By Brian Wandell, Ph.D., Robert F. Dougherty, Ph.D., Michal Ben-Shachar, Ph.D., Gayle K. Deutsch, Ph.D., and Jessica Tsang, Stanford University
We undertook a series of studies to investigate how aesthetic ability and arts education correlate with improvements in children’s reading abilities. As part of this research, we developed new analysis tools for diffusion tensor imaging (DTI), a method that identifies likely connections among brain regions involved in the development of reading skills. We also studied how exposure to the visual arts might relate to phonological awareness (the ability to manipulate speech sounds), which is correlated with reading ability. Further, based on an incidental finding in that research, we initiated a study to explore the relationship between exposure to the visual arts and children’s math calculation abilities.
Our study included 49 children, aged 7 to 12, who were enrolled in an NIH-funded longitudinal study of the development of reading skills and the brain structures associated with this development. To this study, we added measures to determine the effects of arts training on the children’s reading fluency and phonological awareness. We collected data, through parental questionnaires, on the extent of children’s training in visual arts, music, dance, and drama/theater. We also collected parent-reported data on their children’s temperament and openness to experience. We then correlated the arts training data with the children’s test scores in reading fluency and phonological awareness over a three-year period.
The findings indicate that the amount of musical training the children underwent in the first year of our study correlated with the amount of improvements in the children’s reading fluency over the three-year study period.
Additionally, we used DTI in each child to relate these behavioral measurements to individual differences in the anatomy of the corpus callosum- a structure that connects the brain’s left and right hemispheres. We measured the tissue properties of specific connections between brain regions (communication cables) to identify structural differences between strong and weak readers. We found that diffusion in the bundle of axons that connect the brain’s temporal lobes is correlated with all our measures of reading ability; phonological awareness showed the strongest correlation. We also found a weak correlation between visual arts experience and phonological awareness. Finally, based on an incidental finding from this research, preliminary data suggest that exposure to visual arts may be correlated with improvements in children’s math calculation abilities.
Our team had two principal goals as part of the consortium. One was to develop and disseminate tools, to all Consortium grantees, for using diffusion tensor imaging (DTI) of the brain’s white matter. White matter consists of axons (the brain cells’ communication cables), which enable communication from one brain area to another. Our second goal was to investigate how aesthetic ability and arts education correlate with behavioral development of reading skills, and to investigate with DTI how these behavioral measurements correlate with the development of brain structures. To achieve these goals, we integrated this Consortium research into a National Institutes of Health (NIH)-funded longitudinal study of reading development in children ages 7 to 12, in which we assessed development of the children’s reading skills over a three-year period. We have successfully advanced the two Consortium research goals over the past three years.
In the first section of this report, we describe the new behavioral measurements we created for assessing correlations between aesthetic ability and arts education, based on Deutsch’s essential work on acquiring and analyzing arts questionnaires. In the second section, we briefly describe the new tools we developed for analyzing DTI, largely through Akers’ creation of DTI software. In the third section, we describe preliminary results revealing a relationship between brain structure and behaviors that are essential for reading.
One of our major findings from these studies was that the amount of musical training measured in year 1 was significantly correlated with the amount of improvement in Reading Fluency demonstrated in children over the three-year period of our study.
Additionally, we discuss an incidental finding in section four: in our sample, we found a correlation between exposure to the visual arts and improvement in math calculation, as measured by children’s calculation test scores. For this new research direction, we have initiated a collaboration with consortium investigators, Dehaene and Spelke from Harvard, and with our Stanford colleagues in the School of Education, Professor Richard Shavelson and Jessica Tsang.
Study Designs and Results
Section 1: Reading, Arts Education, and Aesthetics
In conjunction with the NIH-funded longitudinal study of reading development, we explored the relationship among training in the arts, children’s temperament, and reading fluency. We developed an arts education questionnaire, and used a modified version of the Child Temperament and Personality Questionnaire (CTPQ) developed by Victor, Rothbart, and Baker (2003). The arts education questionnaire explores four areas of arts education: 1) Visual Arts 2) Music 3) Dance and 4) Drama/Theater, and includes in-school education, as well as formal training and independent practice, in each of these four areas.
