The embodiment of emotion: language use during the feeling of social emotions predicts cortical somatosensory activity.

Abstract

Complex social emotions involve both abstract cognitions and bodily sensations, and individuals may differ on their relative reliance on these. We hypothesized that individuals' descriptions of their feelings during a semi-structured emotion induction interview would reveal two distinct psychological styles-a more abstract, cognitive style and a more body-based, affective style-and that these would be associated with somatosensory neural activity. We examined 28 participants' open-ended verbal responses to admiration- and compassion-provoking narratives in an interview and BOLD activity to the same narratives during subsequent functional magnetic resonance imaging scanning. Consistent with hypotheses, individuals' affective and cognitive word use were stable across emotion conditions, negatively correlated and unrelated to reported emotion strength in the scanner. Greater use of affective relative to cognitive words predicted more activation in SI, SII, middle anterior cingulate cortex and insula during emotion trials. The results suggest that individuals' verbal descriptions of their feelings reflect differential recruitment of neural regions supporting physical body awareness. Although somatosensation has long been recognized as an important component of emotion processing, these results offer 'proof of concept' that individual differences in open-ended speech reflect different processing styles at the neurobiological level. This study also demonstrates SI involvement during social emotional experience.

Full text

Page 1

Embodiment of emotion 1
RUNNING HEAD: EMBODIMENT OF EMOTION















The embodiment of emotion: Language use during the feeling of social emotions predicts
cortical somatosensory activity
Saxbe, D.E., Yang, X.F., Borofsky, L.A., & Immordino-Yang, M.H.
University of Southern California




Corresponding author:
Mary Helen Immordino-Yang
University of Southern California
Assistant Professor of Psychology at the Brain and Creativity Institute,
Assistant Professor of Education at the Rossier School of Education,
Neuroscience Graduate Program Faculty
3641 Watt Way Suite B17 Los Angeles, CA 90089-2520
immordin@usc.edu
(213) 821-2969
© The Author (2012). Published by Oxford University Press. For Permissions, please email:
journals.permissions@oup.com

Social Cognitive and Affective Neuroscience Advance Access published July 13, 2012
by guest on July 15, 2012
http://scan.oxfordjournals.org/
Downloaded from

Page 2

Embodiment of emotion 2
Abstract

Complex social emotions involve both abstract cognitions and bodily sensations, and individuals
may differ on their relative reliance on these. We hypothesized that individuals’ descriptions of
their feelings during a semi-structured emotion induction interview would reveal two distinct
psychological styles – a more abstract, cognitive style, and a more body-based, affective style –
and that these would be associated with somatosensory neural activity. We examined 28
participants’ open-ended verbal responses to admiration- and compassion-provoking narratives
in an interview and BOLD activity to the same narratives during subsequent fMRI scanning.
Consistent with hypotheses, individuals’ affective and cognitive word use were stable across
emotion conditions, negatively correlated, and unrelated to reported emotion strength in the
scanner. Greater use of affective relative to cognitive words predicted more activation in SI, SII,
middle ACC, and insula during emotion trials. Results suggest that individuals’ verbal
descriptions of their feelings reflect differential recruitment of neural regions supporting physical
body awareness. Although somatosensation has long been recognized as an important component
of emotion processing, these results offer ‘proof of concept’ that individual differences in open-
ended speech reflect different processing styles at the neurobiological level. This study also
demonstrates SI involvement during social emotional experience.





