Neural responses to facial attractiveness in the judgments of moral goodness and moral beauty

The judgments of moral goodness and moral beauty objectively refer to the perception and evaluation of moral traits, which are generally influenced by facial attractiveness. For centuries, people have equated beauty with the possession of positive qualities, but it is not clear whether the association between beauty and positive qualities exerts a similarly implicit influence on people's responses to moral goodness and moral beauty, how it affects those responses, and what is the neural basis for such an effect. The present study is the first to examine the neural responses to facial attractiveness in the judgments of moral goodness and moral beauty. We found that beautiful faces in both moral judgments activated the left ventral occipitotemporal cortices sensitive to the geometric configuration of the faces, demonstrating that both moral goodness and moral beauty required the automatic visual analysis of geometrical configuration of attractive faces. In addition, compared to beautiful faces during moral goodness judgment, beautiful faces during moral beauty judgment induced unique activity in the ventral medial prefrontal cortex and midline cortical structures involved in the emotional-valenced information about attractive faces. The opposite comparison elicited specific activity in the left superior temporal cortex and premotor area, which play a critical role in the recognition of facial identity. Our results demonstrated that the neural responses to facial attractiveness in the process of higher order moral decision-makings exhibit both task-general and task-specific characteristics. Our findings contribute to the understanding of the essence of the relationship between morality and aesthetics.


