The descent of words

The word for “sky” in the indigenous Saaroa language of Taiwan is laŋica. Across the South China Sea in the Philippines, the speakers of Ilonggo use laŋit, whereas, on the far-flung islands of the Pacific, Hawaiians say lani and Rarotongans and New Zealand Maori raŋi (1). Systematic sound correspondences between many such words tell us that these languages have evolved from a common ancestor to form part of the Austronesian language family. By meticulously comparing the sounds of words across many languages, linguists can learn about the genealogical relationships between languages and the people who speak them, how sounds change through time and even how long-extinct ancestral languages would have sounded. In PNAS, Bouchard-Côté et al. (2) automate this process by using probabilistic models of sound change to trace the evolution of thousands of words across more than 600 Austronesian languages. The conventional technique for studying language change on the basis of contemporary variation is known as the comparative method (3). This approach identifies shared “cognates” between putatively related languages. Cognates are homologous words of similar meaning that show systematic sound correspondences indicating common ancestry (Fig. 1). Since the 19th century, historical linguists have understood that sound changes occur in a regular but context-sensitive way across the vocabulary of a language. Hence, where Hawaiian has lani and lima for “sky” and “five,” Rarotongan and Maori have raŋi and rima, reflecting a shift in their ancestral lineage from this l sound to r. In deciding whether two words are genuinely cognate, linguists can therefore look beyond superficial similarities by attempting to reconstruct a protolanguage (the common ancestor of the languages in question) and identify regular sound changes acting across the sound systems of its descendants. Reconstructing the descent of words on a language tree. This example shows …