Content uploaded by Victor Johnston
Author content
All content in this area was uploaded by Victor Johnston on Aug 12, 2019
Content may be subject to copyright.
Hormones and Facial Attractiveness
1
Running head: Hormones and Facial Attractiveness.
Male Facial Attractiveness: Evidence for hormone mediated adaptive design.
Victor S. Johnston and Rebecca Hagel
New Mexico State University
Melissa Franklin
University of New Mexico
Bernhard Fink and Karl Grammer
Ludwig-Boltzmann-Institute for Urban Ethology
Vienna, Austria
Correspondence should be addressed to: Victor S. Johnston, Department of Psychology, New
Mexico State University, Las Cruces, 88003 New Mexico, USA
Tel: (505) 646 2800, Fax: (505) 646 6218, email: vic@crl.nmsu.edu
Hormones and Facial Attractiveness
2
Abstract
Experimenters examining male facial attractiveness have concluded that the attractive male
face is (1) an average male face, or (2) a masculinized male face, or (3) a feminized male face.
Others have proposed that symmetry, hormone markers, and the menstrual phase of the
observer, are important variables that influence male attractiveness. This study was designed
to resolve these issues by examining the facial preferences of 42 female volunteers at two
different phases of their menstrual cycle. Preferences were measured using a 40-second
QuickTime movie (1200 frames) that was designed to systematically modify a facial image
from an extreme male to an extreme female configuration. The results indicate that females
exhibit (1) a preference for a male face on the masculine side of average, (2) a shift toward a
more masculine male-face preference during the high-risk phase of their menstrual cycle, (3)
no shift in other facial preferences. An examination of individual differences revealed that
women who scored low on a ‘‘masculinity’’ test (1) showed a larger menstrual shift, (2) had
lower self-esteem and (3) differed in their choice of male faces for dominance and short-term
mates. The results are interpreted as support for a hormonal theory of facial attractiveness.
Keywords: humans; facial attractiveness; sexual selection; mate choice; menstrual cycle;
hormones.
Hormones and Facial Attractiveness
3
Introduction
In a fraction of a second, the brain of a human male or female can ascertain the
physical attractiveness of another person’s face (Johnston & Oliver-Rodriguez, 1997; Oliver-
Rodriguez, Guan, & Johnston, 1999). This remarkable feat appears to depend on the delicate
interplay between physical markers on the face, perhaps fitness indicators (Miller, 2000), and
exquisitely sensitive brains that generate feelings, perhaps fitness monitors (Johnston, 1999),
in response to such signals. Support for this model comes from studies that have used various
image-processing techniques to systematically manipulate the features and proportions of
female facial images and observe the behavioral and/or emotional responses of men and
women exposed to such images. The current experiment attempts to evaluate and refine this
model by examining how the attractiveness of male and female faces varies with both their
displayed hormone markers, and the hormonal state of female viewers, as indicated by their
menstrual phase.
Early studies suggested that the most attractive female face was the average face in a
population (Langlois, & Roggman 1990; Langlois, Roggman, Musselman, & Acton, 1991;
Langlois, Roggman, & Musselman, 1994). Several experimenters, however, have concluded
that the image processing technique used in these studies may be flawed, and proposed that
although the average face is attractive, it is not the most attractive face in a population (Alley
& Cunningham, 1991; Johnston, 2000). Strong support for the ‘‘non-average’’ hypothesis
comes from Perrett, May, and Yoshikawa (1994), who demonstrated that an average face
made by combining random faces is judged to be less attractive than the average of attractive
faces drawn from the same random sample. Indeed, there is now substantial evidence
indicating that attractive female faces are not average, but differ from the average in a
systematic manner. More specifically, they possess a shorter, narrower lower jaw, fuller lips,
and larger eyes than an average face (Cunningham, Roberts, Barbee, Druen, & Wu, 1995;
Hormones and Facial Attractiveness
4
Johnston, & Franklin, 1993; Perrett et al., 1994). These specific markers have been shown to
be effective across cultures (Cunningham et al.,1995; Perrett et al.,1994) and
electrophysiological studies have revealed that they elicit emotional responses in male, but
not female, viewers of female faces (Johnston & Oiver-Rodriguez, 1997). Because pubertal
bone growth (brow ridges and lower jaw) is stimulated by androgens (Tanner, 1978) and lip
fullness parallels estrogen dependent fat deposits elsewhere on the female body (Farkas,
1981), Johnston and Franklin (1993) have hypothesized that an attractive female face may be
displaying hormone markers (high estrogen/ low androgen) that serve as reliable indicators of
fecundity.
In contrast to the research on female facial attractiveness, studies examining the
importance of hormone markers on male faces have produced apparently incompatible
results. For example, although a number of experimenters have demonstrated that women
favor a ‘‘masculinized’’ male face possessing a large jaw, prominent brow ridges, and cheek-
bones (Grammer & Thornhill, 1994; Scheib, Gangestad, & Thornhill, 1999), other studies
have reported that both British and Japanese females prefer a more ‘‘feminized’’ male face
with a shorter than average lower jaw (Perrett, et al., 1998; Penton-Voak, et al.,1999). Still
others have found that a mixture of mature features (large lower jaw, prominent cheekbones
and thick eyebrows) and neotenous features (large eyes and small nose) is the most desirable
configuration of male faces (Cunningham, Barbee, & Pike, 1990). Some of these
inconsistencies across studies may result from the different procedures used to generate the
male facial stimuli. For example, both studies reporting a preference for feminized male faces
used a caricaturing technique to create their masculinized and feminized facial images.
Producing a masculinized male face using this procedure involves an algorithm that further
exaggerates the differences between an average female and an average male face. This
mathematical procedure is based on the assumption that extreme masculine and feminine
Hormones and Facial Attractiveness
5
faces can be produced by a linear extrapolation of the differences between these average
faces. However, given that the majority of consistent male-female differences are the result
of bone-growth, which depends on a complex interaction between androgens, estrogens, and
growth hormone, this linear growth assumption may not be valid (Tanner, 1978; Grumbach,
2000).
