Content uploaded by Aysenil Belger
Author content
All content in this area was uploaded by Aysenil Belger
Content may be subject to copyright.
Human Extrastriate Visual Cortex
and the Perception
of
Faces,
Words, Numbers, and Colors
Truett Allison, Gregory McCarthy, Anna Nobre, Aim
Puce, and Aysenil Belger
Neuropsychology Laboratory,
VA
Medical Center,
West Haven, Connecticut 06516 and Department of
Neurology and Section of Neurosurgery, Yale
University School of Medicine, New Haven,
Connecticut 06510
Electrophysiological correlates of the processing of vi-
sual information were studied in epileptic patients with
electrodes chronically implanted on the surface of
stri-
ate and extrastriate cortex. In separate experiments pa-
tients viewed faces, letter strings (words and non-
words),
numbers, and control stimuli. A negative
potential,
N200, was evoked by
faces,
letter strings, and
numbers, but not by the control stimuli. N200 was re-
corded bilaterally from discrete regions of the fusiform
and inferior temporal
gyri.
These category-specific
face,
letter-string, and number "modules" vary in lo-
cation.
In most cases there was no overlap in the lo-
cation of face and letter-string modules, suggesting a
mosaic of functionally discrete regions. In some cases
letter-string and number N200s were recorded from the
same location, suggesting that these modules may be
less spatially and functionally discrete. Face N200-like
potentials can be recorded from temporal scalp, allow-
ing the possibility of studying early face processing in
normal subjects. Longer-latency face-specific poten-
tials were recorded from the inferior surface of the an-
terior temporal lobe. Potentials evoked by colored
checkerboards were recorded from a region of the fu-
siform gyrus posterior to the fusiform region from which
category-specific N200s were recorded.
These results suggest that there are several pro-
cessing streams in inferior extrastriate cortex. In ad-
dition to object recognition systems previously pro-
posed for faces and words, our preliminary results
suggest a separate system dealing with numbers. Pos-
tulated systems dealing with larger manipulate ob-
jects and animals have not been detected.
Little is known about the manner in which human
extrastriate cortex participates in the processing of
visual information. The ventral or inferior portion of
this region lies buried at the base of the brain over-
lying the cerebellum. It is not exposed by craniotomy
and has thus not been studied by the usual intraoper-
ative localizing techniques of cortical stimulation (e.g.,
to localize language-related cortex;e.g.,Ojemann et al.,
1989),
or evoked potential recording (e.g., to localize
sensorimotor cortex; e.g., Wood et al., 1988). Most of
what we know comes from the study of patients with
naturally occurring lesions, or from inferences derived
from anatomical and physiological studies in nonhu-
man primates.
In the last few years this situation has changed. Pos-
itron emission tomography (PET) now allows the
study of cortical regions participating in specific sen-
sory and cognitive tasks (e.g., Corbetta et al., 1991)-
Previously inaccessible cortex can now be studied
electrophysiologically in conjunction with chronic im-
plantation of electrodes in patients with medically in-
tractable epilepsy who are being evaluated for possi-
ble surgery (e.g., Fried et al., 1991). Here we review
the results of electrophysiological recordings in an ini-
tial series of 34 patients. The results suggest a consid-
erable degree of modularity in the processing of faces,
words, numbers, and colors, resulting in a mosaic of
functionally discrete regions in inferior extrastriate
cortex.
The Perception of Faces
Damage to the occipitotemporal region of the brain,
usually due to occlusion of a branch of the posterior
cerebral artery, may produce an inability to recognize
familiar faces. This condition is called face agnosia or
prosopagnosia (reviewed by Whitely and Warrington,
1977;
Damasio et al., 1982, 1990). Prosopagnosia is
usually produced by bilateral damage but can be pro-
duced by damage to the right hemisphere alone (e.g.,
Michel et al., 1989). A visual field defect, if present,
involves the upper quadrant of the affected field, dem-
onstrating that the essential lesion is located in infe-
rior extrastriate cortex.
We have been studying this conical region in pa-
tients with electrodes on inferior temporal and occip-
ital cortex. Given the clinical literature demonstrating
that this region of cortex is involved in face recogni-
tion, we asked whether viewing faces would generate
Cerebral Cortex Scp/Oct 1994:5:544-554; 1047-3211/94/J4.OO
at University of North Carolina at Chapel Hill on November 1, 2013http://cercor.oxfordjournals.org/Downloaded from at University of North Carolina at Chapel Hill on November 1, 2013http://cercor.oxfordjournals.org/Downloaded from at University of North Carolina at Chapel Hill on November 1, 2013http://cercor.oxfordjournals.org/Downloaded from at University of North Carolina at Chapel Hill on November 1, 2013http://cercor.oxfordjournals.org/Downloaded from at University of North Carolina at Chapel Hill on November 1, 2013http://cercor.oxfordjournals.org/Downloaded from at University of North Carolina at Chapel Hill on November 1, 2013http://cercor.oxfordjournals.org/Downloaded from at University of North Carolina at Chapel Hill on November 1, 2013http://cercor.oxfordjournals.org/Downloaded from at University of North Carolina at Chapel Hill on November 1, 2013http://cercor.oxfordjournals.org/Downloaded from at University of North Carolina at Chapel Hill on November 1, 2013http://cercor.oxfordjournals.org/Downloaded from at University of North Carolina at Chapel Hill on November 1, 2013http://cercor.oxfordjournals.org/Downloaded from at University of North Carolina at Chapel Hill on November 1, 2013http://cercor.oxfordjournals.org/Downloaded from
facescrambled face
carscrambled car
butterfly
Fignre 1. Stimuli used to test fix face-specific neuronal activity.
face-specific neuronal activity. Five types of stimuli
were presented (Fig. 1). Faces were digitally scanned
from a college yearbook, cars were scanned from au-
tomotive magazines, and butterflies were scanned
from a field guide. The patient pressed a key to the
butterflies as a measure of attention. Graphics soft-
ware was used to rearrange the face and car images
to produce equiluminant scrambled images. The im-
ages followed a fixation point, subtended 7.2° X 7.2°
of visual angle, and were displayed for 250 msec at
intervals of
1.8-2.2
sec. Average luminance (101 cd/
m2) was not significantly different for the five types
of images. Each set of 110 stimuli consisted of 25 fac-
es,
cars, and their scrambled equivalents, and 10 but-
terflies. Three stimulus sets were presented; wave-
forms shown were averaged across all stimuli. Field
potentials were recorded from strips of stainless steel
electrodes resting on the cortical surface. Potentials
were recorded (referential to a mastoid) simultaneous-
ly from 32 or 64 locations using filter settings of
0.1
-
100 Hz and digitized at 250 Hz. Electrodes were mag-
netic resonance (MR) imaged to allow precise
localization in relation to sulci and gyri of occipito-
temporal cortex, as illustrated in Figure 2. Electrode
locations were also determined in the coordinate sys-
tem of Talairach and Tournoux (1988).
