Neural fate of seen and unseen faces in visuospatial
neglect: A combined event-related functional
MRI and event-related potential study
Patrik Vuilleumier*†‡§, Noam Sagiv*†¶, Eliot Hazeltine储, Russel A. Poldrack**††, Diane Swick*†, Robert D. Rafal*†‡‡,
and John D. E. Gabrieli**
*Department of Neurology, University of California, Davis, CA 95616; †Veterans Administration Medical Center, Martinez, CA 94553; ¶Department
of Psychology, University of California, Berkeley, CA 94720; 储National Aeronautics and Space Administration–Ames Research Center,
Moffett Field, CA 94035; **Department of Psychology, Stanford University, Stanford, CA 94305; ††Massachusetts General Hospital,
NMR Center, Charlestown, MA 02129; ‡‡School of Psychology, Bangor, Wales LLS7 2DG, United Kingdom; and
‡Institute of Cognitive Neuroscience, University College London, London WC1 3AR, United Kingdom

To compare neural activity produced by visual events that escape
or reach conscious awareness, we used event-related MRI and
evoked potentials in a patient who had neglect and extinction after
focal right parietal damage, but intact visual fields. This neurological disorder entails a loss of awareness for stimuli in the field
contralateral to a brain lesion when stimuli are simultaneously
presented on the ipsilateral side, even though early visual areas
may be intact, and single contralateral stimuli may still be perceived. Functional MRI and event-related potential study were
performed during a task where faces or shapes appeared in the
right, left, or both fields. Unilateral stimuli produced normal
responses in V1 and extrastriate areas. In bilateral events, left faces
that were not perceived still activated right V1 and inferior temporal cortex and evoked nonsignificantly reduced N1 potentials,
with preserved face-specific negative potentials at 170 ms. When
left faces were perceived, the same stimuli produced greater
activity in a distributed network of areas including right V1 and
cuneus, bilateral fusiform gyri, and left parietal cortex. Also,
effective connectivity between visual, parietal, and frontal areas
increased during perception of faces. These results suggest that
activity can occur in V1 and ventral temporal cortex without
awareness, whereas coupling with dorsal parietal and frontal areas
may be critical for such activity to afford conscious perception.

R

ight parietal damage may cause a loss of awareness for
contralateral (left) sensory inputs, such as hemispatial neglect and extinction (1–3). Visual extinction is the failure to
perceive a stimulus in the contralesional field when presented
together with an ipsilesional stimulus (bilateral simultaneous
stimulation, BSS), even though occipital visual areas are intact
and unilateral contralesional stimuli can be perceived when
presented alone. It reflects a deficit of spatial attention toward
the contralesional side, excluding left inputs from awareness in
the presence of competing stimuli (2, 3). Spatial attention
involves a complex neural network centered on the right parietal
lobe (4, 5), but how parietal and related areas interact with
sensory processing in distant cortices is largely unknown.
Here we combined event-related functional MRI (fMRI) and
event-related potentials (ERPs) to study the regional pattern
and temporal course of brain activity produced by seen and
unseen stimuli in a patient with chronic neglect and extinction
caused by parietal damage. In keeping with intact early visual
areas in such patients, behavioral studies suggest that some
residual processing may still occur for contralesional stimuli
without attention, or without awareness, including ‘‘preattentive’’ grouping (e.g., refs. 6 and 7) and semantic priming (e.g., ref.
8). It has been speculated (3, 9) that such effects might relate to
separate cortical visual streams, with temporal areas extracting
object features for identification, and parietal areas encoding
spatial locations and parameters for action (10). Because neglect

www.pnas.org兾cgi兾doi兾10.1073兾pnas.051436898

and extinction follow parietal damage, residual perceptual and
semantic processing still might occur in occipital and temporal
cortex without awareness, in the absence of normal integration
with concomitant processing in parietal regions.
Our study tested this hypothesis by using event-related imaging and electrophysiology measures, which are widely used to
study mechanisms of normal attention (11, 12). There have been
few imaging (e.g., ref. 13) or ERP (e.g., ref. 14) studies in neglect,
and most examined activity at rest or during passive unilateral
visual stimulation, rather than in relation to awareness or
extinction on bilateral stimulation. However, a recent ERP study
(15) found signals evoked by perceived, but not extinguished,
visual stimuli in a parietal patient. By contrast, functional
imaging in another patient (16) showed activation of striate
cortex by extinguished stimuli, although severe extinction on all
bilateral stimuli precluded any comparison with normal perception. In our patient we used both fMRI and ERPs during a
similar extinction task to determine the neural correlates of two
critical conditions: (i) when contralesional stimuli are extinguished, and (ii) when the same stimuli are seen. Stimulus
presentation was arranged so as to obtain a balanced number of
extinguished and seen contralesional events across all bilateral
trials. Like Rees et al. (16), we used face stimuli to exploit
previous knowledge that face processing activates fusiform areas
in temporal cortex (e.g., refs. 17 and 18), and elicits characteristic
potentials 170–200 ms after stimulus onset (e.g., refs. 19–21) in
addition to other visual components such as P1 and N1 (e.g., ref.
11). We reasoned that such responses might help trace the neural
fate of contralesional stimuli (seen or extinguished) at both early
and later processing stages in the visual system.
Methods
Patient. CW is a 67-year-old right-handed male who had right

middle cerebral artery infarct 2 years previously. He has participated in several behavioral studies (e.g., see ref. 22). Anatomical
MRI scans show focal damage in the inferior posterior parietal

