NeuroImage 15, 353–372 (2002)
doi:10.1006/nimg.2001.0982, available online at http://www.idealibrary.com on
Inversion and Contrast Polarity Reversal Affect both Encoding
and Recognition Processes of Unfamiliar Faces:
A Repetition Study Using ERPs
Roxane J. Itier 1 and Margot J. Taylor
CerCo-CNRS, Université Paul Sabatier, 31062 Toulouse, France
Received May 7, 2001
Using ERPs in a face recognition task, we investigated
whether inversion and contrast reversal, which seem to
disrupt different aspects of face configuration, differentially affected encoding and memory for faces. Upright,
inverted, and negative (contrast-reversed) unknown
faces were either immediately repeated (0-lag) or repeated after 1 intervening face (1-lag). The encoding
condition (new) consisted of the first presentation of
items correctly recognized in the two repeated conditions. 0-lag faces were recognized better and faster than
1-lag faces. Inverted and negative pictures elicited
longer reaction times, lower hit rates, and higher false
alarm rates than upright faces. ERP analyses revealed
that negative and inverted faces affected both early (encoding) and late (recognition) stages of face processing.
Early components (N170, VPP) were delayed and enhanced by both inversion and contrast reversal which
also affected P1 and P2 components. Amplitudes were
higher for inverted faces at frontal and parietal sites
from 350 to 600 ms. Priming effects were seen at encoding stages, revealed by shorter latencies and smaller
amplitudes of N170 for repeated stimuli, which did not
differ depending on face type. Repeated faces yielded
more positive amplitudes than new faces from 250 to
450 ms frontally and from 400 to 600 ms parietally.
However, ERP differences revealed that the magnitude of this repetition effect was smaller for negative
and inverted than upright faces at 0-lag but not at
1-lag condition. Thus, face encoding and recognition
processes were affected by inversion and contrast-reversal differently. © 2002 Elsevier Science
INTRODUCTION
Adults have remarkable expertise in learning and
recognizing new faces, yet their recognition perfor1
To whom correspondence should be addressed at CERCO,
CNRS–UMR 5549, Faculté de Médecine de Rangueil, Université
Paul Sabatier, 133, Route de Narbonne, 31062 Toulouse, France.
Fax: ⫹33-562-172-809. E-mail: [email protected].
mance is greatly impaired by face inversion (Farah et
al., 1995a; Hochberg and Galper, 1967; Rhodes et al.,
1993), the decrement in performance being more
marked for faces than many other objects (Yin, 1969).
This disproportional effect of inversion for faces (called
the face inversion effect) is thought to reflect the particularity of faces over objects and hence has been
widely used to study what it is, in the processing of
faces, that makes them special.
In the classification of Diamond and Carey (1986),
human faces share a common configuration (first-order
relational features) formed by horizontally aligned
eyes that are found always above the nose, itself always above the mouth. Second-order relational features defined as “distinctive relations among the elements that define the shared configuration” (Diamond
and Carey, 1986) are the specific arrangements of
these internal features in each face. Inversion has been
used by different behavioral studies to demonstrate the
importance of configural information in normal upright
face processing. For instance, parts of composite faces
are better identified when composites are presented
upside down than when presented right side up (Young
et al., 1987), suggesting that encoding of components of
upright faces is affected by the context of other components (the relations between them) and that inversion
reduces this context effect. Kemp et al. (1990) showed
that detection of spatial displacement of features was
less accurate for inverted faces. Similarly, Rhodes et al.
(1993) found that inversion disrupted the detection of
spatial–relational changes (which they called configuration) more than that of changes in featural information in an old–new recognition task. Inversion also
reduced the ratings of grotesqueness of spatially distorted and thatcherized faces (normal faces with inverted eyes and mouth that look monstrous when presented upright but not when inverted), whereas it did
not change the ratings of posed expressions, supposedly conveyed by marked differences in features (Bartlett and Searcy, 1993). Searcy and Bartlett (1996) also
found that inversion effects on response latencies in a
353
©
1053-8119/02 $35.00
2002 Elsevier Science
All rights reserved.
354
ITIER AND TAYLOR
simultaneous paired comparison task were more pronounced for changes in relational information than for
changes in featural information. In agreement with
this, perceived distinctiveness due to changes in relational information was reduced much more than that
due to changes in local information when faces were
presented upside down, and recognition of these localchanged faces was insensitive to inversion (Leder and
Bruce, 1998).
Thus, these studies, among others, demonstrate that
encoding of upright faces relies on the relational information (or configuration) and that inversion impairs
this encoding more than the encoding of other types of
information such as isolated features (Rhodes et al.,
1993; Bartlett and Searcy, 1993; Searcy and Bartlett,
1996). These results support a dual-mode hypothesis
suggesting that upright faces are processed by two
modes: one mode specialized in encoding the spatial–
relational information and a second mode specialized
in encoding components of a face. In this view, inversion impairs the first mode but not the second one,
resulting in a processing of inverted faces based almost
entirely on their components. The view that inverted
faces are processed somewhat like objects, on the basis
of their parts, is supported by case reports of prosopagnosia, as some prosopagnosic patients are not affected by inversion, while some are even better at recognizing inverted than upright faces (De Gelder and
Rouw, 2000; Farah et al., 1995b).
The holistic hypothesis is another view of upright
face processing and stipulates that upright faces are
encoded as unparsed perceptual wholes, not decomposable into their constituent features (Tanaka and
Farah, 1993). In this view, the holistic processing used
for upright faces would be lost with inversion, and
inverted faces, like objects, would be processed only on
the basis of their parts (Farah et al., 1995a,b). Recent
evidence is, however, in contradiction with the holistic
model of processing faces. Leder and Bruce (2000) demonstrated that the face inversion effect and its size are
determined by the disruption of the relational information processing and not by the disruption of holistic
processing. Leder et al. (2001) confirmed this result
and suggested that the processing of spatial information is sensitive to local properties of the facial features
involved. Furthermore, Cabeza and Kato (2000)
showed that the prototype effect for configural prototypes was eliminated by inversion but that for featural
prototypes was not. In a discrimination task, Freire et
al. (2001) provided direct evidence that the disruption
of configural (relational) information by inversion occurs primarily at encoding stages of face processing,
with no effect of inversion on faces differing on featural
information. In addition, they showed that the delay
between presentation and test did not affect matching
accuracy (for short delays), suggesting that once en-
coded, retention in memory of featural and relational
information is similar.
Thus, there is considerable evidence in the behavioral literature that the inversion effect is mainly due
to the disruption of relational information processing,
principally at the encoding stages of face processing. In
the present study, using the ERP technique, we
wanted to determine if inversion also affects memory
stages of processing and, if it does, if this is due to
impaired encoding.
Like inversion, contrast reversal affects recognition
of faces more than that of objects (Subramaniam and
Biederman, 1997); photo negatives of faces (contrastreversed) are more difficult to recognize than photo
positives (Galper, 1970; Johnston et al., 1992; Kemp et
al., 1990) despite the fact that contrast-polarity reversal preserves all edges and spatial frequencies. This
negative effect could be due to the disruption of 3D
shape-from-shading information (Kemp et al., 1996).
Lewis and Johnston (1997) suggested that contrastreversal of faces disrupted configural information as
increased reaction times were found for judging that
negative thatcherized faces were different from normal
faces. Kemp et al. (1990) found that detection of displacement of features was less accurate for negative
and inverted faces than for upright faces (experiment
1), but not significantly different between inverted and
negative faces. It seems, therefore, that some relational information is disrupted by contrast reversal
(Kemp et al., 1990; Lewis and Johnston, 1997). However, the recent finding by Hole et al. (1999) that similar chimeric-face effects were obtained with both positive and negative upright faces suggests that negative
faces evoke a form of relational processing similar to
that occurring with upright faces. Hole et al. (1999)
argued that there may be two types of relational processing in normal faces, one of which being a holistic
encoding permitting rapid determination that it is a
face that is being perceived and the other one being
relational processing. They argued that negation could
leave the holistic processing intact while disrupting the
relational information, whereas inversion would disrupt both. Furthermore, the fact that inversion and
contrast reversal have additive effects (Kemp et al.,
1990; Lewis and Johnston, 1997) suggests that they
disrupt different configural information. This is
strengthened by the fact that no effect of inversion was
found in experiment 5 of Kemp et al. (1990) using only
a face surround and circles instead of internal features,
whereas a negative effect was found. Thus, there are
contrasting views of what exactly contrast reversal
does on face recognition and what differences exist
between inversion and contrast-reversal effects.
Similarly, the stage at which the negative effect occurs is not well established. As Kemp et al. (1990) had
subjects comparing simultaneously presented faces,
the impairment found with negative faces was percep-
CONFIGURAL CHANGES AND FACE RECOGNITION
tual and not due to memory load. This would imply
that, like inversion, contrast reversal disrupts the encoding stage of face processing. However, Liu and
Chaudhuri (1997) reported that the decrease in performance with negative faces was not due to a deficiency
at the encoding level, but was a problem at the memory
stage.
The behavioral literature thus provides evidence for
perceptual deficiency at the encoding level following
inversion, but is not clear at which stage the deficiency
occurs with contrast-reversed faces and what exactly is
disrupted by contrast reversal.
In addition to the psychological investigations mentioned above, processing of faces has been intensively
studied with various other approaches such as intracranial recordings (Allison et al., 1994; McCarthy et al.,
1999), neuropsychological reports of prosopagnosia
(Damasio et al., 1982; De Gelder and Rouw, 2000), and
neuroimaging studies such as PET and fMRI (Haxby et
al., 1996; Kanwisher et al., 1997; Puce et al., 1995). Of
particular relevance to the present study are the electrophysiological ERP (event-related potential) investigations recording from the surface of the scalp. The
early ERP negative component, N170, largest at posterior temporal sites, has been shown to be face-sensitive (Bentin et al., 1996; Bötzel et al., 1995; George et
al., 1996). Bentin et al. (1996) proposed that it reflected
detection of human physionomic facial features, particularly eyes, as N170 was not as evident to parts of
faces, such as noses, or to animal faces, but was very
prominent to eyes. Eimer (1998) found that the presence or absence of eyes did not affect N170 amplitude.
