香京julia种子在线播放

There are about 376 known extant primate species (Wilson and Reeder, 2005) covering a wide range of morphological variability. For the purpose of this study, we aimed to choose a number of stimuli that would cover as much variability as possible. Thus, we included at least one species from each genus, except for Phaner (Fork-marked Lemur), Procolobus (Olive Colobus), Pseudopotto (False potto) and Simias (Pig-tailed Langur), of which there were no acceptable photographs available at the time of stimuli preparation. We also purposely excluded a human as we did not want to direct the respondents to rank the primate faces in the context of human facial attractiveness. The particular species within genera including similar species were selected based on availability of acceptable photographs. Where there was a high morphological variability within the genus, we included more species (two to eight; e.g., Cercopithecus, Eulemur, Macaca, Saguinus, etc.). The East Javan Langur (Trachypithecus auratus) was included in both black and orange forms. In case of sexually dimorphic species, only males were included. There is a trade-off between the inclusion of both sexes and taxonomic coverage as the number of stimuli need to be limited so that the respondents stay interested and give reliable rankings. Ten species were represented by two different individuals for a control. The random factors were set in a nested hierarchy. The variance of beauty ranking among individuals of the same species was negligible when compared to that between species (VarCorr function in R: Variance nested in the Group (infraorder/superfamily/family): 0.06898805; Genus: 0.10402177; Species: 0.24383308; Residual = individual: 0.05087559). We then assessed correlations between the conspecifics. Spearman’s correlations for all factors (colors, facial measurements, rankings) were high and significant at the p < 0.05 level, except for mouth width and chin (beauty: r = 0.73; human-likeness: r = 0.95). The dataset contained 117 pictures in total (107 after the removal of the control species/individuals, which were not included in the analyses to avoid pseudo-replication). For the full list of included species, see Supplementary Appendix 1.

We collected good quality photographs of primates facing the camera. The main resources were Flickr¹ or Wikimedia Commons² licensed under the Creative Commons license. Supplementary resources were our own photos, photos provided by addressed authors, and books (Rowe, 1996; Mittermeier et al., 2010). For the full list of picture resources, see Supplementary Appendix 1.

Each photograph was modified to show the primate face in a standardized position: the background was cut off and set to white and the face (in the form of a bust, see Figure 1) was rotated so the eyes were intersecting a straight, notional horizontal line. The primate faces were size-adjusted to cover approximately the same space relative to each other on each image. When there were primates showing an emotional expression (e.g., a smile or a frown) or looking sideways, the photos were retouched so the face showed a neutral expression with eyes looking straight to the camera (see Figure 2). Because the used photos could not be standardized under the exact same angle, the primates in the pictures slightly differed in the degree of rotation on both vertical and horizontal axes. Thus, we could not test the effect of symmetry on human rankings of primate facial “beauty” as it clearly corresponded to the rotation of the faces. This rotation had no effect on any of the explained variables (none of the Spearman’s correlations were significant at the p < 0.05 level) and thus was excluded from further analyses.

The recognized taxonomy of Primates consists of seven distinct groups: Lorisoidea (African and Asian prosimians), Lemuroidea (Madagascar prosimians), the Tarsiers, Platyrrhini (New World primates), Cercopithecoidea (Old World monkeys), hylobatids (gibbons), great apes, and humans (Purvis, 1995; Yoder and Irwin, 1999; Pastorini et al., 2001; Geissmann et al., 2004; Mayor et al., 2004; Lei, 2008; Finstermeier et al., 2013). To identify groups of reasonable morphological variability suitable for the purpose of the analysis of human-rated facial attractiveness of primates, we performed the canonical variate analysis (CVA) using the geometrical morphometry data (see below, Section “Shape”). The CVA separated the primates into three distinct groups (see Figure 3), referred to as prosimians, Platyrrhini and Catarrhini in the manuscript. The analysis also confirmed morphological distinctiveness of humans when compared to other primates.

