1. Introduction
Moral emotions are crucial in regulating social interactions, as they promote the welfare of the society or of other people (Haidt et al. 2003). Indeed, they provide the emotional drive to properly behave in social interactions (
Kroll and Egan 2004;
Piretti et al. 2020;
Grecucci et al. 2021), forcing individuals to implement strategies that are optimal over a long period, even though they do not appear functional in the short period (
Ridley 1996;
Sober and Wilson 1998).
It has been proposed that moral cognition depends on prefrontal, temporal and the limbic circuits, associated with the integration of context- independent and -dependent information with the associated emotional reactions (event-feature-emotion complexes model, EFEC) (Moll et al. 2005;
2008). Specifically, the prefrontal cortex seems to be responsible for representing context-dependent knowledge of event sequences (Grafman 1995; Wood and Grafman 2003), the temporal lobes for perceiving social cues and for representing context-independent social semantic knowledge (Lambon Ralph et al. 2017; Olson et al. 2013; Haxby et al. 2000), and the limbic system for the generation of emotional and motivational states (Saper et al. 2000). Hence, according to the EFEC model, the generation of moral emotions, including self-conscious emotions, relies on integrity of a network including prefrontal, temporal and limbic areas (Moll et al. 2005;
2008).
Several studies investigating the neural substrates of moral cognition (Eres et al. 2017; Bzdok et al. 2012), confirmed the anatomical predictions of this model and better defined the topography of the brain areas associated with moral processing. Indeed, they showed that ventromedial and dorsomedial prefrontal cortices (vmPFC and dmPFC), temporo-parietal junction (TPJ), precuneus, posterior cingulate cortex, left amygdala, anterior temporal lobes (ATL) and lateral orbitofrontal cortex were consistently found activated in neuroimaging studies investigating moral processing (Bzdok et al. 2012; Eres et al. 2017;
Grecucci et al. 2021;
Piretti et al. 2020).
Among moral emotions, a sub-group of emotions (e.g., shame, embarrassment, guilt and pride), defined as self-conscious emotions, helps individuals to navigate in the complexities of fitting into groups (
Haidt 2003), satisfying the human need of belonging to social groups (
Baumeister and Leary 1995). Self-conscious emotions are evoked by self-reflection and self-evaluation (
Tangney et al. 2007) and occur when social norms, or agreed-upon social rules, are violated (
Bastin et al. 2016). The provide an immediate feedback that promote inhibition or reinforcement of behaviour based on their positive or valence (
Tangney et al. 2007;
Grecucci et al. 2021;
Piretti et al. 2020). One case in point, is shame that has been proposed as an algorithm the brain uses to inhibit socially and morally unwanted behaviors (
Piretti et al. 2020;
Grecucci et al. 2021).
While the EFEC model might explain the cognitive processes underlying all self-conscious emotions, which are all induced by moral and social norm violation (
Bastin et al. 2016), it does not make any prediction about the different processes that might occur in different types of emotions such as the negative self-conscious emotions.. Indeed, even though shame, embarrassment and guilt are often (but culpably) used interchangeably, they appear to be substantially different (
Gibson 2015). Shame is typically elicited by the belief that the individual’s violation of standards of morality, aesthetics or competence, defines who the individual is (
Wong and Tsai 2007). Hence, it involves the way the individuals perceives themselves and how they believe other people see them and their inadequacy to fulfil social standards (
Tangney et al. 1992). The distinction between shame and embarrassment is still a matter of debate (
for review see Crozier 2014). If on the one hand, embarrassment might be considered as a dimension of shame (
Probyn 2005), on the other, it might represent a distinct emotional entity (
Keltner and Buswell 1997;
Tangney et al. 1996). Embarrassment seems related to trivial social transgressions, occurring suddenly and in public contexts, especially in presence of individuals with equal or higher hierarchical social status. (
Keltner and Buswell 1996;
Tangney et al. 1996;
Buss 2001;
Haidt 2003;
Tangney 2003). Conversely, shame emerges when one perceives personally the serious violation of a moral norm, that might be also experienced in private situations (
Tangney et al. 1996;
Tangney 2003). Furthermore, shame and embarrassment are also distinct in the intensity (i.e., shame is more intense than embarrassment) (
Rochat 2009), in the duration (i.e., shame is more persistent than embarrassment) (
Scheff 1994) and in the focus of attention (i.e., shame affects the self, embarrassment affects the persona, the apparent self). However, these two emotions have also some features in common. They are associated with the same specific physiological reactions (e.g., blushing) (
Sabini and Silver 2005) and the same action tendency, leading people to hide and reduce their social presence and making movement and speech more difficult and less likely (Asendorpf 1990, Keltner and Buswell 1997, Lewis 1993; Miller 1996). However, it has also been reported that, differently from shame, embarrassment leads to reparative behaviours to re-gain social approval (Feinberg et al. 2011;
Keltner and Buswell 1997; Leary et al. 1996). At the neural level, shame was selectively associated with with dlPFC, posterior cingulate cortex and sensory-motor cortex, whereas embarrassment with vlPFC, amygdala and occipital areas, and both emotions with hippocampus and midbrain (Bastin et al. 2014). However, it must be acknowledged that, since the distinction between shame and embarrassment is not sharp, being classified according to the private-public, moral-conventional or low-high intensity dimensions, it is not easy to establish which brain areas are involved in processing these emotions, and which areas might selectively process one of the two emotions.
