intro & lit review
jesswhyte156
JudgingaBookByItsCover-AreFirstImpressionsAccurate-.pdf
University of Colorado, Boulder CU Scholar
Undergraduate Honors Theses Honors Program
Spring 2012
Judging a Book By Its Cover: Are First Impressions Accurate? Tess Adams University of Colorado Boulder
Follow this and additional works at: https://scholar.colorado.edu/honr_theses
This Thesis is brought to you for free and open access by Honors Program at CU Scholar. It has been accepted for inclusion in Undergraduate Honors Theses by an authorized administrator of CU Scholar. For more information, please contact [email protected].
Recommended Citation Adams, Tess, "Judging a Book By Its Cover: Are First Impressions Accurate?" (2012). Undergraduate Honors Theses. 281. https://scholar.colorado.edu/honr_theses/281
JUDGING A BOOK BY ITS COVER 1
Judging a Book By Its Cover: Are First Impressions Accurate?
Tess Adams
Advisor: Dr. Matthew C. Keller
Department of Psychology and Neuroscience
University of Colorado Boulder
Undergraduate Honors Thesis
Committee Members:
Dr. Matthew C. Keller: Department of Psychology and Neuroscience
Dr. Vijay Mittal: Department of Psychology and Neuroscience
Dr. Douglas Duncan: Department of Astrophysical and Planetary Sciences
JUDGING A BOOK BY ITS COVER 2
Abstract
First impressions are integral to human interactions, and
philosophers and scientists have long discussed the idea that the face is a
window into our internal traits. We make judgments of character based on
appearance daily, consciously and subconsciously. Explanations for this
phenomenon include the attractiveness stereotype, self-‐fulfilling prophecies,
or “good genes” hypotheses from evolutionary psychology, but there have
been mixed findings regarding the accuracy of such judgments. The current
study investigates correlations between three subjectively judged “internal”
traits and objective measures of Intelligence, Extraversion, and Neuroticism
on 1600 subjects. We regressed these objective measures on their respective
subjective ratings and controlled for several potential mediating factors. We
found that Intelligence can be judged accurately even when controlling for
potential mediators including attractiveness, SES, and perceived grooming,
and ethnicity. Extraversion can also be judged accurately, but appears to be
mediated by attractiveness, grooming, smiling, and socioeconomic status.
Judgments of Neuroticism, on the other hand, could not be predicted by
subjective ratings. This suggests that we can pick up on valid cues towards a
person’s internal traits without seeing any of their interactions.
JUDGING A BOOK BY ITS COVER 3
Judging a Book By Its Cover: Are First Impressions Accurate?
People make subjective judgments about others on a regular basis,
consciously and subconsciously. But how much information can actually be gleaned
from a glance at a face? The idea that internal traits can be displayed externally
dates back at least to Aristotle, who states, “It is possible to infer character from
features” (Prior Analytics, 2.27). In the late 1700’s, Johann Kaspar Lavater, a Swiss
pastor, published a series of essays on this ideal – known as physiognomy – which
gained a great following into the 19th century. The shape of the nose, the set of the
jaw, the width of the forehead – all were key to understanding whether a person
would be well-‐suited to a particular occupation because those physical traits were
directly linked to intelligence, or kindness, or perseverance. Such judgments are
based on stable traits and facial characteristics, not on fleeting expressions,
emotions, or interactions.
In Darwin’s time, physiognomy was accepted as fact, and he refers to it
throughout his journal in relation to native peoples he met on his travels. Darwin
ran into trouble himself though, when applying to be the “adventurous young man”
accompanying Captain Fitz-‐Roy on the HMS Beagle. “Afterwards… I heard that I had
run a very narrow risk of being rejected, on account of the shape of my nose! He was
an ardent disciple of Lavater, and was convinced that he could judge of a man’s
character by the outline of his features; and he doubted whether anyone with my
nose could possess sufficient energy and determination for the voyage. But I think
he was afterwards well satisfied that my nose had spoken falsely.” (Darwin,
Autobiography, 72).
JUDGING A BOOK BY ITS COVER 4
Physiognomy fell out of favor in the late 19th Century due to its association
with Phrenology – the notion that one’s personality could be found by reading the
bumps on his or her skull, which represented certain areas of the brain being larger
or smaller. Upon opening the skull, scientists discovered that the inside of the skull
is smooth – so bumps could not possibly represent areas of the brain – and thus
phrenology was discredited, and it’s cousin physiognomy along with it.
