Words in melody: an H2
15O PETstudy of brain
activation during singing and speaking
K. J. Je¡ries, J. B. Fritz and A. R. BraunCA
Language Section,NIDCD,NIH, Building10, Room 5N118A, Bethesda,MD 20892,USA
CACorresponding Author: braund@nidcd.nih.gov
Received 5 November 2002; accepted 28 January 2003
DOI:10.1097/01.wnr.0000066198.94941.a4
We used H2
15O PET to characterize the interaction of words and
melodybycomparingbrain activitymeasuredwhile subjects spoke
or sang the words to a familiar song. Relative increases in activity
during speaking vs singing were observed in the left hemisphere, in
classical perisylvian language areas including theposterior superior
temporal gyrus, supramarginal gyrus, and frontal operculum, as
well as in Rolandic cortices and putamen. Relative increases in ac-
tivity during singing were observed in the right hemisphere: these
were maximal in the right anterior superior temporal gyrus and
contiguous portions of the insula; relative increases associated
with singing were also detected in the right anterior middle
temporal gyrus and superior temporal sulcus, medial and dorsolat-
eral prefrontal cortices, mesial temporal cortices and cerebellum,
as well as in Rolandic cortices and nucleus accumbens. These re-
sults indicate that the production of words in song is associated
with activation of regions within right hemisphere areas that are
not mirror-image homologues of left hemisphere perisylvian
language areas, and suggest that multiple neural networks may be
involved in di¡erent aspects of singing.Right hemisphere mechan-
isms may support the £uency-evoking e¡ects of singing in
neurological disorders such as stuttering or aphasia. NeuroReport
14:749^754c 2003 Lippincott Williams &Wilkins.
Key words: Brain; Music; Lateralization; Right hemisphere; Singing
INTRODUCTION
Clinical investigations in the nineteenth century led to the
concept of lateralization of cerebral function, the superior
capacity of each side of the brain to perform distinct classes
of skilled behaviors [1]. Evaluation of patients with focal
brain lesions led to the notion that regions supporting core
language functions (lexical semantics, syntax, and phono-
logy) are located in the left hemisphere. It was subsequently
shown that the right hemisphere may play a greater role in
processing the prosodic, rhythmic, intonational, or melodic
characteristics of speech [2] and song [3].
Indeed, a number of neuropsychological and neuroima-
ging studies support the notion that the right hemisphere
may play a dominant role in musical processing. Patients
with right hemisphere lesions are likely to be impaired in
musical perception and imagery [4–9]. In contrast, aphasic
patients with left hemisphere lesions often show extraordi-
narily well-preserved musical and vocal capabilities [10].
Neuroimaging studies of the intact human brain have also
generally supported this idea of the predominant role of the
right hemisphere in music [11–13]. However, the majority of
these studies have evaluated musical perception; neuroima-
ging studies of musical production have been limited [14],
and only a few have examined singing [15–17].
A more complete understanding of the brain mechanisms
underlying singing, particularly the singing of words, may
help clarify mechanisms that play a role in the pathophy-
siology and treatment of disorders such as developmental
stuttering and certain types of aphasia, in which singing
may enable fluent articulation. The means by which this
interaction of word and melody generates a beneficial effect
is unclear, although it has been hypothesized that right
hemisphere mechanisms may play a compensatory role,
making it possible to articulate lexical items in these
conditions [18]. Such hypotheses are unproven, and the
issue remains controversial.
The generation of words in song, i.e. the interaction of
words and melody, has not yet been evaluated with
neuroimaging methods. Wildgruber et al. [17] examined
covert rather than overt singing; Perry et al. [16] evaluated
overt singing of a single pitch rather than a melody; and
Riecker et al. [15] studied the production of melody, but in
the absence of words. In order to complement and extend
these studies, we used H2
15O PET to directly compare
patterns of cerebral activity in subjects speaking and singing
the words to a familiar song.
