RESEARCH POSTER PRESENTATION DESIGN © 2015

www.PosterPresentations.com

Conclusion

The study of patterns of (segmental) phonemes within the word falls under phonotactics, morpheme (and word) structure constraints, and the study of (morpho)phonological processes. The domain of phonotactics has been taken to be the syllable, supplemented in some studies by sensitivity to word boundary (and sometimes morpheme boundary and even reduplication boundary). Notably, the surveys in WALS (Maddieson 2013), and the World PhonotacticsDatabase (Donohue et al. 2013) are entirely about the structure of the syllable. Any dependency across a word-internal morpheme boundary typically involves alternations and so is accounted for by (morpho)phonological rules (or active constraints), or effects of metrical structure.The standard view (as outlined) tacitly assumes the absence of phonotactic patterns spanning more than one syllable within a morpheme (but not necessarily spanning an entire polysyllabic morpheme). Even so, some such patterns have been long recognised, under the rubrics of vowel (or other) harmony, assimilation (or dissimilation), and OCP effects. Some early statistical studies of connected speech (and writing) were alert to trans-syllabic phoneme patterns insofar as they showed a correlation between word boundaries and the rise and fall of Markovian transition probabilities.

Background

Previousstudies REFERENCESBreen, Gavan. 1981. The Mayi languages of the Queensland gulf country (AIAS new series 29). Canberra: Australian Institute of Aboriginal Studies. Donohue, Mark, Rebecca Hetherington, James McElvenny & Virginia Dawson. 2013. World phonotacticsdatabase. Department of Linguistics, The Australian National University. http://phonotactics.anu.edu.auHercus, Luise A. 1994. A grammar of the Arabana–Wangkangurru language, Lake Eyre basin, South Australia, vol. PL C-128. Canberra: Pacific Linguistics, Research School of Pacific and Asian Studies, Australian National University. doi:10.15144/PL-C128. http://dx.doi.org/10.15144/PL-C128.Macklin-Cordes, Jayden L. 2015. Phylogeny and phonotactics: Quantifying historical signal in sequences of sound. BA(Hons) subthesis. https://www.academia.eduMaddieson, Ian. 2013. Syllable structure. In Matthew S. Dryer & Martin Haspelmath (eds), The world atlas of language structures online, Leipzig: Max Planck Institute for Evolutionary Anthropology. http://wals.info/chapter/12McGregor, William. 1990. A functional grammar of Gooniyandi (Studies in Language Companion Series 22). Amsterdam/Philadelphia: John Benjamins. http://trove.nla.gov.au/work/17301223

Contact david.nash@anu.edu.auhttp://www.anu.edu.au/linguistics/nash/

Some correlation between vowels in adjacent syllables had been noticed in particular Australian languages Gooniyandi (McGregor 1990:88–89) and Ngawun (Breen 1981:28). For (non-PNy) Bardi, Bowern (2012:98-100) showed interaction between V1 and V2: ‘there is a strong tendency for the V2 vowel to be identical to the V1 vowel’ and also found more aCi sequences than expected. Hercus (1994:54) noted the near absence of iCu sequences in Arabana. Macklin-Cordes 2015 studied transition frequencies between adjacent segments for Ngumpin–Yapalanguages (and thus not V/V interaction); later extended to Yolŋu Matha.

AustralianNationalUniversity

DavidNash

‘*iCu’inAustralia:Phonotactics beyondthesyllablelevel

ScopePama-Nyungan languages (and many other Australian languages) conform to the simple assumptions here: • suffixing • initial stress • stems of at least 2 moræ • typically 3 vowels /i,a,u/ in stems (some languages also with /e/, /o/) Other assumptions: vowel length ignored; study confined to first two syllables (trochee); compounds with monosyllabic first element counted only if written as one word

Furthertopics• effect of intervening consonants on VCV patterns: e.g. only i[+labial]u is allowed (as in Parnkala (Thura-Yura), Yapa languages)• diachrony: loanword detection; inference on direction of change; role in Arandic developments• lexical frequency interaction; inter-word interaction: statistics on corpora rather than lexicons• phonetic motivation for interaction?: articulatory, auditory, acquisition; possibly related to word discrimination (‘In short, the evidence that infants rely on transitional probabilities between syllables to segment words from speech is quite convincing.’ Johnson 2012:59)• assessing areal and phylogenetic components of the distribution of phonotactic patterns; see Macklin-Cordes

AcknowledgementsTo LA Hercus, H Koch, J Simpson for discussion (1990).To the compilers of dictionaries used (via OzBib, AILEC).To Claire Bowern for AustralianLanguageLocations

https://zenodo.org/record/848646.To Siva Kalyan and Nay San for R tutorials.

ThisstudyOur study began with the typical 3 vowels /i,a,u/ of stems in Pama-Nyungan languages

neighbouring the Arandic subgroup in central Australia, and revealed a fairly widespread avoidance (more or less) of the high back vowel /u/ following the high front vowel /i/, across intervening (non-labial) consonants C (and independent of metrical structure). In other languages of the region the iCu sequences occur but at a lower frequency than would be expected just from the overall vowel frequencies; also, there is some complementary preference for identical V in adjacent syllables (especially i and u). As a graded property the iCudispreference shows a contiguous areal distribution; there is a similar weaker dispreference in the Western Desert dialect web, and rather different dispreference in other languages further afield, some more strongly avoiding aCu.

In Warlpiri and Warlmanpa (Yapa languages), the sequence iCu occurs only when the consonant is p or w (Nash 1986:73-4). It might be expected that m would also participate in this but there are no examples of an imu (or impu) sequence. In Warlpiri, surface iCu is no more possible across a morpheme boundary than intra-morphemically, as vowel assimilation rules operate to change underlying /iCu/ to iCi or to uCu. In Warlmanpa there are no such vowel assimilation rules and iCu does occur across a morpheme boundary (as also in Arabana-Wangkangurru).

Key(E–O)2 sign(E–O)

E * N/1000

The indicator in each cell is

whereE is ExpectedO is ObservedN is the total Number of words

language name lexicon size

V2

V1

cells are shaded red to greenfor the deviation (under to over) from number of V1-C-V2 expected if V1 and V2 were independent across the lexicon

Pattern persists across lexicon size sampled, e.g.:

Arandic languages: front ~ back distinction largely lost for high vowel

Phonotactic typology can extend to (intramorphemic) vowel patterns beyond the syllable level. Many Australian languages more or less avoid the iCu sequence, and prefer iCi and uCu.

subgroups of Pama-Nyungan

Poster presentation at the 12th Association for Linguistic Typology conference (ALT2017), 13 December 2017