I've been working on the final version of the proto-language for my Central Mountain family, and I'd like to offer the first part of that work here, for comment, before it's finalized.  Pre-Proto-Central Mountain is an internally reconstructed proto-language, taking the family back to its very beginning--or as close to that as can be attained; it is by design a fairly simple agglutinating language--because from it I derive all the complications and oddities of the later languages.  I knew I'd never be satisfied until I could derive all my CM languages from a primordial system with no vowel length and only the most basic sounds; and the result of that desire is Pre-Proto-Central Mountain.

First comes a description of the phonology and basic structure of Pre-Proto-Central Mountain, followed by an account of the sound changes that give rise to Proto-Central Mountain, the ordinary proto-language that gives birth to the later CM languages.  The sound change list is not yet numbered; in place of numbers are (x)s, which will be replaced when the list is deemed final.

--------------------------------------------------------------------------

§1. Central Mountain (CM) was a large family, perhaps largest among its early contemporaries, only a small part of which survived the First Earth.  Our reconstuctions are based, to begin with, on two of the best-known languages, Hlholammelo (Hlh) and Kapakwonak (K), the most ancient attested members of the family.   Comparison of data from more scantily recorded CM dialects has shown that these reconstructions will, in the main, fit all the known languages, and can accordingly be viewed as Proto-Central Mountain (PCM).  

An earlier stage, pre-PCM (pPCM), is investigated through methods of internal reconstruction applied to PCM.  While by no means a blind guess, pPCM may indeed reflect as much speculation as fact, perhaps glossing over (by necessity) substantial irregularities and variations in favor of idealized forms.  Elements and words from pPCM must be viewed with this caveat in mind.  Nevertheless, we regard many, if not most, pPCM forms as probable, while admitting that our internally reconstructed language as a whole is not likely definitive.

PPCM and PCM were alike polysynthetic, non-confugurational and incorporating.  That is to say, words could be made up of a great many component word-forming elements, whose relative order was strictly determined, and which were unintelligible and without meaning if taken out of proper context; while word order in the sentence was largely free.  PPCM was primarily agglutinating, with a few fusional elements; but PCM developed a higher degree of fusion.

A NOTE ON ASTERISKS

For sake of clarity, pPCM forms are marked with a double asterisk (**), while PCM forms bear a single asterisk.

SOUNDS

Transcription follows, in the main, Americanist notation, supplementing with IPA or ad hoc symbols only when the former is inadequate or unclear.  No distinction is drawn between phoneme and phone.

PPCM had five vowels, which apparently did not distinguish quantity: high front **ɩ; mid front **ɛ; low central **a; high back **ʊ; and mid back **ɔ.  The back vowels were rounded; the others were not.  The high and mid vowels were all lax.  

The pPCM simple consonants were unaspirated stops **p, **t, **k; fricatives **s, **h; and nasals **m, **n, **ŋ. The nasal consonants were voiced; the rest were voiceless, probably lenis.  The consonant **ŋ, although alien to the phonologies of the attested languages, is reconstructed based on the need for a segment that shares the nasality of *m and the velarness of *k, both of which are yielded by the segment.

PPCM syllable structure was strict, invariably CV, with a vowel always forming the syllabic peak.  No clusters, consonant or vowel, occurred.  There was a single level of stress, constituted by increased amplitude.  Monosyllabic words all received stress; and disyllabic words were stressed on the first syllable.   In words of three or more syllables, every other syllable was stressed, starting with the first, with the final syllable invariably receiving stress as well.  In polysyllabic words containing an even number of syllables, this resulted in two contiguous intra-word stressed syllables (the final two), a phenomenon which occurred nowhere else in the language.  

THE WORD

Aside from particles, which were uninflected, words in pPCM/PCM fell into a single morphological category whose members could function syntactically as either verbs or nominals.  The pPCM/PCM word had two "slots" or morphological positions, both of which were obligatory: a derivational stem, followed by an inflectional final suffix.  


THE ROOT

As pertains to pPCM at least, the root can be considered an abstract, unanalyzable semantic entity without inherent morphological class and of highly generalized meaning; and from the roots of pPCM we may say are derived the vast majority of pPCM's (and, inevitably, PCM's) actual word-forming elements.  Roots are semantically oriented toward action, motion, process, relationship, quality and quantity; they express nothing that is unambiguously nominal.  Without exception, they follow the pattern CV(CV), with disyllabic roots exhibiting a strict vowel harmony: the central vowel may appear with any other vowel in a disyllabic root, while front vowels may only appear with another front vowel and back vowels may only appear with another back vowel.  This vowel harmony suggests that, at an even earlier stage than pPCM, the disyllabic roots were biconsonantal with identical vowels, which were differentiated as new vocalics developed and as the language grew
 semantically.  The glimpse into the distant past of CM provided by the vowel harmony further suggests that pPCM, for all its phonetic strictures, grew from an even stricter primordial phonology. Most disyllabic roots have two different consonants; but a few contain identical consonants.  No root is longer than two syllables.

