Content uploaded by Marie Luby
Author content
All content in this area was uploaded by Marie Luby
Content may be subject to copyright.
.I~~JKNAL~F
Vol. 7 1. No.
NEI
JROPHYSIOLOGY
2. Febr-uary lW4. l’l.lrlrcYl III ( ‘.S.. I.
RAPID PUBLICATION
Face Recognition in Human Extrastriate Cortex
TRUETT ALLISON, HEIDI GINTER, GREGORY MCCARTHY, ANNA C. NOBRE, AINA PUCE,
MARIE LUBY, AND DENNIS D. SPENCER
Neuropsychologv Laboratory, Veterans Administration Medical Center, West Haven 06516; and Department
of
Newologv and iection ofNeurosurgerv, Ya/e Universit v School o/Medicine, New Haven, Connecticut 06510
e
. w e .
SUMMARY AND CONCLUSIONS
I. Twenty-four patients with electrodes chronically implanted
on the surface of extrastriate visual cortex viewed faces, equilu-
minant scrambled faces, cars, scrambled cars, and butterflies.
2. A surface-negative potential, N200, was evoked by faces but
not by the other categories of stimuli. N200 was recorded only
from small regions of the left and right fusiform and inferior tem-
poral gyri. Electrical stimulation of the same region frequently
produced a temporary inability to name familiar faces.
3. The results suggest that discrete regions of inferior extrastri-
ate visual cortex, varying in location between individuals, are spe-
cialized for the recognition of faces. These “face modules” appear
to be intercalated among other functionally specific small regions.
INTRODUCTION
Recognition of faces is a complex but normally effortless
task. However, patients with lesions in the occipitotem-
poral region of cerebral cortex may lose the ability to recog-
nize familiar faces-even those of family members, famous
faces, and their own mirror-image-suggesting that a re-
gion of visual cortex is specialized for processing facial in-
formation (Bodamer 1947; Damasio et al. 1982; Hecaen
and Angelergues 1962; Whitely and Warrington 1977).
Here we report the first study of face recognition with elec-
trophysiological recordings made directly from the surface
of occipitotemporal cortex. We find that discrete regions of
inferior extrastriate visual cortex are activated by faces but
not by other visual stimuli, suggesting a substantial degree
of anatomic and functional specificity in the processing of
this important class of visual stimuli. A preliminary report
of these results has appeared (Allison et al. 1993b).
METHODS
Recordings were obtained from 24 patients ( 1 1 males, 13 fe-
males, 15-49 years of age) with intractable epilepsy who were
being evaluated for possible surgery (Spencer et al. 1982). With
two exceptions they scored within normal limits on the Benton
face recognition test (Benton et al. 1983).
Under general anesthesia, 8- 12 contact strips of electrodes were
placed subdurally on the cortical surface. The exposed surface of
each stainless steel electrode was 2.2 mm diam; interelectrode
spacing was 1 .O cm. Electrodes were localized by magnetic reso-
nance images using a combination of axial and coronal images
and by images reconstructed along the axis of electrode strips.
Patients were studied 2- 10 days after implantation.
Several categories of black-and-white images were displayed on
a video monitor. I) Photographs of faces were digitally scanned
from a college yearbook. Females and males (cleanshaven) were
equally represented, and none had extraneous features such as
eyeglasses or jewellery. 2) Using graphics software, each photo-
graph was rearranged such that all elements of the original image
were retained (thus the scrambled image was equiluminant with
the original image), but their location and orientation was modi-
fied until the face was unrecognizable. 3) Photographs of front
views of cars were scanned from automotive magazines. 4) Images
of cars were scrambled in the same manner as for faces. 5) Butter-
flies were scanned from a field guide. The average luminance was
not significantly different for faces and scrambled faces (94 t 26
cd/m2) compared with cars and scrambled cars ( 107 ~fr 30 cd/
m2). Each set of 1 10 stimuli consisted of 25 stimuli of each cate-
gory (faces, scrambled faces, cars, and scrambled cars) and 10
butterflies, to which the patient pressed a key to ensure attention.
