Ai 12.0 (still a Lakes Plain language)

[VaptuantaDoi]

Ai

Over the past few years I've been working on and off on an a posteriori Lakes Plain conlang, something that to my knowledge has never been done before. I made a thread about it previously, but I don't like that version any more, and I now have (hopefully) a better understanding of syntax than I did then, so I'm starting afresh. Within this thread I'll be presenting Ai as if it were a real language. Whenever I drop out of this (within a main post other than this one) I'll change to blue text to make it clear.


The Lakes Plain Family

The Lakes Plain languages (henceforth LP) form a language family fairly small in number of speakers, but large in number of languages – it's probably got the smallest average speaker base out of any reasonably-sized, non-moribund primary language family, with about two dozen languages and 8,000 speakers; a few of these are moribund but the majority are fairly vibrant even with only a few hundred speakers. They're spoken in the Lakes Plain region of West Papua (née Irian Jaya), a lowland area of sago swamps around the Tariku river basin, itself a tributary of the Mamberamo. It's fairly remote region, with small villages separated by hours of travel. Most peoples in the area live fairly traditional hunter-gatherer lives with limited agriculture, although recently the work of missionaries has led to some modernisation.
Typologically, the LP languages are remarkable for their small consonant inventories and lack of nasal consonants. Proto-Lakes Plain (PLP) has been reconstructed with an inventory of only *p b t d k, from which the descendants differ only slightly. No LP languages have phonemic nasals, a rarity in the world's languages,¹ and many lack them even phonetically. Descendants have between five and sixteen consonants, the former being Biritai with /b t d ɸ s/ and the latter Abawiri with /b bʷ t d tʷ dʷ d͡ʒ d͡ʒʷ k ɡ kʷ ɡʷ f fʷ s sʷ ɾ/. Other notable features are the presence of tone (phonemic in all languages and batshit in Iau which has up to 19 tonal contrasts), the presence in most languages of extrahigh fricated vowels /i̝ u̝/, fairly isolating morphology, ergative alignment and extensive zero-anaphor. Within the family, Iau is the most notorious; having truncated virtually all words to monosyllables, it gained a large number of tones and several more vowels, while the loss of verbal morphemes has led to an entirely tone-based aspectual system (if you've never read about it you definitely should, it's the least realistic thing I've ever seen in a natlang). Abawiri, the best-described member of the family, can be analysed without the use of grammatical relations, with topic and comment being instead the primary elements of a clause. Biritai has to my knowledge the smallest consonant inventory claimed of any living language with only /b t d ɸ s/.
There's a lot of debate around internal classification, so I'll follow (one of) Duane Clouse's classifications since he did all the work so I reckon he'd know what he was doing. The main division within the family is Tariku vs. Far West (aka Rasawa). The Far West branch contains only three languages (Rasawa, Awera and Saponi), all severely understudied, but notably distinct. The other 20-odd languages are all Tariku, further divided into West Tariku (Tause, Weirate, Deirate, Fayu, Sehudate, Kirikiri, Faia), Central Tariku (Edopi and a big dialect chain centred around Iau), East Tariku (a whole swath of dialect chains with languages roughly along the lines of Doutai/Kai/Waritai/Sikaritai, Biritai/Obokuitai/Eritai, Kwerisa and Papasena), Duvle (which is just Duvle), and East Lakes Plain (Abawiri, Taworta and Dabra). Roughly speaking, Tariku languages can be defined as those that have or had closed syllables at some point, in most cases resulting in extrahigh fricated [i̝ u̝] although in some languages codas remain. Apart from this it's difficult to make sweeping statements about sound changes. East Tariku must have retained closed syllables since Obokuitai, Eritai and Sikaritai all retain some codas, but all the other members of the family have lost them. Central Tariku is sort of generically "innovative", losing consonant clusters, but allowing rare closed syllables of a different origin to those of East Tariku. West Tariku doesn't really have any distinctive changes other than that it lost closed syllables. East Lakes Plain seems to have phonemicised /ɾ/ and two of the languages seem to retain closed syllables; it also seems to share the loss of initial *ku- with Far West, which brings into question its classification, but I defer to consensus here.
The Lakes Plain family is remarkably similar to the Skou languages – compare for instance I’saka's consonant inventory /b t d k ɸ s j w/. They are also tonal and mostly isolating. However, there's a dearth of cognates and it seems proto-Skou was less like PLP than some modern Skou languages are to modern LP languages. A relationship with the East Geelvink Bay languages has been proposed, but they're even more understudied than LP. Also superficially similar is the isolate Keuw.
Within LP, there are only nine languages that have any published material at all. Receiving the most study has been Iau, for which I have a very thorough dictionary and a rough grammar both by Janet Bateman; second is Abawiri, which has a very good grammar by Brendon Yoder; third is Fayu with a short grammar by Paige Maitland. For Doutai, Edopi, Sikaritai and Obokuitai there are phonological sketches (by L & K McAllister, I & S Green, David Martin and D Scott & P B Jenison respectively). There is a short sketch of a pidgin between Duvle and Wano, a Dani language. For Edopi there's also some translation material and discussions of kinship and naming practices, for Kirikiri a short phonological discussion which includes notes on other languages, for Sikaritai a discussion of discourse marking, and for Obokuitai an unpublished grammar and phonology sketch by Jenison partly reproduced in Foley's The Languages and Linguistics of the New Guinea Area. For other languages there are only wordlists and brief mentions in other works. There has only been one attempt at family-level reconstruction, by Duane (±Helja) Clouse, which established the language family and identified key sound changes.