Use of the same lexical material in each case controls for
linguistic features such as syntax or semantics, assuring that
generation of melody, and any potential interaction between
words and melody in singing, remains the principal
variable. We used PET in order to allow continuous vocal
production without confounding artifacts associated with
changes in vocal tract airway volume and resultant
susceptibility effects that can complicate the interpretation
of blood oxygenation level dependent contrast (BOLD) fMRI
studies. Subjects were allowed to sing and speak using
0959-4965c LippincottWilliams &Wilkins Vol 14 No 5 15 April 2003 74 9
BRAIN IMAGING NEUROREPORT
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
natural rate, rhythm, and intonation, and speech and
singing rates were measured and used as covariates in
these analyses. The direct contrast between singing and
speaking was used in order to highlight relative differences
in activity (rather than shared features) in order to test the
following hypotheses: (1) that regions in the left hemisphere
will be more active during speaking, and regions in the right
hemisphere more active during singing, (2) that the latter
will not simply represent homologous portions of the right
perisylvian cortex, but include non-homologous regions
specifically associated with musical production; (3) that
these will include, in addition to more anterior portions of
the auditory association cortices, additional cortical (pre-
frontal, parietal, insular) and subcortical regions (basal
ganglia, cerebellum) that may contribute to the production
of words in song; and finally (4) since our singing task
required subjects to generate complex over-learned melo-
dies, we predicted that additional areas would be recruited,
particularly in the superior temporal gyrus and medial
prefrontal cortex, regions that appear to be involved in
auditory long-term memory.
MATERIALS AND METHODS
Subjects: Subjects included eight females 367 10
(mean7 s.d.) years of age, range 24–50, and 12 males
337 8, range 23–47. All subjects were right handed and free
of medical or psychiatric illness on the basis of history and
physical examination, baseline laboratory evaluation, and
structural MRI. None had received music theory or formal
voice training. Written informed consent was obtained from
all subjects. The study was approved by the NIDCD/
NINDS Institutional Review Board.
Speech and singing tasks: Subjects were scanned while
(1) reciting and (2) singing words to a familiar song. The
sequence of scans was randomized across subjects, all of
whom underwent practice prior to the PET study. Subjects
selected an overlearned song (e.g. happy birthday), and sang
it at a comfortable rate and volume. When the same song was
spoken, subjects were instructed to speak at a natural rate,
keeping rhythm and intonation approximately the same as
during conversational speech. Scanning sessions were re-
corded and transcribed, a computer-generated signal identi-
fying the start of the scan. One-minute samples, from 15 s
prior to 45 s after the start of scan, were used to calculate
speech and singing rates (syllables/s).
Scanning methods: PET scans were performed on a
Scanditronix PC2048-15B tomograph (Uppsala, Sweden)
with axial and in-plane resolution of 6.5 mm. Subjects’ eyes
were patched, and head motion was restricted with a
thermoplastic mask that permitted free movement of
the oral articulators. For each scan, 30 mCi of H2
15O
were injected intravenously. Speech and singing tasks were
initiated 30 s prior to injection of the radiotracer and were
continued throughout the scanning period. Scans com-
menced automatically when the count rate in the brain
reached a threshold value (B20 s after injection) and
continued for 4 min. Studies were separated by 10 min
intervals. Emission data were corrected for attenuation by
means of a transmission scan.
PET data analysis: PET scans were registered and stereo-
taxically normalized using Statistical Parametric Mapping
software (Wellcome Department of Cognitive Neurology,
London, UK). Images were smoothed with a Gaussian filter
(15 15 9 mm in x, y and z axes) to accommodate
intersubject differences in anatomy, and spatially normal-
ized to produce images in a common stereotaxic (Talairach)
space. Differences in global activity were controlled for by
proportional normalization.
Using SPM, we contrasted singing and speaking condi-
tions using a multiple subjects with conditions and
covariates design with rate entered as a confounding
covariate. Tests of significance based on the size of the
activated region [19] were performed; local maxima from
the task contrast were considered most reliable when these
were included with a cluster of significant spatial extent
(instances in which differences did not satisfy this criterion
are so indicated in Table 1).