A second undesirable consequence of caricaturing is its effects on facial symmetry.
Average male and female faces are inevitably low in fluctuating asymmetries (traits that
differ among individuals but are symmetrical at a population level) so morphs generated
between such faces (feminized males and masculinized females) will invariably be quite
symmetrical. It is also inevitable, however, that any small deviations from perfect symmetry
that exist in either the average male or average female face will be substantially exaggerated
in caricatured faces (masculinized males and feminized females). Caricaturing, therefore, is
likely to generate feminized male facial images that are more symmetrical than masculinized
male images. Because facial symmetry has been shown to be correlated with attractiveness
(Grammer & Thornhill, 1994; Mealey, Bridgstock, & Townsend, 1999; Townsend &
Gangestad, 1999; Rhodes, Proffitt, Grady, & Sumich, 1988; Perrett et al., 1999), the observed
preference for feminized over masculinized male facial images may be an artifact arising
from differences in symmetry between these faces. The current study attempts to improve the
research methodology by examining preferences for male (and female) faces that have been
masculinized (or feminized) on the basis of perceived masculinity (or femininity), rather than
employing the caricaturing technique.
Low fluctuating asymmetry (FA) is thought to reflect an ability to resist the harmful
effects of mutations, parasites, and/or toxins during development (Møller & Swaddle, 1997).
Because such resistance is partially heritable (Møller & Thornhill, 1997), there may be
important fitness benefits for females who exhibit a preference for such mates. A variety of
Hormones and Facial Attractiveness
6
experimental findings support this hypothesis. Across species, symmetrical males have
significantly greater mating success (Møller & Thornhill, 1998), and symmetrical men have
been shown to be more desirable and have more sexual opportunities than asymmetrical men
(Thornhill & Gangestad, 1994; Gangestad & Thornhill, 1997). In populations lacking
widespread contraception and modern medicine, these enhanced sexual opportunities can be
translated into more offspring and better health (Waynforth, 1998). However, despite the
correlation between symmetry and attractiveness, it appears that human females may not use
or even perceive fluctuating asymmetries when judging the attractiveness of male faces
(Scheib, Gangestad, & Thornhill, 1999).
Scheib and colleagues (1999) found that the measured FA of male faces was not only
correlated with their attractive ratings, but was also correlated with the attractiveness ratings
of a right or left half-face; circumstances where all cues to bilateral symmetry are absent.
Clearly, their experimental participants did not require cues to bilateral symmetry in order to
make accurate judgments of attractiveness. It appears that the salient cues for attractiveness
are apparent on each half of a male’s face and their presence is correlated with computed FA,
but nor perceived symmetry. An examination of the attractive faces led the authors to
conclude that the specific features correlated with attractiveness and FA were a longer lower
jaw and prominent cheekbones. Keating (1985) also found that the shape of the lower jaw
was an important attribute of male facial attractiveness. Using an Identi-Kit methodology, she
examined the effects of eye-size, lip fullness, brow thickness, and jaw shape, on both
dominance and attractiveness ratings. As she had predicted, the combination of masculine
features (square jaw, narrow eyes, thick eyebrows, and thin lips) enhanced the dominance
ratings of male faces, but only a subset of these attributes (square jaw and thin lips) resulted
in significantly higher attractiveness ratings. It appears that some high testosterone markers
(square jaw), and low estrogen markers (thin lips) influence both the dominance and
Hormones and Facial Attractiveness
7
attractiveness of male faces, but dominance and attractiveness are not identical attributes.
Given these prior results, the current study attempts to clarify the role of hormone markers in
the perception of dominance and attractiveness, in both male and female facial images.
Some of the discrepancy in findings among male attractiveness studies may be a
consequence of differences in the participant populations. One potential source of variance is
the hormonal status of female participants. Penton-Voak, et al. (1999) have shown that
females’ preferences for male faces changed as a function of the viewer’s menstrual phase at
the time of testing. Specifically, females tested during the nine days prior to ovulation (high
conception risk group) preferred a less feminized male face than females tested outside of this
window (low conception risk group). In their most recent study, using 139 participants who
responded to a magazine survey, Penton-Voak and Perrett (2000) reported that women in the
high conception risk group were significantly more likely to prefer a masculine face than
those in the low concept risk group. The authors interpret their findings as evidence for a
conditional mate choice strategy whereby females in the high conception risk group are
exhibiting a preference for male facial cues that signal adaptive heritable genetic
characteristics, such as immunocompetence. However, these menstrual studies have not
shown that the observed change in preference over the menstrual cycle is specific to attractive
male faces. Do females’ preferences for female faces also change? If so, then the observed
effect may simply reflect a general change in mood over the menstrual cycle (Dalton, 1982),
rather than a specific adaptation. Only a specific effect on male facial attractiveness would
provide strong evidence for complex adaptive design (Williams, 1966). To explore the
generality of the menstrual effect, the current study examined how a variety of different facial
preferences (attractive male, attractive female, dominant male, dominant female, etc.) were,
or were not, modified by the hormonal state of female viewers.
Hormones and Facial Attractiveness
8
Methods
Participants
The participants were 42 female volunteers between 18 and 35 years of age (M= 22).
These women were recruited from the undergraduate population at New Mexico State
University in Las Cruces (USA) and the Ludwig-Boltzmann-Institute for Urban Ethology in
Vienna (Austria). Participation in the experiment was limited to volunteers who stated that
they were (1) heterosexual, (2) not currently pregnant or breast feeding a child and (3) not
currently taking any steroid medications or birth control pills. All participants signed an
informed consent document indicating that they were volunteering for an experiment on
facial preferences that would be conducted over two experimental sessions. They were also
informed that they would be asked to provide relevant personal information, all such data
would be confidential, and they could withdraw from the experiment at any time.
Apparatus
In order to measure facial preferences, the method developed by Perrett et al. (1998)
was modified so that average male and female facial images could be masculinized or
feminized to varying degrees on the basis of perceived masculinity and femininity, rather
than the caricature procedure used by these authors. The final experimental tool was in the
form of a 1200-frame QuickTime movie that gradually morphed an extremely masculine
male facial image into to an extremely feminine face, at a rate of 30 frames per second. The
movie was created in four steps (Figure 1).
