ИСТИНА

Each pod of killer whales shares a socially transmitted vocal dialect, which change with time through cultural evolution. We examine macro- and microevolutionary patterns by analyzing vocal repertoires on different levels – from ecotype to matriline – and comparing them with the outcomes of agent-based computer model of dialect evolution. Modeling showed that accumulation of copying errors cannot be solely responsible for the dialect evolution: learning from either mother alone or the entire matriline with calls changing by random errors produced a graded distribution of the call phenotype, without the discrete call types observed in nature. A pattern resembling the dialect diversity observed in the wild arose only when other learning rules – innovation, divergence from kin and error proportional to matriline variance – were applied in combinations. Analysis of the natural dialects confirmed this outcome and suggested additional mechanisms, such as horizontal transmission of call features or saturation of structural changes due to constraints in call change. Bayesian phylogenetic trees of repertoires were generally similar but not necessarily identical to the acoustic similarity trees, suggesting that acoustic similarity is not always explained by common ancestry. Cross-ecotype comparison showed that both low- and high-frequency components of North Pacific transient killer whale calls had significantly lower frequencies than those of the North Pacific resident and North Atlantic populations, suggesting that ecology can constrain call change. In general, we conclude that the cultural evolution of killer whale calls is not a random process driven by steady error accumulation alone: other mechanisms such as innovation and simultaneous divergence and convergence on different scales play an important role. The important practical consequence is that correlation between relatedness and acoustic similarity is not always linear or even proportional and may vary across call types and syllables.