RESULTS
Speech and singing rates differed significantly (speaking,
3.057 0.55 syllables/s; singing, 2.617 0.81 syllables/s,
p¼ 0.01, paired t-test). In order to eliminate the contribution
of rate to apparent task-related differences, it was included
in the SPM contrasts as a confounding covariate.
These contrasts demonstrated that, in general, regions
within the left hemisphere were more active for speaking
and regions in the right were more active for singing. This
hemispheric dissociation was observed both in homologous
and non-homologous regions (Fig. 1; Table 1).
In the prefrontal cortex, the left operculum was more
active for speaking. The left and right medial prefrontal
cortices and contiguous right superior dorsolateral prefron-
tal cortex were more active for singing.
In Rolandic cortices, the secondary somatosensory area
(SII) was more active on the left for speaking, on the right
for singing. Other portions of the left Rolandic cortex,
including both pre- and postcentral gyri, were also more
active for speaking.
In the temporal lobe, posterior portions of the left
superior temporal gyrus (STG), in the vicinity of the planum
temporale were more active for speaking. For singing,
activity was greater in anterior regions on the right,
prominent in the anterior middle temporal gyrus (MTG)
and superior temporal sulcus (STS) and maximal in the
anterior STG and contiguous insula.
Speaking was also associated with relative increases in
activity in the left supramarginal gyrus, and singing with
increases in the right fusiform and parahippocampal gyri,
caudal orbital and posterior cingulate cortices, midline
anterior cerebellum, and bilateral increases in the lingual
gyri and cerebellar vermis.
In the basal ganglia, relative increases in activity in the left
putamen (extending to the anterior insula) and the right
nucleus accumbens were seen during speaking and singing
respectively.
Analysis of spatial extent for speakingsinging revealed a
single cluster of significant spatial extent: cluster 1: 1861
voxels; p (nmax¼k)¼ 0.00002; (left perisylvian, inferior
parietal, rolandic cortices). Analysis of spatial extent for
singingspeaking revealed three significant clusters: cluster
750 Vol 14 No 5 15 April 2003
NEUROREPORT K. J. JEFFRIES, J. B.FRITZ ANDA.R.BRAUN
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
2: 1854 voxels; p(nmax¼ k)¼ 0.00002; (mesial, basal tempor-
al cortex, midline cerebellum); cluster 3: 661 voxels;
p(nmax¼k)¼ 0.04; (right anterior temporal, Rolandic cor-
tices); cluster 4: 817 voxels; p(nmax¼ k)¼ 0.01; (right
dorsolateral prefrontal cortex, medial prefrontal cortices).
DISCUSSION
Our results demonstrate clear differences in the contribu-
tions of the left and right hemispheres to sung and spoken
language: the left hemisphere is relatively more active for
speech, the right more active for singing. Beyond confirming
hemispheric lateralization, we observed conspicuous differ-
ences in the functional nature of regions associated with
either condition.
Clinical correlations: Previous neuropsychological studies
have demonstrated that areas in the right hemisphere are
involved in the perception, imagery, and memory of pitch,
timbre, and melody [5–9], and neuroimaging studies also
emphasize the importance of frontal and temporal areas in
the right hemisphere during musical perception [11–13].
Our results are in agreement with these studies, while
indicating greater activation in areas related to motor
function and vocal self-monitoring during the production
of song. Clinical evidence also supports the notion that the
right hemisphere plays an important role in musical
production, since singing is impaired following lesions,
anesthesia, or transcranial stimulation of the right hemi-
sphere [3,18,20,21].
Previous neuroimaging studies of singing: The design of
the present study differs from earlier neuroimaging studies
that evaluated overt singing [15,16]; but our design also
complements these studies, and our findings both confirm
and extend their results. Perry et al. [16] evaluated singing of
a sustained pitch, comparing this to tone perception. The
use of a perceptual baseline in this contrast effectively
isolates the motor features of singing. In the present study,
the direct comparison of speech and song production
selectively isolates the generation (and self monitoring) of
melody, since essential motor features such as vocalization
and articulation are common to both conditions.