This paper was submitted directly (Track II) to the PNAS office.
Abbreviations: BSS, bilateral simultaneous stimulation; fMRI, functional MRI; ERP, eventrelated potential; RVF, right visual field; LVF, left visual field; Lf, left face; Rf, right face; Ls,
left shape; Rs, right shape.
Parts of these results were reported at the Annual Meeting of the Swiss Society for
Behavioral Neurology in Bern, October 30, 1999, and at the Annual Meeting of the
Cognitive Neuroscience Society in San Francisco, April 8 –10, 2000.
§To whom reprint requests should be addressed at: Institute of Cognitive Neuroscience,

University College, Alexandra House, 17 Queen Square, London WC1 3AR, United Kingdom. E-mail: p.vuilleumier@ucl.ac.uk.
The publication costs of this article were defrayed in part by page charge payment. This
article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C.
§1734 solely to indicate this fact.

PNAS 兩 March 13, 2001 兩 vol. 98 兩 no. 6 兩 3495–3500

NEUROBIOLOGY

Edited by Michael I. Posner, Weill Medical College of Cornell University, New York, NY, and approved January 8, 2001 (received for review
September 12, 2000)

For both fMRI and ERPs, stimulus exposure time was titrated
to ensure adequate performance on unilateral trials together
with a balanced number of extinguished and perceived contralesional stimuli on BSS trials (half of 20 BSS as a criterion during
titration). After titration, stimulus duration was kept fixed at 250
ms for fMRI and 400 ms for ERPs. During experimental
sessions, extinction rate was close to 1兾2 in the ERP study and
slightly greater in the fMRI study (about 2兾3).
fMRI. MRI was performed on a 1.5-T GE Signa MR scanner with

Fig. 1. Structural MRI brain scan of patient CW (axial slices), showing focal
damage in inferior posterior parietal cortex (angular gyrus, thick arrows),
extending into the subcortical white matter of the right hemisphere (dashed
arrows).

cortex, extending into the anterior white matter of the right
hemisphere (Fig. 1). Striate and extrastriate occipital cortex is
intact. Neurological examination shows left-hand weakness with
poor position sense, astereognosis, and left extinction on double
tactile stimulation. Visual fields are full on both sides but left
extinction reliably obtained on bilateral simultaneous stimulation. CW is self-independent, but shows mild left spatial neglect
on standard tests (letter cancellation: 15 left misses兾60 targets;
line bisection: mean 6 cm rightward deviation兾180-cm-long
lines). No neglect is apparent on reading and mental imagery.
Visual extinction was assessed in an event-related design using
the same task during two fMRI sessions and two ERP sessions, all
performed within a 3-month period. Separate training sessions
were given before fMRI and ERP studies to ensure adequate
performance with the task while maintaining central fixation.
Behavioral Task. For both fMRI and ERPs, the stimuli comprised

10 different schematic faces and 10 different meaningless symmetrical shapes on a black background. They were presented on a
computer screen during training and ERPs and reflected from a
liquid crystal display projector onto a mirror inside the scanner
during fMRI. Trials began with a central fixation cross (2 sec),
followed by a brief stimulus in the right (RVF), left (LVF), or both
visual fields (BSS). The bilateral condition (BSS) with a left face
(Lf) and a right shape (Rs) provided the critical experimental trials
for extinction, in which the face could be either perceived or
extinguished by the patient on a given trial. BSS trials were
intermixed with unilateral trials to prevent expectation biases.
The fMRI study consisted of eight runs of pseudorandomized
trials (four runs in each session, 700 events in total). There were six
event types: unilateral trials, including a Lf or right face (Rf) and
a left shape (Ls) or Rs (n ⫽ 50 for each type in each session); blank
catch trials with only the fixation cross and no stimulus in either field
(Fx, n ⫽ 50 each session); and the critical bilateral trials with a Lf
and a Rs (BSSfs, n ⫽ 100 each session). The ERP study included
four blocks of pseudorandomized trials (total 880 events). Again,
there were four possible unilateral trials with either faces (Lf and
Rf, n ⫽ 30 and 30 in each block) or shapes (Ls and Rs, n ⫽ 10 and
40 in each block), as well as the same critical bilateral trials with a
Lf and a Rs (BSSfs, n ⫽ 80 each block). Blank catch trials were
changed to bilateral trials with a shape on both sides (BSSss, n ⫽
30 each block). All stimuli subtended ⬇2° and appeared ⬇6° away
from fixation. CW pressed one of two buttons (with his right hand)
to indicate whether he did or did not see a face. He was instructed
not to guess and to withhold from responding when unsure (3% of
all BSS trials). Response buttons were aligned radially along his
body midline. Average interstimulus interval was 4 sec for fMRI
and 2.5 sec for ERPs.
3496 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.051436898

