#phonemicizing

[hya3rdxn3y]

Kyky Trans dando na esquina da rua, pro boy novinho em Goiania, dlc de Bareback Pov titfucking milf babe novinha safadinha viet nam hay nhat em gai xinh dep Hot Slut Leah Gotti Get Fuck Braquinha linda da bunda grande DANDO gostoso BHPORN Pissen ganz nah drann Skype girl having fun Pussyfucking hardcore Indian masturbating dick moaning loud and cumming heavy

#lichenographical #singeingly #endodontology #postanesthetic #Spartanly #butcher's-broom #deintellectualization #wryness #Sisyphism #otorrhagia #Polygordius #phonemicizing #pterographical #ahey #unfirmness #entrance-denying #proctoptoma #porelike #interopercular #overthrown

[yeli-renrong]

I saw someone on Discord ask why ‘karaoke’ was loaned into English with ao > io (as if the source word were **’karioke’), and with the correct analysis of English phonology, this is easy to explain. It was borrowed before Japanese loans became an orthographic stratum, so the idea is that the second ‘a’ represents FACE, but since it has penultimate stress, the antepenultimate vowel must reduce: //kerejˈʌwkej// > [kerɨjˈʌwkɨj]. 

Cf. ‘Israel’, which is pronounced [ˈɪzrijəl] (representing /izrɨjəl/), but unreduced [ˌɪzˌrajˈɛl] if you’re singing Handel’s Messiah.

Of course, schwa and schwi don’t contrast before /j/, so on some level you could just as well phonemicize these as /kerəjʌwkəj/ and /izrəjəl/. That’d even be diachronically correct! Like, that tiger poem that rhymes ‘eye’ and ‘symmetry’, GVS applied to the unstressed final <y> but vowel reduction reversed it.

[d1squ13tud3]

So language has hiarchiees..... Language also has Ground To Cover. All languages are equally capable of expression. We hold this to be true on humanistic grounds . . . and on the grounds of immigrant children who grow up to be adults without accents.

Whatever number of distinctions there are in language spread out to fill the space and terrain: Phonologically, vowel systems spread out so that the contrast is maximal. Color systems, presumably working on some thread of human psychoperception, spread out so that the contrast is maximal.

Big question must be able — how many nodes of contrast (i.e. distinctions) are optimal? 5 vowel systems are very common. The trapezoid or triangle of the vowel space — optimally distributed with A E I O U? You could do the typologies, examine this . . . Interesting.

Obviously some grammatical distinctions are supremely relevant. Aspect, tense . . . . past/present/future, perfect/imperfect. For instance with a preterite/perfect past distinction, auxiliaries may be used as in English, synthetic tenses as in Spanish, verb stem alternations as in Russian. The grammaticalization of verb stem distinctions (with obviously morphophonological components) in Russian — interesting comparisons/contrasts to synthetic tenses as in Spanish. Obviously I need to study a non Indo-European language to get some more perspective here. But the point is is that the reason tense/aspect/mood is useful as a language for analyses is because these are Generalizations that pop up.

But are the hierarchies rigid? It's said that sometimes, Case hierarchies, Color term hierarchies, etc, are not strictly adhered to — just as sometimes the Vowel Space may not be distributed "optimally." But when the Vowel Space is unstable, we expect it to shift and stabilize,

almost analogously perhaps to a physical system moving towards equilibrium.

The human mind might, as it were, vibrate at a certain frequency, shaking grammatical grains of sand into a recurrent pattern on the table.

Question would be, does non-phonological components of language also need to shift into stable systems? It surely must, although perhaps phonology is itself the ultimate agitator of change: the Germanic stress-first phonological tendency, along with other factors of course, eroded the case system of English.

I'ms ure some Scholar has guessed this before me, but probably Indo-European must have been widely adopted as a second language during migrations, and areal features imposed from pre-IE substrates. We have seen this happen plenty of times in modern ages: isn't ash-tensing in American English due to immigrant phonological adaptations particularly in the Erie Canal region?

So again during Norman-French Old-Norse whatever-else assaults on England, some of the fine phonological distinctions of English must have been worn down (although this is a topic to be explored, because difficult sounds like the interdentals were kept, and in fact during this time the voiced fricatives nevertheless managed to phonemicize themselves, and so on . . .)

