Facial beauty and the late positive component of event-related potentials.

Over the past 10 years, considerable experimental effort has been directed toward determining the facial characteristics that underlie physical attractiveness. Several theories have been proposed to account for these findings (Alley & Cunningham, 1991; Langlois & Roggman, 1990; Thornhill & Gangestad, 1993). Johnston and Franklin (1993) proposed that the perceived attractiveness of a female face is a consequence of sexual selection favoring (a) female facial characteristics that were reliably correlated with fertility and (b) male's affective preferences for such cues. (Although, for any individual female, such cues can be viewed as indicators of fecundity, potential reproductive capacity, the "fertility hypothesis," considers their origin to be a consequence of the differential fertility, actual reproductive capacity, of those who possessed them to different degrees.) In the current study we evaluated the fertility hypothesis by examining the relationship between female facial characteristics and both behavioral and electrophysiological measures of affect, recorded from male volunteers.

Langlois and her colleagues (Langlois & Roggman, 1990; Langlois, Roggman, & Musselman, 1994; Langlois, Roggman, Musselman, & Acton, 1991) were the first to propose that the mathematical average of faces in a population is perceived as attractive. This "average-is-beautiful" position was based on an image-processing technique whereby individual faces were photographed and stored in a computer as a 512 x 512-pixel matrix of grayscale values. The images were then combined by averaging the grayscale values at each point and printing the resulting matrix as a simulated photograph. This procedure revealed that average faces were judged to be more physically attractive than the individual faces that made up the average and, furthermore, the attractiveness of a multi-face composite increased with the number of faces comprising the average (Langlois & Roggman, 1990; Langlois et al., 199]).

Several authors have suggested that artifacts of the averaging procedure, such as the blurring of facial details, the redaction of blemishes, or the enhancement of facial symmetry, may account for some observed effects (Alley & Cunningham, 1991; Thornhill & Gangestad, 1993). A more serious problem for the averaging technique, however, may be the scaling procedure used by the authors. Langlois and Roggman (1990) stated that prior to averaging, individual "faces were adjusted for size and position by matching the location of the eye pupils and the middle of the lip line across all faces" (p. 116). Two steps are required to make this adjustment using image-processing software. First, the size of each individual face must be modified until the horizontal distance between the pupils is the same for all. This is necessary to achieve eye alignment during the averaging of gray-level values. Second, to keep the interpupil distance constant and adjust the eye-lip distance to the same value for each face, it is necessary to scale faces in a vertical direction while holding the horizontal scaling constant. The use of different horizontal and vertical scaling factors inevitably distorts the faces that comprise the average and may account for many apparent differences between the findings of Langlois and Roggman (1990) and several subsequent researchers who have employed different methodologies (Cunningham, Roberts, Barbee, Druen, & Wu, 1995; Ferrario, Sforza, Poggio, & Tartaglia, 1995; Johnston & Franklin, 1993; Perrett, May, & Yoshikawa, 1994).

An alternative procedure for generating attractive faces has been described by Johnston and Franklin (1993). This method employed a genetic algorithm to simulate the evolutionary process and allowed participants to "evolve" their most physically attractive face over a series of generations (Goldberg, 1989; Holland, 1975). The genetic algorithm procedure was first used to examine the facial characteristics of attractive female faces generated by male and female White participants. Anthropometric measurements of the generated faces were compared to the measurements of a normative group of female faces from the local population. The major results are shown in Figure 1. Contrary to the findings of Langlois and Roggman (1990), but consistent with later studies (Cunningham et al., 1995; Ferrario et al., 1995; Perrett et al., 1994), it was found that attractive faces are neither arbitrary nor average, but possess specific facial proportions and some features, such as fullness of the lips, that are significantly different from the population average. These results are best interpreted by considering the evolutionary (distal) and hormonal (proximal) factors that have contributed to male and female facial characteristics over the course of phylogenetic and ontogenetic development.

[Figure 1 ILLUSTRATION OMITTED]

Theoretical Considerations

Anthropologists now believe that as much as 99% of human evolution occurred in small hunter-gatherer territorial groups who lived on the open savanna (Washburn & Harding, 1970). Within such groups, males were predominately hunters or scavengers, and females cared for the young and conducted most of the foraging for vegetable food. Given this long evolutionary history, we should expect many characteristics of male and female, bodies and minds, to reflect the adaptations that were most useful for their biological survival while living under these unique conditions.

An examination of the facial features and proportions of adult males and females reveals many distinctive characteristics, even when adjusted for overall body size. Figure 2 shows the major differences in facial proportions between male and female faces based on the measurement of 67 adult female and 44 adult male faces. On average, males have bushier eyebrows set closer to the eyes, a more pronounced brow ridge, and sunken eyes. Both the nose and the mouth are wider in the average male face, and the lower jaw is both wider and longer than that of the average female face. Although such differences have little relevance in today's world, they may reflect important adaptations and preferences that were established in ancestral environments.

