Activation of Broca's area during the production of spoken and signed language: A combined cytoarchitectonic mapping and PET analysis

Article (PDF Available)inNeuropsychologia 41(14):1868-76 · February 2003with191 Reads
DOI: 10.1016/S0028-3932(03)00125-8 · Source: PubMed
Abstract
Broca's area in the inferior frontal gyrus consists of two cytoarchitectonically defined regions-Brodmann areas (BA) 44 and 45. Combining probabilistic maps of these two areas with functional neuroimaging data obtained using PET, it is shown that BA45, not BA44, is activated by both speech and signing during the production of language narratives in bilingual subjects fluent from early childhood in both American Sign Language (ASL) and English when the generation of complex movements and sounds is taken into account. It is BA44, not BA45, that is activated by the generation of complex articulatory movements of oral/laryngeal or limb musculature. The same patterns of activation are found for oral language production in a group of English speaking monolingual subjects. These findings implicate BA45 as the part of Broca's area that is fundamental to the modality-independent aspects of language generation.
Neuropsychologia 41 (2003) 1868–1876
Activation of Broca’s area during the production of spoken and signed
language: a combined cytoarchitectonic mapping and PET analysis
Barry Horwitz
a,
, Katrin Amunts
b
, Rajan Bhattacharyya
a
, Debra Patkin
a
,
Keith Jeffries
a
, Karl Zilles
b
, Allen R. Braun
a
a
Language Section, National Institute on Deafness and Other Communication Disorders, National Institutes of Health,
Bldg. 10, Rm. 6C420, MSC 1591, Bethesda, MD 20892, USA
b
Research Center, Institute of Medicine, D-52428 Jülich, Germany
Received 30 October 2002; received in revised form 4 March 2003; accepted 20 May 2003
Abstract
Broca’s area in the inferior frontal gyrus consists of two cytoarchitectonically defined regions—Brodmann areas (BA) 44 and 45.
Combining probabilistic maps of these two areas with functional neuroimaging data obtained using PET, it is shown that BA45, not BA44,
is activated by both speech and signing during the production of language narratives in bilingual subjects fluent from early childhood in
both American Sign Language (ASL) and English when the generation of complex movements and sounds is taken into account. It is
BA44, not BA45, that is activated by the generation of complex articulatory movements of oral/laryngeal or limb musculature. The same
patterns of activation are found for oral language production in a group of English speaking monolingual subjects. These findings implicate
BA45 as the part of Broca’s area that is fundamental to the modality-independent aspects of language generation.
Published by Elsevier Ltd.
Keywords: Neuroimaging; American Sign Language; Brain; Human; Brodmann areas
1. Introduction
The name Broca’s area has been applied to the pos-
terior portion of the left inferior frontal gyrus contain-
ing the pars triangularis and the pars opercularis (Broca,
1861; Tomaiuolo et al., 1999). The cytoarchitectonic ar-
eas (Brodmann, 1909) said to comprise this region are
Brodmann areas (BA) 44 and 45 (Uylings et al., 1999).
Damage in the vicinity of this structure often results in a
Broca’s aphasia: an impairment of language characterized
by non-fluent, agrammatical verbal output with relatively
preserved language comprehension (Benson & Geschwind,
1985). Nonetheless, there is much controversy concerning
the relationship between Broca’s area and language-related
function. Lesions restricted to Broca’s area do not always
lead to a Broca’s aphasia (Dronkers, Shapiro, Redfern,
& Knight, 1992; Mohr et al., 1978), and conversely, pa-
tients with a Broca’s aphasia do not always have a lesion
in Broca’s area (Dronkers & Ludy, 1998; Dronkers et al.,
1992). Although numerous functional neuroimaging stud-
ies have reported activation of Broca’s area by language
Corresponding author. Tel.: +1-301-594-7755; fax: +1-301-480-5625.
E-mail address: horwitz@helix.nih.gov (B. Horwitz).
tasks (Caplan, Alpert, Waters, & Olivieri, 2000; Dapretto
& Bookheimer, 1999; Just, Carpenter, Keller, Eddy, &
Thulborn, 1996; Rumsey et al., 1997; Salmelin, Schnitzler,
Schmitz, & Freund, 2000; Zatorre, Evans, Meyer, & Gjedde,
1992), a number of investigators have shown that this area
also can be activated by non-language tasks(Binkofski et al.,
2000; Iacoboni et al., 1999). Even the boundaries of Broca’s
area are ill-defined; a cytoarchitectonic study of BA44 and
BA45 by Amunts et al. (1999) in 10 human brains found
that the cytoarchitectonic boundaries of these areas did not
have a fixed relationship to sulcal landmarks, and that there
was much intersubject variability amongst the 10 brains.
