PII: S0896-6273(03)00293-9 Neuron, Vol. 38, 525–528, May 22, 2003, Copyright 2003 by Cell Press MinireviewBeauty in the Brain of the Beholder groups that are unique to human culture has ensured that culturally specific characteristics are also transmit- ted with facial beauty. Secondary sexual indicators on Carl Senior* Laboratory of Brain and Cognition National Institute of Mental Health male faces do not suggest that the bearer will placeBethesda, Maryland 20892 considerable investment in parental input. Indeed, men with symmetrical faces are less likely to accept fidelity and will invest less time with offspring. When consider-Facial beauty is an honest signal of the genotypic and ing paternity, men with facial characteristics that signalphenotypic quality of the bearer. Beautiful people are a limited opportunity to attract multiple mates would bethus regarded as high-value mates who maximize re- considered to contribute more to reproductive successproductive success by producing viable offspring. than those advertising an ability to attract more thanHere, the functional neuroanatomy of facial beauty is one partner (Penton-Voak et al., 1999). Interestingly,reviewed and placed into the context of the distributed men with neonatal facial characteristics, which advertisemodel for human face perception. A proposed exten- a limited opportunity to attract potential partners, tendsion of the distributed model is provided, which takes to have positive personality traits attributed to them byinto account the neuroanatomy of beautiful face per- female observers that may correlate with subsequentception. behavior. However, these personality traits are not indic- ative of actual behavior and may be fallible. To thisBackground end, female observers may consider male faces to beEvolutionary theory suggests that facial beauty signals attractive if their same-sex peers consider them to bethe optimum reproductive status of the bearer; it thus attractive as well. Thus, by copying the mate choice ofdictates that we are drawn to those who are attractive in same-sex peers, female observers act to increase theorder to maximize our success in reproducing offspring reliability of any assessments of facial attractivenesswith a strong chance of survival (references in Symons, (references in Thornhill and Gangestad, 1999).1979). Take, for example, the effect of estrogen on fe- Clearly, the ethology of beautiful face perception ismale faces; estrogen markers, such as fat deposits in complex, and considering the importance of selectingthe upper cheek and lip area, signal fertility and readi- high-value mates, the distinction needs to be made be-ness for reproductive effort (Perrett et al., 1998). While tween beautiful faces that are rewarding in the sensethese markers are considered attractive by male observ- that they have an adaptive value (i.e., attractive different-ers, the estrogen to testosterone ratio fluctuates with sex faces) and those that are merely aesthetic (i.e., at-age such that female faces “masculinize” as the bearer tractive same-sex faces). An attractive female face maygets older. Estrogen thus provides female faces with a be aesthetic but may also signal good genes that willmarker of youth—a marker that signals the optimal fertil- positively correlate with the genetic strength in any off-ity period for human females. Additionally, in readying spring. However, to male observers, attractive malethe body for reproductive effort, estrogen diverts re- faces cannot constitute such a sociobiological adver- sources away from systemic functions such as cellular tisement and can only be considered aesthetic. Consid- repair mechanisms and immunocompetency. Facial es- ering the adaptive benefit behind perception of differ- trogen markers could thus be considered an honest ent-sex facial beauty and the subsequent motivated signal of the genotypic quality of the bearer, i.e., only behavior that occurs in observers assessing it, the brain those individuals strong enough to promote such a dis- areas implicated in reward function that underpin appe- play can afford to handicap vital functions. The same titive behavior are likely to be active with their percep- principle can be applied to male faces, where testoster- tion. However, considering the sole aesthetic value in one, which diverts resources away from immune func- attractive faces of members of the same sex, natural tion and toward the development of sexually dimorphic selection would have shaped any “beauty detectors” in facial traits, e.g., larger eyebrow ridges, contributes to the brain to become specialized for same- or different- the advertisement of genotypic quality (Grossman, sex faces, but not both. While neuroimaging studies of 1985). Such overt displays of systemic handicap allow facial