Sex Aftereffects 1
Running head: FACE SEX AFTEREFFECTS
TT
ii
tt
ll
ee
::
((
PP
DD
FF
MM
AA
KK
EE
RR
..
E Sex Aftereffects in Face Perception: What You See is Not What You Get
E
PP
SS Jay Friedenberg, Megan Kwasniak, and Christine Haber
))
Manhattan College
PP
rr
ee
vv
ii
ee
ww
::
TT
hh
ii
ss
First Author Contact Information:
E
E
P
P
Jay Friedenberg, Ph.D.
S
S
Associate Professor
Manhattan College
pp
Department of Psychology
ii
Riverdale, NY 10471
cc
tt
uu
Office: (718) 862-7895
rr
Fax: (718) 405-3249
ee
w
w
a
a
s
s
Sex Aftereffects 2
E-mail: Jay.Friedenberg@Manhattan.Edu
Sex Aftereffects 3
Abstract
We demonstrate a face sex aftereffect in which participants who view an
adapting face of one sex judge a subsequently appearing androgynous face as the
opposite sex. Androgynous faces were morphs made by an equal blending of a
male and female face. The aftereffect was replicated in three experiments with
different adapting durations and with both a variable and fixed inter-stimulus
interval. Attractiveness ratings provide a further validation of the phenomenon.
In experiment 3, with near equal numbers of male and female viewers, men who
perceived the morphs as female rated them more attractive than women.
Conversely, women who perceived the morphs as male rated them more
attractive than men.
Sex Aftereffects 4
Sex Aftereffects in Face Perception: What You See Is Not What You Get
In this study, we investigate a sexual aftereffect underlying face
perception. We do this by adapting viewers first to a face of determinate sex and
then have them observe a second androgynous face of indeterminate sex, what
we call a morph. Participants judged the sex and attractiveness of the morph or
of the morph and the adapting face. Morph viewing duration was always fixed,
while adapting duration and the time between presentation of the two faces was
varied across experiments. In what follows, we review the literature on sexual
dimorphism and facial attractiveness and on aftereffects and face perception. We
then describe our technique and make predictions regarding our two dependent
measures of sex judgement and perceived attractiveness.
Sexual Dimorphism and Facial Attractiveness
Mature male and female face differences contain secondary sexual
characteristics produced by hormones like testosterone and estrogen. Adult male
faces demonstrate a larger jawbone, salient cheekbones, and thinner cheeks
(Enlow, 1990). Male faces are also known to have a more protruding browridge
(Berry & Wero, 1993). This sexual dimorphism in facial shape has been linked to
perceived attractiveness. In one evolutionary account, extreme masculine or
feminine characteristics may be considered attractive to members of the opposite
sex because they advertise genetic fitness. For example, testosterone has been
shown to suppress human immune system function (Kanda, Tsuchida, &
Tamaki, 1996). Males with very masculine facial features must therefore have
strong immune systems to buffer such a handicap and may be perceived by
females as good mating partners (Folstad & Karter, 1992). Similarly, secondary
sexual characteristics in females that are estrogen dependent indicate fitness and
are considered attractive (Singh, 1993).
Available research however fails to show that masculine male faces are
consistently considered more attractive. Little & Hancock (2002) constructed
composite male faces made by averaging three, six, or 12 faces together. The
averaging technique simultaneously produced decreased ratings of masculinity
but increased ratings of attractiveness, showing a preference for feminine traits
in their male faces. Swaddle & Reierson (2002) presented a continuum of male
faces ranging from low to high masculinity. Women selected the more masculine
end of the continuum as more dominant but not most attractive. Penton-Voak &
Perrett (2000) showed a female preference for masculine faces, but only when
conception was most likely, during the follicular phase of the menstrual cycle.
They believe women have multiple motives when it comes to selecting a sexual
partner, wanting high reproductive fitness and good genes when conception is
likely, but otherwise desiring a non-dominant partner to assist in child rearing.
Sex Aftereffects 5
Other work demonstrates that women as well as men prefer feminized
faces. Rhodes, Hickford, & Jeffery (2000) found the most attractive male and
female images for Caucasian and Chinese faces were significantly feminized.
Perrett et al. (1998) found that men and women both judge feminized versions of
male faces to be most attractive. Their hypothesized explanation for this is that
feminized faces reflect socially desirable characteristics of warmth, emotionality,
honesty, and cooperativeness. These characteristics should be desirable in a
mating partner by women, but also in a friend or colleague by men.
It is interesting then, to speculate on the perceived attractiveness of
androgynous faces that are equally masculine and feminine. Such faces are the
opposite of those employed in some of the literature, since instead of magnifying
sexually dimorphic characteristics, they reduce them. If androgynous morph
faces of the sort we use here are sexually ambiguous, we would expect lowered
ratings in comparison to faces with a normal degree of masculinity or femininity.
This is because morph faces do not have clear-cut cues guiding reproductive
action. Given the finding that feminized faces are usually preferred by both
sexes, we can make another prediction. Morphs perceived as male but with
accompanying feminine characteristics, should be judged more attractive than
morphs perceived as female with accompanying masculine characteristics. In
other words, a feminized male ought to be better looking than a masculinized
female. We address this prediction in the general discussion where the results of
all three experiments are summarized. Additional predictions regarding the
perceived attractiveness of morphs are made later in the introduction.
