Sex, race, age, and beauty as factors in recognition of faces


Sex, race, age, and beauty
as factors in recognition of faces*

JOHN F. CROSS,t JANE CROSS, and JAMES DALY
Saint Louis University, St. Louis, Missouri 63103

Three hundred Ss, black and white, male and female, at four age levels were
tested for their recognition of 12 photographed faces: black, white, male,
female, child, adolescent, and adult. Females recognized female faces more
frequently than the male faces, while male Ss recognized the male and female
faces with equal facility. Whites recognized the white faces more frequently than
the black faces, while black Ss recognized the black and white faces with equal
facility. Incorrect identifications of 24 faces not actually seen before were
treated separately. Male faces and black faces were misidentified more than
white faces and female faces. The number of false identifications of faces
decreased as Ss increased in age. Perceived beauty in a face facilitated
recognition. There was evidence of large differences in memorability among
individual faces.

adolescents, and adults in their 30s
and 40s. The portraits, each about
1 in. square, were arranged randomly
on an 8 x 11 in. card in a matrix
containing 24 faces, presenting two
faces from each of the 12 age-sex-race
categories. Two such composites were
made. Twelve of the faces, one from
each category, were common to both
cards and each card had 12 additional
faces not appearing on the other card.
The consequence of this arrangement
of stimuli was that the stimulus sample
for recognition included six portraits
of males, of females, of blacks, and of
whites, and four portraits at each age
level. The sample of portraits subject
to misidentification was doubled for
each category of portrait since 12
different faces appeared in the two
matrices.

To many uninitiated observers all
babies tend to look alike, as do most
Orientals. Experience with a group
usually leads the observer to make the
many fine discriminations possible
until he ultimately discovers that the
group is composed of individuals with
infinitely variable and memorable
facial characteristics. Recognition of
faces is a complex perceptual
achievement with practical
applications ranging from
identification of criminals to social
success at cocktail parties.

In the recent literature there are few
studies in this area. Howells (1938)
reported that women were superior to
men and sales personnel superior to
farmers. He found a correlation of
0.27 with IQ of the S. Seeleman
(1940) found that among white Ss,
unfavorable attitudes toward Negroes
were related to poor memory for black
faces. Goldstein and Chance (1965)
report that 9-year-old children found
it easier to recognize younger faces
than those of the same age or older,
but that other age groups of children
were not affected by the age of the
photographed face. Malpass and
Kravitz (1969), working with black
and white college students, found that
white faces were recognized more
frequently than were black faces, but
that there was a significant interaction
between race of the S and race of the
stimulus. Ss having more experience
with the opposite race recognized
faces of that race more frequently
than more racially segregated Ss.

*This research was supported by U.S.
Public Health Service Grant MH 17189-10.
The authors wish to thank the Ss who gave
their time to this project, and the faculty of
each of the participating schools for their
cooperation and help.

tRequests for reprints should be sent to
John F. Cross, who is now Chairman of the
Social Sciences Division, College of the
Virgin Islands, St. Thomas, U.S. Virgin
Islands 00801.

Malpass and Kravitz concluded that
experience with persons of a given race
was related to ability to recognize
faces of that race.

The present study was concerned
with the ability of individuals, black
and white, male and female, children,
adolescents, and adults from racially
segregated and racially integrated
backgrounds, to recognize a wide
range of facial types. The relation of
perceived attractiveness of faces to
their recognizability was investigated
as were differences in recognizability
between individual faces.

METHOD
Subjects

Three hundred Ss, consisting of 80
individuals at each of three age
levels-7, 12, and 17 years-and a
group of 60 adults whose mean age
was 36, were interviewed individually.
In each of the first three age groups
there were 40 black children and 40
white children. Within these groups,
half lived in racially segregated
communities and attended uniracial
schools, while the other half lived in
racially integrated urban
neighborhoods and attended
integrated schools. The adult group
consisted of 40 white adults and 20
black adults from the same
communities and neighborhoods as the
younger Ss, Equal numbers of males
and females were included in each S
group.

Stimulus Material
The photographic portraits used

were selected at random from high
school year books and school portraits
of grade school children and their
teachers. Highly unusual facial
expressions, hair styles, or clothing, or
persons with eyeglasses, were screened.
Black and white faces of both sexes
were represented at three age levels,
7 -year-old children, 13-year-old

Procedure
Before testing, Ss were assigned

randomly to one of two orders, and to
either the male or female examiner.
The S was told that we were studying
beauty, that we were interested in
which faces he thought were pretty or
handsome or beautiful, and that we
were trying to discover what sorts of
faces were attractive to different kinds
of people and whether or not a
person's ideas of beauty changed with
age. The S was not told that he would
be asked to recall any of the faces. The
S was then shown one of the two
group pictures. He was requested to
look at all the faces and to pick which,
if any, he considered pretty,
handsome, or beautiful. The S was
allowed to pick as many faces as he
wished, with as much time as he
needed. If the S acted hastily, he was
gently encouraged to examine all the
faces carefully.