PSEUDO-ROOTS

PPCM pseudo-roots reflect archaic petrified derivational morphology from a very ancient (and perhaps primordial) derivational system that existed in CM prior to the derivational mechanics of pPCM.  Synchronically with respect to pPCM, pseudo-roots are as unanalyzable as roots proper; however, they were originally stems prior to pPCM, but through the loss of the productivity of their derivational infixes and processes they have become frozen forms.  In pPCM, they behave identically to true roots, possessing for the most part generalized meaning, and yielding their own word-forming elements; only through diachronic analysis are their relationships to certain true roots revealed.  While pseudo-roots follow the CV pattern of true roots, they do not always conform to the vowel harmony and may exceed two syllables.

A front-to-back and corresponding central/back-to-front vocalic ablaut, likely in origin a form of sound symbolism, yields pseudo-roots of augmentative and diminutive meanings.  Augmentative ablaut expresses strengthening of meaning, increase in size, greater intensity, the pejorative, worsening of meaning or perversion; diminutive ablaut expresses weakening of meaning, decrease in size, lesser intensity, endearment or bettering of meaning.  Either one can express extremes of meaning. Both literal and figurative augmentative and diminutive derivations are extant.  The ablaut occurs to applicable vowels thus:

Front-to-back (augmentative) ablaut:

**ɩ > **ʊ
**ɛ > **ɔ

Central/back-to-front (diminutive) ablaut:

**a > **ɛ
**ʊ > **ɩ
**ɔ > **ɛ

Another form of probable sound symbolism, involving consonant alternation, also derives augmentative and dimunutive pseudo-roots.  It follows the same general semantics as the vocalic ablaut but is less commonly observed.  The alternations occur to applicable consonants thus:

Fortition (augmentative):

**s > **t
**h > **k

Lenition (diminutive):

**t > **s
**k > **h

Whether the alternations **h ∼ **k and **k ∼ **h involved ʔ and x as intermediates cannot be determined.

Several derivational infixes, which follow the first vowel of a disyllabic root and appear as suffixes on monosyllabic roots, are also seen to yield pseudo-roots:

**-mɛ-: animizer (< root **mɛ 'spiritualness, spiritual power, sacredness, medicine, mysteriousness, animacy')

**-ŋɩ-: consequential (< root **ŋɩ 'do, make, cause, causalness, causality, create')

**-ŋɔ-: extensional (< root **ŋɔ 'sameness, similarity, likeness, commonality')

Due to the extreme remoteness and inaccessability of the prior derivational system in which these infixes were productive, a definitive (nor even sketchy) analysis of that system cannot be offered; only the general semantic effects of the infixes, as they were retained in pPCM at least, can be stated with any degree of accuracy.  As we might expect of the likely primordial times in which it coalesced, this prior derivational system appears to have been semantically much looser and free-form compared to the more consistent system that had replaced it by the time of pPCM; but little more can be said of it, as the aforementioned remnants of it in pPCM offer but a few vague clues. Moreover, pPCM itself appears to have undergone wholesale leveling with the replacement of the prior system by the new ("current") one, making further penetration into the earlier system difficult, if not impossble.

The animizer is an artifact of the introduction of the animacy distinction into CM, which by the time of pPCM was well-established.  Originally apparently a formal means of animizing inanimate initials (see §xx for a discussion of initial animacy and §xx for a related discussion of nominal animacy), it appears on a few pseudo-roots that survived into pPCM; but all these pseudo-roots have undergone significant semantic change, and it is these changes that likely account for their retention.  The consequential and extensional infixes appear more often.  The former derives a pseudo-root that is connected to a true root by means of causality: the pseudo-root owes its existence or substance to the root, or the pseudo-root is so because the root is; or, less typically, the true root is a consequence of the pseudo-root.  The latter derives a pseudo-root that is in some way associated with or is an extension of a true root, most typically resulting in a
 pseudo-root of greater specificity or specialization of meaning than the root.  The association or extension is literal in some cases, figurative in others.

No more than one infix appears on any single pseudo-root; but some pseudo-roots show both an infix and either the ablaut or consonant alternation.  In these cases, where ablaut is involved, the vowel(s) of the original roots and the vowel of the infixes alike undergo change, suggesting that the infixes had become unproductive, and their corresponding pseudo-roots frozen, before the ablaut was applied.

Save where the animizing infix is used, initials deriving from pseudo-roots almost always (except in a few rare cases) share the animacy class of the initials of the true roots from which they come.  Initials from pseudo-roots that undergo only vocalic ablaut or consonantal alternation also share the transitivity,  dynamic/stative class and agreement class of the initials of their associated true roots.  However, initials from pseudo-roots using infixes may not always share the transitivity, dynamic/stative class or agreement class of the initials of their associated true roots.  See §xx for a discussion of these initial distinctions.

In addition, there is a special class of pseudo-roots that are derived by compounding of roots and pseudo-roots; these are the pseudo-roots yielding the (transitive) times numbers 6-20 and the (dynamic) counting numbers 6-20; see §xx for a discussion of numbers in pPCM/PCM.  Though some of these are of considerable length compared to other pseudo-roots, they can rightly be judged as falling into the pseudo-root catgegory, as they yield initials, medials and qualitative/quantitative infixes just as other roots and pseudo-roots do.

