The Role of the Ventrolateral Anterior Temporal Lobes in Social Cognition

Abstract

A key challenge for neurobiological models of social cognition is to elucidate whether brain regions are specialised for that domain. In recent years, discussion surrounding the role of the anterior temporal lobe (ATL) epitomises such debates; some argue it is part of a domain-specific network for social processing, while others claim it is a domain-general hub for semantic representation. In the present study, we used ATL-optimised fMRI to map the contribution of different ATL structures to a variety of paradigms frequently used to probe a crucial social ability, namely theory of mind (ToM). Using multiple tasks enables a clearer attribution of activation to ToM as opposed to idiosyncratic features of stimuli. Further, we directly explored whether these same structures are also activated by a non-social task probing semantic representations. We revealed that common to all of the tasks was activation of a key ventrolateral ATL region that is typically invisible to standard fMRI. This constitutes novel evidence in support of the view that the ventrolateral ATL contributes to social cognition via a domain-general role in the retrieval of conceptual knowledge, and against claims of a specialised social function.

Full text

The Role of the Ventrolateral Anterior Temporal Lobes in Social Cognition
Eva Balgova1, Veronica Diveica1, Jon Walbrin2 & Richard J. Binney1*
1 School of Human and Behavioural Sciences, Bangor University, Gwynedd, Wales
2 Faculdade de Psicologia e de Ciências da Educação, Universidade de Coimbra, Portugal
Abstract
A key challenge for neurobiological models of social cognition is to elucidate whether brain regions are
specialised for that domain. In recent years, discussion surrounding the role of anterior temporal regions
epitomises such debates; some argue the anterior temporal lobe (ATL) is part of a domain-specific
network for social processing, while others claim it comprises a domain-general hub for semantic
representation. In the present study, we used ATL-optimised fMRI to map the contribution of different
ATL structures to a variety of paradigms frequently used to probe a crucial social ability, namely ‘theory
of mind’ (ToM). Using multiple tasks enables a clearer attribution of activation to ToM as opposed to
idiosyncratic features of stimuli. Further, we directly explored whether these same structures are also
activated by a non-social task probing semantic representations. We revealed that common to all of the
tasks was activation of a key ventrolateral ATL region that is often invisible to standard fMRI. This
constitutes novel evidence in support of the view that the ventrolateral ATL contributes to social
cognition via a domain-general role in semantic processing, and against claims of a specialised social
function.
Keywords: semantic memory; social cognition; theory of mind; anterior temporal lobe;
distortion-corrected fMRI
___________________________________________________________________________
*To whom correspondence may be addressed:
Email: R.Binney@Bangor.ac.uk
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

SOCIAL COGNITION & THE ANTERIOR TEMPORAL LOBE 2
Introduction
1
The anterior temporal lobe (ATL) plays a crucial role in support of social cognition
2
(Frith and Frith 2003, 2010; Olson et al. 2013; Binney and Ramsey 2020). Damage to the ATL
3
results in profound and wide-ranging socio-affective deficits in both primates and humans
4
(Klüver and Bucy 1937; Terzian and Dalle Ore 1955; Edwards-Lee et al. 1997; Kumfor and
5
Piguet 2012; Kumfor et al. 2013, 2017; Irish et al. 2014; Binney et al. 2016). Amongst
6
neurotypical samples, the findings of functional neuroimaging studies suggest an almost
7
ubiquitous involvement in the high-level processing of faces and emotions (Wong and Gallate
8
2012; Avidan et al. 2014; Collins and Olson 2014; Collins et al. 2016; Ramot et al. 2019), as
9
well as in more abstracted forms of social processing, such as moral cognition and mental state
10
attribution (also known as theory of mind) (Moll et al. 2005; Schurz et al. 2014; Molenberghs
11
et al. 2016; Diveica et al. 2021).
12
Despite this, across various neurocognitive frameworks of the ‘social brain’, there is no
13
firm consensus regarding the nature of the function that the ATL performs (for a
14
comprehensive review, see Olson et al. 2007; 2013). There are likely two main drivers for this.
15
First, at a glance, it might be difficult to identify a common cognitive process that connects the
16
various social and emotional tasks that implicate the ATL (Olson et al. 2013; Binney and
17
Ramsey 2020). Second, inconsistent definitions of what anatomy it is that constitutes the ATL
18
have greatly contributed to a lack of clarity regarding the locations at which overlap, and
19
divergence of function seemingly occurs. From one perspective, the term “ATL” refers to all
20
cortex comprising the anterior half of the temporal lobe (Binney et al. 2010; Rice, et al. 2015;
21
Binney et al. 2016; Rice et al. 2018), and therefore a large area potentially comprising a number
22
of functionally distinct subregions. However, it has at times been used to more specifically
23
refer to the temporal polar cortex, and the limited boundaries of Brodmann’s area 38 (e.g., Ross
24
and Olson 2010; Simmons et al. 2010). Therefore, the primary aims of the present study were
25
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

to contribute to a more complete description of the ATL subregions engaged in service of social
26
cognitive tasks, and to advance understanding of the nature of their function.
27
One account of the ATL’s role in social tasks is that it stores mental scripts, or schema,
28
that are formed out of prior experiences, and provide a wider context for understanding social
29
interactions (Frith and Frith 2003; Gallagher and Frith 2003). However, until more recently,
30
there has been a lack of direct evidence to support this hypothesis. Moreover, it is unclear as to
31
what extent the proposed social functions played by ATL subregions are distinct from that of
32
more general declarative memory systems. Recent proposals have specifically associated the
33
role of some ATL subregions with the retrieval of social conceptual knowledge, which is
34
posited as a subtype of semantic memory (Zahn et al. 2007; Olson et al. 2013; Binney and
35
Ramsey 2020). Semantic memory is a term used to refer to a long-term store of general
36
conceptual-level knowledge that is involved in transforming sensory inputs into meaningful
37
experiences, and it underpins the ability to recognize and make inferences about objects,
38
people, and events in our environment (Lambon Ralph et al. 2017). Social conceptual
39
knowledge has been defined more distinctively as person-specific knowledge (Simmons et al.
40
2010), but also knowledge about interpersonal relationships, social behaviours, and of more
41
abstract social concepts such as truth and liberty (Zahn et al. 2007; Olson et al. 2013). The
42
claim that the ATL is engaged in retrieving this type of information during social tasks is
43
supported by a functional neuroimaging study that reveals ATL activation that is common to
44
both a social attribution task and a task involving semantic relatedness judgments about socially
45
relevant concepts (Ross and Olson 2010).
46
Moreover, it has been suggested that social conceptual knowledge could have a special,
47
or even privileged status over other categories of semantic information (Zahn et al. 2007; Olson
48
et al. 2013). Indeed, one influential account of the ATL, the social knowledge hypothesis, states
49
that the dorsolateral portion (including the anterior superior and middle temporal gyri) is
50
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

selectively involved in processing social concepts (Olson et al. 2013) This account has some
51
overlap with the social information processing account which claims that the dorsolateral ATL
52
activates to social concepts as part of a larger domain-specific network involved in social
53
information processing (Simmons et al. 2010; Persichetti et al. 2021). Indeed, some fMRI
54
studies have demonstrated a greater response of dorsolateral ATL subregions when semantic
55
judgments made on socially relevant stimuli are compared to similar judgments made on non-
56
social stimuli (Zahn et al. 2007; Ross and Olson 2010; Binney et al. 2016; Rice et al. 2018).
57
Further, proponents of the social knowledge hypothesis argue against a more general role of
58
ATL subregions in semantic processing, and point to the fact that a variety of socially-relevant
59
tasks and stimuli reliably activate the ATL, whereas the majority of functional imaging studies
60
of general semantic processing do not (Olson et al. 2007, 2013; Simmons and Martin 2009;
61
Simmons et al. 2010).
62
However, the ATL is strongly implicated in general semantic processing on the basis
63
of decades of neuropsychological data (Patterson et al. 2007) and a growing body of brain
64
stimulation and electrophysiological studies, as well as functional neuroimaging studies that
65
take special measures to address signal dropout and distortion within this region (Binney et al.
66
2010; Visser et al. 2010; Visser et al. 2010; Lambon Ralph et al. 2017). A critical issue,
67
therefore, is how it is possible to reconcile these two sets of observations within a single unified
68
theory of ATL function.
69
When broadly defined as the anterior half of the temporal lobe, the ATL is comprised
70
of a substantial volume of cortex, amongst which there are numerous subdivisions identifiable
71
on the basis of morphology, cytoarchitecture and connectivity (Ding et al. 2009; Binney et al.
72
2012; Pascual et al. 2015), and it is highly plausible that within it there are either distinct
73
functional parcels (Perischetti et al., 2021) or graded differences in function (Binney et al.,
74
2012; Jackson et al., 2018), including social and semantic function (Olson et al. 2013; Binney
75
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

et al. 2016). Therefore, under what might be called a ‘dual ATL hub account’, social conceptual
76
knowledge could be processed within a distinct location to more general conceptual
77
information (Zahn et al. 2007, 2009). Indeed, while social tasks and semantic judgements on
78
social words activate the dorsolateral ATL (the anterior superior and middle temporal gyri;
79
Zahn et al. 2007; Ross and Olson 2010; Binney et al. 2016; Mellem et al. 2016; Lin, Wang, et
80
al. 2018; Wang et al. 2019; Arioli et al. 2020), the general semantics literature, including data
81
from patients and studies using ATL-optimised fMRI (Binney et al. 2010; Mion et al. 2010;
82
Visser et al. 2010; Lambon Ralph et al. 2017), converges on the ventrolateral ATL (including
83
the anterior fusiform and inferior temporal gyri) as the centre-point of a domain-general
84
conceptual hub.
85
However, two recent studies have demonstrated that using enhanced fMRI techniques
86
greatly affects the patterns of activation observed across the ATL during the processing of
87
social-relevant stimuli and leads to different conclusions (Binney et al. 2016; Rice et al. 2018).
88
In conventional approaches to acquiring fMRI, susceptibility artefacts cause signal loss and
89
image distortion around the location of the ventrolateral ATL, which render the technique
90
effectively blind to activation in this region (Devlin et al. 2000). Spin-echo, and dual-echo
91
echo-planar fMRI, as well as post-acquisition distortion correction techniques, can be used to
92
recover this signal (Embleton et al. 2010; Halai et al. 2014) in which case it becomes clear that
93
the ventrolateral ATL activates strongly during semantic judgements made on both social and
94
non-social stimuli (Hoffman et al. 2015; Binney et al. 2016; Rice et al. 2018). These studies
95
confirmed previous reports by observing that the dorsolateral ATL activates selectively to
96
social semantic stimuli but the omni-category response of the ventrolateral ATL was much
97
greater in magnitude (Binney et al. 2016; Rice et al. 2018). We have argued that these
98
observations are consistent with graded differences in semantic function across a unified
99
representational substrate, as opposed to the ATL comprising two distinct systems for
100
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

processing social and non-social concepts (Binney et al., 2016). According to the graded
101
semantic hub proposal, a large extent of the ATL contributes to semantic processing in the
102
form of a unified representational space, all of which is engaged by the encoding and retrieval
103
of concepts, and by concepts of any kind. The centre of this space exists over the ventrolateral
104
ATL and its engagement during semantic processing is invariant to, for example, idiosyncratic
105
task features, including the modality through which concepts are accessed. Towards the edges
106
of this space, however, there are gradual shifts in semantic function such that it becomes
107
relatively more specialised for encoding certain types of semantic features (Binney et al. 2012;
108
Rice et al. 2015; Bajada et al., 2019). For example, features that are primarily experienced
109
through certain sensorimotor modalities (for a computational exploration of this general
110
hypothesis, see Plaut 2002). Along these lines, the socialness of a concept might reflect
111
semantic features that are uniquely experienced through certain processing streams like those
112
involved processing affect. The dorsolateral ATL in particular may be sensitive to the
113
socialness of concepts because of strong connectivity to medial temporal limbic and frontal
114
limbic regions (via the uncinate fasciculus; Binney et al. 2012; Bajada et al. 2016; Papinutto et
115
al. 2016). By extension of this proposal, we have argued that activation of ATL subregions in
116
service of social cognitive tasks reflects engagement of a domain-general semantic system and
117
that this is centred upon the ventrolateral ATL (Binney and Ramsey 2020).
118
The conclusions that can be drawn from those two particular studies regarding the
119
ATL’s role in social cognition are limited. This is because they used tasks where the demands
120
are primarily semantic in nature and the social relevance of the stimuli may have only been a
121
secondary feature. As such, it remains an open question whether social tasks typically
122
employed in the social neuroscience literature activate the ventrolateral ATL hub. The present
123
study tackles exactly that issue, with a specific focus on mental state attribution or ‘theory of
124
mind’ tasks. We chose this focus because theory of mind (ToM) abilities are considered central
125
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

