UC Berkeley

The topic of this study is grammatical tone (GT), which I define as a tonological operation that is not general across the phonological grammar, and is restricted to the context of a specific morpheme or construction, or a natural class of morphemes or constructions. In typologizing grammatical tone, I frame it in terms of dominance effects (Kiparsky & Halle 1977, Kiparsky 1984, Inkelas 1998), and divide GT into two types. Dominant GT systemically deletes the underlying tone of the target (with or without revaluation by a grammatical tune), while non-dominant GT does not systemically delete it. From a cross-linguistic survey of GT, I develop a typological principle called the dominant GT asymmetry, which states that within a multi-morphemic constituent, the dominant trigger is a dependent (e.g. a modifier of affix), and the target is a lexical head or a dependent structurally closer to the lexical head. In this way, dominance is always directed ‘inward’ within morphological hierarchical structure, supporting earlier statements such as Alderete’s (2001a, 2001b) ‘Strict Base Mutation’.

For any theoretical model of dominant vs. non-dominant GT, I show there are three problems that must be addressed: the origin problem (where does the grammatical tune come from), the erasure problem (why do the underlying tones of the target go unrealized), and the scope problem (what determines where the grammatical tune docks, i.e. its scope). To this end, I develop a novel model which I call Matrix-Basemap Correspondence (an extension of Output-Output Correspondence - Benua 1997) and combine it with Cophonology Theory (Inkelas & Zoll 2007).

Under this theory, the origin problem is attributed to floating tones which are part of the underlying representation of the trigger. A major claim is that there is no representational difference between dominant and non-dominant tone: they both involve floating tonemes undocked to a TBU in the input. I implement my model within Distributed Morphology (Halle & Marantz 1993), whereby triggers of dominant GT are not constructions (as in classic Cophonology Theory – Inkelas 1998), but rather individual vocabulary items (following Sande & Jenks 2017). Dominant triggers have a special cophonology which ranks a constraint enforcing dominance higher than default constraints. This dominant constraint should be understood as a special type of faithfulness: correspondence between a matrix derivation and an abstract basemap consisting of only unvalued tone bearing units, e.g. an input-output basemap //ⒽⓁ + ττ//  \τ́τ̀\. It is to this basemap output that the matrix output must be faithful in tonal shape, resulting in the target’s underlying tone going unrealized (addressing the erasure problem). The central insight here is that dominant GT should be characterized as a special type of paradigm uniformity effect, a hypothesis referred to as dominance as transparadigmatic uniformity. In this way I derive dominance as faithfulness and not as markedness (Inkelas 1998) or a special construction constraint (McPherson & Heath 2016).

Finally, to address the scope problem I develop a theory in which syntactic structure is mapped to a hierarchical morpho-phonological tree via an operation at spell-out called hierarchy exchange. Within instances of dominant GT, a mother node in the morpho-phonological tree consists of the trigger of the grammatical tune (one daughter) and the target (the other daughter). The cophonology of the trigger scopes over the entire sequence, with cophonologies applying cyclically at each node resulting in ‘layers’ of grammatical tone. One advantage of Cophonology Theory is that it is intrinsically cyclic and thus captures the fact that dominance is always inward without stipulation.

A major component of this model is that hierarchy exchange preserves the inside-out derivational history of the syntactic module by referencing asymmetrical c-command. In this way, I conclude that syntax/phonology interface models which appeal to c-command are essentially correct, the most relevant being McPherson (2014) and McPherson & Heath (2016) which derive dominant GT scope via c-command. However, the model proposed in this study differs from them by having only indirect reference to c-command, mediated by hierarchy exchange. I conclude that the real legacy of c-command may not be linearization (as in Kayne 1994), but rather is in delimiting the scope of morphologically-triggered phonological operations.

In short, I posit a model which includes (i) floating tone representation, (ii) output-output correspondence (with abstract basemap induction), (iii) spell-out operations which apply in parallel, (iv) cyclicity within the morpho-phonological module, (v) indirect reference to syntactic structure and syntactic relations (e.g. c-command), and (vi) cophonologies triggered by DM vocabulary items. By combining disparate models and expanding on others, we arrive at a novel account of grammatical tone with extensive empirical coverage.