By the time of pPCM, all of the foregoing infixes and processes giving rise to pseudo-roots had ceased to be productive; the infixes especially would seem to have been unproductive for some time by the era of pPCM.

MORPHOLOGICAL CLASSES

PPCM/PCM's word-forming elements divide into a number of distinct morphological classes. All pPCM word-forming elements save the incorporable nominals are ultimately ascribable to roots (or to roots via pseudo-roots); and multiple elements of differing morphological classes often all have their origins in the same root, and indeed in pPCM are phonetically identical to one another and to their roots of origin.   In PCM, however, many such elements have each developed their own unique phonetic forms from allomorphic variations.  Generally speaking, it is, in pPCM, morphological position (i.e., location in a word) that determines the morphological class of a given element, i.e., its exact meaning and function; we may say that morphological position transforms a pPCM root/pseudo-root into an actual word-forming element of a given class.

The basic morphological classes of pPCM/PCM word-forming elements are prefix, initial, morphological conjunction, medial suffix, final suffix, incorporable nominal affix, infix and objective-subjective suffixes.  Each class save the morphological conjunction divides further into sub-classes.

Prefixes come in three sub-types: the conjunct prefixes, the interrogative prefix  and the vocative prefix.

The initial serves as the core of a word, its broad semantic field or main idea, and occurs in three types: simple initials, complex productive initials and complex unproductive initials.  Simple initials are morphologically simple, phonetically equivalent to a bare root in pPCM.  Complex productive initials are productive constructions that are morphologically complex: they contain either a simple iniitial or a complex unproductive initial and at least one other productive derivational element.  Complex unproductive initials reflect the petrified derivational morphology of pseudo-roots, and are the equivalent of the simple initials of true roots; they behave morphologically in all ways like simple initials.

The morphological conjunction is used to compound multiple initials (see §xx for a discussion of this type of derivation).

Medial suffixes modify initials in various ways, and are phonetically identical to corresponding initials of related meaning in pPCM, and derive from the same roots/pseudo-roots as corresponding initials; and they divide similiarly to the initials, into simple medials, complex productive medials and complex unproductive medials.  The morphological characters of the medial sub-classes match those of similar initial sub-classes.  

Finals are inflectional and give pronominal information of the agent and/or patient; they divide into intransitive objective finals, intransitive subjective finals, transitive dual objective finals, transitive subjective-objective finals, and transitive dual subjective finals.  There is also an abstract final  that appears on certain intransitives and on some transitive passives.

Incorporable nominal affixes distinguish between incorporable participant nominals and incorporable non-participant nominals, the former encompassing agents and patients that are incorporated into a word and the latter all other nominals so incorporated.  Incorporable participant nominals appear as prefixes on the initial, while incorporable non-participant nominals show up as suffixes.

Infixes come in a number of varieties: those applicable to incorporable nominals, the qualitative and quantitative infixes and the possessive infixes; those applicable to initials, the detransitivizing infixes, the negation infix, the causative infix, the reflexive infix, the reciprocal infix and the nominal-related infixes; and those applicable to finals: the ordinal-aspectual-modal infixes and the relational-genitive infixes.  Possessive infixes are applicable to incorporable nominals and finals alike; and some initial infixes are also applicable to medials, though medials have no infixes that are particular to them.  Infixes appear after the first vowel of di- and polysyllabic elements, and as suffixes to monosyllabic ones.

Objective-subjective suffixes appear only in finals and divide into the objective suffix and the subjective suffix, both of which are used to alter the relationship of the agent and patient to the action.

(I'm omitting several sections here dealing with the exact details of stem and final derivation--these sections are very long and complicated, and will be posted separately later.)

PHONETIC DEVELOPMENT

Here the regular historical sound shifts that occurred from pPCM to PCM are sketched.  The changes are formulated and ordered primarily to produce correct forms, and secondarily with an eye toward phonetic naturalness.  The list is by intention chronological, the linearity of which is potentially misleading.  Where one change is not an obvious pre-condition for another, or must logically precede another, it is difficult or impossible to determine which change actually occurred first; in such instances, the list is constructed so as to keep related changes grouped together.  This practice must not be taken to suggest that an earlier listed development necessarily preceded later listed developments temporally, except in cases where those earlier developments had a discernable impact on later ones.  Moreover, no mention is made of the probability of some changes being simultaneous, an exceedingly common phenomenon in the phonological development of
 languages generally.


PPCM > PCM

(x) The dental coronals {t, s, n} palatalize before stressed and unstressed front vowels {ɩ, ɛ}; and after stressed {ɩ, ɛ}, they also progressively palatalize.

{t, s, n} > {tʸ, sʸ, nʸ} / _{ˈɩ, ˈɛ, ɩ, ɛ}
{t, s, n} > {tʸ, sʸ, nʸ} / {ˈɩ, ˈɛ}_


(x) Word-initially and before stressed and unstressed front vowels {ɩ, ɛ}, ŋ is lost. Before a stressed or unstressed rounded vowel {ʊ, ɔ} (other than word-initially), ŋ is assimilated to m; and before a stressed or unstressed central vowel a (other than word-initially), ŋ becomes g, which is then devoiced and merges with k.  With these changes, ŋ ceases to exist.

