Among the original contributions of Martijn Wieling's thesis there is the application and refinement of a graph-theoretic technique, hierarchical bipartite spectral graph partitioning, to phonetic data in order to address an open issue of dialectometry, namely how to cluster dialectal varieties while simultaneously characterizing the underlying linguistic basis.
In principle, the same technique could also be used to reconstruct diachronic change: i.e. a careful investigation of synchronic patterns of linguistic variation with underlying linguistic features can lead to important insights into the comprehension of diachronic phonetic processes. In this talk, starting from the analysis of synchronic patterns of phonetic variation in Tuscany, I will tackle one of the main and most debated features of Tuscan dialects, the phenomenon of spirantization with a specific view to the so-called Tuscan "gorgia" (i.e. voiceless spirantization). In particular, I will show that the newly proposed method of spectral partitioning of bipartite graphs applied to synchronic dialectal data can effectively be used to investigate diachronic phonetic processes.
Starting from a careful analysis of the sound correspondences involving voiceless and voiced stops, the evolution of the spirantization phenomenon is tracked down from different perspectives. First, spirantization is tracked down geographically, across Tuscany from the influential center of Florence to the peripheral areas. Second, it is tracked down phonologically, from voiceless to voiced stops, and within each voicing class from velars to dentals and then to bilabials. Finally, it is tracked down demographically, with young speakers using the most innovative sound correspondences more than old speakers.
The fact that these results are in line with the literature on the topic of Tuscan "gorgia" demonstrates the potential of the method of spectral partitioning of bipartite graphs with respect to the reconstruction of diachronic processes starting from diatopically distributed synchronic dialectal data.