---

1) With two possible exception – Faia has been claimed to have /m n̥/, and no that's not a typo, the alveolar nasal is voiceless. They arise from *bd pd → */bl ɸl/ → */bn ɸn/ → */mn hn/ → /m n̥/. Isn't that delightful? Less interestingly, Rasawa has an adjective suffix -nu which contains the only nasal in the language and probably isn't of native origin.

[Solarius]

I remember seeing the old thread on this and liked it a lot! Excited to see where this goes.

[DV82LECM]

The only further I can see /b t d ɸ s/ going would be the loss of /t/, merging with /s/. At what point would people just be talking static? If there is to be a phonemic bounce-back to a larger inventory, I imagine there could be potentially strange phonemes to derive until a most balanced state, like /m n̥/.

[VaptuantaDoi]

The only further I can see /b t d ɸ s/ going would be the loss of /t/, merging with /s/. At what point would people just be talking static? If there is to be a phonemic bounce-back to a larger inventory, I imagine there could be potentially strange phonemes to derive until a most balanced state, like /m n̥/.

Yeah, I'd be surprised if there were any languages ever to have four consonants. I think when you have five consonants, one being lost or merging would just lose you too many minimal pairs. But hey, Biritai managed to lose /k/, so who knows? LP languages have shown a fair bit of resistance to gaining consonants; most have 6-8. Actually, I'll see how good of an average I can work out:

Code: Select all

PLP       5   p b t d k
(Ai       5   h b   d k       s)

Biritai   5   ɸ b t d         s
Kirikiri  6   ɸ b t d k       s
Obokuitai 6   h b t d k       s
Iau       6   f b t d k       s
Sikaritai 7   ɸ b t d k       s         w
Edopi     9   p b t d k ʤ h   s
                 ⁿt
Doutai    9   p b t d k g     s       j w
Fayu      11  ɸ b t d k ʥ h   s     hɾ
Duvle     12  p b t d k   ɸ β s z ɣ   j w
Abawiri   16  f b t d k g     s     ɾ ʤ
              fʷbʷtʷdʷkʷgʷ    sʷ      ʤʷ

Avg.      9.3
Avg. -Ab  8.6

Unfortunately the phonologies of lots of languages like Eritai, Kaiy, Waritai, Deirate, Weirate, etc. aren't known but I suspect they're to the smaller end. I suspect that a more balanced average would be ca. 8 consonants, still way below what's considered "small" (WALS calls anything <15 consonants small). There's also some aspects of less-described languages which may turn out to be allophonic (like the /p:h/ or /ɸ:h/ contrasts or /hɾ/ in Fayu – I'd be tempted to analyse the latter as /hd/.