Parents reported the number of hours per week their child received formal training outside of school, spent time in independent practice, and received in-school instruction. Parents also rated their child’s skill level. Aesthetic ability was measured by the Child Temperament and Personality Questionnaire. Between year 1 and year 2 of the reading study, we obtained questionnaire data on 49 children (ages 7 through 12). At year 3 of the reading study, we obtained questionnaire data on 41 of these children.
We examined the relationship among behavioral and imaging measures and the Openness to Experience Scale, which is comprised of four factors: 1) Ideas 2) Aesthetics 3) Intellect/Quick to Learn and 4) Perceptual Sensitivity. An exploratory principal components analysis (PCA) was conducted using SPSS (statistical software) to investigate the items from the Child Temperament and Personality Questionnaire in our sample of children (n=49). PCA analysis revealed five components, identified (in order of measured importance) as 1) Intelligence/Quick to Learn, 2) Aesthetics, 3) Perceptual Sensitivity, 4) Distractibility, and 5) Ideas. These explained, respectively, 22.5%, 18.1%, 10.1%, 7.3%, and 5.3% of the variance. The items generally weighted on the same components identified by the original Child Temperament and Personality Questionnaire. Parents’ ratings of their children on the Intelligence/Quick to Learn factor was significantly correlated with several of the IQ and Reading Fluency measures. The correlations ranged from r=.42, (statistically significant at p < .005) to r=.57 (statistically significant at p < .001). Hence, parents’ reports about their children were generally consistent with measurements using standardized tests of IQ and Reading.
The simplest and most compelling observation was this: The amount of musical training measured in year 1 was significantly correlated with the amount of improvement in Reading Fluency between years 1 and 3 (Figure 1.1). We hasten to add that this correlation does not imply that music training caused the reading improvement. It is possible that children who are intellectually capable of pursuing musical training are also ready for reading. The observed correlation should be followed-up with a controlled study to analyze the possibility of a causal connection.
A second, and more surprising, observation was a relationship between early visual arts experience and phonological awareness (attention to patterns of speech sounds). Phonological awareness is an auditory skill that is reliably correlated with reading ability. In the first year of our study, children who had early training in the visual arts had a higher degree of phonological awareness than children with no such training. There was no significant age difference between these groups (mean ages were 10.2 and 10.9). The difference in phonological awareness scores between the two groups, shown in Figure 1.2, was quite large and highly significant (statistically) in the first year of our study. Two years later, however, when the group ages were roughly 12.5y, the difference in phonological awareness scores had disappeared. This phenomenon – an early correlation that dissipates with time – is commonly observed in the developmental literature. The explanation for the disappearance is thought to be that many factors contribute to skill learning. As children develop, they find many ways to master a skill. Hence, the impact of the factor revealed by the early association becomes less and less important with time.
Figure 1.2. Subjects with any visual arts experience prior to age 4 (blue, N=37, 37, 34, for y1, y2, y3 respectively) score higher than those without such training (red; N=11, 9, 7) in the first year (t=2.01, p<0.05). The difference decreases in year 2 and disappears by year 3.
We find this correlation between visual arts training and phonological awareness surprising. Perhaps an explanation can be found in the amount of attention to training provided by the parents, or by attendance to preschool, which provides training in both skills. The visual arts score may be an indicator of intense parental involvement and, with increasing years of schooling, the measured parental effect is reduced. Admittedly, the effect is based on a small sample and may disappear in repeated examination in a larger sample.
As part of our new collaboration with Professor Shavelson, we will extend our analysis of the reading data to item-specific analyses. We plan to complete the study of the full range of reading and general intelligence measures, as well as the correlations with arts experience.
Section 2: Diffusion Analysis and Visualization Tools
Methods for identifying the paths and properties of the large groups of white matter fibers (fascicles), the tracts of brain cell axons that facilitate cellular communication in the human brain, are an important component in understanding human brain organization and development. We have made substantial progress in developing computational algorithms to identify and measure these pathways.