First, sixteen random male and sixteen random female facial images were separately
morphed to produce a composite average male and a composite average female image. The
male photographs were taken from college students between 18 to 26 years of age. The
female pictures were digital images from a CD-ROM by the Japanese artist Akira Gomi
Hormones and Facial Attractiveness
9
(1998), and ranged in age from 18 to 30. All of the photographs were taken under constant
light conditions and showed faces with neutral expression and with no apparent make-up,
facial hair, or adornments (e.g., earrings). Prior to morphing, all pictures were standardized
to the same orientation using the procedure described by Rikowski and Grammer (1999).
Using the ‘‘Facial Explorer’’ program (Grammer, Fieder, & Fink, 1998) the composite
average male and the composite average female image were produced in a single step. [For
details on the morphing algorithm itself see Wolberg (1990), Beier & Neely (1992), Gomes,
Darsa, Costa, & Velho (1999)].
Next, the features and proportions of a perceived masculine and a perceived feminine
face were ‘‘evolved’’ using the FacePrints software program (Johnston, 1994). The details of
this program have been described elsewhere (Johnston & Franklin, 1993). In essence,
FacePrints employs a genetic algorithm that allows participants to search a multidimensional
face-space of more than 34 billion possibilities and find their most masculine or most
feminine facial image. Together with the average male and female faces, these masculine-
male and feminine-female faces defined the features and proportions of the four key data
points needed to construct a movie that slowly morphed a masculine male face to a feminine
female face. However, in order to conceal the locations of the perceived masculine and
feminine faces, both ends of the movie were extrapolated using a five second caricature. That
is, the average male to masculine-male difference was extrapolated to produce an extremely
masculine male facial image and caricaturing the average female to feminine-female
difference produced an extremely feminine face. In the final step, all faces were fitted with
the same androgynous hairstyle and the movie clips were combined into a single QuickTime
movie using Adobe Premiere™. In essence, the movie permitted systematic modification of
the major features and proportions that differentiate human male from female faces following
Hormones and Facial Attractiveness
10
puberty (Farkas, 1981). A user could move back and forward through the movie using both a
slider control and single frame buttons.
---------------------------------
Figure 1 about here.
---------------------------------
Facial symmetry measures were obtained following the procedure developed by
Grammer and Thornhill (1994). Based on the inter-correlations between naiver users, this
procedure has been shown to produce a highly reliable index of bilateral symmetry (Grammer
& Thornhill, 1994; Rikowski & Grammer, 1999). Twelve standard landmarks were
identified on the average and extreme male and female images used in the morph movie.
These landmarks included the innermost and the outermost corners of the eyes and the
leftmost and the rightmost edges of the nose. The points for measuring the cheekbones were
defined as the leftmost and the rightmost pixels of the face on a horizontal line directly
beneath the eyes. The x-coordinates of the jaw and mouth were identified as points on a
horizontal line passing through the corners of the mouth. The ‘‘Facial Explorer’’ program
(Grammer et al., 1998) then measured the midpoints of the six resulting horizontal lines and
computed an index of horizontal symmetry by summing the x-axis differences between the
midpoints. This analysis revealed that (1) the average male face was very symmetrical (FA =
16) and (2) symmetry decreased systematically toward the extreme masculine end of the
movie (FA = 16.5).
Procedure
Each woman was evaluated during two experimental sessions that were exactly two
weeks apart. During session one, she viewed the morph movie and was told how to use the
slider and single step controls in order to find the male or female facial image that was closest
to a specified target face, such as an androgynous face. She was then required to use these
Hormones and Facial Attractiveness
11
controls to find the facial image that most closely resembled a set of target faces. The targets
were, an average-male face (AvM), an average-female face (AvF), an attractive-male face
(AtM), an attractive-female face (AtF), a dominant-looking-male face (DoM), a dominant-
looking-female face DoF), a healthy-looking-male face (HtM), a healthy-looking-female face
(HtF), a masculine-looking-male face (MaM), a feminine-looking-female face (FmF), an
intelligent-looking-male face (ItM), an intelligent-looking-female face (ItF), a good-father-
male face (GfM), a good-mother-female face (GmF), and an androgynous face (Pat). In each
case, a descriptive phrase was used to clarify the desired target face. For example, an average
face was described as "a typical male (female) on the street", a dominant face as that of a
male (female) who was "more likely to give than take orders", an androgynous face as "a face
that could be either male or female", and an attractive face as "the male (female) that you like
best from this range of possibilities". The order of these target faces was counterbalanced
among participants, and between sessions.
As each target face was located, the experimenter noted the corresponding frame
number. When complete, the male faces (including the androgynous face) that had been
selected by the participant were re-displayed in a random order. While each face was visible,
the participant was asked to rate the face on 20 different attributes, using a 7-point Likert
scale for each rating. These attributes were: physically attractive, sexually exciting,
protective, intelligent, coercive, sensitive, impulsive, selfish, trustworthy, good parent for
child, dominant, healthy, masculine, wealthy, volatile, threatening, cooperative, manipulative,
helpful, and controlling. Using the same scale, participants also evaluated the desirability of
each face as a short-term mate (STM) and a long-term mate (LTM).
In session two, each returning woman was asked to select and rate the same set of
faces, using the procedures employed during session one. When these tasks had been
completed, she was required to respond to the 60-item Bem sex-role inventory (BSRI), and
Hormones and Facial Attractiveness
12
the Rosenberg self esteem (SES) questionnaire (Bem, 1974; Rosenberg, 1965). Each woman
then completed a personal history form that included questions concerning the exact date of
her last menses (first day of bleeding), the typical length of her menstrual cycle, regularity of
her cycle, her age, her prior hormone use, and her pregnancy history. Finally, each
participant was given a debriefing statement that explained the purpose of the experiment,
and she was requested to telephone or e-mail the date of the onset of her next menses using
an assigned identification number. The requirement for a post-experiment menses report was
waived for those participants who, on the second session, reported a menses onset between
the two experimental sessions.
