Proto-Hathic had the following consonant inventory:

/p t tɬ k ʔ/
/pʰ tʰ kʰ/
/b d g/
/ɓ ɗ/
/s ɬ ħ h/
/m n ŋ/
/j w/

/ʘ ! ǁ ǂ/, could also appear with aspiration, nasality, or voice

/a e i o u/ + length and rhoticity contrast
/ae ai au əi əu/

/əi əu/ may have been /oi ou/ or somesuch; their exact value cannot be reconstructed.

There was also a pharyngeal element on the vowel, that did not count toward either initial or coda for the purposes of phonotactics. This is written with an underdot.

Proto-Hathic developed into two languages, Cherun and Ngmwragh, after some common changes that completely restructured the sound system.

CVC syllable structure plus preinitials used for grammatical purposes.

Common changes

Strong penultimate stress accent led to syncope of unstressed short vowels and shortening of (rare) unstressed long vowels. Most words became monosyllabic.

Frontness/backness spread from vowels in both directions: the hypothetical words kek and tet merged as cec. This affected all consonants except the resonants /j w/. /a/ counted as front, as did the schwa elements of diphthongs. Diphthongs were considered contours: kauk would give cauk, not cauc.

u unrounded to ɯ.

y w not immediately preceding/following a vowel turn to i u

Proto-Hathic > Cherun

Pharyngeal element broke nonlow vowels, creating the new diphthongs eə iə oə ɯə. oə dissimilated to uə. Initial ħ broke high vowels, i u > əi əɯ, and shifted to h, or x in consonant clusters.

Click loss:
ʘ ! ǁ ǂ > pˠ k tɬ c
ʘʰ !ʰ ǁʰ ǂʰ > pˠʰ kʰ ɬ cʰ
ŋʘ ŋ! ŋǁ ŋǂ > mˠ ŋ n ɲ
gʘ g! gǁ gǂ > bˠ g ɟ l

High vowels get fucked with to make sure i never occurs next to a velar and ɯ never occurs next to a palatal. TODO

bʲ bˠ d ɟ g > vʲ vˠ z ʑ ɣ > fʲ ɸˠ s ɕ x.

Finals get eaten and turn to phonation: creaky, breathy, nasal.

Implosives merge with plain stops but push creaky voice onto the vowel, unless the vowel is nasal, in which case they eject a glottal stop and merge with nasals, so ʔm ʔn.

Not sure if I want to keep cʰ or merge it with c or ɕ.

ae > ɛː. i˞ merges with ɯ˞ in both short and long forms. Short e˞ merges with a˞.

Proto-Hathic > Ngmwragh

Need to do more here so it’s not so much like the protolang, but given the phonology I have now:

bʲ bˠ d ɟ g > v w r j ɣ

Velarized labials shift to labial-velars (could be a retention of original state as opposed to Cherun innovation)

Lose diphthongs somehow and probably laterals too, i dunno

I’m keeping the consonant inventory as small as possible for the purposes of demonstration, but the general principle still holds if it’s expanded.

Vowel:
/ə/

Consonants:
/p t tɕ k/
/pˤ tˤ tʂˤ qˤ/
/j w/

Syllable structure:
C(j/w)V

Vowel allophony:
Pharyngealized consonants lower following vowels, /j/ fronts them, and /w/ backs them. So:

Cə [Cə]
Cjə [Ci]
Cwə [Cu]
Cˤə [Ca]
Cˤjə [Ce]
Cˤwə [Co]

Why analyze it as having only one vowel?
Prefixes! Say you have a word jə, and a prefix pˤ-. The one-vowel analysis makes the best sense of why a word pronounced [i] without the pharyngealized prefix is pronounced [pe], not [pi]. You could also have j/w prefixes, so:

j- + əpə -> [əpə], [ipə]
j + kəpə -> [kəpə], /jəkəpə/ [ikəpə]

But you need to keep at least one vowel in there, so you can distinguish between, for example, /kəjə/ [kə.i] (or [kəji], if you prefer) and /kjə/ [ki].