The S was then given, as an
intervening task, a series of 16 groups
of six faces to look at. judge, and
finally rate for beauty. This procedure
provided data for a separate study of
the perception of beauty in faces, but
served for the present study as a
simulation of the real-world
experience of seeing many faces before
encountering the need to recall a
previously seen face. The S was then
shown the second matrix of portraits
containing 12 faces seen previously
and 12 new faces. S was asked which
of the faces had been shown before.

RESULTS
Results for correct recognitions and

incorrect identifications were analyzed
separately. For the sake of clarity, the
results are presented in four general
areas: (1) the effects and interactions
of sex of the face and age and sex of
the S on recognition and
misidentification of faces; (2) the
effect and interaction of race of face

Perception & Psychophysics, 1971, Vol. 10 (6) Copyright 1971, Psychonomie Society, Inc., Austin, Texas 393



Table 1
Memory for Faces of the Same and Other Sex and the Same and Other Race

Table 2
Memory for Male and Female Faces by Male and Female Ss at Different Ages

Percent Correct Percent Incorrect
Recognition Identification

Age of S

Type of Face 7 12 17 A 7 12 17 A

Male Ss
Male Faces 34 32 36 44 19 15 10 8
Female Faces 33 38 35 41 13 8 4 8
All Faces 34 35 36 42 16 12 7 8

Female Ss
Male Faces 41 34 27 30 26 10 12 10
FemaleFaces 50 50 33 29 19 12 6 4
All Faces 46 42 30 30 22 11 9 7

Perception & Psychophysics, 1971, Vol. 10 (6

Race Variables in Recognition
and False Identification

Recognition of black and white

faces by black and white Ss was
evaluated in a two-way ANOV A with
one repeated measure. Equal numbers
of male and female Ss at four age
levels were included. While neither of
the main effects reached statistical
significance, the interaction of Race of
Face by Race of S was significant at
the .01 level (F = 14.95, df = 1/10).
Black Ss recognized black and white
faces almost equally well, but white Ss
demonstrated poorer recognition of
black faces and better recognition of
white faces, as seen in Table 1. The
data on misidentification indicates
that both black and white Ss made
more false identifications of black
faces than of white faces. The two-way
ANOV A on the misidentification data
revealed the main effect of race of face
to approach significance (F = 4.27,
df = 1/10, P < .10).

Half of our Ss lived in racially
integrated neighborhoods and half in
uniracial communities. The data of
two of the S groups, the 12-year-olds
and the 17 -year-olds, whose level of
integration or segregation would be
most precisely defined in terms of
school population, were combined and
analyzed in terms of the differences in
recognition of black and white faces
relative to the racial environment of
the S. These data are presented in
Table 3.

For black adolescents, racial
environment seemed to have no
measurable effect on recognition
ability. Both integrated and segregated
black adolescents recognized black and
white faces with equal facility. Black
faces were incorrectly identified more
often than white faces. This was true
for both integrated and segregated
black adolescents. Both these results
repeat the overall pattern noted in this
study for black Ss of all ages. For
white adolescents, however, racial
segregation or integration was related
to recognition ability. While both
integrated and segregated white Ss
recognized more white faces than
black faces, the segregated white
adolescents recognized significantly
more white faces than did Ss from

51
57

40
64

53
57

57
56

Percent "Attractive"
Faces Correctly

Recognized

MaleSs
13

7

WhiteSs
12

8

Female Ss
14
10

Black Ss
19
11

Percent
Incorrect

Identification

Sex and Age Effects in
Misidentification of Faces

A three-way ANOV A with two
repeated measures was done for
incorrect identification of male and
female faces by male and female Ss at
four age levels. In Table 1 it can be
seen that both male and female Ss
falsely identified male faces more
frequently than female faces, and that
females made slightly more errors in
identification than did male Ss. In the
ANOV A, for misidentifications,
however, the main effects of sex of
face and sex of S were not statistically
significant, nor was the interaction of
these two variables.