Keywords: Emotion, somatosensory, language, fMRI
Word count: 4336





by guest on July 15, 2012
http://scan.oxfordjournals.org/
Downloaded from

Page 3

Embodiment of emotion 3
The embodiment of emotion: Language use during the feeling of social emotions predicts
cortical somatosensory activity
Emotions involve physiological changes in the brain and body that organize cognition
and can sometimes be felt (Barrett, Mesquita, Ochsner, & Gross, 2007; Damasio, 1994). Thus,
emotional feelings are driven not only by situational knowledge and cognitive assessments, but
by input from bodily reactions (Damasio, 1999; Craig, 2002; Harrison et al., 2011; Herbert &
Pollatos, 2012). These physical sensations underscore the fundamental interdependence between
body and mind, maintained even in psychologically complex emotional states.
Neurobiologically, complex social emotions such as sympathy (Decety & Chaminade, 2003),
moral indignation (Moll, et al., 2005), admiration, and compassion (Immordino-Yang, McColl,
Damasio & Damasio, 2009) have been found to recruit cortical and subcortical structures
involved in sensing and regulating internal organism states, such as the insula and dorsal anterior
middle cingulate cortex (dACC).
The relative contributions of body sensations and cognitive deliberations to emotional
experiences can vary not only by situation, but by individual—different people may rely more or
less on cues from the physical body in processing emotionally evocative information. Variation
in interoceptive ability, or somatosensory awareness for the internal body, has been associated
with variation in individuals’ expressed emotionality (Pollatos, Kirsch, & Schandry, 2005), trait
measures of affective experience (Critchley, Wiens, Rotshtein, O?hman & Dolan, 2004),
feelings of arousal in verbal reports of experienced emotion (Barrett, et al., 2004), and strength
of nonconscious fear priming (Katkin, Wiens, and O?hman, 2001). Areas of the brain
underlying somatosensory and interoceptive processing have also been implicated in individual
differences in traits relevant to social emotions, such as empathy. For example, self-report
by guest on July 15, 2012
http://scan.oxfordjournals.org/
Downloaded from

Page 4

Embodiment of emotion 4
measures of empathy have been associated with neural activation in the anterior insula and
anterior middle cingulate during tasks involving social pain (Eisenberger, 2003) and empathy for
another’s pain (Masten, Morelli, & Eisenberger, 2011; Singer, 2004).
Researchers have distinguished at both the behavioral and neural levels between “bottom-
up” emotion generation – that is, emotions that originate from visceral and autonomic responses
– and “top-down” emotions that result from cognitive processes such as appraisal (e.g., Ochsner
et al., 2009). However, these studies have generally not focused on individual differences and in
fact have actively minimized inter-individual variability, e.g., by instructing participants in how
to reappraise stimuli (e.g., McRae et al., 2011; Ochsner et al., 2009). Nonetheless, a small body
of research has identified that variability in both top-down and bottom-up processes across
individuals exists (e.g., Hamann & Canli, 2004; Ray et al., 2005), suggesting that some people
may spontaneously rely more on body sensation when formulating emotional feelings, while
others may take a more abstract or intellectualized approach. These differing emotion processing
styles may manifest themselves in language: for example, metaphors often include “embodied”
or visceral language (e.g., to grasp an idea or to feel hot with passion) that have been suggested
to rely on sensorimotor simulation (Gallese & Lakoff, 2005).
Therefore, although predominant theories of emotion ascribe an important role to
somatosensory processing in the generation of subjective feeling states, we sought to test
whether individuals would show systematic differences in real-time, open-ended verbal
descriptions of their emotional feelings, and whether such differences would correspond to
differential recruitment of somatosensory processing in the brain during the experience of
complex social emotions. In assessing individuals’ emotional processing styles in their
spontaneous speech, we focused specifically on affective language, i.e. words referencing
by guest on July 15, 2012
http://scan.oxfordjournals.org/
Downloaded from

Page 5

Embodiment of emotion 5
emotional valence, labels, and feeling states. We then examined individual differences in
affective word use relative to the use of cognitive words that referenced mental states and
abstract or cognitive processes. We expected that individuals who used more affectively charged
language to describe their feelings would show more somatosensory activation during emotion
processing, relative to individuals who described their feelings in more cognitive, thought-based
terms. This is the first study, to our knowledge, to combine quantitative analysis of open-ended
speech with fMRI data.
In the current study, quantitative linguistic analysis was applied to transcripts of
participants talking about their emotional responses to a series of narratives, and then language
use was correlated with participants’ subsequent neural responses to the same narratives. The
target emotions were admiration and compassion, two complex social emotions that have been
shown to strongly recruit somatosensory and self-related neural processing (Immordino-Yang et
al., 2009).
Neural Bases of Admiration and Compassion
The current study uses an emotion induction method developed for an earlier study of
social emotions (Immordino-Yang et al., 2009), in which brief narratives describing real people’s
experiences were used to induce varieties of either admiration or compassion in an interview and
then again during fMRI scanning. Narratives were designed to elicit compassion for physical or
for social pain, or admiration for skill or for virtue. A control condition included social narratives
without strong emotional content (beyond being interesting). The four emotion conditions were
found to recruit cortical and subcortical regions associated with interoceptive representation and
homeostatic regulation (e.g., anterior insula, anterior cingulate, hypothalamus, mesencephalon)
relative to control (Immordino-Yang et al., 2009).
by guest on July 15, 2012
http://scan.oxfordjournals.org/
Downloaded from

