Mind the Bar: Using EEG to Trace the Role of Musical Structure in Memory Development

How do musical structure and performance cues aid in memory formation (and associated neural signals) among Western classical music performers when learning music using a musical score? And does the time spent learning that music also influence these memory mechanisms (and associated neural signals)? Research on musical memory has primarily focused on auditory and motor processes, providing little clarification into how those would be embedded within a memory system capable of supporting later recall of specific pieces of music. Hence, aspects of musical semantics and related visual aspects of the Western classical music performer’s experience of learning a piece of music have been under-investigated. Music performers develop explicit knowledge of musical pieces through deliberate and thoughtful practice, with musical structure playing a crucial role in learning and performance (Chaffin et al., 2010; Ginsborg, 2022; Williamon & Valentine, 2002). Experienced music performers report segmenting music into meaningful and individual but linked sections, having the tendency to start and stop at structural points in the pieces that they are learning during rehearsals, and using more analytical approaches when music is longer and more complex (Aiello & Williamon, 2002; Chaffin et al., 2010; Chaffin & Imreh, 1997; Ginsborg, 2022; Hallam, 1997; Reid, 2002; Williamon & Valentine, 2002). This ability to divide music into meaningful chunks seems analogous to the ability of event segmentation, which refers to the cognitive ability to divide our continuous experiences into meaningful ‘chunks’ or events (Zacks, 2020). Changes in location, time or context create event boundaries that help organise information (Clewett et al., 2019; Zacks, 2020). Event segmentation has been investigated using a variety of methods, including image sequences (Griffiths & Fuentemilla, 2020; Güler et al., 2024; Van De Ven et al., 2023), videos (Jafarpour et al., 2022; Sargent et al., 2013; Silva et al., 2019), virtual reality (Horner et al., 2016; Logie & Donaldson, 2021; Radvansky et al., 2010; Van Helvoort et al., 2020), music listening (Büdenbender & Kreutz, 2017; Faraco, 2024; Knösche et al., 2005; Nan et al., 2006; Sridharan et al., 2007), and using sentence and discourse procedures (Cohn et al., 2012; Cunillera et al., 2009; Delogu et al., 2018; Speer et al., 2007; Van Petten & Kutas, 1991; Wang et al., 2008; Weber-Fox & Neville, 2001). The latter research has provided evidence for the role of N400 in event segmentation (Delogu et al., 2018; Kuperberg, 2016; Lau et al., 2008; Zacks et al., 2007). Despite overlapping insights between music psychology and event segmentation theories to date, this connection between theories has been underexplored. More crucially, there is minimal evidence for the neural mechanisms underlying musical segmentation during active learning (as opposed to passive listening). One of the few studies that has investigated musical structure and memory formation in the context of music learning while using neuroimaging techniques has been Williamon & Egner (2004). These authors aimed to provide evidence of the neural underpinnings of musical memory, hypothesising that if musical bars that are structurally important to a piece of music are crucial to aid the encoding and retrieval of a piece of music, then these bars would be processed differently than other bars within the same piece. They asked six classical pianists to learn Bach’s Prelude from Prelude and Fugue in A minor (BWV 889) for two weeks. Following this learning period, they performed a visual recognition task while EEG data was collected from them. The task consisted of observing a musical bar for one second and identifying if the bar had been learned or not. Crucially, the participants did not know that the musical bars being presented differed in their structural importance: 8 bars were structural (e.g., beginning of a phrase) and 24 were non-structural. The authors identified a unique event-related potential (ERP) component (right centro-parietal negativity between 300 ms and 400 ms post-stimuli) for structural bars compared to non-structural bars and not-learned bars. These results seem to suggest an effect of structural importance on musical memory. However, this study has several limitations: the unbalanced stimuli (8 structural bars vs 24 non-structural bars) might have driven the observed differences rather than structural importance per se, and the small sample size (N = 6) limits statistical power and generalizability. Additionally, the lack of control group or condition, as well as the lack of inclusion of structural and non-structural bar distinctions on the not-learned music, make it difficult to draw clear conclusions. In the proposed study, we aim to extend Williamon and Egner's (2004) study with improvements in experimental design. The main aspects of improvements are the following. The stimuli are balanced, with the same number of structural and non-structural bars being presented to participants. The stimuli are also counterbalanced, with three Bach pieces being used as learned, sight-reading or not-learned piece in different combinations per participant. A control condition has also been added, where material is learned but for a much-reduced period of time (sight-reading). With the addition of this condition, this allows for the inclusion of both recognition and source memory scores. Finally, we aim to collect data from a much larger sample of pianists. This research aims to confirm whether there is an observable ERP component related to structural bars, potentially providing evidence for a neuronal mechanism akin to event segmentation. Additionally, it will give initial evidence of the impact of learning (Learned, Sight-read and Not Learned) on these musical structural neuronal mechanisms.