ORIGINAL RESEARCH

published: 16 May 2022

doi: 10.3389/fnhum.2022.866256

Music, Math, and Working Memory:

Magnetoencephalography Mapping
of Brain Activation in Musicians
Ching-I Lu 1* , Margaret Greenwald 1,2 , Yung-Yang Lin 3,4* and Susan M. Bowyer 2,5,6
1
Department of Communication Sciences and Disorders, Wayne State University, Detroit, MI, United States, 2 Department
of Neurology, Wayne State University, Detroit, MI, United States, 3 Institute of Brain Science and Institute of Clinical Medicine,
National Yang Ming Chiao Tung University, Taipei, Taiwan, 4 Department of Critical Care Medicine, Taipei Veterans General
Hospital, Taipei, Taiwan, 5 Department of Neurology, Henry Ford Health System, Detroit, MI, United States, 6 Department
of Physics, Oakland University, Rochester, MI, United States

Musical transposing is highly demanding of working memory, as it involves mentally

converting notes from one musical key (i.e., pitch scale) to another key for singing or
instrumental performance. Because musical transposing involves mental adjustment of
notes up or down by a specific amount, it may share cognitive elements with arithmetical
operations of addition and subtraction. We compared brain activity during high and
low working memory load conditions of musical transposing versus math calculations
Edited by: in classically trained musicians. Magnetoencephalography (MEG) was sensitive to
Vasil Kolev,
Institute of Neurobiology, Bulgarian
differences of task and working memory load. Frontal-occipital connections were highly
Academy of Sciences (BAS), Bulgaria active during transposing, but not during math calculations. Right motor and premotor
Reviewed by: regions were highly active in the more difficult condition of the transposing task. Multiple
Nadia Justel,
frontal lobe regions were highly active across tasks, including the left medial frontal
Interdisciplinary Laboratory
of Cognitive Neuroscience, Argentina area during both transposing and calculation tasks but the right medial frontal area only
Elizabeth W. Pang, during calculations. In the more difficult calculation condition, right temporal regions
University of Toronto, Canada
were highly active. In coherence analyses and neural synchrony analyses, several
*Correspondence:
Ching-I Lu
similarities were seen across calculation tasks; however, latency analyses were sensitive
[email protected] to differences in task complexity across the calculation tasks due to the high temporal
Yung-Yang Lin
resolution of MEG. MEG can be used to examine musical cognition and the neural
[email protected]
consequences of music training. Further systematic study of brain activity during high
Specialty section: versus low memory load conditions of music and other cognitive tasks is needed
This article was submitted to
to illuminate the neural bases of enhanced working memory ability in musicians as
Brain Imaging and Stimulation,
a section of the journal compared to non-musicians.
Frontiers in Human Neuroscience
Keywords: working memory, musical transposing, calculation, music training, magnetoencephalography (MEG)
Received: 31 January 2022
Accepted: 25 March 2022
Published: 16 May 2022
INTRODUCTION
Citation:
Lu C-I, Greenwald M, Lin Y-Y and Working memory is enhanced in musicians as compared to non-musicians (George and Coch,
Bowyer SM (2022) Music, Math,
2011), but neuroimaging studies of musicians have yielded little information about their brain
and Working Memory:
Magnetoencephalography Mapping
activity during cognitive tasks with high versus low working memory load conditions. Similarly,
of Brain Activation in Musicians. there is insufficient detail about high versus low working memory demands of specific music
Front. Hum. Neurosci. 16:866256. tasks in the general debate about potential cognitive effects of music training (Nutley et al., 2014;
doi: 10.3389/fnhum.2022.866256 Swaminathan et al., 2017).

Frontiers in Human Neuroscience | www.frontiersin.org 1 May 2022 | Volume 16 | Article 866256

Lu et al. Music, Math, and Working Memory

Working memory is highly taxed in some music tasks; in occipito-temporal stream of visual-spatial encoding, and to
others, the working memory demand is very low. If music increased frontal lobe activation (Lu et al., 2021).
training includes only low demands on working memory, then Further studies are needed to compare aspects of music,
working memory is not likely to improve from the training, language and mathematical cognition, and brain activity that
nor would it be expected to influence working memory function supports them in musicians and non-musicians. For example,
that supports a different cognitive behavior, such as math or studies using visual music and math tasks are needed to
reading. One music task that is demanding of working memory determine the relative roles of domain-general skills in working
is musical transposing, which involves mentally converting notes memory and spatial attention (Sligte et al., 2009) compared to
from one musical key (i.e., pitch scale) to another key for singing domain-specific cognitive abilities in music or math.
or instrumental performance. Working memory demands are Because musical transposing involves mental calculations
high during musical transposing because the target musical for modifying musical keys, it may share cognitive elements
key must be stored temporarily while notes are manipulated. with arithmetical operations of addition and subtraction. It is
No studies of the cognitive effects of musical training have therefore of interest to compare the brain activity underlying
included transposing in the music training program, though simple arithmetic to that underlying the mental conversion of
this would be one means to assess whether music training transposing in which notes are adjusted up or down by a specific
involving high working memory load would improve working amount. Further, imposing a requirement that participants hold
memory capacity. a cue in working memory to perform the arithmetic and
Ongoing research into the cognitive and neural effects of transposing tasks adds another level of similarity across the
training in music includes studies of experts and non-experts, tasks and a way for working memory load to be manipulated
and the longitudinal effects of exposure and training in children up or down. By studying similarities and differences across the
(Schön et al., 2002; Stewart et al., 2003). Music and math arithmetic and transposing tasks of higher and lower working
are separate cognitive domains, and separable from language memory load, we obtain clues as to how these cognitive tasks
(Ivanova et al., 2020), although there is some evidence they are related and how training in one (e.g., music training) could
may share domain-general structural processing mechanisms possibly affect the other (e.g., performance in math).
with language (Van de Cavey and Hartsuiker, 2016; Nakai In the current paper, we report brain activity in trained
and Okanoya, 2018). There is evidence for a relationship musicians during math calculation tasks compared to musical
between musical achievement and math achievement, usually transposing using MEG. We also examine the effects of working
associative (Holochwost et al., 2017) rather than causal (Hille and memory load across music and calculation tasks.
Schupp, 2015; Wallick, 1998). Recently, Bergee and Weingarten
(2021) controlled for background variables that may influence
achievement in music, math, and reading in children, and MATERIALS AND METHODS
found that musical achievement did relate to math and reading
achievement. However, following a meta-analytic review of Participants
studies of music training, Sala and Gobet (2020) concluded that Twenty-one participants at Taipei Veterans General Hospital,
music training has no impact on non-music cognitive skills and Taipei, Taiwan who were able to read the Western musical
academic achievement. In studies of the effects of music training, notation system completed the study voluntarily with normal
greater specificity is needed in descriptions of the cognitive vision, hearing, motor and cognitive abilities. All of the data
components that are highly active during the music training from one participant had to be discarded due to sleeping,
tasks, and whether the training tasks involve cognitive abilities and individual transposing and calculation task data from two
from domains other than music. participants had to be discarded due to noise; thus, data on
Math, for example, is linked to the musical transposing task all tasks were analyzed for 19 participants. All participants
in that changing from one musical key to another is based on were female classically trained musicians (age range = 19–28;
mental calculations that can involve addition or subtraction skills. x = 23.71) with at least 10 years of musical instrumental training
A varying potential for unidirectional or bidirectional influence (range = 10–22 years; x = 16.71), including reading of standard
of learning in music and math tasks will depend on the cognitive musical notation. Potential participants were not included if
components needed to accomplish each task, and the type and their major instrument was a transposing instrument (i.e., an
degree of overlap between these cognitive components across instrument that produces a higher or lower pitch than is shown in
tasks. Transposing is one example of a music task in which the music notation written for it). For example, if a musicians’ major
influences of music and math during training of transposing may instrument is clarinet in B flat, they would always automatically
be bidirectional. lower two interval pitches while they read the score in G
In previous reports, we have shown that clef. Thus, long-term intensive experience in transposing in
magnetoencephalography (MEG) is sensitive to differences these individuals has the potential to alter patterns of brain
in working memory load (Lu et al., 2019, 2021). By comparing activity during transposing as compared to other musicians. For
brain activity in musical transposing of musical notation versus the musicians who were included in this study, a transposing
sight-reading (in which notes are played as written), we observed instrument was not the major instrument and, based on typical
that the additional mental conversion required for transposing music training in Taiwan in which students are trained to read
was linked to slowed activation of the ventral (fusiform gyrus) transposing scores from approximately the 3rd to the 12th