Does grammatical change always result from phonetic agitation? What determined the grammaticalization of the Unbounded as the English simple present verse the Bounded as the German simple present?

But I remained convinced it must all be hierarchies and continuums, made useful by relations of maximum contrast; and if Optimality Hierarchies are indeed a useful tool here, as I guess they are, then I agree with Prince/Smolensky when they say the constraints necessarily conflict with each other at high, broad levels.

Languages add dimensions of contrast in a systematic way phonologically, unlikely to have a single, odd member of a natural class. Phonemes are generally not chosen arbitrarily, but featurily, in a natural class context. This, perhaps, why Ezh emerges as a phoneme in English, if later than /v/ /eth/ /z/.

This of course why broad/slender consonants, soft/hard consonants — palatalization, nonpalatization (or velarization) — is a broad, natural, very broadly applicable dimension of contrast. As is voicing, nasality (for vowels, but I believe particularly for high vowels), rounding (for vowels, but I believe particularly for high vowels.) Of course, the "particularly for high vowels" stuff makes sense on this part if the vowel space is indeed a triangle (perhaps an inverted pyramid, 4-pyramid, etc). A consequence, probably, of physical considerations with human vocal tract and ears, and perhaps of Sound itself.

HAHAHAHAHAHAH it would be hilarious if cot/caught merger is in part an effect of an instability of the vowel system in terms of having irregular dimensions of distinction (JUST AS IRREGULAR GRAMMAR FEATURES TEND TOWARDS REGULARIZATION.) after Great Vowel shift, there's a long-short pairing in the frontcourt of ride-rid, lede-led, made-mad; very easy, three pairs. But in the back vowel space, we have to deal with "now" as well as "no", "new", "naw" . . . how does foot/strut split factor in? The weird shit with father/bother versus not? In American English, with merged cot-caught, we get a three-pair working in the backcourt as well, LOUD-FOOT, BOOT-BUT, COAT-COT. Odd diphthong out with BOY. Perhaps BIRD-schwa sensible tense-lax distinction. Why does OY exist? Why

[senzacaponecoda]

Something on Tantaftiyy stress

A typical Tantaftiyy word root, verbal or nominal, looked like either CVC or CVCVC

Due to vowel harmony processes when the vowel system collapsed to /a i u ə/ most roots ended up CəCəC. Nouns took case and verbs took conjugations, leading mostly to roots of the form CəCCan or naCəCCan~naCCəCan

While most tri- and quadriliteral roots were formed with consonantal extentions, as often were vowel extensions, -i typically having reductive or diminutive meanings, -u typically having augmentive meanings. Because of anticipatory assimilation, vowel harmony processes, the rise of the definite absolute state of nouns (which didn't mark case and left a root CəCC behind that needed to reform its second schwa), the extensions on nouns metathesized into them, with verbs following suit shortly thereafter due to analogy and the simplifying effects it had on conjugation, leading to marked extensions like CəCCú(wan) >CəCúCan

This marked infix would be stressed, as the distinguishing element from roots without the extensions. And with schwa deleting or merging with short a, this resulting in a vowel template system superficially similar to the Semitic languages.

On the other hand, the language wanted to be moraically balanced, following an original pattern akin to old Egyptian. So vowels in open syllables typically were lengthened. This meant that for original CəCəC-VC roots, you might expect CáCáCán etc. But with schwa deletion, this would result in situations like CəCCán > CaCCán or such. But this phonemicized length (and absorbed the old pitch accent system), before it collapsed into a centralization distinction. This meant that 'overlong' syllables became common. The language wanted to avoid these, so it shortened vowels in regular checked syllables such as case endings, and around preferentially stressed syllables, like the infixed roots. The next preference would be to keep the first open syllable long, which due to schwa deletion and the infixes usually meant the second syllable. So the language fell into an iambic pattern for stress, which would spread by analogy.