Hence, N170, which seems unaffected either by familiarity (Bentin and Deouell, 2000; Eimer, 2000a) or by
directed attention (Séverac-Cauquil et al., 2000), is
likely to reflect face encoding processes, especially the
stages where representations of global face configurations are generated (Eimer, 2000b). Typically, N170 is
delayed (Bentin et al.,1996; Linkenkaer-Hansen et al.,
1998; Rossion et al., 1999a, 2000; Taylor et al., 2001a)
and larger in amplitude (Linkenkaer-Hansen et al.,
1998; Rossion et al., 1999a, 2000; Taylor et al., 2001a)
for inverted faces but not for inverted objects (Rossion
et al., 2000), in favor of special processing of faces
compared to other stimuli. This sensitivity of N170 to
inverted faces has been interpreted as reflecting a disruption of configural information at the encoding stage
of face processing (Rossion et al., 1999a, 2000). To our
knowledge, no study has investigated the effect of contrast reversal on the N170; an effect of contrast reversal on this component would imply that, like the inversion effect, the negative effect takes place at the
encoding stage. Similarly, if no effect of negation is
found, it would mean that this effect takes place at a
memory stage, as claimed by Liu and Chaudhuri
(1997). ERPs could thus help elucidate what is still
unclear in the behavioral literature and could, in ad-
355
dition, provide information on qualitative differences
in processing between inverted and contrast-reversed
faces.
The vertex positive peak (VPP) also appears to respond preferentially to faces (Jeffreys, 1989; Rossion et
al., 1999a,b), but despite its temporal synchronicity
with N170, some authors argue that it is not the positive counterpart of the N170 (Bötzel et al., 1995;
Schendan et al., 1998; Rossion et al., 1999b; Taylor et
al., 1999). Apart from Jeffreys (1993) who studied the
effect of inversion and contrast reversal on the VPP, no
other study has, to our knowledge, investigated contrast reversal on this early component.
Other early ERP components such as P1 and P2 have
also been studied, but their sensitivity to faces remains
controversial (Halit et al., 2000; Linkenkaer-Hansen et
al., 1998; Rossion et al., 1999a,b) and few papers have
investigated effects of configural changes on these components. Thus, in this study, we investigated the effect
of inversion and contrast reversal on all four of these
early components.
ERPs are also a reliable means to assess memory,
and ERP differences are predictive of memory performance, what Paller et al. (1987) termed Dm (difference
in subsequent memory). The Dm, measured as the
amplitude difference at encoding stages between items
that are later recognized versus not recognized, was
first observed for words (Paller et al., 1987), but is also
found for faces, and proved to be memory-related and
not perceptual (Sommer et al., 1991). Amplitude differences are also seen when comparing new items to old
items, a finding referred to as ERP repetition effect or
old–new differences. Repetition effects have been
found for words, drawings of objects (Rugg et al., 1995),
and faces (George et al., 1997; Hepworth et al., 2001;
Schweinberger et al., 1995). Whether ERPs obtained
for repeated items compared to new items were more
positive or negative depended on the electrodes studied. Experiments using faces have found that at parietal and frontal sites, ERPs elicited by repetition were
more positive than ERPs elicited by the first presentation of faces (Munte et al., 1997; Paller et al., 1999,
2000; Sommer et al., 1995, 1997). Hertz et al. (1994)
found a priming effect for an explicit recognition task
but not for an implicit one, and their repetition effect
was seen in more negative-going ERPs recorded at
temporal and occipital electrodes (they did not look at
frontal areas). Face repetition studies in ERPs have
analyzed primarily late components, but there are no
reports of configural effects on these late recognition
potentials.
Thus, there are to date few experiments that elucidate the role of inversion and contrast reversal on early
ERP components reflecting encoding stages, and none
has investigated their possible effects on late components reflecting memory stages. The goals of our study,
consisting of an explicit recognition task of repeated
356
ITIER AND TAYLOR
FIG. 1.
Examples of upright, negative, and inverted faces used in the experiment.
unfamiliar faces presented upright, inverted, and in
negative format, are (i) to investigate the role of contrast reversal on encoding and memory stages of face
processing, (ii) to investigate the role of inversion on
memory as assessed by ERPs (as its role on encoding is
well established), and (iii) to compare effects of inversion and contrast reversal at both stages of processing.
MATERIALS AND METHODS
Subjects
Thirty-six subjects (18 female) participated in the
study. One female was rejected because of poor task
performance and 1 male was rejected because of a
problem in ERP acquisition. The 34 remaining subjects
had normal or corrected-to-normal vision and ranged
from 20 to 33 years (mean age 24.9 years). Two subjects
were left-handed (1 male). The experimental procedure
was approved by the French Comité Opérationnel pour
l’Ethique dans les Sciences de la Vie du CNRS and
informed written consent was obtained from all participants.
Stimuli
Seven-hundred and twenty gray scale pictures of
caucasian male and female faces with neutral expressions were used. All the faces were unknown to the
subjects. Half of the faces were of clean-shaven males.
All faces were without glasses, earrings, jewelry, or
paraphernalia. Inverted and negative (contrast-reversed) pictures were obtained from the upright normal pictures using PhotoShop 5.0 software (see Fig. 1).
Testing consisted of four blocks of trials. Within each
block, three different sets of faces were used: upright,
inverted, and negative faces. Each set contained 60
faces of which 20 were repeated. In each set, half of the
repeated faces were immediately repeated (0-lag condition) and the other half were repeated after one intervening picture (1-lag condition) (we did not use
other lags as performances for inverted and negative
faces in a prestudy using lags of two, dropped to less
than 35% recognition). In each condition, half of the
repeated faces were female. Faces were presented sequentially and all sequences were different. Block order and set order within block were counterbalanced
across subjects with the constraint that stimulus types
were not repeated consecutively. For example, if subject 1 viewed inverted faces in set 1, these faces were
presented as contrast-reversed in set 2 for subject 2
and as upright in set 3 for subject 3. This counterbalancing was in case some series of faces were more
difficult than others and we did not want sets to be
confounded with face type. Faces that appeared in one
format (e.g., upright) were never presented to the same
subject in a different format or in a different block. All
stimuli subtended a visual angle of approximately 11°
⫻ 10° and were centered on the screen.
Procedure
Subjects performed the task in a dimly lit room.
Stimuli were presented on the computer screen at 80
cm from the subjects. The same PC controlled the stimulus presentation and recorded the subjects’ responses
and reaction times (E-Prime v. 4 software).
Participants received a brief training block. Faces
were presented sequentially for 500 ms to allow sufficient encoding time but to minimize eye movements.
The interstimulus interval was 1320 ms, during which
a green cross appeared centered on the screen. Subjects
were instructed to fixate on the cross. Short pauses of
a few minutes were given between sets and between
blocks. The subject’s task was to press the space bar of
the keyboard as fast as possible as soon as he/she
recognized a face that was previously presented. It was
explicitly mentioned to subjects that the repetition of a
picture could occur randomly within four subsequent
pictures only, in order to prevent subjects from trying
357
CONFIGURAL CHANGES AND FACE RECOGNITION
to remember all the faces of a set, but to stay naı̈ve
concerning the 0- and 1-lag conditions.
Electrophysiology
ERPs were recorded from 32 scalp electrodes in an
electrode cap (EasyCap) at a sampling rate of 500 Hz
(NeuroScan, v. 4.0). Electrode impedances were ⱕ5 k⍀.
EEG was amplified with a gain of 500 using a SynAmps system. An average reference montage was used,
with Cz as the reference lead during acquisition; the
average reference was calculated offline and Cz was
reinstated. EOG from the outer canthi and left supraorbital ridge monitored horizontal and vertical eye
movements, respectively. An EOG reduction routine
was run on the continuous files prior to epoching into
1-s sweeps with a 100-ms prestimulus baseline; trials
contaminated with ocular movements were rejected
before averaging. Trials were then averaged according
to the nine possible stimulus categories based on face
type (upright, inverted, or negative) and memory condition (new, 0-lag, or 1-lag) and were digitally filtered
(0.8 – 60 Hz). The “new” faces were defined as faces on
their first presentation which were subsequently repeated and correctly recognized by the subject. New
faces which were never repeated were not included as
we had no subsequent measure as to whether they
were or were not encoded well enough to have been
recognized. It was not possible to analyze the new
trials which were subsequently repeated but not recognized because accuracy was high.
Data Analyses
Hit rates for repeated faces, false alarm rates for new
faces, and mean reaction times (RTs) recorded for hits
were calculated for all categories.
Peak analyses were performed within a ⫾30-ms window around the maximum of the grand average means.
They included four components: P1 (maximal around
120 ms), N170 (maximal around 170 ms), P2 (maximal
around 250 ms), and VPP (maximal around 170 ms).
P1, N170, and P2 were measured at eight electrodes:
the temporal–parietal sites TP9 and TP10, the posterior parietal sites P7 and P8, the occipitoparietal sites
PO9 and PO10, and the occipital O1 and O2 electrodes.
For each subject, the latency of each component was
taken at the electrode where the amplitude was maximum over each hemisphere and the amplitude was
measured at the other electrodes over that hemisphere
at that latency (Picton et al., 2000). The vertex positive
peak VPP was measured at Fz.