Introduction
The face is a highly salient social signal; thus, people pay attention to faces. Facial features that indicate an individual's biological and evolutionary fitness (Darwin 1871) are commonly considered attractive (Perrett et al. 1999;Rhodes et al. 2001). Attractiveness carries positively valenced stereotypes (i.e., the attractiveness halo effect) (Dion et al. 1972;Nisbett and Wilson 1977). Facial attractiveness and the positively valenced stereotypical information with which it is typically associated is believed to come to mind spontaneously upon encountering attractive individuals (Olson and Marshuetz 2005;van Leeuwen and Macrae 2004).
The evaluation of facial attractiveness is fast and automatic and serves as the basis for inferences about a person's socially relevant personal traits (Chatterjee et al. 2009;Locher et al. 1993;Olson and Marshuetz, 2005;Sui and Liu 2009), initially based on the perception of subtle variations in the facial features that have evolutionary significance Zebrowitz et al. 2003).
1 3 Chatterjee et al (2009) found that the core neural system, which is specialized for early visual processing of face perception and face identity (Fang and He 2005;Grill-Spector et al. 2017), mainly including the interior occipital gyrus (IOG) and the fusiform face gyrus (FFA), is also involved in the visual processing of attractive faces before object identification, even when subjects were not explicitly attending to the attractive faces (Kranz and Ishai 2006;Winston et al. 2007;Chatterjee et al. 2009). Iaria et al. (2008) also found that the FFA is activated when making facial attractiveness judgments (Iaria et al. 2008) and that the FFA mainly processes the invariant features (configuration) of the faces. The IOG provides a coarse processing of some aspects of the visual stimulus, then regions more anterior and including visual association cortices (e.g., FFA) construct more extensive, fine-grained perceptual representations that depend more on processing the configuration of the faces (Liu et al. 2010;Pitcher et al. 2011a, b;Steeves et al. 2006). EEG studies have shed light on the temporal dynamics of attractiveness processing, indicating that attractiveness may begin to exert effects as early as the structural encoding stage. Early indices of visual processing, including the early posterior negativity and N170 components, are modulated by facial attractiveness (Marzi and Viggiano, 2010;Zhang and Deng, 2012). The FFA is thought to be the cortical source of the N170 response sensitive to facial attractiveness (Pizzagalli et al. 2002).
The evaluation of facial attractiveness is believed to be an extension of functionally adaptive system for understanding the communicative meaning of emotional expressions, namely the ability to read affective meanings, whose functional role is to prepare one's behavior in relation to the other person (e.g., approach or avoidance) (Todorov 2008). An attractive face is viewed as an approach expression that suggests not only a person's positive affective state, but also a person's capacity to act in a confident, assertive, and friendly way (Montepare and Dobish 2003). Attractive faces seem to be not simply a consequence of aesthetically pleasing characteristics, but rather an integration of multiple cues including physical appearance, inter-personal engagement, and emotional expression. After the early structural encoding stage of face processing, attractiveness has effects in later stages whereby affective and identity information is extracted from faces (Hahn and Perrett 2014). Socially relevant positively valenced information of attractive faces is subsequently labeled by the emotional systems involved in the perception of facial expressions, such as the medial orbitofrontal cortex (mOFC) , to which the occipitotemporal cortices (OTC) have direct projections (Rolls 2007). Previous works (Aharon et al. 2001;Bray and O'Doherty 2007;Cloutier et al. 2008;Hahn and Perrett 2014;Liang et al. 2010;O'Doherty et al. 2003;Winston et al. 2007;Zaki et al. 2011) have shown that a number of brain regions activations, including VTA, vmPFC, basal ganglia, and the NAcc, particularly the mOFC, are involved in the emotional evaluation of attractive faces.
Facial attractiveness, considered to be a fundamental aspect of human social interaction, has a profound influence on human social behaviors (Chatterjee 2015;Kocsor et al. 2013;Maner et al. 2003;Olson and Marshuetz 2005;Todorov et al. 2008;Zebrowitz and Collins 1997). The individual facial attractiveness influences the evaluation and judgment on individual's moral traits, known as the beautyis-good stereotype (Griffin and Langlois 2006). This reliance on aesthetic features to infer moral character suggests a close link between aesthetic and moral valuation (Ferrari et al. 2017). In general, the more attractive a person's face is, the more likely he or she is to be perceived as possessing socially desirable personalities, moral traits, such as kindness, honesty, friendliness and trustworthiness and higher moral standards (Dion et al. 1972;Feingold, 1992;Griffin and Langlois 2006;Langlois et al. 2000;Snyder et al. 1977;Zebrowitz and Montepare 2008), and to be happier and more successful in later life (French 2002;Griffin and Langlois 2006;Langlois et al. 2000;Lewis and Bierly 1990;Rynes and Gerhart 1990). Moral traits judgments from faces may reflect the general evaluation of faces or, at least, approximate this evaluation rather well .
There is good empirical evidence that the evaluations of facial attractiveness and moral traits may rely on similar neural mechanisms (Luo et al. 2019;Tsukiura and Cabeza 2011b;Wang et al. 2015). For instance, Tsukiura and Cabeza (2011a, b) found that activity in the mOFC increased as a function of both perceived attractiveness and moral goodness ratings, and the activations elicited by attractive faces and morally positive behaviors were strongly correlated with each other. Wang et al (2015) identified the common involvement of mOFC in the facial attractiveness and moral beauty judgments. Luo et al (2019) further investigated the neural mechanisms underlying the integrated aesthetics of facial attractiveness and moral beauty, and also found that the activity of mOFC was collectively modulated by facial and moral beauty. In these studies, the researchers reasonably interpret the activation of mOFC as related to the positive emotions triggered by rewarding properties of the stimuli that are engaged automatically when viewing attractive faces or viewing positive moral behaviors.
Although neuroimaging research has greatly enhanced our understanding of responses to facial attractiveness and moral judgments, a closer look reveals limitations of previous literature. First, in previous studies, facial attractiveness and moral judgments were mostly investigated separately, and largely at the level of consciousness (Todorov 2008;Todorov et al. 2008;Tsukiura and Cabeza 2011b;Wang et al. 2015). In fact, the perceptions of facial attractiveness and moral evaluation are always simultaneously 1 3 involved in human beings. In reality, the observers are often unaware of the extent to which the individual's facial attractiveness influences their judgments when they make explicit judgments on an individual's moral behaviors or moral traits (Luo et al. 2019). Both types of perceptions are inherently imbued with affect and, most likely, both serve to infer potential intentions of others (Said et al. 2011;Todorov 2008). As such, it would be of high ecological validity to examine how these forms of information are processed and integrated naturally, which may also provide a window to understand the influence of face appearance on moral judgments. Second, although researchers have shown a profound interaction between facial attractiveness and moral valuation (Dion et al. 1972;Ferrari et al. 2017;Luo et al. 2016;Oosterhof and Todorov 2008;Stolier et al. 2018;Zhang et al. 2014), they seemed to focus much of their discussion on activations within reward networks, they did not further examine the possibility that perceptual responses to attractive faces might trigger the activation of these reward circuits. Third, the shared neural mechanisms between facial attractiveness and moral judgments are not enough to suggest that facial attractiveness contributes equally to different kinds of moral judgments. Nor it is clear how the positively valenced information of attractive faces generates value signals in different sectors of the reward circuit. Studies have shown that the neural responses to attractive faces depend on the task being performed, specifically, the brain regions of attractive faces activated by explicit aesthetic judgments were different from those activated by age judgments unrelated to facial attractiveness (Chatterjee et al 2009Chatterjee and Vartanian 2016;Winston et al. 2007). In other words, the explicit assessment of a face for its attractiveness recruits value signals in the mOFC that are not taken into account when the same face is inspected with respect to other task demands (Chatterjee et al. 2009;Kim et al. 2007).
Moral goodness and moral beauty are two kinds of moral judgments that are often paid attention to in human daily life. They have one thing in common, that is, the object of both judgments is universal human good moral traits (kindness, virtue, etc.) or positive moral behaviors (Cheng et al. 2020). Moral goodness judgment requires the subjects to evaluate how good an individual's positive moral behavior is, whereas moral beauty judgment requires the subjects to assess the inner beauty of the individual who perform the positive moral behavior. Both judgments consist of the appraisal of perceived positive moral traits, such as the virtues of helping others or of sacrificing one's own life for others, and both recruit the inferior temporal cortex (ITC), a ventral visual brain region involved in the perception, recognition and representation of visual objects (Nobre et al. 1994;Vandenberghe et al. 1996), and faces , and the lateral orbitofrontal cortex (OFC), a reward circuit that significantly responds positively to rewarding objects in general (Cheng et al. 2020;Troiani et al. 2016).
Although moral beauty and moral goodness objectively refer to the same human moral acts or virtues, they had differences in the strength of emotional reactions and motivations (Cheng et al. 2020;Diessner et al. 2006Diessner et al. , 2008Keltner and Haidt 2003). When judging a positive moral behavior of an individual as moral goodness, the observer quickly evaluates the level of morality by focusing on the behavior itself, the observation of the behavior and the perception of the visual features of the individual performing the moral behavior (Lieberman 2007), but the observer might remain emotionally unmoved and un-elevated (Cheng et al. 2020;Diessner et al. 2006;Güsewell and Ruch 2012;Wang et al. 2015). The object of moral goodness is experienced as part of the material world under the circumstances of no deeply emotional involvement (Diessner et al. 2008). It seems to be concerned with the psychological processes of physical, visible characteristics and behaviors of one person, which are perceived through sensory patterns and experienced as a part of the physical world. In contrast, when evaluating the inner beauty of a person who performs the same positive moral act as moral goodness, the observer focuses on the internal spiritual, emotional and experiential characters of this person (Haidt 2003) by understanding his or her mental states (i.e., theory of mind, ToM) and vicariously experiencing similar feelings (i.e., empathy) (Cheng et al. 2020). As a kind of aesthetic judgment, moral beauty can generate aesthetic feeling and triggers emotional elevation (Diessner et al. 2006;Haidt 2003Haidt , 2007Keltner and Haidt 2003;Pohling and Diessner 2016). This moral elevation is an approach-oriented emotion, that is, it can trigger a feeling of warmth and expansion and a common eagerness of self-improvement to prompt the observer to do the similar positive moral behavior in the similar situation (Algoe and Haidt 2009;Diessner et al. 2013;Haidt 2003;Haidt and Keltner 2004;Schnall et al. 2010;Shiota et al. 2014;Van dee Vyver andAbrams 2015, 2017). Whether the brain triggers emotional elevation seems to be the most essential difference between moral goodness and moral beauty judgments. Therefore, when assessing the level of moral goodness and moral beauty of a particularly good-looking individual's morally positive behavior, will the attractive appearance of the individual have different effects on the observer's judgments of goodness and beauty? For centuries, people have equated beauty with the possession of positive qualities, but it is not clear whether the association between beauty and positive qualities exerts an implicit influence on people's responses to moral goodness and moral beauty, how it affects, and what is the neural basis for such an effect. Given the introduction above, indicating that moral goodness and moral beauty judgments rely on both common cognitive processes as well as distinct and specific cognitive components, we hypothesize that the visual feature processing of attractive faces is likely to automatically activate the ventral occipitotemporal cortices during the judgments of moral goodness and moral beauty. The lower-level visual analysis of facial attractiveness might play a role in the judgments of moral goodness or moral beauty. In addition, one study showed that brain regions involved in emotional meanings are typically activated by triggering an observer's emotional response (Emery and Amaral 2000). We speculate that the affective meanings of attractive faces may separate out into different patterns of representation. As a moral judgment and a aesthetic judgment that requires deep emotional involvement, attractive faces in moral beauty judgment task would evoke brain regions related to emotional evaluation. But the moral goodness judgment does not require the observer to be emotionally involved, so an individual's attractive face may not activate the brain regions associated with the representation of emotional meanings. Facial attractiveness may influence both types of judgments in similar and distinct ways.

Subjects
Twenty-two (mean age = 21.5 years, s.d. = 1.7) healthy, right-handed female undergraduate students with no history of psychiatric and neurological disorders were recruited as subjects in South China Normal University. Only female subjects were invited to participate in the experiment for the purpose of eliminating possible confusion between the gender of the subjects and the gender of the characters pictured in experimental materials (Cheng et al. 2020;Liang et al. 2010;Luo et al. 2019). More importantly, females are reported to generate more consistent aesthetic ratings across subjects for heterosexual faces than males (Luo et al. 2019;Tsukiura and Cabeza 2011a), and females are more easily moved by scenes of moral judgments than males (Amoyal 2014;Freeman et al. 2009;Haidt and Keltner 2004;Janicke and Oliver 2017;Landis et al. 2009). All subjects were assessed to ensure normal or corrected to normal vision. Informed consent was obtained from all subjects before the experiment according to the requirements of the Research Ethics Review Board of South China Normal University. Subjects received monetary compensation for their participation in the study.