The most striking result of Initial recordings in 24
patients was that a large-amplitude (up to 200 \LV)
negative potential with a peak latency of about 200
msec (N200) was generated by faces but not by the
other categories of stimuli (Fig. 3; see also Allison et
al.,
1994). N200 was generated bilaterally in regions of
the fusiform and inferior temporal gyri indicated by
stippling in Figure 4. The response characteristics of
Cerebral Cortex Sep/Oct 1994, V 4 N 5 545
AB
Figure 2. MR localization
of
cortical surface electrodes (interelectrode distance, 1 cm): oblique axial M) and coronal (fl) images reconstructed along
a
plane
passing through the long axis of the electrode strip. In this example, electrode strip RMT extended from the medial lingual gyrus (location 1) to the lateral lingual
gyrus (location 2), fusiform gyrus (locations 3 and 4), and inferior temporal gyrus (locations 5-7). The electrode artifact is larger than the electrode itself (exposed
surface, 11 mm diameter). C, cuneus;
CBS,
calcarine sulcus;
CG,
cuneate gyrus;
CoS,
collateral sulcus;
FG,
fusiform gyrus;
ITG,
inferior temporal gyrus;
LLG,
lateral
lingual gyrus; MLG, medial lingual gyrus,
OTS,
occipitotemporal sulcus. Adapted from Allison
et
al. (1993).
scrambled (aces
cars
50J/V
400
MSFC800
Figure 3. Examples
of
locally generated field potentials evoked by the stimuli
of
Figure 1. A, Location of recording sites determined as in Figure 2. B, Initial face-
specific potential (N2C0) followed by a later face-specific positivity at 280 msec. (The face-specific positivity at 150 msec was often not present and when present
was usuaBy not face specific.) Patient
SPS;
Talairach coordinates x-34, y-49.
C,
N200 followed by a later negativity at 310 msec. Patient ONB; Talairach coordinates
x-29,
y-40.
D,
Visual evoked potentials recorded from the occipital pole. Patient AHK; Talairach coordinates
x-7,
y-95.
LG,
lingual gyrus;
PG,
parahippocampal gyrus;
other abbreviations are as in Figure
2.
548 Human Extrastrlate Cortex • Allison ct al.
N200 were similar in both hemispheres and in the
fusiform and inferior temporal regions. In a given in-
dividual only a small portion of the total stippled re-
gion was activated; hence, there was considerable in-
terindividual variability in the location of these "face
modules." N200 occurred about 140 msec after the
initial activation of striate cortex by a visual stimulus,
and about 100 msec after initial activation of peristri-
ate cortex (reviewed by Allison et al., 1984). It is the
earliest face-specific activity we have recorded, sug-
gesting that a considerable amount of stimulus pro-
cessing occurs between initial activation of striate cor-
tex and activation of face modules. Cortical
stimulation at sites that generated N200 often pro-
duced a transient prosopagnosia (Allison et al., 1994),
suggesting that face modules are an essential part of
a pathway involved in face recognition.
In PET studies, Sergent et al. (1992) and Haxby et
al.
(1993) used face-matching and face-identification
tasks to identify regions of cortical activation. Of most
relevance in this context are the activated regions of
inferior extrastriate cortex. The centers of activation
in the left and right hemisphere identified by Sergent
et al. (Fig. 4) were close to the centers of the com-
bined fusiform/inferior temporal regions that generate
N200.
The right and left hemisphere anterior fusiform
regions identified by Haxby et
al.
are within the range
of fusiform N200 locations, while their posterior fusi-
form regions are slightly medial to our inferior tem-
poral N200 locations (Fig- 4). Considering the differ-
ences in tasks and techniques among the three
studies, there is good agreement regarding the regions
of inferior cortex involved in face perception. The PET
studies required across-subject averages to determine
regions of activation. It should be possible to image
brain regions involved in face processing in individual
subjects using functional MR techniques (e.g., McCar-
thy et al., 1993). In pilot studies we have imaged the
stippled region of Figure 4 using an echo-planar tech-
nique in conjunction with face and control stimuli
like those of Figure 1. We have not as yet obtained
clear MR activation by faces, perhaps because suscep-
tibility artifacts induced by the ear canals make this
region of inferior cortex difficult to image. Functional
MR studies, particularly in patients who have also
been studied electrophysiologically, will provide an-
other valuable technique for the study of face and ob-
ject recognition.
N200 was followed by later potentials that were
also face specific. There were two types of later activ-
ity: positive potentials with a peak latency of 250-350
msec (Fig. 3/J), or negative potentials with a peak la-
tency of 300-500 msec (Fig. 3C). The behavioral cor-
relates of these potentials are unknown; we surmise
that they will be more affected by attentional and
mnemonic variables than is N200. Visual evoked po-
tentials recorded from striate or peristriate cortex dif-
fered in several respects from the extrastriate poten-
tials:
initial peak latency was shorter (120 msec in the
example shown in Fig. 3-D), an off-potential followed
the initial on-potential by 250 msec (stimulus dura-
tion),
and these potentials were often largest to scram-
Flgure 4. Summary of functional regions
of
inferior extrastnate cortex. Sup-
pling, regions from which category-specific N200 potentials were recorded
to faces (Allison
et
al., 1994), letter strings, and numbers. Crosshatchmg,
regions from which the largest percentage of color potentials were recorded
(Allison et al., 1993).