a custom head coil. High-resolution T1-weighted anatomical images were collected. Functional data were acquired by using a
single-shot gradient-echo spiral pulse sequence (repetition time ⫽
2,000 ms, echo time ⫽ 40 ms), with 18 axial slices (6 mm thick, 1-mm
spacing). Each session consisted of four runs of 261 scans. Motion
was corrected by using AIR 3.0. Scans were realigned, normalized,
and smoothed (6-mm full width, half-maximum Gaussian kernel).
Data were analyzed by using the general linear model in SPM99
(23). The critical bilateral trials (BSSfs) were sorted according to
perceptual report (i.e., Lf perceived or extinguished) and these two
conditions separately modeled. Statistical maps resulted from linear
contrasts between different trial types, thresholded at P ⫽ 0.001
(uncorrected in visual areas, corrected for the whole brain otherwise), and overlaid on coregistered anatomical images averaged
across sessions. We also performed an analysis of effective connectivity (24) to detect voxels in the whole brain where activity
changed as a function of the signal observed in a particular area (i.e.,
V1 or fusiform gyrus, defined by previous contrasts) in the context
of a specific event type (i.e., perception vs. extinction of Lfs). This
was done by fitting the general linear model at every voxel with two
additional regressors: one being the activity at the maxima of
interest (V1 or fusiform), and the second created by multiplying
activity in this maxima with the difference between the regressors
that modeled trials with perception versus trials with extinction. The
multiplication in the second regressor represents the interaction
between activity in the area of interest and the difference in
awareness, and thus indicates significant changes in the coupling of
any distant region with respect to activity in the area of interest on
a trial-by-trial basis.
ERPs. Continuous electroencephalogram (EEG) was recorded

from 31 scalp locations using tin electrodes attached to an elastic
cap, with ground and reference placed on forehead and nose,
respectively (19). Additional electrodes measured horizontal and
vertical electro-oculogram. EEG was sampled at 250 Hz and
amplified 50,000-fold with a frequency bandpass of 0.1–40 Hz.
Electrode impedance was below 5k⍀. A digital bandpass filter of
0.5–20 Hz was applied before analysis. Codes synchronized to
stimulus onset were used to average epochs associated with each
condition. After removing trials with eye blinks, data from the
four blocks were pooled together, randomly reshuffled, and
sorted into eight smaller blocks with an equal number of trials,
so as to obtain repeated measures in the same subject and
exclude any variance due to practice or fatigue (15). Mean
amplitudes were calculated over 10-ms intervals at specific
latency ranges and electrodes, relative to a 50-ms prestimulus
baseline. Different conditions were compared by using ANOVA
and paired t tests on average amplitudes across blocks.
Results
fMRI. The patient correctly detected 98% of unilateral RVF
stimuli and 83% of unilateral LVF stimuli. He withheld from
responding on 7兾200 BSS and 100兾100 catch trials. He extinguished 68% of LVF faces on BSS trials [57兾200 vs. 85兾100 Lfs
seen on bilateral vs. unilateral displays; ␹2(1) ⫽ 85.4, P ⬍ 0.0001].
Hemifield and face-specific responses. Unilateral trials were
used to determine regions activated by LVF stimuli, regardless
of their type, and those activated by seen faces, regardless of
Vuilleumier et al.

Table 1. Regional brain activation in fMRI

Cortical areas
LVF-related stimulation (Lf ⫹ Ls ⬎ Rf ⫹ Rs)
Right striate
Right cuneus
Right inferior temporal gyrus
Right middle temporal gyrus
Left middle temporal gyrus
Face-related activation (Lf ⫹ Rf ⬎ fixation)
Right anterior fusiform gyrus
Right posterior fusiform gyrus
Right inferior temporal gyrus
Right anterior fusiform gyrus
Left anterior fusiform gyrus
Left inferior temporal gyrus
(Lf ⫹ Rf ⬎ Ls ⫹ Rs)
Right anterior fusiform gyrus
Extinguished stimuli (BSSfs ext ⬎ unilat Rs)
Right striate
Right inferior temporal gyrus
Left inferior temporal gyrus
Perceived stimuli (BSSfs perc ⬎ BSSfs ext)
Right striate
Right cuneus
Right cuneus
Right parahipocampal gyrus
Right anterior fusiform gyrus
Right posterior insula
Left pre-cuneus
Left fusiform gyrus
Left parietal (IPS)

x

y

z

T value

4
6
48
64
⫺68

⫺78
⫺52
⫺64
0
0

8
2
6
⫺2
⫺6

3.62
3.70
3.88
4.33
4.03

22
22
40
32
⫺38
⫺40

⫺48
⫺74
⫺70
⫺66
⫺48
⫺70

⫺22
⫺20
⫺16
⫺22
⫺24
⫺22

4.13
3.85
3.97
3.75
4.51
4.00

32

⫺66

⫺22

4.01

0
58
⫺48

⫺80
⫺44
⫺38

12
⫺22
⫺20

2.56
2.51
2.60

8
4
6
14
32
38
⫺20
⫺22
⫺28

⫺94
⫺58
⫺78
⫺56
⫺66
⫺20
⫺66
⫺68
⫺76

0
8
32
⫺6
⫺22
⫺2
20
⫺20
30

3.60
3.47
3.25
3.95
3.57
4.60
3.70
4.69
4.25

Note that coordinates refer to normalized space of the Talairach atlas, but anatomical areas correspond to the
patient’s own brain as defined from his MRI scan and do not necessarily match the same region coordinates in the
Talairach atlas.

