RACE CATEGORIZATION AND FACE RECOGNITION

STAGES IN THE PROCESSING OF LATERALLY

DISPLAYED UNKNOWN FACES
Raymond Bmyer and Myriam Schweich
(University of Louvain, Medical Department, Unite de Neuropsychologie
Experimentale de l'Adulte)

The difficulty in the processing of faces when the stimuli do not issue
from the same race as that of the subject is called the "race effect"
(Brigham, 1986; Bruyer and Dussart, 1985; Lindsay and Wells, 1983; and
Shepherd, 1981, for reviews). Recently, Bruyer and Dussart (1985) inves-
tigated the race effect in a recognition task involving African and Cau-
casian, laterally-displayed face. The "index of race effect" (IRE), essen-
tially, was significant for the white subjects only and was limited to their
right visual field (RVF). Thus, the left hemisphere seems to have diffi-
culties in processing faces of another racial group: the right cerebral
hemisphere was as efficient as the left and remained unaffected by the
"race". According to Bruyer and Dussart, this result could be explained
by, for example, the hypothesis that a familiarity effect was operative.
The main purpose of the present study is to replicate this observation
and to search for possible lateralized effects of familiarization to African
.and Caucasian faces. In Exp. I, white subjects were enrolled in a task
similar to that of Bruyer and Dussart but with a greater number of trials
within each stimuli set: in each condition three successive blocks of 48
trials were administered and analyzed. Bruyer and Dussart used a single
block of 64 trials. The right-sided race effect was expected to be more
important and/ or more lateralized at the beginning than at the end of the
experimental session.
In the study of Bruyer and Dussart (1985) as well as in the present first
experiment, stimuli were blocked as concerns the race: one condition
concerned white faces, another black ones. It was assumed that the sub-
jects did not make any "racial decision", and that the recognition process
was directly activated. Thus, in Exp. II, black and white faces were
randomly intermixed. Therefore, it was assumed that the subjects had to
make a two-stage operation: the racial decision, then the recognition.
We were unable to predict with confidence the kind of main effect of
this change of experiment. Indeed, on the one hand, the need to perform
Cortex (1987) 23, 415-429
416 Raymond Bruyer and Myriam Schweich

two successive operations could increase the difficulty, which would

impair the efficiency. However, it could also facilitate the decision, given
that "race" can be easily detected and that the recognition that follows
would then be facilitated (the number of faces within each race being
lower than in Exp. I). Therefore, we have compared mainly the two
experiments by studying the IREs and lateral differences. A similar pat-
tern of asymmetry concerning the IRE in both experiments could indicate
that a racial decision was unavoidable and automatically triggered even
when unnecessary (Exp. I). Conversely, any change in this pattern could
reasonably be considered due to the racial decision.
In both cases, however, it remained possible that the racial decision
per se was lateralized. To clarify the interpretation, the subjects enrolled
for Exp. II were enrolled in another experiment, a simple racial decision
task. According to recent models (Ellis, 1983, 1986; Hay and Young,
1982), like gender categorization, racial classification would take place
relatively early in the processing and before the decision of familiarity. In
addition, racial (or gender) categorization would closely follow the facial
decision operation. As concerns this facial decision, it seems that both
hemispheres are able to manage the task and that asymmetries appear
during the next, representational step (Hay, 1981 ; Young, Hay and
McWeeny, 1985). As concerns gender categorization, it has been shown
that such a task is faster than any recognition task and performed well by
each hemisphere provided that stimuli are not experimentally degraded
and that individual characteristics of the subjects are considered (Jones,
1979, 1980). This applies for unknown (Hellige and Jonsson, 1985; Hel-
lige, Corwin and Jonsson, 1984; Jones, 1979, 1980; Sergent, 1982a),
familiar (Sergent, 1985), as well as for famous faces (Marzi, Tassinari,
Tressoldi, Barry and Grabowska, 1985). In short, we would like both to
dissociate the recognition from the racial decision stages as responsible
for the lateralization of the race effect reported by Bruyer and Dussart
(1985), and to examine the role of familiarity on this phenomenon.

EXPERIMENT I

The first experiment was essentially the same as that of Bruyer and
Dussart (1985) with slight modifications in order to try to identify a
familiarity effect to explain the lateralization of the race effect.

