Statistical learning and auditory processing in children with music training: An ERP study
Introduction
There is growing evidence that musical training is associated with benefits in not only auditory-related tasks but also other non-trained tasks. For example, musical training is associated with benefits in cognitive processing tasks such as memory, attention, and intelligence (Ho et al., 2003, Strait et al., 2010, Schellenberg, 2011). Some of these effects can also be observed in children who are learning music. For instance, it has been reported that children with music training have enhanced skills in a variety of tasks such as pitch perception (Magne et al., 2006), non-verbal reasoning (Forgeard et al., 2008), executive functions (Moreno et al., 2011), and language-related tasks (Anvari et al., 2002, Moreno et al., 2009, Tsang and Conrad, 2011). In addition, previous studies have reported evidence of plasticity in brain structure and functioning in children with musical experience. When compared to a control group, children with music training had larger auditory evoked potentials (Shahin et al., 2004), larger anterior corpus callosum (Schlaug et al., 2009), as well as larger right precentral gyrus and right primary auditory cortex (Hyde et al., 2009).
Some have suggested that there might be evolutionary links between music and language in the brain (Wallin and Merker, 2001), with considerable overlap between neural structures involved in encoding and perception of music and language (Patel, 2011). Furthermore, both music and language consist of perceptually discrete elements that are combined into structured sequences according to highly complex regularities. It is therefore of interest to investigate if musical experience is associated with processes that facilitate language acquisition. One such cognitive process that facilitates language acquisition is implicit learning of statistical regularities – statistical learning (SL). The term SL was proposed by Saffran et al. (1996a) to describe infants’ ability to identify word boundaries solely from the statistical relationships between syllables in a continuous stream of pseudo-speech. In a series of studies, Saffran and colleagues showed that adults and infants were able to implicitly segment a continuous stream of speech by identifying statistically coherent units (i.e., adjacent units which exhibit strong statistical relationships such as high transitional probability of co-occurrence)(Saffran et al., 1996a, Saffran et al., 1996b). A large body of research in implicit learning has focussed on how knowledge of language and music structures may be acquired implicitly (for reviews see Ettlinger et al., 2011, Rohrmeier and Rebuschat, 2012). In contrast, there is a relatively small body of work on direct comparison of SL abilities in musically trained and untrained individuals. Furthermore, studying these processes in individuals with developing music skills (e.g., children who are participating in musical training) can help us to understand the potential correlates of music practice in other domains of cognition.
Cross-sectional research comparing SL in adult musicians and non-musicians has shown mixed results. Some studies have not shown any group differences on behavioural SL tasks (François and Schön, 2011, Paraskevopoulos et al., 2012) whereas others have shown that adult musicians perform better than non-musicians (Romano Bergstrom et al., 2012, Shook et al., 2013). More recently, Mandikal Vasuki et al. (2016) showed that adult musicians outperformed non-musicians on auditory but not visual statistical learning tasks. Research utilizing electrophysiological measures has shown that musicians are better at learning the statistical structure of a sung language than non-musicians (François and Schön, 2011, François and Schön, 2014). François and Schön, (2011) reported that musicians had a larger N1 familiarity effect indicating that they could recognize statistically coherent (familiar) items better than unfamiliar items when compared with non-musicians. Additionally, event-related potentials (ERPs) recorded during the familiarization of a sung language showed that the N400 learning curve saturates earlier in musicians than non-musicians, indicating faster detection of statistical regularities in musicians (François and Schön, 2014).