Interpreting the pathways, estimated from diffusion tensor imaging (DTI) data, requires not only skilled scientists but also extensive visualization tools. David Akers developed a beautiful graphical interface for selecting, editing, and visualizing DTI white matter fiber tracts. He performed a systematic study of our work flow, and developed a software application that can assist all Consortium investigators who interact with DTI data. Figure 2.1 is from David’s paper (published in the Computer-Human Interaction proceedings) and shows the interface to the software.
Figure 2.1. DTI visualization tool (Akers 2006).
Existing algorithms for tracing white matter tract pathways were analyzed by Anthony Sherbondy, a student in the Electrical Engineering department, together with Robert Dougherty and Professor Wandell. They identified several limitations in these algorithms that cause investigators to miss important fascicles (groups of axon fibers), limiting our ability to estimate the probability that two locations of “gray matter” (regions of nerve cell bodies) are connected, via their axons. Sherbondy, Wandell, and Dougherty developed new probabilistic tracking tools to better estimate these fascicles.
Robert Dougherty, who managed the software development efforts, integrated Aker’s and Sherbondy’s contributions with mrDiffusion, our general laboratory measurement tools for DIT analyses. This package is currently available for download (http://sirl.stanford.edu/software/).
Figure 2.2. DTI tractography and tools (Sherbondy, Akers et al. 2005).
These improved tractography algorithms and analysis tools were integrated into our software for distribution to Consortium investigators as an update to our mrDiffusion package. For instance, Consortium investigators Posner and Neville plan to use our software tools in their research (through efforts by Akers and Dougherty, working with Yalchin Abdullaev, Jolinda Smith and Mark Dow at the University of Oregon Lewis Center for Neuroimaging.) These software development efforts also have been shared with the scientific community broadly. A paper describing the new white matter path-finding methods and algorithms that we developed is currently under review for publication. Further, Akers collaborated with NIH staff to incorporate the DTI visualization methods into the NIH neuroimaging tools, AFNI. This tool is used in hundreds of labs around the world.
Section 3: White Matter Pathways, Reading, and Arts Training
The third component of our research is focused on making specific behavioral and neural measurements. Identifying the main axon fiber tracts in each brain is an important step in tracking brain development. Therefore, Arvel Hernandez developed a protocol for defining specific axonal fiber pathways in each child’s brain and applied that protocol to all the brains.
The survey data from the 49 children who participated in our NIH-supported longitudinal study of reading development inform us about the parents’ and children’s experience with the arts. Described below are results of studies connecting the behavioral measurements with brain structure.
Our initial effort has been to identify the major pathways of axons that pass through the brain’s corpus callosum (Figure 3.1), the structure that connects the brain’s left and right hemispheres. This major commissure, connecting the human brain’s two hemispheres, is comprised of more than 300 million axons. From anterior to posterior (front to back), the fibers in the corpus callosum systematically connect different regions of the brain’s cortex. We have adapted and extended a callosal segmentation scheme proposed by (Huang, Zhang et al. (2005). This method allows us to segment individual corpus callosi and measure development that is localized to specific white matter (axonal) pathways.
Figure 3.1. Segmenting a child’s corpus callosum.
A callosal segmentation in the brain of one child is shown in Figure 3.1. The different colors correspond to the zones where the various axon fiber bundles pass through the callosum. We expect that these zones develop at different rates, and that their diffusion properties will correlate with distinct aspects of cognitive development. For example, visual and auditory information must be integrated rapidly between the brain’s two hemispheres in order to create a unified percept, while axon fibers that communicate between executive frontal lobe control-circuits may not need fast electrical conduction velocities.
Thus, across all humans, the brain’s sensory pathways should exhibit diffusion properties that are indicative of large, fast-conducting axons when compared to the diffusion properties of the executive control pathways. Further, different sets of callosal pathways are likely to be important for different cognitive and behavioral skills. For example, we expect the pathways connecting the brain’s temporal lobes to be important for sound and music perception, and the pathways connecting visual cortices to be important for vision. Thus, individual variation in the diffusion properties in these pathways may vary according to the relevant behavioral measures.
