UC Berkeley

This study addresses one of the most fundamental questions in formal phonology, namely What are the units of phonological representation that grammars manipulate? Linguists have long assumed that segments and binary features are the basic atoms of phonological representations. The present study challenges this assumption by proposing that subsegmental units can be defined with respect to two distinct dimensions of representation: (i) the temporal dimension; and (ii) the spatial dimension, roughly equivalent to the physical magnitude of an articulatory gesture. I draw on two case studies providing instrumental data on two endangered Amazonian languages of Brazil, Panãra [kre] and Kawaiwete [kyz]. Using oral and nasal airflow data, I show that, on the one hand, Panãra exhibits a surface contrast between prenasalized oral stops [nt] and postoralized nasal stops [nt], which crucially differ in the extent of the duration of nasal airflow. On the other hand, Kawaiwete exhibits a distinction between fully oral, partially nasal, and fully nasal vowels, which crucially differ with respect to their degree of opening of the velo-pharyngeal port.

The proposed representational model integrates the basic architecture of Q Theory (Inkelas & Shih 2016; 2017; Shih & Inkelas 2014; 2019) and subfeatural representations (Lionnet 2017) into a single unified framework. On the basis of the Panãra data, I argue for Q theoretic subsegmental representations, which divide the segment into three quantized and linearly ordered subsegments, (q1 q2 q3). This architecture provides the level of granularity necessary to distinguish between post-oralized nasals and pre-nasalized obstruents, where the former is represented with two nasal subsegments followed by one oral subsegment, and the latter is represented with a single nasal subsegment followed by two oral subsegments. Building on these Q theoretic representations, I argue on the basis of Kawaiwete for a scalar decomposition of phonological features, where continuous values can be grouped into one of three possible subfeatural categories: [+F], [?F], and [–F]. In the case of the feature [nasal], I argue for three perceptibly distinct degrees of nasalization: fully nasal [+nasal], partially nasal [?nasal], and fully oral [–nasal]. I show that a model of phonological representations which makes use of both subsegments and subfeatures is not only able to account for the data from both Panãra and Kawaiwete, but can also be extended to account for the full typology of phonological processes involving local nasalization or oralization.