The Impact of Musical Training on Musical Abilities in School-Aged Children. Averil Parker

Transcription

1 The Impact of Musical Training on Musical Abilities in School-Aged Children Averil Parker Presented in Partial Fulfillment for the Requirements for the Degree of Bachelor of Arts Honours Psychology Concordia University April 2017

2 ii Abstract Averil Parker The Impact of Musical Training on Musical Abilities in School-Aged Children There are many different musical abilities which may develop to different degrees in different people. There is no reason to assume that musical training in childhood will impact all musical abilities equally. Two abilities which involve melody perception are melody discrimination (MD), our ability to distinguish between two melodies, and transposed melody discrimination (TMD), our ability to distinguish between two melodies when they are transposed. Researchers have shown that MD and TMD are sensitive to musical training in adult samples. In this study children aged 7 9 years old, with and without musical training, were assessed on tasks measuring MD and TMD in order to assess whether musical training impacts these abilities uniformly, or differently. We found no evidence to support that musical training impacts MD and TMD differently. Abilities involving melody perception, therefore, may be uniformly impacted by musical training in children.

3 iii Acknowledgements It takes a village to raise a scientist, and so I have many people to thank. Thank you to the Natural Sciences and Engineering Council of Canada (NSERC) for contributing funds so that this project, and many others, could take place. A big thank you to my research supervisor, Dr. Virginia Penhune, whose guidance and support is unfailing. I could always count on Dr. Penhune to give me honest feedback and advice, and she was always there for me when I was lost. Thank you to Kierla Ireland, the PhD candidate in our lab with whom I worked closely throughout this project. Kierla is the best team leader I have had the pleasure to work with, who s positive attitude and commitment to realistic goals motivated me to persist through many long nights. Thank you to the Penhune Kids Power Team, I am so glad we all had the opportunity to work side by side. Thank you to team members Genvieve Salendres and Thanya Iyer for all their data collection and recruitment efforts and for their emotional support. Thank you to the members of the Penhune Laboratory for Motor Learning and Neuroplasticity, whose wisdom and feedback was a great asset. Thank you to Sacha Engelhardt for his loving support and patience. Averil Parker, April 2017

4 iv Table of Contents Introduction..1 Method.6 Participants...6 Procedure.6 Measures..8 Statistical analyses.10 Results 11 Group differences...11 Data screening and assumptions ANOVA...12 Correlations 12 Discussion..15 References.. 22 Appendix A: Computer Display for Melody Discrimination Task Appendix B: Factorial Design Summary Statistics Appendix C: Analysis of Variance Source Table.. 27

5 1 The Impact of Musical Training on Different Musical Abilities in School-Aged Children Musical performance requires a wide range of perceptual and cognitive abilities (Sarkamo, Tervaniemi, & Huotilainen, 2013). Two factors that contribute to musical performance are age and training. Young infants and neonates have rudimentary perceptual abilities pertaining to music, such as the ability to discriminate between pitches and between consonant and dissonant intervals (Trainor & Corrigall, 2010). Musical abilities continue to develop into middle childhood as children are exposed to the music of their culture. This development does not depend on formal training, rather, it is the result of regular exposure to music. But there is evidence for an impact of musical training on the development of musical abilities. In longitudinal studies investigating musical ability in children, researchers have found behavioral advantages and structural differences in musicians compared to non-musicians after, but not before, musical training (Hyde et al., 2009; Habibi, Cahn, Damasio, & Damasio, 2016). The objective of the current study is to investigate the impact of musical training on melody perception in school-aged children. Consistent with previous studies (Bartlett & Dowling, 1980; Dowling & Fujitani, 1971; Foster & Zatorre, 2010), our test of melody perception has two conditions which will assess two different abilities: melody discrimination and transposed melody discrimination. Melody discrimination is the ability to perceive differences between two melodies. Transposed melody discrimination is the ability to perceive differences between two melodies where the second one is transposed, that is, the melody is shifted higher or lower in musical space. We hypothesized that musical training will have a greater impact on transposed melody discrimination compared to melody discrimination. Music is organized sound. Basic elements of sound, such as loudness and pitch, are combined in meaningful ways to create higher-order concepts, such as melody (Levitin, 2006a). A melody is made up of tones, which are discrete units of sound. More specifically, a melody is a sequence of tones that unfolds over time. A melody has three basic components: pitch, contour and intervals (Dowling & Fujitani, 1971). Pitch is the degree of highness or lowness of a note (Levitin, 2006a). For example, notes emanating from a tuba will be low in pitch, whereas notes emanating from a piccolo will be high in pitch (Yantis, 2014). Contour is the sequence of ups and downs in a melody (Levitin, 2006a). For example, an ascending C major scale has all ups and the descending form of a musical scale has all downs. An interval is the distance between the pitches of two consecutive notes. Infants have rudimentary perceptual skills related to melody. Indeed, even before birth the fetus responds to changes in pitch of approximately one octave (Trainor & Corrigall, 2010). Young infants can detect changes in pitch as small as one semitone and pitch discrimination reaches adult levels at approximately 8 10 years old. (Stalinski & Schellenberg, 2012). Infants can also perceive consonance