Frontiers in Human Neuroscience | www.frontiersin.org 2 May 2022 | Volume 16 | Article 866256

Lu et al. Music, Math, and Working Memory

grade, the average amount of experience in transposing for the Five-Note Transposing With Treble Clef (5T)
participants included in this study was approximately 10 years. This task was identical to the 1T task except that a sequence
All participants completed the Edinburgh Handedness of five notes was presented for 3,500 ms (Figure 1). During
Inventory (Oldfield, 1971), and laterality quotients indicated each presentation, the participant was to silently transpose
strong right-handed preference for all but one participant, who each written note to the target key and then silently name
was ambidextrous. They also passed the Mini-Mental State the new notes sequentially, with no overt action. This task
Examination (MMSE; Folstein et al., 1975), which screened lasted 6 min 30 s.
for cognitive impairment, and the digit span task (WAIS-III;
Wechsler, 1997), which measured working memory storage Procedure
capacity. All participants were within normal range. None Stimuli were presented electronically using E-Prime Professional
had a history of neurological or psychiatric diseases, or 2.0 software (Psychology Software Tools, Pittsburgh, PA). To
developmental learning difficulties. Each participant provided document participant accuracy in the tasks, behavioral practice
a written informed consent prior the experiment. The study data were collected using overt naming before each participant
was approved by the Taipei Veterans General Hospital Human entered the MEG scanner; over 60 trials of each task, the average
Research Review Board, Taipei. pre-test single accuracy was 99% for 1D, 95% for 5D, 87% for
1T, and 80% for 5T. Silent naming was required inside the
Experimental Design and Stimuli scanner due to decreased signal noise (due to mouth movement)
A two-factor within-participant design was used: 2within (stimuli: in silent naming compared to overt naming. Participant brain
Musical notation versus Digits) × 2within (task: Easier – one waves were monitored during tasks in the scanner to ensure
note or one single-digit versus Difficult-five notes or five-single participant alertness.
digits), with location, amplitude, and latency of activation as Inside the scanner, oral and written instructions were given
dependent variables. Brain activation was observed during the immediately prior to each task, with 2-min breaks between tasks.
four experimental tasks below. We previously discussed the The 1T task was given before the 5T task followed by the 1D
results of the transposing tasks in comparison to musical sight- and 5D tasks. Including screening, practice and experimental
reading (Lu et al., 2021). tasks, the procedure lasted approximately 90 min (Figure 1).
Immediately following the MEG scan, each participant was asked
One Single-Digit Calculation (1D) a general open-ended question as to how they completed the
A written cue indicated a plus (+) or minus (−) single digit from transposing and math tasks.
1 to 5 presented for 1,000 ms randomly (+ 1, + 2, + 3, + 4, + 5,
−1, −2, −3, −4, −5). After a 1,000 ms blank screen, each Data Acquisition and Preprocessing
stimulus (n = 60) was presented randomly for 1,500 ms plus A 306-channel MEG system (Vectorview, Elekta-Neuromag,
1,000 ms ISI. Each stimulus was a single digit from 0 to 9. Helsinki, Finland) was used; this helmet-shaped device covers the
During each presentation, the participant was to silently add to or entire adult head except for the face. Participants were monitored
subtract from the target stimulus based on the plus or minus cue continuously by intercom and camera. During data collection,
and then silently name the correct answer, with no overt action participants were asked to avoid eye and body movements. MEG
(e.g., the cue + 4 followed by the target 8 would be silently named data were recorded with a high pass filter of 0.1 Hz, low pass filter
as 12). This task lasted 4 min 30 s. of 100 Hz, and sampling rate of 508.63 Hz.
MEG signals measured the magnetic fields produced by
Five Single-Digit Calculation (5D) currents fed into four head position indicator coils at known
This task was identical to the one single-digit calculation task scalp locations, two high behind the earlobes and two wide apart
except that a sequence of five digits was presented for 3,500 ms high on the forehead. Coil locations were chosen in relation to
(Figure 1). During each presentation, the participant was to three anatomical landmarks, including left preauricular point,
silently add to or subtract from the target stimulus based on the right preauricular point, and nasion, which were determined
plus or minus cue and then silently name the correct answer, with with three-dimensional digitization. The individual sensors were
no overt action (e.g., the cue + 4 followed by the target 6 4 2 8 5 magnetometers. Head shape was digitized for coregistration to
would be silently named, sequentially, as 10, 8, 6, 12, 9). This task the standard female brain template of T1-weighted MRI. The
lasted 6 min 30 s. MRI scan was performed on a GE 1.5-T, 1-m-bore whole body
magnet. MRI scan parameters were coronal T1 images, 124
One-Note Transposing With Treble Clef (1T) slices, and 256 × 256 matrix including the entire skin surface
A written cue of one of five transposing instruments was of the head. A model of cortical brain surface was created
presented for 1,000 ms pseudo randomly (Clarinet in A, in Eb from this standard MRI and performed in MEG-TOOLS (Moran
or in Bb , French horn in F, and Trumpet in Bb ). After 1,000 ms et al., 2005). The MRI was segmented and brain surface was
blank screen, each stimulus (n = 60) was presented randomly represented by a cortical model of approximately 4,000 dipoles
for 1,500 ms plus 1,000 ms ISI. During each presentation, the each having x, y, and z orientation at each site. Sites were
participant was to silently transpose from the written note to the distributed to represent the same volume of cortical gray matter.
target key and then silently name the new note, with no overt This model was then morphed to fit the digitized head shape
action. This task lasted 4 min 30 s. collected during MEG acquisition.

