Mapping the sequence of brain events in response to disgusting food

Abstract

Warning signals indicating that a food is potentially dangerous may evoke a response that is not limited to the feeling of disgust. We investigated the sequence of brain events in response to visual representations of disgusting food using a dynamic image analysis. Functional MRI was acquired in 30 healthy subjects while they were watching a movie showing disgusting food scenes interspersed with the scenes of appetizing food. Imaging analysis included the identification of the global brain response and the generation of frame‐by‐frame activation maps at the temporal resolution of 2 s. Robust activations were identified in brain structures conventionally associated with the experience of disgust, but our analysis also captured a variety of other brain elements showing distinct temporal evolutions. The earliest events included transient changes in the orbitofrontal cortex and visual areas, followed by a more durable engagement of the periaqueductal gray, a pivotal element in the mediation of responses to threat. A subsequent core phase was characterized by the activation of subcortical and cortical structures directly concerned not only with the emotional dimension of disgust (e.g., amygdala‐hippocampus, insula), but also with the regulation of food intake (e.g., hypothalamus). In a later phase, neural excitement extended to broad cortical areas, the thalamus and cerebellum, and finally to the default mode network that signaled the progressive termination of the evoked response. The response to disgusting food representations is not limited to the emotional domain of disgust, and may sequentially involve a variety of broadly distributed brain networks. Hum Brain Mapp 39:369–380, 2018. © 2017 Wiley Periodicals, Inc.

INTRODUCTION

Animals progressively approach their environment seeking interactions to satisfy their needs, but may quickly withdraw from their surroundings in response to warning signals. In the context of eating behavior, olfactory, gustatory, and visual warning signals may indicate that a food is potentially dangerous by evoking the feeling of disgust as part of a wider experience that also implicates, for example, withdrawal reactions and appetite dampening, threat context learning, arousal heightening, stimulus awareness and appraisal of its significance [Chapman et al., 2013; Lindquist and Barrett, 2012; Peper et al., 2006]. Perhaps for evolutionary reasons, the role of visual (as opposed to gustatory and olfactory) signaling is relatively more relevant in humans than in other mammals.

Disgust has been often contemplated as a basic emotion, particularly in neuroimaging research, and has been experimentally induced using certain (usually visual) stimuli. A large number of studies have used facial expressions of disgust on the basis of their ability to evoke a vicarious experience of disgust. Other studies have used non‐food‐ (e.g., body mutilations) or food‐related pictures to elicit disgust [Lindquist et al., 2012; Wager et al., 2015]. More rarely, movies [Harrison et al., 2010; Karama et al., 2011; Saarimäki et al., 2016; Tettamanti et al., 2012] or direct olfactory stimulation [Meier et al., 2015; Wicker et al., 2003] have been used in such experiments. The brain structures reported as being most consistently associated with disgust include the insula, the amygdala, the frontal cortex, and visual areas [see reviews in Kirby and Robinson, 2016; Lindquist et al., 2012; Murphy et al., 2003; Vytal and Hamann 2010; Wager et al., 2015]. Nevertheless, in the broader perspective of brain and behavioral relationships, there are arguments to suggest that the brain response to food‐related disgust may not be limited to the aforementioned structures.

For example, the periaqueductal gray (PAG) is a key element mediating fight or flight responses to a perceived threat [Mobbs et al., 2007] and in the processing of dangerous painful stimulation [Benarroch, 2012]. As a potent signal of threat, potentially dangerous food representations could arguably activate the PAG. Imaging studies have only exceptionally reported PAG activation in response to disgusting stimuli [Harrison et al., 2010; Karama et al., 2011; Radua et al., 2014]. Simialrly, hypothalamic activation has been rarely discussed in neuroimaging studies of disgust [Karama et al., 2011], despite its central role in the regulation of food intake. The involvement of the orbitofrontal cortex and striatum has also been infrequently emphasized in studies assessing normal populations [Lindquist et al., 2012; Wager et al., 2015], despite their central role in emotional learning processes underlying approach and withdrawal behavior [Peper et al., 2006]. Nevertheless, strong evidence of their relevance in the processing of disgusting stimulation is found in studies assessing altered disgust sensitivity in clinical disorders (see Vicario et al. [2017b] for a recent review).