More recent scientific studies have once again begun looking into whether
subjective impressions based on facial characteristics have any validity. People do
form global and specific trait impressions automatically based on facial structure
(Hassin & Trope, 2000). A study by Willis and Todorov (2006) found that these first
impressions are made after a mere tenth of a second exposure to a face. 42 raters
answered a questionnaire on each of 70 faces (presented as standardized
photographs with neutral expressions). The authors discovered that judgments
made after a 100-‐ms exposure to a face did not differ significantly from those made
with no time constraint. Their result held true for attractiveness, likeability,
trustworthiness, competence, and aggressiveness (Willis, 2006). The results from
Hassin and Willis suggest that we infer personality traits from facial appearance
quickly and uncontrollably, which constantly affects our social interactions whether
we are aware of it or not.
Social psychological studies have drawn attention to the attractiveness
stereotype – a phenomenon known as the “Halo Effect”. The halo effect posits that
we automatically assign positive traits to more attractive people; if a person is
attractive, we also deem them more likely to be nice, intelligent, successful, and
JUDGING A BOOK BY ITS COVER 5
outgoing. Dion et al. point out that physical appearance is the “personal
characteristic most obvious and accessible to others in social interaction”(1971).
The question of the self-‐fulfilling prophecy then arises – do personality traits affect
or reflect appearance, or does appearance mold personality? The authors found that
attractive people were assumed to be more likely to lead happy successful lives, in
all realms from the dating world to the professional world (Dion et al, 1971).
In 2002, Zebrowitz et al. looked into the accuracy of estimating IQ from facial
photos, taking into account several past studies that had mixed results. They
performed a meta-‐analysis on seven perceived intelligence/ measured intelligence
studies from the first half of the twentieth century. Raters rated the intelligence of
subjects from facial photographs. The average “accuracy” (the correlation between
measured intelligence and perceived intelligence) was 0.3, but ranged between 0.07
and 0.7 depending on the study. (Zebrowitz, 2002). The studies took place from
1918 to 2001, with the number of raters ranging from 10 to 1,530, and number of
targets ranging from 10 to 150. The varied characteristics of these studies may
account for the large range in results.
The authors cited the halo effect as a possible explanation for the successful
judgments, but went on to note that evolutionary and social expectancy theories
may predict that attractiveness is associated with actual intelligence (Zebrowitz,
2002). The evolutionary theory would suggest that attractiveness is a way of
broadcasting “good genes”, including higher intelligence – the offspring of more
intelligent mates may be more likely to survive, so traits that display intelligence
would be seen as attractive. Zebrowitz also discussed the potential mediating
JUDGING A BOOK BY ITS COVER 6
effects of grooming, nutrition, and healthcare in the relationship between
attractiveness and intelligence in the context of socioeconomic status.
Socioeconomic status is a good predictor of IQ (citation), and Zebrowitz found that it
is also positively associated with perceptions of attractiveness and intelligence
(Zebrowitz 2002). They suggest that raters used attractiveness to determine SES,
and both attractiveness and SES to determine intelligence. The authors conclude
that people can successfully judge intelligence, and postulate that it is due to the
“valid cue” of attractiveness.
A recent study used composite images to assess accuracy in personality
attribution from looking at faces (Little and Perrett 2007). Many previous studies
regarding personality attribution have involved in-‐person interactions, with or
without verbal communications (rather than photographs only). The ability to
accurately assign personality characteristics without in-‐person interactions is
known as “zero acquaintance”. Little and Perrett stated that this accuracy is found
cross-‐culturally and regardless of medium – photograph, video, or observations.
They formed composite images of people who scored either high or low on self-‐
report measures of personality because the authors thought common characteristics
would be highlighted, and non-‐shared characteristics would disappear by being
averaged out (Little and Perrett, 2007). The photographs used to create the
composites were taken with strict criteria: photos included only the face against a
constant background, and participants posed with neutral expressions, no glasses,
hair pulled back, and clean-‐shaven. The authors found significant agreement
between subjective and self-‐report personality scores for agreeableness,
JUDGING A BOOK BY ITS COVER 7
conscientiousness, and extraversion. These results suggest that faces hold accurate
cues to personality.
Penton-‐Voak et al. (2006) also studied personality judgments from natural
and composite faces. They found that the perceptions of traits formed from
composite faces (based on scoring either high or low on a self-‐report personality
test) were more accurate than those formed from an individual’s face. Perceptions
of extraversion and agreeableness had the strongest relationships with the self-‐
report test measures, and emotional sensitivity had a relationship only in males.
