Attractiveness Learning from Partially Occluded Faces

Attractiveness Learning from Partially Occluded Faces | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Attractiveness Learning from Partially Occluded Faces Natsumi Kubo, Hakudai Makino, Atsunori Ariga This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6903928/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Facial attractiveness tends to be overestimated when it is partially occluded by a sanitary mask ( mask bias ). However, it remains unclear why people have not developed a cognitive strategy to accurately judge the attractiveness of partially occluded faces, despite repeatedly encountering this bias in daily life. Are we simply unable to learn such a strategy? This study examined whether people can learn to adjust their attractiveness judgments of mask faces by receiving repeated feedback on the “true” attractiveness (the uncovered face along with a rating previously assigned by a large majority). Through one week, participants were exposed daily to mask faces followed by feedback. Interestingly, female participants learned to adjust their judgments more quickly than male participants. These findings suggest that people can, to some extent, acquire a strategy for judging facial attractiveness with massed learning. Otherwise, the bias is likely to persist, as it does in everyday settings. Face perception Attractiveness Learning Figures Figure 1 Figure 2 Introduction Even when the whole face is not visible, people can make relatively accurate judgments of facial attractiveness based on limited information (e.g., Miyazaki & Kawahara, 2016 ; Rule et al., 2009 ). Interestingly, judgments of partially occluded faces—such as those seen from behind or covered by a mask—do not fluctuate randomly around the actual attractiveness of the whole face. Instead, they tend to be consistently overestimated, a phenomenon referred to as back-view bias or mask bias (e.g., Hies & Lewis, 2022 ; Ichimura et al., 2021 ; Patel et al., 2020 ; Sugai et al., 2024 ; Yonemura et al., 2013 ; but see Miyazaki & Kawahara, 2016 ). In other words, judgments of attractiveness for partial faces are not merely predictions, but rather appear to reflect a systematic positive bias. In fact, this kind of overestimation is commonly experienced in everyday life (Sudhakar, 2023). Several explanations have been proposed for why people tend to overestimate the attractiveness of a partially occluded face. These include completion based on a positive mental representation of an average or prototypical face (e.g., Kramer & Jones, 2022 ; Orghian & Hidalgo, 2020 ), the occlusion of negative features (e.g., poor symmetry, rough skin, or acne; Kamatani et al., 2021 ; Miyazaki & Kawahara, 2016 ), and strategies related to mate selection (Ichimura et al., 2021 ; Sugai et al., 2024 ). Noting that this bias is more pronounced when men evaluate women’s attractiveness, Ariga and colleagues (Ichimura et al., 2021 ; Sugai et al., 2024 ) viewed the overestimation of attractiveness as a mate selection strategy that increases the likelihood of approaching a partner, particularly for men, who have lower reproductive costs compared to women. These accounts are not mutually exclusive. However, a key question remains: Why does our daily experience not lead to improved accuracy in judging the attractiveness of masked faces? In other words, can we acquire a cognitive strategy for evaluating partially occluded faces by encountering prediction errors—instances where our judgment based on limited information differs from the actual attractiveness of the whole face? As mentioned above, such prediction errors should have occurred repeatedly in everyday life, especially during and after the COVID-19 pandemic, when people were frequently exposed to faces both wearing and unwearing masks. Nevertheless, consistent and robust overestimation has been reported both in laboratory and real-world settings (Sudhakar, 2023). In fact, our ability to recognize masked faces has not improved since the onset of the COVID-19 pandemic, even though we have been repeatedly exposed to both masked and unmasked faces (Freud et al., 2022 ). The present study is the first to address this issue by exploring the learning process involved in attractiveness judgments of partially occluded faces. Specifically, it investigates whether individuals can improve the accuracy of such judgments when provided with short-term feedback (i.e., the “true” attractiveness of the whole face). Preliminary Experiment This preliminary experiment aimed to obtain baseline attractiveness ratings for facial stimuli from a large sample of participants. These ratings were used as the “true” attractiveness scores to provide feedback in the subsequent learning experiment. Ethics Statement All procedures in this study were approved by the Research Ethics Committee of the author’s university (Approval No. 2024-077) and conducted in accordance with the Declaration of Helsinki. Informed consent was obtained from all participants prior to their participation. Method Participants. A total of 1,621 Japanese participants were recruited for the online preliminary experiment. Data from 246 participants were excluded for failing a satisficer, resulting in a final sample of 1,375 participants (674 females; M age = 32.69, SD age = 5.31) used in the analysis. Apparatus and Stimuli. As the preliminary experiment was conducted online, participants used their own devices (PC or tablet). The stimuli consisted of 80 shoulder-up, color headshots of Japanese student models (40 females, aged 18–23) not wearing masks. To maintain a natural appearance, we did not control hairstyles and makeup. However, participants were instructed to maintain a neutral expression and to remove accessories and glasses as much as possible. Photographs were standardized for distance, angle, and luminosity. Each image subtended 434 × 587 pixels. Procedure. Considering the online experiment, the 80 stimulus images were divided into four sets (each including 10 male and 10 female models), and participants were randomly assigned to one of the sets (315–363 participants per set). In each trial, a photograph was randomly selected from the assigned stimulus set. Participants rated the attractiveness of each face on a 7-point scale (1 = not attractive at all, 7 = very attractive). Results The mean attractiveness rating was calculated for each stimulus image separately for male and female participants. Because ratings from male and female participants were highly correlated ( r = .91, p < .001), we collapsed across participant gender and averaged the ratings for each image. These averaged scores were used as the “true” attractiveness values in the subsequent learning experiment (male stimuli: M = 3.02, SD = 0.32; female stimuli: M = 3.62, SD = 0.42). Experiment Method Participants. Twenty Japanese participants were recruited for the experiment. Due to technical or personal issues, four participants did not complete the learning phase. As a result, data from 16 participants (8 females; M age = 21.50, SD age = 1.06) were included in the final analysis. Apparatus and Stimuli. Participants used their own PCs during the learning phase, and a laboratory PC with a 23.8-inch monitor during the test phase. The experiment was controlled using MATLAB (The MathWorks, Inc.). In addition to the 80 no-mask images used in the preliminary experiment, we included 80 photographs of the same individuals wearing a white sanitary mask, all using the same product (referred to as “mask images”). From these 80 models, 60 (30 female) were pseudo-randomly selected as learning stimuli, and the remaining 20 (10 female) were used as test stimuli. Stimulus assignments varied across participants. The selection was made under the constraint that the mean attractiveness scores obtained in the preliminary experiment were balanced between the learning and