We also propose that disgusting food stimulation may activate other elements of the response to threat in the domains of memory encoding, arousal, stimulus appraisal, and attention. Specifically, insofar as attention is concerned, the approach‐withdrawal behavior model predicts a predominant right hemisphere activation in response to alerting signals [Davidson et al., 1990], arguably via the ventral attentional network more recently characterized by functional MRI [Corbetta et al., 2008]. This is in contrast with most neuroimaging research results showing symmetrical frontal lobe responses to disgusting stimuli [Murphy et al., 2003; Wager et al., 2015].

We have assessed brain response to visual representations of disgusting food using a dynamic image analysis to capture the temporal evolution of evoked brain activations. Functional MRI was acquired while healthy participants were watching a movie showing disgusting food scenes interspersed with baseline scenes of appetizing food. The results confirmed strong engagement of the brain elements directly related to disgust. Nonetheless, brain response was not limited to the core disgust system, but involved a variety of functional networks sequentially responding to the potential threat.

METHODS

Participants

Thirty right‐handed healthy participants, including 15 males and 15 females ranging from 19 to 45 years (mean: 27.9 years; SD: 7.8 years) made up the study sample. This sample size was chosen on the basis of previous studies using emotion‐evoking stimulation conducted in our laboratory [Caseras et al., 2013; Pujol et al., 2012, 2013] and in others [Tettamanti et al., 2012; Vytal and Hamann 2010]. An additional sample of 15 right‐handed healthy participants, matched by age and sex to the primary sample, was used as an independent group to explore the activation dynamics of core disgusting regions (see below). This additional sample included 7 males and 8 females ranging from 17 to 37 years (mean: 26.6 years; SD: 6.2 years). A complete medical interview was carried out to exclude subjects with relevant medical or neurological disorders, psychiatric illness, and history of substance abuse. No subject was undergoing medical treatment, suffering from eating disorders or following a vegetarian (or vegan) diet.

This study was conducted in accordance with the principles expressed in the Declaration of Helsinki. The protocol was approved by the Clinical Research Ethical Committee of the Corporació Sanitària Parc Taulí, Sabadell (Barcelona). Written informed consent was obtained from all participants.

Stimulus

A color movie was specifically produced for the study featuring successive dynamic scenes with appetizing food as the baseline (e.g., food items in restaurants or gastronomic scenes) and disgusting food or repulsive eating scenes interspersed at pseudorandom intervals. The duration of stimuli presentation was 6 m in total and included 20 scenes of appetizing food and 8 disgusting scenes. Both stimuli types (baseline and test) were presented concatenated, without null events (i.e., a video scene, positive or negative, was always present in the screen). Each of the disgusting scenes lasted 8–10 s. The experimental paradigm is shown in Figure 1. Disgust‐evoking scenes showed typical repulsive images such as spoiled food (e.g., rotting fruits), unusual food (e.g., a man eating worms), and insects in food (e.g., maggots, cockroaches). The selection of scenes was carried out, on the basis of a consensus reached by researchers, to include two different levels of intensity (moderate and high). A large number (n = 50) of disgust‐evoking scenes was preliminarily selected. Four senior researchers rated the intensity of the elicited feeling of disgust for each scene using a verbal ranging scale from 0 to 100. Four scenes of moderate intensity were selected, rated between 40 and 70 by the four raters, and four scenes of high intensity with ratings >70.

Figure 1.