Materials
As described in our previous study (Cheng et al. 2020), experimental materials consisted of a series of black-andwhite scene drawings depicting morally positive behaviors acted by male characters in daily life. In accordance with previous studies (Luo et al. 2019;Wang et al. 2015), short sentences that described morally positive behavior (i.e., He helps the elderly cross the road, he escorts the injured to the hospital, he visits the orphans in the orphanage, etc.) were collected. Then, black-and-white scene drawings were created based on the content of these sentences. For each scene, a red capital letter "A" was drawn beside the main character to distinguish the character from other figures shown in the scene drawing. And for each scene, there were two versions of the main character's facial attractiveness: a beautiful-face version and an ugly-face version, both of which are based on real male photos of beauty and ugliness. The manipulation of facial attractiveness was not limited to the systematic change of a certain facial feature but rather to the random combination of several facial features (Cui et al. 2019). Our research focused on Fechner's concept of internal psychophysics (the relationship between subjective experience and physical characteristics of neural systems), not on the external psychophysics (the relationship between subjective experience and physical characteristics of stimuli). Although we are interested in the brain areas of neural responses to facial attractiveness on moral goodness and moral beauty, we did not know what specific visual features caused such responses.
Twenty-two female subjects (mean age = 21 years, range 18-24 years; healthy, right-handed college students from South China Normal University with no history of psychiatric and neurological disorders) who had never participated in our similar experiments before, were recruited for the pilot study to validate the scene drawings. The sample size was determined by G*Power (Faul et al. 2007) analysis that would provide sufficient power to detect a medium effect size (f = 0.25). Similar to the method of validating materials used in Cui et al (2019), first, subjects were asked to evaluate the moral level of the main character's behavior described in the scene drawing on a 7-point scale (1 = especially good to 7 = especially bad), with the character's face blurred to prevent the influence of facial attractiveness on the behavior evaluation. According to the results, scene drawings with average scores greater than 4 were excluded to ensure that all the behaviors described in the scene drawings were positive. Second, the same subjects were asked to assess the facial attractiveness of the main character also on a 7-point scale (1 = especially beautiful to 7 = especially ugly), with other parts of the drawings blurred to avoid the impact of the scene and the main character's behavior on the facial attractiveness assessment. Based on the results, drawings with average scores greater than or equal to 4 were excluded for the beautiful-face version scene drawings, and average scores less than or equal to 4 were excluded for the ugly-face version scene drawings. Pilot task orders were counterbalanced across subjects. Finally, a total of 104 scene drawings depicting 4 different moral levels of positive behavior (especially good, rather good, slightly good and neutral), of which each level contained 26 scene drawings. Each scene had one beautiful-face version and one ugly-face version, composing 208 validated scene drawings that were used as the materials for the fMRI experiment (see Fig. 1). Scene drawings were 589 × 500 pixels.
The scene drawings depicting morally positive behaviors were then randomly assigned to two sets used for different moral judgments, respectively, to avoid the repeated use of the same type of stimuli in different tasks. Each set contained 52 scenes drawings, in which each scene had a beautiful-face version and an ugly-face version. The mean rating scores for each moral level of each version of faces in the two sets of materials were calculated and subjected to a 2 (set: the first vs. the second) × 2 (face: beautiful face vs. ugly face) × 4 (moral level: especially good vs. rather good vs. slightly good vs. neutral) repeated measures ANOVA using JASP to examine whether the materials of the two sets were equivalent. Results showed that the main effect of set was not significant, p > 0.05, the main effect of face was not significant, p > 0.05, the main effect of moral level was significant, F(3, 63) = 265.07, p < 0.001, with a reasonable increase observed in the average rating scores as a function of moral levels (m.d. ± s.e. 1.63 ± 0.07, 2.18 ± 0.07, 2.3 ± 0.07 and 3.5 ± 0.07 for level 1, 2, 3 and 4, respectively). The interactions between set and face, between set and level, between face and level were all not significant, ps > 0.05. The three-way interaction among set, face and moral level was also not significant, p > 0.05. Results showed no significant difference between the two sets of materials, which ensured that the two sets were equivalent. These results are significantly important because it indicates that the difference in activation of brain regions involved in the judgments of moral goodness and moral beauty was not caused by differences in experimental stimuli.
Thus, for each judgment task, 52 scene drawings depicting 4 different moral levels (i.e., especially good, rather good and slightly good and neutral) were used that contained 52 beautiful-face scene drawings and 52 ugly-face scene drawings. The allocation of these parts of the materials was balanced across different tasks. The selected stimuli generated the following conditions with an equal number in each: the beautiful-face-version stimuli in moral goodness judgment (BFMG), the ugly-face-version stimuli in moral goodness judgment (UFMG), the beautiful-face-version stimuli in moral beauty judgment (BFMB) and the ugly-face-version stimuli in moral beauty judgment (UFMB). Different moral levels contained in each condition were used as parameters to identify brain regions involved in different judgment tasks, which displayed altered activity simply as a function of the moral ratings in the fMRI data analysis.

Experimental procedure
The fMRI experiment contained two tasks: (i) the moral goodness judgment (MG), (ii) the moral beauty judgment the above shows behavior with different moral levels acted by the main character with a beautiful face (BF); the below shows the same behavior but acted by the main character with an ugly face (UF) 1 3 (MB). Another group of twenty-two female subjects, who had not participated in the pilot experiment nor similar experiments before, were recruited and performed the tasks listed above. in moral goodness judgment task, subjects were asked to look at the scene drawings and rate the level of goodness of the main character's moral behavior on a 4-point scale (1 = especially good, 4 = neutral); in moral beauty judgment task, subjects were instructed to look at the scene drawings and rate the inner beauty of the main character on a 4-point scale (1 = especially beautiful, 4 = neutral). Two equivalent sets that were used separately in moral goodness and moral beauty judgments were balanced across subjects. Two tasks constituted two runs of the whole fMRI experiment. Run orders were counterbalanced across subjects. The order of presentation of experimental stimuli was randomized between subjects in these tasks. During scanning, an event-related design was applied with two runs.
As described in Cheng et al. (2020), a single trial began with the instrument slide, which enlightened the subjects about the specific tasks to perform next. This slide was presented for 4000 ms. Next, one scene drawing was presented with a 4-point scale under it, in which the number indicated the corresponding level of moral behavior or inner beauty of the main character. Subjects were instructed to assess the moral level of the main character's behavior and the level of inner beauty of the main character depicted in the scene drawing as quickly as possible within 4000 ms. Using a response box with buttons labeled 1 to 4, half of the subjects made a left index finger response (i.e., 1 or 2) for the "especially" and "rather" levels of stimuli and a right index finger response (i.e., 3 or 4) for the "slightly" and "neutral" levels. Reverse responses were used for the other half of subjects to balance any influence of movement on data. Once the response was recorded, the rating bar disappeared and was followed by a blank display for a random period ranging between 500 and 5000 ms. If the scene drawing slide was not presented for 4000 ms after subjects pressed the key, the scene drawings disappeared, and a blank display will appear for the remaining time. The maximum duration of the presentation of the scene drawing slide was 4000 ms, based on the results of a pre-behavioral test. Each run was scanned for approximately 13 min. The detailed experimental procedure (See Fig. 2) is as follows: After the fMRI scanning, to examine the efficiency and the stability of facial features of the main character in the scene drawings, subjects were immediately asked to assess the facial attractiveness on a 7-point scale (1 = especially beautiful to 7 = especially ugly). All the stimuli were the same as those applied in fMRI scanning for each subject, except other parts of the drawings blurred to avoid the Fig. 2 The experimental flowchart. Each subject completed 2 judgment tasks which constituted 2 runs of the whole fMRI experiment. Stimuli were a set of scene drawings depicting morally positive behaviors. Subjects were required to rate the level of moral goodness and inner beauty of A's behavior in the scene drawings. Run orders were counterbalanced across subjects. The order of presentation of experimental stimuli was randomized between subjects in two tasks impact of the scene on the facial attractiveness assessment. Like the pilot experiment, scene drawings with average scores greater than or equal to 4 were excluded for the beautiful-face version scene drawings, and average scores less than or equal to 4 were excluded for the ugly-face version scene drawings.