Frontiers in Human Neuroscience | www.frontiersin.org 3 May 2022 | Volume 16 | Article 866256

Lu et al. Music, Math, and Working Memory

FIGURE 1 | Sample stimuli, duration and procedure across the four experimental tasks: (A) the one single-digit calculation task (1D); (B) the five single-digit
calculation task (5D); (C) the one-note transposing task (1T); and (D) the five-note transposing task (5T). All stimuli for 1T and 5T were presented on treble clef and all
cues (transposing instruments) were presented using word form.

Using an independent component analysis (ICA), noise the cortical brain surface was created from an age- appropriate
artifacts due to heart and body movement were eliminated in standard MRI of a female brain, as described above. To calculate
post-processing. Any other artifacts in the data were removed coherence, the MEG data were first divided into 40 (1T or 1D
if needed using singular valued decomposition. Regarding tasks) or 52 (5T or 5D tasks) segments each containing 7.5-s
movement artifact, runs would have been repeated if the coil segments of data, and cortical activity in each segment was
on head positions exceeded 0.5 cm, although this did not occur imaged on the MRI using the MR-FOCUSS imaging technique.
during data acquisition. Data were filtered 3–85 Hz with notch at Using the time sequence of imaged activity, coherence between
60 Hz. The locations of events on trigger and response channels active cortical model sites was calculated for each data segment
were used to select 1.5-s epochs of MEG data to examine average and then averaged for the completed study. In addition, for each
evoked responses during the four experimental tasks. cortical model site, connectivity was quantified by a histogram
of the number of sites to which the site had the same level of
MR-FOCUSS coherence. Statistical analysis of cortical coherence levels (0 to
Event-related cortical activation was studied by averaging all 60 1) were used to quantify differences in network connectivity
trials of the participant’s measured evoked MEG field responses between groups. Changes in coherence and connectivity between
during each task. Data were analyzed using MR-FOCUSS brain regions implicated as having deviant electrophysiological
(current distribution technique; Moran et al., 2005) to localize activity in different tasks within the participants’ brains were
and quantify cortical activation within the brain. The latency quantified and included in further statistical analysis.
(in ms), location and average amplitude of response (nAm/time A region-of-interest (ROI) tool implemented in MEG
point) were extracted from MR-FOCUSS imaging results. MR- Tools was used to identify 54 regions in the brain (27 in
FOCUSS cortical mapping was applied to the interval 0– each hemisphere). MEG Tools uses a non-linear volumetric
1,500 ms after stimulus onset in each experimental task. Selection transformation of the participant’s brain to transform MEG
of significant cortical activation was determined by visually coordinates to standard Talairach or MNI coordinates. This
inspecting imaged MEG solutions overlaid on the anatomical enables the ROI tool to access an atlas of Brodmann’s area
MRI and setting the display threshold to 30% (color coded identifiers and an atlas of cortical structures.
blue) of the maximum cortical source amplitude (color coded
red), and by selecting the high peaks of activity relative to the Neural Synchrony
background brain noise. T-test was used to assess task difference in average coherence
values for each pair of brain regions (N = 1431) (Lajiness-
Coherence Source Imaging O’Neill et al., 2014). A p value was produced for each region
Synchronization of neuronal activity was quantified by pair. Because of the large number of tests being performed
calculating coherence between cortical sites from MEG imaged simultaneously, using a significance level of alphas = 0.1 without
brain activation (Elisevich et al., 2011; Bowyer, 2016). A model of adjusting for multiple testing would lead to a large number of

Frontiers in Human Neuroscience | www.frontiersin.org 4 May 2022 | Volume 16 | Article 866256

Lu et al. Music, Math, and Working Memory

false positive results; therefore, false discovery (FDR) was used to variables. Using general liner model – repeated measure revealed
adjust for multiple testing. Bonferroni adjustments for multiple significant main effects of ROI [F(4,68) = 360.24, p < 0.001],
comparisons aim to control the family wise error rate. From each task [F(1,17) = 51.00, p < 0.001], and loading [F(1,17) = 6.45,
t-test, a t-score was computed according to the method of Efron p = 0.021]. There is an interaction effect between ROI and
to summarize the difference in coherence values between tasks. task [F(4,68) = 67.84, p < 0.001]. Main effects of ROI showed
no significant difference in early latency of activity across the
four tasks (including visual, fusiform, and superior parietal).
RESULTS However, the four tasks showed significantly different latency
at Wernicke’s area [F(372) = 3.64, p = 0.017] and at the stage
of simultaneous visual and frontal activation [F(3,72) = 56.35,
Latency and Amplitude
p < 0.001]. Post hoc comparisons of 1D versus 1T showed a
Cortical mapping using MR-FOCUSS analysis displayed multiple
significant difference at Wernicke’s area [t (18) = 3.47, p = 0.003]
areas of neuronal activity, including visual cortex, fusiform
such that 1T (199 ± 68 ms) was faster than 1D (274 ± 99 ms);
gyrus, superior temporal gyrus (STG), angular gyrus (AG),
for the 1T versus 5T comparison [t (17) = −3.38, p = 0.004],
supramarginal gyrus (SMG; Wernicke’s area included activation
1T (202 ± 69 ms) was faster than 5T (272 ± 61 ms). At the
of the STG, AG, and SMG), the superior parietal gyrus, and
stage of simultaneous visual and frontal activation), a significant
frontal lobe regions. Simultaneous activation of both visual
difference was indicated for the 1D versus 1T comparison [t
and frontal gyri was also measured. Selected images from an
(18) = −10.12, p < 0.001] such that 1D (548 ± 103 ms) was
individual participant are shown as examples in Figures 2, 3.
faster that 1T (968 ± 176 ms), for the 1D versus 5D comparison
Note the latencies for this individual fall within the midrange of
[t (18) = −3.05, p = 0.007] such that 1D (548 ± 103 ms) was faster
the average across all subjects (showing simultaneous frontal and
than 5D (666 ± 113 ms), and for the 5D versus 5T comparison [t
occipital activation occurring earlier for the 5D task than for the
(17) = −8.62, p < 0.001] such that 5D (668 ± 117 ms) was faster
5T task). The temporal resolution of peak activation in these areas
than 5T (1028 ± 125 ms).
during the four tasks is summarized in Table 1.
To better understand the three main effects [ROI (region of
interest), task, memory load and their interaction effects], a three- Coherence Source Imaging
factor repeated-measure analysis was conducted 5within (ROI: We analyzed the entire dataset for each task to find the top
visual versus fusiform versus Wernicke’s area versus superior five highest coherence regions active for all of the participants
parietal versus visual + frontal areas) × 2within (task: calculation combined (Table 2). During 1D, 5D, and 1T tasks, the highest
versus transposing) × 2within (memory loading: one versus coherent region was left parahippocampus; in 5T, the highest
five) with peak latency of activity (visual, fusiform, superior coherent region was right precentral motor cortex. Interestingly,
parietal, Wernicke’s area, and visual + frontal) as dependent the transposing tasks and calculation tasks all engaged the left