For each participant, the date of their last ovulation was computed using either their
post-experimental menses report or the reported menses between their experimental sessions.
Although the duration of the menstrual cycle varies among females, this variance is almost
exclusively confined to the follicular (pre-ovulatory) phase. Thus ovulation is almost exactly
14 days prior to the onset of the next menses, irrespective of the cycle length (Fluhmann,
1957; Matsumoto, Nogami, & Okhuri, 1962; Lein, 1979). In a 28-day cycle ovulation occurs
on about the 14th day; but in a 34-day cycle ovulation occurs on about the 20th day
(Katchadourian, 1980). In the absence of direct hormone measures, this counting backwards
procedure offered the most accurate method for determining the date of ovulation that was
closest to these participant’s test dates. Knowing the date of ovulation allowed each subject’s
menstrual state on the days of testing to be defined in terms of the number of days before or
after ovulation.
Results
For each target face, the participants’ selections made during the follicular phase (pre-
ovulation) were compared to those made during the luteal phase (post-ovulatory) of their
menstrual cycle. A series of paired t-tests revealed that that there was no significant difference
Hormones and Facial Attractiveness
13
in any face choice as a function of whether the woman was in her follicular or luteal menstrual
phase on the day of testing. A second analysis examined only those participants who had been
tested during the time of highest conception risk. Based on the probability of conception over
the menstrual cycle, the time of highest conception risk was defined as the nine days prior to
ovulation (Barrett & Marshall, 1969). Twenty-nine of the women had been tested within this
high-risk window. A within subject analysis of their data revealed that within the high-risk
window, these women selected an attractive male face that was significantly more masculine
than their choice outside of this window (t (28) = 2.20, p = .02). Furthermore, of all the male
and female facial preferences that were collected from these women, the attractive male face
was the only preference that changed significantly with conception risk. These findings
replicate and extend the menstrual shift effect first reported by Penton-Voak et al. (1999). The
observed shift in preference toward a more masculine male face was a change in preference
from a mean frame number of 299 during the low-risk phase, to frame 270 during the high risk
phase; a mean shift of 29 frames of the morph movie.
There was considerable variance in the size of the menstrual shift among the female
participants. These individual differences were clarified by examining the size of the
menstrual shift as a function of participants’ scores on the Bem sex-role inventory (BSRI). On
the BSRI, individuals are classified as androgynous if they score above the median value of
4.9 on both the masculinity (Bem-M) and femininity (Bem-F) scales (Bem, 1974). When
female participants were classified in this manner, the size of the menstrual shift was
significantly larger for the non-androgynous women (56 frames) compared to the androgynous
women (-5 frames); t (27) = 2.55, p = .017. A more detailed analysis revealed that the shift in
preference toward a more masculine male during the high-risk days of the menstrual cycle was
most closely related to a participant’s score on the masculinity scale. That is, there was a
significant inverse correlation between the size of participants’ menstrual shifts and their score
Hormones and Facial Attractiveness
14
on the Bem-M scale (r = -.40, p = .03). Also, the size of the menstrual shift was significantly
different for participants classified as above or below the median on the BEM-M scale (t (27)
2.79: p = .009). The low masculinity group’s average attractive male face changed from frame
298 to frame 245 (a 53 frame shift) whereas the high masculinity group changed from 299 to
315, sixteen frames in the reverse direction (Figure 2).
---------------------------------
Figure 2 about here.
---------------------------------
Participants in the high Bem-M group, also scored higher on the Rosenberg self esteem
questionnaire (t (27) = 2.12, p = .04). Finally, irrespective of menstrual phase, high Bem-M
and low Bem-M women differed on the frame they selected as exemplifying a dominant male
face; (t (27) = 2.46, p = .02. The high Bem-M group selected an extreme masculine face
(frame 6), whereas the low Bem-M women perceived a much less masculinized male face
(frame 63) as depicting a dominant male. This was the only facial preference that differed
between these two groups of participants.
For all 29 participants, the selected mean frame number, over sessions, was computed
for each target face. For male faces, the means were as follows: DoM (43), MaM (115), AtM
(284), HtM (275), GfM (341), ItM (385), AvM (394), Pat (699). The attractive male face
was significantly more masculine than the average male face (t = 4.39 (28); p < .0001) and
significantly different from all other target faces with the exception of the healthy-male face
(t = 0.64 (28); p = .74). A similar analysis of the female target faces revealed the following
mean frame numbers: Pat (699), AvF (925), GmF (959), ItF (995), HtF (1021), FmF (1053),
AtF (1074), DoF (1195). In agreement with prior studies, the attractive-female face was
significantly more feminized than the average-female face (t (28) = 6.93, p < .0001), and
differed from all others except the feminine-female face (t (28) = 1.17, p = .88).
Hormones and Facial Attractiveness
15
Each of the 29 high-risk participants had selected eight male face types (including Pat)
during each experimental session, and subsequently rated each of these faces on twenty
different attribute scales. Averaged across sessions, this procedure yielded 232 male faces
with mean ratings for all twenty attributes. A Principal Components Analysis was performed
on the correlation matrix between the 20 attribute ratings that participants had assigned to
these faces. Three factors accounted for 75 percent of the variance: 46%, 22%, and 7%
respectively. All three factors were rotated using the Varimax method. The rotated factor
pattern indicated that the first factor was heavily loaded with negative attributes. In order of
importance, these were: threatening, volatile, controlling, manipulative, coercive, selfish,
dominant, and impulsive. On this basis, factor one (F1) was labeled the ‘‘Enemy’’ factor. F2
was named the ‘‘Friend’’ factor because, in order of magnitude, it was loaded with the
following positive attributes: helpful, cooperative, trustworthy, good father, wealthy, and
intelligent, all characteristics of a desirable friend. F3 was descriptively called the ‘‘Lover’’
factor. In order of importance, F3 factor-loadings were as follows: physically attractive,
sexually exciting, masculine, healthy, and protective. Figure 3 shows third-order polynomial
curves fitted to the mean factor scores of all three factors, over the first 700 frames of the
morph movie; from the extreme masculine face to the androgynous face. The curve for F1
reveals a progressive increase in perceived threat with increasing masculinity and a
pronounced increase in F1 scores with extreme masculinity. Perceived friendliness declines
rapidly with increasing masculinity, and the Lover factor, F3, peaks close to the location
where masculinized-faces received higher scores on F1 than F2.