Three such sets were presented. The images subtended 7.2
X
7.2’
of visual angle. Stimulus duration was 250 ms, and the interstimu-
lus interval varied randomly between 1.8 and 2.2 s. In separate
recordings, patients viewed content words (e.g., apple) or non-
words (e.g., aplep). Local field potentials generated by these stim-
uli were recorded simultaneously from 32 or 64 locations using a
gain of 10,000 and filter settings of 0. l- 100 Hz ( -3 dB points) and
were digitized at a sampling rate of 250 Hz. Recordings were refer-
ential to a mastoid.
Electrical stimulation of adjacent electrodes was carried out in
14 cases using 5-s trains of constant-current pulses (50 Hz, 0.1 ms
duration). Stimulus intensity was increased in steps between 2
and 10 mA: stimulation was discontinued if afterdischarges were
noted in the electroencephalogram. Patients were not told what
perceptual changes they might experience during stimulation.
However, once a disturbance was noted the stimulus was repeated
and the patient was questioned further about the characteristics of
the change. Most relevant to this study were tests, immediately
before and during stimulation, of object recognition (the patient
viewed drawings of common objects such as a hat or bird), retino-
topic disturbances ( localized phosphenes, usually flickering),
changes in color vision (the patient viewed Ishihara plates), and
tests of the ability to name famous (politicians and celebrities) or
family faces that the patient had quickly and correctly named on
prior testing. The protocols used in this study were approved by
the Human Investigation Committees ofthe West Haven VA Med-
ical Center and Yale University School of Medicine. Informed
consent was obtained.
RESULTS
In patient RCN a strip of electrodes spanned right infe-
rior cortex from the lingual gyrus medially to the inferior
temporal gyrus laterally (Fig. 1,
A
and B). From locations
3 and 4 on the fusiform gyrus a large-amplitude ( 150-200
pV) negativity with a peak latency of -200 ms (N200) was
evoked only by faces but not by other types of stimuli (Fig.
1 C). Similar potentials were recorded from the fusiform
821
ALLISON ET AL.
0
400 800
0
400 800
MSEC MSEC
___._ ____-__---
scramoled faces
------ cars
-------
scrambled cars
5
I
5opJ
FIG.
I.
A : magnetic resonance images ( MRls) of the inferior surface of
the brain in patient RCN. This reconstructed oblique axial view shows the
locations of electrodes of right hemisphere strip RTTP. B: reconstructed
coronal MRI of strip RTTP. C: potentials recorded from strip RTTP.
Reconstructed oblique axial (D) and reconstructed oblique coronal (E)
MRI of strip LPTP. F: potentials recorded from strip LPTP. Stimulus
onset at 0 ms.
gyrus of the left hemisphere (Fig. 1, D-F). N200 was not
generated by equiluminant nonface images (scrambled
faces), by familiar complex objects with some face-like
properties (cars), or by equiluminant scrambled cars. N200
was also not generated by complex biological objects (but-
terflies). However, because butterflies were targets to which
the patient responded with a key press, longer-latency deci-
sion-related potentials (McCarthy et al. 1989) were re-
corded from some locations.
The typical anatomic specificity of N200 is illustrated in
a patient in whom it was recorded only from location LTTP
5 on the inferior temporal gyrus (Fig. 2, A-C). Words and
nonwords (“letterstrings”), in contrast, evoked a large nega-
tivity at - 150 ms only from location LTTP 4 on the fusi-
form gyrus (Fig. 2, A,
B,
and 0). However, this regional
specificity was not absolute; in two of the seven patients in
whom face and letterstring recordings were obtained, N200
was recorded from three sites on the fusiform gyrus, which
also recorded letterstring potentials like those at LTTP 4.