to the construct of social cognition; they are considered as fundamental to successful social
126
interactions, as they enable us to describe, explain and predict behaviour (Frith and Frith 2005;
127
Brüne and Brüne-Cohrs 2006; Apperly 2012; van Hoeck et al. 2014; Heleven and van
128
Overwalle 2018). Neuroimaging studies reliably implicate the right temporo-parietal junction,
129
medial prefrontal cortex and precuneus as part of a core network for ToM (Saxe and Kanwisher
130
2003a; Saxe and Wexler 2005; Saxe 2006; Scholz et al. 2009; Young et al. 2010; Dodell-Feder
131
et al. 2011), whereas the role of the ATL is less clear and appears to be characterised as
132
ancillary by some accounts (van Overwalle 2009; Schurz et al. 2014; Molenberghs et al. 2016).
133
It is possible that a central role of the ventrolateral ATL has gone unnoticed because fMRI
134
studies of ToM typically do not account for technical constraints around this region.
135
We set out to address two key unresolved questions. First, we aimed to determine
136
whether and to what degree different parts of the ATL are activated by established theory of
137
mind tasks. This necessitated two key design elements: (i) the use of dual-echo fMRI and
138
distortion correction to ensure full coverage of the bilateral ATL; and (ii) the use of multiple
139
theory of mind tasks. This second design feature was important because showing common
140
activation across different theory of minds tasks with a variety of stimuli means that we can
141
more confidently assess whether activation can be attributed to ToM ability itself rather than
142
being the result of task demands or idiosyncratic features of the stimuli (Ross and Olson 2010).
143
In fMRI designs, the most commonly used theory of mind tasks include social vignettes,
144
cartoons, and animations that are intended to evoke the attribution of intentions. We used
145
animations as our primary task because they do not directly involve lexical-semantic
146
processing which has been shown to activate the ventrolateral ATL (e. g. Devlin et al. 2000;
147
Spitsyna et al. 2006; Binney et al. 2010; Visser, Embleton, et al. 2010) and would confound
148
our inferences. Moreover, they lend themselves to the creation of a comparable non-social (and
149
non-semantic) control or baseline activation task which is important for controlling for
150
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

attentional and executive demands, as well as perceptual stimulation. We also acquired data
151
during a False Belief task (Dodell-Feder et al. 2011) and a free-viewing animated film (Jacoby
152
et al. 2016) as these are established paradigms for localising the ‘mentalising’ or theory of mind
153
network.
154
Second, we set out to directly assess overlap of ToM related activation with activation
155
evoked by semantic decisions made upon nonverbal, non-social stimuli. Overlap, and
156
particularly overlap in the ventrolateral aspect, would support the hypothesis that activation of
157
the ATL during social tasks reflects the retrieval of semantic knowledge representations (Zahn
158
et al. 2007; Olson et al. 2013; Binney and Ramsey 2020). Specifically, we chose the picture
159
version of the Camel and Cactus task (CCT), which is an established means to engage and
160
measure semantic processing, and has been previously used in neuropsychological, functional
161
imaging and brain stimulation studies (Bozeat et al. 2000; Jefferies and Lambon Ralph 2006;
162
Hoffman et al. 2012; Visser et al. 2012). Further, via these means we were also able to directly
163
test three different accounts of the ATL, amongst a single cohort of participants, as follows:
164
165
1. The Social Knowledge Hypothesis/Social Information Processing Account:
166
according to these accounts, the ATL is part of a domain-specific network involved in
167
processing social information (Simmons et al. 2010), and the dorsolateral ATL is selectively
168
involved in processing concepts with social-emotional content (Olson et al. 2013). Proponents
169
of this account argue against a domain-general role of the ATL in semantic representation
170
(Simmons et al. 2010; Olson et al. 2013). If this is correct, then the ATL would activate for
171
ToM tasks but not during the CCT.
172
2. The Dual ATL Hub account: if a dual hub account is correct then the ToM tasks
173
would exclusively activate the dorsolateral/polar ATL and not the ventrolateral aspect, and the
174
CCT would only activate the ventrolateral ATL.
175
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

3. The Graded ATL Semantic Hub Hypothesis: if the third account, in which the ventral
176
ATL is the centre point of a domain-general hub for both social and non-social semantic
177
processes (Binney et al. 2016), is correct, then the greatest degree of overlap between the ToM
178
tasks and the CCT will be within the ventrolateral portion. We might also observe dorsolateral
179
activation that is more selective to social stimuli.
180
181
Materials and Methods
182
Data Availability statement. Following open science initiatives (Munafò et al. 2017),
183
behavioural and neuroimaging data are openly available on the Open Science Framework
184
project page (https://osf.io/v2gt5/).
185
186
Design Considerations
187
To ensure that the imaging protocol was sensitive to changes in activation across all
188
parts of the ATL, we used a dual-echo gradient-echo echo-planar imaging (EPI) fMRI sequence
189
that is optimised to detect blood oxygen level-dependent (BOLD) signal in areas of the brain
190
that are usually prone to magnetic susceptibility-induced signal loss (Halai et al. 2014, 2015).
191
Dual-echo sequences like the one used here have been shown to have greater ability to detect
192
inferior temporal lobe activation compared to single-echo GE-EPI sequences with a
193
conventional echo time around 30 ms. (Embleton et al. 2010; Halai et al. 2014). Further, to
194
alleviate the impact of geometric distortions and mislocalisations of fMRI signal also caused
195
by magnetic susceptibility artefacts, the dual-echo sequence was combined with a post-
196
acquisition k-space spatial correction. Our choice of distortion correction procedure was based
197
on evidence that that it outperforms more standard corrections that use phase-encoded field
198
maps (Embleton et al. 2010). We also made other methodological decisions such as acquiring
199
with a left-to-right phase-encoding direction that contributes to better quality of data in the
200
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

inferior ATL (Embleton et al. 2010). To demonstrate the effectiveness of these measures, we
201
followed the procedure of previous reports investigating the ATLs (Simmons and Martin 2009;
202
Hoffman et al. 2015) and calculated an average map of temporal signal to noise ratio (tSNR)
203
(see Figure M1). The tSNR map suggests a good signal quality even in the in most ventral
204
proportions of the ATL.
205
Secondly, we adapted stimuli created by Walbrin et al. (2018) to fashion a two
206
alternative force choice (AFC) task involving explicit judgements about the intention of
207
moving shapes (are they behaving in ‘friendly’ or ‘unfriendly’ way?) that are seemingly
208
engaged in goal-directed action. Our task is broadly inspired by the widely used Frith-Happé
209
animations for probing ToM processes (see e.g., Abell et al. 2000; Castelli 2002; Kana et al.
210
2009; Das et al. 2012; Hennion et al. 2016; Synn et al. 2018; Bliksted et al. 2019) but it was
211
designed to be closer to that used in a key study by Ross and Olsen (2010; also see Shultz et
212
al., 2003). We specifically chose Walbrin and colleagues’ stimuli because they offer a higher
213
number of unique trials (impacting sensitivity/power) than other similar stimuli sets, and they
214
are visually well-controlled to minimize the contribution of low-level visual information to
215
brain responses (i.e., they are comprised of visually diverse interactive scenarios that are well-
216
matched for overall motion energy). To control for attentional and executive demands involved
217
in the main task, we reconfigured these stimuli to further create a well-matched perceptual
218
judgment task. See below for further detail regarding the experimental and control task.
219
220
Participants
221
Thirty-one healthy native English speakers took part in the experiment. All participants
222
had normal or corrected-to-normal vision, no history of neurological and psychiatric conditions
223
and were right-handed as established by the Edinburgh Handedness Inventory (Oldfield 1971).
224
Participants provided written informed consent, and the study was approved by the local
225
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

research ethics review committee. Seven participants were excluded because of inadequate task
226
performance (under 70% accuracy) on any one of the social interaction tasks (N=3), or because
227
of failed distortion correction and therefore insufficient data quality (N=4). The final analysed
228
sample comprised of twenty-four participants (12 females, Mage= 22.21, SDage= 2.13).
229
230
Experimental Stimuli and Tasks
231
Theory of Mind (ToM)
232
A total of 126 unique video stimuli designed by Walbrin et al. (2018) were used for the
233
main interaction judgement theory of mind (IJ-ToM) task and its corresponding control task.
234
The IJ-ToM stimuli (N = 63) featured two self-propelled circles representing animate agents
235
that were intentionally interacting and doing so in a co-operative manner in half the trials and
236
a competitive manner in the other half (see Supplementary Figure M2). In Walbrin and
237
colleagues’ original stimulus pool (N=256) half of the scenarios concluded with successful
238
goal outcome (e.g., successfully opening a closed door) and half with unsuccessful goal
239
outcomes. Here we only used a subset of the former. In the IJ-ToM task, participants were
240
instructed to make explicit inferential judgements, via a key press, as to whether the agents’
241
actions towards one another were friendly or unfriendly (unlike the original study that sought
242
to minimize the contribution of ToM judgements and employed a perceptual response task).
243
The associated control task used ‘scrambled’ versions of the interaction stimuli (N = 63) that
244
preserved many of the visual properties but featured altered motion paths such that the shapes
245
did not appear to be intentionally interacting with each other or their environment (see Walbrin
246
et al. 2018). For the present study, these control stimuli were adjusted such that in fifty per cent
247
of trials the speed of motion of one of the two shapes was slower than that of the other. This
248
was done by slowing the frame rate of one of the animation elements (i.e., one of the circles)
249
from 24 to 18 frames per second and removing frames from either the beginning (50%) or end
250
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

of the sequence to maintain the original duration (6 secs). Moreover, we ensured that the more
251
slowly moving object appeared an equal number of times at each relative position on the screen
252
(e.g., left versus right). Participants responded to these stimuli via key press and indicated
253
whether they believed the circles were moving at same or different speeds. Following some
254
initial pilot behavioural testing, the duration of all 126 videos was shortened from 6 to 3 sec to
255
increase task difficulty/eliminate idle time.
256
We also acquired data with two widely used functional localisers for the putative ToM
257
network, namely the False Belief (FB) paradigm (Dodell-Feder et al. 2011) and a more recently
258
validated free-viewing movie paradigm (MOV) (Jacoby et al. 2016). The former is a verbal
259
paradigm which is comprised of two sets of 10 text-based vignettes each of which are presented
260
on screen and followed by true / false questions. One of these sets requires the participant to
261
make inferences about a character’s internal beliefs, and this is contrasted against descriptions
262
of facts about physical events. The MOV paradigm involves passive viewing of a commercial
263
animated film and contrasts BOLD responses to events in which characters are involved in
264
ToM against those in which characters experience physical pain (see Jacoby et al. (2016) for
265
more detail ).
266
267
Non-Verbal Semantic Association
268
Participants also completed a non-verbal version of an established neuropsychological
269
assessment of semantic associative knowledge known as the Camel and Cactus task (CCT;
270
(Bozeat et al. 2000)). This task has been used to engage the semantic network in prior fMRI
271
studies (Visser et al. 2012; Rice et al. 2018). The version used in the present study consisted of
272
36 trials that contained pictorial stimuli and required participants to make semantic associations
273
between a probe object (e.g., a camel) and a target object (e.g., a cactus) that was presented
274
alongside a foil from the same semantic category (e.g., a rose). The CCT was contrasted against
275
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

a perceptual control task (36 trials) that consisted of scrambled versions of the CCT pictures
276
and required participants to identify which of two choice pictures was visually identical to a
277
probe (see more detail in Visser et al. (2012)).
278
279
Experimental Procedure
280
Participants underwent all testing within a single session lasting approximately one
281
hour. Each individual completed three runs of the IJ-ToM procedure reported below, followed
282
by one run of the CCT procedure, two runs of the FB localiser and one run of the MOV
283
localiser. The IJ-ToM task, the CCT and FB task and the corresponding control tasks were
284
presented via E-prime (Psychology Software Tools, 2017) and the MOV task via Psychtoolbox
285
(Brainard 1997; Pelli 1997) software. Behavioural responses were recorded using an MRI
286
compatible response box.
287
288
Interaction Judgement ToM Task
289
The IJ-ToM task and the speed judgement task were paired within a run using an A-
290
rest-B-rest box car block design. Each run contained six blocks per task and three trials per
291
block (18 trials per run per task). There were an equal number of trial types (e.g., cooperative
292
versus competitive) randomly distributed across blocks within a given run. Both types of active
293
blocks were 17.25 secs long and they were separated by blocks of passive fixation lasting 12
294
secs each. Each trial began with a fixation cross (duration = 500ms) which was followed by
295
the target animation (3000ms) and finished with a response cue (three question marks;
296
2000ms). A blank screen occupied an inter stimulus interval of 250ms. Each run lasted 5
297
minutes and 51 secs and consisted of unique sets of animations. The order in which these runs
298
were completed was counterbalanced across participants. Participants also completed three
299
practice blocks for each of the two tasks before the main runs began.
300
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