---------------------------------
Figure 3 about here.
---------------------------------
Hormones and Facial Attractiveness
16
Figure 4 shows third-order polynomial curves fitted to the mean ratings of male faces
with respect to their suitability as a short-term mate (STM) and as a long-term mate (LTM).
These preference curves are plotted separately for the high Bem-M and low Bem-M groups.
For the high Bem-M participants, the LTM and STM rating curves are similar. In both cases,
the preference rating initially increases with increasing masculinization, reaches a maximum
value, and then declines with further masculinization. This same preference pattern is
reflected in the LTM curve of the low-Bem-M participants, but the STM curve for this group
shows a remarkably different pattern. In this case, the desirability of a male as a STM
continues to increase with facial masculinization. That is, women who shift their preference
toward a more masculine male face during the high-risk phase of their menstrual cycle (low
Bem-M group) also prefer short-term mates with very masculine male features. To examine
the relationship between these facial preferences and the personality traits assigned to faces,
LTM and STM ratings were correlated with the F1, F2, and F3 factor scores of the same
faces. For high Bem-M participants, their LTM and STM mean ratings were significantly
correlated with both the Friend (r =.57, p < .001 and r =.44, p < .001, respectively) and Lover
(r = .62, p < .001 and r = .74, p < .001, respectively) factors, but not the Enemy factor.
Similarly, the low Bem-M group’s LTM ratings were significantly correlated with the Friend
(r = .41, p < .001) and Lover (r = .54, p < .001) factors, but their STM ratings were correlated
with the Enemy factor (r = .45, p < .001) and not significantly correlated with the Friend
factor (r = -.08). These relationships indicate that the personality attributes associated with
desirable STMs are quite different between the high Bem-M and low Bem-M participants.
---------------------------------
Figure 4 about here.
---------------------------------
Hormones and Facial Attractiveness
17
Discussion
The morph movie used in the current experiment offers a sensitive tool for examining
facial preferences by providing participants with a choice among male and female facial
images that have been masculinizing and feminizing to varying degrees. Unlike caricaturing
techniques that masculinize or feminize faces on the basis of questionable mathematical
assumptions, the morph movie systematically changes the major facial features and
proportions that are perceived to characterize human maleness and femaleness following
puberty. Furthermore, because such secondary sexual characteristics are mainly a
consequence of different levels of pubertal hormones (Tanner, 1978; Grumbach, 2000), this
methodology provides a basis for interpreting how facial preferences are related to the degree
to which such hormonal markers are displayed on the faces of men and women.
In agreement with prior studies, the current results support the conclusion that women
prefer male faces that are more masculinized than an average male face (Mealey et al., 1999;
Penton-Voak & Perrett, 2000; Thornhill, & Gangestad, 1993). That is, the attractive male
face possesses more extreme testosterone markers, such as a longer, broader lower jaw, and
more pronounced brow ridges and cheekbones than the average male face. These same
hormone markers are also associated with good health. That is, when participants were
required to select a healthy male face, their choice was not significantly different from the
attractive male face. This finding suggests that women consider such testosterone markers to
be an index of good health and that important health considerations may underlie their
aesthetic preference. The close relationship between attractiveness and perceived health is
also evident in the analysis of the participants’ attribute ratings of male faces. On the basis of
their close correlation, the factor analysis grouped together physical attractiveness, sexually
exciting, masculinity, healthy, and protectiveness, as a single factor, F3, that increased
systematically with increasing masculinization (Figure 3). An examination of the F3 factor
Hormones and Facial Attractiveness
18
scores, however, reveals that the relationship between health, attractiveness, and hormone
markers, is not linear. Although the ‘‘Lover’’ scores initially increase with facial
masculinization, they reach a maximum value and then decline with further masculinization.
This observation suggests that either negative attributes are associated with high levels of
testosterone, or positive attributes are perceived to decline under these conditions; both
hypothesis appear to be true. An examination of Figure 3 reveals that the Enemy factor (F1)
increases and the Friend factor (F2) decreases at high levels of masculinity. Clearly the
women in this study viewed pronounced testosterone facial markers to be associated with
dominance, unfriendliness, and a host of negative traits (threatening, volatile, controlling,
manipulative, coercive, and selfish). The causal relationship between testosterone levels and
these behavioral attributes is still controversial (see review by Mazur & Booth, 1998). If such
relationships are valid, however, then the aesthetic preference of human females can be
viewed as an adaptive compromise between the positive attributes associated with higher than
average testosterone (health cues) and the negative attributes associated with more extreme
masculinization.
The changes in females’ preferences over their menstrual phase provide additional
support for an adaptive model of aesthetic preference. First, of all the preferences that were
examined, only the attractive male face varied as a function of menstrual phase. Second, this
shift in preference occurred under very specific circumstances; during the nine days prior to
ovulation, when conception risk is highest. This is also the menstrual phase when there is a
unique hormonal mix, high estrogen levels accompanied by low progesterone levels. Third,
the overall shift in preference was very specific; a more masculine male face was preferred
when conception risk was high. Taken together, these three observations are strong evidence
for adaptive design. It appears that a female’s attraction to the testosterone markers on a
male’s face may be influenced by her estrogen/progesterone ratio. This suggests that the
Hormones and Facial Attractiveness
19
neural mechanism responsible for generating such positive feelings is sensitive to these
circulating hormone levels. The observation that the preference change is restricted to the
time of high conception risk indicates that the "healthy male’’ preference, discussed above,
may involve genetic factors (e.g. immunocompetence) rather than, or in addition to,
phenotypic healthiness. Indeed, Scheib, Gangestad, & Thornhill (1999) have shown that a
generally accepted measure of genotypic quality across species, fluctuating asymmetries, is
correlated with both attractiveness and a composite index of masculine features that included
the length of the lower jaw and prominent cheek-bones. From this perspective, the positive
feelings evoked by testosterone markers are a fitness enhancing adaptation.