field. Several occipital and temporal areas (Table 1), including
right primary visual cortex (V1; Fig. 2A), were activated by LVF
stimulation (Lf ⫹ Ls ⬎ Rf ⫹ Rs), indicating preserved inputs to
visual cortex in the damaged hemisphere. Unilateral RVF
stimuli activated similar left hemisphere areas. Right inferior
temporal gyrus and bilateral fusiform areas were activated by
seen faces, regardless of field (Lf ⫹ Rf ⬎ Fx; Table 1 and Fig.
2B). Right anterior fusiform activity (x y z ⫽ 32 ⫺ 66 ⫺ 22) was
specifically observed for faces as compared with shapes (Lf ⫹
Rf ⬎ Ls ⫹ Rs), like in normal subjects (18).
Seen and unseen faces. The two crucial comparisons concerned
the neural activity distinguishing extinction or perception of a Lf
during bilateral trials. To examine the effect of unseen faces, we
compared bilateral trials with an extinguished face in LVF to those
with a Rs alone (BSSfs extinguished ⬎ Rs), i.e., same awareness but
different stimuli in LVF. This comparison showed no significant
activation at the usual statistical threshold of P ⫽ 0.001 (T ⫽ 3.09).
However, lowering the threshold to P ⫽ 0.005 (T ⫽ 2.56) revealed
three foci of activity: in right striate cortex (V1) and bilateral
postero-inferior temporal gyri (Table 1). These temporal foci were
just lateral to fusiform areas activated by seen faces (Fig. 2C).
Awareness of LVF faces was examined by comparing physically
identical bilateral trials when the face was perceived and when it was
not (BSSfs seen ⬎ BSSfs extinguished), i.e., same stimuli but
different awareness. This revealed increased activity in several areas
of both hemispheres (Table 1 and Fig. 2D), including right V1,
bilateral cuneus, both fusiform gyri, and superior parietal cortex in
the intact left hemisphere (intraparietal sulcus). Fig. 2 E–G shows
the time course of activity in peak voxels of interest across different
Vuilleumier et al.

conditions. Right V1 showed similar activity to seen and unseen
faces in LVF, but no activity when no stimulus was present. By
contrast, whereas both the fusiform and inferior temporal gyri
responded to seen faces (Table 1), only the latter was active when
faces were extinguished.
Finally, we tested for changes in effective connectivity between
areas depending on awareness of LVF faces. A first analysis looked
for voxels showing trial-specific increases that were coupled with
the change of activity found in right V1 during perception as
compared with extinction of a Lf (BSSfs seen ⬎ BSSfs extinguished). Two clusters in left inferior frontal cortex (⫺48 30 ⫺4,
T ⫽ 4.01; ⫺46 24 ⫺22, T ⫽ 3.73; P ⬍ 0.0001) showed such an
increase, indicating a specific interaction between V1 and frontal
areas during trials with awareness (Fig. 3). A second analysis looked
for voxels showing increases coupled with left fusiform activity
during perception versus extinction of faces. Such changes were
found again in left frontal cortex (⫺46 32 ⫺10, T ⫽ 4.15; ⫺44 48
⫺8, T ⫽ 4.05), as well as in inferior and superior left parietal cortex
(⫺66 ⫺42 28, T ⫽ 4.18; ⫺34 ⫺72 46, T ⫽ 4.04), and anterior left
temporal cortex (⫺54 6 ⫺32, T ⫽ 4.43; P ⬍ 0.0001 in all cases). The
superior parietal activation was close to the peak obtained by
contrasting seen vs. extinguished stimuli (Table 1).
ERPs. The patient reported 97% and 96% of unilateral stimuli in

RVF and LVF, respectively. He extinguished 40% of LVF faces in
critical BSSfs trials [115兾120 vs. 192兾320 faces seen in unilateral vs.
bilateral displays; ␹2(1) ⫽ 53.1, P ⬍ 0.0001]. He made only one false
alarm in reporting a face of 120 displays with bilateral shapes.
Hemifield and face-specific responses. Unilateral trials were
used to assess responses to perceived stimuli, regardless of their
PNAS 兩 March 13, 2001 兩 vol. 98 兩 no. 6 兩 3497

NEUROBIOLOGY

Talairach coordinates

Fig. 2. (A–D) Activated areas in fMRI, superimposed on the mean brain scan across sessions (Left and Middle: axial slices; Right: coronal slices; z and y: Talairach
coordinates). Damaged right parietal areas are indicated by black arrows. (A) Retinotopic activation of right V1 by unilateral LVF stimuli. (B) Bilateral activation of
anterior and posterior fusiform gyri by seen faces, irrespective of field. (C) Activation of V1 (Left) and inferior temporal cortex (Middle and Right) by extinguished faces,
lateral to fusiform activation by seen faces in B. (D) Distributed activation produced by seen as compared to extinguished faces in BSS, including right V1, right cuneus,
right insula, and left parietal cortex (Middle and Left). (E–G) Time course of activity poststimulus onset in peak voxel (weighted by smoothing of its surround) across
different conditions. (E) Right V1 voxel activated by LVF stimuli (same as A), showed similar responses to unilateral Lss and Lfs in BSS, regardless of perception or
extinction, but no response to unilateral Rss. (F) Inferior temporal and (G) anterior fusiform voxels were both activated by seen faces irrespective of field (same as B and
C, respectively), with the former responding to seen and extinguished faces in BSS more than to seen Lss, and the latter responding only to seen faces.