6 2 and dissonance. At two months old, infants prefer consonant to dissonance intervals (Stalinski & Schellenberg, 2012), and can more easily detect changes in melodies with consonant than dissonant intervals (Trainor & Corrigall, 2010). Beyond these initial abilities, children become more skilled at processing features of melody specific to their culture with increased exposure to music. For example, adults have difficulty detecting changes in the tuning of melodies constructed from non-native scales, while six month old infants are equally good at detecting changes in the tuning of melodies constructed from native and non-native scales (Stalinski & Schellenberg, 2012). Children become skilled at perceiving features of melodies specific to their culture between the ages of 5 13 years old. This age range, then, is of interest when investigating the development of melody perception. Melody perception involves organizing a sequence of tones by their pitch, contour, and intervals. Tests of melody perception have previously consisted of presenting participants with two melodies and asking them to indicate whether they are the same or different (Bartlett & Dowling, 1980; Dowling & Fujitani, 1971; Foster & Zatorre, 2010). The melodies are either in the same-key (simple condition) or in different keys (transposed condition). The simple condition is a test of melody discrimination (MD) and the transposed condition is a test of transposed melody discrimination (TMD). These two abilities are distinct in the strategies we use to execute them (Dowling & Fujitani, 1971) and in their neural correlates (Zatorre, Evans, & Meyer, 1994; Foster & Zatorre, 2010). Melody discrimination is the ability to tell the difference, or discriminate, between two melodies. Melody discrimination is supported by our ability to remember the absolute value of pitches, and the contour and intervals of a melody (Dowling & Fujitani, 1971). In order to detect a change from one melody to the next, we remember the pitches, contour and intervals in the first melody and compare them to the pitches, contour and intervals in the second melody. Melody discrimination, then, depends on our auditory working memory for pitch as well as mental representations of contour and intervals. Transposed melody discrimination is our ability to discriminate between two melodies when the second melody is transposed higher or lower than the first. Transposition is a musical term which means that a melody has been shifted higher or lower in musical space, the melody can also be said to be in a different key (Levitin, 2006b). Transposition is a common and useful ability. The ability to transpose melodies allows us to recognize melodies as the same even when they begin on different notes (Stalinksi & Schellenberg, 2012). For example, we can recognize the song Happy Birthday regardless of whether it is sung by an adult or a child, a bass or a soprano, etc. That is, we don t perceive the melody as being different if it is started on a higher or lower note, rather, we perceive the melody as being the same. Transposition depends on maintaining a mental representation of the contour and intervals of a melody, while disregarding the absolute values of pitches (Dowling & Fujitani, 1970). We can no longer use

7 3 pitches as a cue, as in MD, because regardless of whether the melodies are the same or different, all the pitches have been changed between the first and second melody. Thus, we rely solely on contour and interval information in order to discriminate between the two melodies. The ability TMD, then, is dependent on our ability to maintain a mental representation of the relative features of a melody, while disregarding its absolute features (i.e. the value of pitches). In summary, the difference in strategies used in MD and TMD is evidence that these are distinct abilities, and not different degrees of the same underlying ability. Researchers in neuroscience provide further evidence for MD and TMD as distinct abilities (Zatorre et al., 1994; Foster & Zatorre, 2010). In a study by Zatorre, Evans and Meyer (1994), adult participants completed three experimental conditions while changes in cerebral blood flow (CBF) were measured with functional positron emission tomography (PET) imaging. In the first condition participants listened to noise bursts, in the second condition they passively listened to melodies, and in the third condition they were asked to make judgements about the melodies they heard (e.g. was the first note higher or lower than the second note). The researchers found that passively listening to a melody was associated with increased CBF in the right superior temporal gyrus (STG). This area of the brain contains the primary auditory cortex, where basic acoustical features of sound are processed (Sarkamo et al., 2013). The STG also contains secondary auditory cortices which showed an increase in CBF when passively listening to a melody compared to noise. When participants were asked to make a judgement about a melody there was an increase in CBF in the right frontal inferior frontal lobe (IFL). The interpretation of these findings was that the right STG is associated with our ability to perceive a melody while the right IFL is associated with auditory working memory, allowing us to make decisions about a melody. Melody discrimination, then, is supported both by the right STG and the right IFL. Whereas MD is associated with increased CBF to the right STG and IFL, Foster and Zatorre (2010) found that, in adults, performance on a TMD task was associated with increased CBF to the intraparietal sulcus (IPS). The IPS is associated with visuospatial ability or, more specifically, with the mental rotation of 3D visual stimuli (Jordan, Heinze, Lutz, Kanowski, & Jäncke, 2001). Foster and Zatorre (2010) hypothesized that activation in the IPS represents the mapping of pitch onto a mental representation of space in order to maintain the relative features of a melody (contour and intervals) while disregarding the absolute features of a melody (pitch). Taken together, these findings suggest that MD is a qualitatively different task than TMD. Musicians have previously been shown to have differences in brain structure (Schlaug et al., 1995) and behavioral advantages on tasks measuring musical ability (Fujioka et al., 2004) compared to non-musicians, and therefore are a group of interest in investigating MD and TMD. Foster and Zatorre

8 4 (2010) administered MD and TMD tasks to adults with and without musical training. They found that musicians outperformed non-musicians on both tasks. Further, they found that 68% of the variance in performance could be accounted for by musical training. These findings suggest that the difference between musicians and non-musicians is due to musical training. Due to the correlational nature of this study, however, it is not possible to rule out the alternative explanation that differences between musicians and non-musicians are due to characteristics which were present before musical training began (e.g. motivation, musical aptitude). In other words, those individuals who were predisposed toward music in the first place may have been those who went on to take music lessons. More convincing evidence of the impact of musical training on musical ability comes from longitudinal studies in children. Hyde et al. (2009) demonstrated that musical training was associated with increases in performance after 15 months of music training. Two groups of children completed tests of musical ability and underwent magnetic resonance imaging (MRI) at two time points approximately 15 months apart. At Time 1 children were 5 6 years old. One group began weekly private lessons (instrumental group) and another group participated in weekly group music classes as a part of the school curriculum (control group). Both groups had no musical training prior to Time 1. Musical ability was assessed with a melody and rhythm discrimination test battery where participants heard melodic and rhythmic phrases and were asked to identify if two phrases were the same or different. The scores from the melody and rhythm discrimination tests were combined together in a single behavioral measure which they termed auditory-musical discrimination. The researchers found that in 15 months the instrumental group progressed more than the control group on the test of auditory-musical discrimination. Additionally, the instrumental group showed greater increases in brain volume in motor areas, the corpus callosum, and the right primary auditory region compared to the control group. Increases in volume of the right auditory area was positively correlated with auditory-musical discrimination scores. These results are strong evidence that musical training impacts musical ability and brain structure in children. The use of a composite measure of musical ability (i.e. musical-auditory discrimination), however, makes it impossible to tease apart the effects of training on different components of musical ability. Results from more recent longitudinal research by Habibi and colleagues (2016) address this problem. Children aged 6 7 years old completed an MD task initially (time 1) and again two years later (time 2). At time 1, one group of children began participation in an orchestra (music group). In a comparison group children participated in another extra-curricular activity (sports group) and in another comparison group, children did not participate in any extra-curricular activities (no-training group). The researchers found that at time 2, children in the music group outperformed both comparison groups on