fMRI data analysis
Data preprocessing and statistical analyses of all images were conducted using SPM12 (Welcome Trust Center for Imaging, London, UK; http:// www. fil. ion. ucl. ac. uk/ spm). For each run of each participant, in preprocessing, images were spatially realigned to the first volume to correct for head movements (no individual run was more than 3.0 mm displacement or 1.0° rotation). These aligned functional images (coregistered with the anatomical image) were then spatially normalized into the Montreal Neurological Institute (MNI) template and resampled with voxel size of 2 × 2 × 2 mm 3 . After that, the data were smoothed with an isotropic full-width half-maximum 6 mm Gaussian kernel.
In 208 scene drawings, the facial attractiveness (beautiful face or ugly face) of the main character was assessed by 22 participants on a 7-point scale (1 = especially beautiful to 7 = especially ugly, see Fig. 2). Based on the results, drawings with average scores greater than or equal to 4 were excluded for the beautiful-face version scene drawings, and average scores less than or equal to 4 were excluded for the ugly-face version scene drawings.
For the statistical analyses of the functional data, condition-related brain activity changes were estimated using a general linear model (GLM) in four conditions at the single subject level, including the beautiful-face-version stimuli in moral goodness judgment (BFMG), the ugly-face-version stimuli in moral goodness judgment (UFMG), the beautiful-face-version stimuli in moral beauty judgment (BFMB) and the ugly-face-version stimuli in moral beauty judgment (UFMB). We defined each experimental condition as a predictor of interest that parametrically modeled moral ratings of the presented items, respectively, and convolved these predictors with SPM12's canonical hemodynamic response function. In the given trails, the duration of a predictor of interest was equal to 4000 ms. Six head-motion parameters estimated during the realignment were included in the model as predictors of no interest. The missed trials were not modeled and thus were included in the baseline. A highpass filter (HPF) with a cutoff period of 128 s was applied to remove low-frequency noise. Note that, the task events were designed to minimize corrections, to estimate the neural responses to facial attractiveness during moral goodness and moral beauty judgment. For each subject, conditionrelated brain regions displaying altered activity simply as a function of moral ratings were identified by contrasting the parametrically modulated predictor of interest against baseline. The four first-level individual contrast images were then fed to the second-level analysis employing a whole-brain random-effects model.
First, we calculated the main effect of beautiful faces in moral goodness and moral beauty judgment tasks, respectively (moral goodness, contrast: BFMG-UFMG; moral beauty, contrast: BFMB-UFMB) and the main effect of ugly faces (moral goodness, contrast: UFMG-BFMG; moral beauty, contrast: UFMB-BFMB) to investigate the neural responses to facial attractiveness in these two judgment tasks. We then performed a conjunction analysis between the above two beautiful face contrasts to identify cortical regions commonly involved in the processing of attractive faces in the judgments of moral goodness and moral beauty. Similarly, we also computed a conjunction analysis between the above two ugly-face contrasts to test whether ugly-face contrast in moral goodness judgment and ugly-face contrast in moral beauty judgment rely on similar neural mechanisms.
Second, we computed the direct comparisons between BFMG and BFMB, between UFMG and UFMB, to identify their unique cortical regions activation patterns. In addition, we also performed a conjunction analysis between the parameter-mediated contrasts of moral goodness (BFMG + UFMG) and moral beauty (BFMB + UFMB) to identify cortical regions commonly involved in moral goodness and moral beauty representation irrespective of facial attractiveness, i.e., beautiful face vs. ugly-face.
Finally, we performed leave-one-subject-out iterative cross-validation (LOOCV) ROI analyses (Kriegeskorte et al. 2009;Pegors et al. 2015;Wang et al. 2015) on the significant clusters resulting from the above whole-brain analysis to test whether activity within any clusters responding to facial attractiveness in the judgments of moral goodness and moral beauty responded differently to beautiful face versus ugly face as a function of judgment tasks. In the LOOCV analyses, on each iteration, data from a single subject were held out as the test set. ROIs were then defined based on a group analysis of the n-1 remaining subjects, using a sphere with 6 mm radius centered on the peak voxel of each significant cluster. Parameters estimates for each task were extracted from these ROIs in the n th subject, and the procedure repeated 22 times. Parameter estimates were then averaged across subjects and were subjected to repeatedmeasures 2 (task: MG VS. MB) × 2 (face: beautiful face VS. ugly face) ANOVAs. This method gave us an estimate of the response in each ROI to beautiful face and ugly face in different judgment tasks, using independent data sets to define the boundary of the cluster and the strength of the effect.
At the group level, the resulting reported voxels are significant at p < 0.05 (FDR-corrected), corrected for multiple comparisons across the whole brain. Clusters below the correction level were neither reported nor visualized. Statistical results were visualized with the MRIcron (https:// people. cas. sc. edu/ rorden/ mricr on) (McCausland Center for Brain Imaging, University of South Carolina) and BrainNetViewer (http:// www. nitrc. org/ proje cts/ bnv/) (Xia et al. 2013).

Results
In the present study, we manipulated two kinds of information for the scene drawings used as experimental stimuli: facial attractiveness and moral behaviors. The data of facial attractiveness was obtained by asking the same participants to assess the level of facial attractiveness of the main character in the same material after the fMRI scanning. The first step in the data analysis was to examine whether the classification of facial attractiveness was effective and stable, and then to investigate whether the judgments of moral goodness and moral beauty were affected by facial attractiveness in the same way.