The callosum itself is a dynamically changing brain region between the ages of seven through the early teens, exhibiting substantial changes in size and shape during this time (e.g., Thompson, Vidal et al. 2001). These morphological changes reflect changes at the cellular level (Aboitiz, Scheibel et al. 1992). We expect that many of these developmental changes are specific to particular pathways at different ages. For example, Thompson et al. suggest that temporal and parietal pathways in the posterior callosum are changing the most in children during their development from 6 to 12 years of age, possibly reflecting the development of complex linguistic and visuo-spatial reasoning ability.
These measurements, however, do not identify which specific axon fiber pathways are developing. Further, there is a distinct gap in the existing measurements. Studies in the brains of people who have died (post-mortem studies) provide excellent data at the cellular level, but behavioral or physiological measurements cannot, of course, be undertaken. Conversely, the gross anatomical studies (at the macroscopic, rather than microscopic, level) such as those of Thompson et al. provide compelling data on living subjects, but cannot go beyond gross morphology (form and structure). To fill this gap, we are using DTI to measure development of the brain’s white matter. DTI shows networks of axons by analyzing the diffusion of water molecules in the brain, which tend to diffuse along the axons. Because the diffusion of water in the brain acts as a probe for microscopic tissue structures, DTI provides a non-invasive window into the cellular changes in humans that occur during development.
Using these tools, we discovered that strong and weak readers differ at a particular location within the posterior segment of the corpus callosum. The diffusivity of water in the direction perpendicular to the callosal fibers is highly correlated with phonological awareness and reading skill (Ben-Shachar, Dougherty et al. 2007; Dougherty, Ben-Shachar et al. 2007).
We discovered that diffusion in the axon fiber bundles that connect the brain’s temporal lobes is most strongly correlated with phonological awareness, an auditory skill that is reliably correlated with reading ability. Children with better phonological awareness skills have fewer and perhaps larger axons passing through the part of the callosum that connects the temporal lobes. While the arts training measures did not correlate with the first year callosal measurements, the behavioral correlation between music training and reading fluency improvement described in our studies above suggests that we may detect a relationship between music training and the development of the diffusion properties. We are currently in the process of making such measurements.
Section 4: Future directions- arts training and math abilities
Our group has acquired longitudinal imaging and behavioral measurements on a group of children with a wide range of reading skills. That study, funded in large part by the NIH, is now at its final year of data acquisition. In analyzing the Dana-supported study results of the arts questionnaires with our behavioral measures, we measured variables that were not part of the NIH study. As a result, we noticed an interesting relationship between visual arts experience and math skills (measured by the Woodcock–Johnson III Calculation test). The correlation between visual art experience and Woodcock-Johnson result is shown in Figure 4.1.
We also identified a moderate correlation (r=0.40, p<0.01) between music experience outside of school and how well children could remember a series of numbers (CTOPP Memory for Digits test). That test is a reliable measure of working (short-term) memory, which is important for cognitive functions including calculations with larger numbers.
Figure 4.1. Visual art experience is correlated with math skills. The horizontal axis shows a weekly average of hours spent on visual art activities in school by y1. The vertical axis shows an age standardized score of calculation measured in year 1 (Woodcock-Johnson III). Pearson r = 0.34 (± 0.14), p=0.017, N=48. The gray line is fitted to the data points minimizing the mean least square error.
Our estimation of math skills was limited to one test, which was administered only in the first year of our study, given that our original focus had been on reading development. To further explore the relationship between arts experience, math skills, and brain structures, we are measuring performance on a variety of math skills. Ultimately, we will examine the relationship between these measures and the longitudinal DTI measurements and arts questionnaires.
The new mathematical skills measurements are based on two major findings on arithmetic in the adult brain. First, it has been shown that recalling answers from memory to arithmetic questions and fully calculating answers are dissociable functions (Dehaene and Cohen, 1997). Second, adults use different brain areas to estimate arithmetic sums compared with calculating exact answers (Dehaene, Spelke et al., 1999). These findings suggest that three different types of arithmetic processing exist in the adult brain: recall, estimation, and exact calculation.