Page 6

Embodiment of emotion 6
The current study employs a new sample and explores whether individuals’ emotional
language is linked with neural activations during emotion processing in regions of the brain that
sense the body. We examine participants’ use of affective words relative to cognitive words
while verbally describing their emotional responses to the narratives, and correlate individual
differences in word use with differences in BOLD.
Quantitative Analysis of Natural Language Using LIWC
Quantitative analysis of natural language has been used in a variety of studies to explore
individual differences in speech and writing. The current study employs the text analysis
program LIWC (Linguistic Inquiry and Word Count; Pennebaker, Booth, and Francis, 2007),
which contains a dictionary of over 2000 words in 70 language categories and counts the
frequency of words in each category. LIWC has been used to examine many different kinds of
written and oral language samples, from expressive writing to diaries to course assignments to
marital interactions (Pennebaker & Francis, 1996; Pennebaker & King, 1999; Sillars, Shellen,
McIntosh, & Pomegranate, 1997). LIWC analyses have found evidence of individual differences
in “linguistic style” that appear stable within individuals across time and topic. Even spontaneous
word use in everyday conversation exhibits within-person consistency over time (Mehl &
Pennebaker, 2003). LIWC word counts have been correlated with personality, projective
measures, and observed behavior (Pennebaker & King, 1999; Tausczik & Pennebaker, 2010).
This study focused on the “cognitive words” and “affect words” categories, situating each
participant in terms of his or her relative use of the words in each category as a proportion of
total words spoken. The “cognitive words” category includes 730 words such as “think,”
“know,” “assume,” “should,” and “acknowledge” – that is, words reflecting mental states and
abstract thinking. Cognitive word use appears to vary by both personality and situational factors.
by guest on July 15, 2012
http://scan.oxfordjournals.org/
Downloaded from

Page 7

Embodiment of emotion 7
For example, in an analysis of 2004 presidential campaign speeches, Dick Cheney and John
Kerry used more cognitive words than George Bush or John Edwards (Slatcher, Chung,
Pennebaker, & Stone, 2007); in Beatles lyrics, John Lennon used more cognitive words than
Paul McCartney (Petrie, Pennebaker, & Sivertsen, 2008). Cognitive word use may increase
during crises, perhaps reflecting attempts to make sense of an event through strategies such as
distancing or reappraisal. Rudy Giuliani’s speeches included more cognitive words during
difficult times (like during the dissolution of his marriage and after the September 11 attacks;
Pennebaker and Lay, 2002) and cognitive words have been found to increase during and after
relationship break-ups (Boals & Klein, 2005) and after disasters (Gortner & Pennebaker, 2003).
The current study aimed to hold situational factors constant throughout the interview so that
individual differences in speech style could emerge for later correlation with BOLD data.
The other category of interest was “affect” words, including 915 words reflecting both
positive and negative emotions such as “happy,” “inspiring,” “crying,” “abandon,” and “cruel.”
Use of affect words has been associated with personality test measures such as the Five Factor
Inventory (Lee et al., 2007). Fiction writers use more affect words in interviews than physicists
(Djikic, Oatley, & Peterson, 2006). Rates of affect words typically increase over baseline during
expressive writing and emotional disclosure exercises (e.g., D’Souza et al., 2008). Since our
protocol involves emotional disclosure, we expected that it would elicit affect word use among
our participants, but to varying degrees depending on individual differences.
The current study investigated whether participants showed individual differences in the
words they used to describe complex social emotions. We expected that cognitive word use
would reflect a more abstract, deliberative emotional style and that affect words would reflect
more embodied, physical engagement with emotion, and that these styles would correlate with
by guest on July 15, 2012
http://scan.oxfordjournals.org/
Downloaded from