Behavioral results
To examine the efficiency and the stability of facial attractiveness classification in the scene drawings, one independent-sample t-test was conducted on the rating scores of facial attractiveness, which were obtained in the pilot experiment and after the fMRI scanning, separately. The results showed that there was no significant difference in scores between the two experiments (t = 1.344, p > 0.05, Cohen's d = 0.16). This result indicated the validity of categorizing experimental stimuli into two versions about the facial attractiveness and the stability of subjects' assessment of facial attractiveness of the main character in scene drawings. In the next behavioral data analysis, task scores for half of the subjects were converted to the same standard as the other half of the subjects (i.e., reverse coded), with largest number representing a neutral level.
The average rating scores and the average response times (RTs) of ratings for each level of moral behaviors of the main character with different facial attractiveness were calculated in moral goodness (MG) and moral beauty (MB) judgments of the fMRI experiment and subjected to a 2 (task: MG vs. MB) × 2 (face: beautiful face vs. ugly face) × 4 (moral level: especially good/beautiful vs. rather good/beautiful vs. slightly good/beautiful vs. neutral) repeated measures ANOVA using JASP. The Greenhouse-Geisser correction for non-sphericity of data was applied as necessary. The uncorrected degree of freedom, corrected p values and effect sizes (partial η 2 ) are reported. The p values of the pairwise comparisons were adjusted using a Bonferroni correction. For all analyses, the significance level was set to 0.05. The results of average rating scores revealed that the main effect of task was marginally significant, F(1, 21) = 3.72, p = 0.07, partial η 2 = 0.15, showing that compared with moral goodness, positive moral behaviors were more likely to be rated beautiful (m.d. ± s.e. MB:2.31 ± 0.1; MG:2.39 ± 0.11); the main effect of face was marginally significant, F(1, 21) = 3.85, p = 0.06, partial η 2 = 0.16, showing that the main character with beautiful face was rated as behaving more beautiful or better (m.d. ± s.e. BF:2.33 ± 0.11; UF:2.36 ± 0.11) than ugly face; the main effect of moral level was significant, F(3, 63) = 150.85, p < 0.001, partial η 2 = 0.88, with a reasonable increase observed in the average rating scores as a function of moral levels (m.d. ± s.e. 1.54 ± 0.09, 2.12 ± 0.12, 2.24 ± 0.12 and 3.5 ± 0.09 for levels 1, 2, 3 and 4, respectively). The interactions between task and face, between task and level were all not significant, ps > 0.05. The interaction between face and moral level was significant, F(3, 63) = 4.9, p < 0.01, post hoc comparisons showed that the behavior of the main character with beautiful face was significantly rated either the most beautiful or the best on the moral level 1, t(21) = 3.35, p = 0.03. In addition, the behavior of the main character with beautiful face also showed a tendency to be more beautiful or better than that with ugly face on the moral level 2 and 3. The threeway interaction among task, face and moral level was not significant, p > 0.05 (see Fig. 3).
The analysis of the response times (RTs) of ratings revealed that the main effects of task and face were all not significant, ps > 0.05. The main effect of moral level was significant, F(3, 63) = 9.21, p < 0.001, with the fastest responses observed for the "especially" level of moral behaviors (m.d. ± s.e. 1845.81 ± 72.51 for level 1) and the slowest responses observed for the "neutral" level (m.d. ± s.e. 2132.15 ± 80 for level 4). Importantly, the interaction between both variables was not significant, ps > 0.05. This result is very important because it indicates that fMRI analyses focused on the activity of brain regions involved in the processing of facial attractiveness occurred in the judgment tasks of moral goodness and moral beauty, and were not confounded by differences in task difficulty.

fMRI results
To investigate the neural responses to facial attractiveness (beautiful faces, BFs and ugly faces, UFs), that occurred in the judgment tasks of moral goodness and moral beauty, our main analysis in GLM included 4 critical morally positive events (i.e., BFMG, UFMG, BFMB and UFMB) which parametrically modeled the moral ratings of the presented items, respectively.
First, compared to UFs, the condition of BFs in moral goodness judgment task significantly activated the left ventral occipitotemporal cortex (OTC), left superior temporal cortex (STC), left anterior temporal lobe (ATL) and left precentral gyrus/inferior frontal gyrus (IFG) (see Fig. 4A and Table 1), the opposite contrast showed no significant region; whereas compared to UFs, the condition of BFs in moral beauty judgment task significantly activated the left ventral occipitotemporal visual cortex (the left IOG and left FFA/lingual gyrus), left PCC/precuneus and vmPFC (refers to the ACC extended to mOFC) (see Fig. 4B and Table 1), the opposite contrast showed that the condition of UFs significantly activated the medial PFC (i.e., ACC), left middle frontal gyrus (MFG) and left temporo-parietal junction (TPJ) (see Fig. 4C and Table 1). Our results showed that the BFs in the judgment tasks of moral goodness and moral beauty commonly recruited 2 similar neural activations: left FFA (Temporal Fusiform Cortex, posterior division) and left IOG (Occipital Face Area, OFA). The conjunction analysis revealed significant overlap in the left IOG/OFA. We noticed that the conjunction analysis also activated the FFA, but did not reach a statistical significance, thus, we applied a small volume correction (SVC) to further check this issue. When performing the SVC, we defined a 10-mm radius sphere centered on the peak voxel of the left FFA. We found significant activation for the conjunction analysis in this area (see Fig. 5A and Table 1). To confirm a similar role of the left OTC on attractive faces in the judgments of moral goodness and moral beauty, we also calculated the correlation between the left IOG activations and between the left FFA activations under the conditions of attractive face for both tasks across subjects (see Fig. 5B). In the cluster of left IOG, we found a highly significant positive correlation (r = 0.69, p < 0.001), indicating that subjects with stronger IOG activation for moral goodness ratings also showed stronger IOG activation for moral beauty ratings. But left FFA correlations did not reach to be statistically significant and only just showed a positive trend (r = 0.31, p = 0.16).
Second, compared to BFs in moral goodness judgment, the BFs in moral beauty judgment uniquely activated the cingulate cortex, including the left PCC/precuneus and ACC extending to mOFC. The opposite contrast showed that the BFs in moral goodness judgment uniquely activated the left promotor area extending to the opercular part of the frontal area (precentral gyrus/IFG). In addition, compared to UFs in moral goodness judgment, the UFs in moral beauty judgment uniquely activated the left TPJ. And no differential activation was found for the opposite contrast (see Fig. 5C and Table 1).