So a root CəCəC might have the associate accusative nouns (definite, indefinite, pl def, pl indef)

0: lan-CaCáC , CaCáCan, lan-uCCáC, uCCáCan

u: lan-CaCúC , CaCúCan, lan-uCCúC, uCCúCan

i: lan- CaCíC , CaCíCan, lan-uCCíC, uCCíCan.

and associate 2nd m singular verbs (imperfective, perfective)

0: śi-CCáC-akú, CíCC-ikú

u: śi-CCúC-akú, CiCúC-ikú

i: śi-CCíC-akú, CiCíC-ikú

The old way (which derived from a prescribed analogy from the naturally evolved way) to form the associated construct stem in between Góoreta and Tantaftiyy was to prefix a- and mutate all other vowels (by vowel harmony) to their open counterparts, and make all syllables open. So the CəCəC roots would become aCəCəCa > áCəCəCa > áCaCáCa, áCaCáCawá, áCaCáCayá. But this form was hard to analyze, becoming often áCaCúCa, áCaCíCa, etc. Not to mention, this put adjacent stressed syllables next to each other in biliteral roots and their analogues. Since the other forms of the noun generally had a CCVC shape as well, and schwa syncopation was a well established process, analogy turned the construct state into aCCVCa.

Other affixes, such as the participle m-, the causative s-, the inchoatives n- or -n, the feminine -t or t- or t-t, the nisba -iyy, and other morphemes, would interfere with this pattern however. At the same time, reduplicated roots, and inherited mono-, bi-, quadro-, and quintoliteral roots didn't necessarily want to play ball, not to mention the phonetic considerations of the 'weak' consonants, such as the semivowels (what exactly does úw mean, anyways? /u:/? /u:w/?)

[linguisten]

Call for Papers

As linguistic research expands to cover an ever increasing number of languages and subfields, various collaborations have been carried out to standardize terminology and glossing practices. This has led to a wealth of useful resources such as Glottopedia or the Leipzig Glossing Rules, with various other projects currently underway. The Japanese Linguistics Internationalization Committee (JLINC), established in April 2019, aims to compile a list of English, German, French, Chinese and Korean translations for Japanese linguistics terminology as well as a set of interlinear glossing rules and phonemicizations for Japonic (Japanese-Ryūkyūan) languages and dialects.

The goal of LiTGaP 2020 is to bring together researchers from various fields of linguistics to discuss issues related to linguistic terminology and translations thereof, as well as the methodology of interlinear glossing and phonemicization. Papers on linguistic terminology from all languages and subfields as well as all aspects of glossing and phonemicization related to any language are welcome.

This year’s conference will focus particularly on the history of linguistic terminology and interlinear glossing. While the interlinear morphemic gloss is a relatively recent phenomenon, the practice of glossing texts with lexical, morphosyntactic and phonetic annotations is by no means a new phenomenon, with glossing traditions dating back centuries if not millennia throughout many cultures. This year, we welcome papers not only on the modern interlinear morphemic gloss, but glossing at large as a cross-cultural, diachronic phenomenon.

Papers are to be 20 minutes long with 10 minutes for Q&A. Papers are to be delivered in English or Japanese. Other languages may also be accepted on a case by case basis (please contact us in advance). If presenting in a language other than English, we ask that you provide handouts or a PowerPoint presentation in English.

Please send your title and abstract in English or Japanese (500 words or 800 characters max) to Matthew Zisk ([email protected]) by October 15, 2019.

Limited financial assistance for travel expenses and/or accommodation will be available for speakers from outside of Japan. Please let us know if you require financial assistance when you apply.

This conference is funded in part by Grant-in-Aid for Scientific Research (B) #19H01265: Development of a multilingual dictionary of Japanese linguistics terminology and a glossing standard for Japonic languages (PI: Matthew Zisk).

#linguistics #linguistic terminology #glottopedia #lingblr #cfp

[wardoftheedgeloaves]

The Germanic languages could be analyzed as having a series of presigmatized stops (possibly inherited unchanged from PIE; cf. NEC unit *st); if we do this but don’t accept unit geminates, Anatolian could be the only branch containing languages with <15 consonants. (Possible counterexamples: creoles, dialects of Spanish, maybe some extinct Italic languages.)

The maximum Spanish phonology seems to be nineteen consonants: /p t tʃ k b~β d~ð g~ɣ m n ɲ f θ s x r ɾ l ʝ ʎ/. Most dialects have s-θ and ʝ~ʎ merger, making seventeen. I don’t think any single dialect of Spanish gets under fifteen, but there are enough dialects with additional mergers that it’s easy to postulate one getting there:

- many young Argentines unpack /ɲ/ to [ni̯]

- some varieties merge /f/ with /x/

- Argentina and Uruguay often have [ʃ] for /ʝ/ + /ʎ/; deaffricated [ʃ] for /tʃ/ occurs a number of places, including Chile. I don’t know if there are any varieties that reflect all of /tʃ ʝ ʎ/ as [ʃ].

