ABSTRACT
Community detection procedures find groups of speakers with similar network positions. Applying a community detection routine for bipartite networks (QuanBiMo) to the Raleigh network data allows linguistic comparisons across groups of speakers, in contrast with the relational analyses in Chapters 3 and 4. The modules (clusters) yielded by the QuanBiMo routine correspond to speakers who attended the same or similar sets of schools. Two hypotheses concerning the linguistic effects of module (network cluster), age, sex, and occupation are tested.
In Generation 1, the SVS had little or no association with socioeconomic status, as linguistic change mostly had not yet begun, and there are few significant differences among modules. In Generation 2, effects of occupation and module begin to emerge. In Central and East Raleigh modules, women are more Southern than men, whereas in North Raleigh, where exposure to non-SVS forms and their positive symbolic value is greater, women are less Southern than men. In Generation 3, white collar females surpass males in the retreat from the SVS, while blue collar females retain the SVS while showing change across apparent time.
