research and developement

1. Faces and sex
Thomas R. Stewart, Ph.D.
Center for Policy Research
Rockefeller College of Public Affairs and Policy
University at Albany
State University of New York
T.STEWART@ALBANY.EDU
Public Administration and Policy
PAD634 Seminar in Judgment and Decision Behavior
Copyright © Thomas R. Stewart

2. 2
Citation
Lynda G. Boothroyd, Benedict C. Jones,
D. Michael Burt, Lisa M. DeBruine,
David I. Perret. Facial correlates of
sociosexuality. Evolution and Human
Behavior xx (2008, in press)
Yahoo news article

3. 3
Sociosexual Orientation Inventory (SOI)
Sociosexual Orientation Inventory (SOI)
developed by Simpson and Gangestad (1991).
The SOI was developed to assess individual
propensity to engage in short term sexual
relationships. A high score on the SOI indicates
an ‘unrestricted’ sociosexuality (i.e., being more
open to short-term sexual relationships) while a
low score indicates a ‘restricted’ sociosexuality
(i.e., being less willing to engage in short-term
sexual relationships).
(see link to the questionnaire on the wiki)

4. 4
Study 1
Study 1 used facial composites of unrestricted (open to
short-term relationships) and restricted (prefer long-
term relationships) individuals. The principle of
averaging images means that randomly varying traits
across two groups will tend towards average in
composites, while traits that are significantly different
between groups will be more clear in composites
(Perrett et al., 1994). Thus, this method is ideal for
detecting subtle structural differences between
groups (e.g., a larger, squarer jaw in the case of
masculine males) and for assessing whether groups
of observers are sensitive to these differences. (p. 2)

5. 5
Male composite pictures from the
paper
Which is
“unrestricted”
composite?
Click here for
answer

6. 6
Caption – male on right is composite of
unrestricted males
Fig. 1. Extended forced choice testing
program as used in Sample 1b
displaying the Male Batch 1 pair
composite images of restricted (left) and
unrestricted (right) individuals. In
Sample 1a, a similar program was used
with the labels next to the buttons.
Return to figure

7. 7
Female composite pictures from the
paper
Which is “unrestricted”
composite? Click here
for anser
Which is more attractive?

8. 8
Caption – Female on right is composite of
unrestricted males
• Fig. 2. Female Batch 1 composites of
restricted (left) and unrestricted (right)
individuals.
Return to figure

9. 9
Study 1 results
Both male and female observers considered the
unrestricted sociosexuality female composites more
attractive than the more restricted female composites
(males assessing longterm: mean=3.976, t20=2.63,
p<.05; males assessing shortterm: mean=3.929,
t20=2.50, pb.05; females assessing same sex
attractiveness: mean=4.238, t20=3.49, p<.01). Female
observers considered composites of restricted males to
be more attractive than composites of unrestricted males
for long-term relationships (mean=2.952, t20=2.20,
p<.05) but showed no significant preference for short-
term relationships (t20=0.55). Male observers had no
preference either between unrestricted and restricted
male composites (t20=0.04). (p. 4)

10. 10
Study 2
… despite the advantages of composites, as
discussed in Study 1 above, one might argue
that composites are too controlled when
considering behavioural attributions and that
observers must be able to detect differences
in real faces if we are to claim that they are
capable of differentiating between social
partners in reality. Therefore, Study 2 was
conducted using real individual faces rather
than composites. (p. 5)

11. 11
Study 2 effect sizes
(underlining added)
There were significant positive correlations between all
perceived restrictedness (i.e., apparent SOI) ratings and all
components of the SOI scores of the photographed
individuals (all rs>0.27, all p<.07, two-tailed), with the exception
of reported and rated number of sexual partners within the last
year, neither of which correlated with any other measure (all
p<.25). Furthermore, individuals' full SOI scores were
significantly correlated with the mean rated SOI across all four
questions (rs=0.367, p<.05). The correlation between SOI score
and mean perceived SOI remained significant after controlling
for individual's age and sex, same-sex ratings of
attractiveness, and masculinity/femininity. Furthermore,
results remained significant when male and female observers'
judgments were analysed separately. There was a trend for
higher SOI to be associated with greater perceived femininity
in female faces (rs=0.399, p=.098, n=23) but no further
significant correlations. (p 5)

