CorticalActivity Related toAccuracy of Letter Recognition1
A. S. Garrett,*D. L. Flowers,† J . R. Absher,† F. H. Fahey,† H. D. Gage,† J . W. Keyes,†
L. J . Porrino,† and F. B. Wood†
*University of California at Davis, Davis, California 95616; and †WakeForest University School of Medicine,
Winston-Salem, North Carolina 27157-1043
ReceivedApril 6, 1999
Previous imaging and neurophysiological studies
have suggested that the posterior inferior temporal
region participates in tasks requiring the recognition
of objects, including faces, words, and letters; however,
the relationship between accuracy of recognition and
activity in that region has not been systematically
investigated. In this study, positron emission tomogra-
phy was used to estimate glucose metabolism in 60
normal adults performing a computer-generated letter-
recognition task. Both a region of interest and a voxel-
based method of analysis, with subject state and trait
variables statistically controlled, found task accuracy
to be: (1) negatively related to metabolism in the left
ventrolateral inferior temporal occipital cortex (Brod-
mann’s area 37, or ventrolateral BA 37) and (2) posi-
tively related to metabolism in a region of the right
ventrolateral frontal cortex (Brodmann’s areas 47 and
11, or right BA 47/11). L eft ventrolateral BA 37 was
significantly related both to hits and to false alarms,
whereas the right BA 47/11 finding was related only to
false alarms. T he results were taken as supporting an
automaticity mechanism for left ventrolateral BA 37,
whereby task accuracy was associated with automatic
letter recognition and in turn to reduced metabolism
in this extrastriate area. T he right BA 47/11 finding was
interpreted as reflecting a separate component of task
accuracy, associated with selectivity of attention
broadly and with inhibition of erroneous responding
in particular. T he findings are interpreted as support-
ing the need for control of variance due to subject and
task variables, not only in correlational but also in
?2000 Academic Press
The processing of certain visual functions of the
perception of discrete objects is widely recognized tobe
mediated by a ventral visual processing system that
originates in striateandextrastriatevisual cortex itself
and extends through the lower lateral and inferior
occipital–temporal cortical region. Also known as the
‘‘object’’ or ‘‘what is it’’ pathway (Ungerleider and Mish-
kin, 1982), it is distinguished from a dorsal ‘‘location’’or
‘‘whereis it’’pathway coursing upward intotheparietal
region. Successive regions in the ventral pathway are
usually considered to involve progressively elaborated
aspects of visual processing (Mishkin et al., 1983;
Gaffan et al., 1986; Maunsell and Newsome, 1987;
Felleman and Van Essen, 1991; Young, 1992), and
within this putative hierarchy the posterior inferior
temporal portion (Brodmann’s area 37) is selectively
involved in object recognition, as first demonstrated in
monkeys (Schwartz et al., 1983; Desimone et al., 1984,
1991; Tanaka et al., 1991; Perrett et al., 1987).
Studies in human subjects also show a posterior
inferior temporal role in object recognition and object
naming (Ungerleider and Haxby, 1994; Price et al.,
1996; Ungerleider et al., 1998; Moore and Price, 1999).
Functional magnetic resonance imaging (fMRI) and
positron emission tomography studies (PET) of re-
gional cerebral blood flow (rCBF) show activation of BA
37 by tasks involving face (Haxby et al., 1991; Sergent
et al., 1992), visual patterns (Roland and Gulyas, 1995;
Schacter et al., 1995; Kawashima et al., 1998), and
objects viewed from different perspectives (Kosslyn et
al., 1994). From a clinical perspective, intraoperative
electrophysiological recordings in the human extrastri-
ate cortex of epileptic patients carried out by Allison et
al. (1994) have shown that cells in the bilateral inferior
temporo-occipital cortex, in the region of BA 37, re-
spond selectively to the presentation of faces, letter
strings, and numbers. Also, patients with damage to
the occipito-temporal region often are unable to recog-
nizefamiliar faces (Damasioet al., 1982), words (Binder
and Mohr, 1992), or symbolic representations of words
(Soma et al., 1989).
Studies of nonpatient volunteers have demonstrated
that the left BA 37 role in object recognition also
includes language-relevant objects such as words, pho-
netically regular nonwords, and letter strings. For
1This work was supported by National Institute of Child Health
and Development PHS Grant P01 HD 21887.
NeuroImage 11, 111–123 (2000)
doi:10.1006/nimg.1999.0528, availableonlineat http://www.idealibrary.com on
Copyright?2000 by Academic Press
All rights of reproduction in any form reserved.
example, fMRI has shown left BA 37 activation by
generation of rhymes and semantic categories (Shay-
witz et al., 1995) and viewing letter strings (Puce et al.,
1996), and PET studies of rCBF have shown activation
of portions of left BA 37 by visual presentation of words
andnonwords (Petersen et al., 1989, 1990) or by viewing
or reading words (Bookheimer et al., 1995; Buchel et al.,
1998, Moore and Price, 1999). Activation in the ventrolat-
eral portion ofBA 37iscommon tomost ofthesestudies.
While the above studies indicate a role for BA 37 in
object recognition, it is unclear how that role relates to
accuracy of task performance. If the activation is due
solely tothestimuli involved, without regard tosuccess
in recognizing them, then accuracy of performance in
recognizing those stimuli should have no relation to
activation. On the other hand, if activation is depen-
dent on actual object recognition itself, as indexed by
task accuracy, then the activation should be propor-
tional totask accuracy. However, accurateperformance
is often less effortful than inaccurate performance (see
is to any substantial extent proportional to the effort
involved, and if poor performanceindexes greater effort or
difficulty, thentheactivationmight begreater withinaccu-
rate performance than with more accurate performance.
Also the roles of demographic variables, such as age and
gender, or extraneous task-related variables, such as state
anxiety, remain tobeclarified.
The present study addresses the above questions by
using a correlational approach that tests the relation-
ship of ventrolateral BA 37 activity, during letter
recognition, totask performance accuracy, with control
for variancein age, gender, andstateanxiety. Of course,
other regions may be expected tobe involved in various
of these relationships, but we offer no specific a priori
hypotheses about these except totest several appropri-
ate candidate regions of interest. This multivariate
correlational approach requires larger sample sizes in
order toachieve adequate degrees of freedom (approxi-
mately 10 subjects for each variable including global
activation, according to the conservative criterion of
Harris, 1975), so the present study invests scanning
resources across a large sample of N ? 60, using a
single activation condition. This between-subjects ap-
proach, as in a classic study of individual differences,
contrasts with the classic subtraction rationale in which
thescanning resources areinvestedwithin subjects across
METHODS AND MATERIALS
Sixty healthy adults (50% male, 18% non-Caucasian
all of whom were African American) were recruited
from the surrounding communities by advertisement
and word of mouth. The subjects ranged in age from 20
to 66 years (mean 40.6, standard deviation 12.3). The
Briggs and Nebes (1975) modification of the Annett
Handedness Inventory classified 51 of the 60 subjects
as strongly preferring the right hand, 3 as strongly
preferring the left hand, and 6 as not strongly prefer-
ring either hand. Table 1 provides further descriptive
information of thesample.
