User:Alexlin01/NH cognates/Phonology

All are tentative.

Proto-North Halmahera
1 This is Voorhoeve's *S, but I personally tend toward a voiced phoneme.

Reflexes
All are tentative and subject to significant revision.

1 Lost phonemic status without having been lost phonetically. 2 Should not be too old of a stage; may even be able to call it "Old Tobelo" if any historical attestations exist.

Modole

 * 1) Glottal-Stop Insertion Rule (in PNeNH)
 * 2) Loss of all glottal stops (🇨🇬, 🇨🇬)
 * 3) *k → ʔ (fairly recent, even Ternate loans have ʔ in place of k)

Tabaru

 * 1) *c → t
 * 2) *l → Ø (🇨🇬 from 🇨🇬)

Sahuic

 * 1) *k → ʔ
 * 2) *p1 → w
 * 3) * 🇨🇬, 🇨🇬, and 🇨🇬 (all borrowed - inherited terms shouldn't have /k/) all compare with Proto-Sangiric *kapuna ("dog")

*p1 vs. *p2
The phoneme /*p1/ may have been [pʷ] and /*p2/ may have been [p]. Evidence for /*p1/ = [pʷ]:

Alternations between *p1 and *p2
Note the alternations (with similar, if not identical, meanings) between the pair *'p1a.lus and *'p2a.lus and the pair *'p1o.kor and *'p2o.kor. Also, a number of languages have an irregular development of *'p1 to p or *'p2 to f/h/w.

*Z

 * 1) Consonant temporarily denoted /*Z/ (since I can only think of something like /ʐ/ or /ʂ/ that could give rise to these), with the consonant correspondences:
 * 2) * Galela /s/, Tobelo /h/, Tabaru /s/, Modole /h/; Loloda /d͡ʒ/, Pagu /l/, Sahu /r/; Ternate /r/, Tidore /r/, W. Makian /d/
 * 3) * PGL: /*Z/
 * 4) ** Pre-Galela: merged with /*s/
 * 5) ** Pre-Loloda: merged with /*ḋ/
 * 6) *** Loloda: /d͡ʒ/
 * 7) * PSahuic: merged with /*r/
 * 8) * Pre-Tobelo: merged with /*s/
 * 9) * Pre-Tabaru: merged with /*s/
 * 10) * Pre-Pagu: merged with /*r/ or /*l/ (depending on whether Pagu /r/ descends from PNH or is a reintroduction after merging with /l/)
 * 11) * Pre-Modole: merged with /*s/
 * 12) * PTT: merged with /*r/
 * 13) * Old W. Makian: merged with /*d/?

*c (and *j)
Consonant temporarily denoted /*c/ that develops into /t/ (Galela, Loloda, Tabaru, Tobelo, Pagu, Modole) and /c/ (Sahu, Ternate, Tidore), with W. Makian as yet uncertain.

*ʔ
May not be phonemic (see Kotynski's paper), and for now is phonetically indicated wherever possible. Initial-position reconstruction for nouns is almost entirely dependent on the Sahu reflex and for verbs can be adduced from how its N- prefixed form surfaces (with initial g indicating a glottal stop and initial ng indicating a plain vowel - as per both Hayami-Allen and Kotynski). Tobelo may also be helpful, but needs more investigation. Medially, Galela and Tabaru reflexes also presumably continue PNNH *ʔ.

Specifically, initial glottal stops in Tabaru reflect both PNNH vowel-initials and ʔ-initials. A lack of initial ʔ in Tabaru is evidence of a lost consonant (PNH *h, *l, *w, and sometimes ḋ). Galela appears to introduce a non-phonemic ʔ before every intial stressed vowel.

It is possible that *ʔ may have originally been /*q/. This could explain its *N- prefixed realization *g, if the process were *N- + *q → *ɢ and *ɢ merged with *g while *q developed into a glottal stop. May also explain the errant realizations of *'ʔa.ŋe(r) ("three"), with *ʔ surfacing as both a glottal stop and as k.

*L?
And so we have the following distribution


 * but see also *moleun, *p2ilok, *p2olikur


 * While these may be conditioned instances of /l/ by neighboring /o/, /u/, or /a/, there are many terms (like *ŋolot, *p1ola, *laḋem, *lakem, *p2olotek), that suggest /l/ contrasts from this /L/.
 * Also, Fortgens, in his Tabaru Grammar, notes "Veel Tab. woorden hebben een l tegen die van het S.-W. een d." (pg. 323). Fortgens hadn't yet discovered the implosives in Sahu.
 * Interesting to note is how many of the above terms are likely borrowed from Austronesian.