Mean ERP amplitudes were also calculated for each
of the nine categories in order to compare memory
effects across conditions and stimuli, within seven
50-ms windows starting from 250 to 600 ms (Fig. 2).
The analyses included frontal sites Fp1, Fp2, Fz, F3,
and F4, and parietal sites POz and Pz.
Repeated measures analyses of variance (ANOVA)
were conducted using Greenhouse–Geisser adjusted
degrees of freedom and post hoc t statistics with Bonferroni corrections for multiple comparisons. For latency analyses, type (3), condition (3), and hemisphere
(2) were intrasubject factors. For amplitude analyses,
electrode was added to intrasubject factors (number of
electrodes varied according to the components studied).
Only P ⬍ 0.01 values were considered significant.
In order to assess the amplitude of memory effects,
ERP differences (0-lag minus new and 1-lag minus
new) and their scalp potential maps were created (using ERPA software (Perrin et al., 1989; www.pavux.
com/erpa)). Finally, using BESA (Scherg, 1990; www.
besa.de), source analysis of ERPs obtained for 0-lag
and 1-lag conditions was modeled on grand averages.
RESULTS
Behavioral Data
Behavioural results are summarized in Table 1.
RT analyses revealed a main effect of stimulus type
(F(2, 65.3) ⫽ 14.87, P ⬍ 0.0001) due to upright faces
eliciting faster responses than inverted (P ⬍ 0.0001)
and negative (P ⬍ 0.0001) faces, which did not differ
between themselves. Immediately repeated faces (0lag condition) elicited faster RTs than did faces repeated after one picture (1-lag condition) as shown by a
main effect of condition (F(1, 33) ⫽ 71.81, P ⬍ 0.0001).
No interactions were found.
For hit rates, a main effect of stimulus type (F(1.6,
54.3) ⫽ 41.71, P ⬍ 0.0001) was due to upright faces
being better recognized (87.3%) than inverted (78%)
and negative (79.2%) faces (P ⬍ 0.0001 for each comparison); no significant difference between inverted
and negative stimuli was found. Higher hit rates were
found for 0-lag (95%) compared to 1-lag (67.9%) faces
(F(1, 33) ⫽ 167.71, P ⬍ 0.0001). A type ⫻ condition
interaction (F(1.8, 58.7) ⫽ 12.90, P ⬍ 0.0001) was due
to greater decreases in performances at condition 1-lag
for inverted and negative faces than for upright faces
(Table 1).
The analysis of false alarm rates revealed an effect of
stimulus type (F(1.9, 61.8) ⫽ 23.37, P ⬍ 0.0001) that
was due to upright faces eliciting lower false alarm
rates than inverted faces (P ⬍ 0.0001) and negative
faces (P ⬍ 0.0001), which did not differ between themselves.
FIG. 2. Grand averaged ERPs overplotted for 0-lag, 1-lag, and new upright faces, at most electrodes sites with the peaks measured
indicated. Vertical lines represent the limits of the period during which mean amplitudes were analyzed (250 – 600 ms).
358
ITIER AND TAYLOR
FIG. 3. P1, N170, and P2 components displayed for upright, inverted, and negative faces at temporoparietal, parietal, occipitoparietal,
and occipital sites where these components were measured. VPP is displayed at Fz. Left hemisphere sites are on the left.
360
ITIER AND TAYLOR
TABLE 1
Mean Reaction Time (ms), Hit Rate (%), and False Alarm Rate (%) by Stimulus Type, for Repeated Faces and New Faces
(Standard Error in Parentheses)
Stimulus type
0-lag
1-lag
0-lag
1-lag
False
alarm rate
(%)
Upright faces
Negative faces
Inverted faces
540 (12)
576 (13)
573 (11)
623 (14)
650 (17)
659 (17)
98.10 (0.54)
94.67 (1.30)
92.35 (1.15)
76.55 (2.38)
63.64 (2.72)
63.57 (2.91)
1.68 (0.35)
3.55 (0.43)
4.01 (0.48)
Reaction time (ms)
Electrophysiological Data
Effects of stimulus type and memory condition, on
P1, VPP, N170, and P2 components are shown in Figs.
3 and 4.
Peak Analyses
P1. For P1 latency, only a main effect of stimulus
type (F(1.7, 56.1) ⫽ 15.91, P ⬍ 0.0001) was found;
inverted faces eliciting longer P1s compared to negative (P ⬍ 0.0001) and upright faces (P ⬍ 0.001) with no
difference between the latter two.
P1 amplitude showed a main effect of stimulus type
(F(2, 63.9) ⫽ 13.29, P ⬍ 0.0001), due to smaller P1s for
negative than inverted (P ⬍ 0.0001) and upright faces
(P ⬍ 0.005), but this effect was only seen at occipital
sites, shown by a type ⫻ electrode interaction (F(3.2,
102.8) ⫽ 22.33, P ⬍ 0.0001). No difference between
upright and inverted stimuli was found. No significant
differences in amplitude or latency were seen when
memory conditions were compared.
VPP. A stimulus type main effect (F(1.8, 61.0) ⫽
36.70, P ⬍ 0.0001) was due to longer VPP latencies for
inverted and negative faces compared to upright faces
(P ⬍ 0.0001); VPP was not significantly different between negative and inverted stimuli. Memory condition did not affect VPP latency.
For VPP amplitude, a main effect of stimulus type
was found (F(1.8, 58.4) ⫽ 38.22, P ⬍ 0.0001) (Fig. 3),
with larger amplitudes for negative faces than inverted
(P ⬍ 0.0001) and upright (P ⬍ 0.0001) faces, which did
not differ significantly. In addition, a main effect of
memory condition (F(1.9, 61.6) ⫽ 5.40, P ⬍ 0.01) was
due to a smaller VPP amplitude for 0-lag faces compared to new faces (P ⬍ 0.005) (Fig. 4).
N170. N170 latency showed a main effect of stimulus type (F(2, 64.6) ⫽ 57.97, P ⬍ 0.0001) due to inverted faces eliciting longer N170s than negative faces
(P ⬍ 0.01) and upright faces (P ⬍ 0.0001). The latter
two also differed, with N170 latency being longer to
negative than to upright faces (P ⬍ 0.0001) (see Fig. 3).
A memory condition main effect (F(1.7, 56) ⫽ 9.84, P ⬍
0.0001) was due to shorter N170 latencies for faces
when they were repeated (P ⬍ 0.005 for both), with no
significant difference between 1- and 0-lag stimuli.
Hit rate (%)
N170 amplitudes were larger (more negative) for
contrast-reversed than inverted faces, which in turn
were larger than upright faces (P ⬍ 0.0001 for each
comparison), as revealed by a main effect of stimulus
type (F(1.9, 61.9) ⫽ 71.04, P ⬍ 0.0001) (Fig. 3). An
electrode effect (F(1.8, 60.4) ⫽ 28.06, P ⬍ 0.0001) for
amplitudes was due to N170s being smallest at occipital electrodes (Figs. 3 and 4). A main effect of memory
condition (F(1.9, 63.1) ⫽ 6.82, P ⬍ 0.005) was due to
immediately repeated faces eliciting smaller N170s
than new faces (P ⬍ 0.005); this effect was most salient
at temporoparietal and occipitoparietal sites as revealed by a condition ⫻ electrode interaction (F(3.3,
108.6) ⫽ 10.20, P ⬍ 0.01), although visible at all sites
(Fig. 4). The effect of hemisphere was examined, but
only a trend was seen (F(1, 33) ⫽ 5.30, P ⬍ 0.05), due
to N170s being slightly larger over the right hemisphere.
As P1 had shown effects of face type, we performed
peak-to-peak analyses to verify that the significant
differences obtained for N170 between inverted and
negative faces were not attributable to these early P1
differences. Thus, we took the latency and amplitude
differences between N170 and P1 for each subject (for
amplitudes, maximum values for each component were
used) and performed 3 (type) ⫻ 3 (memory condition) ⫻
2 (hemisphere) ANOVAs. For latencies, a main effect of
type (F(1.9, 61.6) ⫽ 15.18, P ⬍ 0.0001) was due to
inverted and negative faces eliciting larger latency differences compared to upright faces (P ⬍ 0.005). However, there was no significant difference between inverted and negative faces (P ⬍ 0.47). The same type
effect was found for amplitudes (F(1.7, 55.9) ⫽ 16.48,
P ⬍ 0.0001) and was due to larger amplitude differences for negative and inverted compared to upright
faces (P ⬍ 0.01 and P ⬍ 0.0001, respectively). Again, no
effect between inverted and negative faces was found
(P ⬍ 0.07). The fact that differences between inverted
and negative faces were no longer significant using a
peak-to-peak analysis demonstrates that differences in
latencies and amplitudes between these two face types
were due to P1 effects. Therefore, inverted and negative faces both delay and enhance N170 similarly. The
effects of condition found with this peak-to-peak analysis did not change from the initial analyses.