Figure 3 summarizes the locations from which N200 was
recorded across patients. With one exception in the left lat-
eral lingual gyrus the active sites were in the fusiform and
inferior temporal gyri (a total of 626 inferior surface loca-
tions were studied). The borders and centers of the active
regions are given in Talairach and Tournoux ( 198 8 ) coordi-
nates in Table 1. No active sites were found on the lateral
surface of the temporal lobes (291 locations), occipital
lobes ( 175 locations), parietal lobes ( 164 locations), or
frontal lobes ( 155 locations). Nor were active sites found
on the occipital mesial wall (cuneus and precuneus; 16 lo-
cations) or white matter (depth probes; 73 locations).
Across patients the peak latency of N200 was 193 + 18 ms
(left hemisphere), and 192 + 19 ms (right hemisphere).
N200 did not vary significantly in latency, amplitude, or
responsiveness to the categories of stimuli among the four
regions (left and right fusiform and inferior temporal gyri)
from which it was recorded.
Cortical stimulation produced in seven patients a tempo-
rary inability to name faces that the patient had previously
identified correctly. A striking example was seen in patient
RCN,
a state attorney who knew the governor well. Shown
a photograph of the governor during stimulation at LPTP
3-4 (Fig. 2,
D
and E), he hesitated and said “President
Bush?’ Upon termination of stimulation he gave the
correct name and was surprised at his inability to do so
earlier. Results of stimulation of this electrode strip are
given in Table 2. Object naming was not affected in this and
other cases in which stimulation induced a transient inabil-
ity to identify familiar faces. However, stimulation occa-
sionally affected other functions at the same locations that
produced an inability to name faces. In patient
DNB,
for
example, N200 was recorded from electrodes LPTP 2 and 3
(located on the fusiform gyrus). Stimulation of these elec-
trodes produced only an inability to name famous faces,
whereas stimulation of electrodes 3 and 4 (on the inferior
temporal gyrus) produced an inability to name famous
faces and desaturation of color. In patient
BME, N200
was
recorded only from electrode RPTP 3 (on the fusiform
gyrus); stimulation of electrodes 2-3 and 3-4 produced only
an inability to name family faces. However, stimulation of
electrodes 1 (on the lateral lingual gyrus) and 2 (on the
fusiform gyrus) produced. both a mild inability to name
family faces and retinotopically appropriate colored phos-
phenes; this was the only exception to the finding that stimu-
lation sites that produced an inability to name familiar
faces included at least one electrode that also recorded an
FACE RECOGNITION
A
6
823
FIG.
2. Axial (A) and coronal (B) MRIs of strip LTTP in patient MKY.
C and D: potentials recorded from strip LTTP.
6
, &OP”
0
400
800 0 400 800
MSEC
MSEC
N200. For the locations from which N200 was recorded
and which were stimulated, inability to name famous faces
occurred in 86% (6 / 7 ) of cases in the left hemisphere and
43% (3 /7) of cases in the right hemisphere. Patients did not
report distorted perception of the face; they simply could
not attach a name to it.
DISCUSSION
The major conclusion to be drawn from these recordings
is that discrete regions of inferior extrastriate visual cortex
are activated specifically by faces and not by the other cate-
gories of images tested. It is not possible to state categori-
cally that no other type of visual stimulus would be effective
in activating the same regions. For example, it is possible
that other human body parts (e.g., hands) would also be
effective; this and other categories of objects need to be
tested. However, it is clear that N200 is not due to a lumi-
nance or luminance-contrast effect, nor is it evoked by
complex biological or nonbiological objects per se.
The active region of the fusiform gyrus is somewhat ante-
rior to the posterior portion of the gyrus, which appears on
anatomic (Clarke and Miklossy 1990), metabolic (Cor-
betta et al. 199 1; Lueck et al. 1989)) and electrophysiologi-
cal (Allison et al. 1993a) grounds to be involved in the
perception of color. Comparison with the latter study shows
that across patients there is overlap in the location of color
and face regions. However, within a patient no overlap has
been observed. In the 11 patients of the present study who
were also tested with color recordings, no location that gen-
824
ALLISON ET AL.