test sets and across participants. Procedure. This experiment consisted of two phases: a learning phase and a test phase. In the learning phase, each participant was presented with a mask image randomly selected from the learning stimuli and asked to evaluate its facial attractiveness by predicting the appearance of the face without a mask. Participants rated the attractiveness on a visual analog scale from 1 (not attractive at all) to 7 (very attractive). After the participant clicked on the scale, the image of the same individual without a mask appeared, along with a “true” attractiveness rating that had been measured in the preliminary experiment, serving as feedback (Fig. 1 ). Participants were informed in advance that the feedback score was made by a large majority and were instructed to adjust their own ratings to align with it. They completed 60 learning trials per day for one week (Days 1 to 7) at home, at the same time each day. The order of stimuli was randomized across days and participants, and each stimulus appeared once per day. Following the learning phase, participants took part in the test phase in a laboratory setting. They rated the attractiveness of mask faces from a separate set of test stimuli (not included in the learning phase), using the same 1-to-7 visual analog scale. No feedback was provided during the test phase. The test phase was conducted twice: once immediately after the learning phase (Test 1: Day 8), and once approximately a week later (Test 2: Day 14 or 16, depending on the participant). Results and Discussion The attractiveness rating given by each participant for a masked image was subtracted from the "true" attractiveness rating of the corresponding unmasked image, as determined by the majority in the preliminary experiment. This difference was used as an index of mask bias. This difference was used as an index of mask bias. The mean mask bias was calculated for each participant by stimulus gender and day, and then averaged across participants. Figures 2a (female participants) and 2b (male participants) show the mask bias for each gender of the stimulus models as a function of day (learning and test phases). Because a positive mask bias and its gender difference were predicted by the previous study (Sugai et al., 2024 ), we conducted one-tailed one-sample t -tests (vs. 0) on mask bias values for each day and stimulus gender, separately for each participant gender. Female Participants. Because one participant did not attend Test 2 due to a personal issue, her data were included only in the analyses of the learning phase and Test 1. Female participants showed a significant mask bias for male stimuli only on Day 1, t (7) = 2.828, p = .013, Cohen’s d = 1.000, but no significant bias on subsequent days or for female stimuli. This suggests that learning was completed at this early stage. Interestingly, no significant bias was observed in either of the test phases (Tests 1 and 2), indicating not only that the learned strategy for judging attractiveness based on partially occluded faces transferred to novel stimuli (Test 1), but also that it was retained for approximately one week (Test 2). Male Participants. Because one participant did not attend the test phase due to a personal issue, his data were included only in the analyses of the learning phase. Male participants exhibited a significant mask bias for male stimuli on Days 1, 5, and 7, ts (7) > 2.251, ps 0.796, but not for female stimuli. This suggests that learning was not fully completed for male stimuli within the one-week learning period. Interestingly, a significant mask bias for male stimuli was observed in Test 2 of the test phase, t (6) = 2.166, p = .037, Cohen’s d = 0.819. Baseline Measurement of the Test Stimuli. Bias baselines were measured using naïve participants who took part only in the test phase and did not undergo the learning phase. These participants rated the masked faces using the same 7-point scale as in the preliminary experiment. For female naïve participants ( N = 40, M age = 20.15, SD age = 1.42), baseline bias was significantly greater than zero for both male (0.46) and female (0.64) stimuli, ts (39) > 3.863, ps 0.611. Furthermore, the mask biases shown by female participants who had completed the learning phase were significantly lower than these baseline values, regardless of stimulus gender: ts (7) > 5.836, ps 2.063 on Test 1; ts (6) > 5.159, ps 1.950 on Test 2. In addition, the mask bias for female stimuli on Day 1 was significantly smaller than the bias baseline, t (7) = 5.543, p < .001, Cohen’s d = 1.960, and was not significantly different from zero, suggesting that learning for female stimuli was likely completed within the 60 trials on Day 1. As for male naïve participants ( N = 20, M age = 20.30, SD age = 2.00), the baseline bias was also significantly greater than zero for both male (0.70) and female (0.61) stimuli, ts (19) > 3.322, ps 0.743. Similarly, the test-phase mask biases shown by male participants who underwent the learning phase were significantly lower than baseline values, regardless of stimulus gender, ts (6) > 4.281, ps 1.618. Considering the learning-phase results, male participants appeared to have partially acquired the strategy of judging attractiveness based on partially occluded faces, though learning was incomplete. This partial learning transferred to novel stimuli (Test 1) and persisted for about one week (Test 2). On Day 1, the mask bias for female stimuli was significantly smaller than the baseline, ts (7) > 6.774, ps 2.395, and while bias was not significant (vs. 0) for female stimuli, it remained significant for male stimuli. These findings suggest that learning was likely completed rapidly for female stimuli and partially progressed for male stimuli during the 60 trials on Day 1. Summary of the Results. The central question of the present study was whether individuals can learn a strategy for judging facial attractiveness based on partially occluded faces through repeated exposure to prediction errors—specifically, overestimation. The findings suggest the answer is yes. However, the learning process was influenced by the gender relationship between participants and stimulus models. First, the baseline measurements in the test phase revealed that both female and male participants initially exhibited a mask bias for both male and female stimuli. Second, female participants rapidly completed the learning process for female faces on Day 1 and for male faces by Day 2. Third, male participants, like their female counterparts, quickly learned to judge the attractiveness of female faces on Day 1 but showed only partial progress in learning for male faces, which was not completed within the 7-day period. Fourth, the learned strategy for judging attractiveness based on partially occluded faces was successfully transferred to novel stimuli, both immediately after and one week following the learning phase. Taken together, these results demonstrate two key findings: (1) female participants learned to accurately estimate facial attractiveness through repeated prediction error experiences more effectively than male participants, and (2) the attractiveness of female faces was judged more accurately and earlier after learning compared to male faces. The former finding aligns with previous research suggesting that females are generally more accurate in face perception than males (Lewin & Herlitz, 2002 ; Rehnman & Herlitz, 2007 ). As for the latter, we speculate that female faces may be processed with higher motivation due to their higher average attractiveness compared to male faces (3.62 vs. 3.02 for no-mask faces, as measured in the preliminary experiment). Based on these findings, it is plausible that our daily experience of mask bias reflects a reduced bias due to naturalistic learning over time. Alternatively, it may persist due to a lack of such learning—given that, in