A guide to small consonant inventories

[Pole, the]

RESOLVED WORKSFORME

Ah, now it works. Thanks.

[Nortaneous]

I've been looking up allophony for these. Many of these have allophony like [m] ~ [b] or [l] ~ [r] which means it'll be a bit misleading to assert that the language "lacks /m/" or "lacks /r/".

Doesn't Hawaiian have alveolar~velar variation, at least in the stops?

For /f/, Iau has ɸ~h word-initially, h word-medially, and p word-finally, and I think d varies with implosive ɗ or l before a though I'm running entirely off memory since I can't open the PDF on this computer.

Rotokas has [s] as an allophone of /t/ before /i/.

As for Keuw: voiced stops vary with nasals, voiceless stops can be freely prenasalized, b varies freely with β, and l may be ɾ and usually is after a stop. Apparently the name of the language is usually spelled Kehu.

[Drydic]

I've been looking up allophony for these. Many of these have allophony like [m] ~ [b] or [l] ~ [r] which means it'll be a bit misleading to assert that the language "lacks /m/" or "lacks /r/".

Doesn't Hawaiian have alveolar~velar variation, at least in the stops?

Dialectally. The shift t > k didn't quite make it to some dialects.

[----]

Rotokas has [s] as an allophone of /t/ before /i/.

Gilbertese as well (which should belong here--it only has ten consonants /p pˠ t k m mˠ n ŋ βˠ ɾ/).

[gach]

Some more languages under T with major allophony indicated, all data taken from SIL sheets:

- Tainae with /p t k ʔ d m n f s h/ (10 in total)

- North Tairora with /p t k ʔ mp nt ŋk m n β h~s ɾ~l/ (12 in total)
The allophony of /h/ between [h~s] is apparently free variation but only applies to a "limited number of words".

- Tauade with /p t k m n v l/ (7 in total)
/t/ becomes /ts/ before /i/, /l/ varies between [l~d] before /i/ and [l~r] in other intervocalic environments.

- Tigak with /p t k b g m n ŋ β s ɬ r/ (12 in total)

- Tinputz with /p t k ʔ m n β s h r~l/ (10 in total)

- Tolai with /p t k b d g m n ŋ β r l/ (12 in total)
There's the statement "<s> occurs in some dialects and in loan words". This is probably just a bad way to say that some dialects have /s/ as a native phoneme.

It would still be nice to know what's the real deal with the conflicting data on Taoripi and Koiari. Perhaps other inputs should be checked as well for their integrity.

I'm tempted to attempt some sort of a cluster analysis next (inventories placed at a distance from each other by the number of differences; something like /m/ versus /m~b/ or /r/ versus /ɾ/ might count for less) but I'd have to look for suitable software first.

What sort of clustering did you have in mind? I'm not a heavy user of R but it should be able to offer tools for most of the tricks you'd want to do and the web is full of tutorials for it. In any case I'd start myself by calculating all reasonable inter phoneme probabilities Pr(language has /C₁/|language has /C₂/) just to get some idea of what's there and where it pays off to look more closely.

[Tropylium]

I'm tempted to attempt some sort of a cluster analysis next (inventories placed at a distance from each other by the number of differences; something like /m/ versus /m~b/ or /r/ versus /ɾ/ might count for less) but I'd have to look for suitable software first.

What sort of clustering did you have in mind? I'm not a heavy user of R but it should be able to offer tools for most of the tricks you'd want to do and the web is full of tutorials for it.

I'm mainly planning to dig up languages that are "minimal pairs" or close by with respect to their inventories and seeing if that suffices to build up a web of relations; and if that adds up to some sort of a "hierarchy" in the acquisition of new consonants.

some examples of closely similar phonologies:
– Iquito and Tiriyó differ in that the former has [w] while the latter has [β]; the former has [r] while the latter has [ɾ ~ ɽ]; and the latter includes [ʃ] as an allophone of /s/. These are the ~same inventory differing in 3 phonetic details. Distance = 0.6
– Rarotongan and Tongan only differ in that the former has /r/ while the latter has /l/. These differ only in one feature, and can be considered different inventories differing in 1 consonant. Distance = 1
– Pawnee and Witchita differ in that the former has /p/ while the latter thas /kʷ/; the former has [r] while the latter has [ɾ n]; and the latter has /j/ while the former doesn't. 2 phonological and 1 phonetical differences. Distance = 2.2
— Namia and Sentani differ in that the latter has /f/ and /h/ while the former has /ɾ/. Unlike with Raro/Tongan, these can't really be considered interchangeable though. Distance = 3

[R.Rusanov]

Three differences are less distant than 1?

[Legion]

Sure, if you can't read.

[R.Rusanov]

I fail to see why three distinctions make the same phonology, differing allophonically, while one distinction makes two different and more distant phonologies; I gather this makes me illiterate. Sorry

[Nortaneous]

is r~l not also a phonetic detail

[R.Rusanov]

Yeah, that's basically my question.

[Tropylium]

The data includes several examples of an /r/ ≠ /l/ contrast, versus none of /w/ versus /β/. Hence [±lateral] is a significant feature in the analysis, while [±high] is a redundant one (for [+labial] segments, anyway). Basically the same reason as why we're tracking velar vs. coronal, but not dental vs. alveolar.

And sure there are cases of [r] = [l] too, but this kind of equality is, alas, not transitive. It would be about equally possible to argue for equivalences between /l/ and /d/, or /d/ and /n/, but clearly still /l/ ≠ /n/ etc. And calling this measure a "distance" is really not going to work without the triangle inequality.

