Proto-Gäj language

Proto-Gäj (abbreviated PGäj; also called Proto-Gajic or Proto-Arklobu) is the reconstructed ancestor of the "Gäj" branch of the Koyanic languages.

Vowels
The contrast between long and short vowels was present only in stressed syllables.

The open-mid vowels *ɛ *ɛ̄ *ɔ *ɔ̄ are a hypothetical addition to the Proto-Gäj phonological system which may help to explain certain reflexes in the earliest diverging Gäj language, Zangahan. Proto-Gäj is usually considered with these vowels, since their existence is based mostly on Proto-Koyanic, having apparently merged with the open vowels *æ *ǣ *a *ā respectively in all other Gäj languages.

The long close vowels *ī *ū are thought to have had centralising diphthong qualities [i̯ɨ u̯ʉ] respectively, at least by some late stage of Proto-Gäj.

Evolution
The earliest reconstructible stage of the Gäj branch of the Koyanic languages is Proto-Gäj. This stage of the language had already made several significant changes to the phonology, which can be divided into four roughly chronological stages:


 * Obstruent shifts, including the loss of uvulars and development of voiced plosives and a new set of fricatives.
 * Laryngeal and vowel shifts, primarily laryngeal colouring.
 * Continuant shifts, primarily the loss of voiceless sonorants.
 * Monophthongisation, resulting in the loss of glides in coda clusters.

These four stages are composed of several similar, related changes with no reconstructible internal relative chronology, thus each represents a broad shift in the phonological system.

The following sound changes are reconstructed to have occurred, roughly in the order presented, in the development of Proto-Gäj from its ancestor Proto-Koyanic. In each example given, the start (PKn) and end point (PGäj) for each word is shown, with relevant intermediate steps in bold.

Alternative view of voice alternation
A kind of alternation between voiced and voiceless plosives similar to that shown above is believed to have sometimes affected Proto-Koyanic *z. This can be seen reflected in Proto-Gäj, e.g. in *ākur- "steppe":


 * *ā́qsx₃z- > *ākur- (Note that rhotacisation of *z to *r is a later change)
 * *ā́qsx₃s-B > *āku-B

Some other cases of *z appear uniquely devoiced in Proto-Gäj, e.g. *nǣp- "plain" suggesting ancestral *hn̥x₁ə́spa, where other Koyanic branches reflect *hn̥x₁ə́zpa.

This alternation is difficult to reconcile with the chronology of the sound changes presented above, since if these situations were in fact originally *s which then underwent voicing to *z at the same time as the plosives, this must have occurred before the debuccalisation of *s to *h. This, however, cannot have happened, since ancestral PKn *s shows no such voicing, e.g. in *mē- "liver":


 * *hm̥r̥ə́x₂s- > *mē- (not **mer- as would be expected if *s voiced to *z)
 * *hm̥r̥ə́x₂s-B > *mē-B

The novel reconstruction of Proto-Koyanic suggests that plosives may have already contrasted voice, with aspiration a later development in some branches, specifically noting that it is not necessary to reconstruct aspiration in an ancestor of Proto-Gäj. Additionally, *s and *z are reinterpreted as plosives, *tʲ and *dʲ. Working under this assumption in Gäj, unvoiced plosives would first shift to fricatives (e.g. *tʲ > *θʲ), then voiced plosives would undergo devoicing in certain environments (*dʲ > *tʲ), then all other changes, including the shift *dʲ *tʲ *θʲ > *z *s *s > *r *h *h, occurring at a later time. This alternative view yields identical results in Proto-Gäj:

The case of *nǣp- < *hn̥x₁ə́zpa is considered a regular change where two adjacent voiced plosives were devoiced:
 * *ā́ɢtʲxudʲ- > *ā́gθʲxudʲ- > *ā́kθʲxudʲ- > *ā́ksxuz- > *ākur-
 * *ā́ɢtʲxudʲ-B > *ā́gθʲxudʲ-B > *ā́kθʲxutʲ-B > *ā́ksxus-B > *āxu-B


 * *hnxédʲba > *hnxétʲpa- > *hnxéspa- > *nǣp-

The lack of voicing of PKn *s, NPKn *tʲ, is also naturally resolved:


 * *hmlʲǽtʲ- > *hmlʲǽθʲ- > *hmlʲǽs- > *mē-
 * *hmlʲǽtʲ-B > *hmlʲǽθʲ-B > *hmlʲǽs-B > *mē-B

Laryngeal and vowel shifts
Internal reconstructions of early stages of the Zangahan language have led to the theory that *x₁ did not merge with *x₂ in Gäj, and that x₁ə x₁ə̄ and (x₃)ə (x₃)ə̄ in fact resulted in open-mid vowels ɛ ɛ̄ and ɔ ɔ̄ respectively, which then merge into *æ *ǣ *a *ā in all other Gäj languages. Zangahan is poorly attested, and as such these vowels cannot verifiably be reconstructed in Proto-Gäj, so they are usually not considered.