FIGURE 2 | Sample results of the MEG recordings (MR-FOCUSS analyses) for an individual participant showing simultaneous frontal and occipital activation at
1,244 ms latency in five-note transposing (5T).

Frontiers in Human Neuroscience | www.frontiersin.org 5 May 2022 | Volume 16 | Article 866256

Lu et al. Music, Math, and Working Memory

FIGURE 3 | Sample results of the MEG recordings (MR-FOCUSS analyses) showing simultaneous frontal and occipital activation at 513 ms latency in five single-digit
calculation (5D) for the same participant shown in Figure 2.

TABLE 1 | Temporal resolution of magnetoencephalography (MEG) signals arising from the peak brain activation during each task

Visual cortex Fusiform gyrus Superior parietal gyrus Wernicke’s areaa Frontal and visual cortexb

One single-digit calculation (1D) 84 ± 14 ms 244 ± 131 ms 241 ± 107 ms 274 ± 99 ms 548 ± 103 ms
One-note transposing (1T) 83 ± 20 ms 237 ± 144 ms 216 ± 72 ms 199 ± 68 ms 968 ± 176 ms
Five single-digit calculation (5D) 91 ± 14 ms 236 ± 142 ms 257 ± 91 ms 282 ± 111 ms 666 ± 113 ms
Five-note transposing (5T) 84 ± 16 ms 241 ± 129 ms 270 ± 122 ms 271 ± 60 ms 1019 ± 127 ms
a Bilateralsuperior temporal gyrus (STG), angular gyrus (AG), and supramarginal gyrus (SMG).
b Frontal and visual regions activated simultaneously.
ms, milliseconds.

TABLE 2 | Spatial resolution of magnetoencephalography (MEG) signals arising from the top five highest coherent regions during each task.

The highest region 2nd 3rd 4th 5th

One single-digit calculation (1D) Left parahippocampus Right medial frontal Left medial frontal Left inferior frontal Right superior frontal
One-note transposing (1T) Left parahippocampus Left superior parietal Right superior frontal Right medial orbitofrontal Left medial frontal
Five single-digit calculation (5D) Left parahippocampus Right medial frontal Left medial frontal Right middle temporal Right fusiform
Five-note transposing (5T) Right precentral (BA4) Right superior occipital Right inferior frontal Right precentral (BA6) Left medial frontal

medial frontal area. Also, the calculation tasks had the same In particular, inter-hemispheric network activity differences were
top three highest coherent regions: Left parahippocampus, right identified between limbic system and other regions. The p values
medial frontal, and then left medial frontal. of less than 0.05 identified using this procedure are listed in
Table 3.
Neural Synchrony Analysis Of the 13 pathways that were found to have significant
To identify neuronal networks most strongly activated during differences between the 5D networks versus the 5T networks, the
each task, 1,431 pathway connections were evaluated for their most likely (i.e., most common) was the occipital-limbic system
coherence value in each task, and then a bootstrap method pathway. This is shown in Figure 4, as compared to the less
was used to identify differences (p < 0.05) between tasks. This likely pathways.
provides information on which networks (i.e., two locations) are In comparing the 1D networks versus the 1T networks,
significantly involved for each task compared to the other task. significant differences were found in 3 out of 1,431 pathways.
In comparing the 5D networks versus the 5T networks, In particular, inter-hemispheric network activity differences were
significant differences were found in 13 out of 1,431 pathways. identified between occipital and frontal regions.

Frontiers in Human Neuroscience | www.frontiersin.org 6 May 2022 | Volume 16 | Article 866256

Lu et al. Music, Math, and Working Memory

TABLE 3 | Differences in network activation: Five single-digit calculation (5D) versus five-note transposing (5T).

Path Mean.5D Mean.5T t p.value

L.cingulate_gyrus.L.inferior_frontal_gyrus 0.027 0.021 2.109 0.042

R.cingulate_gyrus.R.gyrus_rectus 0.03 0.02 2.057 0.047
L.inferior_occipital_gyrus.R.parahippocampal_gyrus 0.053 0.073 −2.586 0.014
L.inferior_occipital_gyrus.R.insular_cortex 0.028 0.04 −2.265 0.03
R.parahippocampal_gyrus.R.supramarginal_gyrus 0.037 0.049 −2.192 0.035
L.inferior_occipital_gyrus.R.angular_gyrus 0.163 0.197 −2.18 0.036
L.inferior_occipital_gyrus.R.superior_temporal_gyrus 0.135 0.167 −2.131 0.04
R.insular_cortex.R.lingual_gyrus 0.014 0.02 −2.128 0.04
L.lateral_orbitofrontal_gyrus.R.insular_cortex 0.022 0.03 −2.098 0.043
L.inferior_occipital_gyrus.R.supramarginal_gyrus 0.155 0.187 −2.084 0.044
R.angular_gyrus.R.parahippocampal_gyrus 0.038 0.05 −2.084 0.044
L.lingual_gyrus.R.insular_cortex 0.015 0.021 −2.039 0.049
L.middle_temporal_gyrus.R.insular_cortex 0.024 0.031 −2.027 0.05

The limbic system including insula, putamen, parahippocampal gyrus, caudate, hippocampus.

FIGURE 4 | Number of differences in neural connections: Comparison between five single-digit calculation (5D) versus five-note transposing (5T).