type. Contralesional stimuli (Lf ⫹ Ls) elicited early components
at right occipital sites, with a positive waveform at ⬇80 ms
postonset (P1) and a negative waveform at ⬇130 ms (N1) (Fig.
4A). ANOVA of P1 and N1 amplitudes with the factors hemisphere (right vs. left), field (contralateral vs ipsilateral), and
stimuli (face vs. shape) revealed an interaction of field and
hemisphere [P1: F(1,62) ⫽ 6.33, P ⫽ 0.015; N1: F(1,62) ⫽ 7.95,
P ⫽ 0.006], but no other effect or interaction. Contralateral field
P1 and N1 showed no hemispheric difference [F(1,30) ⫽ 2.14,
P ⫽ 0.15]; right occipital responses to LVF stimuli were similar
to left occipital responses to RVF stimuli [paired comparison
across blocks, P1: t(7) ⫽ 1.21, P ⫽ 0.27; N1: t(7) ⫽ 1.60, P ⫽ 0.15].
This result is consistent with intact LVF input into right extrastriate areas (11). By contrast, ipsilateral field P1 and N1 showed
a significant effect of hemisphere [P1: F(1,30) ⫽ 4.83, P ⫽ 0.04;
N1: F(1,30) ⫽ 5.16, P ⫽ 0.03], with decreased responses on the
damaged side. This finding might reflect abnormal callosal
transfer, as previously found in neglect (15). Responses specific
to faces in the LVF also were examined, by comparing all
bilateral trials with a Lf (BSSfs) vs. Ls (BSSss), regardless of
perception. Although early components did not differ, LVF
faces produced a negative wave ⬇170–180 ms postonset in right
posterior temporal regions (Fig. 4B), consistent with the latency
and location of face-selective signals in normals (N170, e.g., ref.
19). This negativity was replaced by a positivity with shapes in
3498 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.051436898

Fig. 3. Changes in effective connectivity on bilateral trials when LVF faces
were seen as compared with extinguished. Increased activity in striate cortex
covaried with increases in inferior left frontal gyrus (blue-to-green foci),
whereas increased fusiform activity covaried with increases in similar frontal
regions, as well as in superior and inferior parietal cortex, and anterior
temporal cortex (red-to-yellow foci).

Vuilleumier et al.

Fig. 4. ERPs at right occipital (O2), left occipital (O1), right posterior temporal (T6), and midline electrodes (CZ) for different conditions during 300-ms
poststimulus onset. (A) Unilateral stimuli (faces and shapes) in RVF and LVF
evoked an early contralateral occipital positivity at ⬇80 ms (P1) and negativity
at ⬇120 ms (N1). (B) Lfs in bilateral trials (BSSfs) as opposed to Lss (BSSss)
evoked a specific negativity at right posterior temporal site (⬇170 ms postonset) and a positivity at midline (⬇190 ms postonset), but similar occipital N1
(regardless of perception, as in averaged ERPs shown here). Eye movements
(HEOG) were similar and cannot explain this ERP difference. (C) Difference
waveform obtained by subtracting ERPs to Lfs from ERPs to Lss in BSS, clearly
showing a face-specific negativity in right temporal region (N170). (D) Perception versus extinction. Compared with a Rs alone (Rs), bilateral stimuli with
a LVF face (BSSfs) evoked a right occipital N1 and a temporal N170, both when
the face was seen and when it was extinguished. Eye movements were similar
during Rs and extinction trials and cannot explain this ERP difference. A later
positivity at midline (⬇190 ms postonset) occurred only when faces were seen
(P190). Oculographic differences after ⬇200 ms might partly contribute to this
signal.

LVF, with a reliable difference at electrode T6 [F(1,14) ⫽ 8.58,
P ⫽ 0.01], clearly shown by the face-specific waveform obtained
by subtraction (Fig. 4C). In addition, a midline positivity was
evoked by faces but not shapes ⬇190–200 ms postonset (P190).
Seen and unseen faces. Again, the crucial comparisons concerned signals associated with extinction or perception of faces.
Compared with RVF shapes alone (Rs), bilateral trials with a
LVF face (BSSfs) evoked a larger right occipital N1 when the
face was seen, nonsignificantly reduced when the face was
extinguished (Fig. 4D). ANOVA showed marginal effect of
condition (Rs, BSSfs extinguished, BSSfs seen) on right occipital
P1 [F(2,21) ⫽ 3.2, P ⫽ 0.06] but not N1 amplitude [F(2,21) ⬍ 1].
However, planned comparisons indicated greater N1 with faces
in LVF (seen or extinguished) as compared with Rs [t(7) ⫽ 2.30
and 2.35, respectively, P ⫽ 0.05], but no difference between seen
and extinguished faces [t(7) ⫽ 1.48, P ⫽ 0.18]. Compared with
Rs alone, a right occipital P1 occurred with seen faces [t(7) ⫽
2.66, P ⫽ 0.03] but not with extinction [t(7) ⫽ 1.68, P ⫽ 0.14],
although the difference between these two conditions was not
significant [t(7) ⫽ 0.43].
In critical BSSfs trials, there was a similar negativity at
170–180 ms in the right posterior temporal region when the face
was seen and when it was extinguished (Fig. 4D), replaced by a
positive wave with a RVF shape alone. T6 electrode amplitudes
around 170 ms showed a main effect of condition [F(2,21) ⫽
6.82, P ⫽ 0.005], with a reliable difference between Rs alone and
Vuilleumier et al.