12. 12
Judging sexual strategy from faces
- Summary
http://www.boothlab.org/
A research paper currently published in Evolution and Human Behavior shows that we may
be subtly aware of other people’s attitudes to sex. Three groups of undergraduate students
were photographed and completed a questionnaire called the sociosexual orientation
inventory (SOI) which asks about past sexual behaviour (e.g. number of one night stands)
and current attitudes – such as "is sex without love okay"? Observers were then shown
either real individuals faces or ‘average’ images of faces, and it seemed that across the
studies, observers were often able to distinguish between those who scored low on the SOI
(and thus are not generally keen on casual sex) and those with high scores (who tend to
have had more partners and be more comfortable with uncommitted sex).
However, what is far more interesting, is that despite the subtlety of the explicit awareness
of who-thinks/does-what, there is a very strong tendency for women to be attracted to men
who score low on the SOI – i.e. men who are less interested in casual sex. Men have the
opposite preference with female faces; they strongly prefer the ‘high SOI’ women. In fact,
even other women thought that high SOI females were more attractive.
Furthermore, high SOI men were also viewed as looking more masculine. This backs up
previous work which showed that more masculine men were perceived as being less likely
to commit to a long term relationship; now we can see that men who are less likely to be in
a long term relationship (although the questionnaire doesn’t actually ask about that) also
look more masculine. These convergent lines of evidence support the idea that part of the
variation in women’s preferences for male masculinity is due to the negative connotations it
has for long term partnerships.

13. 13
Gangestad et al. (1992) abstract
•Gangestad, S. W., J. A. Simpson, et al. (1992). "Differential accuracy in person
perception across traits: Examination of a functional hypothesis." Journal of Personality
and Social Psychology 62(4): 688-698.
–Although strangers can assess certain traits of unacquainted others
with moderate validity, overall validity is low. Differential validity across
traits may be due to (1) the extent to which targets display valid cues or
(2) the extent to which perceivers validly use cues. A functionalist
perspective suggests that valid cue utilization should vary with how
important the consequences of accurate trait assessment are. It was
predicted from this perspective that perceivers would judge strangers'
sociosexuality more accurately than 3 other traits (social potency, social
closeness, and stress reaction). Perceivers viewed 1-min videotaped
segments of targets being interviewed and rated them on the 4 traits.
Ratings were correlated with target-reported trait measures. As
predicted, perceivers' ratings of male sociosexuality agreed relatively
well with self-reports. This effect was moderated by sex of target and sex
of perceiver.

14. 14
Gangestad et al. (1992) hypothesis
p. 689

15. 15
Gangestad et al. (1992) lens model
diagram
p. 689

16. 16
Gangestad et al. (1992) method
Perceivers viewed 1-min presentations (interviews) of
20 targets in which only visual cues were available to
them. They then rated each target on four traits:
sociosexuality, social potency, social closeness, and
stress reaction. Data were analyzed in two ways.
First, we correlated raters' social judgments with self-
report markers of the traits provided by targets.
Second, we broke overall accuracy into components
specified by the lens model. To do so, we identified
and measured a number of obvious cues that
perceivers could potentially use. (p. 690)

17. 17
Gangestad et al. (1992) advantage of lens
model analysis
Perceivers could have used cues other than
those we measured (e.g., more subtle forms
of nonverbal expressions not reflected in
gross measures such as percentage of time
smiling). Through lens model analyses,
however, we were able to assess the extent
to which perceivers' agreement with target
self-reports were attributable to valid use of
the cues we both did and did not measure. (p.
690)

18. 18
Gangestad et al. (1992)
(behaviors and impressions)
All interviews were viewed by trained raters and
coded for the occurrence of specific behaviors
(e.g., the percentage of time spent smiling, the
percentage of time eye contact with the camera
was maintained) as well as 34 specific
impressions the inteviewees conveyed during
the interview (e.g., how animated they were,
how relaxed they appeared, and how phony
they appeared. (p. 691)

