Rannsachadh na Gàidhlig, University of Edinburgh, Edinburgh, UK.
I provide a new analysis of the ‘tonal accent’ opposition in Gaelic, conceptualizing it as involving lexically specified metrical — in particular foot — structure. As discussed among others by Ternes (1980, 2006), many Gaelic varieties use suprasegmental features such as pitch (e.g. Lewis, Wester Ross) or glottalization (e.g. Jura, Colonsay) to make two kinds of distinctions: disyllabic hiatus vs. heavy syllables (àth vs. atha); and disyllabic vs. svarabhakti words (balg vs. balach). Ternes rightly compares these patterns with tonal accent systems in Germanic, and postulates lexical tone as the key distinction between the two ‘accents’.
Non-tonal analyses have been proposed by authors such as Bosch & de Jong (1997); Smith (1999); Brown (2009); Iosad (2015). In particular, Smith and Iosad offer analyses where the svarabhakti vowel in words like balg is metrically invisible, with consequences for the patterning of tone and glottalization. However, this analysis can only work if svarabhakti vowels are fully predictable synchronically. In this simple form, the analysis is untenable: svarabhakti vowels may be absent in contexts where they are phonologically motivated (e.g. [x] representing preaspiration of [k] does not trigger svarabhakti olc [ɔɫxk], *[ɔɫɔxk]), and conversely svarabhakti-related suprasegmental patterns can appear when they are synchronically unmotivated (e.g. [faɫa.i], [mara.i] with ‘monosyllabic’ accent for falbhaidh, marbhaidh). Similar counterexampels can be adduced for glottalization: compare [marəv] for marbh with svarabhakti-induced lack of glottalization but [marɪç] marbhaidh where the lack of [ʔ] cannot be so explained. Building on recent work in the analysis of tonal accents (e.g. Morén-Duolljá 2013, Köhnlein 2016, Iosad 2016, Kehrein 2017), in this paper I show how these problems can be resolved if both glottalization and tonal accents in Scottish Gaelic are analysed using lexically pre-specified foot structure.