Cognitive Neuroscience and Neuropsychology
Website publication 29 September 1997
NeuroReport 8, 3275–3279 (1997)
WE report results from a patient in whom we obtained
converging evidence from positron emission tomography (PET) and intraoperative stimulation mapping to
support a one-way dissociation between the functional
areas involved in word repetition and synonym generation. Intraoperative stimulation mapping interfered
with synonym generation but did not disturb word
repetition at the same left inferior frontal site at which
a cerebral blood flow (CBF) increase had been observed
for a synonym generation task. The results for this single
subject suggest that the functional areas involved in
different aspects of linguistic processing are dissociable
and that specific disruption under conditions of cortical
stimulation can be correlated with the brain regions
identified via PET as the most active during performance
of a specific task.
Obligatory role of the
LIFG in synonym
generation: evidence
from PET and cortical
stimulation
Denise Klein,CA André Olivier,
Brenda Milner, Robert J. Zatorre,
Ingrid Johnsrude, Ernst Meyer
and Alan C. Evans
Neuropsychology/Cognitive Neuroscience Unit,
Department of Neurology and Neurosurgery
and McConnell Brain Imaging Center, Montreal
Neurological Institute and Hospital, McGill
University, 380l University St., Montreal,
Québec, H3A 2B4, Canada
Key words: Imaging; Language; Positron emission tomography; Stimulation mapping
Introduction
p
The introduction of cerebral blood flow (CBF)
measurement and positron emission tomography
(PET)1 in conjunction with detailed in vivo magnetic
resonance imaging (MRI)2 has permitted the direct
exploration of brain function in the human subject.
Central to PET studies of cognitive function is the
use of tasks that differ only in one or a few operations. By comparing PET images acquired during the
performance of these tasks, it is possible to isolate
the brain areas differentially involved with the operations that distinguish the two tasks.3
Language processes have been extensively studied
in brain-imaging research and the distributed, modular nature of these processes has been described.4
Although such studies have provided preliminary
insights into the neural basis of these different
linguistic functions, the precise determination of
which brain structures mediate a particular cognitive
function is still a source of debate. For example, functional brain imaging studies have consistently demonstrated increased CBF in the left inferior frontal gyrus
(LIFG: Brodmann’s areas 45 and 47) during the
performance of tasks that involve lexical access and
retrieval, such as noun–verb generation5,6 or synonym
© Rapid Science Publishers
CA
Corresponding Author
generation,7 but such observations require validation
from direct experimental evidence, such as can be
obtained from intraoperative stimulation mapping.
Intraoperative corticography in the awake patient8,9
provides a way to map the locations at which interference with language occurs. The pattern of organization derived from electrical stimulation mapping,
like that derived from lesions, depends on the disruption of a behaviour when a particular cortical area
is functionally inactivated. This technique thus
identifies cortical areas that must be active for the
behaviour and thus are essential for the measured
function. In contrast, PET is an activation method,
revealing the network of brain structures involved
in a tested behaviour, but discrimination between
essential and non-essential structures is not possible
using PET alone.10,11 This lack of specificity poses a
particular problem when functional imaging of
language is used as a clinical tool in presurgical planning for interventions anticipated to encroach upon
frontal or temporoparietal regions in the languagedominant hemisphere. Since both techniques for
investigating language organization have some limitations, convergence of the data can throw light on
essential versus participatory areas for a given
language task.
Vol 8 No 15 20 October 1997
3275
D. Klein et al.
1
We report evidence from a patient in whom we
observed consistent localization of function in the left
inferior frontal cortex using two complementary
methods. First, we obtained physiological correlates
of language derived from CBF recording during
synonym generation and word repetition tasks using
PET. Second, we stimulated the exposed cortex
during surgery with the awake patient during performance of tasks similar to those used in the PET study.
11
Subject and Methods
11
11
11
11
1p
Case history: In March 1995, the patient (WW), a
38-year-old right-handed physician presented with a
grand mal seizure and was found to have an oligodendroglioma in the left premotor area. A brain
biopsy, performed elsewhere, was complicated by a
postoperative intracranial hemorrhage that required
craniotomy to evacuate blood. The patient was left
with a right hemiparesis, primarily affecting the arm,
and he also displayed a non-fluent dysphasia and an
apraxia. He was released from hospital in April 1995,
and, after rehabilitation, experienced rapid return
of leg function and a gradual return of speech.
He had good recovery until he experienced his first
post-discharge seizure in July 1995, subsequent to
which he continued to have right focal-motor
seizures every 4–6 weeks, resulting in a post-ictal
Todd’s paralysis of the right hand and speech arrest.
WW was admitted to the Montreal Neurological
Hospital in March 1996 for resection of the tumour.