Materials and Methods

The experiment was conducted with 16 naive, right-handed (self-report)

female subjects with normal or corrected visual acuity, with a non-inverted hand
 Race categorization and face recognition 417

position for writing, and no left-handed near relatives. The mean age was 22.69
years; SD = 1.93. Only white, Belgian subjects were enrolled. In addition, we
selected those who had not been familiarized with the "other race": none had
lived in Africa, and none had black people among her five best friends.
The stimuli were those used by Bruyer and Dussart (1985). Within each
stimulus race (Caucasian vs. African), eight kinds of slides were prepared. There
were four neutral male full-faces, each displayed in the left (L VF) or right visual
field (RVF), under the normal or mirror version. The size of the faces was 4 deg of
angle of the visual field and faces were centered 4 deg left or right of the central
fixation point. The slides were back-projected for 180 msec. A distractor face of
the same ethnic group was systematically simultaneously displayed in the oppo-
site field, with an arrowhead at fixation pointing towards the stimulus. Two
distractors were randomly associated in the various occurrences of stimuli; the
dis tractors were never used as targets.
Each subject was submitted to two experimental conditions according to the
stimulus, race in a balanced design. In both conditions, after preliminary trials for
training, the subject was asked to depress the left button with the left index finger
for two stimuli, the right (right index) for the other two. The button assignment
was balanced across subjects. Each condition was introduced with 16 preliminary
unanalyzed trials and then three blocks of 48 trials followed: in each block, each
face (or its mirror image) appeared six times in each field, in a random
sequence.
The dependent measures were the proportions of errors (out of 24) and the
reaction time of the correct responses (RT) by visual field, block, and stimulus'
race. In addition, an IRE was computed with the two measures, by combining the
scores on the two kinds of stimuli using the formula (other race - same race)/
(other race + same race); the IRE was therefore a number between -1 and + 1
with 0 indicating a lack of race effect. The results have generally been submitted
to analyses of variance (ANOVAs), and the significant effects were then studied
by the Newman-Keuls test (p<0.05).

Results

Table I displays the mean proportion of errors (out of 24) and the
mean RTs in function of the block, visual field, and white vs. black faces,
as well as the combined IREs. These results have been analyzed together
with those of Exp. II (below).

EXPERIMENT II

The goal of this experiment is to engage the subjects in a two-stage

decision task by inserting a preliminary "racial decision" stage before the
recognition per se. Differences in the patterns of results between Exp. I
and Exp. II would be due to this additional stage, which will be studied in
isolation in Exp. IV. The lack of any difference between Exp. I and II,
conversely, would suggest that a "racial decision" stage, although not
 ..,.
IX>

TABLE I
Mean correct RTs (msec), Proportions of Errors out of 24 (between Parentheses) of Exp. I, and the Corresponding IREs.
Means of 16 Subjects
Block 1 Block 2 Block 3
Visual field: Left Right Left Right Left Right ~

White faces 1134 (.068) 1149 (.078) 1003 (.052) 1046 (.039) 926 (.031) 941 (.031) ~
Black faces 1441 (.245) 1412 (.255) 1281 (.229) 1288 (.216) 1233 (.151) 1243 (.185) ~
~
IREs (RT) .1096 .1004 .1160 .0990 .1278 .1243
Il:I
~
~
§
~

~
TABLE II
5°
~

Mean Correct RTs (msec), Proportions of Errors out of 12 (between Parentheses) of Exp. II, and the Corresponding IREs. ~
Means of 16 Subjects ~
§:
Block 1 Block 2 Block 3
Visual field: Left Right Left Right Left Right
White faces 1115 (.094) 1238 (.089) 957 (.036) 1146 (.031) 874 (.021) 868 (.010)
Black faces 1342 (.198) 1569 (.234) 1138 (.089) 1193 (.182) 984 (.062) 1180 (.104)
IREs (RT) .0800 .1066 .0790 .0590 .0540 .1443
 Race categorization and face recognition 419

necessary, was already at work in Exp I. Finally, it could be that both

experiments differ only as a general main effect, which would indicate
that the inserted task either facilitates or impairs the global process.