beauty are scant, recent investigations do suggest observers to formulate judgments of the relative good sex-differentiated responses for perception of beautiful health of the bearer. A further facial marker considered faces, these studies are discussed later (Aharon et al., attractive and indicative of the genetic superiority of 2001; O’Doherty et al., 2003). However, our current un- individuals from either sex is “fluctuating asymmetry”— derstanding of the cognitive neuroanatomy of face per- the extent to which the left half of the face is the same ception does not account for these findings. To better as the right. The optimum development of facial symme- understand facial beauty and the brain, a brief discus- try is impeded by environmental stressors in utero, thus sion of the neuroanatomy of face perception is needed. the more symmetrical a face is the less that person has Neuroanatomy of Face Perception been exposed to developmental stress (Thornhill and The distributed human neural system for face perception Moller, 1997). (distributed face model) is the most contemporary neu- However, the evolution of highly complex social roanatomical model encompassing what is known of face processing to date (Haxby et al., 2000). This model is primarily based on functional neuroimaging data and*Correspondence: carl@codon.nih.gov Neuron 526 Figure 1. Proposed Extension of the Distributed Human Neural System for Face Perception The solid lines represent parts of the existing model, while the broken lines represent the proposed extension. Initial appraisal of facial beauty takes places in the SLEA and the ventral tegmentum (VT), where activity is reflected with a positive signal response in the orbitofrontal gyrus (GOb), nucleus accumbens (NAc), and ventromedial prefrontal cortex (VPFC) for rewarding beauty only. However, deactivity in the SLEA results in concomitant deactivitation in the NAc during perception of aesthetic beauty. parses out the various processes involved in face per- rior insula. One area described in the distributed face model, the amygdala, has rich connections to manyception into two main systems. The first of these is the core system that is involved in visual analysis of the other areas implicated in reward and is thus pertinent to our discussion of the cognitive neuroanatomy of facialface. The second is the extended system, which projects out from the core system and computes the various beauty. Classically, the amygdala is thought to process threat-types of information gleaned from the face. Inside both the core and extended systems there is further func- related facial expressions (e.g., fear and anger); how- ever, it also plays a part in mediation of other expres-tional modularity. Within the core system, the visual analysis of the variant aspects of the face (e.g., eye sions and social cognition in general (Adolphs, 2001). Recent studies of the amygdala suggest that it may alsogaze, expression, and lip movement) are computed in parallel to the invariant aspects of the face, such as the process meaning from the face that is unrelated to the facial expression being used as stimuli. This epiphenom-perception of identity. The functional subdivision of the extended system is based on the computation of spatial enal activity may be indicative of how the observer feels. Take, for example, the now robust finding that directattention, prelexical speech perception, emotion, and unique biographical information. Each of the subdivi- eye gaze engenders activity in the observer’s amygdala. While the ability to correctly infer the direction of gaze issions within the extended and core systems compute their respective functions in parallel; however, there is a fundamental human trait, it does, quite paradoxically, have ambiguous meaning. Direct gaze can signal poten-reciprocal feedback between most of the units in the model (see Figure 1). tial threat but it can also indicate interpersonal attraction (Haxby et al., 2002). The fact that presentation of facialUnlike previous cognitive models of face processing, the distributed face model has neurological specificity, expressions of happiness (which are considered to be more attractive than other expressions) and looking atwith each unit in the model mapped to a distinct sub- strate. The core system consists of the inferior occipital people who you love evokes activity in the amygdala suggests that this area of the brain may play a crucialgyri, which mediate perception of early visual features of the face. The inferior occipital gyri project into the role in beautiful face perception (Bartels and Zeki, 2000). The Extended Amygdala and Reward Systemsuperior temporal sulcus (STS) and lateral fusiform gy- rus, which computes variant (expression) and invariant The amygdala is an