Discussion
This study provides combined spatial and temporal analysis of
neural activity evoked by seen and unseen stimuli in neglect,
using both event-related imaging and electrophysiological measures. Our findings have implications not only for theories of
extinction in parietal injury and mechanisms of attention, but
also for the processing of faces, and more generally, for the
neural correlates of visual awareness.
Both fMRI and ERPs showed that contralesional inputs can
still activate striate and extrastriate areas in the damaged hemisphere after parietal lesion, even without awareness. By comparing fMRI signals for shapes presented alone in the normal
RVF and the same shapes presented with an extinguished face
in the LVF (i.e., different stimuli but same percept), we found
that unseen faces activated right visual cortex (V1) and inferior
temporal areas, just lateral to fusiform regions with face-specific
responses. This activation was weak compared with seen stimuli,
but remarkably specific. Moreover, these regions exhibited a
similar time course of activity when the face was extinguished or
seen (Fig. 2). Similarly, ERPs showed that extinguished faces
elicited early occipital responses, with no significant reduction of
N1 relative to seen faces (but no reliable P1). N1 latencies were
shorter than those reported in young normals, on both the intact
and lesioned sides, possibly because of our patient’s age or
neurological status. Unseen faces also evoked a later negativity
in right temporal sites, not evoked by shapes, with the same
latency as the N170 known to be selective for faces (19, 25).
On the other hand, fMRI showed that awareness of physically
identical stimuli in the LVF (as compared with extinction) was
associated with activation of a distributed network including
primary cortex (V1), cuneus, bilateral fusiform gyri, and the
intact left parietal cortex. ERPs also were modulated by awareness, with nonsignificant increases of early occipital P1, and later
anterior responses (⬇200 ms at midline).
In agreement with our results, another event-related fMRI study
of visual extinction (16) found preserved activation of striate cortex
by unseen stimuli. Those authors also reported category-specific
response to extinguished faces (as compared with pictures of
houses) but only when examining a small predefined region of
interest in the right fusiform gyrus and at low statistical threshold.
Their patient missed contralesional stimuli on all bilateral trials,
precluding a comparison with perceived stimuli. By contrast, we
were able to demonstrate distinct patterns of activity associated
with seen and unseen faces and found no evidence of unconscious
activation in face-specific fusiform areas. Other positron-emission
tomography or fMRI studies of tactile extinction yielded conflicting
results, with either preserved§§ or decreased (27) activity of primary
sensory cortex. However, their blocked design precluded assess§§Beversdorf, D. Q., Anderson, J. M., Auerbach, E. J., Briggs, R. W., Hughes, J. D., Crosson,

B. & Heilman, K. M. (1999) Neurology 52, Suppl. 2., A232 (abstr.).

PNAS 兩 March 13, 2001 兩 vol. 98 兩 no. 6 兩 3499

NEUROBIOLOGY

seen or extinguished faces in BSSfs [t(7) ⫽ 2.86 and 3.14,
respectively, P ⫽ 0.02], but no difference between seen and
extinguished trials [t(7) ⫽ 1.23, P ⫽ 0.26]. Eye movements were
similar on extinction and unilateral right trials (Fig. 4D), ruling
out ERP differences due to saccades.
Compared with extinguished Lfs, BSS trials with seen faces
evoked a later positive wave at 190–200 ms over right central and
midline regions (Fig. 4D). The effect of condition (Rs, BSSfs
extinguished, BSSfs perceived) was highly significant [F(2,21) ⫽
7.23, P ⫽ 0.004]. Paired comparisons with Rs indicated that P190
occurred with seen [t(7) ⫽ 3.63, P ⫽ 0.008] but not extinguished
faces [t(7) ⫽ 0.76, P ⫽ 0.47], whereas perception and extinction
differed from each other [t(7) ⫽ 2.92, P ⫽ 0.02]. However, eye
position tended to differ around 200 ms and may have contributed to the latter differences in ERP.