Materials and Methods

Another series of 16 female subjects was enrolled, with identical criteria as for
Exp. I. The mean age was 22.44 years; SD = 1.17.
The stimuli were the same as those of Exp. I. However, black and white faces
were randomly intermixed and the subjects had to recognize one .black and one
white face by depressing a given button, and the other black and the other white
face by depressing the other button. Thus, the initial material for preliminary
learning was a set of two white and two black faces. Consequently, 50% of the
initial trials were maintained, to design a single experiment with three blocks of
48 trials plus 16 items for familiarization. The discarded trials were retained for
Exp. IV. The procedure was identical to that of Exp. I, with only one condition.
Therefore, the order of conditions was irrelevant.

Results
Table II displays the mean proportion of errors (out of 12) and RTs by
experimental condition, as for Exp. I.

Analysis of Exps. I and II

Errors
Since mean error rates as low as I or 2% appeared in at least some cells
,of Tables I and II, it was doubtfull that homogeneity of variances applied.
Therefore, ANOVA has not been performed on errors, and the IREs have
not been derived and analyzed.
We can observe, at a purely descriptive level: (a) an advantage of white
over black faces in every experiment X field X block combination; (b) an
improvement of accuracy with familiarization (blocks), for black faces
only; (c) a superiority of Exp. II over Exp. I, for black faces only; and (d)
an advantage of the LVF over the RVF, for black faces only, which
resulted more from Exp. II than from Exp. I. The combination of obser-
vations (c) and (d) would suggest an higher race effect in the RVF than in
the LVF, especially during Exp. I.

RTs
The experiment X field X race X block ANOVA on RTs revealed
three significant main effects and one significant first-order interaction.
Correct responses were faster to white (1033 msec) than black stimuli
420 Raymond Bruyer and Myriam Schweich

Fig. I - The experiment

X field significant interaction
RT(msec) for the RTs in Exps. I and
/I.
1200 Exp. II
Exp. I

1100

1000
1'"
Left Right
field field

(1275: F= 36.4; dJ. = 1, 30; p< <0.001), and to stimuli displayed in the
LVF (1119 msec) than to those shown in the RVF (1189: F= 6.1; d.f. = 1 ,
30; p<0.025); the block significant main effect (F= 30.33; dJ. = 2,60;
p< <0.001) evidenced that all the differences were significant among the
three stages of the experimental session (1300 vs. 1132 vs. 1031 msec). The
significant experiment X field interaction (F= 4.43; dJ. = 1, 30;
p<0.05) is illustrated by Figure 1. The field effect was significant in Exp.
II only, and an experiment effect appeared in the LVF only.
The IREs derived from the RTs were analyzed by an experiment X
field X block ANOVA. A significant, first-order, experiment X field
interaction appeared (F= 4.64; dJ.= 1,30; p<0.05: Figure 2). The IRE
of the LVF stimuli during Exp. II significantly differed from the other
three, with no difference among these three. The sole other effect was a
tendency toward an experiment X block interaction (F= 2.68; dJ. = 2,
60; p<O.I), which did not contribute to the interpretation.
In addition, the 12 IREs were tested against zero (Student t-tests). In
Exp. I, blocks 2 and 3, the IRE of each hemifield was significant at
p<O.OOI ; for the first block, the IRE was significant at p<O.OOI for the
RVF and at p<0.02 for the LVF. In Exp. II, LVF, the IREs were
significant at p<O.OI, p<0.02, and p<0.05, for blocks 2, 3, and 1 res-
pectively; for stimuli displayed in the RVF, the IRE was significant at
p<O.OOI during the third block, and at p<O.OI during the first; in the
second block, the IRE did not reach a statistically significant leveL
 Race categorization and face recognition 421

Fig. 2 - The experiment -

X field significant interaction IRE
for the IREs derived from
RTs in Exps. I and /I.
.12
.11 Exp.l
Exp . l1