Experimental paradigm. Blue represents the mean MRI signal time course from the frontal cortex, insula, amygdala, and fusiform gyrus obtained in a 15‐participant independent sample undergoing identical assessment. Red represents the baseline/activation regressor, both nonadjusted (top) and adjusted (bottom), using the actual dynamic information. The numbers correspond to the scans. [Color figure can be viewed at http://wileyonlinelibrary.com]

During scanning, participants passively viewed the 6 m movie with MRI‐compatible high‐resolution goggles (VisuaStim Digital System, Resonance Technology Inc., Northridge, CA). They were instructed to simply attend and watch the movies. No active task was required during fMRI scanning. Movies were presented without sound to avoid attentional confounds. Immediately after the MRI session, participants were requested to rate the intensity (0–100) of the feeling of disgust elicited by each scene during the imaging session. The subjective rating of all participants confirmed that 4 out of 8 disgusting scenes evoked moderate subjective disgust (group mean ± SD scores; 67.0 ± 30.1, 58.9 ± 30.1, 59.1 ± 31.2, 57.8 ± 33.4) and 4 out of 8 evoked high subjective disgust (group mean ± SD scores; 72.3 ± 32.6, 75.7 ± 31.7, 77.8 ± 30.3, 83.1 ± 27.0). There was no significant difference for any individual disgusting scene rating between the main study sample and the additional study sample.

MRI Acquisition

Participants underwent functional MRI 1 h after receiving a standardized lunch to maximally control for the experimental context and to enhance the repulsive effects in the satiate state [Meier et al., 2015].

A 1.5 T Signa Excite system (General Electric, Milwaukee, WI) equipped with an eight‐channel phased‐array head coil and single‐shot echoplanar imaging (EPI) software was used. The functional sequence consisted of gradient recalled acquisition in the steady state emphasizing blood‐oxygen‐level dependent (BOLD) contrast (time of repetition [TR], 2,000 ms; time of echo [TE], 50 ms; pulse angle, 90°) within a field of view of 24 cm, with a 64 × 64 pixel matrix, and a slice thickness of 4 mm (interslice gap, 1.5 mm). Twenty‐two interleaved slices were prescribed parallel to the anterior–posterior commissure line covering the whole‐brain. The first four (additional) images in each run were discarded to allow magnetization to reach equilibrium.

Image Preprocessing

Imaging data were processed using MATLAB version 2011b (The MathWorks Inc, Natick, MA) and Statistical Parametric Mapping software (SPM8; The Wellcome Department of Imaging Neuroscience, London). Preprocessing involved motion correction, spatial normalization and smoothing using a Gaussian filter (full‐width half‐maximum, 8 mm). Data were normalized to the standard SPM‐EPI template and resliced to 2 mm isotropic resolution in Montreal Neurological Institute (MNI) space. All image sequences were inspected for potential acquisition and normalization artifacts.

Control of potential head motion effects

To control for the effects of head motion, we adopted the following approach: (i) time series were aligned to the first image volume in each participant using a least squares minimization and a 6‐parameter (rigid body) spatial transformation. (ii) We included 6 motion‐related regressors and estimates of global brain signal fluctuations as confounding variables in our first‐level (single‐subject) analyses. (iii) Within‐subject, censoring‐based MRI signal artifact removal (scrubbing) [Power et al., 2014] was used to discard motion‐affected volumes. For each participant, interframe motion measurements [Pujol et al., 2014] served as an index of data quality to flag volumes of suspect quality across the run. At points with interframe motion >0.2 mm, that corresponding volume, the immediately preceding and the succeeding two volumes were discarded. Using this procedure, a mean of 2.7 (1.5%) volumes (range 0–26) from the total 180 volumes that are included in the fMRI sequence were removed. (iv) The remaining potential motion effects were controlled by including a motion summary measurement for each participant as a covariate in the group analyses in SPM [Pujol et al., 2014].