To date, only one study has investigated SL in children learning music (François et al., 2013). François et al. (2013) investigated auditory SL (aSL) using a longitudinal approach in 8-year old children. Twenty-four children were pseudo-randomly divided into two groups in which each group received either music or painting training for a period of two years. Two teachers, specifically trained in music or painting (one for each type of training), were recruited for the study. Each group was trained for 45 min, twice a week in the first year and once a week in the second year. Stimuli used for assessing SL were 4 tri-syllabic ‘words’ and each word was sung on a particular melodic contour (e.g., the word “pymiso” was sung as B3 E4 F4). Both behavioural and electrophysiological measures were used to assess SL. A two-alternative forced choice (AFC) task was used to assess SL behaviourally. Four part-words were created for the 2 AFC task. Each word (referred to as a familiar word because it was presented during familiarization) was paired with a part-word (referred to as an unfamiliar word because it was not presented during familiarization). While aSL improved over time in children learning music, the performance of the children learning painting remained at chance level on the behavioural measure. In addition, the familiarity effect (difference in ERPs for familiar and unfamiliar words in the 450–550 ms latency range) was larger in the music group than the painting group. The familiarity effect was interpreted as evidence that children learning music could better identify statistical regularities. The authors concluded that the enhancements in aSL may be attributed to better rhythm perception in children with music training. The longitudinal design of this study supports the possibility that music training enhances SL. However, the question of causality remains an open empirical question. It is also possible that children with higher SL may be attracted to musicianship.
To the best of our knowledge, no study has taken an individual differences approach to assess whether higher SL is associated with higher musical abilities such as rhythm or melody discrimination. Furthermore, very little is known about visual statistical learning (vSL) in children with music training, although there are a couple of studies that have shown enhanced visual sequence learning in adult musicians (Romano Bergstrom et al., 2012, Anaya et al., 2017). In the present study, we used a cross-sectional approach to examine the association between SL (both aSL and vSL) and musical training in children.
Behavioural measures of SL are limited by their inability to establish the time course of SL in the brain – they cannot distinguish whether extraction of statistical information was achieved via fast online segmentation or via more slow and variable segmentation processes. To understand the brain dynamics as SL proceeds, we need to measure brain responses like the N1 and the N400 as the stimuli containing regularities are presented during an SL task. These brain responses occur within a few hundred milliseconds of the stimuli being presented (Näätänen and Picton, 1987, Kutas and Federmeier, 2011). This necessitates the use of techniques like electroencephalography (EEG) which are capable of detecting these responses with a suitable temporal resolution such as within milliseconds (Menon et al., 1997). Thus, investigating SL using neurophysiological measures in conjunction with behavioural measures helps us to identify neural correlates of SL. In this study, we used both behavioural and online electrophysiological measures to assess SL.
Previous studies have reported online ERP measures of aSL and vSL in non-musician adults (Abla et al., 2008, Abla and Okanoya, 2009). In these studies, an embedded triplet task was used to assess SL. The embedded triplet task comprised a familiarization and a test phase. During familiarization, a stream of concatenated triplets was presented to the participants. Items within a triplet had a higher chance of co-occurrence or higher transitional probability and were therefore more predictable. Abla and colleagues reported a triplet onset effect (in the N1 and N400 response) as an index for online measure of segmentation during familiarization. The ERPs were larger in less predictable positions (initial stimulus of a triplet) than in more predictable positions (final stimulus of a triplet) which was referred to as a triplet onset effect. Individuals who performed better in the test phase of the SL task showed a larger triplet onset effect.
More recently, our group obtained an online measure of SL using ERPs during aSL and vSL tasks in adult musicians and non-musicians (Mandikal Vasuki et al., 2017). Musicians, when compared to non-musicians, obtained higher scores on the behavioural aSL measure and showed early N1 and N400 triplet onset effects in the online measure. In contrast, no group differences were seen in the behavioural or online measure of vSL. The findings support the hypothesis that musicians may be faster than non-musicians at extraction of statistical regularities in the auditory domain.
In the present study, our approach was to examine an array of auditory and cognitive behaviours (including SL) in children who had been enrolled in private music lessons for at least 1.5 years prior to participation. A second group of children, matched in chronological age, socio-economic background, and parents’ education level, served as a comparison group. Importantly, we investigated individual differences among children to assess whether higher SL is associated with higher music skills. In order to understand brain dynamics as the process of learning unfolds, we obtained an online measure of segmentation during the aSL and vSL tasks using ERPs. In the present study, we used similar paradigms and the same principles as Abla et al. (2008), to record and analyse the ERP data. In the current study, as in Mandikal Vasuki et al. (2017), we used tones and picture stimuli to evaluate aSL and vSL respectively.