Kannow is pretty exclusively head-marking, so nothing about alignment is marked on the nouns. On the verb, there are three possible roles: agent, patient, and what I call “secondary patient”, glossed D , because I don’t have a better name for it. Marking is a bit ridiculous, and changes depending on how many arguments the verb takes, so I’ll detail each separately. (Zero-argument verbs are impossible.)

One argument: The split is agent/patient and encodes volition: if the subject intended the action, it’s marked as the agent, and if not, it’s marked as the patient.

Erlgòsow.
[ʔeɵ̯ʂtɬkʷoə̯səʍ]
er-lgòs-ow
PST-fall-C1.A.SG
“He intentionally fell down.”

Erlgòsón.
[ʔeɵ̯ʂtɬkʷɵsəo̯nʷ]
er-lgòs-n`
PST-fall-C1.P.SG
“He accidentally fell down.”

Two arguments: The split is primary/secondary patient and encodes whether the object of the action was affected by it. In “I read the book”, the book would be in the secondary, unless I wrote notes in the margins or something. But sometimes this acts like the Finnish telicity distinction: (class 2 is the noun class for most animals, so ‘animal’ is implicit)

Ercagtow.
[ʔeʂcʰag̚θəʍ]
er-cag-t-ow
PST-shoot-C2.P.SG-C1.A.SG
“He shot the animal.”

Ercagtʼow.
[ʔeʂcʰag̚tʼəʍ]
er-cag-tʼ-ow
PST-shoot-C2.D.SG-C1.A.SG
“He shot at the animal .”

Three arguments: No split, because there’s no room for arguments to move.

——

Miar, on the other hand, keeps the “affected” distinction for all numbers of arguments, to the point of adding dummy arguments onto verbs that look like they should take only one. Affected arguments are marked as primary patients, even if they are also subjects; unaffected arguments are not double-marked, in the case of subjects, or marked as secondary instead of primary, in the case of objects. (In terms of affix order in the case of dummy marking, the subject comes before the object, and the object comes before the third argument.) Also, since, unlike in Kannow, a verb can have more than one argument in each role, word order, which is fairly standard nominative-accusative (specifically, nom-acc-dat), is slightly unimportant — but only slightly, since ambiguity is rare. (note: first-person is glossed as 1P, noun class x is glossed as Cx, also mind the harmony processes)

Rƞÿtu.
[ʐɳɜtʰɔ]
r-ñ-ÿ-to
1P.P-PST-1P.S-wash
“I washed myself.”

Qañäsö.
[qʰaɲɛtsʰʏ]
qa-ñ-ÿ-to
1P.P-PST-1P.S-wash
“I tried to wash myself, but it didn’t do any good.”

Rokñÿzÿy latʼol.
[ɽʊxɲɜdzɜj dɮatʼʊɮ]
r-ok-ñ-ÿ-dÿr latʼol
1P.P-C8.D-PST-1P.S-read book
“I read the book; the book was not affected by this action, but I was.”

Okñÿzÿy latʼol.
[ʊxɲɜdzɜj dɮatʼʊɮ]
ok-ñ-ÿ-dÿr latʼol
C8.D-PST-1P.S-read book
“I read the book; the book was not affected by this action, and neither was I. ”

Rölñÿzÿy latʼol.
[ɽʊɮɲɜdzɜj dɮatʼʊɮ]
r-öl-ñ-ÿ-dÿr latʼol
1P.P-C8.P-PST-1P.S-read book
“I read the book; the book was affected by this action, and so was I.”

Ölñÿzÿy latʼol.
[ʊɮɲɜdzɜj dɮatʼʊɮ]
öl-ñ-ÿ-dÿr latʼol
C8.P-PST-1P.S-read book
“I read the book; the book was affected by this action, but I was not. ”

I’m not going to give examples for three arguments because it’d take fucking forever, but you get the idea.

edit: Hahaha oh man I fucked up the vowel harmony and then went back to fix it and fucked it up again what am I even doing.

