You are on page 1of 13

Neuropsychologia 171 (2022) 108240

Contents lists available at ScienceDirect

Neuropsychologia
journal homepage: www.elsevier.com/locate/neuropsychologia

Neural differences in social and figurative language processing on the


autism spectrum
William W. Graves a, *, Hillary Levinson a, Linsah Coulanges b, Shannon Cahalan a, Daniel Cruz c,
Catherine Sancimino d, Vanessa H. Bal e, Miriam Rosenberg-Lee a
a
Department of Psychology, Rutgers University – Newark, Newark, NJ, USA
b
Department of Psychology, University of Pittsburgh, Pittsburgh, PA, USA
c
Psychology, Hackensack Meridian Health – Mountainside Medical Center, Montclair, NJ, USA
d
Department of Pediatric Neuroscience, Albert Einstein College of Medicine, Bronx, NY, USA
e
Graduate School of Applied and Professional Psychology, Rutgers University – New Brunswick, Piscataway, NJ, USA

A R T I C L E I N F O A B S T R A C T

Keywords: Individuals on the autism spectrum often have trouble with social and figurative language. As social language is
MeSH terms) often figurative, it can be challenging to disentangle the cognitive and neural sources of these difficulties. Neural
Humans systems for social cognition and language comprehension overlap in areas involved in retrieving linguistic
Language
meaning (semantics), such as the anterior temporal lobe (ATL), ventro-medial prefrontal cortex (vmPFC), pos­
Autism spectrum disorder
Brain
terior cingulate cortex (PCC), and posterior middle temporal gyrus (pMTG). Using adjective-noun phrases, we
Magnetic resonance imaging manipulated social/nonsocial and figurative/literal dimensions, which we expected to activate distinct but
Semantics overlapping regions. We hypothesized that activation differences in the group with autism (AUT) would be
Social cognition greater for more social and figurative stimuli. During fMRI, participants in the AUT group (N = 19) and those in
the non-autistic comparison (NAC) group (N = 22) made familiarity judgments to 192 phrases in a balanced 2 ×
2 (social/nonsocial x figurative/literal) design. Social phrases activated the PCC in all participants, but only the
NAC group activated the vmPFC. Figurative phrases were rated as more literal by the AUT group, with the
figurative-literal phrase contrast showing no activation in the AUT group, but activating the PCC and right pMTG
in the NAC group. The one significant group-level neural difference was for the social-figurative condition
predicted to be most different between groups: greater activation for the AUT group in the right ATL. Differences
in the right ATL and pMTG in the AUT group suggest altered engagement of right homologues of the canonical
semantic network being recruited for processing combined social and figurative language.

1. Introduction et al., 2008; Phan et al., 2021). Abstract language contains both social
content and varying degrees of figurative or literal content (Borghi et al.,
Individuals on the autism spectrum, including those without a his­ 2017). Notably, the neural systems found to modulate with different
tory of language delays, have been found to exhibit difficulties com­ degrees of social content are largely the same as those that respond to
prehending some kinds of abstract language such as metaphor (Vicente different degrees of semantic content during language comprehension
and Falkum, 2021; Vulchanova et al., 2015). Whether these difficulties (Smallwood et al., 2021). In this study we sought to better understand
arise primarily from a linguistic source, or relate to other core features of the neural underpinnings of language comprehension differences in
autism, such as difficulties with social cognition, has been challenging to autism by determining whether there are distinct neuro-cognitive
pinpoint. This challenge arises from the fact that abstract language and mechanisms behind social and figurative language processing.
metaphor often contain social content. Like difficulties with social
processing, abstract language comprehension difficulties in autism have
been found both in terms of behavioral performance (Kalandadze et al., 1.1. Social language processing
2018; Vulchanova et al., 2015) and differences in neural systems (Groen
While there are wide-ranging individual differences, in general,

* Corresponding author.
E-mail address: william.graves@rutgers.edu (W.W. Graves).

https://doi.org/10.1016/j.neuropsychologia.2022.108240
Received 2 March 2022; Received in revised form 13 April 2022; Accepted 18 April 2022
Available online 21 April 2022
0028-3932/© 2022 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).
W.W. Graves et al. Neuropsychologia 171 (2022) 108240

individuals with autism are often considered to have challenges in inferior frontal cortex. However, in their meta-analysis, the left but not
comprehending social content, including social language (Groen et al., right ATL was activated for figurative language (Rapp et al., 2012).
2008; Kalandadze et al., 2018; Levy and Perry, 2011; Tager-Flusberg Considered together with studies pointing to a prominent role of the ATL
et al., 2005). Those without a history of measurable adverse effects on in social (Olson et al., 2007; Ross and Olson, 2010) and semantic
language development or verbal intelligence can have quite strong cognition (Patterson et al., 2007; Visser et al., 2010), particularly on the
general language and reading skills. Yet these same individuals may left when language is involved (Binder et al., 2009; Rice et al., 2015;
stumble when it comes to socially focused language such as irony, Wang et al., 2010), we can predict a divergence from activation in the
sarcasm, or pragmatics that depend on inferring the mental states of left ATL if there is divergent processing of social and figurative language
others (Landa and Goldberg, 2005; Williams et al., 2015). in individuals with autism.
The neural underpinnings of challenges with social language in Regarding the typical neural patterns related to semantic processing,
autism point to differences largely within the same midline and lateral it is notable that these regions reside in higher-order association
temporo-parietal areas that are typically activated for processing social cortices, rather than more primary sensory or motor cortices. By
language. This network of areas includes anterior midline structures contrast, a pattern seen in the handful of studies of semantic processing
such as the medial prefrontal cortex (mPFC), posterior midline struc­ in autism is that participants on the spectrum activate sensory-related
tures such as the posterior cingulate cortex (PCC) and precuneus, the areas such as the lateral occipital cortex. For the same semantic pro­
anterior temporal lobe (ATL), and lateral structures such as the posterior cessing conditions, activation for non-autistic participants is restricted
middle temporal gyrus (pMTG) and angular gyrus (AG) (Denny et al., to areas of association cortex that have been independently related to
2012; Molenberghs et al., 2016; Olson et al., 2007; Van Overwalle, semantics (Gaffrey et al., 2007; Harris et al., 2006).
2009). Largely overlapping with these areas of the mentalizing network More generally, a recent meta-analysis of studies examining the
are areas consistently found to activate for language comprehension neural basis of semantic processing in autism suggested only limited
(semantic processing), including the mPFC, PCC, ATL, pMTG, and AG findings to be consistent across studies (Phan et al., 2021). Whether their
(Binder and Desai, 2011; Binder et al., 1999, 2009; Mattheiss et al., paucity of results was due to a lack of relevant studies (i.e., N = 11 in the
2018). The goals of the current study were twofold. One was to deter­ meta-analysis), or spatial heterogeneity of functional neural organiza­
mine whether neural areas responding to social or figurative content will tion in autism, was not clear. Nevertheless, a finding that did consis­
differ when those factors are orthogonally manipulated. The other was tently emerge was greater activation for the autistic than non-autistic
to determine if neural areas responding to manipulation of social or groups in the left pMTG bordering on the posterior superior temporal
figurative content will differ for participants with compared to those gyrus (pSTG). This region is at a nexus point between the semantic
without autism. network that includes pMTG and the language network that includes
pSTG associated with phonological (speech sound or auditory word
1.2. Figurative language processing form) processing (Graves et al., 2008; Vigneau et al., 2006). The pSTG is
also among the areas consistently under-activated for readers with
Pursued largely independently from research on the neural basis of dyslexia compared to typical readers (Richlan et al., 2009), suggesting
social comprehension differences in autism, there is also a considerable that the autistic group may have been relying more than the non-autistic
body of empirical research on the neural basis of general language group on phonology as a route to semantic processing. Altogether, these
comprehension differences in autism (Harris et al., 2006; Kana et al., findings raise the possibility that the typical neural networks underlying
2006; Knaus et al., 2008; Phan et al., 2021; Sahyoun et al., 2010), with semantics are at least partly distinct from the neural substrates involved
fewer studies focusing on neural differences in figurative language such in semantic processing for individuals with autism.
as metaphor (Chouinard et al., 2017) or puns (Kana and Wadsworth,
2012). The central research question of the current study is to determine 1.3. Social and figurative language processing
whether there is an interaction between processing social content and
processing figurative content. More specifically, it may be the case that, Social and figurative language use are generally folded into the label
for autistic adults, social processing is mediated by different neural re­ of language pragmatics, yet pragmatics extends beyond the simple
gions than in non-autistic adults, particularly when social information adjective-noun stimuli considered here to also include wider linguistic
occurs in figurative contexts. Such a finding would help clarify the na­ constructs such as sentence-level syntax, multi-sentence discourse, and
ture of differences in abstract language comprehension that have been prosody (Bach, 2004; Pylkkänen and McElree, 2006; Tager-Flusberg
reported for individuals with autism. et al., 2005). In an effort to distinguish the neuro-cognitive effects of
In terms of predictions for figurative compared to literal content, social and figurative content, we use relatively simple stimuli that we
studies have demonstrated a great deal of variation with regard to the can carefully control to orthogonalize the social and figurative factors.
neural loci identified, even at the level of the hemisphere (Brownell Indeed, disentangling the distinct contributions of correlated factors is a
et al., 1990; Eviatar and Just, 2006; Gagnon et al., 2003). One source of well-established approach in the cognitive neuroscience of language.
this variation may relate to the different types of non-literal language For a related example, cognitive and computational models of reading
used in different studies. This possibility is reinforced by two different are broken down into visual (orthographic), auditory (phonological),
large-scale meta-analyses of functional neuroimaging studies in and concept (semantic) components (Harm and Seidenberg, 2004; Plaut
non-autistic participants processing figurative (non-literal) language. et al., 1996; Seidenberg and McClelland, 1989). Even though orthog­
One (Bohrn et al., 2012) revealed that areas of the semantic network raphy and phonology are highly correlated in alphabetic languages,
such as the medial and lateral parietal and lateral temporal areas are examining their separate mechanisms has been critically helpful in
consistently activated for processing literal language, while areas acti­ teasing apart the corresponding neural systems involved in reading
vated for different kinds of non-literal language such as metaphor, irony, (Graves et al., 2010; Taylor et al., 2013). Here we take a similar
and sarcasm were more heterogenous, and in turn less consistent, across approach to neural systems related to pragmatic language in autism by
the cortex. There were, however, a few areas activated for figurative carefully constructing and selecting stimuli to test for potentially distinct
compared to literal language processing that included the bilateral neural processing of its social and figurative aspects.
inferior frontal and anterior temporal cortices. The other meta-analysis In studying the possibly separate and intersecting neural effects of
(Rapp et al., 2012) similarly combined the various types of non-literal varying social and figurative content, we note that there have been
language and contrasted them with literal language. Their findings relatively few studies examining neural responses to manipulating both
were similar to Bohrn et al. in revealing bilateral (but greater on the left) of these factors. One of the few related studies of which we are aware
activation across studies for figurative compared to literal language in used photographic stimuli displaying social situations in which speech