The aim of the present study was to compare musically trained and untrained children on measures of auditory processing, cognitive processing, as well as behavioural and online SL. Based on the aforementioned literature, we hypothesized that children with music training would outperform their peers on the auditory and cognitive processing tasks used in the current study. With regard to the SL tasks, we hypothesized that children with music training would have higher scores on the behavioural SL measures, and larger triplet onset effects on the online ERP measures of learning during the SL tasks.
Section snippets
Participants
Fifty children (N = 50, 23 girls) were recruited through advertisements on websites and group emails. Among them, 25 were musically trained (mean age = 10.5 years old, SD = 0.9 years; 14 girls) and 25 were not musically trained (mean age = 10.1 years old, SD = 0.9 years; 9 girls). There was no significant difference in age between the two groups [t (48) = 1.4, p = 0.2, d = 0.44]. The musically trained and untrained children are henceforth referred as musicians and non-musicians respectively. Musicians had an
Auditory processing
Comparing the performance of musicians and non-musicians on measures of auditory processing using Pillai’s trace, a significant effect of group was observed [V = 0.5, F (6, 43) = 8.5, p < 0.001]. After correcting for multiple comparisons, separate univariate analysis of variance (ANOVAs) on the dependent variables revealed a significant effect of group on melody discrimination [F (1, 48) = 10.1, p = 0.003, ηp2 = 0.17], rhythm discrimination [F (1, 48) = 7.8, p = 0.007, ηp2 = 0.14], overall music skill score [F
Discussion
In the present study, we compared child musicians and non-musicians on an array of auditory processing, cognitive processing, and SL measures. We also obtained behavioural and online measures of aSL and vSL in these samples of participants. We hypothesized that musicians would outperform non-musicians on auditory and cognitive processing measures. We also hypothesized that musicians would score higher on the behavioural SL tasks and show larger triplet onset effects in the online ERP tasks.
Our
Acknowledgements
This work was completed as part of doctoral dissertation of the first author at Macquarie University. The work was funded by the Australia Awards Scholarships and the HEARing CRC, established and supported under the Cooperative Research Centres Program – Business Australia. This work was also supported by the Australian Research Council Centre of Excellence for Cognition and its Disorders (CE110001021). Arciuli was supported by a mid-career Future Fellowship granted by the Australian Research
References (78)
- et al.
Visual statistical learning of shape sequences: an ERP study
Neurosci Res
(2009) - et al.
Relations among musical skills, phonological processing, and early reading ability in preschool children
J Exp Child Psychol
(2002) - et al.
Statistical learning is lasting and consistent over time
Neurosci Lett
(2012) - et al.
Effect of musical expertise on visuospatial abilities: evidence from reaction times and mental imagery
Brain Cogn
(2004) - et al.
Implicit statistical learning in language processing: word predictability is the key
Cognition
(2010) - et al.
Rhythmic cues to speech segmentation: evidence from juncture misperception
J Mem Lang
(1992) - et al.
The perception of rhythmic units in speech by infants and adults
J Mem Lang
(1997) - et al.
Is statistical learning constrained by lower level perceptual organization?
Cognition
(2013) - et al.
Neural sensitivity to statistical regularities as a fundamental biological process that underlies auditory learning: the role of musical practice
Hear Res
(2014) - et al.
A comparative study of different artefact removal algorithms for EEG signals acquired during functional MRI
Neuroimage
(2007)
Endogeneous brain potentials associated with selective auditory attention
Electroencephalogr Clin Neurophysiol
Procedural learning is impaired in dyslexia: evidence from a meta-analysis of serial reaction time studies
Res Dev Disabil
Does combing the scalp reduce scalp electrode impedances?