FIG. 3. Summary of locations from which a surface-negative potential
(N2OO) was (0) and was not (- - -) recorded. The inferior view was made
of a brain obtained at autopsy. Both hemispheres are shown as identical,
but there is considerable variation in the morphology of this cortex. cs,
collateral sulcus; fg, fusiform gyrus: itg. inferior temporal gyrus; lg, lingual
gyrus: ots, occipitotemporal sulcus.
erated an N200 also showed a significant color effect. On
the other hand, the results of cortical stimulation summa-
rized above occasionally suggested either some overlap of
these functional regions or current spread to nearby func-
tionally different regions. Taken as a whole the results indi-
cate that I) neurons responsive to color are located mainly
in the posterior fusiform gyrus, 2) neurons responsive to
faces are located mainly in the middle portion of the fusi-
form gyrus, and 3) a border region is occasionally respon-
sive to both kinds of stimuli.
A region of the inferior temporal gyrus, posterior to the
face region of fusiform gyrus and lateral to the color region
of the posterior fusiform gyrus, also responded selectively
to faces (Figs. 2 and 3: Table 1). Whether this region is
anatomically contiguous and functionally similar to the fu-
siform region is unclear. N200 did not differ in its charac-
teristics in inferior temporal compared with fusiform sites,
but electrode sites in this region of the inferior temporal
gyrus were relatively sparse. Although the entire region
from which N200 was recorded was fairly large, only a
small portion of it appears to be activated in a given individ-
ual. This implies a considerable degree of individual vari-
ability in the location of “face modules.” Stimulation of
perisylvian cortex also suggests individual variability in the
TABLE 1.
Tulairwh cwrdinutcs of’ N200 locat ions
Center Range
Location .Y
!’
.Y
Y
Left fusiform -34 -38 -231-46
-241-57
Right fusiform 33 -46 26139 -231-63
Left inferior temporal -48 -73 -431-Q -56/-80
Right inferior temporal 45 -58 42149
-421-7 1
TABLE
2. Rmdts of’I.rwticd stirmlution (IO mA) in putimt RCN
Electrodes
Location Effect
LPTP l-2 fg
No retinotopic effect; color perception normal;
identification of famous faces normal
LPTP 2-3 fg
No retinotopic effect; color perception normal;
unable to name Pres. Kennedy, hesitation
in naming Benjamin Franklin; object
naming normal
LPTP 3-4 fg. itg
No retinotopic effect; color perception normal:
unable to name Michael Jackson; misnamed
state governor as Pres. Bush; object naming
normal
LPTP 4-5 itg No retinotopic effect; color perception normal:
stimulation terminated due to afterdischarge
in EEG
For electrode locations see Fig. 1, D and E. fg, fusiform gyrus: itg, infe-
rior temporal gyrus; EEG, electroencephalogram.
location of discrete language-related regions (Ojemann et
al. 1989).
N200 occurs - 100 ms after PlOO, the first prominent
potential recorded from the posterior scalp and thought to
be generated in occipital extrastriate cortex (Halliday and
Michael 1970; Jeffreys and Axford 1972). In our recordings
the faces were unfamiliar to patients, and no explicit identi-
fication or memory task was involved. N200 may thus re-
flect an early stage in the processing of facial information,
such as the initial “structural encoding” stage in the model
of Bruce and Young ( 1986) or the “template formation”
stage in the model of Damasio et al. ( 1982), in which tem-
plates of newly encountered faces are thought to be formed
and stored in the lingual and fusiform gyri. Identification
and naming of faces presumably occurs at later stages of
processing. The transient disruption of face naming pro-
duced by cortical stimulation (Table 2) may be due to
disruption of the flow of information to these modules, be-
cause stimulation did not appear to affect the perception of
faces.