umbrella term for a functionally and anatomically heterogeneous collection of nuclei that re-(identity) facial aspects, respectively. The lateral fusi- form then projects forward along the anterior temporal side in the superiomedial aspect of the anterior temporal lobes. The dense reciprocal interconnections that con-cortex, where personal semantic information regarding the face is represented, thus instigating familiarity. In vey information to and from a wide variety of neural areas ensure that the amygdala (or amygdaloid complex) isparallel to this, the superior temporal sulcus has recipro- cal projections to the intrapariatal sulcus, where spatial well situated to play an important role in the computa- tions that underpin the perception of beautiful faces. Theattention from the face is processed. The ability to shift spatial attention from facial cues, such as interpretation various nuclei that constitute the amygdala are defined according to the particular functional criteria adopted,of gaze direction, is a fundamental skill allowing suc- cessful interaction in complex social groups. The supe- but it is generally agreed that there are two main clusters, these being the older and smaller corticomedial grouprior temporal sulcus also works in concert with regions within the superior temporal gyrus that are responsive to and the basolateral group (references in Aggleton, 2000). The sublenticular (below the lentiform body) extendedmovements unique to speech. Finally, there is reciprocal connectivity between the superior temporal sulcus and amygdala of the basal forebrain (SLEA) projects from the corticomedial group into the stria terminalis to formregions of the brain involved in emotion processing, such as the amygdala, orbitofrontal gyrus, and the ante- an extension of the “core amygdala.” The stria terminalis Minireview 527 project into the caudate body and to the nucleus accum- While male observers judged both beautiful male and female faces to be attractive, they only found the beauti-bens. Due to the relay nature of the amygdala, the nu- cleus accumbens thus enjoys reciprocal connectivity ful female faces to be rewarding to look at and rated them as more attractive with multiple presentations.between the insular cortex, the hippocampal formation, the medial prefrontal cortex, as well as the orbitofrontal Moreover, even though beautiful male and female faces were rated to be equally attractive, the male observersgyrus (Heimer et al., 1997). Further evidence showing projections from the corticomedial group to the dopa- only expended effort (via multiple key presses) to gaze longer at the beautiful female faces, showing that onlyminergic brain reward areas does suggest that it may play an initial role in perception of beautiful faces, with these faces were rewarding. To investigate the neural mechanisms involved in per-the brain reward system being “activated” at subse- quent stages (Fudge and Haber, 2000). ception of rewarding and aesthetic facial beauty, a third cohort of male observers were presented with picturesThe areas of the brain that mediate reward are diverse but are linked together through the medial forebrain of male and female faces (both average and attractive) in a block design imaging experiment. To ensure that thebundle—a complex structure consisting of approxi- mately 60 fiber bundles that project through the lateral fMRI results were not contaminated by the attentional modulation of gaze, all material was presented to thehypothalamus toward the rostral basal forebrain. The ventral tegmentum is a dopaminergic site that is con- subjects tachistoscopically (200 ms), and each picture was separated with an interstimulus period of 3800 ms.nected by a descending component of the medial fore- brain bundle. The ventral tegmentum also projects to Several regions of interest were first defined by com- paring engendered activity in all experimental condi-other areas of the brain, such as the nucleus accumbens and the orbitofrontal gyrus, which are now seen to play tions with any activity revealed within the control period (which only consisted of a fixation cross). This analysisa role in the brain reward process (Rolls, 2000). These systems would have evolved to ensure the appetitive served to localize the regions of interest that were sub- sequently studied for the specific experimental effects.drive toward stimuli that were beneficial to our survival as a species. In light of the adaptive benefits behind Presenting male subjects with beautiful female faces (which would involve both reward and aesthetics) en-beautiful faces, one would hypothesize involvement of these substrates. Recently, functional magnetic reso- gendered significant activity in both the right orbitofron- tal gyrus and bilateral nucleus accumbens. This