Phonology
This page gives a basic overview of Mirja's phonology. See the referenced pages for more detailed descriptions of individual phenomena.
Segmental phoneme inventory
Mirja has the following segmental phonemes.
Consonants
Sounds in parentheses appear only as the result of phonological or morphophonemic processes, and do not appear in the underlying representation of any morphemes.
| Bilabial | Labiodental | Apical | Palatal | Velar | Glottal | ||
|---|---|---|---|---|---|---|---|
| Obstruent | Voiceless stop | h[n 1] | t | k | |||
| Voiceless fricative | (ɸ~w̥) | (f) | s (θ) | x | h | ||
| Voiced | v | d~ð | g~ɣ | ||||
| Nasal | m | n | (ŋ) | ||||
| Liquid | Lateral | l ɬ | |||||
| Rhotic | ɾ ɾ̥ | ||||||
| Approximant | w (ɸ~w̥) | j (j̥) | |||||
- ↑ Underlyingly /p/ for some purposes, but surfaces as [h] in isolation
/r̥/ is often pronounced closer to something like [ʂ], and /j̥/ is likely best understood phonetically as [ç].
Consonant mutations
There are three processes that can change one consonant into another. The most straightforward is a dissimilation process with /s/: the second of two consecutive syllables with /s/ will replace the /s/ with /θ/:
su-s
3sg-TOP
va-t
do.it-PAST
'They did it'
The second is a merger process when two sufficiently similar consonants come together. Usually, when two consonants come together underlyingly, the result is either gemination (if they're identical) or the insertion of an epenthetic vowel between them (if they're different). However, certain pairs of different consonants merge into a geminate when they come together. In the table below, these otherwise unexpected mergers are given in bold. (Expected geminates are non-bold, and epenthetically separated pairs are indicated by '--'.)
| Second phoneme → | t | l | n | k |
|---|---|---|---|---|
| First phoneme ↓ | ||||
| t | tt | ll | nn | -- |
| l | ll | ll | nn | -- |
| n | nn | ll | nn | ŋŋ |
An example of this is given below.
don-l
drink-NEG
'didn't drink'
TODO: when three come together (e.g. don-l-t)
The third and most significant process that transmutes phonemes is topicalisation. Mirja's basic topic marker is a feature bundle that docks to the final consonant of a word and transforms it into a voiceless continuant:
no-rV-*
1sg-OBL-TOP
karu
is.visible
'I see it'
Vowels
| Front | Non-front | ||
|---|---|---|---|
| Unrounded | Rounded | ||
| High | i | ɨ | u |
| Low | e | a | o |
In addition to the above vowel phonemes, Mirja has two different kinds of automatic vowel insertion. One is simple epenthesis - /a ɨ/ are inserted to break up consonant clusters that can't be resolved by the above merger processes (/a/ if the preceding vowel is low; /ɨ/ if it's high). The other is 'empty vowel slots' - certain morphemes have segments specified for being A Vowel, but with no information about which vowel. In these cases, the previous vowel is copied entirely.
Diphthongs must share rounding and height across both elements; as a result, /ɨi ae/ are the only two valid diphthongs. When morphology causes any other vowels to come together, the result is a long version of the second vowel:
e-ima
eat-morning
'eat (some) breakfast'
Suprasegmentals
Mirja has both phonemic tone and an active (non-phonemic) stress system, which interact in complex ways. In general, tones are assigned starting at the left edge of the morpheme they belong to, then stress is assigned in ways that partially depend on tone (but also on syllable weight), and then tones can be moved to better line up with the resulting stress locations.
Mirja has four phonemic tone melodies: /L H LH HL/. MORE.
Stress assignment
STUFF
Tonological processes
/H/ spread
A high tone will spread rightwards until it encounters some other tone or the end of its phonological word.
EXAMPLES.
= /HH/ blocking
Mirja disallows two marked high tones from being adjacent to each other, and will insert a low tone between them to break them up. This can look somewhat different depending on the context it happens in. In an underlying non-final /HH/ sequence, the inserted L will push the H rightwards to create [HLH]. In an underlying /HØ...H/ sequence where the adjacency is due to the above spread rule, the L is placed to the left of the final H, resulting in an [HH...LH] sequence - in other words, the H spreads rightwards until it hits the automatic L in front of the next H. In an underlying /HH#/ sequence, where there's no room to reassign the H rightwards due to the word boundary immediately following, the H simply ends up off the end of the word entirely and is either outright deleted or left floating with no further evidence of its continued presence - in other words, resulting in an [HL#] sequence on the surface.
EXAMPLES.
Contouring and merging
STUFF