.09
.08
.07
1"
Left Right
field field

Discussion

We note first that even if statistical analyses of accuracy were not

allowed, the two dependent measures agreed in their interpretative mean-
ing, so no speed-accuracy tradeoff can be suspected. Second, if we con-
sider the main effects, the results were, to a large extent, as expected: our
white subjects processed white stimuli better than black; the recognition
of faces was easier in the LVF than in the RVF; and the performance
improved from the first through the third block of trials.
- A first purpose of the present study was to test the familiarity inter-
pretation of the laterality of the race effect found by Bruyer and Dussart
(1985). In our Exp. I, which was the same as that made by Bruyer and
Dussart except in length, and in-Exp. II also, we noted that advantage of
white over black stimuli was observed within each block, and that the
block effect was limited to the black faces in errors (but observed for black
and white faces in RTs). More importantly, there was no sign of lateral-
ization of this race X block interaction, no block X field interaction, and
all the IREs of Exp. I differed significantly from zero. Thus, with a
lengthened experiment, the signs of lateralized IRE disappeared. An
evidence favoring our hypothesis could be found in the decreasing dif-
ference of accuracy between white and black stimuli from the beginning
up to the end of the experimental session. Nevertheless, it is a weak
argument, given the lack of interactions with the visual field. Therefore,
422 Raymond Bruyer and Myriam Schweich

there are discrepancies between our first experiment and the results
reported bu Bruyer and Dussart as well as, evidently, no sign of reduction
of the lateralization of IREs with practice. Moreover, no sign of lateral-
ized practice effect appeared in our second, more similar to daily life
conditions, experiment.
As a transition to the second point of this Discussion, we note, in spite
of the preceding comments, a locus of agreement between the present
results and those of Bruyer and Dussart (1985). The descriptive analysis of
accuracy suggested an higher race effect in the RVF that in the LVF, a
phenomenon that was precisely obtained in the previous study. In addi-
tion, this race-by-field "interaction" seems to be larger in Exp. I (than
in Exp. II), i.e., in the condition that was the most similar to that of Bruyer
and Dussart (1985). So, the current data tend to replicate the previous
ones. In addition, we note that this lateralization of the race effect was
very more slight than in the previous study, which could suggest that it was
at least in part due to familiarization (in the present study the experi-
mental session was lengthened).
The other purpose of this study was the search for lateralized differ-
ences between Exp. I and Exp. II. We recall that in Exp. I the subjects had
to recognize faces within a predefined set of uniracial stimuli, while in
Exp. II the recognition necessarily followed an early racial decision
operation. There appeared a clear LVF superiority in Exp. II with no
lateral difference in Exp. I: this was observed for errors and for RTs, as
well as in the study of the IREs derived from the RTs. In addition, the
examination of errors reveals that this LVF advantage in Exp. II was only
due to the black stimuli. This last observation, therefore, qualified the r~ce
X field descriptive interaction evidencing a LVF superiority for black
faces: indeed, this effect was only present in Exp. II. Thus, in Exp. II,
stimuli (especially black) were better recognized in the LVF than in the
right, and the race effect was larger in the RVF than in the left. Moreover,
the experiment effect was limited to the LVF for each dependent measure
and the derived IREs. Since this effect systematically favored Exp. II, it
therefore appeared that the addition of an early, racial-decision task
facilitated the processing but in the LVF only.
However, it was possible that the lateralization observed in Exp. II but
not in Exp. I was not due to the recognition stage per se. Indeed, the
preliminary race classification could be lateralized (better performed in
the LVF than in the right), this asymmetry being then transferred to the
subsequent stages (for this notion of transmitted asymmetry see Mosco-
vitch, 1979). Experiment IV was planned to test this possibility.
Before to start Exp. IV, however, an additional control was needed).

lWe thank an anonymous reviewer who suggested this experiment.

 Race categorization and face recognition 423

Exp. II was better performed than Exp. I, for faces shown in the LVF. This
could agree with Sergent and Bindra's (1981) considerations about the
possible lateralized effect of task difficulty. In Exp. II, provided the race
categorization has been correctly performed (which will be examined in
Exp. IV), the subject had to discriminate between two faces only (vs. four
faces in Exp. I). Therefore, it could be that this reduction in task requi-
rements had induced the LVF superiority in Exp. II. This hypothesis was
empirically studied in Exp. III.

EXPERIMENT III

Rationale and Methods

Exp. III has been designed in combining the procedure of Exps. I and II. Like
in Exp. I, the recognition stage without preliminary race classification operation
has been tested in Exp. III. However, like in Exp. II, the subjects of Exp. III had to
discriminate between two faces only, within each kind of stimuli (black,
white).
A new series of 24 female subjects was enrolled, with identical criteria as for
Exps. I and II (mean age = 22.54 years; SD= 1.14).
Exp. III was identical to Exp I, with some modifications. Within each racial
group of stimuli, the six possible pairs of faces have been used and the subjects
had to depress the left or the right button according to the single laterally
displayed face. The button assignment, the used pair face and the order of
conditions (whites first vs. blacks first) were balanced across subjects, with one
subjects in each of the 2 X 6 X 2 combinations.