9 5 tests of MD. This is evidence that the impact of musical training is specific and not due to cognitive benefits accrued by simply any enriching activity, such as sports. Taken together, the studies by Hyde and colleagues (2009) and Habibi and colleagues (2016) provide evidence that musical training improves MD in children. To our knowledge there is only one developmental study investigating transposition ability in children. Sutherland, Paus, and Zatorre (2013) administered a TMD task to adolescents aged 14.5 years old. The participants were non-musicians and so it was not possible to determine the impact of musical training. They did find that, with an average of 56.7% correct on the task, adolescents performed similarly to the non-musician adults in Foster and Zatorre s (2010) study. These findings demonstrate that, although difficult, it is possible for adolescents to solve the TMD task and that performance reaches adult level by at least age 14.5 years old in non-musicians. The objective of the current study was to replicate, in children aged 7 14 years old, the behavioral differences found between adult musicians and non-musicians on different tests of musical ability (Foster & Zatorre, 2010). Specifically, we assessed whether or not the advantages on MD and TMD tasks seen in adult musicians compared to non-musicians are seen in children with musical training compared to those without training. Further, we assessed whether musical training impacts MD and TMD differently. We predicted that (1) scores on both tasks will increase with age; (2) MD scores will be higher than TMD scores; (3) children with musical training (musicians) will outperform children without training (non-musicians); and (4) the difference between musicians and non-musicians will be greater for the TMD task than for the MD task.

10 6 Method Participants Children (N = 174, 94 girls) aged 7 14 years old (M = 10.26, SD = 1.81) were recruited from science and music camps in Montreal, Ottawa and Waterloo. Participants were recruited as a part of a larger study investigating the impact of early and late training on musical ability in childhood. Parents signed a consent form and children gave verbal assent to participate in this study. Participants received a gift card and a small toy as compensation for their time. Children were categorized as musicians (n = 117) or non-musicians (n = 57). Using guidelines set in place by Skoe and Kraus (2013), a child was originally considered a musician if that child had 3 consecutive years of music lessons and practice. A non-musician was a child with 2 consecutive years of weekly music lessons and practice. Music lessons were weekly, private music lessons of at least 30 minutes in duration, taught by an expert. Music practice had to occur at least once a week outside of lessons and could be structured (e.g. using a book) or unstructured (e.g. playing along with favorite songs). In our sample we found a small number of children with between two and three years of music lessons and practice. In order to include these children in our sample we changed our operational definition of a musician to a child with 2.5 years of music lessons and practice, and of non-musician to a child with 2.5 years of music lessons and practice. Given that including these children results in an average years of music lessons for musicians (M = 4.19, SD = 1.09) and non-musicians (M = 0.49, SD =.78) which was within our original parameters (i.e. 3 years of music lessons for musicians and 2 years of music lessons for nonmusicians), we didn t expect that including these children will change the overall pattern of our results. The number of non-musicians aged years was too small to represent a meaningful group, so those children were excluded from our analyses. Our final sample was thus reduced to 94 participants (44 boys, 50 girls), divided into three age groups of 7 year olds (n = 26), 8 year olds (n = 31) and 9 year olds (n = 37) with an average age of 8.53 years old (SD =.94). A total of 50 children were musicians and 44 children were non-musicians. We calculated a priori power using G*Power 3.1 (Faul et al., 2009). In order to achieve a power of.80, we would have needed to collect data from 108 participants. With 94 participants, we concluded that the statistical power in the current study is lower than what would be ideal, but not by a drastic amount. Participant characteristics are reported in Table 1. Mother s level of education was used as a proxy for socioeconomic status (SES) of the child. In our sample mother s had an average of years of education (SD = 2.91) which is equivalent to a Master s degree (see Survey of Musical Interests, below). Ethical approval was granted by the institutional review board of Concordia University. Procedure

11 Table 1 7 Participant characteristics for musicians and non-musicians. Musicians M (SD) / n (%) Years of Music lessons 4.19 (1.09) Mother s highest level of education High school 1 (2.08) College 1 (2.08) Bachelor s 18 (37.50) Master s 16 (33.33) Ph.D. 9 (18.75) M.D. 3 (6.25) Gender Boys 17 (34.00) Girls 33 (66.00) Non-musicians M (SD) / n (%) Years of music lessons 0.49 (.78) Mother s highest level of education High school 0 (0.00) College 3 (7.31) Bachelor s 13 (31.71) Master s 16 (39.02) Ph.D. 7 (17.07) M.D. 2 (4.88) Gender Boys 27 (61.36) Girls 17 (38.64) Note. Mother s highest level of education achieved was used as a proxy for socioeconomic status (SES).

12 8 The researchers tested each child individually. The total testing duration was minutes. Participants were offered breaks and snacks on an individual basis, according to their needs. As a part of the larger study, children were administered a battery of tests assessing various aspects of musical and cognitive ability. Only the measures used for this thesis will be described below. The MD and TMD tasks were administered on a laptop which ran presentation software (Neurobehavioural Systems, To keep the storyline consistent (described below) the MD task was always administered before the TMD task. Auditory stimuli were presented binaurally via headphones. Measures Survey of Musical Interests. Parents completed the Survery of Musical Interests which assessed demographics and the child s history of musical experience (Desrochers, Comeau, Jardaneh, & Green- Demers, 2006). Questionnaire items assessed age of onset of training, months of consecutive music lessons, and hours of weekly practice. An item which assessed the mother s highest level of education was used as a proxy to measure SES of the child. Options were High school, College, Bachelor s, Master s, Ph.D., or M.D. For ease of analysis, we converted this data into number of years of education of the mother. Specifically, High school was 12 years, College was 14 years, Bachelor s was 16 years, Master s was 18 years, and Ph.D./M.D. was 23 years. Discrimination Tasks. The Melody Discrimination and Transposed Melody Discrimination tasks were adapted from an adult version of the tasks created by Foster and Zatorre (2010). In both tasks, children were presented with two melodies and were asked to indicate whether the second melody was the same as or different from the first. Melodies were in the western major scale and contained between five and 11 notes. Each note was 320 ms in duration. Pitches in the melodies were between C4 (middle C on a piano) and E6. To make these tasks more fun and interesting, and to help children understand the instructions, children were given a storyline. We assessed the internal consistency of scores using the Kuder-Richardson formula (KR-20), which is a Cronbach s alpha for a dichotomous outcome variable. We calculated the KR-20 for scores of 7-, 8-, and 9-year-old musician and non-musicians, for each task. For MD, musicians scores in different age groups ranged from and non-musicians scores ranged from For TMD, musicians scores in different age groups ranged from and nonmusicians scores ranged from These scores indicate acceptable reliability, with the exception of the scores of non-musicians on the TMD task. This limitation will be discussed in the discussion section. Children were cued with a visual display on the computer screen see Figure A (Appendix A) and told that Teacher Elephant was teaching her students how to sing. Teacher Elephant always sang the first melody. The Echoing Elephant repeated the melody back exactly as he heard it whereas the