2
W.W. Graves et al. Neuropsychologia 171 (2022) 108240

bubbles (as seen, for example, in comic books) contained either ironic We designed the current study to bring additional insight into how
statements or statements that were only sensible if interpreted idio­ social and figurative language domains are typically organized in the
matically (Wakusawa et al., 2007). These stimuli were presented in fMRI brain, and how they respond in a population with known but under­
alternating between blocks where the phrases were to be interpreted specified processing differences. Building on the theoretical and
based on the pictured situational context or interpreted literally. The empirical background described above, we predict anterior and poste­
idiom condition activated the vmPFC only in the social context. The rior midline and lateral temporo-parietal areas previously found to be
irony condition activated the dorso-medial PFC (dmPFC) and vmPFC in related to social and semantic cognition to differ in the group with
the social context but not the literal context, and the right ATL in both autism, particularly for areas such as the ATL and pMTG related to
contexts. This study suggests that activation in the dmPFC, vmPFC, and combining concepts in these domains. Maximal differences are expected
right ATL may relate to interactions between degrees of social and when processing phrases that are both social and figurative, as these
figurative context. domains are often challenging for individuals with autism.
Another relevant study considered predictions for separate neural
activations to a social compared to a semantic manipulation for the same 2. Material and methods
single words or (two-word) phrases. Specifically, Lin and colleagues
crossed amount of social content with a semantics-related factor such as 2.1. Participants
degree of either imageability (i.e., subjective ratings of the degree to
which words call to mind an image) or meaning plausibility. They found Interested participants who self-reported to be on the autism spec­
adjacent and partially overlapping activation for social and semantic trum (AUT group) were recruited by advertisement and word-of-mouth
factors in regions of bilateral inferior parietal lobule and pMTG, and contacts from communities in central and northern New Jersey. These
activation for social content alone in bilateral ATL (Lin et al., 2018, included disability service offices in colleges and universities. Non-
2020). These findings are relevant to the current study in that both use autistic comparison (NAC group) participants were recruited for the
short, carefully controlled phrases to enable manipulation of social study through advertisement in the local university community and the
content separately from a potentially related linguistic factor. Critically psychology research participant pool. All participants provided written,
however, the current study, aside from using English instead of Chinese informed consent as approved by the Rutgers University Institutional
phrases, manipulates the degree of figurativeness. In this way, we can Review Board (Rutgers IRB). Participants were compensated $15 per
investigate the effects of two factors that are both relevant to compre­ hour for behavior-only sessions and $30 per hour for MRI sessions.
hension difficulties in autism, socialness and figurativeness. The po­ Participants consisted of 41 individuals either with AUT (N = 19,
tential neural differences in responses to these factors among autistic mean age = 21.3) or without (N = 22, mean age = 21.3). Participants in
individuals have until now been largely unexplored. the NAC group met the following criteria: 1) no documented history of
psychiatric or neurologic disorders, 2) right-handed, 3) English learned
1.4. Current study as a primary language no later than five years old, 4) no metal embedded
in soft tissue or other contraindications to MRI, and 5) minimum full-
Issues with pragmatic language that involve aspects of higher-level, scale IQ score of 85 on the Wechsler Abbreviated Scale of Intelligence
inferential meaning are typically studied using longer stimuli, as in – II, as estimated by the vocabulary and matrix reasoning sub-tests
passage comprehension. However, use of such stimuli in the functional (FSIQ-2 from the WASI-II; Wechsler, 2011).
magnetic resonance imaging environment can be challenging for a Participants in the AUT group met the same study criteria as the NAC
number of reasons. For example, small differences in perceptual span of group, except that they had a primary diagnosis of autism spectrum
stimuli, or uncontrolled differences in form or meaning at the individual disorder. Diagnostic status was confirmed with objective psychometric
word, sentence, or discourse level can lead to unexpected neural effects instruments and evaluation, including the Autism Diagnostic Observa­
and difficulties interpreting contrasts between conditions (Vigliocco tion Schedule – Second Edition (ADOS-2) Module 4 (Hus and Lord,
et al., 2012), as our own pilot investigations also revealed. The challenge 2014), administered either by a licensed clinical psychologist who was
then is to elicit social and figurative language processing in a way that research-reliable or a doctoral student supervised by such a clinician.
also allows for exerting experimental control over the factors being To verify that the groups were matched on the relevant factors
investigated. (summarized in Table 1), we performed t-tests of mean differences be­
We address this challenge by using two-word, adjective-nouns tween groups. Results did not reveal any reliable differences (all p > 0.1)
phrases. Studies using short phrases and even single words have been between the groups with regard to age (t (39) = 0.72) or FSIQ2 (t (39) =
shown to elicit differences in semantic processing between groups of 1.11). The verbal comprehension index (VCI) and perceptual reasoning
individuals with and without autism (Kamio et al., 2007; Norbury and index (PRI) were calculated for all participants but one in the AUT group
Nation, 2011). Specifically, we use two-word (adjective-noun) phrases who only completed the vocabulary and matrix reasoning subtests. The
rather than single-word or discourse-level stimuli. We chose this VCI (t (38) = 1.65) and PRI (t (38) = 1.32) also did not significantly
approach because single words are unlikely to provide sufficient context differ between groups (p > 0.1). Gender ratios were also found to not be
to enable separate manipulation of both social and figurative content. reliably different between groups by χ2 testing (χ2 = 0.40, p > 0.1).
Conversely, discourse-level stimuli afford less control, and the typical
length of these stimuli would make it difficult to include sufficient Table 1
numbers of stimuli during fMRI to be sensitive to the relatively subtle Characteristics of participants in the two study groups. Numbers in parentheses
manipulations used here. Moreover, making judgments to two-word are standard deviations. FSIQ2 was available for all participants and is based on
phrases is a well-established task we have used in previous studies for the vocabulary and matrix reasoning subtests of the WASI. VIQ and PIQ were
probing the neural basis of combinatorial semantic processing, or con­ available for all NAC participants and 18 participants in the AUT group. Groups
ceptual combination (Graves et al., 2010; Graves et al., 2013). Within did not reliably differ on any of these measures (ps > 0.1).
the more extensive semantic network identified in previous studies AUT NAC
(Binder et al., 2009), semantic processing of two-word phrases has been
N 19 22
particularly shown to activate areas of ventro-medial prefrontal cortex Gender 15 M, 4 F 19 M, 3 F
(vmPFC) and anterior temporal lobe (ATL), along with both left and Age 20.8 (1.9) 21.3 (2.3)
right-sided posterior middle temporal and angular gyri (pMTG and AG) FSIQ2 114.2 (12.8) 110.2 (10.4)
(Boylan et al., 2017; Graessner et al., 2021; Graves et al., 2010; VIQ 109.8 (12.4) 104.5 (8.2)
PIQ 111.3 (13.0) 106.2 (11.2)
Pylkkänen, 2019).