The exact weighing was an impressionistic thing though; I suppose e.g. having [ʃ] as a variant of /s/ could perhaps be more reasonably taken for something like 0.5 instead, since there indeed are cases of /s/ ≠ /ʃ/.

[Tropylium]

and sure it would be handy if this data could be simply treated as an (approximately 25-dimensional) set of independent variables, but no — the distance of "+/s/ -/ʃ/" from "-/s/ +/ʃ/" should clearly be considered smaller than a distance like that of "+/n/ -/ʔ/" from "-/n/ +/ʔ/". And the fact that we can still "walk" thru nearby phonemes to almost anywhere else (e.g. /n/ is close to /d/ is close to /t/ is close to /k/ is close to /ʔ/) means that it's also not even possible to consider e.g. the sibilant dimensions as a single isolated "bundle".

This is also why I don't plan on calculating differences between widely different phonologies, just finding the closest couple neighbors for everything.

[Nortaneous]

updated chart

[Nortaneous]

Updated again. Does anyone know what's going on with Mekeo? Blust doesn't. I've listed all the inventories he gives in his book on Austronesian (including the one for general Mekeo, which differs from the one I already had) but it could probably be narrowed down.

[cromulant]

Karajá. Really now, the fuck is this supposed to be?.

The voiceless plosive: /k/
The voiced plosives: /b d/
The implosive: /ɗ/
The affricates: /t͡ʃ d͡ʒ/
The fricatives: /θ ʃ h/
The lateral: /l/
The tap: /ɾ/
The glide: /w/

Can't tell if serious...

[cromulant]

And here we have its sister lang, Apinayé, also with 12: /p t c k ʔ m n ɲ ŋ v r z /.

I repeat: /z/ without /s/! Apinayé's only fricatives are voiced!

Strange things happen in the Amazon...

[Acid Badger]

Strange things happen in the Amazon...

What I always wonder though is whether such strange inventories come into existence because of gaps in the available data. The better studied languages down there look much more sane, and even some analyses of Pirahã's inventory don't look that weird.

[Sakir]

Perhaps, due to an abundance of humor, they talk weird in front of the intrepid field linguist.

[kodé]

Karajá. Really now, the fuck is this supposed to be?.

The voiceless plosive: /k/
The voiced plosives: /b d/
The implosive: /ɗ/
The affricates: /t͡ʃ d͡ʒ/
The fricatives: /θ ʃ h/
The lateral: /l/
The tap: /ɾ/
The glide: /w/

Can't tell if serious...

Not really that strange: if /d ɗ/ is really /t d/ on the level of abstract features, then you get /b t d t͡ʃ d͡ʒ k/, which is completely reasonable. Having the two non-glottal fricatives be /θ ʃ/ isn't that strange either: you can see them as underlyingly coronal fricatives with a +/- anterior contrast. The only really weird thing here is the lack of nasal phonemes; I'd be surprised if there weren't at least some nasalized allophones of something.

[sangi39]

The only really weird thing here is the lack of nasal phonemes; I'd be surprised if there weren't at least some nasalized allophones of something.

According to the linked Wikipedia article, there are:

. /a/ is nasalized word initially and when preceded by /h/ or a voiced stop: /aθi/ → [ãθi] 'grass', /ɔha/ → [ɔhã] 'armadillo'; this in turn nasalizes a preceding /b/ or /d/: /bahadu/ → [mãhãdu] 'group', /dadi/ → [nãdi] 'my mother'

[cromulant]

Not really that strange: if /d ɗ/ is really /t d/ on the level of abstract features, then you get /b t d t͡ʃ d͡ʒ k/, which is completely reasonable.

/b t d t͡ʃ d͡ʒ k/ is indeed less weird...but I have not seen that cited as Karajá's plosive inventory anywhere, and it seems to be a pretty well-documented language.

And apparently, men's speech lacks [k] and certain instances of [ t͡ʃ]!

Having the two non-glottal fricatives be /θ ʃ/ isn't that strange either: you can see them as underlyingly coronal fricatives with a +/- anterior contrast.

Yeah, that's very mild weirdness. It didn't even register with me after being so shocked by the plosives.

The only really weird thing here is the lack of nasal phonemes; I'd be surprised if there weren't at least some nasalized allophones of something.

There are: numerous Amazonian (also African) languages are said to "lack nasal phonemes;" nasals are allophones of voiced stops after nasal vowels. (Or, voiced stops are plosives allophones of nasals after oral vowels. But my impression is that the voiced stop allophones are usually taken as primary).

[----]

And apparently, men's speech lacks [k] and certain instances of [ t͡ʃ]!

That was the most spectacular thing about it to me--if some men really do get rid of *all* instances of [k], then for some speakers the language has no unvoiced plosives, unless you count affricates.

[2+3 clusivity]

As for Karajá, Dixon and Aikenvald's The Amazon Languages (p, 176) has the same consonant inventory as shown on Wiki. They cite fortune '73.

[Nortaneous]

Updated. Going through all the organized phonology datas on SIL PNG -- done through C as of now.

Yeah, that's very mild weirdness. It didn't even register with me after being so shocked by the plosives.

Also shows up in Asia -- Zhuang has a fricative inventory of /f β θ ɕ ɣ h/ and Saisiyat /θ ʂ ħ/. I think Kazakh alveolar fricatives have dental thibilant allophones -- either that or the one Kazakh I've met had a lisp.

Nasalization only of /a/ in Karaja is interesting -- /a/ is apparently always nasalized in Iau, which otherwise has no nasality whatsoever anywhere in it.