The patient provided informed consent for the
imaging and mapping procedures, which were
performed according to institution-reviewed medicoethical guidelines.
Presurgical mapping with PET: In our laboratory
we have observed, both in a group study and in each
of seven normal volunteer subjects that, subtraction
of word repetition from synonym generation (using
a repeated design) consistently elicits strongest CBF
changes in the left inferior frontal region.12 Because
of the reliability of these results, these two tasks were
used as a reference for direct examination of the relationship between morphology and function in patient
WW, whose focal brain lesion was located in the left
superior frontal cortex. Although several processes
are shared by the two tasks, we were interested in
the additional processes of lexical search and retrieval,
required for synonym generation but not for word
repetition.
For each of three scans, 15 individual words were
presented binaurally through insert earphones
(Eartone 3A), and the patient was instructed to say
aloud a synonym (e.g. leap–jump; weep–cry). Three
separate word lists, matched item by item on a range
3276 Vol 8 No 15 20 October 1997
of psycholinguistic variables (word length, syllable
number, frequency and part of speech), were used
for each scan, and stimuli were presented every
4.2 s, for a total scanning time of 60 s. Similarly, three
word-repetition scans, in which WW repeated heard
words, served as baseline scans. The lights were
dimmed for the duration of the scans, which were
presented in a counterbalanced order, and WW was
instructed to keep his eyes closed throughout each
scan. Accuracy of responses during the scans was
monitored. WW repeated all the words correctly but
was slightly less accurate at synonym generation,
although he was able to perform the task (list 1, 11/15
correct; list 2, 9/15 correct; list 3, 14/15 correct).
Standard PET scanning methods were used in
conjunction with the 15O-labeled water bolus injection technique.1 Images were acquired on a Scanditronix PC-2048 system, which produces 15 image
slices at an intrinsic resolution of 5.0 3 5.0 3
6.0 mm.13 PET images were reconstructed using
an 18 mm Hanning filter. PET data were then
normalized for global CBF value, averaged for each
activation state, and the mean CBF-change images
were obtained.13 These volumes were converted to a
t-statistic volume by dividing each voxel by the
pooled standard deviation in normalized CBF for all
intracerebral voxels.14 An MRI volume (160 sagittal
slices, 1 mm thick) was obtained, MRI and PET
volumes were co-registered, and the dataset was
linearly re-sampled into a stereotaxic coordinate
system.15,16 The presence of significant focal changes
was tested by a method based on 3-D Gaussian
random-field theory.14 Values with a criterion of t >
3.5 were deemed statistically significant (p < 0.0002,
two-tailed, uncorrected).
Regions of significant CBF increase are shown in
Table 1. When word repetition was subtracted from
synonym generation, a series of strong activations
were observed anteriorly in the left inferior-frontal
gyrus (areas 45 and 47), as predicted from our studies
of normal right-handed subjects.7,12 In addition, bilateral activations were observed in the orbito-frontal
region (Brodmann’s area 11) and in the middle frontal
gyrus (area 10). These findings suggest major
language representation in the left hemisphere, with
the strongest CBF changes related to synonym generation being inferior to the lesion (see Fig. 1). For
comparison, this left inferior frontal focus (marked
with an * in Table 1) is located very near the foci of
our two previous studies of normal volunteers.7,12
Intraoperative stimulation mapping: The operation
involved reopening of the left frontal craniotomy, a
lesionectomy and corticectomy in the intermediate
frontal area. Cortical stimulation mapping was performed with the goal of stimulating sites previously
Obligatory role of the LIFG in synonym generation
Table 1. Blood flow increases: Synonym generation
minus Word repetition
Stereotaxic
coordinates (mm)
Region
t-value
Brodmann
area
x
y
z
11/47
10/47
45/47
10
–16
–24
–47
–47
–17
–70
32
56
25
53
–14
–40
–26
–15
–14
–9
2
–8
6.43a
6.16a
5.87*
4.54a
4.35
4.10
25
40
40
39
1
53
25
39
20
39
–18
–21
–2
-6
3
5.57
4.72
4.6
3.78
4.15
Left hemisphere
Orbital frontal
Inferior frontal
Middle frontal
Thalamus
Middle temporal
Right hemisphere
Orbital frontal
Middle frontal
Insular/frontal opercular
Cingulate (midline)
21
11
11
10
47
Activation foci represent peaks of statistically significant
increases in normalized CBF. The stereotaxic coordinates16
are: x, medial to lateral distance relative to the midline
(positive = right); y, anterior–posterior distance relative
to the anterior commissure (positive = anterior); z is the
superior–inferior distance relative to the anterior commisssure line (positive = superior).
a
Due to the location of the craniotomy, these activation foci
were not accessible during the operation and were therefore not sampled by stimulation mapping.