J Neurosci Methods
Musicians' edge: a comparison of auditory processing, cognitive abilities and statistical learning
Hear Res
Nonparametric statistical testing of EEG-and MEG-data
J Neurosci Methods
The assessment and analysis of handedness: the Edinburgh inventory
Neuropsychologia
Statistical learning effects in musicians and non-musicians: an MEG study
Neuropsychologia
Unsupervised statistical learning underpins computational, behavioural, and neural manifestations of musical expectation
Neuroimage
Attention, probability, and task demands as determinants of P300 latency from auditory stimuli
Electroencephalogr Clin Neurophysiol
Statistical learning of tone sequences by human infants and adults
Cognition
Word segmentation: the role of distributional cues
J Mem Lang
Event-related potentials index segmentation of nonsense sounds
Neuropsychologia
An ERP study of continuous speech processing: I. Segmentation, semantics, and syntax in native speakers
Brain Res Cogn Brain Res
Musical experience shapes top-down auditory mechanisms: evidence from masking and auditory attention performance
Hear Res
The musical ear test, a new reliable test for measuring musical competence
Learn Indiv Differ
On-line assessment of statistical learning by event-related potentials
J Cogn Neurosci
Visual-spatial sequence learning and memory in trained musicians
Psychol Music
Statistical learning in typically developing children: the role of age and speed of stimulus presentation
Dev Sci
Statistical learning is related to reading ability in children and adults
Cogn Sci
Musical experience, auditory perception and reading-related skills in children
PLoS ONE
Controlling the false discovery rate: a practical and powerful approach to multiple testing
J R Stat Soc Ser B Stat Methodol
TONI-4, test of nonverbal intelligence
Executive function, visual attention and the cocktail party problem in musicians and non-musicians
PLoS ONE
Implicit memory in music and language
Front Psychol
Statistical learning in children with specific language impairment
J Speech Lang Hear Res
Short-term music instruction and young children's developmental music aptitude
J Res Mus Edu
Practicing a musical instrument in childhood is associated with enhanced verbal ability and nonverbal reasoning
PLoS ONE
Music training for the development of speech segmentation
Cereb Cortex
Musical expertise boosts implicit learning of both musical and linguistic structures
Cereb Cortex
Cited by (22)
Is there such a thing as a ‘good statistical learner’?
2022, Trends in Cognitive SciencesCitation Excerpt :Several studies (listed in Table 1) have found significant positive correlations between performance on an SL task and a range of cognitive skills. Importantly, some of these correlations were observed when the same task predicted different functions in different modalities (e.g., a similar visual embedded pattern learning task with alien-like figures correlates with both reading abilities [28] and musical skills [29]). This suggests that a given SL task reflects a general ability for learning regularities, so that it can simultaneously predict performance across different cognitive domains.
Maternal stress predicts neural responses during auditory statistical learning in 26-month-old children: An event-related potential study
2021, CognitionCitation Excerpt :However growing evidence suggests that individual differences in experience and development are associated with modulations of neural processes underlying statistical learning (in both the auditory and visual domains). For example, both children and adults with, as opposed to without, “enriched” auditory experience (e.g., musical training), have been found to show a larger word onset effect (Mandikal Vasuki et al., 2017a, 2017b). In contrast, children with developmental language disorders (SLI and developmental dyslexia) have been found to perform less proficiently on statistical learning tasks (e.g., showing slower reaction times, or poorer recognition memory for familiarized items) when compared to children with typical language development, and it has been proposed that these differences might account for the language difficulties observed in these children throughout development (Gabay et al., 2015; Obeid et al., 2016; Saffran, 2018).
Measuring children's auditory statistical learning via serial recall
2020, Journal of Experimental Child PsychologyCitation Excerpt :Some research has shown that processing- and reflection-based measurements are not always associated, most prominently for linguistic ASL (Batterink et al., 2015; Isbilen et al., 2017, 2020). By contrast, investigations of nonlinguistic ASL in adults (Mandikal Vasuki et al., 2017b) and in children (Mandikal Vasuki et al., 2017a) show a closer alignment of results from processing- and reflection-based measures. In the case of language-like stimuli, the results suggest that ASL leads to knowledge that can be explicitly reflected on but that its relationship to accrued implicit knowledge may be weak.
Musical expertise facilitates statistical learning of rhythm and the perceptive uncertainty: A cross-cultural study
2020, NeuropsychologiaCitation Excerpt :To the best of our knowledge, however, few studies addressed how SL of “temporal” patterns such as rhythm and meter is reflected in neurophysiological responses. Furthermore, although it has been revealed that “spectral” SL is facilitated by musical training (Paraskevopoulos et al., 2012; Rao et al., 2017), there are few neurophysiological studies that have investigated whether musical training modulates temporal SL capacities and the uncertainty encoding of a rhythm sequence, and how musical culture tune them. The temporal as well as spectral information plays an important role in comprehension of auditory structure such as language and music.