[Figure 2 ILLUSTRATION OMITTED]

The adult male face and body can be seen to possess many characteristics that would have been functionally useful during his long ancestral history as a hunter or scavenger. The enlarged openings of the mouth and nostrils provide effective passageways for the rapid transport of air to and from the lungs. This enhanced air flow, together with the larger vital capacity of the lungs, provide the necessary supply of oxygen to support the higher metabolic rate and higher hemoglobin count required for the efficient use of his larger muscle mass. Hunting also requires a high degree of energy expenditure and inevitable sweating from the brow, and other regions of the body, to regulate body temperature. Large bushy eyebrows set close to the eye on a protruding brow ridge provide an effective method for excluding sweat from the eye sockets as well as providing protection from an overhead sun; as a consequence, male eyes appear narrower than female eyes. In ancestral environments, adult females who were attracted to such physical characteristics would have enjoyed some enhanced fitness because their offspring would reap the benefits of these successful adaptations. From this perspective, it is not surprising that many of these adaptations appear at puberty.

At puberty, the face and body of males and females undergo a metamorphosis induced by steroid hormones in the presence of growth hormone (Tanner, 1978). Non-reproductive boys and girls enter puberty with almost identical proportions of muscle, fat, and bone but exit their metamorphosis as mature reproductives with very different body shapes and compositions (Forbes, 1975). Following puberty, a young woman has gained about 35 pounds of body fat, twice as much as a young man. In contrast, the male has acquired about one and a half times as much muscle and skeletal mass. Facial and axillary hair are now evident in the male, and physiologically, his metabolic rate, maximum [VO.sub.2], and lung capacity are all higher than that of females (Klafs & Arnheim, 1981). In addition, the male body undergoes a pronounced adolescent growth spurt controlled by androgens, especially testosterone, in the presence of growth hormone (Tanner, 1978). For females, presumably the small amount of androgens released from the adrenal glands is largely responsible for this adolescent growth (Tanner, 1978). This difference in androgen levels between the sexes appears to underlie the taller height and longer lower jaw region found in adult males compared to adult females (Figure 3).

[Figure 3 ILLUSTRATION OMITTED]

For females, estrogens at puberty regulate the accumulation of fat deposits throughout the body, resulting in a marked transformation of the body contours and an increase in the fullness of the lips, which reach their maximum size about 14.5 years of age (Farkas, 1981). By the end of puberty, the female mammary glands have undergone radical modifications in their structure, and strategic fat deposits have resulted in pronounced changes to her body contours. Johnston and Franklin (1993) proposed that shorter than average lower face proportions and fuller than average lips in female faces can serve as reliable indicators of high fecundity resulting from high estrogen and low androgen exposure at puberty. Ancestral males who found such facial features to be physically attractive would have enjoyed higher reproductive success than those with other emotional preferences. This "fertility hypothesis" of female beauty is supported by the observation that these facial attributes are considered to be physically attractive across many different cultures (Cunningham et al., 1995; Perrett et al., 1994). Furthermore, the most attractive female body shape is also correlated with high estrogen and low androgen exposure (Singh, 1993a,b,c), and deviations from this shape are associated with a marked reduction in fertility, as measured by ease of artificial insemination (Zaastra et al., 1993). The fertility hypothesis proposes that the attractiveness of female faces and bodies is a consequence of sexual selection on the basis of observable cues that are reliably correlated with fertility.

The current study was designed to expand on these findings by systematically manipulating the proportions and features of faces using image-processing software and examining the effects of these manipulations on behavioral ratings of attractiveness and a late positive component (LPC) of event-related potentials (ERPs) that has been shown to be a reliable correlate of affect.

Electrophysiological Correlates of Affect

Since an LPC, with a parietal (Pz) greater than central (Cz) greater than frontal (Fz) scalp distribution, was first reported by Sutton, Braren, Zubin, and John (1965), considerable experimental effort has been directed toward uncovering the psychological correlates of this component. Two properties of the LPC were initially established: its inverse relationship with subjective stimulus probability and its proportional relationship with "task relevance," as defined by the experimenter's instructions (Donchin & Coles, 1988). A large number of experiments, however, have shown that the LPC amplitude also varies with affect (Begleiter, Gross, & Kissin, 1967; Begleiter, Porjesz, & Chou, 1983; Farwell & Donchin, 1991; Johnston, Burleson, & Miller, 1987; Johnston, Miller, & Burleson, 1986; Kostandov, 1975; Simons, 1982). Johnston et al. (1986) demonstrated that when stimulus probability and classical "task relevance" were both held constant, the LPC varied with the "affective value" of the stimulus to the observer; stimuli with high negative or high positive affective value elicit larger LPCs than stimuli rated as emotionally neutral. Larger LPCs for negative and positive stimuli, compared to neutral stimuli, have been confirmed by other investigators (Naumann, Bartussek, Diedrich, Vogelbacher, & Mehrtens, 1993). Crites (in press) measured ERPs to positive, neutral, and negative stimuli in participants who were instructed either to report accurately their preferences or report falsely that neutral or negative stimuli were positive. When participants were accurate, their LPC amplitude was consistent with their preferences. However, when participants gave false reports, the LPC amplitude was consistent with the actual affective value of the stimulus rather than the reported preference.