Functional MRI Data Analysis

Global brain response

To obtain a basic map of the global brain response to disgusting food representations, a boxcar regressor was generated considering a baseline condition and an activation condition including the eight disgusting scenes (Fig. 1). As both the time required by the stimuli to evoke feelings of disgust and the duration of the evoked response were a priori unknown in our experiment, this basic regressor model was adjusted using dynamic information from the 15‐participant independent sample experiment. In each of these participants, the functional MRI signal time courses from the 4 core regions showing the most consistent activation in previous research (frontal cortex MNI‐transformed x = 44, y = 11, z = 25, insula x = −36, y = 23, z = 1, amygdala x = 26, y = −3, z = −19 and fusiform gyrus x = −42, y = −61, z = −15 [Kirby and Robinson, 2016]) were averaged to obtain a single activation time course representing activation in core disgust regions. For each location, the region was defined as a 3.5 mm radial sphere. Region mean signal was extracted at each time point using MarsBaR region of interest toolbox in MNI stereotaxic space [Brett et al., 2002]. Our baseline‐activation regressor was then modified using this dynamic information to adjust both activation onset and response duration (Fig. 1). On average across the 8 scenes, disgust stimulation activated these regions with a 5.6 s delay and the activation persisted for 12.1 s after stimulus cessation (2.6 s and 9.1 s, respectively, if a conventional hemodynamic delay of 3 s is subtracted). The total mean duration of core region activation was 15 s.

The dynamically adjusted regressor, therefore, served to carry out the individual analyses in the main study sample. The resulting first‐level SPM contrast images were carried forward to group‐level random‐effects analyses. One‐sample t‐test designs were used to generate both a global map (baseline vs disgust) and a map of the contrast between the 4 scenes generating moderate disgust and the 4 scenes generating high disgust (implicit modeling of baseline). In addition, activation maps were compared with their mirror (flipped) version to test for potential asymmetries.

Frame‐by‐frame activation maps

This analysis aimed to characterize the temporal evolution of brain activations evoked by disgust stimulation using data averaged from the 8 disgusting scenes featured in the functional MRI run. The procedures are described in a previous study [Lopez‐Sola et al., 2010]. We employed the finite impulse response (FIR) analysis approach [Dale and Buckner, 1997] to obtain 12 activation maps covering the activation cycle with a temporal resolution of 2 s (1 scan). Specifically, functional MRI time‐series were modeled using 12 boxcar regressors corresponding to 12 consecutive scans covering an activation window of 24 s starting from the first scan after stimulus onset (time 0) in each of the 8 disgusting scenes (i.e., each boxcar regressor contained one scan from each disgusting scene). The model included an implicit baseline of 176 s (88 scans) in total. For each subject, SPM contrast images were calculated for the 12 regressors expressing the relative BOLD signal change from baseline throughout the activation cycle. The contrast images were then entered in 12 group random‐effects analyses (one‐sample t tests) to generate whole‐brain activation t‐statistic maps for each scan (TR).

The 12 activation maps from the group analyses were used to create movie sequences that dynamically illustrated the temporal evolution of brain responses within representative brain views. The transition between consecutive images was displayed using a previously described method [Pujol et al., 2006] involving the composition of a 15‐step morphing sequence (using Fantamorph software, v. 5.0, Abrosoft, Devon, UK). A separate FIR analysis was performed to obtain one additional frame at the end of the activation cycle (frame 13) to better illustrate the activation termination specifically in the figures.

To graphically represent group activation time courses for the main brain areas, we plotted activation measurements (t values obtained from the region coordinate showing peak activation across the cycle) against 12 time points (scans).

Thresholding criteria

A very high threshold (p < 0.05 whole‐brain voxelwise family‐wise error (FWE) corrected, t > 5.7) was used to select the principal elements of the global response to disgusting food representations. In the frame by frame mapping and the analysis comparing scenes of high vs moderate disgust intensity, clusters >2.1 ml (264 voxels) at a height threshold of P < 0.005 were considered, which satisfied the FWE rate correction of P _FWE < 0.05, according to Monte‐Carlo simulations.

RESULTS

Global Brain Response to Disgusting Food Representations

A global brain activation map was generated based on the functional MRI signal contrast between baseline and disgust condition. As both the time required by the stimuli to evoke feelings of disgust and the duration of the evoked response were a priori unknown, the analysis (task regressor) was adjusted to the actual activation dynamics of core regions characterized using an independent experiment (see Methods and Fig. 1).