everyday life, we rarely encounter prediction errors across 60 individuals per day or view the same individuals repeatedly over seven consecutive days. In either case, the present study provides new evidence that individuals can learn to judge facial attractiveness based on partial information to some extent through a one-week, massed learning protocol. Otherwise, the bias observed in daily life is likely to remain uncorrected. Gender Differences in the Learning Process. Previous research by Ariga and colleagues has shown that overestimation in attractiveness judgments for invisible faces varies by participant gender (Ichimura et al., 2021 ; Sugai et al., 2024 ). Specifically, male participants tended to overestimate female facial attractiveness more than female participants did for male faces. This overestimation was reduced when male participants were explicitly instructed to evaluate female faces as if they were judging friends rather than potential romantic partners. Ariga and colleagues interpreted these findings from the perspective of reproductive cost. In general, males have lower reproductive costs than females (Bateman, 1948 ), leading them to pursue greater reproductive opportunities while minimizing Type II errors, according to error management theory (Haselton & Buss, 2000 ). This tendency may drive the overestimation of female facial attractiveness. In contrast, females bear higher reproductive costs, prompting more cautious judgments and a tendency to avoid Type I errors, where overestimation may be meaningless or even detrimental. In the present study, the finding that female participants more effectively reduced overestimation when provided with feedback may reflect an adaptive strategy aligned with mate-choice processes. However, it is notable that, contrary to previous findings (Sugai et al., 2024 ), male participants did not show increased mask bias when evaluating female faces. Instead, they exhibited a relatively persistent mask bias, particularly for male stimuli. These unexpected results may stem from task demands. In the current study, participants were asked to align their attractiveness judgments with scores provided by the majority, while in the previous study, judgments were made based on personal preference. Furthermore, according to the reproductive cost account (Ichimura et al., 2021 ; Sugai et al., 2024 ), males may be less motivated to accurately judge female facial attractiveness in everyday contexts, which could explain the incomplete learning observed in the male participants in this study. What Did They Learn? One might argue that the participants in this study adjusted their rating scores a posteriori , after making an initial attractiveness judgment. In this interpretation, they would have learned how much to modify their ratings to approximate the “true” attractiveness scores, calculated retrospectively. However, this explanation seems unlikely. If participants had simply employed a calculation strategy, both male and female participants should have demonstrated similar learning processes. Yet, the results revealed gender differences in learning. Moreover, the reduction in mask bias observed during the learning phase persisted into the test phase with novel stimuli, even after a week. This persistence makes it unlikely that participants merely memorized the faces or their associated “true” attractiveness scores. Furthermore, given the difficulty of recognizing masked faces (Freud et al., 2022 ), it is improbable that participants relied on simple memorization. It is more plausible that participants acquired some form of cognitive strategy for judging attractiveness based on partially occluded faces. One possibility is that they learned to infer the attractiveness of a whole face using only limited, visible facial features. A second possibility is that they developed the ability to complete the occluded facial area through amodal completion. A third possibility is that they learned to evaluate attractiveness based solely on the visible facial parts while ignoring the occluded regions. Indeed, previous research has shown that the eyes significantly contribute to overall facial attractiveness ratings (Saegusa & Watanabe, 2016 ). Since the current data do not allow us to distinguish between these possibilities, future studies are necessary to investigate this issue. Interestingly, previous research has shown that adapting to distorted faces can shift the criteria for what is perceived as normal and attractive to align with the distortion (Rhodes et al., 2003 ). If a similar adaptation had occurred for mask faces during the current learning phase, the mask bias would have persisted or even increased. However, this was not the case. Thus, it is notable that participants learned a strategy to adjust their attractiveness judgments of masked faces despite potential adaptation effects. In sum, what is clear for now is that people are capable of learning a cognitive strategy to estimate facial attractiveness with some degree of accuracy—regardless of whether such learning occurs in everyday life. Declarations Conflict of interest The authors declare that they have no conflict of interest. Funding statement The present study was supported by a Chuo University Grant for Special Research to AA. Consent for Publication All participants obtained written informed consent before the experiment. Author Contribution All authors planned this study. N.K. and H.M. conducted the experiments. N.K. analyzed the data and wrote the draft of the paper. All authors reviewed the manuscript. This study was conducted under A.A. supervision. Data Availability All the data have been uploaded and are available in the OSF repository: https://osf.io/g78nd/ References Bateman, A. J. (1948). Intra-sexual selection in drosophila. Heredity , 2 (3), 349–368. Freud, E., Di Giammarino, D., Stajduhar, A., Rosenbaum, R. S., Avidan, G., & Ganel, T. (2022). Recognition of masked faces in the era of the pandemic: No improvement despite extensive natural exposure. Psychological Science , 33 (10), 1635–1650. Haselton, M. G., & Buss, D. M. (2000). Error management theory: a new perspective on biases in cross-sex mind reading. Journal of personality and social psychology , 78 (1), 81. Hies, O., & Lewis, M. B. (2022). Beyond the beauty of occlusion: Medical masks increase facial attractiveness more than other face coverings. Cognitive Research: Principles and Implications , 7 , 1–6. https://doi.org/10.1186/s41235-021-00351-9 Ichimura, F., Moriwaki, M., & Ariga, A. (2021). Romantic bias in judging the attractiveness of faces from the back. Journal of Nonverbal Behavior , 45 , 339–350. https://doi.org/10.1007/s10919-021-00361-7 Kamatani, M., Ito, M., Miyazaki, Y., & Kawahara, J. I. (2021). Effects of masks worn to protect against COVID-19 on the perception of facial attractiveness. i-Perception , 12 , 1–14. https://doi.org/10.1177/20416695211027920 Kramer, R. S. S., & Jones, A. L. (2022). Incomplete faces are completed using a more average face. Cognitive Research: Principles and Implications , 7 (1), 1–12. https://doi.org/10.1186/s41235-022-00429-y Lewin, C., & Herlitz, A. (2002). Sex differences in face recognition–women's faces make the difference. Brain and cognition , 50 (1), 121–128. https://doi.org/10.1016/s0278-2626(02)00016-7 Miyazaki, Y., & Kawahara, J. I. (2016). The mask bias on perceived facial attractiveness. Japanese Psychological Research , 58 , 261–272. https://doi.org/10.1111/jpr.12116 Orghian, D., & Hidalgo, C. A. (2020). Humans judge faces in incomplete photographs as physically more attractive. Scientific Reports , 10 (1), 110. Patel, V., Mazzaferro, D. M., Sarwer, D. B., & Bartlett, S. P. (2020). Beauty and the mask. Plastic and Reconstructive Surgery Global Open , 8 , e3048. https://doi.org/10.1097/GOX.0000000000003048 Rehnman, J., & Herlitz, A. (2007). Women remember more faces than men do. Acta