The experimental participants showing the largest menstrual shift in their attractive
male preference were those scoring lowest on androgyny; more specifically, lowest on the
Bem masculinity scale (Bem-M). The observed correlation between Bem-M scores and the
size of the shift suggests that this relationship should be viewed as a continuum rather than a
dichotomy. For the purpose of discussion, however, it is useful to view the behaviors of these
two groups as representing two different reproductive strategies. In contrast to the high Bem-
M group, whose STM and LTM preferences are consistent and relatively stable across their
menstrual cycle, the low BEM-M participants shift their preference toward a more
masculinized male for both STMs, and during the high conception risk phase of their cycle.
That is, the low Bem-M group may be characterized as employing an opportunistic strategy
rather than a stable mate-choice strategy. An opportunistic strategy offers a solution to what
Cashdan (1996) describes as a woman’s conflict between finding a mate who will invest
(LTM) and securing good genes for her offspring (STM); different males are preferred for
different functions. It has been suggested that such mating strategies arise as a function of the
security of attachment to primary caregivers during childhood (Bowlby, 1969, 1973, 1980;
Draper & Harpending, 1982). Girls, for example, who grow up without fathers, are more
Hormones and Facial Attractiveness
20
likely to mature earlier, exhibit ‘‘precocious’’ sexuality, have low self-esteem, and have
difficulty forming long-term relationships (Chisholm, 1993; Jones et al 1972; Moffitt, Caspi
& Belsky, 1992; Surbey, 1990; Draper & Harpending, 1982). Although the current research
findings do not reflect directly on these hypotheses, it is of interest that the ‘‘opportunistic‘‘
females were lower in self-esteem and more sensitive to male dominance cues than those
with a stable mate-choice strategy. Perhaps father-daughter bonding enhances a female’s
self-esteem and reduces her sensitivity to male dominance cues, while a lack of attachment
has the reverse effect. Future research should attempt to further characterize these
‘‘opportunistic’’ females by directly measuring their parental bonding and their performance
on cognitive tasks, such as mental rotations, that could provide a second measure of
‘‘masculinization’’ for comparison with their low Bem-M scores. In this regard, it is
noteworthy that males’ performance on such mental rotation tasks has already been shown to
decline with father absence during childhood (Draper and Harpending , 1982).
In the current study, participants (a) expressed a preference for male faces that were
more masculine than the average, (b) shifted their preference toward a more masculine face
during periods of high conception risk, and in some cases (c) rated very masculine faces as
most desirable for short-term mates. All such preferences and shifts in preference, however,
were toward male faces that were less symmetrical than the average male face. This is
convincing evidence that participants’ choices were strongly influenced by hormone markers,
and symmetry, as a measure of immunocompetence, can not account for these preference
shifts. It is still possible, however, that the facial hormone markers serve as an honest index
of immunocompetence and explain why females are attracted to such features, but there may
be another reason for this preference. This alternative hypothesis arises from an examination
of the specific facial and bodily features that are influenced by testosterone at puberty.
Hormones and Facial Attractiveness
21
Boys and girls enter puberty with almost identical proportions of muscle, fat, and bone,
but they exit this critical phase of development as reproductive adults with completely
different body shapes and compositions. During puberty a male’s body undergoes a
pronounced adolescent growth spurt regulated by testosterone and aromatized testosterone, in
the presence of growth hormone (Tanner, 1978; Grumbach, 2000). By the end of puberty,
men have about 1.5 times the skeletal and muscle mass of women, whereas women have
stored twice as much body fat as men (Forbes, 1975). Facial changes parallel these
modifications to a male’s body. On average, men have more pronounced brow ridges,
sunken eyes, and bushier eyebrows set closer to the eyes (Farkas, 1981). Both the nose and
the mouth are wider in the male face, while the lower jaw is both wider and longer than that
of an average female face (Farkas, 1981). All of these male characteristics appear to have
little relevance in today’s world, but they would have been useful during his long ancestral
history as a hunter. The enlarged openings of the mouth and nostrils provide effective
passageways for the rapid transport of air to and from the lungs. This enhanced airflow,
together with the larger vital capacity of his lungs, is necessary for an adequate supply of
oxygen to support the higher metabolic rate and hemoglobin level required for the efficient
use of his larger muscle mass. These attributes are clearly advantageous for a hunting life
style. Less obvious are the adaptations around the eyes. Hunting requires a high degree of
energy expenditure that inevitably entails profuse sweating from the brow and other regions
of the body in order to regulate body temperature. Large bushy eyebrows set close to the eye
on protruding brow ridges provide an effective method for excluding sweat from the eye
sockets as well as offering protection from an overhead sun. Given these considerations, a
female’s preference for a more masculinized male during periods of high conception risk may
reflect an ancestral adaptive strategy for acquiring the genes of a healthy hunter; a direct
benefit to her male offspring.
Hormones and Facial Attractiveness
22
The healthy hunter hypothesis gains additional support from examining the origins of
other human aesthetic preferences. Why does sugar have a pleasant taste, or rotten eggs have
an unpleasant odor, or tissue damage feel unpleasant? These evaluative sensory feelings are
present at birth, but they do not reflect the inherent properties of events in the external world.
That is, hydrogen sulfide gas does not possess an unpleasant odor, rather, it is the human
brain that has evolved to generate an unpleasant sensation for an event that is biologically
dangerous; in this case, contaminated food. Indeed, Johnston (1999) has argued that all
pleasant or unpleasant sensory feelings (and emotions) have evolved as proximate evaluations
of events in the physical (or social) world that consistently posed threats or benefits to gene
survival, in ancestral environments. In some instances, such as the pleasant taste of sugar,
these preferences may have little current value, and may even be maladaptive in the modern
world. Prior to sugar refineries, however, individuals with a sugar preference ate ripe fruit,
the major available source of sugars, and benefited from both a healthy diet and a rich energy
supply for manufacturing ATP. When viewed from this perspective, male facial beauty may
be an evaluative feeling evoked by attributes that were more important in a hunter-gatherer
era than in today’s environment. Indeed, it is difficult to comprehend the modern female’s
preference for males with prominent gluteus maximus muscles (buttock muscles), without
considering the ancestral environment when such preferences had clear survival value.