If the difference between shame and embarrassment is not as clear-cut, the distinction between guilt and the other two emotions is more evident. Guilt occurs when the violation of social norms induces harm or suffering to other individuals (
Hoffman 1982;
Grecucci et al. 2021;
Piretti et al. 2020), typically in a relationship or among members of the same group (
Fiske 1991). Differently from shame and embarrassment in which, respectively, the self and the persona are perceived as defective, in guilt a specific action is typically perceived as wrong (
Hoffman 1982;
Lewis 1971;
Lewis et al. 1993). The occurrence of guilt induces remorse and behavioural responses that aim to repair the wrong action (
Tangney et al. 2007). This difference in the focus, self-oriented and other-oriented, for shame and guilt respectively, has important consequences on empathy for other people: while guilt tends to increase the empathic concern towards other people, empathic responses seem to be disrupted by self-oriented distress associated with shame (
Tangney et al. 2007).
Table 1.
Differences between shame and guilt.
Table 1.
Differences between shame and guilt.
|
SHAME |
GUILT |
Target |
What we are: related to the entire self. ”I’m bad” |
What we do: related to specific behaviours “What I did has been bad” |
Level |
Interpersonal – it occurs only with others |
Intrapsychic – it occurs alone |
Emotional activation |
Painful |
Less painful |
Emotional perception |
Difficult to recognize |
Easy to recognize |
Action tendency |
Motivates hiding and inhibition |
Motivates reparation to the situation |
Relation with aggression, hostility, violence, externalization |
Increased for shame-proneness individuals |
Decreased for guilt- proneness individuals |
Scapegoat |
Blame mainly others |
Blame myself |
Responsibility |
Deflected outward |
Accepted |
In a review, Bastin and collaborators (2014) suggested that guilt processing was selectively associated with ventral ACC, precuneus, premotor and posterior temporal areas. In addition, both guilt and shame processing were associated with anterior insula and dACC, and that both guilt and embarrassment processing with dorsomedial prefrontal cortex (dmPFC), vlPFC and anterior temporal lobe (ATL) (Bastin et al. 2014). In addition, a recent meta-analysis (
Gifuni et al. 2017) partially confirmed the guilt neural substrates proposed by Bastin and collaborators (2014), reporting the activation of precuneus, dorsal ACC, dmPFC, and posterior temporal areas, in association with guilt processing (
Gifuni et al. 2017).
However, it is worth noting that studies investigating self-conscious emotions used heterogeneous methods that prevent any firm conclusions from being drawn. For this reason, we run a meta-analysis study including neuroimaging research on the neural substrates of negative self-conscious emotions, i.e., to pinpoint brain areas consistently associated with shame, embarrassment and guilt processing. We predicted that shame, embarrassment and guilt may show different brain activations mirroring behavioural differences related to the emotions, together with some shared activations in light of their moral-self-conscious nature.
2. Materials and Methods
In order to find studies investigating the neural underpinnings of shame, embarrassment and guilt we conducted a research on PubMed (
https://www.ncbi.nlm.nih.gov/pubmed/) using the terms ((“fMRI” OR “functional magnetic resonance imaging” OR “PET”) AND (“shame” OR “embarrassment” OR “guilt” OR “moral emotions” OR “self-conscious emotions” OR “moral violations” OR “social standard violation”)), and setting a range of dates between January 1st 1995 and December 14th 2018. This research identified 123 studies.