The relationship between LPC and affect has been further supported by a number of recent studies demonstrating that both the affective value of stimuli and their LPC amplitude can be influenced by the biological state of the observer. Warren and McDonough (in press) demonstrated greater LPCs to smoking-related stimuli in smokers than in nonsmokers, an effect not found for neutral stimuli. Johnston and Wang (19911) examined the hypothesis that both the LPC and affective value vary with biological state, by making use of the natural fluctuations in hormone, levels over the menstrual cycle. They demonstrated that identical stimuli (pictures of babies, models, skin pathology, etc.) shown under different hormonal states were rated differently and also elicited different LPC amplitudes. As stimuli became more or less important in different biological states, this change in affective value was reflected in their LPC amplitude, leading to an interaction between the LPC elicited by the pictures and the hormonal status, as indexed by menstrual phase. These results are strong support for the hypothesis that this LPC varies with the "affective value" of the stimulus to the observer (Johnston et al., 1986).

The current study was designed to examine the relationship between behavioral ratings of affect and the LPC elicited by male and female faces that vary in physical attractiveness. The affective value hypothesis generates specific predictions concerning the amplitude of LPCs elicited by female faces that are identical in features but differ in the proportions of the lower face. If shorter lower face proportions enhance the attractiveness of any female face, then such faces should elicit larger LPCs than identical faces with average proportions under conditions when all stimulus faces are equally probable and task relevant. More specifically, the affective value hypothesis predicts that (a) female faces will elicit larger LPCs than male faces in an experimental group of male participants and (b) female, but not male, faces with short proportions will elicit a larger LPC component than faces with the same features but in the proportions of the average female face in the population. The inclusion of male faces is required because it can be argued that even though all stimuli are equally probable within an experimental session, a larger LPC to a female face with shorter proportions may reflect the lower "real world" probability of this configuration; that is, these shorter proportions are less probable than average proportions in the general population. If this is the case, however, then male faces with shorter than average male proportions should also elicit larger LPCs. Finally, the affective value hypothesis predicts that (c) the amplitude of the LPC will be correlated with participants' facial beauty ratings. To eliminate the possibility that such a correlation is a result of differential cognitive processing, it was necessary that the nature of the experiment be concealed and the collection of beauty ratings be delayed until all ERP data have been collected.

Method

Participants

The participants were 25 male student volunteers (ages 18-25 years) who received experimental credit for their participation. They were told that the experiment was concerned with recording their "brain-waves" to human faces, and they could withdraw from the experiment at any time, with full experimental credit. The consent form also requested participants to excuse themselves from the experiment, with full experimental credit, if they had a sexual preference other than heterosexual. No participants withdrew for any reason.

Measures

FacePrints computer software (Caldwell & Johnston, 1991) was used to generate a set of 64 gray-scale facial images (32 male, 32 female). The configuration of the faces was based on prior studies establishing the average (Farkas, 1981; Johnston & Franklin, 1993) and attractive features and proportions of female faces (Johnston & Franklin, 1993; Perrett et al., 1994). All combinations of two hair types (black hairstyle/blond hairstyle), two eye types (wide/narrow), two chin types (narrow/broad), two lip types (full/narrow), and two lower face proportions (short/long) were generated for both male and female facial images. Figure 4 shows two female faces with blond hair, wide eyes, narrow chin, and full lips that differ only in proportions. One face has the proportions of an average female face; the other has the short lower face proportions that have been shown to be attractive. Figure 4 also shows two male control faces, one with average male proportions, and the other modified from the average male face by the same degree as the modification made to the female faces. The stimuli subtended a 10-degree angle of visual arc when they were back projected onto a translucent Plexiglas screen from a Kodak carousel projector located in a room adjacent to the electrically shielded experimental chamber.

[Figure 4 ILLUSTRATION OMITTED]

Grass Model 7P122 DC amplifiers with upper and lower half-amplitude frequency responses set at 60 Hz and 0.04 Hz, respectively, were used to amplify the EEG signals. The amplified signals were collected by an IBM/XT computer fitted with a Tecmar Labmaster A/D board. The computer controlled the slide projector (used for displaying the stimuli), the button panel (used by participants to initiate a trial), and the tone generator (used to signal the beginning and the end of each trial).

Procedure

Each participant was fitted with three active non-polarizing (Ag/AgCl) electrodes at scalp locations Fz, Cz, and Pz, referenced to linked earlobes (A1 and A2). Eye-movements were monitored by recording the voltage between two electrodes located above and below the left orbit. These recordings were used to eliminate any ERPs contaminated with blinks or eye movements that exceeded 10 degrees of visual arc. The resistance of all scalp electrodes was adjusted until they were less than 10K.

After electrode placement, each man was seated in an electrically shielded, semi-dark room with a projection screen directly in front of him and a button panel adjacent to his right hand. He was instructed that he would hear a 500-Hz tone for 100-ms at the beginning of each trial. This tone indicated when he could press a button to initiate the presentation of a stimulus slide. After a 1,000-ms pause, 1 picture slide was projected for 1 second. At that point, the projector advanced to the next slide, and a tone indicated that the next trial could now be initiated. ERPs were recorded for 1,070 ms, starting 50 ms prior to the onset of each visual stimulus, and digitized at a rate of 100 points/second.

