Expertise in processing musical notation. An investigation of chunking, working memory, and eye movements

Lörch, Lucas

PDF
Lörch Dissertation - Expertise in Processing Musical Notation.pdf - Veröffentlichte Version
Download (983kB)

URN:	urn:nbn:de:bsz:180-madoc-638969
Dokumenttyp:	Dissertation
Erscheinungsjahr:	2023
Ort der Veröffentlichung:	Mannheim
Hochschule:	Universität Mannheim
Gutachter:	Münzer, Stefan
Sprache der Veröffentlichung:	Englisch
Einrichtung:	Fakultät für Sozialwissenschaften > Bildungspsychologie (Münzer 2012-)
Fachgebiet:	150 Psychologie
Freie Schlagwörter (Englisch):	musical expertise , expert memory , working memory , sight-reading , eye movements , computational modeling
Abstract:	Expertise develops as a result of deliberate practice, that is, as a result of activities that are designed with the primary purpose of attaining and improving skills. As a by-product of deliberate practice, experts acquire superior memory for information from their domain. This superior memory is enabled by chunking and by the rapid access to long-term memory. Chunking denotes a process of searching for known sequences in encoded stimuli and, if found, recoding them as single units. The present work investigated expertise in the cognitive processing of musical notation. More specifically, the three studies of the present work respectively examined (1) chunking in short-term memory for musical note symbols, (2) working memory for musical note symbols in the context of the Time-Based Resource Sharing model, and (3) eye movements during sight-reading, i.e., during the instrumental performance of unfamiliar notated melodies. The first study was based on the finding that chunking in short-term memory for musical notation is supported by systematic compared to random tonal structures. To extend this finding, I analyzed which specific features of systematic tonal structures support chunking. To this end, I designed a serial recall task with single quarter note symbols as memoranda. The number of memoranda and the tonal structure of the note sequences were varied within-participants. Chunking-supportive tonal structures resided in a clearly recognizable tonal context, i.e., contained only notes from the scale defined by the first note of the sequence, and contained meaningful melodic cells with clear labels. In the first simple span experiment, these melodic cells were major triads, while in the second, they were based on authentic cadences. Note sequences that had a chunking-obstructive tonal structure did not reside in a clearly recognizable tonal context and did not contain melodic cells with clear labels. Participants’ musical expertise was indicated by the Gold-MSI questionnaire. I expected that participants on a higher expertise level would recall notes more accurately, that chunking-supportive sequences would be recalled more accurately, and that the advantage in chunking-supportive sequences would be larger for participants on a higher expertise level. Analyses provided evidence for both expected main effects. Probably due to a ceiling effect, the expected interaction was only found in the long sequences containing authentic cadences. I conclude that, given that a note sequence has a systematic tonal structure, the combination of a clear tonal context with meaningful melodic cells might support chunking. The second study investigated expert working memory in the context of the Time-Based Resource Sharing (TBRS) model. This theoretical model assumes that working memory involves a rapid switching between maintaining already encoded information and processing new stimuli. It was translated into a computational model that simulates recall in complex span tasks. In a recent version of this computational mode (TBRSC), a chunking mechanism was added. Although various theories, such as template theory or long-term working memory theory, conceptualize working memory as being inherently influenced by expertise, TBRS does not account for expertise differences. Thus, using a newly developed complex span task for musical notation, I investigated how expertise might be conceptualized in the context of the TBRS model. In the task, participants had to memorize single quarter notes for serial recall of pitch. In between the visual presentation of each note symbol, they had to perform a distractor task, i.e., they had to sight-read a short melody. To manipulate the potential for chunking, the tonal structure of the sequences of to-be-remembered notes was varied. The sequences contained melodic cells that were either more meaningful (major triads) or less meaningful (arbitrary trichords). The complex span task was completed by a music student and a hobby musician sub-sample. Using the Gold-MSI questionnaire, a higher-expertise and a lower-expertise group was created in both sub-samples. Recall was simulated using TBRSC. Estimates for certain parameters, namely the strength of encoding (R), the chunk search duration (cSD), the probability of chunk retrieval (PCR), and the time used for distractor processing (Ta), were compared between expertise groups. I expected to find evidence for experts’ more rapid LTM access (i.e., a larger R and a smaller cSD) and more reliable chunking (i.e., a larger PCR). This expectation was supported in the hobby musician sub- sample. In addition, it was found that lower-expertise hobby musicians might have spent less time on the processing of distractors. They might have compensated inefficient memory processes by increasing refresh times during the distractor task. Parameter differences in the music student sub- sample were marginal, which probably was due to a ceiling effect. I conclude that expert working memory in the TBRS model might be conceptualized by rapid LTM access which increases refresh times, and by reliable chunking which increases the efficiency of refreshing. In the domain of musical sight-reading, there is some initial evidence that sight-reading accuracy, i.e., how accurately notes are performed on an instrument, might be associated with eye movements. This was found for measures such as pupil size, the size of the eye-hand span, or overall gaze duration. However, a systematic investigation of the association between eye movements and performance accuracy during sight-reading is still missing. I developed a software tool, the MidiAnalyzer, to assess the accuracy of experimental MIDI (Musical Instrument Digital Interface) data. Using eye movement and MIDI data from the sight-reading task that was embedded in the complex span task, I investigated how the number and duration of fixations during sight-reading was associated with the accuracy of pitch and note onset. In the statistical analyses, I controlled for the effect of certain covariates, namely musical expertise, practice, and features of notes. Musical expertise was indicated by the Gold-MSI questionnaire. Practice was represented by the number of trials participants had already completed. Rhythmical features of notes were accounted for by the type of note pair (eighth-eighth, eighth-quarter, quarter-quarter, quarter-eighth) that participants were currently reading. Results showed that the number of fixations was negatively associated with the accuracy of note onset in both the reading of the whole melody and the reading of note pairs. I conclude that reading with fewer fixations might require less cognitive resources as eye movement planning and information integration is less demanding. The saved cognitive resources might be used to increase the accuracy of the instrumental performance. In summary, the present work provided evidence for experts’ superior processing of musical notation in short-term memory, working memory, and sight-reading. It highlighted the central role of chunking in expert memory, and showed that chunking is a process with both universal and domain-specific aspects. In addition, the present work demonstrated that unraveling the role of eye movements in accurate sight-reading is substantial for the comprehensive understanding of this skill.