Aftereffects and Face Perception
Aftereffects are a widespread phenomenon in visual perception. Examples
include color aftereffects between the color opponents red-green and blue-
yellow, which allows for finer discrimination between wavelengths (Abramov &
Gordon, 1994). In motion perception aftereffects exist between motions in
opposite directions. Downward motion for instance, has an opponent relation to
upward motion (Hershenson, 1993). This helps to discriminate between motions
in different directions. Aftereffects like these can be the result of opponent-
process mechanisms where receptors for opposite attributes mutually inhibit one
another. If the neural receptor for one attribute is adapted, it becomes fatigued
and temporarily looses the ability to inhibit its opponent that becomes relatively
more active.
Aftereffects have been found for more complex visual forms as well,
although it is not clear if opponent-process mechanisms underlie them. Webster
and MacLin (1999) adapted subjects to frontal views of faces that were expanded
or contracted relative to a midpoint on the nose. Test faces then appeared
distorted in the opposite direction: if the initial face was expanded, the test face
appeared contracted and vice versa. These aftereffects transferred across
Sex Aftereffects 6
different faces, but were less pronounced when the adapting and test faces were
at different orientations, i.e. the effect was reduced between faces viewed upright
and then inverted.
Visual aftereffects for other facial attributes have also been discovered.
Leopold et al. (2001) found adaptation to a specific face identity caused an
average face to take on the "opposite" identity. They had participants adapt to a
test face and then measured identification accuracy for other faces along a
trajectory of morphed faces. The trajectory consisted of a path through a multi-
dimensional space with a mean or average face in the middle, an original face on
one side of the mean, and an "anti-face" on the other side. Identification
performance was shifted toward the original face after adaptation to the anti-
face, in comparison to adaptation to a non-matching face on another trajectory or
to a no adaptation condition. In addition Kaping et al. (2002) reports that
adaptation to a particular emotional expression (happy or sad), race (Caucasian
or Japanese), or sex (male or female) could cause a prior neutral image to acquire
the opposite characteristics.
Rhodes et al. (2003) demonstrated that contrast effects occur also for
judgments of averageness and attractiveness. They adapted viewers to faces that
were expanded or contracted relative to an average face. The participant's
consideration of what faces were average and attractive shifted in the direction of
the adaptation. These results suggest that adaptation renormalizes the face space
or dimension of facial features so that what is considered average or "normal" is
now centered closer to the adapting stimulus. They hypothesize that this type of
mechanism may have evolved to allow individuals to adjust to local population
characteristics, since non-migratory communities may be best suited to resident
environmental conditions (Rhodes et al., 2003). Alternatively, Rhodes mentions
that this sort of mechanism may be more general, reflecting the extraction of
prototypes useful for recognizing and categorizing objects (Rosch, 1973).
As Kaping et al (2002) have shown there are aftereffects for facial sexual
characteristics. The sex of faces is an extremely consequential feature. It has
specific functional implications for social interaction. For instance, knowing
someone's sex informs us whether this person poses a threat or may be a
potential reproductive partner. Cases of sexual ambiguity demonstrate just how
important this is. If we encounter an androgynous person whose sex cannot be
determined, we are at a loss as to how to behave towards them.
An aftereffect for the dimension of face sexuality has interesting
implications. Adapting to male faces would make them the new average.
Subsequently, an androgynous face with mixed features would now be judged
female, since the average has been moved toward the male end of the
continuum. Conversely, adaptation to female faces would shift the average
toward the feminine side causing androgynous faces to be perceived as male.
The consequence of this is a perceptual adaptation to the local distribution of
sexual characteristics. Time spent in a predominately male-looking social group
Sex Aftereffects 7
would predispose an individual to perceive a slightly feminine face as female.
Time spent among a female-looking group would additionally cause faces with
small amounts of masculinity to appear male. From a functional standpoint, this
adaptation would facilitate appropriate social interaction within the group.
In this study, we predict the presence of such a face sex aftereffect. We
demonstrate that continued exposure to a face of one sex produces increased
perceptual salience for the opposite sex. Continued exposure to a male face for
example produces increased salience of female face attributes. Conversely,
adaptation to a female face results in increased perceptual salience for male facial
characteristics.
Technique and Predictions
We test this aftereffect hypothesis in the current study using an adaptation
technique. Adapting faces, which are unaltered individual male or female faces,
are presented to observers for a given duration. Participants are then shown
androgynous faces. These androgynous faces contain the face previously viewed
combined with another face of the opposite sex never seen in the experiment. We
call these sexually ambiguous faces morphs, since they are constructed by an
equally weighted average of a single male and female face. If there is a face sex
aftereffect, exposure to a male adapting face should increase saliency of female
faces and vice versa. This can be measured in the observer's subsequent
judgement of the morph. If they are exposed to male adapting faces, they ought
to perceive the morphs as more female. Should they view female adapting faces,
they instead ought to see the morph as male.
In this investigation we ask observers to also rate the perceived
attractiveness, both of the adapting faces and the morphs. If there is no
aftereffect, then morph faces, with indeterminate sexual characteristics, should
be rated less attractive than unaltered adapting faces. If there is a face sex
aftereffect, then morphs will have a sexual identity and attractiveness ratings to
them should vary depending on sex of the observer. We expect that morphs
perceived as male will be judged more attractive by females and vice versa.
Finally, as discussed above we predict that morphs considered as male will be
judged better looking by both sexes.
Experiment 1
Method
Participants.
A total of 56 students from Manhattan College and the College of Mount
Saint Vincent participated. There were 16 males and 40 females from three
Sex Aftereffects 8
undergraduate classes. Average age was 19 years. All received extra credit for
their participation and had vision that was normal or corrected to normal.
Apparatus.
Adapting faces were obtained from a University of Michigan Internet face
database, now defunct. The faces were of undergraduate engineering students.
