Thesis Open Access
The patterning of /v/ in Russian and other languages has long posed a problem for phonological theory because of its ambiguous classification with respect to the feature [sonorant]: like obstruents, /v/ participates as a target of final devoicing and regressive voicing assimilation, but like sonorants, fails to trigger regressive voicing assimilation.
In this dissertation I tackle the problem of Russian /v/ by situating it in a broader cross-linguistic landscape, in terms of its acoustic properties, its relationship to other non-sibilant voiced fricatives, or voiced spirants /B, D, G/, and its phonological typology. The empirical results of the dissertation challenge the binary division between obstruents and sonorants, and suggest that a finer distinction is required.
Chapter 2 and Chapter 3 focus on the phonetic characteristics of /v/. Chapter 2 investigates the relationship between voicing and frication type in four languages—English, Greek, Serbian, and Russian—using measures that distinguish between voicing and frication type: duration, harmonicity, and spectral centroid. Two environments, word-initial stressed and word-medial unstressed, are used in order to assess the effect of word-internal prosodic factors. I argue based on these results that although /v/ has the highest harmonicity in each language, the relationship between voicing and frication type is revealed by spectral centroid, and that the relationship is independent in English and Greek, but not in Serbian; in Russian the results depend on environment.
Chapter 3 tests whether there exists a correlation between phonological and phonetic identity, using Greek, in which /v/ patterns as an obstruent, Serbian, in which /v/ patterns as a sonorant, and Russian, in which /v/ is ambiguous. A partial correlation between phonetic and phonological identity was found, with the relativized centroid values for Greek and Serbian /v/ being distinguished in both environments. Russian and Greek do not differ significantly in the WIS environment, while Russian and Serbian do not differ significantly in the WMU environment.
Chapter 4 turns to investigating the distribution in consonant inventories and implicational relations of voicing in spirants and sibilants. The results of this investigation show that voiced spirants and voiced sibilants do not share the same cross-linguistic distribution: voiced spirants frequently surface in small inventories, but only if unpaired, while the correlation between the presence of voiced sibilants (which only occur with a voiceless counterpart) and inventory size is linear. Voiced spirants also tend to appear in inventories without a voicing contrast in the stops, unlike voiced sibilants.
Chapter 5 argues that Russian /v/ is featurally specified as [+sonorant, +obstruent], and that regressive voicing assimilation is inherently asymmetric: all segments that are [+obstruent] are targets of devoicing under regressive voicing assimilation and final devoicing, but only segments that are exclusively obstruents trigger RVA. A typology of phonological patterning of /v/ under [obstruent] and [sonorant] is proposed.
Chapter 6 considers the quantal nature of [obstruent] and [sonorant], and sketches a model that combines Quantal Theory (Stevens, 1989) and Emergent Feature Theory (Mielke, 2008). It is argued that by expanding the set of quantal domains, both clear cases of feature assignment and ambiguous cases can be understood within the same framework.