Disentangling stress and pitch accent: A typology of prominence at different prosodic levels

There are numerous studies devoted to the typological examination of stress (e.g. Hyman 1977, Liberman & Prince 1977, Hayes 1980, 1995, Prince 1983, Halle & Vergnaud 1987, Goedemans and van der Hulst 1996, Elenbaas & Kager 1999, Gordon 2002, Hyde 2002, Heinz 2009, van der Hulst & Goedemans 2009, van der Hulst et al 2010). This research program has produced many interesting observations and theoretical treatments of stress. Nevertheless, despite the advances it has made, stress typology is subject to limitations that make its findings difficult to situate within a broader theory of linguistic prominence encompassing intonation and prosodic constituency. In particular, even early stress typologists (e.g. Hyman 1977) observed that most typological and theoretical work on stress is based on descriptions gleaned from grammars, for which it is plausible, even likely, that the reported stress patterns are based on words uttered in isolation, where the word is equivalent to an utterance. In such cases, the reported stress patterns more accurately reflect phraseor utterance-level prominence rather than true word-level stress. This chapter represents a preliminary attempt to tease apart word-level stress from prominence associated with larger prosodic units with an eye toward creating a typological database of both types of prominence and their relationship to each other. In particular, we focus here on the ability of peripheral syllables to bear prominence at both the word and phrase level. While this endeavor might seem premature due to the relative paucity of thorough descriptions that disambiguate prominence attributed to different constituents, there is a clear trend toward increasing awareness among linguists of the important distinction between wordlevel stress and higher-level prominence. Grammars are thus more likely nowadays to provide explicit, even if cursory, statements about whether their description of prominence refers to word-level stress or phrase-level prominence. Genetti’s (2007) grammar of the Tibeto-Burman language Dolakha Newar provides an example of a recent grammar with a detailed prosodic description featuring a five-page description of stress, a 22-page chapter on prosody, including phrasal accent and intonation, as well as several representative pitch and intensity traces. Closely intertwined with this movement toward greater clarity of prosodic descriptions is the now ubiquitous availability of free acoustic analysis software, in particular, Praat (Boersma & Weenink 2010), which allows for both rapid confirmation of impressionistic judgments about stress and phrasal prominence and the possibility of more systematic quantitative study of acoustic correlates of prominence of different types. Some grammars now include representative displays such as pitch traces, waveforms, intensity curves, that provide some sense of how prominence is acoustically manifested. The incorporation of phonetic data of this sort is