361
CONFIGURAL CHANGES AND FACE RECOGNITION
TABLE 2
Significant Main Effects of Stimulus Type and Memory Condition on Mean Amplitudes Analyzed over Seven 50-ms Time
Windows at Frontal (Fp1, Fp2, Fz, F3, F4) and Parietal (Pz, POz) Sites, with F and P Values
Time windows (ms)
250–300
Frontal sites
Effect of
stimulus type
Effect of
memory
Parietal sites
Effect of
stimulus type
Effect of
memory
300–350
350–400
400–450
450–500
F ⫽ 27, 11
P****
0-L ⬎ 1-L ⬎
NEW
F ⫽ 21, 1
P****
i ⬎ n ⬎ up
F ⫽ 32, 33
P****
0-L and 1-L
⬎ NEW
F ⫽ 24, 76
P****
i ⬎ n ⬎ up
F ⫽ 16, 18
P****
0-L and 1-L
⬎ NEW
F ⫽ 31, 35
P****
i ⬎ n ⬎ up
F ⫽ 7, 42
P**
0-L and 1-L
⬎ NEW
F ⫽ 20, 53
P****
i ⬎ n ⬎ up
—
—
F ⫽ 8, 55
P***
i ⬎ n and
up
F ⫽ 6, 67
P**
i⬎n
F ⫽ 5, 57
P*
i⬎n
F ⫽ 5, 4
P*
i ⬎ up
F ⫽ 5, 2
P*
NEW ⬎ 0-L
—
F ⫽ 11, 59
P****
i ⬎ n and
up
F ⫽ 10, 78
P****
0-L ⬎ NEW
and 1-L
F ⫽ 13, 02
P****
0-L ⬎ NEW
and 1-L
F ⫽ 11, 82
P****
0-L ⬎ NEW
and 1-L
F ⫽ 5, 12
P*
0-L and 1-L
⬎ NEW
—
—
—
500–550
F ⫽ 9, 77
P****
i and n ⬎ up
—
550–600
F ⫽ 5, 87
P***
i ⬎ up
—
Note. up, upright faces; i, inverted faces; n, negative faces. 0-L, condition 0-lag; 1-L, condition 1-lag; NEW, condition new (encoding). Main
effects P values: *P ⬍ 0.01; **P ⬍ 0.005; ***P ⬍ 0.001; ****P ⬍ 0.0001. Post hoc tests are also reported (e.g., 0-L ⬎ 1-L means that the main
effect of memory is due to 0-lag faces eliciting a greater mean amplitude than 1-lag face).
P2. Stimulus type did not affect P2 latency. A main
effect of memory condition (F(2, 65.3) ⫽ 18.13, P ⬍
0.0001) was due to P2 being longer for new faces than
0-lag (P ⬍ 0.001) and 1-lag faces (P ⬍ 0.005); no difference was seen between the two lags for repeated
faces. A hemisphere effect (F(1, 33) ⫽ 11.17, P ⬍ 0.005)
was seen, due to shorter P2 latencies over the left than
the right hemisphere.
An electrode main effect was seen for P2 amplitude
(F(2, 66.8) ⫽ 67.11, P ⬍ 0.0001) as P2 was largest at
occipital sites and smallest at temporoparietal sites
(Fig. 3). A main effect of stimulus type (F(1.8, 58) ⫽
5.39, P ⬍ 0.01), which was due to negative faces eliciting a larger P2 than upright faces (P ⬍ 0.005), was
most salient at parietal and occipital sites, seen in a
type ⫻ electrode interaction (F(4.2, 139) ⫽ 4.56, P ⬍
0.001). No effect of memory condition was found.
Peak-to-peak analyses were also performed between
P2 and P1, to ensure that type effects found on P2
amplitude were not influenced by early P1 effects. No
new results were found; thus effects seen on P2 were
not due to those found with P1.
As encoding components and all behavioral measures showed effects of face type, we also performed
correlation analyses to see if a deficient encoding could
be related directly to decreased performances and increased reaction times. No correlations were found between differences between face types for hits and RTs
(0-lag and 1-lag conditions) and differences between
face types for P1 and N170 amplitudes or latencies
(“new” condition). These are in accordance with Rossion et al. (1999a) who also did not find any correlations
between accuracy, RTs, and N170 latency or amplitude
differences.
Mean Amplitudes over Time Windows (see Fig. 5)
Mean amplitude over seven 50-ms windows (250 –
600 ms) for all three face types and memory conditions
are depicted in Fig. 5. Table 2 shows the corresponding
F and P values.
Frontal sites. Inverted stimuli led to the most positive amplitudes from 300 to 600 ms (Fig. 5a) followed
by negative and then upright faces. A condition main
effect was found from 250 to 450 ms due to repeated
faces eliciting more positive amplitudes than new faces
(Fig. 5b). A type ⫻ condition interaction was found
between 400 and 450 ms (F(42.5, 8.65) ⫽ 4.91, P ⫽
0.005): the mean amplitudes were ordered 1-lag ⬎
0-lag ⬎ new for inverted and negative faces whereas
for upright faces, the order was 0-lag ⬎ 1-lag ⬎ new.
No electrode effects were found.
Parietal sites. Mean amplitude at parietal sites was
affected by stimulus type 50 ms after frontal sites,
beginning at 350 ms (Table 2, Fig. 5c). Again, inverted
faces elicited the largest amplitudes while the amplitudes for negative and upright faces was almost identical throughout the time windows. The memory condition effect appeared later as well, beginning at 400
ms and lasting until 600 ms (Fig. 5d). A condition ⫻
FIG. 4. Grand averaged ERPs as a function of memory condition (new, 0-lag, and 1-lag). P1, N170, and P2 components are displayed for
upright faces at temporoparietal, parietal, occipitoparietal, and occipital sites for both hemispheres and VPP at Fz. Left hemisphere sites are
on the left.
CONFIGURAL CHANGES AND FACE RECOGNITION
363
FIG. 6. Scalp distributions obtained for 0-lag minus new ERP differences. All three face types (upright, negative, and inverted) are
presented at latencies 230, 292, 354, 414, 476, 538, and 598 ms with a ⫾3.7-␮V amplitude scale. Note the greater frontal and parietal
activation for upright faces.
FIG. 7. Scalp distributions obtained for 1-lag minus new ERP differences. All three types (upright, negative, and inverted) are presented
at latencies 230, 292, 354, 414, 476, 538, and 598 ms with a ⫾2.6-␮V amplitude scale. Note the change in amplitude scale compared to Fig.
6 and the absence of parietal activation.
364
ITIER AND TAYLOR
FIG. 5. Mean amplitudes obtained over seven 50-ms windows at frontal (a, b) and parietal (c, d) sites as a function of face type (a, c)
(upright, inverted, and negative) and memory condition (b, d) (new, 0-lag, and 1-lag). Note the largest difference between repeated (0-lag,
1-lag) faces and new faces frontally (b) from 250-450 ms, whereas for the parietal sites (d) the largest difference is due to greater activation
to 0-lag than either 1-lag or new faces, from 400 to 600 ms. Significance levels of these effects are in Table 2.
electrode interaction was found from 450 to 550 ms
(P ⬍ 0.001) and was due to a greater amplitude difference between Pz and POz at condition 0-lag compared
to new and 1-lag conditions. No type ⫻ condition interactions were found.
ERP Differences
To further explore memory effects, 0-lag and 1-lag
stimuli ERPs were subtracted from new-faces ERPs,
for each stimulus type. Topographical maps of those
differences were obtained using ERPA software. As
shown in Fig. 6, an increased positivity appeared in
frontal areas around 300 ms for the 0-lag-minus-new
difference and lasted until 450 ms for upright faces. It
was accompanied by increased negativity bilaterally in
occipitotemporal regions around 300 ms. This frontal
activity was larger for upright faces than negative and
inverted faces, for which it disappeared more rapidly
(around 350 ms). A relative increase in positivity in
parietal areas could be observed around 480 ms for
upright faces; it was very small for negative faces and
completely absent for inverted faces. Figure 7 shows
the 1-lag-minus-new difference map, where the increased amplitude is much smaller than for the 0-lagminus-new difference map (note the change in amplitude scale). The same frontal increase in positivity was
seen from around 300 to 450 ms but was more promi-
nent for negative faces in this repetition condition.
Increased amplitudes frontally also lasted longer for
inverted faces, up to 500 ms. Comparable parietal activity was not seen except for upright faces where it
was very small and late compared to the parietal increase in positivity displayed in Fig. 6. Figure 8 represents the mean amplitude differences at frontal sites
over the 400- to 450-ms window where a type ⫻ condi-
FIG. 8. Mean amplitude differences between memory conditions
(0-lag minus new, 1-lag minus new), from 400 to 450 ms at frontal
sites (Fp1, Fp2, Fz, F3, F4) as a function of face type (inverted,
negative, upright).
CONFIGURAL CHANGES AND FACE RECOGNITION
365
tion interaction was found. For upright faces, there
was greater increased positivity frontally with faces
repeated immediately than with faces repeated after
1-lag, but this pattern was reversed for the inverted
and negative faces.
present even for immediately repeated stimuli. Moreover, using only one intervening face (1-lag), performances dropped from 98 to 76% for upright faces and
from about 93 to 63% for inverted and negative faces,
reflecting the difficulty of the task.
Dipole Localization
Effects of Stimulus Type
Source analysis using BESA99 software was performed on the grand averages. Figures 9b and 9c represent dipoles obtained for upright faces for 0-lag and
1-lag conditions, respectively. Dipoles were constrained in symmetry on time periods chosen as a
function of the component studied: P1 was fitted between 90 and 120 ms, N170 between 140 and 190 ms,
and P2 between 190 and 260 ms. As VPP occurred at
the same latency range as N170, we introduced two
dipole pairs on this fitting period to account for the
VPP, but we did not obtain stable solutions across
conditions. Therefore, we kept only the pair of dipoles
for N170. We fitted the later frontal activity on a 270to 380-ms period using a regional source and the remaining activation was fitted between 400 and 600 ms
with one further pair of dipoles. Each dipole pair was
fitted independently from the others initially, and then
the regional source was fitted when all dipoles were on.
A pair of dipoles was obtained in the parietooccipital
cortex for P1 component. N170 was modeled by a pair
of dipoles localized ventrally and posteriorly. P2 was
modeled by parietooccipital dipoles situated more laterally and dorsally than P1 dipoles. The frontal activation observed was best modeled by a regional source
situated orbitofrontally. Very similar dipoles were
found for inverted and negative faces and did not differ
markedly between memory conditions as can be seen in
Figs. 9b and 9c. The only dipoles that differed between
memory condition were the pair modeling the late ERP
activity (from 400 to 600 ms). They were situated posteriorly but were more ventral and lateral in the 0-lag
condition and more medial and superior in the 1-lag
condition.