ERPs were collected from each participant during three experimental sessions, with a five-minute break between sessions to rearrange the stimulus slides. A session involved the presentation of all 64 picture slides with a different pseudo-random order between sessions and participants. Following the third session, participants were required to rate each of the 64 slides on a 5-point scale of physical attractiveness. They were instructed to use the terms most beautiful and least beautiful as end points when rating slides of female faces and the terms most handsome and least handsome when rating male faces. Participants performed no experimental task other than viewing the stimulus material.

Results

The data were initially analyzed using two within-subject ANOVAs; facial features serving as the independent variables and beauty/handsomeness ratings, or measures of the LPC, serving as dependent variables. The effects of facial feature combinations on ratings of physical attractiveness and LPC measures are shown in Table 1. For clarity, the behavioral and ERP results will be discussed in two sections, organized by facial features.

Table 1 F-ratios for Within-Subject ANOVA for Physical Attractiveness Rating and LPC Factor Score

                           Rating of Physical        LPC Factor
Effect                       Attractiveness       Score at Pz Site

Face Sex                       27.96(***)               0.02
Prop                           14.40(***)               0.02
Hair                           46.83(***)               7.46(**)
Eyes                            4.96(*)                 0.03
Lips                            4.50(*)                 2.89
Chin                            0.31                    0.19
Prop X Hair                     0.38                    0.32
Prop X Eyes                     0.09                    0.15
Prop X Lips                     9.46(**)                6.18(*)
Prop X Chin                     0.55                    7.16
Hair X Eyes                     0.61                    0.54
Hair X Lips                     1.95                    0.11
Hair X Chin                     3.95                    0.28
Eyes X Lips                    10.53(**)                5.90(*)
Eyes X Chin                     2.61                    0.41
Lips X Chin                     0.00                    1.30

Face Sex X Prop                16.50(***)               4.78(*)
Face Sex X Hair                 0.88                    2.44
Face Sex X Eyes                 4.48(*)                 0.09
Face Sex X Lips                 1.08                    0.54
Face Sex X Chin                12.04(**)                0.24

Face Sex X Prop X Hair          0.52                    2.66
Face Sex X Prop X Eyes          0.85                    0.97
Face Sex X Prop X Lips          0.79                    0.42
Face Sex X Prop X Chin         10.94(**)                0.77
Face Sex X Hair X Eyes          0.00                    0.02
Face Sex X Hair X Lips          0.17                    0.99
Face Sex X Hair X Chin          0.10                    0.09
Face Sex X Eyes X Lips         15.29(***)               1.55
Face Sex X Eyes X Chin          1.02                    3.41
Face Sex X Lips X Chin          0.10                    7.98(*)

Note: Degrees of freedom for all main, two-way, and three-way interactions are 1 and 24.

(*) p < .05

(**) p < .01

(***) p < .001

Behavioral Data

Proportions and hair. An examination of the men's behavioral ratings revealed a strong Face Sex X Proportion interaction. This interaction was clarified by independent ANOVAs on the ratings of male and female faces, separately. Men found the short lower face proportions of women's faces to be more attractive (M = 3.18) than average-face proportions (M = 2.82), F (1, 24) = 19.74, p [is less than] .001. However, proportions did not produce any significant rating differences in the case of male faces. It appears that the effect of proportions on behavioral ratings is mainly a consequence of the higher ratings given to female faces that possessed shorter than average lower jaw proportions. An analysis of the behavioral ratings also revealed a significant main effect of hair color. Faces with blond hair were considered more attractive (M = 3.03) on average than faces with black hair (M = 2.14) for both male and female faces.

Eyes and lips. A main effect of lips indicated that participants considered full lips (M = 2.63) more attractive than narrow lips (M = 2.53). However, a Lips X Proportion interaction revealed that lip effects were only significant within the context of average facial proportions. Lip type had no effect on faces with short proportions.

A statistically reliable Face Sex X Eyes interaction revealed that wide eyes (M = 3.18) were considered more attractive than narrow eyes in female faces (M = 2.81), but this was not the case for male faces. A strong three-way Face Sex X Eyes X Lips interaction refined the picture by revealing that the wide eyes beauty effect on female faces was substantially enhanced by full lips, compared to narrow lips. These discriminations were not reflected in the attractiveness ratings of male faces, where the Eyes X Lips interaction was not significant. For male faces, the effects of eyes did not depend on whether the lips were full or narrow. As a result, the overall two-way Eyes X Lips effect across face sexes was basically a reflection of the female simple interaction profile.

Chin. This is the feature whose effect on the attractiveness ratings was most different between male and female faces. A Face Sex X Chin interaction was due to participants rating narrow chins (M = 3.05) as more attractive than broad chins (M = 2.95) on female faces, but broad chins more attractive (M = 2.24) than narrow chins (M = 2.10) on male faces. Facial proportions were again useful as a context for evaluating these effects. An analysis of a Face Sex X Proportions X Chin interaction revealed that the Face Sex X Chin effect described previously was statistically reliable when faces had short proportions but not when faces had average proportions.

