Epenthetic Dorsals in Khalkha

Many of the modern Mongolic languages display alternations between /g/ and zero. These cases are among a very small number of putative dorsal stop epenthesis patterns known in the phonological literature, and epenthesis of these consonants is considered in some theories to be highly marked or even impossible (see e.g. Lombardi 2002, de Lacy 2006, Uffmann 2014). In recent years, multiple accounts of these /g/-zero alternations (GZA) have been proposed, some of them attributing the observed patterns to allomorphy (de Lacy & Kingston 2013) or a phonological process of ‘splitting’ (Staroverov 2014) rather than true epenthesis. In the present work, we describe GZA from both diachronic and synchronic perspectives with the aim of documenting its historical development and providing an insightful analysis of the process as it occurs synchronically in Khalkha Mongolian. We argue that GZA arose diachronically from reanalysis of the lenition and eventual loss of intervocalic *g, which took place prior to the breakup of Common Mongolic. We establish that GZA should not be characterized as a synchronic process of /g/-deletion, though it arose from the aftermath of such a process historically. We further detail the shortcomings of describing GZA in terms of allomorphy, as well as the theoretical and empirical problems with treating consonant-zero alternations as splitting rather than insertion. We conclude that the Khalkha pattern is most insightfully described as epenthesis of a phoneme /g/ with varied surface manifestations. Since both markedness-based and splitting-based accounts face difficulties with GZA and other patterns of its type, these patterns may require a return to a constraint-based approach without a fixed markedness hierarchy, or a rule-based account such as the one we present.