We found a very robust activation pattern. Regions with highly significant activations included the insula/frontal operculum, amygdala, lateral frontal cortex, and fusiform gyrus (core regions), although similarly robust activations were found in the hypothalamus, PAG, hippocampus, basal ganglia, superior temporal/supramarginal gyri, visual areas, and the cerebellum (Fig. 2). The activations were bilateral with the exception of the lateral frontal and temporo‐parietal cortices, showing significant right hemisphere predominance (Table 1). Areas of deactivations were also identified involving sensorimotor cortices and the supplementary motor area (Table 1 and Fig. 3).

Figure 2.

Global brain response to disgusting food representations. Brain regions showing the most robust response to disgusting food scenes compared with scenes of appetizing food. The analysis (regressor model) was adjusted to the actual activation dynamics of core regions characterized using an independent experiment. The right hemisphere corresponds to the right side of coronal views. [Color figure can be viewed at http://wileyonlinelibrary.com]

Table 1.

Global brain response to disgusting food

Brain activations a	x y z	t
R lateral frontal cortex	46 14 24	10.3
L lateral frontal cortex	−44 10 20	7.4
R frontal operculum/anterior insula	48 32 0	9.9
L frontal operculum/anterior insula	−56 36 −2	8.9
R superior temporal/supramarginal gyri	50 −50 10	13.2
R Amygdala	26 −2 −26	13.5
L Amygdala	−20 −2 −26	10.6
R Hippocampus	28 −16 −10	10.1
L Hippocampus	−28 −18 −12	11.2
Hypothalamus	−4 −4 0	9.4
Periaqueductal gray (PAG)	−4 −28 −8	9.8
R lenticular nucleus	14 4 8	9.7
L lenticular nucleus	−12 2 8	9.1
R extra‐striate visual cortex/fusiform gyrus	44 −86 −2	14.1
L extra‐striate visual cortex/fusiform gyrus	−42 −80 6	14.0
Primary visual cortex	0 −90 12	11.2
Cerebellum	−10 −78 −18	8.7
Right > left activation asymmetry b
Lateral frontal cortex (1447 voxels)	−56 10 46	6.2
Superior temporal/supramarginal gyri (2846 voxels)	−52 −42 14	6.3
Brain deactivations**
Sensorimotor cortex	−24 −44 64	7.1
Supplementary motor area	2 −18 60	7.0

x y z, coordinates (mm) given in Montreal Neurological Institute (MNI) space. Statistics at threshold P_FWE < 0.05 whole‐brain voxelwise corrected.

^aNumber of activation (8 mm³) voxels above threshold: 107,160 voxels.

^bNumber of deactivation voxels above threshold: 14,724 voxels.

Figure 3.

Brain areas showing significant deactivation (baseline > disgusting condition) in the global analysis. The right hemisphere corresponds to the right side of coronal and axial views. [Color figure can be viewed at http://wileyonlinelibrary.com]

A primary analysis of the correlation between the subjective scores of generated disgust and brain activation showed no significant findings. Nevertheless, in the analysis comparing brain activation generated by the scenes evoking high disgust with those evoking moderate disgust, the intensity of the subjective disgust was associated with higher activation in the PAG, right amygdala, left insula, occipital, and posterior temporal areas, in addition to the anterior cingulate cortex (ACC) and adjacent medial frontal cortex. On the other hand, the intensity of subjective disgust was associated with lower activation in the orbitofrontal cortex and parietal cortex (Table 2 and Fig. 4).

Table 2.