C,
centers
of
PET activation in a color-attend task (Cor-
betta et
al.,
1991); H, centers of PET activation in a face-matching task (Haxby
et al., 1993);
I
centers
of
PET activation in
a
color task (Lueck
et
al., 1989),
5, centers of PET activation in a face-identification task (Sergent et al., 1992)
bled images, probably due to additional luminance-
contrast borders created by scrambling (Fig- 1)-
There is considerable evidence that the left and
right hemispheres process face information differently
(reviewed by Rhodes, 1993). Faces presented in the
left visual field in the usual upright orientation are
identified more rapidly (e.g., Geffen et al., 1971) and
more accurately (e.g., Leehey et al., 1978) than when
presented in the right visual field. However, the left
visual field (right hemisphere) advantage may disap-
pear when inverted faces are presented (Leehey et al.,
1978),
suggesting that the right hemisphere tends to
use configural features to form a gestalt of the face
whereas the left hemisphere tends to use a feature-
detection strategy Hemispheric differences in face
processing are also seen electrophysiologically (Fig.
5).
In this patient electrodes were located nearly sym-
metrically in the left and right fusiform gyrus. Upright
and inverted faces evoked essentially identical N200s
in the left hemisphere (Fig. 5B). In contrast, N200 re-
corded from the right hemisphere was considerably
smaller and 20 msec later when evoked by inverted
faces compared to upright faces (Fig. 5C). These ef-
fects are not due to inverted images per se, since up-
right and inverted cars evoked similar potentials in
both hemispheres. These effects were obtained with
faces presented centrally; how N200 and other face-
specific activities behave when stimuli are presented
Cerebral Cortex Scp/Oct 1994, V 4 N 5 547
inverted faces
inverted cars
N20050J/V
400800
figure 5. Interhemispheric differences
in
processing upright
and
inverted
faces.
A
Location
of
recording sites. Patient MBN.
B,
Potentials recorded
from
the
left hemisphere (Talairach coordinates x-43, y-51). C, Potentials
re-
corded from
the
right hemisphere (Talairach coordinates x33, y-47).
in the left and right visual fields remains to be deter-
mined.
In monkeys, face-selective neurons in the upper
wall of the superior temporal sulcus also show an in-
crease in latency when activated by inverted faces
compared with upright faces (Perrett et a!., 1988).
A
more direct comparison with our recordings might be
with single-unit recordings in ventral inferior temporal
cortex (area TE or ATT), which on anatomical and
physiological grounds (Desimone,
1991;
Tanaka et al.,
1991;Heywood and Cowey, 1992;Perrett et al., 1992;
Gross et al., 1993) is likely to be the monkey homolog
of the stippled region of Figure
4.
However, face-selec-
tive neurons in the ventral region (e.g., Yamane et al.,
1988;
Tanaka et al., 1991) have apparently not been
faces
scrambled faces
cars
400
MSEC800
figure 6. Face-specific potentials recorded from
the
anterior temporal lobe
A Location
of
recordings sites.
B,
Anterior negative potential. Patient
SGI;
Talairach coordinates x26,
yl
C, Anterior positive potential. Patient CLR;
Ta-
lairach coordinates x24,
y-8.
tested for orientation selectivity. In humans, a scalp-
positive potential in the 170-200 msec latency range
is also smaller and delayed when evoked by inverted
compared with upright faces (Jeffreys,
1993).
This po-
tential, best recorded from the vertex and larger to
face than to nonface stimuli (e.g., Seeck and Griisser,
1992;
Jeffreys,
1993),
may in part reflect the same neu-
ronal activity as does N200. An equivalent dipole
source in inferior cortex generating a surface-negative
potential (in this case N200) would generate an equiv-
alent P2OO as recorded above inferior cortex (in this
case from the upper
scalp),
similar to the topographic
pattern described for
auditor}'
potentials generated in
the superior temporal plane (Wood and Wolpaw,
1982).
In addition to the face-specific potentials recorded
from the stippled region of Figure 4, long-latency face-
specific potentials were recorded from the anterior
portion of inferior temporal cortex
(Fig. 6).
They were
548 Human Extrastriatc Cortex • Allison et al.
variable in latency and waveform, and were negative
(Fig.
6ff) or positive
(Tig. 6C).
They were also variable
in location but seem to be concentrated in or near
entorhinal cortex, as in Figure 6, suggesting that they
reflect later stages of mnemonic or associative pro-
cessing of face information. These face-specinc poten-
tials are not P300-like potentials elicited by targets (re-
viewed by McCarthy et al., 1989) since at these
locations such potentials were not evoked by target
butterflies (Fig. 6B,C)
Shlomo Bentin at Hebrew University is studying
scalp-recorded potentials evoked by the stimuli of
Fig-
ure 1. A face-specific negativity at about 170 msec is
recorded from temporal scalp locations T, and T6, and
a slightly later positivity is recorded from the vertex
(C,) at about 180 msec
(Fig. 7).
The negativity is larger
over the right hemisphere (T6), perhaps reflecting the
greater role of the right hemisphere in processing face
information (Geffen et al.,
1971;
Whitely and Warring-
ton, 1977; Leehey et al., 1978; Michel et al., 1989;
Rhodes, 1993). While the scalp negativity and N200
are similar enough to suggest their correspondence,
their relationship is not altogether
clear.
T,
and T6 elec-
trodes are typically located over the posterior middle
temporal gyrus (Homan et al., 1987) and are therefore
not well placed to record potentials generated in in-
ferior cortex. Furthermore, we have never recorded a
face-specific N200 from the lateral surface of the tem-
poral lobe. The positivity recorded from the vertex
may reflect the same scalp-positive potential recorded
in other studies of face processing (e.g., Seeck and
Griisser, 1992; Jeffreys, 1993). The relationship be-
tween N200 and scalp-recorded potentials remains to
be clarified. In any case, the ability to record N200 or
N200-like face-specific potentials from the scalp is im-
portant because it should be possible to study face
processing developmental!^ in patients with proso-
pagnosia or other conditions that affect face recogni-
tion, and in other settings in which intracraniaJ re-
cordings cannot be obtained.
The Perception of Words
We have been studying potentials generated by con-
tent words (e.g., apple), by legal nonwords or pseu-
dowords (e.g., glimp), and by illegal nonwords (e.g.,
xpczt).