Of the 3 pathways that were found to have significant transposing of high and low working memory load (5D versus
differences between the 1D networks versus the 1T networks, 5T, 1D versus 1T; Figure 5). Of the pathways found to have
the most likely (i.e., most common) was the frontal-occipital significant differences between the two tasks of high working
pathway, especially in right inferior frontal gyrus. The negative memory load (5D versus 5T), the most likely pathways were
t values showed that the 1T networks are more active than 1D inter-hemispheric.
networks between the occipital and frontal regions (Table 4). After the MEG session, each participant responded to an
The 1D networks versus the 5D networks were compared. informal query about their strategy in the experimental tasks. The
Significant differences were found in none of 1,431 pathways. participants reported that they responded one-by-one to the five
In neural synchrony analysis, 2 of the 1,431 pathways digits or notes presented in each 5D and 5T stimulus; further
differed significantly between the 5T versus 1T. Strong network information about individual strategy was not obtained.
differences between these transposing tasks were observed in
connections from left superior occipital gyrus to left frontal
regions (Table 5). As indicated by the positive t values, the 5T DISCUSSION
task involved these pathways significantly more than 1T.
We also observed the following intra- and inter-hemispheric MEG is sensitive to differences in brain activation in musical
differences in coherence across tasks of calculation versus transposing versus digit calculation, as shown in this study, and

Frontiers in Human Neuroscience | www.frontiersin.org 7 May 2022 | Volume 16 | Article 866256

Lu et al. Music, Math, and Working Memory

TABLE 4 | Differences in network activation: One single-digit calculation (1D) multiple elements that indicate clef, staff, and key signature,
versus one note transposing (1T).
whereas in calculation the viewed digits form a simpler display. In
Path Mean.1D Mean.1T T p.value transposing, high demands on visual working memory and visual
perception may have required greater support from visual cortex
L.inferior_occipital_gyrus. 0.185 0.22 −2.244 0.031 and its interactions with frontal lobe systems.
R.inferior_frontal_gyrus
The 5T task involved left occipital-frontal network pathways
L.superior_occipital_gyrus. 0.133 0.163 −2.074 0.045
significantly more than the 1T, and the top four most active
R.inferior_frontal_gyrus
regions in 5T were in frontal or occipital areas. Greater
L.middle_occipital_gyrus. 0.178 0.21 −2.062 0.046
R.inferior_frontal_gyrus visual complexity of the visual pre-cue cannot account for
increased frontal-occipital activation in the transposing tasks
versus calculation, because previously we identified high activity
TABLE 5 | Differences in network activation: Five-note transposing (5T) versus
one-note transposing (1T). in frontal-occipital areas in both high and low memory load
conditions of musical sight-reading wherein there is no pre-cue
Path Mean.5T Mean.1T T p.value (Lu et al., 2021). Musical sight-reading stimuli were similar in
L.inferior_frontal_gyrus. 0.147 0.121 2.877 0.007
complexity to transposing stimuli; as in the 5T task, the high-
L.superior_occipital_gyrus and low-load sight-reading tasks resulted in high activity in
L.precentral_gyrus. 0.17 0.147 2.025 0.05 frontal lobe areas and right superior occipital gyrus. In contrast,
L.superior_occipital_gyrus occipital cortex and frontal-occipital connections were not highly
activated in the calculation task.
Interestingly, among the top three most highly active areas
can be used to examine the neural correlates of musical and in 5T were right primary motor cortex (BA4) and right
mathematical cognition and the consequences of music training. premotor cortex and supplementary motor area (BA6), possibly
The patterns of brain activation observed here are influenced by reflecting motor planning or programming of hand movement
working memory load and task type. (consciously or unconsciously) when musical notation was
Transposing differed from calculation in frontal-occipital viewed. Right precentral cortex was also found to be highly
activation. The simultaneous frontal and occipital activation active in both high- and low-load sight-reading (Lu et al.,
occurred significantly more slowly during transposing compared 2021). High activity in the motor cortex was not evident during
to calculation (1T slower than 1D; 5T slower than 5D). calculation, suggesting the digit stimuli did not result in initiation
Neural synchrony analyses revealed more frontal-occipital neural of motor programming.
connections active in 5T than 5D, and in 1T than in 1D. Frontal- In both the coherence analyses (in top three most highly
occipital interactions support visual perception (Ruff et al., 2006) active regions) and the neural synchrony analyses, the calculation
and visual working memory (Barton and Brewer, 2013). In the tasks were found to have similar patterns of brain activation.
current study, the notes to be transposed are perceived among However, in latency analysis there was a significantly slower

FIGURE 5 | Number of differences in neural connections: Intra-right hemisphere, intra-left hemisphere and inter-hemispheric cortical differences in tasks of high
working memory load (5D versus 5T), low working memory load (1D versus 1T), and digits only (5D versus 1D).

Frontiers in Human Neuroscience | www.frontiersin.org 8 May 2022 | Volume 16 | Article 866256