Potential correspondences

 * 1) Final (ignoring the echo) Paḏisua ʔ and Tala'i ḏ
 * 2) Occasional Northern (Tabaru, Tobelo, Galela, etc.) s and Ternate/Tidore/Sahu c when the more usual correspondence is t and c

"Improper Retentions" in non-TT/Sahuic "inherited" vocabulary
Note the comparanda where the Ternate term (in its present form) cannot be the direct donor of the Northern forms.

Possible explanations include:
 * 1) Borrowings from Sahu, if we assume the Sahu terms to be inherited
 * 2) Borrowings from Pre-Ternate-Tidore (before the loss of finals in TT)
 * 3) * These loans would have reintroduced palatals to the Northern languages
 * 4) * This presumes that this Pre-TT period coincided with the period after which the native palatals had been lost in the Northern languages
 * 5) * If the kingdom of Jailolo actually had the influence that legend gives it, then this early Jailolo language (Old Jailolo, to distinguish from what Katarabumi might have written down?) might be the most likely candidate, possibly equivalent to PTT or a separate branch within TT

Notice also that (as yet) these all could be culture loans (even the numeral 1000).

Almost Reconstructible Terms

 * *gani and *gaʔani ("louse")
 * *gani (all but Tabaru)
 * *gaʔani (Tabaru)
 * *roŋa and *romaŋ ("name")
 * *roŋa (Ternate, Tidore, Galela, Tabaru, Modole - possibly borrowed into the last three)
 * *romaŋ (Sahu, Gamkonora, Tobelo (B & H), Loloda, Pagu)
 * *suramo and *sauramo ("fog, mist")
 * *suramo (Ternate, Tidore [soramo])
 * *sauramo (Tobelo (H & B), Modole, Pagu, West Makian)
 * sairamo (Sahu)

Intervocalic /k/ and its absence
Interestingly, in van der Crab's glossary from the 1862, he lists many of these Tidore terms with an 'h' in place of a 'k' (laho). However, he write "lao" for 🇨🇬. I did though get the impression that this wordlist was not produced by him, but compiled from other sources. And, de Clercq (1890) writes simply "lau". This list contains (I think) all the examples of k-elision in Tidore found in the data I have (Kamus Bahasa Tidore, etc.). But, the examples given for k-retention are only a small subset of the ones in the data. There appears to be no pattern; k is lost and retained in the same vocalic environments and k is lost and retained in core and non-core vocabulary (and morphemes).
 * The NH cognate is Ternate by default and specified when not Ternate.

*N- prefixation

 * *N- + *p1 ⇒ *b
 * *N- + *p2 ⇒ *b
 * *N- + *t ⇒ *d
 * *N- + *c ⇒ *j
 * *N- + *ḋ ⇒ *ɲ
 * *N- + *k ⇒ *g
 * *N- + *ʔ ⇒ *g
 * *N- + *w ⇒ *ŋ
 * *N- + *h ⇒ *ŋ
 * *N- + *∅ ⇒ *ŋ

Pre-Nasalization hypothesis
Purely conjectural: voiced stops may have been originally prenasalized. Evidence:
 * N- prefixation (be it actually ŋ- prefixation or otherwise): (for example) *N + *p = *Np → *ᵐb → b (in reflexes) works a little more nicely than without the prenasalization hypothesis. Additionally, *N + *ʔ = *Nʔ may plausibly have become *g (but this is far from clear phonetically).
 * Supports *tofʔaŋe → *topʔaŋe but *tofdiŋ (/*tofⁿdiŋ/) → *tomdiŋ (since otherwise we should expect **tobdiŋ; however, note WM tepedingi).
 * Could support certain old loans such as kambing → *kabi(n/ŋ) and tanduk → *taduk. But, we see the same simplification behavior in *gutiŋ and 🇨🇬 (from rantai). If nasalization was spontaneous in these Malay terms (like empat) and the NH forms are loaned from non-nasalized donors, then only do these loans support prenasalization.
 * Areally appears to be well-supported.
 * Evidence against:
 * 🇨🇬 → 🇨🇬 (though this one would have gone through North Moluccan Malay)
 * 🇨🇬 → 🇨🇬 (though this one would have gone through North Moluccan Malay)
 * 🇨🇬 → 🇨🇬 (though this one would have gone through North Moluccan Malay)