13 9 Forgetful Monkey always made a little mistake. Children were asked if the second melody was sung by the Echoing Elephant (same) or the Forgetful Monkey (different) and they indicated their response by clicking the left or right button on a computer mouse. Their response was recorded as correct or incorrect by the computer. For each task, children performed four practice items before performing the test trials. Two items were guided by the researcher, who gave the child feedback. Then children completed two practice items by themselves. During the practice items headphones were off and unplugged, and during test trials headphones were on and plugged in to the computer. Children completed two blocks of 15 test trials each for a total of 30 trials per task. After each item (practice and test trials), the computer screen displayed either incorrect or correct, giving the participants immediate feedback on their performance. In the MD task a total of 16 trials were different and 14 trials were same. Melodies in the different trials were changed by one note. The altered note was shifted up or down up to five semitones. Key and contour were maintained in the different melodies. In the TMD task a total of 17 trials were different and 13 were same. The pitches of the second melody were shifted (i.e. transposed) four semitones higher in both same and different trials. In different trials, aside from the transposition, one note was altered by one semitone to a pitch outside the melody s new key. Contour was maintained in different melodies. Because all the pitches were changed between the first and second melodies, and the contour was the same for both the first and second melodies, pitch and contour were removed as cues. Participants, then, had only the intervals between the notes as a basis for discriminating between the two melodies, making our test of TMD a challenging task. This was intentional, to create a task which was sensitive to a wide range of musical experience. The goal was to prevent a ceiling effect in musicians (who may find the task too easy) and a floor effect in non-musicians (who may find the task too hard; Foster, 2015 personal communication). The storyline was modified in the TMD task. Children were told that Teacher Elephant is still there, but now she is teaching Baby Elephant and Baby Monkey how to sing, and they sing in a much higher voice. This was to explain to the children why the second melody sounds higher than the first melody, and also to help them understand that they should ignore that feature of the stimulus and focus on whether or not there are any additional changes between the first and second melody. Note that it made sense to administer the TMD task after the MD task in order to make salient the small size of the students, lending credibility to their status as babies and therefore increasing the ease of understanding the instructions. However, this opens our study up to possible practice effects. This limitation will be discussed in the discussion section.

14 10 Performance on both tasks was assessed with proportion correct. Each trial was recorded by the computer as a 0 (incorrect) or a 1 (correct). Correct trials were divided by the total number of trials on that block. The final score for each task was an average of the scores from the two blocks. Higher scores indicated better performance on the task. The average proportion correct for each task was the raw data for analyses. An item analysis was conducted to address the unequal distribution of same and different trials. We retained 10 same trials and 10 different trials for each task which had the highest point biserial correlations. Statistical analyses All analyses were performed in SPSS except eta-squared which was estimated using JASP ( A mixed ANOVA was used to test the hypothesis that musical training impacts different musical abilities differently. Specifically, we expected that the difference between musicians and non-musicians would vary across MD and TMD tasks. Bivariate correlations were calculated in order to investigate the impact of duration of training on performance on the task. Specifically, we expected that as duration of training increases, performance on the task will also increase. It is important to note that age and duration of training are confounded in this study. Specifically, as children get older they have also been taking music lessons for longer. In order to investigate the relation between duration of training and performance on the task in a way that accounts for age, we created age-based norm scores. These were z-scores such that each age group (7, 8, and 9 years olds) had a mean score of zero with a standard deviation of one, for each task. Importantly, we used the means and standard deviations (in raw scores) of non-musicians when we calculated the z-scores of musicians. Thus, non-musicians were a reference group, and musicians z-scores could be interpreted as how a musician s performance compares to the average non-musician relative to their age group. Because we were assessing the impact of musical training on performance, only musicians scores were included in the correlations. We used a one-tailed test because we had an a priori prediction about the direction of the relationship between musical training and performance on the task. Specifically, that there would be a positive relation between these variables.