Lu et al. Music, Math, and Working Memory

peak of simultaneous frontal and occipital activation in 5D (Hayashi et al., 2000) observed activation of right frontal areas
compared to 1D. This likely reflects increased working memory during subtraction but not multiplication, which they attributed
demands in 5D compared to 1D, detected by the high temporal to greater working memory demands in their subtraction task.
resolution of MEG. In the current study, the task design was the same for the
Parahippocampal gyrus, part of medial temporal lobe and the transposing and calculation tasks; nevertheless, differences in
limbic system, was the most highly active brain area for the visual display and the nature of mental conversion needed to
calculation tasks and 1T. This area is involved in visuospatial solve each item may have placed more demands on working
processing and other cognitive functions such as memory memory in transposing compared to math calculations.
(Aminoff et al., 2013). In neural synchrony comparisons, the Along with left and right medial frontal areas, several
5D task involved the occipital-limbic pathway significantly more other frontal lobe regions were highly active in the current
than did 5T. Limbic activity was not as prominent in the 5T task, experimental tasks. The right inferior frontal lobe has been
which had high activations in other regions such as right motor described as important for supporting mathematical subtraction
and premotor cortex. (Kong et al., 2005). These results, along with high activity of the
Multiple distributed processes and brain regions have been right inferior frontal area in the 5T task, occipital connections
described as contributing to working memory; for example, to right inferior frontal cortex in the 1T task (shown in neural
mental representations encoding visual feature information synchrony analysis), and high activity in left inferior frontal
in temporal-occipital cortex (ventral visual pathway), spatial area in the 1D task suggest that these inferior frontal areas
information in frontal-parietal cortex (dorsal visual pathway), should be examined further in comparisons of neural activity
and information such as behavioral significance in frontal cortex in music versus math. The specific roles of the right superior
(Eriksson et al., 2015; Ren et al., 2019). The right medial temporal frontal gyrus (highly active in both 1D and 1T), and the right
lobe provides important support for memory (Jeneson and medial orbitofrontal area (highly active in 1T) in working
Squire, 2012), and the fusiform gyrus is involved in higher visual memory also need further study as they relate to musical
perception and memory (Weiner et al., 2017). In the current and mathematical cognition. Given that activation patterns of
study, when the calculation task involved more numbers to be frontal lobe and frontal-posterior networks can be influenced
added or subtracted the participants relied more on right medial by specific modifications to task design within domain (e.g.,
temporal lobe and right fusiform gyrus, both highly active in differential effects of a pre-cue versus a post-cue; Ruff et al.,
5D but not in 1D. Previously, we (Lu et al., 2021) attributed 2007), more research is needed to define how task adjustments
slowed fusiform activation in transposing compared to musical within and across music and math tasks may affect similarities
sight-reading to the additional mental conversion required for and differences in corresponding neural activity.
transposing; this may require more ventral stream than dorsal Neural responses elicited during the calculation tasks in the
stream processing. Interestingly, in the current study with all current study are affected by the high working memory demands
four tasks more highly demanding of working memory than the of the tasks. Using an arithmetic equation verification task
musical sight-reading task, there is no significant difference in in which working memory demands were limited, Rosenberg-
time course of fusiform activation. Lee and colleagues (Rosenberg-Lee et al., 2011) assessed neural
Superior parietal cortex is important for spatial encoding responses in tasks of addition, subtraction, multiplication and
(Stewart, 2005), and was highly active in the left hemisphere division. Among the brain regions they identified as involved
during 1T in the current study. As part of a dorsal “where” stream in these tasks, the posterior parietal cortex (PPC) was shown
of spatial encoding, this region is involved in encoding stimulus to be critically involved in representing cognitive processes of
location, whereas a ventral “what” stream that includes the retrieval, calculation and inversion differentially involved in these
fusiform appears to encode features for stimulus identification tasks. Others have identified the PPC as important in supporting
(Mishkin and Ungerleider, 1982; Goodale and Milner, 1992). mathematical functions (Dehaene and Cohen, 1997). In the
Reading music may involve interaction between the dorsal and current study, MEG was sensitive to differences in high and low
ventral streams (Mongelli et al., 2017). Previously, we (Lu et al., working memory load across tasks, as described above. However,
2019) identified bilateral activation of superior parietal cortex all of the tasks were demanding of working memory and did
during silent reading of English letters and musical notes. In the result in high activity in brain regions supporting working
current study this region likely played a role across tasks but was memory and, overall, relatively lower activation of brain regions
not prominently activated in 5T, 5D and 1D. supporting other cognitive elements of the tasks.
The left medial frontal area was highly active during all In the present investigation, though participants were required
four experimental tasks; interestingly, right medial frontal area to silently name the solutions to each transposing or math
was highly active during calculation tasks but not during problem (e.g., “B” or “7” for the 1T and 1D tasks, and sequentially
transposing tasks. Functions of frontal lobe regions may overlap during the 5T and 5D tasks), left hemisphere activation was not
in supporting working memory (Duncan and Owen, 2000); dominant during these tasks. This is in contrast to the results of
within this distribution of function the right medial frontal laterality index analyses in our previous study (Lu et al., 2019)
area may have some relative specialization for supporting in which silent naming of English letters and silent reading of
calculation. Differences in working memory demands across notes in musical sight-reading both resulted in left hemisphere-
tasks have influenced efforts to define frontal lobe specialization dominant activation. We (Lu et al., 2019) hypothesized that
in mathematical tasks; for example, Hayashi and colleagues those results reflected left hemisphere phonological activation

Frontiers in Human Neuroscience | www.frontiersin.org 9 May 2022 | Volume 16 | Article 866256

Lu et al. Music, Math, and Working Memory

to support silent naming based on prior findings that MEG cognitive task demands on working memory and other cognitive
neural activity supporting silent naming is lateralized to the left subcomponents will be needed to capture how the brain supports
hemisphere in right-handed individuals (Bowyer et al., 2004). music and math functions in experts and non-experts, and the
In the current study, results from the three analysis methods cognitive effects of training.
used do not show that brain activation is more lateralized to the
left hemisphere for the transposing or calculation tasks. This is
another example of how high task demands on working memory CONCLUSION
here may have eclipsed other cognitive elements of the tasks that
can be observed using MEG when the working memory task MEG was sensitive to differences in working memory load
demands are lower. during musical transposing and calculation tasks in a group
Expert musicians, and students being trained in musical of classically trained musicians. Frontal-occipital connections
transposing, may employ one or more strategies to accomplish were highly active during the transposing tasks, but not during
transposing during ongoing performance. We previously the math calculation tasks. Right temporal regions were highly
described potential differences in musical transposing strategies active in the more difficult condition of the calculation task.
that may include reliance on auditory imagery, visual-spatial Multiple frontal lobe regions were highly active across tasks;
imagery, or both (Lu et al., 2021). The activation of motor notably, the left medial frontal area was highly active in all four
cortex during the 5T task in the current study may reflect tasks, but the right medial frontal area was highly active only
a motor strategy in transposing. It remains unclear whether during calculations. Right motor and premotor regions were
this motor activation reflects motor intention or occurred highly active in the more difficult condition of the transposing
unconsciously in these highly trained musicians presented with task but not during calculations. Coherence analyses and neural
musical notation. Further research is warranted into individual synchrony analyses yielded several similarities in brain activation
transposing strategies used by musicians, and by students across the calculation tasks, but latency analyses were sensitive to
during their development of transposing skills. The possibility of differences in task complexity across the two tasks due to the high
strategic motor involvement in transposing could be examined temporal resolution of MEG. As was done in the current study,
for potential differences in motor activation in response to future studies should compare brain activity in cognitive tasks
notes presented in the treble clef (associated with right hand involving higher versus lower working memory load. Systematic
movement in piano) versus bass clef (associated with left hand manipulations to task demands on working memory and other
movement in piano). cognitive elements of music and math tasks will be necessary to
Current research into whether music training can affect specify the brain regions supporting elements of these tasks in
math achievement includes hypotheses about the effects of experts and non-experts. MEG is sensitive to these effects and
music training on domain-general processes such as working can be used to examine the neural correlates of musical and
memory (Eriksson et al., 2015). These effects have been examined mathematical cognition and the consequences of music training.
in relation to subtypes of working memory; for example,
Roden and colleagues (Roden et al., 2012) found that music
training resulted in no improvement in visual memory, but that DATA AVAILABILITY STATEMENT
verbal memory improved. Simmons and colleagues (Simmons
The datasets presented in this article are not readily available as
et al., 2012) described that subcomponents of working memory
the format is unique. Requests to access the datasets should be
have different relationships with different mathematical skills.
directed to C-IL, [email protected].
Continued research into cognitive elements involved in easier
and harder iterations of music and math tasks (and cognitive
elements shared across tasks) will influence efforts to specify ETHICS STATEMENT
the neural correlates of these cognitive subcomponents. MEG
tasks with lower working memory load may allow observation The studies involving human participants were reviewed and
of neural activity supporting specific cognitive elements that approved by Taipei Veterans General Hospital Human Research
are not seen when working memory load is high; however, an Review Board, Taipei. The patients/participants provided their
advantage of using tasks with high memory load is that they written informed consent to participate in this study.
may be closer to conditions in the real world. In real world
performance of music or math, working memory demands will
vary with the task but will be generally high. For example, AUTHOR CONTRIBUTIONS
during real world musical performance, the broader influences
of context and attentional demands along with the musician’s C-IL, MG, and SB contributed to conception and design of
work to interpret meaning and emotion all contribute to the study. C-IL and Y-YL contributed to study procedures, and
task complexity. With increasing expertise in music or math, collection and organization of data. C-IL and SB processed the
information processing load will be reduced in some aspects, data. C-IL performed the statistical analyses. C-IL and MG wrote
such as greater ability to use context to anticipate continuations sections of the manuscript. MG wrote the first draft of the
in music (Waters et al., 1997). Along with ongoing advances in manuscript. All authors contributed to manuscript revision, read,
technology and data analysis methods, systematic adjustments of and approved the submitted version.