Age of S was a highly significant
variable (F = 24.88, df = 3/30,
p < .01). For both male and female Ss
the tendency was for fewer false
identifications with increasing age.
Although male and female Ss were
similar in their performance at age
levels of 12, 17, and adult, the
7-year-old girls made many more
errors in identification than did
7 -year-old boys. This is largely the
basis of the significant Age of S by Sex
of S interaction (F = 3.45, df = 3/30,
p<.05).

Percent
Correct

Type of Face Recognition

Male Faces 36
Female Faces 37

MaleFaces 33
Female Faces 43

Black Faces 39
WhiteFaces 40

Black Faces 27
WhiteFaces 45

394

and race of the S on recognition and
misidentification; (3) the relation
between perceived beauty of the face
and recognition; and (4) differences
between individual faces.
Generalizations concerning the facial
variables are limited because of the
small sample of portraits used.

Effects of Sex and Age
on Recognition

Percent correct recognition of the
male and female faces by male and
female Ss at different ages was
analyzed in a three-way ANOV A with
two repeated measures (Winer, 1962,
p.328). Each cell of the 2 by 2 by 4
(sex of face and sex of S and age of S)
analysis contained the proportion of
correct recognitions for each of the six
faces used. Both black and white Ss
were included. While female Ss
(N =150) recognized more faces than
did male Ss (N =150), and while the
female faces (N = 6) were recognized
more frequently than the male faces
(N = 6), neither sex of S nor sex of
face were significant main effects. The
Sex of Face by Sex of S interaction
was significant beyond the .01 level
(F = 11.06, df = 1/10). This result is
more easily understood by referring to
the percent of correct recognitions of
male and female faces presented in
Table 1. Male Ss recognized the male
and female faces with equal facility,
while female Ss recognized the male
faces less often than did males and
recognized the female faces more
often.

The age of S main effect was not
statistically significant (F = 1.77,
df = 3/30, p < .2). When male and
female results are seen separately in
Table 2, there is seen a slight
improvement in recognition with age
for males, while for females there is a
slight decrease in correct recognitions
with increasing age. This Age of S by
Sex of S interaction was statistically
significant (F = 8.64, df = 3/30,
p < .01).



Table 3
Memory for Black and White Faces by Segregated and Integrated Black and Wh;t" Ss

racially integrated schools (Z = 3.07,
p < .01). Predictably, segregated white
adolescents made more false
identifications of black faces than of
white faces. The incorrect
identifications of the integrated white
Ss, however, were equally divided
between black and white faces, an
exception to the otherwise consistent
finding of more false identification of
black than of white faces.

Effect of Perceived Beauty
on Recognition of Faces

The initial task presented to each S
was to choose from the first matrix of
faces any which he considered pretty,
handsome, or beautiful. Table 1
indicates the percentage ~f correct
recognitions of faces by Ss who
reported those faces attractive in
comparison to the percentage of
correct recognition of all Ss. In every
case, a higher percentage of Ss who
perceived beauty in a face were
subsequently able to recognize the
face.

Range of Recognizability
of Various Faces

For each portrait the total number
of correct identifications possible was
300. The most frequently recognized
face, that of the 7-year-old white girl
was recognized by slightly more than
half the Ss. The face least frequently
recognized, that of the 12-year-old
black male, was recognized by only
17% of the Ss. The mean number of Ss
correctly recognizing a picture was
113, and the range was from 51 to
156. For 22 of the 24 faces which
could be misidentified, the range was
from 4 to 25 false identifications out
of a possible 150. The two notable
exceptions were a 12-year-old black
male face, which was falsely identified
by 40 Ss, and an adult black male face,
which was falsely identified by 47 Ss.

Although earlier work (Howells,
1938) led us to expect that females
would be better at recognizing faces
than males, we found females superior
only in the case of the female faces.
This result might be specific to the six
female faces in oor sample, but this
seems unlikely ir w of the range of
age and the race d.. ~rence in both our
Ss and our photos. Possibly female Ss

white faces, blacks would have had
ample opportunity to learn the
discriminanda for both racial types,
while whites would have not. The
smaller difference in recognition
efficiency between black and white
faces found for the integrated white
adolescent Ss as compared to the
segregated whites is consistent with
this interpretation. It remains for
future research to establish that there
are in fact differences in the
discriminanda, be they specific or
more Gestalt-like, between racial
types. An alternative explanation of
these results might be made strictly in
terms of motivational factors,
considering the relative importance to
the S groups of the ability to
distinguish within the groups of faces
in question.