The effects of face type were seen at all peaks studied
for both amplitude and latency (except for P2 where it
was seen only for amplitude) and also for the mean
amplitudes measured frontally and parietally.
P1. P1 is widely reported as the earliest index of
endogenous processing of visual stimuli (Mangun,
1995). Most of the studies on face perception and recognition have not analyzed possible effects on P1. In
addition, among those who studied this component,
results are contradictory: some have found P1 latency
and amplitude not affected by inversion (Rossion et al.,
1999a) while others have found that P1 was longer and
larger for inverted faces (Linkenkaer-Hansen et al.,
1998), as found in the present study. Similarly, a
shorter P1 latency for faces compared to inverted faces
was also found in adults and children (Taylor et al.,
2001a), which was independent of attentional instructions and underlined the salience of face stimuli across
a wide age range even at this very early latency.
It has been suggested that processing of faces or
complex visual stimuli occurs as early as 120 ms (Debruille et al., 1998; Linkenkaer-Hansen et al., 1998)
which is in agreement with Halit et al. (2000) showing
an increase in P1 amplitude with stretching facial features. Those authors suggested that this effect on P1
amplitude could be due to psychophysical differences
in the stimuli or to top-down or attentional modulations. In our experiment, there was no mean luminance
difference between face types, although the local luminance on the face only (regardless of hair and background) was likely diminished for negative stimuli.
Thus, our reduced P1 amplitude with negative faces
may be due to changes in contrast and local luminance,
in agreement with the hypothesis that effects on P1
could reflect physical variations in the stimuli used.
However, the increase in P1 latency found only for
inverted faces could also reflect the disruption of configural changes on very early processes. If, like the
behavioral literature suggests, contrast reversal disrupts relational processing but not a holistic processing, whereas inversion disrupts both (Hole et al., 1999),
then the increased latency in P1 found for inverted but
not for negative faces could reflect the disruption of
this holistic processing by inversion. This type effect as
early as the P1 component could thus index the disruption of the perception of a face as a face when the face
is upside down, whereas the face would still be judged
as a face even when contrast is changed.
P1 does not appear to be face-specific (Rossion et al.,
DISCUSSION
In this paper we found that repetition of faces and
changes in their configuration affected not only behavioral measures of reaction times and hit rates, but also
affected both early and late event-related potentials.
Our decrease in recognition rates and increase in reaction times and false alarm rates for inverted and
negative faces are in agreement with the behavioral
literature and confirm the difficulty of recognizing
faces in such formats. Furthermore, the difficulty
seemed equal for inversion and contrast reversal as no
significant differences between those types were found
in any of the behavioral measures obtained. The difference in performances between face types was
366
ITIER AND TAYLOR
FIG. 9. Source analyses of ERP potentials using BESA for upright faces, showing the source waveforms (a) and the dipole locations for
ERPs for 0-lag (residual variance 2.49%) (b) and 1-lag faces (residual variance 2.17%) (c).
1999b; Taylor et al., 2001a), but our results are in
agreement with the suggestion that it nevertheless
reflects very early detection of configural changes such
as that occurring with face inversion, as well as psychophysical differences such as changes in luminance
and contrast. The pair of dipoles modeling P1 was
localized in the parietooccipital cortex, a finding in
accordance with the P1 generator found in the literature (Hillyard and Anllo-Vento, 1998; Rossion et al.,
1999b) and consistent with activation of extrastriate
visual cortex, where attentional modulation and physical features detection are possible.
N170. The N170 was affected both in amplitude
and in latency by stimulus type. It was longest for
inverted faces but largest for negative faces. However,
when peak-to-peak analyses were performed between
N170 and P1 components, differences between inverted and negative stimuli were no longer found.
Therefore, both inversion and contrast-reversal increase N170 latency and amplitude by the same
amount. ERP studies have shown an increase in N170
latency and amplitude for inverted faces (Edmonds et
al., 1997; Linkenkaer-Hansen et al., 1998; Rossion et
al., 1999a, 2000) compared to upright faces; Bentin et
al. (1996) found only an increase in latency with inversion. To our knowledge, it is the first time that contrast-reversal effects are observed on the N170 and
directly compared to inversion effects on this same
component.
It is likely that N170 reflects early structural encoding stages and is thus sensitive to changes in configuration, as latency increases in N170 have been found
for other types of configural disruption (George et al.,
1996), when one or several features were removed (Eimer, 1998; Bentin et al., 1996) or when only features
were used (Jemel et al., 1999; Séverac-Cauquil et al.,
2000; Taylor et al., 2001a,b). Thus, any changes in
configuration increase N170 latency compared to normal upright faces. It has been suggested that the reduced configural information with inversion slows
down processing and increases the difficulty of facial
encoding (McCarthy et al., 1999; Rossion et al., 1999a,
2000), resulting in a latency and amplitude increase.
As for scrambled faces (George et al., 1996), amplitude
increase with inversion could be the result of a longlasting temporal negativity associated with difficulty
superimposed on the N170 component or it could be
due to inverted faces recruiting face-sensitive brain
areas as well as more general areas involved in object
recognition (Haxby et al., 1999). These explanations
could apply for negative format faces as well, since they
also yielded N170 latency and amplitude increases.
The source analyses of N170 yielded a pair of dipoles
situated in the ventral pathway of the cortex, in the
CONFIGURAL CHANGES AND FACE RECOGNITION
region of the fusiform gyrus, a finding in accordance
with fMRI and PET studies showing specific activation
of the fusiform during face perception and recognition
tasks (Haxby et al., 1994; Kanwisher et al., 1997; McCarthy et al., 1997; Puce et al., 1995; Sergent et al.,
1992).
Our results support the hypothesis that N170 reflects face encoding processes (Bentin et al., 1996; Eimer, 1998; Eimer, 2000b) and, as we found increases in
N170 latency and amplitude for inverted and negative
faces compared to upright faces, that both these face
manipulations affect this encoding. Therefore, like inversion, contrast reversal disrupts the encoding stage
of face processing, a result in agreement with Kemp et
al. (1990) and in contradiction with Liu and Chaudhuri
(1997) who claimed that only memory stages were affected by contrast reversal.
VPP. Consistent with other studies, VPP was delayed for inverted (Jeffreys, 1989, 1993; Rossion et al.,
1999a) and negative faces (Jeffreys, 1989, 1993) compared to upright faces. However, our finding of VPP
amplitude being largest for negative faces and smallest
for upright faces contrasts with the report by Jeffreys
et al. (1993), who found the maximum amplitude for
upright faces, and their conclusion that VPP does not
depend on stimulus identification. In a recent study,
Taylor et al. (2001b) found that VPP was not affected
by changing the direction of gaze in faces; subtle
changes in faces thus seem not sufficient to affect VPP
whereas greater changes, such as contrast-polarity reversal or inversion of a face, can influence this component. Eyes-only stimuli (Séverac-Cauquil et al., 2000)
as well as scrambled faces (internal face features rearranged) (Yamamoto and Kashikura, 1999) also produce
longer latency VPPs than do upright faces. Our results
are thus in agreement with the suggestions that VPP
could be sensitive to configural changes and index
early configural processing, like N170, even though the
question of whether the two components originate from
the same source or not is still unresolved (Bötzel et al.,
1995; Schendan et al., 1998; Rossion et al., 1999b).
P2. P2 latency and amplitude were not affected by
inversion, as also reported by Rossion et al. (1999a),
nor was P2 latency influenced by contrast reversal.
However, P2 amplitude was larger for negative than
upright faces over occipital sites. The peak-to-peak
analysis revealed that this effect was not due to the
smaller P1 amplitude found earlier. What processes P2
indexes is not clear in the literature; Halit et al. (2000)
suggested that P2 reflected facial encoding of identity
although this explanation does not account for our
results. As the P2 source was modeled by a pair of
parietooccipital dipoles laterally and dorsally situated,
P2 could also reflect activation of extrastriate visual
cortex areas where physical features detection is pos-
367
sible and could thus be sensitive to physical changes of
pictures such as contrast reversal.
In summary, for the early ERP components, we can
conclude that inversion and contrast reversal affect
face encoding stages differently. The effects found on
P1 amplitude for contrast-reversed faces could reflect
the disruption of physical aspects of those pictures
(luminance and contrast per se). Increased P1 latency
for inverted faces only could be due to the disruption of
a holistic processing due to inversion (Hole et al., 1999);
it thus seems that it takes longer to perceive a face as
a face when it is upside down, whereas it does not when
it is in negative format. In contrast, increased latencies
and amplitudes of N170 could reflect the disruption of
relational features supposed to occur for both inversion
and contrast reversal (Hole et al., 1999). This latter
encoding stage would be similarly disrupted for inversion and contrast reversal. These differences between
inversion and contrast reversal at early stages of processing are reinforced by their additive effect found on
the VPP by Jeffreys (1993).