Psychologia , 124 (3), 344–355. https://doi.org/10.1016/j.actpsy.2006.04.004 Rhodes, G., Jeffery, L., Watson, T. L., Clifford, C. W., & Nakayama, K. (2003). Fitting the mind to the world: Face adaptation and attractiveness aftereffects. Psychological science , 14 (6), 558–566. Rule, N. O., Ambady, N., & Adams, R. B. (2009). Personality in perspective: Judgmental consistency across orientations of the face. Perception , 38 , 1688–1699. https://doi.org/10.1068/p6384 Saegusa, C., & Watanabe, K. (2016). Judgments of facial attractiveness as a combination of facial parts information over time: Social and aesthetic factors. Journal of experimental psychology Human perception and performance , 42 (2), 173–179. https://doi.org/10.1037/xhp0000149 Sudhakar, S. (2023, February 7). ‘Unattractive individuals’ wear masks more often than others: Study. New York Post . https://nypost.com/2023/02/07/unattractive-individuals-more-likely-to-wear-masks-study/ Sugai, M., Yonemitsu, F., & Ariga, A. (2024). Romantic bias in judging the attractiveness of faces wearing masks. i-Perception , 15 (5), 1–9. Yonemura, K., Ono, F., & Watanabe, K. (2013). Back view of beauty: A bias in attractiveness judgment. Perception , 42 , 95–102. https://doi.org/10.1068/p7356 Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6903928","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":475403496,"identity":"30d4d11a-4384-4f82-960c-a665e52ae021","order_by":0,"name":"Natsumi Kubo","email":"","orcid":"","institution":"Chuo University","correspondingAuthor":false,"prefix":"","firstName":"Natsumi","middleName":"","lastName":"Kubo","suffix":""},{"id":475403497,"identity":"2e784da5-9c17-49c3-b8ee-0418e8c7fc87","order_by":1,"name":"Hakudai Makino","email":"","orcid":"","institution":"Chuo University","correspondingAuthor":false,"prefix":"","firstName":"Hakudai","middleName":"","lastName":"Makino","suffix":""},{"id":475403498,"identity":"ec3204dd-bc80-4baf-bce2-9288c71e8b6c","order_by":2,"name":"Atsunori Ariga","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABD0lEQVRIiWNgGAWjYJCCAwwMbDzy7A0MDIwNcEE2Qlr4ZAx7DpCgBQjkbBhuJKBowQ34288+PPilxoyHcebzCww/dxyWY5BIYPzwg4EvD5cWiTPpBodljqXxsEvnFDD2njlsDNTCLNnDwFaM2yNpDIcl2I7xMM7OSWBmbDucuB/oQmmgXxJxOVL+/DOgln//eRhunoFoaQDa8hufFoMbaQwHP7ax8TDcYD8A08KG1xbDG0BbGPvYeAx7chgO9ralGzPwPGyz7DHA7Re582nMH398Y7OXZz/+8MHPNms5Bvbkwzd+VBzDGWIgwMwDpngMDkD4oNgxOJaATwvjDzDF/gBZsAavllEwCkbBKBhRAAB1DFY4JXMn2AAAAABJRU5ErkJggg==","orcid":"","institution":"Chuo University","correspondingAuthor":true,"prefix":"","firstName":"Atsunori","middleName":"","lastName":"Ariga","suffix":""}],"badges":[],"createdAt":"2025-06-16 09:23:33","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6903928/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6903928/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":85723493,"identity":"b582d057-43c1-4c82-843d-1f4cf617f360","added_by":"auto","created_at":"2025-07-01 06:07:58","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":27569,"visible":true,"origin":"","legend":"\u003cp\u003eA schematic illustration of the learning and test phases in Experiment\u003c/p\u003e","description":"","filename":"Picture1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-6903928/v1/552f5ddfbdbfc8d67c1d9722.jpg"},{"id":85724179,"identity":"f494fa31-7319-4b11-ad23-919c0f368439","added_by":"auto","created_at":"2025-07-01 06:15:58","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":116138,"visible":true,"origin":"","legend":"\u003cp\u003eResults of Experiment (a: female participants, b: male participants).\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-6903928/v1/0201f6dbf492a8706eb94033.png"},{"id":89249321,"identity":"af362f53-4f78-424c-a4fb-af2221fe337e","added_by":"auto","created_at":"2025-08-18 03:01:39","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":532388,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6903928/v1/16f96f4c-3b38-498e-917e-a1d7e09fa067.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Attractiveness Learning from Partially Occluded Faces","fulltext":[{"header":"Introduction","content":"\u003cp\u003eEven when the whole face is not visible, people can make relatively accurate judgments of facial attractiveness based on limited information (e.g., Miyazaki \u0026amp; Kawahara, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Rule et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2009\u003c/span\u003e). Interestingly, judgments of partially occluded faces\u0026mdash;such as those seen from behind or covered by a mask\u0026mdash;do not fluctuate randomly around the actual attractiveness of the whole face. Instead, they tend to be consistently overestimated, a phenomenon referred to as \u003cem\u003eback-view bias\u003c/em\u003e or \u003cem\u003emask bias\u003c/em\u003e (e.g., Hies \u0026amp; Lewis, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Ichimura et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Patel et al., \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Sugai et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Yonemura et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; but see Miyazaki \u0026amp; Kawahara, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). In other words, judgments of attractiveness for partial faces are not merely predictions, but rather appear to reflect a systematic positive bias. In fact, this kind of overestimation is commonly experienced in everyday life (Sudhakar, 2023).\u003c/p\u003e \u003cp\u003eSeveral explanations have been proposed for why people tend to overestimate the attractiveness of a partially occluded face. These include completion based on a positive mental representation of an average or prototypical face (e.g., Kramer \u0026amp; Jones, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Orghian \u0026amp; Hidalgo, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2020\u003c/span\u003e), the occlusion of negative features (e.g., poor symmetry, rough skin, or acne; Kamatani et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Miyazaki \u0026amp; Kawahara, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2016\u003c/span\u003e), and strategies related to mate selection (Ichimura et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Sugai et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Noting that this bias is more pronounced when men evaluate women\u0026rsquo;s attractiveness, Ariga and colleagues (Ichimura et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Sugai et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) viewed the overestimation of attractiveness as a mate selection strategy that increases the likelihood of approaching a partner, particularly for men, who have lower reproductive costs compared to women. These accounts are not mutually exclusive. However, a key question remains: Why does our daily experience not lead to improved accuracy in judging the attractiveness of masked faces? In other words, can we acquire a cognitive strategy for evaluating partially occluded faces by encountering prediction errors\u0026mdash;instances where our judgment based on limited information differs from the actual attractiveness of the whole face? As mentioned above, such prediction errors should have occurred repeatedly in everyday life, especially during and after the COVID-19 pandemic, when people were frequently exposed to faces both wearing and unwearing masks. Nevertheless, consistent and robust overestimation has been reported both in laboratory and real-world settings (Sudhakar, 2023). In fact, our ability to recognize masked faces has not improved since the onset of the COVID-19 pandemic, even though we have been repeatedly exposed to both masked and unmasked faces (Freud et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). The present study is the first to address this issue by exploring the learning process involved in attractiveness judgments of partially occluded faces. Specifically, it investigates whether individuals can improve the accuracy of such judgments when provided with short-term feedback (i.e., the \u0026ldquo;true\u0026rdquo; attractiveness of the whole face).