Those used in the experiment were selected on the following criteria: a full
frontal view, no obvious blemishes, no glasses or facial hair, and the absence of
strong shadows or reflections. The images were in color and cropped below the
chin and above the hairline. Facial expression varied slightly.
There were a total of 56 unique adapting faces. An adapting face was
unaltered, it did not undergo any stimulus manipulation. Adapting faces were
presented to observers but also used to generate the morphs. An attempt was
made to balance adapting faces with respect to ethnicity. This resulted in 14
Asian, 20 Black, and 22 Caucasian faces. Given our sample, there were an
insufficient number of adapting faces to create Hispanic morphs, so Hispanic
faces were not used.
We created morph faces by equally combining a single male and a single
female adapting face. This yielded 28 total morphs, seven Asians, 10 Blacks, and
11 Caucasians. The two single adapting faces were superimposed and blended to
create a 50% composite image using the Morpher 2.0 program for the Mac
(Fujimiya, 1997). In Morpher, the user selects nodes, which are corresponding
points on the two input images at high regions of contrast. These nodes on each
image are then connected with lines called boundaries. Morpher uses the
corresponding boundary information on each image to create a composite image,
which is weighted pixel average.
All faces, adapting stimuli and morphs, were presented in color on an
Apple iBook laptop computer. Face dimensions were 4.0 cm in width and 4.5 cm
in height. Participants made judgements at an average viewing distance of 45 cm.
Each face was centered in the middle of the 35.5 cm (14 inch) diagonal length
screen. A 50% homogenous gray background filled the remainder of the viewing
region.
Procedure.
Data collection from the three classes took place in one classroom. Each
class was run in one or two sessions of approximately 15 students each.
Participants first read and signed a consent form then read a detailed set of
written instructions. After this, the research assistant explained the instructions
to them verbally and answered any questions they had about the procedure.
The three conditions in the study were Male Adapt, with male adapt faces
followed by morphs, Female Adapt, with female adapts preceding morphs, and
Sex Aftereffects 9
the No Adapt condition, with morph faces only. The students were randomly
assigned to one of these three conditions, but did not know which condition they
were in. Instructions for each condition were the same, to simply view each face
that appeared and evaluate it, first with respect to sex, then with respect to its
perceived attractiveness.
Each participant viewed the experiment on his or her own monitor. In the
No Adapt condition, morphs appeared for seven seconds. A response screen
then appeared prompting the participant to push the "M" key if they thought it
was a male face and the "F" key if it was perceived as female. A second response
screen then appeared prompting the participant to enter a number between "1"
and "7" to indicate the perceived attractiveness of the face, with higher numbers
indicating greater perceived attractiveness. Both response screens were response
terminated and remained up for as long as it took an individual to decide. The
time to make each decision was recorded. Order of faces was randomized.
The two adapt conditions contained twice as many trials as the No Adapt
conditions, because participants first viewed and evaluated an adapting face and
then its corresponding morph. An adapting face appeared for seven seconds,
followed by the sex and attractiveness rating screens. Then in the next trial the
morph face appeared also for seven seconds and followed by the two rating
screens. The adapt face that preceded a morph was always one of the two
adapting faces that were used to construct it. In the male adapt condition it was
the male face. In the female adapt condition it was the female face. The order of
these adapt-morph trial pairs was randomized.
Results and Discussion
The primary independent variable was Adapt Type, with three levels.
These were Male Adapt, Female Adapt, and the No Adapt control. Adapt Type
was a between-subjects variable. There were two major dependent variables. The
first was sex judgement. This was the observer's perception of whether the faces
were male or female. The second were attractiveness ratings. This was the
perceived attractiveness of the faces based on the seven-point rating scale. The
results of each dependent variable are summarized in separate sections below.
The data was first screened for outliers. All responses greater than seven
seconds, which constituted less than one percent of the data, were excluded from
subsequent analyses.
Sex Judgements.
Chi-square tests were performed separately for morph and adapt faces.
This was done because all three conditions contain morphs but only the male and
female adapt conditions have adapting faces. There was a significant effect of
Adapt Type on the perceived sex of morph faces 2 (2, N = 1,549) = 72.43, p <
Sex Aftereffects 10
0.01. Observers judged the morphs in the No Adapt condition ambiguous with
respect to sex, with a 55% - 45% male - female split. These data demonstrate that
the morph faces were perceived as androgynous, with perhaps a slight male bias.
The results also indicate the presence of a face sex aftereffect. Sixty percent of the
morphs were judged female when preceded by a male adapt, while 67.3% of
them were perceived male when they followed a female adapt. These data are
consistent with our hypothesis that exposure to the adapting faces may fatigue a
sex-specific face representation, allowing the opposite sex characteristics of the
morph to become salient.
As expected, there was a highly significant effect of Adapt Type on sex
judgements for the adapt faces, 2 (1, N = 925) = 812.81, p < 0.01. Adapting faces
were very unambiguous with respect to sex with 97.8% responding male in Male
Adapt and 95.8% responding female for Female Adapt. This demonstrates that
our unaltered adapting stimuli were categorically perceived as distinctly male or
female. Table 1 shows the percentage of male and female responses in the Adapt
Type conditions for both types of face.
Attractiveness Ratings.
As before, analyses were performed separately for adapts and adapting
faces. A one-way ANOVA showed there was a significant effect of Adapt Type
on perceived attractiveness for morphs F(2, 53) = 44.23, p < 0.01. Morphs in the
Female Adapt group were considered most attractive (M = 3.41, SE = 0.14).