Subjects were included if they had no history of
neurological disease, head injury, diabetes or other
metabolic disease, heart disease, drug or alcohol abuse,
seizures, liver disease, or glaucoma or current or recent
use of psychoactive or metabolically relevant medica-
tions. Subjects were also excluded if their urine drug
screen on the day of scanning showed any evidence of
illegal or centrally acting (e.g., antihistimine) drug use.
All subjects included in the sample had the absence of
spike discharges or observable slow wave activity on
resting EEG, had normal MRI of the brain, and had
normal fasting blood glucose levels. By the criteria of
the Schedule for Affective Disorders and Schizophre-
nia—Lifetime Version (Endicott and Spitzer, 1978),
subjects were excluded for histories of bipolar affective
disorder or schizophrenia or for current unipolar affec-
tive disorder. Subjects were selected without reference
toreading ability or Wechsler Adult Intelligence Test—
Revised subtests (Wechsler, 1981), and their scores on
single-word reading (Letter–Word Identification Sub-
test of the Woodcock J ohnson—Revised; Woodcock and
J ohnson, 1989) and selected Wechsler subtests were
typical of normal adults, as shown in Table1.
All subjects were scanned in the morning and were
instructed toabstain from nicotineand caffeinefor 24 h
prior to the study and from food or drink except water
after midnight on the day of the study. Subjects were
paid for their participation, the study was approved by
TABL E 1
Demographic and Task Performance Characteristics
Mean Minimum Maximum
Percentage false alarms
Number of trials
WAIS-R block design
WAIS-R digit span
WJ -R Word ID
Note. STAI-S, State–Trait Anxiety Inventory of Spielberger;
WAIS-R, Wechsler Adult Intelligence Scale, Revised; WJ -R Word ID,
Woodcock J ohnson—Revised Letter–Word Identification Task.
GARRETT ET AL.
themedical school’s Institutional Review Board, andall
subjects gavetheir written consent.
Consistent positioning in the PET and MRI scanners
was facilitated by individually molded thermosetting
plastic masks along with fiducial markers on themask.
Prior to FDG uptake, subjects practiced a computer-
generated letter-recognition task (described below) un-
til confident of their performanceof thetask and until a
criterion scorewas achieved (greater than 75% hits and
less than 25% false alarms during the trial period).
Subjects wore glasses if needed, and none complained
of problems seeing the stimuli. State anxiety (Spiel-
berger et al., 1983) was measured just prior to the 10
mCi FDG bolus injection, which was delivered into the
right arm antecubital vein, through an intravenous (iv)
line previously placed but immediately removed after
injection. Blood samples were withdrawn every 15 s
over the first 2 min and at minutes 3, 4, 6, 8, 13, 18, 28,
and 38 and following scanning, through a previously
placed indwelling iv in the left arm, which had been
heated to 110°F. The subject performed the computer
task for 35 min, voided, and returned to the PET
scanner for acquisition of theemission scan.
Letter Task during FDG Uptake
Stimuli for the letter-recognition task included 12
letters (6 uppercase and 6 lowercase; for example, K, V,
h, and c) and 12 unfamiliar characters (for example, \,
_0, ¥, and ?) presented in either black or white inside a
magenta box at the center of a computer monitor. The
target was any of the 12 letters regardless of color,
nontarget characters were any of the 12 unfamiliar
characters. Probability of target versus nontarget items
was 50%. The stimuli subtended a 0.7° visual angle.
Each stimulus was flashed for 50 ms. A black dot
fixation point was displayed continuously in the center
of thescreen within themagenta box.
The subject controlled the pace of the game by
pressing and holding down the mouse button tostart a
trial and lifting the right finger from the mouse button
as quickly as possible when a target stimulus was
presented. Subjects again depressed the mouse button
to start the next trial. If no finger lift response was
made, and the button remained down, then the next
stimulus was presented at a random 1.5 to 2.0 s after
The computer provided auditory and visual feedback
for both correct and incorrect responses. The differen-
tial auditory feedback was a short, higher pitched tone
or ‘‘beep’’ for correct and a moderately lower pitched
tone or ‘‘boop’’ for incorrect. Tones were of equal inten-
sity. Visually, several feedbacks were given, as follows.
A correct response caused two dots, one below and one
above the fixation point, to move by a small increment
toward the fixation point. When five correct responses
wereaccumulated, then thedots ended up closeoutside
the magenta box in middle field and a message was
printed on the screen informing the subject of the
award of bonus points for that group of correct re-
sponses. The dots were then reset. Incorrect responses
caused a printed statement to appear notifying the
subject of the mistake, whether miss or false alarm.
Finally, after every 17 trials, a rest message was
printed that informed the subject of a brief (3 s) rest
before resuming the task. A critical reaction time was
set at 900 ms (i.e., target relevant responses after this
time were recorded as misses). Percentages of hits and
false alarms were used to calculate d-prime, the mea-
sure of signal detection accuracy (Green and Swets,
A 1.5-T GE Signa MRI scanner (GE Medical Systems,
Milwaukee, WI) was used to acquire axial 3D spoiled
gradient-echo T1-weighted images (2.5-mm thickness
with no interslice gap, TR ? 45 ms, TE ? 5 ms, flip
angle ? 45°, NEX ? 1). This anatomic image was used
tolocalize regions of interest (ROIs) on the PET image.
All scans in this study were performed using a
Siemens/CTI 951/31 ECAT scanner which has a resolu-
tion of approximately 6.0 mm in all axes. The attenua-
tion-corrected emission data were reconstructed with
filtered back-projection using a Hann filter with a 0.4
Quantitative images of glucose metabolic rate were
generated by applying the method of Phelps et al.
(1979), and the reconstructed PET images were regis-
tered by trained investigators to their respective MRI
images using the ‘‘Register’’ program (Neelin et al.,
1993), validated by Woods (1996). The MRI was placed
in stereotaxic space (Talairach and Tournoux, 1988)
using an algorithm developed by Louis Collins of
Montreal Neurological Institute (Collins et al., 1994)
and the PET data were resampled tocorrespond tothe
PET Regionsof Interest Analysis
Gemini, a locally developed software package, simul-
taneously displays the MRI and its coregistered PET
study, allowing any point to be simultaneously viewed
in three orthogonal planes (transverse, sagittal, and
coronal). It alsoallows theplacement of a spherical ROI
to be guided by viewing all three planes simulta-
MRI segmentation to isolate the functionally rel-
evant (Kadekaroet al., 1985; J ulianoet al., 1981, 1983)
CORTICAL ACTIVITY RELATED TO TASK ACCURACY
gray matter signal intensity boundaries was accom-
plished by placing small spherical ROIs, of diameter
0.67 cm, bilaterally in the genu of the corpus callosum
(representing white matter), the head of the caudate
nucleus (representing gray matter), and the anterior
cerebral ventricles (representing CSF). Relative homo-
geneity of tissue type was achieved by accepting only
spheres whose distribution of pixel values had a coeffi-
cient of variation (standard deviation as percentage of
the mean) not exceeding 5% for gray and white matter
or 12% for cerebrospinal fluid. The average of the
median pixel values in the two spheres for each tissue
type was then taken as the criterion pixel intensity for
that tissue. Midpoints between the gray and the white
and between the gray and the CSF tissue criteria then
defined the thresholds for segmenting the MRI pixels
into tissue types. Gemini then allowed into the analy-
ses only the PET voxels whose MRI intensities were
within the gray matter thresholds. The interscorer
reliability (A.S.G. and D.L.F.) for both upper and lower
gray matter thresholds, calculated as described for 15
randomly selected MRIs, was 0.99.