\u003c/p\u003e"},{"header":"Preliminary Experiment","content":"\u003cp\u003eThis preliminary experiment aimed to obtain baseline attractiveness ratings for facial stimuli from a large sample of participants. These ratings were used as the \u0026ldquo;true\u0026rdquo; attractiveness scores to provide feedback in the subsequent learning experiment.\u003c/p\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eEthics Statement\u003c/h2\u003e \u003cp\u003eAll procedures in this study were approved by the Research Ethics Committee of the author\u0026rsquo;s university (Approval No. 2024-077) and conducted in accordance with the Declaration of Helsinki. Informed consent was obtained from all participants prior to their participation.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eMethod\u003c/h3\u003e\n\u003cp\u003e \u003cem\u003eParticipants.\u003c/em\u003e A total of 1,621 Japanese participants were recruited for the online preliminary experiment. Data from 246 participants were excluded for failing a satisficer, resulting in a final sample of 1,375 participants (674 females; \u003cem\u003eM\u003c/em\u003e\u003csub\u003eage\u003c/sub\u003e = 32.69, \u003cem\u003eSD\u003c/em\u003e\u003csub\u003eage\u003c/sub\u003e = 5.31) used in the analysis.\u003c/p\u003e \u003cp\u003e \u003cem\u003eApparatus and Stimuli.\u003c/em\u003e As the preliminary experiment was conducted online, participants used their own devices (PC or tablet). The stimuli consisted of 80 shoulder-up, color headshots of Japanese student models (40 females, aged 18\u0026ndash;23) not wearing masks. To maintain a natural appearance, we did not control hairstyles and makeup. However, participants were instructed to maintain a neutral expression and to remove accessories and glasses as much as possible. Photographs were standardized for distance, angle, and luminosity. Each image subtended 434 \u0026times; 587 pixels.\u003c/p\u003e \u003cp\u003e \u003cem\u003eProcedure.\u003c/em\u003e Considering the online experiment, the 80 stimulus images were divided into four sets (each including 10 male and 10 female models), and participants were randomly assigned to one of the sets (315\u0026ndash;363 participants per set). In each trial, a photograph was randomly selected from the assigned stimulus set. Participants rated the attractiveness of each face on a 7-point scale (1\u0026thinsp;=\u0026thinsp;not attractive at all, 7\u0026thinsp;=\u0026thinsp;very attractive).\u003c/p\u003e\n\u003ch3\u003eResults\u003c/h3\u003e\n\u003cp\u003eThe mean attractiveness rating was calculated for each stimulus image separately for male and female participants. Because ratings from male and female participants were highly correlated (\u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;.91, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001), we collapsed across participant gender and averaged the ratings for each image. These averaged scores were used as the \u0026ldquo;true\u0026rdquo; attractiveness values in the subsequent learning experiment (male stimuli: \u003cem\u003eM\u003c/em\u003e\u0026thinsp;=\u0026thinsp;3.02, \u003cem\u003eSD\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.32; female stimuli: \u003cem\u003eM\u003c/em\u003e\u0026thinsp;=\u0026thinsp;3.62, \u003cem\u003eSD\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.42).\u003c/p\u003e"},{"header":"Experiment","content":"\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003eMethod\u003c/h2\u003e \u003cp\u003e \u003cem\u003eParticipants.\u003c/em\u003e Twenty Japanese participants were recruited for the experiment. Due to technical or personal issues, four participants did not complete the learning phase. As a result, data from 16 participants (8 females; \u003cem\u003eM\u003c/em\u003e\u003csub\u003eage\u003c/sub\u003e = 21.50, \u003cem\u003eSD\u003c/em\u003e\u003csub\u003eage\u003c/sub\u003e = 1.06) were included in the final analysis.\u003c/p\u003e \u003cp\u003e \u003cem\u003eApparatus and Stimuli.\u003c/em\u003e Participants used their own PCs during the learning phase, and a laboratory PC with a 23.8-inch monitor during the test phase. The experiment was controlled using MATLAB (The MathWorks, Inc.).\u003c/p\u003e \u003cp\u003eIn addition to the 80 no-mask images used in the preliminary experiment, we included 80 photographs of the same individuals wearing a white sanitary mask, all using the same product (referred to as \u0026ldquo;mask images\u0026rdquo;). From these 80 models, 60 (30 female) were pseudo-randomly selected as learning stimuli, and the remaining 20 (10 female) were used as test stimuli. Stimulus assignments varied across participants. The selection was made under the constraint that the mean attractiveness scores obtained in the preliminary experiment were balanced between the learning and test sets and across participants.\u003c/p\u003e \u003cp\u003e \u003cem\u003eProcedure.\u003c/em\u003e This experiment consisted of two phases: a learning phase and a test phase. In the learning phase, each participant was presented with a mask image randomly selected from the learning stimuli and asked to evaluate its facial attractiveness by predicting the appearance of the face without a mask. Participants rated the attractiveness on a visual analog scale from 1 (not attractive at all) to 7 (very attractive). After the participant clicked on the scale, the image of the same individual without a mask appeared, along with a \u0026ldquo;true\u0026rdquo; attractiveness rating that had been measured in the preliminary experiment, serving as feedback (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Participants were informed in advance that the feedback score was made by a large majority and were instructed to adjust their own ratings to align with it. They completed 60 learning trials per day for one week (Days 1 to 7) at home, at the same time each day. The order of stimuli was randomized across days and participants, and each stimulus appeared once per day.\u003c/p\u003e \u003cp\u003eFollowing the learning phase, participants took part in the test phase in a laboratory setting. They rated the attractiveness of mask faces from a separate set of test stimuli (not included in the learning phase), using the same 1-to-7 visual analog scale. No feedback was provided during the test phase. The test phase was conducted twice: once immediately after the learning phase (Test 1: Day 8), and once approximately a week later (Test 2: Day 14 or 16, depending on the participant).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Results and Discussion","content":"\u003cp\u003eThe attractiveness rating given by each participant for a masked image was subtracted from the \"true\" attractiveness rating of the corresponding unmasked image, as determined by the majority in the preliminary experiment. This difference was used as an index of mask bias. This difference was used as an index of mask bias. The mean mask bias was calculated for each participant by stimulus gender and day, and then averaged across participants. Figures\u0026nbsp;2a (female participants) and 2b (male participants) show the mask bias for each gender of the stimulus models as a function of day (learning and test phases). Because a positive mask bias and its gender difference were predicted by the previous study (Sugai et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), we conducted one-tailed one-sample \u003cem\u003et\u003c/em\u003e-tests (vs. 0) on mask bias values for each day and stimulus gender, separately for each participant gender.\u003c/p\u003e \u003cp\u003e \u003cem\u003eFemale Participants.