Mapping out the Source domain

This paper takes a strictly empirical approach to the encoding of spatial notions in the four ancient Indo-European languages Ancient Greek, Latin, Gothic and Old Church Slavonic. By generating semantic maps on the basis of parallel corpus data, without any semantic pre-analysis, we use methods well tested in typology to study the basic divisions in the spatial domain in the four closely-related languages, and to determine the finer subdivisions within the Source domain. We find that the four languages are similar, but clearly independent of each other, each carving up the spatial domain in different ways. We also find substantial encoding overlaps between the Source and Location domains.