Page 8

Embodiment of emotion 8
patterns of neural activity during narratives designed to elicit feelings of admiration and
compassion, but not control narratives.
Hypotheses
This study has two parts: first, we explored whether the cognitive and affect words
participants used to describe their feelings showed consistent individual differences, and we
examined the relationships between these categories of word use. We hypothesized that the word
use categories would show intra-individual stability across the variety of emotions induced
during the interview and that they would be negatively correlated with each other.
Next, we probed the correspondence of these categories to BOLD activity. We
hypothesized that people who used more affect words relative to cognitive words would show
higher BOLD activity in areas of the brain associated with bodily sensation, including the
somatosensory cortices, insula, and anterior cingulate cortex. We expected that these results
would reflect a genuine difference in processing style rather than an overall increase in
emotionality; as such, we expected word use to be uncorrelated with button-press ratings of
emotion strength collected during fMRI scanning. We also expected that these results would
reflect a difference specific to emotion processing, and as such that the associations with BOLD
would not hold during relatively unemotional control social processing.
Methods
Participants
Twenty-eight right-handed, healthy, native English-speaking participants (16 female, 12
male) were recruited for the study. Participants were all students or staff at a large private
university on the U.S. West Coast. The average age was 21.29 years (Range: 18-27, SD = 2.65).
by guest on July 15, 2012
http://scan.oxfordjournals.org/
Downloaded from

Page 9

Embodiment of emotion 9
Fourteen participants identified as Caucasian-American, 12 as Asian-American, one as Latino-
American, and one as African-American.
Procedures
Before entering the scanner, participants took part in a two-hour, one-on-one interview
with an experimenter, following the method described in Immordino-Yang et al., 2009. The
experimenter presented fifty true narratives depicting real people (not actors or celebrities) using
a scripted verbal description supplemented by video and audio clips shown on a laptop. Fifty
narratives (10 per condition), each taking 60-90 seconds to recount, were presented to elicit: 1)
Admiration for virtue (narratives involved demonstrations of marked self-sacrifice and
dedication to helping others); 2) Admiration for skill (narratives involved demonstrations of
exceptional talents in athletics, the arts, or other domains); 3) Compassion for social pain
(narratives involved situations of bereavement, social rejection, and other forms of psychological
pain); 4) Compassion for physical pain (narratives depicted accidental bodily injuries, e.g., sports
accidents); and 5) Control social processing (narratives involved interesting situations with real
people, such as a man describing his adventures abroad, that were not as emotionally evocative
as narratives in the other categories). Length and use of video vs. audio stimuli did not differ
systematically between conditions. After each narrative presentation, participants were asked
“How does this person’s story make you feel?” as an open-ended prompt to describe their
emotional responses. Participants were not told the emotion categories in the experiment and
were encouraged to be as honest and open as possible. These interviews were videotaped and
transcribed, and transcriptions were independently verified.
Following the interview, participants entered the MRI scanner, where they watched 5-
second ‘reminder’ video clips excerpted from each of the narratives presented in the interview,
by guest on July 15, 2012
http://scan.oxfordjournals.org/
Downloaded from

Page 10

Embodiment of emotion 10
followed by 13 seconds of gray screen and two seconds of fixation. During each scanning trial,
participants rated the strength of their real-time emotional responses to the narrative stimulus
using a button box (first finger = not emotional, second finger = moderately emotional, third
finger = strongly emotional, fourth finger = overwhelmingly emotional). Participants could
respond with their button-press rating at any time during the 18-second trial. The scanning
session comprised four functional runs of approximately 9 minutes each, with each narrative
presented twice (but only once in a given run), for a total of 100 trials. In both the interview and
scanning, narratives were shown in pseudorandom order to counterbalance one-back presentation
history so that no one condition systematically preceded another. Order of runs during scanning
was also counterbalanced.
Word Use Analyses
Verified transcripts of the interviews were edited to remove the experimenter’s speech
and transcription notes, so that only participants’ speech remained. Transcripts were also edited
to remove filler words and phrases (e.g., “like,” “you know,” “I mean”). The edited transcripts
were processed using LIWC, which automatically generates word count rates in multiple
categories, including the “cognitive words” and “affect words” categories. LIWC data reflect the
percentage of total words devoted to each category; for example, a mean of 10 would indicate
that 10% of the words spoken by a participant corresponded to that category.
Functional Imaging Data Acquisition and Analysis
Whole brain images were acquired using a Siemens 3 Tesla MAGNETON TIM Trio
scanner with a 12-channel matrix head coil. Functional scans were acquired using a T2*
weighted Echo Planar (EPI) sequence (TR = 2 sec, TE = 30 ms, flip angle = 90°) with a voxel
resolution of 3mm × 3mm × 4.5mm. Thirty-two continuous transverse slices were continuously
by guest on July 15, 2012
http://scan.oxfordjournals.org/
Downloaded from