Disgusting feeling intensity effect on brain activation

Scenes of high > moderate disgust	Number of voxels (ml)	x y z	t
Periaqueductal gray (PAG)	131* (1.0)	−2 −28 −8	3.2
R amygdala	185* (1.5)	24 −2 −24	3.7
L insula	1184 (9.5)	−38 22 10	4.4
L posterior temporal cortex	3165 (25.3)	−50 −76 12	9.1
R posterior temporal cortex	4297 (34.4)	46 −48 14	6.4
Visual cortex	4617 (36.9)	−6 −76 2	8.8
Anterior cingulate cortex (ACC)	794 (6.3)	4 16 28	4.2
Medial frontal cortex	1072 (8.6)	+8 48 26	4.7
Scenes of high < moderate disgust
Orbitofrontal cortex	3915 (31.3)	20 46 −14	5.1
Parietal cortex	8271 (66.2)	60 −32 54	5.3

x y z, coordinates (mm) given in Montreal Neurological Institute (MNI) space. Statistics at corrected threshold P_FWE < 0.05 estimated using Monte‐Carlo simulations, with the exception of small brain areas* (amygdala and PAG) showing subthreshold results at the whole‐brain correction level.

Figure 4.

Brain activation during high versus moderate disgust. Regions showing higher (top) and lower (bottom) activation during scenes evoking high disgust are illustrated. The right hemisphere corresponds to the right side of coronal and axial views. [Color figure can be viewed at http://wileyonlinelibrary.com]

Temporal Evolution of the Brain Response to Disgusting Food Representations

A frame‐by‐frame analysis of brain activation identified (i) a number of early and, in part, transitory activations, (ii) the subsequent progressive recruitment of subcortical and cortical core disgust structures, and (iii) a late activation of large‐scale brain networks.

The earliest changes were identified in the second frame of the functional MRI acquisition (in the interval 2–4 s after stimulus onset). At this point, activation in the primary visual cortex began (Fig. 5), coinciding with transient activation in a small part of the left ventral basal ganglia extending to the amygdala (Supporting Information, Video 1). Interestingly, the fusiform gyrus was deactivated during this interval, but not at later stages, when it was robustly activated.

Figure 5.

Temporal evolution of the response to disgusting food representations. Each image column summarizes the events of representative moments from the earliest to the latest phases. The right hemisphere corresponds to the right side of coronal and axial views. *Corresponds to the frame 4–6 s. [Color figure can be viewed at http://wileyonlinelibrary.com]

The next frame (4–6 s) was characterized by the extensive, albeit brief, activation of the medial orbitofrontal cortex lasting one frame (Fig. 5) and by activation onset in the right posterior temporal/supramarginal gyrus cortex.

The core phase of the brain response to disgusting stimuli was characterized by the recruitment of the PAG, insula, fusiform gyri (changing from deactivation to activation), and right frontal cortex, all beginning in the 6–8 s frame and, subsequently, the hypothalamus, hippocampus, and amygdala starting in the 8–10 s frame (Figs. 5 and 6, and Supporting Information, Video 1). Activation in these regions (and in visual areas) persisted beyond stimulus cessation and vanished at different moments, with the PAG and hypothalamus showing the shortest duration (significant until the 14–16 s frame and 16–18 s frame, respectively) and the insula and amygdala showing the largest activations (both significant until the 20–22 s frame) (Fig. 6).

Figure 6.

Graphic illustration showing the temporal evolution of the response to disgusting food representations. Statistic t values of the activations (axis y) are plotted against time (axis x) expressed as 2 s frames starting from stimulus onset (frame 1). The bold line in axis x indicates the stimulus duration (8–10 s) of the disgusting scenes. Data from the 8 scenes were averaged to illustrate the activation cycle. [Color figure can be viewed at http://wileyonlinelibrary.com]

The late phase of the brain response was characterized by the activation of a set of additional brain structures starting in the frames where the core structures showed maximal activation (10–12 s to 14–16 s) and persisting broadly until the end of our temporal window (Fig. 6). In this late stage, significant changes appeared firstly in regions relevant to brain arousal (including the thalamus, basal ganglia, and cerebellum) and in an extensive part of the cerebral cortex (Figs. 5 and 6, and Supporting Information, Video 1). Finally, regions pertaining to the so‐called default mode network were the latest areas incorporated in the activation pattern (i.e., angular gyrus, posterior cingulate cortex (PCC)/precuneus, and medial frontal cortex).