Letter strings were displayed for 500 msec at
intervals of 2.1-2.2 sec as red characters on a white
rectangular background, and subtended 5-10° of vi-
sual angle horizontally and 0.5-1.0° vertically. Re-
cording conditions were otherwise as described
above.
A
variety of word-related potentials were seen
in several cortical regions (Nobre and McCarthy,
1994).
Of most relevance to this review, focal nega-
tive potentials generated by word-related stimuli
were recorded from the same regions of the fusiform
and inferior temporal gyri that generated the face
N200.
The word-related N200 was evoked equally
well by content words and by legal and illegal non-
words (Fig. SB), that is, by any letter string. Face and
letter-string N200s were similar in waveform, but let-
ter-string N200s were typically shorter in peak laten-
cy by about 30 msec. Electrodes that recorded a let-
faces
scrambled faces
cars
scrambled cars
10;/V
400
MSEC800
Figure 7. Potentials recorded from the vertex (C,|, and from left (7,) and right
(7,1 posterior temporal scalp, averaged across 10 subjects.
ter-string N200 did not as a rule also record a face
N200,
suggesting a considerable degree of anatomical
and functional specificity of "letter-string modules"
in extrastriate cortex. As in the case of face N200s,
letter-string N200s were often followed by longer-la-
tency letter-string-specific potentials (e.g., Fig. SB,
the negativity at about 300 msec). Long-latency po-
tentials evoked by letter-strings were also recorded
from more anterior regions of the inferior temporal
lobe,
particularly near entorhinal cortex. This activity
is modified by word type and semantic priming (No-
bre and McCarthy, 1994).
The Perception
of
Numbers
Puce et al. (1994) recorded potentials generated in a
task in which patients learned to associate words
(concrete nouns and adjectives) with ideograms (Jap-
anese characters). In a distractor task simple arithme-
tic expressions (e.g., 4 + 3 = ?) were presented. Sur-
prisingly, these equations evoked an N200 in discrete
regions of extrastriate cortex. Words and ideograms
usually did not evoke an N200 at the same locations,
but an arithmetic operation was required and might
have been important in evoking the
potential.
To
char-
acterize this activity more systematically we have pre-
sented single or paired Arabic numbers and six types
of control
stimuli.
Stimuli were displayed for 300 msec
Cerebral Cortex Sep/Oct 1994, V 4 N 5 549
content words
faces
nonwords
pseudowords
arabic numbers
concrete nouns
faces
illegal nonwords
false fonts
+
N200
0400
MSEC
T
50pV
800
Figure 8. Potentials evoked
by
letter strings and numberi
A,
Location
of
recording sites. B, Letter-string N200. Patient
MKY;
Talairach coordinates x-45,
y-55.
C,
Number N200. Patient AFO; Talairach coordinates x-26, y-38.
at intervals
of
1.4-1.6
sec as
white characters
on a
black background, and subtended
1-12° of
visual
an-
gle horizontally
and 1.2°
vertically. Recording condi-
tions were otherwise
as
described above. Figure
8C
shows potentials recorded from
a
location
on the fu-
siform gyms.
A
"number
N2OO"
was recorded
to Ara-
bic numbers
but not to
faces, letter strings,
or
false
fonts (illegal number-like
or
letter-like characters). Ar-
abic numbers presented alone therefore sufficed
to
generate
a
number N200.
In
this recording
the re-
sponse selectivity
to
numbers
is
clear. However,
in
some cases number
and
letter-string N200s were
re-
corded from
the
same location. The degree
of
spatial
and functional differentiation
of
number
and
letter-
string modules remains
to be
determined.
All
locations
to date from which a number N200 has been recorded
were within
the
stippled region
of
Figure
4.
The Perception
of
Color
Damage
to
occipitotemporal cortex
may
produce
a
selective deficit
in the
perception
of
color. This
con-
dition
is
referred
to as a
cerebral
or
central achroma-
topsia
to
indicate that
the
color blindness
is
cerebral
rather than retinal
in
origin (reviewed
by
Damasio
et
al.,
1980; Zeki, 1990). Achromatopsia
is
often,
but not
necessarily, accompanied by
an
upper quadrant defect
in
the
affected half field, demonstrating that
the es-
sential lesion
is
located
in
inferior extrastriate cortex.
We have examined
the
involvement
of
this region
of cortex
in
color vision
by
recording potentials
evoked
by red and
blue checkerboards (Allison
et al.,
1993).
The checkerboards were displayed
at
intervals
of 3.5-4.0
sec in an
adaptation stimulus (1000 msec
duration)-test stimulus
(100
msec duration) design,
and subtended 21°
of
visual
angle horizontally and
16°
vertically. Recording conditions were otherwise
as de-
scribed above. Regions showing
a
high percentage
of
statistically significant color potentials
are
indicated
by crosshatching
in
Figure
4. The
color region
in-
volves
the
posterior portion
of
the fusiform gyrus and
may extend into
the
lateral portion
of the
lingual
gyrus,
regions that have been implicated
in the per-
ception
of
color since Verrey's description
of an ach-
romatopsia
in 1887 (see
Zeki, 1990). Electrical stim-
ulation
of the
same locations that show
a
significant
color effect often produces colored phosphenes
or,
less commonly, desaturation
of
color.
By
contrast,
in
our experience stimulation
of
striate cortex rarely pro-
duces color effects. We have
not
detected longer-la-
tency color-specific potentials
in
more anterior
regions
of
the temporal lobe. With
the
proper exper-
imental conditions such activity might
be
recordable,
for example,
in the
region
of the
left anterior lingual
gyrus thought
by
Damasio
and
Damasio (1983)
to be
involved
in
color naming.
Two PET studies
of
color perception have been
car-
ried out. Lueck
et al.
(1989) presented colored
and
equiluminant gray-scale stimuli,
and
activated
a pos-
terior region
of
fusiform gyrus (Fig.
4)
similar
to the
region
we
found electrophysiologically. Corbetta
et al.
(1991) presented stimuli that varied along several
di-
mensions including color, and found that
an
attend-to-
color condition activated
a
slightly more posterior
re-
gion
(Fig. 4).