3
W.W. Graves et al. Neuropsychologia 171 (2022) 108240

2.2. Experiment stimuli and design or concreteness (for matching phrases on a general semantic factor, as
described below) using a 1–7 Likert-type scale. Examples of phrases not
The stimuli consisted of two-word, adjective-noun phrases. Each included for rating were given with anchor points illustrating that
phrase was rated as high or low on socialness and high and low on higher numbers corresponded to greater levels figurativeness, social­
figurativeness, and selected to create a fully crossed 2 × 2 design. Social ness, or concreteness. Data integrity was checked by comparing each
phrases were defined as using language that “relates to interactions of participant’s ratings to the mean rating from the remaining participants
people, groups of people, or society as well as reflections on interactions for the same items. Participants whose ratings were more than 2.5 SD
with other people.” Figurative phrases were defined as “departing from from the mean were excluded. This resulted in each phrase being rated
the usual or most basic sense of the words.” There were 48 phrases per by at least 27 participants.
cell (social figurative, social literal, nonsocial figurative, nonsocial Using these stimulus characteristics and ratings, we selected phrases
literal) for a total of 192 phrases. Characteristics of these phrases, that clearly and reliably differed in figurativeness and socialness across
including examples and relevant factors on which they are matched, are the four conditions. These conditions were also statistically matched for
shown in Table 2. surface characteristics (number of letters) along with factors related to
To generate these stimuli, we first conducted a norming study. Using familiarity (frequency of adjective, noun, and whole-phrase occurrence)
the SUBTLEX-US database, we considered words with concreteness and semantics (concreteness). Selection of phrase stimuli to meet these
ratings (Brysbaert et al., 2014) and whose dominant part of speech was criteria was performed using the Stochastic Optimization of Stimuli
either adjective or noun. We then selected the 100 most frequent ad­ software (Armstrong et al., 2012). Characteristics for the final set of 192
jectives and 400 most frequent nouns, and paired them in order to derive phrases are summarized in Table 2. Phrases that were bimodally
40,000 possible pairs. Initial winnowing of this list was performed by distributed to reliably differ on socialness (t (190) = 34.1, p < 0.001) did
cross-referencing with the Google bigrams database (Brants and Franz, not reliably differ in figurativeness (t (190) = 0.7, p > 0.5), and phrases
2006). Here bigrams refers to two-word combinations, from which we bimodally distributed for figurativeness (t (190) = 36.9, p < 0.001) did
selected only those from our list that appeared more than 1000 times on not reliably differ in socialness (t (190) = 0.6, p > 0.5). The other factors
the Internet by the year 2006. This resulted in 5000 pairs, which were being controlled for (Table 2, below the double line) did not reliably
manually inspected for likelihood of cleanly fitting into one of the cells differ across conditions (all ps > 0.1). The full set of stimuli and their
in our 2 × 2 design. Use of this database-driven approach made clear relevant characteristics are provided as a comma-separated-values file in
that meaningful figurative phrases are unlikely to emerge by this supplementary material.
method. We therefore supplemented figurative phrases by consulting
the free online Farlex Idioms and Slang Dictionary (2017). This yielded a 2.3. Image acquisition
set of 1084 candidate phrases that were submitted for independent
rating as described next. All brain scans were acquired at the Rutgers University Brain Im­
To select the final set of phrase stimuli for the fMRI experiment, we aging Center (Newark, NJ) using a 3 T S TIM Trio with a 12-channel
obtained ratings from a separate group of participants recruited using head coil. Following acquisition of initial localizer slices in each of the
Amazon Mechanical Turk. To ensure high-quality ratings, we restricted axial, sagittal, and coronal planes, a high-resolution 3D anatomical
participants to those in the United States who had a record of completing (MPRAGE) scan was acquired using the following parameters: TR =
10,000 other assignments with a 99% acceptance rate from other re­ 1900 ms, TE = 2.52 ms, flip angle 9◦ , image matrix = 256 × 256, field of
questors. Potential raters were told to accept the assignment only if they view = 256 mm, number of slices = 176, slice thickness = 1 mm, voxel
considered English their native or first language. Raters then completed size = 1 mm isotropic. Next fMRI data were acquired as T2*-weighted
a short demographic questionnaire, and anyone reporting an age below gradient-echo single-shot echo planar images (EPI) with the following
18 years or a language other than English as their first/native language parameters: TR = 2000 ms, TE = 25 ms, flip angle = 77◦ , image matrix
was excluded from providing ratings. Participants gave written = 64 × 64, field of view = 195 mm, number of slices = 35, slice
informed consent in accordance with the Rutgers IRB and were thickness = 3.0 mm with interleaved acquisition, voxel size = 3 mm
compensated $4.00 per half hour. Each participant was given a list of isotropic. These were acquired during four runs of 3 min 28 s each,
360 unique phrases and asked to rate them on figurativeness, socialness, during which participants performed the task described next. Stimuli
were viewed through a prism mirror mounted to the head coil, and
finger-press responses were recorded using an MRI-compatible button
Table 2
box.
Rows show each factor either being manipulated (social and figurative, above
the double line, with examples in parentheses) or being controlled (remaining
rows below the double line). Columns show each cell in the completely crossed 2.4. In-scanner task
2 × 2 factorial design. Cells contain means (with standard deviations) for each
value. The values for social, figurative, and concrete come from subjective rat­ As the experiment was meant to probe neural responses to orthog­
ings and are therefore unitless. Frequencies are log-transformed number of onally manipulated social and figurative content of phrases, we chose
occurrences. the familiarity judgment task as a way of prompting participants to
Social Nonsocial attend to the meaning of the phrases without directing them to focus
specifically on the social or figurative aspects of their meaning. We also
Figurative Literal Figurative Literal
(blue blood) (gentle (sharp taste) (sore assumed the groups would have roughly equal access to their own fa­
hug) throat) miliarity judgments, thereby eliciting phrase comprehension while
Social 4.95 (0.66) 4.96 (0.86) 2.15 (0.28) 1.90 (0.39)
minimizing the involvement of potentially domain-general performance
Figurative 5.73 (0.46) 2.75 (0.58) 5.62 (0.41) 2.54 (0.57) differences such as attention or time-on-task.
Concrete 3.98 (1.74) 4.41 (1.54) 3.87 (1.43) 4.06 (1.43) Participants were instructed to press a button on a handheld MRI
Number of 9.77 (2.41) 10.77 9.98 (2.27) 10.04 compatible two-button box to indicate whether each phrase they were
letters (2.73) (2.32)
seeing was familiar or unfamiliar. Familiar phrases were defined in the
Phrase 10.52 (1.62) 10.64 10.56 (1.81) 10.87
frequency (1.81) (1.59) instructions as ones they “use or hear often”. Following an initial 8 s of
Adjective 7.94 (1.70) 8.06 (1.69) 7.74 (2.03) 7.71 (1.17) fixation at the beginning of each run to allow EPI saturation effects to
frequency stabilize, either a fixation or a trial would appear. During trials, the
Noun 8.05 (1.50) 8.56 (1.76) 8.05 (1.67) 8.08 (1.76) stimulus (phrase) was displayed for 1.5 s, followed by a fixation cross for
frequency
0.5 s. The screen would then return to a fixation cross for a randomly

4
W.W. Graves et al. Neuropsychologia 171 (2022) 108240

jittered inter-trial interval (mean: 2.5 s). Reaction time (RT) was thresholds were set to p < 0.005, and mapwise cluster corrected to p
recorded as the time from stimulus onset to a button press. < 0.05.
A potential concern about using familiarity judgment as the in-
scanner task is that it does not explicitly probe the degree to which 3. Results
participants comprehended the phrases in social or figurative terms. To
address this, a subset of participants in the AUT (N = 14) and NAC (N = 3.1. Behavioral
15) groups rated the 192 phrases seen during scanning in terms of so­
cial/nonsocial and figurative/literal content. This outside-the-scanner Task responses during scanning were not reliably different between
task followed the same procedure as described above for Amazon Me­ the groups. Participants in the AUT group had slightly higher rates of
chanical Turk. judging the phrases to be familiar (81%) compared to the NAC group
(78%), but this was not a statistically significant difference (F1,39 = 1.16,
2.5. Data analysis p = 0.29). The 2 × 2 × 2 ANOVAS (Fig. 1A) did reveal that during
scanning, literal phrases were judged to be familiar (F1,39 = 215.7, p <
Analysis of behavioral data from the in-scanner task was conducted 0.001) more often (85.9% familiar) than were figurative phrases (72.3%
as a 2 × 2 (within-subjects for social and figurative) x 2 (between familiar). However, there was no main effect of group, and these ratings
groups) ANOVA to test for potential group differences in rate of did not interact with group status. Similarly, social phrases were judged
“familiar” responses, with subjects as the random factor. The same to be familiar more often (84.9%) than were nonsocial phrases (73.3%,
ANOVA setup was also performed with RT as the dependent variable. F1,39 = 206.9, p < 0.001). Here too, there was no main effect of group,
Functional neuroimaging data were analyzed using AFNI software nor was there an interaction with group status. There was, however, an
(Cox, 1996; Cox and Hyde, 1997) within custom Linux shell scripts interaction between the social and figurative factors (F1,39 = 34.3, p <
(tc-shell under Ubuntu 18.04). The reconstructed data were converted 0.001), such that phrases that were both nonsocial and figurative were
from DICOM to AFNI HEAD/BRIK format. Motion correction and judged on average to be less familiar (64% familiar) than were the other
anatomical alignment to the functional data were performed using the three conditions (ranging from 81% familiar for social/figurative
AFNI script, align_epi_anat.py (Saad et al., 2009). TRs with motion phrases to 89% familiar for social/literal phrases, with nonsocial/literal
displacement greater than 10% from the mean were determined using phrases falling in between at 83% familiar).
3dToutcount and censored from further analysis. A nuisance regressor Regarding response times for making these judgments, the AUT
was also created to model non-physiological signal fluctuation using the group averaged 990 ms (ms), while the NAC group averaged 970 ms.
first principal component (calculated using the AFNI program 3 dpc) of These means were also not reliably different between the groups (F1,39
the signal from the lateral ventricles. = 0.29, p = 0.60). Overall, the RT data followed a pattern similar to the
To ensure that any neural differences found between groups could one described above for the rate at which phrases were judged to be
not be due to differences in amount of motion during fMRI, we per­ familiar. That is, participants were faster overall to make judgments to
formed motion correction and analyzed the resulting motion parameters literal phrases (961 ms) than to figurative ones (994 ms). This main
as part of the standard AFNI processing pipeline (Saad et al., 2009). effect of figurativeness on RT was statistically significant (F1,39 = 26.3,
Specifically, six motion parameters were calculated when registering p < 0.001), but did not interact with group (Fig. 1B). Similarly, social
each image volume in the fMRI time series to a central volume: Three phrases were judged more quickly on average (958 ms) than were
directions of rotation, and three directions of displacement. The tem­ nonsocial phrases (997 ms). This main effect of phrase socialness was
poral derivative of this 6-element vector for a given time point relative statistically reliable (F1,39 = 58.4, p < 0.001), but did not interact with
to the previous one was calculated to determine the amount of motion group. Unlike the pattern above for rate of familiarity, the RT for making
from one time point to the next. The derivative vectors were then those ratings did not show an interaction between levels of socialness
combined into a single term by taking the Euclidean (L2) norm, yielding and figurativeness.
one motion value per time point. The mean across time points was the To determine whether the groups were comprehending the phrases
mean motion for each participant. As these values were quite small differently in terms of social or figurative content, a subset (71%, AUT N
overall (AUT: 0.08 mm, NAC: 0.09 mm), and no participant in either = 14, NAC N = 15) of participants who underwent scanning also
group exceeded 0.22 mm of displacement from one frame to the next, we completed post-scan ratings of the phrases on Likert-type scales of
did not apply a cutoff threshold for participant inclusion. Motion did not figurative/literal and social/nonsocial. There was a significant interac­
significantly differ between groups (t39 = 0.74, p = 0.47). tion of figurative/literal ratings with group (Fig. 2), such that the AUT
First-level regression analysis of individual participant data was group rated the figurative phrases as more literal than the NAC group
performed using 3dDeconvolve, with seven nuisance covariates and six did (F1,29 = 9.06, p = 0.003). There were no group differences in the
explanatory variables of interest. Of the seven nuisance covariates, 6 social/nonsocial ratings (F1,29 = 0.01, p = 0.937).
were motion parameters, 3 for rotation in the pitch, roll, and yaw di­
rections, and 3 for displacement in those directions, along with the 3.2. Functional neuroimaging
physiological noise regressor from the lateral ventricles as described
above. The six explanatory variables were 4 for the cells in the 2 × 2 We tested fMRI data for a main effect of social content by contrasting
design: social/figurative, social/literal, nonsocial/figurative, nonsocial/ neural responses to social minus nonsocial phrases. Both groups showed
literal, along with a term modeling any null responses, and a continuous, reliable activation for this contrast. Social phrases showed activation
z-score normalized regressor for trial RT. The resulting output images relative to nonsocial phrases primarily in midline structures, including
were then nonlinearly warped to Talairach space (the TT_N27 brain bilateral PCC, precuneus, and dmPFC (Fig. 3, Table 3). Additionally,
space; Lancaster et al., 2000) using diffeomorphic techniques in only the NAC group showed activation for social phrases in the vmPFC.
3dQwarp and smoothed with a 5 mm FWHM kernel using the AFNI Although this apparent difference between groups did not reach statis­
program 3dmerge. tical significance in the whole brain analysis, a plot of the relative ac­
Second-level analysis of the NAC and AUT groups was performed tivations in the vmPFC (Fig. 3C) revealed a numerically larger difference
with the AFNI program 3dttest++ to compare first-level contrasts be­ between activation for social and nonsocial phrases for the NAC
tween groups at each voxel. The “-Clustsim” option was used to imple­ compared to the AUT group.
ment current best-practice approaches to map-wise cluster corrections We also tested for main effects in the contrast between figurative and
for multiple comparisons that also take into account the smoothness of literal phrases. The NAC group showed activation for literal phrases
the data (Cox et al., 2017). For all contrasts reported, voxel-level primarily in the bilateral precuneus and PCC, and right pMTG (Fig. 4A,