So basically, I don't think there's much incentive for a big boost in phonemes. The two "large" inventories (Duvle and Abawiri) both gained a whole lot of consonants at one time – Abawiri with a rounding contrast (only 9 consonants otherwise) and Duvle through mass frication (pi̝ bi̝ ti̝ di̝ ki̝ → ɸ β s z ɣ), rather than just phonemicising extra consonants piecemeal.

[Porphyrogenitos]

I'm reading Yoder's grammar of Abawiri right now, and wow, it's fascinating how it really doesn't seem to have any marking of grammatical relations at all. Topics, comments, and focused elements can all be semantic agents, patients, locations, and other roles.

I will also point out that all of the attested Lakes Plain microinventories still hold to the typological generalization whereby pairs of consonants of the same PoA are distinguished by MoA, something that seems to be an organizing principle for even the smallest inventories - e.g. Rokotas with the voiceless/voiced contrast or Hawaiian with /m/ vs. /p/, /n/ vs. /l/, and /h/ vs. /ʔ/. It looks like the Lakes Plain languages all have at least two or three such contrasts, while the proposed Ai inventory would only be hanging onto one such contrast, /d/ vs. /s/. It seems likely that the language would hold onto this contrast, and even that it might have evolved in a way that retained (or redeveloped) at least one other MoA contrast at the same PoA.

[VaptuantaDoi]

I will also point out that all of the attested Lakes Plain microinventories still hold to the typological generalization whereby pairs of consonants of the same PoA are distinguished by MoA, something that seems to be an organizing principle for even the smallest inventories - e.g. Rokotas with the voiceless/voiced contrast or Hawaiian with /m/ vs. /p/, /n/ vs. /l/, and /h/ vs. /ʔ/. It looks like the Lakes Plain languages all have at least two or three such contrasts, while the proposed Ai inventory would only be hanging onto one such contrast, /d/ vs. /s/. It seems likely that the language would hold onto this contrast, and even that it might have evolved in a way that retained (or redeveloped) at least one other MoA contrast at the same PoA.

There's Obokuitai with /b, t d s, k, h/, i.e. only one PoA contrasts MoAs, but it's three-way. You could analyse Ai like this:

Code: Select all

     PER  CTR  RAD
-VC   k    s    h
+VC   b    d

Or like this

Code: Select all

     LAB  COR  VEL
PLOS  b    d    k
CONT  h    s

(/h/ has a labial allophone [ɸ] which could be taken as the underlying form)

It just depends on how strict you are with what determines the "same" POA.

[VaptuantaDoi]

Phonology

I've finally pretty much settled on Ai phonology. It's fairly similar to other (natlang) Lakes Plain languages but with a few extra shifts to add some spice. The tone system is my own but it's vaguely similar to that of Abawiri or the possibly-related Skou languages.


Ai has five consonants and seven vowels which combined to produce CV and V syllables. Each syllable has a surface tone; these are determined by word-level tonal melodies. Allophony is fairly extensive as is typical of a small consonant inventory. Notable features include the lack of nasals either phonemically or phonetically, and the areally unusual appearance of central vowels.


Consonants

Ai has only five consonants:

/b d k/ b d k
/s h/ s h

As with some other Lakes Plain languages like Obokuitai, there is only an MOA contrast in coronal phonemes. Unlike all other Lakes Plain languages, voicing is not contrastive on stops (or any other phonemes), although each segment is fairly consistently either voiced (in the case of /b d/) or voiceless (in the case of /k s h/). A simpler analysis only posits three POAs and two MOAs:

Code: Select all

     PER  CTR  RAD
-VC   k    s    h
+VC   b    d

This inventory differs from Ai's neighbour Obokuitai by lacking a voiceless coronal plosive */t/. It seems that Ai underwent a chain shift of *t k → k ∅:

​ ​ ​ ​ ​ ​ ​ PLP *kdiCV "banana" → Ai /ɞdi/ (cf. Obokuitai /kdíd/)
​ ​ ​ ​ ​ ​ ​ *kubadi "fly (insect)" → Ai /ubadɞ/ (cf. Ob. /kuâde/)
​ ​ ​ ​ ​ ​ ​ *touCV "breast" → Ai /kɔʉ̯/ (Ob. /tóud/)
​ ​ ​ ​ ​ ​ ​ *tau "person" → Ai /ka/ (Ob. /tá/)

The consonants /b d/ for a natural class of [+voiced] or arguably [+cont]; they have lenided allophones [β l~ɾ] in post-vocalic position:

​ ​ ​ ​ ​ ​ ​ ö́bö̀ /ɞ́bɞ/ → [ɞ́βɞ̂] "leech"
​ ​ ​ ​ ​ ​ ​ álau /adau̯/ → [álàu̯] ~ [áɾàu̯] "wing"

Interestingly this lenition is not triggered by the non-back high vowels /i ʉ/. Following /i/, /d/ is instead palatalised to [d͡ʑ], while /ib ʉb ʉd/ retain their stopped values.