Three-dimensional ROI spheres, chosen a priori to
represent candidate areas of general interest, were
located on the MRI image, according torules presented
in Table 2, which also shows their average location in
Talairach space(for a morecompletedescription of ROI
location rules, see Wood and Flowers, 1999). Separate
TABL E 2
Regions of Interest
Region of interest
Size of spherical ROI and instruction
tolocate the ROI
(x, y, z)
Calcarine fissure 2.0 cm. Locate the calcarine fissure on the sagittal view nearest midline. Posi-
tion the sphere entirely within the occipital pole, its center in the calcarine
fissure. On coronal view, move the center laterally in the calcarine toa point
1 cm from midline.
2.5 cm. Find the most lateral sagittal plane passing through the inferior tem-
poral sulcus and the preoccipital incisura (occipital notch), which usually
T-intersects the occipitotemporal sulcus on the ventral surface. Locate that
intersection by paging through the axial planes in this region. Center the
sphere at this intersection, moving it vertically toensure it is wholly within
2.0 cm. On saggital views, locate the superior temporal sulcus (STS). Page
medially toits termination or that of its superior (angular) branch and
verify on the axial view. Page down toan axial plane that is 1 cm below the
STS terminus, and center the sphere in the STS. Page down farther if nec-
essary toplace sphere wholly in the cerebrum.
2.5 cm. Observe the coronal view tangent tothe genu of the corpus callosum; it
displays the superior, middle (when present), and inferior frontal sulci
(IFS). The IFS is easily located as a nearly horizontal feature and can be
verified from the sagittal view. Center the sphere in the IFS on this plane,
moving it medially until sphere is wholly within the cerebrum.
2.5 cm. Beginning from a low axial view, page upward until the orbital frontal
bone is visualized around the eyes. There observe a coronal plane through
both orbits and locate the most medial ventral gyrus, the gyrus rectus. The
next gyrus laterally will be the orbital frontal gyrus (OFG). Center the
sphere in the gyrus sothat the edges of the sphere include both sulci sur-
rounding the gyrus. Adjust it vertically toavoid ocular muscle. The ROI
often includes other short sulci.
2.0 cm. Select the axial slice showing the head of the caudate at its widest.
Center the sphere on this plane and adjust from all views, sothe sphere
does not infringe on the putamen. It may extend intothe lateral ventricle.
3.0 cm. Using all views, position the sphere within the thalamus, tangent to
midline and excluding all adjacent basal ganglia structures. The sphere
may include portions of the ventricles.
2.5 cm. Center the sphere midway between the hemispheres, on the line of
intersection of the axial and coronal planes tangent tothe midline dorsal
and anterior surfaces of the callosum.
Left: ?10, ?82, 8
Right: 11, ?82, 8
Brodmann area 37
Left: ?46, ?68, ?11
Right: 48, ?63, ?12
Left: ?46, ?58, 41
Right: 47, ?53, 42
Left: ?40, 35, 12
Right: 44, 35, 12
Orbital frontalLeft: ?30, 40, ?9
Right: 32, 39, ?9
Dorsal caudateLeft: ?10, 15, 4
Right: 11, 15, 4
ThalamusLeft: ?11, ?16, 8
Right: 11, ?16, 8
Anterior cingulate(Mid:) 0, 33, 37
Note. Anatomical description for positioning on high-resolution MRI, sizes, and average locations of their centers in the stereotaxic space of
Talairach and Tournoux (1988); x, sagittal plane (negative indicates left hemisphere); y, coronal plane (negative indicates posterior to a
vertical plane through the anterior commissure); z, axial plane (negative indicates inferior to a horizontal plane through the AC-PC line);
GARRETT ET AL.
observers (A.S.G. vs D.L.F.) had especially close agree-
ment (between 0 and 3 mm difference in any plane) for
all regions except theangular gyri (between 4.5 and 6.0
mm absolute difference in the x and y planes) and the
orbital frontal gyri (between 4.5 and 6.0 mm absolute
difference in the y plane). Placement of the ventrolat-
eral temporal occipital BA 37 sphere is shown in Fig. 1.
Histograms of voxel intensity on the PET scan were
taken for each spherical ROI, and the 95th percentile
was taken as the glucose metabolic value for each ROI,
consistent with our prior phantom studies showing it to
be the most accurate and reliable measure of metabolic
intensity in a region of known intensity (Fahey et al.,
1998). Similar high-percentile methods have been used
with FDG PET data (Moeller et al., 1987) and15O PET
data (Raichle et al., 1994). Reliability coefficients for
ROI values calculated as described were between r ?
0.87 and r ? 0.99 except for the right inferior frontal
(r ? 0.75) and right BA 37 (r ? 0.81).
Data were analyzed in two ways, both employing
general linear models (GLM) to predict glucose meta-
bolic rate from task accuracy, controlling for individual
differences variables in age, gender, and state anxiety
and for whole brain metabolism (Moeller and Strother,
1991; Friston et al., 1990, 1991). The first method
employed the ROI calculations described in detail
above using Statistical Analysis Software (SAS Insti-
tute, Carey, NC) toconstruct thegeneral linear models.
The second method, voxel-based Statistical Parametric
F IG. 1.
PET image. Images areoriented by radiological convention (left is right).
A three-dimensional ROI sphere(2.75 cm in diameter) encloses left Brodmann’s area 37 on an MRI (on theleft) and a coregistered
CORTICAL ACTIVITY RELATED TO TASK ACCURACY
Mapping (SPM), was carried out using SPM96 software
(Friston et al., 1991, 1994, 1995; Worsley et al., 1992),
after conversion to ANALYZE format (Mayo Clinic,
Rochester, Minnesota). Conversion to standard space
(Talairach and Tournoux, 1988) was done using a
locally constructed templatethat is theaverageof 40 of
the 60 normal glucose metabolized PET brains. The
normalized PE T images were then conventionally
smoothed using a 13-mm (FWHM) Gaussian kernel.
In the SPM multisubject, single-condition design,
subject and covariate effects were estimated according
to the general linear model at every voxel exceeding
80% of themean global gray matter threshold.A height
threshold of P ? 0.001, uncorrected, was set toidentify
clusters. Inasmuch as the ROI analyses had already
constrained the total volume to prespecified areas, a
small volume correction was employed (Worsley, 1996).
General Linear Modelsof Regional Metabolism
Pearson correlations between task variables and
subject variables arelisted in Table3.
The formal GLM predicted metabolism for each ROI
from d-prime(task accuracy) age, gender, stateanxiety,
andwholebrain averagemetabolism. Table4 lists theF
values and significance levels for each variable in each
A significant F valueindicates uniquecontribution of
that variabletotheregional metabolic-dependent mea-
sure after all other sources of variance in the model
havebeen taken intoaccount (TypeIII sums ofsquares).
