Abstract
Male leopard seals make stylised sequences of vocalization units for the purposes of mate attraction and territorial signaling during their breeding season. For our study population of 26 leopard seals in Eastern Antarctica, these underwater displays consist of a common ‘alphabet’ of five discrete sounds. While the patterning of these sequences varies, the frequency and temporal characteristics of these sounds remain consistent across the population, suggesting that the combination and ordering of sounds is crucial for encoding information. In this study we used 1st Order Markov models of individual acoustic sequences based on empirically observed probability distributions and a maximum likelihood estimator to classify sets of ‘unknown’ calling bouts to their producer. We found that this classifier was able to attribute the unknown bouts to the correct seals more often than expected by chance. This result implies that the acoustic sequences could be used by the male seals to broadcast their individual identity to potential mates and/or rival males.