Electrophysiological Data

For each participant the data set contained 192 average waveforms (2 face sexes X 2 hair types X 2 lower face proportions X 2 eye features X 2 lip types X 2 chins X 3 electrode sites), collected over the three experimental sessions. For all 25 participants, this produced 4,800 ERPs, which were analyzed using a Principal Components Analysis and Varimax Rotation (Donchin & Heffley, 1978). The purpose of this analysis was to provide a set of uncorrelated factors, thought to reflect the activity of underlying physiological generators, that could be used to compare waveforms. The major advantage of this method, over baseline to peak amplitude analysis, is its ability to discriminate between temporally overlapping factors (Donchin & Heffley, 1978). Following subtraction of the pre-stimulus baseline, a Principal Components Analysis was performed on the covariance matrix of all 4,800 ERPs. Four factors were rotated using the Varimax method. After rotation, these factors accounted for 23%, 19%, 16%, and 14% of the variance, respectively. The temporal and spatial distribution of these factors was used to associate them with ERP components. Factor 1 was positive, started late, and loaded maximally at 900 milliseconds after stimulus onset; it had a Fz [is greater than] Cz [is greater than] Pz scalp distribution and was identified as the slow wave (Ruchkin, Johnson, Mahaffey, & Sutton, 1988). Factor 2 was positive with a maximum loading at about 550 ms after stimulus onset; its scalp distribution followed the typical Pz [is greater than] Cz [is greater than] Fz pattern of the classic P300 wave (Sutton et al., 1965). However, to avoid the controversy concerning whether such a component should be labeled as P300, which normally peaks about 300 ms after a simple stimulus presentation, it was identified only as a late positive component with a P300 scalp distribution. Finally, Factor 3, also a positive component, had a maximum loading 200 ms after stimulus onset and was identified as the P2 component.

Average ratings and ERPs were computed over all 25 participants for each female face, generating a data set of 32 average waveforms and ratings. A correlation was found between beauty ratings and LPC factor scores at the parietal electrode site, r = 46, p [is less than] .01 (Figure 5). This systematic relationship between the behavioral and electrophysiological variables was only found for the female data, the LPC factor, and the parietal site, where it is typical for the LPC to have maximum sensitivity. Neither LPC measures at any other electrode site (including the eye-movement channel) nor any other factor scores showed a significant correlation with attractiveness rating. As a consequence, the ANOVA results shown in Table 1 parallel the behavioral analysis, with the LPC factor scores at the parietal electrode site serving as the dependent variable.

[Figure 5 ILLUSTRATION OMITTED]

Hair and proportions. A main effect of hair type indicated that blond hairstyles elicited larger LPC scores than black hairstyles. A significant Face Sex X Proportions interaction revealed that the effect of lower face proportion was opposite for male and female faces (Figure 6). Short female proportions elicited larger LPC scores than average female proportions, but average male proportions elicited larger LPC scores than short male proportions. In the data for male faces, however, there was an interaction between proportions and chin shape, F(1, 24) = 7.83, p [is less than] .01. Broad chins elicited larger LPCs than narrow chins, irrespective of proportions, but for narrow chins the average proportions elicited larger LPCs than short proportions.

[Figure 6 ILLUSTRATION OMITTED]

Eyes and lips. A Lips X Proportion interaction indicated the fullness of lips had opposite effects for each proportion. Full lips elicited larger LPC scores on faces with short proportions, but narrow lips evoked larger scores in faces with average proportions. A significant Eyes X Lips interaction also revealed a dependency of the eye effects on lip type. Wide eyes elicited a greater LPC score than narrow eyes on faces with full lips but a smaller score on faces with narrow lips.

Chin. A three-way Face Sex X Chin X Lips interaction revealed differences in the processing of chin and lip combinations for male and female faces. The combination of full lips and a narrow chin elicited the largest LPC score in female faces, but full lips with a broad chin was most effective in male faces. Unlike the behavioral data, however, none of the simple two-way interactions was found to be statistically significant.

Discussion

The current study was designed to examine the effects of non-average female facial features and proportions on both the behavioral and electrophysiological responses; recorded from men. The study was not designed to explore the full range of male facial features that contribute to attractiveness; male faces were included as controls to eliminate alternative interpretations of the electrophysiological responses elicited by the female faces. As a consequence, the discussion w al focus on female facial characteristics and only those aspects of male features that appear to be justified, given the limitations of the experimental design.

Langlois and Roggman (1990) proposed that cross-cultural preferences for attractive faces may be accounted for by evolutionary pressures that favor characteristics close to the mean of the population or by cognitive processes that favor prototypical category members. These two positions are not mutually exclusive. From an evolutionary viewpoint, any facial characteristic that enhances biological survival will increase in frequency over generations as a consequence of natural selection. For example, wide, flat noses are best in the tropics, where they need only contribute a small proportion of the required moisture and warming to the hot steamy air; a much larger nose is required to condition the air in a cold, and climate (Morris, 1967). It follows that the average facial features of any breeding population should evolve, over generations, toward the most desirable configuration for survival in the local environment. Average faces are also symmetrical. It has been argued that the degree of symmetry in a face may reveal the ability to canalize development in the face of stress, that is, the ability to resist parasites during development (Watson & Thornhill, 1994). Individuals who are attracted to an average face would gain enhanced fitness because theft offspring would reap the benefits of these successful adaptations. Natural selection theory, therefore, predicts that individuals should be attracted to the average facial characteristics in a breeding population and consider such characteristics to be beautiful. This distal design can be efficiently implemented by a proximal cognitive mechanism whereby individuals are attracted to a prototypical face based on theft environmental experience.