There was a high degree of consensus in ratings, but the authors stated that the
overall validity of the judgments were “unclear and somewhat controversial”
(Penton-‐Voak, 2006).
The current study seeks to test the validity of the link between perceived
internal traits and their objectively measured counterparts. We investigated three
traits: Intelligence, Extraversion, and Neuroticism. Each of these has a subjective
and objective measure, the subjective measures being perceived intelligence,
perceived extraversion, and perceived emotional sensitivity, and the objective
measures being IQ scores and self-‐report personality test scores. First, we tested to
see if the results of subjective impressions correlate with objective measures of the
same trait. We then examined whether any relationships remained after controlling
for potential mediating variables such as attractiveness, sex, grooming, ethnicity,
and socioeconomic status. A positive correlation remaining after regression would
suggest that there is merit in subjective judgments of traits above and beyond
information gleaned from the potential mediating variables.
JUDGING A BOOK BY ITS COVER 8
Method
Participants
The samples for this study were drawn from two twin databases, with 1599
total subjects. The larger set of twins (n= 1357) is from the Genetic Epidemiology
department at the Queensland Institute of Medical Research (QIMR) in Brisbane,
Australia. The other set is from the Longitudinal Twin Study (LTS) at the Institute
for Behavioral Genetics in Boulder, Colorado (n=242).
The gender ratio within both sets is about equal, with 54.7% female and 45.3% male
twins. The twin’s ages range from 15 to 23 years old at the time of the photograph.
Objective data for all twins was collected prior to the current study by the
researchers at QIMR and LTS. To be included in our analyses, each subject needed a
photograph and data regarding IQ, personality test scores, age, height, weight, and
sex. Any twin sets without a full complement of data were excluded from the study.
Photographs
Photographs for LTS twins were cropped from the photographs previously
obtained for the data set. Subjects were taken into a photo room and asked to
remove their shoes, glasses, jackets, and other distracting apparel. Four
photographs were taken against a one-‐inch grid: two full body and two with head
and shoulders only. Participants were asked to maintain a neutral expression.
Finished photos were 29.5 KB in JPG format, and cropped to include face and hair
only.
Photographs of the QIMR subjects were not as tightly controlled because
photographs were intended for identification purposes rather than for subjective
JUDGING A BOOK BY ITS COVER 9
ratings. Participants were allowed jewelry, makeup, jackets, headbands, glasses, etc.
The shots were taken from the shoulder up.
In both sets we excluded photographs in which the subjects were blinking, or
turning their heads. We tilted to upright any photographs in which the subjects
were tilting their heads in Adobe Photoshop in order to maintain continuity
between photographs for the raters. All photos were cropped by research assistants
to include face only, from just below the chin to just above the hair.
Rating Procedure
Ratings for each subjective trait were carried out in the same way. A
computer program displayed photos as a slide show with 50 subjects at a time. Each
group was gender-‐specific, and groups alternated between male and female. The
first slide of each group displayed instructions “In a moment, you are going to rate
the following group of faces on (Trait). But first you will see a slideshow of all the
faces. Use this time to get a sense of the range and variation among the faces for the
trait of (Trait).” In order to obtain a standard distribution of the trait within each
group, raters viewed each face for 2 seconds in a slideshow without rating. Raters
were instructed to make distribution of scores among each set of fifty approximately
uniform. After the slideshow, a screen with the definition of the trait was shown
prior to rating to remind research assistants of what to focus on when assigning
subjective ratings to faces. An example slide (shown with composite face, not one of
our subjects) is shown in Figure 1. Results from each rater’s subjective impressions
were averaged, and the mean was used in the correlation against the actual score.
Cronbach’s α was used to measure inter-‐rater reliability, or how consistent
JUDGING A BOOK BY ITS COVER 10
raters were in their ratings of each subject. It is defined as α = k c/ v+ (k-‐1)c ,
where c is the average of the unique ratings covariances, �v is the average of
the �unique variances and k is the number of raters. See table 1 for Cronbach’s α for
each subjectively measured trait.
Table 1 – Inter-‐rater Reliability
Trait Cronbach’s α Number of Raters
Average Correlation between raters
Intelligence 0.60 7 r = 0.18
Extraversion 0.90 11 r = 0.47
Neuroticism 0.57 10 r = 0.13
Attractiveness 0.87 8 r = 0.45
Grooming 0.70 2 r = 0.54
Smiling 0.90 2 r = 0.82
Acne 0.77 2 r = 0.62
JUDGING A BOOK BY ITS COVER 11
Figure 1
(Composite face obtained from Google Images)
Perceived Intelligence
Perceived intelligence ratings were gathered from four female and three male
raters. Raters were undergraduate research assistants from the University of
Colorado at Boulder. Raters were given the following instruction: “Rate this face’s
intelligence on a scale from 1-‐7, 7 being the most intelligent”.