Subsequently, we refined our research by applying the following criteria:
- 1)
paper originally published in English;
- 2)
fMRI or PET studies including task-related whole brain analyses. Studies reporting region of interest (ROI analyses, resting-state fMRI analyses, diffusion tensor imaging (DTI) or voxel-based morphometry (VBM) were excluded;
- 3)
participants were healthy adults: In case of studies involving neurological or psychiatric patients, children or adolescents, we considered only contrasts involving healthy controls, if reported;
- 4)
Studies investigating the neural underpinnings of shame and guilt were included into two different sets, for two distinct meta-analyses. Specifically, we included studies contrasting shame/embarrassment vs. neutral or other emotional conditions and guilt vs. neutral or other emotional conditions. Studies failing to distinguish embarrassment/shame and guilt were excluded.
Since the difference between shame and embarrassment is not clear-cut, as they can be classified according to different criteria, and since the same physiological reactions and the same action tendencies, and their distinction is still a matter of debate, we decided to include in the same set both shame and embarrassment.
This method allowed us to identify 15 studies for the shame/embarrassment set (168 foci, 373 total subjects) and 17 studies for the guilt set (123 foci, 367 total subjects) (see Table 3). The most used paradigm in the studies analysed was emotion induction through verbal scripts (shame/embarrassment = 5; guilt = 7), pictures (shame/embarrassment = 5), both scripts and pictures (guilt = 3), vignettes (shame/embarrassment = 3) or movies (guilt = 1), while a few studies used the recollection of autobiographical memories through verbal scripts (shame/embarrassment = 1; guilt = 3), interpersonal games (shame/embarrassment = 1, guilt = 3), or implicit association task (guilt = 1).
Table 2.
Studies investigating shame/embarrassment and guilt brain processing.
Table 2.
Studies investigating shame/embarrassment and guilt brain processing.
Subset |
Authors |
Paradigm |
Stimulus type |
Contrasts |
Foci |
Subjects (Females) |
Shame/ |
Bas-Hoogendam et al. 2017 |
Induction |
Verbal scripts |
Unintentional violations > neutral |
5 |
21(15) |
embarrassment |
Berthoz et al. 2002 |
Induction |
Verbal scripts |
Unintentional violations > normal |
15 |
12(0) |
|
Finger et al. 2006 |
Induction |
Verbal scripts |
Moral and social with audience > social and neutral without audience |
2 |
16(-) |
|
Krach et al. 2011 |
Induction |
Vignettes |
Vicarious embarrassment > neutral |
9 |
32(17) |
|
Krach et al. 2015 |
Induction |
Vignettes |
Social pain > social neutral |
17 |
16(0) |
|
Laneri et al. 2017 |
Induction |
Vignettes |
Empathic embarrassment > neutral |
14 |
51(21) |
|
Melchers et al. 2015 |
Induction |
Pictures |
Vicarious embarrassment > neutral |
6 |
60(39) |
|
Michl et al. 2012 |
Induction |
Verbal scripts |
Shame > neutral |
10 |
14(7) |
|
Morita et al. 2008 |
Induction |
self- and other-faces |
Self-face > other-face |
9 |
19(10) |
|
Morita et al. 2012 |
Induction |
self- and other-faces |
Self-face > other-face |
29 |
15(2) |
|
Morita et al. 2014 |
Induction |
self- and other-faces |
Self-face > other-face |
17 |
32(16) |
|
Morita et al. 2016 |
Induction |
self- and other-faces |
Self-face > other-face |
13 |
18(0) |
|
Paulus et al. 2015 |
Induction |
Vignettes |
Positive correlation of vicarious embarrassment |
11 |
32(17) |
|
Paulus et al. 2018 |
Induction |
Vignettes |
Fremdscham > neutral |
15 |
34(0) |
|
Takahashi et al. 2004 |
Induction |
Verbal scripts |
Embarrassment > neutral |
10 |
19(9) |
|
Wagner et al. 2011 |
Recollection |
Verbal scripts |
Shame > neutral |
10 |
15(15) |
|
Zhu et al. 2018 |
Interpersonal game |
Pictorial stimuli (dots) |
Shame > happiness |
2 |
30(17) |
Guilt |
Basile B et al. 2011 |
Induction |
Verbal and facial stimuli |
Guilt > anger and sadness |
3 |
22(13) |
|
Finger et al. 2006 |
Induction |
Verbal scripts |
Moral > social and neutral |
5 |
16(-) |
|
Fourie et al. 2014 |
implicit association task |
verbal and facial stimuli |
Prejudice feedback > neutral feedback |
5 |
22(22) |
|
Gilead et al. 2016 |
Induction |
Verbal scripts |
Guilt > anger, joy, pride |
10 |
19(14) |
|
Gradin et al. 2016 |
Interpersonal game |
Verbal |
Defection > cooperation |
6 |
25(17) |
|
Green et al. 2012 |
Induction |
Verbal scripts |
Guilt > indignation (Within HC) |
7 |
22(18) |
|
Kédia et al. 2008 |
Induction |
Verbal scripts |
Guilt > self-anger |
4 |
29(14) |
|