Even after statistical control, d-prime contributed sig-
nificantly—and inversely—to the variance in left lat-
eral BA 37 metabolism (P ? 0.0005). D-prime also
negatively predicted left angular gyrus metabolism
while it positively predicted left thalamus and right
orbital frontal metabolism (all at P ? 0.05). No other
ROI was significantly related to task accuracy. As a
special test for possible confounding effects of handed-
ness, rate of stimulus presentation, and number of
printed error messages delivered to the subjects, we
alsoproduced subsequent GLMs for these four regions,
adding each of the potential confounding variables
separately. These confirmed that the addition of hand-
edness or stimulus presentation ratehad littleeffect on
the relationship of d-prime to regional metabolism.
Adding the number of printed error messages as a
statistical covariate control did not change the area 37
finding, but did abolish, i.e., explain, the right orbital
frontal (BA 47/11) relationship to d-prime. Table 5
summarizes these effects on the d-prime affect in the
basic general linear model.
The calculation of glucose metabolic rate is exponen-
tially weighted from the moment of injection and is
therefore dominated by the earlier minutes of the
uptake period. We addressed this issue by examining
data from the 37 subjects in the experiment on whom
we had individual trial data. These subjects did not
differ from the total 60 in any demographic or task
variables or in the regional metabolic findings. Within
the 37 subjects, d-prime for the first 40 or 80 trials was
significantly related(r ? 0.64, P ? 0.0001 andr ? 0.77,
TABL E 3
Pearson Product-Moment Correlations between Task
Performance Variables and Individual Difference Variables
Task variableAge Gender
% False alarms
(items per minute)
Note. Gender is coded as male, 1; female, 2. STAI-S, State–Trait
Anxiety Inventory of Spielberger. ‘‘?’’ indicates an inverse correla-
tion. Nonsignificant correlations with task variables areomitted.
* P ? 0.05.
** P ? 0.005.
*** P ? 0.0001.
TABL E 4
F Values (Type III Sums of Squares) for Linear Regres-
sion Models of ROI Metabolism Predicted by Individual
Individual difference variables included
in each general linear model
L BA 37
R BA 37
L inf frontal
R inf frontal
L angular G
R angular G
Note. ‘‘?’’indicates an inverserelation. L, left; R, right; Inf, inferior;
G, gyrus;Ant cing, anterior cingulate.
P ? 0.05 not shown.
* P ? 0.05.
** P ? 0.005.
*** P ? 0.0005.
GARRETT ET AL.
P ? 0.0001, respectively) to d-prime over the entire
task period. More importantly, all relations to task
accuracy were the same if the d-prime accuracy score
was calculated only from the first 40 or 80 trials (on
average, the first 2.5 or 5 min). There was a tendency
for d-prime toimprove over trials, but measures of this
improvement showed norelation tothe metabolic find-
Table 6 summarizes the correlations between ROI-
calculated glucose metabolism and components of the
task performance calculations, including percentage of
hits and false alarms, after partialling for age, gender,
anxiety, and global metabolism.
Scatter plots of left ventrolateral BA 37 and BA 47/11
metabolism as a function of d-prime task accuracy are
shown in Figs. 2 and 3, respectively. Values shown are
studentized residuals (z scores) after partialling for
total metabolism, age, gender, and stateanxiety.
Statistical Parametric Mapping Result
With the same covariance for age, gender, state
anxiety, and global metabolic rate, the SPM ANCOVAs
(Friston et al., 1990) corroborated the ROI findings for
left ventrolateral BA 37 and right orbital BA 47/11
(z ? 3.32, P ? 0.05 and z ? 3.73, P ? 0.005, respec-
tively). These results are summarized in Table 7. Both
the positive and the negative contrasts are shown in
the ‘‘glass brain’’ renderings in Fig. 4. Notably, the left
angular gyrus and thalamus ROI findings were not
duplicated by SPM.
ROI a posteriori testing was carried out by placing
2-cm spheres centered on the maximum voxels of the
two additional clusters in the right frontal lobe identi-
fied with SPM. The center of the a priori selected
inferior frontal 3.0-cm ROI was separated from the two
a posteriori SPM maximumvoxels within BA 47 andBA
11 by 2.6 and 2.8 cm in the axial direction, respectively.
Neither a posteriori ROI reachedsignificance(P ? 0.10)
in its prediction of task accuracy.
As a final step, a cross correlation was carried out in
SPM to determine the shared variance between left
lateral BA 37 and all other voxels. The opposing
correlations between d-prime and left lateral BA 37
(negative) and between d-prime and right orbital BA
47/11(positive) naturally predicteda reciprocal relation-
ship between these two regions, but the calculation
served to determine its strength and it also explored
thepossibility of other statistical connectivity toBA 37.
For consistency, the calculation covaried for age, gen-
der, stateanxiety, andmean global metabolicrate. Only
the expected negative correlation between the two
regions was found (F ? 3.42; P ? 0.01, corrected for
small volumeof a 2.0-cm sphere).
In brief, both the ROI and the voxel-based mapping
(SPM) methods converged to demonstrate an inverse
relationship between left inferior posterior temporal
occipital cortex (left ventrolateral BA 37) and task
accuracy as measured by d-prime and separately by
hits andfalsealarms. Therelationshipwas statistically
independent of age, gender, state anxiety, handedness,
rateofsubject-pacedstimulus presentation, andprinted
feedback during the task. Right orbital frontal cortex
(BA 47/11) activity was also positively related to task
accuracy, but only to false alarms and not hits. The
relationship was independent of age, gender, state
anxiety, handedness, and stimulus presentation rate,
TABL E 5
F (P) Values for the Independent Contribution to Variance
in Region of Interest Activation by Task Accuracy (D-Prime)
After Systematically Adding Rate of Stimulus Presentation,
Handedness, or Number of Feedback Messages to the Basic
General Linear Model (Predicting Regional Metabolism from
D-Prime with Age, Gender, StateAnxiety, and Global Activa-
tion Included as Covariates)a
Region of interest
Left angular gyrus
13.99 (0.0005) 14.60 (0.0004) 12.93 (0.0007)
5.09 (0.0282)4.65 (0.0356)
4.01 (0.0504) 4.27 (0.0438)
6.10 (0.0168) 5.20 (0.0267)
aF (P) in the basic model for d-prime prediction of BA 37 is 14.61
(P ? 0.0003) and for BA 47/11 is 4.80 (P ? 0.0329).
TABL E 6
Pearson Product-Moment Correlations between Task Per-
formance Variables and Regional Glucose Metabolism in
Regions of Interest, Controlling for Age, Sex, State Anxiety,
and Global Metabolism
Region of interest
% False alarms
Note. ‘‘?’’ indicates an inversecorrelation.
* P ? 0.05.
** P ? 0.005.
*** P ? 0.0005.
All other correlations areP ? 0.05.