Fig. 3
Results obtained from the behavioral data. Mean rating scores are presented as a function of moral level and facial attractiveness in the judgments of moral goodness and moral beauty Third, the conjunction analysis between the parametermediated contrasts of moral goodness and moral beauty irrespective of facial attractiveness revealed significant overlap in the right lateral OFC (see Fig. 6A). Further, we performed a LOOCV ROI analysis on this cluster to test whether activity responding to both moral goodness and moral beauty appreciations responded differently to beautiful face versus ugly face in those two different judgment tasks. It was found that there was not any significant effect, demonstrating that the activity in cortical region commonly involved in the judgments of moral goodness and moral beauty was not affected by facial attractiveness (see Fig. 6B).
To assess whether any of the above significant clusters responded selectively to beautiful or ugly faces in different judgment tasks, we performed a cross-validation analysis in which we iteratively defined the ROIs on n-1 subjects and then extracted parameters estimates for the "left-out" subject. Repeated measures ANOVAs indicated that there were significant effects of BFs over the left FFA and IOG/ OFA, with no significant interactions for either moral goodness or moral beauty judgments. These results suggest that the left FFA (Fig. 7A) and IOG/OFA (Fig. 7B) respond similarly to variation in BFs for these two judgment tasks (FFA, F(1, 21) = 5.37, p = 0.03, partial η 2 = 0.22; IOG/OFA, F(1, 21) = 4.45, p = 0.05, partial η 2 = 0.18), demonstrating the general role of the ventral occipitotemporal cortex in the automatic processing of BFs in both the judgments of moral goodness and moral beauty. In contrast, within the ROIs of the left PCC/precuneus (Fig. 7C) and vmPFC (i.e., ACC extending to mOFC) (Fig. 7D), the main effects of task and face were all significant (PCC/precuneus, task: F(1, 21) = 7.58, p = 0.01, partial η 2 = 0.27; face: F(1, 21) = 21.29, p < 0.001, partial η 2 = 0.5; ACC/mOFC, task: F(1, 21) = 14.98, p < 0.001, partial η 2 = 0.42; face: F(1, 21) = 3.82, p = 0.06, partial η 2 = 0.15), and more importantly, the interactions between face and task were significant (PCC/precuneus, F(1, 21) = 4.04, p = 0.05, partial η 2 = 0.2; ACC/mOFC, F(1, 21) = 4.75, p = 0.04, partial η 2 = 0.2), which were mainly driven by higher activity in BFs in moral beauty judgment than UFs in moral beauty judgment (PCC/precuneus, t(21) = 4.53, p < 0.001; ACC/mOFC, t(21) = 3.12, p < 0.001), and higher than BFs in moral goodness judgment (PCC/precuneus, t(21) = 3.41, p < 0.01; ACC/mOFC, t(21) = 4.16, p < 0.001). These results demonstrated the unique roles of the left PCC/precuneus and ACC/mOFC in the processing of BFs in moral beauty judgment. Although the ACC was also significantly activated by ugly faces in moral beauty judgment, the ROI results showed that the activity of beautiful faces was significantly higher than that of ugly faces. Within the left TPJ (Fig. 7E), a significant effect for task-the moral beauty judgment elicited greater activity than the moral goodness judgment (F(1, 21) = 7.55, p = 0.01, partial η 2 = 0.26)-was found. No interaction between task and face was found in this area, but there was obviously a trend for higher activity in UFs than BFs in moral beauty judgment. Within the left MFG (Fig. 7F), there was significant effect of task-the moral beauty judgment elicited greater activity than the moral goodness judgment (F(1, 21) = 10.89, p < 0.01, partial η 2 = 0.33), and significant interaction (F(1, 21) = 7.48, p = 0.01, partial η 2 = 0.25), which was mainly driven by higher activity in UFs in moral beauty judgment than BFs in moral These results demonstrated that the left TPJ and MFG might play a critical role in the processing of UFs in moral beauty judgment. Within the ROI of the left precentral gyrus/IFG (Fig. 7G), significant effects for task and face were found, with higher activity in BFs in moral goodness judgment, but no significant interaction. This result indicated the unique role of left premotor area/IFG in the processing of BFs in moral goodness judgment.

Discussion
The judgments of moral goodness and moral beauty objectively refer to the evaluation of moral behaviors or moral traits (i.e., kindness, virtue, etc.), which is generally influenced by facial attractiveness. But it is not clear whether facial attractiveness has similar or different influence on the judgments of moral goodness and moral beauty. In the current study, we investigated the neural responses to facial attractiveness under two conditions of higher order moral decision-makings: moral goodness judgment and moral beauty judgment. The same subjects were required to evaluate the level of goodness about the positive moral behavior and the level of inner beauty of the person doing that same behavior with different levels of facial attractiveness (beautiful or ugly face). During the whole experiment, subjects were not instructed to pay attention to an individual's face, but were asked to make moral judgments based on this individual's moral behavior. So, what we're talking about is whether the subjects' judgments on the moral behavior of an individual is affected by the individual's facial attractiveness, with the focus on the influence of implicit perception of facial attractiveness on higher social decisions, which is more in line with natural situations and reality in terms of decision-making and has certain ecological validity. We have four main findings. First, behavioral results indicated that the morally positive behaviors performed by shows differences between facial attractiveness in moral goodness and moral beauty judgments. Red region indicates that beautiful faces in moral goodness judgment elicited greater activity than beautiful faces in moral beauty judgment; Blue regions indicate that beautiful faces in moral beauty judgment elicited greater activity than beautiful faces in moral goodness judgment; Green region indicates that ugly faces in moral beauty judgment elicited greater activity than ugly faces in moral goodness judgment. Activation maps are shown at a wholebrain level of p < 0.05, FDR corrected for multiple comparisons Fig. 6 The image on the left A shows a conjunction effect of moral goodness and moral beauty. The right image B shows mean parameter estimates for beautiful faces and ugly faces in moral goodness and moral beauty judgments within the right inferior OFC. Error bars represent standard error of means. Activation maps are shown at a whole-brain level of p < 0.05, FDR corrected for multiple comparisons individuals with beautiful faces were more likely to be evaluated as morally good and morally beautiful than behaviors performed by individuals with ugly faces, as indexed by the rating scores but not response time. Second, neuroimaging results showed that, similar to beautiful faces in moral goodness judgment trials, beautiful faces in moral beauty judgment trials recruited an overlapping of the left ventral occipitotemporal extrastriate cortices (OTC) sensitive to the early perceptual processing of faces, including inferior occipital gyrus (IOG) and fusiform face area (FFA). Third, compared to beautiful faces during moral goodness judgment, beautiful faces during moral beauty judgment induced specific activity in the ventral medial prefrontal cortex (vmPFC) (i.e., ACC extending to mOFC) and left PCC/precuneus. The opposite comparison elicited specific activity in the left premotor area extending to the opercular part of the inferior frontal gyrus (IFG). Moreover, compared to ugly faces during moral goodness judgment, ugly faces during moral beauty judgment induced specific activity in the left temporo-parietal junction (TPJ). The opposite comparison elicited no significant cluster. Fourth, the right lateral orbitofrontal cortex (OFC) was commonly involved in the representations of moral goodness and moral beauty, but was not regulated by facial attractiveness. Our results demonstrated that the neural responses to facial attractiveness in the process of higher moral decision-makings exhibit both task-general and task-specific characteristics. Facial attractiveness contributes similarly and differently to the judgments of moral goodness and moral beauty.

Shared response to attractive faces in the judgments of moral goodness and moral beauty
In the current study, we hypothesized that the judgments of moral goodness and moral beauty would be associated with enhanced activations in regions involved in the early perceptual processing of attractive faces, given that visual features pertaining to facial attractiveness would be apprehended automatically (Hung et al. 2016;Palermo and Rhodes 2007) at a glance (Olson and Marshuetz 2005) and influence people's social judgments made about others (Chatterjee 2003(Chatterjee , 2015Chatterjee et al. 2009;Chatterjee and Vartanian 2016;Moonja Park Kim and Rosenberg 1980;Todorov et al. 2008;Winston et al. 2007). Consistent with our expectation, we found that the left ventral occipitotemporal visual cortices, including IOG and FFA, responded to the beautiful faces independent of judgment task, showing a similar magnitude of activation during both judgments of moral goodness and moral beauty.
Previous studies confirmed that the apprehension of attractive faces is associated with an identifiable neural response, specifically, attractive faces automatically activate the ventral occipitotemporal cortices, particularly sensitive to face perception, including IOG and FFA (Chatterjee et al. 2009;Kranz and Ishai 2006;Winston et al. 2007). An attractive face is a salient social signal that reflect the overall effect of all physical attributes of a face (Rhodes 2006). Regions of the occipital and posterior temporal visual cortices involved in the lower-level visual processing of perceptual features of faces play a critical role in the perceptual processing of attractive faces (Chatterjee et al. 2009). Recent fMRI studies also showed that the cognitive and neural mechanisms involved in the perceptual analyses of faces and attractive faces were considerably overlapping (Calder and Young 2005;Fisher et al. 2016;Li et al. 2019;Redfern and Benton 2017;Sun et al. 2015). The face-selective neurons in the occipitotemporal cortical visual areas are tuned to the identity of faces (Eick et al. 2020), and have representations that are invariant with respect to, for example, retinal position, size and even view. These invariant representations are ideally suited to provide the inputs to brain regions such as mOFC that is involved in social communication and emotional behaviors.
Combining our behavioral findings, that individuals with a beautiful face who perform moral behaviors are more likely to be judged as morally good and internally beautiful, compared with individuals with an ugly face who perform the same behaviors, our imaging result, that the neural activity involved in the perceptual apprehension of attractive faces was activated in both judgment tasks of moral goodness and moral beauty, demonstrates that both the judgments of moral goodness and moral beauty require the automatic visual analysis of the geometrical configuration of attractive faces of individuals performing moral behaviors, before attractive information can be transmitted to the extended emotional system involving in emotional representations. Our result is consistent with the previous findings that the occipital and posterior temporal visual cortices are responsive to the perceptual processing of attractive faces and remain present even when subjects are not attending explicitly to facial attractiveness (Chatterjee et al. 2009;Winston et al. 2007). We suggest that the perceptual processing of visual features of attractive faces is equally crucial for both forms of moral decision-making. It is in line with the view that the perceptual response to attractive faces involves a pattern of domain nonspecific regional activations (Chatterjee et al. 2009).