Morphs in the Male Adapt condition were rated least attractive (M = 2.63, SE =
0.13). Morphs in the No Adapt control fell in between (M = 3.14, SE = 0.08) All
pair-wise comparisons between these means were significant by Fisher's
Protected Least Significant Difference (PLSD) at an alpha level of 0.05. Notice
that morph faces alone in the No Adapt condition were not rated as less
attractive than all of the adapt faces.
An explanation for the Adapt Type variable results emerges when the
data are broken down by sex of the participant. Women judged morphs in the
Female Adapt group as more attractive (M = 3.59, SE = 0.07) than men (M = 3.04,
SE = 0.12). Because the female adapts masculinized the morphs, heterosexual
women should see them as more attractive. Since there are more than twice as
many women as men in our sample, female judgements are probably exerting a
stronger influence on the data, raising the overall mean in this condition. In the
Male Adapt condition, feminization of the morphs would produce an opposite
effect. Here we would expect men to rate the morphs higher. In fact they don't,
but the mean difference between men (M = 2.49, SE = 0.10) and women (M =
2.67, SE = 0.07) is much smaller. This may be due to the limited number of male
participants. In experiments 2 and 3, with a more equal number of observers
from either sex, differences in responding to these conditions disappears.
Sex Aftereffects 11
The separate analysis for adapting faces showed greater perceived
attractiveness in the Female Adapt condition (M = 3.35, SE = 0.06) than for Male
Adapt (M = 2.85, SE = 0.06). This was a significant difference t(32) = 28.54, p <
0.01. When subsequently broken down by sex of participant, it turns out this
larger mean comes from female (M = 3.55, SE = 0.07) rather than male (M = 2.89,
SE = 0.12) observers.
Experiment 2
Aftereffect phenomena require exposure to an adapting stimulus. But the
actual amount of adaptation needed to elicit aftereffects varies. In some cases, it
is only a few seconds. Webster and MacLin (1999) obtained figural aftereffects for
faces with just a few seconds of exposure time. So did Rhodes et al. (2003). In
others, like chromatic adaptation, exposure time may run 45 seconds or higher
(Uchikawa, K, Uchikawa, H., & Boynton, R. M., 1989).
In experiment 1 we obtained aftereffects with as little as seven seconds of
adaptation time. In the next experiment we employ adaptation times both below
and above seven seconds to see if the strength of the aftereffect changes. In
particular, we wish to see if there is a minimum exposure time necessary to elicit
the phenomenon and whether it will increase with an increase in adapting
stimulus view time. Hershenson (1993) found that motion aftereffects are more
powerful the longer one looks at the inducing stimulus. If the face aftereffect is
governed by the same mechanisms as in these well-known aftereffects, we could
expect greater adaptation to produce a stronger face sex aftereffect.
In experiment 2 we employed a short adaptation duration of two seconds,
five seconds below that used in experiment 1 and a longer adapting duration of
12 seconds, five seconds above that used previously. It is unclear what to predict
for the 2-second condition. If this viewing time is too short it may be insufficient
to elicit any type of adaptation and there will be no effect. A 12-second exposure
though, should allow for greater adaptation and so produce an aftereffect
stronger than that seen in experiment 1.
Method
Participants.
A different set of 56 Manhattan College and College of Mount Saint
Vincent undergraduate students participated for extra class credit. There were 27
males and 29 females from three separate psychology courses. Average age was
19.6 years. Vision was normal or corrected to normal.
Procedure.
Sex Aftereffects 12
Unless otherwise stated, the stimuli and procedure used here were
identical to that of experiment 1. There were four between-subject conditions:
Male Adapt 2-seconds, Male Adapt 12-seconds, Female Adapt 2-seconds, and
Female Adapt 12-seconds. These times were deliberately set to be symmetrically
five seconds below and above the experiment 1 duration. Viewing duration of
the morph faces that followed was seven seconds, the same as in the previous
study. Response screens for sex judgement and attractiveness ratings were
response terminated as before. In the male adapt conditions, participants saw
only the male adapt faces. In the female adapt conditions they saw only the
female adapt faces. Order of the adapt-morph trial pairs was randomized.
Participants were randomly assigned to one of the four conditions and were
naive with respect to each condition. Running time for the 2-second versions was
20 minutes. For the 12-second versions, it was 25 minutes.
Results
There were two primary independent variables. These were Adapt Sex
with two levels, Male Adapt and Female Adapt. This was crossed with Adapt
Duration, also with two levels, 2-seconds and 12-seconds. Dependent measures
were the same as in experiment 1. Responses greater than seven seconds were
considered outliers and removed from all analyses. They made up less than one
percent of the data.
Sex Judgements.
The morph data were examined first. There was a significant effect of the
Adapt Sex variable 2 (3, N = 1,607) = 44.75, p < 0.01. As expected, the
distribution of male and female responses to morphs were about equal in the
short viewing conditions. For Male Adapt 2-seconds and Female Adapt 2-
seconds, percent responding to either sex was very close to an even 50 - 50 split.
In the longer viewing conditions, there was again evidence of a face sex
aftereffect. For the Male Adapt 12-second condition, 63.5% of the morphs were
perceived as female. In the Female Adapt 12-second condition, 58.5% of the
morphs were seen as male. Table 2 shows the percentage of male and female
responses for the Adapt Sex and Adapt Duration variables for both types of face.
The Adapt Sex variable had a significant effect on responding to adapting
faces faces as well, 2 (3, N = 1,604) = 1,409.63, p < 0.01. As expected, the adapts
were unambiguous. There was near unanimity in perception of male faces as
male and female faces as female. There was no discernible difference in
responding to these faces as a function of duration, i.e., participants were not
substantially more accurate at longer durations.