ŋ > ø / #_
ŋ > ø / _{ˈɩ, ˈɛ, ɩ, ɛ}
ŋ > m / _{ˈʊ, ˈɔ, ʊ, ɔ}
ŋ > k  / _{ˈa, a}


(x) With the consonant loss in change (x), compensatory lengthening rules first apply.  These rules become a persistent feature of CM and give rise to its three-way vowel length distinction, apparently without any pre-existing length from an independent source, which is ordinarily taken as a requirement for compensatory lengthening to occur.  The compensatory lengthening rules work to preserve overall length following the shortening or loss of sounds.  The rules can be summarized by assigning length values to various types of segments, and by providing for the shift of a segment's length value (in whole or in part) to another segment (typically to a vowel) when the segment is lost or shortened.  Relative to short (i.e., unlengthened) vowels, which equal 1, simple consonants have a length of 0.5; half-long vowels a length of 1.5; and long vowels a length of 2.  The vocalic length values correspond approximately to actual relative vowel durations--i.e., a
 half-long vowel is held about 1.5 times as long as a short vowel, and a long vowel about twice as long.  In this case, the first vowel preceding a lost ŋ will become half-long if it is short; or if this vowel is half-long, it will become long.  If there is no preceding vowel, the first vowel following a lost ŋ will become half-long if it is short, or long if it is half-long.  These processes occur from left to right across the word until all lost consonants are compensated for.


(x) It is likely, but not demonstratable, that half-long and long vowels in unstressed syllables briefly take on a weaker secondary stress, until the shift of stress patterns in change (x) below.  However, no trace of such secondary stress remains after the shift, and this secondary stress, if it existed, seems to have exerted no discernable phonetic influence.


(x) The persistent relationship between extended vowel length and tenseness among the high and mid vowels first appears.  Half-long and long {ɩ, ɛ, ʊ, ɔ} become tense.  Hereafter, half-long and long high and mid vowels will always be realized as tense, while their short counterparts continue to always be realized as lax.

{ɩˑ, ɛˑ, ʊˑ, ɔˑ, ɩː, ɛː, ʊː, ɔː} > {iˑ, eˑ, uˑ, oˑ, iː, eː, uː, oː}


(x) Unstressed short vowels {ɩ, ɛ, a, ʊ, ɔ} are syncopated.

V > ø


(x) Compensatory lengthening applies with the vowel loss in change (x).  The first vowel preceding a syncopated vowel will become long if it is short; if the first preceding vowel is non-short, the first vowel following a syncopated vowel will become long if it is short; if the first following vowel is non-short, the first vowel preceding a syncopated vowel will become long if it is half-long, and if it is not, the first vowel following a syncopated vowel will become long if it is half-long.   In cases where there is no following vowel and this rule calls for lengthening of the first following vowel, that part of the rule is ignored.  The length of any syncopated vowel that cannot be compensated for (in whole or in part) by the foregoing is simply lost. These processes occur from left to right across the word until all syncopated vowels are compensated for, to the degree that they can be.


(x) Syllable length sandhi rules first apply.  Like compensatory lengthening, these become a persistent feature of CM, but they are ranked above compensatory lengthening and prevent it (and other changes) from occuring when the rules would be violated.  The rules prohibit non-short syllables from occuring consecutively in a word, where the length of a syllable is determined by the length of its vowel; thus, a short syllable must always intervene between two non-short syllables.  The sandhi processes can be summarized as follows, using the length values of segments as discussed above: sequences of two or more consecutive non-short syllables in a word are divided into groups of two non-short syllables each, from left to right across a word, with no syllable belonging to more than one group, and with any leftover non-short syllable momentarily ignored with respect to these disyllable groups.  The first syllable in the first (leftmost) disyllable group loses
 one or more increments of its length, such that it becomes short, and the lost length is then shifted to an adjoining syllable--preferably, to the preceding syllable, if there is one, or to the following syllable, if there is not, or if the preceding syllable cannot receive the shifted length (for either of the reasons described hereafter) or can receive only part of the shifted length while the following syllable can receive all of it; shift of length will not occur, however, if it results in a sequence of two consecutive non-short syllables, and/or if the syllable that would receive the shifted length is already long.  Short and half-long syllables may receive only enough shifted length to make them long, and they may receive the entire quantity of shifted length, if able, or only part of it; and any length that cannot be shifted as described above is simply lost.  Once the first disyllable group is thus altered, these processes are repeated with the
 second disyllable group (if there is one), and so on, from left to right across the word.  If, after all groups are altered,  there remains any sequence of two consecutive non-short syllabes, these syllables are treated as their own distinct group, irrespective of and separate from the other previous groups, and are subjected to the foregoing processes.


(x) Half-long and long high and mid vowels become tense; short high and mid vowels become lax.