Somewhat surprisingly, there don’t seem to be any dialects (known to Wikipedia) that have any unconditional mergers involving /l r ɾ/--the closest is some varieties of Caribbean Spanish where “strong” /r/ is uvular and sometimes devoiced, but this does not merge with /x/.

Oscan had fifteen (really fourteen): /p t k b d g m n f s x~h v~w r l ts/. /v~w/ is from *w. I had to run to Buck for <z> /ts/: it’s apparently an orthographic convention and always continues earlier *t + *s or inserted in medial *ns > -nts-, so it doesn’t seem to be truly phonemic. Gaulish also had fourteen consonants, removing the /f/ and /x/ of Oscan (the latter exists in the Gaulish script but is at least in some dialects only a preconsonantal allophone of /k/, sometimes phonemicized later) and adding /j/ (really phonemic? probably only to the extent it was in Oscan) and /ts/ (from *-ds- and *-st-).

Interestingly, Oscan also appears to have had undergone yroldism, as original *tl is continued as /kl/ (it becomes geminate /ll/ in Latin).

[scientificalnonsense]

Infinite Alphabet

My friend and I just got into a very trippy conversation about parallel universes and linguistics, sparked by the screenshot of a post by user @georgemallory:

“nothing will fuck you up as much as the realization that there’s no real reason the alphabet needs to be in order”

This is, in fact, very true: the sequence of letters in the English alphabet (or indeed, any alphabet) is, to my knowledge, totally arbitrary. (I will come around to this point later.)

What is interesting to note is that the present arbitrary sequence of the alphabet is uniquely processed in our brain because of the conditioning we have all received from a very young age.

ABCDEFGHIJKLMNOPQRSTUVWXYZ

QWERTYUIOPASDFGHJKLZXCVBNM

Now. Which of those did you process as individual letters…

and which did you attempt to process as a word?

If you’re like almost every human being that understands English, you processed the top sequence as the alphabet and the bottom sequence you attempted to phonemicize in your head like any other word. We are just conditioned to recognize that first string as alphabetical, in the “right” order, and we instantly recognize it as something to be processed letter by letter. But the second string has the exact same letters, and yet because it’s not in the “right” alphabetical order, we attempt to process it as a word.

From this point the discussion moved from psychology to the more complex stuff: math. My friend speculated what the linguistic and psychological consequences would be if young English-learners were made to learn all the different sequences of the same English alphabet. I did some quick calculations with the permutation formula and proved that this task would be nearly impossible:

nPk= n!/(n-k)!

If we have 26 letters and we are choosing 26 at a time, then both n and k are equal to 26: 

26P26= 26!/(26-26)!

26P26 = 4.032914611266 e26

In layman notation, that is a whopping 403,291,461,126,600,000,000,000,000 (~four hundred and three septillion) arbitrary variations of the letter sequence in the English alphabet.

And this, dear readers, is where shit got incredibly trippy.

As I have said in previous posts and will doubtless repeat over and over again in future ones: String Theory, and to some extent quantum theory as a whole, allows for infinite parallel universes. The thing about these infinite parallel universes is that the vast majority of them only differ very slightly from one another.

Like, say, having a different sequence of letters in the alphabet.

Imagine now that there are 403,291,461,126,600,000,000,000,000 parallel universes where the only thing that is different is the sequence of letters in their human alphabets.

This led to the suggestion that perhaps there are also parallel universes where the arrangement of the characters is identical to ours, but the names given to each letter are shifted around. That is another four hundred and three septillion possible alternate realities. You can see how quickly even tiny differences add up to an infinite number of parallel universes.

#string theory #alphabet #linguistics #psychology #trippy

[linguistlist-blog]

Calls: First International Conference on Linguistic Terminology, Glossing and Phonemicization

Call for Papers: Papers on linguistic terminology from all languages and subfields as well as all aspects of glossing and phonemicization related to any language are welcome. This year’s conference will focus particularly on the history of linguistic terminology and interlinear glossing. While the interlinear morphemic gloss is a relatively recent phenomenon, the practice of glossing texts with lexical, morphosyntactic and phonetic annotations is by no means a new phenomenon, with http://dlvr.it/RBxpmt

[linguainfo]

Calls: First International Conference on Linguistic Terminology, Glossing and Phonemicization

http://linguistlist.org/issues/30/30-3230.html?utm_source=dlvr.it&utm_medium=tumblr

[yeli-renrong]

Is contrastive syllabification possible?