​ ​ ​ ​ ​ ​ ​ uíbekúlúà /uíbɛku·dúa/ → [ùíbɛ̄kúlúâ] "suck up, slurp"
​ ​ ​ ​ ​ ​ ​ ádǜdebà → /âdʉdɛ·bá/ → [álʉ̂dɛ̀bâ] "play"
​ ​ ​ ​ ​ ​ ​ áide /ai̯dɛ/ → [ái̯d͡ʑɛ̀] "yesterday"

Voiced phonemes are also allophonically imploded word-initially:

​ ​ ​ ​ ​ ​ ​ bédâ /bɛ̂da/ → [ɓɛ́la᷈] "sweet potato"
​ ​ ​ ​ ​ ​ ​ dâ /dâ/ → [ɗa᷈] "this"

/h/ can be labialised to [ɸ] in the environment of rounded vowels, especially /u/. This is more common in the dialect of the eastern village, Usabiri (the native name being Úsàbö́dö, from whence may be taken the alternative name Usabiritai for either this dialect or the Ai language as a whole) where Biritai and Iau influence is more prominent.

​ ​ ​ ​ ​ ​ ​ óhoi /ɔhɔi̯/ → [ɔ́hɔ̀i̯], [ɔ́ɸɔ̀i̯] "hair"

The remaining phonemes /k s/ are mostly invariable. However a plosive pronunciation of /s/ as [t] is heard not infrequently, especially in the speech of women, before any vowel other than /i/.

​ ​ ​ ​ ​ ​ ​ soi /sɔi̯/ → [​sɔ̀i̯], [tɔ̀i̯] "dry"
​ ​ ​ ​ ​ ​ ​ sì /sí/ → [​sî] "gecko" (*[tî])

In recent borrowings /t/ is assimilated as /s/ rather than historical /k/:

​ ​ ​ ​ ​ ​ ​ Indonesian tembakau → söbákàu [sɞ̀βákâu̯] "tobacco"

/k/ is quite strongly aspirated in all positions and may even be fricated to [x].

​ ​ ​ ​ ​ ​ ​ káahu /kaahu/ → [kʰáàhù], [xáàhù] "star"


Vowels

Ai has seven vowels.

/i ʉ u/ i ü u
/ɛ ɞ ɔ/ e ö o
/a/ a

This makes it unusual in having more vowel phonemes than consonant phonemes; it has a consonant-vowel ratio of 0.71. Within Lakes Plain this is also found in Biritai (0.71), Iau (0.75) and Kirikiri (0.86); it is also seen in some completely unrelated languages like Puinave (0.73) and Karajá (0.69), both from South America.

Lacking from Ai are the extrahigh fricated vowels /i̝ u̝/ found either phonemically or phonetically in most Lakes Plain languages. These are instead reflected as normal high vowels /i ʉ/ while original high vowels are seen as /ɞ u/. Similar changes of /i̝ i/ → /i ɪ/ and /u̝/ → /y/ are seen in Duvle and Abawiri respectively.

​ ​ ​ ​ ​ ​ ​ PLP *kdiCV "banana" → Ai /ɞ́di/ (cf. Kirikiri /kdi̝/)
​ ​ ​ ​ ​ ​ ​ *kuda "tree" → Ai /ʉ́/ (cf. Kaiy /ku̝/, Abawiri /yɾɛ/)
​ ​ ​ ​ ​ ​ ​ *ti "string bag" → Ai /kɞ́/ (cf. Obokuitai /tí/)
​ ​ ​ ​ ​ ​ ​ Ai /su/ "kangaroo" (cf. Obokuitai, Doutai /su/)

There is considerably less vowel allophony due to the restricted space available to each phoneme. /i ʉ/ tend to be slightly higher than /u/, and /ɔ/ may be as low as [ɒ]. /ɞ/ varies between [ɞ ~ ɵ ~ œ ~ ø] and /ʉ/ between [ʉ ~ y]. The diphthongs /ɛi̯ ɛʉ̯ ɛu̯ ai̯ aʉ̯ au̯ ɔi̯ ɔʉ̯ ɔu̯/, and in the Usabiritai dialect only /aɞ̯/. Of these /ɛʉ̯ ɛu̯ ɔu̯/ are fairly rare.

