Diren Darbaz
Pronouns: She/Her
About
My research project
Musicalising Birdsong: Building a Zygonic-Inspired, Expectation-Based Grammar for the Analysis, Transcription, and Composition of Bengalese Finch Vocal SequencesThis project investigates the syntax of Bengalese finch song as a way of informing musical practice. It focuses on how syllable and rest categories are organised in sequence: what follows what, which sequences remain stable, and where variation occurs. These relationships reveal recurring patterns in the birds’ vocal behaviour and provide the basis for birdsong-derived templates for analysis and composition.
The research draws on an existing annotated long-form Bengalese finch song corpus, including syllable labels, onset-offset data and source recordings. It examines how syllables, rests and bouts (continuous stretches of singing separated by silences of 300 ms or more) are organised across the dataset, with particular attention to repeated patterns, branching points, transition probabilities and the distribution of vocal units across individual birds and the wider corpus. A central part of the method is the classification of within-bout gaps into rest categories. While prior birdsong research has examined silences between vocal events in terms of their duration and their relationship to vocal sequencing, this project classifies these gaps as rests and treats them as sequential units in their own right, so that they can be analysed alongside vocal events.
These syllable and rest features are used to build probabilistic summaries of the song’s sequential organisation, including Markov-style transition profiles and rest-class models that show how timing contributes to sequence structure. Music-theoretical analysis is then used as a way of interpreting these patterns without imposing human musical intention onto the birds. Ockelford’s Zygonic Theory and related expectation-based approaches provide a framework for describing repetition, continuation, return and variation in the data. The aim is not to prove that birdsong is music, but to show how certain kinds of avian vocal structure can be described in terms that are meaningful for musical analysis while remaining anchored in biological and behavioural evidence.
The central contribution lies in translating these analyses into birdsong-derived pattern templates that can inform compositional practice. These templates are designed to show observed syllable and rest sequences, their probabilities, and their different levels of stability or flexibility across the corpus. They are intended to support the creation of new musical works based on non-human vocal logic, which will be demonstrated through compositional applications developed as part of the project. These applications will operate at different interpretative levels, ranging from relatively close realisations of the observed data to freer creative transformations of its underlying principles. A generative stepper engine developed as part of the project will form one practical tool within this process, using observed syllable and rest-transition probabilities to generate bout-like sequences and test how the analytical patterns can be realised in sound. In this way, the project does not treat birdsong only as sonic inspiration, but as a structured source of musical organisation that can inform composition and potentially offer a comparative lens for thinking about existing human music.
Supervisors
This project investigates the syntax of Bengalese finch song as a way of informing musical practice. It focuses on how syllable and rest categories are organised in sequence: what follows what, which sequences remain stable, and where variation occurs. These relationships reveal recurring patterns in the birds’ vocal behaviour and provide the basis for birdsong-derived templates for analysis and composition.
The research draws on an existing annotated long-form Bengalese finch song corpus, including syllable labels, onset-offset data and source recordings. It examines how syllables, rests and bouts (continuous stretches of singing separated by silences of 300 ms or more) are organised across the dataset, with particular attention to repeated patterns, branching points, transition probabilities and the distribution of vocal units across individual birds and the wider corpus. A central part of the method is the classification of within-bout gaps into rest categories. While prior birdsong research has examined silences between vocal events in terms of their duration and their relationship to vocal sequencing, this project classifies these gaps as rests and treats them as sequential units in their own right, so that they can be analysed alongside vocal events.
These syllable and rest features are used to build probabilistic summaries of the song’s sequential organisation, including Markov-style transition profiles and rest-class models that show how timing contributes to sequence structure. Music-theoretical analysis is then used as a way of interpreting these patterns without imposing human musical intention onto the birds. Ockelford’s Zygonic Theory and related expectation-based approaches provide a framework for describing repetition, continuation, return and variation in the data. The aim is not to prove that birdsong is music, but to show how certain kinds of avian vocal structure can be described in terms that are meaningful for musical analysis while remaining anchored in biological and behavioural evidence.
The central contribution lies in translating these analyses into birdsong-derived pattern templates that can inform compositional practice. These templates are designed to show observed syllable and rest sequences, their probabilities, and their different levels of stability or flexibility across the corpus. They are intended to support the creation of new musical works based on non-human vocal logic, which will be demonstrated through compositional applications developed as part of the project. These applications will operate at different interpretative levels, ranging from relatively close realisations of the observed data to freer creative transformations of its underlying principles. A generative stepper engine developed as part of the project will form one practical tool within this process, using observed syllable and rest-transition probabilities to generate bout-like sequences and test how the analytical patterns can be realised in sound. In this way, the project does not treat birdsong only as sonic inspiration, but as a structured source of musical organisation that can inform composition and potentially offer a comparative lens for thinking about existing human music.