Results
Errors were rare (about 2.5%) and not studied. The correct RTs have
been submitted to the race X field X block ANOVA. It appeared a
tendency towards a practice effect (F = 2.63; d.f. = 2,46; 0.05 <p<O.I):
responses were faster in the third (685 msec) and second blocks (698) that
in the first (780), with no difference between the second and the third. The
block effect did not significantly interact with the other factors.
White faces (659 msec) were processed faster than the black (783
msec: F= 13.67; dJ.= 1,46; p<O.OOI), and stimuli were better recog-
nized in the LVF (685 msec) than in the RF (757 msec: F= 7.54; dJ. = 1,
46; p<O.OI). The race X field significant interaction (F= 4.5; dJ. = 1,
46; P < 0.05) revealed only that the field effect was higher for black (741 vs
825 msec) than for white faces (630 vs. 688 msec). Note however that the
race effect was maintained in both fields, as was the field effects in both
races.
All IREs differed significantly from zero (p<0.02), and the field X
block ANOVA confirmed that the race effect tended to be higher in the
424 Raymond Bruyer and Myriam Schweich

RVF than in the LVF (.0809 and .0905: F=3.l2; dJ.=I, 46;
0.1 >p>0.05).

Discussion
The lateral differences recorded in Exp. II (with no asymmetry in Exp.
I) were reproduced here without any early race decision. Therefore, there
are reasons to think that differences between Exp. I and II were more
related with the within-race difficulty than with the need for an early
cognitive process of classification.
This early operation alone will finally be studied in Exp. IV.

EXPERIMENT IV

Rationale and Methods

In order to clarify the possible effect of a preliminary race classification, this

task alone has been required from the subjects of Exp. II.
The material of Exp. I not retained for Exp. II was used in Exp. IV. The
procedure was identical, but the subjects had to categorize the faces as black vs.
white by using the pushbuttons. The button assignment was balanced across
subjects, as was the order of experiments (II vs. IV). In Exp. IV, the distractors
displayed in the contralateral field were removed, since they were systematically
of the same race as that of the target.

Results and Analysis

Errors were rare (1.3% of the responses), so the fixed, race X field X
block ANOVA was not performed due to the floor-effects. For the same
reason, IREs were not computed, analyzed in a field X block ANOVA or
tested against zero.
The RTs are shown in Table III; they were submitted to the race X
field X block ANOVA. The sole significant result was a practice effect
(F = 5.39; dJ. = 2,30; p<O.OI): correct responses were faster in the third
(609 msec) and second blocks (613) than in the first (652), with no
difference between the second and the third. The IREs derived from the
RTs were not compared in a field X block ANOVA, since neither differed
from zero (Hests). This last observation, in addition, indicates that res-
ponses were not biased toward a specific kind of stimuli.

Discussion
As expected, the racial decision task was easier (Exp. I: 13% errors,
Exp. II: 9.5%, Exp. III: 2.5%, Exp. IV: 1.3%) and faster than the previous
 Race categorization and face recognition 425

TABLE III

Mean Correct RTs (msec) and Their Corresponding IREs for Exp. IV Means of
16 Subjects
Block 1 Block 2 Block 3
Visual field Left Right Left Right Left Right
White faces 639 669 621 623 594 616
Black faces 645 653 618 589 621 605
IREs -.0028 -.0079 -.0004 -.0171 -.0202 -.0090

