Fine-scale song exchange in two neighbouring populations of humpback whales

Abstract

The songs of humpback whales (Megaptera novaeagliae), a patterned vocal display usually associated with breeding, serve as an excellent model of large-scale cultural transmission in a nonhuman species. These songs are considered to be particularly complex due to their well-known hierarchical structure and constantly changing nature. Two types of changes occur in the songs of the whales in the South Pacific meta-population: small, progressive evolutionary changes and radical cultural revolutions in which the song’s pattern is entirely replaced during a single breeding season. While patterns of song change and the transmission of these changes among populations have been studied on a broad scale, fine-scale examination of the song’s fundamental components (i.e. unit sequences) is lacking. This thesis aims to investigate fine-scale song structure in order to better understand song learning mechanisms. Firstly, intra-population song structures were examined within the east Australian song over time. Secondly, inter-population learning was assessed during the song’s transmission to the neighbouring population that winters off New Caledonia.Several quantitative methods were developed or applied here to identify song features at the unit sequence level. An acoustic dictionary of the song repertoire was created using a selforganising map, which quantified similarity between units and provided a robust means of classifying units. Metrics for measuring song complexity were modified to address the unique hierarchical structure of humpback whale song and quantify complexity at multiple levels in the song. An analysis of network structure, previously only applied to human language and birdsong, was used to measure the connectivity of units within song arrangements. These methods were applied to the transcription of songs over 13 consecutive years in east Australia and two consecutive years with two song types in New Caledonia, allowing for fine-scale features to be described and quantified. To understand song learning within a single population, fine-scale structural features of the east Australian song were quantified over the same period. The complexity of both theme and unit sequences fluctuated over time, showing a clear relationship with both progressive, evolutionary changes and revolutionary events. Songs became increasingly more complex as the songs evolved through small additions of novel material, but were simplified when cultural revolutions replaced the song pattern. However, these fluctuations were not reflected in underlying network structures or second order entropy of each song. Every song displayed a degree of “small-world” network structure characterised by clusters of highly connected units. Additionally, they contained fine-scale structures such as deterministic transition patterns (i.e. doublet and triplet repetitions), redundant unit usage, and predictable unit arrangements. These features were always present despite marked differences in songs from year to year and are likely to facilitate song learning. How structural features changed during transmission between east Australia and New Caledonia provided insight into inter-population social learning. Each individual theme maintained or increased its complexity during transmission, but the song’s arrangements and the underlying small-world structure remained largely unchanged. Songs were therefore learned across populations with high fidelity and without needing to be simplified. Such precision probably requires a mechanism of song transmission with a high degree of acoustic contact. Recent studies have suggested that such acoustic contact may occur on the shared Antarctic feeding grounds or individual movement between populations.This thesis is the first study to quantify the fine-scale structural and syntactic features of humpback whale song, and does so over multiple years and across two adjacent populations. These results indicate that song learning is probably influenced by a combination of conformity to an underlying structural template and small individual additions of novel material that increase complexity in the song’s arrangement. Widespread rules of song structure should be explored in future studies by examining these fine-scale song features in other populations or ocean basins. Features identified here such as small-world networks and central ‘hub’ units also occur in other complex vocal learning displays, such as birdsong and human language. These commonalities suggest a convergence of certain features in vocal learning across multiple taxa. Given how integral social learning is within human evolution and culture, comparative information on social learning across non-human species is a key factor in further understanding our own cultural evolution.