Based on these findings, we have piloted a task that may clarify how different arithmetic processes develop. In the task, participants see an arithmetic problem on a computer screen along with two answer options. They select the correct answer by pushing a button corresponding to the side of the screen on which the answer appears. We present three types of problems in blocks: simple multiplication and additions problems that can be answered by recalling memorized facts; two-digit addition problems in which the participant selects the exact answer; and two-digit addition problems in which the participant chooses the answer option that best approximates the right answer (estimation task). In addition to the computerized task, we have included a standard age-normed assessment of math ability, the WRAT-4 (Wide Range Achievement Test 4). This will be a measure of math achievement as it is traditionally tested.
Our pilot data from the computer task show that children spend different amounts of time answering the different types of problems, such that the problems recalled from memory take the shortest amount of time, the estimation problems the second shortest amount of time, and the exact calculation problems the longest.
As we continue to explore mathematical skills, we will be able to analyze whether these differences develop with age, how they relate to performance on other cognitive measures, and how they relate to white matter brain structures. The specific neuroanatomical hypothesis we are investigating concerns the anterior segment of the arcuate fasciculus (see Figure 4.2). This is a large pathway that connects cortical regions consistently activated in arithmetic tasks. This bilateral fiber tract may be important for working memory as well as for mathematical skills. We are in the processing of identifying this tract in the brains of all the children in our study.
Fig. 4.2. Dorsal fibers of the arcuate fasciculus connecting the inferior frontal cortex with inferior parietal and posterior temporal cortex in a single child (7 year old female good reader). The anterior segment (cyan) is found bilaterally, while the long segment (orange) is found in the left hemisphere only.
Our Consortium research resulted in three advances. First, we developed and distributed advanced software tools for analyzing and visualizing DTI data. These tools can be used to study the neural basis for the development of cognitive skills. Second, our data on the relationship between arts training and cognitive development revealed a correlation between the amount of music training and the amount of improvement in reading fluency in children. Third, we discovered a correlation between brain structure—the diffusion properties of a part of the corpus callosum—and reading ability as well as phonological awareness, an auditory skill that is closely related to reading ability.
Additionally, we have preliminary data on a relationship between mathematical skills and the arts. We have noticed a surprising correlation between visual arts training and math calculation skill, as measured by the Woodcock-Johnson III Calculation test. We have piloted a new set of experimental studies, based on the cognitive hypotheses from Dehaene and Spelke, and are undertaking studies to determine whether the anterior segment of the arcuate fasciculus might be a biological indicator of mathematical skill development, and whether this same segment may be encouraged to develop by training in the visual arts.
back to top
Aboitiz, F., A.B. Scheibel, et al. (1992). “Individual differences in brain asymmetries and fiber composition in the human corpus callosum.” Brain Res 598(1-2): 154-61.
Akers, D. (2006). CINCH: A Cooperatively Designed Marking Interface for 3D Pathway Selection. UIST Montreux, Switzerland.
Ben-Shachar, M., R.F. Dougherty, et al. (2007). “White matter pathways in reading.” Curr Opin Neurobiol 17(2): 258-70.
Dehaene, S. and L. Cohen (1997). “Cerebral pathways for calculation: double dissociation between rote verbal and quantitative knowledge of arithmetic.” Cortex 33(2): 219-50.
Dehaene, S., E. Spelke, et al. (1999). “Sources of mathematical thinking: behavioral and brain-imaging evidence.” Science 284(5416): 970-4.
Dougherty, R.F., M. Ben-Shachar, et al. (2007). “Temporal-callosal pathway diffusivity predicts phonological skills in children.” Proc Natl Acad Sci USA 104(20): 8556-61.
Huang, H., J. Zhang, et al. (2005). “DTI tractography based parcellation of white matter: application to the mid-sagittal morphology of corpus callosum.” Neuroimage 26(1): 195-205.
Sherbondy, A., D. Akers, et al. (2005). “Exploring connectivity of the brain’s white matter with dynamic queries.” IEEE Trans Vis Comput Graph 11(4): 419-30.
Thompson, P.M., C. Vidal, et al. (2001). “Mapping adolescent brain change reveals dynamic wave of accelerated gray matter loss in very early-onset schizophrenia.” Proc Natl Acad Sci U S A 98(20): 11650-5.
Victor, J., M. Rothbart, et al. (2003). Developing an Integrated Assessment of Children’s Temperament and Personality. Biennial meeting of the Society for Research in Children
back to top