Abstract
Storytelling is intrinsically human, found in every known human population, and exclusive to the human species. From folktales to movies to the grand narratives of human history, stories affect lives and shape communities, playing a critical role in social learning, cultural inheritance, community building, and identity formation. Passed down from generation to generation and shared across geographical, linguistic, and cultural boundaries, stories evolve from one telling to the next, creating lineages shaped by both preservation and innovation. During the settlement of Polynesia stories migrated with the people, voyaging across the Pacific, linking one island to another in a genealogical network of narrative.
This thesis explores vertical and horizontal transmission of the Polynesian story of Māui stealing fire using computational phylogenetic methods adapted from evolutionary biology. Sixty variants of the story were sourced from texts collected at the beginning of the colonial era and coded for the presence or absence of 618 narrative motifs to quantify the narrative distance between the different versions. This information was then used to create a phylogenetic network to visualise the relationships between the variants based on similarity and variation.
The results show that post-settlement interaction networks in Polynesia operated like single populations, sometimes exhibiting as much diversity within islands as between connected islands. In this context, horizontal transmission only really occurred between non-contiguous areas and was difficult to distinguish from vertical transmission.
This research was also a test of the application of NeighborNet phylogenetic networks to visualise vertical and horizontal transmission in narrative data. The results were mixed. The network showed regional patterning of diversity consistent with current models of Polynesian settlement and interaction, however the use of distance data rather than trait data to build the graph reduces the explanatory power of the result. Specifically, the relationships between taxa are based on resemblance, not strict homology, so the graph cannot be used to infer the chronology, direction, or sequence of evolution. While NeighborNet graphs are a useful tool for exploring reticulation and conflict in cultural data as precursor or follow-up to cladistic analysis, they are somewhat limited as an investigative tool on their own.