\u003c/em\u003e Because one participant did not attend Test 2 due to a personal issue, her data were included only in the analyses of the learning phase and Test 1. Female participants showed a significant mask bias for male stimuli only on Day 1, \u003cem\u003et\u003c/em\u003e(7)\u0026thinsp;=\u0026thinsp;2.828, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;.013, Cohen\u0026rsquo;s \u003cem\u003ed\u003c/em\u003e\u0026thinsp;=\u0026thinsp;1.000, but no significant bias on subsequent days or for female stimuli. This suggests that learning was completed at this early stage. Interestingly, no significant bias was observed in either of the test phases (Tests 1 and 2), indicating not only that the learned strategy for judging attractiveness based on partially occluded faces transferred to novel stimuli (Test 1), but also that it was retained for approximately one week (Test 2).\u003c/p\u003e \u003cp\u003e \u003cem\u003eMale Participants.\u003c/em\u003e Because one participant did not attend the test phase due to a personal issue, his data were included only in the analyses of the learning phase. Male participants exhibited a significant mask bias for male stimuli on Days 1, 5, and 7, \u003cem\u003ets\u003c/em\u003e(7)\u0026thinsp;\u0026gt;\u0026thinsp;2.251, \u003cem\u003eps\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.030, Cohen\u0026rsquo;s \u003cem\u003eds\u003c/em\u003e\u0026thinsp;\u0026gt;\u0026thinsp;0.796, but not for female stimuli. This suggests that learning was not fully completed for male stimuli within the one-week learning period. Interestingly, a significant mask bias for male stimuli was observed in Test 2 of the test phase, \u003cem\u003et\u003c/em\u003e(6)\u0026thinsp;=\u0026thinsp;2.166, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;.037, Cohen\u0026rsquo;s \u003cem\u003ed\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.819.\u003c/p\u003e \u003cp\u003e \u003cem\u003eBaseline Measurement of the Test Stimuli.\u003c/em\u003e Bias baselines were measured using na\u0026iuml;ve participants who took part only in the test phase and did not undergo the learning phase. These participants rated the masked faces using the same 7-point scale as in the preliminary experiment. For female na\u0026iuml;ve participants (\u003cem\u003eN\u003c/em\u003e\u0026thinsp;=\u0026thinsp;40, \u003cem\u003eM\u003c/em\u003e\u003csub\u003eage\u003c/sub\u003e = 20.15, \u003cem\u003eSD\u003c/em\u003e\u003csub\u003eage\u003c/sub\u003e = 1.42), baseline bias was significantly greater than zero for both male (0.46) and female (0.64) stimuli, \u003cem\u003ets\u003c/em\u003e(39)\u0026thinsp;\u0026gt;\u0026thinsp;3.863, \u003cem\u003eps\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001, Cohen\u0026rsquo;s \u003cem\u003eds\u003c/em\u003e\u0026thinsp;\u0026gt;\u0026thinsp;0.611. Furthermore, the mask biases shown by female participants who had completed the learning phase were significantly lower than these baseline values, regardless of stimulus gender: \u003cem\u003ets\u003c/em\u003e(7)\u0026thinsp;\u0026gt;\u0026thinsp;5.836, \u003cem\u003eps\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001, Cohen\u0026rsquo;s \u003cem\u003eds\u003c/em\u003e\u0026thinsp;\u0026gt;\u0026thinsp;2.063 on Test 1; \u003cem\u003ets\u003c/em\u003e(6)\u0026thinsp;\u0026gt;\u0026thinsp;5.159, \u003cem\u003eps\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.002, Cohen\u0026rsquo;s \u003cem\u003eds\u003c/em\u003e\u0026thinsp;\u0026gt;\u0026thinsp;1.950 on Test 2. In addition, the mask bias for female stimuli on Day 1 was significantly smaller than the bias baseline, \u003cem\u003et\u003c/em\u003e(7)\u0026thinsp;=\u0026thinsp;5.543, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001, Cohen\u0026rsquo;s \u003cem\u003ed\u003c/em\u003e\u0026thinsp;=\u0026thinsp;1.960, and was not significantly different from zero, suggesting that learning for female stimuli was likely completed within the 60 trials on Day 1.\u003c/p\u003e \u003cp\u003eAs for male na\u0026iuml;ve participants (\u003cem\u003eN\u003c/em\u003e\u0026thinsp;=\u0026thinsp;20, \u003cem\u003eM\u003c/em\u003e\u003csub\u003eage\u003c/sub\u003e = 20.30, \u003cem\u003eSD\u003c/em\u003e\u003csub\u003eage\u003c/sub\u003e = 2.00), the baseline bias was also significantly greater than zero for both male (0.70) and female (0.61) stimuli, \u003cem\u003ets\u003c/em\u003e(19)\u0026thinsp;\u0026gt;\u0026thinsp;3.322, \u003cem\u003eps\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.002, Cohen\u0026rsquo;s \u003cem\u003eds\u003c/em\u003e\u0026thinsp;\u0026gt;\u0026thinsp;0.743. Similarly, the test-phase mask biases shown by male participants who underwent the learning phase were significantly lower than baseline values, regardless of stimulus gender, \u003cem\u003ets\u003c/em\u003e(6)\u0026thinsp;\u0026gt;\u0026thinsp;4.281, \u003cem\u003eps\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.005, Cohen\u0026rsquo;s \u003cem\u003eds\u003c/em\u003e\u0026thinsp;\u0026gt;\u0026thinsp;1.618. Considering the learning-phase results, male participants appeared to have partially acquired the strategy of judging attractiveness based on partially occluded faces, though learning was incomplete. This partial learning transferred to novel stimuli (Test 1) and persisted for about one week (Test 2). On Day 1, the mask bias for female stimuli was significantly smaller than the baseline, \u003cem\u003ets\u003c/em\u003e(7)\u0026thinsp;\u0026gt;\u0026thinsp;6.774, \u003cem\u003eps\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001, Cohen\u0026rsquo;s \u003cem\u003eds\u003c/em\u003e\u0026thinsp;\u0026gt;\u0026thinsp;2.395, and while bias was not significant (vs. 0) for female stimuli, it remained significant for male stimuli. These findings suggest that learning was likely completed rapidly for female stimuli and partially progressed for male stimuli during the 60 trials on Day 1.\u003c/p\u003e \u003cp\u003e \u003cem\u003eSummary of the Results.\u003c/em\u003e The central question of the present study was whether individuals can learn a strategy for judging facial attractiveness based on partially occluded faces through repeated exposure to prediction errors\u0026mdash;specifically, overestimation. The findings suggest the answer is yes. However, the learning process was influenced by the gender relationship between participants and stimulus models.\u003c/p\u003e \u003cp\u003eFirst, the baseline measurements in the test phase revealed that both female and male participants initially exhibited a mask bias for both male and female stimuli. Second, female participants rapidly completed the learning process for female faces on Day 1 and for male faces by Day 2. Third, male participants, like their female counterparts, quickly learned to judge the attractiveness of female faces on Day 1 but showed only partial progress in learning for male faces, which was not completed within the 7-day period. Fourth, the learned strategy for judging attractiveness based on partially occluded faces was successfully transferred to novel stimuli, both immediately after and one week following the learning phase. Taken together, these results demonstrate two key findings: (1) female participants learned to accurately estimate facial attractiveness through repeated prediction error experiences more effectively than male participants, and (2) the attractiveness of female faces was judged more accurately and earlier after learning compared to male faces. The former finding aligns with previous research suggesting that females are generally more accurate in face perception than males (Lewin \u0026amp; Herlitz, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2002\u003c/span\u003e; Rehnman \u0026amp; Herlitz, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2007\u003c/span\u003e). As for the latter, we speculate that female faces may be processed with higher motivation due to their higher average attractiveness compared to male faces (3.62 vs. 3.02 for no-mask faces, as measured in the preliminary experiment). Based on these findings, it is plausible that our daily experience of mask bias reflects a reduced bias due to naturalistic learning over time. Alternatively, it may persist due to a lack of such learning\u0026mdash;given that, in everyday life, we rarely encounter prediction errors across 60 individuals per day or view the same individuals repeatedly over seven consecutive days. In either case, the present study provides new evidence that individuals can learn to judge facial attractiveness based on partial information to some extent through a one-week, massed learning protocol. Otherwise, the bias observed in daily life is likely to remain uncorrected.