These data require some discussion. Our first major prediction, that a
short duration would not elicit the face aftereffect, was fulfilled. Responding in
Sex Aftereffects 13
these cases was identical to a 50 - 50 split. This is consistent with the assumption
than an internal face representation must be adapted to produce the aftereffect.
Two seconds of viewing time is apparently too little time to do so. Our second
prediction, that the aftereffect should be stronger for longer adapting durations,
was not entirely fulfilled. In experiment 1, the Male Adapt condition produced a
60% female response to morphs. The current data show a slight increase to
63.5%. In experiment 1, there were 67.3% male responses to morphs in the
Female Adapt condition. In the current study with longer durations this drops to
58.5%. If one collapses across Adapt Sex, there was an overall greater sex
aftereffect in the first study where duration was seven seconds.
Why is this the case? Twelve seconds is still a fairly short adaptation time
compared to other perceptual aftereffect phenomena. These results suggest a fast
adapting system for faces, one that peaks somewhere between the two values
tested here, but which may decline shortly thereafter. A fast adapting system
may match the rapid pace of social discourse. Important linguistic ideas are often
communicated in just a few seconds. Sex appropriate discourse may thus also
require rapid sexual discrimination. We cannot distinguish here whether this
result is due to a fast-acting adaptation or simply an artifact of the way in which
the visual system works. The exact time-course of this effect warrants additional
investigation. We study it further in the next experiment.
Attractiveness Ratings.
First of all, we note that morph faces overall were not rated significantly
less attractive than adapts, again suggesting that morph faces are not entirely
sexually indeterminate, as would be expected if there were a face sex aftereffect
at work. A (2 x 2) factorial ANOVA with Adapt Sex and Adapt Duration as
factors was performed next for the attractiveness ratings. This was done for
morph faces first. There was no effect of Adapt Sex. Ratings were not
significantly greater for female adapts. There was a significant effect of Adapt
Duration F(1, 44) = 3.57, p < 0.05, with morphs in the longer 12-second duration
judged more attractive (M = 2.91, SE = 0.04) than those in the 2-second duration
(M = 2.84, SE = 0.05). The Adapt Sex by Adapt Duration interaction was not
significant.
This same analysis was next applied to the adapt faces. Adapt Sex and the
Adapt Sex by Adapt Duration interaction were not significant. There was again a
main effect of Adapt Duration F(1,44) = 4.56, p < 0.05. Adapts viewed for 12
seconds were judged more attractive (M = 2.99, SE = 0.04) than those viewed for
two seconds (M = 2.84, SE = 0.05).
Why are the adapting faces more attractive when viewed longer?
According to the mere exposure effect, liking for novel stimuli, including human
faces, increases with repeated exposure (Moreland & Zajonc, 1982). Longer
viewing times may increase familiarity and therefore perceived attractiveness in
Sex Aftereffects 14
these faces. But how can we explain this same result for morphs? They were
viewed for the same time in both the short and long conditions. It is possible that
ratings for adapts may generalize to morphs. When an adapting face is judged
better looking, it increases the probability that the morph, which resembles it,
will also be considered better looking.
Experiment 3
In the first two experiments, observers judged both faces. This was
necessary since it allowed us to compare the relationship between them.
However, judgements to the first face produced a variable time interval between
it and the second face. Across all viewers in experiment 1 average decision time
for the two response screens was a little over two seconds. This variability makes
it difficult to ascertain a pure effect of adapting duration. Although the first face
appears for a fixed time, its subsequent influence may fade, especially at a longer
inter-stimulus interval (ISI).
To remedy this situation, we conduct a third study where the duration
between adapt offset and morph onset is fixed. In fact, we eliminate any ISI
altogether with the morph appearing immediately after the adapt. We predict the
elimination of the ISI should produce stronger aftereffects. We employ the same
adapting durations as in the previous studies: two, seven and 12 seconds.
Experiment 3 thus serves as a controlled replication of the first two experiments.
In addition, an attempt was made this time to deliberately equate the
number of male and female participants in each condition, allowing us to
examine the interaction of this subject variable. The inclusion of participant sex
in the analysis also gives us a means to validate the opposite sex effect. If male
adapts induce perception of a female morph, then heterosexual male viewers
should rate these morphs higher in attractiveness. Conversely, female adapts
ought to induce higher attractiveness ratings of morphs by heterosexual female
viewers if they are perceived as male.
Method
Participants.
A total of 82 undergraduates from Manhattan College and the College of
Mount Saint Vincent participated for extra class credit. There were 36 males and
46 females. Average age was 19.4 years. Vision was normal or corrected to
normal.
Procedure.
Sex Aftereffects 15
Adapts and morphs were the same as those used in the first two studies.
Participants were instructed to view both faces in a trial but to evaluate only the
second face of each pair. Adapt duration varied depending on condition. Morphs
immediately superceded adapting faces. Morph duration remained at seven
seconds. The two response screens followed presentation of the morphs.
Six conditions resulted from the crossing of Adapt Duration with three
levels (2-seconds, 7-seconds, and 12-seconds) and Adapt Sex with two levels
(Male Adapt and Female Adapt). Male and female observers were separately and
randomly assigned to these six conditions. On average there were seven
members of each sex in each of these six groups. All other procedural aspects
matched those used earlier.
Results and Discussion
Adapt Duration and Adapt Sex were the independent variables. Sex
judgement, attractiveness ratings, and RT were dependents and are reported in
each section below. Participant Sex was a subject variable. Outliers were screened
following the same criteria as before. The data in this experiment are responses
to morph faces only.