{ɩˑ, ɛˑ, ʊˑ, ɔˑ, ɩː, ɛː, ʊː, ɔː} > {iˑ, eˑ, uˑ, oˑ, iː, eː, uː, oː}
{i, e, u, o} > {ɩ, ɛ, ʊ, ɔ}


(x) Stress patterns shift, perhaps in response to some of the previous changes.  Monosyllabic words no longer receive stress; and disyllabic words become stressed on the second syllable.  In words of three or more syllables, every other syllable is still stressed, but starting with the second.  In all syllabic lengths of word, half-long and long syllables are all likewise stressed, and the syllable following a half-long or long syllable is always treated like the beginning of a new word--i.e., the stress count starts over, so that the syllable immediately following a half-long or long syllable is left unstressed, the syllable after that is stressed, and so on. In words of three or more syllables, final syllables are stressed only if they receive stress by the foregoing patterns. These stress rules become a persistent feature, but they are ranked below both the syllable length sandhi and compensatory lengthening rules and may be violated by either of
 them, or by certain other changes; but they are consistently reapplied following violation by any change.


(x) VC monosyllables metathesize to CV.

#{V, Vˑ, Vː}C# > #C{V, Vˑ, Vː}#


(x) Word-final consonants drop.  Because the monosyllables in which this change occurs are already long, compensatory lengthening does not occur, and consequently the length of these dropped consonants is simply lost.

C > ø / _#



(x) Consonant clusters of stop + nasal and nasal + stop undergo assimilation.  Where palatalized consonants are assimilated to sounds that do not possess palatalized varieties, they lose palatalization.

pn > tn
pnʸ > tnʸ
tm > pm
km > pm
kn > tn
knʸ > tnʸ
mt > nt
mk > nk
mtʸ > ntʸ
np > mp
tʸm > pm
nʸp > mp



(x) Consonant clusters of stop + nasal and {s, sʸ} + nasal undergo metathesis.

pm > mp
tn > nt
tnʸ > nʸt
sm > ms
sn > ns
snʸ > nʸs
tʸn > ntʸ
tʸnʸ > nʸtʸ
sʸm > msʸ
sʸn > nsʸ
sʸnʸ > nʸsʸ




(x) Assimilation occurs in consonant clusters with palatalized consonants: {t, s, n} become palatalized when they precede a palatalized consonant, while {tʸ, sʸ, nʸ} are depalatalized when they precede a non-palatalized consonant.

{t, s, n} > {tʸ, sʸ, nʸ} / _{tʸ, sʸ, nʸ}
{tʸ, sʸ, nʸ} > {t, s, n} / _C(-ʸ)




(x) Dental stop plus dental fricative consonant clusters become affricates.

ts > c
tʸsʸ > cʸ




(x) The cluster hh reduces to h.

h > ø / h_



(x) Compensatory lengthening applies with the loss of h in change (x) above, though its application is constrained by the syllable length sandhi rules.  If the first vowel preceding a lost h is half-long, the vowel becomes long; if it is not, but the first vowel after a lost h is half-long, that vowel will become long instead.  This applies from left to right across the word wherever possible; and the length of a dropped h that cannot be compensated for in this way is simply lost.




(x) Geminate consonant clusters become long consonants (approximately twice the length of a normal consonant).

pp > pː
tt > tː
kk > kː
ss > sː
mm > mː
nn > nː
tʸtʸ > tʸː
sʸsʸ > sʸː
nʸnʸ > nʸː




(x) A glottal stop replaces the extended duration of the long consonants.

pː > pʔ
tː > tʔ
kː > kʔ
sː > sʔ
mː > mʔ
nː > nʔ
tʸː > tʸʔ
sʸː > sʸʔ
nʸː > nʸʔ




(x) ʔ merges with h as h.

ʔ > h



(x) Clusters of stop + h become aspirates.

ph > pʰ
th > tʰ
kh > kʰ
tʸh > tʸʰ




(x) h drops after another consonant.

h > ø / C_




(x) Compensatory lengthening applies with the loss of h in change (x) above, though its application is constrained by the syllable length sandhi rules.  If the first vowel preceding a lost h is half-long, the vowel becomes long; if it is not, but the first vowel after a lost h is half-long, that vowel will become long instead.  This applies from left to right across the word wherever possible; and the length of a dropped h that cannot be compensated for in this way is simply lost.





(x) Short unstressed front and central vowels are assimilated between velar and glottal consonants.

ɩ > ʊ / {h, k, kʰ}_h
ɛ > ɔ / {h, k, kʰ}_h
a > ɔ / {h, k, kʰ}_h
ɩ > ɛ / {h, k, kʰ}_{k, kʰ}
ɛ > a / {h, k, kʰ}_{k, kʰ}




(x) Short unstressed front vowels are assimilated between bilabial consonants.

ɩ > ʊ / {p, pʰ, m}_{p, pʰ, m}
ɛ > ɔ / {p, pʰ, m}_{p, pʰ, m}




(x) Word-initial and intervocal h that is onset in unstressed syllables is lost.

h > ø / #_V
h > ø / {ˈV, ˈVˑ, ˈVː}_V



(x) Compensatory lengthening applies with the loss of h in change (x) above, though its application is constrained by the syllable length sandhi rules.  If the first vowel preceding a lost h is half-long, the vowel becomes long; if it is not (or if the lost h was word-initial) but the first vowel after the lost h is half-long, that vowel will become long instead.  This applies from left to right across the word wherever possible.  In unstressed monosyllables where h is lost, a short vowel following a lost h lengthens to half-long. The length of a dropped h that cannot be compensated for in these ways is simply lost.



