Consonants
The underlying (=morphophonemic) consonant inventory is as follows:
|m n ŋ|
|p t s k|
|β ɹ z ɰ|
|r j w|
Phonemically there are several other consonants, but they can be analysed as arising from the application of phonological rules:
- The retroflex consonants /ʈ ɳʈ ʂ/ can be straightforwardly analysed as clusters /tr ntr sr/.
- The palatals /ɲ c ɲc ɕ/ are in complementary distribution with the alveolars and velars after /i j/. It appears that all palatals can be analysed as either an alveolar or velar consonant which has been palatalised by a preceding /i/ or /j/. (But it may have disappeared on the surface: e.g. |kəeintəna| /ˈkeːɲcəna/, though this is uncommon.)
- You could analyse nasal+stop consonants as prenasalised on the basis that /ɳ/ occurs only in the cluster /ɳʈ/. I don’t see this as a particularly compelling analysis, but I’ll use it nonetheless.
/m n ɲ ŋ/
/p t ʈ c k/
/mp nt ɳʈ ɲc ŋk/
/s ʂ ɕ/
/β z ɹ ɰ/
/r j w/
Vowels
There are seven short vowels and five long vowels:
/a e i o u ə ɨ/
/aː eː iː oː uː/
Phonotactics and stress
Maximal underlying syllable structure is ((C₁)C₂(r))V(C₃). All consonants but /ŋ/ are allowed in the onset; the coda can be either a nasal or a stop. In onset clusters, C₂ can be a stop or /s ɰ/; if it is a stop, C₁ may be a homorganic nasal.
Primary and secondary stress are predictable from morphophonemic representation but become unpredictable if given only the surface form. In stressed syllables (of either type), the nucleus may only be a long vowel or one of /a i/. You need not represent stress in a romanisation.
Sample
/naˈɳʈaːməˌnuːɹəˌnaː.it ˈwaβeˌweːjem ‖ sitˈniː ˈarəˌeːweˌruːɹamˌkaŋ ˈarəˌeːraruˌɹaːm ‖ niːm ˈjeːrasaˌɹeːwat | woˈɹuː | ˈpaŋk.siˌjaː | sɨˈmpiɹiˌjam | zaˈɹeː rəˈweːweˌβaːɰaˌβaːɰɨ | ˈɹeːɰep.kɨ ‖ ˈwaβeˌreːraniˌɲeːm ‖ ˈsitni 44 ˈjeːjatnuˌʂaːɰiːɲatˌkaːŋɹam ‖ naː kaˈβaːɰəŋ ˈwiːreˌaːŋkrəm ‖ ɰaˈɹaŋkəŋ ˈtajəˌsaβiˌneːwanaˌniːcəɹaːm ‖ ˈwiːmatˌki ˈiːwiˌŋkraɹaˌɰiːzəm ˈweːreriˌtiːɕo/