IQ
IQ scores were obtained from the existing QIMR and LTS twin data sets. Scores are
from the Weschler Adult Intelligence Scale (WAIS).
JUDGING A BOOK BY ITS COVER 12
Perceived Extraversion
Nine female and two male raters rated each face for extraversion. The instructions
read: “Rate this face’s extraversion on a scale of 1-‐7, 7 being most extroverted.” In
order to give a working definition of extraversion that was consistent across raters,
raters were given the following: Extroverted people are more likely to be energetic,
assertive, sociable, talkative, stimulation-‐seeking, action-‐oriented, and enthusiastic.
Introverts tend to be reserved, and have a preference for quieter, less stimulating
environments. Raters also had selections from the JEPQ surveys for extraversion to
give them a more comprehensive working definition of extraversion.
Self-‐Report Extraversion
Self-‐report extraversion scores were obtained from the Junior Eysenck Personality
Questionnaire (JEPQ) personality test, conducted by LTS and QIMR prior to the
current study.
Perceived Neuroticism
Eight female and two male raters rated each face for neuroticism. Raters were asked
to rate “emotional sensitivity” rather than “neuroticism” to avoid bias from the
colloquial use of “neurotic” in society. Instructions read: “Rate this face’s emotional
sensitivity (prone to anxiety, depression, etc.,) on a scale from 1-‐7, 7 being most
emotionally sensitive.” The working definition of emotional sensitivity is the
tendency to easily experience negative emotions. The opposite end of the spectrum
would be emotional stability – people with high emotional stability are calm, less
easily upset, and less likely to experience negative feelings such as anxiety,
JUDGING A BOOK BY ITS COVER 13
depression, self-‐consciousness, and vulnerability. Selections from the JEPQ test were
supplied for neuroticism as well to allow raters to be more consistent.
Self-‐Report Neuroticism
Self-‐report neuroticism scores were obtained from the JEPQ personality test,
conducted by LTS and QIMR prior to this study.
Control Variables
The following variables were examined as potential mediators for any correlations
between subjective and objective measures of the three traits. Undergraduate raters
viewed slideshows as discussed above. The following variables were collected prior
to the current study.
Attractiveness
Eight undergraduate research assistants rated attractiveness on a scale from 1-‐7,
1 being “low attractiveness” and 7 being “high attractiveness“.
Smiling
Photos were rated (n=2 raters) on a scale from one to three, one being “No
Smile”, two being “Partial Smile”, and three being “Full Smile”.
Grooming
Raters (n=2) were asked to decide how much effort the subject had put into their
appearance that day. Photos were rated on a scale of 1-‐7, 1 being “Un-‐groomed”,
and 7 being “Well Groomed”. Grooming is related to attractiveness and may
contribute to the halo effect.
JUDGING A BOOK BY ITS COVER 14
Acne
Photos were rated (n=2 raters) on a scale from 1-‐7, 1 being “No Acne” and 7
being “Heavy Acne”.
Socioeconomic Status
The American Psychological Association defines socioeconomic status as “the
social standing or class of an individual or group. It is often measured as a
combination of education, income and occupation.” (American Psychological
Association, 2012). Research has shown that Socioeconomic Status and IQ are
positively correlated – a higher SES predicts a higher IQ and vice-‐versa. We
therefore controlled for SES to test whether it mediated any of the relationships
between subjective and objective/self-‐report measures.
Genomic Principal Component Scores
Although our sample was almost exclusively Caucasian, we wanted to assess
whether subtle ethnic differences might mediate any potential effects observed.
To do this, we included genomic principal components in our regression
analyses to control for any subtle ethnicity differences between subjects that
may have mediated our results. Both QIMR and LTS twins had been previously
genotyped on genome-‐wide platforms for unrelated studies. Such genome-‐wide
data can be used to accurately estimate subtle ethnic differences between people
using a principal components analysis conducted on the genomic relationship
matrix (derived from ~ 100,000 single nucleotide polymorphisms in roughly
linkage equilibrium). We included the first five principal components as
covariates in our regression analyses.
