Sperm whale social groups can be assigned to vocal clans based on their production of codas, short stereotyped patterns of clicks. It is currently unclear whether genetic variation could account for these behavioural differences. We studied mitochondrial DNA (mtDNA) variation among sympatric vocal clans in the Pacific Ocean, using sequences extracted from sloughed skin samples. We sampled 194 individuals from 30 social groups belonging to one of three vocal clans. As in previous studies of sperm whales, mtDNA control region diversity was low (π = 0.003), with just 14 haplotypes present in our sample. Both hierarchical AMOVAs and partial Mantel tests showed that vocal clan was a more important factor in matrilineal population genetic structure than geography, even though our sampling spanned thousands of kilometres. The variance component attributed to vocal dialects (7.7%) was an order of magnitude higher than those previously reported in birds, while the variance component attributed to geographic area was negligible. Despite this, the two most common haplotypes were present in significant quantities in each clan, meaning that variation in the control region cannot account for behavioural variation between clans, and instead parallels the situation in humans where parent-offspring transmission of language variation has resulted in correlations with neutral genes. Our results also raise questions for the management of sperm whale populations, which has traditionally been based on dividing populations into geographic 'stocks', suggesting that culturally-defined vocal clans may be more appropriate management units.

A method for the automatic classification of free-ranging delphinid vocalizations is presented. The vocalizations of short-beaked and long-beaked common ͑Delphinus delphis and Delphinus capensis͒, Pacific white-sided ͑Lagenorhynchus obliquidens͒, and bottlenose ͑Tursiops truncatus͒ dolphins were recorded in a pelagic environment of the Southern California Bight and the Gulf of California over a period of 4 years. Cepstral feature vectors are extracted from call data which contain simultaneous overlapping whistles, burst-pulses, and clicks from a single species. These features are grouped into multisecond segments. A portion of the data is used to train Gaussian mixture models of varying orders for each species. The remaining call data are used to test the performance of the models. Species are predicted based upon probabilistic measures of model similarity with test segment groups having durations between 1 and 25 s. For this data set, 256 mixture Gaussian mixture models and segments of at least 10 s of call data resulted in the best classification results. The classifier predicts the species of groups with 67%-75% accuracy depending upon the partitioning of the training and test data.