Preferences for each of the primate faces were assessed using an online survey following Frynta et al. (2010) and Lišková and Frynta (2013). The respondents (n = 286, 91 men and 199 women) were Czech citizens, 15–69 years old (mean age was 22.7 years). Their task was to rate each of the faces on a scale (1–7 Likert scale; 1 = the most “beautiful,” 7 = the least “beautiful” or “ugly”) according to their perceived “beauty.” The photographs, resized to 360 × 540 pixels, were presented one by one on a computer screen in a random order. Prior to the presentation of the stimuli, the respondents were able to see the whole variability of the stimuli in the form of thumbnail-sized preview pictures (160 × 240 pixels). After that, the respondents started to score the pictures. The whole set was divided into groups of 39 photos, and after evaluating each of the groups, the respondents were allowed to take a rest, although the majority of the respondents finished the scoring without the need of a break. In total, all 286 respondents rated all 117 pictures.

In order to quantify and test congruence in species ranking provided by different respondents, we adopted a two-way, consistency, average-measures intra-class correlation (ICC; McGraw and Wong, 1996; Hallgren, 2012) computed in R (irr package). Principal component analysis (PCA) was performed to visualize the multivariate structure of the data sets and to extract uncorrelated axes for further analyses. MANOVA and General Linear Models (LMs) were applied to test the effects of independent explanatory variables. Full LMs were further reduced according to Akaike criterion until log-likelihood tests revealed a significant comparison between the full and reduced models. Mann–Whitney test was used as a non-parametric alternative for variables deviating from normality (raw sores). The contribution of the explanatory variables (constrains) to the attractiveness rating of the primate faces was examined in redundancy analysis (RDA) as implemented in the R package vegan (Oksanen et al., 2017). RDA is a multivariate direct gradient method. It extracts and summarizes the variation in a set of response variables (subjective evaluation of primate beauty) that can be explained by a set of explanatory variables (see Section “Explanatory Variables”). This analysis permits to plot both response and explanatory variables to a space defined by the extracted gradients and enables to detect redundancy (i.e., shared variability) between sets of response and explanatory variables. Statistical significance of the gradients was confirmed by permutation tests. Most of the calculations were performed in R (R Development Core Team, 2010) and Statistica 9.1. (StatSoft, 2010).

This study was carried out in accordance with the recommendations of Institutional Review Board (IRB), Faculty of Sciences, Charles University in Prague, approval no. 2013/7, with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the IRB.

Results of the ranking procedure revealed considerable congruence among the respondents. Although the reliability of the individual rankings was quite low (ICC = 0.147, 0.204, 0.182 for men, women, and pooled data, respectively, with all p < 0.001), the ICC for the average-measures was in an excellent range: ICC = 0.940 for men, 0.981 for women, and 0.985 for the pooled data (Shrout and Fleiss, 1979; Cicchetti, 1994). These results indicate that there was a high degree of agreement within the group of the respondents and suggest that preferences for primate faces were rated similarly. Also the correlation between ranks provided by male and female respondents was very high: r² = 81.6%, p < 0.001. Multivariate analysis of variance revealed no effect of age (Wilks = 0.6191, F_{171, 109} = 0.97, p = 0.575) or age × gender interaction (Wilks = 0.5823, F_{171, 109} = 1.13, p = 0.2436), nevertheless, a small, but significant effect of gender (Wilks = 0.5384, F_{171, 109} = 1.35, p = 0.041) was found. To identify the species that substantially contributed to the gender differences, we performed Mann–Whitney U tests comparing the raw ranks of each species in male and female respondents; the levels of significance were Bonferroni-corrected. Men significantly differed in their preferences from women in only five cases, all of which were preferred by men more than by women: de Brazza’s Monkey (Cercopithecus neglectus), Patas Monkey (Erythrocebus patas), Humboldt’s Woolly Monkey (Lagothrix lagotricha), Drill (Mandrillus leucophaeus), and the Sumatran orangutan (Pongo abelii). Because the gender differences were small and involved only 5 out of 107 examined species of primates, we decided to pool the genders in further analyses concerning the means or multivariate axes (RDA) computed from the preference ranks. Both of these methods extract the agreement among respondents and thus further blend the minor effect of gender.

The primates whose faces were rated as the most “beautiful” were mostly prosimians: the top winners were the Black-and-white Ruffed Lemur (Varecia variegata), Ring-tailed Lemur (Lemur catta) and Southern Lesser Galago (Galago moholi). Moreover, little monkeys such as the marmosets (Callitrichinae) were favorite among the respondents, together with apes such as the Agile Gibbon (Hylobates agilis) and Bonobo (Pan paniscus). In contrast, the Proboscis Monkey (Nasalis larvatus) or the Bald Uacari (Cacajao calvus) were rated as the “least beautiful” (or “ugly”).