5
W.W. Graves et al. Neuropsychologia 171 (2022) 108240

Fig. 1. Performance on the in-scanner task. There were no differences between groups in terms of proportion of phrases judged to be familiar (A) or response times
(RT) to those make those judgments (B). Significant effects of phrase type are described in the text of the Results section. AUT: autism, NAC: non-autism comparison.

Table 3). No brain areas showed significant activation for this contrast in
the AUT group (Fig. 4B, Table 3). To illustrate the directions of activa­
tion for the conditions making up this contrast, BOLD signal parameter
estimates are plotted separately for the figurative and literal conditions
in each group (Fig. 4C). As shown in the plot, the difference between the
figurative and literal conditions is greater for the NAC group than the
AUT group. Although this apparent group difference does not reach
statistical significance when correcting for multiple comparisons at the
whole brain level, it is notable that activation differences for figurative
compared to literal phrases appear robustly in the NAC group and not at
all in the AUT group. The full 2 × 2 × 2 ANOVA for the neural results did
not yield any significant between-group effects.
Rounding out the planned comparisons, we considered each of the
four cells in the 2 × 2 design. Reliable group differences were found for
activation in the social-figurative cell in the regions of the bilateral mid-
cingulate extending to supplementary motor area, and the right ATL
(Fig. 5A, Table 4). The ATL activation was of particular interest due to its
known role in semantics (as summarized in the Introduction). Plotting
the activation in this region for each group revealed deactivation for the
NAC relative to AUT group (Fig. 5B).
To better understand the nature of the right ATL group result, we
performed an exploratory follow-up analysis of social/nonsocial and
figurative/literal main effects in which we kept the voxel-level threshold
at p < 0.005, but relaxed the extent threshold so that it was no longer
mapwise corrected to p < 0.05. This revealed a trend toward a main
effect of socialness in which social phrases showed numerically greater
activation than non-social phrases in the left ATL for the NAC group only
(Fig. 6). No such trend was found for main effects of figurativeness in the
left ATL.

4. Discussion

The overall aim of this study was to determine the neural basis of
Fig. 2. Post-scan ratings in terms of figurative/literal on a Likert-type scale for differences in abstract language processing among individuals with
the 192 phrases seen during scanning. The ratings were obtained from a subset autism, separating components related to degree of figurative or social
of the scanned participants in the autism (AUT, N = 14) and non-autism language content. Main effects of the figurative-literal contrast were
comparison (NAC, N = 15) groups. The AUT group rated figurative phrases found in several semantics-related areas such as PCC and pMTG.
as significantly more literal than did the NAC group.

6
W.W. Graves et al. Neuropsychologia 171 (2022) 108240

Table 3
Peak coordinates for the main effects of the contrast between literal and figu­
rative phrases (upper rows), and the contrast between social and nonsocial
phrases (lower rows), separated by group. L = left, R = right.
Location of extreme point Cluster size X Y Z z-
(mm3) score

Social > nonsocial


NAC group
Bilateral medial prefrontal 9260
L medial superior frontal gyrus 0 51 37 4.56
L gyrus rectus 0 57 − 6 4.45
Bilateral medial parietal 6719
L posterior cingulate − 5 − 51 26 5.28
AUT group
Bilateral medial parietal 6002
L posterior cingulate 0 − 50 27 4.62
Bilateral medial prefrontal 4156
R medial superior frontal gyrus 2 59 17 4.20
Literal > figurative
NAC group
Bilateral medial parieto- 30,216
occipital
R precuneus 7 − 54 32 5.14
L cuneus − 1 − 84 31 4.34
R superior parietal lobule 20 − 64 54 4.17
L superior parietal lobule − 28 − 74 37 4.12
L occipital pole 0 − 89 − 6 4.06
Right posterior temporal and 6577
inferior parietal
R posterior middle temporal 57 − 49 5 4.87
gyrus
AUT group
None
Figurative > literal
None
R medial superior frontal gyrus 2 43 42 3.77

vmPFC, was activated in the NAC group only. Critically, the social and
figurative cell of the 2 × 2 factorial design that was expected to show the
greatest group differences indeed showed a reliable difference in group
activation in the right ATL.

4.1. Social-nonsocial contrast

As social impairments constitute a critical criterion for AUT diag­


nosis, we expected neural differences in the processing of social
compared to non-social phrases. While there were indeed numerical
differences between the groups in results of this social-nonsocial
contrast in the vmPFC, it is also striking that both groups showed acti­
vation in areas of the PCC and dmPFC that have been found in previous
studies of social cognition (Denny et al., 2012). This pattern of results,
along with the fact that both groups endorsed similar proportions of the
social and nonsocial phrases as being familiar, suggests that group dif­
ferences due to overall social content may be minimal for these rela­
tively simple stimuli.
Of note, however, is the marginally significant group differences for
the social-nonsocial contrast in the vmPFC. While the vmPFC is
considered part of the brain network for social cognition, recent studies
have also aimed to determine whether different parts of that network
play somewhat different roles in social cognition. For example, a recent
Fig. 3. Social minus nonsocial contrast for the NAC group (A), AUT group (B),
and parameter estimates in vmPFC (C) illustrating the relative effects in the
large-scale meta-analysis of neural components of social cognition found
contrast for both groups. the vmPFC to be related to verbally mediated judgments of peoples
beliefs or traits (Schurz et al., 2021). A separate line of research with
brain lesion-deficit analyses has linked the vmPFC to valuation of
Strikingly, this pattern was detected for the NAC group only. This result
reward-related signals, potentially in relation to signals from other parts
aligns with the findings that participants in the AUT group rated figu­
of the body, such as signals that are detected from autonomic skin
rative phrases as more literal than did the NAC group. Main effects of the
conductance responses (Bechara et al., 1996). This interpretation is
social-nonsocial contrast, however, were found for both groups in
compatible with previous lesion-deficit studies showing that damage to
bilateral PCC and dmPFC, consistent with previous studies of social
the vmPFC can lead to atypical social decisions (Koenigs and Tranel,
stimulus processing. Another region expected for this contrast, the
2007; Koenigs et al., 2007). There is also a separate line of research on

7
W.W. Graves et al. Neuropsychologia 171 (2022) 108240

Fig. 5. Significant group differences in the social, figurative condition in the


right ATL (A), along with plots of size and direction of effects (B).