ment of single events with or without awareness. A single case study
using ERPs (15) failed to demonstrate early visual responses
(P1-N1) to extinguished stimuli. However, stimuli were small
light-emitting diodes in far peripheral field, presumably evoking too
weak activity in the visual system to allow reliable ERP signals
without attention, and recordings were made at parietal, rather
than occipital sites. In our patient, unseen faces elicited an early
occipital N1 response and also a later face negativity in the temporal
region.
Neural activity evoked by stimuli without awareness has been
found in normal subjects (28) or in hemianopic patients with
blindsight (29, 30). Blindsight has raised questions as to the role
of area V1 in visual awareness (31, 32) because its destruction
suppresses vision despite residual unconscious performance. We
found that V1 was still activated by extinguished stimuli, without
affording awareness (see also ref. 16). Yet, V1 was further
activated when contralesional faces were seen, as compared with
activity elicited by unseen stimuli. This finding may suggest some
critical interaction with modulatory or reentrant signals from
other brain areas (33). In our patient, fMRI analysis of effective
connectivity revealed that increases in V1 and fusiform activity
during awareness of faces were coupled with increases in a
network of frontal and parietal areas that are implicated in
spatial attention in normals (4, 12, 34). These results also
converge with recent evidence that attentional modulation can
operate both at early and late stages of vision (11, 35, 36).
Preserved cortical activation by unseen faces in our study and
others (16) illuminates the neural basis of residual processing in
neglect and extinction, implicating object segmentation (6, 7, 37)
or semantic analysis (8) in occipital and temporal cortex. However, this result does not imply that activation in the ventral
stream is fully intact (38). Unseen faces evoked a distinctive ERP
negativity and fMRI activity in the inferior temporal gyrus, but
only seen faces activated medial fusiform areas (Fig. 2). These
areas may subserve different aspects of face processing (39, 40).
Inferior temporal cortex may encode face structure and be
automatically engaged (41), whereas fusiform cortex may extract
specific face traits (18) and is modulated by attention (42) or
explicit detection of faces (43, 44).

This anatomical dissociation has implications for the generators
of face-specific ERPs that occurred for extinguished faces. In
normals, faces evoke a temporal negativity and a midline positivity
between 170 and 190 ms (19, 20). Our data suggest that these
components may reflect distinct sources (45, 46). N170 potentials
might index an early stage of face processing in temporal cortex (19,
46), whereas positive potentials might involve a distinct stage in
fusiform and anterior regions (20, 21, 45). Consistent with this
dissociation, N170 is not modulated by attention in normals (47).
It has been posited that visual awareness may depend on the
ventral stream (10) and兾or striate cortex (31), whereas dorsal
parietal areas operate without awareness. However, our results and
others (16, 30) show that striate and ventral activity is not sufficient
to support awareness without some interaction with parietal, and
perhaps frontal, areas. We found parietal activation in the intact
hemisphere only when contralesional faces were seen (as opposed
to extinguished), and a coupling of superior and inferior parietal
regions with the fusiform gyrus specific to such events. We cannot
entirely exclude a role of differential eye movements, but these
parietal regions and cuneus have been implicated in distinct aspect
of attention across a variety of tasks irrespective of eye movements
(34, 48). Moreover, detection of a face in degraded stimuli is
accompanied by long-distance coupling between parietal and occipitotemporal regions (43, 49). We suggest that visual awareness
may relate to the integration of processing along both dorsal and
ventral streams, binding ‘‘what’’ and ‘‘where’’ information about
stimuli to afford selective attention and directed action (26).
Damage to inferior parietal regions serving to link the dorsal and
ventral streams might result in neglect and extinction, whereas
residual processing might still occur in occipital and temporal areas
without awareness.

1. Heilman, K. M., Watson, R. T. & Valenstein, E. (1993) in Neglect and Related Disorders, eds.
Heilman, K. M. & Valenstein, E. (Oxford Univ. Press, New York), pp. 279–336.
2. Rafal, R. D. (1994) Curr. Opin. Neurobiol. 4, 231–236.
3. Driver, J. & Mattingley, J. B. (1998) Nat. Neurosci. 1, 17–22.
4. Mesulam, M. M. (1981) Ann. Neurol. 4, 309–325.
5. Posner, M. I. & Petersen, S. (1990) Annu. Rev. Neurosci. 13, 25–42.
6. Ward, R., Goodrich, S. & Driver, J. (1994) Visual Cognit. 1, 101–129.
7. Vuilleumier, P. & Landis, T. (1998) NeuroReport 9, 2481–2484.
8. Berti, A. & Rizzolatti, G. (1992) J. Cognit. Neurosci. 4, 345–351.
9. Driver, J. & Vuilleumier, P. (2001) Cognition, 79, 39–88.
10. Milner, A. D. & Goodale, M. A. (1995) The Visual Brain in Action (Oxford Univ. Press,
Oxford).
11. Hillyard, S. A., Teder-Sälerärvi, W. A. & Münte, T. F. (1998) Curr. Opin. Neurobiol. 8, 202–210.
12. Corbetta, M. & Shulman, G. L. (1998) Philos. Trans. R. Soc. London B 353, 1353–1362.
13. Perani, D., Vallar, G., Paulesu, E., Alberoni, M. & Fazio, F. (1993) Neuropsychologia 31,
2115–2125.
14. Vallar, G., Sandroni, P., Rusconi, M. L. & Barbieri, S. (1991) Neurology 41, 1918–1922.
15. Marzi, C., Girelli, M., Miniussi, C., Smania, N. & Maravita, A. (2000) J. Cognit. Neurosci.
12, 869–877.
16. Rees, G., Wojciulik, E., Clarke, K., Husain, M., Frith, C. D. & Driver, J. (2000) Brain 123,
1624–1633.
17. Sergent, J., Ohta, S. & Macdonald, B. (1992) Brain 155, 15–36.
18. Kanwisher, N., McDermott, J. & Chun, M. M. (1996) J. Neurosci. 17, 4302–4311.
19. Bentin, S., Allison, T., Puce, A., Perez, E. & McCarthy, G. (1996) J. Cognit. Neurosci. 8, 551–565.
20. Jeffreys, D. A. (1996) Visual Cognit. 3, 1–38.
21. Allison, T., Puce, A., Spencer, D. & McCarthy, G. (1999) Cereb. Cortex 9, 415–430.
22. Vuilleumier, P. & Rafal, R. (1999) Nat. Neurosci. 2, 783–784.
23. Friston, K. J., Holmes, A. P., Worsley, K. J., Poline, J. B., Frith, C. D. & Frackowiak, R. S.
(1995) Hum. Brain Mapp. 2, 189–210.
24. Friston, K. (1994) Hum. Brain Mapp. 2, 56–78.
25. Eimer, M. (2000) Cognit. Neuropsychol. 17, 103–116.
26. Robertson, L., Treisman, A., Friedman-Hill, S. & Grabowecky, M. (1997) J. Cognit. Neurosci.
9, 295–317.
27. Remy, P., Zilbovicius, M., Degos, J. D., Bachoud-Levy, A. C., Rancurel, G., Cesaro, P. &
Samson, Y. (1999) Neurology 52, 571–577.