Frontiers in Human Neuroscience | www.frontiersin.org 10 May 2022 | Volume 16 | Article 866256

Lu et al. Music, Math, and Working Memory

REFERENCES Mishkin, M., and Ungerleider, L. G. (1982). Contribution of striate inputs to the
visuospatial functions of parieto-preoccipital cortex in monkeys. Behav. Brain
Aminoff, E. M., Kveraga, K., and Bar, M. (2013). The role of the parahippocampal Res. 6, 57–77. doi: 10.1016/0166-4328(82)90081-X
cortex in cognition. Trends Cogn. Sci. 17, 379–390. doi: 10.1016/j.tics.2013. Mongelli, V., Dehaene, S., Vinckier, F., Peretz, I., Bartolomeo, P., and Cohen, L.
06.009 (2017). Music and words in the visual cortex: the impact of musical expertise.
Barton, B., and Brewer, A. A. (2013). Visual working memory in human cortex. Cortex 86, 260–274. doi: 10.1016/j.cortex.2016.05.016
Psychology(Irvine) 4, 655–662. doi: 10.4236/psych.2013.48093 Moran, J. E., Bowyer, S. M., and Tepley, N. (2005). Multi-resolution FOCUSS:
Bergee, M., and Weingarten, K. M. (2021). Multilevel models of the relationship a source imaging technique applied to MEG data. Brain Topogr. 18, 1–17.
between music achievement and reading and math achievement. J. Res. Music doi: 10.1007/s10548-005-7896-x
Educ. 68, 398–418. doi: 10.1177/0022429420941432 Nakai, T., and Okanoya, K. (2018). Neural evidence of cross-domain structural
Bowyer, S. M. (2016). Coherence a measure of the brain networks: past and present. interaction between language and arithmetic. Sci. Rep. 8:12873. doi: 10.1038/
Neuropsychiatr. Electrophysiol. 2:1. doi: 10.1186/s40810-015-0015-7 s41598-018-31279-8
Bowyer, S. M., Moran, J. E., Mason, K. M., Constantinou, J. E., Smith, B. J., Barkley, Nutley, S. B., Darki, F., and Klingberg, T. (2014). Music practice is associated with
G. L., et al. (2004). MEG localization of language-specific cortex utilizing MR- development of working memory during childhood and adolescence. Front.
FOCUSS. Neurology 62, 2247–2255. PMIsD:NOPMID doi: 10.1212/01.wnl. Hum. Neurosci. 7:926. doi: 10.3389/fnhum.2013.00926
0000130385.21160.7a Oldfield, R. C. (1971). The assessment and analysis of handedness: the edinburgh
Dehaene, S., and Cohen, L. (1997). Cerebral pathways for calculation: double inventory. Neuropsychologia 9, 97–113. doi: 10.1016/0028-3932(71)90067-4
dissociation between rote verbal and quantitative knowledge of arithmetic. Ren, Z., Zhang, Y., He, H., Feng, Q., Bi, T., and Qui, J. (2019). The different brain
Cortex 33, 219–250. doi: 10.1016/s0010-9452(08)70002-9 mechanisms of object and spatial working memory: voxel-based morphometry
Duncan, J., and Owen, A. M. (2000). Common regions of the human frontal and resting-state functional connectivity. Front. Hum. Neurosci. 13:248. doi:
lobe recruited by diverse cognitive demands. Trends Neurosci. 23, 475–483. 10.3389/fnhum.2019.00248
doi: 10.1016/S0166-2236(00)01633-7 Roden, I., Kreutz, G., and Bongard, S. (2012). Effects of a school-based instrumental
Elisevich, K., Shukla, N., Moran, J. E., Smith, B., Schultz, L., Mason, K., et al. music program on verbal and visual memory in primary school children: a
(2011). An assessment of MEG coherence imaging in the study of temporal lobe longitudinal study. Front. Psychol. 3:572. doi: 10.3389/fpsyg.2012.00572
epilepsy. Epilepsia 52, 1110–1119. doi: 10.1111/j.1528-1167.2011.02990.x Rosenberg-Lee, M., Chang, T. T., Young, C. B., Wu, S., and Menon, V. (2011).
Eriksson, J., Vogel, E. K., Lansner, A., Bergstrom, F., and Nyberg, L. (2015). Functional dissociations between four basic arithmetic operations in the human
Neurocognitive architecture of working memory. Neuron 88, 33–46. doi: 10. posterior parietal cortex: a cytoarchitectonic mapping study. Neuropsychologia
1016/j.neuron.2015.09.020 49, 2592–2608. doi: 10.1016/j.neuropsychologia.2011.04.035
Folstein, M. F., Folstein, S. E., and McHugh, P. R. (1975). Mini-Mental state: a Ruff, C. C., Blankenburg, F., Bjoertomt, O., Bestmann, S., Freeman, E., Haynes,
practical method for grading the cognitive state of patients for the clinician. J. D., et al. (2006). Concurrent TMS-fMRI and psychophysics reveal frontal
J. Psychiatr. Res. 12, 189–198. doi: 10.1016/0022-3956(75)90026-6 influences on human retinotopic visual cortex. Curr. Biol. 8, 1479–1488. doi:
George, E. M., and Coch, D. (2011). Music training and working memory: an ERP 10.1016/j.cub.2006.06.057
study. Neuropsychologia 49, 1083–1094. doi: 10.1016/j.neuropsychologia.2011. Ruff, C. C., Kristjansson, A., and Driver, J. (2007). Readout from iconic memory
02.001 and selective spatial attention involve similar neural processes. Psychol. Sci. 18,
Goodale, M. A., and Milner, A. D. (1992). Separate visual pathways for perception 901–909. doi: 10.1111/j.1467-9280.2007.01998.x
and action. Trends Neurosci. 15, 20–25. doi: 10.1016/0166-2236(92)90344-8 Sala, G., and Gobet, F. (2020). Cognitive and academic benefits of music training
Hayashi, N., Ishii, K., Kitagaki, H., and Kazui, H. (2000). Regional differences with children: a multilevel meta-analysis. Mem. Cognit. 48, 1429–1441. doi:
in cerebral blood flow during recitation of the multiplication table and actual 10.3758/s13421-020-01060-2
calculation: a positron emission tomography study. J. Neurol. Sci. 176, 102–108. Schön, D., Anton, J. L., Roth, M., and Besson, M. (2002). An fMRI study of music
doi: 10.1016/S0022-510X(00)00323-3 sight-reading. Neuro Rep. 13, 2285–2289. doi: 10.1097/01.wnr.0000044224.
Hille, A., and Schupp, J. (2015). How learning a musical instrument affects the 79663.f5
development of skills. Econ. Educ. Rev. 44, 56–82. doi: 10.1016/j.econedurev. Simmons, F., Willis, C., and Adams, A. (2012). Different components of working
2014.10.007 memory have different relationships with different mathematical skills. J. Exp.
Holochwost, S. J., Propper, C. B., Wolf, D. P., Willoughby, M. T., Fisher, K. R., Child Psychol. 111, 139–155. doi: 10.1016/j.jecp.2011.08.011
Kolacz, J., et al. (2017). Music education, academic achievement, and executive Sligte, L. G., Scholte, H. S., and Lamme, V. A. F. (2009). V4 activity predicts
functions. Psychol. Aesthet. Creat. Arts 11, 147–166. doi: 10.1037/aca0000112 the strength of visual short-term memory representations. J. Neurosci. 29,
Ivanova, A. A., Srikant, S., Sueoka, Y., Kean, H. H., Dhamala, R., O’Reilly, 7432–7438. doi: 10.1523/JNEUROSCI.0784-09.2009
U., et al. (2020). Comprehension of computer code relies primarily on Stewart, L. (2005). A neurocognitive approach to music reading. Ann. N. Y. Acad.
domain-general executive brain regions. Elife 9:e58906. doi: 10.7554/eLife.5 Sci. 1060, 377–386.
8906 Stewart, L., Henson, R., Kampe, K., Walsh, V., Turner, R., and Frith, U. (2003).
Jeneson, A., and Squire, L. (2012). Working memory, long-term memory, Brain changes after learning to read and play music. Neuroimage 20, 71–83.
and medial temporal lobe function. Learn. Mem. 19, 15–25. doi: doi: 10.1016/S1053-8119(03)00248-9
10.1101/lm.024018.111 Swaminathan, S., Schellenberg, E., and Khalil, S. (2017). Revisiting the
Kong, J., Wang, C., Kwong, K., Vangel, M., Chua, E., and Gollub, R. (2005). The association between music lessons and intelligence: training effects
neural substrate of arithmetic operations and procedure complexity. Cogn.Brain or music aptitude? Intelligence 62, 119–124. doi: 10.1016/j.intell.2017.
Res. 22, 397–405. doi: 10.1016/j.cogbrainres.2004.09.011 03.005
Lajiness-O’Neill, R., Richard, A. E., Moran, J. E., Olszewski, A., Pawluk, Van de Cavey, J., and Hartsuiker, R. (2016). Is there a domain-general cognitive
L., Jacobson, D., et al. (2014). Neural synchrony examined with sequencing system? Evidence from structural priming across music, math,
magnetoencephalography (MEG) during eye gaze processing in autism action descriptions, and language. Cognition 146, 172–184. doi: 10.1016/j.
spectrum disorders: preliminary findings. J. Neurodev. Disord. 6:15. doi: 10. cognition.2015.09.013
1186/1866-1955-6-15 Wallick, M. (1998). A comparison of ohio proficiency test results between fourth-
Lu, C., Greenwald, M. L., Lin, Y., and Bowyer, S. M. (2019). Reading musical grade pullout students and those of matched ability. J. Res. Music Educ. 46,
notation versus english letters: mapping brain activation with MEG. Psychol. 239–247. doi: 10.2307/3345626
Music 47, 255–269. doi: 10.1177/0305735617744886 Waters, A. J., Underwood, G., and Findlay, J. M. (1997).
Lu, C., Greenwald, M. L., Lin, Y., and Bowyer, S. M. (2021). Musical transposing Studying expertise in music reading: use of a pattern-matching
versus sight-reading: mapping brain activation with Magnetoencephalography paradigm. Percept. Psychophys. 59, 477–488. doi: 10.3758/BF032
(MEG). Psychol. Music 49, 581–599. doi: 10.1177/0305735619883692 11857