Glottal Stop Insertion Rule
Kotynski's glottal insertion rule: "Insert glottal stop before vowel-initial roots where that vowel is stressed." He argues that it applies to PNH (what I'm calling PNNH - Proto Nuclear North Halmaheran, since West Makian needs better data). I propose that Kotynski's glottal insertion rule apply only to Proto-Northeast North Halmaheran (PNeNH). Specifically, PTT and PSahuic appear to have preserved a distinction between plain vowel initials and glottal stop initials, suggesting that /*ʔ/ was at least partly phonemic.

I tentatively propose a reconstruction *ca.'i.ʔit ("to hurry, quicken, hasten") with descendants including 🇨🇬, 🇨🇬, and 🇨🇬. It appears as though the sequence *iʔi was reduced to *i in both PNeNH and PTT, but not in PSahuic. Comparing with *ti.ʔi.kit ("to cough"), with reflexes 🇨🇬, 🇨🇬, 🇨🇬, and 🇨🇬, the difference seems to be which syllable carries the stress. A stressed glottal stop may be preserved, whereas an unstressed one in a V1.ʔV1 sequence may not be.

Palatalization hypothesis
The proto-consonant pairs p1, p2; t, c; d, j; n, ɲ; and l, L could all represent plain vs. palatalized consonants. In particular, it could be: It's unclear if PNNH *b and *m had plain and palatalized realizations. It's also possible that this could extend somehow to the proto-consonants *S (/sʲ/?) and *ḋ (/gʲ/?), but more research is necessary. It also may shed light on the (tentative) relationship between PNNH *ʔ and Tidore y.
 * p2 = /p/; p1 = /pʲ/
 * t = /t/; c = /tʲ/
 * d = /d/; j = /dʲ/
 * n = /n/; ɲ = /nʲ/
 * L = /l/; l = /lʲ/

If this is the case, then it was likely not triggered by back vowels, since Ternate, Tidore, and Sahu all maintain distinctions between sequences such as nyi and ni, ci and ti, etc.

Stressed Vowel Lowering Hypothesis
All branches (Sahuic more than others) of Nuclear NH demonstrate considerable vowel assimilation, and primarily this appears to be the stressed vowel assimilating to the following vowel. What might facilitate this is lowering of the primary (and possibly secondary) stressed vowel in a word. Particularly, I posit that all branches of Nuclear NH experienced something resembling the following lowerings: It is possible that /i/ lowered to /ɪ/, but I currently lack evidence of /i/ assimilation. In fact, almost all (all but *'i.gon) reconstructions with primary stress on /i/ have no other vowels (they're all /i/), which is extremely suspicious.
 * /'o/ → ['ɔ]
 * /'u/ → ['ʊ]
 * /'e/ → ['ɛ]
 * Actually, Van der Roest (his entries are more phonetic than phonemic) writes Tobelo (Boeng) hedete as the reflex of PNNH *'si.det, suggesting exactly that.

These lowerings may be indicated in Baarda's and others' vowel diacritics, but I have not figured out their system.

Additionally, there is evidence that the vowel directly preceding a stressed vowel may assimilate to the stressed vowel: In many other cases, the preceding vowel is reconstructed to already be the same as the stressed vowel (*be.'Ze.Zoŋ, *'bo'ʔo.car, *p2e.'re.ki, *ti.'ʔi.kit, etc.), and so potential historical dissimilarities are currently undetectable. This may indicate that, if the stressed vowel lowering did in fact occur, it also occurred on the preceding vowel.
 * PNNH *bo.'te.kon → PNeNH *be.'te.kon, 🇨🇬
 * PNNH *gol.'ci.p1ir → PNeNH *gi.'ci.p1ir, 🇨🇬
 * PNNH *mo.'na.ram → PNeNH *ma.'na.ram
 * PNNH *p1o.'ḋe.ka → PNeNH *p1e.'ḋe.ka
 * PNNH *p2o.'Li.kur → PNeNH *p2i.'li.kur

Could this also just be lowering of the first vowel? Uncertain.