28. Sobel, N., Prabhakaran, V., Hartley, C. A., Desmond, J. E., Glover, G. H., Sullivan, E. V.
& Gabrieli, J. D. E. (1999) Brain 122, 209–217.
29. Sahraie, A., Weiskrantz, L., Barbur, J. L., Simmons, A., Williams, S. C. R. & Brammer, M. J.
(1997) Proc. Natl. Acad. Sci. USA 94, 9406–9411.
30. Baseler, H. A., Morland, A. B. & Wandell, B. A. (1999) J. Neurosci. 19, 2619–1627.
31. Weiskrantz, L. (1986) Blindsight: A Case Study and Implications (Oxford Univ. Press,
Oxford).
32. Crick, F. & Koch, C. (1995) Nature (London) 375, 121–123.
33. Corthout, E., Uttl, B., Ziemann, U., Cowey, A. & Hallett, M. (1999) Neuropsychologia 37,
137–145.
34. Corbetta, M., Kincade, J. M., Ollinger, J. M., McAvoy, M. P. & Shulman, G. L. (2000) Nat.
Neurosci. 3, 292–297.
35. Ghandi, S. P., Heeger, D. J. & Boyton, G. M. (1999) Proc. Natl. Acad. Sci. USA 96,
3314–3319.
36. Somers, D. C., Dale, A. M., Seiffert, A. E. & Tootell, R. B. H. (1999) Proc. Natl. Acad. Sci.
USA 96, 1663–1668.
37. Driver, J., Baylis, G. C. & Rafal, R. D. (1993) Nature (London) 360, 73–75.
38. Farah, M. J., Monheit, M. A., Brunn, J. L. & Wallace, M. A. (1991) Neuropsychologia 29,
949–958.
39. George, N., Dolan, R. J., Fink, G. R., Baylis, G. C., Russell, C. & Driver, J. (1999) Nat.
Neurosci. 2, 574–580.
40. Haxby, J. V., Hoffman, E. A. & Gobbini, M. I. (2000) Trends Cognit. Neurosci. 4, 223–232.
41. Grill-Spector, K., Kushnir, T., Edeleman, S., Avidan, G., Itchak, Y. & Malach, R. (1999)
Neuron 24, 187–203.
42. Wojciulik, E., Kanwisher, N. & Driver, J. (1998) J. Neurophysiol. 79, 1574–1578.
43. Dolan, R. J., Fink, G. R., Rolls, E., Booth, M., Holmes, A., Frackowiak, R. S. J. & Friston,
K. J. (1997) Nature (London) 389, 596–599.
44. Tong, F., Nakayama, K., Vaughan, J. T. & Kanwisher, N. (1998) Neuron 21, 753–759.
45. George, N., Evans, J., Fiori, N., Davidoff, J. & Renault, B. (1996) Cognit. Brain. Res. 4, 65–76.
46. Bentin, S. & Deouell, L. Y. (2000) Cognit. Neuropsychol. 17, 35–54.
47. Séverac Cauquil, A., Edmonds, G. E. & Taylor, M. J. (2000) NeuroReport 11, 2167–2171.
48. Wojciulik, E. & Kanwisher, N. (1999) Neuron 23, 747–764.
49. Rodriguez, E., George, N., Lachaux, J.-P., Martinerie, J., Renault, B. & Varela, F. J. (1999)
Nature (London) 397, 430–433.

3500 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.051436898

Many thanks to J. Armony for invaluable help in statistical parameter
mapping analysis, as well as to T. Landis, P. Halligan, J. Driver, R. Dolan,
N. Kanwisher, and S. Schwartz for helpful discussions. This work was
supported by the Swiss National Science Foundation (P.V.), McDonnellPew Program in Cognitive Neuroscience (R.A.P. and D.S.), the National
Institute on Deafness and Other Communication Disorders (D.S.), and
the U.S. Public Health Service (R.D.R.).

Vuilleumier et al.