Because BA44 is just anterior to the mouth area of the
motor strip, it had been widely conjectured that the loca-
tion of Broca’s area reflects the involvement of this cortex
with the motor articulatory aspects of speaking (Goodglass,
1993) (however, see Corina et al., 1999 for evidence counter
to this view). However, several studies of fluent speakers of
sign language have shown that Broca’s area seems to play
a crucial role in language production, even though the pro-
duction in signers relies primarily on the use of arms and
hands (Braun, Guillemin, Hosey, & Varga, 2001; Hickok,
Bellugi, & Klima, 1998; McGuire et al., 1997; Petitto et al.,
2000).
0028-3932/$ see front matter. Published by Elsevier Ltd.
doi:10.1016/S0028-3932(03)00125-8
B. Horwitz et al./Neuropsychologia 41 (2003) 1868–1876 1869
This paper examines the role that the two cytoarchi-
tectonic subdivisions of Broca’s region play in language
production in hearing subjects fluent in both American Sign
Language (ASL) and English. By examining languages
whose modes of comprehension-production are dramati-
cally different (i.e. auditory-oral for spoken languages such
as English, visual-gestural for ASL), any conclusions we
draw should pertain to the most general aspects of language
production.
Probabilistic maps of the two Brodmann areas constitut-
ing Broca’s area were obtained by Amunts et al. (1999).
These maps were derived by histological analysis that de-
termined the locations and spatial extents of BA44 and
BA45 in the left and right hemispheres in serial coronal
sections of 10 individual brains, as mentioned above, af-
ter which the cytoarchitectonic data were transposed to a
common space so as to form a probabilistic atlas, where
voxel value indicates the percentage of brains having that
location as BA44 or BA45. We combined these maps with
functional brain activation data obtained by positron emis-
sion tomography (PET) (Braun et al., 2001), thus allowing
us to determine the probability that BA44 or BA45 was ac-
tivated during language production tasks relative to motor
control tasks. Most neuroimaging studies of language em-
ploy relatively simple tasks in which specific linguistic el-
ements can be isolated and individually studied. We chose
to have our subjects “tell a story” while they were being
scanned, thus allowing us to examine language production
in connected discourse as it is used in everyday life. In this
way, the interpretation of any differences in the activation of
BA44 and BA45 will not depend upon any single linguistic
element.
2. Materials and methods
2.1. Positron emission tomography
2.1.1. Subjects
The PET data used in this study were obtained from two
groups of subjects: (1) Bilinguals—11 healthy volunteers
(six males and five females, age range: 28–56 years) who
were the adult children of deaf parents and were fluent in
both English and ASL (Braun et al., 2001); these subjects
were exposed to both English and ASL as native languages
from birth, and continued to use both languages daily at
the time of the study; (2) Monolinguals—20 healthy vol-
unteers (12 males and 8 females, age range: 23–50 years)
who were native speakers of English and had no knowl-
edge of ASL (Braun et al., 1997). All subjects were right
handed, and were free of medical or neuropsychiatric dis-
orders. These studies were conducted under a protocol ap-
proved by the NIDCD-NINDS IRB (NIH 92-DC-0178).
Written informed consent was obtained according to the
declaration of Helsinki. Subjects were compensated for par-
ticipating. Details about subjects and procedures, and por-
tions of the PET data, can be found in Braun et al. (2001,
1997).
2.1.2. Tasks
In the studies of Braun et al. (1997, 2001), a number of
experimental conditions were employed, of which five are
considered here: (1) REST—each subject lay quietly in the
scanner with his/her eyes patched (eyes remained covered
for all conditions); (2) SPEECH—each subject was told to
recount spontaneously a remembered incident from his life
(e.g. some event that occurred during a vacation) using vo-
cal speech with normal speech rate, intonation and rhythm;
(3) ASL—same as (2), except the subject was instructed
to use ASL rather than spoken English to produce a spon-
taneous narrative using normal production rate, extent of
signing space, and rhythm; (4) ORAL
CTRL—each sub-
ject produced self-generated laryngeal and oral articulatory
movements and associated sounds devoid of linguistic con-
tent employing all of the muscle groups activated during
speech; (5) LIMB
CTRL—subjects made self-generated bi-
lateral, non-symmetrical, non-routinized movements of the
hands and arms, along with simple movements of the up-
per and lower face, similar in rate and range as those used
during signing, but lacking linguistic content.