Vowel hiatus is not uncommon. Low-tone /i u/ may become non-syllabic [j w] in hiatus (/u/ is considerably more common than /i/ in this position)

​ ​ ​ ​ ​ ​ ​ uíbekúlúà /uíbɛku·dúa/ → [ùíbɛ̄kúlúâ] ~ [wíbɛ̄kúlúâ] "suck up, slurp"

Note that the precise shift of high front vowels was a merger of *i̝ i → /i/ following another vowel, *i̝ i → /id i/ before another vowel, and *i̝ i → /i ɞ/ otherwise. Hence why hiatus /ɞ/ is rare and diphthongs don't end in /ɞ/ (except the additional Usabiritai diphthong).


Tone

Tone is assigned at the level of the morpheme. All morphemes are either unmarked for tone, or marked for H or F tone. When unmarked (free) morphemes are produced in isolation, L tone is assigned to all syllables; when attached to another morpheme they assimilate to its tone pattern. Otherwise multiple tone patterns can occur within a multimorphemic word. H morphemes take H tone on all syllables.

​ ​ ​ ​ ​ ​ ​ |hada| L → /haLdaL/ "land"
​ ​ ​ ​ ​ ​ ​ |idɛdɞ| H → /iHdɛHdɞH/ "sing"

Monosyllabic and bisyllabic F morphemes take falling tone on the final syllable, and high on the penult. Morphemes with three or more syllables take falling on the penult, low on the ult and high on all pre-penults.

​ ​ ​ ​ ​ ​ ​ |kɛ| F → /kɛF/ "fish"
​ ​ ​ ​ ​ ​ ​ |ukɞ| F → /uHkɞF/ "grasshopper"
​ ​ ​ ​ ​ ​ ​ |ubadɞ| F → /uHbaFdɞL/ "fly (insect)"

At this point all the tones which have been assigned to syllables are underlying. Three rules are then applied to obtain the surface forms – firstly, word-final lowering; secondly, prevention of HH sequences, and thirdly prevention of LL sequences.

1. L H F → L HL LHL / _ #
2. H → L / _ H ! / _ .
3. L → H / [L .] _ L ! _ .

Rule (1) shifts falling tones to rising-falling and high tones to falling word-finally. Low tones are unaffected. Note that this process applies word-finally rather than morpheme-finally, so a tone pattern within a word behaves differently.

​ ​ ​ ​ ​ ​ ​ |ubadɞ| F → |uHbaFdɞL| → [úβâlɞ̀] "fly (insect)"
​ ​ ​ ​ ​ ​ ​ |hɞdi| H → |hɞHdiH| → [hɞ́lî] "follow"
​ ​ ​ ​ ​ ​ ​ |ɔdɔ| F → |ɔHdɔF| → [ɔ́lɔ᷈] "old"
​ ​ ​ ​ ​ ​ ​ |morpheme

Rule (2) lowers an H tone preceding another H tone. This applies left-to-right recursively, so that any sequence of H tones is changed to a series of L tones followed by a single H tone. Note that this does not apply over a morpheme boundary (notated above as "."), which is the only position where HH sequences can occur.

