Proceedings of the Annual Meeting of the Cognitive Science Society

It has been argued that semantic categories across languagesreflect pressure for efficient communication. Recently, thisidea has been cast in terms of a general information-theoreticprinciple of efficiency, the Information Bottleneck (IB) prin-ciple, and it has been shown that this principle accounts forthe emergence and evolution of named color categories acrosslanguages, including soft structure and patterns of inconsistentnaming. However, it is not yet clear to what extent this ac-count generalizes to semantic domains other than color. Herewe show that it generalizes to two qualitatively different se-mantic domains: names for containers, and for animals. First,we show that container naming in Dutch and French is near-optimal in the IB sense, and that IB broadly accounts for softcategories and inconsistent naming patterns in both languages.Second, we show that a hierarchy of animal categories derivedfrom IB captures cross-linguistic tendencies in the growth ofanimal taxonomies. Taken together, these findings suggest thatfundamental information-theoretic principles of efficient cod-ing may shape semantic categories across languages and acrossdomains.