UC San Diego

Cross-linguistically, there are a variety of attested non-phonological conditions on

phonotactics and alternations. The extreme end of this are exceptions, whose idiosyncratic

behaviors are unattributable to any morphological, morphosyntactic, or semantic

class. Studying exceptions is challenging because they require a substantive language

description to even be identified, and because their identification can appear to subvert

descriptive and analytical generalizations about the grammar. As a result, exceptionality

research often focuses on major world languages, and there is considerable contention surrounding

whether exceptions are wholly phonological, extragrammatical, or something

in between.

This dissertation addresses these gaps via an in-depth case study of segmental phonology

in Mushunguli (Somali Chizigula, Kizigua; Narrow Bantu, G.31), an endangered,

under-described language spoken by members of the Somali Bantu diaspora. This case

study, drawn from original fieldwork conducted in 2011-2012, includes a description and

analysis of the hiatus resolution and onset structure conspiracies of Mushunguli, the former

of which exhibits what appear to be four operations (asymmetric coalescence, glide

formation, secondary articulation, and elision) in identical morphosyntactic contexts.

Situated within this discussion are three exceptional patterns, each representing a

separate typological instantiation of exceptional blocking: a set of high vowel-initial stems

that exclusively block coalescence, but not other applicable repairs (simple blocking); a prefix

and a verb root which unexpectedly undergo otherwise unattested palatalization in

lieu of elision (walljumping); and a set of roots that exceptionally block all forms of hiatus

resolution (total non-participation). Adopting lexically-indexed constraints in Stratal Optimality

Theory as a means of capturing these patterns reveals complex interdependencies

between exceptions and regular forms in Mushunguli, with the form and behavior of one

exception crucially determining the forms and behaviors of other exceptional and regular

patterns. This suggests that exceptions are lexical but not extragrammatical, instead

playing an important role in the grammar as reifying and reinforcing agents.

The study concludes by examining alternative representational analyses of the

Mushunguli exceptions, including whole-segment absolute neutralization and underspecification.

While these approaches are sometimes capable of capturing the exceptional

patterns, they ultimately struggle to unify or situate them with respect to the grammar as

a whole.