15 11 Results Group differences. We performed a goodness of fit chi-square test and independent samples t- tests to determine whether there were any differences in group characteristics between musicians and non-musicians. An independent samples t-test revealed that there were no statistically significant differences in SES between musicians and non-musicians, t (87) =.20, p =.85, d =.04. A chi-square test revealed statistically significant group differences in gender distribution, (1, N = 94) = 7.04, p =.01, =.27, such that there were more girl musicians (n = 33, 66.00%) than girl non-musicians (n = 17, 38.64%) and more boy non-musicians (n = 27, 61.36%) than boy musicians (n = 17, 34.00%). An independent samples t-test revealed no statistically significant differences in performance between boys and girls on either the MD, t (91) =.62, p =.54, d =.13, or TMD, t (91) =.95, p =.35, d =.21, tasks. We found no reason, then, to control for either SES or gender in subsequent analyses. Data screening and assumptions. We calculated the descriptive statistics (mean, standard deviation, and minimum and maximum values) for performance on the MD and TMD tasks and for age. We found that there were no out-of-range, incorrect or implausible values so we concluded that there were no data entry errors. There were no participants with z-scores beyond an absolute value of 3, and so we concluded that there were no outliers. There was one participant with missing data, and they were excluded from the analyses listwise, reducing the sample from 94 to 93 participants. Assumptions for a mixed ANOVA include independence, normality, sphericity and homogeneity of variance. There were no issues with independence, normality, or sphericity. We visually inspected the boxplots of performance on the MD and TMD tasks for 7-, 8-, and 9-year-old musicians and nonmusicians to determine if the scores had equal variances. Seven year-old musicians had a larger variance in scores relative to the other groups. For a more rigorous investigation of variance, we looked at the numerical value of the variances (s 2 ) for 7-, 8-, and 9-year-old musicians and non-musicians, for each task. Consistent with the boxplots, 7-year-old musicians varied more in their performance on the MD task (s 2 =.0361), relative to other groups (s 2 = ). Unequal variances may increase the chance of detecting a significant difference where there is none, so we should interpret the results of significance tests cautiously. This limitation will be discussed further in the discussion section. Prior to running bivariate correlations, variables of interest were assessed for normality and heteroscedasticity. No skew index exceeded an absolute value of 3, and no kurtosis value exceeded an absolute value of 10 and so we concluded that there were no severe departures from normality. Upon visual examination of boxplots, it was apparent that lessons in months was much more variable than either performance on the MD task (in z-scores) or performance on the TMD task (in z-scores). To remedy this issue, we transformed lessons in units of months to lessons in units of standard deviation

16 12 (i.e. z-scores), thereby making the variances of the variables of interest more similar. Analysis of variance. To investigate whether or not musical training impacts MD and TMD differently we conducted a mixed analyses of variance with two between-subjects factors: age (7-, 8- or,9-year old) and group (musician or non-musician); and one within-subjects factor: task (MD and TMD). The outcome variable was proportion correct on the task. Means and standard deviations are presented in Table B (Appendix B). There was a statistically significant main effect of group (F (1, 87) = 39.56, p >.001, η 2 =.28) such that, across tasks and across ages, musicians outperformed nonmusicians. There was a statistically significant main effect of task (F (1, 87) = 46.46, p >.001, η 2 =.34) such that, across ages and across musicians and non-musicians, MD was an easier task and yielded higher scores than TMD. There was a statistically significant main effect of age (F (2, 87) = 4.60, p =.01, η 2 =.06). Pair-wise comparisons revealed that 7 year olds performed statistically significantly worse than 9 year olds (p =.02, d =.80). The group by task interaction was not statistically significant (F (1, 87) =.39, p =.54, η 2 =.003, see Figure 1). This result means that musical training did not impact MD and TMD tasks differently and, therefore, our main hypothesis was not supported. There was a statistically significant group by age interaction (F (2, 87) = 3.37, p =.04, η 2 =.05, see Figure 2). Pairwise comparisons revealed that 7-year-old musicians performed statistically significantly worse than 9- year-old musicians (p =.01, d =.87) while there were no statistically significant differences in performance across age groups of non-musicians. This suggests that for musicians, performance on both tasks increases with increasing age, while the performance of non-musicians does not increase with age. Correlations. To further investigate the impact of musical training on musical ability we conducted bivariate correlations between musicians performance on the tasks and duration of training. We found a statistically significant correlation between performance on the MD task in units of standard deviation and duration of musical training in units of standard deviation (r (47) =.27, p =.03), such that as duration of training increased performance on the task also increased. We also found a statistically significant correlation between performance on the TMD task and duration of musical training (r (47) =.25, p =.045) such that as duration of musical training increased, performance on the task also increased.

17 Proportion correct MD TMD Musician Non-musician Figure 1. Depicts a bar graph illustrating that the difference between musicians and non-musicians did not vary across tasks. Our main hypothesis, that musical training impacts MD and TMD differently, was not supported. Error bars are standard error of the mean.

18 Proportion correct * Musician Non-Musician Age in years Figure 2. Depicts a bar graph illustrating a group by age interaction. Musician 9-year-olds outperformed musician 7-year-olds while there were no age-related differences in performance for non-musicians. This suggests that musicians increased in performance as they get older while non-musicians did not. Error bars are standard error of the mean.

19 15 Discussion In this study we assessed the performance of children with and without musical training on tasks measuring MD and TMD in order to determine whether musical training impacts these tasks uniformly, or differently. We predicted that that MD would be an easier task than TMD, and this was supported in this study. This is consistent with previous research investigating these tasks in adults (Foster & Zatorre, 2010). As mentioned above, the TMD task was likely more difficult because all the melodies were created to maintain the contour between the first and second melodies. Removing contour forced participants to focus solely on interval structure as the differentiating feature between melodies. Contour is a salient cue (Dowling, 1978). Our results suggest that in children, as in adults, interval structure is a more difficult feature to discriminate than melodic contour. We predicted that musicians would outperform non-musicians, and this was supported in this study. This is consistent with the results of Foster and Zatorre (2010), who showed that adult musicians outperformed non-musicians on both tasks. Thus the differences between musicians and non-musicians observed in adults are preserved in children. Additionally, approximately 4 years of musical training was sufficient to produce these differences. This is consistent with the results of a longitudinal study by Hyde and colleagues (2009) who demonstrated that children with 15 months of musical training had changes in brain structure and musical ability, including MD, compared to children with no training. This is the first study, to our knowledge, demonstrating that the TMD task is also sensitive to musical training in childhood. Our main hypothesis, that musical training would impact MD and TMD differently, was not supported in this study. This was evidenced by a non-significant task by group interaction, with an effect size which indicated little or no practical significance. It may be that MD and TMD are too similar to be differentially impacted by musical training. We argued that because these tasks have different cognitive and neurological correlates, they might develop differently. Melody discrimination involves the use of pitch, contour and intervals as auditory cues (Dowling & Fujitani, 1971), and is supported by the right superior temporal gyrus and the right inferior frontal lobe (Zatorre et al., 1994). Transposed melody discrimination relies solely on contour and interval structure as auditory cues (Dowling & Fujitani, 1971), and is supported by the intraparietal sulcus (Foster & Zatorre, 2010). Nevertheless, MD and TMD are both perceptual abilities, and they both involve perceiving melodies specifically. In this study the impact of musical training on melody perception was not specific to a certain type of melody, but rather was robust across different types of melodies. Therefore, MD and TMD may represent different levels of difficulty within the same underlying construct rather than distinct abilities unto themselves. Interestingly, musicians increased in performance with increasing age, while non-musicians did