Table 4
Peak coordinates of the between-group activation differences for social, figu­
rative phrases.
Fig. 4. Figurative minus literal contrast for the NAC group (A), AUT group (B), Location of extreme point Cluster size X Y Z z-
and parameter estimates in right pMTG (C) illustrating the relative effects in the (mm3) score
contrast for both groups. AUT > NAC
Bilateral cingulate and 3190
aspects of AUT related to atypical awareness or perception of one’s own supplementary motor
L middle cingulate − 11 3 36 4.32
body (Schauder et al., 2015). To the extent that autonomic or other R supplementary motor area 5 − 9 50 4.05
sensory signals are integrated with ongoing cognitive processes during R anterior temporal lobe 2199
tasks such as the one used here, it is relevant to note that the vmPFC has R temporal pole 34 8 − 23 4.89
been hypothesized to be related to multisensory integration issues in
autism (Martínez-Sanchis, 2014).

8
W.W. Graves et al. Neuropsychologia 171 (2022) 108240

To distinguish the roles of the ATL and AG in conceptual combina­


tion, one theory emphasizes the distinction between taxonomic and
thematic relations (Mirman et al., 2017). Taxonomic relations share
category-related features (e.g., dog—bear, or coffee—tea), whereas
thematic relations co-occur in similar events, contexts, or scenarios (e.g.,
dog—leash, coffee—spoon). Reviews have summarized evidence
pointing to the ATL as more associated with taxonomic relations, and the
AG as more associated with thematic relations (Davis and Yee, 2019;
Mirman et al., 2017). For example, survivors of stable, focal lesions to
the left ATL exhibit deficits primarily in processing taxonomically
related items, while participants with lesions to the left AG have deficits
in processing thematically related items (Schwartz et al., 2011). In terms
of the current study, stimuli consisted of adjective-noun phrases that
were not selected in terms of taxonomic or thematic relations. Of the
two, however, adjective-noun phrases would seem to be more analogous
to thematic relations, as the constituent words combine to form an
overall concept where the first word is descriptive and has no require­
ment to be taxonomically related to the second. Therefore, it may be that
the right AG activation found here relates to processing phrases con­
taining especially clear, meaningful descriptions.
Of particular importance to the aims of the current study is the lack
of any significant differences between figurative and literal phrases in
the AUT group. While the magnitude of this contrast did not formally
survive whole-brain correction for multiple comparisons, the fact that
the NAC group activated numerous expected areas for this contrast and
the AUT group significantly activated none is nonetheless striking.
While we are not aware of functional neuroimaging studies specifically
examining responses in participants with AUT to figurative compared to
literal phrases, two previous studies are relevant. One fMRI study con­
Fig. 6. Exploratory follow-up analysis showing trends in the left ATL (A) for
trasted neural responses in AUT and NAC participants to metaphor
both groups toward a main effect of socialness (B). Although it did not survive
processing. They found greater activation in the AUT group in visual
extent thresholding, there was a clear trend toward greater activation for social
compared to non-social phrases in the NAC group compared to the AUT group. association cortex, thalamus, and cerebellum (Chouinard et al., 2017). A
second fMRI study compared neural responses to verbal humor (puns)
between AUT and NAC groups, finding greater activation in the AUT
4.2. Figurative-literal contrast
group for puns in the bilateral inferior parietal lobule, right hippocam­
pus and middle frontal gyrus (Kana and Wadsworth, 2012). Although
The overall pattern of results for this contrast was that the NAC group
these studies did find activation for their groups of AUT participants in
showed greater activation for the literal condition in the figurative
their respective non-literal language conditions, they were in entirely
minus literal contrast in the right pMTG, extending inferiorly into the
non-overlapping areas, making it difficult to draw specific
inferior temporal gyrus and superiorly into the AG, along with bilateral
between-study conclusions about what brain areas to expect to be
activation of the PCC. These results largely replicate those of a previous
activated. A salient potential source of this variability could be the de­
study from one of us (WG). In that study, straightforwardly meaningful
gree of matching between groups. Participants in the AUT group from
two-word phrases (e.g., “ski jacket”) were contrasted with the reverse of
Kana and Wadsworth (2012) had significantly lower verbal IQ overall
each phrase (e.g., “jacket ski”) that were independently judged to be
compared to the NAC group (103.7 compared to 114.5), while Choui­
largely meaningless. If the literal phrases in the current study are
nard et al. (2017) matched the groups in their study for non-verbal IQ
considered to be more straightforwardly meaningful than the figurative
but did not report verbal IQ. We suggest that careful matching of groups
phrases (e.g., “cold air” compared to “cold turkey”), then the current
for both verbal and non-verbal IQ, as in the current study, may help
study can be seen as at least partially replicating our previous finding in
increase consistency of findings by reducing uncontrolled variability.
which more meaningful phrases activated the right pMTG, extending to
Relative to neural activation studies, figurative language processing
AG, and bilateral PCC (Graves et al., 2010). Similar findings have been
in AUT has been studied more extensively in behavioral studies. Such
reported for conventional (easily interpretable) compared to novel (less
studies were compiled in a recent meta-analysis, which concluded that
interpretable) metaphors in German noun-noun phrases that also acti­
figurative language processing is indeed atypical in AUT (Kalandadze
vated the bilateral PCC, right pMTG and AG (Forgács et al., 2012).
et al., 2018). Relatedly, participants with AUT have been shown to have
While the above pattern of straightforward, meaningful two-word
difficulty calling to mind the multiple meanings of homographs as a
phrases activating the right pMTG and bilateral PCC may be repli­
function of context (e.g., “In her eye/dress, there was a big tear.”; Happé,
cable, other studies report the left ATL to be prominently involved in
1997). While such issues with lexical ambiguity are not always found in
combining verbal concepts (Baron and Osherson, 2011; Baron et al.,
AUT groups (Norbury, 2005), focusing on a single meaning, rather than
2010; Bemis and Pylkkänen, 2011; Li and Pylkkänen, 2021; Pylkkänen,
modulating meaning activation by context, is broadly consistent with
2019; Westerlund and Pylkkänen, 2014). Studies have also found both
the “weak coherence” theory of cognition in autism (Happé and Frith,
the left ATL and bilateral AG to be modulated by conceptual combina­
2006). This theory would presumably predict retrieval of a single,
tion (Boylan et al., 2017), while still others have reported activation for
usually literal, meaning regardless of small differences in linguistic
meaningful two-word conceptual combination in the AG bilaterally but
context. In the current adjective-noun phrase task, this may have led the
not ATL (Graessner et al., 2021; Lin et al., 2020). These varied findings
AUT group to treat literal and figurative phrases in much the same way,
leave open the question of the relative contributions of the ATL and AG
thereby attenuating the neural differences between them. Indeed, rat­
to conceptual combination in minimal phrase units such as those used
ings from a large subset of participants in the current study showed that
here.
the AUT group interpreted the figurative phrases as more being literal