301
Camel and Cactus Task
302
The CCT and the corresponding perceptual identity matching control task were
303
alternated within a single run using a blocked design. There were 9 blocks per task, each
304
consisting of four trials (totalling 36 trials per task) and lasting 20 secs. A trial began with a
305
fixation cross (500ms) followed by a stimulus triad (4500ms). Participants responded via key
306
press while the probe and choice items were on screen. Active blocks were separated by brief
307
rest blocks lasting 4000ms and, overall, the run lasted for 7 mins and 12 secs.
308
309
False Belief and animated movie localisers
310
Each run of the false belief localiser lasted 4 minutes and 32 seconds and consisted of
311
10 trials of belief vignettes and 10 trials of the fact vignettes. Finally, the passive MOV
312
scanning run lasted 5 mins and 59 seconds including a fixation period of 10 secs prior to the
313
beginning of the movie. Further details regarding these paradigms are reported by Jacoby et al.
314
(2016).
315
316
Imaging Acquisition
317
All imaging was performed on a 3T Phillips Achieva MRI scanner with a 32-element
318
SENSE head coil using a 2.5 sense factor for image acquisition. The parameters of the dual-
319
echo gradient-echo EPI fMRI sequence were the following: 31 axial slices covering the whole
320
brain and obtained in an ascending sequential order with a first echo time (TE) = 12ms and
321
second TE = 35ms, repetition time (TR) = 2000ms, flip angle = 85°, FOV (mm) = 240 × 240
322
× 124, slice thickness = 4 mm, no interslice gap, reconstructed voxel size (mm) = 2.5 × 2.5 and
323
reconstruction matrix = 96 × 96. Prior to image acquisition for each run, we acquired five
324
dummy scans to allow the initial magnetisation to stabilise. This was followed by acquiring
325
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

177 volumes for each IJ-ToM task run, 218 volumes for the CCT task run, 138 volumes for
326
each FB task run and 180 volumes for the MOV task run. Adhering to the distortion-correction
327
method, we acquired these functional runs with a single direction k space traversal in the left-
328
right phase-encoding direction. We also acquired a short EPI “pre-scan” with the participants
329
at rest. The parameters of the pre-scan matched the functional scans except that it included
330
interleaved dual direction k space traversals. This gave 10 pairs of images with opposing direction
331
distortions (10 left-right and 10 right-left) which were to be used in the distortion correction
332
procedure described below. To check the quality of the distortion corrected images, we
333
obtained a high resolution T2-weighted scan consisting of 36 slices covering the whole brain,
334
with TR= 17ms, TE= 89ms; reconstructed voxel size (mm) = 0.45 x 0.45 x 4; reconstruction
335
matrix= 512 x512. Additionally, we used a T1-weighted 3D imaging sequence to acquire an
336
anatomical scan, consisting of 175 slices covering the whole brain, for use in spatial
337
normalisation procedures. The parameters of this scan were as follows: P reduction (RL)
338
SENSE factor of 2 and S reduction (FH) SENSE factor of 1, TR = 18ms, TE = 3.4ms, 8° flip
339
angle, reconstructed voxel size (mm) = 0.94 × 0.94 × 1.00 and reconstruction matrix = 240 ×
340
240.
341
342
Data Analysis
343
Behavioural Data
344
Incorrectly answered trials, missed trials and trials with response latencies that were
345
two standard deviations above or below the participant’s task mean were excluded from
346
analyses of behavioural data. Task performance was assessed in terms of both accuracy and
347
decision times and compared using paired-sample T-tests. Average decision times per block of
348
each task were also calculated so that they could be used as regressors of no interest in fMRI
349
analyses.
350
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

351
Distortion Correction and fMRI pre-processing
352
A spatial remapping correction was computed separately for images acquired at the
353
long and the short echo time, and using a method reported elsewhere (Embleton et al. 2010).
354
This was implemented via in-house MATLAB script (available upon request) as well as
355
SPM12’s (Statistical Parametric Mapping software; Wellcome Trust Centre for Neuroimaging,
356
London, UK) 6-parameter rigid body registration algorithm. Briefly, in the first step, each
357
functional volume was registered to the mean of the 10 pre-scan volumes acquired at the same
358
echo time. Although this initial step was taken primarily as part of the distortion correction
359
procedure, it also functioned to correct the time-series for differences in subject positioning in
360
between functional runs and for minor motion artefacts within a run. Next, one spatial
361
transformation matrix per echo time was calculated from opposingly-distorted pre-scan
362
images. These transformations consisted of the remapping necessary to correct geometric
363
distortion and were applied to each of the main functional volumes. This resulted in two
364
motion- and distortion-corrected time-series per run (one per echo) which were subsequently
365
combined at each timepoint using a simple linear average of image pairs.
366
All of the remaining pre-processing steps and analyses were carried out using SPM12.
367
Slice-timing correction referenced to the middle slice was performed on the distortion- and
368
motion-corrected images. The T1-weighted anatomical scan was co-registered to a mean of the
369
functional images using a 6-parameter rigid-body transform, and then SPM12’s unified
370
segmentation and normalisation procedure and the DARTEL (diffeomorphic anatomical
371
registration though an exponentiated lie algebra; (Ashburner 2007)) toolbox were used to
372
estimate a spatial transform to register the structural image to Montreal Neurological Institute
373
(MNI) standard stereotaxic space. This transform was subsequently applied to the co-registered
374
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

functional volumes which were resampled to a 3 x 3x 3 mm voxel size and smoothed with an
375
8 mm full-width half-maximum Gaussian filter.
376
377
fMRI Statistical Analysis
378
Data were analyzed using the general linear model approach (GLM). At the within-subject
379
level, a fixed effect analysis was carried out upon each task pair (e.g., the interaction judgement
380
task and the perceptual control task), incorporating all functional runs within a single GLM.
381
Block onsets and durations were modelled with a boxcar function and convolved with the
382
canonical hemodynamic response function. A high pass filter with a cut off of 128s was also
383
applied. The extracted motion parameters were entered into the model as regressors of no
384
interest. Decision time data were also modelled to account for differences in task difficulty.
385
Due to the block design employed, there was a single value for each epoch of a task which was
386
the average of response times across the trials. These average decision times for each block
387
were mean centred. To avoid false positive activations in the surrounding CSF due to
388
physiological noise, we used an explicit mask restricted to cerebral tissue that was created from
389
tissue segments generated by DARTEL in MNI space and binarised with a 0.4 threshold.
390
At the level of multi-subject analyses, we first examined activation during the IJ-ToM
391
task at the whole brain level. To complement the whole-brain topographic representation of the
392
data, we used a priori regions of interest (ROI), defined on the basis of independent datasets,
393
to extract and quantify the magnitude of activation within different ATL subregions. This was
394
done using the SPM MarsBar toolbox (Brett et al. 2002). A key aim of this study was also to
395
assess whether parts of the ATL are commonly activated by different types of behavioural
396
paradigm used to localise the putative ToM network. To do this we performed a formal
397
conjunction analysis (Price and Friston 1997; Nichols et al. 2005a). In addition to our
398
interacting geometric shapes paradigm, this included a version of False Belief task (Dodell-
399
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

Feder et al. 2011) which is comprised of verbal vignettes, and a free-viewing movie paradigm
400
(Jacoby et al. 2016). This analysis was an important step because it would enable us to home
401
in on those regional activations that are a feature of ToM abilities irrespective of the manner in
402
which they are probed. Moreover, if one is to compare ToM tasks that are qualitatively very
403
different from one another, it is possible to attribute the common activations much more
404
convincingly to the particular cognitive process of interest, as opposed to similarities in
405
physical stimulus properties or peripheral elements of the task demands (Friston et al. 1999).
406
Using a further conjunction analysis we also explored overlap between the IJ-ToM task and a
407
nonverbal semantic association task. This enabled us to test the hypothesis that activation of
408
the ATL during social tasks reflects the retrieval of semantic knowledge representations (Zahn
409
et al. 2007; Olson et al. 2013; Binney and Ramsey 2020).
410
Whole-brain multi-subject random effects analyses were conducted on each of the
411
following contrasts of interest: IJ-TOM task: interaction > speed judgements, interaction
412
judgements > rest, speed judgements > interaction judgements; CCT task: semantic >
413
perceptual judgements; FB task: false belief > false fact judgements; MOV task: mentalizing
414
> pain. One-sample t-tests were performed on all sets of contrast images following application
415
of the same explicit mask as used in the single subject analyses. The resulting statistical maps
416
were assessed for cluster-wise significance using a cluster-defining voxel-height threshold of
417
p < .001 uncorrected, and family-wise error (FWE) corrected cluster extent threshold at p < .05
418
(calculated per SPM12 under the random field theory framework; see details regarding
419
smoothness of data, the search volumes and RESELS in Supplementary Table M1).
420
Thresholded maps were overlaid on a MNI152 template brain using MRIcroGL
421
(https://www.nitrc.org/projects/mricrogl). We used an AAL atlas implemented in R label4MRI
422
package (https://github.com/yunshiuan/label4MRI) to guide the labelling of peak co-ordinates
423
in the output tables.
424
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