\u003c/p\u003e \u003cp\u003e \u003cem\u003eGender Differences in the Learning Process.\u003c/em\u003e Previous research by Ariga and colleagues has shown that overestimation in attractiveness judgments for invisible faces varies by participant gender (Ichimura et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Sugai et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Specifically, male participants tended to overestimate female facial attractiveness more than female participants did for male faces. This overestimation was reduced when male participants were explicitly instructed to evaluate female faces as if they were judging friends rather than potential romantic partners. Ariga and colleagues interpreted these findings from the perspective of reproductive cost. In general, males have lower reproductive costs than females (Bateman, \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1948\u003c/span\u003e), leading them to pursue greater reproductive opportunities while minimizing Type II errors, according to error management theory (Haselton \u0026amp; Buss, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2000\u003c/span\u003e). This tendency may drive the overestimation of female facial attractiveness. In contrast, females bear higher reproductive costs, prompting more cautious judgments and a tendency to avoid Type I errors, where overestimation may be meaningless or even detrimental. In the present study, the finding that female participants more effectively reduced overestimation when provided with feedback may reflect an adaptive strategy aligned with mate-choice processes. However, it is notable that, contrary to previous findings (Sugai et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), male participants did not show increased mask bias when evaluating female faces. Instead, they exhibited a relatively persistent mask bias, particularly for male stimuli. These unexpected results may stem from task demands. In the current study, participants were asked to align their attractiveness judgments with scores provided by the majority, while in the previous study, judgments were made based on personal preference. Furthermore, according to the reproductive cost account (Ichimura et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Sugai et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), males may be less motivated to accurately judge female facial attractiveness in everyday contexts, which could explain the incomplete learning observed in the male participants in this study.\u003c/p\u003e \u003cp\u003e \u003cem\u003eWhat Did They Learn?\u003c/em\u003e One might argue that the participants in this study adjusted their rating scores \u003cem\u003ea posteriori\u003c/em\u003e, after making an initial attractiveness judgment. In this interpretation, they would have learned how much to modify their ratings to approximate the \u0026ldquo;true\u0026rdquo; attractiveness scores, calculated retrospectively. However, this explanation seems unlikely. If participants had simply employed a calculation strategy, both male and female participants should have demonstrated similar learning processes. Yet, the results revealed gender differences in learning.\u003c/p\u003e \u003cp\u003eMoreover, the reduction in mask bias observed during the learning phase persisted into the test phase with novel stimuli, even after a week. This persistence makes it unlikely that participants merely memorized the faces or their associated \u0026ldquo;true\u0026rdquo; attractiveness scores. Furthermore, given the difficulty of recognizing masked faces (Freud et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), it is improbable that participants relied on simple memorization.\u003c/p\u003e \u003cp\u003eIt is more plausible that participants acquired some form of cognitive strategy for judging attractiveness based on partially occluded faces. One possibility is that they learned to infer the attractiveness of a whole face using only limited, visible facial features. A second possibility is that they developed the ability to complete the occluded facial area through amodal completion. A third possibility is that they learned to evaluate attractiveness based solely on the visible facial parts while ignoring the occluded regions. Indeed, previous research has shown that the eyes significantly contribute to overall facial attractiveness ratings (Saegusa \u0026amp; Watanabe, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). Since the current data do not allow us to distinguish between these possibilities, future studies are necessary to investigate this issue.\u003c/p\u003e \u003cp\u003eInterestingly, previous research has shown that adapting to distorted faces can shift the criteria for what is perceived as normal and attractive to align with the distortion (Rhodes et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2003\u003c/span\u003e). If a similar adaptation had occurred for mask faces during the current learning phase, the mask bias would have persisted or even increased. However, this was not the case. Thus, it is notable that participants learned a strategy to adjust their attractiveness judgments of masked faces despite potential adaptation effects. In sum, what is clear for now is that people are capable of learning a cognitive strategy to estimate facial attractiveness with some degree of accuracy\u0026mdash;regardless of whether such learning occurs in everyday life.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eConflict of interest\u003c/h2\u003e\n\u003cp\u003eThe authors declare that they have no conflict of interest.\u003c/p\u003e\n\u003ch2\u003eFunding statement\u003c/h2\u003e\n\u003cp\u003eThe present study was supported by a Chuo University Grant for Special Research to AA.\u003c/p\u003e\n\u003ch2\u003eConsent for Publication\u003c/h2\u003e\n\u003cp\u003eAll participants obtained written informed consent before the experiment.\u003c/p\u003e\n\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\n\u003cp\u003eAll authors planned this study. N.K. and H.M. conducted the experiments. N.K. analyzed the data and wrote the draft of the paper. All authors reviewed the manuscript. This study was conducted under A.A. supervision.\u003c/p\u003e\n\u003ch2\u003eData Availability\u003c/h2\u003e\n\u003cp\u003eAll the data have been uploaded and are available in the OSF repository: https://osf.io/g78nd/\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eBateman, A. J. (1948). Intra-sexual selection in drosophila. \u003cem\u003eHeredity\u003c/em\u003e, \u003cem\u003e2\u003c/em\u003e(3), 349\u0026ndash;368.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFreud, E., Di Giammarino, D., Stajduhar, A., Rosenbaum, R. S., Avidan, G., \u0026amp; Ganel, T. (2022). Recognition of masked faces in the era of the pandemic: No improvement despite extensive natural exposure. \u003cem\u003ePsychological Science\u003c/em\u003e, \u003cem\u003e33\u003c/em\u003e(10), 1635\u0026ndash;1650.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHaselton, M. G., \u0026amp; Buss, D. M. (2000). Error management theory: a new perspective on biases in cross-sex mind reading. \u003cem\u003eJournal of personality and social psychology\u003c/em\u003e, \u003cem\u003e78\u003c/em\u003e(1), 81.