Riecker et al. [15] studied overt production of melody, but in
the absence of words, comparing recitation of months of the
year to production of a non-lyrical tune. In our study we
Table 1. Regions in which normalized rCBF rates are greater during speaking vs singing or during singing vs speaking are shown, along with Z scores
representing localmaxzima, and Talairach coordinates.
Region Brodmann
area
Speaking Singing
Left hemisphere Right hemisphere Left hemisphere Right hemisphere
Z score x y z Z score x y z Z score x y z Z score x y z
Subcortical
Cerebellum
Cerebellar vermis 3.03 10 72 8b 3.12 2 52 8b
Anterior cerebellum 3.09 16 36 20b
Basal ganglia
Putamen/insula 2.99 28 16 4a
Nucleus accumbens 3.52 8 16 8
Prefrontal
Middle frontal operculum 45 3.19 44 24 8a
Superior frontal operculum 44/6 3.40 50 6 24a
Dorsolateral prefrontal
cortex
8 3.37 26 36 40d
Medial prefrontal cortex 9/10 3.65 6 58 20d 3.39 18 52 32d
Perirolandic
SII 43 3.16 52 10 20a 3.68 46 18 16c
Postcentral gyrus 3,1,2 3.68 48 22 32a
Precentral gyrus 6 3.03 48 2 24a
Temporal
Anterior STG/insula 22 4.20 48 6 0c
Anterior MTG/STS 21 3.34 58 8 4c
Posterior STG/PT 22 3.13 54 50 20a
Basal temporal, occipital
Fusiform/parahippocampal
gyrus
37 3.03 20 52 4b
Cuneus/lingual gyrus 18 2.99 2 84 4b 3.09 2 88 8b
Parietal
Supramarginal gyrus 3.34 60 38 24a
Proisocortical
Posterior cingulate 23/31 3.17 14 52 16b
Caudal orbital cortex 25/32 3.09 2 16 8
Superscripts indicate clusters of signi¢cant spatial extent (see text) inwhich localmaxima are included.
aCluster1,1861voxels; p(nmaxZk)¼ 0.0002;
bCluster 2,1854 voxels; p(nmaxZk)¼ 0.0002.
cCluster 3, 661voxels; p(nmaxZk)¼ 0.04;
dCluster 3, 817 voxels; p(nmaxZk)¼ 0.01.
Vol 14 No 5 15 April 2003 751
PET STUDYOF SINGINGAND SPEAKING NEUROREPORT
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
evaluate the production of song in words. Use of the same
lexical material in both conditions insures that differences in
brain activation do not reflect task-related linguistic features,
but are due instead to the interaction of word and melody. As
noted earlier, this may have clinical relevance for disorders in
which singing words enables fluent articulation.
Hemispheric lateralization: Regions more active during
speaking vs singing words were lateralized to the left
hemisphere and include areas that are constituents of the
classical perisylvian language system, i.e. left frontal
operculum, posterior superior temporal gyrus and planum
temporale [22,23]. Speaking was also associated with
increased activity in the left supramarginal gyrus, which
may constitute an auditory motor interface for spoken
language production [24].
Singing, on the other hand, was associated with greater
activity in a wide array of regions within the right
hemisphere. Some of these represent homologues of left
hemisphere areas in which activity was greater during
speech. However, the majority, e.g., right prefrontal,
temporal, paralimbic, and subcortical areas, do not, and
may instead constitute elements of a functionally distinct
system within the right hemisphere.
These patterns may be interpreted in a number of ways.
The relative increases in left hemisphere perisylvian activity
may not simply reflect enhanced activity in these regions
during speaking of words, but may instead represent
relative decreases in activity manifest when words are
sung. That is, singing may involve selective activation of
right hemisphere regions and concomitant suppression of
activity in left hemisphere perisylvian areas that are
normally more active during speech production. Conver-
sely, activity in the right hemisphere system may be
suppressed when the left hemisphere is more strongly
engaged in the production of spoken language [25].