DISCUSSION

Although the disgusting food representations evoked robust activation in brain structures conventionally associated with disgust, our dynamic approach also captured a variety of other brain elements showing distinct temporal evolutions. The early phase was characterized by initial changes in visual areas and transient orbitofrontal cortex activation, followed by a more durable engagement of the PAG and a right‐hemisphere attentional network. In the core phase, activations were identified in subcortical and cortical structures more directly concerned with disgust. In the later phase, neural excitement extended to broad cortical areas, the thalamus and cerebellum, and finally to the default mode network that signaled the progressive termination of the evoked brain response.

Significant activations were observed as early as 2–4 s after stimulus onset. If a conventional time to activation onset of 3 s is considered, then the data would indicate that the neural events commenced immediately after stimulation began and presumably prior to full disgust awareness. The amygdala may automatically react to alerting signals before their conscious perception [Phelps and LeDoux, 2005]. We found only marginal evidence of amygdala involvement at the very early stage, perhaps as a result of the relatively low temporal resolution of functional MRI. Nevertheless, as part of the fast response to salient stimuli, the amygdala in turn may quickly activate both primary sensory areas enhancing perception [Phelps and LeDoux, 2005] and the orbitofrontal cortex that codes the aversive (or incentive) value of the stimulus [Dagher, 2012]. We found very early changes in both primary visual areas and the orbitofrontal cortex presumably contributing to automatic processes during potential danger detection.

The lateral–ventral frontal cortex and superior temporal/supramarginal gyri were activated thereafter with exquisite right‐hemisphere predominance initially. These areas are main components of the so called ventral attentional network, which has a relevant role in the conscious detection of salient stimuli [Corbetta et al., 2008]. The temporal evolution of the activation observed in this network gives further support to the proposed right‐hemisphere lateralization of withdrawal responses to potential danger [Davidson et al., 1990]. However, similar areas of the left hemisphere were activated in later phases as part of the broad expansion of brain activation. We suspect that such a generalization of cortical activation may, to some extent, account for the lack of frontal activation asymmetry in functional MRI studies adopting conventional analyses [Wager et al., 2015].

A number of structures of the limbic system were robustly activated in the core phase including the PAG, which is a pivotal element in the mediation of responses to potential danger, the hypothalamus, the main center for regulating food intake, the amygdala as an ingredient common to all basic emotions and the anterior insula with the adjacent frontal operculum and lateral orbitofrontal cortex. The insula with surrounding frontal cortex are the best candidates to account for disgust awareness as a bodily sensation typically implying unpleasant feelings in the oral cavity and abdomen [Nummenmaa et al., 2014]. The primary gustatory cortex is precisely located in the anterior insula and frontal operculum [Scott and Plata‐Salamán, 1999], for both palatable and aversive (disgusting) gustatory stimulation [Vincis and Fontanini, 2016]. Hunger is another oral and abdominal sensation with a cortical representation in the anterior insula and surrounding frontal cortex [Dagher, 2012]. It is also relevant to mention that disgust (like nausea) may be physiologically elicited as a sensory phenomenon by physical or chemical stimulation of the throat and stomach [Chapman and Anderson, 2012].

In the late phase, activation extended bilaterally to large cortical areas in the frontal cortex and inferior parietal and posterior temporal cortices, and to the thalamus and cerebellum. All these changes overlapped with activation in core structures, with late‐stage activation involving a vast brain extension. The combination of widespread cortical activation with activation in the densely connected thalamus and cerebellum may be interpreted as reflecting brain arousal or general excitement resulting from emotionally powerful stimulation [Calabrò et al., 2015].

The latest elements incorporated in the activation pattern pertained to the default mode network. Importantly, the activation concurred with still present activation in core structures, such as the amygdala and insula, and they cannot therefore be easily interpreted as the expression of higher activation at rest (i.e., as a deactivation phenomenon). Indeed, the default mode network did not appear in any frame‐by‐frame deactivation contrast in our analysis. All in all, the activation of the default mode network coincided with the progressive termination of the brain response to disgusting food representations. It has been proposed that the default mode network has a direct role in regulating attention [Leech and Sharp, 2014] which could favor disengagement from the focus of attention in our experiment.

