Tid: Torsdag 3 maj kl. 15-17
Plats: C307, Södra huset, Frescati


In recent years, the familiar notion of the family tree in comparative linguistics has been challenged by methods of representing relations among languages coming from biology, including networks such as NeighborNet (se illustration below). Unlike traditional family trees, such networks can show conflicts in phylogenetic signals, represented by boxes embedded in the network. The talk will focus on such networks and on recent, numeral methods for measuring the amount of conflicting signals (reticulation) associated with a language or a phylogeny, i.e., the degree to which languages deviate from a tree-like evolutionary model. Given these methods and the dataset of the ASJP project, with word lists from well over one half of the world’s languages,1 we can now answer questions such as the following: are larger or older families more reticulate than smaller or younger ones? Are dialects or emerging languages more reticulate than languages at large? Are languages without close relatives within a family (such as Armenian within Indo-European) more or less reticulate than is normally the case? Using reticulation metrics it is also possible to explain some of the differences between language classifications based on traditional methods in historical linguistics and classifications based on a recent computational method that uses phonological distances among word pairs as its input.

Hjärtligt välkomna!
Ljuba Veselinova