Page 11

Embodiment of emotion 11
acquired to cover the whole brain and brain stem, with breaks between runs. Anatomical images
were acquired using a magnetization prepared rapid acquisition gradient (MPRAGE) sequence
(TI = 900 ms, TR = 1950 ms, TE = 2.26 ms, flip angle = 7°) with an isotropic voxel resolution of
1mm.
Data were processed using SPM8 (Wellcome Department of Cognitive Neurology,
London, UK) in MATLAB 2009b (MathWorks, Inc.). Functional images were slice-timing
corrected, motion corrected and co-registered to the anatomical image. Anatomical images were
segmented and normalized to MNI space (Montreal Neurological Institute) using tissue
probabilistic maps (segmentation, SPM8). The same normalization transformation was applied to
the functional images, which were then smoothed using an 8-mm FWHM Gaussian kernel. Pre-
processed BOLD time series were subjected to high-pass filtering (cut-off period 128 seconds)
and session-specific grand mean scaling, and were corrected for serial correlation (AR[1]
model).
Following Immordino-Yang et al., 2009, each condition was modeled using a finite
impulse response function (FIR) with 9 time bins (each bin corresponding to a 2-second TR) to
capture the complex neural activity during the 18-second trial. Only emotion trials in which
participants indicated feeling emotional (2nd-4th finger button press) and control trials in which
participants reported not feeing emotional (1st finger button press) were included. Excluded trials
were modeled as a separate condition of no interest.
Fixed-effect models were run at the individual level. The sums of parameter estimates
corresponding to the 4th-8th TRs (6-16 seconds post-trial onset, the time window previously
shown in Immordino-Yang et al. (2009) to capture the emotion-related BOLD responses in this
task) were used to create contrast maps. To test our main neural hypothesis, we calculated a
by guest on July 15, 2012
http://scan.oxfordjournals.org/
Downloaded from

Page 12

Embodiment of emotion 12
combined contrast of all emotion conditions versus implicit baseline, and entered individuals’
contrast maps into a group-level whole-brain regression analysis using the affect-cognitive
speech variable as a covariate (weighted at ‘1’). We thresholded the resulting contrast map using
q(FDR) < 0.05, which resulted in a t threshold of t = 2.62. Resulting maps were displayed on a 3-
dimensional template brain; we confirmed the anatomical localization of our results with an
expert in neuroanatomy, Hanna Damasio. See Table 1 and Figure 1 for results.
To ensure that our main finding was not driven by a subset of emotion conditions, we
repeated the analysis with separate contrasts for each emotion condition versus implicit baseline.
We applied the critical t threshold from the main contrast (i.e. all emotions vs. baseline, t = 2.62)
and examined our hypothesized regions of interest. See Table 2 for results.
Results
Word Use
About 7% of the words spoken by participants during the interview were categorized as
affect words (M = 6.71, Range: 4.41-10.01, SD = 1.39). Cognitive words comprised about 20%
of participants’ speech (M = 21.28, Range: 17.67-24.49, SD = 1.88). We examined the valence
of affect words used by participants and found that use of negative and of positive affect words
were correlated, r(26) = 0.47, p = .01.
Cognitive and affect word use rates displayed high intra-individual stability across the
five conditions (four emotion conditions and control; Cronbach’s alphas: .80 for cognitive and
.83 for affect words). When cognitive words across the conditions were entered into Principal
Component Analysis, a single factor emerged (Eigenvalue = 2.83); a single factor also emerged
for the five conditions of affect words (Eigenvalue = 3.01). Therefore, as expected, both
by guest on July 15, 2012
http://scan.oxfordjournals.org/
Downloaded from