According to Martin 1991, the valid word initials in Sikaritai are [ɸ~h b w t d~l s k kʷ ɸɾ bɾ tr sɾ kɾ]. The inventory of valid word-medial syllable initials replaces [d~l] with [ɾ d]. These contrast: [kódá] ‘later’, [kóɾɛ̀] ‘lake’.

(Cɣ consonant clusters also exist, but the ɣ bears tone, so these are best treated as allophonically reduced CVk sequences and will be ignored here.)

The unit consonants are easy to account for: /f b w t d s k kʷ/. Martin unifies the [ɾ] of clusters with /d/, giving /fd bd td sd kd/. But what about the word-medial consonants?

Sikaritai permits two final consonants, [d̚ g̚], which Martin calls /d k/. When a word ending in [d̚] takes a vowel-initial suffix, the [d̚] becomes [d]: [píd-kò] ‘sudah dipotong’, [pìd-ó-wà] ‘potong’. (These are both glossed as English ‘cut’; the Malay gloss indicates the difference in aspect.)

So Martin proposes contrastive syllabification: /kód.á/ ‘later’ vs. /kó.dɛ̀/ ‘lake’. This is optimal from the perspective of segmental economy, but it violates established principles of phonemicization: if there are minimal (or near-minimal) pairs between two phones, they should be assigned to different phonemes.

In English, /h/ can only appear in onset position and /ŋ/ can only appear in coda position. These could be unified as /ꜧ/, and English could be said to have contrastive syllabification, but we don’t do that -- they’re different enough sounds that they’re assigned different phonemes anyway.

(Given that both consonants have further distributional restrictions, I’m actually not sure whether contrastive syllabification would even be required for unification. /h/ can, I think, only occur word-initially or as the onset of a stressed syllable, and /ŋ/ has some bizarre set of restrictions involving alternation with /ŋg/. The heng phoneme would obviously be [h] word-initially, so the question is whether /ŋ./ could appear immediately before a stressed vowel.)

But the same could be done for Sikaritai [-d-] vs. [-ɾ-]! There’s no reason coda [-d̚] has to be the same thing as onset [d-]; in fact, there’s no reason the onset and coda inventories of Sikaritai should be analyzed as overlapping at all. The language could just as well be described as having an onset inventory of /f b w t r s k kʷ/ + /fr br tr sr kr/ (with word-initial fortition of /r/ to [d-]) and a coda inventory of /d g/. So the minimal pairs would still be CVC.V vs. CV.CV, but the syllabification would be automatic: /kódá/ with /d/ which can only occur in coda, /kórɛ̀/ with /r/ which can only occur in onset. Even if Martin’s CVC.V vs. CV.CV analysis is found to correspond to indisputable psychological reality for the speakers of Sikaritai, by way of language games, superintelligent oracles, or whatever other contrivance, it would still be possible, principled, and reasonable to construct the phonemic analysis in a manner that avoids contrastive syllabification! And the same is true for Kiparsky’s examples of Finnish /hææt/ vs. /hæ.æt/ and /ha.uis.sa/ vs. /hau.is.sa/ and Japanese /an.i/ vs. /a.ni/ vs. /an.ni/: why not /hæːt/ and /hææt/, /haui̯ssa/ and /hau̯issa/, and /aɴi/, /ani/, and /aɴni/?

Are we sure that contrastive syllabification isn’t definitionally impossible?

#linguistics

[yeli-renrong]

@possessivesuffix​:

Contrastive syllabification as syllable boundary (hiatus) versus zero (long vowel/diphthong) is reasonably well known, constrastive syllable boundary placement across a consonant group rather less so. Even the supposed English examples I’ve seen seem like a “weird way to get rid of” some nascent distinctions like [t] – [ɾ].

Similar adjustments could be proposed fairly rampantly really whenever a language lenites some phoneme *X medially into a new sound, and then reintroduces the old sound from some cluster or by some sort of compounding. E.g. for Middle English we could suggest that medial [v ð z] are /.f .θ .s/, while medial [f θ s], which in the first place originate from Old English /ff θθ ss/, are /f. θ. s./.