When overviewed within the particular groups, all three groups included both attractive and unattractive species (see Figure 6). However, the prosimians were rated as significantly more attractive than the other groups (post-hoc Tukey test, p < 0.01). The particular cues affecting the respondents’ decision and the relation to the uncanny valley theory is discussed below in the respective sections.

The inner facial parts represent one of the strongest factors determining the beauty of the species within the group of the primates most similar to us, i.e., the Catarrhini. Thus, the size of eyes, interocular length, mouth width, and length from the nose to the mouth (or eyes to the mouth) are strong cues that our respondents use as a guide when ranking the “beauty” of Catarrhine faces. Consequently, although the respondents were not instructed to categorize the primates (the scoring procedure in our experiment instructed the respondents to assign scores to the pictured faces according to the subjectively perceived beauty), the RDA2 axis shows that the respondents still categorized the ranked subjects, and this categorization was mainly based on the inner facial parts of the primate faces. This phenomenon is often reported in studies focused on human perception of animal beauty (e.g., snakes; Marešová et al., 2009; Landová et al., 2012; birds: Lišková et al., 2015) and resembles the task recognized as unsupervised human categorization (Pothos and Chater, 2002; Pothos and Close, 2008).

In literature, the understanding of the role of inner and outer facial features is unclear. Some authors claim that young children mostly use the outer facial features as the cues for facial recognition, and then this pattern switches to the “adult version,” in which the faces are recognized using the inner features (Campbell et al., 1995; Turati et al., 2006). Other authors argue that both children and adults use inner facial features for the recognition of familiar faces, but outer facial features for recognition of the unfamiliar ones (Ellis et al., 1979; Want et al., 2003; Bonner and Burton, 2004; Ge et al., 2008). Our results show that the inner facial features are not only used for categorization of the primates, but also play a very important role in the assessment of the facial beauty of the Catarrhine primates. Outer facial features are used to a much less extent, but also appear to contribute to the assessment of primate beauty (see Figure 7).

Our results show that two colors affect the attractiveness of primate faces: the bluish tint (in Catarrhines and the full picture set) and the yellowish brown color (Platyrrhines). In literature, colors do play a role in the assessment of attractiveness, especially the red color, which is important for both humans and non-human primates. Human faces exhibiting brighter red are perceived as healthier and more attractive (Re et al., 2011). Female Rhesus Macaques prefer males with redder faces (Waitt et al., 2003; but see Waitt et al., 2006, where this preference only applied to red hindquarters). Moreover, red clothing or even extraneous red (for example, red background of a presented picture stimulus) is perceived as more attractive by both human respondents (Elliot et al., 2010) and non-human primates (Hughes et al., 2015).

We examined the full variability of colors present in the picture set of primates, i.e., not only red, but also orange, yellow, and the bluish tint. Within our examined picture set, only three primates possessed bright red coloration of the face (The Bald Uacari, Silvery Marmoset, and Japanese Macaque). Thus, we instead tested the effect of the overall presence of the red color, mostly expressed as darker red or reddish brown fur color. However, we found no effect of this color on human preferences. The only color that positively affected human decisions toward all primates (regardless of the particular groups) was blue—the same color that is, within the context of facial attractiveness, usually perceived negatively, as blue, pale faces indicate low oxygenation and poor health (Stadie, 1919). However, blue is very often reported as the most preferred color in other studies examining human rating of animal beauty, e.g., parrots (Frynta et al., 2010), birds (Lišková and Frynta, 2013; Lišková et al., 2015), and even snakes (Ptáčková et al., 2017). In the latter case, the color was present only in the form of a bluish tint. Similarly, in this paper, this bluish tint affected the overall evaluation of the beauty and was rather the effect of the photos than the primate species themselves.