​ ​ ​ ​ ​ ​ ​ |ʉ-kɔi̯dɔ-i-ɛ| H → |ʉHkɔi̯HdɔHiHɛH| → |ʉHkɔi̯LdɔHiHɛH| → [ʉ́·kɔ̄i̯d͡ʑɔ́·í·ɛ̂] "will close it up"

Rule (3) repairs LL sequences where this will not produce HH sequences (i.e. Ai has a constraint against LL sequences which is trumped by a constraint against HH sequences). An L tone is raised to an H tone when the following tone is low and the previous is not high (invariably this means it is low); or when the following tone is low and it (the first L tone) is preceded by a morpheme boundary. Again this is left-to-right recursive does not apply over morpheme boundaries, so L.L remains. Since it is blocked by an adjacent H tone, sequences of LL are still permitted, although sequences of LLL are not. This means a sequence of L tones will become an alternating sequence of L and H tones:

​ ​ ​ ​ ​ ​ ​ $.$.$.$.$.$.$H → *H.H.H.H.H.H.H → *L.L.L.L.L.L.H → L.H.L.H.L.L.H
​ ​ ​ ​ ​ ​ ​ $.$.$.$.$.$.$.$.$.$.$.$H → *H.H.H.H.H.H.H.H.H.H.H.H → *L.L.L.L.L.L.L.L.L.L.L.H → L.H.L.H.L.H.L.H.L.H.L.H

Allophonically there is a further rule which produces mid tones in predictable enivronments. Unlike the segmental rules, this applies at a word level irrespective of morpheme boundaries.

• L → M / H (.) _ (.) H

Tone is romanised with diacritical marks; unmarked for low or mid tone, an acute for high tone, a grave for falling tone and a circumflex for rising-falling.

​ ​ ​ ​ ​ ​ ​ [​sɔ̀i̯] → soi "dry"
​ ​ ​ ​ ​ ​ ​ [úβâlɞ̀] → úbàdö "fly (insect)"
​ ​ ​ ​ ​ ​ ​ [ɔ́lɔ᷈] → ódô "old"


Syllable structure

Ai syllables are of the form (C)V(V). Onsets may consist of zero or a single consonant; nuclei may be either a single vowel or a diphthong. In addition every syllable bears a surface-level tone.

​ ​ ​ ​ ​ ​ ​ V:​ ​ ​ ​ ​ ​ ​ ö [ɞ̀] "1SG pronoun"
​ ​ ​ ​ ​ ​ ​ VV:​ ​ ​ ​ ​ ai [ài̯] "1PL pronoun"
​ ​ ​ ​ ​ ​ ​ CV:​ ​ ​ ​ ​ ​ku [kʰù] "centipede"
​ ​ ​ ​ ​ ​ ​ CVV:​ ​ ​ kêi [kʰɛ᷈i̯] "child"

---

That should be pretty much it! Comments and questions always welcome; I suspect I haven't explained the tone system all that well, so I might cover it in more detail in another post.

[Creyeditor]

Nice tone section. I think the melody mapping and the melody inventory is very naturalistic. I also like the low raising at the word level and the word-final low boundary tone. I was wondering if similar stuff happens across word boundaries or at the end of a phrase.
Your tone dissimilations are a bit typologically unusual, I would say. High tone melodies 'spread-out' across a morpheme are counted as a single high tone for the purposes of dissimilation in many languages (e.g. Shona). You doing the opposite is a nice little quirk that gives a very Papuan feel.

[VaptuantaDoi]

Nice tone section. I think the melody mapping and the melody inventory is very naturalistic. I also like the low raising at the word level and the word-final low boundary tone. I was wondering if similar stuff happens across word boundaries or at the end of a phrase.
Your tone dissimilations are a bit typologically unusual, I would say. High tone melodies 'spread-out' across a morpheme are counted as a single high tone for the purposes of dissimilation in many languages (e.g. Shona). You doing the opposite is a nice little quirk that gives a very Papuan feel.

Thanks! Basically I was trying to extend a general Skou/Lakes Plain tone system, suited to isolating ~ mildly agglutinative languages, to the borderline-polysynthetic Ai. My idea is that the pseudo-obligatory contour principle is kind of a hang-over from Ai's less synthetic past.

I'm gonna work out higher-level tone behaviour, but I'll have to read up on other north Papuan languages first. I also might add some more tone melodies – Obokuitai (e.g.) has eleven word-level tone patterns and Ai having only three might be a bit too much of a collapse (or it might not be; I'll be making up like 80% of the vocabulary anyway).

[Creyeditor]

I like the three melody inventory. It gives a nice pitch-accent like touch.

[Arayaz]

Really liking this so far! Excited to see where it goes.