. https​://doi.org/10.1016/j.neuro​biore​v.2012.02.007. synaesthesia: A review. Journal of Neuropsychology, 5, 214–242.
Kadosh, R. C., & Henik, A. (2007). Can synaesthesia research inform https​://doi.org/10.1111/j.1748-6653.2011.02006​.x.
cognitive science? Trends in Cognitive Sciences, 11, 177–184. https​ Schauder, K. B., & Bennetto, L. (2016). Toward an interdisciplinary
://doi.org/10.1016/j.tics.2007.01.003. understanding of sensory dysfunction in autism spectrum disorder:
Kanner, L. (1943). Autistic disturbances of affective contact. Nerv- An integration of the neural and symptom literatures. Frontiers in
ous Child: Journal of Psychopathology, Psychotherapy, Mental Neuroscience, 10, 268. https​://doi.org/10.3389/fnins​.2016.00268​.
Hygiene, and Guidance of the Child, 2, 217–250. Simmons, D. R., Robertson, A. E., McKay, L. S., Toal, E., McAleer,
King, D. J., Hodgekins, J., Chouinard, P. A., Chouinard, V.-A., & Speran- P., & Pollick, F. E. (2009). Vision in autism spectrum disorders.
dio, I. (2017). A review of abnormalities in the perception of visual Vision Research, 49, 2705–2739. https​://doi.org/10.1016/j.visre​
illusions in schizophrenia. Psychonomic Bulletin & Review, 24, s.2009.08.005.
734–751. https​://doi.org/10.3758/s1342​3-016-1168-5. Simner, J., Mulvenna, C., Sagiv, N., Tsakaniok, E., Witherby, S. A.,
Koldewyn, K., Jiang, Y., Weigelt, S., & Kanwisher, N. (2013). Global/ Fraser, C., et al. (2006). Synaesthesia: The prevalence of atypical
local processing in autism: Not a disability, but a disinclination. cross-modal experiences. Perception, 35, 1024–1033. https​://doi.
Journal of Autism and Developmental Disorders, 43, 2329–2340. org/10.1068/p5469​.
https​://doi.org/10.1007/s1080​3-013-1777-z. Sutherland, A., & Crewther, D. P. (2010). Magnocellular visual evoked
Lodi-Smith, J., Rodgers, J. D., Cunningham, S. A., Lopata, C., & Tho- potential delay with high autism spectrum quotient yields a neural
meer, M. L. (2018). Meta-analysis of Big Five personality traits mechanism for altered perception. Brain, 133, 2089–2097. https​://
in autism spectrum disorder. Autism. https​://doi.org/10.1177/13623​ doi.org/10.1093/brain​/awq12​2.
61318​76657​1. Van der Hallen, R., Evers, K., Brewaeys, K., Van den Noortgate, W., &
Manning, C., Charman, T., & Pellicano, E. (2015). Brief report: Coherent Wagemans, J. (2015). Global processing takes time: A meta-analysis
motion processing in autism: Is dot lifetime an important parameter? on local-global visual processing in ASD. Psychological Bulletin,
Journal of Autism and Developmental Disorders, 45, 2252–2258. 141, 549–573. https​://doi.org/10.1037/bul00​00004​.
https​://doi.org/10.1007/s1080​3-015-2365-1. Van Leeuwen, T. M., & Dingemanse, M. (2016). Colour associations in
Manning, C., Morgan, M. J., Allen, C. T. W., & Pellicano, E. (2017). synaesthetes and nonsynaesthetes: A large-scale study in Dutch. Per-
Susceptibility to Ebbinghaus and Müller-Leyer illusions in autis- ception, 45, 333–334. https​://doi.org/10.1177/03010​06616​67127​3.
tic children: A comparison of three different methods. Molecular Van Leeuwen, T. M., Hagoort, P., & Händel, B. F. (2013). Real color
Autism. https​://doi.org/10.1186/s1322​9-017-0127-y. captures attention and overrides spatial cues in grapheme-color syn-
Mealor, A. D., Simner, J., Rothen, N., Carmichael, D. A., & Ward, J. esthetes but not in controls. Neuropsychologia, 51, 1802–1813. https​
(2016). Different dimensions of cognitive style in typical and atypi- ://doi.org/10.1016/j.neuro​psych​ologi​a.2013.06.024.
cal cognition: New evidence and a new measurement tool. PLoS Van Leeuwen, T. M., Van Petersen, E., Burghoorn, F., Dingemanse, M.,
ONE. https​://doi.org/10.1371/journ​al.pone.01554​83. & Van Lier, R. (2019). Autistic traits in synaesthesia: Atypical sen-
Muth, A., Hönekopp, J., & Falter, C. M. (2014). Visuo-spatial perfor- sory sensitivity and enhanced perception of details. Philosophical
mance in autism: A meta-analysis. Journal of Autism and Devel- Transactions of the Royal Society B: Biological Sciences. https​://doi.
opmental Disorders, 44, 3245–3263. https​://doi.org/10.1007/s1080​ org/10.1098/rstb.2019.0024.
3-014-2188-5. Ward, J. (2019). Synaesthesia: A distinct entity that is an emergent feature
Navon, D. (1977). Forest before trees: The precedence of global features of adaptive neurocognitive differences. Philosophical Transactions
in visual perception. Cognitive Psychology, 9, 353–383. https​://doi. of the Royal Society B: Biological Sciences. Retrieved from http://
org/10.1016/0010-0285(77)90012​-3. sro.susse​x.ac.uk/id/eprin​t/82842​/
Neufeld, J., Roy, M., Zapf, A., Sinke, C., Emrich, H. M., Prox-Vagedes, Ward, J., Brown, P., Sherwood, J., & Simner, J. (2018). An autistic-like
V., et al. (2013). Is synesthesia more common in patients with profile of attention and perception in synaesthesia. Cortex, 107,
Asperger syndrome? Frontiers in Human Neuroscience. https​://doi. 121–130. https​://doi.org/10.1016/j.corte​x.2017.10.008.
org/10.3389/fnhum​.2013.00847​. Ward, J., Hoadley, C., Hughes, J. E. A., Smith, P., Allison, C., Baron-
Newsome, W. T., & Paré, E. B. (1988). A selective impairment of motion Cohen, S., et al. (2017). Atypical sensory sentivity as a shared fea-
perception following lesions of the middle temporal visual area ture between synaesthesia and autism. Scientific Reports, 7, 41155.
(MT). The Journal of Neuroscience, 8, 2201–2211. Retrieved from https​://doi.org/10.1038/srep4​1155.
http://www.jneur​osci.org Wickham, H. (2016). ggplot2: Elegant graphics for data analysis [Com-
Pellicano, E., Maybery, M., Durkin, K., & Maley, A. (2006). Multiple puter software]. Retrieved from http://ggplo​t2.org
cognitive capabilities/deficits in children with autism spectrum dis-
order: ‘Weak’ central coherence and its relationship to theory of

13
Journal of Autism and Developmental Disorders (2020) 50:12–29 29

Witkin, H. A., Ottman, R. K., Raskin, E., & Karp, S. A. (1971). A manual analysis. International Journal of Applied Science and Technology,
for the embedded figures test. Palo Alto, California, CA: Consulting 4, 9–19. Retrieved from http://www.ijast​net.com
Psychologists Press.
Yoo, W., Mayberry, R., Bae, S., Singh, K., He, Q. P., & Lilliard, J. W. Jr. Publisher’s Note Springer Nature remains neutral with regard to
(2014). A study of effects of multicollinearity in the multivariable jurisdictional claims in published maps and institutional affiliations.

13