Two groups have studied face recognition by positron
emission tomography (PET) using a variety of face-match-
ing and face-discrimination tasks (Haxby et al. 1993; Ser-
gent et al. 1992). The tasks used by Sergent et al. activated
several cortical regions. Of most relevance here, the centers
of activation of the left and right fusiform gyrus were poste-
rior and lateral to the N200 fusiform locations but were
very close to the centers of the combined fusiform and infe-
rior temporal locations, suggesting that the active PET re-
gions were a combination of the fusiform and inferior tem-
poral regions from which N200 was recorded. The centers
of the active regions of the left and right anterior fusiform
gyrus reported by Haxby et al. are within the range of N200
locations. Their posterior fusiform active regions are
slightly medial to the regions of inferior temporal cortex
from which N200 was recorded. Both PET studies found
additional activation of inferior occipital cortex. This activ-
ity is apparently not manifested as N200 potentials; how-
ever, we had few electrode sites in this region (Fig. 3 ).
It has been argued that face recognition is carried out by
special brain mechanisms (Bodamer 1947; Farah 1990;
Hecaen and Angelergues 1962; Whitely and Warrington
1977). However, face recognition can be thought of simply
FACE RECOGNITION 825
as an important example of a general process of object recog-
nition (Critchley 1986; Damasio et al. 1982), based on stud-
ies of patients who have lost the ability to identify familiar
faces and often exhibit other types of visual agnosia as well.
Our results suggest that small regions of the fusiform and
inferior temporal gyri are indeed devoted specifically to
face recognition. These face modules may be homologous
with regions of monkey temporal cortex that respond prefer-
entially to faces (e.g., Bruce et al. 198 1; Desimone 199 1;
Harries and Perrett 199 1) and may be intercalated with
other functionally specific regions (e.g., a “letterstring mod-
ule” at location 4 of Fig. 2). Given the apparent mosaic of
functions parcelled out to small regions of this cortex, it is
not surprising that bilateral large lesions might disturb iden-
tification of various categories of objects, giving rise to the
impression that face recognition is part of a unitary process
of object recognition.
We thank H. Cohen, C. Faulkner, J. Jasiorkowski, and K. McCarthy for
assistance.
This work was supported by the Department of Veterans Affairs and
National Institute of Mental Health Grant MH-05286.
Address for reprint requests: T. Allison, Neuropsychology Laboratory
1 16B1, VA Medical Center, West Haven, CT 065 16.
Present addresses: H. Ginter, University of Maryland Medical School,
Baltimore, MD 2 1202; A. C. Nobre, Dept. of Neurology, Beth Israel Hospi-
tal, Boston, MA 022 15.
Received 30 August 1993; accepted in final form 4 November 1993.
REFERENCES
ALLISON, T., BEGLEITER, A., MCCARTHY, G., ROESSLER, E., NOBRE,
A. C., AND SPENCER, D. D. Electrophysiological studies of color process-
ing in human visual
cortex. Elc~troenc~~phal~~~r. C/in. Nezuvphysiol. 88:
343-355, 1993a.
ALLISON, T., GINTER, H., MCCARTHY, G., NOBRE, A. C., PUCE, A., LUBY,
M., MCCARTHY, K., AND SPENCER, D. D. Electrophysiological studies
of face recognition in human extrastriate cortex. Sot.
Nc>zrrosci. Ahstr.
19: 976, 1993b.
BENTON, A. L., HAMSHER, K. DES., VARNEY, N. R., AND SPREEN, 0.
Contributions to N~~zrr(~ps~~~hol~)~ic~ai As.ressmc>nt. New
York: Oxford
University Press, 1983, 146 pp.
BODAMER, J. Die Prosop-agnosie: die Agnosie des Physiognomieerken-
nens.
Arch. Psyhiatr. Nuvclnkrank.
179: 6-54, 1947.
BRUCE, C. J., DESIMONE, R., AND GROSS, C. G. Visual properties of neu-
rons in a polysensory area in superior temporal sulcus of the macaque. J.
NezuvphJuGl. 46: 369-384,
198 1.
BRUCE, V. AND YOUNG, A. Understanding face recognition.
Br. J. Psycho/.
77: 305-327, 1986.
CRITCHLEY, M.
The Citadel c!f‘thd> Scvscs. New
York: Raven Press, 1986,
277 pp.
CLARKE, S. AND MIKLOSSY, J. Occipital cortex in man: organization of
callosal connections, related myelo- and cytoarchitecture, and putative
boundaries of functional visual areas. J. ~~~rnp.