I don’t see how that Middle English example holds -- syllable-initial voicing was dialectally limited in MidE -- but I think this general principle does. I’m not sure if this analysis would hold up to scrutiny, but offhand I think English [t]~[ɾ] could come down to syllabification, e.g. autism [ˈɔˌtʰɪ.zəm] vs. Chartism [ˈtʃɑɹɾˌɪz.əm]. When I thought of the example word autism I expected this could be reduced to the interaction between syllabification and stress, but now I’m not so sure; I think the secondary stresses in autism and Chartism are the same, and avoiding contrastive syllabification would require a morphological analysis. (cf. finger vs. singer)

But I think there’s a general principle that could be extracted here. Let’s say a language has phone X in unambiguous syllable-initial position (e.g. word-initially and word-medial cluster-finally), phone Y in unambiguous syllable-final position (e.g. word-finally and word-medial cluster-initially), and X~Y contrasts intervocalically. We’d like to be able to unify X and Y, but the intervocalic contrasts still need to be explained. Under the conventional rules of phonemicization, we’d posit both /X/ and /Y/, but we could instead posit one phoneme -- let’s call it /Z/ -- and contrastive syllabification: /$Z/ [X], /Z$/ [Y].

As an example, let’s say we have a language with CVC syllable structure, and the phones [l] and [ɹ]. [l] appears word-initially and cluster-finally (e.g. [lapom], [meklep]); [ɹ] appears word-finally and cluster-initially (e.g. [dakuɹ], [noɹpak]), and [l] and [ɹ] contrast intervocalically and only intervocalically (e.g. [kolo] vs. [koɹo]). Under the accepted conventions of phonemic analysis, we’d have /lapom/, /meklep/, /dakuɹ/, /noɹpak/, /kolo/, and /koɹo/, with the distributional constraint that /l/ cannot appear syllable-finally and /r/ cannot appear word-initially or after a consonant; but we could just as well unify them as /L/ and have /Lapom/, /mek.Lep/, /Lakud/, /noL.pak/, and, with contrastive syllabification, /ko.Lo/ and /koL.o/.

But... we’d really like to have some morphological tests to confirm our phonemicization. How would this work? If we have a root /taL/ [taɹ] and an affix /-a/, we have the form /taL-a/. If this produces /taL.a/ [taɹ.a], doesn’t this instead confirm the two-phoneme analysis? So we’d like it to instead produce /ta.La/ [ta.la]... but this involves resyllabification!

Under the two-phoneme analysis, this would probably be taken as a morphophonemic rule, and indeed it’s not unheard of for morphophonemic rules to conspire to avoid sequences that are permissible in roots.

[yeli-renrong]

I tried to reconstruct Proto-Mekeo from Alan Jones’s pan-dialectal wordlist, but I couldn’t find any sound changes for West Mekeo. Other than loss of *r, which is a cover symbol for the correspondence NW 0 : W 0 : N 0/l : E l, but Jones says this /l/ is intrusive, and instances of “*r” in N could just be loans from E. (N lapu-lapu <? POc *qapuk.)

In NW, *b > β - also allegedly *l > e but I think that loses you more regularity than it gets with Jones’s wordlist, so absent better information I’d call it sporadic. In N, p > f. In E, p k > f ʔ, b g w > p k f.

That doesn’t line up with the attested inventories, but neither do Jones’s wordlists. N should only have six consonants, /b k m ŋ v l/, but if I run the slightly phonemicized (e.g. [ts n] > /k ŋ/ since they’re the palatalized allophones) data through a frequency counter, I get /ŋ b k m g f l/ with >10 occurrences and /w p v t n ʔ β/ with <10. (Some of that, at least, is not failure of phonemicization; the word for ‘red’ is perfectly regular except that it has to be reconstructed as *bitoŋa.)

[v] looks like it should be /b/ -- velo ‘good’ < *belo, ivi ‘river’ ~ W ubi, NW ui (there’s some vacillation between i/u, o/u, and e/i, so *ubi/*ibi works here), and... ivi ‘water’!

[p] corresponds to E [p] in ‘dust’ and ‘heart’ and to E [f] in ‘few’ and ‘hold’. Clearly irregular, possibly errors of transcription or hearing in all cases.

[w] is entirely regular but could potentially be an error for a vowel.