20 16 not, as evident in a significant task by group interaction. The effect size for this interaction was small. This may be due to the small age range investigated in this study (7 9 year olds). It is possible that had we examined a larger range of age groups (e.g year olds) with more years of musical training, we would have observed a larger effect. Despite the small effect size, we deemed it worthwhile to explore possible explanations for the age by group interaction. One explanation is that age is a proxy for maturation or development, and that children increase in performance with increasing age. Indeed, the brain s response to auditory stimuli develops between 8 10 years old and continues to mature until years old (Ponton, Eggermont, Khosla, Kwong, & Don, 2002). Maturation did not appear to affect the results of our study, however. Children without musical training did not improve on the task with increasing age, so, in the age range we investigated, maturation did not affect performance. This is interesting because it means that it is more likely that the differences between musically trained and untrained children are actually related to musical training. To further examine the effect of training, we calculated correlations between duration of training and musicians performance on the task, using age-based norm scores. These effects were significant, but small. In other words, the duration of training has a statistically significant but practically speaking small impact on performance on the tasks. This is in contrast to studies with adults where musical training accounted for 68% of the variance in performance on musical tasks (Foster & Zatorre, 2010). This discrepancy may be due to the fact that the children in this study had an average of approximately 4 years of musical training, while in Foster and Zatorre s (2010) study, musicians were adults and had a minimum of 8 years of musical training and an average of 17,300 hours of lifetime musical practice. It may be that 4 years of training in musicians is not sufficient to produce large differences in performance compared participants with no training. Nevertheless, our results suggest that musical training has some effect on task performance after only 3 4 years. However, other factors are certainly involved. Some of these factors are discussed next. Firstly, our groups were self-selected and so we cannot rule out the possibility that pre-existing differences, such as genetics, may account for the differences in performance between musicians and non-musicians. Indeed, researchers have found that there is a genetic component to musical ability (see Tan, McPherson, Peretz, Berkovic, & Wilson, 2014). For example, Oikkonen and colleagues (2015) found an association between music perception and genes involved in the development of the inferior colliculus (IC). This is relevant to melody perception specifically because the IC is important for tonotopic organization, where different pitches are mapped onto different neuronal cells (Ress & Chandrasekaran, 2013). Individual differences in genetics which influence the development of the auditory pathway may account for a child s initial perceptual abilities and influence the ease with which

21 17 auditory skills are acquired. In our sample, it is possible that children with heightened perceptual abilities in the first place were those who pursued musical training, while those who had difficulty with musical activities, did not. Differences in genetics may also play a role in how musical abilities are expressed at different ages. Alternatively, it may be that taking music lessons interacts with the developing brain to produce changes in performance at different ages. For instance, the brains response to auditory stimuli changes between 8 and 10 years old and continue to mature until years old (Ponton et al., 2002). Using electroencephalogram (EEG) data, Ponton and colleagues (2002) found that the P1 response to auditory stimuli decreases in amplitude with age, while the N1 response emerges between 8 and 10 years old and increases in amplitude thereafter. Habibi and colleagues (2016) demonstrated that musical training can interact with the development of the brain s response to auditory stimuli and produce changes in musical ability. In their study, musicians had an accelerated maturation of the auditory brain response after only two years of musical training compared to control groups. Specifically, children who began musical training between 6 7 years old had larger decreases in amplitude of the P1 response relative to control groups after two years. Further, 76% of music group participants had developed the N1 response, compared to 46 53% of participants in control groups. These neurological differences were also reflected in an enhanced performance on a task measuring MD in the music group relative to controls. Thus, musical training altered the developmental trajectory of the brain s response to auditory stimuli, and produced a difference in performance. In our study, musical training may have interacted with normal development to produce differences in performance at different ages. With on-going training, these children might come to display the kinds of behavioral and brain differences observed in adults in other studies. This study has several limitations which should be taken into consideration. We used a crosssectional design and we are therefore limited in our ability to make inferences about the impact of musical training over time. Although it is unlikely that our study suffers from reverse causation (i.e. that increased performance on a task caused increases in musical training), our study is correlational in nature and therefore is subject to the third variable problem. Some other variable not measured may account for the differences between musicians and non-musicians, and not musical training per se. For example, it may be that musicians are simply exposed to music more than non-musicians, and this heightened exposure may account for improved melody discrimination abilities. Another limitation is that our groups were self-selected. That is, children were not randomly assigned to either a musician or non-musician group. We cannot guarantee, then, that these two groups do not differ on some important variable that may impact their performance on the tasks. Indeed,

22 18 children who pursue music lessons seem to be different from those who do not take lessons on demographic, personality and cognitive variables (Corrigall & Schellenberg, 2015). Children in our sample, however, did not differ on a range of demographic variables which may be thought to be related to the likelihood of entering musical training, such as SES. Children who identify as musicians may be more motivated to achieve on a musical task in order to maintain a consistent self-concept. Indeed, cognitive dissonance theory posits that we are motivated to maintain a consistent self-image in order to avoid the uncomfortable psychological state (i.e. cognitive dissonance) which arises when we behave in ways which are inconsistent with our self-concept (Festinger, 1957). We used a convenience sampling method so our groups were not equivalent in size, age or gender, and our findings may not apply to the wider population. Indeed, in our sample there were more girls in the musician group and more boys in the non-musician group, however we did not find any gender differences in performance on the tasks. Nevertheless, without randomization our results are less generalizable to the larger population. It is perhaps unsurprising given that we recruited children from music and science camps that our sample is from a relatively high SES group, as measured by mother s highest level of education attained. In particular, we should be cautious when generalizing to samples of lower SES. Another limitation of this study was heteroscedasticity. When sample sizes are unequal across groups an ANOVA is not robust against violations of assumptions (Field, 2009). In our study, 7-year-old musicians were the smallest group size and were more variable than other groups. If smaller groups have larger variances than larger groups, then the F-ratio is more liberal and more prone to type I errors (Field, 2009). In other words, the F-ratio is more likely to detect a difference where there is none. This is of particular importance in this study because we did find a statistically significant difference between 7- year-old and 9-year-old musicians, such that 9-year-old musicians outperformed 7-year-old musicians. This result is consistent with the our correlational finding that performance increases as a function of increased musical training, given that with increasing age musicians also accumulate more years of training. That 9-year-old musicians outperformed 7-year-old musicians is also consistent with other longitudinal studies in children indicating that MD ability improves more in children who receive musical training relative to those who don t (Hyde et al., 2009; Habibi et al., 2016). Nevertheless, this result should be replicated in other samples who are tested on MD and TMD tasks before we draw firm conclusions about the impact of musical training over time. Overall our scores demonstrated acceptable reliability, except for non-musician scores for the TMD task. As mentioned above, in the TMD task contour is maintained in different trials which makes discriminating between two melodies more difficult. It is possible that for non-musicians the differences

