Background

Language Family: Sino-Tibetan / Tibeto-Burman / Himalayish / Mahakiranti / Kham-Magar-Chepang-Sunwari / Sunwari

Phonology

/ɖ/, /tʃʰ/, and aspirated voiced stops occur in Nepali loanwords (Borchers 2008, 24).
/ʈ/ and /ʈʰ/ also occur almost exclusively in loanwords, but there exists a small set of native Sunwar words that feature them (Borchers 2008, 34–35).
/ɓ/ once existed in Sunwar, but has since been lost; however, it existed recently enough that contemporary speakers remember the sound change as having occurred. In different areas, the phoneme has moved to either [b] or [w] (Borchers 2008, 33–34).
There exists a retroflex consonant that only occurs as the second sound in a consonant cluster - different speakers realize it as [ɭ] or [ɽ], and perhaps also [ɖ]. Although these were probably a single phoneme in an older form of Sunwar, they have largely diverged between different areas where Sunwar is spoken; I will be following one theory proposed by Borchers (2008), treating the modern retroflex consonants as allophones of /l/ and /ɾ/ respectively (pp. 36-38).
/ʔ/ is only phonemic in rare cases (followed by stops in the codas of syllables with long vowels) (Borchers 2008, 39–40).
Rapacha (2013) argues for alveolar rather than postalveolar affricates (p. 69).

	Place of Articulation
Manner of Articulation	Labial	Dental	Alveolar	Postalveolar	Retroflex	Palatal	Velar	Glottal
Stops	p pʰ b	t̪ t̪ʰ d̪			ʈ ʈʰ		k kʰ ɡ	ʔ
Affricates				tʃ dʒ
Fricatives			s	ʃ				h
Nasals	m		n				ŋ
Flaps			ɾ
Approximants	w	l̪			ɭ	j
Note: For phonemes that share a cell, those on the left are voiceless and those on the right are voiced. Where three phonemes share a cell, they are ordered voiceless, aspirated, and voiced.

Vowel length is phonemic for /a/ and /u/ (Borchers 2008, 26).
Rapacha (2013) argues for /ə/ as a sixth vowel in Sunwar (pp. 57, 59-60).
All vowels, except for /uː/, nasalize contrastively; however, Borchers (2008) theorizes that this contrast may be gradually disappearing in Sunwar (pp. 23, 27-28).
Tone exists in Sunwar, but Borchers (2008) argues that it is predictable, and states that he was unable to identify any minimal pairs that differ only in tone (pp. 28-30). Similarly, Rapacha (2013) remarks that tone contrasts are typically neutralized in the dialect areas he studied (p. 107). For this reason, I have chosen to omit tone in my ruleset.
The Sunwar diphthongs may occur as sequences of vowels when they exist across morpheme boundaries (Borchers 2008, 31). Given this, I will not include them below.

	Front	Central	Back
High	i ĩ		u ũ uː
Mid	e ẽ		o õ
Low		a ã aː ãː
Note: Vowels with the diacritic (ː) are long. Vowels with the diacritic (◌̃) are nasal.

The glottal stop is not consistently represented in the orthography, compromising the language. Some authors write it by repeating the vowel while adding the virama diacritic (्) to the second; others represent the glottal stop with the character for /k/, and others don’t include the glottal stop at all (Borchers 2008, 49). Because of the relatively rare nature of /ʔ/, I have opted to proceed with the language anyway. In the rule set, I have opted to represent the glottal stop with the first method listed above. However, even with this, /ʔ/ does not appear in the Crúbadán corpus.
/ɭ/ and /ʃ/ are not always represented orthographically as well (Borchers 2008, 49). At least for /ɭ/, accounts explain that it may be represented by the same grapheme as /ɾ/, or even /ɖ/ or /l̪/. Below I have listed that both /ɾ/ and /ɭ/ are transcribed by the same grapheme, compromising the language. I have defaulted to /ɾ/ in the rules, meaning that /ɭ/ is not represented in the Crúbadán corpus.
The aspirated phonemes present only in loanwords are not reflected in the ruleset.

Consonant Grapheme	Phoneme	Comment
क	/ka/
ख	/kʰa/
ग	/ɡa/
घ	/ɡʱa/	only in loanwords
ङ	/ŋa/
च	/tʃa/
छ	/tʃʰa/	only in loanwords
ज	/dʒa/
झ	/dʒʱa/	only in loanwords
ट	/ʈa/	rare
ठ	/ʈʰa/	rare
ड	/ɖa/	only in loanwords
ढ	/ɖʱa/	only in loanwords
त	/t̪a/
थ	/t̪ʰa/
द	/d̪a/
ध	/d̪ʱa/	only in loanwords
न	/na/
प	/pa/
फ	/pʰa/
ब	/ba/
भ	/bʱa/	only in loanwords
म	/ma/
य	/ja/
र	/ɾa/; /ɭa/	/ɾa/: default in the rules
ल	/l̪a/
व	/wa/
स	/sa/
ष; श	/ʃa/
ह	/ha/
Vowel Grapheme
अ	/a/
आ	/aː/
इ; ई	/i/	vowel length is not contrastive for /i/
उ	/u/
ऊ	/uː/
ए	/e/
ओ	/o/
Diacritic
्		called a virama, this diacritic removes the inherent vowel or indicates a glottal stop when following duplicate vowels (see comment above)
ा	/aː/
ि; ी	/i/	vowel length is not contrastive for /i/
ु	/u/
ू	/uː/
े	/e/
ो	/o/
ं		called a candrabindu, this diacritic marks nasalization of the preceding vowel

The syllable structure of Sunwar is (C)(C)V(V)(C)(C) (Baskaran 2009, 629–30). Syllables must have at least one consonant, and in most cases can contain a maximum of four sounds.

Some speakers spirantize /pʰ/ to [ɸ] (Borchers 2008, 32).
Syllable-final /n/ and /ŋ/ are often realized weakly, and simply nasalize the preceding vowel (ibid.).

In fast speech, the short vowel /a/ tends to back to [ɑ], and /e/ tends to lower to [ɛ] (Borchers 2008, 25).
/i/ lengthens to [iː] in some word classes (Borchers 2008, 26–27).
Word-initial /w/ may gain aspiration (Borchers 2008, 33).
/ɾ/ trills to [r] before palatal or velar consonants (Borchers 2008, 35).
Epenthesis of [x] occurs between /ɾ/ and /tʃ/ (Borchers 2008, 39).
Word-final vowels may delete in fast speech (Rapacha 2013, 68).
Postvocalic /l/ velarizes to [ɫ] when followed by another /l/ (Rapacha 2013, 83).
Consonant gemination in Sunwar occurs, most commonly with /b/, /tʃ/, /k/, /l̪/, /m/, /n/, /p/, /s/, and /t̪/ (Rapacha 2013, 94).

Baskaran, S. Ganesh. 2009. “Sunwar.” In Sikkim: Part-I, 614–74. Language Division Office of the Registrar General & Census Commissioner.

Borchers, Dörte. 2008. A Grammar of Sunwar: Descriptive Grammar, Paradigms, Texts and Glossary. Brill.

Rapacha, Lal B. 2013. “A Descriptive Grammar of Kirānti-Kõits.” PhD thesis, Jawaharlal Nehru University.