The bilingual subjects were scanned during all five condi-
tions; the monolingual subjects were scanned during REST,
SPEECH, and ORAL
CTRL. The sequence of tasks was
randomized across subjects; the stories that the bilingual
subjects recounted for one language (English or ASL) were
continuations of those produced for the other language. Sub-
jects underwent training in all tasks before PET scanning.
2.1.3. PET scanning and data analysis
Details of the scanning protocol were presented in Braun
et al. (1997, 2001). In brief, data were acquired on a Scan-
ditronix PC2048-15B tomograph (Uppsala, Sweden): 15
contiguous slices with a resolution of 6.5mm FWHM in
all directions. A transmission scan was used to correct for
attenuation. For each scan 30mCi of H
2
O
15
was injected
intravenously. Regional brain radioactivity concentration
was used as an index of regional cerebral blood flow (rCBF)
(Fox, Mintun, Raichle, & Herscovitch, 1984).
Data obtained from PET were analyzed using SPM
(Wellcome Department of Cognitive Neurology, Lon-
don, UK, http://www.fil.ion.ucl.ac.uk/spm/). Preprocess-
ing steps included image registration, spatial smoothing
(15mm × 15mm × 9 mm), and spatial normalization into
the stereotactic space of the Talairach & Tournoux atlas
(1988). In this paper we are only interested in activations in
the region of the inferior frontal gyrus and so we limited our
statistical analysis solely to this area. We performed a series
of pairwise SPM contrasts and retained all voxels with a Z
exceeding 2.33 (which corresponds to P<0.01, one-tailed,
uncorrected). Use of this threshold for our analysis is con-
servative because it protects against claims about differences
in activation for ASL and SPEECH in BA44 and BA45.
1870 B. Horwitz et al. / Neuropsychologia 41 (2003) 1868–1876
2.2. Cytoarchitectonic data analysis
2.2.1. Cytoarchitectonic analysis and probabilistic
mapping
The cytoarchitectonic analysis was presented in detail in
Amunts et al. (1999). Briefly, cytoarchitectonic mapping of
BA44 and BA45 of 10 individual brains (five males, five
females) was accomplished using an observer-independent
method (Schleicher, Amunts, Geyer, Morosan, & Zilles,
1998) that evaluates the gray level index representing vol-
ume fraction of cell bodies along vertical trajectories from
the cortical surface to the white matter. Statistically signifi-
cant changes in the laminar distribution of the gray level in-
dex are used to detect transitions between cytoarchitectonic
areas.
A high resolution MR of each of the 10 individual brains
was transformed (Schormann & Zilles, 1998) into the stan-
dardized space of the European Computerized Human Brain
Database (ECHBD) (Roland & Zilles, 1996), which uses
the coordinates of Talairach space. The probabilistic value
for a particular Brodmann area for each voxel in this space
was computed as the percentage of the 10 subjects having
their cytoarchitectonic data for that Brodmann area mapped
to that voxel. For our purposes, we divided each of the two
Brodmann regions (BA44 and BA45) into a core area con-
Fig. 1. Activations of BA45 (top row) and BA44 (bottom row) during production of language narratives compared to a motor control task. Shown are
representative horizontal slices (left side of each image corresponds to the left side of the brain; the level in mm superior to the AC-PC plane (z-coordinate
of Talairach & Tournoux, 1988) is indicated on each slice). Images displayed in the two columns on the left are from the bilingual (English and ASL)
subjects, and those in the column on the right are from the monolingual English speakers. Voxels in dark blue correspond to core parts of the specific
Brodmann area, those in light blue to peripheral voxels. Voxels significantly more active in one condition compared to a second (Z>2.33) are shown in
green. Voxels in the peripheral part of a Brodmann area that had a significant PET activation are displayed in red, and core voxels that were significantly
activated are shown in yellow.
sisting of voxels that were found in the majority of post-
mortem brains (five or more for BA45; four or more for
BA44, since there was no voxel common to more than eight
brains) and a peripheral area consisting of voxels which
were found in a minority of the postmortem brains (images
of the cytoarchitectonic data can be found at http://www.fz-
juelich.de/ime/ProbabilityMaps
eng.html).