Lateralization in homologous brain regions: The homo-
logous regions within left and right hemispheres that were
more active for speaking or singing are most closely
associated with sensorimotor function.
In the Rolandic cortices, maximal differences were found
in the secondary somatosensory cortex (SII, BA 43) in which
activity was greater in the left hemisphere for speaking, in
the right for singing. The idea that lateralization for speech
and singing is manifest at the level of primary sensorimotor
cortices is consistent with all three of the neuroimaging
studies cited above [15–17]. Task related differences were
also seen in homologous portions of the insula, with the
right more active for singing, the left more active for speech.
Both Riecker et al. [15] and Perry et al. [16] reported
differences in insular activity related to singing: Perry et al.
reported bilateral insular activation when simple singing
was compared with complex tone perception, consistent
with a contrast highlighting motor activity per se. Our
results, on the other hand, are more consistent with those of
Speaking - Singing
Singing - Speaking
z = +8 +15 +20 +25 +38
z = −12 0 +16 +25 +38
+4.0
+2.0
Fig. 1. Brain maps depicting di¡erences in regional cerebral blood £ow (rCBF) during speaking and singing the words to a familiar song.The top row
illustrates signi¢cant elevations in rCBF during speaking (vs singing); the bottom row, elevations during singing (vs speaking). Statistical parametric maps
resulting from these analyses are displayed on a standardizedMRI scan, whichwas transformed linearly into the same stereotaxic (Talairach) space as the
SPM {z} data. Scans are displayed using neurological convention (left hemisphere is represented on the left). Planes of section relative to the anterior
commissural-posterior commissural line are indicated for each contrast.Values are Z-scores representing the signi¢cance level of voxel-wise changes in
normalized rCBF for each contrast.The range of scores is coded in the accompanying color tables. Signi¢cant di¡erences correspond to local maxima
summarized inTable1.
752 Vol 14 No 5 15 April 2003
NEUROREPORT K. J. JEFFRIES, J. B.FRITZ ANDA.R.BRAUN
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Reicker et al., who also reported left insular activation for
speech and right insular activation for singing.
The insula is an important convergence zone, that appears
to be involved in relatively direct processing of auditory
input and serves as a parallel relay from temporal to frontal
motor and other higher order association areas [26]. One
such projection, to the frontal operculum, may, in the left
hemisphere, subserve the role proposed for the insula in
speech planning and articulation [27]. Our results suggest
that the right insula might similarly participate in coordina-
tion of the oral-articulatory and laryngeal musculature for
the generation of words in melody.
The possibility that singing words is associated with more
direct control of oral motor processes by the right hemi-
sphere could account for the fact that singing can enable
fluent articulation in developmental stuttering (there is
evidence that a shift to right insular activation may underlie
fluent speech production in these individuals [28]) as well as
in patients with Broca’s aphasia resulting from left
perisylvian lesions that include the insula.
Lateralization in non-homologous brain regions: Latera-
lized differences between singing and speaking were more
numerous in non-homologous regions of the brain. For
example, such differences were manifest in distinct areas
along the anterior-posterior axis of the temporal lobe. While
activity in the left posterior superior temporal regions was
greater for speaking, right anterior temporal regions, includ-
ing both STG, MTG, and intervening STS, were more active
for singing. These observations are consistent with Parsons
[14], who reported activation of right anterior temporal areas
during musical (keyboard) performance. In addition, Perry
et al. [16] reported right lateralized activity in the auditory
cortices during simple singing, although the local maxima in
that study were more posterior than we observed here.
The anterior portions of temporal cortex, including the STS
(lateralized to the right hemisphere, in approximately the
regions we report) have been shown to be activated for voice
perception [29,30]. However, since voice is produced in both
of our conditions, our results suggest that right anterior
temporal regions may be more actively involved in self-
monitoring of sung as opposed to spoken words, perhaps
playing a role in the intricate feedback control of pitch and
rhythmic phrasing during the production of melody.
