The moraic mess that is Danish

Dual weight criteria and mora recursion

(with Yonatan Goldshtein)

In moraic theory, particular syllables of a languages are typically defined as either heavy of light for all phonological rules (Hyman 1985). However, some languages seem to treat the same syllables as light for some rules and heavy for others (e.g. Davis 2011:129-132). To account for such behaviour, Hayes (1995) proposes that moraic structure can be organized as a layered grid, where the bottom layer, typically involved in syllable-internal patterns, underlies the top layer, typically involved in higher-level prosodic structure. The present paper argues that Danish is just such a language: distinguishing two separate criteria for assigning weight to syllables allows us to understand the phonology of stød. We implement a Stratal Phonology analysis (Bermúdez-Otero 2018) by reimagining Hayes’ moraic grid representation as recursion at the level of the mora (Iosad 2013). Specifically, we suggest that maximal moraic projections are involved in syllable-level processes (Hayes’ bottom layer) and minimal moraic projections are involved in the processes assigning stød (Hayes’ top layer).

Previous models of syllable weight in Danish have focused either on syllable-internal structure (e.g. Vazquez-Larruscaín 2021), or on the distribution of stød (e.g. Basbøll 2005, 2008). The key fact of syllable structure is that the maximal syllable in Danish is CVXsC: host [hoːˀst] ‘cough’, tekst [ˈtɛkst] ‘text’. In this account, the ‘core’ syllable is maximally bimoraic, /sC/ is the maximal ‘appendix’, and moraic structure is assigned without regard to sonority.

The key fact of the phonology of stød is that the second mora needs to be sonorous for it. When this condition is met, stød is contrastive: kvart [ˈkhvɑːt] ‘quarter’ vs. start [ˈstɑːˀt] ‘start’. In syllables with only one sonorous mora, (kat [ˈkhæt] ‘cat’), the contrast is neutralized. Basbøll argues that only sonorant segments can project morae, and further, that stød is the default prosody of the second mora of heavy syllables. In this analysis, a syllable with two sonorous morae that does not have stød needs special marking. In particular, word-final sonorants can be specified as extra-prosodic: in this case, they do not contribute to moraic structure, inhibiting stød (ven [ˈvɛn] ‘friend’), as long as they remain word-final. The addition of (most) suffixes causes them to undergo default mora assignment, which results in the appearance of stød (vennen [ˈvɛnˀn̩] ‘the friend’).

The paper develops a model that integrates these two main insights. Vazquez-Larruscaín’s (2021) phonotactic generalizations apply to maximal moraic projections, which are constructed without reference to sonority. Mora recursion, by contrast, is only licensed by high sonority: therefore, obstruents are unable to support non-minimal projections and cannot carry stød. Sonorants do project recursive morae by default, but can be blocked by faithfulness from doing so, accounting to lexical marking of non-stød, as in Basbøll’s model. We show how constraint interaction in a Stratal OT grammar derives the contextual nature of this faithfulness effect. As a result, our model retains the full empirical coverage of the competing approaches whilst simplifying the necessary lexical specification and capturing the morphological influences on stød assignment (Goldshtein 2023, Goldshtein & Iosad 2022).

← Length in strange places

|  All presentations  |

In search of (lost?) contact-induced change in phonology →