Facial beauty is a positive feeling generated within the brains of men and women, in
response to a complex configuration of visual cues displayed on the faces of males and
females. It now appears that such features and proportions are predominantly the result of
pubertal hormones, and these hormone markers may indicate fitness enhancing attributes that
are, or were, of importance to the opposite sex. It is also apparent that the hormonal state of
viewers, as determined by their sex and/or reproductive status, may modulate the intensity of
the feelings evoked by these hormone markers. When viewed within this framework, facial
Hormones and Facial Attractiveness
23
beauty is not a trivial matter, but rather, it is the product of the co-evolution of fitness cues
(hormone markers) and feelings (fitness monitors) that have, or had, important reproductive
consequences for both males and females.
Hormones and Facial Attractiveness
24
References
Alley, T. R., & Cunningham, M. R. (1991). Averaged faces are attractive, but very
attractive faces are not average. Psychological Science, 2(2), 123-125.
Barrett, J. C. & Marshall, J. (1969). The risk of conception on different days of the
menstrual cycle. Population Studies, 23, 455-461.
Beier, T. & Neely, S. (1992). Feature-based image metamorphosis. Computer
Graphics, 26, (2), 35-42.
Bem, S. L. (1974). The measurement of psychological androgyny. Journal of
Consulting and Clinical Psychology, 42, 155-162.
Bowlby, J. (1969). Attachment and loss. Vol. 1. Attachment. New York: Basic Books.
Bowlby, J. (1973). Attachment and loss. Vol. 2. Separation: Anxiety and anger. New
York: Basic Books.
Bowlby, J. (1980). Attachment and loss. Vol. 3. Loss. New York: Basic Books.
Cashdan, E. (1996). Women's mating strategies. Evolutionary Anthropology, 5(4),
134-143.
Chisholm, J. S. (1993). Death, hope, and sex. Current Anthropology, 34(1), 1-24.
Cunningham, M.R., Barbee, A. P., & Pike, C. L. (1990). What do women want?
Facialmetric assessment of multiple motives in the perception of male facial physical
attractiveness. Journal of Personality and Social Psychology, 59(1), 61-72.
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 attraction. Journal of Personality and Social
Psychology, 68(2), 261-279.
Dalton, K. (1982). Premenstrual tension: An overview. In R. C. Friedman (Ed.),
Behavior and the menstrual cycle (pp. 276-290). New York: Dekker.
Hormones and Facial Attractiveness
25
Draper, P., & Harpending, H. (1982). Father absence and reproductive strategy: An
evolutionary perspective. Journal of Anthropological Research, 38, 255-273.
Farkas, L.G. (1981). Anthropometry of the head and face in medicine. New York:
Elsevier.
Fluhmann, C. F. (1957). Irregularities of the stages of the menstrual cycles. Western
Journal of Surgery, Obstetrics, and Gynecology, 65, 265-277.
Forbes, G. B. (1975). Puberty: Body composition. In S. R. Berenson (ed,), Puberty
(pp. 132-145). Stenfert-Kroese: Leiden.
Gangestad, S. W., & Thornhill, R. (1997) The evolutionary psychology of extrapair
sex: The role of fluctuating asymmetry. Evolution and Human Behavior, 18, 69-88.
Gangestad, S. W., Thornhill, R., & Yeo, R. A. (1994). Facial attractiveness,
developmental stability, and fluctuating asymmetry. Ethology and Sociobiology, 15, 73-85.
Gomes, J., Darsa, L., Costa, B. & Velho, L. (1999). Warping and morphing of
graphical objects. San Francisco: Morgan Kaufmann Publishers.
Gomi, A. (1998). Americans 1.0, Los Angeles 1994 [CD-ROM]. Bunkasha, Japan:
Digitalogue.
Grammer, K., Fieder, M. & Fink, B. (1998). Facial Explorer (Version 1.0.) [Computer
software]. Vienna: Ludwig-Boltzmann-Institute for Urban Ethology.
Grammer, K. & Thornhill, R. (1994). Human (Homo sapiens) facial attractiveness and
sexual selection: The role of symmetry and averageness. Journal of Comparative Psychology,
108, 3, 233-242.
Grumbach, M. M. (2000). Pubertal maturation in aromatase deficiency and resistance
to estrogen. In J. P. Bourguignon & T. M. Plant (eds), The onset of puberty in perspective (pp.
247-267). Amsterdam: Elsevier.
Hormones and Facial Attractiveness
26
Johnston, V. S. (1994). Methods and apparatus for generating composites of human
faces. U.S. Patent no. 5,375,195.
Johnston, V. S. (1999). Why we feel: The science of human emotions. Perseus Press:
Reading, Mass.
Johnston, V. S. (2000). Female facial beauty: The fertility hypothesis. Pragmatics &
Cognition, 8(1), 107-122.
Johnston, V. S., & Franklin, M. (1993). Is beauty in the eye of the beholder? Ethology
and Sociobiology, 14, 183-199.
Johnston, V. S. & Oliver-Rodriguez, J. C. (1997). Facial beauty and the late positive
component of event-related potentials. The Journal of Sex Research, 34 (2), 188-198.
Jones, B., Leeton, J., McLeod, I., & Wood, C. (1972). Factors influencing the age of
menarche in a lower socio-economic group in Melbourne. Medical Journal of Australia, 21,
533-535.
Katchadourian, H.A. (1980). Fundamentals of human sexuality (3rd ed.). New York:
Holt, Rinehart, and Winston.
Keating, C. F. (1985). Gender and Physiognomy of dominance and attractiveness.
Social Psychology Quarterly, 48 (1), 61-70.