A similar left-right dichotomy was detected in non-
homologous, functionally distinct portions of the prefrontal
cortex. While left opercular areas were more active for
speaking, their homologues in the right hemisphere did not
appear to be more active for singing (unlike the results of
the Perry et al. study). Instead, we found that singing words
was associated with increased activity in the medial
prefrontal cortex (MPF) and contiguous portions of the
right superior dorsolateral prefrontal cortex.
The dorsolateral prefrontal cortex plays a role in temporal
sequencing of behavior [31] and, in the right hemisphere,
could support a timing mechanism for production of words
in song. With respect to the medial prefrontal cortex, it is
interesting that the adjacent paramedian cortices, with
which the MPF may share certain functional characteristics,
have been implicated in control of phonation: the anterior
cingulate cortex, for example, appears to regulate elicitation
of species-specific calls in lower mammalian species [32]
and stimulation of the contiguous pre-supplementary motor
area produces vocalization in humans [33,34]. If the MPF
plays a similar role in humans, relative increases in activity
in this region may be related to more precise control of
phonation required during singing. In this context it is
interesting that the MPF receives dense auditory projections
from the anterior STG [35], which we have shown is also
significantly more active during singing. In addition, both
regions have been implicated in auditory memory [36].
Hence another possibility is that the activation of the MPF
may reflect its role in auditory memory for the familiar
melodies chosen by our subjects. Interestingly, this portion
of the prefrontal cortex also has the richest connections with
the limbic system [37].
Involvement of the limbic system in singing is also
supported by findings in the basal ganglia. In these nuclei,
which have been shown to play a role in speech motor
control [38], we observed relative elevations in the left
hemisphere for speech, and in the right for singing. However,
for speech, relative increases were found in the left
dorsal putamen, which lies at the center of the motor circuit
[39]. Singing was associated with increases in activity, not in
the contralateral putamen, but in the right ventral striatum,
particularly in the nucleus accumbens, at the center of the
cortico-limbic circuitry. This area has been shown to be
activated during pleasurable emotional states induced by
familiar, self-selected musical passages [40] which may have
been evoked by the songs chosen by our subjects.
Lateralized increases in activity selectively associated with
singing: A role for limbic structures is also supported by
activation of paralimbic regions, including the right para-
hippocampal gyrus, during singing. In the monkey, this region
is known to receive the largest auditory cortical input of all the
mesial temporal areas [41]. It is interesting in this context that
the hippocampus and parahippocampal gyri may play a
role in the detection of musical consonances and dissonances
[42,43], suggesting that they may participate in self-monitoring
during the production of song. Relative increases associated
with singing were found in other limbic-related regions,
including the right posterior cingulate cortex which, in the
monkey has been shown to receive a large projection from
secondary auditory association areas as well [44].
It should be noted that a subset of the mesial temporal
regions in which relative increases were detected during
singing, i.e. fusiform and lingual gyri, also constitute extra-
striate visual association areas, consistent with extrastriate
activations reported by Perry et al., who suggest that cross-
modal processes might be involved in the production of song.
The importance of the cerebellum in singing has been
established clinically: patients with cerebellar lesions often
demonstrate unsteady vocal pitch and impairments in the
perception of temporal features of auditory stimuli [45]. Our
results indicate that the cerebellar vermis was selectively
more active during singing, an observation consistent with
the findings of Perry et al. (but inconsistent with Riecker
et al., who observed reciprocal activations for singing and
speaking in more lateral portions of the cerebellum).
It has been suggested [46] that the vermis provides a
circuitry though which sensory systems extract temporal
information, enabling precisely timed motor responses. The
singing of words (rather than the generation of melody
Vol 14 No 5 15 April 2003 753
PET STUDYOF SINGINGAND SPEAKING NEUROREPORT
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
alone as in the Riecker et al. study), requires continuous
alteration of voiced and voiceless segments and precise
adjustments of stress, pitch, volume, and rhythm, features
that may be mediated by, and place greater demands upon,
the cerebellar circuitry.