However, despite its appeal, natural selection alone cannot account for the consistent experimental finding that non-average facial proportions are considered to be attractive across cultures (Cunningham et al., 1995; Perrett et al., 1994). The observation that very attractive faces are systematically different from the average in many different cultures, even when exposure to Western media is minimal (Cunningham et al., 1995), is better explained by the theory of sexual selection. From this viewpoint, a characteristic that is attractive to the opposite sex may enhance reproductive success even though the exhibited feature may be detrimental to personal survival. A classic example of sexual selection is the peacock's tail (Cronin, 1991). Even though a large, colorful tail may be dangerous to a peacock's personal survival, it continues to become more elaborate over generations if it serves as a signal for an attribute that is highly attractive to the opposite sex (Hamilton, 1982). Under these conditions, the tail length grows until its average benefit (in gained reproductions) is equal to its average cost (in lost reproductions). The average tail in the population is then a compromise between the influences of natural and sexual selection, whereas the most attractive tail possesses the extreme characteristics that are most desirable to the opposite sex.

Where facial preferences are concerned, sexual selection theory may explain why an average face is considered to be a beautiful face (Langlois & Roggman, 1990; Langlois et al., 1991) but not the most beautiful face in any cultural group (Barber, 1995; Cunningham et al., 1995; Johnston & Franklin, 1994; Perrett et al., 1994). The most attractive faces exhibit extreme characteristics that are systematically different from the average, and the direction of these differences is consistent across cultures. Cunningham et al. (1995) demonstrated a high correlation in attractiveness ratings across groups of Asian, Hispanic, and White participants, when assessing the physical attractiveness of Asian, Hispanic, Black, and White female faces. These results indicate that participants can detect cues for beauty that have wide generality across cultures. The same authors provide convincing evidence that such cues do not depend upon exposure to Western media. All of these observations are consistent with a sexual selection viewpoint on female facial beauty. The nature of these non-average facial features provides some insight into their potential functional significance.

The behavioral ratings in the current study indicate that men find shorter lower face measurements to enhance the attractiveness of female faces, but not male faces, when all other facial attributes are identical. The electrophysiological data provide parallel results. Here a significant sex by proportion interaction reveals that the effect of lower face proportions are opposite for male and female faces (Figure 6). Short female proportions elicit larger LPC scores than average female proportions, but the reverse is true for male faces. To the extent that these proportions are a product of androgen levels, as they appear to be (Figure 3), then the data are consistent with a male preference for female characteristics that reflect low androgen levels. The data are also compatible with testosterone effects being considered attractive in male faces. At the parietal electrode site, male broad chins elicited larger LPCs than narrow chins, irrespective of proportions, but average proportions elicited larger LPCs than short proportions on male faces with narrow chins. That is, for male faces, length may be irrelevant when the chin is broad (one effect of testosterone), but when the chin is narrow, then a longer chin (a second testosterone effect) may have more affective value.

All stimuli and stimulus categories were equally probable and task relevant, so the observed differences in the LPC cannot be attributed to these variables. Although it could be argued that female faces with shorter than average lower face proportions have a lower probability of occurrence in the real world, the failure to find any such effect on the LPC elicited by the shorter than average male proportions makes such an interpretation unconvincing. In contrast, if the amplitude of the LPC is viewed as reflecting the affective value of stimuli, the pattern of behavioral and electrophysiological results discussed previously is highly consistent and in accordance with theoretical expectations. The high correlation between parietal LPC and the beauty rating of female faces provides additional support for the affective value hypothesis (Figure 5).

The effects of lip and eye size on behavioral and LPC measures are again fairly consistent between the behavioral and ERP measures. In general, full lips are considered to be more attractive than narrow lips, irrespective of the sex of the face. Although full lips do not further enhance the already high beauty rating of faces with smaller lower face proportions, they do produce a significant increase in beauty when the lower face is in average proportions. The ERP data also reveal larger LPC scores to full lips compared to narrow lips, but only on faces with smaller lower face proportions. Full lips may be an important variable influencing physical attractiveness, particularly when they occur in conjunction with short lower jaw proportions. These findings are consistent with the fertility hypothesis proposed by Johnston and Franklin (1993). They found that lips and proportions were the major factors contributing to female facial attractiveness and suggested that these attributes can serve as indicators of a hormonal configuration associated with high fertility.

In agreement with prior studies (Perrett et al., 1994), wide eyes also enhance the beauty rating of female faces, but not male faces. This exclusively female beauty effect is more pronounced when wide eyes are found in conjunction with full lips. Similarly, wide eyes elicit a larger LPC score than narrow eyes on faces with full lips, but a lower score on faces with narrow lips, once again revealing a close conformity between the behavioral and electrophysiological measures. As noted previously, male eyes appear narrower because of the growth of the brow ridge. More open eyes, like a shorter lower jaw, may also serve as an index of lower than average androgen levels. In keeping with the fertility hypothesis, the effect of wide eyes on beauty is only found on female faces and is most pronounced when found in conjunction with estrogen effects (full lips).