Sex Judgements.
Chi square analyses for sex judgement were performed separately for
Male Adapt and Female Adapt. There was a significant effect of Adapt Duration
for female faces, X2 (2, N = 1,204) = 10.53, p < 0.01. The effect for male faces was
marginally significant, X2 (2, N = 1,118) = 3.82, p = 0.14. For the 2-second duration
for male and female adapts, responding to morphs approximated a 50 - 50 split.
For the longer duration of 7-seconds and 12-seconds, there is a clear aftereffect at
work in each case approximating a two-thirds split. Table 3 shows percent
responding to male and female adapts for each level of Adapt Duration.
In comparing these values to those obtained with a variable ISI, one sees
little difference. In experiment 2, the 2-second duration with a variable ISI
produced responding indistinguishable from a 50 - 50 split, with deviations away
from this of no more than two percent. In the current study with a fixed and
immediate ISI, the deviations from a perfect 50 - 50 split are a bit larger,
bordering on five percent. These numbers however do not differ substantially
from the No Adapt control data in experiment 1, where there was a 55 - 45 male -
female split. We conclude therefore that a 2-second adapt duration, even with a
zero ISI is insufficient to elicit the effect.
There is also little difference in the judgement data when comparing the
longer adapting durations. The 7-second conditions for male and female adapts
in experiment 1 are identical to those obtained here. For the 12-second duration,
the differences in responding are also negligible, within a few percentage points.
One can conclude from these results that the variable ISIs in the first two
Sex Aftereffects 16
experiments were of such a short nature that they did not allow for any
meaningful loss of the effect.
Additional chi square analyses were next performed for Adapt Duration
and Participant Sex. There were no major differences in the pattern of statistical
significance or of responding by male or female observers. This suggests that the
mechanism underlying the aftereffect is the same in both sexes. If the putative
reason for such a mechanism is the ability to make fine-tuned discriminations
between members of each sex so as to functionally adjust one's social behavior,
then there is no reason to expect any such difference. Members of both sexes
have an equally important need to recognize and identify the sex to which a
given individual belongs. The primary reasons for this identification are the
same in men and women, namely threat and mate assessment.
To summarize, the data from this experiment replicate quite well the
equivalent conditions from experiments 1 and 2. The pattern of response for
judging the sex of morph faces is the same in all cases. There is an absence of any
aftereffect at short 2-second intervals, an apparent asymptote at 7-seconds, and a
leveling off of the effect at adapt durations greater than this. In addition, there is
no discernible difference in the face sex aftereffect for male and female observers.
Attractiveness Ratings.
A factorial ANOVA with Adapt Sex and Participant Sex as factors was
calculated in order to examine any interaction. There was no main effect of
Adapt Sex. There was a significant main effect of Participant Sex, F(1, 54) = 5.70,
p < 0.05. Female observers judged the morphs more attractive (M = 3.02, SE =
0.04) than males (M = 2.87, SE = 0.04). The interaction between Adapt Sex and
Participant Sex was also significant, F(1, 54) = 54.36, p < 0.01.
The shape of the interaction between sex of adapt and participant is in the
form of a cross over. In the Female Adapt condition where morphs are perceived
as male, females rate them more attractive (M = 3.24, SE = 0.05) than males (M =
2.70, SE = 0.06). In the Male Adapt condition where morphs are perceived as
female, males rate them higher (M = 3.05, SE = 0.06) than females (M = 2.75, SE =
0.05). This finding nicely validates the aftereffect. It shows that in a primarily
heterosexual subject pool, morphs biased toward male are considered more
attractive by females. Conversely, morphs biased toward female are rated more
attractive by males. This validation is only appropriate here in this experiment,
where the number of observers of each sex were equated.
General Discussion
The Face Sex Aftereffect
Sex Aftereffects 17
In the introduction, we hypothesized the existence of a face sex aftereffect,
where exposure to a face of one sex would make a subsequently appearing
androgynous morph appear to be the opposite sex. The two instances of this are
viewing a male adapting face and perceiving the morph as female and viewing a
female adapting face and perceiving the morph as male. These specific
predictions were confirmed in three separate experiments under a variety of
different conditions, showing the phenomenon is fairly robust.
We also manipulated duration of the adapting face, expecting that longer
durations would produce a stronger aftereffect. This prediction was partly
confirmed. The results of experiments 1 and 2 show no aftereffect at a short 2-
second duration, a strong effect at a moderate 7-second duration, but no
continued increase in the magnitude of the effect at a longer 12-second duration.
We conclude that 2-seconds is insufficient time to produce the effect but that it
may occur quite quickly thereafter, reaching a peak only a few seconds later. The
speed of this adaptation, as discussed earlier, may be related to the rapid
requirement for sexual identification in social discourse. Future research might
investigate aftereffect strength as a function of shorter intervals in the range we
used here as well as at longer durations beyond 12-seconds.
In the first two experiments, observers judged adapting faces and morphs.
This introduced a variable ISI between faces that could have mitigated the
aftereffect. To see if this was the case we conducted a third experiment with a
fixed and immediate ISI. The magnitude of the aftereffect remained unchanged,
showing that a spacing of several seconds in between faces is not enough to
diminish the phenomenon. Experiments 1 and 2 did not control for sex of the
observer. So in experiment 3 we tested equal numbers of males and females,
finding no difference in the aftereffect for participant sex. We conclude based on
these results that the aftereffect is the same in both sexes because its functional
significance for both sexes is the same.
We reiterate the reason for such a mechanism as stated in the introduction.
Appropriate social interaction requires unambiguous sexual identification.