JUDGING A BOOK BY ITS COVER 15
Statistical Analyses
To examine the relationships between perceived and measured evaluations
of intelligence, extraversion, and neuroticism, we performed correlation and
regression analyses using the R statistical package. Because the subjects were all
twins and siblings, statistical tests conducted on the entire sample would yield
biased (too low) p-‐values due to the dependencies in the data. We therefore split the
subjects into two groups such that only one family member was randomly selected
to be in each dataset. All analyses were run twice – once with group one (n= 730)
and once with group two (n= 717) – thus creating an in-‐study pseudo-‐replication.
The datasets were not truly independent because the second sample contained
individuals from the same family (co-‐twins or siblings), and twins and siblings are
inherently similar (especially in the case of monozygotic twins – these subjects look
identical and will most likely receive similar ratings).
Beyond the initial correlations between the rated trait and its objectively
measured counterpart, we performed a multiple regression analysis to control for
factors that might explain the basic correlations. These factors are age, sex,
grooming, smiling, BMI, acne, socioeconomic status, and ethnicity (as defined
above).
JUDGING A BOOK BY ITS COVER 16
Results
Intelligence
The zero-‐order correlation between Perceived Intelligence and Measured IQ
was r = 0.161 (p = 9e-‐6, df=758) for group one and r = .105 (p< 0.006, df =691) for
group two. When we look at Measured IQ as predicted by Subjective Intelligence
score accounting for attractiveness, age, sex, grooming, smiling, BMI, and acne, the
partial correlation increased, with an r of 0.37, (p =2.28e-‐5, df=693) for group one
and r= 0.32, (p = 1.26e-‐11, df =651) for group two. None of the variables we
controlled for had significant effects. We residualized IQ based on Principal
Components in order to leave only the portion of intelligence not related to genetic
differences. Residual ratings are the degree to which the predicted rating varies
from the actual rating. After taking principal components and SES into account, the
correlation remains about the same, r = 0.356 (p = 5.081e-‐9, df = 434).
Table 2 -‐ Bivariate correlations between subjective intelligence and potential mediating factors IQ Subj. Int Groom Smile Acne Attr. SEI
IQ 1.00 0.10 0.03 0.08 -‐0.01 0.02 0.27
Subj. Int. 0.10 1.00 0.13 0.36 -‐0.04 0.28 0.14
Groom 0.03 0.13 1.00 0.05 -‐0.27 0.63 0.08
Smile 0.08 0.36 0.05 1.00 0.00 0.09 0.06
Acne -‐0.01 -‐0.04 -‐0.27 0.00 1.00 -‐0.40 -‐0.08
Attr. 0.02 0.28 0.63 0.09 -‐0.40 1.00 0.14
SEI 0.27 0.14 0.08 0.06 -‐0.08 0.14 1.00
JUDGING A BOOK BY ITS COVER 17
Predicting IQ From Subjective Intelligence Rating
Table 3 – Bivariate correlations between subjective extraversion and potential mediating factors JEPQ Score Subj.
Extr.
Groom Smile Acne Attr. SEI
JEPQ Score 1.00 0.13 0.12 0.01 -‐0.02 0.14 -‐0.04
Subj. Extr. 0.13 1.00 0.40 0.64 -‐0.13 0.51 0.13
Groom 0.12 0.40 1.00 0.05 -‐0.27 0.63 0.08
Smile 0.01 0.64 0.05 1.00 0.00 0.09 0.06
Acne -‐0.02 -‐0.13 -‐0.27 0.00 1.00 -‐0.40 -‐0.08
Attr. 0.14 0.51 0.63 0.09 -‐0.40 1.00 0.14
SEI -‐0.04 0.13 0.08 0.06 -‐0.08 0.14 1.00
JUDGING A BOOK BY ITS COVER 18
Extraversion
The zero-‐order correlation between Perceived Extraversion and Measured
Extraversion from the Junior Eysenck Personality Questionnaire (JEPQ) is r= 0.163,
(p = 0.0003004, df =488) for group one. For group two, r = 0.196 (p = 1.656e-‐05, df
=473). When we examined Measured Extraversion as predicted by the Perceived
Extraversion score accounting for grooming, smiling, and attractiveness, the
correlation lost much of its significance, suggesting that those mediators were a very
strong influence in the raters’ perceptions of extraversion (r=0.111, df= 485,
p=0.01452). When socioeconomic status was included, the correlation dropped to r
= 0.08187075 (df = 369, p = 0.1154). Socioeconomic status appears influence
subjective extraversion separately from grooming, smiling, and attractiveness,
which can be viewed as a “self-‐presentation” effect.
