Clicks and codas

[Mâq Lar]

The few natlangs that have click consonants have them as word onsets (in all such languages) and word medially (in a few of those languages), but never in coda position. Why? I've never come across an actual explanation, but the obvious possible reasons seem to be
1. articulatory - it's too hard to make clicks in coda positions
2. perceptibility - they don't sound like they belong to the previous syllable and so interfere with perception of word boundaries
3. sonority - clicks are (apparently) the least sonorous of all sounds, and the sonority sequencing principle (or an extension thereof) won't allow codas to fall to such a low level of sonority. Perhaps connected with the point about them not being easily perceived as belonging to the previous word
4. historical - maybe they originated from a reduction of clusters that couldn't occur in coda position
5. coincidental - languages with clicks tend to have few codas anyway, so it might just be that they are low on the list of good codas and so never appear even though it's not in general impossible

Any ideas?

[Xephyr]

Khoesan languages are very restricted when it comes to what consonants are allowed where. IIRC none of them allow word-final consonants, and very very few consonants are allowed intervocalically. Initials, however, are often very complex even when talking about non-clicks. So the reason clicks have a restricted distribution is because consonants in Khoesan have a restricted distribution, and if you ask me speculating beyond this point is pointless.

[Nortaneous]

Do any languages with clicks allow word-final consonants?

edit: Sandawe, apparently.

[Mâq Lar]

So in that case is it fair to say that clicks are not specifically restricted in codas any more than other consonants are?

And that presumably then they would pattern phonologically as obstruents of low(est) sonority?

[Xephyr]

Talking about it in terms of sonority isn't very useful, either, since Khoesan languages are unusual in that area as well. For instance, they allow complex contour-segments that resemble (and probably descend from) consonant clusters in word-initial position despite elsewhere having strict CV syllable structure, which is atypical.

[Mâq Lar]

I didn't mean analysing the distribution in Khoesan or other click languages in terms of sonority hierarchy; I meant rather that in the (admittedly tiny) sample of languages in which they exist, they don't really behave specially (particularly in regards to being admissible in a coda), so there's no reason to suspect that if they were more widespread they wouldn't just pattern as any other kind of consonant, and so presumably as a consonant of low sonority.

Also, I'm not sure what you mean about complex contours being unusual with respect to strict CV syllable structure; you mean they don't allow consonant clusters but allow complex contours (presumably the mixed voicing combinations and the contour clicks) which might have developed from them? I'm not sure why that should be atypical - clusters reduced to something closer to CV, giving a complex C (or is that the atypical part since usually languages delete to simplify?). Also, at least some Khoesan and other click languages do allow onset clusters though, and AFAIK the clusters don't violate the sonority sequencing principle. Have I misunderstood something?

[Xephyr]

Also, I'm not sure what you mean about complex contours being unusual with respect to strict CV syllable structure; you mean they don't allow consonant clusters but allow complex contours (presumably the mixed voicing combinations and the contour clicks) which might have developed from them?

I mean that in Khoesan the distinction between "consonant cluster" and "contour segment" becomes very blurry.

I'm not sure why that should be atypical - clusters reduced to something closer to CV, giving a complex C (or is that the atypical part since usually languages delete to simplify?).

The complex C's are only present word-initially, though. Usually if a language only tentatively allows consonant clusters then they are ambisyllabic, combining a codic "VC" with an initial "CV", which Khoesan disallows.

Also, at least some Khoesan and other click languages do allow onset clusters though, and AFAIK the clusters don't violate the sonority sequencing principle. Have I misunderstood something?

It's atypical in that the only initial clusters (if they are clusters) allowed are obstruent-obstruent clusters like <tχ>, <gkx'>, or <gǂq>, without allowing any clusters (initial or otherwise) that are obstruent-sonorant or sonorant-obstruent.

[Mâq Lar]

Ah yes, I see what you mean.

Although I'm not sure that sonority is not a useful idea when it comes to Khoesan; I googled and found this PhD thesis on Khoekhoe; the author describes the restrictions on root initial and and root medial consonants in terms of sonority: root initial, clitic initial and root medial consonants in both Khoekhoe and ǃXóõ are correlated with 'strength' in roughly that order. So root initial is the only position with the 'strongest' (least sonorant) sounds - clicks - and it lacks the most sonorant sounds (approximant). Medial position is the reverse - it only has more sonorant sounds (nasals and approximants). (Clitic initial position is somewhere between the two - onsets don't allow the highest and lowest sonority consonants, so no clicks or approximants). (The table from the paper is hiding in the 'more')

More: show

image.jpeg (203.59 KiB) Viewed 4223 times

Sandawe and Zulu allow all consonants both initially and medially in roots (although clicks are more common in initial than medial position), but not in affixes - the author uses this to point out that the restrictions on clicks are language specific, rather than inherent to clicks themselves. She analyses Khoekhoe as having a strong preference for marked features to appear at the left-edge of prosodically words - so clicks, contour tones, and the citation form of tone melodies (rather than the less marked sandhi form) appear initially in content words, and not medially or in affixes

So it seems that clicks just pattern as rare, marked, extra-low sonority/extra-high strength consonants; with the preference for initial position over medial and the non-occurring coda position coming from the cross linguistically common sonority shape of syllables (onsets are preferably low sonority, codas are high) and the language-specific preference of Khoesan for marked features at left-edges.

Theoretically then, a language could be expected ('expected' is a strong word; it just wouldn't be an impossibility) to have coda clicks (albeit probably with lower frequency than the other consonants) if it had very permissive phonotactics across the board - no restrictions on minimum sonority of codas in particular, or an Arrernte-like preference for codas maybe. The only problem I could see with that would be perceptive - a listener might naturally parse coda clicks as onsets of the following word. Maybe it would only be practical in a language with strict CV(C) order, or where coda clicks appear pre-pausally, with allophones in other positions. Hmmmmmm

[Salmoneus]

I've not read up on it, but to me the restrictions on khoisan clicks seem to heavily suggest something like SEA first-vowel reduction, with subsequent transformation of the cluster into a click. Something like kata > kta > |a. [this would also help in producing more complicated tone systems]

[Mâq Lar]

That's generally assumed to be their origin, yes. Cross-morpheme /t.k/ clusters in German are occasionally realised as clicks

Although that wouldn't entirely explain the other marked features surfacing at the left edges of words though

[alynnidalar]

Cross-morpheme /t.k/ clusters in German are occasionally realised as clicks

Really? That's quite interesting! Do you have any further reading on this or links to video/audio where this is happening?

[Mâq Lar]

Really? That's quite interesting! Do you have any further reading on this or links to video/audio where this is happening?

Just this paper. Maybe Google or YouTube would turn up some recordings

[Nortaneous]

That's not "generally assumed" as in "it's generally assumed that PIE had three series of stops". It's "generally assumed" as in "it's generally assumed that lightning comes from Thor, dunno man, you have any better ideas?".