Anatomical evidence (Clarke
and
Miklossy,
1990)
also suggests that
the
posterior fusi-
form gyrus
may be, or be a
part
of, the
human
ho-
molog
of
area V4
in
monkeys,
a
region implicated
in
the perception
of
color
and
form (reviewed
by
Zeki,
1990).
Many neurons
in
V4 are selective
for
color rath-
er than wavelength,
in the
sense diat their responses
correlate with
the
human perception
of
color; that
is,
they exhibit color constancy (Zeki, 1990). Responses
of neurons
in VI do not
exhibit this property. Thus,
according
to
Zeki, V4
is the
initial region
of
cortex
in
which color
is
treated
as
color, rather than wave-
length composition.
General Discussion
Early
Visual
Processing in Extrastriate Cortex
Even
at the
preliminary stage
of
investigation
re-
viewed here,
it
appears that four processing streams
559 Human Extra5triatc Cortex
•
Allison
et a].
can be discerned subsequent to initial processing in
striate and peristriate cortex. Three deal with specific
categories of objects (if faces, letter strings, and num-
bers can be considered "objects"), and the fourth
deals with color.
Faces
The stippled region of Figure 4 appears to be the ini-
tial region of cortex in which faces are treated as fac-
es ("this is a face"), rather than as, for example, an
irregular ellipse containing two ovals, a vertical line,
and a horizontal line. This assertion is based on the
internal evidence of specificity reviewed above, and
on the plausible (albeit unproven) assumption that
the neurons generating N200 are functionally similar
to the neurons in monkey temporal cortex that re-
spond selectively to faces or face components (re-
viewed by Desimone,
1991;
Harries and Perrett, 1991;
Perrett et al., 1992; Gross et al., 1993). Specifically, the
human neurons may be equivalent to neurons in the
anterior ventral portion of monkey inferior temporal
cortex, the first cortical locus along the ventral path-
way that contains face-selective neurons (Tanaka et al.,
1991;
Perrett et al., 1992; Gross et al., 1993). To take
the argument a step farther, face modules may be
functionally similar to the patches of monkey cortex
containing a relatively large percentage of face-selec-
tive neurons (Harries and Perrett, 1991). We cannot
estimate the size of a face module with any certainty
since usually only one electrode strip (oriented me-
diolaterally, or obliquely as in Fig. 2) happens to be
placed on the relevant cortex. However, N200 is usu-
ally recorded from only one electrode (and from not
more than two adjacent electrodes) of a strip, sug-
gesting that in the mediolateral dimension the module
is no larger than 1-2 cm wide.
Letter strings
The stippled region of Figure 4 may be the initial re-
gion of cortex in which letter strings are treated as
words or possible words ("this is a string of letters
that may be a word") rather than as a configuration
of straight and curved lines. Surprisingly, letter-string
modules treat words, pseudowords, and nonwords
equally, at least during the initial activity recorded as
a letter-string N200
(Fig.
8B).
This suggests that lexical
and semantic processing occurs later or, alternatively,
that a letter-string N200 reflects activity at an earlier
"pre-letter-string" stage of processing. Recordings us-
ing progressively simplified and decomposed stimuli
are needed to determine the necessary and sufficient
conditions to evoke a letter-string N200 (for the same
strategy applied to recordings in monkey inferior tem-
poral cortex, see Gross et al., 1972; Tanaka, 1993).
Numbers
Only recently have we begun recording potentials
evoked by numbers and arithmetic equations. How-
ever, preliminary results
(Fig.
8C) suggest that number
N200s have the same general properties as face and
letter-string N200s, and are also generated in the stip-
pled region of Figure
4.
A
number module may be the
first cortical region that treats numbers as numbers
("this is a number or arithmetic expression"). As in
the case of letter strings, additional recordings using
a variety of number and decomposed number-like
stimuli will be required to determine the response
properties of the presumptive number modules.
It is tempting to speculate that some cases of word
alexia and number alexia (reviewed by Damasio and
Damasio, 1983; Levin et al., 1993) are manifestations
of damage to letter-string and number modules, re-
spectively. If the initial processing of faces, letter-
strings, and numbers occurs in the same general re-
gion of extrastriate cortex, as our results suggest, one
would expect lesions of the region to produce cor-
related deficits in these functions. Such cases are the
exception rather than the rule. While it is true that
word and number alexia often occur together, their
association with prosopagnosia is infrequent. For ex-
ample, in 16 patients with alexia studied by Damasio
and Damasio (1983), only two had an associated pro-
sopagnosia. Such dissociations may simply reflect vari-
ability in the location of modules (discussed below).
Further work is needed to clarify the relationship be-
tween the neurophysiological and neuropsychological
results.
Later Visual Processing
Later stages of visual processing are believed to occur
in the anterior temporal lobe, based on several con-
siderations including the anterior projections of the
ventral pathway in monkeys (Van Essen et al., 1992;
Gross et al., 1993), single-unit recordings in humans
(Heit et al., 1988) and monkeys (Rolls, 1992), and PET
studies in humans (Sergent et al., 1992; Mclntosh et
al.,
1994). The anterior category-specific potentials
(e.g.,
Fig.
6) may reflect some portion of this later pro-
cessing. The apparent concentration of such locations
near entorhinal cortex suggests that the activity may
be related to mnemonic processes ("this is Bill Clin-
ton's face"). Later processing may also occur in mod-
ules and may be reflected in the potentials following
N200 (e.g., Fig. 3fi,C). These potentials might be
evoked by later feedforward activity in structures that
project to the fusiform/inferior temporal region, for
example, the human homolog of V4 that in the mon-
key provides the main visual input to inferior tempo-
ral cortex (Boussaoud et al.,
1991;
Gross et al., 1993).
Alternatively they might be evoked by feedback activ-
ity from downstream structures (Van Essen et al.,
1992).
Why Are Object Modules So Variable
in Location?
In a given patient we can predict that face, letter-
string, and number modules, if found, will be located
somewhere within the stippled regions of Figure 4.
But as far as we know there is no way of predicting
where within the region any module might be. Why
is this?