Michl et al. 2012 |
Induction |
Verbal scripts |
Guilt > neutral |
19 |
14(7) |
|
Molenberghs et al. 2015 |
Induction |
Video |
Civilians > Soldiers |
3 |
48(24) |
|
Morey et al. 2012 |
Induction |
Verbal scripts |
Positive correlation of guilt |
6 |
16(0) |
|
Peth et al. 2015 |
Recollection |
Verbal |
Guilty action > neutral |
10 |
20(6) |
|
Shin et al. 2000 |
Recollection |
Verbal scripts |
Guilt > neutral |
8 |
8(0) |
|
Takahashi et al. 2004 |
Induction |
Verbal scripts |
Guilt > neutral |
5 |
19(9) |
|
Ty et al. 2017 |
Induction |
Verbal and pictorial stimuli |
Restitution > harm |
1 |
18(9) |
|
Wagner et al. 2011 |
Recollection |
Verbal scripts |
Guilt > neutral |
24 |
15(15) |
|
Yu et al. 2014 |
Interpersonal game |
Pictorial stimuli (dots) |
Self-incorrect > both incorrect |
1 |
24(11) |
|
Zhu et al. 2018 |
Interpersonal game |
Pictorial stimuli (dots) |
Guilt > happiness |
5 |
30(17) |
2.1. Statistical Analysis
Analyses were conducted using the software GingerALE v3.0.2 (
http://brainmap.org/). The activation likelihood estimation method, implemented in the software (
Eickhoff et al. 2009;
2012;
Turkeltaub et al. 2012), uses probability theory to define the spatial convergence of foci reported in the selected studies. Specifically, a Gaussian blur with an empirically-derived full-width half maximum (dependent on the number of participants included in the study) is applied to each focus from a single study. Then, all the foci from a single study are represented in a modelled activation map and voxel-wise ALE scores are computed combining all the individual maps. To distinguish between true convergence of foci from random noise a permutation test is applied. We adopted the method described by
Turkeltaub et al. (
2012) that minimizes within-study effects, preventing the summation of foci of the same experiment that are placed close to each other. For studies reporting between-subjects contrasts, we used the number of participants included in the smallest group as the total number of study participants.
The analyses were performed on studies’ coordinate in Talaraich space. So, in case coordinate were reported in MNI space we converted them to Talaraich space, using the coordinate converter of the GingerALE software, while we kept the same coordinates in studies reporting results in Talaraich space. For each set of studies, we performed the meta-analysis applying a cluster-level family-wise error correction using an uncorrected p-value < .001 for individual voxels, 1000 permutations and a cluster-level threshold of p < .05, as suggested by Eickhoff and collaborators (2016).
Finally, we performed further analyses. We run 1) a conjunction analysis aiming to elucidate common neural activations of shame/embarrassment and guilt; 2) a subtraction analysis in order to highlight specific neural activations of either shame/embarrassment or guilt.
Subtraction analyses were performed subtracting one of the outputs of the previous analyses (ALE images) to the other (i.e., Shame/Embarrassment vs. Guilt, Guilt vs. Shame/Embarrassment). Since the two sets of studies differ in the sample size, GingerALE software computes a simulation of data randomly pooling the original data and then creating two new sets of the same size of the original datasets. For each new dataset, an ALE image is created and then subtracted to the other. These simulated images are compared with the real observed data. After 104 permutations, a voxelwise P-value image reveals for each voxel, where the real data is located in the distribution of all the possible values (for that specific voxel). Values are converted into z-scores. Subtraction analyses results are presented with a threshold of p < .05 uncorrected and a cluster size > 200 mm3, since input data for these contast analyses were already corrected for multiple comparisons, as in previous studies (
Eickoff et al. 2012, Laird et al. 2005;
Zmigrod et al. 2016). Results are visualized using MricoGL (
https://www.mccauslandcenter.sc.edu/mricrogl).
Author Contributions
Conceptualization, L.P., E.P., A.G.; methodology, L.P., A.G.; software, L.P., E.P.; validation, all the authors; formal analysis, L.P., E.P., C.G., A.G.; investigation, L.P., E.P., A.G.; resources, L.P., A.G.; data curation, L.P., E.P., A.G.; writing—original draft preparation, L.P., E.P., A.G.; writing—review and editing, C.G., R.I.R., R.J.; visualization, L.P., E.P.; supervision, A.G.; project administration, A.G.; funding acquisition, L.P. All authors have read and agreed to the published version of the manuscript.”