9
W.W. Graves et al. Neuropsychologia 171 (2022) 108240

than did the NAC group. Further support for the interpretation that greater activation of the
Before turning to the between-groups result, we note that the within- right ATL for social-figurative phrases in the AUT group represents
group results for the social-nonsocial contrast and the NAC result for the atypical recruitment of semantics comes from an exploratory follow-up
figurative-literal contrast appear to partially overlap the putative default analysis. By keeping the voxel-level threshold constant but relaxing the
mode network, particularly in the PCC. The default mode network is so spatial extent threshold, we identified a set of voxels in the left ATL
called because it is more active at rest than during task performance. (Fig. 6A) that showed a trend toward greater activation for social
Whether the function of that network relates to particular cognitive compared to non-social phrases in the NAC group only (Fig. 6B). This is
processes or a metabolically determined “baseline activation” is highly consistent with previous studies showing neurotypical left ATL
controversial (Buckner et al., 2008; Raichle, 2015; Smallwood et al., involvement in social semantics (Lin et al., 2018, 2020; Simmons et al.,
2021). As there is no true baseline in fMRI (Gusnard and Raichle, 2001), 2010). While this follow-up analysis was strictly exploratory, it lends
studies often report inter-trial intervals or blocks of no task as an implicit qualitative support to the interpretation that the AUT group atypically
“resting” baseline. While we have followed that convention in the bar recruited ATL nodes of the semantic network. The novel finding from the
graphs to illustrate the relative pattern of activation, the activation maps current study is that processing for social concepts in the ATL shifts
themselves are based on direct contrasts between conditions. For rightward in the AUT group, particularly when those concepts also
example, Fig. 3 shows activation greater for social than nonsocial, and contain figurative content.
Fig. 4 shows activation greater for literal than figurative. Which side of Another, non-exclusive interpretation of the greater activation of the
these contrasts should a priori be considered baseline or default is not right ATL by the AUT group for the social/figurative phrases relates to
obvious. Instead, we have focused on the separate but related question the putative role of the ATL in processing taxonomic, rather than the­
of what kinds of cognitive processes are typically associated with acti­ matic, relations. While the set of phrase stimuli used here were not
vations in this set of brain areas. There is a large body of evidence selected in terms of taxonomic or thematic relations, a preference for
showing that these areas process semantics (Binder et al., 2009; Lanzoni taxonomic interpretations, whether implicit or explicit, could poten­
et al., 2020; Mattheiss et al., 2018; Murphy et al., 2018; Wirth et al., tially explain the greater recruitment of the ATL by the AUT group.
2011), and our stimuli were designed to modulate processing within the Indeed, directly testing this possibility could be a fruitful direction for
semantic system. Viewed this way, the activation patterns shown in future study, as we are aware of no existing studies to specifically
Figs. 3 and 4 comport with the interpretation that the stimuli and compare neural responses to taxonomic and thematic processing across
experimental design did indeed succeed in modulating the target neural AUT and NAC groups.
and cognitive systems.
4.4. Potential limitations
4.3. Social and figurative phrases
There is an inherent trade-off between careful sample selection and
Greater activation in the AUT group for phrases that were both social generalization to wider populations. In this the current study is no
and figurative was found in a cluster spanning the mid-cingulate cortex exception. Testing a sample where most participants were un­
and supplementary motor area (SMA), along with a separate cluster in dergraduates helped to ensure that participants would be generally
the right ATL. The mid-cingulate and SMA are functionally associated familiar with the kinds of tasks they would be asked to perform for the
with a larger network of fronto-parietal regions found to activate across study, as well as more generally being able to schedule and keep ap­
domains for tasks requiring attention to external stimuli relative to rest pointments, and tolerate the MRI environment for an extended time.
(Duncan, 2010; Fedorenko et al., 2013). While such tasks have been While these abilities were necessary for study completion, our choice of
found to include those relevant to social cognition, such as making social participants also means that the results may not generalize to samples
preference decisions (Chen et al., 2010), the broadly domain-general who could not perform these tasks. Conversely, most participants in both
nature of responses found in these regions makes it unlikely to be spe­ groups tested here seem to be doing well in college overall, at least in
cific to the social and figurative stimuli considered here. Additionally, terms of mainlining successful enrollment. This raises the possibility that
the proportion of phrases judged to be familiar, and the RT for those the greater activation seen in the AUT group in the right ATL for the
responses, were not reliably different between groups. That plus the fact social/figurative condition could relate to the process of managing to
that trial-level RT was included in the fMRI analyses suggests that the deal with a world that is not particularly attuned to neuro-diversity.
results do not simply reflect domain-general effects such as time-on-task, A second concern that arises with many studies involving special
executive functioning, or working memory demands. populations is sample size. While it is true that more participants is
The social and figurative condition was the cell of our 2 × 2 factorial generally better, note that the current group of participants with autism
design predicted to differ most between the groups. By placing demands (N = 19) and the overall sample size (N = 41) is larger than the two most
simultaneously on social and linguistic comprehension, we expected relevant functional neuroimaging studies of non-literal language in AUT
brain response in this condition to reveal areas where the AUT group that we were able to identify in the literature, with Ns of 16 (Kana and
was processing these stimuli differently from the NAC group, presum­ Wadsworth, 2012) and 12 (Chouinard et al., 2017) in their AUT groups.
ably reflecting a difference in social and semantic networks. Greater To address the related concern of potential false-positive results, we took
activation in the AUT group for the right rather than left ATL is care to use current best practices to minimize potential false-positives in
consistent with this expectation. Meta-analyses of semantic processing our whole-brain analyses (Cox et al., 2017; Eklund et al., 2016). This
have found the left more than right ATL to be typically activated (Visser may have come with the trade-off of limiting sensitivity to the poten­
et al., 2010; Wang et al., 2010), particularly for tasks involving language tially subtle differences that might be expected between groups that are
stimuli (Binder et al., 2009; Rice et al., 2015). The left/right distinction as well matched as those in the current study. Ultimately, however, we
in these meta-analyses is relative, however, as activation in both ATLs acknowledge that this relatively limited sample size warrants caution in
was found, but with greater activation in left ATL (Binder et al., 2009; interpreting how generalizable the findings may be to new samples.
Rice et al., 2015; Visser et al., 2010). Only one of these meta-analyses Additionally, we choose to use adjective-noun pairs as stimuli (full
separately analyzed neural responses to social stimuli. They found set provided as supplementary material) instead of more linguistically
bilateral ATL activation across studies with stimuli aimed at social complex sentences or even discourse-level materials. In doing so, we
concepts (Rice et al., 2015). Therefore, activation of the right ATL, emphasized experimental control at the necessary expense of ecological
rather than left or bilateral, by the AUT group for social and figurative richness. It is therefore unknown at this point whether the patterns seen
phrases suggests that this group recruited key areas of the semantic in the current results would obtain with more complex or less tightly
processing network in a relatively atypical fashion. controlled stimuli.

10
W.W. Graves et al. Neuropsychologia 171 (2022) 108240

Finally, while we made every effort to ensure that levels of social and research assistants who helped with recruitment and data collection.
figurative content were orthogonally manipulated in our stimuli, this Funding for this study was provided by a Rutgers University Busch
does not change the fact that these types of content typically co-occur in Biomedical Grant to MRL and WWG.
everyday language. Such co-occurrences may lead participants to make
automatic associations between social and figurative content. Viewed Appendix A. Supplementary data
another way, however, these possible automatic associations can actu­
ally be seen as a potential source of the observed effects. That is, stan­ Supplementary data to this article can be found online at https://doi.
dard conceptual networks activated by social or figurative concepts are org/10.1016/j.neuropsychologia.2022.108240.
expected to spread activation to other related social or figurative con­
cepts, as in classically conceived spreading activation networks References
(reviewed by Ober and Shenaut, 2006). However, in the current case the
AUT group was interpreting the figurative phrases more literally, and Armstrong, B.C., Watson, C.E., Plaut, D.C., 2012. SOS! an algorithm and software for the
stochastic optimization of stimuli. Behav. Res. Methods 44 (3), 675–705.
neurally activating areas not typically engaged in verbal semantic pro­ Bach, K., 2004. Pragmatics and the philosophy of language. In: Horn, L.R., Ward, G.
cessing. Viewing this as resulting from a difference in the spread of (Eds.), The Handbook of Pragmatics. John Wiley & Sons, Incorporated, pp. 463–487.
activation among nodes of a semantic network organized through a Baron, S.G., Osherson, D., 2011. Evidence for conceptual combination in the left anterior
temporal lobe. Neuroimage 55, 1847–1852.
lifetime of associations that may be subtly different in the AUT Baron, S.G., Thompson-Schill, S.L., Weber, M., Osherson, D., 2010. An early stage of
compared to NAC groups may be a helpful way to interpret the source of conceptual combination: superimposition of constituent concepts in left
the differences reported here. anterolateral temporal lobe. Cognit. Neurosci. 1 (1), 44–51.
Bechara, A., Tranel, D., Damasio, H., Damasio, A.R., 1996. Failure to respond
autonomically to anticipated future outcomes following damage to prefrontal cortex.
4.5. Conclusion Cerebr. Cortex 6 (2), 215–225.
Bemis, D.K., Pylkkänen, L., 2011. Simple composition: a Magnetoencephalography
investigation into the comprehension of minimal linguistic phrases. J. Neurosci. 31
This study addressed the question of whether the apparent diffi­ (8), 2801–2814.
culties individuals with AUT can have in processing pragmatic aspects of Binder, J.R., Desai, R.H., 2011. The neurobiology of semantic memory. Trends Cognit.
language, such as social and figurative content, is related to differences Sci. 15 (11), 527–536.
Binder, J.R., Desai, R.H., Graves, W.W., Conant, L.L., 2009. Where is the semantic
in activation of brain areas related to processing social and figurative
system? A critical review and meta-analysis of 120 functional neuroimaging studies.
content. Participants with AUT showed neural activations broadly Cerebr. Cortex 19, 2767–2796. https://doi.org/10.1093/cercor/bhp055.
similar to the NAC group when processing phrases in terms of overall Binder, J.R., Frost, J.A., Hammeke, T.A., Bellgowan, P.S.F., Rao, S.M., Cox, R.W., 1999.
social or figurative content. There were, however, clear trends toward Conceptual processing during the conscious resting state: a functional MRI study.
J. Cognit. Neurosci. 11 (1), 80–93.
differences in key areas of the social cognition network, where only the Bohrn, I.C., Altmann, U., Jacobs, A.M., 2012. Looking at the brains behind figurative
NAC participants activated vmPFC for more social phrases, and only the language—a quantitative meta-analysis of neuroimaging studies on metaphor,
NAC group showed differences in activation between figurative and idiom, and irony processing. Neuropsychologia 50 (11), 2669–2683.
Borghi, A.M., Binkofski, F., Castelfranchi, C., Cimatti, F., Scorolli, C., Tummolini, L.,
literal phrases. Critically, there was a significant difference in the right 2017. The challenge of abstract concepts. Psychol. Bull. 143 (3), 263.
ATL between the groups for phrases that were expected to be processed Boylan, C., Trueswell, J.C., Thompson-Schill, S.L., 2017. Relational vs. attributive
with maximal difference between groups – those that were both social interpretation of nominal compounds differentially engages angular gyrus and
anterior temporal lobe. Brain Lang. 169, 8–21.
and figurative. We suggest that the partially distinct sets of brain areas Brants, T., Franz, A., 2006. Web 1T 5-gram Version 1. Linguistic Data Consortium,
shown here to respond to social and figurative content are atypically Philadelphia.
organized, and that activation in the right ATL in particular may be Brownell, H.H., Simpson, T.L., Bihrle, A.M., Potter, H.H., Gardner, H., 1990.
Appreciation of metaphoric alternative word meanings by left and right brain-
related to some of the challenges individuals with AUT often report in damaged patients. Neuropsychologia 28 (4), 375–383.
processing social and figurative language. The greater activation in the Brysbaert, M., Warriner, A.B., Kuperman, V., 2014. Concreteness ratings for 40 thousand
AUT group for social/figurative phrases in the ATL also raises the pos­ generally known English word lemmas. Behav. Res. Methods 46 (3), 904–911.
Buckner, R.L., Andrews-Hanna, J.R., Schacter, D.L., 2008. The brain’s default network:
sibility that the AUT group may be processing the phrases in a taxonomic
anatomy, function, and relevance to disease. Ann. N. Y. Acad. Sci. 1124, 1–38.
fashion. An implication of this for future exploration is that school-age Chen, A.C., Welsh, R.C., Liberzon, I., Taylor, S.F., 2010. ‘Do I like this person?’A network
individuals with autism may be more readily taught social and figura­ analysis of midline cortex during a social preference task. Neuroimage 51 (2),
tive content if it is presented in a taxonomic rather than thematic 930–939.
Chouinard, B., Volden, J., Cribben, I., Cummine, J., 2017. Neurological evaluation of the
fashion. selection stage of metaphor comprehension in individuals with and without autism
spectrum disorder. Neuroscience 361, 19–33.
Credit author statement Cox, R.W., 1996. AFNI: software for analysis and visualization of functional magnetic
resonance neuroimages. Comput. Biomed. Res. 29, 162–173.
Cox, R.W., Chen, G., Glen, D.R., Reynolds, R.C., Taylor, P.A., 2017. FMRI clustering in
William W. Graves: Conceptualization, Methodology, Formal AFNI: false-positive rates redux. Brain Connect. 7 (3), 152–171. https://doi.org/
analysis, Resources, Writing – original draft, Writing – review & editing, 10.1089/brain.2016.0475.
Cox, R.W., Hyde, J.S., 1997. Software tools for analysis and visualization of fMRI data.
Visualization, Supervision, Project administration, Funding acquisition. NMR Biomed.: An International Journal Devoted to the Development and
Hillary Levinson: Conceptualization, Methodology, Formal analysis, Application of Magnetic Resonance In Vivo 10 (4-5), 171–178.
Investigation, Data curation, Writing – original draft, Project adminis­ Davis, C.P., Yee, E., 2019. Features, labels, space, and time: factors supporting taxonomic
relationships in the anterior temporal lobe and thematic relationships in the angular
tration. Linsah Coulanges: Investigation, Data curation, Writing – re­ gyrus. Language, Cognition and Neuroscience 34 (10), 1347–1357.
view & editing, Project administration. Shannon Cahalan: Denny, B.T., Kober, H., Wager, T.D., Ochsner, K.N., 2012. A meta-analysis of functional
Investigation, Data curation, Writing – review & editing, Project neuroimaging studies of self-and other judgments reveals a spatial gradient for
mentalizing in medial prefrontal cortex. J. Cognit. Neurosci. 24 (8), 1742–1752.
administration. Daniel Cruz: Investigation, Writing – review & editing.
Duncan, J., 2010. The multiple-demand (MD) system of the primate brain: mental
Catherine Sancimino: Investigation, Writing – review & editing. programs for intelligent behaviour. Trends Cognit. Sci. 14 (4), 172–179.
Vanessa H. Bal: Investigation, Writing – review & editing, Supervision. Eklund, A., Nichols, T.E., Knutsson, H., 2016. Cluster failure: why fMRI inferences for
Miriam Rosenberg-Lee: Conceptualization, Methodology, Writing – spatial extent have inflated false-positive rates. Proc. Natl. Acad. Sci. U.S.A. 113
(33), 7900–7905.
review & editing, Supervision, Project administration, Funding Eviatar, Z., Just, M.A., 2006. Brain correlates of discourse processing: an fMRI
acquisition. investigation of irony and conventional metaphor comprehension. Neuropsychologia
44 (12), 2348–2359.
Fedorenko, E., Duncan, J., Kanwisher, N., 2013. Broad domain generality in focal regions
Acknowledgements of frontal and parietal cortex. Proc. Natl. Acad. Sci. U.S.A. 110 (41), 16616–16621.