23 19 between the two transposed melodies is too subtle to detect, resulting in a floor effect which would account for the poor reliability. This is consistent with our finding that non-musicians had an average of.51 proportion correct on the TMD task, which is to say they performed at chance levels. This suggests that the difficulty of the task may be too high and therefore inappropriate for young nonmusicians. In future versions of this task, an easier level of difficulty could be added by incorporating different trials in which contour is violated, which should be a salient enough cue for non-musicians to detect. In this study there was a possibility of practice effects, as the MD task was always presented before the TMD task. As previously mentioned, the TMD task is intentionally more difficult than the MD task. Practice on the MD task was intended to help children to perform at their best on the TMD task. However, this introduced the methodological problem of practice effects. Given that TMD performance was at chance levels for non-musicians, it appears that children did not benefit from repeated presentations in order to better their performance on the TMD task relative to the MD task. Nevertheless, we can t guarantee that performance on the TMD task wouldn t be different if it were presented first instead of second. Despite these limitations this study has wider implications. We found that musicians outperformed non-musicians on both MD and TMD abilities, which suggests that both abilities are sensitive to musical training in school-aged children. It is therefore appropriate to include exercises that target both abilities in a music-class curriculum for school-aged children. In particular, given that nonmusicians performed so poorly on the TMD task across studies and ages (Sutherland et al., 2013, Foster & Zatorre, 2010), it may be that musical training is particularly important for the development of this skill. Music educators may be tempted to develop a more rigorous and targeted training regimen in order to develop TMD, given that it is a more difficult ability to master. However, our results show that this may not be necessary. We found that musical training impacts both MD and TMD abilities similarly. It appears, then, that rather than requiring special attention, TMD benefits from normal training. We therefore find no reason to recommend that music educators devote themselves to developing teaching methods for TMD above and beyond what is already being taught. This study opens up several possibilities for future directions. We created age-based norm scores so that a child s performance was compared to that of other children in their age group. In a future study we could use these norm scores to determine what is an average, high, or poor score for a child relative to their age group. We also calculated the z-scores of musicians using the mean and standard deviation of non-musicians. This makes it possible to quantify, in units of standard deviation, how a musician s performance compares to the average non-musician, relative to their age-group. Future researchers could

24 20 do the same for adolescent or adult musicians compared to same age non-musicians. In this way the difference between musicians and non-musicians could be compared across studies, despite differences in developmental stages. This study also generates new questions for future researchers to investigate. We found that two abilities of melody perception (MD and TMD) are similarly impacted by musical ability, but many more abilities remain. Musicality is a heterogeneous construct comprising many distinct performance and perceptual abilities (Levitin, 2012). To continue investigations into the impact of musical training on different musical abilities, future researchers could compare the impact of musical training on rhythmic versus melodic perception, for example distinguishing between different rhythms versus distinguishing between different types of melodies (e.g. same-key and different-key melodies). Additionally, researchers could compare perceptual versus performance abilities, for example, singing a melody versus discriminating between melodies. Researchers could also investigate the impact of different training paradigms on different musical abilities. For example, western classical music, which is the method under which our participants were trained, emphasizes harmonic complexity (Schellenberg et al., 2005), while African music emphasizes rhythmic complexity (Chernoff, 1997; Temperley, 2000). It would be interesting to do a cross-cultural study on the effects of these different training paradigms on different musical abilities. One could hypothesize that more rhythmically inclined training regimens would develop rhythmic perception and performance, perhaps to the detriment of melodic perceptual and performancebased abilities. Similarly, training with an emphasis on melody may improve melodic perceptual and performance abilities with a cost to rhythm-based abilities. Jazz music is a blend of European and African-American influences (Stump, 2000). One may hypothesize, then, that jazz training results in a well-rounded profile of musical abilities. In terms of methodology, wherever possible, future researchers should use longitudinal designs in order to make more confident claims about the impact of musical training over time. They should use random assignment in order to account for individual differences. Following Habibi and colleagues (2016), researchers should also use an inactive and active control, such as a sports or science extracurricular activity, so that they can be confident that the difference between musicians and nonmusicians is due to musical training specifically and not enrichment in general. In conclusion, musicians were equally advantaged on tasks measuring MD and TMD compared to non-musicians. In other words, abilities related to listening to and making decisions about melodies appear to be similarly impacted by musical training. The impact of musical training on melody perception, then, is robust across different types of melodies. Musical training appears to be particularly

25 21 important, however, for the development of TMD, as children without musical training perform at around chance level on that task. As a group, musicians seem to increase in performance as they get older while non-musicians do not. This may be due to an increase in musical training with increasing age, pre-existing individual differences such as genetics, or an interaction between the two. Future investigators could assess the relative importance of musical training for rhythmic compared to melodic perceptual abilities, or perceptual compared to performance abilities. These findings would be relevant to music educators, who could make more informed decisions regarding what to include, focus on, or leave out of a music-school curriculum for school-aged children.