When analyzing the particular groups, the GLM revealed that the attractiveness of the New World monkeys was positively affected by the yellowish brown color. This is in agreement with the perception of human faces as more attractive (Stephen et al., 2012) and healthier (and thus more attractive: Stephen et al., 2009) show that the respondents increased yellow color together with red and overall lightness when aiming to create a healthy-looking human face. Even papers dealing with animal attractiveness report preference for yellow color in some cases, especially when rating the beauty of birds (Lišková and Frynta, 2013; Lišková et al., 2015). However, animal attractiveness is usually mainly determined by the pattern and achromatic contrasts (Lišková and Frynta, 2013; Lišková et al., 2015). Similarly, the only other variable next to the yellow color that explained the attractiveness ratings of the New World primates was the mean lightness—the respondents rated darker monkeys as more beautiful. This agrees with the animal studies but contrasts human-facial studies in which either lighter (Van den Berghe and Frost, 1986; Stephen et al., 2009; Coetzee et al., 2012) or medium-toned, but not pale or black faces are rated as more attractive (Frisby, 2006; Stephen et al., 2012, but see Fink et al., 2001).

Although the pattern variable was dropped out from the final LMs, the multivariate analysis shows that, at least to some extent, it negatively affects the evaluation of overall primate facial attractiveness. This seems to contradict some of the papers that report positive effect of pattern to the evaluation of animal (Lišková et al., 2015; Ptáčková et al., 2017) and mammalian (Landová et al., in prep.) beauty. However, it is the highly contrasting pattern of large spots, stripes, and other marks, that positively affects the perceived animal beauty; not the diminutive unevenness of the fur color (i.e., agouti-type fur coloration), which is what the pattern variable used in this study corresponded to and which was perceived negatively. The primates possessing contrasting patches of fur coloration (e.g., the Black and White Ruffed Lemur and the Ring-tailed Lemur) were still considered the most beautiful.

Homogenous skin color distribution and surface topography (wrinkles), signs of health and age, also affect human preferences for attractive conspecific faces (Samson et al., 2010) and could possibly affect the preferences of primate faces as well. However, our set of stimuli was not controlled for the age of the depicted individuals (they were all adults of unspecified age) and thus we could not test the effect of the features affecting perception of age. Moreover, majority of the species included in the study possessed a face that was fully covered by fur. A carefully designed experiment with more uniform stimuli varying only in facial surface topography (e.g., faces of chimpanzees of varying age, or a manipulated picture set) would be needed to examine this interesting question.

There is not much variability within the prosimians in SSD, as most species have genders of similar size (Dixson, 2012). Platyrrhine primates differ more (variance in LG reaches from -0.72 in the Brown Capuchin to 0.31 in the Lemurine Night Monkey), but still lack the most prominent facial extremities typical for many male-larger Catarrhine species such as the Western Gorilla, Mandrill, Drill, Orangutans, Golden Snub-nosed Monkey, Proboscis Monkey, Patas Monkey, Gelada, or the Lion-tailed Macaque. Thus, it is not surprising that the degree of sexual dimorphism only affects the beauty within the Catarrhines—the male-larger species are perceived as less beautiful. In other words, the respondents rate conspicuous facial features (“extremities”) negatively. Similarly, many researchers agree that distinctive features (caricatures) usually help for better recognition of individual faces (Rhodes et al., 1987; Mauro and Kubovy, 1992; Lee et al., 2000; but see Hancock and Little, 2011), but are rated negatively when being evaluated for attractiveness (Deffenbacher et al., 1998). In other words, conspicuous features are usually rated as unattractive as opposed to average features (e.g., Rhodes and Tremewan, 1996; O’Toole et al., 1999). However, preference for facial averageness may be based on a more general principle as other objects such as dogs, birds, fish, or cars were also found to be preferred when they were of an average shape (Halberstadt and Rhodes, 2000, 2003).