Finally, blond hair generated both higher behavior ratings and larger LPC factor scores for both male and female faces. This correspondence further supports the interpretation of the LPC as a correlate of affective value but should not be considered as evidence for hair color being an important aspect of beauty; hair color was confounded with hair style in the current experiment.

Taken together, the behavioral and electrophysiological responses to female faces support the prior findings that shorter lower jaw proportions, full lips, and wide eyes all contribute to female facial attractiveness. The observation that these non-average features appear to enhance female beauty across cultures and are most pronounced following puberty suggests that they have an important biological function. The fertility hypothesis proposes that such female facial characteristics are a consequence of sexual selection favoring cues that were reliable indicators of female fertility. For humans, the importance of exhibiting and detecting such cues may lie in the rapid decline of female fertility with

age (Jones, 1995). From an examination of fertility rates in human populations that do not use contraceptives, Henry (1961) noted that female fertility rate declines to 85% between 30 and 34 and drops to 35% by age 40-44, compared with the fertility of women between 20 and 24 years of age. Given this relatively small window of reproductive opportunity, it would have been advantageous for ancestral males to evolve sensitivity to, and select mates on the basis of, cues that were reliably correlated with fertility.

Using human faces as experimental stimuli, we examined three specific predictions generated by the "affective value" hypothesis of LPCs (Johnston & Wang, 1991). Two of the three predictions were supported: (a) female, but not male, faces with short lower jaw proportions elicited a larger LPC than faces with average proportions and (b) the amplitude of the LPC was correlated with participants' ratings of physical attractiveness. The third prediction, that female faces would elicit larger LPCs than male faces in an experimental group of male participants, was in the predicted direction, although not statistically significant. Current studies underway in our laboratory, using female participants, may yet provide support for this prediction. As well as the predicted effects, the other observed changes in the LPC with facial features present an intricate pattern of results that is not easily explained within the standard cognitive framework (Donchin & Coles, 1988) but is consistent with the "affective value" hypothesis proposed by Johnston and Wang (1991) on the basis of ERPs elicited by biologically relevant stimuli.

Electrophysiologists, by confining their stimulus material to laboratory-generated signals (tones and flashes), may be needlessly sacrificing ecological validity in the search for experimental control. Worse, by ignoring the processing of real-world stimuli, this approach may give rise to general models of information processing that have little or no relevance to human performance in the real world. Our approach in the current study has been to regard the human brain as possessing a collection of adaptations that have evolved in response to the practical real-world problems that were repeatedly encountered in ancestral environments. From this perspective, an examination of the electro-physiological responses to biologically relevant stimuli is of paramount concern.

References

Alley, T. R., & Cunningham, M. R. (1991). Averaged faces are attractive, but very attractive faces are not average. Psychological Science, 2, 123-125.

Barber, N. (1995). The evolutionary psychology of physical attractiveness: Sexual selection and human morphology. Ethology and Sociobiology, 16, 395-424.

Begleiter, H., Gross, M. M., & Kissin, B. (1967). Evoked cortical responses to affective visual stimuli. Psychophysiology, 3, 336-344.

Begleiter, H., Porjesz, B., & Chou, C. L. (1983). P3 and stimulus incentive value. Psychophysiology, 20, 95-101.

Caldwell, C., & Johnston, V. S. (1991). Tracking a criminal suspect through "face-space" with a genetic algorithm. In R. Belew & L. Booker (Eds.), Proceedings of the fourth international conference on genetic algorithms (pp. 416-421). San Mateo, CA: Morgan Kaufmann.

Crites, S. L. (in press). Bioelectric echoes from evaluative categorization: II. A late positive brain potential that varies as a function of attitude registration rather than attitude report. Journal of Personality and Social Psychology.

Cronin, H. (1991). The ant and the peacock. Cambridge, England: Cambridge University Press.

Cunningham, M. R., Roberts, A. R., Barbee, A. P., Druen, P. B., & Wu, C. (1995). "Their ideas of beauty are, on the whole, the same as ours": Consistency and variability in the cross-cultural perception of female physical attractiveness. Journal of Personality and Social Psychology, 68, 261-279.

Donchin, E., & Coles, M. G. H. (1988). Is the P300 component a manifestation of context updating? Behavioral and Brain Sciences, 11, 357-374.

Donchin, E., & Heffley, E. F., III. (1978). Multivariate analysis of event-related potential data: A tutorial review. In D. Otto (Ed.), Multidisciplinary perspectives in event-related brain potential research (pp. 555-572). Washington, DC: US Government Printing Office.

Farkas, G. F. (1981). Anthropometry of the head and face in medicine. New York: Elsevier.

Farwell, L. A., & Donchin, E. (1991). The truth will out: Interrogative polygraphy ("lie detection") with event-related potentials. Psychophysiology, 28, 531-547.

Ferrario, V. F., Sforza, C., Poggio, C. E., & Tartaglia, G. (1995). Facial morphometry of television actresses compared with normal women. Journal of Oral and Maxillo-facial Surgery, 53, 1008-1015.

Forbes, G. B. (1975). Puberty: Body composition. In S. R. Berenson (Ed.), Puberty (pp. 132-145). Leiden: Stenfert-Kroese.