Frontal and parietal sites. Late ERP components
were also affected by face types, as shown in Figs. 5a
and 5c and Table 2. Mean amplitude was increased for
inverted compared to upright and negative faces, from
300 to 600 ms at frontal sites (Fig. 5a) and from 350 to
600 ms at parietal sites (Fig. 5c). Negative faces
yielded a greater positivity than upright faces only
frontally between 300 and 500 ms. The maximum difference in activity among the three face types was seen
between 350 and 400 ms at frontal sites (Fig. 5a). At
parietal sites, differences were no longer seen between
negative and upright faces (Fig. 5c). As in early components, the difficulty of the task could explain an
extra processing for unusual stimuli, compared to
usual upright faces that necessitate the least activation. However, as no differences in behavioral measures were found between inverted and negative stimuli, it suggests that these faces are of similar level of
difficulty and thus cannot account for the frontal effects. Configural differences are a more likely explanation. When faces are upright and in usual positive
format, the activity appears to be minimum whereas
when the face is inverted, increased positivity is maximum. Negative faces, however, are unusual due to the
contrast reversal, but the holistic configuration is preserved (Hole et al., 1999), and only relational information between facial features seems disrupted, leading
to a weaker increase in positivity than for inverted
faces. It seems that the more the face is disrupted, the
larger the increase in positivity. Our experiment does
not allow us to determine if these differences found at
frontal and parietal sites on the “new” trials are only
due to the disruption of encoding stages, as reflected by
effects of type on P1 and N170, or if they are due to
368
ITIER AND TAYLOR
additional disruption of processing at later stages, such
as entry into memory.
Thus, the only conclusion that can be drawn at this
stage is that inversion and contrast-reversal affect late
potentials of “new” trials differently, as seen in different magnitude of increased positivity at frontal and
parietal sites.
Effects of Memory Condition (Repetition Effects)
Repetition effects on early components. Repetition
effects were seen as early as the N170 and on all
subsequent components analyzed, a finding in agreement with Schweinberger et al. (1995) who reported
early priming effects on components found between
130 and 160 ms (p. 725). Repetition affected N170 by
decreasing its amplitude for 0-lag faces and its latency
for 0-lag and 1-lag faces compared to new faces. Eimer
(2000a) reported no repetition effect on N170 amplitude (latency was not measured) using repetition lags
greater than 0 (immediate repetition was not used);
however, Campanella et al. (2000) found an amplitude
reduction of N170 in a matching task when the face
was identical, presumably due to identity priming effect. It is possible that such a repetition effect is only
seen on amplitude for immediate repetition and not for
longer lags, an idea supported recently by the findings
of Guillaume and Tiberghien (2001). In the present
study, repetition affected latencies also, as both 0-lag
and 1-lag stimuli elicited shorter N170 than new stimuli. It is possible that these small but consistent effects
were not significant in other papers because of smaller
numbers of subjects and were significant here as we
tested 34 subjects. Further studies need to investigate
whether these effects disappear with longer lags. VPP
also showed a repetition effect, as its amplitude was
smaller for immediately repeated faces than for new
faces, whereas this was not seen in studies investigating familiarity or recognition effects on VPP (Rossion et
al., 1999b).
The lack of type ⫻ memory condition interaction
suggests that all face types were influenced by repetition similarly. These repetition effects, independent of
stimulus type, appear to be priming effects, which are
largely perceptually driven. Priming effects are thus
seen on encoding stages regardless of stimulus type,
over occipitotemporal areas. These effects, reflected as
decreases in latencies and amplitudes of N170 and
VPP components, are in agreement with the hypothesis that because of priming, more rapid and less activation of perceptual areas is required for the processing of recently presented information (Buckner et al.,
1995); with face stimuli, those perceptual areas would
be the fusiform gyrus and other ventral temporaloccipital areas.
Repetition effects on late components. The stimulus
repetition yielded greater positivity from 250 to 450 ms
frontally and from 400 to 600 ms parietally. These two
increases in positivity can clearly be seen in Figs. 6 and
7 which display the topographies of ERP differences
between repeated and new faces for all three face
types. They appear independent as the topography was
different as well as their occurrence in time. The frontal positivity was best modeled by a regional source
situated orbitofrontally (Fig. 9) while the parietal activity was modeled by a pair of dipoles situated occipitoparietally.
The parietal positivity could reflect a P3 component
as seen on Pz electrode on Fig. 2. However, the greater
positivity observed for 0-lag faces compared to new and
1-lag faces is unlikely due to the probability of appearance of repeated faces as 0-lag and 1-lag faces were
equiprobable. Furthermore, the parietal activity disappeared for the 1-lag condition for inverted and negative
faces and this cannot be accounted for by the probability of target appearance as this probability was the
same for upright, inverted, and negative faces. In addition, as we measured repetition effects for successful
recognition and given that a motor response was required for all three types of faces, the parietal activation cannot be attributed to either motor action or
anticipation. Indeed, as seen on Fig. 6, the parietal
positivity is observed for upright and negative faces
only, not for inverted faces, despite behavioral responses for all three face types. In contrast, this parietal activity likely reflects the old–new effect reported
at parietal and central sites in the literature (Bentin
and McCarthy, 1994; Paller et al., 2000; Sommer et al.,
1997) around 400 – 600 ms. This old–new effect is
thought to reflect stages of recognition that are not
domain-specific (Munte et al., 1997). The fact that it is
absent for inverted faces only (Fig. 6) is in accordance
with the proposal that the parietal old–new effect indexes recollection in a graded fashion (Wilding, 2000):
the better the recollection, the larger the old–new effect. It also agrees with the absence of any parietal
activity for the 1-lag-minus-new condition (Fig. 7) that
can be interpreted as reflecting a less efficient recollection, supported by the decreases in hit rates and increases in RTs for 1-lag compared to 0-lag condition.
Our results show maximum activation in frontal areas, in agreement with a frontal-positive Dm for explicit subsequent retrieval and intentional learning
found by Paller (1990) and increased frontal activation
with retrieval found by Ranganath and Paller (1999).
Wilding (2000) also found a frontal old–new effect. We
thus attribute our frontal repetition effect to the explicit recognition of faces. This increase in ERP amplitude at frontal sites is consistent with intracranial
recordings in frontal areas that revealed the involvement of prefrontal structures in memory-related modulation recorded on the scalp (Guillem et al., 1995). It is
also in agreement with fMRI and PET studies reporting increased activation in prefrontal cortex (PFC) ar-
CONFIGURAL CHANGES AND FACE RECOGNITION
eas during face-matching and face recognition tasks in
addition to the expected ventral occipitotemporal regions (Courtney et al., 1996, 1997; Haxby et al., 1996;
Ungerleider et al., 1998). The PFC is also involved in
recognition of words or objects other than faces (Cohen
et al., 1997; McCarthy et al., 1996; Schacter et al., 1995;
Shallice et al., 1994; Rugg et al., 1999) as well as in the
maintenance of an active representation of items in
working memory (Courtney et al., 1997) and is therefore general to working memory tasks and not specific
to face material (although specific PFC areas could be
activated by specific working memory tasks, see McCarthy et al. (1996) and Ungerleider et al. (1998)). The
frontal source we obtained for repeated items (Figs. 9b
and 9c) is consistent with the involvement of PFC in
working memory. It should be noted that more negative-going ERPs for repeated faces bilaterally over occipitotemporal areas occurred at the same time as the
frontal increase in positivity (around 300 ms on Figs. 6
and 7 and on electrodes P7 and P8 on Fig. 4). This is
consistent with the negative-going priming effects for
repeated items found by Hertz et al. (1994). It could
reflect increase in reactivation of the areas involved in
face recognition (fusiform gyrus in particular) while
frontal activation would reflect general working memory processes (Haxby et al., 1996; Courtney et al., 1997)
as well as explicit recollection.
When comparing face type topographies for the differential activity between 0-lag and new condition (Fig.
6), it is clear that the activity at frontal sites was
smaller for negative and especially inverted faces compared to that for upright faces. Like the parietal old–
new effect, this smaller frontal activity could reflect the
less efficient recollection observed for inverted and negative faces compared to upright faces. The fact that
frontal activity observed for 1-lag-minus-new condition
(Fig. 7) was generally smaller than that for 0-lag-minus-new condition (compare the amplitude scales) supports the idea of less efficient recollection processes for
condition 1-lag. However, as seen on Fig. 7, the increased positivity at frontal sites for the 1-lag-minusnew condition seemed larger for negative faces compared to upright and inverted faces. Furthermore, the
frontal activity lasted longer for inverted and negative
faces. This could reflect, in addition to recognition processes, the retrieval effort (Rugg and Wilding, 2000)
that would be more difficult for inverted and contrastreversed faces than for upright faces.
As mentioned, no type ⫻ condition interactions were
found for the priming effects observed on early components, showing that the priming was similar for all
three types of faces. However, on later ERPs, such an
interaction was found between 400 and 450 ms, due to
an amplitude increase from 0-lag to 1-lag for inverted
and negative faces but a decrease for upright faces.
This, again, could reflect greater retrieval effort (Rugg
and Wilding, 2000) for inverted and negative faces
369
compared to upright faces, in addition to working
memory processes.
Thus, although our experiment did not allow the
separation of priming from recollection, the repetition
effects observed on early components are likely perceptual priming effects as they were small, maximal for
0-lag faces, and similar across the three types of faces.
In contrast, the frontal and parietal activities occurring later likely reflected working memory processes
and explicit recognition. As the immediate repetition
was the easiest condition, the frontal activity differences between face types for the 0-lag-minus-new condition can be explained by a better recollection for
upright faces compared to the other two types. For the
1-lag-minus-new condition, however, it is likely that
the effort of retrieval due to the 1-lag condition as well
as to the face type difficulty was added to the recollection processes observed at frontal sites. A limit of our
experiment lies in the fact that effects of recognition
and that of retrieval effort are not dissociable; it is
therefore, difficult to know what the differences between face types are at these different memory stages.
CONCLUSIONS
We have found that contrast-reversal delayed and
enhanced N170 component similarly to inversion.
These effects suggest that, like inversion, contrast reversal disrupts face encoding processes. Delayed latency of the P1 was also found, but only for inverted
faces. Therefore, in conjunction with the behavioral
literature, we suggest that the perception of a face as a
face, i.e., the holistic processing, is the first step of face
processing, is reflected by P1, and is disrupted by inversion. The next step would be the relational processing of internal features enabling the encoding of face
identity, reflected by N170 and disrupted by both inversion and contrast reversal.