Continuant shifts
The main change to the continuant inventory is the loss of contrastive voicelessness through various shifts, including the loss of *h.

Monophthongisation
Monophthongisation occurred very late in the history of Proto-Gäj, with the earliest diverging Gäj languages still reflecting some evidence of the original offglides. If Proto-Gäj did indeed have the open-mid vowels *ɛ *ɛ̄ *ɔ *ɔ̄ as one theory suggests, they would have merged into *æ *ǣ *a *ā respectively around the time of this monophthongisation.

Stem mutation
A key characteristic of Gäj morphology is stem mutation, where the underlying form of a noun or verb stem is mutated to a slightly different form in certain inflections. The two forms of the stem are known as basic and mutated.

Basic stem
The basic stem is used in the perfective and irrealis verb conjugations and the ergative, absolutive, accusative and dative nominal cases. This is the unmarked form of the stem which may take any vowel and any final consonant. When vowel-initial suffixes are attached to vowel-final basic stems, *-h- is commonly inserted between the stem and the suffix.

Mutated stem
The mutated stem is used in the imperfective verb conjugation and the possessive, genitive and ablative nominal cases.

In the mutated stem, any stem-final obstruent is regularly devoiced:


 * *-b *-d *-j *-g > *-p *-t *-c *-k

Additionally, *-r is deleted in the mutated stem, because the vast majority of stem-final *-r originated from *-z which then rhotacised to *-r after *-z > *-s devoicing and subsequent loss of *-s occurred. This regular alternation was so common that it was analogically applied to forms which originally had no sibilant, such as PKn *phə̄́x₃r "voice" reflected by the PGäj stems *fōr-/*fō-.

Some stems feature one of four underlying diphthongs, *-iy, *-uw, *-ey and *-ow, which lose the glide in the mutated form, for example, *gey- "to rest" has the mutated stem form *ge-. This alternation is the result of late Gäj monophthongisation, not reflected in the Zangahan language.

The Namb languages went on to further morphologically separate the two stem forms via ablaut in the mutated stem. For more information, see here.

The semantic association between the cases was strengthened in many daughter languages, in which the two stem forms of many words became separated not only grammatically, but lexically, undergoing different semantic shifts and resulting in separate words.

Some examples of consonant-final and vowel-final verb stems are shown below.

Nouns
Proto-Gäj sees a development of the animacy system, splitting Proto-Koyanic's three inflectional paradigms (animate singular, animate plural, inanimate) into four:


 * Countable singular
 * Countable plural
 * Uncountable
 * Locative

The countable noun class generally contains all objects with a discrete, countable form. Human referents generally did not mark the ergative, instead using a nominative-accusative alignment with the absolutive case taking the nominative role, resulting in a split ergative alignment. The possessive case is also distinguished from the genitive for humans only.

Uncountable nouns had no grammatical number and used a simple system of ergative-absolutive alignment with additional dative and genitive cases. This category of nouns generally consisted of things with no discrete form, such as substances, colours, sensations, and weather events, but also includes group entities such as peoples.

Locatives end in *-θu and have two additional marked cases for lative (motion towards) and ablative (motion away from). This category includes proper nouns, which in Gäj are not personal names, but rather names of places, celestial objects, months, time periods etc. The lative suffix *-xu derives from Proto-Koyanic *-ukʰū which allowed voicing of a preceding plain obstruent, thus attaches to the basic stem.

Verbs
Verbs had three conjugation paradigms: perfective and irrealis, which used the basic stem, and the imperfective with the mutated stem.

Each conjugation has a set of suffixes which mark the person of the subject and a set of prefixes and infixes which mark tense, aspect and mood information and show agreement with the plurality of the subject. The combination of suffixes and prefixes allows all six person and number combinations to be distinguished.

The imperfective suffixes appear to be descended from PKn infinitive series {*-kʰi, *-kʰīkʰ, *-ʡī} with an unknown suffix *-r and irregular infixing on the second person suffix *-xi⟨r⟩x. This unusual development represents the only finite verb forms developed from the PKn infinitives present in the Koyanic language family.

The irrealis suffixes are the reflexes of the reconstructed suffixes of the PKn conditional series, reflecting only the singular person forms {*-(ə)ncʰūʡ, *-(ə)ntx₂əyu, *-(ə)nqtx₂əʡ} and dropping the plurals.

Perfective conjugation
The prefixes of the perfective conjugation are single syllables composed of two obligatory parts, an initial consonant which carries tense and person information, and a rhyme which may further specify the person.