ones (about 1175, 1130,721, and 625 msec for Exps. I, II, III, and IV,
respectively). This observation supports the intuitive notion that race
classification is an early stage in the processing, precedes the recognition
stage, and can therefore be applied to both known and unknown faces.
Similar considerations were suggested by recent models of face recogni-
tion (e.g. Ellis, 1983; Hay and Young, 1982) and have already been found
in studies dealing with male/ female classification of known and unknown
faces (e.g. Marzi et al., 1985; Sergent, 1985).
No significant IRE emerged with this task, which is not surprising,
given the race effect has usually been observed in recognition tasks and, in
some cases, in discrimination tasks (Sherped, 1981, for a review). There
was no reason, a priori, to suppose that white and black stimuli would
differ in a task like the one used here.
Finally, we noted a significant practice effect, but no sign of lateral-
ization. The race categorization in the LVF was as fast as in the RVF. This
could again support the notion that such a decision is a preliminary one,
and not subject to asymmetrical processing (for sex categorization and
similar results, see Hellige and Jonsson, 1985; Marzi et al., 1985; Sergent,
1985). More importantly, together with the results of Exp. III, it indicates
that the results of Exps. I and II cannot be interpreted by an asymmetry in
the early operations that would then be transferred to the subsequent
stages.

GENERAL DISCUSSION

In Bruyer and Dussart's study (1985), there was no sign of asymmetry

of performance in the recognition of black and white faces, but the IRE
derived from the RTs was significant in the RVF only. The authors
hypothesized a familiarization effect to account for this observation. In
order to test this hypothesis, we designed an experiment that replicated
that of Bruyer and Dussart but with more trials (a single block of 64 trials
vs. three blocks of 48 items). We replicated the lack of asymmetry of
426 Raymond Bruyer and Myriam Schweich

performance. In addition, the race effect was observed in each block of trials
but the IRE remained significant from the beginning to the end of the
experimental session, with no sign of lateralization; more precisely, it was
significant in both visual fields. Finally, a familiarization effect appeared in
both fields, which tended to be more important for black than for white
stimuli. Therefore, we essentially reproduced the observations by Bruyer
and Dussart, the race-effect now being observed in each hemifield.
This result could be considered as not supporting the familiarization
hypothesis. However, the interpretation of a lack of effect is always
hazardous. In particular, it could be that various kinds of familiarity must
be distinguished (as suggested elsewhere: Bruyer, Rectem and Dupuis,
1986). Bruyer and Dussart (1985) considered familiarization as the degree
of daily-life exposure to faces of the other race. In support of this con-
ception, they did not find signs of lateralization of the IREs in their black
subjects who lived in Belgium and, therefore, had been "familiarized"
with white faces (while their white subjects, like ours, had not been
familiarized with blacks, i.e., they had no black friends and had not stayed
in Africa). In our first experiment, we were probably dealing with another
kind of familiarization, namely "experimental familiarization", i.e. the
effect of seing a small set of previously unknown faces several times. These
differences in the meaning of the "familiarity" could explain the empirical
data. Experiments are needed in which the other kind of familiarity would
be tested: it would be useful to enroll subjects with various degrees of
"daily-life familiarity" with black faces, in order to study the correlations
between the IREs and the level of familiarity, or to compare a group of
familiarized to a group of unfamiliarized subjects.
In our first experiment, as in the study by Bruyer and Dussart (1985),
the trials were blocked as regards the race of the stimuli. Consequently,
the "racial decision" stage was not needed, and the subjects were engaged
in the recognition process only. In Exp. II, as in daily life conditions, black
and white faces were mixed so that the subj ects were requested to perform
a two-stage process: a racial classification operation and then the recog-
nition processing. This time an advantage of the LVF appeared (especially
for black faces), the IRE was higher in the right than in the LVF (as in the
study by Bruyer and Dussart), and the IREs decreased from the first to the
third block in the LVF only.
This last observation runs counter to the familiarization hypothesis,
which predicted a reduction of the IRE with practice, and reinforces the
intuitive notion that another kind of familiarity should have been exam-
ined. However, this observation, together with the asymmetry of IRE,
supports the previous study in that the left hemisphere could be less
adapted to process unusual stimuli. It could be that the right brain is
particularly involved in the processing of tasks for which no automatized
 Race categorization and face recognition 427

