The ER source data comes from Erich
Round and
should be cited as:
Round, Erich R. 2019. Phonemic inventories of Australia.
Notes on the inventories and their compilation from Erich are below. The data are available in phoible long format in ER_inventories.tsv and contain only phonemes, with no information on allophones or linguistic tone.
Phonemic analysis is a valuable prelude to cross-linguistic comparison. Nevertheless, phonemicization is not a deterministic process and linguistic practice is not uniform. This has certain consequences for the compilation and use of cross-linguistic summary datasets such as this one. The following notes are arranged by topic and highlight how these issues have been accommodated in the compilation of the Australian dataset.
The labeling of phonemes acts to group phonemes, implicitly or explicitly, into classes (e.g. ‘the stops’), yet phonemes may class together differently with respect to allophony vs phonological alternations vs phonotactics and even vs their history. Consequently, analysts are often forced to choose between multiple conceivable classifications, each of which captures only some of the facts. When they do so, one analyst’s criteria may differ from those of another. As a result, some cross-doculectal variation in phoneme labels may reflect more a variation among analysts’ necessary choices between multiple, defensible options, than empirical differences among languages. Where this appears to be the case, I have in some instances relabelled phonemes relative to source documents, to promote cross-linguistic comparability. For example, stops in languages with no contrast for voicing or fortition are labelled here with plain, voiceless IPA symbols, following widespread phonological practice. More cases are noted below.
The task of dividing words into segments can itself present challenges, since in this task phonemic analysis becomes interdependent with the analysis of phonotactics. The phonotactics of Australian languages are highly constrained (Hamilton 1996), and this has enabled researchers to converge on broadly similar criteria for segmentation, enhancing the comparability of cross-linguistic data. However, in cases where languages contain unusual phonotactic sequences at the phonetic level, the analytical response in terms of segmentation and phonemicization has been more varied. By comparing across sources, it is possible to discern certain recurrent patterns of analytic practice. The next set of topics records attempts to accommodate these issues in the Australian dataset. The aim has been to bring to the fore empirical variation, over the variability among analytic choices.
Some sources analyse a phonotactically unusual phonetic sequence [X+Y], particularly ones which descend historically from [X] or [Y], as a complex phonemic segment /XY/; others analyse it as a phonemic sequence /X/+/Y/. However, no complex segment /XY/ posited for an Australian language is contrastive with /X/+/Y/, at least under some, reasonable set of assumptions about the rest of the inventory. Consequently, for the purposes of enhanced cross-linguistic comparability, all such sequences are treated in the dataset as two-segment phonemic sequences, /X/+/Y/. Notes on specific types: (1) A minority of Australian languages permit phonetic stop+nasal or stop+lateral sequences only at homorganic places of articulation. In original sources these are analysed as sequences (e.g. Sommer 1969 on Kunjen) or single complex segments (e.g. Breen 1981 on Arandic languages). (2) Almost all Australian languages permit phonetic nasal+stop sequences at homorganic places of articulation, but word-initially they are rare. In original sources they are analysed as sequences (e.g. Dixon 1991 on Mbabaram) or single complex segments (e.g. Crowley 1981 on Mpakwithi). (3) Few Australian languages permit phonetic stop+liquid sequences. The sequence [tr] in particular is sometimes analysed in original sources as a single complex segment (e.g. Hale 1997 on Linngithigh). (4) In Australian languages [w] often colors neighboring vowels; furthermore, most Australian languages permit [C+w] phonetic sequences, and these may be temporally coarticulated (e.g. Round 2009:65 on Kayardild). In some languages, these facts are analysed with phonemic /C/+/w/ sequences (e.g. Round 2009), but occasionally with complex, labialized consonants (Dixon 1991 on Mbabaram). (5) Harvey (2011) presents arguments for analysing ‘pre-palatalized’ complex segments in Arandic languages as /j/+/C/.
All Australian languages contrast either one or two ‘apical’ places of articulation, where the primary constriction is formed with the tongue tip; and either one or two ‘laminal’ places of articulation, with constriction formed by the tongue blade. In this dataset, laminals are labelled either as dental (with the IPA bridge diacritic) or pre-palatal (with IPA curl). Apicals are plain /t/, /n/ etc. for apical alveolars, IPA retroflex for retroflexes, and for languages with no alveolar–retroflex contrast, the single apical place is explicitly labelled as apical (with the IPA inverted bridge diacritic).
Some Australian languages contrast two series of stops, typically matched at all superlaryngeal places of articulation, and sources may phonemicize them in a number of ways. For the purposes of enhanced cross-linguistic comparability, this dataset follows Butcher (2004, 2012) and phonemicizes them as fortis and lenis, with no voicing contrast.
Some Australian languages have an extra, voiced lenis, stop-like or tap/flap-like phoneme solely at the apical place(s) of articulation. Sources may phonemicize these as voiced/lenis stops or as taps/flaps (see Breen 1997 for discussion). For cross-linguistic comparability, they are phonemicized here as tap/flaps.
editor's note: The retroflex lateral flap that was originally represented
as ɺ̢ (027A+0322, the alveolar lateral flap + retroflex hook) in Dr.
Round's data has been changed by the PHOIBLE editors to be represented as ɺ̠̺
(027A+0320+033A, the alveolar lateral flap with retraction and apical
diacritics).
Most sources phonemicize long vowels as a single long vowel segment. Some phonemicize them as two, adjacent short vowels (e.g. McDonald & Wurm 1979:5 for Wangkumara [a:]). The phonemicization here is as the single, long segment.
Some sources analyse the unpredictable presence/absence of a glottal stop as a prosodic feature (e.g. Waters 1980 for Djinang) and others as a phonemic segment. The representation here is as a phonemic segment.