* See Fig. 1.
p
identified by PET as subserving a particular linguistic
function.
The t-statistic images generated from the PET
study were transformed back into the dimensions of
the original MRI image, using the inverse of the
matrix used for stereotaxic transformation. In the
operating room, an image-guided surgical system was
used (The Viewing Wand, ISG Technologies Inc.
Toronto) to display the patient’s MRI and PET
images along with a representation of an intraoperative probe whose position is tracked in real time
during the operation.17 Using a landmark-matching
algorithm, the patient’s position was coregistered with
the computer image with an accuracy of the order of
1 mm.17 Stimulation mapping of language was carried
out after the lesion had been located and after the
sensorimotor cortex had been identified (Fig. 1: tags
visible in the lower central areas 1,2,3,7 and 8).
For language mapping, trains of 60 Hz, 2 ms total
duration biphasic square wave pulses from a constant
current stimulator were delivered across 1 mm
bipolar electrodes separated by 5 mm. The current
level was selected as the largest current not evoking
after-discharges in the area of cortex to be sampled.
Due to the extent of the craniotomy and time
limitations on stimulation mapping in the operating
room, there was a trade-off between number of
samples, the number of behaviours and the number
of sites where stimulation effects on these behaviours
could be assessed. Neither the patient nor the
examiner was able to detect when stimulation was
taking place, or which location was being probed.
Stimulation was applied to several areas in the left
frontal cortex, but the anterior LIFG was specifically
targeted, because this was the site of strongest PET
CBF change observed for our normal volunteer
subjects,12 and also because it was a site of strong
CBF change observed for WW (Table 1). Due to the
location of the craniotomy, several activation foci in
the left frontal region were not accessible during the
operation and were therefore not sampled with stimulation mapping (see Table 1).
To allow for direct comparison between the two
techniques, similar tasks were used in the operating
room to those used with PET. The patient was
presented orally with a word and was then required
to repeat it and to generate its synonym within a
single trial (e.g. leap–leap; jump). Stimulation began
at the onset of the word to be repeated and continued
for approximately 5 s. When stimulation was applied
at 2.0 V over the anterior LIFG (i.e. F3), WW could
repeat words, but he could not produce a synonym.
Out of the five items presented at this specific site
(which corresponded to the area of increased activity
observed with PET), he was always able to repeat
the word but was never able to produce a synonym.
WW had previously been able to provide the
synonyms for all of these words when no stimulation was applied. Stimulation at this same location
also resulted in a failure to produce a response when
he was presented with an item for picture naming,
but he was able to count correctly from 1 to 20. The
sites of stimulation were marked with tags and,
after stimulation, the location relative to the PET activation was recorded with the Viewing Wand (see
Fig. 1). Importantly, there were several other neighbouring sites in the anterior inferior frontal region
that were sampled, but where no interference was
noted. Comprehension was intact throughout intraoperative testing.
In postoperative debriefing, the patient reported
that during the stimulation procedure, in the instances
where stimulation interfered with lexical search and
retrieval, he had known the item that he wanted to
say, but could not produce the response. He termed
it ‘thought blocking’. When tested one day postoperatively, he could again produce all the synonyms
correctly and could identify which items he had
experienced difficulty with in the operating room.
Discussion
The study of the organization of language in the
human brain, and the hypotheses derived therefrom, are influenced by the specific limitations and
Vol 8 No 15 20 October 1997
3277
D. Klein et al.
1
11
11
11
11
11
1p
FIG. 1. PET and stimulation-mapping data are displayed for WW performing the synonym generation task. Top left: PET subtraction image
superimposed upon WW’s MRI, resampled into a stereotaxic coordinate system15,16 to allow comparison and localization with reference to
our dataset of normal volunteer subjects.7,12 Several predictable increases in CBF were observed in the left inferior frontal gyrus, inferior to
the lesion, in an area similar to that activated in normal subjects.7,12 The peak visible in the image corresponds with Brodmann’s area 45/47
in Table 1, and is located 5.8 mm from the focus reported in Ref. 7 (yellow square) and 3.00 mm from the focus in Ref. 12 (red square). It
is also the location where interference with synonym generation, but not word repetition, was observed during cortical stimulation. The
other panels show the data superimposed on the MRI in native space. Top right: Averaged PET subtraction image. Bottom left: Responses
elicited by stimulation. The yellow x’s show the exact position where stimulation interfered with speech and function. Tags S1 and S2 refer
to interference with synonym generation, while the numbers 1,2,3,7 and 8 refer to sites where movements and dysesthetic sensations were
evoked by stimulation. Both the yellow x’s and the white tags were inputted during WW’s surgery, using the Viewing Wand. Bottom right:
PET and stimulation-mapping responses superimposed.
strengths of the different techniques used to study
such processes; consequently there is much to benefit
from the corroboration of findings with different
methods. In this individual case, electrical stimulation of neural tissue at the site of the CBF increase
related to synonym generation in PET produced a
reliable interference with the performance of the same
task. This interference was both functionally and
anatomically specific, being observed for synonym
generation, but not for repeating words, and was
elicited by stimulation of a discrete area of cortex.
