Revisiting Southern Gallo-Romance from a complexity theory standpoint: Occitan

In this paper, the inner structure of the Occitan dialect network is revisited in the light of a range of cumulative (Ward’s method) vs. reductive (Complete linkage, Groupe Average, Weighted Average) hierarchical algorithms provided by Gabmap, an Online dialectometric application for calculating distance/similarities by edit distance (Levenshtein algorithm). Reticularity of the Occitan geolinguistic space is addressed through connectograms using Gephi, and Multidimensional Scaling is also used to some extent. After sketching the canonical classifications of the Occitan geolinguistic space (Bec, Ronjat), providing the “eponymous dialects”, we explore the deep patterns of this diasystem, bringing to light a hierarchy of systemic entities constituting an array of “invisible dialects”, corresponding to entities of various size and functions ( macrodialects, dialects, subdialects, varieties, hubs, small worlds, buffer zones, default dialects ). The approach is based on concrete linguistic data from the THESOC database (Université de Nice/CNRS), contrasting the major isoglosses (macrodialectal features) with the “intricate variables”, i.e., segmental nexi , extracting data relating to strategic points in the complex dialectal network from reductionist algorithms.