Texas Music Education Research

21

Neural Basis of the Comprehension of

Peter T. Fox and Larry M. Parsons

Donald A. Hodges
University of Texas at San Antonio

Scientists have been interested for some time in the phenomenon of music in the brain, and a
body of research is beginning to emerge (see Hodges, 1996). Only recently, however, have very
powerful technologies been developed that allow for the monitoring of brain activations during
musical tasks. Magnetic resonance imaging (MRI) and positron emission tomography (PET) are
two such advances. MRI provides detailed information about anatomical structures, whereas
PET provides information about function. By using the two in combination, one can map
function onto location and thus learn what is going on and where it is taking place.
In a previous experiment, we used MRI and PET to monitor brain activations during piano
performance (Fox, Parsons, Hodges, & Sergent, 1995). Eight professional pianists performed
three tasks by (a) playing a portion of Bach’s Italian Concerto from memory (Bach), (b) playing
scales up and down the keyboard (Scales), and (c) laying quietly (Rest). The study produced
four findings:
1. During the performance of Bach, the primary brain activation was in motor areas
having to do with control of complex movement patterns (i.e., moving the arms,
hands, and fingers).
2. Pianists showed strong areas of brain deactivation; that is, there were clearly
identified areas of the brain that were less active while playing Bach than they
were at rest.
3. Certain areas of the brain were more active on the left side of the brain during
Scales and others were more active on the right side during Bach.
4. There are auditory projection areas (related to hearing) in the cerebellum
Rather than continue to investigate motoric aspects of musical performance, we were interested
in more cognitive aspects. Therefore, The purpose of this experiment was to determine neural
mechanisms involved in melodic, harmonic, and rhythmic error detection.
Fox, Parsons, & Hodges
22

Method

Subjects for this experiment were eight professors of music from a state university music
department. Each subject had an earned doctorate in music and had had many years of
professional experience. All were right-handed males. Pilot testing indicated that all were able
to do the required tasks at a high level of competency.
Subjects were asked to compare aural and visual versions of Bach chorales. Using Finale, a
computer-notation program, eight Bach chorales were notated SATB in open score (i.e., with
each voice—soprano, alto, tenor, bass—on a different staff). A second version of each chorale
was prepared with the intentionally-implanted errors. MIDI files were created to allow playback
of the error-filled version. Finale allows one to view the correctly notated version, while
listening to the modified version.
Subjects completed 5 tasks, each one repeated for a total of 10 trials. There were 2 trials each
of:
Rest—subjects merely remained passive during the scan time with no particular
mental focus; this was used to establish a baseline of brain activations.
Passive Listening—subjects listened to a Bach chorale and looked at a fixation point.
This engaged auditory, but not visual, systems.
Melodic Errors—subjects read a correctly notated Bach chorale while they heard a
performance with intentionally implanted errors in the melody. They were asked
to covertly identify these melodic errors.
Rhythmic Errors—subjects read a correctly notated Bach chorale while they heard a
performance with intentionally implanted rhythmic errors. They covertly
identified rhythmic errors that could appear in any of the four voices.
Harmonic Errors—subjects read a correctly notated Bach chorale while they heard a
performance with intentionally implanted errors in any of four voices. Once
again, they were asked to covertly identify these errors.
A crucial aspect was the difficulty level of the tasks. Errors that were too easy to detect (e.g.,
non-diatonic tones) or too difficult (e.g., a simple revoicing of a chord) were used sparingly. An
overall task that was too easy would not engage musical processing systems sufficiently. An
impossibly difficult task might likewise insufficiently engage the relevant systems (e.g., if a
subject merely gave up and did not try). A pilot study was conducted to arrive at the appropriate
level of difficulty.
The pilot study provided data demonstrating that the subjects were able to do the task. This
was important, because during the PET monitoring subjects would to make no overt actions.
Rather, in the melodic, harmonic, and rhythmic trials, subjects mentally recognized when there
were mismatches between what they saw and what they heard. This was done so that motor
systems in the brain would not be activated.
During the collection of MRI data, subjects lay on a narrow bed with a plastic “cage” placed
over the head, and were then pulled into a tube-like device for approximately 45 minutes. They
were given a call button to push in case of a panic attack. A research assistant monitored the
entire procedure via a television hook-up and could speak with the subject through headphones.
Subjects simply had to lie passively; no tasks were required. Subjects wore padding over the
ears to lessen the noise of the MRI equipment.
PET data were obtained in another set of rooms. Subjects again reclined on a narrow bed.
This time, however, the top of the head was at the opening of a large donut-shaped machine.
 Texas Music Education Research
23

Subjects wore a plastic face mask that had an opening so that they could easily view a computer
monitor suspended overhead. The purpose of the mask was to eliminate extraneous head
movements.
An IV was inserted, which allowed for a minute amount of radioactively-tagged water to be
introduced into the blood stream. Each dose was started just as each new task was begun. Each
task lasted for a little more than a minute. This allowed for an uptake of the radioactive material
into the brain via the bloodstream. Since more active parts of the brain need more oxygen, PET
can trace where the blood is drawn; this is known as regional cerebral blood flow (RCBF). Data
are reported both numerically and visually in the form of a color-coded brain map indicating
various areas of activation (more active than at rest) and deactivation (less active than at rest).
PET data were mapped onto MRI data for precise location. Data analyses were done by a
subtractive technique. This means that activations in one task were subtracted from activations
in another. For example, Melodic Error Detection minus Passive Listening shows brain
activations necessary for melodic error detection beyond what is required for merely listening to
music.