[β] occurs once and corresponds to E [p], so it’s probably an error for /b/.

[ʔ] appears in eŋaiaeŋaʔiŋa ‘that’, which looks like garbage.

But it really looks to me like there’s a k/g contrast, which would put every dialect at seven or eight consonants: NW /p β k g m ŋ l/, W /p b k g m ŋ l/, N /f b k g m ŋ l/, E /f p ʔ k m ŋ l/, possibly + /w/ in all but E.

If E is the only dialect with a consonant merger, that’d make it the least useful for external comparison (anyone have a dictionary of Motu?), but of course it’s the standard dialect. Jones cites some forms from the nearby language Kuni to show that one dialect apparently had unconditional *d > g (maybe also *t > k), but it’d be nice to have outside corroboration. OTOH, ear is W aiŋa <? POc *taliŋa.

A quick search for potential POc cognates:

bite: gaŋa < *kaRati- bone: uŋia < *suRi breast: kuku < *susu come: mai < *lako mai drink: iŋu < *inum dust: N lapu-lapu < *qapuk ear: aiŋa < *taliŋa eat: aŋi < *kani eye: ma < *mata father: ama < *tama-ña feather: bui-ŋa < *pulu five: ima < *lima four: baŋi < *pat(i) (ba-ŋi?) fruit: bua < *puaq good: lobia < *ma-pia (*lo-pia?) good: belo < *paliji I: E lau < *au kill: au-buŋu-a < *puŋuq leaf: ŋaŋau < *rau (*ra~rau?) live: mauŋi < *maqurip liver: ake < *qate louse: E uʔu < *kutu moon: puia < *pulan name: aga < *qajan nose: gu < *ijuŋ road: gia < *jalan see: ia < *kita smoke: agu < *qasu suck: mika < *miji they: ia < *ira three: oio < *tolu

Not only are most of these probably wrong, a lot of them are mutually exclusive! Better methods will be necessary to figure out how the Mekeo consonant inventory happened - comparison with more closely related languages, or getting ABVD into a format that can actually be queried, or something.

[wardoftheedgeloaves]

Has anybody actually written an article trying to debunk the Automated Similarity Judgment Program? 

It shouldn’t be that difficult to do this. We know of groups of clearly related languages that are separated at a vast remove, such as Wiyot/Yurok/Algonquian--or, e.g., Irish, Bengali and Dutch. Just for fun, let’s compare the words from one to ten in these three (arbitrarily using Connacht for Irish) converting them to the heavily simplified phonemicization that the ASJP uses:

Irish: /in do tri TEh3r kuT Se SExt oxt ni dEh/

Dutch: /en tue dri vir vEif zEs zEv3n oxt nex3n tin/

Bengali: /Ek dui tin Car pEC Coj SEt Et noj doS/

I mean...eh OK the similarities are still pretty obvious. You’d get Irish and Dutch grouped together, I’m pretty sure. Some words, like “four” and “five”, don’t look at all similar--and good luck with trying to get correspondences out. (Original word-final *ŋ gives n in Irish and Dutch and k in Bengali...)

(but the ASJP uses a forty-word list with a few words on it that are prone to replacement, such as pronouns. A disproof-of-concept shouldn’t be that difficult.)

More broadly though there’s a point here about how much gets lost when you have a family with a small number of members separated by enormous time differences. Yurok /Wiyot/Algonquian are probably not that much older a family than Indo-European and may even be younger--but in the total absence of other branches or premodern documentation, comparative work is extremely difficult when it isn’t impossible.

[yeli-renrong]

This paper contains an overview of the difficulties with Sino-Tibetan prefixes. There are a lot of them! For example, in the root *kwəy ‘dog’, the Chin languages reinterpreted the initial *kw- cluster as a preinitial + initial sequence *k-wəy, and then dropped the preinitial, giving forms like Lushai ui. Karenic also did this, but instead of dropping the preinitial, it... substituted another one, and then re-monosyllabized the root, giving thwi.

And, according to the paper, in some roots that began with *pV, the initial *p- excreted an allophonic [w], which was phonemicized, and then the *pw- cluster was reanalyzed as a preinitial + initial sequence *p-w-. 

Apparently there are a lot of words in Sino-Tibetan that show variance between *p- and *w-, and there’s no consistent cross-family pattern to it at all.

#linguistics