Proto-Miar vowel system: a e o i ɨ u + length distinction. Height harmony: a e o ~ i ɨ u, operates on both short and long vowels. (natlang precedent: Kusunda)

a o u > ɛ ø y everywhere but backing environments (dentals, retroflexes, uvulars, velars maybe?) — triggers eː ɨː > ea̯ ɯi̯

Length system collapses, producing a system of:

High: ɛ e ø a ea̯ o
Low: i ɨ y ɯi̯ ɨ u

Obviously these are hard to distinguish, so they get kicked apart a bit:

High: ɛ ɪ ø a e ʊ
Low: i ʌ y ɨ ʌ o

Not sure what to do with diphthongs yet.

Cerrais Kett bléng /bleŋ/ is cognate to Continental Kett ḵálň /kʷáʎɲ/.

How?

The Proto-Kett form is wkę́léng /ɸkǽléŋ/. Both languages share ɸk > kʷ and æ > a.

Insular Kett (of which Cerrais Kett is a dialect) turns the tone system into a stress system, with stress on the last high tone, and then kʷ > p > b (in order to avoid a merger with existing p, which develops aspiration and develops further to ɸ in some dialects), giving baléng in the standard. Then, in the Cerrais dialect, the unstressed vowel drops.

Continental Kett retains the tone system, but turns high-toned vowels to low-toned ones adjacent to another high-toned vowel and between two voiced consonants. ŋ palatalizes to ɲ after front vowels, and then low-toned vowels drop between two consonants of the same POA when an invalid cluster would not be created. Then the palatalization of ɲ spreads to l, giving ʎ.

——

Insular Kett antéyc /n̩ˈdeɥt͡s/ is cognate to Continental Kett örműh /ø̀rmýx/.

How?

The Proto-Kett form is arzemúsk /ɒ̀ɽèmúsk/. The aforementioned shift, æ > a, triggered a chain shift in both languages, shifting a o u > o u y (note that a is actually more like ɒ); however, in Insular Kett, o later merged back into a. Also, Insular merges ɽ into t, instead of r, as Continental does. Insular has a process of metathesis, similar to that of Comanche, where sequences of the form V₁C₁V₂C₂V₃ shift to V₁C₂C₁V₃. Also, high vowels break into e+semivowel clusters in Insular, which has only two phonemic vowels.

The two dialects also reduce the final consonant cluster in different ways. Insular drops the last stop in word-final fricative+stop clusters (and then shifts many final s to ts); Continental drops the s and fricates the stop.

To finish the sound changes for Continental: the low-tone vowel-dropping process mentioned above applies, and regressive backness-based vowel harmony applies.

A list of sound changes for the two cognate pairs below the fold.

Read more …

Note: The formatting is ass because Tumblr is a worthless joke of a site. I’ve tried to fix it God only knows how many times now, and each time it mangles it. I’ll probably move at some point, because getting formatting to work on this godawful pile of steaming shit is much more of a headache than it’s worth. May David Karp be violently anally sodomized with every interface design book known to man.

Verb prefixes:

The screeve prefix is usually (I’ve fucked it up many times before) glossed in the following order: finiteness, evidentiality, tense, mood, voice.

Finiteness:

• F,FINITE- finite (default)
• NF, NONF- nonfinite

Evidentiality:

• I, INFER - inferential evidentiality 
• S, SENS - sensory evidentiality (default)

Tense:

• RP- remote past
• NRP- non-remote past
• P, PRES - present (default)
• NRF- non-remote future
• RF- remote future

Mood:

• INDIC- indicative (default)
• C,COND- conditional
• O,OPT- optative

Voice:

• ACT- active (default)
• P,PASS- passive
• CA, CAUS - causative

There are a few other things that can be prefixed to the verb: manner markers and incorporators (haven’t got many of them yet, so I’ll hold off on describing them), the plural patient prefix (which doesn’t introduce any abbreviations not used elsewhere), and the question marker k-, glossedQ.

Verb suffixes:

This is just person marking. There are two forms of the person-marking gloss: one for first- and second-person affixes, and another for third-person affixes, which mark for noun class.