CORTICAL ACTIVITY RELATED TO TASK ACCURACY
but was abolished (explained) by including feedback in
the statistical model. SPM analysis alsofound statisti-
cal connectivity between left ventrolateral BA 37 and
activation in three right frontal regions. The most
ventral of these corresponded to the orbital frontal
location. The other two did not show a relationship to
The inverse relationship between task performance
and left ventrolateral BA 37 metabolism plausibly
suggests that inefficient, or at least less automatized,
task performance is more metabolically demanding
(either in terms of magnitude or spatial extent) of the
posterior neural mechanisms than efficient, automatic
task performance. The demonstration of inverse rela-
tionships between regional activation and task perfor-
manceis not novel (Wood et al., 1980; Haier et al., 1992;
J enkins et al., 1994; Schlaug et al., 1994; Grady et al.,
1996; see also the review by Wood, 1990). That the
finding survives all relevant statistical controls sug-
gests that it must relate to task performance specifi-
cally. Descriptively, at least, an automaticity ex-
planation whereby better task performance is less
F IG. 2.
studentized residuals (z scores) after partialling for total metabolism, age, gender, and stateanxiety, as provided in theGLMs in thetext.
Scatter plot and regression line of left ventrolateral BA 37 metabolism over d-prime task accuracy. Values are plotted as
F IG. 3.
studentized residuals (z scores) after partially for total metabolism, age, gender, and stateanxiety, as provided in theGLMs in thetext.
Scatter plot and regression line of right orbitofrontal BA 47/11 metabolism over d-prime task accuracy. Values are plotted as
GARRETT ET AL.
effortful and less metabolically demanding, is difficult
Of course, ventrolateral BA 37 is not the only ventral
extrastriate region to be implicated in lexical process-
ing and discrimination. For example, Petersen et al.
(1990) showed both lateral and medial extrastriate
activation by words and nonwords compared tofixation
control, but the effect disappeared in the Howard et al.
study (1992) when the control task was unpronounce-
able letter strings or false fonts. In contrast, activation
of the ventrolateral posterior temporo-occipital visual
processing stream, near or including our ventrolateral
BA 37 ROI, does survivea number of different compari-
son conditions, as in simple viewing of real words vs
viewing of face or texture controls (Puce et al., 1996) or
overt discriminations such as real words vs letter
strings (Buchel et al., 1998).Activation in this area also
appears to involve tasks in which silent or overt
naming of visual objects is required (Bookheimer et al.,
1995; Price et al., 1996; Moore and Price, 1999). To be
sure, objects other than lexical ones also activate this
region—e.g., pattern matching vs pattern scanning
(Kawashima et al., 1998). Consistent with the above,
then, theventrolateral BA 37 activity weobservein our
study isplausibly relevant todiscrimination task perfor-
mance, moresothan medial extrastriateactivity would
be. Furthermore, the fact that ventrolateral BA 37
activation is inversetotask performanceis explainable
by the naming data reviewed above (Price et al., 1996;
MooreandPrice, 1999): if subjects adopteda strategy of
even subvocal naming of thestimuli (Bookheimer et al.,
1995), and if that strategy is more often adopted by
individuals having more difficulty with the task, then
the excess ventrolateral BA 37 activity with poor
The role of the right orbital cortex in this letter
recognition task was not explicitly predicted. Neverthe-
less, right frontal lateral and orbital activity has been
positively related topercentage of correct responses on
a face recognition task (Grady et al., 1996), to perfor-
mance on a Stroop task (Bench et al., 1993), to task
performance feedback (Elliot et al., 1997), and to the
number of successive correct responses while shifting
attention between color and shape (Nagahama et al.,
TABL E 7
Stereotaxic Location and Confidence Levels for Maximum
Voxels Within Clusters Identified by SPM in Predicted Loca-
tions (P ? 0.001 for Height, Uncorrected, with Small Volume
x, y, z
3.7340, 46, ?14 47/11** Right orbital frontal
?46, ?58, ?12 37* Left temporal occipital
Note. Gray matter threshold at 0.80 of mean global metabolic rate.
ANCOVA normalization for mean global metabolism; confounding
covariates age, gender, and stateanxiety.
* P ? 0.05.
** P ? 0.005.
F IG. 4.
normal adults during a letter-identification task with age, gender, and state anxiety covaried. Voxel values are normalized by global mean
metabolism. The orthogonal set on the left shows higher glucose metabolic rate with better performance (d-prime), and the set on the right
shows lower glucosemetabolic raterelativetobetter performance.
Statistical parametric maps (thresholded at P ? 0.001, uncorrected) predicting normalized glucose metabolic rate in N ? 60
CORTICAL ACTIVITY RELATED TO TASK ACCURACY
1998). Of special note is the fact that the first two of
these studies showed reciprocal decreases in other
regions, including extrastriatevisual cortex.
Reciprocities of activation (some regions increasing
while others decrease) are elsewhere familiar in the
literature. For example, reciprocal changes in activa-
tion with diminishing difficulty over trials of practice
on motor tasks have been demonstrated (Grafton et al.,
1992, 1994; Schlaug et al., 1994; J enkins et al., 1994).
Other studies involving identification of object or word
stimuli have suggested an inverse response between
left posterior cortical and frontal areas. Howard et al.
(1992) found a reduction in right inferior prefrontal
cortex activation (and in other right hemisphere areas)
and increased activation of the left posterior middle
temporal gyrus during a single-word reading task
when the control task was viewing false font strings
while repeating a single word (thus controlling for
processing of visual patterns, articulation, and audi-
tory feedback of subject’s own voice). Buckner et al.
(1996) alsoreported opposing anterior versus posterior
activation, but in the reverse direction with increased
right frontal and decreased bilateral parietal activity,
when subjects performedan auditory semanticmemory
task wherein they recalled practiced word–word or
picture–word pairs when the control task was the
simplerepetition of words.
Another example of reciprocity—demonstrating neu-
ral system responses to both increases and decreases
in difficulty—occurs in a cleverly designed study by
Raichle et al. (1994), who examined cortical activation
during verb generation to visually presented nouns
(contrasted with simply reading a string of nouns).
They foundincreasedactivation in left posterior tempo-
ral cortices (including BA 37) and other regions (ante-
rior cingulate, left prefrontal, andright cerebellum) but
concomitant decreased activation in bilateral Sylvian
insular cortex and the left medial extrastriate cortex.
After 15 min of practice, however, activity in these
regions reversed while task performance, as measured
by reaction time, decreased simultaneously. Introduc-
ing a novel list of nouns returned most of these regions
(and reaction time) to the ‘‘naive’’ state, leading these
investigators to conclude that separate circuits serve
more effortful and automatic response choice selection
in a word generation task when visual stimuli areused.
Notably, however, themethod of averaging scans across
subjects did not allow for a direct analysis of the effect
of individual task accuracy on focal neuronal activity.
A relationship similar to that of the present study
was reported by Grady et al. (1996) who found that,
compared toa sensorimotor control task (viewing noise
patterns), the task of matching progressively degraded
faces was related to progressive increases in right
frontal BA 9/46 and concomitant decreases in medial
striate and bilateral fusiform (BA 19/37) regions. Accu-
rate judgments during this self-paced task correspond-
ingly declined, as might be expected. It is tempting to
infer a relationship between task accuracy and specific
cortical changes; however, unexpectedly—and consis-
tent with the present findings—a direct correlation
between task performance and cortical blood flow dur-
ing the high degradation condition revealed positive
correlations between accuracy and a right frontal re-
gion (BA 45) and negative correlations between accu-
racy and left prefrontal cortex and striate cortex.