Although synonym generation is a more complicated
3278 Vol 8 No 15 20 October 1997
task than word repetition, and hence may be more
easily disrupted by some general interfering factor,
the differential effect was observed at a specific site
and not at other sites. The interference with synonym
generation but not repetition at this particular site
supports the idea that different cortical locations in
an individual seem primarily concerned with specific
language-related functions. The anterior LIFG thus
appears to be critical for synonym generation but not
for word repetition, which is consistent with the
numerous reports of CBF increases in this region on
tasks requiring lexical search and retrieval.5–7,12
Obligatory role of the LIFG in synonym generation
PET and intraoperative stimulation both have
limitations as mapping techniques. PET is unable to
clarify the relationship between involved and essential cortex because all the cortical regions involved
during task performance are detected. In contrast,
intraoperative cortical stimulation mapping may
accomplish this goal, but it has limited cortical
sampling. Nevertheless, in the case of WW, comparison of these two techniques established a functional
anatomic relationship, and allowed for the identification of a cortical area essential for a particular facet
of language. In so doing, the potential importance of
activation mapping in guiding neurosurgical interventions was further underlined.
Conclusion
We present a case study of a patient who developed
seizures secondary to a left frontal tumour, and in
whom we had the opportunity to map the area of
the left frontal cortex required for synonym generation that we had previously localized by PET activation studies in normal volunteer subjects. At
surgery, stimulation of the same area of the anterior
left inferior frontal gyrus consistently interfered with
synonym generation but not with word repetition,
supporting the idea of a spatial correspondence
between PET activation foci and functional regions
established by other measures. This finding suggests
that there are dissociable brain regions involved in
different language functions.
References
1. Raichle ME, Martin WRW, Herscovitch P et al. J Nucl Med 24, 790–798
(1983).
2. Evans AC, Marrett S, Torrescorzo J et al. J Cerebr Blood Flow Metab 11,
A69–A78 (1991).
3. Posner MI, Petersen SE, Fox PT and Raichle ME. Science 240, 1627–1631
(1988).
4. Raichle ME. Annu Rev Psychol 45, 333–356 (1994).
5. Petersen SE, Fox PT, Posner MI et al. Nature 331, 585–589 (1988).
6. Raichle ME, Fiez JA, Videen TO et al. Cerebr Cortex 4, 8–26 (1994).
7. Klein D, Milner B, Zatorre RJ et al. Proc Natl Acad Sci USA 92, 2899–2903
(1995).
8. Ojemann GA. J Neurosci 11, 2281–2287 (1991).
9. Penfield W and Roberts L. Speech and Brain Mechanisms. Princeton:
Princeton University Press, 1959.
10. Steinmetz H and Seitz RJ. Neuropsychologia 29, 1149–1161 (1991).
11. Bookheimer SY, Zeffiro TA, Blaxton T et al. Neurology 48, 1056–1065 (1997).
12. Klein D, Zatorre RJ, Milner B et al. NeuroImage 3, S444 (1996).
13. Evans AC, Thompson CJ, Marrett S et al. IEEE Trans Med Imaging 10,
90–98 (1991).
14. Worsley KJ, Evans AC, Marrett S and Neelin P. J Cerebr Blood Flow Metab
12, 900–918 (1992).
15. Evans AC, Marrettt S, Neelin P et al. NeuroImage 1, 43–53 (1992).
16. Talairach J and Tournoux P. Co-Planar Stereotaxic Atlas of the Human
Brain. 3-Dimensional Proportional System: An Approach to Cerebral
Imaging. New York: Thieme Medical Publishers, 1988.
17. Peters T, Davey B, Munger P et al. IEEE Trans Med Imaging 15, 121–128
(1996).
ACKNOWLEDGEMENTS: We thank patient WW for his cooperation and consent.
Supported by grants from the Medical Research Council of Canada (MT2624,
SP-30 and MT11541), the McDonnell-Pew Program in Cognitive Neuroscience,
the Isaac Walter Killam Fellowship Fund of the MNI and the Fonds de la
Recherche en Santé du Québec.
Received 24 July 1997;
accepted 17 August 1997
p
Vol 8 No 15 20 October 1997
3279