Within the ROI analysis, (Brett et al. 2002) two ATL subregions were explored in each
425
hemisphere. A ventrolateral ATL ROI was defined by peak coordinates of activation reported
426
by an independent study of non-verbal semantic processing (Visser et al. 2012) [MNI: +/- 57,
427
-15, -24]. We also examined a polar ATL ROI which was defined on the basis of activation
428
tuned towards socially relevant semantic stimuli as reported by Binney et al. (2016) [MNI: +/-
429
48, 9, -39]. Furthermore, so that we could compare the degree of ATL activation to that of a
430
more established ToM region, we defined a third ROI on the basis of ToM-related TPJ
431
activation reported by Saxe and Kanwisher (2003b) [+/- 54, -60, 21]. These sets of coordinates
432
defined a centre of mass for spheres with a radius of 10mm (See Figure 1 panel B for an
433
illustration of ROI locations). Per subject, a single summary statistic was calculated to represent
434
activation across all the voxels in an ROI (the mean of the parameter estimates) for the IJ-ToM
435
task relative to the speed judgment control task. One-sample t-tests were then performed to
436
assess group-level significance. To control for multiple comparisons, p-values were Bonferroni
437
corrected on the basis of the number of ROIs (multiplied by 6) as implemented in MarsBar.
438
We also conducted planned comparisons between ROIs in each hemisphere, and between
439
hemispheric homologue regions, using paired t-tests. For the conjunction analyses, we used a
440
p< .001 uncorrected voxel height threshold to be achieved by each contrast independently prior
441
to conjunction (Price and Friston 1997; Nichols et al. 2005a).
442
443
Results
444
Behavioural Data
445
Mean accuracy and decision times for all tasks are displayed in Table 1. Performance
446
on the animated interaction friendliness judgement (IJ ToM task) was more accurate than on
447
the speed judgement control task (t (23) = 7.50, p < .001, Cohen’s d = 1.53), and decision times
448
were also faster (t (23) = -3.08, p = .005, d = 0.63). Performance during semantic association
449
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

judgements (CCT task) was less accurate than performance in the perceptual identity matching
450
control task (t (23) = -8.83, p< .005, d = -1.80), although there was no significant difference in
451
the latency of decision times (t (23) = -0.65, p= .522, d= -0.13). Accuracy across the false belief
452
and false facts judgments (FB task) was comparable (t (23) = 0.77, p = .450, d = 0.16) although
453
decision times were faster in the false fact task (t (23) = 2.73, p = .012, d = 0.56).
454
Table 1. Behavioural data
Task
Accuracy
(%)
Decision time
(ms)
Interaction Friendliness Judgement
96.37 (04.35)
468.38 (125.44)
Perceptual Speed Judgement
84.88 (07.62)
524.25 (159.84)
Semantic Association Judgement
79.75 (17.84)
1491.92 (382.65)
Perceptual Identity Matching
92.59 (19.96)
1529.76 (460.57)
False Belief Judgement
70.42 (19.67)
2779.56(385.47)
False Fact Judgement
67.92 (15.87)
2560.31(354.64)
Standard deviations stated in parentheses
Activation During a Social Attribution Task Given Full Temporal Lobe Coverage
455
A whole brain univariate analysis contrasting social interaction friendliness judgments
456
with the matched speed judgement task revealed robust bilateral ATL activation that was
457
centred over the ventrolateral aspects in both hemispheres (see Figure 1, panel A and Table
458
2). In the left hemisphere, this extended from the ventrolateral temporopolar cortex (BA38),
459
along the inferior middle temporal gyrus and inferior temporal gyrus (ITG), to approximately
460
halfway along the temporal lobe (y ≈ -17). This included a maxima that is notably similar in
461
location (MNI coordinates x = -54, y = 6, z = -39) to that identified in association with
462
processing of abstract social concepts (relative to matched abstract non-social concepts; x = -
463
54, y = 9, z = -33 and animal function concepts; x= -48, y= 9, z= -39) by Binney et al. (2016).
464
The same cluster also extended more posteriorly upon the basal surface and along the
465
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

fusiform/lingual and posterior inferior temporal gyri. It also traversed up into the parietal lobe
466
and the intraparietal sulcus. In the right hemisphere, ATL activation also covered much of the
467
ventrolateral surface (particularly the polar cortex and the anterior-most portion of the middle
468
temporal gyrus (MTG) but extended less posteriorly (to y ≈ -11) than it did in the left.
469
Outside of the ATL, and as expected, this contrast also revealed activation amongst key
470
nodes of the putative ToM network, including the temporoparietal junction (TPJ), the medial
471
prefrontal cortex (mPFC) and the precuneus (Frith and Frith 2003, 2006; Saxe and Kanwisher
472
2003b; van Overwalle 2009; Jacoby et al. 2016). TPJ activation was observed in both
473
hemispheres at the position of the posterior superior temporal sulcus/gyrus (STS/STG) and, in
474
the right hemisphere, it extended to more posterior regions (at y ≈ -54) that are frequently
475
emphasized in landmark studies (Saxe and Kanwisher 2003b; Saxe and Powell 2006) and large
476
scale meta-analyses (Schurz et al. 2014; Molenberghs et al. 2016) of the theory of mind
477
network. Further activation was revealed in the left posterior MTG, the left insula, and bilateral
478
temporooccipital and cerebellar regions.
479
Activation during the social interaction friendliness judgments was also contrasted with
480
passive fixation/rest. There was notably little activation in the ATLs, except for a small cluster
481
in the left superior temporal pole (see Supplementary Figure 1 and Supplementary Table
482
R1). This is consistent with the idea that there may be automatic semantic activation (e.g.,
483
mind-wandering, episodic recall and socially oriented thoughts) during periods of passive
484
fixation, and it demonstrates the importance of using active baseline tasks for detecting ATL
485
activation which has been highlighted in prior meta-analyses and empirical investigations
486
(Binder et al. 1999, 2009; Visser, Jefferies, et al. 2010; Andrews-Hanna et al. 2014).
487
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

Figure 1. Panel A. Cortical regions activated during the main experimental ToM task (the interaction judgement),
relative to the speed judgement control task. The statistical map was thresholded with an uncorrected voxel height
threshold of p < .001 and a family wise error corrected minimum cluster extent threshold (k= 152) at p < .05.
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

Cross-sections were chosen to display the location of activation found in key studies investigating ToM processing
(Saxe & Kanwisher, 2003; right TPJ [51, -54, 27]), semantic processing of social concepts (Binney, Hoffman &
Lambon Ralph, 2016; left TP [-48, 9, -39]) and general semantic processing (Visser, Jefferies, Embleton &
Lambon Ralph, 2012; left inferior ATL [-57, -15, -24]). Panel B. Summary of the ROI analyses comparing the
magnitude of activation for the interaction judgement ToM task (relative to that during speed judgments control
task). An asterisk denotes a significant effect at p < .05 after Bonferroni correction. Numerical p-values are
displayed where comparisons yielded a p-value greater than .05 but less than .1. TP = temporal pole, vATL =
ventrolateral anterior temporal lobe, TPJ = temporo-parietal junction, L = left, R = right.
Outside of this region there was robust bilateral fronto-parietal activation including of
488
the bilateral TPJ and the ventrolateral prefrontal cortex, and activation of the mPFC, the
489
precuneus, and temporooccipital and cerebellar regions. The contrast revealing greater
490
activation for the speed judgment relative to the social attribution task is reported in
491
Supplementary Figure 2 and Supplementary Table R2 and revealed the right middle frontal
492
gyrus and a number of midline structures.
493
Table 2 Significant activation clusters in the social interaction judgement > speed judgement
contrast (p < .05, FWE-corrected, corresponding to an extent threshold of k = 152 following
a cluster-defining threshold of p< .001, uncorrected)
Cluster Name and Location of Maxima
Cluster Extent (voxels)
Peak (Z)
MNI Coordinates (mm)
x
y
z
L temporal – parietal – occipital
2242
anterior ITG / sulcus
5.72
-57
-6
-30
anterior ITG
5.26
-54
6
-39
posterior ITG
5.07
-51
-51
-24
precuneus
4.78
-12
-60
42
posterior ITG
4.76
-45
-60
-9
inferior parietal lobule
4.68
-39
-51
45
precuneus
4.49
-9
-75
48
middle/anterior ITG
4.42
-48
-21
-27
posterior MTG
4.23
-48
-54
3
cerebellum
4.21
-51
-51
-36
posterior MTG
4.10
-39
-60
15
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

anterior MTG/TP
4.04
-42
18
-42
R temporal – parietal – occipital
1708
inferior occipital gyrus
5.52
48
-63
-12
occipital pole
5.01
30
-90
-6
middle occipital gyrus
4.74
33
-75
3
posterior MTG
4.39
63
-36
-9
cerebellum
4.35
36
-42
-33
cerebellum
4.29
27
-78
-42
cerebellum
4.28
36
-84
-39
posterior STG/TPJ
4.26
57
-39
21
cerebellum
4.24
42
-48
-30
posterior STS/TPJ
4.22
42
-60
18
cerebellum
4.21
45
-51
-42
middle occipital gyrus
4.05
39
-81
12
Bilateral frontal
1017
L anterior SFG
5.46
-3
63
21
R anterior orbital gyrus
5.30
9
45
-24
R anterior gyrus rectus
5.21
6
51
-18
L middle SFG
5.08
-12
60
27
L middle SFG
4.93
-12
57
36
R middle SFG
4.81
3
60
30
R anterior mPFC
4.77
9
63
-9
R anterior mPFC
4.77
3
57
-6
R anterior mPFC
3.45
3
60
6
L temporal-parietal
362
superior parietal lobule
4.65
-54
-27
18
middle STG
3.60
-63
-15
9
R anterior temporal
160
anterior MTG
4.57
60
6
-33
anterior ITG
4.54
51
12
-45
anterior MTG/TP
3.97
51
18
-39
anterior MTG/TP
3.54
45
24
-39
anterior MTG
3.37
60
-9
-24
L dorsal frontal
410
middle superior frontal sulcus
4.38
-24
3
60
posterior superior frontal sulcus
4.29
-24
0
48
posterior SFG
4.15
-12
-12
78
posterior superior frontal sulcus
4.01
-30
-6
63
posterior SFG
3.30
-21
27
60
R parietal
152
superior postcentral gyrus
4.29
24
-42
57
superior parietal lobule
4.09
18
-57
60
precuneus
3.32
9
-63
54
The table shows up to 12 local maxima per cluster more than 8.0 mm apart. L= left; R= right; ITG =
inferior temporal gyrus; TP= temporal pole; MTG= middle temporal gyrus; TPJ= temporo-parietal junction;
AG= angular gyrus; SFG = superior frontal gyrus; mPFC= medial frontal cortex; STS= superior temporal sulcus
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

We used an a priori ROI-based approach to compare the magnitude of regional
494
responses to the social attribution task both within each hemisphere and between hemispheric
495
homologues. We focused upon two key ATL subregions, the temporopolar cortex and the
496
posteriorly adjacent ventrolateral surface, as well as temporoparietal cortex (i.e., the TPJ)
497
frequently implicated in theory of mind. The positions of these ROIs and the results are
498
displayed in Figure 1, panel B. Bonferroni-corrected one-sample T-tests revealed significant
499
activation during social interaction judgements in the left vATL (t (24) = 5.09, Cohen’s d=
500
1.04), temporal pole (t (24) = 3.77, Cohen’s d= .77 ) and TPJ (t (24) = .20 , Cohen’s d= .04 )
501
and also the right vATL (t (24) = 3.57, Cohen’s d= .73), temporal pole (t (24) = 4.03, Cohen’s
502
d= .82) and TPJ (t (24) = .17, Cohen’s d= .04) (all p < .005). Numerically speaking, across all
503
the ROIs, the left vATL revealed the largest effect size, and the TPJ showed the weakest effects.
504
Planned statistical comparisons (see Supplementary Table R3) confirmed greater activation
505
in the left as compared to the right vATL (t (24) = 2.45, p = .02, Cohen’s d =.50). There were
506
no other significant pairwise differences.
507
508
Common Activation of the ATL Across Three Different ToM Paradigms
509
In the subsequent analysis, we aimed to map out subregions of the bilateral ATL in
510
which there is overlapping activation between some of the different types of behavioural
511
paradigm used to localise the putative ToM network (Dodell-Feder et al. 2011; Jacoby et al.
512
2016). The results of independent whole-brain analyses contrasting two further ToM tasks (the
513
False Belief task and the free-viewing movie paradigm) with their respective control tasks are
514
reported in Supplementary Figures 3 and 4 and Supplementary Tables R4 and R5. We
515
formally assessed activation overlap between the three ToM tasks using a conjunction analysis
516
performed across the whole brain (Nichols et al. 2005b). For complete visualisation of the
517
results and to capture the full extent of both the overlap and divergence in the topography of
518
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