Although the present study does not systematically
manipulate quality of the stimuli, targets are made
challenging to detect by their 50-ms duration and by
restricting response time, whereas the Grady study did
not constrain stimulus duration or response time.
Thus, both studies point totheinterpretativeerror that
could be made if the relationship between strength of
performance and regional activity is not specifically
examined. Of course, forces driving these similar find-
ings may not bethesame.
Even though working memory was not explictly
elicited by the present task, if a memory strategy were
to be adopted by the subject—such as maintaining a
representation of ‘‘letters’’ or ‘‘nonletter characters’’—
performance might be enhanced. Consider the findings
of McIntosh et al. (1996) that as retention delay was
increased on a delayed match-to-sample face percep-
tion task there was a decrease in striate and ventral
extrastriate area activity as measured by cerebral
blood flow, interpreted as reflecting transient visual
perceptual processes, while right prefrontal activity
increased, interpreted as a greater reliance in the more
sustained task of holding the object’s icon in working
memory. Right frontal BA 47/11 activation has also
been reported in response to delayed match-to-sample
tasks involving face matching (Haxby et al., 1994,
1995). Thus, if thesubjects in thepresent study differed
in their relianceon a strategy that minimized theeffort
in earlier sensory processing and maximized the effort
in maintaining a working memory representation of
the target stimuli, and if that strategy engendered
better performance, then thereciprocity ofleft extrastri-
ate and right orbital activation could be explained.
Against this explanation, however, is the fact that the
degree of a subject’s improvement in performance over
trials—which would be expected to reflect an increas-
ingly available working memory of the target stimuli—
was not related to localized brain activity in our study.
Perhaps a more satisfying interpretation of the pre-
sent results relates totask demands, which in this case
require subjects to withhold responses to particular
stimuli as well as to emit responses to target stimuli.
Notably, in the present study orbital metabolism is
inversely related to the percentage of false alarms but
not significantly related to percentage of hits. In con-
trast, BA 37 is relatedtoboth hits andfalsealarms. The
GARRETT ET AL.
orbital frontal component, therefore, may be particu-
larly related to the allocation of resources to inhibit
impulsive responses, whereas the ventrolateral BA 37
component may be more directly related tobasic target
vs nontarget discriminations. That would also explain
why statistical control for the number of printed feed-
back messages, which weremorehighly correlatedwith
variancein errors than in hits, would abolish or explain
theright orbital frontal finding.
Consistent with our a priori decision tointerpret only
those sites corroborated both by ROI and by SPM, we
have declined to interpret the angular gyrus and
thalamic findings, which should then be the object of
Our technique of validating ROI methods by SPM is
similar, albeit directionally reversed, tothat of Koeppet
al. (1997). They propose that, because of SPM’s moder-
ate liability to partial volume effects, the conservative
best approach would be to use SPM to generate candi-
date maps and then test them by an ROI method. In
this case, we had a priori candidate regions, toconfirm
Finally, the role of stimulus familiarity in object
recognition is not addressed by this study. Hits and
misses reflect responses to letters, while false alarms
and correct rejections are responses to characters.
However, since left ventrolateral BA 37 metabolism is
greater for any correct decision, whether hits or correct
rejections, then familiarity is not confounded with the
decision process itself. However, this explanation must
be tested further. The length of the glucose uptake
period (35 min) may allow the subject to gain enough
experience with the foil characters to consider them
familiar; however, noposttest of this was administered.
The present study reduces unexplained variance by
accounting for age, gender, and state anxiety. This is
variance that would otherwise be treated as error
variance, so some inconsistencies across studies in the
literature may be explained by these sources of vari-
ance. On the other hand, despite these controls, there
remains a significant proportion of unexplained vari-
ance in localized activation, in the present study and
other studies. Futurestrategies for exploring this ques-
tion should include both a correlational approach that
addresses other demographic or state variables and
also a subtraction approach that uses appropriate
baseline controls. Both strategies remain important,
since the subtraction strategy would still be subject to
confounds due to such factors as accuracy and state
anxiety, which could vary not only between subjects but
The present study, therefore, suggests the following
methodological considerations. Individual task perfor-
mance measures should be considered in interpreting
cortical activation. Subject samples need to be of suffi-
cient size to allow for statistical control of subject
variables. Interactions among task variables, subject
variables, and localized brain activity are not only
possible but should be expected. Finally, inverse corre-
lations and reciprocities of localized brain activity
should not be overlooked, and data analytic strategies
should routinely examinefor them.
Theauthors thank Cathy Eades for her invaluablehelp in develop-
ing and refining the Gemini software used to measure the PET data
and alsoDennis Vickland, NataliePrice, and Beth Harkness for their
Allison, T., McCarthy, G., Nobre, A., Puce, A., and Belger, A. 1994.
Human extrastriate visual cortex and the perception of faces,
words, numbers, and colors. Cereb. Cortex 4:544–554.
Bench, C. J ., Frith, C. D., Grasby, P. M., Frackowiak, R. S., andDolan,
R. J . 1993. Investigations of the functional anatomy of attention
using theStroop test. Neuropsychologia 31:907–922.
Binder, J . R., and Mohr, J . P. 1992. Thetopography of callosal reading
pathways. Brain 115:1807–1826.
Bookheimer, S. Y., Zeffiro, T. A., Blaxton, T., Gaillard, W., and
Theodore, W. 1995. Regional cerebral blood flow during object
naming and word reading. Hum. Brain Mapp. 3:93–106.
Briggs, G. G., and Nebes, R. D. 1975. Patterns of hand preferencein a
student population. Cortex 11:230–238.
Buckner, R. L., Raichle, M. E., Meizin, F. M., andPetersen, S. E. 1996.
Functional anatomic studies of memory retrieval for auditory
words and visual pictures. J . Neurosci. 16:6219–6235.
Buchel, C., Price, C., and Kriston, K. 1998. A multimodal language
region in theventral visual pathway. Nature394:274–277.
Collins, D. L., Neelin, P., Peters, T. M., and Evans, A. C. 1994.
Automatic 3D intersubject registration of MR volumetric data in
standardizedTalairach space. J . Comput. Assisted Tomogr. 18:192–
Courtney, S. M., Ungerleider, L. G., Keil, K., and Haxby, J . V. 1996.
Object andspatial visual working memory activateseparateneural
systems in human cortex. Cereb. Cortex 6:39–49.
Damasio, A. R., Damasio, H., and Van Hoesen, G. W. 1982. Prosopag-
nosia: Anatomic basis and behavioral mechanisms. Neurology
Desimone, R. 1991. Face-selective cells in the temporal cortex of
Desimone, R., Albright, T. D., Gross, C. G., and Bruce, C. 1984.
Stimulus selective properties of inferior temporal neurons in the
macaque. J . Neurosci. 4:2051–2062.
Elliot, R., and Dolan, R. J . 1998. Activation of different anterior
cingulate foci in association with hypothesis testing and response
Elliot, R., Frith, C. D., and Dolan, R. J . 1997. Differential neural
response to positive and negative feedback in planning and guess-
ing tasks. Neuropsychologia 35:1395–404.