(x) High and mid short vowels that became half-long in change (x) above become tense.

{ɩˑ, ɛˑ, ʊˑ, ɔˑ} > {iˑ, eˑ, uˑ, oˑ}


(x) A previously unstressed monosyllable whose vowel lengthened to half-long to compensate for a lost h in change (x) above becomes stressed.





(x) In words of three or more syllables, sequences of long syllable + short syllable + long syllable become long syllable + short syllable + half-long syllable.  Because any shifting of this lost length to a syllable adjoining the half-long syllable would create a sequence of two consecutive non-short syllables, the lost length is not compensated for and is simply lost.  This change occurs from left to right across a word until all such syllabic sequences are altered.





(x) Long vowels become half-long before a consonant cluster. Because any shifting of this lost length to a syllable adjoining the half-long vowel would create a sequence of two consecutive non-short syllables, the lost length is not compensated for and is simply lost.

ˈVː > ˈVˑ / _CC





(x) Half-long vowels become short before a consonant cluster.

ˈVˑ > ˈV / _CC





(x) Compensatory lengthening applies with the shortening of half-long vowels in change (x) above.  The compensatory lengthening here is doubly constrained: by both the syllable length sandhi rules and by the prohibition on non-short vowels before consonant clusters.  The first vowel preceding a shortened vowel will become half-long if it is short, provided that that first preceding vowel is not before a consonant cluster and that making it half-long does not result in a sequence of two consecutive non-short syllables; if the first preceding vowel is half-long, it will become long.  If there is no preceding vowel, or if it is before a consonant cluster, or if lengthening it results in a sequence of two consecutive non-short syllables, the first vowel following a shortened vowel will become half-long if it is short, provided that that first following vowel is not before a consonant cluster and that making it half-long does not result in a sequence of two
 consecutive non-short syllables; if the first following vowel is half-long, it will become long.  Any lost length that cannot be compensated for as described above is simply lost.  These changes occur from left to right across a word until all compensatable shortened vowels are compensated for.





(x) Half-long and long high and mid vowels become tense; short high and mid vowels become lax.

{ɩˑ, ɛˑ, ʊˑ, ɔˑ, ɩː, ɛː, ʊː, ɔː} > {iˑ, eˑ, uˑ, oˑ, iː, eː, uː, oː}
{i, e, u, o} > {ɩ, ɛ, ʊ, ɔ}




(X) Stress is redistributed according to the rules in change (x).




(x) Assimilation occurs in clusters of vowels in hiatus: a short stressed or unstressed vowel is assimilated by a following vowel.  This assimilation occurs from left to right across a word.  Short vowels assimilate as follows:

{ˈa, a} {ˈɩ, ɩ, iˑ, iː} > {ˈɛ , ɛ} {ˈɩ, ɩ, iˑ, iː}
{ˈa, a} {ˈʊ, ʊ, uˑ, uː} > {ˈɔ, ɔ} {ˈʊ, ʊ, uˑ, uː}
{ˈɛ , ɛ} {ˈʊ, ʊ, uˑ, uː} > {ˈɩ, ɩ} {ˈʊ, ʊ, uˑ, uː}
{ˈɩ, ɩ} {ˈa, a, aˑ, aː} > {ˈɛ , ɛ} {ˈa, a, aˑ, aː}
{ˈɩ, ɩ} {ˈɔ, ɔ, oˑ, oː} > {ˈɛ , ɛ} {ˈɔ, ɔ, oˑ, oː}
{ˈɔ, ɔ} {ˈɩ, ɩ, iˑ, iː} > {ˈʊ, ʊ} {ˈɩ, ɩ, iˑ, iː}
{ˈʊ, ʊ} {ˈa, a, aˑ, aː} > {ˈɔ, ɔ} {ˈa, a, aˑ, aː}
{ˈʊ, ʊ} {ˈɛ, ɛ, eˑ, eː} > {ˈɔ, ɔ} {ˈɛ, ɛ, eˑ, eː}







(x) Unstressed short high vowels {ɩ, ʊ} become non-syllabic glides {y, w} before other vowels.  If a resulting y appears directly after a palatalized consonant, a short ɩ is inserted between the consonant and the y, as part of the change.  Then, if no such short ɩ is inserted and the contraction of {ɩ, ʊ} to {y, w} results in a sequence of two consecutive non-short syllables, a short ɩ is inserted after y and a short ʊ after w, as part of the change. Where this series of changes occurs more than once per word, it does so from left to right across the word.






(x) Compensatory lengthening occurs with the contraction of the high vowels into glides in change (x) above, but only if no short vowel was inserted as part of the change.  The compensatory lengthening here is doubly constrained: by both the syllable length sandhi rules and by the prohibition on non-short vowels before consonant clusters, where glides are not considered consonants.  The first vowel preceding a glide will become half-long if it is short, provided that that first preceding vowel is not before a consonant cluster and that making it half-long does not result in a sequence of two consecutive non-short syllables; if the first preceding vowel is half-long, it will become long.  If there is no preceding vowel, or if it is before a consonant cluster, or if lengthening it results in a sequence of two consecutive non-short syllables, the first vowel following a glide will become half-long if it is short, provided that that first following vowel is
 not before a consonant cluster and that making it half-long does not result in a sequence of two consecutive non-short syllables; if the first following vowel is half-long, it will become long.  Any lost length that cannot be compensated for as described above is simply lost.  These changes occur from left to right across a word until all compensatable lost length is compensated for.