We gratefully acknowledge the study participants, and the numerous

11
W.W. Graves et al. Neuropsychologia 171 (2022) 108240

Forgács, B., Bohrn, I., Baudewig, J., Hofmann, M.J., Pléh, C., Jacobs, A.M., 2012. Neural Mattheiss, S.R., Levinson, H., Graves, W.W., 2018. Duality of function: activation for
correlates of combinatorial semantic processing of literal and figurative noun noun meaningless nonwords and semantic codes in the same brain areas. Cerebr. Cortex
compound words. Neuroimage 63 (3), 1432–1442. 28 (7), 2516–2524. https://doi.org/10.1093/cercor/bhy053.
Gaffrey, M.S., Kleinhans, N.M., Haist, F., Akshoomoff, N., Campbell, A., Courchesne, E., Mirman, D., Landrigan, J.-F., Britt, A.E., 2017. Taxonomic and thematic semantic
Müller, R.-A., 2007. A typical participation of visual cortex during word processing systems. Psychol. Bull. 143 (5), 499.
in autism: an fMRI study of semantic decision. Neuropsychologia 45, 1672–1684. Molenberghs, P., Johnson, H., Henry, J.D., Mattingley, J.B., 2016. Understanding the
Gagnon, L., Goulet, P., Giroux, F., Joanette, Y., 2003. Processing of metaphoric and non- minds of others: a neuroimaging meta-analysis. Neurosci. Biobehav. Rev. 65,
metaphoric alternative meanings of words after right-and left-hemispheric lesion. 276–291.
Brain Lang. 87 (2), 217–226. Murphy, C., Jefferies, E., Rueschemeyer, S.-A., Sormaz, M., Wang, H.-t., Margulies, D.S.,
Graessner, A., Zaccarella, E., Hartwigsen, G., 2021. Differential contributions of left- Smallwood, J., 2018. Distant from input: evidence of regions within the default
hemispheric language regions to basic semantic composition. Brain Struct. Funct. mode network supporting perceptually-decoupled and conceptually-guided
226 (2), 501–518. cognition. Neuroimage 171, 393–401.
Graves, W.W., Binder, J.R., Desai, R.H., Conant, L.L., Seidenberg, M.S., 2010a. Neural Norbury, C., 2005. Barking up the wrong tree? Lexical ambiguity resolution in children
correlates of implicit and explicit combinatorial semantic processing. Neuroimage with language impairments and autistic spectrum disorders. J. Exp. Child Psychol.
53, 638–646. https://doi.org/10.1016/j.neuroimage.2010.06.055. 90 (2), 142–171.
Graves, W.W., Binder, J.R., Seidenberg, M.S., 2013. Noun-noun combination: Norbury, C., Nation, K., 2011. Understanding variability in reading comprehension in
meaningfulness ratings and lexical statistics for 2,160 word pairs. Behav. Res. adolescents with autism spectrum disorders: interactions with language status and
Methods 45 (2), 463–469. decoding skill. Sci. Stud. Read. 15 (3), 191–210.
Graves, W.W., Grabowski, T.J., Mehta, S., Gupta, P., 2008. Left posterior superior Ober, B.A., Shenaut, G.K., 2006. Semantic memory. In: Handbook of Psycholinguistics.
temporal gyrus participates specifically in accessing lexical phonology. J. Cognit. Elsevier, pp. 403–453.
Neurosci. 20, 1698–1710. Olson, I.R., Plotzker, A., Ezzyat, Y., 2007. The enigmatic temporal pole: a review of
Groen, W.B., Zwiers, M.P., van der Gaag, R.-J., Buitelaar, J.K., 2008. The phenotype and findings on social and emotional processing. Brain 130, 1718–1731.
neural correlates of language in autism: an integrative review. Neurosci. Biobehav. Patterson, K., Nestor, P.J., Rogers, T.T., 2007. Where do you know what you know? The
Rev. 32 (8), 1416–1425. representation of semantic knowledge in the human brain. Nat. Rev. Neurosci. 8,
Gusnard, D.A., Raichle, M.E., 2001. Searching for a baseline: functional imaging and the 976–987.
resting brain. Nat. Rev. Neurosci. 2, 685–694. Phan, L., Tariq, A., Lam, G., Pang, E.W., Alain, C., 2021. The neurobiology of semantic
Happé, F., 1997. Central coherence and theory of mind in autism: reading homographs in processing in autism spectrum disorder: an activation likelihood estimation analysis.
context. Br. J. Dev. Psychol. 15 (1), 1–12. J. Autism Dev. Disord. 51 (9), 3266–3279.
Happé, F., Frith, U., 2006. The weak coherence account: detail-focused cognitive style in Plaut, D.C., McClelland, J.L., Seidenberg, M.S., Patterson, K., 1996. Understanding
autism spectrum disorders. J. Autism Dev. Disord. 36 (1), 5–25. normal and impaired word reading: computational principles in quasi-regular
Harm, M.W., Seidenberg, M.S., 2004. Computing the meanings of words in reading: domains. Psychol. Rev. 103 (1), 56–115.
cooperative division of labor between visual and phonological processes. Psychol. Pylkkänen, L., 2019. The neural basis of combinatory syntax and semantics. Science 366
Rev. 111 (3), 662–720. (6461), 62–66. https://doi.org/10.1126/science.aax0050.
Harris, G.J., Chabris, C.F., Clark, J., Urban, T., Aharon, I., Steele, S., McGrath, L., Pylkkänen, L., McElree, B., 2006. The syntax-semantics interface: on-line composition of
Condouris, K., Tager-Flusberg, H., 2006. Brain activation during semantic processing sentence meaning. In: Handbook of Psycholinguistics. Elsevier, pp. 539–579.
in autism spectrum disorders via functional magnetic resonance imaging. Brain Raichle, M.E., 2015. The brain’s default mode network. Annu. Rev. Neurosci. 38,
Cognit. 61 (1), 54–68. 433–447.
Hus, V., Lord, C., 2014. The autism diagnostic observation schedule, Module 4: revised Rapp, A.M., Mutschler, D.E., Erb, M., 2012. Where in the brain is nonliteral language? A
algorithm and standardized severity scores. J. Autism Dev. Disord. 44 (8), coordinate-based meta-analysis of functional magnetic resonance imaging studies.
1996–2012. https://doi.org/10.1007/s10803-014-2080-3. Neuroimage 63 (1), 600–610.
International, F., 2017. The Farlex Idioms and Slang Dictionary. CreateSpace Rice, G.E., Lambon Ralph, M.A., Hoffman, P., 2015. The roles of left versus right anterior
Independent Publishing Platform. https://books.google.com/books?id=N3Z temporal lobes in conceptual knowledge: an ALE meta-analysis of 97 functional
stAEACAAJ. neuroimaging studies. Cerebr. Cortex 25 (11), 4374–4391.
Kalandadze, T., Norbury, C., Nærland, T., Næss, K.-A.B., 2018. Figurative language Richlan, F., Kronbichler, M., Wimmer, H., 2009. Functional abnormalities in the dyslexic
comprehension in individuals with autism spectrum disorder: a meta-analytic brain: a quantitative meta-analysis of neuroimaging studies. Hum. Brain Mapp. 30
review. Autism 22 (2), 99–117. (10), 3299–3308.
Kamio, Y., Robins, D., Kelley, E., Swainson, B., Fein, D., 2007. Atypical lexical/semantic Ross, L.A., Olson, I.R., 2010. Social cognition and the anterior temporal lobes.
processing in high-functioning autism spectrum disorders without early language Neuroimage 49, 3452–3462.
delay. J. Autism Dev. Disord. 37, 1116–1122. Saad, Z.S., Glen, D.R., Chen, G., Beauchamp, M.S., Desai, R., Cox, R.W., 2009. A new
Kana, R.K., Keller, T.A., Cherkassky, V.L., Minshew, N.J., Just, M.A., 2006. Sentence method for improving functional-to-structural MRI alignment using local Pearson