Langlois, J. H., & Roggman, L. A. (1990). Attractive faces are only average.
Psychological Review, 1(2), 115-121.
Langlois, J. H., 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(5), 354-357.
Langlois, J. H., Roggman, L. A., & Musselman, L. (1994). What is average and what is
not average about attractive faces? Psychological Science, 5(4), 214-220.
Lein, A. (1979). The cycling female: Her menstrual rhythm. San Francisco: Freeman.
Hormones and Facial Attractiveness
27
Matsumoto, S.Y., Nogami, Y., & Ohkuri, S.(1962). Statistical studies on menstruation:
A criticism on the definition of normal menstruation. Gunma Journal of Medical Science, 11,
297-309.
Mazur, A. & Booth, A. (1998). Testosterone and dominance in men. Behavioral and
Brain Sciences, 21, 353-363.
Mealey, L., Bridgstock, R., & Townsend, G. C. (1999). Symmetry and perceived facial
attractiveness. Journal of Personality and Social Psychology, 76(1), 151-158.
Miller, G. (2000). The mating mind: How sexual choice shaped the evolution of
human nature. Heinenmann.
Moffitt, T. E., Caspi, A., & Belsky, J. (1992). Childhood experience and the onset of
menarche: A test of a sociobiological model. Childhood Development, 63, 47-58.
Møller, A. P., & Swaddle, J. P. (1997). Asymmetry, developmental stability, and
evolution. Oxford.
Møller, A. P., & Thornhill, R. (1998). Male parental care, differential parental
investment by females and sexual selection. Animal Behavior, 55(6), 1507-1515,
Møller, A. P. & Thornhill, R. (1997). Bilateral symmetry and sexual selection : a meta-
analysis. American Naturalist, 151, 174-192.
Oliver-Rodriguez, J. C., Guan, J., & Johnston, V.S. (1999). Gender differences in late
Positive components evoked by Human Faces. Psychophysiology, 36, 176-185.
Penton-Voak, I. S., Perrett, D. J. ,Castles, D. L., Kobayashi, T., Burt, D. M., Murray,
L. K., and Minamisawa, R. (1999). Menstrual cycle alters face preference. Nature, 399, 741-
742.
Penton-Voak, I. S., & Perrett, D. I. (2000). Female preference for male faces change
cyclically: Further evidence. Evolution and Human Behavior, 21, 39-48.
Hormones and Facial Attractiveness
28
Perrett, D. I., Burt, M. D. Penton-Voak, I. S. Lee, K. J. Rowland, D. A. & Edwards, R.
(1999). Symmetry and human facial attractiveness. Evolution and Human Behavior, 20, 295-
307.
Perrett, D. I., Lee, K. J., Penton-Voak, I., Rowland, D., Yoshikawa, S., Burt, D. M.,
Henzi, S. P., Castles, D. L., & Akamatsu, S. (1998). Effects of sexual dimorphism on facial
attractiveness. Nature, 394, 884-887.
Perrett, D. I., May, K. A., & Yoshikawa, S. (1994). Facial shape and judgments of
female attractiveness. Nature, 368(6468), 239-242.
Rikowski, A., & Grammer, K. (1999). Human body odour, symmetry and attractive-
ness. Proceedings of the Royal Society of London, Series B, 266, 869-874.
Rosengerg, M. (1965). Society and the adolescent self image. New Jersey: Princeton
University Press.
Scheib, J. E., Gangestad, S. W., & Thornhill, R. (1999). Facial attractiveness,
symmetry and cues of good genes. Proceedings of the Royal Society of London B 266, 1913-
1917.
Surbey, M. (1990). Family composition, stress, and human menarche. In F. Bercovitch
& T. Zeigler (Eds.), The socioendocrinology of primate reproduction. New York: Liss.
Tanner, J.M. (1978). Feotus into man: Physical growth from conception to maturity.
London: Open Books.
Thornhill, R., & Gangestad, S. W. (1994). Human fluctuating asymmetry and sexual
behavior. Psychological Science, 5, 297-302.
Thornhill, R., & Gangestad, S. W. (1993). Human facial beauty: Averageness,
symmetry, and parasite resistance. Human Nature, 4(3), 237-269
Waynforth, D. (1998). Fluctuating asymmetry and human male life-history traits in
rural Belize. Proceedings of the Royal Society of London B 265, 1497-1501.
Hormones and Facial Attractiveness
29
Williams, C. G. (1966). Adaptation and natural selection: A critique of some
evolutionary thought. Princeton: Princeton University Press.
Wolberg, G. (1990). Digital Image Warping. Los Alamitos, California: IEEE
Computer Society Press.
Hormones and Facial Attractiveness
30
Figure Captions
Figure 1. Schematic diagram of the steps used to produce the facial morph movie. 1.
Simultaneous morphs produce average male and female faces. 2. Faceprints program
produces features and proportions of masculine and feminine faces. 3. Carricatures produce
extreme male and extreme female faces. 4. Three morphs produce movie clips that are
combined into a single movie.
Figure 2. From left to right; the perceived average male face (frame 394) and the attractive
male face when conception risk is low (frame 298), and when conception risk is high (frame
245). The originals were color images.
Figure 3. Third order polynomial curves fitted to the mean factor scores of all three factors
(Enemy, Friend and Lover) over the first 700 frames (masculine images) of the morph movie.
For reference, the positions of the masculine male (MaM) and average male (AvM) are
marked.
Figure 4. Mean ratings of faces for desirability as a short-term mate (STM) and a long-term
mate (LTM), for participants who scored high or low on the Bem masculinity scale.
3
2
1
o
-1
0 100 200 300 400 500 600 700
F1: Enemy
F2: Friend
F3: Lover
Movie Frames
Mean Factor Scores
MaM AvM
0100 200 300 400 500 600 700
1
2
3
4
5
6
7
Movie Frames
Mean Rating
0100 200 300 400 500 600 700
1
2
3
4
5
6
7
Movie Frames
Mean Rating
High Bem-M Group Low Bem-M Group
STM
LTM
STM
LTM