Figure 2. Panel A. Topological overlap of cortical regions activated by the interaction judgement > speed
judgement contrast, the false belief story > photograph contrast, and the mentalising > pain contrast from the
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

free-viewing movie localiser. Each of the three statistical maps were independently thresholded with an
uncorrected voxel height threshold of p < .001 and then overlaid within MRICron using additive colour blending.
White patches indicate three-way overlap between all three ToM contrasts. Panel B. Topological overlap of
cortical regions activated by the interaction judgement > speed judgement contrast (red), and the nonverbal
semantic association (Camel and Cactus task) > perceptual judgement contrast (green). The two statistical maps
were independently thresholded with an uncorrected voxel height threshold of p < .001 and then overlaid within
MRICron using additive colour blending. Yellow patches indicate overlap between theory of mind and general
semantic processing. Cross-sections were chosen to display the location of activation found in key studies
investigating ToM processing (Saxe & Kanwisher, 2003; right TPJ [51, -54, 27]), semantic processing of social
concepts (Binney, Hoffman & Lambon Ralph, 2016; left TP [-48, 9, -39]) and general semantic processing
(Visser, Jefferies, Embleton & Lambon Ralph, 2012; left inferior ATL [-57, -15, -24]), as well one further key
area of 3-way overlap (y = -7).
activation, the three whole brain activation maps are overlaid on each other in Figure 2 panel
519
A, whereas a map limited to the formal statistical conjunction can be found in Supplementary
520
Figure 5 and Supplementary Table R6. Regarding ATL activation, the conjunction analysis
521
revealed three-way overlap between the ToM tasks exclusively within the left ventrolateral
522
ATL. This extended over the anterior ITG and MTG from about y ≈ -7 to y ≈ 9 and is also
523
strikingly similar to ATL regions reported as activated by social concepts by (Binney, et al.
524
2016). As would be expected from prior literature, 3-way overlap was also observed in the
525
mPFC and bilateral TPJ.
526
527
ATL Activation Common to both ToM and General Semantic Processing
528
Finally, we performed a conjunction analysis aimed at identifying any potential overlap
529
between ATL regions engaged by theory of mind tasks and those engaged by general semantic
530
processing. The same sample of participants and the same ATL-optimised dual-echo imaging
531
sequence were used to acquire fMRI data while individuals completed a nonverbal semantic
532
association task. The result of an independent whole-brain analysis contrasting this task with
533
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

a matched control task is reported in Supplementary Figure 6 and Supplementary Tables
534
R7. This contrast was entered into a whole brain conjunction analysis along with the interacting
535
geometric shapes paradigm. The full extent of overlap and divergence between ToM activation
536
and general semantic activation is displayed in Figure 2 panel B, while the results of the formal
537
statistical conjunction are found in Supplementary Figure 7 and Supplementary Table R8.
538
Both theory of mind and general semantics activated the left ventrolateral ATL. Specifically,
539
there was a cluster of 114 commonly activated voxels in the left ventral ATL with the activation
540
starting to converge at y ≈ -15, showing the most robust overlap at y ≈ -7, and still overlapping
541
in inferior polar regions at y ≈ 9. There was a further common ATL activation (extent = 32
542
voxels) within the left medial temporal pole. On the basis of this analysis, the right ATL
543
appeared only to be activated by the IJ-ToM task. Outside of the ATL region, there was also
544
overlap in the left pMTG and TPJ region, as well as the left mPFC and bilateral inferior
545
temporo-occipital regions.
546
547
Discussion
548
The present study was aimed at evaluating alternative accounts of the role of the anterior
549
temporal lobes (ATL) in social cognition. According to the social knowledge hypothesis and
550
the Social Information Processing Account, this region serves a domain-specific role in
551
processing socially-relevant semantic information, and its proponents have argued against any
552
domain-general role in semantic processing (Simmons et al. 2010; Olson et al. 2013). Further,
553
this proposal particularly pinpoints the dorsolateral subregions of this relatively large and
554
structurally heterogenous area (Zahn et al. 2007; Ross and Olson 2010). Another hypothesis,
555
which we refer to as the ‘dual hub’ account, distinguishes between two separate ATL hubs,
556
one for social semantic and one for general semantic processing. Alternatively, the graded
557
semantic hub hypothesis, holds that the ATL is a unified domain-general conceptual hub
558
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

involved in the representation of all manner of conceptual-level knowledge (Binney, et al.
559
2016; Lambon Ralph et al. 2017). According to this account, the ventrolateral ATL is a critical
560
centre-point for general semantic knowledge representation. Other ATL sub-regions, including
561
the dorsolateral surface and the poles, are characterised as having connectivity-driven graded
562
variations in semantic function, including a ‘sensitivity’ to information that is perceived
563
primarily within certain sensorimotor modalities and/or has a particular behavioural (e.g.,
564
social) relevance (Plaut 2002; Visser and Lambon Ralph 2011; Binney et al. 2012; Binney et
565
al. 2016). The key findings of the present study were as follows:
566
1. By using distortion-corrected dual-echo fMRI, we were able to confirm, within a
567
whole-brain analysis, that the bilateral ventrolateral ATL is engaged by a nonverbal
568
task of the kind that has frequently been employed to localise the putative theory of
569
mind network. A region of interest analysis revealed that the left ventrolateral ATL was
570
more activated than the right homologue, and both these regions were as robustly
571
activated as the TPJ which is another key node in the theory of mind network.
572
2. Moreover, the left ventrolateral ATL activation was confirmed as a key feature of
573
theory of mind by the fact that it was activated robustly across three different paradigms
574
employing a range of verbal and nonverbal stimuli.
575
3. Finally, the left ATL activation associated with theory of mind greatly overlapped with
576
that evoked by semantic association judgements performed on non-social picture
577
stimuli.
578
Overall, these findings support the hypothesis that the ATL is a domain-general conceptual
579
hub and suggest that its contribution to social cognition is related to the retrieval of a general
580
class of semantic knowledge representations (Binney and Ramsey 2020). These findings are
581
not compatible with the social knowledge hypothesis nor the dual hub hypothesis
582
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

The Functional Contribution of ATL Subregions to Social and Semantic Cognition
583
A link between certain parts of the ATL (e.g., temporopolar cortex; for a review see
584
Olson et al. 2013) and social cognition has been recognised for well over a century, owed in
585
part to the acclaimed work of Brown and Schafer (1888) and, later, Klüver and Bucy (1937)
586
who performed bilateral ATL resection in non-human primates. These investigations are best
587
known for the profound post-operative changes in social behaviour, including emotional
588
blunting and hypersexuality. However, Klüver and Bucy’s primary aims were to establish
589
whether these bilateral lesions led to high-level perceptual deficits, namely visual and auditory
590
associative agnosias or, as referred to by these authors, ‘psychic blindness’. Indeed, this set of
591
studies detail a broad symptom complex that was chiefly characterised by a failure to generate
592
the meaning of visual and auditory stimuli. Therefore, it appears that their subjects were
593
exhibiting multimodal semantic deficits that might explain, and not just co-present with, the
594
social-affective disturbances.
595
In more recent years, the social neurosciences have seen another rise in interest
596
regarding the specific role played by the ATL (for a review see Olson et al. 2013). In particular,
597
there emerged the social knowledge hypothesis, which states that this region supports a domain-
598
specific class of semantic knowledge: social concepts (Zahn et al. 2007; Ross and Olson 2010;
599
Simmons et al. 2010). Although this account acknowledges supporting evidence from within
600
comparative and behavioural neurology, it is primarily based on functional neuroimaging data
601
which specifically points to the dorsolateral and polar ATL (also see Zahn et al., 2007).
602
Another long-standing series of studies have implicated the ATL in more general forms
603
of semantic processing (Lambon Ralph et al. 2017). These include detailed neuropsychological
604
investigations of a disorder known as semantic dementia (SD). The SD syndrome falls within
605
the spectrum of frontotemporal dementia and exhibits relatively focal atrophy and
606
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

hypometabolism centred on the bilateral anterior temporal lobes (Mummery et al. 2000; Nestor
607
et al. 2006). This is coupled with a progressive, central impairment of semantic memory that
608
is evident in both expressive and receptive semantic tasks, and across all modalities including
609
spoken and written language, object use, picture-based tasks, environmental sound tasks, and
610
in olfaction and taste (Hodges and Patterson 2007; Luzzi et al. 2007; Patterson et al. 2007;
611
Piwnica Worms et al. 2010). Moreover, this human disorder displays striking parallels to the
612
observations of Klüver and Bucy, in that the multimodal semantic deficit is accompanied by a
613
range of socio-affective deficits, which include impaired emotion recognition and empathy,
614
impaired capacity for ToM, and a loss of person-specific knowledge (Edwards-Lee et al. 1997;
615
Binney et al. 2016; Snowden et al. 2018; Ding et al. 2020). This patient evidence is bolstered
616
by a now extensive set of multi-method studies that used electrophysiological recordings,
617
neurostimulation techniques (TMS/tDCS) and/or functional neuroimaging in neurotypical
618
samples (Marinkovic et al. 2003; Pobric et al. 2008; Binney et al. 2010; Chan et al. 2011;
619
Binney and Lambon Ralph 2015; Shimotake et al. 2015) all of which point to a role of the ATL
620
in general semantic processing. However, as compared to the social knowledge hypothesis, this
621
literature has converged upon a different subregion, the ventrolateral ATL, as the critical
622
substrate for semantic knowledge representation. This includes the findings of ATL-optimised
623
fMRI studies and the data from SD which reveals that the ventrolateral ATL is, alongside the
624
temporopolar regions, the most atrophied ATL subregion in this disorder (Galton et al. 2001;
625
Binney et al. 2010; Mion et al. 2010). Moreover, it is noteworthy that Klüver and Bucy (1939)
626
also remarked that the symptoms they observed in non-human primate’s failed to appear after
627
resections limited to the dorsolateral convolutions of the temporal lobe. Nor did they present
628
after severing connections of the temporal lobe to the frontal or to the occipital lobes.
629
The findings of the present study are most compatible with this second set of
630
observations and implicate the ventrolateral ATL in both social and general semantic
631
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

processing. To our knowledge, they represent the first firm demonstration using fMRI of
632
ventrolateral ATL activation during the types of social (and more specifically, theory of mind)
633
paradigms that are typically employed in the social neuroscience literature. This ATL
634
subregion is frequently missing from fMRI studies probing theory of mind because of
635
methodological considerations we were able to overcome (see below). The fact that three very
636
different theory of mind paradigms evoked ventrolateral ATL activation suggest that it is a
637
feature of ToM irrespective of the paradigm with which it is probed and therefore that it reflects
638
a core cognitive component of theory of mind. Moreover, the fact that this activation
639
overlapped directly with that evoked by a set of non-social semantic judgements is consistent
640
with the claim that engagement of the ATL by social tasks reflects access to a general class of
641
domain-general conceptual representations (Binney and Ramsey 2020).
642
Our results complement recent studies that found evidence of a role of the left
643
ventrolateral ATL in accessing abstract social concepts (Binney et al. 2016; Rice et al. 2018)
644
as well as other forms of social conceptual knowledge such as person semantics (Rice et al.
645
2018). Moreover, this region responds to meaningful stimuli across number of domains
646
including faces, bodies, objects, and linguistic stimuli (Visser et al. 2012; Avidan et al. 2014;
647
Collins and Olson 2014; Harry et al. 2016; Ramot et al. 2019). This, alongside the fact that we
648
were able to directly demonstrate ventrolateral ATL activation common to both nonverbal (the
649
interacting shapes task) and verbal (the false belief vignettes) theory of mind tasks is consistent
650
with the notion that the ventrolateral ATL is a supramodal hub engaged in semantic retrieval
651
irrespective of the sensory, motor or linguistic modality through which concepts are probed
652
(Lambon Ralph et al. 2017).
653
The dorsolateral ATL subregion previously implicated in the representation of social
654
conceptual knowledge (e.g., Zahn et al. 2007) was notably absent within our main set of
655
contrasts. This was unexpected and the reasons are unclear. It is important to note however that
656
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

an absence of evidence for activation within this and other ATL subregions does not necessarily
657
preclude the possibility that they also have a role in theory of mind and/or general semantic
658
processing. Instead, it is possible that at the level of stringent thresholding that we employed
659
in this study we have only been able to identify the strongest peaks of activation, or functional
660
epicenters. In this case, it might be that activation in other ATL subregions only becomes
661
detectable at a supra-threshold level when there is increased power from higher levels of
662
sampling or when more finely tuned contrasts are used to evoke greater effects sizes. We did
663
contrast semantic judgments made on social and non-social stimuli in two prior studies, and
664
these revealed a sensitivity of activation to social stimuli in the dorsolateral ATL (Binney et al.
665
2016; Rice et al. 2018). However, these findings do not support a dual hub account of the ATL
666
in which there are discrete functional subdivisions for difference classes of concept. Instead,
667
they were in alignment with a ‘graded hub’ account in which the whole ATL comprises a single
668
semantic hub but it has graded subspecialisations towards certain types of conceptual
669
information (Plaut 2002; Binney et al. 2012; Rice et al. 2015). This is because the adjacent
670
ventrolateral ATL responded equally to both the social and non-social stimuli, and to a much
671
greater extent than the dorsolateral subregion. According to graded hub hypothesis, the
672
ventrolateral ATL region is the centre-point of the hub and has a modality/domain/category-
673
general semantic function. The sensitivity of the dorsolateral/polar ATL to social stimuli may
674
follow from this subregion’s close proximity to and strong connectivity with the limbic system
675
(via the uncinate fasciculus; Binney et al. 2012; Papinutto et al. 2016; Bajada et al. 2017), and
676
could reflect a specialisation in the assimilation of, for example, emotion-related or
677
interoceptive information into coherent semantic representations (Olson et al. 2007; Vigliocco
678
et al. 2014; Rice et al. 2015).
679
A clear difference the way in which the ATL was engaged by the semantic judgements
680
and the theory of mind tasks is that the latter appears far more bilateral. Moreover, ToM elicited
681
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