analysisnance imaging (fMRI) was applied in an effort to under- stand more about the processes mediating perception also revealed activity, albeit at a lower threshold, in the ventral tegmentum and SLEA. The brain areas revealedof beauty (Aharon et al., 2001; O’Doherty et al., 2003). The results of these investigations place facial beauty with the analysis of the rewarding beautiful faces repre- sent a subcortical distributed system, with each compo-alongside emotion and speech as a fundamental condi- tion that can be read from the face and also suggest that nent contributing a specific role to perception of facial beauty.aesthetic and rewarding properties of beautiful faces are processed separately. Lesions to the sublenticular nucleus reduce the re- warding effect of stimulation in the medial forebrain bun-Neuroimaging Studies of Facial Beauty In the Aharon et al. (2001) study, a behavioral task was dle (Arvanitogiannis et al., 1996). The SLEA and the ven- tral tegmentum share a common pathway through thefirst given to a group of heterosexual male observers who where asked to rate the attractiveness of average medial forebrain bundle, and activity has been revealed in the dopaminergic neurons of the ventral tegmentumand beautiful male and female faces. This task revealed a clear difference between the beautiful and average to salient sensory events which have no relationship to reward (Horvitz, 2000). It may be the case that the SLEAsets, regardless of the sex of the face. However, this difference was modulated by the amount of time that and the ventral tegmentum operate together to form an “appraisal unit” that is the first, and common stage, inthe observer saw faces belonging to a particular group. The beautiful female faces were rated as more attractive perception of beautiful faces. On the other hand, the orbitofrontal gyrus is specifically active when the subjectwith each presentation, while the average female faces were rated as less attractive. On the other hand, the is anticipating a reward outcome that could contain ei- ther a positive or negative valence (Schultz and Dicken-beautiful and average male faces did not reveal such a separation of ratings with subsequent presentations. A son, 2000). The dopaminergic meso-accumbens path- way plays a central role in the initiation of motivatedsecond cohort of male observers underwent another behavioral task where they could manipulate the length behavior, with the nucleus accumbens involved in the assessment of reward expectancy (Breiter and Rosen,of time that a picture of a face remained on the screen. The duration of presentation could be extended or short- 1999). Thus, these areas of the heterosexual male brain encompass the appraisal, anticipation, and assessmentened when the subject pressed a specific key on a key- board. It was hypothesized that male observers would of the reward value inherent in beautiful female faces. Aharon et al. (2001) then investigated the effects ofmake multiple keypresses to gaze longer at those faces that were rewarding (i.e., the attractive female faces). the aesthetics of facial beauty only (i.e., beautiful male faces versus average male faces). As before, the ventralPerhaps unsurprisingly, the results confirmed this hy- pothesis, with male observers choosing to gaze longer tegmentum was active; however, the SLEA revealed a negative signal response (deactivation), suggesting thatat the beautiful female faces. There was no difference between the length of time the observers chose to look these areas may respond together in an additive manner for aesthetics in rewarding faces only. In addition toat the average female and male faces and the attractive male faces. These results converge on a behavioral dis- this, a bilateral deactivation was revealed in the nucleus accumbens. The pattern of activity revealed in the nu-sociation of the qualities that constitute facial beauty. Neuron 528 cleus accumbens for aesthetic beauty is supported by proposed module for the rewarding beautiful faces also studies of reward expectancy. For example, a decrease represents activity in the orbitofrontal gyrus and ventro- in the fMRI signal from baseline is revealed when an medial prefrontal cortex, as activity here was specific expected reward is not delivered (Knutson et al., 2001). to this set of faces only (see Figure 1). It would seem likely that the bilateral deactivation in the One caveat has to be stressed: as noted above, few nucleus accumbens for the Aharon et al. (2001) study neuroimaging studies on facial beauty exist, and as signaled an absence of reward expectancy toward the such, the proposed extensions of the distributed face beautiful male faces. model will have to remain provisional until the necessary In a later fMRI (event related) experiment, the cortical empirical investigations confirming their robust validity response to male and female beautiful faces for observ- are carried out. ers of both sexes was studied (O’Doherty et al., 2003). Selected ReadingIn this study, no activity in the nucleus accumbens was revealed when male and female subjects were shown Adolphs, R. (2001). Curr. Opin. Neurobiol. 