Multiple networks for singing: The various cortical and
subcortical regions that are activated during singing may in
fact represent elements of large-scale, distributed networks
that support the production of words in melody. Although
the present analysis does not demonstrate that regions are
functionally coupled in such a fashion, they can be
heuristically grouped into systems that appear to play a
role in: (1) fine motor control of intonation, pitch and
volume during the generation of melody (medial prefrontal
cortex, Rolandic cortices, anterior insula, and cerebellum)
[14–17,26,32]; (2) auditory feedback for self-monitoring of
pitch, volume and rhythmic phrasing (parahippocampal
and anterior middle temporal cortices, anterior STG, STS,
and cerebellum) [15,16]; (3) melodic memory (anterior STG,
parahippocampal and other mesial temporal cortices,
medial prefrontal cortex, and posterior cingulate cortex)
[7,35,36,41]; (4) emotional responses generated during song
production (parahippocampal gyrus, nucleus accumbens,
and associated limbic regions) [40,43]. Additional studies
will be necessary to test the validity and generality of the
hypothesis that these regions operate as functional networks
underlying the production of song.
CONCLUSION
Singing words and speaking words are associated with
lateralized differences in cerebral activity: regions in the
right hemisphere are more active during singing, regions in
the left hemisphere more active during speaking. Relative
increases are detected in homologous portions of the left
and right hemispheres, in regions typically associated with
sensorimotor function. These asymmetries suggest that oral-
laryngeal motor activity may be more directly controlled by
regions in the right hemisphere (including Rolandic and
insular cortices) when words are sung. This pattern may
provide insight into the nature of neurological conditions
such as stuttering and aphasia in which singing can induce
fluency. Singing also appears to engage right hemisphere
systems that are not homologues of left hemisphere motor
or language areas. For example, while speaking is associated
with greater activity in left perisylvian regions, singing is
associated with increased activity in right anterior temporal,
prefrontal, and paralimbic cortices (regions which are also
anatomically interconnected). Thus, functionally distinct
networks within the right hemisphere may underlie
production of words in melody. Regions such as the right
anterior STS and cerebellar vermis may be involved in self-
monitoring and feedback guided regulation of singing,
which may require more precise adjustments of vocal pitch,
volume and rhythmic phrasing. Activation of these regions
may also support the fluency-inducing effects of words
produced in melody.
REFERENCES
1. Geschwind N and Galaburda AM. Cerebral Lateralization: Biological
Mechanisms, Associations, and Pathology. Cambridge, MA: MIT Press;
1987, p. 5.
2. Ross ED. Neurol Clin 11, 9–23 (1993).
3. Alexander MP, Benson DF and Stuss DT. Brain Lang 37, 656–691 (1989).
4. Milner B. Laterality effects in audition. In: Mountcastle VB (ed.).
Interhemispheric Relations and Cerebral Dominance. Baltimore: Johns
Hopkins Press; 1962, pp. 177–195.
5. Robin DA, Tranel D and Damasio H. Brain Lang 39, 539–555 (1990).
6. Liegeois-Chauvel C, Peretz I, Babai M et al. Brain 121, 1853–1867 (1998).
7. Zatorre RJ and Samson S. Brain 114, 2403–2417 (1991).
8. Zatorre RJ and Halpern AR. Neuropsychologia 31, 221–232 (1993).
9. Samson S and Zatorre RJ. Neuropsychologia 32, 231–240 (1994).
10. Kaplan JA and Gardner H. Artistry after unilateral brain disease. In:
Goodglass H and Damasio AR (eds). Language, Aphasia and Related
Disorders. Amsterdam: Elsevier; 1990, pp. 141–155.