(x) Half-long and long high and mid vowels become tense.

{ɩˑ, ɛˑ, ʊˑ, ɔˑ, ɩː, ɛː, ʊː, ɔː} > {iˑ, eˑ, uˑ, oˑ, iː, eː, uː, oː}





(X) Stress is redistributed according to the rules in change (x).







(x) Unstressed short high vowels {ɩ, ʊ} become non-syllabic glides {y, w} after other vowels, save word-finally.  If the contraction of {ɩ, ʊ} to {y, w} results in a sequence of two consecutive non-short syllables, a short ɩ is inserted after y and a short ʊ after w, as part of the change.   Where this series of changes occurs more than once per word, it does so from left to right across the word.






(x) Compensatory lengthening occurs with the contraction of the high vowels into glides in change (x) above, but only if no short vowel was inserted as part of the change.  The compensatory lengthening here is doubly constrained: by both the syllable length sandhi rules and by the prohibition on non-short vowels before consonant clusters, where glides are not considered consonants.  The first vowel preceding a glide will become half-long if it is short, provided that that first preceding vowel is not before a consonant cluster and that making it half-long does not result in a sequence of two consecutive non-short syllables; if the first preceding vowel is half-long, it will become long.  If there is no preceding vowel, or if it is before a consonant cluster, or if lengthening it results in a sequence of two consecutive non-short syllables, the first vowel following a glide will become half-long if it is short, provided that that first following vowel is
 not before a consonant cluster and that making it half-long does not result in a sequence of two consecutive non-short syllables; if the first following vowel is half-long, it will become long.  Any lost length that cannot be compensated for as described above is simply lost.  These changes occur from left to right across a word until all compensatable lost length is compensated for.






(x) Half-long and long high and mid vowels become tense.

{ɩˑ, ɛˑ, ʊˑ, ɔˑ, ɩː, ɛː, ʊː, ɔː} > {iˑ, eˑ, uˑ, oˑ, iː, eː, uː, oː}





(X) Stress is redistributed according to the rules in change (x).






(x) Remaining clusters of two qualtitatively identical short stressed or unstressed vowels in hiatus will become a long vowel, regardless of where the cluster occurs, unless this results in a sequence of two consecutive non-short syllables.  This change may occur more than once per word, and when it does, it occurs from left to right across the word.

ɩɩ > ɩː
ɛɛ > ɛː
ʊʊ > ʊː
ɔɔ > ɔː
aa > aː







(x) Long high and mid vowels become tense.

{ɩː, ɛː, ʊː, ɔː} > {iː, eː, uː, oː}





(X) Stress is redistributed according to the rules in change (x).









(x) In remaining clusters of two qualitatively identical or nearly-identical stressed or unstressed vowels in hiatus, a glide is inserted between the vowels, regardless of where the cluster occurs.  The quality of this glide is determined by the vowels in the cluster: a y is inserted when the vowels are front, and a w inserted when the vowels are non-front.  This change may occur more than once per word, and when it does, it occurs from left to right across the word.

{a, aˑ, aː} {a, aˑ, aː} > {a, aˑ, aː}w{a, aˑ, aː}
{ɛ, eˑ, eː} {ɛ, eˑ, eː} > {ɛ, eˑ, eː}y{ɛ, eˑ, eː}
{ɩ , iˑ, iː} {ɩ , iˑ, iː} > {ɩ , iˑ, iː}y{ɩ , iˑ, iː}
{ɔ, oˑ, oː} {ɔ, oˑ, oː} > {ɔ, oˑ, oː}w{ɔ, oˑ, oː}
{ʊ, uˑ, uː} {ʊ, uˑ, uː} > {ʊ, uˑ, uː}w{ʊ, uˑ, uː}









(x) In a remaining cluster of three or four stressed or unstressed vowels in hiatus, a glide is inserted after the second vowel in the cluster.  The quality of this glide is determined by the second vowel in the cluster: a y is inserted when the second vowel is front, and a w inserted when the second vowel is non-front.