bilateral ATL activation regardless of the verbal/non-verbal nature of the stimuli. The role of
682
the ATL in semantic cognition is proposed to be bilateral although, again, perhaps with graded
683
specialisations towards processing verbal semantic information in the left hemisphere (Lambon
684
Ralph et al. 2001; Rice et al. 2015). The role of the ATL in social cognition has been ascribed
685
with a right lateralisation within some accounts drawing mainly on neuropsychological patient
686
evidence (Gorno-tempini et al. 2003; Zahn et al. 2009; Irish et al. 2014; Gainotti 2015; Rice et
687
al. 2017; Borghesani et al. 2019) although the fMRI studies reviewed above (Ross and Olson
688
2010; Binney et al. 2016; Rice et al. 2018) indicate bilateral involvement (also see Rice,
689
Lambon Ralph, et al. 2015; Pobric et al. 2016; Lin et al. 2018; Arioli et al. 2020; Catricalà et
690
al. 2020). Contrary to all these prior findings, the present ROI results are suggestive of a greater
691
response in the left as compared to the right ATL. The reasons for this are unclear, and an
692
interesting aim for future neuroimaging studies is to explore factors (e.g., task; stimulus
693
modality) that could potentially drive differences in the activation of bilateral ATL subregions
694
both in the context of social and general semantic tasks.
695
696
The status of the ATL in neurobiological accounts of social cognition
697
Animal ablation studies (Brown and Schafer 1888; Klüver and Bucy 1937) and case
698
descriptions of the profound consequences for humans of focal ATL lesions (Terzian and Dalle
699
Ore 1955) and degeneration (e.g. Edwards-Lee et al. 1997) provided some relatively early clues
700
as to the importance of the anterior temporal cortex for socio-affective competences.
701
Nonetheless, the ATL often does not feature prominently within contemporary neurobiological
702
frameworks for understanding social behaviour (Decety and Lamm 2007; Lieberman 2007;
703
Adolphs 2009; van Overwalle 2009; Spunt and Adolphs 2017). It is overshadowed by
704
prefrontal, medial and lateral temporoparietal regions, and seemingly attributed with an
705
ancillary status. This could be due, at least in part, to the predominance of fMRI in the social
706
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

neurosciences and the fact that this technique is typically blind to activation in a significant
707
proportion of this region (Devlin 2002). Inconsistencies in the presence and location of ATL
708
activation across various social domains, relative to the TPJ for example, could explain a
709
modest appetite for further exploring the region’s contribution.
710
Here, and in two prior ATL-optimised fMRI studies (Binney et al. 2016; Rice et al.
711
2018), we have shown that when steps are taken to alleviate the technical limitations of the
712
fMRI technique, robust ATL activations are observed across a variety of social stimuli and
713
social tasks. Activation also occurs in a ventrolateral ATL region that is one of the most
714
affected in patients with both striking semantic and social impairments (Binney et al. 2010;
715
Binney et al. 2016; Kumfor et al. 2016). Moreover, in the present study, we have demonstrated
716
that left ventrolateral ATL activation is at least as robust, in extent and magnitude, as that of
717
another key social region (the TPJ), and at least as consistent across different tasks and stimuli.
718
Overall, we interpret this as initial evidence from neurotypical samples to complement that
719
obtained from patient studies, that the ventrolateral ATL is of equal functional import to social
720
cognition as other key nodes of the ‘social brain’ (such as the TPJ, the mPFC and the
721
precuneus).
722
Several authors have argued that progress in social neuroscience theory will rapidly
723
accelerate if it embraces established and detailed models from within other more general
724
domains of cognition (Spunt and Adolphs 2017; Amodio 2019; Ramsey and Ward 2020).
725
Taking a similar perspective, we have recently proposed that a unifying feature amongst many
726
forms of social cognitive processing is the retrieval of conceptual knowledge, and that it could
727
be productive to understand social cognition to essentially be an example of semantic cognition
728
(Binney and Ramsey 2020). This would appear a reasonable viewpoint given that social
729
interaction is, at its core, a process of meaningful exchange between persons. The main
730
practical implication of this proposal, at least for the present discussion, is that social and
731
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

semantic cognition rely on the same cognitive and brain mechanisms, and this positions the
732
ventrolateral ATL at the heart of social cognition. According to this framework, other key
733
nodes of the ‘social brain’, including the mPFC and the TPJ, could also serve a domain-general
734
role rather than one that is specialised towards processing social information (van Overwalle
735
2009; Seghier et al. 2010; Cabeza et al. 2012; Bzdok et al. 2016; Humphreys et al. 2020;
736
Diveica et al. 2021). In summary, we argue that there is a growing need to re-evaluate the
737
relative contribution of all these regions, as well as develop a better understanding of the way
738
they interact in service of social cognition.
739
740
Conclusions and Future Directions
741
In conclusion, our findings support the claim that the ventrolateral ATL is an important
742
contributor to social cognition and point to a specific role as a domain-general hub for
743
conceptual knowledge representations that help inform our understanding of others and guide
744
our own meaning-driven social behaviours. A key methodological determinant underpinning
745
these findings was the use of a neuroimaging technique that maximises the signal obtained
746
from across the entire ATL region. However, the present study is also limited by its
747
methodology. To a large extent, fMRI remains the predominant mode of investigation in the
748
social neurosciences. However, it cannot be escaped that the inferences it allows are merely
749
correlational and not at all causal. For this reason, the field needs to increasingly turn to patient
750
models such as stroke, temporal lobe epilepsy, and frontotemporal dementia (Kumfor et al.
751
2017; Rankin 2020, 2021), as well as non-invasive techniques, such as transcranial magnetic
752
stimulation, that can be used to more directly probe the neural architecture of cognition in
753
neurological healthy samples. This will enable us to get a firmer grasp on key questions
754
including those regarding the laterality of function within the ATL and the TPJ, as well as the
755
functional necessity of distinct subregions.
756
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

Acknowledgements: The authors would like to thank Jordan Bryne and Taylor Baumler for
757
their assistance with data collection and recruitment, and Paul Downing and Kami Koldewyn
758
for their comments on a previous version of this manuscript.
759
References
Abell F, Happé F, Frith U. 2000. Do triangles play tricks? Attribution of mental states to animated shapes in
760
normal and abnormal development. Cognitive Development. 15:1–16.
761
Adolphs R. 2009. The Social Brain: Neural Basis of Social Knowledge. Annual Review of Psychology. 60:693–
762
716.
763
Amodio DM. 2019. Social Cognition 2.0: An Interactive Memory Systems Account. Trends in Cognitive
764
Sciences. 23:21–33.
765
Andrews-Hanna JR, Smallwood J, Spreng RN. 2014. The default network and self-generated thought: component
766
processes, dynamic control, and clinical relevance. Annals of the New York Academy of Sciences. 1316:29–
767
52.
768
Apperly IA. 2012. What is “theory of mind”? Concepts, cognitive processes and individual differences. Quarterly
769
Journal of Experimental Psychology. 65:825–839.
770
Arioli M, Gianelli C, Canessa N. 2020. Neural representation of social concepts: a coordinate-based meta-analysis
771
of fMRI studies. Brain Imaging and Behavior.
772
Ashburner J. 2007. A fast diffeomorphic image registration algorithm. NeuroImage. 38:95–113.
773
Avidan G, Tanzer M, Hadj-Bouziane F, Liu N, Ungerleider LG, Behrmann M. 2014. Selective dissociation
774
between core and extended regions of the face processing network in congenital prosopagnosia. Cerebral
775
Cortex. 24:1565–1578.
776
Bajada CJ, Haroon HA, Azadbakht H, Parker GJM, Lambon Ralph MA, Cloutman LL. 2017. The tract
777
terminations in the temporal lobe: Their location and associated functions. Cortex. 97:277–290.
778
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

Binder JR, Desai RH, Graves WW, Conant LL. 2009. Where is the semantic system? A critical review and meta-
779
analysis of 120 functional neuroimaging studies. Cerebral Cortex. 19:2767–2796.
780
Binder JR, Frost JA, Hammeke TA, Bellgowan PSF, Rao SM, Cox RW. 1999. Conceptual Processing during the
781
Conscious Resting State: A Functional MRI Study. Journal of Cognitive Neuroscience. 11:80–93.
782
Binney RJ, Embleton K v., Jefferies E, Parker GJMM, Lambon Ralph MA. 2010. The ventral and inferolateral
783
aspects of the anterior temporal lobe are crucial in semantic memory: Evidence from a novel direct
784
comparison of distortion-corrected fMRI, rTMS, and semantic dementia. Cerebral Cortex.
785
Binney RJ, Henry ML, Babiak M, Pressman PS, Santos-Santos MA, Narvid J, Mandelli ML, Strain PJ, Miller
786
BL, Rankin KP, Rosen HJ, Gorno-Tempini ML. 2016. Reading words and other people: A comparison of
787
exception word, familiar face and affect processing in the left and right temporal variants of primary
788
progressive aphasia. Cortex. 82:147–163.
789
Binney RJ, Hoffman P, Lambon Ralph MA. 2016. Mapping the Multiple Graded Contributions of the Anterior
790
Temporal Lobe Representational Hub to Abstract and Social Concepts: Evidence from Distortion-corrected
791
fMRI. Cerebral Cortex. 26:4227–4241.
792
Binney RJ, Lambon Ralph MA. 2015. Using a combination of fMRI and anterior temporal lobe rTMS to measure
793
intrinsic and induced activation changes across the semantic cognition network. Neuropsychologia. 76:170–
794
181.
795
Binney RJ, Parker GJM, Lambon Ralph MA. 2012. Convergent Connectivity and Graded Specialization in the
796
Rostral Human Temporal Lobe as Revealed by Diffusion-Weighted Imaging Probabilistic Tractography.
797
Journal of Cognitive Neuroscience. 24:1998–2014.
798
Binney RJ, Ramsey R. 2020. Social Semantics: The role of conceptual knowledge and cognitive control in a
799
neurobiological model of the social brain. Neuroscience and Biobehavioral Reviews.
800
Bliksted V, Frith C, Videbech P, Fagerlund B, Emborg C, Simonsen A, Roepstorff A, Campbell-Meiklejohn D.
801
2019. Hyper- and Hypomentalizing in Patients with First-Episode Schizophrenia: fMRI and Behavioral
802
Studies. Schizophrenia Bulletin. 45:377–385.
803
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