We also found repetition effects on early components
that were similar for all three face types that we interpreted as perceptual priming. In contrast, the frontal
and parietal old–new effects believed to reflect working
memory and recognition processes differed depending
on face type. The parietal old–new effect was maximal
for upright faces and nonexistent for inverted faces at
condition 0-lag. It was also nonexistent at condition
1-lag for all face types, reflecting the more difficult
recollection processes with this repetition. The larger
frontal activity likely reflected the involvement of PFC
in working memory and recognition processes. At condition 0-lag, it was largest for upright faces and smallest for inverted faces. Although at 1-lag it was generally smaller and delayed than at 0-lag, it was largest to
negative faces and smallest to upright faces. We interpreted this last result as reflecting retrieval effort and
difficulty due to inversion and contrast reversal, in
addition to recognition processes and working memory.
370
ITIER AND TAYLOR
Thus, inversion and contrast-reversal affect, albeit differently, both encoding and memory stages of face processing.
ACKNOWLEDGMENTS
This research was supported by a grant from the Fondation des
Aveugles et Handicapés Visuels de France (F.A.F). We thank anonymous reviewers for helpful comments and suggestions.
REFERENCES
Allison, T., Ginter, H., McCarthy, G., Nobre, A. C., Puce, A., Luby,
M., and Spencer, D. D. 1994. Face recognition in human extrastriate cortex. J. Neurophysiol. 71: 821– 825.
Bartlett, J. C., and Searcy, J. 1993. Inversion and configuration of
faces. Cogn. Psychol. 25: 281–316.
Bentin, S., Allison, T., Puce, A., Perez, E., and McCarthy, G. 1996.
Electrophysiological studies of face perception in humans. J. Cogn.
Neurosci. 8(6):551–565.
Bentin, S., and Deouell, L. Y. 2000. Structural encoding and identification in face processing: ERP evidence for separate mechanisms.
Cogn. Neuropsychol. 17: 35–54.
Bentin, S., and McCarthy, G. 1994. The effect of immediate stimulus
repetition on reaction time and event-related potentials in tasks of
different complexity. J. Exp. Psychol. 20: 130 –149.
Bötzel, K., Schulze, S., and Stodieck, S. R. 1995. Scalp topography
and analysis of intracranial sources of face-evoked potentials. Exp.
Brain Res. 104: 135–143.
Buckner, R. L., Petersen, S. E., Ojemann, J. G., Miezin, F. M.,
Squire, L. R., and Raichle, M. E. 1995. Functional anatomical
studies of explicit and implicit memory retrieval tasks. J. Neurosci. 15: 12–29.
Cabeza, R., and Kato, T. 2000. Features are also important: Contributions of featural and configural processing to face recognition.
Psychol. Sci. 11: 429 – 433.
Campanella, S., Hanoteau, C., Depy, D., Rossion, B., Bruyer, R.,
Crommelinck, M., and Guerit, J. M. 2000. Right N170 modulation
in a face discrimination task: An account for categorical perception
of familiar faces. Psychophysiology 37: 796 – 806.
Cohen, J. D., Perlstein, W. M., Braver, T. S., Nystrom, L. E., Noll,
D. C., Jonides, J., and Smith, E. E. 1997. Temporal dynamics of
brain activation during a working memory task. Nature 386: 604 –
608.
Courtney, S. M., Ungerleider, L. G., Keil, K., and Haxby, J. V. 1996.
Object and spatial visual working memory activate separate neural systems in human cortex. Cereb. Cortex 6: 39 – 49.
Courtney, S. M., Ungerleider, L. G., Keil, K., and Haxby, J. V. 1997.
Transient and sustained activity in a distributed neural system for
human working memory. Nature 386: 608 – 611.
Damasio, A. R., Damasio, H., and Van Hoesen, G. W. 1982. Prosopagnosia: Anatomic basis and behavioral mechanisms. Neurology
32: 331–341.
De Gelder, B., and Rouw, R. 2000. Configural face processes in
acquired and developmental prosopagnosia: Evidence for two separate face systems? NeuroReport 11: 3145–3150.
Debruille, J. B., Guillem, F., and Renault, B. 1998. ERPs and chronometry of face recognition: Following-up Seeck et al. and George
et al. NeuroReport 9: 3349 –3353.
Diamond, R., and Carey, S. 1986. Why faces are and are not special:
An effect of expertise. J. Exp. Psychol. Gen. 115: 107–117.
Edmonds, G., Pang, E. W., Smith, M. L., McCarthy, G., and Taylor,
M. J. 1997. Hemispheric asymmetries in facial processing: Sex
differences. Soc. Neurosci. Abstr. 23: 2112.
Eimer, M. 1998. Does the face-specific N170 component reflect the
activity of a specialized eye processor? NeuroReport 9: 2945–2948.
Eimer, M. 2000a. Event-related brain potentials distinguish processing stages involved in face perception and recognition. Clin. Neurophysiol. 111: 694 –705.
Eimer, M. 2000b. The face-specific N170 component reflects late
stages in the structural encoding of faces. NeuroReport 11: 2319 –
2324.
Farah, M. J., Tanaka, J. W., and Drain, H. M. 1995a. What causes
the face inversion effect? J. Exp. Psychol. Hum. Percept. Perform.
21: 628 – 634.
Farah, M. J., Wilson, K. D., Drain, H. M., and Tanaka, J. R. 1995b.
The inverted face inversion effect in prosopagnosia: Evidence for
mandatory, face-specific perceptual mechanisms. Vis. Res. 35:
2089 –2093.
Freire, A., Lee, K., and Symons, L. A. 2000. The face-inversion effect
as a deficit in the encoding of configural information: Direct evidence. Perception 29: 159 –170.
Galper, R. E. 1970. Recognition of faces in photographic negative.
Psychonom. Sci. 19: 207–208.
George, N., Evans, J., Fiori, N., Davidoff, J., and Renault, B. 1996.
Brain events related to normal and moderately scrambled faces.
Cogn. Brain Res. 4: 65–76.
George, N., Jemel, B., Fiori, N., and Renault, B. 1997. Face and
shape repetition effects in humans: A spatio-temporal ERP study.
NeuroReport 8: 1417–1423.
Guillaume, F., and Tiberghien, G. 2001. An event-related potential
study of contextual modifications in a face recognition task. NeuroReport 12: 1209 –1216.
Guillem, F., N’Kaoua, B., Rougier, A., and Claverie, B. 1995. Intracranial topography of event-related potentials (N400/P600) elicited
during a continuous recognition memory task. Psychophysiology
32: 382–392.
Halit, H., de Haan, M., and Johnson, M. H. 2000. Modulation of
event-related potentials by prototypical and atypical faces. NeuroReport 11: 1871–1875.
Haxby, J. V., Horwitz, B., Ungerleider, L. G., Maisog, J. M., Pietrini,
P., and Grady, C. L. 1994. The functional organization of human
extrastriate cortex: A PET-rCBF study of selective attention to
faces and locations. J. Neurosci. 14: 6336 – 6353.
Haxby, J. V., Ungerleider, L. G., Clark, V. P., Schouten, J. L., Hoffman, E. A., and Martin, A. 1999. The effect of face inversion on
activity in human neural systems for face and object perception.
Neuron 22: 189 –199.
Haxby, J. V., Ungerleider, L. G., Horwitz, B., Maisog, J. M., Rapoport, S. I., and Grady, C. L. 1996. Face encoding and recognition
in the human brain. Proc. Natl. Acad. Sci. USA 93: 922–927.
Hepworth, S. L., Rovet, J. F., and Taylor, M. J. 2001. Neurophysiological correlates of verbal and nonverbal short-term memory in
children: repetition of words and faces. Psychophysiology 38: 594 –
600.
Hertz, S., Porjesz, B., Begleiter, H., and Chorlian, D. 1994. Eventrelated potentials to faces: The effects of priming and recognition.
Electroencephalogr. Clin. Neurophysiol. 92: 342–351.
Hillyard, S. A., and Anllo-Vento, L. 1998. Event-related brain potentials in the study of visual selective attention. Proc. Natl. Acad.
Sci. USA 95: 781–787.
Hochberg, J., and Galper, R. E. 1967. Recognition of faces. I. An
exploratory study. Psychonom. Sci. 9: 619 – 620.
CONFIGURAL CHANGES AND FACE RECOGNITION
Hole, G. J., George, P., and Dunsmore, V. 1999. Evidence for holistic
processing of faces viewed as photographic negatives. Perception
28: 341–359.
Jeffreys, D. A. 1989. A face-responsive potential recorded from the
human scalp. Exp. Brain Res. 78: 193–202.
Jeffreys, D. A. 1993. The influence of stimulus orientation on the
vertex positive scalp potential evoked by faces. Exp. Brain Res. 96:
163–172.
Jemel, B., George, N., Chaby, L., Fiori, N., and Renault, B. 1999.
Differential processing of part-to-whole and part-to-part face priming: An ERP study. NeuroReport 10: 1069 –1075.
Johnston, A., Hill, H., and Carman, N. 1992. Recognising faces:
Effects of lighting direction, inversion, and brightness reversal.
Perception 21: 365–375.
Kanwisher, N., McDermott, J., and Chun, M. M. 1997. The fusiform
face area: A module in human extrastriate cortex specialized for
face perception. J. Neurosci. 17: 4302– 4311.
Kemp, R., McManus, C., and Pigott, T. 1990. Sensitivity to the
displacement of facial features in negative and inverted images.
Perception 19: 531–543.
Kemp, R., Pike, G., White, P., and Musselman, A. 1996. Perception
and recognition of normal and negative faces: The role of shape
from shading and pigmentation cues. Perception 25: 37–52.