Adaptation to a local population of sexual face characteristics normalizes an
individual's responses to that group. This reduces ambiguity and promotes
adaptive sex-specific social behavior. Our data suggest that time spent among
males for example will renormalize an internal face sex dimension so that the
mean is now shifted closer to males. This makes subsequently appearing faces
with smaller amounts of femininity appear more female. Men under these
conditions would be more likely to mate with these females, a response
promoting reproduction and group survival. Women under these conditions
would also perceive these faces as more female and react in a socially
appropriate manner, perhaps through increased cooperative interaction that
would also promote group survival. If this explanation is correct, then the face
normalization process we report here may be the result of evolutionary selection
pressures.
Sex Aftereffects 18
What kind of mechanism underlies this phenomenon? Eifuku et. al. (2004)
have found that the anterior inferior temporal gyrus (ITG) in macaque monkeys
is involved in judgments of facial identity in conjunction with the superior
temporal sulcus (STS). In humans the STS is implicated in processing of facial
movement and expression both of which are social cues (Allison, Puce, &
McCarthy, 2000). It is possible neurons in these or other regions code for sexual
identity of faces and that these cells might have opponent relations. If this were
the case, activation in the cell population coding for one sex might suppress
activity in neurons coding for the opposite sex. This could mediate the effect we
have discovered here.
We want to emphasize that the above explanation is speculative. This
investigation does not address the neural substrate of the phenomenon. It is still
not known if the aftereffect is mitigated by mutual inhibition between neural face
detectors or is the result of some other process. This study only demonstrates the
existence of the aftereffect and does not attempt to explicate the brain
mechanisms that give rise to it. There is in fact some doubt about the exact nature
of any such mechanism. The effect seems to be mediated by high-level neurons
that are not tied to retinotopic input, since Rhodes et al. (2003) obtained
aftereffects with a 90 change in orientation between test and adapting faces.
Another question is whether the phenomenon is even specific to faces at all. The
face adaptation effects obtained here and elsewhere may be the result of a
general-purpose perceptual process that mediates a wide variety of different
stimuli.
Attractiveness Ratings
We made several speculations with regard to attractiveness in the
introduction. First, we stated that morphs might not be rated less attractive than
the adapting faces if they were perceived as male or female. The data confirm
this. In the first two experiments, there was no significant difference in ratings
between adapts and morphs. This is what one would expect if the morphs were
not perceived as androgynous because of the influence of the preceding faces.
Second, we used the dependent measure of attractiveness to validate the
face sex aftereffect. Assuming a primarily heterosexual subject pool, we
predicted that males should judge morphs higher in the Male Adapt condition
where the morphs are perceived as female. Alternatively, females should judge
morphs higher in the Female Adapt condition that biases morphs as male.
Overall though, there should be no difference in ratings for the Adapt Sex
variable across both sexes, since opposing male and female preferences should
cancel each other out. These predictions were confirmed in experiments 2 and 3
where approximately equal numbers of male and female viewers were tested.
They were to some extent also corroborated in experiment 1. In that study,
Sex Aftereffects 19
ratings were higher in the Female Adapt condition that seems to have been due
to the large number of female viewers.
In the introduction, we predicted that morphs perceived as male would be
judged more attractive than morphs perceived as female. The literature shows
observers consider feminized faces more attractive, perhaps because these
features signal cooperativeness and other socially desirable characteristics. Was
this the case in our study? In experiments 1 and 3, there was a slight tendency to
judge morphs perceived as male more attractive. In experiment 2, this trend was
reversed. In all cases, the mean differences were not significant. We conclude
then that morphs of the sort we use here are not considered more attractive when
judged male.
Future Prospects
There are a few limitations in the current study. The aftereffect
demonstrated here is specific to the faces we used. In our stimuli, the adapting
face constituted half of the morph. It remains to be seen if the aftereffect is more
general, linked to a generic representation of maleness or femaleness and not to
any other particular facial characteristics. Recall that the figural aftereffects
obtained by Webster and MacLin (1999) transferred across different faces. These
results imply that aftereffects for sexual identity may also be general. In any
event, this hypothesis can be tested by constructing morphs from male and
female faces that are not used as adapting stimuli in the same experiment. If the
aftereffect is undiminished then the neural representations being adapted are not
tied to any individual configuration of facial features and instead code for an
abstract form of sexual identity. If the aftereffect is diminished, one could then
vary the similarity of the adapting face to its morph counterpart to see which face
attributes the effect may be linked. This would be an interesting investigation, as
it addresses the question of whether sexual identity is a separable or integral face
dimension (Garner, 1974).
We should mention here that morphs have limited ecological validity,
since faces of this sort are so rarely encountered in everyday life. It is of course
possible to evaluate the aftereffect without using morphs at all. One could do this
by presenting a male or female adapt and then following it with a normal
unaltered male or female face. Participants instead of judging sex would rate the
masculinity or femininity of the second face. Our data predict that male adapting
faces would feminize the ensuing faces while female adapts would masculinize
them.
Sex Aftereffects 20
References
Abramov, I., & Gordon, J. (1994). Color appearance: On seeing red, or
yellow, or green, or blue. Annual Review of Psychology, 45, 451-485.
Allison, T., Puce, A., & McCarthy, G. (2000). Social perception from visual cues:
Role of the STS region. Trends in Cognitive Sciences, 4, 267-278.
Berry, D. S., & Wero, J. L. F. (1993). Accuracy in face perception: A view
from ecological psychology. Journal of Personality, 61, 497-520.