Borghesani V, Narvid J, Battistella G, Shwe W, Watson C, Binney RJ, Sturm V, Miller Z, Mandelli ML, Miller
804
B, Gorno-Tempini ML. 2019. “Looks familiar, but I do not know who she is”: The role of the anterior right
805
temporal lobe in famous face recognition. Cortex.
806
Bozeat S, Lambon Ralph MA, Patterson K, Garrard P, Hodges JR. 2000. Non-verbal semantic impairment in
807
semantic dementia. Neuropsychologia. 38:1207–1215.
808
Brainard DH. 1997. The Psychophysics Toolbox. Spatial Vision. 10:433–436.
809
Brett Anton J. L. Valabregue R. & Poline J. B. M, Brett, M., Anton, J. L., Valabregue, R., & Poline JB. 2002.
810
Region of interest analysis using the MarsBar toolbox for SPM 99. Neuroimage. 16.
811
Brown S, Sharpey-Schafer EA. 1888. XI. An investigation into the functions of the occipital and temporal lobes
812
of the monkey’s brain. Philosophical Transactions of the Royal Society of London(B). 303–327.
813
Brüne M, Brüne-Cohrs U. 2006. Theory of mind-evolution, ontogeny, brain mechanisms and psychopathology.
814
Neuroscience and Biobehavioral Reviews.
815
Bzdok D, Hartwigsen G, Reid A, Laird AR, Fox PT, Eickhoff SB. 2016. Left inferior parietal lobe engagement
816
in social cognition and language. Neuroscience and Biobehavioral Reviews. 68:319–334.
817
Cabeza R, Ciaramelli E, Moscovitch M. 2012. Cognitive contributions of the ventral parietal cortex: an integrative
818
theoretical account. Trends in Cognitive Sciences. 16:338–352.
819
Castelli F. 2002. Autism, Asperger syndrome and brain mechanisms for the attribution of mental states to animated
820
shapes. Brain. 125:1839–1849.
821
Catricalà E, Conca F, Fertonani A, Miniussi C, Cappa SF. 2020. State-dependent TMS reveals the differential
822
contribution of ATL and IPS to the representation of abstract concepts related to social and quantity
823
knowledge. Cortex. 123:30–41.
824
Chan AM, Baker JM, Eskandar E, Schomer D, Ulbert I, Marinkovic K, Cash SS, Halgren E. 2011. First-pass
825
selectivity for semantic categories in human anteroventral temporal lobe. Journal of Neuroscience.
826
31:18119–18129.
827
Collins JA, Koski JE, Olson IR. 2016. More Than Meets the Eye: The Merging of Perceptual and Conceptual
828
Knowledge in the Anterior Temporal Face Area. Frontiers in Human Neuroscience. 10:189.
829
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

Collins JA, Olson IR. 2014. Knowledge is power: How conceptual knowledge transforms visual cognition.
830
Psychonomic Bulletin and Review. 21:843–860.
831
Das P, Lagopoulos J, Coulston CM, Henderson AF, Malhi GS. 2012. Mentalizing impairment in schizophrenia:
832
A functional MRI study. Schizophrenia Research. 134:158–164.
833
Decety J, Lamm C. 2007. The role of the right temporoparietal junction in social interaction: How low-level
834
computational processes contribute to meta-cognition. Neuroscientist. 13:580–593.
835
Devlin J. 2002. Is there an anatomical basis for category-specificity? Semantic memory studies in PET and fMRI.
836
Neuropsychologia. 40:54–75.
837
Devlin JT, Russell RP, Davis MH, Price CJ, Wilson J, Moss HE, Matthews PM, Tyler LK. 2000. Susceptibility-
838
Induced Loss of Signal: Comparing PET and fMRI on a Semantic Task. NeuroImage. 11:589–600.
839
Ding H, Jiang X, Member S, Shuai B, Liu AQ, Wang G, Member S. 2020. and Multi-Path Decoding. 29:3520–
840
3533.
841
Ding S-L, van Hoesen GW, Cassell MD, Poremba A. 2009. Parcellation of human temporal polar cortex: A
842
combined analysis of multiple cytoarchitectonic, chemoarchitectonic, and pathological markers. The
843
Journal of Comparative Neurology. 514:595–623.
844
Diveica V, Koldweyn K, Binney RJ. 2021. Establishing a Role of the Semantic Control Network in Social
845
Cognitive Processing: A Meta-analysis of Functional Neuroimaging Studies. bioRxiv.
846
Dodell-Feder D, Koster-Hale J, Bedny M, Saxe R. 2011. FMRI item analysis in a theory of mind task.
847
NeuroImage. 55:705–712.
848
Edwards-Lee T, Miller BL, Benson DF, Cummings JL, Russell GL, Boone K, Mena I. 1997. The temporal variant
849
of frontotemporal dementia. Brain. 120:1027–1040.
850
Embleton K v., Haroon HA, Morris DM, Ralph MAL, Parker GJMM. 2010. Distortion correction for diffusion-
851
weighted MRI tractography and fMRI in the temporal lobes. Human Brain Mapping. 31:1570–1587.
852
Frith C, Frith U. 2005. Theory of mind. Current Biology. 15:R644–R645.
853
Frith CD, Frith U. 2006. The Neural Basis of Mentalizing. Neuron. 50:531–534.
854
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

Frith U, Frith C. 2010. The social brain: allowing humans to boldly go where no other species has been.
855
Philosophical Transactions of the Royal Society B: Biological Sciences. 365:165–176.
856
Frith U, Frith CD. 2003. Development and neurophysiology of mentalizing. Philosophical Transactions of the
857
Royal Society of London Series B: Biological Sciences. 358:459–473.
858
Gainotti G. 2015. Is the difference between right and left ATLs due to the distinction between general and social
859
cognition or between verbal and non-verbal representations? Neuroscience and Biobehavioral Reviews.
860
51:296–312.
861
Gallagher HL, Frith CD. 2003. Functional imaging of ‘theory of mind.’ Trends in Cognitive Sciences. 7:77–83.
862
Galton CJ, Patterson K, Graham K, Lambon-Ralph MA, Williams G, Antoun N, Sahakian BJ, Hodges JR. 2001.
863
Differing patterns of temporal atrophy in Alzheimer’s disease and semantic dementia. Neurology. 57:216–
864
225.
865
Gorno-tempini ML, Rankin KP, Woolley JD, Rosen HJ, Phengrasamy L, Miller BL. 2003. Right Temporal
866
Atrophy.
867
Halai AD, Parkes LM, Welbourne SR. 2015. Dual-echo fMRI can detect activations in inferior temporal lobe
868
during intelligible speech comprehension. NeuroImage. 122:214–221.
869
Halai AD, Welbourne SR, Embleton K, Parkes LM. 2014. A comparison of dual gradient-echo and spin-echo
870
fMRI of the inferior temporal lobe. Human Brain Mapping. 35:4118–4128.
871
Harry BB, Umla-Runge K, Lawrence AD, Graham KS, Downing PE. 2016. Evidence for integrated visual face
872
and body representations in the anterior temporal lobes. Journal of Cognitive Neuroscience.
873
Heleven E, van Overwalle F. 2018. The neural basis of representing others’ inner states. Current Opinion in
874
Psychology.
875
Hennion S, Delbeuck X, Koelkebeck K, Brion M, Tyvaert L, Plomhause L, Derambure P, Lopes R, Szurhaj W.
876
2016. A functional magnetic resonance imaging investigation of theory of mind impairments in patients
877
with temporal lobe epilepsy. Neuropsychologia. 93:271–279.
878
Hodges JR, Patterson K. 2007. Semantic dementia: a unique clinicopathological syndrome. The Lancet
879
Neurology. 6:1004–1014.
880
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

Hoffman P, Binney RJ, Lambon Ralph MA. 2015. Differing contributions of inferior prefrontal and anterior
881
temporal cortex to concrete and abstract conceptual knowledge. Cortex.
882
Hoffman P, Pobric G, Drakesmith M, Lambon Ralph MA. 2012. Posterior middle temporal gyrus is involved in
883
verbal and non-verbal semantic cognition: Evidence from rTMS. Aphasiology. 26:1119–1130.
884
Humphreys GF, Jackson RL, Lambon Ralph MA. 2020. Overarching Principles and Dimensions of the Functional
885
Organization in the Inferior Parietal Cortex. Cerebral Cortex. 30:5639–5653.
886
Irish M, Hodges JR, Piguet O. 2014. Right anterior temporal lobe dysfunction underlies theory of mind
887
impairments in semantic dementia. Brain. 137:1241–1253.
888
Jacoby N, Bruneau E, Koster-Hale J, Saxe R. 2016. Localizing Pain Matrix and Theory of Mind networks with
889
both verbal and non-verbal stimuli. NeuroImage. 126:39–48.
890
Jefferies E, Lambon Ralph MA. 2006. Semantic impairment in stroke aphasia versus semantic dementia: A case-
891
series comparison. Brain. 129:2132–2147.
892
Kana RK, Keller T, Cherkassky VL, Minshew NJ, Just MA. 2009. Atypical frontal-posterior synchronization of
893
Theory of Mind regions in autism during mental state attribution. Social Neuroscience. 4:135–152.
894
Klüver H, Bucy PC. 1937. “ Psychic blindness” and other symptoms following bilateral temporal lobectomy in
895
Rhesus monkeys. American Journal of Physiology. 119:3552–353.
896
Kumfor F, Hazelton JL, de Winter F-L, de Langavant LC, van den Stock J. 2017. Clinical Studies of Social
897
Neuroscience: A Lesion Model Approach. In: Neuroscience and Social Science. Cham: Springer
898
International Publishing. p. 255–296.
899
Kumfor F, Honan C, McDonald S, Hazelton JL, Hodges JR, Piguet O. 2017. Assessing the “social brain” in
900
dementia: Applying TASIT-S. Cortex. 93:166–177.
901
Kumfor F, Irish M, Hodges JR, Piguet O. 2013. Discrete Neural Correlates for the Recognition of Negative
902
Emotions: Insights from Frontotemporal Dementia. PLoS ONE. 8.
903
Kumfor F, Landin-Romero R, Devenney E, Hutchings R, Grasso R, Hodges JR, Piguet O. 2016. On the right
904
side? A longitudinal study of left-versus right-lateralized semantic dementia. Brain. 139:986–998.
905
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

Kumfor F, Piguet O. 2012. Disturbance of emotion processing in frontotemporal dementia: A synthesis of
906
cognitive and neuroimaging findings. Neuropsychology Review.
907
Lambon Ralph MA, Mcclelland JL, Patterson K, Galton CJ, Hodges JR. 2001. No right to speak? The relationship
908
between object naming and semantic impairment: Neuropsychological evidence and a computational model.
909
Journal of Cognitive Neuroscience. 13:341–356.
910
Lieberman MD. 2007. Social Cognitive Neuroscience: A Review of Core Processes. Annual Review of
911
Psychology. 58:259–289.
912
Lin N, Wang X, Xu Y, Wang X, Hua H, Zhao Y, Li X. 2018. Fine Subdivisions of the Semantic Network
913
Supporting Social and Sensory–Motor Semantic Processing. Cerebral Cortex. 28:2699–2710.
914
Lin N, Yang X, Li J, Wang S, Hua H, Ma Y, Li X. 2018. Neural correlates of three cognitive processes involved
915
in theory of mind and discourse comprehension. Cognitive, Affective and Behavioral Neuroscience.
916
18:273–283.
917
Luzzi S, Snowden JS, Neary D, Coccia M, Provinciali L, Lambon Ralph MA. 2007. Distinct patterns of olfactory
918
impairment in Alzheimer’s disease, semantic dementia, frontotemporal dementia, and corticobasal
919
degeneration. Neuropsychologia. 45:1823–1831.
920
Marinkovic K, Dhond RP, Dale AM, Glessner M, Carr V, Halgren E. 2003. Spatiotemporal Dynamics of
921
Modality-Specific and Supramodal Word Processing The importance of the left anterior temporal cortex
922
for. Neuron. 38:487–497.
923
Mellem MS, Jasmin KM, Peng C, Martin A. 2016. Sentence processing in anterior superior temporal cortex shows
924
a social-emotional bias. Neuropsychologia. 89:217–224.
925
Mion M, Patterson K, Acosta-Cabronero J, Pengas G, Izquierdo-Garcia D, Hong YT, Fryer TD, Williams GB,
926
Hodges JR, Nestor PJ. 2010. What the left and right anterior fusiform gyri tell us about semantic memory.
927
Brain. 133:3256–3268.
928
Molenberghs P, Johnson H, Henry JD, Mattingley JB. 2016. Understanding the minds of others: A neuroimaging
929
meta-analysis. Neuroscience and Biobehavioral Reviews. 65:276–291.
930
Moll J, Zahn R, de Oliveira-Souza R, Krueger F, Grafman J. 2005. Opinion: The neural basis of human moral
931
cognition. Nature Reviews Neuroscience.
932
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