Predicting JEPQ Scores from Subjectively Rated Extraversion
JUDGING A BOOK BY ITS COVER 19
Neuroticism
We found no correlation between measured JEPQ Neuroticism scores and Subjective
Emotional Sensitivity scores. The correlation was r =0.003 (df = 1063, p-‐value =
0.928). After controlling for BMI, sex, zygosity, age, grooming, smiling, acne, and
socioeconomic status, the correlation remains insignificant: r = -‐0.003 (df = 369, p-‐
value = 0.9608). The bivariate correlations suggest that subjective neuroticism is
influenced negatively by grooming, smiling, and attractiveness, and positively by
acne. However, none of the mediators correlate with self-‐report neuroticism score
to a meaningful degree.
Predicting JEPQ Scores from Subjective Neuroticism
JUDGING A BOOK BY ITS COVER 20
Table 4 – Bivariate correlations between subjective neuroticism and potential mediating factors
JEPQ Score Subj. Neur. Groom Smile Acne Attr. SEI
JEPQ Score 1.00 -‐0.05 0.00 0.06 -‐0.08 0.03 -‐0.06
Subj. Neur. -‐0.05 1.00 -‐0.35 -‐0.39 0.29 -‐0.43 -‐0.07
Groom 0.00 -‐0.35 1.00 0.05 -‐0.27 0.63 0.08
Smile 0.06 -‐0.39 0.05 1.00 0.00 0.09 0.06
Acne -‐0.08 0.29 -‐0.27 0.00 1.00 -‐0.40 -‐0.08
Attr. 0.03 -‐0.43 0.63 0.09 -‐0.40 1.00 0.14
SEI -‐0.06 -‐0.07 0.08 0.06 -‐0.08 0.14 1.00
Discussion
Our study builds on previous findings, and show that raters can judge
‘internal’ behavioral traits at levels above chance simply from brief assessments of
photographs. Our results further suggest that the halo effect cannot explain these
judgments because the correlation between subjective and objective measures of
intelligence increased when we controlled for attractiveness. Clearly, raters can
derive information from a face that is not mediated by traditional physical
attractiveness.
Although the correlations between the subjective and objective measures are
small, the p-‐values are still statistically significant, meaning these results are very
unlikely to have arisen by chance. Small correlations imply that accurate
information about extraversion and intelligence is available in photographs of
JUDGING A BOOK BY ITS COVER 21
people's faces, but there is not much of it. A large number of observations increase
the power to detect small but real effects, as well as the likelihood that these results
accurately estimate the relationships between people's internal traits and
observers' ability to assess them in photographs.
For intelligence, the objective-‐subjective correlation (r = 0.316) is consistent
with the result from Zebrowitz’s meta-‐analysis, which found that the average
correlation between perceived and actual intelligence across studies is 0.3.
Zebrowitz concluded that the correlation is likely due to the halo effect, meaning
attractiveness can explain the accuracy in predicting IQ from subjective intelligence
(2002). However, our results showed no significant effect of attractiveness in
predicting IQ.
Extraversion is predictable from our subjective impressions, but seems to be
partially due to observations about grooming, smiling, and attractiveness. The zero-‐
order correlation, r = 0.188 (p=3.692e-‐05, df =473), was significant, and such
mediators as BMI, age, and sex, do not have any significant effects. However,
controlling for smiling, grooming, and attractiveness diminishes the significance
substantially, suggesting that these three factors are potential mediators of the
subjective-‐objective extraversion relationship. These factors can be thought of
together as a “self-‐presentation” variable that raters reported using as criteria for
making their ratings. Since each of these factors were related to both subjective
extraversion and objective extraversion, the relationship goes down between those
two after controlling for smiling, grooming, or attractiveness.
JUDGING A BOOK BY ITS COVER 22
Raters reported using smiling and grooming as criteria for extraversion, but
attractiveness may play a more subconscious role. Attractiveness is the “personal
characteristic most obvious and accessible”, according to attractiveness stereotype
research (Dion, 1971). Our research therefore does supply some support for a “halo
effect” of extroversion – people can indeed guess at someone’s level of extraversion
from their appearance, but this appears largely to be due to the fact that attractive,
groomed people who smile are more likely to be extraverted. When predicting
Subjective Extraversion from self-‐report extraversion along with other factors, the
JEPQ score had a small significant effect, but most of the variance was due to
grooming, attractiveness, and degree of smile. There was also a large effect from
socioeconomic status (SES), which was slightly related to attractiveness (r=0.14),
but not to grooming or smiling. Apparent SES therefore may be included in the halo
effect: more attractive people are expected to have a higher SES and vice versa.