comprehension in autism: thinking in pictures with decreased functional correlation. Neuroimage 44, 839–848.
connectivity. Brain 129, 2484–2493. Sahyoun, C.P., Belliveau, J.W., Soulières, I., Schwartz, S., Mody, M., 2010. Neuroimaging
Kana, R.K., Wadsworth, H.M., 2012. The archeologist’s career ended in ruins”: of the functional and structural networks underlying visuospatial vs. linguistic
hemispheric differences in pun comprehension in autism. Neuroimage 62 (1), 77–86. reasoning in high-functioning autism. Neuropsychologia 48 (1), 86–95.
Knaus, T.A., Silver, A.M., Lindgren, K.A., Hadjikhani, N., Tager-Flusberg, H., 2008. fMRI Schauder, K.B., Mash, L.E., Bryant, L.K., Cascio, C.J., 2015. Interoceptive ability and
activation during a language task in adolescents with ASD. J. Int. Neuropsychol. Soc. body awareness in autism spectrum disorder. J. Exp. Child Psychol. 131, 193–200.
14 (6), 967–979. Schurz, M., Radua, J., Tholen, M.G., Maliske, L., Margulies, D.S., Mars, R.B., Sallet, J.,
Koenigs, M., Tranel, D., 2007. Irrational economic decision-making after ventromedial Kanske, P., 2021. Toward a hierarchical model of social cognition: a neuroimaging
prefrontal damage: evidence from the Ultimatum Game. J. Neurosci. 27 (4), meta-analysis and integrative review of empathy and theory of mind. Psychol. Bull.
951–956. 147 (3), 293.
Koenigs, M., Young, L., Adolphs, R., Tranel, D., Cushman, F., Hauser, M., Damasio, A., Schwartz, M.F., Kimberg, D.Y., Walker, G.M., Brecher, A., Faseyitan, O.K., Dell, G.S.,
2007. Damage to the prefrontal cortex increases utilitarian moral judgements. Mirman, D., Coslett, H.B., 2011. Neuroanatomical dissociation for taxonomic and
Nature 446 (7138), 908–911. thematic knowledge in the human brain. Proc. Natl. Acad. Sci. U.S.A. 108 (20),
Lancaster, J.L., Woldorff, M.G., Parsons, L.M., Liotti, M., Freitas, C.S., Rainey, L., 8520–8524.
Kochunov, P.V., Nickerson, D., Mikiten, S.A., Fox, P.T., 2000. Automated Talairach Seidenberg, M.S., McClelland, J.L., 1989. A distributed, developmental model of word
atlas labels for functional brain mapping. Hum. Brain Mapp. 10, 120–131. recognition and naming. Psychol. Rev. 96 (4), 523–568.
Landa, R.J., Goldberg, M.C., 2005. Language, social, and executive functions in high Simmons, W.K., Reddish, M., Bellgowan, P.S., Martin, A., 2010. The selectivity and
functioning autism: a continuum of performance. J. Autism Dev. Disord. 35 (5), functional connectivity of the anterior temporal lobes. Cerebr. Cortex 20, 813–825.
557–573. https://doi.org/10.1007/s10803-005-0001-1. Smallwood, J., Bernhardt, B.C., Leech, R., Bzdok, D., Jefferies, E., Margulies, D.S., 2021.
Lanzoni, L., Ravasio, D., Thompson, H., Vatansever, D., Margulies, D., Smallwood, J., The default mode network in cognition: a topographical perspective. Nat. Rev.
Jefferies, E., 2020. The role of default mode network in semantic cue integration. Neurosci. 1–11.
Neuroimage 219, 117019. Tager-Flusberg, H., Paul, R., Lord, C., 2005. language and communication in autism. In:
Levy, A., Perry, A., 2011. Outcomes in adolescents and adults wiht autism: a review of Volkmar, F.R., Paul, R., Klin, A., Cohen, D. (Eds.), Handbook of Autism and
the literature. Research in Autism Spectrum Disorders 5, 1271–1282. Pervasive Developmental Disorders: Diagnosis, Development, Neurobiology, and
Li, J., Pylkkänen, L., 2021. Disentangling semantic composition and semantic association Behavior. John Wiley & Sons Inc, pp. 335–364. https://doi.org/10.1002/
in the left temporal lobe. J. Neurosci. 41, 6526–6538. 9780470939345.ch12.
Lin, N., Wang, X., Xu, Y., Wang, X., Hua, H., Zhao, Y., Li, X., 2018. Fine Subdivisions of Taylor, J.S.H., Rastle, K., Davis, M.H., 2013. Can cognitive models explain brain
the semantic network supporting social and sensory–motor semantic processing. activation during word and pseudoword reading? A meta-analysis of 36
Cerebr. Cortex 28 (8), 2699–2710. https://doi.org/10.1093/cercor/bhx148. neuroimaging studies. Psychol. Bull. 139 (4), 766–791. https://doi.org/10.1037/
Lin, N., Xu, Y., Yang, H., Zhang, G., Zhang, M., Wang, S., Hua, H., Li, X., 2020. a0030266.
Dissociating the neural correlates of the sociality and plausibility effects in simple Van Overwalle, F., 2009. Social cognition and the brain: a meta-analysis. Hum. Brain
conceptual combination. Brain Struct. Funct. 225 (3), 995–1008. Mapp. 30, 829–858.
Martínez-Sanchis, S., 2014. Neurobiological foundations of multisensory integration in Vicente, A., Falkum, I.L., 2021. Accounting for the Preference for Literal Meanings in
people with autism spectrum disorders: the role of the medial prefrontal cortex. Autism Spectrum Conditions. Mind & Language.
Front. Hum. Neurosci. 8, 970. Vigliocco, G., Tranel, D., Druks, J., 2012. Language Production. The Cambridge
handbook of psycholinguistics, p. 443.

12
W.W. Graves et al. Neuropsychologia 171 (2022) 108240

Vigneau, M., Beaucousin, V., Hervé, P.Y., Duffau, H., Crivello, F., Houdé, O., Mazoyer, B., Westerlund, M., Pylkkänen, L., 2014. The role of the left anterior temporal lobe in
Tzourio-Mazoyer, N., 2006. Meta-analyzing left hemisphere language areas: semantic composition vs. semantic memory. Neuropsychologia 57, 59–70.
phonology, semantics, and sentence processing. Neuroimage 30, 1414–1432. Williams, D.L., Minshew, N.J., Goldstein, G., 2015. Further understanding of complex
Visser, M., Jefferies, E., Lambon Ralph, M.A., 2010. Semantic processing in the anterior information processing in verbal adolescents and adults with autism spectrum
temporal lobes: a meta-analysis of the functional neuroimaging literature. J. Cognit. disorders. Autism 19 (7), 859–867.
Neurosci. 22 (6), 1083–1094. Wirth, M., Jann, K., Dierks, T., Federspiel, A., Wiest, R., Horn, H., 2011. Semantic
Vulchanova, M., Saldaña, D., Chahboun, S., Vulchanov, V., 2015. Figurative language memory involvement in the default mode network: a functional neuroimaging study
processing in atypical populations: the ASD perspective. Front. Hum. Neurosci. 9, 24. using independent component analysis. Neuroimage 54 (4), 3057–3066.
Wakusawa, K., Sugiura, M., Sassa, Y., Jeong, H., Horie, K., Sato, S., Yokoyama, H.,
Tsuchiya, S., Inuma, K., Kawashima, R., 2007. Comprehension of implicit meanings
in social situations involving irony: a functional MRI study. Neuroimage 37 (4),
Further reading
1417–1426.
Wang, J., Conder, J.A., Blitzer, D.N., Shinkareva, S.V., 2010. Neural representation of Graves, W.W., Desai, R., Humphries, C., Seidenberg, M.S., Binder, J.R., 2010b. Neural
abstract and concrete concepts: a meta-analysis of neuroimaging studies. Hum. Brain systems for reading aloud: a multiparametric approach. Cerebr. Cortex 20,
Mapp. 31, 1459–1468. 1799–1815. https://doi.org/10.1093/cercor/bhp245.
Wechsler, D., 2011. Wechsler Abbreviated Scale of Intelligence, second ed. The
Psychological Corporation.

13

You might also like