Mummery CJ, Patterson K, Price CJ, Ashburner J, Frackowiak RSJ, Hodges JR. 2000. A voxel-based
933
morphometry study of semantic dementia: Relationship between temporal lobe atrophy and semantic
934
memory. Annals of Neurology. 47:36–45.
935
Munafò MR, Nosek BA, Bishop DVM, Button KS, Chambers CD, Percie Du Sert N, Simonsohn U, Wagenmakers
936
EJ, Ware JJ, Ioannidis JPA. 2017. A manifesto for reproducible science. Nature Human Behaviour.
937
Nestor PJ, Fryer TD, Hodges JR. 2006. Declarative memory impairments in Alzheimer’s disease and semantic
938
dementia. NeuroImage. 30:1010–1020.
939
Nichols T, Brett M, Andersson J, Wager T, Poline JB. 2005a. Valid conjunction inference with the minimum
940
statistic. NeuroImage. 25:653–660.
941
Nichols T, Brett M, Andersson J, Wager T, Poline JB. 2005b. Valid conjunction inference with the minimum
942
statistic. NeuroImage. 25:653–660.
943
Oldfield RC. 1971. The assessment and analysis of handedness: The Edinburgh inventory. Neuropsychologia.
944
9:97–113.
945
Olson IR, McCoy D, Klobusicky E, Ross LA. 2013. Social cognition and the anterior temporal lobes: A review
946
and theoretical framework. Social Cognitive and Affective Neuroscience.
947
Olson IR, Plotzker A, Ezzyat Y. 2007. The Enigmatic temporal pole: A review of findings on social and emotional
948
processing. Brain.
949
Papinutto N, Galantucci S, Mandelli ML, Gesierich B, Jovicich J, Caverzasi E, Henry RG, Seeley WW, Miller
950
BL, Shapiro KA, Gorno-Tempini ML. 2016. Structural connectivity of the human anterior temporal lobe:
951
A diffusion magnetic resonance imaging study. Human Brain Mapping. 37:2210–2222.
952
Pascual B, Masdeu JC, Hollenbeck M, Makris N, Insausti R, Ding SL, Dickerson BC. 2015. Large-scale brain
953
networks of the human left temporal pole: A functional connectivity MRI study. Cerebral Cortex. 25:680–
954
702.
955
Patterson K, Nestor PJ, Rogers TT. 2007. Where do you know what you know? The representation of semantic
956
knowledge in the human brain. Nature Reviews Neuroscience.
957
Pelli DG. 1997. The VideoToolbox software for visual psychophysics: transforming numbers into movies. Spatial
958
Vision. 10:437–442.
959
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

Persichetti AS, Denning JM, Gotts SJ, Martin A. 2021. A Data-Driven Functional Mapping of the Anterior
960
Temporal Lobes. The Journal of Neuroscience. 41:6038–6049.
961
Piwnica Worms KE, Omar R, Hailstone JC, Warren JD. 2010. Flavour processing in semantic dementia. Cortex.
962
46:761–768.
963
Plaut DC. 2002. Graded modality-specific specialisation in semantics: A computational account of optic aphasia,
964
Cognitive Neuropsychology.
965
Pobric G, Jefferies E, Ralph AL. 2008. Anterior temporal lobes and non-verbal semantic processing: Brain
966
Stimulation. 1:307.
967
Pobric G, Ralph MAL, Zahn R. 2016. Hemispheric specialization within the superior anterior temporal cortex for
968
social and nonsocial concepts. Journal of Cognitive Neuroscience. 28:351–360.
969
Price CJ, Friston KJ. 1997. Cognitive conjunction: A new approach to brain activation experiments. NeuroImage.
970
5:261–270.
971
Ralph MAL, Jefferies E, Patterson K, Rogers TT. 2017. The neural and computational bases of semantic
972
cognition. Nature Reviews Neuroscience. 18:42–55.
973
Ramot M, Walsh C, Martin A. 2019. Multifaceted integration: Memory for faces is subserved by widespread
974
connections between visual, memory, auditory, and social networks. Journal of Neuroscience. 39:4976–
975
4985.
976
Ramsey R, Ward R. 2020. Putting the Nonsocial Into Social Neuroscience: A Role for Domain-General Priority
977
Maps During Social Interactions. Perspectives on Psychological Science. 15:1076–1094.
978
Rankin KP. 2020. Brain Networks Supporting Social Cognition in Dementia. Current Behavioral Neuroscience
979
Reports. 203–211.
980
Rankin KP. 2021. Measuring Behavior and Social Cognition in FTLD. In: Frontotemporal Dementias. Springer.
981
p. 51–65.
982
Rice GE, Caswell H, Moore P, Hoffman P, Lambon Ralph MA. 2017. The effects of left versus right anterior
983
temporal lobe resection on semantic processing of words, objects, and faces. Journal of the Neurological
984
Sciences. 381:684.
985
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

Rice GE, Hoffman P, Binney RJ, Lambon Ralph MA. 2018. Concrete versus abstract forms of social concept: An
986
fMRI comparison of knowledge about people versus social terms. Philosophical Transactions of the Royal
987
Society B: Biological Sciences. 373:20170136.
988
Rice GE, Hoffman P, Lambon Ralph MA. 2015. Graded specialization within and between the anterior temporal
989
lobes. Annals of the New York Academy of Sciences. 1359:84–97.
990
Rice GE, Ralph MAL, Hoffman P. 2015. The roles of left versus right anterior temporal lobes in conceptual
991
knowledge: An ALE meta-analysis of 97 functional neuroimaging studies. Cerebral Cortex. 25:4374–4391.
992
Ross LA, Olson IR. 2010. Social cognition and the anterior temporal lobes. NeuroImage. 49:3452–3462.
993
Saxe R. 2006. Why and how to study Theory of Mind with fMRI. Brain Research. 1079:57–65.
994
Saxe R, Kanwisher N. 2003a. People thinking about thinking people. The role of the temporo-parietal junction in
995
&quot;theory of mind&quot;. NeuroImage. 19:1835–1842.
996
Saxe R, Kanwisher N. 2003b. People thinking about thinking peopleThe role of the temporo-parietal junction in
997
“theory of mind.” NeuroImage. 19:1835–1842.
998
Saxe R, Powell LJ. 2006. It’s the Thought That Counts: Specific Brain Regions for One Component of Theory of
999
Mind, Source: Psychological Science.
1000
Saxe R, Wexler A. 2005. Making sense of another mind: the role of the right temporo-parietal junction.
1001
Neuropsychologia. 43:1391–1399.
1002
Scholz J, Triantafyllou C, Whitfield-Gabrieli S, Brown EN, Saxe R. 2009. Distinct regions of right temporo-
1003
parietal junction are selective for theory of mind and exogenous attention. PLoS ONE. 4.
1004
Schurz M, Radua J, Aichhorn M, Richlan F, Perner J. 2014. Fractionating theory of mind: A meta-analysis of
1005
functional brain imaging studies. Neuroscience & Biobehavioral Reviews. 42:9–34.
1006
Seghier ML, Fagan E, Price CJ. 2010. Functional Subdivisions in the Left Angular Gyrus Where the Semantic
1007
System Meets and Diverges from the Default Network. Journal of Neuroscience. 30:16809–16817.
1008
Shimotake A, Matsumoto R, Ueno T, Kunieda T, Saito S, Hoffman P, Kikuchi T, Fukuyama H, Miyamoto S,
1009
Takahashi R, Ikeda A, Lambon Ralph MA. 2015. Direct exploration of the role of the ventral anterior
1010
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

temporal lobe in semantic memory: Cortical stimulation and local field potential evidence from subdural
1011
grid electrodes. Cerebral Cortex. 25:3802–3817.
1012
Simmons WK, Martin A. 2009. The anterior temporal lobes and the functional architecture of semantic memory.
1013
Journal of the International Neuropsychological Society. 15:645–649.
1014
Simmons WK, Reddish M, Bellgowan PSFF, Martin A. 2010. The selectivity and functional connectivity of the
1015
anterior temporal lobes. Cerebral Cortex. 20:813–825.
1016
Snowden JS, Harris JM, Thompson JC, Kobylecki C, Jones M, Richardson AM, Neary D. 2018. Semantic
1017
dementia and the left and right temporal lobes. Cortex. 107:188–203.
1018
Spitsyna G, Warren JE, Scott SK, Turkheimer FE, Wise RJS. 2006. Converging language streams in the human
1019
temporal lobe. Journal of Neuroscience. 26:7328–7336.
1020
Spunt RP, Adolphs R. 2017. A new look at domain specificity: Insights from social neuroscience. Nature Reviews
1021
Neuroscience.
1022
Synn A, Mothakunnel A, Kumfor F, Chen Y, Piguet O, Hodges JR, Irish M. 2018. Mental states in moving shapes:
1023
distinct cortical and subcortical contributions to theory of mind impairments in dementia. Journal of
1024
Alzheimer’s Disease. 61:521–535.
1025
Terzian H, Dalle Ore G. 1955. Syndrome of Klüver and Bucy, reproduced in man by bilateral removal of the
1026
temporal lobes. Neurology.
1027
van Hoeck N, Begtas E, Steen J, Kestemont J, Vandekerckhove M, van Overwalle F. 2014. False belief and
1028
counterfactual reasoning in a social environment. NeuroImage. 90:315–325.
1029
van Overwalle F. 2009. Social cognition and the brain: A meta-analysis. Human Brain Mapping. 30:829–858.
1030
Vigliocco G, Kousta S-TT, della Rosa PA, Vinson DP, Tettamanti M, Devlin JT, Cappa SF. 2014. The neural
1031
representation of abstract words: The role of emotion. Cerebral Cortex. 24:1767–1777.
1032
Visser M, Embleton K v., Jefferies E, Parker GJ, Ralph MAL. 2010. The inferior, anterior temporal lobes and
1033
semantic memory clarified: Novel evidence from distortion-corrected fMRI. Neuropsychologia. 48:1689–
1034
1696.
1035
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint

Visser M, Jefferies E, Embleton K v., Lambon Ralph MA. 2012. Both the Middle Temporal Gyrus and the Ventral
1036
Anterior Temporal Area Are Crucial for Multimodal Semantic Processing: Distortion-corrected fMRI
1037
Evidence for a Double Gradient of Information Convergence in the Temporal Lobes. Journal of Cognitive
1038
Neuroscience. 24:1766–1778.
1039
Visser M, Jefferies E, Lambon Ralph MA, Ralph L, Visser M, Jefferies E, Lambon Ralph MA. 2010. Semantic
1040
Processing in the Anterior Temporal Lobes : A Meta-analysis of the Functional. Journal of Cognitive
1041
Neuroscience. 22:1083–1094.
1042
Visser M, Lambon Ralph MA. 2011. Differential Contributions of Bilateral Ventral Anterior Temporal Lobe and
1043
Left Anterior Superior Temporal Gyrus to Semantic Processes. Journal of Cognitive Neuroscience.
1044
23:3121–3131.
1045
Walbrin J, Downing P, Koldewyn K. 2018. Neural responses to visually observed social interactions.
1046
Neuropsychologia. 112:31–39.
1047
Wang X, Wang B, Bi Y. 2019. Close yet independent: Dissociation of social from valence and abstract semantic
1048
dimensions in the left anterior temporal lobe. Human Brain Mapping. 40:4759–4776.
1049
Wong C, Gallate J. 2012. The function of the anterior temporal lobe: A review of the empirical evidence. Brain
1050
Research. 1449:94–116.
1051
Young L, Dodell-Feder D, Saxe R. 2010. What gets the attention of the temporo-parietal junction? An fMRI
1052
investigation of attention and theory of mind. Neuropsychologia. 48:2658–2664.
1053
Zahn R, Moll J, Iyengar V, Huey ED, Tierney M, Krueger F, Grafman J. 2009. Social conceptual impairments in
1054
frontotemporal lobar degeneration with right anterior temporal hypometabolism. Brain. 132:604–616.
1055
Zahn R, Moll J, Krueger F, Huey ED, Garrido G, Grafman J. 2007. Social concepts are represented in the superior
1056
anterior temporal cortex. Proceedings of the National Academy of Sciences. 104:6430–6435.
1057
1058
.CC-BY 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted November 10, 2021. ; https://doi.org/10.1101/2021.09.10.459496doi: bioRxiv preprint