Observing clothing, jewelry, and hairstyle may have contributed to higher ratings of
both attractiveness and subjective extraversion.
Correlations between subjective and objective neuroticism measures seem to
be due entirely to chance (p values were not significant). Pervious studies have
found significance in judging emotional sensitivity in males but not in females
(Penton-‐Voak, 2006). However, the present findings did not reveal any significant
sex differences for neuroticism.
Cronbach’s α measures for inter-‐rater reliability were excellent for
attractiveness, smiling, and extraversion (α = 0.87-‐0.9), good for grooming and acne
(α = 0.7-‐0.8), and decent for intelligence and neuroticism (α =0.6). This makes sense
JUDGING A BOOK BY ITS COVER 23
given the raters’ descriptions of their rating methodology. Smiling is more objective
than subjective – “Is this subject smiling?” does not leave much room for
interpretation, so α ≈ 0.9 is expected. Attractiveness (α =0.87) is more subjective,
but previous research has shown that evaluations of facial attractiveness are highly
consistent across raters, even cross-‐culturally (Langlois et al., 2000) or in young
infants (Langlois et al., 1987; Slater et al., 1998).
Raters reported using similar criteria in making their ratings of subjective
extraversion, which likely contributes to the high degree of agreement (α =0.9).
High extraversion ratings were given for subjects with confident expressions in the
eyes, genuine smiles, piercings, and wild hairstyles. Low extraversion ratings were
assigned to subjects who were timid, anxious-‐looking, or un-‐groomed.
Grooming and acne were each rated by only two raters, so we can expect
these α measures to be lower. Acne, the less subjective of the two variables, had an α
of 0.77. Grooming (α = 0.7) is difficult to define, and easy to confuse with
attractiveness. Grooming was defined to be a measure of how much effort the
person put into their appearance that morning, to try to avoid conflating grooming
with attractiveness and vice-‐versa. Different standards of grooming were allowed in
the LTS and QIMR sets, but when sample was controlled for, there we no significant
differences between them.
Intelligence and Neuroticism have the lowest reliability (α = 0.6 and α =0.57
respectively). Raters reported judging intelligence based on a “gut instinct”. In
rating high emotional sensitivity, raters paid the most attention to the expression in
the eyes and the degree of smile. Low emotional sensitivity ratings were given to
JUDGING A BOOK BY ITS COVER 24
individuals with more defined facial features (e.g. square jaw), smaller eyes,
confident body language, and genuine smile. Despite similar criteria, the α measure
for Neuroticism is still low (though not unacceptable).
The true correlations between subjective measures of intelligence and
neuroticism may have been higher if there had been more raters, and thus more
reliable subjective impressions. Individuals are not very accurate in their subjective
ratings, but once aggregated, statistically significant correlations can be observed.
For example, an individual rater’s subjective assessment of intelligence only
correlates slightly with IQ (r= 0.), but the overall correlation once the subjective
ratings are averaged between all the raters is r = 0.3. We did have a large number of
subjects (n=1600), but not all data was available for all subjects, so the sample size
was reduced. Splitting the data into two groups to preserve independence also
decreased the sample size, and thus the power of our observations. However, the
current study had a large number of target faces compared to other studies, and
although some subjective measures were not as reliable as we would have liked, the
large sample size provided some compensation.
The main limitation to the current study was the lack of standardization in
the photographs used for ratings. Although the photos were controlled for LTS
twins, the larger proportion of the sample (QIMR) had photographs for
identification purposes that were not intended for use in subjective ratings.
Participants were not asked to maintain neutral expressions, or refrain from
grooming, and smiling and grooming had very large effects in perceived personality
traits. Grooming is a difficult factor to assess regardless – some people may
JUDGING A BOOK BY ITS COVER 25
inherently appear more groomed if more attractive, or appear more attractive if
they groom regularly.
The current study has determined that there is in fact something in the face
besides attractiveness that displays internal traits, the next step is to examine what
it is. Perhaps by measuring certain facial features and correlating them with our
subjective impressions of traits we will discover a key to phenotypic displays of
personality. People must be picking up on some facial cue in order to develop the
“gut reaction” described by the raters, especially in the domain of intelligence. A
large sample with controlled photos is necessary to exclude factors such as smiling.
With today’s new technologies and analytic methods, people’s early fascination with
judging character from features deserves a second chance.