Our deactivation pattern was dominated by changes notably restricted to the sensorimotor cortex and supplementary motor area (Table 1 and Fig. 3). Somatic (and visceral) feelings are part of emotions and each emotion shows a particular bodily map [Nummenmaa et al., 2014]. Disgust is mostly characterized by sensations in the digestive system and around the throat region, which is consistent with activations observed in the anterior insula and frontal operculum. However, there is an emotion‐related inhibitory control from the PAG, amygdala, and insula on the motor cortex and supplementary motor area in situations of threat [Benarroch, 2012; Hassa et al., 2017; Vicario et al., 2017a], which arguably may contribute to account for deactivations in our study. Of relevance, a recent transcranial magnetic stimulation study has demonstrated motor cortex inhibition at the tongue representation specifically in response to revulsion‐eliciting food pictures and disgusted facial expressions [Vicario et al., 2017a].

One limitation in our study is the lack of information on the disgust propensity and sensitivity of our participants, as the intensity of brain response to disgust‐evoking stimuli may be directly related to such a propensity. Another limitation may be the scarce information regarding the subjective feelings that were actually generated. Ideally, a continuous rating of temporal evolution of evoked disgust would have allowed us to establish a closer correspondence between feelings and their cerebral representation. Moreover, the feelings evoked by the scenes of appetizing food used as baseline in the movie sequence were not specifically evaluated. It is known that disgust stimuli affect the subjective evaluation of positive stimuli if they are in close neighborhood [Borg et al., 2016; Legget et al., 2015]. Although we obtained robust brain activations in the contrast between disgusting and appetizing scenes, the interaction between both stimuli types was not controlled in our study.

A concern in our study, and generally in studies using subjective ratings, is the low consistency of subjective scores across individuals. We found no significant associations between subjective ratings of disgust and the global activation pattern, which may reflect such a limitation. However, the use of group average data in the strategy of comparing activation between high and moderate disgusting scenes may, to some extent, circumvent this limitation as was the case in pain studies [Wager et al., 2013]. This analysis confirmed a correspondence between the intensity of subjective disgust and activation in relevant core elements such as the insula, amygdala and PAG, in addition to the ACC (a key site for the conscious appraisal of threat [Harrison et al., 2015]) and visual areas. Importantly, higher disgust intensity was associated with lower activation in the orbitofrontal cortex in this analysis, which may express a dual effect of disgusting stimulation on this brain area, with initial enhancement of (aversion‐related) activity and later inhibition of (appetite‐related) activity during the most intense scenes.

Our results may serve to further characterize the pattern of brain activation associated with disgust, although the data perhaps more relevantly suggest that the response to potent disgusting stimuli may not be limited to an evoked emotion, but implies a wider experience. Our functional MRI study, therefore, may contribute to the literature on disgust processing by summarizing all the major elements of such an experience. The use of a dynamic image analysis was critical to capture their temporal evolution.

Described in short, a rapid response in brain systems ready to detect potential danger signals may influence primary areas to enhance stimulus perception (early activation in visual areas) and code the aversive value of the stimulus (orbitofrontal cortex). Disgusting food representations are capable of generating alert (and engage the specialized ventral attentional network) and prepare to react to threat (PAG activation). Core disgust is an emotionally colored (amygdala) sensory phenomenon (insula and surrounding frontal cortex) that influences structures regulating food intake (hypothalamus). Emotionally powerful stimulation generates brain excitement (broad cortical and thalamic activation) until the response terminates coinciding with activation in brain areas with a role in regulating the focus of attention (default mode network). Our contribution may be one example of how a complex mental state is constructed from the interaction of broadly distributed brain networks supporting basic mental operations [Lindquist and Barrett, 2012].

POTENTIAL CONFLICTS OF INTEREST

We have no competing interests.

Supporting information

Supporting Information

Supporting Information Movie