Different response to attractive faces in the judgments of moral goodness and moral beauty
In addition to the overlapping brain regions (IOG and FFA) activated by the beautiful faces, we also found that beautiful faces activated unique brain regions for each task, specifically, in moral goodness judgment, the left superior temporal cortex (STC), anterior temporal lobule (ATL) and premotor area extending to the opercular part of the inferior frontal gyrus (IFG) showed a significant increase in activation for beautiful faces compared to ugly faces; the left premotor area extending to the opercular part of the IFG was more activated by beautiful faces compared with the activity of beautiful faces in moral beauty judgment; whereas in moral beauty judgment, the midline cortical structures (MCS), including vmPFC/ACC/mOFC, PCC and precuneus, showed a significant increase in activation for beautiful faces. That beautiful faces in different judgment tasks activated different brain regions might imply that the representation of emotional aspects of attractive faces is task-specific, that is, different moral judgments might interpret emotional signals expressed by attractive faces in different ways.
Early perceptual processing of the attractive faces needs to be linked to many other brain structures which are mainly involved in emotional representations, to accomplish the evaluation of attractive faces. Because attractiveness has effects in later stages whereby identity and affective qualities are extracted from faces after the early structural encoding stage of face perception (Hahn and Perrett 2014). Calder and Young (2005) proposed that the invariant and variable features (geometrical configuration) of faces may be encoded by the same perceptual characterization system, followed by separation . Separate representations are constructed that make explicit information useful for recognizing facial identity or recognizing the affective qualities of faces at a step that is subsequent to perceptual processing. Recognizing actions (variant changes in facial expressions) and emotions from perceptual information in faces form the basis for understanding the intentions and feelings of others and allow us to successfully navigate our social world (McArthur and Baron 1983;Thornhill and Gangestad 1999).

Neural responses to attractive faces in moral goodness judgment task
The recognition of identity is based on the perception of aspects of facial structure that are invariant across changes in expression and other movements of the eyes and mouth, generally recruits the FFA, which is especially involved in representing the static features of faces, and the superior temporal sulcus (STS) especially involved in representing the dynamic, changeable features of faces (Bruce and Young 1986;Hasselmo et al. 1989;Haxby et al. 2000Haxby et al. , 2002Hoffman and Haxby 2000;Winston et al. 2004). STS and FFA neurons could be form detectors sensitive to particular configurations of the face surface, together, encoding a sequence of face configurations corresponding to a particular facial expression, act as a motion pattern detector (Giese and Poggio 2003;Said et al. 2011). An additional face-selective region has been identified in the anterior temporal lobule (ATL) (Jonas et al. 2016), associated with dynamic aspects of face perception (Calder et al. 2007;Pitcher et al. 2011a, b). In addition to the STC, an extended network for facial expression perception reaches the frontal operculum and premotor cortex. IFG and nearby motor areas show activity during the perception of facial expression and may have a causal role in their recognition (Pitcher et al. 2008;Pitcher, et al. 2011a, b). The STC serves as the visual input of the dynamics of the facial expression, and posterior IFG and adjacent ventral premotor cortex, which generates higher-level motor plans of expression action (Dijksterhuis and Bargh 2001;Iacoboni and Dapretto 2006;Monfardini et al. 2013;Yang et al. 2015). The recognition of facial identity is particularly well suited for the mirror neurons hypothesis, as it is known that humans mimic the facial expression of people they are interaction with and produce micro-expressions when simply looking at expressive face images (Dijksterhuis and Bargh 2001;Dimberg1982;Dimberg et al. 2000;Monfardini et al. 2013). When perceiving and decoding other individual's expression signals and inferring their underlying intentions, the observer may try to understand what the actions would mean if they themselves were to perform the same actions; the role of the observer's self is much more prominent and the understanding of others' actions is relatively self-based (Iacoboni and Dapretto 2006;Yang et al. 2015).
Our results showed that the brain networks of beautiful faces in moral goodness judgment task appeared to be consistent with the neural activity in the recognition of facial identity. The attractive face of the main character who performs a moral behavior captures the observer's attention (Chen et al. 2012;Leder et al. 2010), and the perceptual processing of attractive face might rely heavily on the observer's self. The observation and understanding of the action here should refer to the visual features of attractive faces, rather than the moral action itself, because individuals with an ugly face also perform the same moral behavior, but there are no significant brain regions activated. We noticed that in moral goodness judgment task, information about the perceptual processing of beautiful faces was not transmitted to the brain regions responsible for processing the emotional aspects of attractive faces, for example, the vmPFC, and particularly the mOFC. This result is consistent with the affectivemeaning-centered view of vmPFC, that is, vmPFC and its subcortical connections are not essential for simple forms of affects, valuation and affective learning, but are essential when conceptual information drives affective physiological and behavioral responses (Benoit et al. 2014;Roy et al. 2012). Indeed, the appreciation of attractive faces unfolds as a neurobiological process in which sensory information is projected to the reward circuit, which generates levels of pleasure (Skov and Nadal 2020). Although the perceptual representation of a face is obviously important to how it is appreciated, its positive emotional impact is not inherent to this perceptual representation. Rather, emotional values must be understood as a gloss that applied onto sensation by the mesocorticolimbic system (Ellingsen et al. 2015). We speculate that in moral goodness judgment, the observers were likely to observe and imitate the specific facial features and expressions to mediate the influence of attractive faces on the moral goodness judgment, and did not concern themselves with the emotional meanings of attractive faces.