operators are available -amI that the left intervenes with well-established
routines, as suggested by Goldberg and Costa (1981). In addition, the
"classic" LVF advantage for face recognition emerged in this experiment
for a task needing a previous racial decision.
But this asymmetry cannot be attributed to the transmission of an
asymmetry from this early stage (Moscovith, 1979), since Exp. IV did not
reveal any sign of lateralization. In addition, Exp. II was easier and faster
than Exp. I for stimuli displayed in the LVF only. This result could be
linked to the considerations of Sergent and Bindra (1981). Indeed, once
the racial decision has been accomplished - and the results of Exp. IV
indicate that this operation is made rapidly and with few errors - the
subjects had only to discriminate among two faces inside the identified
racial group, an operation that is classically better made in the left than in
the RVF, at least with photographed, easily discriminable faces. Thus, the
differences of laterality between Exp. I (no asymmetry) and Exp. II (LVF
advantage) can be attributed to the reduction of the difficulty of the task:
within a given racial group of stimuli, the subject had to discriminate
between four faces in Exp. I, but between two in Exp. II. This argument is
strongly strengthened by the results of Exp. III. In this last condition, as in
Exp. II, the LVF advantage emerged while no early race decision was
needed. Therefore, by combining this observation with the results of Exp.
IV, it becomes obvious that the laterality effect in Exp. II (together with
the lack of asymmetry in Exp. I) resulted more from the "within race"
reduction of the task difficulty, than from the early racial decision oper-
ation.
A last comment will concern the LVF superiority that was especially
due to the black stimuli. This observation could be linked with the con-
siderations of Sergent (1982b, 1983) who suggests that the right hemi-
tspheric competence improves (relatively to that of the left) when the
luminance of the material is reduced: obviously, our black faces were
naturally less luminous than the white. However, there are three empirical
arguments that run against such a simple, unifactorial interpretation.
First, this hemifield X luminance interaction is usually evidenced with
RTs, while we noted it in accurancy, not RTs. Second, this LVF advantage
for black stimuli was observed in Exp. II only, yet the very same material
was used in Exp. I. However, this phenomenon was more expected in Exp.
II since the subjects had first to make an early racial categorization that
can be considered as a luminance discrimination task. But the fourth
experimental result deserves this interpretation: indeed, when the racial
decision alone was examined (Exp. IV), there was no sign of race X field
interaction and no sign of asymmetry.
We conclude that we have examined a two-step process of face recog-
nition in which each step, examined separately, is not lateralized but in
428 Raymond Bruyer and Myriam Schweich

which the combination reveals a right hemisphere superiority (and a

smaller race effects, with some familiarization effect, within this right
hemisphere). Therefore, it might be that some intermediate cognitive
operations appeared between the racial categorization and the recogni-
tion, that were responsible for the laterality effects and that remain to be
identified. The large difference in speed of processing between Exp. II
(1130 msec) and Exp. IV (625 msec) can be considered empirical evidence
favoring the existence of some such unknown intermediate operations;
nevertheless, this difference must not be overestimated since the subjects
had to process the directional information furnished by the arrow at the
fixation point only in Exp. II. However, some methodological qualifica-
tions are in order that invite us to experiment further. First, we recall the
various notions of face familiarity. Second, we have indicated the passage
from a discrimination between four (blocked) faces in Exp. I to the
discrimination between two in Exp. II (provided the racial classification
has been correct ely made). Third, we note that in Exps. II and IV the
subjects could base the racial decision on a "simple" dark/light discrim-
ination. Thus, experiments with Caucasian vs. Oriental faces could well be
fruitful.

ABSTRACT

Bruyer and Dussart (1985) have recently shown that the "race effect", i.e. the
difficulty in recognizing faces issuing from an ethnic group different from that of
the subject, is limited to the right visual field. They suspected familiarization to be
responsible for this asymmetry. In Exp. I, we tested this hypothesis by repeating
the experiment of Bruyer and Dussart with a greater number of trials. A sample of
16 subjects were given the task of recognizing black and white faces laterally
displayed for 180 msec. No laterality effect appeared, and the race effect was
observed to an equal degree in each hemifield at all stages of the experimental
session. It could thus be that various kinds of familiarization must be distin-
guished. In Exp. II, with 16 new subjects, the black and white faces were mixed so
that the subjects had first to perform racial categorization, then a recognition.
This time, an advantage of the left field appeared, the race effect was larger in the
right than in the left field, and the race effect decreased with familiarization in the
left field only. Two complementary experiments with 24 and 16 subjects showed
that this phenomenon was not explainable by laterality effects in the early racial
decision operation, but well by a lateralized effect of tasks requirements.