Children begin to deal widi written words and
numbers at the age of three or four, and an argument
could be made that letter-string and number modules
Cerebral Cortex Sep/Oct 1994, V 4 N 5 551
faces
cars
scrambled care
butterflies
5O.yV
400
MSEC800
Fgnre 9. Car-specific potential. A, Location of electrode she. B, Posftivrty at
320 msec evoked by cars. Patient BME; Talairach coordinates x27, y-3.
are formed catch-as-catch-can from "leftover" patches
of fusiform/inferior temporal cortex. But such an ar-
gument fails for face modules. Newborns attend to fac-
es more than to nonface stimuli of equal complexity,
and over their first 6 months infants become increas-
ingly skilled at discriminating faces (reviewed by Ca-
rey, 1992; Ellis, 1992). In addition, inferior temporal
neurons of infant monkeys show the same selectivity
for faces as do neurons of adult monkeys (Rodman et
al.,
1993).
Yet despite the availability of regions for the
establishment of anatomically fixed face modules,
such a deterministic process apparent!)' does not oc-
cur.
No individual would have had letter-string or num-
ber modules until very late (a few thousand years ago)
in human evolution, and few individuals had them un-
til a few hundred years
ago,
before the advent of wide-
spread literacy. In preliterate humans what occupied
the patches of cortex now occupied in most of us by
letter-string and number modules? Perhaps they were
occupied by modules for edible plants, or for danger-
ous animals, or more generally for whatever objects
were important for
survival.
Similarly,
it is possible that
category-specific modules develop in adult humans
depending on their interests and occupation. The
dairy fanner who developed a prosopagnosia and was
also unable to identify his cows (Bornstein et al.,
1969) may have had a "cow module" (operationally
defined as a region of fusiform/inferior temporal cor-
tex that generates a cow-specific N200). There is a
surprising degree of plasticity in adult sensory cortex
(reviewed by Merzenich et al., 1988; Kaas, 1991; Gil-
bert, 1993); perhaps the same principles of dynamic
reorganization apply to an even greater extent in ex-
trastriate cortex.
Of course, it is easy to take this line of reasoning to
improbable lengths. One would not want to posit a
module for each of the hundreds of categories of ob-
jects (typewriters, shoes, hands . . .) that we all rec-
ognize. It may be the case that a fairly small number
of object modules is essential for that object category
but also participates in the recognition of other ob-
jects.
This brings us to our final topic.
How Many Object Recognition Systems
Are There?
In my view, It is a vain pursuit to seek to isolate var-
ious subtypes of visual agnosia, e.g., for colours or for
faces.
But surely, the pinnacle of absurdity Is found in
the alleged dichotomy between an agnosia for inani-
mate as opposed to animate objects.
As
lief might one
seek to distinguish an agnosia for knives as against
forks,
which could equally well be claimed from the
case notes of some patients. ...
I
hazard the plea . .
for the abandonment of the term "agnosia" suggested
in 1891 by Freud. Though at the time it might have
been appropriate, subsequent writers have debased
the currency.—Macdonald Critchley
A persistent issue in the study of object recognition
is the structure of the system or systems involved. In
physiological terms, how many categories of objects
are important enough to be processed by dedicated
pathways and cortical regions? On the one hand are
those who would agree with Critchley (1965) that
there exists but one unitary system of object recog-
nition. On the other hand one cannot fail to be im-
pressed by the category-specific deficits (some per-
haps of the knife and fork variety, but others more
compelling) seen in patients with brain lesions. Over
the past 20 years many anatomical and neurophysio-
logical studies in monkeys demonstrate that after ini-
tial analysis in striate (VI) and peristriate (V2) cortex,
specialized processing of visual information is subcon-
tracted out to various regions of extrastriate cortex.
These processing streams can be conceptualized
broadly,
for example, Ungerleider and Mishkin's(1982)
dorsal "where is it?" stream and ventral "what is it?"
stream, or in baroque profusion, for example, the doz-
ens of specialized areas and hundreds of pathways
summarized by Van Essen et al. (1992). Furthermore,
psychophysical studies in humans suggest the exis-
tence of "channels," hypothetical entities having non-
overlapping sensitivities to classes of visual stimuli
and operating more or less independently (reviewed
by Regan, 1989). Thus, in the quarter century since
Critchley's jeremiad the atmosphere has changed. We
are now accustomed to thinking of parallel pathways
acting on specific subsets of sensory information.
Farah's (1990) review of the visual agnosias led her
to the conclusion that only two systems are required
552 Human Extrastrfcite Cortex • Alltson ct al
for object recognition: one is required primarily for
written words and to a lesser extent for other objects,
and another is required primarily for faces and to a
lesser extent for other objects. The letter-string and
face recordings described above support this scheme,
but in addition suggest a third system dealing with
numbers.
If Farah's is the most parsimonious model of object
recognition (given that more than one system is pos-
tulated in the first place), Konorski (1967) by contrast
conceives of nine "gnostic fields" based primarily on
dissociations observed in patients. One of his catego-
ries is faces, consistent with our recordings. Another
category is "signs," which includes words and num-
bers;
our results suggest a subdivision of this category
as noted above. Two of Konorski's categories are "larg-
er partially manipulable objects" including cars, and
"animated objects" including
animals.
We
have record-
ed car-specific potentials in two patients; an example
is shown in Figure 9. These potentials consisted of a
broad, late positivity, and were not recorded from the
stippled region of Figure 4. We have never recorded
an N200 to cars or butterflies despite the fact that
these stimuli were included in all the face and number
recordings and most of the letter-string
recordings.
We
conclude either that "larger partially manipulable ob-
jects"
and "animated objects" do not exist as gnostic
fields (by our operational definition, modules gener-
ating a category-specific N200), or that we have yet
to identify effective objects within these valid fields.
However, the point to be emphasized is that these and
other models of object recognition can be tested em-
pirically with recordings like those reviewed here.
N200 is a recording of postsynaptic potentials in hun-
dreds or thousands of neurons with varying selectivi-
ties,
and in that respect human recordings cannot
achieve the degree of specificity obtainable in single-
unit recordings in monkeys. On the other hand, hu-
man subjects bring to these tasks a tremendous
amount of prior knowledge of objects and language
skills that we hope to exploit to further the under-
standing of object recognition.
