ИСТИНА

Killer whale groups have unique repertoires of stereotyped calls – vocal dialects, which are transmitted by vocal learning rather than genetically. It is generally accepted that the dialects pass mostly vertically from mother to offspring and change with time due to accumulation of random errors. To test this assumption, we compared the similarity of 10 syllables in 3 call types (K1, K5 and K7) across 14 families of resident killer whales from Avacha Gulf, Kamchatka, Russian Far East. The similarity was calculated using dynamic time warping algorithm in MATLAB. We found that similarity patterns varied for different syllables even within the same call type, and correlations of distance matrices in most cases were non-significant. We created the random evolution model in MATLAB to test if these patterns corresponded with this model. The model simulated the evolution of dialects of three groups. On the first stage, one group evolved independently and two evolved together. On the second stage, all three groups evolved independently. If the duration ratio of both stages was 1:1, the two related groups were more similar to each other than to the third group in 46% of cases (random 33,3%). If the duration of the second stage decreased related to the first stage, this percent increased, e.g. it was 54% when the ratio was 2:1 and 80% when the ratio was 20:1. However, the percent decreased if the variability of the simulated call traits was limited. Our results suggest that the observed correlation between similarity patterns of syllables is lower than predicted by the random evolution model, but this can be caused by different reasons, including the structural limitation of call variability.