Abstract
The expansion and diversification of individual speaker groups throughout history have led to the fact that today we speak different languages in different places. These processes depend, among other things, on landscape features such as natural barriers.
This paper examines the influence of rivers and mountains as natural barriers on the spatial location of language areas. South America serves as the study area. For this purpose, the language areas of 295 languages were interpolated with Voronoi tessellation (VT) and the influence of landscape variables on the weighting of VT was investigated. Three models were tested an unweighted VT, a VT considering rivers as weights and a VT considering mountains as weights. Known point geometries of languages from two different sources were used as seed points. These are, on the one hand Glottolog, which is freely available and on the other hand Ethnologue, which requires a license for access.
The generated surfaces were then compared for their spatial similarity to corresponding polygons of a reference dataset using the Jaccard Index(JI). Subsequently, the JI of the individual language areas were compared with each other to make statements about the influence of the corresponding landscape variable in the weighting of the VT.
The VT weighted with mountains showed only minimal improvements compared to the unweighted VT. The small differences suggest that mountains can only be considered as a driving factor for the separation of two language areas to a limited extent. When weighted with rivers, the results were significantly worse than for unweighted VT, which contradicts the assumption that rivers are a natural barrier with respect to the dispersion of speaker groups. Due to the limitation in the methodology, the choice of seed points did not play a decisive role.