19. 19
Gangestad et. al. results p. 693

20. 20
Gangestad et. al. results p. 693
(physical attractiveness partialed out)

21. 21
Gangestad et al. (1992) impression
factors
• The resultant factors included:
• (a) Social Engagement (marked by high ratings on
engaging, witty, competent, stimulating, disclosing,
drawing attention to self, dominant, captivating, and
skilled, and low ratings on dull);
• (b) Interest (marked by high ratings on flirtatious, inviting,
conveys interest, and bold);
• (c) Comfort (marked by high ratings on relaxed and low
ratings on anxious, self-conscious, inhibited, and shy);
• (d) Provocativeness (marked by high ratings on seductive,
provocative, and sexually appealing); and
• (e) Pretentiousness (marked by high ratings on arrogant
and phony, and low ratings on nice). (p. 691)

22. 22
Gangestad et al. (1992) cues
Two sorts of behavioral cues were measured. First, we
assessed 11 dimensions of the targets' nonverbal behavior
displayed during the presentation (e.g., percentage of time
spent smiling, percentage of time maintaining eye contact,
and number of laughs). Five principal components are
known to account for the bulk of the variance underlying
these 11 behavioral indicators. We treated these
components as composite behavioral cues. Second, we
assessed global impressions created by the targets during
their presentations. Independent raters evaluated each
target on a heterogeneous set of 34 adjective descriptors
(e.g., animated, inhibited, and relaxed). Five interpretable
factors underlie these descrip tors (Simpson et al., 199 l),
and we treated these five dimensions as impressionistic
cues.

23. 23
Gangestad et. al.
results p. 694
(lens model analysis)
(rc is G)
… we broke overall accuracy
into components specified by
the lens model. To do so, we
identified and measured a
number of obvious cues that
perceivers could potentially
use. (p. 690)
Cues included 5 impression
factors and 5 non-verbal
behavioral factors (Table 1).
Note on Table 4

24. 24
Note on table 4
Return to Table 4

25. 25
Discussion (p. 695)
We predicted that, relative to three other prominent traits, social
perceivers could validly assess the trait of sociosexuality on the basis of
brief exposure to previously unacquainted individuals. Results supported
our predictions. In general, social perceivers' ratings of sociosexuality
corresponded with target self-reports to a greater extent than did their
ratings of social potency, social closeness, and stress reaction.
It seems reasonable to suggest that the enhanced target-perceiver
agreement on sociosexuality was in fact attributable to relatively superior
validity of perceivers' ratings. The obvious alternative explanation that
enhanced agreement was due to relatively superior validity of target self-
reports on sociosexuality seems implausible. The reliabilities of the target
self-reports on social potency, social closeness, and stress reaction all
exceeded the reliability of the SOI, and although reliability does not imply
validity, all three trait measures have been shown to relate substantially to
peer report.
In addition to the overall superiority of perceiver ratings of sociosexuality,
however, we found that perceiver-target agreement on sociosexuality was
moderated by sex of target and sex of perceiver. Perceivers could assess
male sociosexuality with greater accuracy than female sociosexuality
Furthermore, male perceivers demonstrated superior assessment of
sociosexuality relative to female perceivers.

26. 26
Discussion – lens model (p. 695)
Lens model analyses designed to identify the locus of
enhanced perceiver-target agreement on sociosexuality
revealed that perceiver assessments of sociosexuality
were not superior because perceivers utilized cues we
measured more validly (even though perceivers did
utilize these cues in valid ways when assessing male
targets). Rather, the superiority of perceivers' ratings of
sociosexuality primarily stemmed from their utilization
of cues not captured by our measures. The explicit
cues we assessed were relatively gross motor ones
(e.g, time spent smiling and time leaning forward).
Perceivers may have utilized cues involving subtle
microfacial expressions and body movements to validly
assess sociosexuality