Notes
This work
was
supported
by the
Veterans Administration
and
by
NIMH Grant MH-05286.
We
thank J
Claxton.J.
Jasior-
kowski,
and
K. McCarthy
for
assistance. M. Luby kindly
pro-
vided electrode localization.
We
are grateful
to
Drs.
D. D.
and
S. S. Spencer
and the
members
of
the Yale/VA Epilepsy Sur-
gery Program
for
their cooperation
in the
studies reviewed
here.
Correspondence should
be
addressed
to
Dr.
Truett Allison,
Neuropsychology Laboratory
1I6B1,
VA
Medical Center,
West Haven, CT 06516.
References
Allison
T,
Hume AL, Wood CC, Goff WR (1984) Develop-
mental
and
aging changes
in
somatosensory, auditory and
visual evoked potentials. Electroencephalogr Clin Neuro-
physiol 58:14-24.
Allison
T,
Begleiter
A,
McCarthy
G,
Roessler
E,
Nobre
AC,
Spencer DD (1993) Qectrophysiological studies
of
color
processing
in
human visual cortex. Electroencephalogr
Clin Neurophysiol 88:343-355.
Allison
T,
Ginter
H,
McCarthy
G,
Nobre
A,
Puce
A,
Luby
M,
Spencer
DD
(1994) Face recognition
in
human extra-
striate cortex. J Neurophysiol 71:821-825.
Bornstein
B,
Stroka H, Munitz
H
(1969) Prosopagnosia with
animal face agnosia. Cortex 5:164-169.
Boussaoud D, Desimone R, Ungerleider
LG
(1991) Visual
to-
pography
of
area TEO
in the
macaque.) Comp Neurol
306:554-575.
Carey S (1992) Becoming
a
face expert. Philos Trans R
Soc
Lond [Biol] 335:95-103.
Clarke S, Miklossy
J
(1990) Occipital cortex
in
man: orga-
nization
of
callosal connections, related myelo-
and cy-
toarchitecture,
and
putative boundaries
of
functional
vi-
sual
areas.
J
Comp Neurol 298188-214.
Corbetta M, Miezin FM, Dobmeyer S, Shulman GL, Peterson
SE (1991) Selective
and
divided attention during visual
discriminations
of
shape, color,
and
speed: functional
anatomy
by
positron emission tomography.
J
Neurosci
11:2383-2402.
Critchley
M
(1965)
An
appraisal
of
visual (object) agnosia
a tribute
to
Professor Otto Potzl. In: Transactions
of the
8th International Congress on Neurology,
Vol
5
(Konstans
J, ed). Amsterdam: Elsevier. Reprinted
in: The
citadel
of
the senses (Critchley
M,
ed). New York: Raven, 1986.
Damasio AR, Damaslo H (1983) The anatomic basis
of
pure
alexia. Neurology 33:1573-1583.
Damasio
A,
Yamada T, Damasio
D,
Corbett
J,
McKee J (1980)
Central achromatopsia: behavioral, anatomic, and physio-
logic aspects Neurology 30:1064-1071.
Damasio AR, Damasio
H,
Van Hoesen GW (1982) Prosopag-
nosia: anatomic basis
and
behavioral mechanisms
Neu-
rology 32:331-341.
Damasio AR, Tranel D, Damasio
H
(1990) Face agnosia
and
the neural substrates
of
memory. Annu Rev Neurosci
13:
89-109.
Desimone R (1991) Face-selective cells
in the
temporal cor-
tex
of
monkeys. J Cognit Neurosci 3:1-8.
FJlis HD (1992)
The
development
of
face processing skills.
Philos Trans R
Soc
Lond [Biol] 335:105-111.
Farah MJ (1990) Visual agnosia Cambridge,
MA:
MIT Press.
Fried
I,
Katz A, McCarthy
G,
Sass KJ, Williamson P, Spencer
SS,
Spencer
DD
(1991) Functional organization
of hu-
man supplementary motor cortex studied
by
electrical
stimulation. J Neurosci 11:3656-3666.
Geffen G, Bradshaw
JL,
Wallace
G
(1971) Interhemispheric
effects
on
reaction time
to
verbal
and
nonverbal visual
stimuli.
J
Exp
Psychol 87:415-422.
Gilbert CD (1993) Circuitry, architecture, and functional dy-
namics
of
visual cortex. Cereb Cortex 3:373-386.
Gross
CG,
Rocha-Miranda
CE,
Bender
DB
(1972) Visual prop-
erties
of
neurons
in
inferior temporal cortex
of the ma-
caque.
J
Neurophysiol 35:96-111.
Gross CG, Rodman HR, Gochin PM, Colombo
MW
(1993)
Inferior temporal cortex
as a
pattern recognition device.
In: Proceedings
of
the
3d
NEC Conference Research Sym-
posium, Computational learning
and
cognition (Baum
E,
cd),
pp
44-73.
Philadelphia:SIAM.
Harries MH, Perrett
Dl
(1991) Visual processing
of
faces
in
temporal cortex: physiological evidence
for
modular
or-
ganization
and
possible anatomical correlates.
)
Cognit
Neurosci 3:9-24.
Haxby JV, Grady
CL,
Horwitz B, Salerno
J,
Ungcrieider
LG,
Mishkln M, Schapiro MB (1993) Dissociation
of
object
and spatial visual processing pathways
In
human extra-
striate cortex.
In:
Functional organisation
of the
human
visual cortex (Gulyas B, Ottoson D, Roland PE, eds),
pp
329-340. Oxford: Pergamon.
Hcit G, Smith ME, Halgren
E
(1988) Neural encoding
of
in-
dividual words and faces by the human hippocampus and
amygdala. Nature 333:773-775.
Heywood CA, Cowey
A
(1992)
The
role
of the
"face-cell"
area
in the
discrimination
and
recognition
of
faces
by
monkeys. Philos Trans R Soc Lond (Biol] 335:31-38.
Homan
RW,
Herman J, Purdy
P
(1987) Cerebral location
of
international
10-20
system electrode placements. Elec-
troencephalogr Clin Neurophysiol 66:376-382.