Leder, H., and Bruce, V. 1998. Local and relational aspects of face
distinctiveness. Q. J. Exp. Psychol. A 51: 449 – 473.
Leder, H., and Bruce, V. 2000. When inverted faces are recognized:
The role of configural information in face recognition. Q. J. Exp.
Psychol. A 53: 513–536.
Leder, H., Candrian, G., Huber, O., and Bruce, V. 2001. Configural
features in the context of upright and inverted faces. Perception
30: 73– 83.
Lewis, M. B., and Johnston, R. A. 1997. The Thatcher illusion as a
test of configural disruption. Perception 26: 225–227.
Linkenkaer-Hansen, K., Palva, J. M., Sams, M., Hietanen, J. K.,
Aronen, H. J., and Ilmoniemi, R. J. 1998. Face-selective processing
in human extrastriate cortex around 120 ms after stimulus onset
revealed by magneto- and electroencephalography. Neurosci. Lett.
253: 147–150.
Liu, C. H., and Chaudhuri, A. 1997. Face recognition with multi-tone
and two-tone photographic negatives. Perception 26: 1289 –1296.
Mangun, G. R. 1995. Neural mechanisms of visual selective attention. Psychophysiology 32: 4 –18.
McCarthy, G., Puce, A., Belger, A., and Allison, T. 1999. Electrophysiological studies of human face perception. II. Response properties
of face-specific potentials generated in occipitotemporal cortex.
Cereb. Cortex 9: 431– 444.
McCarthy, G., Puce, A., Constable, R. T., Krystal, J. H., Gore, J. C.,
and Goldman-Rakic, P. 1996. Activation of human prefrontal cortex during spatial and nonspatial working memory tasks measured by functional MRI. Cereb. Cortex 6: 600 – 611.
McCarthy, G., Puce, A., Gore, J. C., and Allison, T. 1997. Facespecific processing in the human fusiform gyrus. J. Cogn. Neurosci.
9: 604 – 609.
Munte, T. F., Brack, M., Grootheer, O., Wieringa, B. M., Matzke, M.,
and Johannes, S. 1997. Event-related brain potentials to unfamiliar faces in explicit and implicit memory tasks. Neurosci. Res. 28:
223–233.
Paller, K. A. 1990. Recall and stem-completion priming have different electrophysiological correlates and are modified differentially
by directed forgetting. J. Exp. Psychol. Learn. Mem. Cogn. 16:
1021–1032.
Paller, K. A., Bozic, V. S., Ranganath, C., Grabowecky, M., and
Yamada, S. 1999. Brain waves following remembered faces index
conscious recollection. Cogn. Brain Res. 7: 519 –531.
371
Paller, K. A., Gonsalves, B., Grabowecky, M., Bozic, V. S., and
Yamada, S. 2000. Electrophysiological correlates of recollecting
faces of known and unknown individuals. NeuroImage 11: 98 –110.
Paller, K. A., Kutas, M., and Mayes, A. R. 1987. Neural correlates of
encoding in an incidental learning paradigm. Electroencephalogr.
Clin. Neurophysiol. 67: 360 –371.
Perrin, F., Pernier, J., Bertrand, O., and Echallier, J. F. 1989. Spherical splines for scalp potential and current density mapping. Electroencephalogr. Clin. Neurophysiol. 72: 184 –187.
Picton, T. W., Bentin, S., Berg, P., Donchin, E., Hillyard, S. A.,
Johnson, R., Jr., Miller, G. A., Ritter, W., Ruchkin, D. S., Rugg,
M. D., and Taylor, M. J. 2000. Guidelines for using human eventrelated potentials to study cognition: Recording standards and
publication criteria. Psychophysiology 37: 127–152.
Puce, A., Allison, T., Gore, J. C., and McCarthy, G. 1995. Facesensitive regions in human extrastriate cortex studied by functional MRI. J. Neurophysiol. 74: 1192–1199.
Ranganath, C., and Paller, K. A. 1999. Frontal brain potentials
during recognition are modulated by requirements to retrieve perceptual detail. Neuron 22: 605– 613.
Rhodes, G., Brake, S., and Atkinson, A. P. 1993. What’s lost in
inverted faces? Cognition 47: 25–57.
Rossion, B., Campanella, S., Gomez, C. M., Delinte, A., Debatisse, D.,
Liard, L., Dubois, S., Bruyer, R., Crommelinck, M., and Guerit,
J. M. 1999b. Task modulation of brain activity related to familiar
and unfamiliar face processing: An ERP study. Clin. Neurophysiol.
110: 449 – 462.
Rossion, B., Delvenne, J. F., Debatisse, D., Goffaux, V., Bruyer, R.,
Crommelinck, M., and Guerit, J. M. 1999a. Spatio-temporal localization of the face inversion effect: An event-related potentials
study. Biol. Psychol. 50: 173–189.
Rossion, B., Gauthier, I., Tarr, M. J., Despland, P., Bruyer, R.,
Linotte, S., and Crommelinck, M. 2000. The N170 occipito-temporal component is delayed and enhanced to inverted faces but not to
inverted objects: An electrophysiological account of face-specific
processes in the human brain. NeuroReport 11: 69 –74.
Rugg, M. D., Fletcher, P. C., Chua, P. M., and Dolan, R. J. 1999. The
role of the prefrontal cortex in recognition memory and memory for
source: An fMRI study. NeuroImage 10: 520 –529.
Rugg, M. D., Soardi, M., and Doyle, M. C. 1995. Modulation of
event-related potentials by the repetition of drawings of novel
objects. Cogn. Brain Res. 3: 17–24.
Rugg, M. D., and Wilding, E. L. 2000. Retrieval processing and
episodic memory. Trends Cogn. Sci. 4: 108 –115.
Schacter, D. L., Reiman, E., Uecker, A., Polster, M. R., Yun, L. S.,
and Cooper, L. A. 1995. Brain regions associated with retrieval
of structurally coherent visual information. Nature 376: 587–
590.
Schendan, H. E., Ganis, G., and Kutas, M. 1998. Neurophysiological
evidence for visual perceptual categorization of words and faces
within 150 ms. Psychophysiology 35: 240 –251.
Scherg, M. 1990. Fundamentals of dipole source potential analysis.
In Auditory Evoked Magnetic and Electric Potentials (F. Grandori,
M. Hoke, and G. L. Romani, Eds.), Adv. Audiol., pp. 40 – 69. Basel,
Karger.
Schweinburger, S., Pfütze, E.-M., and Sommer, W. 1995. Repetition
priming and associative priming of face recognition: Evidence from
event-related potentials. J. Exp. Psychol. Learn. Mem. Cogn. 21:
722–736.
Searcy, J. H., and Bartlett, J. C. 1996. Inversion and processing of
component and spatial-relational information in faces. J. Exp.
Psychol. Hum. Percept. Perform. 22: 904 –915.
Sergent, J., Ohta, S., and MacDonald, B. 1992. Functional neuroanatomy of face and object processing: A positron emission tomography study. Brain 115(Pt 1):15–36.
372
ITIER AND TAYLOR
Séverac-Cauquil, A. S., Edmonds, G. E., and Taylor, M. J. 2000. Is
the face-sensitive N170 the only ERP not affected by selective
attention? NeuroReport 11: 2167–2171.
Shallice, T., Fletcher, P., Frith, C. D., Grasby, P., Frackowiak,
R. S., and Dolan, R. J. 1994. Brain regions associated with
acquisition and retrieval of verbal episodic memory. Nature 368:
633– 635.
Sommer, W., Heinz, A., Leuthold, H., Matt, J., and Schweinberger,
S. R. 1995. Metamemory, distinctiveness, and event-related potentials in recognition memory for faces. Mem. Cognit. 23: 1–11.
Sommer, W., Komoss, E., and Schweinberger, S. R. 1997. Differential
localization of brain systems subserving memory for names and
faces in normal subjects with event-related potentials. Electroencephalogr. Clin. Neurophysiol. 102: 192–199.
Sommer, W., Schweinberger, S. R., and Matt, J. 1991. Human brain
potential correlates of face encoding into memory. Electroencephalogr. Clin. Neurophysiol. 79: 457– 463.
Subramaniam, S., and Biederman, I. 1997. Does the contrast reversal affect object identification? Invest. Ophtalmol. Vis. Sci. 38: 998.
Tanaka, J. W., and Farah, M. J. 1993. Parts and wholes in face
recognition. Q. J. Exp. Psychol. 2: 225–245.
Taylor, M. J., Edmonds, G. E., McCarthy, G., and Allison, T. 2001a.
Eyes first! Eye processing develops before face processing in children. NeuroReport 12: 1671–1676.
Taylor, M. J., Itier, R. J., Allison, T., and Edmonds, G. E. 2001b.
Direction of gaze effects on early face processing: eyes-only versus
full faces. Cogn. Brain Res. 10: 333–340.
Taylor, M. J., McCarthy, G., Saliba, E., and Degiovanni, E. 1999.
ERP evidence of developmental changes in processing of faces.
Clin. Neurophysiol. 110: 910 –915.
Ungerleider, L. G., Courtney, S. M., and Haxby, J. V. 1998. A neural
system for human visual working memory. Proc. Natl. Acad. Sci.
USA 95: 883– 890.
Wilding, E. L. 2000. In what way does the parietal ERP old/new
effect index recollection? Int. J. Psychophysiol. 35: 81– 87.
Yamamoto, S., and Kashikura, K. 1999. Speed of face recognition in
humans: An event-related potentials study. NeuroReport 10:
3531–3534.
Yin, R. K. 1969. Looking at upside-down faces. J. Exp. Psychol. 81:
141–145.
Young, A. W., Hellawell, D., and Hay, D. C. 1987. Configurational
information in face perception. Perception 16: 747–759.