11. Halpern AR and Zatorre RJ. Cerebr Cortex 9, 697–704 (1999).
12. Zatorre RJ, Evans AC and Meyer E. J Neurosci 14, 1908–1919 (1994).
13. Zatorre RJ, Halpern AR, Perry DW et al. J. Cogn Neurosci 8, 29–46 (1996).
14. Parsons LM. Ann NY Acad Sci 930, 211–231 (2001).
15. Riecker A, Ackermann H, Wildgruber D et al. Neuroreport 11, 1997–2000
(2000).
16. Perry DW, Zatorre RJ, Petrides M et al. Neuroreport 10, 3453–3458
(1999).
17. Wildgruber D, Ackermann H, Klose U et al. Neuroreport 7, 2791–2795
(1996).
18. Cadalbert A, Landis T, Regard M et al. J Clin Exp Neuropsychol 16, 664–670
(1994).
19. Friston KJ, Worsley KJ, Frackowiak RS et al. Hum Brain Mapp 1, 210–220
(1994).
20. Henson RA. Amusia. In: Vinken PJ, Bruyn GW and Klawans HL (eds).
Clinical Neuropsychology. Amsterdam: Elsevier; 1985, pp. 483–490.
21. Epstein CM, Meador KJ, Loring DW et al. Clin Neurophysiol 110, 1073–1079
(1999).
22. Geschwind N. Brain 88, 237–294 (1965).
23. Geschwind N. Specializations of the human brain. In: Geschwind N (ed.).
The Brain. San Francisco: W.H. Freeman; 1979.
24. Hickok G. J Psycholinguist Res 30, 225–235 (2001).
25. Numminen J, Salmelin R and Hari R. Neurosci Lett 265, 119–122 (1999).
26. Mufson EJ and Mesulam MM. J Comp Neurol 212, 23–37 (1982).
27. Dronkers NF. Nature 384, 159–161 (1996).
28. Braun AR, Varga M, Stager S et al. Brain 120, 761–784 (1997).
29. Belin P, Zatorre RJ, Lafaille P et al. Nature 403, 309–312 (2000).
30. Scott SK, Blank CC, Rosen S et al. Brain 123 Pt 12, 2400–2406 (2000).
31. Ferreira CT, Verin M, Pillon B et al. Cortex 34, 83–98 (1998).
32. Jurgens U. Neurosci Biobehav Rev 26, 235–258 (2002).
33. Penfield W and Roberts L. Speech and Brain–-Mechanisms. Princeton, NJ:
Princeton University Press; 1959, p. 286.
34. Fried I, Katz A, McCarthy G et al. J Neurosci 11, 3656–3666 (1991).
35. Munoz M, Mishkin M and Saunders RC. Soc Neurosci Abstr 27, 1415
(2001).
36. Barbas H, Ghashghaei H, Dombrowski SM et al. J Comp Neurol 410,
343–367 (1999).
37. Ongur D and Price JL. Cerebr Cortex 10, 206–219 (2000).
38. Murdoch BE. Folia Phoniatr Logop 53, 233–251 (2001).
39. Alexander GE, Crutcher MD and DeLong MR. Prog Brain Res 85, 119–146
(1990).
40. Blood AJ and Zatorre RJ. Proc Natl Acad Sci USA 98, 11818–11823
(2001).
41. Suzuki WA and Amaral DG. J Comp Neurol 350, 497–533 (1994).
42. Wieser HG and Mazzola G. Neuropsychologia 24, 805–812 (1986).
43. Blood AJ, Zatorre RJ, Bermudez P et al. Nature Neurosci 2, 382–387
(1999).
44. Yukie M. Neurosci Res 22, 179–187 (1995).
45. Ivry R and Keele S. J Cogn Neurosci 1, 136–152 (1989).
46. Penhune VB, Zattore RJ and Evans AC. J Cogn Neurosci 10, 752–765
(1998).
754 Vol 14 No 5 15 April 2003
NEUROREPORT K. J. JEFFRIES, J. B.FRITZ ANDA.R.BRAUN
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.