The weird natlang phoneme thread

[gach]

there's no way to tell the difference between units and clusters from a language-internal basis, so if you make that decision, it has to be done based on cross-linguistic sanity

wat

There are plenty of languages where there are e.g. strong distributional arguments for whether a given segment is a cluster or not.

I think that's meant for Khoisan languages specifically. There the analysis of some complex onsets as clusters or single phonemes might be unresolvable without consulting cross linguistic typological arguments or asking what analysis your professor wants you to follow so that he'll grant you funding for the next year as well.

Honestly I think that most of the Khoisan complex onsets are indeed best to be analysed as single consonant phonemes. Still, there are some cases where I'm very tempted to try a cluster analysis, like the Naro fricated consonants. It could be possible to analyse these as Cx clusters so that /x/ gets an uvular realisation after other consonants. There are some problems with that road but I confess preferring a splitter analysis over eagerly positing yet more funky phonemes that my favourite language has and yours doesn't. I guess that's a symptom of knowing too many particle physicists who like to solve problems by proposing new particles or fields.

I'd like to know if any natlangs out there: ...

-allow geminate consonants in clusters, e.g., /m:b/, /nz:/.

In Finnish nasal/liquid+obstruent clusters distinguish between short and geminated obstruents:

lanka "string" ~ ankka "duck"
anti "yield" ~ Antti male personal name
alta "from below" ~ taltta "chisel"
kansa "people" ~ kanssa "with"
irti "loose" ~ pirtti "cabin"

[Dē Graut Bʉr]

English's got /h~ŋ/.

[Pogostick Man]

English's got /h~ŋ/.

Coda rhinoglottophilia!

[Shemtov]

There's a Dravidian language called Toda that has a Retroflex Lateral Fricative.

[CatDoom]

There's a Dravidian language called Toda that has a Retroflex Lateral Fricative.

Toda has a lot more than that going for it; it's a pretty unusual language in a lot of ways, and I love it. In addition to a dozen fricatives and two lateral approximants, it has six different rhotic trills, which includes two that are partly retroflexed and four that seem to be partly distinguished from one another by a +/-ATR feature.

[Tropylium]

I'd like to know if any natlangs out there: ...

-allow geminate consonants in clusters, e.g., /m:b/, /nz:/.

In Finnish nasal/liquid+obstruent clusters distinguish between short and geminated obstruents:

And in several Sami languages you can find the reverse, with gemination contrasted in the coda position (though I think this is only found in consonant gradation, not in different roots entirely). Some example accusativ : nominativ pairs from Pite Sami (of which a snazzy grammar has been released recently):
/paːrʰko/ : /paːrːʰko/ 'bark'
/luakta/ : /luakːta/ 'bay'

For some varieties, I've seen it proposed to analyze long preaspirated consonants as clusters ([ʰpː] = /hːp/ etc.), so that would answer also your first question.

---

On the topic of weird natlang allophones, I recently learned that Surgut Khanty realizes the sequence /wu/ as [ʁʷu]. Seems similar to the common thing of realizing /ji/ as [ʝi], but with uvularness as an extra complication.

[gach]

And in several Sami languages you can find the reverse, with gemination contrasted in the coda position (though I think this is only found in consonant gradation, not in different roots entirely).

Come to think of it, I can't think of anything else except for the development of gradation that would have given rise to these stretched codas.

For some varieties, I've seen it proposed to analyze long preaspirated consonants as clusters ([ʰpː] = /hːp/ etc.), so that would answer also your first question.

At least for me when I listen to North Saami being spoken, the preaspiration appears as a long [h:].

[Tropylium]

And in several Sami languages you can find the reverse, with gemination contrasted in the coda position (though I think this is only found in consonant gradation, not in different roots entirely).

Come to think of it, I can't think of anything else except for the development of gradation that would have given rise to these stretched codas.

They come from that, yes. I am open to the idea that there might exist two originally homophonic roots of which one is however now found only in the strong grade and another only in the weak grade — in which case these would be an example of a coda gemination contrast that is purely lexical and not morphophonological.

[Clearsand]

There's a Dravidian language called Toda that has a Retroflex Lateral Fricative.

That is a delicious phone.