2.2.2. Combining PET and cytoarchitectonic data
The PET data were resliced so that their voxel size
matched the volumes containing the cytoarchitectonic data
(1mm × 1mm× 1mm). The PET data were then mapped
onto the cytoarchitectonic volumes. Differences between
the various templates used for standardized space are
small in the region around Broca’s area [e.g. see (Brett,
Christoff, Cusack, & Lancaster, 2001)(http://www.mrc-
cbu.cam.ac.uk/imaging/mnispace.html) for a comparison of
the brain template of the Montreal Neurological Institute
(MNI) and the Talairach atlas; Indefrey et al. (2001) discuss
the MNI template as compared to the ECHBD template].
Our findings will be illustrated in Figs. 1 and 2, where
the following convention is employed. The cytoarchitectonic
data are shown in blue; dark blue corresponds to core voxels
and light blue to peripheral voxels. Voxels that were signifi-
cantly activated in one task condition relative to another are
B. Horwitz et al./Neuropsychologia 41 (2003) 1868–1876 1871
Fig. 2. Activation of BA45 (top) and BA44 (bottom) comparing each motor control task to a resting condition. See caption to Fig. 1 for the conventions
used.
shown in green. PET activations in core voxels are shown
in yellow and in peripheral voxels in red.
The number and percentage of core voxels that were acti-
vated by the various contrasts of interest in each hemisphere
were computed.
3. Results
Our results are presented in Tables 1 and 2. Shown
are the Talairach z (dorsal–ventral with respect to the
Table 1
Number and percentage of activated core BA45 voxels
Task comparisons Left hemisphere Right hemisphere
Core slices
activated
Total no. of activated core voxels
(percentage of total) (total no. of
core voxels = 1053)
Core slices
activated
Total no. of activated core voxels
(percentage of total) (total no. of
core voxels = 1493)
Bilinguals
SPEECH–ORAL
CTRL 15–24mm 186 (18%) None 0 (0%)
ASL–LIMB
CTRL 7–21mm 346 (33%) None 0 (0%)
ASL–SPEECH None 0 (0%) None 0 (0%)
SPEECH–ASL None 0 (0%) None 0 (0%)
ORAL
CTRL–REST 24–26mm 21 (2.0%) None 0 (0%)
LIMB
CTRL–REST None 0 (0%) None 0 (0%)
Monolinguals
SPEECH–ORAL
CTRL 14–30mm 392 (37%) None 0 (0%)
ORAL
CTRL–REST 16–18, 24mm 16 (1.5%) None 0 (0%)
anterior–posterior (AC–PC) commissure plane)-coordinates
for the slices in which core BA45 (Table 1) and core BA44
(Table 2) were activated in each hemisphere in the vari-
ous contrasts. Also shown are the number and percentage
of core voxels activated for the two cytoarchitectonically
defined regions. Results are presented separately for the
bilingual and monolingual groups.
For BA45 we found consistent core activation in the left
hemisphere when language production, whether produced
verbally or using ASL, was compared to the appropri-
ate motor control task, as illustrated in Fig. 1 (images
1872 B. Horwitz et al. / Neuropsychologia 41 (2003) 1868–1876
Table 2
Number and percentage of activated core BA44 voxels
Task comparisons Left hemisphere Right hemisphere
Core slices
activated
Total no. of activated core voxels
(percentage of total) (total no. of
core voxels = 2193)
Core slices
activated
Total no. of activated core voxels
(percentage of total) (total no. of
core voxels = 1405)
Bilinguals
SPEECH–ORAL
CTRL 32–36mm 16 (0.7%) None 0 (0%)
ASL–LIMB
CTRL 7–21mm 134 (6%) None 0 (0%)
ASL–SPEECH None 0 (0%) None 0 (0%)
SPEECH–ASL None 0 (0%) None 0 (0%)
ORAL
CTRL–REST 6–36mm 1068 (49%) 9–30mm 587 (42%)
LIMB
CTRL–REST 12–29mm 314 (14%) 17–19mm 7 (0.5%)
Monolinguals
SPEECH–ORAL
CTRL 26mm 6 (0.3%) None 0 (0%)
ORAL
CTRL–REST 6–36mm 1402 (64%) 9–28mm 380 (27%)
showing the results for all brain slices for each contrast can
be found at http://www.nidcd.nih.gov/research/scientists/
horwitzb
supp.asp). This was the case for both the bilin-
gual subjects and for the monolingual subjects. Not a single
BA45 core voxel in the right hemisphere was activated in
these contrasts. Although ASL, compared to its motor con-
trol task, appeared to activate almost twice the number of
core voxels as did SPEECH compared to its motor control
task in the bilingual subjects, when the two language pro-
duction tasks (ASL and SPEECH) were compared directly
to each other, no core voxels in BA45 were significantly
more active for one language relative to the other in either
hemisphere.