Frontiers in Human Neuroscience | www.frontiersin.org 11 May 2022 | Volume 16 | Article 866256

Lu et al. Music, Math, and Working Memory

Wechsler, D. (1997). Wechsler Adult Intelligence Scale, 3rd Edn. San Antonio, TX: Publisher’s Note: All claims expressed in this article are solely those of the authors
The Psychological Corp. and do not necessarily represent those of their affiliated organizations, or those of
Weiner, K. S., Barnett, M. A., Lorenz, S., Caspers, J., Stigliani, A., Amunts, K., et al. the publisher, the editors and the reviewers. Any product that may be evaluated in
(2017). The cytoarchitecture of domain-specific regions in human high-level this article, or claim that may be made by its manufacturer, is not guaranteed or
visual cortex. Cereb. Cortex 27, 146–161. doi: 10.1093/cercor/bhw361 endorsed by the publisher.

Conflict of Interest: The authors declare that the research was conducted in the Copyright © 2022 Lu, Greenwald, Lin and Bowyer. This is an open-access article
absence of any commercial or financial relationships that could be construed as a distributed under the terms of the Creative Commons Attribution License (CC BY).
potential conflict of interest. The use, distribution or reproduction in other forums is permitted, provided the
original author(s) and the copyright owner(s) are credited and that the original
The reviewer EWP declared a past co-authorship with one of the authors SB publication in this journal is cited, in accordance with accepted academic practice. No
to the handling editor. use, distribution or reproduction is permitted which does not comply with these terms.

Frontiers in Human Neuroscience | www.frontiersin.org 12 May 2022 | Volume 16 | Article 866256