REFERENCES

BRIGHAM, J.e. The influence of race on face recognition. In H.D. Ellis, M.A. Jeeves, F.
Newcombe and A.W. Young (Eds.), Aspects of Face Processing. Dordrecht: Nijhoff,
1986.
BRUYER, R., and DUSSART, T. Lateral differences in the race effect in face recognition.
International Journal of Neuroscience, 28: 61-72, 1985.
 Race categorization and face recognition 429

BRUYER, R., RECTEM, D., and DUPUIS, M. Various kinds of face familiarity and a short
report on a case of prosopagnosia. Psychologica Belgica, 26: 221-225, 1986.
ELLIS, H.D. The role of the right hemisphere in face perception. In AW. Young (Ed.),
Functions of the Right Cerebral Hemisphere. London: Academic Press, 1983.
ELLIS, H.D. Introduction: processes underlying face recognition. In R. Bruyer (Ed.), The
Neuropsychology of Face Perception and Facial Expression. Hillsdale: Erlbaum,
1986.
GOLDBERG, E., and COSTA, L.D. Hemispheric differences in the acquisition and use of
descriptive systems. Brain and Language, 14: 144-173, 1981.
HAY, D.C. Asymmetries in face processing: evidence for a right hemisphere perceptual
advantage. Quarterly Journal of Experimental Psychology, 33A: 267-274, 1981.
HAY, D.C., and YOUNG, AW. The human face. In AW. Ellis (Ed.), Normality and
Pathology in Cognitive Functions. London: Academic Press, 1982.
HELLIGE, J.B., and JONSSON, J.E. Effects of stimulus duration on processing lateralized
faces. Bulletin of the Psychonomic Society, 23: 401-403, 1985.
HELLIGE, J.B., CORWIN, W.H., and JONSSON, J.E. Effects of perceptual quality on the
processing of human faces presented to the left and right cerebral hemisphere. Journal
of Experimental Psychology: Human Perception and Performance, 10: 90-107, 1984.
JONES, B. Sex and visual field effects on accuracy and decision making when subjects
classify male and female faces. Cortex, 15: 551-560, 1979.
JONES, B. Sex and handedness as factors in visual field organization for a categorization
task. Journal of Experimental Psychology: Human Perception and Performance, 6:
494-500, 1980.
LINDSAY, R.C.L., and WELLS, G.L. What do we really know about cross-race eyewitness
identification? In S.M.A Lloyd-Bostock and B.R. Clifford (Eds.) Evaluating Witness
Evidence: Recent Psychological Research and New Perspectives. New York: Wiley,
1983.
MARZI, C.A, TASSINARI, G., TRESSOLDI, P.E., BARRY, c., and GRABOWSKA, A Hemis-
pheric asymmetry in face perception tasks of different cognitive· requirement. Human
Neurobiology, 4: 15-20, 1985.
MOSCOVITCH, M. Information processing and the cerebral hemispheres. In M.S. Gazzan-
iga (Ed.), Handbook of Behavioral Neurobiology; 2: Neuropsychology. New York:
Plenum Press, 1979.
SERGENT, J. Theoretical and methodological consequences of variations in exposure
duration in visual laterality studies. Perception and Psychophysics, 31: 451-461,
1982a.
SERGENT, J. Influence of luminance on hemispheric processing. Bulletin of the Psychon-
omic Society, 20: 221-233, 1982b.
SERGENT, J. Role of the input in visual hemispheric asymmetries. Psychological Bulletin,
93: 481-512, 1983.
SERGENT, J. Influence of task and input factors on hemispheric involvement in face
processing. Journal of Experimental Psychology: Human Perception and Performance,
11: 846-861, 1985.
SERGENT, J., and BINDRA, D. Differential hemispheric processing of faces: methodolog-
ical considerations and reinterpretation. Psychological Bullettin, 89: 541-554, 1981.
SHEPHERD, J. Social factors in face recognition. In G. Davies, H. Ellis and J. Shepherd
(Eds.), Perceiving and Remembering Faces. London: Academic Press, 1981.
YOUNG, AW., HAY, D.C., and MCWEENY, K.H. Right cerebral hemisphere superiority
for constructing facial representation. Neuropsychologia, 23: 195-202, 1985.

R. Bruyer, UCL-NEXA, Avenue Hippocrate, 5545, B-1200 Bruxelles, Belgium.