26 22 References Bartlett, J. C., & Dowling, W. J. (1980). Recognition of transposed melodies: A key-distance effect in developmental perspective. Journal of Experimental Psychology: Human Perception and Performance, 6, Retrieved from Corrigall, K. A., & Schellenberg, G. (2015). Predicting who takes music lessons: Parent and child characteristics. Frontiers In Psychology, 6, doi: /fpsyg Chernoff, J. (1997). 'Hearing' in west African idioms. The World Of Music, 39, Retrieved from Desrochers, A., Comeau, G., Jardaneh, N., & Green-Demers, I. (2006). L élaboration d une échelle pour mesurer la motivation chez les jeunes élèves en piano. Revue de recherche en éducation musicale, 24, Retrieved from Dowling, W. J., & Fujitani, D. S. (1971). Contour, interval, and pitch recognition in memory for melodies. Journal of the Acoustical Society of America, 49, doi: / Dowling, W. J. (1978). Scale and contour: Two components of a theory of memory for melodies. Psychological review, 85, doi: / x Faul, F., Erdfelder, E., Buchner, A., & Lang, A.-G. (2009). Statistical power analyses using G*Power 3.1: Tests for correlation and regression analyses. Behavior Research Methods, 41, doi: /BRM Field, A. (2009). The theory behind ANOVA. In M. Carmichael & E. Jenner (Eds.), Discovering statistics using SPSS, third edition (pp ). Thousand Oaks, CA: Sage publications. Festinger, L. (1957). A theory of cognitive dissonance. Stanford, CA: Stanford University Press. Foster, N. E., & Zatorre, R. J. (2010). A role for the intraparietal sulcus in transforming musical pitch information. Cerebral Cortex, 20, doi: /cercor/bhp199 Fujioka, T., Trainor, L. J., Ross, B., Kakigi, R., & Pantev, C. (2004). Musical training enhances automatic encoding of melodic contour and interval structure. Journal of cognitive neuroscience, 16, doi: / Habibi, A., Cahn, B. R., Damasio, A., & Damasio, H. (2016). Neural correlates of accelerate auditory processing in children engaged in music training. Developmental Cognitive Neuroscience, 21, doi: /j.dcn Hyde, K. L., Lerch, J., Norton, A., Forgeard, M., Winner, E., Evans, A. C., & Schlaug, G. (2009). Musical training shapes structural brain development. Journal of Neuroscience, 29, doi: /jneurosci

27 23 Jordan, K., Heinze, H. J., Lutz, K., Kanowski, M., & Jäncke, L. (2001). Cortical activations during the mental rotation of different visual objects. Neuroimage, 13, doi: /nimg Levitin, D. J. (2006a). What is Music?. In This is your brain on music (pp ). New York, NY: Dutton/Penguin Books Levitin, D. J. (2006b). Foot Tapping. In This is your brain on music (pp. 78). New York, NY: Dutton/Penguin Books Levitin, D. J. (2012). What does it mean to be musical?. Neuron, 73, doi: /j.neuron Oikkonen, J., Huang, Y., Onkamo, P., Ukkola-Vuoti, L., Raijas, P., Karma, K., Vieland, V. J., & Järvelä, I. (2015). A genome-wide linkage and association study of musical aptitude identifies loci containing genes related to inner ear development and neurocognitive functions. Molecular Psychiatry, 20, doi: /mp Ress, D., & Chandrasekaran, B. (2013). Tonotopic organization in the depth of human inferior colliculus. Frontiers in Human Neuroscience, 7, doi: /fnhum Sarkamo, T., Tervaniemi, M., & Huotilainen, M. (2013). Music perception and cognition: development, neural basis, and rehabilitative use of music. Wiley Interdisciplinary Reviews: Cognitive Science, 4, doi: /wcs.1237 Schellenberg, E. G., Bigand, E., Poulin-Charronnat, B., Garnier, C., & Stevens, C. (2005). Children's implicit knowledge of harmony in western music. Developmental Science, 8, doi: /j x Schlaug, G., Jäncke, L., Huang, Y., Staiger, J.F., & Steinmetz, H. (1995) Increased corpus callosum size in musicians. Neuropsychologia, 33, doi:1.1016/ (95) Skoe, E., & Kraus, N. (2013). Musical training heightens auditory brainstem function during sensitive periods in development. Frontiers in Psychology, 4. doi: /fpsyg Stalinski, S. M., & Schellenberg, E. G. (2012). Music cognition: a developmental perspective. Topics in Cognitive Science, 4, doi: /j x Stump, R. (. (2000). Place and innovation in popular music: The bebop revolution in jazz. Journal Of Cultural Geography, 18, Retrieved from Sutherland, M. E., Paus, T., & Zatorre, R. J. (2013). Neuroanatomical correlates of musical transposition in adolescents: a longitudinal approach. Frontiers in Systems Neuroscience, 7. doi: /fnsys

28 24 Tan, Y. T., McPherson, G. E., Peretz, I., Berkovic, S. F., & Wilson, S. J. (2014). The genetic basis of musical ability. Frontiers In Psychology, 5, doi: /fpsyg Temperley, D. (2000). Meter and grouping in African music: a view from music theory. Ethnomusicology, 44(1), Retrieved from Trainor, L. J., & Corrigall, K. A. (2010). Music acquisition and effects of musical experience. In Music Perception (pp ). Springer New York. doi: / _4 Trehub, S. E. (2003). The developmental origins of musicality. Nature Neuroscience, 6, doi: /nn1084 Yantis, S. (2014). Sound and the Ears. In Sensation and Perception (pp ). New York, NY: Worth Publishers Zatorre, R. J., Evans, A. C., & Meyer, E. (1994). Neural mechanisms underlying melodic perception and memory for pitch. Journal of Neuroscience, 14, Retrieved from Zatorre, R. J. (2003). Absolute pitch: A model for understanding the influence of genes and development on neural and cognitive function. Nature Neuroscience, 6, doi: /nn1085

29 Appendix A: Computer Display for Melody Discrimination Task Figure A 25 Note. Depicts the visual display of Teacher Elephant teaching her students how to sing, which guided participants through the tasks. Presented in color during testing.