Results

The results show that melodic, rhythmic, and harmonic error detection involved many regions
of the brain in a distributed network. Score-reading and listening for melody, harmony, and
rhythm were supported by processes in both cerebral hemispheres. Melody activated each
hemisphere equally, whereas harmony and rhythm activated more of the left than the right
hemisphere. All tasks activated a lateral premotor area, although at different heights of the
brain—rhythm superior, harmony middle, and melody inferior. Also, the three tasks activated
the bilateral, more strongly left, inferior lateral frontal cortex, although again in different
subareas within that region. These two areas are known to be involved in sequencing motor
behavior and may in this context be involved in translating the notes to motor-auditory codes.
The inferior parietal region was also activated bilaterally in each task, with the rhythm and
melody conditions activating those areas more strongly in the right hemisphere and the harmony
conditions activating those areas more strongly in the left hemisphere. This activation could be
involved in performing a mapping between music notation and its corresponding sounds or
melody.
During the melody condition, there were strong activations in bilateral superior temporal and
bilateral, strongly right, middle temporal areas. These superior and middle temporal areas are
secondary auditory association areas. There was no activation in temporal areas for harmony;
and the rhythm condition produced only minor, left inferior activation and bilateral middle
temporal activation. Superior temporal activation for melodic, but not rhythmic or harmony
processing, is consistent with neuropsychological data showing that superior temporal lesions
spare rhythmic processing but cause impaired melodic perception. Furthermore, activation of
these secondary auditory regions bilaterally is consistent with other neuroimaging data indicating
bilateral activation of those areas during the perception or performance of musical pieces and
scales (Fox, Parsons, Hodges, & Sergent, 1995).
The melody condition also produced major activation foci in bilateral subgyral medial frontal
area, which is perhaps related to attentional processing. In the rhythm task, only modestly-sized
foci appeared outside the cerebellum, where the majority of the activation was present. The
former activation foci were in bilateral mostly left anterior cingulate, an area likely related to
attentional processing. The rhythm condition also activated bilateral thalamus, possibly relaying
Fox, Parsons, & Hodges
24

sensory processing information with respect to stimulus timing between cerebellar and cerebral
cortical regions.
Overall, the observed distributed pattern of activations outline the neural systems subserving
musical comprehension. This variation in location of activation suggests that the different
regions or subareas perform different functions for different tasks.
The results also show that melodic, harmonic, and rhythmic error detection were each
supported by a unique neural network. While the network for each of the tasks was discrete
from the others, each condition frequently activated different sub-areas of the same major brain
area. The data also revealed an area in the right half of the brain (fusiform gyrus) that interprets
written notes (and passages of notes) and that corresponds in location to the area in the left half
of the brain known to interpret written letters and words.
There was strong activation in the cerebellum, regardless of the aspect of music to which
subjects attended. Such activation, in the absence of motor behavior, contradicts the
predominant theory of the function of the cerebellum (namely, that it coordinates fine motor
behavior). Activation of the cerebellum in the rhythmic condition was twice that in the melody
and harmony conditions. This association of the cerebellum with the rhythm task is consistent
with neurological data suggesting cerebellar lesions cause impaired perception of temporal order
(Shepherd, 1994).

Discussion

These results should put to rest any lingering notions that there is a single “music center” in the
brain, that music is only in the right side of the brain, or any other such simplistic notions.
Instead, this supports the contention that music involves “an extensive neural system (or
systems) involving widely distributed, but locally specialized, regions of the brain” (Hodges,
1996, p. 260). Imagine the complexity involved when an experienced conductor, for example, is
listening for melodic, rhythmic, and harmonic errors, as well as for many other problems (e.g.,
intonation, tone quality, dynamics, phrasing, articulation). One can only wonder at the
activations in Barenboim's brain while conducting a Mahler symphony.

References

Fox, P., Parsons, L., Hodges, D., & Sergent, J. (1995). Piano performance from memory: A PET Study. Research
presented at the Human Brain Mapping Conference, Paris.
Hodges, D. (1996). Neuromusical research: A review of the literature. In D. Hodges (Ed.), Handbook of Music
Psychology (pp. 197-284). San Antonio, TX: IMR Press.
Shepherd, G. (1994). Neurobiology (3rd ed.). New York: Oxford University Press.