The findings in BA44 are considerably differentthan those
in BA45. Essentially, very few voxels were activated in ei-
ther hemisphere when SPEECH was contrasted to its motor
control task in both the bilingual and monolingual subjects
(see Table 2 and Fig. 1). There was a small amount of left
BA44 activation for ASL compared to its control task, but,
as a percentage of the total number of core voxels, it was
only one-third that seen for BA45 (an examination of the im-
ages presented in the supplementary material indicates that
the PET voxels that extend into the core part of BA44 for
the ASL-control task contrast arise from the posterior part
of the PET activation centered on BA45). As was the case
for BA45, not a single right hemisphere BA44 core voxel
was activated in either group by either language production
task. When the two language production tasks (ASL and
SPEECH) were compared directly to each other in the bilin-
gual group, no core voxels in BA44 were significantly more
active in either hemisphere for one language versus the other.
The reason for the lack of significant BA44 activation
is clear if we contrast each of the motor control tasks
against REST (see Tables 1 and 2). Although few BA45
core voxels show significant activation, there is extensive
core BA44 activation in the left hemisphere in both bilin-
gual and monolingual subjects. This is especially the case
for the oral motor control task, for which there is also an
extensive activation in core BA44 in the right hemisphere.
Fig. 2 illustrates these findings.
4. Discussion
In this study, we combined probabilistic cytoarchitectonic
maps of Brodmann areas 44 and 45, the major constituents
of Broca’s area, with PET activation data obtained during
language production from subjects who were bilingual for
both spoken English and American Sign Language. We di-
vided each cytoarchitectonic area into core and peripheral
portions, with the core parts consisting of those voxels oc-
curring in the majority of the mapped brains. We found con-
sistent activation of the core part of BA45 for both ASL
and SPEECH in the left hemisphere. There were no signifi-
cant differences in any core or peripheral voxels contrasting
ASL and SPEECH directly, suggesting that the same parts
of BA45 are used for both sign and speech. There was very
little, if any, BA45 core activation during the motor con-
trol tasks, when compared to rest. Comparable results were
found for a group of monolingual speakers. For BA44 in
the left hemisphere, very little core activation was found for
SPEECH in the bilingual and monolingual groups, but there
was a small amount of core BA44 activation for ASL. How-
ever, there was extensive activation of the core BA44 when
comparing both motor control tasks to rest. These results
provide important information about the role of these two
cytoarchitectonic subdivisions in language production.
The advantages of using a probabilistic atlas to better de-
fine the spatial location of PET or fMRI activations has been
commented on by a number of researchers (Binkofski et al.,
2000; Indefrey et al., 2001; Morosan et al., 2001; Roland
& Zilles, 1998). It has been argued that cytoarchitecton-
ics represents a better structural correlate for defining brain
functional fields than does gyral/sulcal location, even though
cytoarchitectonics itself is a limited measure and likely to
be supplemented when more information about the regional
B. Horwitz et al./Neuropsychologia 41 (2003) 1868–1876 1873
distribution of numerous neurochemical markers (e.g. recep-
tors) becomes available (for reviews, see Roland & Zilles,
1998 and Zilles et al., 2002). However, at present cytoar-
chitectonic designations can only be obtained in autopsied
tissue. The probabilistic brain atlas approach, therefore, rep-
resents a workable compromise for assigning cytoarchitec-
tonic designations to brain locations in subjects undergoing
functional brain imaging studies.
We have two key results. First, we found extensive
involvement of BA45 and little or no involvement of