Cerebral Cortex Scp/Oct
1994, V 4 N 5 553
Jeffreys DA (1993) The influence of stimulus orientation on
the vertex positive scalp potential evoked by faces. Exp
Brain Res 96:163-172.
Kaas JH (1991) Plasticity of sensory and motor maps in
adult mammals. Annu Rev Neurosci 14:137-167.
Konorski J (1967) Integrative activity of the brain. Chicago:
The University of Chicago.
Leehey S, Carey S, Diamond R, Cahn A (1978) Upright and
inverted faces: the right hemisphere knows the differ-
ence.
Cortex 14:411-419.
Levin HS, Goldstein FC, Spiers PA (1993) Acalculia. In- Clin-
ical neuropsychology (Heilman
KM,
Valenstein
E,
eds), pp
4O9-460. London: Oxford UP
Lueck CJ, Zeki S, Friston KJ, Cope P, Cunningham VJ, Lam-
mertsma
AA,
Kennard C, Frackowiak RSJ (1989) The col-
our centre in the cerebral cortex of man. Nature 340:
386-389.
McCarthy G, Wood CC, Williamson PD, Spencer DD (1989)
Task-dependent field potentials in human hippocampal
formation. J Neurosci 9:4253-4268
McCarthy G, Blamire AM, Rothman DL, Gruetter R, Shulman
RG (1993) Echo-planar magnetic resonance imaging
studies of frontal cortex activation during word genera-
tion in humans. Proc Natl Acad Sci USA 90:4952-4956.
Mclntosh AR, Grady CL, Ungerleider
LG,
Haxby
JV,
Rapoport
SI,
Horwitz B (1994) Network analyse of cortical visual
pathways mapped with
PET.
J
Neurosci 14:655-666.
Merzenich MM, Recanzone
G,
Jenkins WM, Allard TT, Nudo
RJ (1988) Cortical representational plasticity. In: Neuro-
biology of neocortex (Rakic
P,
Singer W, eds), pp 42-67.
Berlin: Wiley.
Michel
F,
Poncet M, Signoret JL (1989) Les lesions respons-
ables de la prosopagnosie sontelles toujours bilateraies?
Rev Neurol (Paris) 146:764-770.
Nobre
AC,
McCarthy G (1994) Language-related field poten-
tials in the anterior-medial temporal
lobe:
effects of word-
type and semantic
priming.
J
Neurosci, in press.
Ojemann
G,
Ojemann J, Lertich E, Berger
M
(1989) Cortical
language localization in left, dominant hemisphere: an
electrical stimulation mapping investigation in 117 pa-
tients.
J
Neurosurg 71:316-326.
Perrett Dl, Mlstlin AJ, Chitty AJ, Smith PAJ, Potter DD, Broen-
nimann R, Harries
M
(1988) Specialized face processing
and hemispheric asymmetry in man and monkey evi-
dence from single unit and reaction time studies. Behav
Brain Res 29:245-258.
Perrett Dl, Hietanen JK, Oram MW, Benson PJ (1992) Orga-
nization and functions of cells responsive to faces in the
temporal cortex. Philos Trans R Soc Lond [Biol] 335:23-
30
Puce A, Nobre
AC,
Allison T, Luby
ML,
McCarthy
KM,
McCar-
thy G (1994) Event-related potential correlates of visual
recognition modules in extrastriate cortex. Paper pre-
sented at the first annual meeting of the Cognitive Neu-
roscience Society, San Francisco.
Regan D (1989) Human brain electrophysiology: evoked po-
tentials and evoked magnetic fields in science and med-
icine New York: Qsevier.
Rhodes G (1993) Configural coding, expertise, and the right
hemisphere advantage for face recognition. Brain Cognit
22:19-41.
Rodman HR, 6 Scalaidhe SR Gross CG (1993) Response
properties of neurons in temporal cortical visual areas of
infant
monkeys.
J
Neurophysiol 70:1115-1136.
Rolls ET (1992) Neurophysiology and functions of the pri-
mate amygdala. In: The amygdala (Aggleton JP, ed), pp
143-165-
New York: Wiley.
Seeck M, Grusser OJ (1992) Category-related components
in visual evoked
potentials:
photographs of faces, persons,
flowers and tools as stimuli. Exp Brain Res 92:338-349.
Sergent J, Ohta S, MacDonald B (1992) Functional neuro-
anatomy of face and object processing: a positron emis-
sion tomography study. Brain 115:15-36.
Talairach J, Tournoux P (1988) Co-planar stereotaxic atlas
of the human brain. New York: Thieme
Tanaka K (1993) Neuronal mechanisms of object recogni-
tion. Science 262:685-688.
Tanaka K, Saito HA, Fukada
Y,
Moriya
M
(1991) Coding the
visual image of objects in the inferior temporal cortex of
the macaque monkey. J Neurophysiol 66:170-189.
Ungerleider LG, Mishkin M (1982) Two cortical visual sys-
tems.
In: Analysis of visual behavior Ongle DJ, Goodale
MA,
Mansfield
RJW,
eds). Cambridge,
MA:
MIT Press.
Van Essen DC, Anderson CH, Felleman DJ (1992) Informa-
tion processing in the primate visual system: an integrat-
ed systems perspective. Science 255:419-423-
WhitelyAM.WarringtonEK (1977) Prosopagnosia:
a
clinical,
psychological, and anatomical study of three patients. J
Neurol Neurosurg Psychiatry 40:395-403-
Wood CC, Wolpaw JR (1982) Scalp distribution of human
auditory evoked potentials. II. Evidence for overlapping
sources and involvement of auditory cortex. FJectroen-
cephalogr Clin Neurophysiol 54:25-38.
Wood
CC,
Spencer
DD,
Allison
T,
McCarthy
G,
Williamson PD,
Goff WR (1988) Localization of human sensorimotor
cortex during surgery by cortical surface recording of
somatosensory evoked potentials. J Neurosurg 68:99-
111.
Yamane S, Kaji S, Kawano K (1988) What facial features
activate face neurons in the inferotemporal cortex? Exp
Brain Res 73-209-214.
Zeki SM (1990) A century of cerebral achromatopsia. Brain
1131721-1777.
554 Human Extrastriate Cortex • Allison et a).