Neural responses to attractive faces in moral beauty judgment task
The recognition of emotional meanings is derived from facial features that resemble facial expressions signaling approach/avoidance behaviors , and is generally modulated by a network of structures, including mOFC, cingulate cortex and ventral striatum (Maddock et al. 2003;Rolls 2019;Vuilleumier et al. 2001). Most brain structures that participate in the recognition of emotions or the social significance of faces involve both perceptual processing-identifying the geometric configuration of facial features to discriminate among different stimuli on the basis of their appearance-and recognition of the emotional meaning of a stimulus-knowing that a certain expression signals an emotion (Adolphs 2002). in moral beauty judgment, beautiful faces of the main characters who perform morally positive behaviors did not activate the superior temporal cortex (STC) and mirror nervous system, but significantly evoked the vmPFC (i.e., ACC extending to the mOFC) and left PCC/precuneus, which seemed to be consistent with the neural activity in the recognition of emotional aspects of attractive faces.
Work from fMRI studies has consistently shown that the vmPFC, particularly the mOFC, is relevant to the evaluation of attractive faces (Aharon et al. 2001;Bray and O'Doherty 2007;Cloutier et al. 2008;Hampshire et al. 2012;Ishizu and Zeki 2011;Kawabata and Zeki 2004;Kawasaki et al. 2001;Liang et al. 2010;Marinkovic et al. 2000;O'Doherty et al. 2003;Smith et al. 2016;Cabeza 2011a, 2011b;Winston et al. 2007;Zaki et al. 2011). The vmPFC has been shown not only to underlie subjective experiences of emotional value, but also, more specifically, to underlie valuation that incorporates conceptual information to produce "affective meaning" (Aydogan et al. 2018;Roy et al. 2012;Winecoff et al. 2013). Neurons located in vmPFC, especially the mOFC, appear to integrate information from different sources (Knutson and Genevsky 2018) and drive affective physiological and behavioral responses (Roy et al. 2012). The vmPFC also plays an important role in emotional regulation of behavior, particularly at the intersection of emotion and decision-making. Damaged to the vmPFC, results in both blunted emotional responses, demonstrated by reduced psychophysiological responses to emotional stimuli, and abnormal decision-making (Rolls and Grabenhorst 2008). The extensive connections between occipitotemporal visual regions and the prefrontal cortex (Seltzer and Pandya 1 3 1989) provide for an effective source of perceptual input to the vmPFC and also make it plausible that the vmPFC could play a role in recognizing facial emotion via direct modulation of activity in occipitotemporal cortex via feedback (Adolphs 2002). In addition, there is evidence that the recognition and experience of emotion are closely related, and in principle, consistent with the idea that the mOFC participates in emotion recognition via triggering an emotional response in the observer, and more importantly, this process depends on the context and the task (Emery and Amaral 2000). That brain regions associated with reward information respond to attractive faces suggests physical attractiveness may hold incentive salience (Berridge and Robinson 2003) and influence the observer's perception of moral beauty.
Moreover, a number of fMRI studies have found that the midline cortical structures (MCS), including ACC, PCC and precuneus, are related to the self-referential processing, and self-referential thought, related to hopes and aspirations, tends to be associated with the anterior midline regions whereas self-referential thought related to duties and obligations is associated with the posterior midline regions (Johnson et al. 2006;Nejad et al. 2013). The result that the ACC, PCC and precuneus tended to respond more strongly to attractive faces than unattractive faces in moral beauty judgment suggests that MCS activations might be involved in the higher order processing of self and positive emotionalrelated information (Quevedo et al. 2018;Sugiura 2015) and this processing is more likely to occur in moral beauty judgments than moral goodness judgment. Once again, this pattern of activations bears a strong resemblance to that associated with positive emotional expressions (Dolan et al. 1996;Kesler-West et al. 2001;Phillips et al. 1998). The highly consistent presence of these areas suggests that under certain tasks, attractive faces modulate activity in emotional meanings regions of the brain. This result is very important. For a long time, the academic has proposed that the generation of aesthetic feelings is influenced by the external image of the object, while the generation of moral sense is not affected (Zaidel and Nadal 2011), but this is only from a speculative perspective. The results of the present study not only provide empirical evidence for this important principle, but uncover this critical difference only exist in the late encoding stage of object perception.
In addition, it is important to note that we found that the common neural representations of moral goodness and moral beauty was localized in the right lateral OFC, but its activity was not affected by facial attractiveness in either of judgment tasks. This result is consistent with some findings of the OFC's role in social reward value, that lateral portions of the OFC are significantly more responsive to rewarding objects in general, while medial portions are involved in processing the emotional aspects of faces, including emotional valence and facial attractiveness (Aharon et al. 2001;Chatterjee et al. 2009;Cloutier et al. 2008;Ishai 2007;Kawabata and Zeki 2004;Liang et al. 2010;Pegors et al. 2015;Troiani et al. 2016). This result is also in line with the view that the brain requires a domain-general common currency signal, localizes in the ventrolateral PFC, when comparing and deciding between different rewards (Delgado 2007;Levy and Glimcher 2012;Peters and Buchel 2009;Sescousse et al. 2015), whereas adaptive behavior may require a domainspecific value signal generally located in the central OFC (Heinzelmann et al. 2020;Howard et al. 2015;Howard and Kahnt 2017). In the present study, subjects were asked to compare and decide between different morally positive behaviors, and they were not directed to attend to the faces, which are value signals of adaptive behaviors, of individuals who perform moral behaviors. Although some studies have found that the lateral OFC also responds to beautiful faces (left latOFC: Winston et al. 2007;right latOFC: Tsukiura and Cabeza 2011a, b), this activity for face attractiveness may only arise when participants are explicitly evaluating face attractiveness (Pegors et al. 2015). Other studies have found that the role of the lateral OFC in aesthetic preference is not limited to the representation of the reward value of stimuli regarded as beautiful, it also seems to play a role in the processing of stimuli rated as not beautiful (Munar et al. 2012).
In brief, the present study directly compared the facial attractiveness under two conditions of higher order moral decision-makings: moral judgment and moral beauty judgment and examined the commonalities and differences in the neural responses to facial attractiveness during these two processes through an experimental approach. However, there are also several limitations in this study. First, we must admit that scene drawings used as experimental materials could lower the ecological validity of the study. Although the drawings give us a high level of control, in "real life," we look at human faces in action; even looking at a still photo drops ecological validity, but has more than a cartoon drawing. Second, we only recruited female subjects to participate in the experiment. Although we were doing it to eliminate the possible confusion between the gender of the subjects and the gender of the protagonist in the experimental stimulus, we must admit that this is also a limitation of the current study. Therefore, in future studies, we should recruit male subjects to complete different types of moral judgments and further explore whether there are gender differences in the neural mechanisms of facial attraction's influence on higher order moral decision-making. This will help us determine the extent to which we hold people responsible for those positive moral actions and reactions.

Conclusion
The present study aimed to directly compare the neural responses to facial attractiveness in the judgments of moral goodness and moral beauty. With this goal, scene drawings were chosen as suitable experimental materials for they have a significant advantage of simultaneously providing information on both characters' external facial attractiveness and moral behaviors. In our experiment, the perceptual responses to attractive faces across left inferior occipital gyrus (IOG) and fusiform face area (FFA) contributed similarly to the judgments of moral goodness and moral beauty; however, the emotional responses to attractive faces might have contributed differently to these two types of judgments. Attractive faces might be perceived by observers primarily by observed and imitated particular facial expressions when they were required to evaluate the goodness level of morally positive behaviors, while attractive faces were likely to be perceived primarily by eliciting the observers' emotional response or engagement driven by relevant conceptual information when subjects were asked to assess their inner beauty, so, emotional representations of attractive faces recruited different brain regions between the two types of judgments. We suggest that the neural responses to attractive faces involves patterns of task-general and taskspecific regional activations. Our findings contribute to the understanding of the relationship between the relationship between the sense of morality and aesthetic and the perception of face.