Social learning is key in the development of both human and non-human animal societies. Here, we provide quantitative evidence that supports the existence of social learning in sperm whales across socio-cultural barriers, based on acoustic data from locations in the Pacific and Atlantic Oceans. Sperm whale populations have traditionally been partitioned into clans based on their vocal repertoire (what they say) of rhythmically patterned clicks (codas), and in particular their identity codas, which serve as symbolic markers for each clan. However, identity codas account for between 35% and 60% of all codas vocalized depending on the different clans. We introduce a computational modeling approach that recovers clan structure and shows new evidence of social learning across clans from the internal temporal structure of non-identity codas, the remaining fraction of codas. The proposed method is based on vocal style, which encodes how sperm whales assemble individual clicks into codas. Specifically, we modeled clicking pattern data using generative models based on variable length Markov chains, producing what we term "subcoda trees". Based on our results, we propose here a new concept of vocal identity, which consists of both vocal repertoire and style. We show that (i) style-delimited clans are similar to repertoire-delimited clans, and that (ii) sympatry increases vocal style similarity between clans for non-identity codas, but has no significant effect on identity codas. This implies that different clans who geographically overlap have similar styles for most codas, which in turn implies social learning across cultural boundaries. More broadly, the proposed method provides a new framework for comparing communication systems of other animal species, with potential implications for our understanding of cultural transmission in animal societies.

翻译：社会学习是人类及非人类动物社会发展的关键。本文基于太平洋和大西洋多个地点的声学数据，提供了支持抹香鲸跨社会文化屏障进行社会学习的定量证据。抹香鲸种群传统上根据其发声库（即它们"说什么"）——由节奏性模式的咔嗒声（尾音编码）组成，尤其是身份尾音编码（作为每个族群的符号标记）——被划分为不同族群。然而，身份尾音编码在不同族群中占所有尾音编码发声量的35%至60%。我们引入一种计算建模方法，该方法可恢复族群结构，并基于非身份尾音编码（剩余部分尾音编码）的内部时间结构，提供了跨族群社会学习的新证据。所提方法基于发声风格，该风格编码了抹香鲸如何将单个咔嗒声组合成尾音编码。具体而言，我们采用基于可变长度马尔可夫链的生成模型对咔嗒声模式数据建模，生成了我们称之为"子尾音树"的结构。基于研究结果，我们在此提出发声身份的新概念，该概念包含发声库与发声风格两个层面。研究表明：（i）按风格划分的族群与按发声库划分的族群高度相似；（ii）同域分布会增强不同族群间非身份尾音编码的发声风格相似性，但对身份尾音编码无显著影响。这意味着地理重叠的不同族群在大部分尾音编码上具有相似风格，进而暗示了跨文化边界的社会学习。更广泛而言，所提方法为比较其他动物物种的通信系统提供了新框架，对理解动物社会的文化传播具有潜在启示意义。