Our study found that intelligence can be judged accurately even when
controlling for potential mediators including attractiveness, SES, and perceived
grooming. Extraversion can also be judged accurately, but appears to be mediated
by attractiveness, grooming, and smiling. Judgments of Neuroticism, on the other
hand, could not be predicted by subjective ratings. This suggests that humans can
pick up on valid cues towards a person’s internal traits without observing any of
their interactions. Since we make judgments about personality from facial
characteristics every day, the study of personality attribution from facial features –
new physiognomy – is certainly worth further study.
JUDGING A BOOK BY ITS COVER 26
REFERENCES
Barlow, Nora ed. 1958. The autobiography of Charles Darwin 1809-‐1882. With the
original omissions restored. Edited and with appendix and notes by his grand-‐
daughter Nora Barlow. London: Collins
Borman, W. C. (1977). Consistency of rating accuracy and rating errors in the
judgment of human performance. Organiza-‐tional Behavior and Human
Decision Processes, 20(1): 238– 252.
DeYoung, C.G., et al. (2008). Externalizing behavior and the higher order factors of
the big five. Journal of Abnormal Psychology, 117(4): 947-‐953.
doi:10.1037/a0013742
Hassin, R., Trope, Y. (2000). Facing faces: Studies on the cognitive aspects of
physiognomy. Journal of Personality and Social Psychology, 78(5): 837-‐852.
doi: 10.1037/0022-‐3514.78.5.837
Little, A.C., Perrett, D.I. (2007) Using composite images to assess accuracy in
personality attribution to faces. British Journal of Psychology, 98: 111-‐126.
doi:10.1348/000712606X109648
McCartney, K., Harris, M.J., & Bernieri, F. (1990) Growing up and growing apart: A
developmental meta-‐analysis of twin studies. Psychological Bulletin, 107(2):
226-‐237. doi: 10.1037/0033-‐2909.107.2.226
Nisbett, R.E., Wilson, T.D. (1977). The halo effect: Evidence for unconscious
alteration of judgments. Journal of Personality and Social Psychology, 35(4):
JUDGING A BOOK BY ITS COVER 27
250-‐256. doi:10.1037/0022-‐3514.35.4.250 Retrieved from:
http://psycnet.apa.org/journals/psp/35/4/250/
Passini, F. T., Norman, W. T. (1966). A universal conception of personality structure?
Jounal of Personality and Social Psychology, 4(1): 44-‐49.
doi:10.1037/h0023519 Retrieved from:
http://psycnet.apa.org/journals/psp/4/1/44
Penton-‐Voak, I., et al. (2006) Personality judgments from natural and composite
facial images: More evidence for a “kernel of truth” in social perception.
Social Cognition, 24(5): 607-‐640. doi: 10.1521/soco.2006.24.5.607
Sheppard, L.D. et al. (2011). The effect of target attractiveness and rating method on
the accuracy of trait ratings. Journal of Personnel Psychology, 10(1):24–33
doi: 10.1027/1866-‐5888/a000030
Thomas, J.C., Meeke, H. (2010). Rater error. Corsini Encyclopedia of Psychology.
doi:10.1002/9780470479216.corpsy0774
Willis, J., Todorov, A. (2006). First impressions: Making up your mind after a 100-‐ms
exposure to a face. Psychological Science, 17(7): 592-‐598. doi:
10.1111/j.1467-‐9280.2006.01750.x
Wright, M. (2001). Genetics of cognition: Outline of a collaborative twin study. Twin
Research, 4(1): 58-‐46. DOI: http://dx.doi.org/10.1375/1369052012146
Yzerbyt, V.Y., Kervyn, N., & Judd, C. (2008). Compensation versus halo: The unique
relations between the fundamental dimensions of social judgment.
Personality and Social Psychology Bulletin, 35(8): 1110-‐1123.
JUDGING A BOOK BY ITS COVER 28
doi: 10.1177/0146167208318602 Retrieved from:
http://onlinelibrary.wiley.com/doi/10.1002/9780470479216.corpsy0774/f
ull
Zebrowitz, L. A., Collins, M. A., & Dutta, R. (1998). The relationship between
appearance and personality across the life span. Personality and Social
Psychology Bulletin, 24(1): 736–749.
Zebrowitz, L. A., et al. (2002). Looking smart and looking good: Facial cues to
intelligence and their origins. Personality and Social Psychology Bulletin,
28(2): 238-‐249.
- University of Colorado, Boulder
- CU Scholar
- Spring 2012
- Judging a Book By Its Cover: Are First Impressions Accurate?
- Tess Adams
- Recommended Citation
- Microsoft Word - Adams - Thesis Final Two.docx