The uncanny valley theory describes an empirical rule first mentioned in an essay by a robotics professor Masahiro Mori in 1970 (and later re-published in English for a broader audience to see; Mori et al., 2012). Mori hypothesized that if a robot resembled humans in appearance, people would feel affinity toward it, up to the point where it was too similar to humans—almost undistinguishable. At that point, people would experience a negative, eerie-like sensation, and he called this descent of affinity the “uncanny valley.” This relationship of human-likeness and attractiveness (in sense of positive affinity toward an object; for a relationship between the emotional ratings of eeriness and attractiveness, see Burleigh et al., 2013) was later tested in a number of papers, which found evidence in support of the uncanny valley theory (e.g., Seyama and Nagayama, 2007; MacDorman et al., 2009; Mitchell et al., 2011). Steckenfinger and Ghazanfar (2009) described a support for this phenomenon even in macaque monkeys, which preferred realistic and stylized macaque faces over faces very close to realism. The reason behind uncanny valley is unclear; often disputed mechanisms include the atypical feature hypothesis or the category conflict hypothesis (Burleigh et al., 2013). In the first case, the effect of uncanny valley is present when evaluating pictures that include an abnormal feature, such as bizarre eyes (Seyama and Nagayama, 2007). In the second case, the uncanny valley negatively affects stimuli containing features belonging to multiple categories, eliciting discomfort because it is evaluated as ambiguous and confusing (Saygin et al., 2011). Once the respondents cease to recognize the conflicting object as “human,” the attractiveness returns back to the linear character of the observed attractiveness (or eeriness).

In this matter, our results may seem contradictive as overally the most preferred primates were the prosimians, which were rated as the least human-like. However, within the group of Catarrhine primates, i.e., the group phylogenetically closest to humans (which also includes species most similar to humans, see Figure 6B), human-likeness positively affected the rated facial attractiveness. Thus, it is reasonable to examine whether the effect of uncanny valley can be applied to the results within this group. The graph clearly shows that some of the most human-like species do “fall into the valley,” but when looked at in more detail, there are exceptions to this uncanny valley rule. Some primates sharing similar rates of human-likeness fall into the notional valley, while others remain attractive. The unattractive, yet human-like primates, are represented by species such as the Orangutans, Proboscis Monkey, and the Drill, and the reason may be the presence of the abnormal (distinctive) features, as discussed above in Section “SSD, Averageness, and Facial Extremities”. Interesting fact is that at least some of these features, such as the prolonged nose of the Proboscis monkey, are not perceived as unattractive per se: for example, the elephants, elephant shrews, coatis or tapirs all possess a prolonged nose, but they were all rated as very or fairly attractive in a previous study (Frynta et al., 2013). Thus, these results cannot be interpreted simply as a preference for average feature size, but rather as a preference for average feature size when present on a human or human-like animal. Possibly, our complex neural system for facial recognition causes the judgements of “beauty” to be far more strict when judging “humans” (including human-like objects and animals) than different objects (Hanson, 2005). The uncanny valley phenomenon can be in fact linked to the expertise to human faces: the reason why the uncanny effect is so widespread is because every human has an expertise in recognizing humans; however, it is possible that the same effect might be observed in other types of expertises, as these display similar behavioral and neuropsychological pattern (Diamond and Carey, 1986; Xu, 2005; Dufour and Petit, 2010).

Previously, some researchers showed that even distinct, extreme features (up to some point) can be perceived as more attractive than average, if the exaggeration is based on attractive features. Such feature then may represent a super-stimulus, which is a concept derived from ethological studies, describing an object that contains features more accentuated than natural stimulus, which elicits a response more strongly than the stimulus for which it evolved (Tinbergen, 1951). For example, Perrett et al. (1994) report that enhanced female features, including higher cheek bones, thinner jaws, and larger eyes, were rated as more attractive than average (also see Perrett et al., 1998), and similar results were found by Jones and Hill (1993), who found that enlarged proportion of eye width to face height (i.e., feminine/neotenous feature) was preferred more than the average proportions. Could the presence of enhanced attractive features be the reason why in our data, some human-like primates are attractive above the linear relationship between human-likeness and attractiveness (see Figure 8)? For example, the Agile Gibbon (H. agilis) does have a contrasting, black-and-white face, a feature found to be very attractive in human preferences for animals (Frynta et al., 2013; Lišková et al., 2015). It would be interesting to examine this phenomenon in another study with more controlled, manipulated stimuli.

The support for the uncanny valley theory is ambiguous in our study. Some of the most human-like primates do fall into the notional “valley”; however, some other primates overcome it. Thus, our results rather show that it is the “atypical feature present on a human-like object” that makes some of the primates to be rated very negatively. However, this does not necessarily neglect the uncanny valley. Rather, it supports the atypical feature hypothesis of the uncanny valley theory (Seyama and Nagayama, 2007; MacDorman et al., 2009).