Ncurol. 298:
188-2 14,
1990.
CORBETTA, M., MIEZIN, F. M., DOBMEYER, S., SHULMAN, G. L., AND
PETERSON, S. E. Selective and divided attention during visual discrimi-
nation of shape, color, and speed: functional anatomy by positron emis-
sion tomography. J.
Nc~zuosb.
1 1: 2383-2402, 199 1.
DAMASIO, A. R., DAMASIO, H., AND VAN HOESEN, G. W. Prosopagnosia:
anatomic basis and behavioral mechanisms.
Neurollgy 32: 33
l-34 1,
1982.
DESIMONE, R. Face-selective cells in the temporal cortex of monkeys. J.
Cogn. Ncurosci. 3:
l-8, 199 1.
FARAH, M. J.
Visual Agnosia.
Cambridge: MIT Press, 1990, 184 pp.
HALLIDAY, A. M. AND MICHAEL, W. F. Changes in pattern-evoked re-
sponses in man associated with the vertical and horizontal meridians of
the visual field. J.
Physiol. Land. 208: 499-5
13, 1970.
HARRIES, M. H. AND PERRETT, D. I. Visual processing of faces in temporal
cortex: physiological evidence for a modular organization and possible
anatomical correlates. J.
C’ogn. Nuu-osci. 3: 9-24,
199 1.
HAXBY, J. V., GRADY, C. L., HORWITZ, B., SALERNO, J., UNGERLEIDER,
L. G., MISHKIN, M., AND SCHAPIRO, M. B. Dissociation of object and
spatial visual processing pathways in human extrastriate cortex. In:
Functional Organisation cfthc Human Visual Cortex,
edited by B. Gu-
lyas, D. Ottoson, and P. E. Roland. Oxford: Pergamon Press, 1993, p.
329-340.
HECAEN, H. AND ANGELERGUES, R. Agnosia for faces (prosopagnosia).
Arc-h. Ncurol. 7: 24-32, 1962.
JEFFREYS, D. A. AND AXFORD, J. G. Source locations of pattern-specific
components of human visual evoked potentials. II. Component of ex-
trastriate cortical origin.
E-K-~. Brain Rc1.s.
16: 22-40, 1972.
LUECK, C. J., ZEKI, S., FRISTON, K. J., DEIBER, M. P., COPE, P., CUN-
NINGHAM, V. J., LAMMERTSMA, A. A., KENNARD, C., AND FRACKO-
WIAK, R. S. J. The colour centre in the cerebral cortex of man.
Nature
Lond. 340: 386-389,
1989.
MCCARTHY, G., WOOD, C. C., WILLIAMSON, P. D., AND SPENCER, D. D.
Task-dependent field potentials in human hippocampal formation. J.
Nclzuosc*i. 9: 4253-4268,
1989.
OJEMANN, G., OJEMANN, J., LETTICH, E., AND BERGER, M. Cortical lan-
guage localization in left, dominant hemisphere: an electrical stimula-
tion mapping investigation in 1 17 patients. J.
Nezrrosurg. 7
1: 3 16-326,
1989.
SERGENT, J., OHTA, S., AND MACDONALD, B. Functional neuroanatomy
of face and object processing: a positron emission tomography study.
Brain
115: 15-36, 1992.
SPENCER, S. S., SPENCER, D. D., WILLIAMSON, P. D., AND MATTSON,
R. H. The localizing value of depth electroencephalography in 32 refrac-
tory patients.
Ann. Nclrrol.
12: 248-253, 1982.
TALAIRACH, J. AND TOURNOUX, P.
Co-planar Stcrcotasic’ Atlas c!f’ the
Human Brain. New
York: Thieme Publishers, 1988, 122 pp.
WHITELY, A. M. AND WARRINGTON, E. K. Prosopagnosia: a clinical, psy-
chological, and anatomical study of three patients. J.
Nuu~~I. Ncjuro-
surg. P.syl>hiatr. 40: 395-403,
1977.