First- and second-person affixes are of the form #N.R, where # is the person, N is the number, and R is the role. Third-person affixes are of the form Cc.R.N, where C is the letter C,c is the class number (1 through 9), R is the role, and N is the number. In general, first- and second-person affixes use the one-letter form of the number (so1S.A, 2P.D,etc.), and third-person affixes use the two-letter form.

Person:

• 1- first person
• 2- second person

Number:

• S,SG- singular
• P,PL- plural

Role:

• A- agent
• P- patient
• D- as-yet unnamed third role (D for Dative, one of its main uses)

Noun prefixes:

The only noun prefix so far is the possessive prefix, used to show that the noun is owned or possessed (no, not in that sense!) by something else. That, like a verb class marker, is glossed two different ways: #N.POSS if the possessor is first- or second-person, and Cc.N.POSSif the possessor is third-person. (POSScan be abbreviated toP; no ambiguity will arise, since the other thing writtenPcan only occur on verbs.)

Noun suffixes:

The only noun suffix so far is the class/definiteness marker. This is writtenCc.D.N,where c is class, D is definiteness, and N is number. (Note that definite nouns do not mark for number, so number is only glossed on indefinites.)

Definiteness:

• D,DEF- definite
• I, IND, INDEF - indefinite

Kannow conditionals come in two forms: monoclausal forms, when the apodosis (then-statement) doesn’t introduce any core arguments not contained in the prothasis (if-statement), and polyclausal forms, when it does.

The monoclausal forms consist of one clause: a serialization of the main verb, put in the conditional, for the prothasis, and a nonfinite form of the result-verb, for the apodosis. For example:

“If you fuck a goat, it will not be happy.”
Ņeřbamatk kìlimt gòtoŗ.
[ŋ̟eʂˈbɑmɑθkʰ ˈkʷʰidɮimtʰ ˈgʷotʰər̥]
ņeř-bama-t-g kì-lim-t gòt-ŗ
RF.I.C-fuck-C2.P.SG-2S.A NF.NP.MID-be_happy-C2.P.SG goat-C2.IND.SG

Bolded in the gloss are the relevant parts: the conditional (C) main verb (note that conditionality is marked on the screeve prefix, which also encodes tense (remote future, the default tense for irrealis statements), evidentiality (inferential, for obvious reasons), and voice (active, unmarked), and the nonfinite (NF) form (todo: figure out why I’m calling them that) immediately following it, containing the apodosis, “it will not be happy”. Note that “goat” follows both verbs in the complex.

And an example of the polyclausal form:

“If you fuck a goat, your mother will not be happy.”
Ņeřbamatk gòtoŗ, ņeclimín tgèsow.
[ŋ̟eʂˈbɑmɑθkʰ ˈgʷotʰər̥ ŋ̟eçˈtɬʰimiy̯nʷ ˈθgʷesəʍ]
ņeř-bama-t-g gòt-ŗ, ņec-lim-nʷ tʷ-gès-ow    
RF.I.C-fuck-C2.P.SG-2S.A goat-C2.IND.SG, RF.I.M-be_happy-C2.P.SG 2S.POSS-mother-C2.DEF

Again, the conditional is bolded; also bolded is the screeve prefix of the second verb, which begins a clause of its own, hence the name of the construction.

Obviously there’s a bunch of weird TAM shit, but I can’t be assed with that right now. I haven’t even codified the rules for mediopassives and agent/patient/as-yet unnamed third person slot marking yet, so I’ll probably need to redo these examples someday.

b d ɟ g > w ð j ɣ / V_V
p t c k > b d ɟ g / V_V
j ɣ w > z j ɣ
ð > v (at this point, there’d probably be interference from a neighboring lang, since the coronal stops have to be alveolar for the next change)
b d ɟ g > v z j ɣ / V_V

So intervocalically, *p *d are reflected as /v/, *t *ɟ as /z/, *c *g as /j/, and *k *b as /ɣ/.

(Source: incatena.org)