Endicott, J ., and Spitzer, R. L. 1978. A diagnostic interview: The
Schedule for Affective Disorders and Schizophrenia. Arch. Gen.
Fahey, F. H., Wood, F. W., Flowers, D. L., Eades, C. G., Gage, H. D.,
and Harkness, B. A. 1998. Evaluation of brain activation in FDG
PET studies. J . Comput. Assisted Tomogr. 22:953–961.
Felleman, D. J ., and Van Essen, D. C. 1991. Distributed hierarchical
processing in theprimatecerebral cortex. Cereb. Cortex 1:1–47.
CORTICAL ACTIVITY RELATED TO TASK ACCURACY
Friston, K. J ., Frith, C. D., Liddle, P. F., Dolan, R. J ., Lammertsma,
A. A., and Frackowiak, R. S. K. 1990. The relationship between
global and local changes in PET scans. J . Cereb. Blood Flow Metab.
Friston, K. J ., Frith, C. D., Liddle, P. F., and Frackowiak, R. S. J .
1991. Comparing functional (PET) images: The assessment of
significant change. J . Cereb. Blood Flow Metab. 11:690–699.
Friston, K. J ., Worsley, K. J ., Frackowiak, R. S. J ., Mazziotta, J . C.,
and Evans, A. C. 1994. Assessing the significance of focal activa-
tions using their spatial extent. Hum. Brain Mapp. 1:214–220.
Friston, K. J ., Holmes, A. P., Worsley, K. J ., Poline, J . B., Frith, C. D.,
and Frackowiak, R. S. J . 1995. Statistical parametric maps in
functional imaging:A general approach. Hum. Brain Mapp. 2:189–
Gaffan, D., Harrison, S., and Gaffan, E. A. 1986. Visual identification
following infero-temporal ablation in the monkey. J . Exp. Psychol.
Grady, C. L., Horwitz, B., Pietrini, P., Mentis, M. J ., Ungerleider,
L. G., Rapoport, S. I., and Haxby, J . V. 1996. Effect of task difficulty
on cerebral blood flow during perceptual matching of faces. Hum.
Brain Mapp. 4:227–239.
Grafton, S. T., Mazziotta, J . C., Presty, S., Friston, K. J ., Frackowiak,
R. S. J ., and Phelps, M. E. 1992. Functional anatomy of human
procedural learning determined with regional cerebral blood flow
and PET. J . Neurosci. 12:2542–2548.
Grafton, S. T., Woods, R. P., and Tyszka, M. 1994. Functional imaging
of procedural motor learning: Relating cerebral blood flow with
individual subject performance. Hum. Brain Mapp. 1:221–234.
Green, D. M., and Swets, J . A. 1966. Signal Detection Theory and
Psychophysics. Wiley, New York.
Haier, R. J ., Siegel, B. V., J r., MacLachlan, A., Soderling, E., Lotten-
berg, S., and Buchsbaum, M. S. 1992. Regional glucose metabolic
changes after learning a complex visuospatial/motor task: A posi-
tron emission tomographic study. Brain Res. 570:134–143.
Harris, R. J . 1975. A Primer of Multivariate Statistics. Academic
Press, New York.
Haxby, J . V., Grady, C. L., Horwitz, B., Ungerleider, L. G., Mishkin,
M., Carson, R. E., Herscovitch, P., Schapiro, M. B., and Rapoport,
S. I. 1991. Dissociation of object and spatial visual processing
pathways in human extrastriatecortex. Proc. Natl. Acad. Sci. USA
Haxby, J . V., Horwitz, B., Ungerleider, L. G., Maisog, J . M., Pietrini,
P., and Grady, C. L. 1994. The functional organization of human
extrastriate cortex: A PET–rCBF study of selective attention to
faces and locations. J . Neurosci. 14:6336–6353.
Haxby, J . V., Ungerleider, L. G., Horwitz, B., Rapoport, S. I., and
Grady, C. L. 1995. Hemispheric differences in neural systems for
face working memory: A PET–rCBF study. Hum. Brain Mapp.
Howard, D., Patterson, K., Wise, R., Brown, D., Friston, K., Weiller,
C., and Frackowiak, R. 1992. The cortical localization of the
lexicons. Positron emission tomography evidence. Brain 115:1769–
J enkins, I. H., Brooks, D. J ., Nixon, P. D., Frackowiak, R. S. J ., and
Passingham, R. E. 1994. Motor sequence learning: A study with
positron emission tomography. J . Neurosci. 14:3775–3790.
J uliano, S. L., Hand, P. J ., and Whitsel, B. L. 1981. Patterns of
increasedmetabolicactivity in somatosensory cortex of themonkey
Macaca fascicularis, subjectedtocontrolledcutaneous stimulation:
A 2-deoxyglucosestudy. J . Neurophysiol. 46:1260–1284.
J uliano, S. L., Whitsel, B. L., and Hand, P. J . 1983. Patterns of
metabolic activity in cytoarchitectural area SII and surrounding
cortical fields of themonkey. J . Neurophysiol. 50:961–980.
Kadekaro, M., Grane,A. M., and Sokoloff, L. 1985. Differential effects
of electrical stimulation of sciatic nerve on metabolic activity in
spinal cord and dorsal root ganglion in the rat. Proc. Natl. Acad.
Sci. USA 82:2337–6013.
Kawashima, R., Satoh, K., Goto, R., Inoue, D., Itoh, M., and Fukuda,
H. 1998. The role of the left inferior temporal cortex for visual
pattern discrimination—A PET study. NeuroReport 9:1581–1586.
Koepp, M. J ., Labbe, D., Richardson, M. P., Brooks, B. J ., Paesschen,
W., Cunningham, B. J ., andDuncan, J . S. 1987. Regional hippocam-
pal [11C] flumazenil PET in temporal lobe epilepsy with unilateral
and bilateral hippocampal sclerosis. Brain 120:1865–1876.
Kosslyn, S. M.,Alpert, N. M., Thompson, W. L., Chabris, C. F., Rauch,
S. L., and Anderson, A. K. 1994. Identifying objects seen from
different viewpoints.A PET investigation. Brain 117:1055–1071.
Maunsell, J . H. R., and Newsome, W. T. 1987. Visual processing in
monkey extrastriatecortex. Annu. Rev. Neurosci. 10:363–401.
McIntosh, A. R., Grady, C. L., Haxby, J . V., Ungerleider, L. G.,
Rapoport, S. I., and Horwitz, B. 1996. Changes in limbic and
prefrontal functional interactions in a working memory task for
faces. Cereb. Cortex 6:571–584.
Mishkin, M., Ungerleider, L. G., and Macko, K.A. 1983. Object vision
and spatial vision: Twocortical pathways. Trends Neurosci. 6:414–
Moeller, J . R., Strother, S. C., Sidtis, J . J ., and Rottenberg, D.A. 1987.
Scaled subprofile model: A statistical approach to the analysis of
functional patterns in positron emission tomographic data. J .
Cereb. Blood Flow Metab. 7:649–658.
Moeller, J . R., and Strother, S. C. 1991. A regional covariance
approach totheanalysis of functional patterns in positron emission
tomographic data. J . Cereb. Blood Flow Metab. 11:A121–A135.
