Revisiting Phonological Generalizations in Australian Languages*

1.1 Aims and background Australian languages are famous for their near-uniform phonemic inventories. Authors such as Busby (1980); Dixon (1980); Hamilton (1996); Dixon (2002); Butcher (2006), amongst others, have emphasized the similarity of phoneme inventories across the continent, using it as evidence of intensive lexical diffusion and linguistic convergence (Dixon 1997, Dixon 2002:547). Cross-linguistic surveys of Australian languages such as those just mentioned have repeatedly published claims regarding the patterns found in the ‘typical’ Australian language, as well as other phonological generalizations meant to apply more or less equally to all of the approximately 360 languages of the continent. The apparent uniformity of Australian languages also stands out in worldwide typological surveys (Maddieson, 1984; Mielke, 2008; Hunley et al., 2012). Moreover, Australianists have often assumed that uniformity in the phonemic inventory is coupled with uniform phonotactics (Dixon, 2002:547). Otherwise unqualified statements about uniformity in inventory and phonotactics are also easily found in reference grammars of languages in the region (for one example among many, compare Goddard 1985:21, 43, 66, 323). This ‘typical’ inventory is given in Figure 1. This assumption is, in itself, surprising, given that there is no general assumption in phonology that associates uniformity in inventory size or composition with phonotactic generalizations such as syllable structure constraints or segment frequencies. Such phonological uniformity, if real, is surprising and unusual given the country’s phylogenetic diversity. Pace Dixon (2002), there are approximately 28 phylic families (Dixon, 1980; O’Grady et al., 1966), and the largest of these, Pama-Nyungan, has approximately 30 low-level subgroups.1 Lexical differentiation among Pama-Nyungan subgroups is extensive and rates of lexical replacement are high (Bowern & Atkinson, 2012). However, contra Dixon (2001, 2002), rates of borrowing are, for the most part, low (Bowern et al., 2011; Bowern & Atkinson, 2012). The only ways for this large number of languages to be so similar would be through intensive, continent-wide language contact; shallow linguistic relationships; or extreme conservatism in sound change. None of these assumptions is particularly plausible. Bowern et al. (2011) show that levels of language contact are variable and in scale with other languages of the world. Bowern & Atkinson (2012) find high rates of lexical replacement, implying that the linguistic relationships among subgroups of PamaNyungan are not particularly shallow.2 Sound change conservatism has not been investigated but publications such as Alpher (2004); Koch (1996, 1997), and others, provide evidence for sound change in these languages. The claims of uniformity should thus be treated with skepticism. In this paper, we revisit claims of uniformity in Australian languages by testing variation in aspects of the phoneme inventories and phonotactics of a representative sample of languages. We do this by utilizing