Eifuku, S., De Souza, W. C., Tamura, R., Nishijo, H., & Ono, T. (2004). Neuronal
correlates of face identification in the monkey anterior temporal cortical
areas. Journal of Neurophysiology, 91, 358-371.
Enlow, D. H. (1990). Facial growth (3rd ed.). Philadelphia: Harcourt
Brace Jovanovich.
Folstad, I., & Karter, A. J. (1992). Parasites, bright males and the
immunocompetence handicap. American Naturalist, 139, 603-622.
Fujimiya, M. (1997). Morpher for Macintosh. Retrieved August 23, 2001,
from http://www.asahi-net.or.jp/~FX6M-FJMY/mop00e.html.
Garner, W. (1974). The processing of information and structure. New Jersey:
Lawrence Erlbaum.
Hershenson, M. (1993). Linear and rotational motion aftereffects as a
function of inspection duration. Vision Research, 33(14), 1913-1919.
Kanda, N., Tsuchida, T., & Tamaki, K. (1996). Testosterone inhibits
immunoglobulin production by human peripheral blood
Sex Aftereffects 21
mononuclear cells. Clinical and Experimental Immunology, 106, 410-
415.
Kaping, D., Bilson, A.C., & Webster, M. A. (2002, November). Adaptation and
categorical judgments of faces. Poster presented at the Object Perception
and Memory Conference, Kansas City, MO.
Leopold, D. A., O'Toole, A.J., Vetter, T., & Blanz, V. (2001). Prototype-referenced
shape encoding revealed by high-level aftereffects. Nature Neuroscience,
4, 89-94.
Little, A. C., & Hancock, P. J. B. (2002). The role of masculinity and
distinctiveness in judgments of human male facial attractiveness.
British Journal of Pychology, 93, 451-464.
Moreland, R. L., & Zajonc, R. B. (1982). Exposure effects in person
perception: Familiarity, similarity, and attraction. Journal of
Experimental Social Psychology, 18, 395-415.
Penton-Voak, I., Perrett, D. (2000). Female preference for male faces
changes cyclically-further evidence. Evolution and Human
Behaviour, 21, 39-48.
Perrett, D. I., Lee, K. J., Penton-Voak, I. S, Rowland, D. R., 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.
Sex Aftereffects 22
Rhodes, G., Hickford, C., & Jeffery, L. (2000). Sex-typicality and
attractiveness: Are supermale and superfemale faces
superattractive? British Journal of Psychology, 91, 125-140.
Rhodes, G., Jeffery, L., Watson, T. L., Clifford, C. W. G., & Nakayama, K. (2003).
Fitting the mind to the world: Face adaptation and attractiveness
aftereffects. Psychological Science, 14(6), 558-566.
Russell, J. A., & Carroll, J. M. (1999). On the bipolarity of positive and
negative affect. Psychological Bulletin, 125(1), 3-30.
Singh, D. (1993). Body shape and women's attractiveness-the critical role
of waist-to-hip ratio. Human Nature, 4, 297-321.
Swaddle, J. P., & Reierson, G. W. (2002). Testosterone increases perceived
dominance but not attractiveness in human males. Proceedings of
the Royal Society of London. Series B., 269, 2285-2289.
Uchikawa, K., Uchikawa, H., & Boynton, R. M. (1989). Partial color
constancy of isolated surface colors examined by a color-naming
method. Perception, 18, 83-91.
Webster, M. A., & MacLin, O. H. (1999). Figural aftereffects in the
perception of faces. Psychonomic Bulletin & Review, 6, 647-653.
Sex Aftereffects 23
Author Note
Jay Friedenberg, Megan Kwasniak, and Christine Haber, Department of
Psychology, Manhattan College.
Preliminary results of the first two experiments were presented at the 43rd
Annual Meeting of the Psychonomic Society, November 21 – November 24, 2002,
in Kansas City, Missouri.
Correspondence concerning this article should be addressed to Jay
Friedenberg, Department of Psychology, Manhattan College, Manhattan College
Parkway, Riverdale, New York 10471. Electronic mail may be sent via Internet to
Jay.Friedenberg@Manhattan.Edu
Sex Aftereffects 24
Table 1
Percent responding to morph and adapting faces in experiment 1 for type of
adapt
______________________________________________________________________
Adapt Type
____________
Response Female Adapt Male Adapt No Adapt
_____________________________________________________________________
Morph Faces
Male 67.4 39.9 55.2
Female 32.6 60.1 44.8
Adapt Faces
Male 4.2 97.9 N.A.
Female 95.8 2.1 N.A.
_____________________________________________________________________
Sex Aftereffects 25
Table 2
Percent responding to morph and adapting faces in experiment 2 as a function of
adapt sex and duration
_____________________________________________________________________
Adapt Sex
Female Adapt Male Adapt
______________ ____________
Adapt Duration 2 12 2 12
(seconds)
_____________________________________________________________________
Morph Faces
Male 48.2 58.5 50.5 36.6
Female 51.8 42.2 49.5 63.4
Adapt Faces
Male 2.7 4.7 97.0 97.9
Female 97.3 95.3 3.0 2.1
_____________________________________________________________________
Sex Aftereffects 26
Table 3
Percent responding to morph faces in experiment 3 as a function of adapt sex and
duration
______________________________________________________________________
Adapt Sex
Female Adapt Male Adapt
_____________ ____________
Adapt Duration 2 7 12 2 7 12
(seconds)
_____________________________________________________________________
Response
Male 56.5 66.7 65.4 45.2 39.5 38.7
Female 43.5 33.3 34.6 54.8 60.5 61.3
_____________________________________________________________________
Sex Aftereffects 27
Sex Aftereffects 28
Sex Aftereffects 29