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHies, O., \u0026amp; Lewis, M. B. (2022). Beyond the beauty of occlusion: Medical masks increase facial attractiveness more than other face coverings. \u003cem\u003eCognitive Research: Principles and Implications\u003c/em\u003e, \u003cem\u003e7\u003c/em\u003e, 1\u0026ndash;6. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s41235-021-00351-9\u003c/span\u003e\u003cspan address=\"10.1186/s41235-021-00351-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIchimura, F., Moriwaki, M., \u0026amp; Ariga, A. (2021). Romantic bias in judging the attractiveness of faces from the back. \u003cem\u003eJournal of Nonverbal Behavior\u003c/em\u003e, \u003cem\u003e45\u003c/em\u003e, 339\u0026ndash;350. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10919-021-00361-7\u003c/span\u003e\u003cspan address=\"10.1007/s10919-021-00361-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKamatani, M., Ito, M., Miyazaki, Y., \u0026amp; Kawahara, J. I. (2021). Effects of masks worn to protect against COVID-19 on the perception of facial attractiveness. \u003cem\u003ei-Perception\u003c/em\u003e, \u003cem\u003e12\u003c/em\u003e, 1\u0026ndash;14. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/20416695211027920\u003c/span\u003e\u003cspan address=\"10.1177/20416695211027920\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKramer, R. S. S., \u0026amp; Jones, A. L. (2022). Incomplete faces are completed using a more average face. \u003cem\u003eCognitive Research: Principles and Implications\u003c/em\u003e, \u003cem\u003e7\u003c/em\u003e(1), 1\u0026ndash;12. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s41235-022-00429-y\u003c/span\u003e\u003cspan address=\"10.1186/s41235-022-00429-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLewin, C., \u0026amp; Herlitz, A. (2002). Sex differences in face recognition\u0026ndash;women's faces make the difference. \u003cem\u003eBrain and cognition\u003c/em\u003e, \u003cem\u003e50\u003c/em\u003e(1), 121\u0026ndash;128. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/s0278-2626(02)00016-7\u003c/span\u003e\u003cspan address=\"10.1016/s0278-2626(02)00016-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMiyazaki, Y., \u0026amp; Kawahara, J. I. (2016). The mask bias on perceived facial attractiveness. \u003cem\u003eJapanese Psychological Research\u003c/em\u003e, \u003cem\u003e58\u003c/em\u003e, 261\u0026ndash;272. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/jpr.12116\u003c/span\u003e\u003cspan address=\"10.1111/jpr.12116\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOrghian, D., \u0026amp; Hidalgo, C. A. (2020). Humans judge faces in incomplete photographs as physically more attractive. \u003cem\u003eScientific Reports\u003c/em\u003e, \u003cem\u003e10\u003c/em\u003e(1), 110.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePatel, V., Mazzaferro, D. M., Sarwer, D. B., \u0026amp; Bartlett, S. P. (2020). Beauty and the mask. \u003cem\u003ePlastic and Reconstructive Surgery Global Open\u003c/em\u003e, \u003cem\u003e8\u003c/em\u003e, e3048. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1097/GOX.0000000000003048\u003c/span\u003e\u003cspan address=\"10.1097/GOX.0000000000003048\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRehnman, J., \u0026amp; Herlitz, A. (2007). Women remember more faces than men do. \u003cem\u003eActa Psychologia\u003c/em\u003e, \u003cem\u003e124\u003c/em\u003e(3), 344\u0026ndash;355. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.actpsy.2006.04.004\u003c/span\u003e\u003cspan address=\"10.1016/j.actpsy.2006.04.004\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRhodes, G., Jeffery, L., Watson, T. L., Clifford, C. W., \u0026amp; Nakayama, K. (2003). Fitting the mind to the world: Face adaptation and attractiveness aftereffects. \u003cem\u003ePsychological science\u003c/em\u003e, \u003cem\u003e14\u003c/em\u003e(6), 558\u0026ndash;566.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRule, N. O., Ambady, N., \u0026amp; Adams, R. B. (2009). Personality in perspective: Judgmental consistency across orientations of the face. \u003cem\u003ePerception\u003c/em\u003e, \u003cem\u003e38\u003c/em\u003e, 1688\u0026ndash;1699. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1068/p6384\u003c/span\u003e\u003cspan address=\"10.1068/p6384\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSaegusa, C., \u0026amp; Watanabe, K. (2016). Judgments of facial attractiveness as a combination of facial parts information over time: Social and aesthetic factors. \u003cem\u003eJournal of experimental psychology Human perception and performance\u003c/em\u003e, \u003cem\u003e42\u003c/em\u003e(2), 173\u0026ndash;179. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1037/xhp0000149\u003c/span\u003e\u003cspan address=\"10.1037/xhp0000149\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSudhakar, S. (2023, February 7). \u0026lsquo;Unattractive individuals\u0026rsquo; wear masks more often than others: Study. \u003cem\u003eNew York Post\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://nypost.com/2023/02/07/unattractive-individuals-more-likely-to-wear-masks-study/\u003c/span\u003e\u003cspan address=\"https://nypost.com/2023/02/07/unattractive-individuals-more-likely-to-wear-masks-study/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSugai, M., Yonemitsu, F., \u0026amp; Ariga, A. (2024). Romantic bias in judging the attractiveness of faces wearing masks. \u003cem\u003ei-Perception\u003c/em\u003e, \u003cem\u003e15\u003c/em\u003e(5), 1\u0026ndash;9.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYonemura, K., Ono, F., \u0026amp; Watanabe, K. (2013). Back view of beauty: A bias in attractiveness judgment. \u003cem\u003ePerception\u003c/em\u003e, \u003cem\u003e42\u003c/em\u003e, 95\u0026ndash;102. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1068/p7356\u003c/span\u003e\u003cspan address=\"10.1068/p7356\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Face perception, Attractiveness, Learning","lastPublishedDoi":"10.21203/rs.3.rs-6903928/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6903928/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eFacial attractiveness tends to be overestimated when it is partially occluded by a sanitary mask (\u003cem\u003emask bias\u003c/em\u003e). However, it remains unclear why people have not developed a cognitive strategy to accurately judge the attractiveness of partially occluded faces, despite repeatedly encountering this bias in daily life. Are we simply unable to learn such a strategy? This study examined whether people can learn to adjust their attractiveness judgments of mask faces by receiving repeated feedback on the \u0026ldquo;true\u0026rdquo; attractiveness (the uncovered face along with a rating previously assigned by a large majority). Through one week, participants were exposed daily to mask faces followed by feedback. Interestingly, female participants learned to adjust their judgments more quickly than male participants. These findings suggest that people can, to some extent, acquire a strategy for judging facial attractiveness with massed learning. Otherwise, the bias is likely to persist, as it does in everyday settings.\u003c/p\u003e","manuscriptTitle":"Attractiveness Learning from Partially Occluded Faces","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-07-01 06:07:53","doi":"10.21203/rs.3.rs-6903928/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"8330597b-0b78-49b3-b1c6-9319e05993ca","owner":[],"postedDate":"July 1st, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-08-18T02:53:33+00:00","versionOfRecord":[],"versionCreatedAt":"2025-07-01 06:07:53","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6903928","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6903928","identity":"rs-6903928","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}