Goldberg, D. E. (1989). Genetic algorithms in search, optimization and machine learning. Reading, MA: Addison Wesley.

Hamilton, W. D. (1982). Heritable true fitness and bright birds: A role for parasites? Science, 218, 384-387.

Henry, L. (1961). Some data on natural fertility. Eugenics Quarterly, 8, 81-91.

Holland, J. H. (1975). Adaptation in natural and artificial systems. Ann Arbor: University of Michigan Press.

Johnston, V. S., Burleson, M. H., & Miller, D. R. (1987). Emotional value and late positive components of ERPs. In R. Johnson, Jr., R. Parasuraman, & J. W. Rohrbaugh (Eds.), Current trends in event-related potential research, electroencephalography and clinical neurophysiology, supplement 40 (pp. 198-203). Amsterdam: Elsevier.

Johnston, V. S., & Franklin, M. (1993). Is beauty in the eye of the beholder? Ethology and Sociobiology, 14, 183-199.

Johnston, V. S., Miller, D. R., & Burleson. M. H. (1986). Multiple P3s to emotional stimuli and their theoretical significance. Psychophysiology, 23, 684-694.

Johnston, V. S., & Wang, X. T. (1991). The relationship between menstrual phase and the P3 component of ERPs. Psychophysiology, 28, 400-409.

Jones, D. (1995). Sexual selection, physical attractiveness, and facial neoteny. Current Anthropology, 36, 723-748.

Klafs, C. E., & Arnheim, D. D. (1981). Modern principles of athletic training (5th ed.). London: Mosby.

Kostandov, E. A. (1975). Negative emotions and perception. In L. Levi (Ed.), Emotions, -Their parameters and measurement (pp. 405-420). New York: Raven.

Langlois, J. H., & Roggman, L. A. (1990). Attractive faces are only average. Psychological Science, 1, 115-121.

Langois, J. H., Roggman, L. A., & Musselman, L. (1994). What is average and what is not average about attractive faces? Psychological Science, 5, 214-220.

Langlois, J. R., Roggman, L. A., Musselman, L., & Acton, S. (1991). A picture is worth a thousand words: Reply to "On the difficulty of averaging faces." Psychological Science, 2, 354-357.

Morris, D. (1967). The naked ape: A zoologist's study of the human animal. New York: McGraw-Hill.

Naumann, E., Bartussek, D., Diedrich, O., Vogelbacher, D., & Mehrtens, S. (1993). Emotionality and the late positive complex of the event-related potential. In H. J. Heinze, T. F. Munte, & G. R. Mangun (Eds.), New developments in event-related potentials (pp. 157-162). Boston: Birkhauser.

Perrett, D. I., May, K. A., & Yoshikawa, S. (1994). Facial shape and judgments of female attractiveness. Nature, 368, 239-242.

Ruchkin, D. S., Johnson, R., Jr., Mahaffey, D., & Sutton, S. (1988). Towards a functional categorization of slow waves. Psychophysiology, 25, 339-353.

Simons, R. F. (1982). Physical anhedonia and future psychopathology: An electrical continuity? Psychophysiology, 19, 433-440.

Singh, D. (1993a). Adaptive significance of female physical attractiveness: Role of waist-to-hip ratio. Journal of Personality and Social Psychology, 65, 293-307.

Singh, D. (1993b). Body shape and woman's attractiveness: The critical role of waist-to-hip ratio. Human Nature, 4, 297-321.

Singh, D. (1993c). Is thin really beautiful and good? Relationship between waist-to-hip ratio (WHR) and female attractiveness. Journal of Personality and Individual Differences, 16, 123-132.

Sutton, S., Braren, M., Zubin, J., & John, E. R. (1965). Evoked potential correlates of stimulus uncertainty. Science, 150, 1187-1188.

Tanner, J. M. (1978). Foetus into man: Physical growth from conception to maturity. London: Open Books.

Thornhill, R., & Gangestad, S. W. (1993). Human facial beauty: Averageness, symmetry and parasite resistance. Human Nature, 4, 237-269.

Warren, C. A., & McDonough, B. E. (in press). Reactivity of smoker and nonsmoker brain evoked potentials to smoking cues. Addictive Behaviors.

Washburn, S. L., & Harding, R. S. (1970). Evolution of primate behavior. In F. O. Schmitt (Ed.), Neural and behavioral evolution. Neurosciences: Second study program (pp. 39-47). New York: Rockefeller University Press.

Watson, P. J., & Thornhill, R. (1994). Fluctuating asymmetry and sexual selection. Tree, 9, 21-25.

Zaastra, B. M., Seidell, J. C., Van Noord, P. A. H., teVelde, E. R., Habbema, J. D. F., et al. (1993). Fat and female fecundity: Prospective study of effect of body fat distribution and conception rates. British Medical Journal, 306, 484-487.

Manuscript accepted April 15, 1996

This manuscript is based on the report, "The Biological Basis of Facial Beauty," presented at the International Conference on the Biological Basis of Sexual Orientation and Sex-Typical Behavior Patterns, Minot State University, Minot, ND, May 25-27, 1995.

Address reprint requests to Victor S. Johnston, Ph.D., Department of Psychology, New Mexico State University, Las Cruces, NM 88003. E-mail: [e-mail].