11, 231–239. attractive and average faces of both sexes. However, Aggleton, J.P. (2000). The Amygdala: A Functional Analysis (Oxford:as Aharon et al. (2001) showed a positive and negative Oxford University Press). signal change for attractive female or male faces, the Aharon, I., Etcoff, N., Ariely, D., Chabris, C.F., O’Conner, E., and presence of activity in the nucleus accumbens cannot Breiter, H.C. (2001). Neuron 32, 537–551. be considered an artifact brought about with the use of Arvanitogiannis, A., Waraczynski, M., and Shizgal, P. (1996). Physiol. a block experimental design. It remains to be tested Behav. 59, 795–806. whether or not deactivation in the nucleus accumbens Bartels, A., and Zeki, S. (2000). Neuroreport 11, 3829–3834. when observers are shown attractive same-sex faces is Breiter, H.C., and Rosen, B.R. (1999). Ann. N Y Acad. Sci. 877, unique to the heterosexual male population. Interest- 523–547. ingly, this study also revealed activity within the ventro- Fudge, J.L., and Haber, S.N. (2000). Neuroscience 97, 479–494. medial prefrontal cortex for heterosexual male observ- Grossman, C.J. (1985). Science 227, 257–261. ers viewing attractive female faces. Given the role that Haxby, J.V., Hoffman, E.A., and Gobbini, I. (2000). Trends Cogn. Sci. the ventromedial prefrontal cortex may play in social 4, 223–233. reasoning, it seems likely that it works in concert with Haxby, J.V., Hoffman, E.A., and Gobbini, I. (2002). Biol. Psychiatry the orbitofrontal gyrus to compute reward expectancy in 51, 59–67. a possible social context (references in Adolphs, 2001). Heimer, L., Alheid, G.F., de Olmos, J., Groenewegen, H.J., Haber, Even so, the pattern of activity revealed for the analysis S.N., Harlan, R.E., and Zahm, D.S. (1997). J. Neuropsychiatry Clin. of aesthetic and rewarding beautiful faces is not ac- Neurosci. 9, 354–381. counted for in the distributed face model; modifications Horvitz, J.C. (2000). Neuroscience 96, 651–656. of the extended system of this model are proposed to Knutson, B., Adams, C.M., Fong, G.W., and Hommer, D. (2001). J. account for these findings. Neurosci. 21, 159. Beauty in the Brain O’Doherty, J., Winston, J., Critchley, H., Perrett, D., Burt, D.M., and Dolan, R.J. (2003). Neuropsychologia 41, 147–155.The data to hand suggest that the SLEA and the ventral tegmentum serve a pivotal role in the initial appraisal Penton-Voak, I.S., Perrett, D.I., Castles, D.L., Kobayshi, T., Burt, D.M., Murray, L.K., and Minamisawa, R. (1999). Nature 24, 741–742.of beautiful faces. As described above, the amygdala Perrett, D.I., Lee, K.J., Penton-Voak, I., Rowland, D., Yoshikawa, S.,receives input from the superior temporal sulcus. How- Burt, D.M., Henzi, S.P., Castles, D.L., and Akamatsu, S. (1998). Na-ever, as the computations carried out by the SLEA and ture 394, 826–827.ventral tegmentum, and the subsequent projections to Rolls, E.T. (2000). Cereb. Cortex 10, 284–294.the nucleus accumbens, are distinct from emotion pro- Schultz, W., and Dickenson, A. (2000). Annu. Rev. Neurosci. 23,cessing per se, they are represented alongside the 473–500.“emotion module” in the distributed face model and not Symons, D. (1979). The Evolution of Human Sexuality (Oxford: Ox-projected out from within. The projection from the STS ford University Press).to the SLEA and ventral tegmentum is supported by Thornhill, R., and Gangestad, S.W. (1999). Trends Cogn. Sci. 3,work showing modulation of the brain reward system by 452–460.eye gaze, a manipulation previously found to engender Thornhill, R., and Moller, A.P. (1997). Biol. Rev. Camb. Philos. Soc.activity in the anterior bank of the STS (Haxby et al., 72, 497–528. 2002). When subjects are presented with beautiful faces that are rewarding, engendered activity in the sublenticular nucleus is mirrored with activity in the ventral tegmen- tum, orbitofrontal gyrus, and nucleus accumbens. How- ever, beautiful faces that are merely aesthestic engender a negative signal change (deactivation) in the SLEA and the nucleus accumbens. Due to the opposite directions of signal response in the nucleus accumbens for percep- tion of aesthetic and rewarding beautiful faces, these computations are represented separately. Therefore, projections from the module representing the SLEA/ven- tral tegmentum diverge into two separate pathways— the first represents the processing of rewarding beautiful faces, and the second represents the aesthetic. The