The greater memorability of
attractive faces is another result
requiring further research to determine
the relative weight of the various
factors which may account for it.
Simple explanation in terms of
differences in exposure time are ruled
out. Our Ss often spent more time
scrutinizing faces of doubtful beauty,
finally deciding that most of these
were not really pretty or handsome,
than they spent examining obviously
(to them) attractive faces.
Instructional set probably played a
part in the results. While Ss did not
expect to be asked later to recognize
specific faces, they were set for an
examination of their personal
standards of beauty. While they
generally spent less time looking at the
attractive faces, they may have given
them their more active attention,
especially as Gestalten, which would
enhance later recall of the face as a
whole.

The differences in memorability of
individual faces was a notable result of
our study. Out of our sample of 12
f aces, the most memorable was
recognized three times as often as the
least. Of the 24 which could be
incorrectly identified, the most
mistaken face was falsely identified 12
times as often as the least. When we
arrayed these faces in the order of
their memorability and asked 15 of
our colleagues to identify the continua
on which the faces were ordered, they
could detect none. When told the basis
of the ordering, 14 out of 15 correctly
identified the high and low end of the
array of faces ordered by hit
frequency and 13 out of 15 did so for
the faces ordered by false alarm
frequency. At this point they were
prepared to suggest that the
well-remembered faces were less
"ordinary" looking, while the more
misidentified faces were more
..ordinary, " which suggests an
interference theory of facial

Percent
Incorrect

Identification

Segregated Black Ss
40 13
41 9

Segregated White Ss
25 14
48 4

Percent
Correct

Recognition

are more attentive to female faces
because of the emphasis placed on
attractive female facial characteristics.
The results for incorrect
identifications showed few differences
between male and female faces. This
finding is not inconsistent with the
hypothesis of sensitivity of females to
female faces since, for faces not
previously seen and consequently not
scrutinized, no sex of S difference
would be expected.

While percent correct recognition
did not change with age, holding
around 40%, performance improved
steadily by virtue of a decreasing false
alarm rate up to the adult level. The
percent correct recognition and
incorrect identification is most likely
quite specific to the instructions and
testing conditions used, but the trend
of performance related to age of S is
probably a widely generalizable
finding. The most obvious
interpretation is that younger Ss guess
more than older. Another explanation
is that younger Ss may have less ability
to recall the circumstances under
which a face was seen and
consequently confused the faces
shown them in the matrix with faces
seen prior to the interview or with
faces in the interpolated series. If
young children are less capable of
making the fine discriminations which
differentiate faces, the probability of
such confusion would be increased.
Thus, at least three factors may
contribute to performance in
recognition or misidentification of
faces: willingness to guess, ability to
make discriminations, and recall of the
circumstances of the prior perception.

The effect of race on recognition of
faces is interpretable in the light of
differences between the black and the
white experience in the United States
at the time that we did the testing
(summer and autumn of 1969).
American blacks, in their work and via
television and movies, are virtually
assured of exposure to whites. For a
majority of whites the situation is
reversed. These differences in exposure
rates could account in part at least for
the results obtained. If those details of
physiognomy which provide the most
salient diseriminanda between black
faces are different from those between

Percent
Incorrect

Identification

Integrated White Ss
22 8
34 9

Integrated Black Ss
38 13
38 5

Percent
Correct

Recognition

Black Faces
White Faces

Black Faces
White Faces

Type of Face

Perception & Psychophysics, 1971, Vol. 10 (6) 395



memorability, More typical faces are
more easily confused with previously
seen faces. An interference theory
would similarly account for the greater
memorability of the more than
typically attractive faces.

A final note of concern is directed
to the finding that 2 of our 24 faces
were misidentified by so many Ss.
These faces did not remind our Ss of
any well-known persons, which might
account for our result. The social
significance of the finding rests in the

possible consequences for persons who
are atypically misidentifiable, say, for
example, in a police line-up. It is
interesting that the faces in question
were both black males. The need for
further investigation of this issue is
apparent.

REFERENCES
GOLDSTEIN, A. G .. & CHANCE, J. E.

Recognition of children's faces.
Perceptual & Motor Skills, 1965, 20,
547-548.

HOWELLS, T. H. A study of ability to
recognize faces. Journal of Abnormal &

Social Psychology, 1938, 33, 124-127.
MALPASS, n, S., & KRAVITZ, J.

Recognition for faces of own and other
race. Journal of Personality & Social
Psychology, 1969, 13, 330-335.

SEELEMAN, V. The influence of attitude
upon remembering of pictorial material.
Archives of Psychology, 1940, 36,
258-268.

WINER, B. J. Statistical principles in
experimental design. New York:
McGraw-Hill, 1962.

(Accepted for publication May 19, 1971.)

396 Perception & Psychophysics, 1971, Vol. 10 (6)