LAUSR

Acoustic signals provide a model system for investigating the evolutionary processes that shape phenotypic diversity. Here we aim to investigate patterns of variation in tarsier duet phrases recorded from two mainland sites and eight offshore islands in North Sulawesi, Indonesia. We extracted features from the spectrograms of tarsier duet phrases and used a two-pronged approach to investigate patterns of variation in our data set. Previous work on other tarsier groups indicates that differences in duet features are related to species differentiation and our data set contained recordings from at least four genetically distinct tarsier groups. First, we used unsupervised clustering to test for the tendency to cluster in male and female tarsier duet contributions. Male and female tarsier duet contributions exhibit substantial differences in structure, so we analyzed them separately. For both males and females, we expected to see four unique clusters that correspond to the four genetically distinct tarsier groups. Second, we used a multivariate, variance components model to estimate how variance in tarsier male or female features was partitioned across three levels: within-individual, between-individual, and between-site. We did not find evidence for four unique clusters, as our unsupervised clustering results indicated that both male and female duets formed only two discrete clusters. Our use of the multivariate, variance components modeling indicated that certain features of male and female duets from Sangihe and Siau islands showed further separation of when compared with the other tarsier groups. The results of our study indicate that tarsier duets recorded from three islands in Bunaken National Park-which were recently identified as a genetically-distinct group-were not different from the other mainland Sulawesi sites. Little is known about the processes that lead to divergence in acoustic signals or the rate of phenotypic change, and future work comparing acoustic and genetic distance across tarsier populations will be informative.