Moore, C. J ., and Price, C. J . 1998. Threedistinct ventral occipitotem-
poral regions for reading and object naming. NeuroImage 10:181–
Nagahama, Y., Sadato, N., Yamauchi, H., Katsumi, Y., Hayashi, T.,
Fukuyama, H., Kimura, J ., Shibasaki, H., and Yonekura, Y. 1998.
Neural activity during attention shifts between object features.
Neelin, P., Crossman, J ., Hawkes, D. J ., Ma,Y., andEvans,A. C. 1993.
Validation of an MRI/PET landmark registration method using 3D
simulated PET images and point simulations. Comput. Med. Imag.
Perrett, D. I., Mistlin, A. J ., and Chitty, A. J . 1987. Visual neurones
responsivetofaces. Trends Neurosci. 10:358–364.
Petersen, S. E., Fox, P. T., Posner, M. I., Mintun, M., and Raichle,
M. E. 1989. Positron emission tomographic studies of the process-
ing of singlewords. J . Cognit. Neurosci. 1:153–170.
Petersen, S. E., Fox, P. T., Snyder, A. Z., and Raichle, M. E. 1990.
Activation of extrastriateand frontal cortical areas by visual words
and word-likestimuli. Science249:1041–1044.
Phelps, M. E., Huang, S. C., Hoffman, E. J ., Selin, C., Sokoloff, L., and
Kuhl, D. E. 1979. Tomographic measurement of local cerebral
glucose metabolic rate in humans with [18F]2-fluoro-2-deoxy-2-D-
glucose: Validation of method. Ann. Neurol. 6:371–388.
Price, C. J ., Moore, C. J ., Humphreys, G. W., Frackowiak, R. S. J ., and
Friston, K. J . 1996. The neural regions sustaining object recogni-
tion and naming. Proc. R. Soc. London 262:1501–1507.
Puce, A., Allison, T., Asgari, M., Gore, J . C., and McCarthy, G. 1996.
Differential sensitivity of human visual cortex to faces, letter-
strings, and textures: A functional magnetic resonance imaging
study. J . Neurosci. 16:5205–5215.
Raichle, M. E., Fiez, J . A., Videen, T. O., MacLeod, A.-M. K., Pardo,
J . V., Fox, P. T., and Petersen, S. E. 1994. Practice-related changes
in human brain functional anatomy during nonmotor learning.
Cereb. Cortex 4:8–26.
Roland, P. E., and Gulyas, B. 1995. Visual memory, visual imagery,
GARRETT ET AL.
and visual recognition of large field patterns by the human brain: Download full-text
Functional anatomy by positron emission tomography. Cereb.
Schacter, D. L., Reiman, E., Uecker, A., Polster, M. R., Yun, L. S., and
Cooper, L. A. 1995. Brain regions associated with retrieval of
structurally coherent visual information. Nature376:587–590.
Schlaug, G., Knorr, U., and Seitz, R. J . 1994. Inter-subject variability
of cerebral activations in acquiring a motor skill: A study with
positron emission tomography. Exp. Brain Res. 98:523–534.
Schwartz, E. L., Desimone, R., Albright, T. D., and Gross, C. G. 1983.
Shaperecognition and inferior temporal neurons. Proc. Natl. Acad.
Sci. USA 80:5776–5778.
Sergent, J ., Ohta, S., and MacDonald, B. 1992. Functional neuro-
anatomy offaceandobject processing.A positron emission tomogra-
phy study. Brain 115:15–36.
Shaywitz, B. A., Pugh, K. R., Constable, R. T., Shaywitz, S. E.,
Bronen, R. A., Fulbright, R. K., Shankweiler, D. P., Katz, L.,
Fletcher, J . M., Skudlarski, P., and Gore, J . C. 1995. Localization of
semantic processing using functional magnetic resonanceimaging.
Hum. Brain Mapp. 2:149–158.
Soma, Y., Sugishita, M., Kitamura, K., Maruyama, S., and Imanaga,
H. 1989. Lexical agraphia in the J apanese language. Brain 112:
Spielberger, C. D., Gorsuch, R. L., Luchene, R., Bagg, P. R., and
J acobs, G. A. 1983. Manual for the State–Trait Anxiety Inventory
(STAI)—Form Y. Consulting Psychology Press, PaloAlto, CA.
Talairach, J ., and Tournoux, P. 1988. Coplanar Stereotaxic Atlas of
theHuman Brain. Thieme, New York.
Tanaka, K., Saito, H., Fukuda,Y., andMoriya, M. 1991. Coding visual
images of objects in the infero-temporal cortex of the macaque
monkey. J . Neurophysiol. 66:179–189.
Ungerleider, L. G., Courtney, S. M., and Haxby, J . V. 1998. A neural
system for human visual working memory. Proc. Natl. Acad. Sci.
Ungerleider, L. G., and Haxby, J . V. 1994. ‘What’ and ‘where’ in the
human brain. Curr. Opin. Neurobiol. 4:157–165.
Ungerleider, L. G., and Mishkin, M. 1982. Two cortical visual
systems. In Analysis of Visual Behavior (D. J . Ingle, M.A. Goodale,
and R. J . W. Mansfield, Eds.), pp. 549–586. MIT Press, Cambridge,
Wechsler, D. 1981. Wechsler Adult Intelligence Scale—Revised. Psy-
chological Corp., New York.
Wood, F. 1990. Functional neuroimaging in neurobehavioral re-
search. In Neuromethods, Vol. 17, Neuropsychology (A. A. Boulton,
G. B. Gaker, and M. Hiscock, Eds.). Humana Press, Clifton, NJ .
Wood, F.,Armentrout, R., Toole, J ., McHenry, L., and Stump, D. 1980.
Regional cerebral blood flow during rest and memory activation in
a patient with global amnesia. Brain Lang. 9:124–136.
Wood, F. B., and Flowers, D. L. 1999. Functional neuroanatomy of
dyslexic subtypes: A survey of 43 candidate regions with a factor
analytic validation across 100 cases. In Reading and Attention
Disorders: Neurobiological Correlates (D. D. Duane, Ed.), pp.
131–161. Yorkton Press, Parkton.
Woodcock, R. W., and J ohnson, M. B. 1989. Woodcock–J ohnson
Psychoeducational Battery—Revised. DLM Teaching Resources,
Woods, R. P. 1996. Correlation of brain structure and function. In
Brain Mapping: The Methods (A. W. Toga and J . C. Mazziotta,
Eds.), pp. 313–341.Academic Press, San Diego.
Worsley, K. J ., Evans, A. C., Marrett, S., and Neelin, P. 1992. A
three-dimensional statistical analysis for rCBF activation studies
in human brain. J . Cereb. Blood Flow Metab. 12:900–918.
Worsley, K. J ., Marrett, S., Neelin, P., Vandal,A. C., Friston, K. J ., and
Evans, A. C. 1996. A unified statistical approach for determining
significant signals in images of cerebral activation. Hum. Brain
Young, M. P. 1992. Objective analysis of the topological organization
of theprimatecortical visual system. Nature358:152–154.
CORTICAL ACTIVITY RELATED TO TASK ACCURACY