{V, Vˑ, Vː} {a, aˑ, aː} {V, Vˑ, Vː} ( {V, Vˑ, Vː} ) > {V, Vˑ, Vː} {a, aˑ, aː}w{V, Vˑ, Vː} ( {V, Vˑ, Vː} )
{V, Vˑ, Vː} {ɛ, eˑ, eː} {V, Vˑ, Vː} ( {V, Vˑ, Vː} ) > {V, Vˑ, Vː} {ɛ, eˑ, eː}y{V, Vˑ, Vː} ( {V, Vˑ, Vː} )
{V, Vˑ, Vː} {ɩ , iˑ, iː} {V, Vˑ, Vː} ( {V, Vˑ, Vː} ) > {V, Vˑ, Vː} {ɩ , iˑ, iː}y{V, Vˑ, Vː} ( {V, Vˑ, Vː} )
{V, Vˑ, Vː} {ɔ, oˑ, oː} {V, Vˑ, Vː} ( {V, Vˑ, Vː} ) > {V, Vˑ, Vː} {ɔ, oˑ, oː}w{V, Vˑ, Vː} ( {V, Vˑ, Vː} )
{V, Vˑ, Vː} {ʊ, uˑ, uː} {V, Vˑ, Vː} ( {V, Vˑ, Vː} ) > {V, Vˑ, Vː} {ʊ, uˑ, uː}w{V, Vˑ, Vː} ( {V, Vˑ, Vː} )







(x) A glide is inserted before a stressed or unstressed high vowel word-initially.

{ɩ , iˑ, iː} > y{ɩ , iˑ, iː} / #_
{ʊ, uˑ, uː} > w{ʊ, uˑ, uː} / #_





(x) Between the two stressed or unstressed vowels of remaining word-initial and word-final clusters of vowels in hiatus, a glide is inserted.  The quality of this glide is determined by the first vowel in the cluster: a y is inserted when the first vowel is front, and a w inserted when the first vowel is non-front.

{a, aˑ, aː} {V, Vˑ, Vː} > {a, aˑ, aː}w{V, Vˑ, Vː}  / #_
{ɛ, eˑ, eː} {V, Vˑ, Vː} > {ɛ, eˑ, eː}y{V, Vˑ, Vː} / #_
{ɔ, oˑ, oː} {V, Vˑ, Vː} > {ɔ, oˑ, oː}w{V, Vˑ, Vː} / #_

{a, aˑ, aː} {V, Vˑ, Vː} > {a, aˑ, aː}w{V, Vˑ, Vː}  / _#
{ɛ, eˑ, eː} {V, Vˑ, Vː} > {ɛ, eˑ, eː}y{V, Vˑ, Vː} / _#
{ɔ, oˑ, oː} {V, Vˑ, Vː} > {ɔ, oˑ, oː}w{V, Vˑ, Vː} / _#
{ɩ , iˑ, iː} {V, Vˑ, Vː} > {ɩ , iˑ, iː}y{V, Vˑ, Vː} / _#
{ʊ, uˑ, uː} {V, Vˑ, Vː} > {ʊ, uˑ, uː}w{V, Vˑ, Vː} / _#




(x) The remaining high vowels are lowered to mid, regardless of length or stress.  With this change, CM assumes the basic tri-vowel system that it will long retain.

{ɩ , iˑ, iː} > {ɛ, eˑ, eː}
{ʊ, uˑ, uː} > {ɔ, oˑ, oː}





(x) In clusters of two qualitatively identical or nearly-identical stressed or unstressed vowels in hiatus created by change (x) above, the short vowel of the cluster dissimilates; if both vowels are short, the second dissimilates.

ɛɛ > ɛa
ɔɔ > ɔa

ɛ{eˑ, eː} > a{eˑ, eː}
ɔ{oˑ, oː} > a{oˑ, oː}

{eˑ, eː}ɛ > {eˑ, eː}a
{oˑ, oː}ɔ > {oˑ, oː}a





(x) Aspirated stops lose their aspiration before {y, w}.

Cʰ > C / _{y, w}






(x) Stressed and unstressed ɛ becomes a in the initial syllable of a word (or in the only syllable of a monosyllable).

{ˈɛ, ɛ} > {ˈa, a} / #(C)(C)_






(x) Sequences of w + nasal undergo metathesis before a vowel.

wm > mw / _V
wn > nw / _V
wnʸ > nʸw / _V






(x) Nasals are assimilated before w word-initially or after a vowel, glide or h.

#nw > #mw
#nʸw > #mw

{V, w, y, h}nw > {V, w, y, h}mw
{V, w, y, h}nʸw > {V, w, y, h}mw






(x) Long word-final vowels become half-long in words of two or more syllables. Because any shifting of this lost length to a preceding short syllable would create a sequence of two consecutive non-short syllables, the lost length is not compensated for and is simply lost.

ˈVː > ˈVˑ / V(C)(C)(C)(C)_#





(x) Half-long word-final vowels become short in words of two or more syllables.

ˈVˑ > ˈV / V(C)(C)(C)(C)_#





(x) Compensatory lengthening occurs with the shortening of word-final vowels in change (x) above. The compensatory lengthening here is doubly constrained: by both the syllable length sandhi rules and by the prohibition on non-short vowels before consonant clusters, where glides are not considered consonants.  The first vowel preceding a shortened word-final vowel will be short; and this vowel will become half-long to compensate for the lost length if its lengthening does not result in a sequence of two consecutive non-short syllables and if it does not precede a consonant cluster.  If either or both of these conditions hold, however, the lost length of a shortened word-final vowel is simply lost.






(x) Half-long mid vowels become tense; short mid vowels become lax.

{ɛˑ, ɔˑ} > {eˑ, oˑ}
{e, o} > {ɛ, ɔ}





(X) Stress is redistributed according to the rules in change (x).