Page 13

Embodiment of emotion 13
cognitive and affect word use appear to show individual stability across the different emotions
induced during the interview.
Given the high stability of word use across conditions, affect and cognitive words were
totaled across all conditions. As shown in Figure 1, the total counts of affect words were
negatively correlated with totals for cognitive words, r(26) = -0.43, p = .02. Because of this
finding, counts for these two categories of words were combined into a single variable situating
each individual in terms of his or her relative use of affective vs. cognitive speech. This variable,
which had a mean of 0.0 (Range: -2.78-3.25, SD = 1.69), was created by calculating z-scores
across participants for both the cognitive words and the affect words categories and then
subtracting the cognitive words z-score from the affect words z-score. This variable was then
correlated with BOLD (see below).
Neither affect words, cognitive words, nor the combined variable, correlated with
participants’ age, gender, or ethnicity, with the exception that women used more cognitive words
(at a marginal level of significance; r(26) = .35, p = .07).
No significant associations emerged between button-press ratings of experienced emotion
strength in the scanner and cognitive or affect word use, or the combined variable. Correlation
coefficients ranged from .03 to .18 and all p values exceeded .35.
Main effect of emotion versus baseline on BOLD
We examined contrasts of each emotion condition versus baseline and the control
condition versus baseline. In each case we found extensive suprathreshold activations that
covered virtually the entire brain including our hypothesized regions, and no significant
deactivations. Therefore, any positive correlations between word use and BOLD reflect a
by guest on July 15, 2012
http://scan.oxfordjournals.org/
Downloaded from

Page 14

Embodiment of emotion 14
relationship with the positive-going BOLD signal and not with a lessening of BOLD
suppression.
Associations between word use and BOLD
Neuroimaging results revealed significant associations in our hypothesized regions of
interest between the combined word use variable and activation during each of the emotion
conditions versus baseline, as well as with the combined emotion conditions versus baseline.
In each emotion condition, BOLD activity in the primary and secondary somatosensory
cortices (SI and SII), the insula, and the anterior cingulate cortex (ACC) was associated with the
combined word use variable (see Table 1). The locations of peak positive association in each
contrast generally did not overlap. Together these results suggest that our findings are not driven
by neural responses to a particular type of social emotion, e.g. pain-based or pleasurable.
Table 2 presents all suprathreshold clusters that correlated with the speech variable for
the contrast of the combined emotion conditions versus baseline. Figure 2 presents representative
images of this result.
The combined word use variable was not associated with the BOLD responses to the
control narratives, with the exception of one cluster in SII (x=68, y=-24, z=16, Z = 3.14),
suggesting that our effects are specific to emotion processing, not processing of narratives more
generally (e.g., driven by mnemonic demands).
Discussion
This study examined spontaneous affective and cognitive language use during emotion
induction interviews and associations with activation in somatosensensory (including
interoceptive) brain regions. Rates of cognitive words and affect words showed stability within
individuals across the different varieties of emotions induced, suggesting that these word rates
by guest on July 15, 2012
http://scan.oxfordjournals.org/
Downloaded from

Page 15

Embodiment of emotion 15
reflect reliable individual differences in emotion processing. Across participants, use of cognitive
words was negatively associated with use of affect words, and people who used more affect
words, relative to cognitive words, showed greater activation in somatosensory areas during
emotion. Importantly, cognitive and affect word use rates were not associated with button-press
ratings of emotion strength in the scanner and did not predict brain activation during control
social processing (in which participants reported feeling unemotional). Together these findings
suggest that these word use patterns reflect distinct styles of emotion processing, and not simply
the intensity of individuals’ emotions in relation to our stimuli. In addition, the lack of
relationship between BOLD and speech style during the control condition suggests that our
effect is specific to emotion and not attributable to memory or other cognitive processing that
would be equivalent between the emotion and control conditions of our experiment.
The areas of activation that correlated with affective, relative to cognitive, language use
included the SI, SII, insula, and middle ACC – all somatosensory areas that have been associated
with the processing of emotions. The insula is involved in interoceptive body awareness and thus
in feeling emotions (e.g., Craig, 2009; Damasio, 1999; 2005; Damasio et al. 2000; Harrison et
al., 2011), including social emotions such as empathy for others’ pain (Immordino-Yang, et al.,
2009; Singer, 2006). The middle ACC is involved in the perception of bodily pain (Talbot, et al.,
1991; Baliki et al., 2006) and related to individual differences in emotional awareness (e.g., Lane
et al., 1998; McRae, et al., 2008). Somatosensory cortices have been implicated in emotional
judgments of other people (Adolphs, Damasio, Tranel, Cooper, & Damasio, 2000), and insula
lesions have been related to difficulty discerning disgust from others’ facial expressions (Calder,
Keane, Manes, Antoun, & Young, 2000). Our findings suggest that these somatosensory areas
are recruited particularly strongly during emotion processing by some people, possibly to create
by guest on July 15, 2012
http://scan.oxfordjournals.org/
Downloaded from